id stringlengths 3 9 | source stringclasses 1 value | version stringclasses 1 value | text stringlengths 1.54k 298k | added stringdate 1993-11-25 05:05:38 2024-09-20 15:30:25 | created stringdate 1-01-01 00:00:00 2024-07-31 00:00:00 | metadata dict |
|---|---|---|---|---|---|---|
247465975 | pes2o/s2orc | v3-fos-license | Addressing Syntax-Based Semantic Complementation: Incorporating Entity and Soft Dependency Constraints into Metonymy Resolution
: State-of-the-art methods for metonymy resolution (MR) consider the sentential context by modeling the entire sentence. However, entity representation, or syntactic structure that are informative may be beneficial for identifying metonymy. Other approaches only using deep neural network fail to capture such information. To leverage both entity and syntax constraints, this paper proposes a robust model EBAGCN for metonymy resolution. First, this work extracts syntactic dependency relations under the guidance of syntactic knowledge. Then the work constructs a neural network to incorporate both entity representation and syntactic structure into better resolution representations. In this way, the proposed model alleviates the impact of noisy information from entire sentences and breaks the limit of performance on the complicated texts. Experiments on the SemEval and ReLocaR dataset show that the proposed model significantly outperforms the state-of-the-art method BERT by more than 4%. Ablation tests demonstrate that leveraging these two types of constraints benefits fine pre-trained language models in the MR task.
Introduction
Metonymy is a common figurative language phenomenon that refers to substituting the name of a thing using one of its closely associated attributes (e.g., producer-for-product, place-for-event, place-for-inhabitant). This linguistic phenomenon [1,2] is pervasive in daily life and literature. For example, named entities in the text are often used in a metonymy manner to imply an irregular denotation.
Identifying ambiguities in metonymy is a fundamental process in many NLP applications such as relation extraction [3], machine comprehension [4], and other downstream tasks [5]. The following sentence shows an example of metonymy: "Last year, Ma acquired Eleme". In this example, the meaning of "Ma" under this particular circumstance has been changed. It is more appropriate to interpret "Ma" as "Ma's company Alibaba" instead of the literate concept of "a famous entrepreneur".
In literature, conventional methods for MR mainly rely on the features derived from lexical resources such as dictionaries, taggers, parsers, and WordNet or other handcrafted lexical resources. At present, more researchers are focusing on deep neural networks (DNN) [6,7], which is becoming the mainstream approach to handle various tasks in NLP, including metonymy resolution [8]. DNN models can effectively encode all words in a sentence in a sequential way to obtain the semantic representation of the text to easily capture contextual information throughout the whole sentence and achieve stateof-the-art performance. Moreover, pre-trained language representations demonstrated efficiency in improving many NLP tasks, e.g., text classification [9], event extraction [10], and relation extraction [11]. Benefiting from the contextual representations, these models significantly surpass competitive neural models in many NLP tasks [12,13], for example Figure 1. A metonymy example on the geographical parsing problem. According to linguistic valency in English grammar, the meaning of metonymic entity "Ma" is mainly decided by the root of the sentences, i.e., the verb "acquired". Dependency relations are the basis of understanding metonymy.
Considering entity awareness and dependency information both benefit MR, this paper presents EBAGCN (Entity BERT with Attention-guided GCN) to leverage entity constraints from sequence-based pre-trained language representations and soft dependency constraints from dependency trees at the same time.
As a result, the proposed approach is superior in capturing the semantics of the sentence and the long-distance dependency relations, which better benefits MR. To summarize, the main contributions of this paper are as follows: • propose a novel method for metonymy resolution relying on recent advances on pre-trained language representations that integrate entity knowledge and significantly improve the accuracy. • incorporate attention-guided GCN to MR with hard /soft dependency constraints, which imposes the pre-trained language representations with prior syntactic knowledge. • experiments on two benchmark datasets show that the proposed model EBAGCN is significantly superior to previous works and improves the BERT-based baseline systems.
Related Work
Metonymy Resolution Analysis of metonymy as a linguistic phenomenon dates back to feature-based methods [17,18]. Ref. [19] use an SVM with a robust knowledge base built from Wikipedia to remain the best result in all feature-based methods. These methods inevitably suffer from error propagation because of their dependence on the manual feature extraction. Furthermore, the work takes much effort, while helpful information is hard to be caught. As a result, the performance is not satisfied.
Recently, the majority of works for MR have focused on the deep neural network (DNN). Ref. [8] propose PreWin based on Long Short Term Memory (LSTM) architecture. This work is the first to integrate structural knowledge into MR. In contrast, the effect is limited because of the setting of the predicate window, which only retains the words around the predicate while losing much important information in that process. Other approaches [20] leverage NER and POS features on LSTM to enrich the representation of tokens. However, these methods only improve slightly since they only pay attention to the independent tokens and ignore the relations between tokens in a sentence that is beneficial to MR.
Pre-trained language models have shown great success in many NLP tasks. In particular, BERT, proposed by [16], shows a significant impact. Intuitively, it is natural to introduce pre-trained models to MR. These models vastly outperform the conventional DNN models and reach the top of the leaderboard in MR. Nevertheless, pre-trained models encode the whole sentence to catch contextual features, leading to the ignorance of syntactic features in sentences. Ref. [21] fine-tunes BERT with target word masking and data augmentation to detect metonymy more accurately.
In addition to the context information provided by the sequence-based model, [22] pays attention to the entity word. They disambiguate every word in a sentence by reformulating metonymy detection as a sequence labeling task and investigate the impact of entity and context on metonymy detection.
Dependency Constraints Integration The research on MR so far has made limited application of dependency trees. However, research on other NLP classification tasks widely employ dependency information. Differing from traditional sequence-based models, dependency-based models integrate dependency information [23], taking advantage of seizing dependency relations that are obscure from the surface form alone.
As the effect of dependency information is widely recognized, more attention is paid to pruning strategies (i.e., how to distill syntactic information from dependency trees efficiently). Ref. [3] use the shortest dependency path between the entities in the full tree. Ref. [24] apply graph convolutional networks [25] model on a pruned tree and a novel pruning strategy to the input trees by retaining words immediately around the shortest path between entities among which a relation might hold. Although these hard-pruning methods remove irrelevant relations efficiently based on predefined rules, they suffer from eliminating useful information wrongly at the same time.
More recently, ref. [26] proposed AGGCN and employed a soft-pruning strategy. The method enables the dependency relations to have weights to balance relevant and irrelevant information with multi-head attention mechanism. Ref. [27] proposed a dependencydriven approach for relation extraction with attentive graph convolutional networks (A-GCN). In this approach, an attention mechanism in graph convolutional networks is applied to different contextual words in the dependency tree obtained from an off-the-shelf dependency parser, to distinguish the importance of different word dependencies.
Pre-trained Model This is the idea of pre-training, originated in the field of computer vision, and then developed into NLP. A pre-trained word vector is the most common application of pre-training in NLP. The annotated corpus is very limited in many NLP tasks, which is not enough to train excellent word vectors. Therefore, large-scale unannotated corpus unrelated to the current task is usually used for pre-training word vectors. At present, many deep learning models tend to use pre-trained word vectors (such as Word2Vec [28] and GloVe [29], etc.) for initialization to accelerate the convergence speed of the network.
To consider contextual information when setting word vector, pre-trained models such as Context2Vec [30], ELMo [15] were developed and achieved good results. BERT is a model training directly on deep Transformer network. The best results were achieved in many downstream tasks of NLP through pre-training and fine-tuning [31]. Different from other deep learning models, BERT adjusts the context at all levels jointly before training to obtain bi-directional representation of each token. BERT solves the representation difficulty to a large extent by fine-tuning the output of a specific task. Compared with recurrent neural networks, BERT relying on Transformer can capture long-distance dependencies more effectively and have a more accurate semantic understanding of each token in the current context.
Graph Convolutional Network DNN models have achieved great success in both CV and NLP. As a representative model of deep learning, the convolutional neural network can solve regular spatial structures. While much data does not have structure, the graph convolutional (GCN) network arises at a historic moment. GCN is a widely used architecture to encode the information in a graph, where in each GCN layer, information in each node communicates to its neighbors through the connections between them. The effectiveness of GCN models to encode the contextual information over a graph of an input sentence has been demonstrated by many previous studies [32,33].
The Proposed Model
This section presents the basic components used for constructing the model. The overall architecture of the proposed model is shown in Figure 2.
BERT Encoding Unit
The pre-trained language model BERT is a multi-layer bidirectional transformer encoder designed to pre-train deep bidirectional representations by conditioning both left and right context. This unit uses the BERT encoder to model sentences and output fine-tuned contextual representations. It takes as input sentence S, and computes for each token a context-aware representation. Concretely, the input packs as [CLS, S t , SEP], where CLS is a special token for classification; S t is the token sequence of S generated by a WordPiece Tokenizer; SEP is the token indicating the end of a sentence. For each hidden representation h 0 i at the index i, initial token embedding s tok i is concatenated with positional embedding s pos i After going through N successive Transformer encoder blocks, the encoder generates context-aware representations for each token to be the output of this unit, represented as h N i :
Syntactic Integration Unit
As is shown in Figure 3, the Syntactic Integration Unit is designed to integrate syntax into BERT and is the most crucial component of this approach. In a multi-layer GCN, the node representation h (l) i is produced by applying a graph convolution operation in layers from 1 to l − 1, described as follows: where W (l) represents the weight matrix, b (l) stands for the bias vector, and ρ is an activation. h (l−1) and h (l) are the hidden state in prior and current layer, respectively. Syntactic Integration Unit contains attention guided layer, densely connected layer, and linear combination layer.
Attention Guided Layer Most existing methods adopt hard dependency relations (i.e., 1, 0 denote relation exists or not) to impose syntactic constraints. However, these methods require the pre-defined pruning strategy based on expert experience and simply set the dependency relations considered "irrelevant" as zero-weight (not attended). These rules may bias representations, especially toward a larger dependency graph. Reversely, the attention guided layer helps to launch the "soft pruning" strategy. This layer generally generates the attention guided adjacency matrix A (t) whose weights range from 0 to 1 by multi-head attention [34]. The shape of A (t) is the same as the original adjacency matrix A for convenience. Precisely, A (t) is calculated as follows: where Q, K, V are, respectively, query, key, value in multi-head attention, Q, K are both equal to the input representation R (m−1) (i.e., output of the prior module), d is the dimension of R (m−1) , W Q i and W K i are both learnable parameters ∈ R d×d , A (t) is the t-th attention guided adjacency matrix corresponding to the t-th head.
In this way, the attention guided layer outputs a large fully connected graph to reallocate the importance of each dependency relation rather than pruning the graph into a smaller structure as tradition.
Densely Connected Layer
This layer helps to learn more local and non-local information and train a deeper model using densely connected operations. Each densely connected layer has L sub-layers. L is a hyper-parameter for each module. These sub-layers are placed in regular sequence, and each sub-layer takes all preceding sub-layers' output as input. The structure of Densely Connected Layer is shown in Figure 4. g (l) j is calculated in this layer, which is defined as the concatenation of the initial representation and the representations produced in each preceding sub-layer: where x j is initial input representation, h are the outputs of all preceding sublayers. In addition, the dimension of representations in these sub-layers is shrunk to improve the parameter efficiency, i.e., d hidden = d/L, where L is the number of sub-layers, d is the input dimension. For example, the number of sub-layer is 2 and input dimension is 1024, d hidden = d/L = 512. Then a new representation whose dimension is 1024(512 × 2) is formed by concatenating all these sub-layer outputs. N densely connected layers compute N adjacency matrixes produced by attention guided layer. The GCN computation for each sub-layer should be modified because of the application of multi-head attention: t are learnable weights and bias, which are selected by t and associated with the attention guided adjacency matrix A (t) .
Linear Combination Layer
In this layer, the final output is obtained by combining representations output by N Densely Connected Layer corresponding to N heads: where h out ∈ R d is the combined representation of N heads as well as the output of the module. W out and b out are learnable weights and bias.
Joint Unit
In Joint Unit, the context representation and entity representation are united to form the final joint representation for MR.
Context Representation BERT encoder produces the final hidden state sequence H corresponding to the task-oriented embedding of each token. According to the BERT mechanism, the representation H 0 output by the special token "[CLS]" serves as the pooled representation of the whole sentence. Therefore, H 0 serves to represent the aggregate sequence as context representation.
Entity Representation To help the model capture the clues of entities and enhance the expression ability, entity indicator is inserted at the beginning and end of the entity.
The entity represents as follows: suppose that H m . . . H n are the hidden states of entity E output by Syntactic Integration Unit (m, n represent the start index and end index of the entity, respectively), an average operation is applied to obtain a final representation: Representation Integration For classifying, model concatenate H 0 and H e and consecutively apply two full connected layers with activation.
Classifier Unit
A softmax layer is applied to produce a probability distribution p(y|x, θ): θ refers all learnable parameters in the network W ∈ R d h ×d h * 2 , W * ∈ R r×d h , where r is the number of classification types, d h is the dim of BERT representation.
Dataset
The experiments are conducted on two publicly available benchmarks: the SemEval2007 [18] and ReLocaR [8] datasets. Unlike WiMCor [35] and GWN [36] that contains huge amount of instances, SemEval2007 and ReLocaR are relatively smaller. The samples lay into two classes: literal and metonymic. SemEval contains 925 training and 908 test instances, while ReLocaR comprises a train (1026 samples) and a test (1000 samples) dataset. The class distribution of SemEval is approx 80% literal, 20% metonymic. To eliminate the high class bias of SemEval, the class distribution of ReLocaR sets to be 50% literal, 50% metonymic.
Since the MR task is still in its infancy, there is no available MR dataset for Chinese. English datasets SemEval and ReLocaR are employed to construct a Chinese MR dataset through text translation, manual adjustment, and labeling.
1.
Text translation: examples of SemEval and ReLocaR are translated using API on the Internet and finally obtain independent Chinese samples; 2.
Manual adjustment: Considering the poor quality of the dataset obtained from API, all examples are corrected and well selected to meet the Chinese expression norms; 3.
Labeling: Inserting a pair of indicators to mark the entity.
After the above steps, an MR dataset called CMR in Chinese is constructed to verify the model performance on Chinese texts. Finally, the dataset contains 1986 entity-tagged instances, of which 1192 are randomly divided as a training set and 794 as a test set. Each instance contains a sentence with the entity tag and a classification tag of literal or metonymic.
Dependency Parsing
Given a sentence from the dataset, first the sentence is tokenized with the tokenization tool "jieba". Then, dependency parsing is launched for the tokenization list by the tool of Stanford CoreNLP [37]. After the dependency graph is output by dependency parsing, the dependency relations are first encoded into an adjacency matrix A. In particular, if there is a dependency edge existing between node i and j, then A ij = 1 and A ji = 1, otherwise A ij = 0 and A ji = 0.
Model Construction
The proposed models are encoded with Python 3.6 and deep learning framework PyTorch 1.1. They are trained on a Tesla v100-16GB GPU. EBAGCN requires approx. 1.5 times GPU memory compared with vanilla BERT-LARGE.
Given a sentence S with an entity E, MR aims to predict whether E is a metonymic entity nominal. The key idea of EBAGCN is to enhance BERT representation with structural knowledge from dependency trees and entities. Generally, the entire sentence will first go through the BERT Encoding Unit to obtain the deep bidirectional representation for each token. Then, launching dependency parsing to extract the dependency relations from each sentence. Subsequently, both deep bidirectional representations and dependency relations are fed into the Syntactic Integration Unit. The achieved vector representations are enriched by syntactic knowledge and integrated with context representation in Joint Unit. Finally, the fused embedding is served to produce a final prediction distribution in the Classifier Unit.
The evaluation for experiments are accuracy, precision, recall, and F1.
• accuracy The probability of the total sample that predicted correct results; • precision The probability of actually being a positive sample of all the predicted positive samples; • recall The probability of being predicted to be a positive sample in a sample that is actually positive; • F1 The F1 score is balanced by taking into account both accuracy and recall. The expression of F1 score is:
Models
The baseline models used in the experiment are listed below. SVM+Wikipedia: SVM+Wikipedia is the previous SOTA statistical model. It applies SVM with Wikipedia's network of categories and articles to automatically discover new relations and their instances.
LSTM and BiLSTM: LSTM is one of the most potent dynamic classifiers publicly known [38]. Because of the featured memory function of remembering last hidden states, it achieves promising results and is widely used in various NLP tasks. Moreover, BiL-STM improves the token representation by being aware of the conditions from both directions [39], making contextual reasoning available. Additionally, two kinds of representations, GloVe [29] and ELMo [15] are performed separately to ensure a credible model result.
Paragraph, Immediate, and PreWin: These three models are primarily built upon BiLSTM. They simultaneously encode tokens as word embeddings and dependency tags as one-hot vectors (5-10 tokens in general). The difference between them is in the way of picking tokens. Immediate-y selects y number of words on the left and right side of the entity as input to the model [40,41]. The Paragraph model extends the Immediate model by taking the 50 words from the side of each entity as the input to the classifier. The PreWin model relies on a predicate window consisting of a direct vocabulary around the recognized predicate to eliminate noise over a long distance. fastText: FastText is a tool that computes the word vector and classifies the text without great academic innovation. However, its advantages are obvious. In text classification, fastText can achieve performance similar to the deep network with little training cost.
CNN: The experiment applies the classical CNN model target for the text classification, which is composed of input layer, convolution layer, pooling layer, and softmax layer. Since the whole model adapts to the text (rather than CNN's traditional application: image), some adjustments are made to adapt the NLP task.
BiLSTM+Att: BiLSTM+Att applies an attention layer on the BiLSTM to increase the representation ability.
BERT: BERT is a language model trained on deep Transformer networks which performs NLP tasks well through pre-training and fine-tuning, and achieves the best results in many downstream tasks of NLP [31,42].
Main Result on SemEval and ReLocaR
On SemEval and ReLocaR, this approach compares with the feature-based model, deep neural network model and pre-trained language model. Table 1 reports the results. Table 1. Result on English datasets. On both benchmarks, the F1-score for either class and the overall accuracy are given in this table. "L" and "M" denote literal and metonymic class, respectively. +NER+POS means integrating both NER and POS features with the baseline model. In general, due to the application of advanced pre-trained language representation along with soft dependency clues, EBAGCN obtains best results. Three self-constructed models are compared in the experiment: Entity BERT (BERT model integrating the entity information by joint representation but discarding the syntactic constraints), EBGCN (Entity BERT with GCN, apply normal GCN without attention to impose hard syntactic constraints on Entity BERT), EBAGCN (Entity BERT with Attention-guided GCN, apply attention-guided GCN to impose soft syntactic constraints on Entity BERT).
MODEL
The result in Table 1 shows that the models in this work significantly outperform previous SOTA model SVM+Wikipedia which is based on feature engineering. They also surpass all the DNN models including LSTM, BiLSTM, and PreWin, even if they incorporate POS and NER features to enrich the representation. This result illustrates that the conventional features cannot provide enough contextual information. Entity BERT and BERT are both pre-trained language models. However, there are significant differences in effect due to the incorporation of entity constraint, which greatly improves the accuracy of the model. Furthermore, the three models also perform differently. The experiment on EBGCN produces a decent accuracy that is 0.3% and 0.2% higher than Entity BERT on SemEval and ReLocaR, which illustrates that the application of GCN helps improve performance by catching the ignored information from syntax. Moreover, EBAGCN obtains an improvement of 0.7% and 0.2% compared with EBGCN in terms of accuracy. This fact provides ample proof that the introduction of the multi-head attention mechanism assists GCNs in learning better information aggregations by simultaneously pruning irrelevant information and emphasizing dominating relations concerning indicators such as verbs in a soft method.
The table also gives the F1-score for literal and metonymic results, respectively. The consequence shows that EBAGCN achieves the best F1-score on SemEval (metonymic accounts for 20%) and ReLocaR (metonymic accounts for 50%), which suggests that EBAGCN is adaptive in various class distributions.
To be specific, Entity BERT uses the BERT-based neural network to aggregate information about context and entity semantics to form better semantic vector representation. In this way, the model leverages entity words and enhances the interaction between entity words and context information, thus improving the accuracy and recall rate of MR. The improved result of Entity BERT verifies the importance of entity information and proves that Entity BERT can effectively solve the missing entity information in metonymy.
In the framework of cognitive linguistics, syntactic structures are considered to contain important information. Existing models based on DNN scan the information encoding of the whole sentence sequence and compress it into vector expression, which cannot capture the syntactic structure that plays an important role in the transmission of natural language information. In addition, all syntactic dependencies are added to the syntactic representation vector with the same weight, so it is impossible to distinguish the contribution of each dependency. Thus, EBAGCN steps further to leverage syntax knowledge selectively. The weight allocation system for syntax dependencies of EBAGCN do not just resist noise interference, but also improve the accuracy of MR. Finally, the proposed EBAGCN can effectively solve the low accuracy of long and difficult sentences as well as key word recognition.
Main Result on CMR
Unlike the English dataset ReLocaR and SemEval, Chinese text is harder to understand. Table 2 gives the result on CMR. As can be seen from the experimental results, fastText, CNN, and BiLSTM+Att have little gap in the Chinese MR task. BERT greatly improves the performance by relying on the powerful ability of the pre-trained model, and the Entity BERT proposed in this paper achieves a better result by reinforcing the entity information. Compared with Entity BERT, EBGCN is about 1.2% higher in accuracy, showing a huge performance improvement and proving the significance of dependency knowledge. However, taking advantage of syntactic noise elimination, EBAGCN obtains the SOTA result on CMR, proving the validity of this work.
Entity Ablation Experiment
The validity of the Entity BERT was demonstrated in the main result above. This experiment mines further to understand the specific contributions of each module besides the pre-trained BERT component. Take Entity BERT as baseline, this work proposes three additional models: • Entity BERT NO-SEP-NO-ENT discard both entity representation H e and the entity indicators around the entity, i.e., only representation corresponding to "[CLS]" is used for classification; • Entity BERT NO-SEP only discard the entity representation H e but reserve entity indicators around the entity; • Entity BERT NO-ENT only discard the entity indicators around the entity but reserve entity representation H e . Table 3 shows the results of the ablation experiment. From the table, all three methods perform worse than Entity BERT. Among them, Entity BERT NO-SEP-NO-ENT performs the worst, proving that both entity indicator and entity representation make great contributions to the model. The meaning of using the entity indicator is to integrate entity location information into the BERT pre-trained model. On the other hand, entity representation further enriches the information and helps the model achieve high accuracy. Entity BERT enhances the influence of entity information on discriminant results and provides a solid foundation for accurately representing the joint embedding of entity and context. In MR task, most of the key information is focused on entity. The integrity of entity representation largely determines the performance of the model. Making full use of the semantic, location and structural information of entity words can effectively reduce the influence of noise. Therefore, as entity information in metonymy text is difficult to be extracted and represented, a fusion perception method is applied in this paper, which can effectively improve the accuracy and efficiency of MR under the joint guidance of entity and context knowledge. Based on the above fashion, this paper puts forward Entity BERT that jointly training BERT with entity and context. In that case, Entity BERT makes use of the key information of the entity to reduce the influence of the noise in the sentence, Thus, the accuracy and recall rate are greatly improved.
Entity Contribution Verification
To further study the contribution of entity information on MR, this experiment inputs a single entity representation H e into the BERT model without using contextual information representation H 0 . This work maps semantic representation of several entities into Figure 5.
As shown in the left picture of Figure 5, before BERT is fine-tuned, the semantic representation of the entity is not far from the original, indicating that entity information is very sparse and weak without entity fine-tuning. However, as shown in the right figure, after BERT fine-tuning, metonymic and literal entities are divided into two clusters, which shows that the fine-tuned model can judge the metonymy.
The existing deep learning model depends on the context, whose representation is formed by inputting all the tokens in the whole sentence into the fully connected layer, inevitably containing a lot of noise. This experiment proves that integrating entity in-formation into the language model helps the task of MR and verifies the rationality of Entity BERT. The causes of poor model performance may be more than one. For example, long sentences are likely to affect the accuracy of classification for the following reasons:
UK
• Contextual meanings for long sentences are more difficult to capture and represent. • The position of key tokens, such as a predicate, is noisy and, therefore, difficult to determine.
Intuitively, lacking model interpretability of non-local syntactic relations, sequencebased models such as Entity BERT cannot sufficiently capture long-distance dependence. Thus, the accuracy of Entity BERT drops fiercely as predicted as shown in Figure 6 when the sentence length grows. However, such a performance degradation can be alleviated using EBAGCN, which suggests that catching a mass of non-local syntactic relations helps the proposed model accurately infer the meaning of the entity, especially in longer sentences. The proposed EBAGCN solve the problem of complex sentence patterns and difficult syntactic understanding. By means of GCN based on the attention mechanism, the semantic and syntactic representation of the context is jointly trained. GCN effectively help integrate syntactic knowledge into vector representation, while attention mechanism highlights the expression of key information in dependency, which eliminates the syntactic noise to a certain extent.
Attention Visualization
This experiment provides a case study, using a motivating example that is correctly classified by EBAGCN but misclassified by Entity BERT, to vividly show the effectiveness of the proposed model. Given the sentence "He later went to report Malaysia for one year", people can easily distinguish "Malaysia" as a metonymic entity nominal by extending "Malaysia" as a concept of "a big event in Malaysia". Nevertheless, from the semantic perspective independently, the verb phrase "went to" is such a strong indicator that Entity BERT is prone to recognize "Malaysia" as a literal territory (for the regular usage of "went to someplace") falsely and overlooks the true predicate "report". How EBAGCN resolves the problems mentioned above is explained by visualizing the attention weights in the model.
First, the attention matrix of Transformer encoder blocks is compared in BERT Encoding Unit to display the syntactic integration's contribution to the whole model. Figure 7a,b shows that the weight for tokens in Entity BERT is more decentralized, while EBAGCN concentrates on "report" and "Malaysia" rather than "went to" thanks to the application of syntactic features, indicating that with the help of syntactic component, EBAGCN can better pick valid token and discard the irrelevant even misleading chunks. The Proposed Model section shows that the Attention Guided Layer transfers the hard dependency matrix into an attention guided matrix, which enables the syntactic component to select relevant syntactic constraints efficiently. Thus, this work further displays the attention guided matrix to demonstrate the superiority of soft dependency relations. As shown in Figure 7c, after launching the multi-head mechanism, despite the existence of dependency relations for the prepositional phrase "for one year", the weights for these relations are pretty futile compared with the main clause that includes verb and other determining features for judging relations in prepositional phrase useless to the MR task. This approach sets the model free from pre-defined pruning strategy and automatically obtains high-quality relations.
Findings and Limitations
The main findings of this paper are as follows: • integrating entity clues into the MR model better helps the model to represent an entity completely. • MR depends on syntax to clarify the sentence structure. Under these circumstances, GCN inputs syntactic knowledge into the model completely and efficiently. Furthermore, soft pruning strategy helps to remove the noise in syntactic dependency relations. • the proposed method can be generalized for different word-level NLP tasks, such as event extraction and relation extraction, indicating the great prospects of this research.
However, there are also some limitations need to be solved in the future: • dependency directions and types (such as nsubj, nmod) should be added to the model to make the model more robust. • the constraints mined in this paper all come from sentence. However, many metonymies are proper nouns that denote existing well-known works or events, suffering from the limitations of lacking "real world" and common sense knowledge to access the referents. External knowledge bases could be an appropriate choice to overcome this problem.
Conclusions
This paper proposed an entity-and syntax-aware approach to metonymy resolution based on pre-trained BERT, which substantially outperforms existing methods over a broad range of datasets. This work further demonstrated that the proposed end-to-end metonymy resolution model can improve the performance of more complicated and longer sentences. The experiments also evaluated the performance of different models in hard/soft dependency setting, and showed the proposed method to generalize the best. This constraint integration method can be applied to tasks beyond metonymy resolution. Numerous word-level classification tasks such as relation classification lack high-quality, balanced datasets. Thus, the proposed approach can be applied to these tasks to contribute to the research community. | 2022-03-16T15:20:35.771Z | 2022-03-12T00:00:00.000 | {
"year": 2022,
"sha1": "739d38295b35ee14efae2f8fa236945e566728ac",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1999-5903/14/3/85/pdf",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "9b2202efad380ef88b0f4aa592388cade054b818",
"s2fieldsofstudy": [
"Computer Science",
"Linguistics"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
16593618 | pes2o/s2orc | v3-fos-license | Germinoma with Involvement of Midline and Off-Midline Intracranial Structures
Germinomas are malignant intracranial germ tumors, usually found in suprasellar regions. Less than 10% are localized in off-middle structures, and synchronous involvement of both structures has only exceptionally been published. A case of an 18-year-old male patient with progressive right-sided hemiparesis and panhypopituitarism was reviewed. Brain MRI showed a solid mass involving pituitary and hypothalamus with thickening of pituitary stalk, high intensity lesions on T2-weighted imaging in left internal capsule, caudate nucleus, globus pallidus, and mild atrophy of the left internal capsule and cerebral peduncle. Nonadenomatous lesions were considered in the differential diagnosis. Alfa-fetoprotein (AFP) levels were negative in both serum and cerebrospinal fluid (CSF), while β-human chorionic gonadotrophin (β-HCG) levels were slightly increased in CSF. A transsphenoidal biopsy identified a germinoma. Four cycles of chemotherapy with bleomicine, etoposide, and cysplatin were given, followed by radiotherapy, but patients died due to a recidiva. Conclusion. Germinoma must be considered in patients with insipidus diabetes with a sellar mass with thickening of pituitary stalk; and ectopic germinoma must be suspected in patients with slowly progressive hemiparesis with cerebral hemiatrophy. Even with a rare condition, colocalization of midline and off-midline germinoma must be suspected in the presence of these typical signs of both localizations.
Introduction
Germ cell tumors (GCT) represent approximately 3% of neoplasms in children's cancer registries [1]. They constitute 0.1 to 2.4% of all childhood intracranial tumors in North America and Europe, while they account for almost 2.1 to 9.5% in Japan and the Far East [2,3]. Central nervous system germ cell tumors (CNSGCT) are rare and most of them occur in patients under 20 years of age [1,4].
CNSGCTs have been classified in "secreting" and "nonsecreting" tumors. Secreting tumors are defined as those presenting with an elevated CSF AFP ≥ 10 ng/mL or above the local laboratory's normal range and/or a CSF -HCG level ≥ 50 IU/l or greater than the accepted laboratory normal range. This has been shown to be related to prognosis and treatment response [1]. Brain germinomas are usually serologically negative for these markers [5].
The most common sites of involvement of intracranial germinomas are the pineal or suprasellar regions, while some patients have both localizations at the time of diagnosis [1,4]. Off-midline germinomas arising in the basal ganglia, thalami, and internal capsule, also called ectopic germinomas, are rare entities representing only 5 to 10% of all CNSGNC [4,6].
We describe an unusual case of a male patient with a germinoma with a synchronous involvement of midline and offmidline structures. Such a case has only been described twice to our knowledge [4,7].
Case Report
An 18-year-old male patient was admitted to the hospital with psychomotor excitement, polyuria, polydipsia, vomits, and a seven-month history of progressive right-sided hemiparesis with dystonia. He had poor school performance, anxiety, and 2 Case Reports in Endocrinology emotional lability for the last two years. Brain nonenhanced computed tomography done seven months before admission was normal. On physical examination, he was in a poor general condition, pale, with a low low body mass index (14,4 kg/m 2 ). Blood pressure was 90/60 mmHcg, with a poor response to fluid administration. He had a Tanner-stage 3, with 6 mL testis, pubic hair: G-2, and axillary hair: G-2. On neurological examination he presented a right hemiparesis with hyperreflexia and dystonia. The presence of polyuria with low urine density in association with hypernatremia suggested the diagnosis of diabetes insipidus (DI). The refractory arterial hypotension suggested adrenal insufficiency. Intravenous hydrocortisone was then started. Afterwards, levothyroxine and oral desmopressin acetate were added. Laboratory examination confirmed the diagnosis of hypopituitarism (Table 1). Clinical response was evident, with a dramatic improvement after hormonal substitution. However, normal natremia levels were difficult to achieve, and desmopressin dose was adjusted. Right-sided hemiparesis persisted, and a program of physical rehabilitation was promptly started. Vomits also persisted, although less frequently.
Brain magnetic resonance imaging (MRI) showed a solid mass with homogeneous enhancement after gadolinium, involving the pituitary and the hypothalamus with marked thickening of the pituitary stalk. The posterior pituitary hyperintensity was absent on T1-weighted images ( Figure 1). A high signal intensity lesion on T2-weighted images was evident in the left internal capsule, corona radiata, caudate nucleus, and globus pallidus bilaterally ( Figure 2). Mild atrophy of the left internal capsule and homolateral cerebral peduncle was also evident ( Figure 3).
With the suspicion of germinoma, a lumbar punction was indicated. Differential diagnosis with other nonadenomatous lesions was also considered: normal thoracic TC and angiotensin converting enzyme ruled out sarcoidosis; skull and long bones X-rays were normal, without typical lytic lesions of Langerhans cell histiocytosis. The CSF examination did not show atypical cells. AFP levels were negative in both serum and CSF, while HCG was slightly increased only in CSF. A transsphenoidal biopsy identified a pure germinoma. A spine MRI excluded metastatic lesions.
A treatment of four cycles of chemotherapy every three weeks with bleomicine, etoposide, and cysplatin was given.
An MRI study a month after chemotherapy showed a complete response with disappearance of the pituitary and suprasellar mass (Figure 4), while a reduction in off-midline white matter lesions was also evident. Three months later MRI had no changes, showing no evidence of tumoral recurrence.
After an improvement in his general condition, whole brain radiation therapy was indicated nine months after chemotherapy. Unfortunately, he died after a thalamic recurrence shortly after concluding this last treatment.
Discussion
Germinomas are the most common and least malignant intracranial germ tumors, usually found in the pineal and suprasellar regions. Five to 10% of the GCTs are ectopic, being localized in off-middle structures, like the thalamus, basal ganglia and internal capsule. Synchronous involvement of midline and off-midline structures, as described in this case, has only exceptionally been published [4,7]. Clinical presentation depends upon the size and the localization of the tumor. Patients with suprasellar GNC usually present hypothalamic-pituitary dysfunction, being DI one of the most common symptoms, isolated or in association with other hormone deficiencies. Ophthalmic abnormalities such as bilateral hemianopsia may also be present [4]. In our case hypopituitarism with DI was found. Suprasellar germinoma must be suspected in all young patients with isolated DI or in association with other pituitary deficits even when neurological and ophthalmological symptoms are absent [5]. Nevertheless, differential diagnosis with nonadenomatous inflammatory (infundibuloneurohypophysitis, sarcoidosis, and Wegener granulomatosis), neoplastic (Langerhans cell histyocitosis, craniopharyngiomas, metastases, leukemias, lymphomas, and brain tumors), or infectious (tuberculosis) lesions involving pituitary stalk must be considered [6,7]. It is important to emphasize that the presence of DI almost rules out a pituitary adenoma. Clinical manifestations of ectopic germinomas are insidious. Slowly progressive hemiparesis and neuropsychiatric symptoms such as dementia, psychosis, or cognitive decline with poor school performance are usually found; cognitive abnormalities are usually one of the earliest manifestations [4,[8][9][10][11][12][13][14]. In one of the most recently published series including 20 patients with basal ganglia or thalamic germinomas, all of them had hemiparesis at the time of diagnosis and 45% had cognitive decline [9]. Duration of clinical symptoms ranged from 1 month to 4.5 years, with a mean period of 1.5 years [11]. In accordance with the aforementioned, in our patient a progressive cognitive impairment and a seven-month history of progressive neurological deficit were evident.
Neuroimaging studies are useful in differential diagnoses. But measurement of serum and CSF tumor markers and/or histological studies are required for the confirmation of the diagnoses of germinoma. In this case the MRI showed a solid sellar and suprasellar mass with marked thickening of the pituitary stalk tumor (in accordance with panhypopituitarism), in addition to high signal intensity lesions in the left internal capsule, caudade nucleus, and both globus pallidus with atrophy of the left internal capsule and cerebral peduncle, confirming the synchronous involvement of midline and off-midline structures (in accordance with hemiparesis and 4 Case Reports in Endocrinology cognitive decline). In the early course of the disease, before the appearance of motor symptoms, MRI changes are not so evident. The earliest and most common feature on MRI is the atrophy of the basal ganglia as well as the presence of subtle signal intensity changes as hyperintensity on T1and T2-weighted images [13]. In our patient, the synchronous presence of the sellar mass was a clue for the presumptive diagnosis of ectopic germinoma. A tumor biopsy is required for the diagnosis, except in cases where tumor markers are elevated [1]. In our case, both tumor markers were negative in serum and -HCG was slightly increased in CSF. A transsphenoidal biopsy settled the diagnosis of a pure germinoma being negative for both -HCG and AFP, but positive for placental alkaline phosphatase (PLAP). However, biopsy-proven germinomas can have nongerminomatous elements among the unbiopsied sites and nonsecreting tumors can also have nongerminomatous components with a less favorable prognosis [14]. This was probably the case in our patient, considering his bad evolution in a short period of time.
Early diagnosis is very important because a delay in treatment can result in more severe neurologic deficits, as observed in our case. It has been demonstrated that, except for patients with small tumors, pituitary dysfunction before treatment persists or even worsens after tumor remission, mainly after radiotherapy. The earlier diagnosis and the prompt starting of treatment, before irreversible pituitaryhypothalamic damage occurs, contribute to improving the outcome of pituitary function in patients with neurohypophyseal germinomas [10]. In our case, even when only chemotherapy was given initially, panhypopituitarism and hemiparesis persisted after treatment.
The optimal management strategy for CNSGCTs remains unsettled due to a lack of prospective trials, mainly due to the infrequency of these tumors [14]. Germinomas are extremely sensitive to both irradiation and platinum-based chemotherapy but the recurrence rate after initial therapy may be approximately 10% or higher [1,[14][15][16][17]. Standard treatment for germinomas has been craniospinal irradiation (CSI) with survival rates of more than 90%. In order to avoid relapses, high dose radiotherapy delivered to the whole ventricle or a larger field is necessary. In an attempt to reduce the morbidity of CSI, cooperative groups had investigated the feasibility of a sequential treatment of chemotherapy followed by focal irradiation [15,16] or even chemotherapy alone. This last approach was tested by the International Central Nervous System Germ Cell Tumor Study Group, reporting a high rate of complete response but with a high rate of relapse [17]. However, response to irradiation after recurrence is usually very good [18,19]. Chemotherapy combined with reduced-dose radiation therapy has shown promising results in the tumor control [14][15][16][17][20][21][22][23][24]. But longer follow-up periods are necessary to draw firm conclusions regarding the superiority of this treatment over standard-dose [25], considering that a late recurrence is not a rare event [26].
In our patient, a treatment of four cycles of chemotherapy with bleomicine, etoposide, and cysplatin was chosen, with a very satisfactory initial response. Coadjuvant radiation therapy was administered nine months after, but he died because of a thalamic recurrence of the tumor.
In conclusion, sellar germinomas must be ruled out in all young patients with isolated DI or in association with other pituitary deficits. Ectopic germinoma must be suspected in patients with insidious neuropsychiatric symptoms and progressive hemiparesis, particularly if it is associated with subtle focal lesions in the basal ganglia and cerebral hemiatrophy. Though infrequent, involvement of both midline and off-midline structures may be present. A prompt diagnosis can avoid clinical sequelae and diminishes longterm impairment. The optimal therapeutic strategy for CNS-GCTs is not established yet, and an individualized approach is recommended. | 2018-04-03T05:56:05.661Z | 2014-02-09T00:00:00.000 | {
"year": 2014,
"sha1": "ee5c465d92261deb154caf037b27eee10e768976",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/crie/2014/936937.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "82f261fcafbb279efd2cbd4007671dc1e95c70b2",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
261164566 | pes2o/s2orc | v3-fos-license | Case Report: Missing zinc finger domains: hemophagocytic lymphohistiocytosis in a GATA2 deficiency patient triggered by non-tuberculous mycobacteriosis
Haploinsufficiency of GATA2, also known as GATA2 deficiency, leads to a wide spectrum of clinical manifestations. Here we described another 28-year-old man with a GATA2 variant who also suffered from hemophagocytic lymphohistiocytosis(HLH), who was finally diagnosed with HLH triggered by Mycobacterium avium bloodstream infection due to primary immunodeficiency. We reviewed GATA2 deficiency patients with HLH and found that GATA2 variants causing loss of zinc finger domains were associated with HLH, and erythema nodosa might be an accompanying symptom.
Introduction
Hemophagocytic lymphohistiocytosis (HLH) is a life-threatening syndrome due to immune responses of activated macrophages and lymphocytes, with common clinical features including fever, splenomegaly, cytopenia, elevated aminotransferase and ferritin levels (1).Driven by the aberrant immune response, HLH can occur at all ages.The products of HLH-related genes are involved in cell-mediated cytotoxicity and lymphocytes activation/survival, and immunocompromised patients are susceptible to HLH (1).As an important determinant of multilineage hematopoiesis, GATA-binding protein 2 (GATA2) is a member of the GATA family of transcription factors.
Caused by heterozygous loss-of-function GATA2 gene variants, haploinsufficiency of GATA2, also known as GATA2 deficiency, leads to a wide spectrum of clinical manifestations including mycobacterial infections, viral infections, bone marrow failure, leukemia and lymphedema (2).It has been reported that nontuberculous mycobacterial infections were the most common infections, while disseminated mycobacteriosis was rare in childhood but becomes more frequent with age, due to the diminishing hematopoietic function of the bone marrow (3)(4)(5).GATA2 deficiency has highly variable penetrance, and infections or other factors may affect epigenetic mechanisms to trigger pathogenesis and alter penetrance (6).
In 2021, our group reported a 17-year-old Chinese Han woman with a heterozygous GATA2 variant, who had recurrent HLH and erythema nodosa (7).Here we described another 28-year-old man with a GATA2 variant who also suffered from HLH triggered by non-tuberculous mycobacterial infection, and further functional evaluation was conducted.
Case description
A 28-year-old Chinese Han man was admitted with a history of intermittent fever for 3 years and erythema nodosa for 7 months.He had the first unprovoked onset of fever with a maximum temperature of 40°C lasting for two weeks.He had a second flare after two months and went to the local hospital, where he was diagnosed with hemophagocytosis and right inferior lobar pneumonia, based on bone marrow smear and chest computed tomography (CT).In 2021, he developed bilateral parotid enlargement, recurrent fever, and erythema nodosa.He was treated with various anti-infective agents, including moxifloxacin, linezolid, vancomycin, voriconazole, and cefoperazone-sulbactam, but the fever continued, mostly in the afternoon and at night.
The physical examination revealed the presence of papules on the trunk and upper extremities.His complete blood count indicated pancytopenia with WBC (0.6-1.3)×10 9 /L, neutrophils (0.3-0.8)×10 9 /L, lymphocytes 0.2×10 9 /L, monocytes 0.00×10 9 /L, hemoglobin (72-90)g/ L, and platelets (43-64)×10 9 /L.NK cells activity decreased (1.36%, normal range: ≥15.11%), and sCD25 was 6696pg/ml (normal range: <6400pg/ml).His CD107a expression in NK and CTL cells was within the normal range.His NK cell DPerforin was 72.56%, which was lower than the normal level.He also had elevated levels of IL-5 (76.2pg/ml, normal range: 0-17pg/ml), IFN-g (106.3pg/ml,normal range:0-95pg/ ml), ferritin (895ng/ml, normal range: 80-130ng/ml), C-reactive protein (CRP) (107mg/L, normal range: 0-8mg/L), and erythrocyte sedimentation rate (ESR) (75mm/h, normal range: 0-15mm/h).Antiphospholipid antibodies, antinuclear antibodies and antineutrophil cytoplasmic antibodies were all negative.Multiple lesions in the bilateral lung fields and multiple hypermetabolic lymphadenopathies in the hilum, mediastinum, and supraclavicular area and enlarged spleen were found by CT and PET/CT (Figure 1A).Mediastinum lymph node biopsy revealed a necrotic background with some plasma cells, eosinophils and lymphocytes infiltration.The pathological biopsy of subcutaneous nodules revealed lymphocyte infiltration scattered around the superficial small blood vessels of the dermis and collagen fiber proliferation.A bone marrow biopsy showed proliferation of erythroid series, predominantly of intermediate or late erythroblasts.In local hospital, the patient underwent blood cultures (including cultures for slowly-growing bacteria/mycobacteria), respiratory infection antigen IgM antibody, T-SPOT.TB, G test, GM test, Mycoplasma pneumoniae serological test, CMV-DNA and EBV-DNA, and the results were all negative.Bronchial brushing including Xpert, cryptococcal antigen, Aspergillus antigen, tuberculous smear, Fungal smear, TB-DNA, GM test were all negative.Next-generation sequencing (NGS) technology was conducted in BALF and detected Klebsiella pneumoniae.A diagnosis of HLH was made based on HLH-2004 criteria (8).He had no relevant family history.He had the heterozygous nonsense variant of the GATA2 gene c.599dupG, p.S201*.His father and sister carried the wild type of the GATA2 gene, and his mother died in an accident (Figure 1B).
Therapies, changes of his complete blood count, ferritin and lactate dehydrogenase during the hospitalization were shown in Figures 1C, D. His fever and erythema nodosa temporarily improved but relapsed after dexamethasone withdrawal.There are multiple pieces of evidence confirming the effectiveness of ruxolitinib combined with dexamethasone in the treatment of HLH (9, 10), therefore ruxolitinib was added in his treatment with dexamethasone.The blood culture at three weeks after admission revealed the presence of mycobacteria.He was finally diagnosed with HLH triggered by Mycobacterium avium bloodstream infection due to primary immunodeficiency caused by germline GATA2 deficiency.His symptoms did not improve despite 4-drug antituberculosis therapy, combined with dexamethasone and ruxolitinib.Bone marrow biopsy showed the hematopoietic tissue was decreased and myeloid/erythroid ratio was reduced, no typical tuberculosis granuloma or multinucleated giant cells observed.The last bone marrow smear revealed hypoplastic myelopoiesis, myeloid/erythroid ratio=0.63:1,intermediate or late erythroblasts ratio increased.Two weeks after discharge, he was hospitalized in the local hospital due to short of breath, and chest CT scan showed emerging ground glass opacities.He was given antibiotics but died of severe pulmonary infection one month later.
Discussion
GATA2 presents two highly conserved zinc finger domains, playing a critical role during erythroid maturation and hematopoietic development.The GATA2 variant in our patient creates stop codons that result in the absence of both two zinc finger domains (Figure 1E), which is the location of most pathogenic variants (6).Neither the proband's father nor his sister carried this variant, suggesting the GATA2 variant might be a de novo mutation.This mutation site has not been previously reported to be associated with GATA2 deficiency and HLH.
Current reports of HLH caused by GATA2 deficiency remain scarce, and we would venture to hypothesize that most variants of these HLH patients affected zinc finger domain function or resulted in loss of the zinc finger domain due to premature termination of translation, and the HLH onset was induced by the infections of bacteria or virus (7,(11)(12)(13)(14)(15)(16)(17) (Table 1).HLH occurs most commonly in infants and children, and some adults develop the disease may due to mutations with partial residual protein function that compensate for some immunological deficiencies (18).Most of the GATA2 deficiency patients presented with HLH in adulthood (11,12,(15)(16)(17), which can be explained by partial defect in GATA2 protein function.4 of 9 patients simultaneously experienced two zinc finger domain deletions all developed HLH after the age of 16, and were usually accompanied by rash, pancytopenia, splenomegaly and lymphadenopathy (7,12,16), further summary of clinical features requires the increase in the number of cases.Although it was not mentioned in other previous case reports, erythema nodosa appeared in two HLH patients with GATA2 deficiency identified as different infections in our center (7).As we know, erythema nodosa may indicate an underlying infection with non-tuberculous mycobacteria or fungi in GATA2 deficiency (4, 19), but the patient's nodule biopsy did not reveal evidence of infection.Hence, we assumed erythema nodosa as inflammatory reaction in GATA2 deficiency patients with HLH.Furthermore, it is known that NK cells develop and function under the influence of GATA2.While GATA2 haploinsufficiency leads to a specific loss of CD56 bright NK cells (20), lacking activity of NK cells of this patient be related to the absence of two zinc finger domains, with a reduced function of perforin release.The survival and differentiation potential of NK cells lacking perforin is higher (21), mutations in the perforin-coding gene cause familial HLH (22).The patient's phenotype may suggest that the defect in the GATA2 zinc finger domain and the function of perforin secretion in NK cells needs further investigation.It was also noted that among the four nonsense mutations in the case reports, three caused loss of both zinc finger domains in GATA2 (7,16), while one caused loss of C-terminal zinc-finger domain (17).It has been reported that 56% of GATA2 deficiency patients had lung involvement, and nontuberculous mycobacteria was the most common pathogen associated with chronic infection (23).The changes in bone marrow smear in the later stage suggest that HLH may eventually develop into MDS.Taken together, we suggest that primary immunodeficiency such as GATA2 deficiency especially patients who lack zinc finger domains should be considered in HLH patients, erythema nodosa might be an accompanying symptom, and opportunity infections should be highly suspected especially in those affecting cellular immunity (23).
RNA-seq was used to analyze gene expression in the proband's peripheral blood mononuclear cells (Figures 1F, G), revealing that four of the top ten differentially expressed genes SLC6A8, FAM210B, FHDC1 and GMPR were the targets genes of the GATA2 transcription factor in the datasets from the ENCODE Transcription Factor Targets dataset (24, 25), which shed light on the importance of GATA2 in his phenotype.Given that the patient had an HLH phenotype, we found that it had been reported SLC6A8 could mediate creatine to activate macrophages (26).Plasma levels of his inflammatory cytokines were significantly higher than those of the healthy controls (Figure 1H).Further experiments on GATA2 could not be performed because the patient was deceased.
In two cases from our center, GATA2 deficiency-associated infections were likely to be the occult cause of HLH (7).Neither patient received alloHSCT for various considerations, and respectively died at 1 and 6 months after discharge.It has been documented that alloHSCT may be a modality of a cure for GATA2 deficiency-related HLH, regardless of the presence of active infection (16).The various treatment modalities often resulted in a poor prognosis, suggesting that perhaps early alloHSCT is the treatment of choice.Therefore, having dealt with this severe disease, the only thing can be done is to save time.Early diagnosis and prompt treatment might improve the prognosis of these patients.
FIGURE 1 (A) Chest CT showing enlarged mediastinal lymph nodes (yellow arrow) and PET/CT with hypermetabolic mediastinal lymph nodes (red arrow).(B)Pedigree chart of the patient.(C) Trends in red blood cells (RBC), white blood cells (WBC), platelet count (PLT) levels and treatments during the patient's hospital stay.(D) Changes in ferritin (Fer) and lactate dehydrogenase (LDH) during the hospitalization.(E) GATA2 variants sites of two GATA2 deficiency-related HLH patients in our center (Protter, http://wlab.ethz.ch/protter).Heatmap (F) and volcano plot (G) for differentially expressed genes identified comparing patient and healthy controls (HCs) (volcano plot was performed using the OmicStudio tools at https://www.omicstudio.cn).(H) The comparisons of plasma levels of inflammatory cytokines between the patient and HCs.**P ≤ 0.01; *P ≤ 0.05.
TABLE 1
Summary of GATA2 deficiency patients with HLH. | 2023-08-26T15:07:58.933Z | 2023-08-23T00:00:00.000 | {
"year": 2023,
"sha1": "c45522e49ba91a6306b03a8ab1472bb73948ac66",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fimmu.2023.1191757/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "6df3e9ff7f7a2cd04bcc093782e960d06a92d310",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
55341454 | pes2o/s2orc | v3-fos-license | Review of Ph.D Studies in Humanities: Updating New Horizons in Education
The context of information society, technological development and new societal challenges demands new frameworks for critical analysis of the knowledge production as well as for the geopolitical democratization of their distribution. In our article “Cultural processes, social change and new horizons in education,” published in Procedia. Social and Behavioral Sciences, 174, we presented the educational development work done in the Doctorate in Creation and Culture Studies of the Universidad de las Américas Puebla (UDLAP), Mexico, in order to face the contemporary demands for understanding and studying processes of knowledge construction, transmission and diffusion beyond the academic standards of use and epistemologies of text. This focus has been especially important for the program, as it is based on the collaboration on the transdisciplinary basis between Arts and Humanities and Social Sciences, having some extensions to ITC engineering and Natural Sciences through its research projects and creative practices. This text is a review of the above-mentioned article but also an updating of that program presentation due to the recent curriculum renovation and with it, integration of new research foci.
Introduction
One of the characteristics of the new Ph.D. programs in humanities has been their trend to widen the Eurocentric shallow sense of history, seeking for knowledges sprung from the unseen roots of the Others forgotten by the historical amnesia of the Western universities [1,2]. The response to that has been the emergence of programs with 'disobedient' epistemic frameworks; the colonial axis of power-knowledge-culture, which had reduced entire people to objects of study, and others to hegemonic producers of knowledge, had now to be re-valued by themselves by socialcultural struggle and by epistemological and aesthetic innovation [3].
In the case of high level research, knowledge production and doctoral education, challenges we are continuously facing are complex, beginning with the need to understand the social tendencies of the contexts in which we are working as well as the usability and impact of the produced knowledge in the contemporary societies and how new forms of knowledge production should be inserted to institutionalized structures such as universities. Thus, in order to contribute to the alternative practices in knowledge production in communities of emergent economies such as Mexico, since 2006 the Universidad de las Américas Puebla (UDLAP) has offered a transdisciplinary doctorate program in Creation and Culture Studies with the participation of a wide range of researchers from art, architecture and urban studies to anthropology, political sciences and psychology. This recently revised and re-accredited program by the National Council for Science and Technology, CONACYT, has maintained its unusual pedagogical focus through the insertion of students to institutional research projects and scaffolding their academic performance through research groups thus consolidating the basic goals of collaboration of the program: Unusual pedagogical approach through this tutorial system provides scaffolding by faculty together with a whole group of associated researchers and fellow students, detonate a high level, trans-disciplinary research in order to detect, explore and potentiate alternative knowledges as vehicles to lead to a social change and to a socio-culturally sustainable development,
Review of Ph.D Studies in Humanities: Updating New Horizons in Education Abstract
The context of information society, technological development and new societal challenges demands new frameworks for critical analysis of the knowledge production as well as for the geopolitical democratization of their distribution. In our article "Cultural processes, social change and new horizons in education," published in Procedia. Social and Behavioral Sciences, 174, we presented the educational development work done in the Doctorate in Creation and Culture Studies of the Universidad de las Américas Puebla (UDLAP), Mexico, in order to face the contemporary demands for understanding and studying processes of knowledge construction, transmission and diffusion beyond the academic standards of use and epistemologies of text. This focus has been especially important for the program, as it is based on the collaboration on the transdisciplinary basis between Arts and Humanities and Social Sciences, having some extensions to ITC engineering and Natural Sciences through its research projects and creative practices. This text is a review of the above-mentioned article but also an updating of that program presentation due to the recent curriculum renovation and with it, integration of new research foci.
Keywords: Psychology; Political sciences; Anthropology; Humanities; Natural sciences action, thinking and feeling not known before; the wideness and velocity of the cybernetic web, the scope of technology of mass media and the human environment manipulated by commercial strategies and the recent collapse of paradigms that until today had given form to the individual and to the collective creativity and innovation, demands for a formation of researchers and creators who are able to respond and give consistence to our time proposing spaces for cultural dialogue and critique and new forms for creative knowledge production and transmission [6].
Program Updates and Renewed Research Lines
To reach these goals, in Arts and Humanities research and education it has been especially critical to be able to define what 'viability' and 'usability' of knowledge mean. Though, in Humanities as in other areas related to creative processes with non-rational foci, the criteria of legitimation of knowledge is changing, thanks to the transdisciplinary collaboration. The study of different experiences of alternative knowledge production and new creative practices related to spatial and sensorial interventions, performances, experimentation with materials, has been managed in the Doctorate in Creation and Cultural Studies and in the institutional research project 'Epistemologies beyond the Text. Cultural Practices in the Information Age' linked to it through research lines 'Expanded Scriptures' focused mainly on ITC and new media in art, culture and society, and 'Subaltern Knowledges' focused on the study of the subaltern Others on the postcolonial basis. As examples the results, some dissertation projects can be mentioned, such as 'The Theatre of the Oppressed: An Aesthetic Practice Where the Subject Becomes Visible' [7], in which Theatre of the Oppressed (TO) of Augusto Boal [8] was used as the main methodology applied to explore through theatre and in creative ways some daily expressions of patriarchate, homo/lesbophobia, employment discrimination and even power relations existing in prison between internals beyond the control of the wards; 'Which Reality? Interactions at the Southern Border of Mexico, as a Projection of an Artistic Creation Practice' [9], aimed to be at the same time a creation process and an analysis of an artistic practice as a vehicle for generating and understanding knowledge [10] based on the diversity of cultural and social manifestations along the southern border of Mexico, where the border line not only marks a geopolitical limit between two nation-states (Mexico and Guatemala), but also emphasizes their differences and all they have in common through a palpable repertoires of mobility of people, things and events [11], and 'Behind the Walls: Democracy, Political and Cultural Management in Mexico and Ecuador 2007-2013' [12] exploring cultural management, democratization, cultural policy and cultural democracy [13] as concepts that shape practices of street art and graffiti as non-hegemonic experiences in institutional spaces and in which the author suggests that both street art and graffiti should be considered beyond their traditional location as part of youth culture and their relation with gangs and urban criminality [1].
In 2016-2017, due to the request of the National Council for Science and Technology, CONACYT, and its National Postgraduate as detonators of cultural practices and knowledge production beyond the Occidental canon [1].
Transdisciplinarity and Education for Development in Ph.D. Studies in Humanities
In the academic world, the intellectual authority has been evidenced through high quality research validated by institutionalized systems and practices as gatekeepers for the knowledge production, through different categories of academically validated products. Results of the research work should though also have an impact beyond the academic circles, in wider social contexts through concrete indicators of their viability and usability as knowledge. In the ideal case, these external indicators should match the consensus reached during the critical disciplinary debate inside the academic circles [4].
In order to have an effective focus on the current societal problems and demands for development in emergent economies, the trans-disciplinary institutional research project accompanying the Ph.D. program has constructed its line of thought and that of knowledge production, on the postcolonial and postmodern socio-cultural basis aimed to nurture the postgraduate program. The institutional research project itself, to which the dissertation work of each student has been inserted and through which it has been constantly monitored, has resulted an innovative pedagogic mechanism through which students have been involved in active research work guided by an academic tutor and thus effectively introduced to research methodologies and theories straight from the beginning [1].
In the program framework, the student is not only a qualified researcher, but also a social innovator, creator and experimenter, being an interpreter and transmitter of socially constructed knowledge gathered through the interaction between students and human communities. The cooperation between researchers, creators and communities in terms of recollection, analysis and study of alternative knowledge production requires a special kind of creativity and sensibility able to capture meanings and individual and collective manifestations as these present forms beyond the canonical validation of knowledge by cultural norms. Thus, these hybrid, sensuous and critically reflective cultural forms demand new, innovative research initiatives that challenge the current stiff university system demanding substantial changes in the traditional role of research and education in humanities; for example, Belonging Bologna-seminar report [5] emphasizes that art, design and culture education and research should always be connected to and reflecting the events of the outside world [1].
Taking into account the above-mentioned observations, the Ph.D. program in Creation and Culture Theories proposes to introduce its students to the incredible whirl of social, political, scientific and artistic events worldwide considering that the velocity of these events has exceeded our capacity to elaborate responsible theoretic reflection, to understand their origin and causes and their impact and future consequences in societies. Thus, the analysis of cultural production has been exceeded by modes of
ACTA PSYCHOPATHOLOGICA ISSN 2469-6676
Programs of Excellence, the Ph.D. program was critically analyzed and renovated to respond to the current conditions in the area of Latin American doctoral education. Thus, the program is defining itself now as a program immersed in current lines of thought worldwide in order to be an authentic school for contemporary thinking and as a leader in the formation of new generations of thinkers, creators, promoters, analysts and opinion leaders able to understand and enhance the present conditions having the multiple and heterogeneous areas as the platform for the cultural production [6].
The renovated and reinforced program is aiming at educating researchers and creators able to trigger cultural and societal innovations that permit to rethink contemporary culture through critical creative practices and knowledge production with a guaranteed influence on diverse socio-cultural environments through interdisciplinary dialogue in which socio-cultural knowledge is used to analyze objects of study from different perspectives [6]. In order to follow the past successful framework of an institutional research project as an all covering umbrella for all the studies of the program, a new research framework initiative is under work. The proposed topic for this projects is defined as "Technologies of Knowledge, Globalization and New Cultural Practices,' exploring expanded epistemologies with in order to study transformations in the knowledge formats, their usability and viability in the globalizing contexts and the specific conditions in which these transformations are manifesting themselves and diffused worldwide due to the demand for constructing, transmitting but also for manipulating, possessing and controlling new knowledge and creative and innovative practices. Graduate studies impact several disciplines to improve the environment and influence a change in the society in which they develop. Researchers, professors and graduates develop new ideas on artistic and cultural approaches: graffiti as a cultural form of resistance [12] and the philosophical categories of insignificant or the non-interrupting strategy of repetition and passivity that reveals new approaches on cultural, social and political resistances [14]. Others like posthegemony and everyday life are also forms of cultural resistance that help draw the necessary social changes [15].
The renovated program and this new initiative is divided into three research lines and research groups denominated 'Art, technology and knowledge' focusing on the study of the current art theory that involves the creator and art researcher to discussions of their socio-cultural environment impacted by technological practices of the knowledge production. The line 'Globalizations, governability and mobility' explores the knowledge production in terms of the analysis of forms of governability manipulating us but also manipulated by us. Through these framework topics such as Nation-state, global cities, bio-and necro-politics, open government and 'fluent planning' may be explored. And finally, the line 'Intersectionalities: gender, bodies and spaces' propose to analyze the complexity of the world and of the human experience focusing on the study of discrimination, injustice and identities. Covered by these three frameworks, new dissertation projects have been or are being developed. On the other hand, migratory movements from Latin America to the United States and Canada have triggered interesting transcultural situations in which local expressions and identities have been transported to new places beyond nation borders impacting the culture of places of destination of the migrants, as has happened in the case of 'chicanos', Mexican origin people in the United States. The Ph.D. dissertation project entitled 'Mexican Digital Diaspora: Transnationalism, ICT and Social Networks in North America' [16], explores new perspectives and paradigms about transnationalism within the Mexican migration context. Taking into account transnationalism that consists of those processes forged by migrants through which they maintain simultaneous and complex social relations between their locations of origin and their current place of living [17], the project explores the migrant users' performativity in Internet and particularly the increasing use of Social Networks (like Facebook or Twitter). Specifically, in the area of health, this epistemological vision influence in modifying the natural history of the disease in various pathologies that afflict Third World countries. For example: "Illegality", Flexible Accumulation and Health: Return Migration of Sick, Exhausted and Dying Workers; Society for Applied Anthropology is one of the approaches that allow us to recommend migratory public policies that affect a better quality of life and socio-political development [18].
The new hybrid cultures do not only emerge in the new living places of the migrants but also have a transformative effect in their communities of origin, triggering cultural and social transformations, in ways of life and popular culture. We should also mention the dissertation project 'The Burdens of Chastity' [19], studying the clashes between the LGBT communities and the society in Mexico and projects still under work such as 'Biopolitics and Dirty War in Mexico: Construction of Precarious Subjectivities' about the political conflict between the government and diverse resistance movements during the socio-political turbulences in Mexico in 1960's, when disappearances and political murders were the vehicle for the government to maintain under control the anti-governmental movements impacting the subjectivities of the period [20,21].
Conclusion
It is important to continue highlighting our intention to follow what Water Mignolo has pointed out about the knowledge production and transmission of knowledge as vehicles in the peace making and construction of identity in the postcolonial Latin America as a geopolitical project. Thus, the contemporary higher education and research carried out in the emerging economies of the previous colonies such as Mexico should be able to develop an amplified potential to produce updated knowledge beyond traditional institutions of knowledge construction and epistemological production. Countries like Mexico, aspiring of occupying a much stronger and visible participation to the international politics and culture, needs to trigger a societal change in its own territory in order to solve its great inequalities. In this task all the creative, cultural, historical and human resources should be used including research and project development. As we have demonstrated, new kind of research projects in which different disciplines have joined their forces in order to explore a wide range of problematic of our globalizing world on the local level, can go beyond their accustomed limits in order to impact local conditions through locally adapted knowledge production. It is important to create solutions and strategical models for the transformation of social realities in conflictive places through recognizable languages and meanings, be they textual or nontextual, in order to reformulate the socio-cultural identities transforming them to detonators of the societal change through culturally sustainable processes. Thus, transdisciplinary research and postgraduate education produces researchers, creators, teachers and socio-cultural promoters dedicated to the study of how the globalization, technology and cultural processes define the knowledge production processes as well as the character of our postmodern knowledge and of the institutions and institutional connections and networks that manipulate and operate it.
Retaking once again the words of Walter Mignolo, as in our article in Procedia, Social and Behavioral Sciences, we continue reaffirming: And I say the 'humanities,' and not just 'the humanists' […] Since all the knowledge and understanding is human understanding (from genomics to dance, from electric engineering to literature, from mathematical models in economy to political economy), every scholar, academic, and scientist has a responsibility towards the humanities; in other words, he or she has critical, ethical, and political responsibilities in the production, dissemination, transformation, and enactment of knowledge. The humanities can no longer afford to be what they have been for the past sixty years: a 'complement' to the 'efficiency' of 'serious' technological knowledge that guarantees a constant progress of humanity as a whole and a sublime 'enrichment' of human beings as Human Being. | 2019-06-13T13:17:16.881Z | 2018-01-01T00:00:00.000 | {
"year": 2018,
"sha1": "cd94d73896df319e11ca1a0746e5fd6c3dfca5d8",
"oa_license": "CCBY",
"oa_url": "http://psychopathology.imedpub.com/review-of-phd-studies-in-humanities-updating-new-horizons-in-education.pdf",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "afe6a64cc801fefd9d5adde4bbe0c593abb462d5",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": []
} |
238215330 | pes2o/s2orc | v3-fos-license | Optical excitations of Skyrmions, knotted solitons, and defects in atoms
Analogies between non-trivial topologies of matter and light have inspired numerous studies, including defect formation in structured light and topological photonic band-structures. Three-dimensional topological objects of localized particle-like nature attract broad interest across discipline boundaries from elementary particle physics and cosmology to condensed matter physics. Here we show how simple structured light beams can be transformed into optical excitations of atoms with considerably more complex topologies representing three-dimensional particle-like Skyrmions. This construction can also be described in terms of linked Hopf maps, analogous to knotted solitons of the Skyrme-Faddeev model. We identify the transverse polarization density current as the effective magnetic gauge potential for the Chern-Simons helicity term. While we prepare simpler two-dimensional baby-Skyrmions and singular defects using the traditional Stokes vectors on the Poincar\'e sphere for light, particle-like topologies can only be achieved in the full optical hypersphere description that no longer discards the variation of the total electromagnetic phase of vibration.
I. INTRODUCTION
Topologically non-trivial defects, textures, and knots have inspired physicists since the days of Kelvin [1]. They are remarkably ubiquitous throughout physics, spanning a vast range of energy scales from cosmology and elementary particle physics, to superconductors, superfluidity, and liquid crystals. The universal nature of topological stability in such diverse areas provides unprecedented opportunities to use experimentally accessible laboratory systems as emulators even of cosmology and high-energy physics where the experimental evidence is absent [2]. In recent years, the experimental study of topological defects and textures in structured optical fields has emerged as one of the most promising areas to engineer and detect topologically non-trivial characteristics [3], including singularities of the phase or polarization that may form knotted or linked geometries [4][5][6][7] or Möbius strips [8]. Another line of research on non-trivial topologies of light has focused on photonic band structures [9], analogous to electronic band structures in crystals.
For the particular case of baby-Skyrmions in optical fields, field profiles are usually analyzed using the Stokes vector, i.e., a point on the Poincaré sphere, corresponding to the coherent, transverse polarization state at each point in the field. However, going beyond these more easily observable parameters, the full topology of the field configurations, crucially, also depends on the spatial variation of the total phase of vibration on the polarization ellipse [55], which is the sum of the phases of the electric field components, and is not represented by the Stokes vector. This complete topology is then described by the optical hypersphere S 3 (unit sphere in 4D) [33], allowing, e.g., for full 3D particle-like topologies of light.
Here we utilize simple configurations of structured light fields to show how these can lead to optical excitations in atomic media of comparable or considerably more complex topologies. Baby-Skyrmions, represented by full Poincaré beams [56] in light fields, can straightforwardly be transferred to optical excitations, and therefore frozen and stored, in strongly confined oblate atomic ensembles. We consider a J = 0 → J = 1 transition that can form, e.g., in 88 Sr very long-lived excitations. By going beyond the Stokes representation of light beams to incorporate the full degrees of freedom of the field amplitudes, where we no longer discard the spatial variation of the sum of the phases for the two field components, we can form 3D particle-like Skyrmions, localized in space. We identify the transverse polarization density of the atoms as a synthetic magnetic vector potential of the 3D Skyrmions with non-trivial helicity. While constructing such an object directly in a light beam is quite challenging even for modern structured light engineering [33], we show how appropriately adjusting the lightmatter coupling provides a solution with simple copropagating beams. For this solution, we then formulate the Stokes representation to provide precisely a Hopf fibration between the optical hypersphere and the Poincaré sphere, representing knotted solitons or Hopfions, analogous to the knotted solitons in the Skyrme-Faddeev model [11,25], and show the linked and trefoil knot Hopfion preimages of the Poincaré sphere. While such objects are non-singular, we also show how singular defects can be transferred from light to optical excitations. For systems where the light scattering is strong, and light mediates dipole-dipole interactions between the atoms, we remarkably find that singular defects can even exist as collective excitation eigenmodes. These behave as spatially delocalized 'superatoms', exhibiting their own collective resonance linewidth and line shift.
II. BABY-SKYRMIONS
We first show how to prepare 2D baby-Skyrmions in an atomic ensemble. A non-singular topological texture can be constructed by letting a (pseudo-)spin orient into a localized structure that points in every direction somewhere within a 2D plane, but takes a uniform constant value everywhere sufficiently far away from the origin, independently of the direction. The plane can then be compactified to a unit sphere S 2 and the orientations of the spin on the 2D plane can be characterized by S 2 → S 2 mappings. Such mappings can take topologically non-trivial values, associated with the existence of baby-Skyrmions, also frequently called non-singular vortices.
For optical fields, the state is most commonly characterized on the S 2 Poincaré sphere by an easily observable Stokes vector S [55,57], and the S 2 → S 2 mapping defining the baby-Skyrmion topology counts the number of times the object wraps over S 2 , where ijk denotes a completely antisymmetric Levi-Civita tensor. A field configuration that satisfies a nontrivial winding W = 1 can be achieved using a superposition of a Gaussian and Laguerre-Gaussian (LG) beam, with wavevector k and frequency ω = c|k| = ck. Working with slowly-varying amplitudes for the light and atoms by factoring out the fast-rotating term exp(−iωt), the positive frequency component of the field, E(r), is given by Here U l,p (w 0 ) are the LG modes with azimuthal quantum number l, radial quantum number p, and focused beam width w 0 [3]. The light field of Eq. (2), now a full Poincaré beam [56], contains a Néel type baby-Skyrmion whose optical polarization we have defined here in the linearê x,y basis, instead of the commonly used circular basis [48,52,54], because the linear basis is physically relevant when manipulating the atomic transition, as discussed in Sec. III A.
Topologically non-trivial fields in this simple example can be straightforwardly transformed to optical excitations in atomic ensembles. We consider a |J = 0, m = 0 → |J = 1, m = υ transition which can in alkaline-earth-metal-like atoms be very narrow, forming long-lived excitations. For instance, the 88 Sr clock transition 1 S 0 → 3 P 0 has a linewidth controllable by a magnetic field, with the transition entirely forbidden at the zero field. We create a non-singular topological texture of the optical excitation by considering an oblate ensemble of atoms, strongly confined along the light propagation direction (z axis). We write the optical excitation as an electric polarization density, or the density of electromagnetic vibration in atoms, with the slowly-varying positive frequency component υ on atom j, located at r j , is given in terms of the reduced dipole matrix element D and the excitation amplitudes P (j) υ , with the unit vectorsê ± = ∓(ê x ± iê y )/ √ 2 andê 0 =ê z . Light couples to P via the atomic polarizability, α = −D 2 /[ 0 (∆ + iγ)], according to P(r) = 0 αE(r), where γ denotes the resonance linewidth of the atom and ∆ is the detuning of the laser frequency from the atomic resonance. The excitation is then re-emitted back to the light field, where the scattered light amplitude is given by d 3 r G(r − r )P(r ) and G(r)d denotes the dipole radiation at r from an oscillating dipole d at the origin [58].
For describing the topology of the optical excitation, we define a pseudo-spinor in terms of the normalized transverse atomic polarization densities, where, as for the light field in Eq. (2), we work in a linear rather than circular basis, withP j = P j /|P| and the longitudinal component P z = 0. We can then define the corresponding atomic Stokes vector where σ j are the Pauli matrices. In Fig. 1, we show the baby-Skyrmion configuration generated by the field in Eq. (2). The atomic Stokes vector, Eq. (4), now has a fountain-like structure, , and takes a uniform value S = (0, 0, −1) sufficiently far away from the center of the object. It is easy to verify that the winding number Eq. (1) for S integrates to W = 1, and that the same topological structure in the incident field is excited in the atomic polarization density. The principle of creating a baby-Skyrmion is therefore closely related to the studies of analogous objects in exciton-polariton systems [47,48].
A. 3D Skyrmions
We now show how 3D Skyrmionic structures can be constructed by considering the full complex nature of the electric polarization. The Stokes vector representation of the Poincaré sphere for the light field amplitudes or optical excitations in atoms [Eq. (4)] does not provide the full field description, as the texture may also exhibit nontrivial, non-uniform spatial variation of the total phase of the two field components, which is discarded. A more complete description of the field topology can instead be obtained using the optical hypersphere S 3 [33].
The field parametrization in S 3 permits considerably more complex, particle-like objects, localized in 3D physical space. Compactifying the real 3D space, such that the fields are assumed to take the same value far away from the particle, independently of the direction, allows us to describe the topology by S 3 → S 3 mappings. Such mappings can be characterized by distinct topological equivalence classes, identified by the third homotopy group elements Π 3 (S 3 ) = Z. Non-trivial objects whose S 3 mappings wrap over the order parameter space an integer number of times represent topologically non-trivial solu-tions, originally introduced by Skyrme [10].
We now parametrize the atomic polarization spinor on the S 3 optical hypersphere by writing it as a fourcomponent unit vectorn = (n 1 , n 2 , n 3 , n 4 ), and takinĝ wheren is represented by the hyperspherical angles 0 < ψ, β ≤ π and 0 < η ≤ 2π. The integer topological charge of the 3D Skyrmion (known in high-energy physics as the baryon number [14]), is found then by counting the number of timesn wraps over S 3 , and is therefore analogous to the linking number density in (super)fluids [16], where J is replaced by the (super)fluid velocity, and to the Chern-Simons term for the magnetic helicity [59], in which case J represents the gauge potential for the magnetic field (Note that the sign of the winding numbers may vary depending on the orientations of the coordinates and the mappings). To understand the structure of the Skyrmion in Eq. (5), we consider a simple analytic mapping from 3D Euclidean real space to the optical hypersphere [22] with η = pφ, β = θ and ψ = qς(r), finding that Eq. (6) integrates to give a topological charge B = pq, where the monotonic function ς(r) satisfies ς(0) = 0 and ς → π sufficiently far from the origin. The first spinor component vanishes along the z axis, and now forms a multiply-quantized vortex line with a winding number p. The second component vanishes at the circles θ = π/2, r = ς −1 [(n − 1/2)π/q] for n = 1, . . . , q, withP y ∼ −δr − iδθ in the circle vicinity, and hence forms q concentric vortex rings with different radii. The vortex line threads the vortex rings, and has a non-vanishing density confined inside the toroidal regions around the vortex ring singularities, such that the Skyrmion is spatially localized, forming a particlelike object. Any continuous deformation of Eq. (5) conserves the discrete topological charge; a 3D Skyrmion with B = pq can also be constructed by taking any combination of singly-and multiply-quantized lines (rings) with total winding q (p), located in the components of P x (P y ), where the lines thread through the rings. Forming such a structure in the polarization density using electromagnetic fields in free space alone is a rather challenging task of structured light engineering [33]. However, we can here exploit the properties of the light-matter coupling to simplify the field profiles considerably. To create the Skyrmion, we take a coherent superposition of copropagating light beams where for the LG beam we now choose l = 1 to form a B = 1 Skyrmion, although we consider higher-order charges in the next section. For the Gaussian beams of unequal focusing, the parameter c = exp(−ρ 2 0 /w 2 1 + ρ 2 0 /w 2 2 ) defines the circular radius ρ 0 in the z = 0 plane of minimum focusing at which they interfere destructively. Destructive interference outside the ring is prevented due to diffraction. Diffraction also leads to variation of the phase (Gouy phase), such that U 0,0 (w 1 ) − cU 0,0 (w 2 ) ∼ (ρ − ρ 0 ) + iζz in the zero field ring vicinity. The ypolarized light component now forms a singular vortex ring [60] with a 2π phase winding, analogously toP y of Eq. (5) for q = −1, and a vortex core anisotropy The x-polarized light component exhibits a singular vortex line, analogously toP x of Eq. (5) for p = −1, where the LG beam has an intensity that reaches its maximum in the z = 0 plane at ρ = w x / √ 2, coinciding with the vortex ring singularity. However, the intensity is not confined along the z direction as required by the Skyrmion solution Eq. (5). In order to achieve the desired profile, we can utilize the light-matter coupling, which can be selectively turned on around the z = 0 plane only, to confine P x . This can be achieved by controlling the m = 0 quadratic Zeeman level shift, either by magnetic fields, or ac Stark shifts of lasers or microwaves [61].
In Fig. 2(a), we show the topological charge density B for the 3D Skyrmion constructed using the field in Eq. (8), [where we have ignored any contribution from the beam phase factor exp(ikz)], and the confinement of P x , achieved using spatially dependent level shifts ∆ x (r) = δ[1 − exp(−z 2 /10w 2 x )] in P(r) = 0 α(∆ υ )E(r). We consider long-lived excitations with extremely narrow linewidth, so typically δ γ, and we take δ/γ = 200. The topological charge density shows the localization of the Skyrmion, with the density concentrated at the origin, and also in two rings where the gradient of P x and P y becomes large from the applied level shifts and vortex ring phase winding, respectively. Changing the vortex ring core anisotropy, Eq. (9), which has the value ζ = 0.08 in Fig. 2(a), increases the concentration in the rings for a more anisotropic core. We find the corresponding transverse polarization density current J [ Fig. 2(b)], which represents the synthetic magnetic vector potential with an integer linking number, has a large magnitude where the charge density is highly concentrated. At the charge density rings, J flows radially inwards or outwards, while closer to the origin, J flows almost entirely along the ±x directions.
B. Knotted solitons
We have shown how 3D particle-like objects can be prepared by going beyond the Stokes vector representation used to describe baby-Skyrmions in Sec. II and parametrizing the optical excitations on S 3 . However, we can also construct particle-like 3D objects using the Poincaré sphere, instead of the full optical hypersphere. The advantages of our choice of representation for the optical hypersphere in Eq. (5) become apparent when we formulate the S 3 → S 2 transformation from the optical hypersphere to the Stokes vector precisely as a Hopf fibration [33,62,63].
The Hopf fibration, initially of purely mathematical interest, arises naturally in field theories. In the Skyrme-Faddeev model, 3D topological objects known as Hopfions are classified by an integer-valued Hopf charge [25][26][27][28][29]. Considerable interest in these systems was generated by the observations that the stable solutions may exhibit knots. The Hopf map of the vectorn on S 3 to a vector h = (h 1 , h 2 , h 3 ) on S 2 is given by where the mapping falls into distinct topological equivalence classes Π 3 (S 2 ) = Z, characterized by the integer Hopf charge, Q H . Upon substituting the expressions forn in terms ofP, the mapping indeed returns the atomic Stokes vector, Eq. (4). Applying the Hopf map of Eq. (10) to the B = 1 Skyrmion of Fig. 2(a), we obtain a Hopfion with charge Q H = 1, shown in Fig. 2(c) by the field profileĥ. The particle-like nature of the Hopfion is clearly visible, where the full 3D spin texture is localized around the origin. At large distances in any direction from the center, and along the vortex line whereP x vanishes, we haveĥ = (0, 0, −1), while at the vortex ring withP y = 0,ĥ = (0, 0, 1). The topological structure of the Hopfion is revealed when considering the reduction in dimensionality of the parameter space under the Hopf map of Eq. (10), where multiple points on S 3 map to the same point on S 2 . These points form closed curves in real space, known as Hopfion preimages, which interlink an integer number of times, as the preimages of the Hopfion introduced in Fig. 2(c) show in Fig. 3(a). The linking number is given by the Hopf charge, Q H , which can be shown [64] to be equal to the 3D Skyrmion charge, Eq. (6). Therefore, we can increase the preimage interlinking by increasing the total winding of vortex rings and lines, as discussed in Sec. III A. Multiply-quantized vortex lines in P x can easily be prepared by changing the beam orbital angular momentum in Eq. (8). Choosing l = 2, we form a Q H = 2 Hopfion with real space preimages that interlink twice, as shown in Fig. 3(b).
Here we show how we can even prepare Hopfions that have the highly sought-after knotted structure, provided that the total winding of vortex rings is increased. This is more complicated than preparing higher quantized vortex lines, not least because multiply-quantized optical vortex rings are forbidden in paraxial light beams [65]. However, we overcome this limitation by two alternative strategies. The first is to create the Hopfion using the field in Eq. (8), but where the y-polarized light component is now chosen to drive a two-photon transition, with the beam wavelength doubled. Each photon excites a single vortex ring, such that P y ∝ [U 0,0 (w 1 )−cU 0,0 (w 2 )] 2 , therefore forming a doubly-quantized ring [60], with P y ∼ [(ρ − ρ 0 ) + iζz] 2 in the ring vicinity. Using then l = 3 for the LG beam in Eq. (8) to prepare a triply-quantized vortex line that threads the vortex ring, we create a Hopfion that has a real space preimage of a trefoil knot, Fig. 3(c). An alternative method is to choose the y-polarized light component in Eq. (8) to be structured according to the techniques of Refs. [4,65], where using the superposition of LG modes, gives a configuration of four coaxial vortex rings (two oppositely winding in the focal plane, one above and one below the focal plane), through which the triply-quantized vortex line threads and creates a Hopfion with a trefoil knot preimage, Fig. 3(d).
IV. SINGULAR DEFECTS
Until now, we have considered non-singular topological textures where the orientation of the spin is well defined everywhere in space. We now show how it is also possible to form singular defects for which the spinor becomes ill-defined at a finite number of points. Structured light fields that exhibit singularities can create singular defects in atomic optical excitations by analogous principles to non-singular textures. To form 2D optical point defects in oblate atomic ensembles strongly confined along the zdirection, we consider a full Poincaré beam profile formed by a superposition of two LG beams with opposite orbital angular momenta [48], where ϕ = 0 (ϕ = π) results in an azimuthal (radial) singular vortex in the light field.
For a dominant incident field, the singular configuration can be transferred onto the atomic polarization density without the need for any applied level shifts, as in Sec. II. Such a configuration eventually radiates at the single-atom decay rate. However, remarkably, we find that specific defect structures can be highly robust and stable even in the strongly interacting limit where the incident field is no longer dominant in the atomic ensemble. These structures therefore represent spatially delocalized coherent 'superatoms' that extend over the sample. In a cold and dense atomic ensemble, resonant incident light can scatter strongly, mediating dipole-dipole interactions between the atoms. In Fig. 4, we show the real components of the steady-state polarization density for interacting atoms driven by the field in Eq. (12), with w 0 /λ = 2.77, in the limit of low light intensity where individual atoms respond to light as classical linear oscillators [66,67]. For an atom spacing a/λ = 0.5, the intensity of light scattered between nearest-neighbor atoms at the center of the lattice, I scat , is much larger than the maximum intensity of the incident field, I inc , with I scat /I inc 2. Therefore the atoms no longer emit light independently, but instead exhibit collective optical excitations, together with collective resonance linewidths and line shifts. Despite the presence of strong collective behavior, the optical excitations in Fig. 4 show clear vortex-like structures similar to the incident field. To understand this behavior, we calculate the collective excitation eigenmodes of the interacting system.
We find that the system supports several collective eigenmodes with singular defects in the real components of the atomic polarization amplitudes. The resulting sta- tionary excitations in Fig. 4 consist almost solely of a single collective excitation with an azimuthal (radial) defect, with a well-defined resonance linewidth and line shift, where the eigenmode occupation [68] reaches 99% at the eigenmode resonance, ∆/γ = 0.90 (∆/γ = 0.89). S 1 → S 1 mappings determine the winding number (Poincaré index) for a singular topological defect, as the count of the net total change in the real components of the polarization density orientation around a closed loop, with Q = 1 for the azimuthal and radial vortices. In an infinite system, the collective eigenmodes are real, and the system exhibits true topological defects. However, for the small atomic ensembles considered here, the imaginary components of the eigenmodes do not entirely vanish, e.g., the azimuthal vortex eigenmode appearing in the stationary excitation of Fig. 4(a) has a small 3% contribution to the total polarization density amplitude from the imaginary part. In comparison, the full stationary excitation has a 2% contribution from the imaginary part.
V. CONCLUDING REMARKS 3D particle-like topological objects have inspired research across a wide range of different disciplines. The ideas originate from Kelvin, who proposed how vortex strings forming closed loops, links, and knots could explain the structure of atoms [1]. To transfer such universal concepts to light and optical excitations, standard textbook representations of field amplitudes in terms of the Stokes vector on the Poincaré sphere fall dramatically short of the goal. This is because particle-like topologies can only be achieved in the complete optical hypersphere description, where variation of the total electromagnetic phase of vibration is retained. Here we have constructed a comprehensive platform of topologically non-trivial optical excitations of atoms, induced by light. The resulting amplitudes of electronic vibrations have been shown to exhibit substantially more complex topologies than the incident light creating them. In addition, this allows topological objects to be stored in excitations in highly controllable quantum systems with long lifetimes.
The proposed setup potentially paves the way for applications in future quantum simulators. The Skyrme model of 3D particle-like objects [10] is not only an elegant mathematical construction, but also simulates a low-energy limit of QCD where baryons are described by the quantised states of classical soliton solutions [69]. By describing these field configurations using linked Hopf maps, the particle-like objects take the form of links and knots, analogous to knotted solitons of the Skyrme-Faddeev model [25][26][27][28][29], and representing physical realisations of Kelvin's ideas in optical excitations. | 2021-09-30T01:16:25.751Z | 2021-09-28T00:00:00.000 | {
"year": 2021,
"sha1": "a5e1cc818038fa8c616cf94b0123e26ceaeb15ab",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "a5e1cc818038fa8c616cf94b0123e26ceaeb15ab",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
118587208 | pes2o/s2orc | v3-fos-license | Non-leptonic weak processes in spin-one color superconducting quark matter
The non-leptonic weak processes $s+u\to u+d$ and $u+d \to s+u$ are known to dominate the dissipation mechanism responsible for the viscosity of strange quark matter in its normal phase. The rates of such processes remain unknown for many color superconducting phases of quark matter. In this paper, we partially fill up the gap by calculating the difference of the rates of the two non-leptonic weak processes in four transverse spin-one color superconducting phases of quark matter (slightly) out of $\beta$-equilibrium. The four phases studied are the color-spin locked phase, the polar phase, the planar phase and the {\em A}-phase. In the limit of vanishing color superconducting gap, we reproduce the known results in the normal phase. In the general case, the rates are suppressed relative to the normal phase. The degree of the suppression is determined by the structure of the gap function in momentum space, which in turn is determined by the pairing pattern of quarks. At low temperatures, the rate is dominated by the ungapped modes. In this limit, the strongest suppression of the rate occurs in the color-spin-locked phase, and the weakest is in the polar phase and the {\em A}-phase.
I. INTRODUCTION
The interior of neutron stars is made of very dense baryonic matter. Currently our knowledge regarding the actual state of such matter is incomplete. One commonly accepted hypotheses is that the densest regions inside neutron stars are made of quark matter [1]. Moreover, such quark matter may be a color superconductor [2,3]. (For reviews on color superconductivity see for example Refs. [4,5,6,7,8,9,10,11,12].) From the viewpoint of basic research, it is of fundamental importance to test this hypothesis empirically.
The way to test the idea regarding the presence of quark matter inside stars is to make predictions regarding physics processes that affect observable features of stars and then test them against the stellar data. One class of physics properties that are substantially modified by the presence of color superconducting quark matter is related to the rates of weak processes. Such processes, for example, affect the cooling rates [13] and the suppression of the rotational (r-mode) instabilities [14] in stars. The latter in particular is determined by the viscous properties of dense matter [15].
Theoretically, the ground state of baryonic matter at very high density corresponds to the color-flavor-locked (CFL) phase of quark matter [16]. In this phase, quarks of all three colors and all three flavors participate in spinzero Cooper pairing on equal footing. The rates of the weak processes and some of their effects on the physical * Electronic address: igor.shovkovy@asu.edu properties of the CFL phase of quark matter have been discussed in Refs. [17,18,19,20,21].
With decreasing the density, the CFL phase should break up. This is due to the disruptive effects of a large difference between the masses of the strange quark and the light (up and down) quarks [22]. Such a difference leads to a mismatch between the Fermi momenta of quarks and, therefore, spoils the "democratic" pairing of the CFL phase. When the CFL phase breaks up, another type of spin-zero color superconductivity, the socalled two-flavor color superconducting (2SC) phase [23], can still be possible. In the 2SC phase, strange quarks do not participate in pairing. Also, up and down quarks of one color remain unpaired. Some weak processes in the 2SC phase and their effects on the physical properties have been studied in Refs. [24,25].
It is important to mention that matter inside stars is neutral (at least on average) and in β-equilibrium. Enforcing these two conditions affects the pairing between quarks and may disrupt the usual formation of cross-flavor spin-zero Cooper pairs [22]. In this case, the ground state of matter can be in other forms, for example, such as stable variants of crystalline [26], gapless [27,28], or other exotic phases [29,30]. When spin-zero pairing cannot occur, the ground state can be in one of the spinone color superconducting phases, in which same flavor quarks combine to form Cooper pairs [31,32,33,34,35].
Compared to the spin-zero case, the energy gap in spinone color superconductors is likely to be about two orders of magnitude smaller. This means that the actual value of the gap may be somewhere in the range from 0.01 MeV to 1 MeV. It appears that even such relatively small gaps can substantially affect the cooling rate of a quark star [36]. By the same token, such gaps can strongly modify the rates of the non-leptonic weak processes and, thus, affect the viscosity of stellar quark matter.
The bulk viscosity in the normal phase of three-flavor quark matter is usually dominated by the non-leptonic weak processes [37,38,39,40,41,42]. (The corresponding processes are diagrammatically shown in Fig. 1.) It was argued in Ref. [43], however, that the interplay between the Urca and non-leptonic processes may be rather involved even in the normal phase of quark matter. Indeed, because of the resonance-like dynamics responsible for the bulk viscosity and because of a subtle interference between the two types of the weak processes, a larger rate of the non-leptonic processes may not automatically mean its dominant role. In fact, it was shown that the contributions of the two types of weak processes are not separable and that, at low frequencies relevant for some pulsars, taking into account the Urca processes may substantially modify the result [43].
Currently it remains unknown how a similar interplay between the two types of weak processes is realized in spin-one color superconducting phases. Primarily, this is because the rates of the non-leptonic weak processes in the corresponding phases have not been calculated.
(Note that the rates of the Urca processes in several spin-one color superconducting phases were obtained in Refs. [36,44].) The purpose of this paper is to study the corresponding non-leptonic rates.
The rest of the paper is organized as follows. The derivation of a general expression for the non-leptonic rate, based on the Kadanoff-Baym formalism [46], is presented in the next section. The structure of the quark propagators in spin-one color superconducting phases is described in Subsec. II A. This is used in Subsec. II B to derive the imaginary part of the W -boson polarization tensor, which is the key ingredient in the expression for the rate. The net rate of the d-quark production (i.e., the difference of the rates of s+u → u+d and u+d → s+u) in the case of a small deviation from chemical equilibrium is obtained in Sec. III. There we also present the numerical results for each of the following spin-one color superconducting phases: the CSL phase (Subsec. III A), the polar phase (Subsec. III B), the A-phase (Subsec. III C), and the planar phase (Subsec. III D). In Sec. IV, we discuss the main results and their physical meaning. Two Appendices at the end of the paper contain some details, used in the derivation of the rate.
FIG. 2: Feynman diagram for the d-quark self-energy. The particle four-momenta are shown in parenthesis next to the particle names.
II. FORMALISM
In order to calculate the rates of the non-leptonic processes, we use the same approach as in Refs. [25,36,43,44,45]. It is based on the Kadanoff-Baym formalism [46]. The starting point of the analysis is the general Kadanoff-Baym equation for the Green functions (propagators) of the down (or strange) quarks. After applying the conventional gradient expansion close to equilibrium, we derive the following kinetic equation for the d-quark Green function: Here we denote the quark four-momenta by capital letters, e.g., P = (p 0 , p), where p 0 is the energy and p is the three-momentum. The structure of the quark Green's functions S < (P 1 ) and S > (P 1 ) in spin-one color superconducting phases will be discussed in the next subsection. To leading order, the quark self-energies Σ < (P 1 ) and Σ > (P 1 ) are given by the Feynman diagram in Fig. 2. This translates into the following explicit expression: where, by definition, M W and Q = P 1 − P 4 are the mass and the four-momentum of the W -boson, respectively. (Note that the large hierarchy between the Wboson mass and a typical momentum transfer Q 1 MeV justifies the approximation in which the W -boson propagator is replaced by 1/M 2 W .) As seen from the diagram in Fig. 2, the expression for the polarization tensor of the W -boson is given by In the Nambu-Gorkov notation used here, the explicit form of the (tree-level) vertices for the weak processes d ↔ u + W − and s ↔ u + W − reads [25] Γ µ ud/us,± = e V ud/us These are given in terms of the elements of the Cabibbo-Kobayashi-Maskawa matrix V ud and V us , and the weak mixing angle θ W . By construction, the τ -matrices operate in flavor space (u, d, s) and have the following form: By making use of Eqs. (1) and (2), the kinetic equation takes the following form: The physical meaning of the expression on the left hand side of this equation is the time derivative of the d-quark distribution function. By integrating this over the complete phase space, we obtain the net rate of the d-quark production: Then, by making use of the kinetic equation (1), we derive This rate should be non-vanishing only if the rates of the two non-leptonic weak processes u + s → d + u and d + u → u + s differ. In β-equilibrium, in particular, the latter two should be equal and the net rate of the d-quark production should vanish. The corresponding state of equilibrium in dense quark matter is reached when the chemical potentials of all three quark flavors are equal, i.e., µ u = µ d = µ s . (For simplicity, here it is assumed that all three quark flavors are approximately massless and, therefore, that the electrical neutrality of quark matter is achieved without the need for the electrons.) When the system is forced out of equilibrium, e.g., during the density oscillations caused by the collective modes of stellar matter, small deviations from β-equilibrium are induced. For our purposes, the corresponding state can be described by the following set of the chemical potentials: µ u = µ d = µ and µ s = µ + δµ, where δµ is a small parameter that characterizes the magnitude of the departure from the equilibrium state. Out of equilibrium, the net production of d-quarks may be nonzero. For example, if δµ > 0 (δµ < 0) the system has a deficit (an excess) of the down quarks and an excess (a deficit) of the strange quarks. Then, one of the weak processes, i.e., u + s → d + u (d + u → u + s), will start to dominate over the other in order to restore the equilibrium. The net rate of the d-quark (or equivalently s-quark) production characterizes how quickly this happens.
In order to calculate the net rate of d-quark production, however, one needs to know the explicit structure of the quark propagators in the specific spin-one color superconducting phases. The knowledge of the quark propagators is also needed for the calculation of the polarization tensors Π < µν (Q) and Π > µν (Q). These are discussed in the next two subsections.
A. Quark propagator in a spin-one color superconductor
In general, Cooper pairs in spin-one color superconducting phases are given by diquarks in a color antitriplet (antisymmetric) and spin triplet (symmetric) state. Depending on a specific color-spin structure, which is determined by the alignments of the antitriplet in color space and the triplet in spin (coordinate) space, many inequivalent color superconducting phases may form.
Each phase is unambiguously specified by the structure of the gap matrix, which is commonly written in the following form [32]: where φ e (P ) is the gap function. The Dirac matrices Λ e p ≡ (1 + eγ 0 γ ·p)/2, with e = ±, are the projectors onto the positive and negative energy states. The color structure of Φ(P ) + is determined by where (J i ) jk = −iǫ ijk are the antisymmetric matrices in color space,p ≡ p/p is the unit vector in the direction of the quasiparticle three-momentum p, and γ j ⊥ ≡ γ j − p j (γ ·p). The explicit form of the 3 × 3 matrix ∆ ij and the value of the angular parameter θ determine specific phases of superconducting matter. Among them, there is a number of inert and noninert spin-one phases [35], which are naturally characterized by the continuous and discrete symmetries preserved in the ground state.
In the two special cases, θ = 0 and θ = π/2, the corresponding phases are called longitudinal and transverse, respectively. In this paper we focus on the transverse phases (θ = π/2), in which only quarks of the opposite chiralities pair and which have lower free energies than the longitudinal phases [32]. To further constrain the large number of possibilities, we concentrate only on the following four most popular ones: the color-spin locked phase (CSL), the A-phase, the polar phase and the planar phase.
The structure of the matrices ∆ ij and M p for the mentioned four phases are quoted in the first two rows of Tab. I (for more details see Refs. [32]). In the corresponding ground states, the original symmetry SU(3) c ×SO(3) J ×U(1) em of one-flavor quark matter breaks down to [31,32,33,34,35] respectively. In spin-one color superconductors, there is no crossflavor pairing and, therefore, the quark propagator is diagonal in flavor space, i.e., The Nambu-Gorkov structure of each flavor-diagonal element is given by where f = u, d, s. The normal (diagonal) and anomalous (off-diagonal) components of the Nambu-Gorkov propagator have the following structure [32]: Here, r labels different quasiparticle excitations in colorsuperconducting quark matter. The matrices P − p,r and P + p,r are the projectors onto the subspaces spanned by the eigenvectors of M p M † p and γ 0 M † p M p γ 0 , respectively. The explicit form of the projectors for each phase can be found in Ref. [36]. It should be noted that both matrices M p M † p and γ 0 M † p M p γ 0 have the same set of eigenvalues λ p,r , The list of all eigenvalues as well as their degeneracies are given in the last three rows of Tab. I. Each of the eigenvalues determines a quark quasiparticle with the following dispersion relation: The separate components of the propagators in subspaces spanned by the eigenvectors, see Eqs. (13), (14) and (15), can be conveniently rewritten in terms of the corresponding distribution functions f (ǫ p,r,f ) and the Bogoliubov coefficients B ± p,r,f , i.e., The Bogoliubov coefficients and the fermion distribution function are defined as follows: CSL phase planar phase polar phase [Note that f (−ǫ) = 1 − f (ǫ).] The quark propagators in Eq. (12) can now be used to derive the general expressions for Π < µν (Q) and Π > µν (Q). This is done in the next subsection. The results are then used to calculate the rate Γ d in Eq. (8).
B. W -boson polarization tensor
The W -boson polarization tensor is given in terms of the quark propagators in Eq. (3). By taking into account the Nambu-Gorkov and flavor structure of the weak interaction vertices in Eq. (4), as well as the quark propagator in Eq. (12), we derive where we introduced the notation P 3 ≡ P 2 +Q. Note that the anomalous (off-diagonal) elements of the Nambu-Gorkov propagators dropped out from the result. This is the consequence of the electric charge conservation.
In calculations, this comes about as a result of the specific flavor structure of the weak interaction vertices in Eq. (4). One can further simplify the result for the polarization tensor in Eq. (25) by noticing that the two terms on the right hand side are equal. From physical viewpoint, this is related to the fact that the two terms are the charge-conjugate contributions of each other. After taking this into, we arrive at the following expression for the polarization tensor: Then, by using the explicit structure of the normal components of the u-and s-quark propagators, defined in Eq. (13), we obtain This can be rewritten in an equivalent form as where, by definition, the tensor T rr ′ µν (p,p ′ ) is given by the following trace (in color and Dirac spaces): This trace was calculated for each of the four spin-one color superconducting phases in Ref. [36]. For convenience, the corresponding results are also quoted in Appendix A. Finally, by making use of Eqs. (19) and (20), we arrive at the following expression for the W -boson polarization tensor: Here we denote δµ ≡ µ s − µ u and assume that the upper (lower) sign corresponds to Π < (Π > ). It should be mentioned that one of the δ-functions was used to perform the integration over p 2,0 .
III. CALCULATION OF THE RATE
In this section we derive a general expression for the net rate of the d-quark production in spin-one color superconducting quark matter close to chemical equilibrium. The corresponding rate is formally defined by Eq. (8). By making use of the quark propagators and the W -boson polarization tensor, derived in the previous section, we obtain where we used the following results for the traces: As in the calculation of the polarization tensor, the anomalous (off-diagonal) Nambu-Gorkov components of quark propagators did not contribute to these traces. This is the consequence of the specific flavor structure of the weak interaction vertices (4). After making use of Eqs. (19), (20) and (30), we obtain In derivation, we used the definition of the Fermi constant in terms of the W -boson mass, and the following Lorentz contraction: where ω rr ′ (p,p ′ ) denotes a color trace that involves a pair of quasiparticles (r and r ′ ) with the given directions of their three-momenta (p andp ′ ) in a specific spin-one color superconducting phase. The corresponding traces for all four phases are listed in Appendix A.
Formally, the expression in Eq. (34) gives the net rate of the d-quark production in quark matter away from chemical equilibrium. The first term in the brackets describes the production of d-quarks due to s + u → u + d, while the second one describes the annihilation of dquarks due to u + d → s + u.
Here it might be instructive to note that the above expression for the rate Γ d resembles the general result for the net rate of the d-quark production in the normal phase of strange quark matter [47]. The key difference comes from the presence of the Bogoliubov coefficients B p,r,f and the ω rr ′ (p,p ′ ) functions that account for a non-trivial quark structure of the quasiparticles in spinone color superconductors. Naturally, when such quasi-particles are the asymptotic states for the weak processes, the amplitude is not the same as in the normal phase.
The degree of departure from β-equilibrium and, thus, the net rate is controlled by the parameter δµ = µ s − µ d . When δµ = 0, the expression in the square brackets of Eq. (34) vanishes and Γ d = 0. When δµ = 0, on the other hand, one has Γ d ≃ λδµ (37) to leading order in small δµ [48]. Note that the overall sign was chosen so that λ is positive definite. (Recall that a positive δµ means an excess of strange quarks, which should drive a net production of d-quarks, while a negative δµ means a deficit of strange quarks, which will be produced by annihilating some d-quarks.) From the general expression in Eq. (34), we derive λ = 5λ 0 2 11 π 5 µ 5 T 3 r1r2r3r4 e1e2e3e4 where is the corresponding λ-rate in the normal phase of strange quark matter [47].
A. Analysis of the rate in CSL phase
Out of the four spin-one color superconducting phases studied in this paper, the CSL phase is special. This is the only phase in which the dispersion relations of quasiparticles are isotropic. As a result, the corresponding rate is the easiest to calculate. In this subsection, we analyze the λ-rate in the CSL phase in detail.
Let us start by noting that the explicit form of the ω rr ′ (p,p ′ )-functions in the CSL phase is given by (See Appendix A and Ref. [36].) By making use of these expressions and introducing the following notation for the angular integrals: we arrive at the following representation for the λ-rate in the CSL phase: ,u )f (e 3 ǫ p3,r3,s )f (e 4 ǫ p4,r4,u )δ(e 1 ǫ p1,r1,d + e 2 ǫ p2,r2,u − e 3 ǫ p3,r3,s − e 4 ǫ p4,r4,u ). (44) Here we took into account that, to leading order in inverse powers of µ, the absolute values of the quark threemomenta can be approximated by µ. In the same approximation, the explicit form of functions F r1r2r3r4 are given in Appendix B. All of them are proportional to 1/µ 3 . This factor cancels out with the overall µ 3 in Eq. (44). In order to perform the remaining numerical integrations, it is convenient to introduce new dimensionless integration variables x i = (p i −µ)/T instead of p i (i = 1, 2, 3, 4). The integration over x 4 is done explicitly by making use of the δ-function. The remaining three-dimensional integration is done numerically, using a Monte-Carlo method. One finds that the ratio λ (CSL) /λ 0 is a function of a single dimensionless ratio, φ/T . Before proceeding to the numerical results, it is instructive to analyze the limiting case of low temperatures (or alternatively very large φ/T ). In this limit, only the ungapped r = 2 quasiparticle modes should contribute to the rate. The corresponding contribution is easy to obtain analytically, i.e., The subleading correction to this result is suppressed by an exponentially small factor exp(− √ 2φ/T ). (Note that √ 2 in the exponent is connected with the conventional choice of the CSL gap, which is √ 2φ rather than φ.) It might be instructive to mention that the asymptotic value in Eq. (45) is substantially smaller than λ 0 /9, which is the corresponding contribution of a single ungapped mode in the normal phase. The additional suppression comes from the functions ω 22 (p 4 ,p 1 ) and ω 22 (p 3 ,p 2 ) which modify the amplitude of the weak processes with respect to the normal phase. Except for the special case of collinear processes (i.e.,p 4 parallel top 1 andp 3 parallel top 2 ), the corresponding ω-functions are less than 1, see Eq. (42). Interestingly, this kind of suppression is a unique property of the non-leptonic rates and is not seen in analogous Urca rates because the latter are dominated by the collinear processes [36,44].
All our numerical results for the λ-rates as a function of φ/T are shown in Fig. 3 [49]. In the case of the CSL phase (black points and the interpolating line in Fig. 3), we used the Mathematica's adaptive quasi-Monte-Carlo method to calculate the λ-rate. In order to improve the efficiency of the method, we partitioned the range of integration for each of the three dimensionless integration variables x i = (p i − µ)/T into several (up to 6) nonoverlapping regions. This approach insures that the main contribution, coming from a close neighborhood of the Fermi sphere, is not lost in the integration over a formally very large phase space.
As seen from Fig. 3, the numerical results smoothly interpolate between the value of the rate in the normal phase λ 0 and the asymptotic value of the rate due to the CSL ungapped modes, given by Eq. (45).
B. Analysis of the rate in polar phase
Unlike the CSL phase, the polar phase is not isotropic. However, it is the simplest one among the other three phases. While the dispersions relations of its quasiparticles depend on the angle θ p between the momentum p and a fixed z-drection, its ω rr ′ (p,p ′ )-functions are independent of the quasiparticle momenta, i.e., ω rr ′ (p,p ′ ) = n r δ rr ′ , with n 1 = 2 and n 2 = 1, see Appendix A. Taking this into account, the corresponding λ-rate takes a simple form: ,u )f (e 3 ǫ p3,r2,s )f (e 4 ǫ p4,r1,u ) × δ(e 1 ǫ p1,r1,d + e 2 ǫ p2,r2,u − e 3 ǫ p3,r2,s − e 4 ǫ p4,r1,u ). (47) By making use of the first δ-function, we easily perform the integration over p 4 . We can also perform the integration over one of the remaining polar coordinates. This is possible because the integrand depends only on the two independent combinations of the polar angles, i.e.,φ 1 = ϕ 1 − ϕ 3 andφ 2 = ϕ 2 − ϕ 3 . By usingφ 1 andφ 2 as new integration variables (for simplicity of notation, the tildes are dropped in the following), we see that the integrand is independent of the variable ϕ 3 . Finally, by approximating p 2 1 p 2 2 p 2 3 ≃ µ 6 in the integration measure, we rewrite the expression for the rate as follows: where the new integration variables are x i = (p i − µ)/T and ξ i = cos θ pi . By definition, the dimensionless energy is with λ ξ,1 = 1 − ξ 2 and λ ξ,2 = 0, and the new Bogoliubov coefficients are Note that the expressions for x 4 , ξ 4 and cos θ 34 in Eq. (48) are given by where p 4 = |p 1 + p 2 − p 3 | is a function of x i (i = 1, 2, 3) and the three cosine functions, In the calculation, we used a customized Monte-Carlo method in order to improve the statistical error of the integration over x i (with i = 1, 2, 3). To this end, we used a special type of importance sampling, which is motivated by the fact that the main contribution to the rate should come from the region near the Fermi surface. In order to implement this, we utilized random variables distributed according to the Gaussian distribution [50]: where x 0 and σ are the mean and the width of the distribution, respectively. This was applied to the numerical integration over the dimensionless variables x i = (p i − µ)/T (i = 1, 2, 3), in which case we took x 0 = 0 and σ = 3. In order to generate independent variables (e.g., x 1 and x 2 ), distributed according to Eq. (53), we applied the Box-Muller transform, where u 1 and u 2 are two independent variables, uniformly distributed in the range from 0 to 1. In our numerical calculation, we also used a Gaussian function to approximate the δ-function responsible for the energy conservation in the expression for the rate (48). For this purpose, we used the width of the distribution σ 0 = 0.2. This appeared to be sufficiently small to avoid strong violations of the energy conservation in the weak processes and, at the same time, sufficiently large to use in a Monte-Carlo integration with the number of (eight-dimensional) random points on the order of 10 6 (in a Mathematica code) or 10 7 (in a Fortran/C++ code).
The numerical results for the λ-rate in the polar phase are shown by the blue squares (and the interpolating line) in Fig. 3. At vanishing φ/T , the rate coincides with that in the normal phase. At asymptotically large value of φ/T , on the other hand, the rate approaches λ 0 /9. This value is marked by the purple dashed line in the figure. Theoretically, the rate is dominated by the ungapped modes (r 1 = r 2 = 2) in the φ/T → ∞ limit. The corresponding contribution can be obtained by analytical methods as follows. We start by pointing that the Bogoliubov coefficients for the ungapped modes are equal to the unit step functions: B ei xi,ξi,2 ≡ Θ(−e i x i ), where by definition Θ(x) = 1 for x ≥ 0 and Θ(x) = 0 otherwise. Since these Bogoliubov coefficients are nonzero only for e i = sign(−x i ), each sums over e i effectively reduces to a single contribution. By taking this into account and making use of the result for the angular integration, K 0 , defined in Appendix B, we derive ) (e x1 + 1)(e x2 + 1)(e −x3 + 1)(e −x4 + 1) It should be noted that the numerical results for the polar phase in Fig. 3 approach this asymptotic value very slowly. We can speculate that this indicates a weak (probably, power-law) suppression of the contribution of the gapped (mixed with ungapped) modes to the rate.
The key feature responsible for this behavior in the polar phase is the presence of gapless nodes at θ p = 0 and θ p = π in the dispersion relation of the gapped modes. As we shall see below, the same qualitative property is shared by the A-phase, whose gapped modes also have a node at θ p = π. In contrast, the rates in the CSL and planar phases, whose gapped modes have no gapless nodes, show asymptotes that are consistent with the rapid, exponential approach to their asymptotic values.
C. Analysis of the rate in A-phase
The analysis in the A-phase of spin-one color superconducting matter can be performed along the same lines as in the polar phase. The apparent complication of the A-phase is the existence of three, rather than two distinct quasiparticle excitations. However, it appears that the contributions of the two gapped modes (r = 1, 2) can be replaced by a single contribution of a modified mode with the energy ǫ p = (p − µ) 2 + |φ| 2 λ p where λ p ≡ (1 + cos θ p ) 2 (cf. the dispersion relations of the modes r = 1, 2 in Tab. I). This alternative representation is possible because of the special, separable structure of the corresponding ω rr ′ (p,p ′ ) functions in the A-phase. As seen from the expressions in Eq. (A7), the mode r = 1 contributes only when cos θ p of the corresponding quasiparticle is positive, while the mode r = 2 contributes only when cos θ p is negative. Then, when the contributions are nonvanishing, one always gets ω rr ′ (p,p ′ ) = 2. By also noting that the corresponding eigenvalues λ p,1 = (1 + | cos θ p |) 2 for cos θ p > 0 (57) and λ p,2 = (1 − | cos θ p |) 2 for cos θ p < 0 (58) formally take the same form, i.e., λ p ≡ (1 + cos θ p ) 2 , we conclude that the sum over the original modes r = 1, 2 in the rate can indeed be replaced by a single contribution of the modified mode as defined above. By making use of this observation, the general expression for the rate in the A-phase takes the form, which is similar to that in the polar phase, see Eq. (47), but with a different dispersion relation of the (modified) gapped mode. By using a Monte-Carlo algorithm as in the previous case, we perform a numerical calculation of the λ-rate in the A-phase. The corresponding results are shown by red stars (and the interpolating line) in Fig. 3. In the limit of large φ/T , the rate is saturated by the contribution of ungapped modes, which is the same as in the polar phase, namely λ 0 /9. The derivation of this asymptotic expression is the same as in the polar phase. The corresponding value is marked by the purple dashed line in the figure. A slow (probably, power-law) approach of the asymptotic value at φ/T → ∞ is again associated with the presence of a gapless node (at θ p = π) in the dispersion relation of the (modified) gapped quasiparticles.
D. Analysis of the rate in planar phase
The calculation of the rate in the planar case requires the largest amount of computer time. One of the main reasons for that is the much more complicated expressions for the ω rr ′ (p,p ′ )-functions (see Appendix A). The numerical results for the λ-rate in the planar phase are shown by green diamonds (and the interpolating line) in Fig. 3. The asymptotic value of the rate at large φ/T was extracted only numerically. By taking into account possible systematic errors (e.g., due to the overall normalization of the rate that may differ by up to 15% from the analytical estimate (39) in the normal phase), we estimate λ (planar) ≃ (0.038 ± 0.003)λ 0 for φ/T → ∞. Note that this is smaller than λ 0 /9, which is the contribution of a single mode in the normal phase. As in the CSL phase, in the planar phase the additional suppression comes from the ω-functions for the ungapped modes.
IV. DISCUSSION
In this paper we derived the near-equilibrium rates of the net d-quark production (or equivalently the λ-rates) due to the non-leptonic weak processes (i.e., the difference of the rates of u + d → u + s and u + s → u + d) in spin-one color-superconducting strange quark matter at high density. The main numerical results are presented in Fig. 3.
In the limit of φ/T = 0, which is same as the normal (unpaired) phase of strange quark matter, our results reproduce the known result of Ref. [47]. The effect of color superconductivity is to suppress these rates. The degree of the suppression depends on the details of the specific spin-one phases. To large extent, this is controlled by the value of the energy gap (more precisely, φ/T ) as well as its functional dependence on the direction of the quasiparticle momentum. At very large φ/T (or equivalently in the limit of low temperatures), the λ-rates approach fixed values, which are determined by the contribution of the ungapped modes alone. The corresponding limiting value is the smallest in the CSL phase. It is less than a third of the "canonical" value λ 0 /9 due to a single ungapped mode in the normal phase of matter. The additional suppression comes from the modification of the quasiparticles due to color superconductivity. A similar observation applies to the planar phase. The rates in the other two phases, i.e., the polar and A-phase, approach the asymptotic values equal to λ 0 /9.
The numerical results for the λ-rates in Fig. 3 also indicate that the asymptotic approach to the limiting values can be qualitatively different in spin-one color superconducting phases. In the case of the polar and A-phase, the approach seems to follow a power law. In contrast, the approach appears to be exponential in the case of the CSL and planar phase. This qualitative difference can be easily understood. The power law is the consequence of the presence of gapless nodes in the dispersion relations of the gapped quasiparticles in the polar and A-phase (the nodes are located at θ p = 0 and θ p = π in the polar phase, and at θ p = π in the A-phase). In the CSL and planar phase, the approach to the asymptotic value at φ/T → ∞ is exponential because no gapless nodes are found in their gapped quasiparticles. (Note that a similar observation regarding the rates of the semi-leptonic processes was made in Ref. [36,44].) The results for the rates of non-leptonic weak processes, presented here, is an important ingredient for the calculation of the bulk viscosity of spin-one colorsuperconducting strange quark matter. If such matter is present inside neutron stars, its viscosity will be one of the mechanisms responsible for damping of the stellar r-mode instabilities [14].
APPENDIX A: COLOR AND DIRAC TRACES
In this appendix, we write down the explicit expressions for the tensor T rr ′ µν (p,p ′ ), defined by Eq. (29). The corresponding results were obtained in Ref. [36]. In general, one finds that where ω rr ′ (p,p ′ ) are the functions determined by a specific color-spin structure of the gap matrix, and The explicit form of all the components of this tensor can be also found in Ref. [36]. It is more important for us here to note that the following result for the contraction of this tensor with itself is valid: (A3) Since an essential information regarding spin-one colorsuperconducting phases is carried by the ω rr ′ (p,p ′ ) functions, we also quote them here. (For more details, see Ref. [36].) In the polar phase, the ω rr ′ (p,p ′ ) functions do not depend on the quark momenta. They are given by the following expressions: In the planar phase, the explicit form of the ω rr ′ (p,p ′ ) functions reads where .
(A6) In the A-phase, there are three different quasiparticle branches (r = 1, 2, 3). Consequently, there are more ω rr ′ (p,p ′ ) functions, i.e., Finally, in the CSL phase, the corresponding functions are In the calculation of the λ-rate in the CSL phase, there are four different types of angular integrations over the phase space of quark momenta. Thus, the results for the F r1r2r3r4 functions, formally defined by Eq. (43) in the main text, have the following general structure: where the four types of angular integrals are given by: The result for K 0 was obtained in Ref. [25]. It reads L 0 (p 12 , P 12 , p 34 , P 34 ) , where p ij ≡ |p i − p j |, P ij ≡ p i + p j , and which is given in terms of To leading order in powers of large µ, this result simplifies to L 0 (0, 2µ, 0, 2µ) = 2 8 µ 5 15 . (B9) By making use of Eq. (B6), therefore, we obtain K 0 ≃ 4π 3 µ 8 L 0 (0, 2µ, 0, 2µ) =
Calculation of K2
As is easy to see, the expression for K 2 can be obtained from K 1 by the following exchange of variables: p 1 ↔ p 2 and p 3 ↔ p 4 . Thus, the result reads K 2 = π 3 2p 2 1 p 2 2 p 2 3 p 2 4 L 2 (p 12 , P 12 , p 34 , P 34 ) , where and (B22) We also find that K 2 is identical to K 1 to leading order in powers of large µ, i.e., K 2 ≃ 2 10 π 3 35µ 3 . (B23)
Calculation of K3
Now we calculate the angular integral K 3 . We start by using the same approach as in the calculation of K 1 , To calculate the integral over Ω 1 , we fix the coordinate system so that the z-axis coincides with the direction of P.
After integration, we obtain | 2009-12-19T16:31:50.000Z | 2009-12-19T00:00:00.000 | {
"year": 2009,
"sha1": "926184770eac9a1ce3081f327afb634402f6c7b7",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0912.3851",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "926184770eac9a1ce3081f327afb634402f6c7b7",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
202550527 | pes2o/s2orc | v3-fos-license | Maternal food allergy is associated with daughters’ menarche in early adolescence
Rationale Associations between allergic disease and puberty amongst females have been widely studied. However, this association has received less attention in multigenerational populations. To this end, we sought to examine maternal allergic disease status ever, and daughters’ menarche. Methods In a cohort of children born in 1995, in Manitoba, Canada, we considered maternal allergic disease ever to daughters’ age 7–8 years, and daughters’ menarche at ages 12–14 years. We included all participants for whom we had information on both the exposure and the outcome of those eligible. Data were analysed using descriptive statistics and logistic regression, with adjustment for confounding variables. Results Overall, the prevalences of maternal allergic diseases were 28.6% for asthma 18.8% for food allergy, 27.3% for eczema and 45.5% for rhinitis. By age 12–14 years, 41.6% (64/159) girls had reached menarche. Maternal food allergy was significantly associated with daughters’ menarche (OR 4.39, 95% CI 1.51–12.73), whereas no association was found for maternal asthma, eczema or rhinitis. With consideration to comorbid disease, a combination of maternal asthma + food allergy was associated with daughters’ menarche by age 12–14 years (OR 6.41; 95% CI 1.32–31.01). Conclusions Maternal food allergy ever is associated with daughters’ menarche by age 12–14 years.
A gender switch in allergic disease has been noted during the pubertal years [1]. Less is known about the effect of allergic disease on timing of pubertal development. In one Swedish study, no clear associations were found between asthma, including timing of onset and phenotypes, and pubertal staging [2]. To our knowledge, no studies have examined maternal allergic disease and the timing of daughters' menarche.
To investigate this knowledge gap, we used data from 154 mother-daughter dyads from the Study of Allergy, Genes and the Environment (SAGE) [3], a general population-based cohort of children at high-and low risk for asthma. In this exploratory analysis, data on maternal allergic disease (self-reported asthma, food allergy, eczema and/or rhinitis) were collected several years prior to daughters' menarche. Thus, this study design provided the ability to estimate the impact of maternal allergic disease and daughters' subsequent age of menarche. As maternal stress is associated with early pubertal onset for their daughters [4], we also considered maternal depression shortly after their daughters' births in our analyses. Our primary aim was to examine the association between individual maternal allergic diseases and daughters' menarche. Our secondary aim was to consider the timing of disease onset and disease comorbidities, and daughters' menarche.
Briefly, in 2002, 723 children born in Manitoba, Canada in 1995 were recruited to SAGE. Children and their families participated in al assessment and completed questionnaires when the children were ages 7-9 years (baseline), ages 10-11 (late childhood) and 12-14 years (adolescence; 68% retention). At baseline, mothers reported if they had experienced any previous symptoms of asthma, food allergy, eczema or rhinitis. Asthma was further dichotomised as childhood onset (≤ 12 years) vs. post-pubertal onset (13+ years). At the adolescent visit, At baseline, mothers provided information on breastfeeding, birthweight, gestational age, and maternal smoking and education. Additionally, mothers reported household income, which we dichotomised at $50,000, to approximately align with the median Manitoba income at baseline. In late childhood, mothers reported whether they felt depressed or hopeless following their daughter's birth in 1995. Possible answers to this question were dichotomised as no vs. yes.
In adolescence, girls' heights and weights were measured in triplicate by research staff. The mean measures were taken, from which body mass index (BMI) was calculated. As hip and waist measures (in centimeters), converted to waist-hip ratio, did not substantially alter point estimates (< 0.10) or change statistical significance compared to analyses in which BMI was considered, we present only the results in which BMI was considered.
Data were described using n, %, mean, and 95% confidence intervals (95% CI). Analytic statistics included logistic regression, reported as odds ratios (OR) and corresponding 95% CI. Potential confounding variables were identified using directed acyclic graphs [4], and considered in partially and fully adjusted models. Statistical significance was set at p < 0.05. Data were analysed using Stata 13.1 (College Station, TX). Ethical permission was granted by the University of Manitoba Health Research Ethics Board (HS14742(HS2002:078)).
Of the 470 participants seen in adolescence, 203 were girls, for whom menarche data were available for 154 (75.9%) This constituted our study population. Mothers reported predominantly Caucasian ethnicity, and the majority had post-secondary education and had breastfed their daughters (Table 1). Approximately 30% (44/154) of mothers had asthma, of whom 38.4% (17/44) had pre-pubertal asthma. Other allergic diseases were also common.
No associations were found between maternal allergic disease and daughters' thelarche ( Table 2). In contrast, in unadjusted and partially adjusted models, maternal asthma trended towards an association with daughters' menarche by the adolescent visit, whereas this association was significant for maternal food allergy (Table 2). In models adjusted for all covariates except maternal depression, the statistically significant association between maternal food allergy and daughters' menarche persisted (OR 3.02; 95% CI 1.15-7.93; p < 0.03). In contrast, neither maternal eczema nor rhinitis were associated with daughters' menarche. Adding the covariate, maternal depression, insubstantially altered the corresponding point estimates, thereby further strengthening the results. The difference in findings between thelarche and menarche may be partly attributable to differences in reporting (daughter vs. mother, respectively), as reflected by a moderate correlation between these variables (r 0.498).
Given the null findings between maternal allergic disease and daughters' thelarche, we performed no further similar analyses. However, we did consider timing of maternal asthma onset and one vs. both of these allergic diseases, in association with daughters' menarche. Nearly all (95.5%; 42/44) mothers with asthma reported the age at which they had their first asthma exacerbation. No associations were found between prevs post-pubertal first maternal asthma exacerbation and daughters' menarche (Table 3). Similarly, no statistically significant associations were found between maternal asthma or food allergy ever and daughters' menarche in partially adjusted and fully adjusted models. Both maternal asthma and food allergy increased the odds of daughters' menarche more than five-fold (fully adjusted: OR 5.71; 95% CI 1.20-27.3). Although the numbers for some sub-analyses were small, the point estimates were similar to those from analyses of the entire study population, indicating a robust association. Moreover, these analyses were robust to adjustment for BMI, which is also associated with early puberty [5]. Data on maternal allergic disease were based on self-report, not clinical testing. However, by using maternal data from baseline and reports of daughters' pubertal development at later ages, we eliminate any potential reporting bias. We acknowledge that out outcome, daughters' menarche, was reported by mothers rather than the girls themselves. However, any differences in classification of menarche are likely to be non-differential. In addition, we were unable to consider maternal age at menarche, as these data were not collected in our study. To our knowledge, this is the first study on maternal food allergy and daughters' pubertal development, and highlights the need to consider the impact of maternal allergic disease using multigenerational studies. Although there are diverging results as to which parent's allergic disease status confers greater risk [6,7], a greater maternal impact may be attributable to their role as the parent of origin, as well as environmental and immunological interactions with the offspring during pregnancy and birth. Accordingly, we restricted our analyses to mother-daughter dyads. There is substantial, but collectively inconclusive evidence surrounding allergic disease and puberty [6,8], asthma and subsequent menarche [7,9]. Whereas biological plausibility has been described for this association in a single generation, it remains unclear why maternal food allergy is associated with daughters' menarche in early adolescence. Maternal atopic disease, especially food allergy, may well indicate the start of a multigenerational cascade of chronic, inflammatory disease. As such, this observation warrants further investigation as early menarche increase the daughters' risk of other chronic conditions, including type 2 diabetes [10] and cardiovascular disease [11]. Likewise, physicians treating girls whose mothers have food allergy may wish to be mindful of early menarche, and carefully monitor factors, such as body weight and blood glucose, which increase the risk of cardiovascular disease.
In conclusion, our study demonstrates an association between maternal food allergy alone, or in combination with asthma, and daughters' menarche in early adolescence.
Abbreviations BMI: body mass index; OR: odds ratio; SAGE: study of allergy, genes and the environment; 95% CI: 95th percent confidence interval. | 2019-09-12T13:28:22.172Z | 2019-09-11T00:00:00.000 | {
"year": 2019,
"sha1": "3e980f523589f22142d52c4a0e2686fe9fdca266",
"oa_license": "CCBY",
"oa_url": "https://aacijournal.biomedcentral.com/track/pdf/10.1186/s13223-019-0371-0",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8b3f5f4d73695a1e06cc858c8ae1ecdd36924175",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
232081704 | pes2o/s2orc | v3-fos-license | A Tad-like apparatus is required for contact-dependent prey killing in predatory social bacteria
Myxococcus xanthus, a soil bacterium, predates collectively using motility to invade prey colonies. Prey lysis is mostly thought to rely on secreted factors, cocktails of antibiotics and enzymes, and direct contact with Myxococcus cells. In this study, we show that on surfaces the coupling of A-motility and contact-dependent killing is the central predatory mechanism driving effective prey colony invasion and consumption. At the molecular level, contact-dependent killing involves a newly discovered type IV filament-like machinery (Kil) that both promotes motility arrest and prey cell plasmolysis. In this process, Kil proteins assemble at the predator-prey contact site, suggesting that they allow tight contact with prey cells for their intoxication. Kil-like systems form a new class of Tad-like machineries in predatory bacteria, suggesting a conserved function in predator-prey interactions. This study further reveals a novel cell-cell interaction function for bacterial pili-like assemblages.
Introduction
Bacterial predators have evolved strategies to consume other microbes as a nutrient source. Despite the suspected importance of predation on microbial ecology (Mu et al., 2020), a limited number of bacterial species are currently reported as predatory. Amongst them, obligate intracellular predators collectively known as BALOs (e.g. Bdellovibrio bacteriovorus) (Mu et al., 2020), penetrate the bacterial prey cell wall and multiply in the periplasm, escaping and killing the host bacteria (Laloux, 2019). Quite differently, facultative predators (meaning that they can be cultured in absence of prey if nutrient media are provided, that is Myxococcus, Lysobacter, and Herpetosiphon, Mu et al., 2020) attack their preys extracellularly, presumably by secreting antimicrobial substances and digesting the resulting products. Among these organisms and studied here, Myxococcus xanthus, a delta-proteobacterium, is of particular interest because it uses large-scale collective movements to attack prey bacteria in a so-called 'wolf-pack' mechanism (Thiery and Kaimer, 2020).
A tremendous body of work describes how Myxococcus cells move and respond to signals in pure culture . In contrast, mechanistic studies of the predatory cycle have been limited. Currently, it is considered that coordinated group movements allow Myxococcus cells to invade prey colonies and consume them via the secretion of a number of diffusible factors, extracellular enzymes, antibiotics, and outer membrane vesicles (Thiery and Kaimer, 2020;Pérez et al., 2016;Xiao et al., 2011). While each of these processes could contribute to predation, evidence for their requirement is still missing (Thiery and Kaimer, 2020). In addition, Myxococcus cells have also been observed to kill prey cells upon contact, an intriguing process during which single motile Myxococcus cells were observed to stop and induce E. coli plasmolysis within a few minutes (Zhang et al., 2020b). Myxococcus pauses were more frequent with live E. coli cells implying prey detection but the mechanism at work and the exact relevance of prey-contact killing for predation remain unclear. Potential contact-dependent mechanisms have been described in Myxococcus, including Type VI secretion (Troselj et al., 2018) and Outer Membrane Exchange (OME, Sah and Wall, 2020). However, while these processes have been implicated in Kin recognition and homeostatic regulations within Myxococcus colonies (Troselj et al., 2018;Sah and Wall, 2020), they remain to be clearly implicated in predation. In this study, we analyzed the importance of motility and contact-dependent killing in the Myxococcus predation cycle.
To explore these central questions, we first developed a sufficiently resolved imaging assay where the Myxococcus predation cycle can be imaged stably at the single-cell level over periods of time encompassing several hours with a temporal resolution of seconds. The exact methodology underlying this technique is described in a dedicated manuscript (Panigrahi, 2020); briefly, the system relates predatory patterns observed at the mesoscale with single-cell resolution, obtained by zooming in and out on the same microscopy specimen (Figure 1). Here, we employed it to study how Myxococcus cells invade and grow over Escherichia coli prey cells during the initial invasion stage ( Figure 1, Video 1). Figure 1. A-motility is required for invasion of prey colonies. Colony plate assays showing invasion of an E. coli prey colony (dotted line) 48 hr after plating by WT (a, Video 1), A + S -(b) and A -S + (c, Video 2) strains. Scale bar = 2 mm. (a1) Zoom of the invasion front. Myxococcus single cells are labeled with mCherry. Blue arrows show the movement of 'arrowhead' cell groups as they invade prey colonies. White arrows point to A-motile single cells that penetrated the prey colony. Scale bar = 10 mm. See associated Video 1 for the full time lapse. (b1) Zoom of the invasion front formed by A + Scells. The A-motile Myxococcus cells can infiltrate the prey colony and kill prey cells. Scale bar = 10 mm. (c1) Zoom of the invasion front formed by A -S + cells. Note that the S-motile Myxococcus cells come in contact with the prey colony, but in absence of A-motility, the predatory cells fail to infiltrate the colony and remain stuck at the border. Scale bar = 10 mm. See associated Video 2 for the full time lapse.
A-motility is required for prey colony invasion
Although the function of motility in prey invasion is generally accepted, Myxococcus xanthus possesses two independent motility systems and the relative contribution of each system to the invasion process is unknown. Social (S)-motility is a form of bacterial 'twitching' motility that uses so-called Type IV pili (TFP) acting at the bacterial pole (Mercier et al., 2020). In this process, polymerized TFPs act like 'grappling hooks' that retract and pull the cell forward. S-motility promotes the coordinated movements of Myxococcus cells within large cell groups due to interaction with a self-secreted extracellular matrix formed of Exo-Polysaccharide (EPS) (Hu et al., 2016;Li et al., 2003;Islam et al., 2020). A(Adventurous)-motility promotes the movement of Myxococcus single cells at the colony edges. A-motility is driven by a mobile cell-envelope motor complex (named Agl-Glt) that traffics in helical trajectories along the cell axis, driving rotational propulsion of the cell when it becomes tethered to the underlying surface at so-called bacterial Focal Adhesions (bFAs) (Faure et al., 2016). We tested the relative contribution of each motility system to prey invasion by comparing the relative predatory performances of WT, A + S -(DpilA, Sun et al., 2011) and A -S + (-DaglQ, Sun et al., 2011) strains ( Figure 1). Interestingly, although A + Scells were defective in the late developmental steps (fruiting body formation), they were still proficient at prey invasion ( Figure 1b). On the contrary, the A -S + strain was very defective at prey colony invasion (Figure 1c). Zooming at the prey colony border, it was apparent that the A -S + cells were able to expand and contact the prey colony, but they were unable to penetrate it efficiently, suggesting that Type IV pili on their own are not sufficient for invasion ( Figure 1c, Video 2). Conversely, A-motile cells were observed to penetrate the tightly-knitted E. coli colony with single Myxococcus cells moving into the prey colony, followed by larger cell groups (Figure 1a). Similar motility requirements were also observed in a predatory assay where predatory and prey cells are pre-mixed before they are spotted on an agar surface (see Figure 2 and Figure 2-figure supplement 1). Thus, A-motility is the main driver of prey invasion on surfaces. Video 1. Invasion of an E. coli colony by WT Myxococcus cells. This movie was taken at the interface between the two colonies during invasion. The movie is an 8x compression of an original movie that was shot for 10 hr with a frame taken every 30 s at Â40 magnification. To facilitate Myxococcus cells tracking, the wild-type strain was labeled with the mCherry fluorescent protein. https://elifesciences.org/articles/72409#video1 Video 2. A-motility is required for prey invasion. This movie was taken at the interface between the two colonies during invasion. The movie is an 8x compression of an original movie that was shot for 10 hr with a frame taken every 30 s at Â40 magnification.
Invading Myxococcus cells kill prey cells upon contact
To further determine how A-motile cells invade the prey colony, we shot single cell time-lapse movies of the invasion process. First, we localized a bFA marker, the AglZ protein (Mignot et al., 2007) fused to Neon-Green (AglZ-NG) in Myxococcus cells as they penetrate the prey colony. AglZ-NG binds to the cytoplasmic face of the Agl-Glt complex and has long been used as a bFA localization marker; it generally forms fixed fluorescent clusters on the ventral side of the cell that retain fixed positions in gliding cells (Mignot et al., 2007). As Myxococcus cells invaded prey colonies, they often formed 'arrow-shaped' cell groups, in which the cells within the arrow assembled focal adhesions ( Figure 2a, Video 3). E. coli cells lysed in the vicinity of the Myxococcus cells, suggesting that a contact-dependent killing mechanism (as reported by Zhang et al., 2020b) occurs during prey colony invasion (Figure 2a). To observe this activity directly, we set up a Myxococcus -E. coli interaction microscopy assay where predator -prey interactions can be easily studied, isolated from a larger multicellular context (see Materials and methods). In this system, A-motile Myxococcus cells were observed to mark a pause and disassemble bFAs when contacting E. coli cells (Figure 2b, Video 4, further quantified below); this pause was invariably followed by the rapid death of E. coli, as detected by the instantaneous dispersal of a cytosolic fluorescent protein (mCherry or GFP, Figure 2b-c, observed in n=20 cells). This observation suggests that the killing occurs by plasmolysis, a process which is likely to be the same as that described by Zhang et al., 2020b. To demonstrate this, we mixed Myxococcus cells with E. coli cells in which peptidoglycan (PG) had been labeled by fluorescent D-amino Acids (TADA Faure et al., 2016). TADA is covalently incorporated into the PG pentapeptide backbone and it does not diffuse Shown is an AglZ-YFP expressing Myxococcus cell establishing contact with an mCherry-expressing E. coli cell (overlay and phase contrast image). Note that the Myxococcus cell resumes movement and thus re-initiates bFA (white arrowheads) formation immediately after E. coli cell lysis. See associated Video 4 for the full time lapse. Scale bar = 2 mm. (c) Myxococcus (outlined in white) provokes E. coli plasmolysis. Top: shown is a GFP-expressing E. coli cell lysing in contact with a Myxococcus cell. GFP fluorescence remains stable for 5 min after contact and becomes undetectable instantaneously, suggesting plasmolysis of the E. coli cell. Scale bar = 2 mm. Bottom: graphic representation of fluorescence intensity loss upon prey lysis. (d,e) Myxococcus contact provokes local degradation of the E. coli peptidoglycan. (d) E. coli PG was labeled covalently with the fluorescent D-amino acid TADA. Two E. coli cells lyse upon contact. Holes in the PG-labelling are observed at the contact sites (white arrows). Note that evidence for plasmolysis and local IM membrane contraction is visible by phase contrast for the lower E. coli cell (dark arrow). Scale bar = 2 mm. (e) Kymograph of TADA-labeling corresponding to the upper E. coli cell. At time 0, which corresponds to the detection of cell lysis, a hole is detected at the contact site and propagates bi-directionally from the initial site showing that the prey cell wall is degraded in time after cell death. Scale bar = 1 mm. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. E. coil loss of fluorescence during contact-dependent lysis (Figure 2c). laterally (Kuru et al., 2012). We first observed contraction of the E. coli cytosolic dense region at the pole by phase contrast (Figure 2d), which was followed by the appearance of a dark area in the PG TADA staining exactly at the predator-prey contact site ( Figure 2d). It is unlikely that this dark area forms due to the new incorporation of unlabeled prey PG, because it was detected immediately upon prey cell death and propagated bi-directionally afterwards (Figure 2d-e). Thus, these observations suggest that, upon contact, Myxococcus induces degradation of the E. coli PG, which provokes cell lysis due to loss of turgor pressure and hyper osmotic shock (Zhang et al., 2020b). The bi-directional propagation of PG hydrolysis (as detected by loss of TADA signal) suggests that PG hydrolysis could be driven by the activity of PG hydrolase(s) disseminating from the predator-prey contact site.
A predicted Tad-pilus is required for contact-dependent killing We next aimed to identify the molecular system that underlies contact-dependent killing. Although motility appears to be essential during the predation process ( Figure 2-figure supplement 1), at the microscopic level, direct transplantation of A -S -(DaglQ DpilA) in E. coli prey colonies still exhibit contact-dependent killing ( Figure 2-figure supplement 2), demonstrating that the killing activity is not carried by the motility complexes themselves. Myxococcus xanthus also expresses a functional Type VI secretion system (T6SS), which appears to act as a factor modulating population homeostasis and mediating Kin discrimination between M. xanthus strains (Troselj et al., 2018;Vassallo et al., 2020). A T6SS deletion strain (Dt6ss) had no observable defect in contact-dependent killing of prey cells ( Figure 2-figure supplements 1 and 3). In addition, the Myxococcus T6SS assembled in a prey-independent manner as observed using a functional VipA-GFP strain that marks the T6SS contractile sheath (Brunet et al., 2013;Figure 2-figure supplements 4-5), confirming that T6SS is not involved in predatory killing on surfaces.
To identify the genes (directly or indirectly) involved in the contact-dependent killing mechanism, we designed an assay where contact-dependent killing can be directly monitored in liquid cultures and observed via a simple colorimetric assay. In this system, the lysis of E. coli cells can be directly monitored when intracellular b-galactosidase is released in buffer containing ChloroPhenol Red-b-D-Galactopyranoside (CPRG), which acts as a substrate for the enzyme and generates a dark red hydrolysis reaction product (Paradis-Bleau et al., 2014). Indeed, while Myxococcus or E. coli cells incubated alone did not produce color during a 120 hr incubation, their mixing produced red color indicative of E. coli lysis after 24 hr (Figure 2-figure supplement 6). In this assay, a t6SS mutant was still able to lyse E. coli cells, demonstrating that predation is not T6SS-dependent (Figure 2figure supplements 6-7). CPRG hydrolysis was not detected when Myxococcus and E. coli were separated by a semi-permeable membrane that allows diffusion of soluble molecules, showing that the assay reports contact-dependent killing (Figure 2-figure supplement 6). In this liquid assay, the Myxococcus -E. coli contacts are very distinct from contacts on solid surfaces and thus, the genetic requirements are likely quite distinct. Indeed, in the liquid CPRG assay, we observed that TFPs are essential for killing while the Agl/Glt system is dispensable (Figure 2-figure supplement 7). In this condition, TFPs promote a prey-induced aggregation of cells (Figure 2-figure supplement 8) and thus probably mediate the necessary tight contacts between Myxococcus and E. coli cells. As shown below, the killing process itself is the same in liquid as the one observed on surfaces and it is not directly mediated by the TFPs.
Given the probable indirect effect of TFPs, we next searched additional systems involved in CPRG contact-dependent killing. Using a targeted approach, we tested the effect of mutations in genome annotated cell-envelope complexes on contact-dependent killing in liquid cultures. Doing so, we identified two critical genetic regions, the MXAN_3102-3108 and the MXAN_4648-4661 regions ( Figure 3). Functional annotations indicate that both genetic regions carry a complementary set of genes encoding proteins that assemble a so-called Tight adherence (Tad) pilus. Bacterial Tad pili are members of the type IV filament superfamily (also including Type IV pili, a and b types, and Type II secretion systems) and extrude polymeric pilin filaments assembled via inner membrane associated motors through an OM secretin (Denise et al., 2019). Tad pili have been generally involved in bacterial adhesion and more recently, in contact-dependent regulation of adhesion (Ellison et al., 2017). The MXAN_3102-3108 cluster genes with annotated functions encode a predicted pre-pilin peptidase (CpaA and renamed KilA), a secretin homolog (CpaC/KilC) and a cytoplasmic hexameric ATPase (CpaF/KilF) (following the Caulobacter crescentus Tad pilus encoding cpa genes nomenclature, see . A Tad-like apparatus is required for prey recognition and contact-dependent killing. (a) Model structure of the Kil system following bioinformatics predictions. Annotated cluster 1 and cluster 2 genes are shown together with the possible localization of their protein products. Dark triangles indicate the genes that were deleted in this study. (b) kil mutants are impaired in E. coli lysis in liquid. Kinetics of CPRG-hydrolysis by the b-Galactosidase (expressed as Miller Units) observed after co-incubation of Myxococcus wild-type (WT) or the kil mutants with E. coli for 24 hr. M. xanthus and E. coli alone were used as negative controls. This experiment was performed independently four times. (c) The percentage of contacts with E. coli leading to a pause in motility was calculated for M. xanthus wild-type (from five independent predation movies, number of contacts observed n=807) and the kil mutants (number of contacts observed for DkilC: n=1780; DkilH: n=1219; DkilG: n=1141; DkilB: n=842; DkilK: n=710; DkilKLM: n=1446) (d) The percentage of contacts with E. coli leading to cell lysis was also estimated. In panels (b), (c), and (d), error bars represent the standard deviation of the mean. One-way ANOVA statistical analysis followed by Dunnett's posttest was performed to evaluate if the differences observed, relative to wild-type, were significant (*: p 0.05, **: p 0.01, ****: p 0.0001) or not (ns: p>0.05). The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. CPRG assay ( Figure 3b). Source data 2. Counting percentage of contacts with a prey leading to motility pauses and prey cell lysis (Figure 3c and d). However, the splitting of Tad homologs in distinct genetic clusters is a unique situation (Denise et al., 2019) and asks whether these genes encode proteins involved in the same function.
Expression analysis suggests that the cluster 1 and cluster 2 genes are expressed together and induced in starvation conditions ( Figure 3-figure supplement 2). We systematically deleted all the predicted Tad components in cluster 1 and 2 alone or in combination and measured the ability of each mutant to lyse E. coli in the CPRG colorimetric assay ( Figure 3b). All the predicted core genes, IM platform, OM secretin and associated CpaB homolog are essential for prey lysis, with the exception of the putative pre-pilin peptidase, KilA. Deletion of the genes encoding predicted pseudopilins KilL and M did not affect E. coli killing; in these conditions, pilin fibers are only partially required because deletion of KilK, the major pilin subunit, reduces the lytic activity significantly but not fully (Figure 3b). Given that the genes are organized into potential operon structures, we confirmed that the CPRG-killing phenotypes of predicted cluster 1 and cluster 2 core genes were not caused by potential polar effects (Figure 3-figure supplement 3). In liquid, predicted core gene mutants had the same propensity as wild-type to form biofilms in presence of the prey suggesting that they act downstream in the interaction process (Figure 2-figure supplement 8).
We next tested whether liquid killing and contact-dependent killing on surfaces reflected the same process. For this, we analyzed selected kil mutants: the predicted secretin (KilC), the IM platform proteins (KilH and KilG), the OM-CpaB homolog (KilB), and the pilin and pseudopilins (KilK, L, M) in contact-dependent killing at the single cell level. Prey recognition is first revealed by the induction of a motility pause upon prey cell contact (see Figure 2b). This recognition was severely impaired although not fully in secretin (DkilC), IM platform protein (DkilG) and triple pilin (DkilKMN) mutants ( Figure 3c,~8% of the contacts led to motility pauses vs~30% for the WT). In contrast, recognition was not impaired to significant levels in IM platform protein (DkilH), CpaB-homolog (DkilB) and pilin (DkilK) mutants ( Figure 3c). The potential basis of this differential impact is further analyzed in the discussion. On the contrary, prey cell plasmolysis was dramatically impacted in all predicted core components (~2% of the contacts led to prey lysis vs~26% for the WT), the only exception being the single pilin (DkilK) mutant in which prey cell lysis was reduced but still present (~13%, Figure 3d). Deletion of all three genes encoding pilin-like proteins were nevertheless affected in prey cell killing to levels observed in core component mutants. This is not observed to such extent in the CPRG assay, which could be explained by different cell-cell interaction requirements perhaps compensated by TFPs in liquid cultures. Given the prominent role of the pilins at the single cell level, the predicted pre-pilin peptidase KilA would have been expected to be essential. However, expression of the kilA gene is very low under all tested conditions (Figure 3-figure supplement 2). Prepilin peptidases are known to be promiscuous (Berry and Pelicic, 2015) and thus another peptidase (i.e. PilD, the Type IV pilus peptidase, Friedrich et al., 2014) could also process the Kil-associated pilins. This hypothesis could, however, not be tested because PilD appears essential for reasons that remain to be determined (Friedrich et al., 2014). Altogether, the data supports that the proteins from the two clusters function in starvation conditions and that they could make up a Tad-like core structure, for prey cell recognition, regulating motility in contact with prey cells, and prey killing, allowing contact-dependent plasmolysis.
Kil proteins assemble at contact sites and mediate motility regulation and killing
We next determined if the Kil proteins indeed form a single Tad-like system in contact with prey cells. To do so, the predicted ATPase (KilF) (Figure 3a) was N-terminally fused to the Neon Green (NG) and expressed from the native chromosomal locus (Figure 4a). The corresponding fusion appeared fully functional (Figure 4-figure supplement 1). In absence of prey cells, NG-KilF was diffuse in the cytoplasm. Remarkably, when Myxoccocus cells established contact with prey cells, NG-KilF rapidly formed a fluorescent-bright cluster at the prey contact site. Cluster formation was invariably followed by a motility pause and cell lysis. Observed clusters did not localize to any specific , but nevertheless KilG-NG clusters could also be observed, forming at the prey contact site immediately followed by cell lysis (Figure 4b, Video 6). These results strongly suggest that a Tad-apparatus assembled from the products of the cluster 1 and cluster 2 genes.
Additional non-core proteins are also recruited at the contact sites: downstream from kilF and likely co-transcribed, the MXAN_3108 gene (kilD, Figure 3a and The percentage of contacts with E. coli leading to NG-KilD foci formation was calculated for M. xanthus wild-type (from five independent predation movies, number of contacts observed n=807) and the kil mutants (number of contacts observed for DkilC: n=1780; DkilH: n=1219; DkilG: n=1141; DkilB: n=842; DkilK: n=710; DkilKLM: n=1446). (f) The percentage of NG-KilD foci associated with a motility pause was also estimated for M. xanthus WT (from five independent predation movies, number of NG-KilD foci observed n=198) and the kil mutants (number of NG-KilD foci observed for DkilC: n=320; DkilH: n=270; DkilG: n=251; DkilB: n=215; DkilK: n=94; DkilKLM: n=355). (g) The percentage of NG-KilD foci leading to E. coli lysis was estimated as well. In panels (d), (e), and (f), error bars represent the standard deviation to the mean. One-way ANOVA statistical analysis followed by Dunnett's posttest was performed to evaluate if the differences observed, relative to wild-type, were significant (*: p 0.05, **: p 0.01, ****: p 0.0001) or not (ns: p>0.05). The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Counting percentage of contacts with a prey leading to NG-KilD foci formation and counting percentage of NG-KilD foci associated with motility pause and prey cell lysis (Figure 4e, f and g). Video 5. NG-KilF cluster formation in contact with E. coli prey cells. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact with E. coli cells. The movie was shot at x100 magnification objective for 30 min. Pictures were taken every 30 s. https://elifesciences.org/articles/72409#video5 Video 6. KilG-NG cluster formation in contact with E. coli prey cells. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact with E. coli cells. The movie was shot at Â100 magnification objective for 9 min. Pictures were taken every 30 s. NG-KilD fusion was fully functional, also forming a fluorescent-bright cluster at a prey contact site, followed by motility arrest and prey cell lysis (Figure 4c and Figure 4-figure supplement 1, Video 7). As this protein is the most downstream component of the cluster 1 region, which facilitates further genetic manipulations (see below), we used it as a reporter for Kil system-associated functions. First, to confirm that prey intoxication occurs at sites where the Kil proteins are recruited, we imaged NG-KilD in the presence of E. coli cells labeled with TADA. As expected, PG degradation was detected at the points where the clusters are formed, showing that cluster formation correlates with contact dependent killing (Figure 4d). Using cluster assembly as a proxy for activation of the Kil system, we measured that killing is observed within~2 min after assembly, a rapid effect which suggests that Kil system assembly is tightly connected to a prey cell lytic activity (Figure 4-figure supplement 3).
We next used NG-KilD as a proxy to monitor the function of the Kil Tad apparatus in prey recognition and killing. For this, NG-KilD was stably expressed from the native chromosomal locus in different genetic backgrounds (Figure 4-figure supplement 4). In WT cells, NG-KilD clusters only formed in the presence of prey cells and~30% contacts were productive for cluster formation (Figure 4e). In kil mutants, NG-KilD clusters still formed upon prey cell contact with a minor reduction (up to twofold in DkilC and DkilK), suggesting that the Tad-like apparatus is not directly responsible for initial prey cell sensing (Figure 4e, Video 8). Nevertheless, cluster assembly was highly correlated to motility pauses ( Figure 4f); which was impaired (up to 60%) in the kil mutants (except in the pilin, DkilK ) and most strongly in the DkilC (secretin), DkilG (IM platform) and triple pilin (DkilKLM) mutants (Figure 4f). Strikingly and contrarily to WT (and except for the major pilin (kilK) mutant), cluster formation was not followed by cell lysis in all kil mutants (or very rarely, 4% of the time vs more than 80% in WT, Figure 4g). Altogether, these results indicate the Tad-like Kil system is dispensable for immediate prey recognition, but functions downstream to induce a motility pause and critically, provoke prey cell lysis.
The kil apparatus is central for Myxococcus predation
We next tested the exact contribution of the kil genes to predation and prey consumption ( Figure 5). This question is especially relevant because a number of mechanisms have been proposed to contribute to Myxococcus predation and all involve the extracellular secretion of toxic cargos (Sah and Wall, 2020;Mercier et al., 2020;Hu et al., 2016). In pure cultures, deletion of the kil genes is not linked to detectable motility and growth phenotypes, suggesting that the Tad-like Kil system mostly operates in predatory context ( Figure 5-figure supplements 1-2). Critically, core kil mutants where unable to predate colonies on plate, which could be fully complemented when the corresponding kil genes were expressed ectopically ( Figure 5a). When observed by time lapse, a kil mutant (here DkilACF) can invade a prey colony, but no prey killing is observed, showing that the prey killing phenotype is indeed due to the loss of contact-dependent killing (Figure 5b, Video 9). To measure the impact of this defect quantitatively, we developed a flow cytometry assay that directly measures the relative proportion of Myxococcus cells and E. coli cells in the prey colony across time (Figure 5c, see methods). In this assay, we observed that WT Myxococcus cells completely take over the E. coli population after 72 hr (Figure 5c). In contrast, the E. coli population remained fully viable when in contact with the DkilACF triple mutant, even after 72 hr (Figure 5c). Predatory-null phenotypes were also observed in absence of selected Tad structural components, including the secretin (KilC), the Video 7. NG-KilD cluster formation in contact with E. coli prey cells. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact with E. coli cells. The movie was shot at x100 magnification objective for 15 min. Pictures were taken every 30 s. https://elifesciences.org/articles/72409#video7 ATPase (KilF) and the IM platform protein (KilH) (Figure 5d). A partial defect was observed in the pilin (DkilK) mutant but a triple pilin deletion mutant (DkilKLM) was, however, completely deficient ( Figure 5d).
To further test whether Kil-dependent prey killing provides the necessary nutrient source, we directly imaged Myxococcus growth in prey colonies, tracking single cells over the course of 6 hr (see Materials and methods). This analysis revealed that invading Myxococcus cell grow actively during prey invasion. The Myxococcus cell cycle could be imaged directly in single cells: cell size increased linearly up to a certain length, which was followed by a motility pause and cytokinesis (Figure 5e, Video 10). The daughter cells immediately resumed growth at the same speed (Figure 5e). Cell size and cell age are therefore linearly correlated allowing estimation of a~5.5 hr generation time from a compilation of traces (Figure 5f, n=16). When the DkilACF mutant was similarly observed, cell size tended to decrease with time and cell division was not observed (Figure 5ef, n=20). Cell shortening could be a consequence of starvation, as observed for example in Bacillus subtilis (Weart et al., 2007) (although this remains to be documented in Myxococcus). Taken together, these results demonstrate the central function of the Kil Tad apparatus in prey killing and consumption.
The Kil system promotes killing of phylogenetically diverse prey bacteria
Myxococcus is a versatile predator and can attack and digest a large number of preys (Morgan et al., 2010;Müller et al., 2016). We therefore tested if the Kil system also mediates predation by contact-dependent killing of other bacterial species. To this aim, we tested evolutionarily distant preys: diderm bacteria (Caulobacter crescentus, Salmonella typhimurium and Pseudomonas aeruginosa), and a monoderm bacterium (Bacillus subtilis). In agar plate assays, M. xanthus was able to lyse all tested preys, except P. aeruginosa (Figure 6a-f). When the Kil system was deleted, the predation ability of M. xanthus was severely diminished in all cases (Figure 6a-f). Consistently, Myxococcus assembled NG-KilD clusters in contact with Caulobacter, Salmonella, and Bacillus cells, which in all cases led to cell plasmolysis (Figure 6g-i, Videos 11-13). Myxococcus cells were however unable to form lethal clusters when mixed with Pseudomonas cells (tested for n=150 contacts, an example is shown in Video 14), suggesting that the Kil system does not assemble and thus, some bacteria can evade the prey recognition mechanism.
The kil genes are present in other predatory delta-proteobacteria We next explored bacterial genomes for the presence of kil-like genes. Phylogenetic analysis indicates that the ATPase (KilF), IM platform proteins (KilH and KilG) and CpaB protein (KilB) share similar evolutionary trajectories, allowing the construction of a well-supported phylogenetic tree based on a supermatrix (Figure 7, see Materials and methods). This analysis reveals that Kil-like systems are indeed related to Tad systems (i.e. Tad systems from alpha-proteobacteria, Figure 7) but their distribution appears to be restricted to the delta-proteobacteria, specifically in Myxococcales, in Bdellovibrionales and in the recently discovered Bradymonadales. Remarkably, these bacteria are all predatory and thus, given that their predicted Kil machineries are very similar to the Myxococcus Kil system, they could perform similar functions (Figure 7, Supplementary file 2). This is possible because Bradymonadales are also thought to predate by surface motility and extracellular prey attack (Mu et al., 2020). In addition, while at first glance it may seem that Bdellovibrio species have a distinct predatory cycle, penetrating the prey cell to actively replicate in their periplasmic space (Laloux, 2019); this cycle involves a number of processes that are similar to Myxobacteria: Bdellovibrio cells also attack prey cells using gliding motility (Lambert et al., 2011) and attach to them using Type IV pili and a number of common regulatory proteins (Milner et al., 2014). Prey cell penetration follows from the ability of the predatory cell to drill a hole into the prey PG at the attachment site (Kuru et al., 2017). Although this remains to be proven directly, genetic evidence has shown that the Bdellovibrio Kil homologs are important for prey invasion and attachment (Avidan et al., 2017;Duncan et al., 2019).
Discussion
Prior to this work, Myxococcus predation was thought to involve motility, contact-dependent killing (Zhang et al., 2020b), secreted proteins, Outer Membrane Vesicles (OMVs) and antibiotics (i.e. Myxovirescin and Myxoprincomide) to kill and digest preys extracellularly (Thiery and Kaimer, 2020;Pérez et al., 2016). While a contribution of these processes is not to be ruled out, they are most likely involved in prey cell digestion (for example by degradative enzymes) rather than killing (Thiery and Kaimer, 2020), and we show here that contact-dependent killing is the major prey killing mechanism. In Myxococcus, contact-dependent killing can be mediated by several processes, now including T6SS, OME (Outer Membrane Exchange) and Kil. We exclude a function for the T6SS, for which a role in Myxococcus interspecies interactions has yet to be demonstrated. Rather, it appears that together with OME, Type-VI secretion controls a phenomenon called social compatibility, in which the exchange of toxins between Myxococcus cells prevents immune cells from mixing with non-immune cells (Vassallo et al., 2020). We have not tested a possible function of OME in prey killing because OME allows transfer of OM protein and lipids between Myxococcus cells when contact is established between identical outer membrane receptors, TraA (Sah and Wall, 2020). OME is therefore a process that only occurs between Myxococcus strains to mediate social compatibility when SitA lipoprotein toxins are delivered to non-immune TraA-carrying Myxococcus target cells (Vassallo et al., 2017).
The Kil system is both required for contact-dependent killing in liquid and on surfaces. Remarkably, proteins belonging to each motility systems show distinct requirements in liquid or on solid media. In liquid, Type-IV pili mediate prey-induced biofilm formation, which likely brings Myxococcus in close contact with the prey cells. This intriguing process likely requires EPS (since pilA mutants also lack EPS, Black et al., 2006), which deserves further exploration. On surfaces, likely a more biologically relevant context, contact-dependent killing is coupled to A-motility to penetrate prey colonies and interact with individual prey cells. The prey recognition mechanism is especially intriguing because dynamic assembly of a Tad-like system at the prey contact site is a novel observation; in general, these machineries and other Type-IV filamentous systems (Denise et al., 2019), such as TFPs tend to assemble at fixed cellular sites, often a cell pole (Mercier et al., 2020;Ellison et al., 2017;Koch et al., 2021). NG-KilD clusters do not require a functional Tad-like system to form in contact with prey cells, suggesting that prey contact induces Tad assembly via an upstream signaling cascade. Such sensory system could be encoded within the clusters 1 and 2, which contain a large number of conserved genes with unknown predicted functions (up to 11 proteins of unknown functions just considering cluster 1 and 2, Figure 3a, Supplementary file 1). In particular, the large number of predicted proteins with FHA type domains (Almawi et al., 2017; Supplementary file 1) suggests a function in a potential signaling cascade. In Pseudomonas aeruginosa, FHA domain-proteins act downstream from a phosphorylation cascade triggered by contact, allowing Pseudomonas to fire its T6SS upon contact . This mechanism is triggered by general perturbation of the Pseudomonas membrane , which could also be the case for the Kil system. Kil assembly is provoked both by monoderm and diderm bacteria, which suggests that preyspecific determinants are unlikely. Recognition is nevertheless non-universal and does not occur in contact with Pseudomonas or Myxococcus itself. Therefore, evasion mechanisms must exist, perhaps in the form of genetic determinants that shield cells from recognition.
The Kil Tad-like system itself is required to pause A-motility and for prey cell killing. Motility regulation could be indirect because differential effects are observed depending on kil gene deletions (Figures 3 and 4), suggesting that assembly of a functional Tad apparatus is not strictly required for Video 8. NG-KilD clusters form in a DkilC mutant but no motility pauses and prey cell lysis can be observed. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact E. coli cells. The movie was shot at Â100 magnification objective for 5.5 min. Pictures were taken every 30 s. https://elifesciences.org/articles/72409#video8 Figure 5. The kil genes are required for M.xanthus nutrition over prey cells. (a) The Kil system is essential for predation. Core deletion mutants in Tadlike genes, DkilC (Secretin), DkilF (ATPase), DkilH (IM platform), and DkilG (IM platform) were mixed with E. coli and spotted on CF agar plates (+ 0.07% glucose). After 24 hr of incubation, the mutants carrying the empty vector (EV) pSWU19 were strongly deficient in predation. The same kil mutants ectopically expressing the different kil genes under the control of the pilA promoter presented a restored predation phenotype similar to the WT-EV control. (b) A kil mutant can invade but cannot lyse E. coli prey colonies. mCherry-labeled WT and triple kilACF mutant are shown for comparison. Note that invading WT cells form corridors (yellow arrowheads) in the prey colony and ghost E. coli cells as well as cell debris (blue arrowheads) are left Figure 5 continued on next page regulation. In contrast, prey killing requires a functional Tad apparatus. In particular, the pilin proteins are required during prey invasion but they are dispensable (partially) in liquid cultures showing that they do not promote toxicity. In liquid, direct contacts may be enforced by TFPs in the biofilm, perhaps rendering the Tad pilins partially redundant. Tad pilin function nevertheless becomes essential on surfaces when A-motility is active. How the pilins organize to form polymers and whether they do, remains to be determined; the lack of the major pilin (KilK) is compensated by the remaining pseudo-pilins KilL and M, which is somewhat surprising given that pseudopilins are generally considered to prime assembly of major pilin polymers (Denise et al., 2019). It is currently unclear if the Kil system is also a toxinsecretion device; for example, if it also functions as a Type II secretion system. Alternatively, the Kil complex might recruit a toxin delivery system at the prey contact site. This latter hypothesis is in fact suggested by the remaining low (but still detectable) contact-dependent toxicity of the kil mutants (Figures 3 and 4). Given that Myxococcus induces prey PG degradation locally, we hypothesize that a secreted cell wall hydrolase becomes active at the prey contact site. This is not unprecedented: Bdellovibrio cells secrete a sophisticated set of PG modifying enzymes, D,Dendopeptidases (Lerner et al., 2012), L,D transpeptidases (Kuru et al., 2017) and Lysozymelike enzymes (Harding et al., 2020) to penetrate prey cells, carve them into bdelloplasts and escape. In Myxococcus, deleting potential D,Dendopeptidases (Zhang et al., 2020a) (DdacB) did not affect predation (Figure 7-figure supplement 1) which might not be surprising given behind the infiltrating Myxococcus cells. In contrast, while the DkilACF penetrates the prey colony, corridors and prey ghost cells are not observed. Scale bar = 10 mm. See corresponding Video 9 for the full time lapse. (c) The kil genes are essential for prey killing. E. coli mCherry cells were measured by flow cytometry at time 0, 24, 48, and 72 hr after the onset of predation. The E. coli survival index was calculated by dividing the percentage of 'E. coli events' at t=24, 48, or 72 hr by the percentage of 'E. coli events' at the beginning of the experiment (t=0). This experiment was independently performed three times (n=6 per strain and time point). For each sample, 500,000 events were analyzed. Each data point indicates the mean ± the standard deviation. For each time point, unpaired t-test (with Welch's correction) statistical analysis was performed to evaluate if the differences observed, relative to wild-type, were significant (***: p 0.001) or not (ns: p>0.05). (d) E. coli survival in presence of the different kil mutant strains at 48 hr. E. coli mCherry cells were measured by flow cytometry at time 0 and 48 hr after predation. This experiment was independently performed three times (n=9 per strain and time point). Events were counted as in (c) and each data point indicates the mean ± the standard deviation. One-way ANOVA statistical analysis followed by Dunnett's posttest was performed to evaluate if the differences observed, relative to wild-type, were significant (****: p 0.0001). (e, f) The kil genes are essential for Myxococcus growth on prey. (e) Cell growth during invasion. Cell length is a function of cell age during invasion and can be monitored over time in WT cells (in red). In contrast, cell length tends to decrease in a DkilACF mutant ( Video 9. A DkilACF still invades but does not kill E. coli prey cells. This movie was taken at the interface between the two colonies during invasion. The movie is a 4x compression of an original movie that was shot for 4.5 hr with a frame taken every 30 s at Â40 magnification. To facilitate tracking of Myxococcus DkilACF, cells are labeled with the mCherry fluorescent protein. https://elifesciences.org/articles/72409#video9 that Myxococcus simply lyses its preys while Bdellovibrio needs to penetrate them while avoiding their lysis to support its intracellular cycle. The Myxococcus toxin remains to be discovered, bearing in mind, that similar to synergistic toxic T6SS effectors (LaCourse et al., 2018), several toxic effectors could be injected, perhaps explaining how Myxococcus is able to kill both monoderm and diderm preys.
The Myxobacteria are potential keystone taxa in the soil microbial food web (Petters et al., 2021), meaning that Kil-dependent mechanisms could have a major impact in shaping soil ecosystems. While the Kil proteins are most similar to proteins from Tad systems, there are a number of key differences that suggest profound diversification: (i), the Kil system involves a single ATPase and other Tad proteins such as assembly proteins TadG, RcpB, and pilotin TadD are missing Denise et al., 2019;(ii), several Kil proteins have unique signatures, the large number of associated genes of unknown function; in particular, the over-representation of associated FHA domain proteins, including the central hexameric ATPase KilF itself fused to an N-terminal FHA domain. The KilC secretin is also uniquely short and lacks the N0 domain, canonically found in secretin proteins (Tosi et al., 2014), which could be linked to increased propensity for dynamic recruitment at prey contact sites. Future studies of the Kil machinery could therefore reveal how the contact-dependent properties of Tad pili were adapted to prey cell interaction and intoxication, likely a key evolutionary process in predatory bacteria.
Materials and methods
Bacterial strains, growth conditions, motility plates, genetic constructs, and western blotting See Supplementary files 3, 4 and 5 corresponding to Table 3: strains, Table 4: plasmids and Table 5: primers. E. coli cells were grown under standard laboratory conditions in Luria-Bertani (LB) broth supplemented with antibiotics, if necessary. M. xanthus strains were grown at 32˚C in CYE (Casitone Yeast Extract) rich media as previously described (Zhang et al., 2020a). S. enterica Typhimurium, B. subtilis, and P. aeruginosa were grown overnight at 37˚C in LB. C. crescentus strain NA1000 was grown overnight at 30˚C in liquid PYE (Peptone Yeast Extract). Motility plate assays were conducted as previously described on soft (0.3%) or hard (1.5%) agar CYE plates (Bustamante et al., 2004).
The deletion strains and the strains expressing the different Neon Green fusions were obtained using a double-recombination strategy as previously described (Bustamante et al., 2004;Shaner et al., 2013). Briefly, the kil deletion alleles (carrying~700-nucleotide long 5' and 3' flanking sequences of M. xanthus locus tags) were Gibson assembled into the suicide plasmid pBJ114 (galK, Kan R ) and used for allelic exchange. Plasmids were introduced in M. xanthus by electroporation. After selection, clones containing the deletion alleles were identified by PCR. Using the same strategy, 'Neon Green fusion' alleles were introduced at kilD and kilF loci. The corresponding strains expressed, under the control of their native promoters, N-terminal Neon Green fusions of KilD and KilF.
For complementation of DkilC, DkilF, DkilG, and DkilH strains, we used the pSWU19 plasmid (Kan R ) allowing ectopic expression of the corresponding genes from the pilA promoter at Mx8-att site. The same strains transformed with the empty vector were used as controls.
Video 10. Predatory cells division and tracking during invasion of prey colony. To follow cell growth and division at the single cell level during prey invasion, WT cells were mixed with a WT strain expressing the mCherry at a 50:1 ratio and imaged every 30 s at Â40 magnification for up to 10 hr within non-labeled prey colonies. Cell growth was measured by fitting cell contours to medial axis model followed by tracking under Microbe-J. Real time of the track for the example cell: 95 min.
https://elifesciences.org/articles/72409#video10 To evaluate if M. xanthus kil mutant had lost the ability to lyse by direct contact different preys, prey-cell suspensions were directly mixed with M. xanthus WT or DkilACF and spotted on CF agar (+ 0.07% glucose). After 24 and 48 hr of incubation, pictures of the spots corresponding to the different predator/prey couples were taken. Note that Pseudomonas aeruginosa is resistant in this assay. Scale bar = 3.5 mm. (b, c, d, e, f) Prey cell survival upon predation was evaluated by CFU counting. The different preys (Kan R ) were mixed with M. xanthus WT (blue circles) or DkilACF (orange circles) strains and spotted on CF agar (+ 0.07% glucose). Spots were harvested after 0, 8, 24, and 48 hr of predation, serially diluted and platted on agar plates with kanamycin for CFU counting. The prey alone (black circles) was used as a control. Two experimental replicates were used per time point. This experiment was independently performed three times. Error bars represent the standard deviation to the mean. To express KilG C-terminally fused to Neon Green, a pSWU19-PpilA-kilG-NG was created and transformed in the DkilG strain.
Western blotting was performed as previously described (Bustamante et al., 2004) using a commercial polyclonal anti Neon-Green antibody (Chromotek).
Growth rate comparison in liquid cultures
To compare growth rates of M. xanthus WT and DkilACF strains, overnight CYE cultures were used to inoculate 25 ml of CYE at OD 600 = 0.05. Cultures were then incubated at 32˚C with a shaking speed of 160 rpm. To avoid measuring cell densities at night, a second set of cultures were inoculated 12 hr later at OD 600 = 0.05. Every 4 hr, 1 ml sample of each culture was used to measure optical densities at 600 nm with a spectrophotometer. The different measurements were then combined into a single growth curve. This experiment was performed with three independent cultures per strain.
Predation assay on agar plates Prey colony invasion on CF agar plates M. xanthus and E. coli were respectively grown overnight in 20 ml of CYE at 32˚C and in 20 ml of LB at 37˚C. The next day, cells were pelleted and resuspended in CF medium (MOPS 10 mM pH 7.6; KH 2 PO 4 1 mM; MgSO 4 8 mM; (NH 4 ) 2 SO 4 0.02%; Na citrate 0.2%; Bacto Casitone 0.015%) to a final OD 600 of 5. 10 ml of M. xanthus and prey cell suspensions were then spotted next to each other (leaving less than 1 mm gap between each spot) on CF 1.5% agar plates with or without 0.07% glucose (to allow minimal growth of the prey cells) and incubated at 32˚C. After 48 hr incubation, pictures of the plates were taken using a Nikon Olympus SZ61 binocular loupe (x10 magnification) equipped with a camera and an oblique filter. ImageJ software was used to measure the surface of the prey spot lysed by M. xanthus.
Spotting predator-prey mixes on CF agar plates
To force the contact between M. xanthus and a prey, mixes of predator/prey were made and spotted of CF agar plates. 200 ml of a prey cell suspension (in CF, OD 600 = 5)were mixed with 25 ml of a M. xanthus cell suspension (in CF, OD 600 = 5) and 10 ml of this mix were spotted on CF agar plates supplemented with 0.07% glucose. As described above, pictures of the plates were taken after 24 hr incubation.
Video 11. NG-KilD cluster formation in contact with Caulobacter crescentus. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact with a C. crescentus cell. The movie was shot at Â100 magnification objective for 7 min. Pictures were taken every 30 s. https://elifesciences.org/articles/72409#video11 Video 12. NG-KilD cluster formation in contact with Salmonella typhimurium. Shown is an overlay of the fluorescence and phase contrast images of a motile Myxococcus cell in predatory contact with an S. enterica Typhimurium cell. The movie was shot at Â100 magnification objective for 20 min. Pictures were taken every 30 s. https://elifesciences.org/articles/72409#video12 Visualizing prey colony invasion and contact-dependent killing by microscopy Prey colony invasion on CF agar pads Prey invasion was imaged by microscopy using the Bacto-Hubble system the specific details of the Method are described elsewhere (Panigrahi, 2020). Briefly, cell suspensions concentrated to OD 600 =5 were spotted at 1 mm distance onto CF 1.5% agar pads and a Gene Frame (Thermo Fisher Scientific) was used to sandwich the pad between the slide and the coverslip and limit evaporation of the sample. Slides were incubated at 32˚C for 6 hr before imaging, allowing Myxococcus and E. coli to form microcolonies. Time-lapse of the predation process was taken at Â40 or Â100 magnification. Movies were taken at the invasion front where Myxococcus cells enter the E. coli colony. To facilitate tracking, M. xanthus cells were labeled with fluorescence (Ducret et al., 2013). Fluorescence images were acquired by microscopy every 30 s for up to 10 hr, at room temperature (see below for experimental details of time lapse acquisitions ).
Spotting predator-prey mixes on CF agar pads
To image contact-dependent killing between M. xanthus and prey cells (E. coli, C. crescentus, B. subtilis, S. typhimurium, and P. aeruginosa), cells were grown as described above, pelleted and resuspended in CF medium to a final OD 600 of 1. Equal volumes of M. xanthus and prey cell suspensions were then mixed together and 1 ml of the mix was spotted on a freshly made CF 1.5% agar pad on a microscope slide. After the spot has dried, the agar pad was covered with a glass coverslip, and incubated in the dark at room temperature for 20-30 min before imaging.
Time-lapse experiments were performed using two automated and inverted epifluorescence microscope: a TE2000-E-PFS (Nikon), with a Â100/1.4 DLL objective and an ORCA Flash 4.0LT camera (Hamamatsu) or a Ti Nikon microscope equipped with an ORCA Flash 4.0LT camera (Hamamatsu). Theses microscopes are equipped with the 'Perfect Focus System' (PFS) that automatically maintains focus so that the point of interest within a specimen is always kept in sharp focus at all times, despite any mechanical or thermal perturbations. Images were recorded with NIS software from Nikon. All fluorescence images were acquired with appropriate filters with a minimal exposure time to minimize photo-bleaching and phototoxicity effects: 30 min long time-lapses (one image acquired every 30 s) of the predation process were taken at x100 magnification. DIA images were acquired using a 5 ms light exposure and GFP fluorescent images were acquired using a 100 ms fluorescence exposure with power intensity set to 50% (excitation wavelength 470 nm) to avoid phototoxicity.
Labelling E. coli cells with the fluorescent D-amino acid TADA Lyophilized TADA (MW = 381.2 g/mol, laboratory stock Faure et al., 2016) was re-suspended in DMSO at 150 mM and conserved at À20˚C. The labeling was performed for 2 hr in the dark at room temperature, using 2 ml of the TADA solution for 1 ml of cells culture (OD 600 = 2). Cells were then washed four times with 1 ml of CF and directly used for predation assays on agar pad.
Image analysis
Image analysis was performed under FIJI (Schindelin et al., 2012) and MicrobeJ an ImageJ plug-in for the analysis of bacterial cells.
Semantic segmentation of Myxococcus cells
This was performed using the newly developed MiSiC system, a deep learning-based bacterial cell segmentation tool (Panigrahi, 2020). The system was used in semantic segmentation mode and annotated manually to reveal E. coli lysing cells.
Kymograph construction
Kymographs were obtained after manual measurements of fluorescence intensities along FIJI handdrawn segments and the FIJI-Plot profile tool. The measurements were then exported into the Prism software (Graphpad, Prism 8) to construct kymographs.
Cell tracking
Cell tracking and associated morphometrics were obtained using MicrobeJ. Image stacks were first processed, stabilized and filtered with a moderate Gaussian blur and cells were detected by thresholding and fitted with the Plug-in 'medial axis' model. Trajectories were systematically verified and corrected by hand when necessary.
Tracking Myxococcus motility pauses , NG-KilD foci formation and prey cell lysis during predation In 30 min time-lapses, contacts between prey cells and Myxococcus cells were scored. Pauses were counted when the predatory cell stopped all movement upon contact with the prey. We also counted if these contacts lead to the formation of NG-KilD foci and to cell lysis. Thus, for a determined E. coli cell, we scored the number of contacts with Myxococcus, the number of pauses these contacts induces in M. xanthus motility, the number of NG-KilD foci formed upon contacts and, ultimately, the lysis of the cell. Five independent movies were analyzed for each strain and the percentage of contacts leading to a pause in motility, NG-KilD foci formation and cell lysis was calculated. We also estimated the percentage of NG-KilD clusters leading to cell lysis.
Tracking cluster time to lysis
Time to lysis measures the elapsed time between cluster appearance to prey cell death. Data were obtained from two biological replicates.
CPRG assay for contact-dependent killing in liquid CPRG assay in 24-well plates M. xanthus and E. coli cultures were grown overnight, pelleted and resuspended in CF at OD 600~5 . 100 ml of M. xanthus cell suspension (WT and mutants) were mixed with 100 ml of E. coli cell suspension in a 24-well plate. In each well, 2 ml of CF medium supplemented with CPRG (Sigma Aldrich, 20 mg/ml) and IPTG (Euromedex, 50 mM) were added to induce lacZ expression. The plates were then incubated at 32˚C with shaking and pictures were taken after 24 and 48 hr of incubation. To test the contact-dependance, a two-chamber assay was carried out in a Corning 24 well-plates containing a 0.4 mm pore polycarbonate membrane insert (Corning Transwell 3413). This membrane is permeable to small metabolites and proteins and impermeable to cells. E. coli cells were inoculated into the top chamber and M. xanthus cells into the bottom chamber.
CPRG assay in 96-well plates
To evaluate the predation efficiency of the different kil mutants, the CPRG assay was adapted as follow: wild-type M. xanthus and the kil mutant strains were grown overnight in 15 ml of CYE. E. coli was grown overnight in 15 ml of LB. The next morning, M. xanthus and E. coli cells were pelleted and resuspended in CF at OD 600 = 0.5 and 10, respectively. To induce expression of the b-galactosidase, IPTG (100 mM final) was added to the E. coli cell suspension.
In a 96-well plate, 100 ml of M. xanthus cell suspension were mixed with 100 ml of E. coli cell suspension. Wells containing only M. xanthus, E. coli or CF were used as controls. The lid of the 96-well plate was then sealed with a breathable tape (Greiner bio-one) and the plate was incubated for 24 hr at 32˚C while shaking at 160 rpm. In this setup, we observed that M. xanthus and E.coli cells aggregate at the bottom of the well and therefore come in direct contact, favoring predation in liquid.
The next day, the plate was centrifuged 10 min at 4800 rpm and 25 ml of the supernatant were transferred in a new 96-well plate containing 125 ml of Z-buffer (Na 2 HPO 4 60 mM, NaH 2 PO 4 40 mM, KCl 10 mM pH7) supplemented with 20 mg/ml of CPRG. After 15-30 min of incubation at 37˚C, the enzymatic reaction was stopped with 65 ml of Na 2 CO 3 (1 M) and the absorbance at 576 nm was measured using a TECAN Spark plate reader.
This experiment was performed independently four times. For Miller unit calculation, after absorbance of the blank (with CF) reaction was subtracted, the absorbances measured at 576 nm were divided by the incubation time and the volume of cell lysate used for reaction. The resulting number was then multiplied by 1000.
Crystal violet biofilm staining
In a 96-well plate, 100 ml of M. xanthus cell suspension (in CF, OD 600 = 0.5) were mixed with 100 ml of E. coli cell suspension (in CF, OD 600 = 10) and incubated for 24 hr at 32˚C while shaking at 160 rpm. The next day, the supernatant was carefully removed and the wells were washed with 200 ml of CF twice. Then, 100 ml of a 0.01% crystal violet solution were added to each well and incubated for 5 min. Wells were washed twice with 200 ml of water before imaging.
Prey CFU counting after predation E. coli, S. typhi, P. aeruginosa, and B. subtilis kanamycin resistant strains were grown at 37˚C in liquid LB supplemented with kanamycin (50 or 10 mg/ml). C. crescentus kanamycin resistant strain was grown at 30˚C in liquid PYE supplemented with kanamycin (25 mg/ml). Wild-type and DkilACF M. xanthus strains were grown at 32˚C in liquid CYE. Cells were then centrifuged and pellets were resuspended in CF at an OD 600 of 5. 25 ml of M. xanthus cell suspensions and 200 ml of prey cell suspensions were then mixed together and 10 ml were spotted on CF agar plates supplemented with 0.07% glucose. After drying, plates were incubated at 32˚C. At 0, 8, 24, and 48 hr time points, spots were harvested with a loop and resuspended in 500 ml of CF. This solution was then used to make 10-fold serial dilutions in a 96-well plate containing CF. At the exception of C. crescentus, 5 ml of each dilution were spotted on LB agar plates supplemented with 10 mg/ml of kanamycin and incubated at 37˚C for 24 hr. C. crescentus dilutions were spotted on PYE agar plates supplemented with 25 mg/ml of kanamycin and incubated at 30˚C for 24 hr. The next day, colony-forming units were counted and the number of prey cells that survived in the predator/prey spot was calculated.
Measurements of E. coli killing by flow cytometry M. xanthus strains (wild-type and kil mutants) constitutively expressing GFP were grown overnight in liquid CYE without antibiotics. E. coli mCherry (prey) was grown overnight in liquid LB supplemented with ampicillin (100 mg/ml). The next morning, optical densities of the cultures were adjusted in CF medium to OD 600 =5. M. xanthus GFP and E. coli mCherry cell suspensions were then spotted onto fresh CF 1.5% agar plates as previously described (Bustamante et al., 2004). Briefly, 10 ml drops of the prey and the predator cell suspensions were placed next to each other and let dry. Inoculated plates were then incubated at 32˚C. Time 0 corresponds to the time at which the prey and the predator spots were set on the CF agar plate. At time 0, 24, 48, and 72 hr (post predation) and for each M. xanthus strain, two predator/prey spot couples were harvested with a loop and resuspended in 750 ml of TPM. To fix the samples, paraformaldehyde (32% in distilled water, Electron Microscopy Sciences) was then added to the samples to a final concentration of 4%. After 10 min incubation at room temperature, samples were centrifuged (8 min, 7500 rpm), cell pellets were then resuspended in fresh TPM and optical densities were adjusted to OD 600~0 .1.
Samples were then analyzed by flow cytometry on a Bio-Rad S3e Cell Sorter and data were processed using ProSort and FlowJo softwares. For each sample, a total population of 500,000 events was used and corresponds to the sum of M. xanthus-GFP and E. coli-mCherry events. A blue laser (488 nm, 100 mW) was used for detection of forward scatter (FSC) and side scatter (SSC) and for excitation of GFP. A yellow-green laser (561 nm, 100 mW) was used for excitation of mCherry. GFP and mCherry signals were collected using, respectively, the emission filters FL1 (525/30 nm) and FL3 (615/25 nm) and a compensation was applied on the mCherry signal. Samples were run using the low-pressure mode (~10,000 particles/s). The density plots obtained (small angle scattering FSC versus wide angle scattering SSC signal) were first gated on the overlapped populations of M. xanthus and E. coli, then filtered to remove the multiple events and finally gated for high FL1 signal (M. xanthus-GFP) and high FL3 signal (E. coli-mCherry).
Bioinformatics analyses Homology search strategy
We used several search strategies to identify all potential homologous proteins of the Kil system: we first used BLAST (Camacho et al., 2009;Altschul et al., 1997) to search for reciprocal best hits (RBH) between the M. xanthus and the B. bacteriovorus and B. Sediminis Kil systems, as well as the C. crescentus Tad system, identifying bona fide orthologs between the three species. We limited the search space to the respective proteomes of the three species. We then used HHPRED (Hildebrand et al., 2009) to search for remotely conserved homologs in B. bacteriovorus using the proteins from the two operons identified in M. xanthus. Finally, we performed domain comparisons between proteins from the B. bacteriovorus and B. sediminis Kil operons and C. crescentus Tad system to identify proteins with similar domain compositions in M. xanthus. Identified orthologs or homologs between the three species, the employed search strategy, as well as resulting e-values are shown in Supplementary file 2 (Table 2): M. xanthus proteins with homologs identified in B. bacteriovorus HD100, C. crescentus CB15 and B. sediminis.
Structure predictions
Tertiary structural models of the secretin and the cytoplasmic ATPase were done using Phyre2 (Kelley et al., 2015) or SWISS-MODEL (Waterhouse et al., 2018), in both cases using default parameters. Quaternary models were generated using SWISS-MODEL. Structural models were displayed using Chimera (Pettersen et al., 2004) and further processed in Illustrator .
Phylogenetic analyses
We used the four well-conserved Kil system components for phylogenetic analysis. To collect species with secretion systems similar to the Kil system, we first used MultiGeneBLAST (Medema et al., 2013) with default parameters. Orthologs of the four proteins from B. bacteriovorus, B. Sediminis and C. crescentus from closely related species were added manually. We aligned each of the four proteins separately using MAFFT (Katoh et al., 2002) and created a supermatrix from the four individual alignments. Gblocks (Katoh et al., 2002) using relaxed parameters was used prior to tree reconstruction to remove badly aligned or extended gap regions. . Alignments of individual trees were also trimmed using Gblocks. PhyML (Guindon et al., 2010) was used for tree reconstruction, using the JTT model and 100 bootstrap iterations. Trees were displayed with Dendroscope (Huson and Scornavacca, 2012) and further processed in Illustrator . The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. The following previously published dataset was used:
Data availability
Author ( | 2021-03-02T14:22:39.861Z | 2021-02-25T00:00:00.000 | {
"year": 2021,
"sha1": "558b7a6c29eed84e5c4537406239912d24518ef9",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.7554/elife.72409",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "23e36323d703904a6f8bac2a9408dcfedd9a3fe9",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
37372136 | pes2o/s2orc | v3-fos-license | Effective Range of Percutaneous Posterior Full-Endoscopic Paramedian Cervical Disc Herniation Discectomy and Indications for Patient Selection.
The objective was to investigate the effective and safe range of paramedian CDH by percutaneous posterior full-endoscopy cervical intervertebral disc nucleus pulposus resection (PPFECD) to provide a reference for indications and patient selection. Sixteen patients with CDH satisfied the inclusion criteria. Before surgery the patients underwent cervical spine MRI, and the distance between the dural sac and herniated disc was measured. An assessment was performed by MRI immediately after surgery, measuring the distance between dural sac and medial border of discectomy (DSMD). The preoperative average distance between the dural sac and peak of the herniated disc (DSPHD) was 3.87 ± 1.32 mm; preoperative average distance between dural sac and medial border of herniated disc (DSMHD) was 6.91 ± 1.21 mm and an average distance of postoperative DSMD was 5.41 ± 1.40 mm. Postoperative VAS of neck and shoulder pain was significantly decreased but JOA was significantly increased in each time point compared with preoperative ones. In summary, the effective range of PPFECD to treat paramedian CDH was 5.41 ± 1.40 mm, indicating that DSMHD and DSPHD were within 6.91 ± 1.21 mm and 3.87 ± 1.32 mm, respectively. PPFECD surgery is, therefore, a safe and effective treatment option for patients with partial paramedian cervical disc herniation.
Introduction
Cervical disc herniation (CDH) is one of the most common degenerative spinal disorders and is characterized by upper extremity pain and neurological deficits. Depending on the site of the intraspinal disc herniation, CDH can be divided into three categories: median, paramedian, and lateral herniations [1]. There are several conservative treatments currently in use which have achieved therapeutic outcomes, such as medication with steroidal and nonsteroidal antiinflammatory drugs and physical therapy; however, surgical interventions can be necessary for patients with severe cervical radiculopathy and myelopathy [2,3]. Posterior laminoforaminotomy access to the cervical spine was developed in the early 1940s [4], whereas anterior access for the operation of cervical disc changes was described in the late 1950s [5]. Additional surgical approaches arising from posterior and anterior access have also been explored, such as anterior cervical decompression without fusion, anterior foraminotomy, posterior microscope-assisted or endoscope assisted "keyhole foraminotomy," and cervical disc replacement [6,7]. The anterior cervical discectomy and fusion (ACDF) procedure is commonly regarded as the most successful approach to CDH treatment since it can maximally attenuate herniated disc compression and maintain anterior stability [8,9]. In most cases, these techniques will result in satisfactory patient outcomes; however, the ACDF procedure, owing to a high fusion rate, is considered the gold standard for the treatment of CDH [10,11]. Given the high frequency of ACDF operations, there have been reports of complications stemming 2 BioMed Research International from this procedure, such as the formation of pseudarthrosis, severe degeneration of adjacent segments, height reduction of the intervertebral space (IVS), and motion loss of the cervical spine fusion [12,13].
In an effort to reduce such surgery-related complications, there has been a push to develop advanced endoscopic techniques, of which full-endoscopic cervical discectomy (FECD) has been extensively used. FECD falls into two main categories: anterior (AFECD) and posterior (PFECD). AFECD is mostly used for patients with median CDH, which can effectively resect a protruded intervertebral disc to decompress the spinal cord. However, this is a risky approach given the proximity of major blood vessels and nerves [14]. Furthermore, damaged intervertebral disc tissues stemming from AFECD can lead to spinal instability and negatively impact postoperative recovery. By comparison, PFECD is generally considered a safer procedure for patients with the lateral CDH. One of the key advantages of PFECD is that it tends not to disturb the intervertebral disc [15,16]. In this report, we demonstrate that PFECD can be used to treat paramedian CDH when a partially protruding nucleus pulposus, close to the front of the cord, is properly removed. The rationale for this study was to investigate the effective and safe range of paramedian cervical discectomy with percutaneous posterior full-endoscopy, which can serve as a reference for its surgical indications.
Patient Characteristics.
Sixteen patients (seven female, nine male) were recruited for this study. The patients presented with paramedian CDH, as defined by a paramedian herniation that pressed the spinal cord unilaterally and deformed it into a comma shape, pressing the spinal cord and nerve root and then generating myeloradiculopathy symptom [1]. All the patients, whose ages ranged from 26 to 62 years (mean: 42 years), underwent percutaneous posterior FECD between August 2015 and September 2016. The reported duration of pain ranged from 1 to 78 months (mean: 13 months), and preoperative neurologic presentation included myelopathy in ten patients and myeloradiculopathy in six patients. The operations targeted levels C4-C5 in one patients, C5-C6 in six patients, and C6-C7 in nine patients.
Inclusion
Criteria. The inclusion criteria were as follows: (1) having received failed conservative treatment lasting for more than 4 weeks or symptoms deteriorating to the extent of becoming unbearable; (2) neurological symptoms (myelopathy and/or myeloradiculopathy) consistent with the preoperative magnetic resonance image; and (3) single-level paramedian disc herniation. The exclusion criteria were as follows: (1) clear segmental instabilities or deformities; (2) cervical intervertebral disc with calcification; (3) isolated neck pain for which the cause could not be determined by magnetic resonance imaging (MRI); (4) foraminal stenosis without disc herniation; (5) multiple-level disc herniation; (6) previous surgery at the same segment; and (7) a suspected infection or tumor in the cervical spine.
Preoperative Evaluation.
The examinations were performed by two surgeons with experience in this technique. The vertical distance between the lateral border of the dural sac and peak of the herniated disc (DSPHD) and the distance between the lateral border of the dural sac and the intersection of the dural sac and medial border of the herniated disc (DSMHD) were recorded by two independent doctors from MRI images ( Figure 1). The final values were calculated as the average of triplicate measurements from each doctor. In addition, the Visual Analog Scale (VAS) was used to determine neck and arm pain and the modified Japanese Orthopedic Association (JOA) scoring system to determine functional status.
Operative Technique.
Operations were performed under general anesthesia with the patients placed in a prone position with heightening at the chest to keep the neck flexed. The patients' shoulders were immobilized with tape and the arms placed caudally on the body with gentle tension to aid the fluoroscopic visualization of cervical levels. The line of spinal joints was marked using posterior-anterior radiography guidance, whereas the operations were guided by lateral radiography. Once the location of the cervical segment had been accurately determined, a skin incision was made and a dilator with a 6.9 mm outer diameter bluntly inserted into the facet joint. The operation sheath was inserted via the beveled opening of the operation performed under visual control and the site was irrigated continuously with 0.9% saline solution. The facet joint was completely exposed and grinded with a high-speed grinding drill. The lateral ligamentum flavum was resected to expand the intervertebral foramen to allow the endoscope to penetrate into the spinal canal, after which the herniated disc tissue was resected. The nucleus pulposus of the cervical intervertebral disc was ablated by radiofrequency (RF) ablation; the mobilization of the nerves root was repeatedly checked ( Figure 2). Finally, all instruments were removed and the incision closed by suturing. Operation times, bleed volumes, and intraoperative complications for each patient were recorded.
2.5. Follow-Up. All patients were followed up 3, 28, 90, and 180 days after surgery, each patient receiving a questionnaire by mail four working days ahead of their attendance at the clinic. The follow-up examinations were conducted by two physicians, neither of whom had been involved in the operations. Besides general parameters, other relevant information was collected using the following evaluations: the modified Macnab criteria were used to evaluate the postoperative outcomes, whereas VAS and JOA scores were recorded at the final follow-up visit [17,18]. MRI scans were taken of each patient at day 3, and the distance between the lateral border of the dural sac and the intersection of the dural sac and medial border of discectomy (DSMD) were measured from the MRIs of the cervical spine in the transverse plane ( Figure 3). These measurements were done in triplicate by two doctors who had not been involved with the operations, and the average value of these six measurements constituted the final data points. 2.6. Statistical Analysis. Statistical analysis was performed using the Statistical Package for the Social Sciences (ver. 18.0, SPSS, Chicago, IL, USA). The Tamhane test and Dunnett test were applied to compare pre-and postoperative VAS and JOA scores at various times. The differences between pre-and postoperative distance measurements were analyzed using a paired sample -test. In all analyses, a probability < 0.05 was considered significant. Results were presented as a mean ± standard deviation.
Perioperative Complications.
None of the patients experienced any preoperative or postoperative complications, such as postoperative bleeding, injury to the nerve or dura, damage to the spinal cord with hemi-/paraparesis or paralysis of the upper extremities. There were not any complications from infection, spondylodiscitis, or thrombosis. Deterioration of existing symptoms was not observed in any of the patients. Table 1: Pre-and postoperation measurements of distances between the dural sac and herniated disc and distances between the dural sac and the medial border of discectomy (mean ± SD).
Indicators
DSPHD (mm) DSMHD (mm) DSMD (mm) Measured value 3.87 ± 1.32 6.91 ± 1.21 5.41 ± 1.40 DSPHD: the vertical distance between lateral border of dural sac and peak of herniated disc; DSMHD: the vertical distance between lateral border of dural sac and intersection of dural sac and medial border of herniated disc; DSMD: the vertical distance between lateral border of dural sac and intersection of dural sac and medial border of discectomy.
significant difference (P < 0.05). The difference in JOA score at day 3, when compared with JOA scores at days 28, 90, and 180, was significant (P < 0.05), whereas the differences amongst the JOA scores at days 28, 90, and 180 failed to reach significance (P > 0.05) ( Table 2). Applying the modified Macnab criteria [19] to evaluate the curative effect 6 months after surgery, 13 cases were deemed to be excellent and 3 cases deemed to be good.
Discussion
Surgical intervention represents the most efficacious clinical alternative for CDH cases that fail to respond to conservative treatments. An operation by a surgeon can effectively relieve the spinal cord or nerve root compression and promote the recovery of its function, thus achieving a significant improvement in clinical symptoms [19][20][21]. In recent years, the application of minimally invasive surgical techniques has reduced many of the unfavorable factors associated with traditional open surgery, such as tissue trauma and excessive bleeding and considerable risk of nervous, parenchymal, and vascular lesions which was associated with an increased hospital stay [22][23][24]. Additional benefits include reduced recovery time in bed after surgery and lower incidents of severe complications, such as hypostatic pneumonia or deep venous thrombosis of the lower limbs. There is less impact on the muscles and nuchal ligaments attached to the vertebral plate and spinous process via sequential dilation, which would otherwise need to be isolated during the operation [22,25]. This results in reduced postoperative scar tissue formation which in the past could lead to persistent pain and discomfort in the back of the neck and a faster return to work [7,26,27]. With percutaneous posterior endoscopic cervical intervertebral disc nucleus pulposus resections, the inner margin of the articular process is grinded with an abrasive drill to open a hole into the spinal canal. This greatly reduces damage to posterior ligaments, muscles, and bone, while retaining maximal biomechanical stability of the cervical vertebrae. A follow-up study of 87 patients found that two years after receiving percutaneous PFECD (PPFECD) operations, 87.4% of the patients reported no recurrence of neck or shoulder pain, and only 9.2% experienced occasional pain. Although the decompression results of PPFECD were similar to conventional ACDF, the operation-related traumatization was reduced [28]. PPFECD is now considered a safer and more effective treatment for cervical intervertebral disc herniation when compared with conventional ACDF and has advantages when compared with the anterior percutaneous endoscopic cervical intervertebral disc nucleus pulposus extirpation operation (PAFECD), such as reduction of the volume of disc removal, the length of hospital stays, and the postoperative radiographical changes [29]. The latter procedure requires the surgeon to go through the intervertebral disc, leading to unavoidable damage to the intervertebral disc which can cause issues such as postoperative accelerated disc degeneration, cervical instability, and loss of physiological flexibility. A two-year follow-up study of 103 patients having received APECD found that up to 12% of the patients had significantly decreased intervertebral height, increased incidence of cervical kyphosis, and occasional arm pain [30].
By comparison, with PPFECD there was no intervertebral disc damage and less damage to the surrounding muscles, ligaments, and zygapophysis due to the use of an intraoperative channel through the back of the neck. The result was no aggravation of cervical kyphosis deformity or postoperative decrease in intervertebral space height [31]. In the present study, sixteen patients underwent PPFECD surgery to remedy paramedian cervical disc herniation. All sixteen operations were successful and proceeded without any complications. The curative effects were deemed to be satisfactory, as measured by all patients reporting an absence or significantly reduced postoperative neck pain, eliminating the need for oral analgesics by the end of a 6-month follow-up study.
Previously, PPFECD has been applied to CDH patients whose symptoms included radiculopathy with upper extremity numbness and pain [16]. Since the use of this technique is less common to treat multisegmental CDH, we limited our study to patients with single segmental cervical intervertebral disc herniation. By applying a gentle and intermittent surgical procedure, we could remove the cervical intervertebral disc protruding from the inner side of the dural sac and compressing the spinal cord; this further expanded the resection range. In the observation period following the operation, symptoms of numbness and fatigue of one side of the body and upper and lower limbs and walking instability improved significantly. Our results indicate that PPFECD is an effective treatment option for spinal disorders caused by CDH which offers significant scope to investigate the resection range of PPFECD.
We selected the edge of dural sac as a position marker for primary reasons: (i) it has a clear boundary which makes it a good point of reference and (ii) the dural sac is rarely affected by the surgery so that risk of damage to the spinal cord is negligible. The removal of the cervical intervertebral disc protruding from inside the dural sac can increase the risk of spinal cord injury, due to stretching of the sac. Consequently, one of the key objectives of this study was to explore the range of resection possible with the PPFECD technique on cervical intervertebral disc herniating from inside the dural sac.
In the postoperation follow-up, we found that the distance between the edge of the dural sac and the inside edge of the intervertebral disc was significantly smaller than between the edge of the dural sac and the inside edge of the herniated disc. In other words, the resected amount of actual intervertebral disc tissues was less than that of the preoperative measurements. A possible explanation for this could be that the nucleus pulposus, not otherwise protruding from the disc, was RF ablated, causing a decrease in the pressure inside the intervertebral disc thereby retracting the remaining herniated disc. Postoperative symptoms of all patients were improved at all cervical levels, and cervical MRIs showed no evidence of protrusions compressing the spinal cord or nerves in any of the patient. At the final 6month follow-up, there were no reported complications, such as spinal cord injury or dural sac rupture. Based on the analysis of all the experimental data, we propose that the application of PPFECD for cervical disc herniation up to 6.91 ± 1.21 mm and peak of herniated disc up to 3.87 ± 1.32 mm is safe and effective and that the safe resection range of cervical intervertebral disc using PPFECD is up to 5.41 ± 1.40 mm within the border of the dural sac.
In our study, all patients recovered fully from the operation, experiencing no serious or even minor complications and, most importantly, preoperative symptoms of pain, numbness, and fatigue were completely relieved at the six-month endpoint of the study. There were no reported symptoms of spinal cord injury, such as sensory disturbance, muscle weakness, pathological reflex, or defecation incontinence. PPFECD offers advantages such as smaller incisions, less tissue damage, and adequate nerve root decompression which result in faster postoperative recovery, with fewer complications, shorter hospital stays, and, therefore, reduced cost. These are benefits that point to PPFECD as a safe and effective surgical procedure for treating cervical disc herniation. With a more in-depth understanding of spinal anatomy and refinement of surgical techniques, the indication and application scope of PPFECD could be expanded even further, thus exploring new space for the treatment of paramedian type CDH.
Conclusion
In this study, we have found that the effective range was 5.41 ± 1.40 mm from the border of the dural sac for percutaneous posterior full-endoscopic cervical disc herniation discectomy treated paramedian cervical disc. For patients with partial paramedian cervical disc herniation was up to 6.91 ± 1.21 mm and peak of herniated disc up to 3.87 ± 1.32 mm within the lateral border of dural sac. PPFECD surgery is a safe and effective treatment option.
Conflicts of Interest
No financial conflicts of interest exist. | 2018-04-03T05:48:59.092Z | 2017-10-26T00:00:00.000 | {
"year": 2017,
"sha1": "b4599d8b4fc906028ce72c6f1f7250497b670866",
"oa_license": "CCBY",
"oa_url": "http://downloads.hindawi.com/journals/bmri/2017/3610385.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "b8196f606e0510699a980f5d331c7cfe7a2197b8",
"s2fieldsofstudy": [
"Engineering",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
257437040 | pes2o/s2orc | v3-fos-license | Observations of Player (de)Selection Within a Professional UK Soccer Academy
The present study engaged in an ethnographical observation of the processes used to determine player (de)selections within a professional academy. English category-2 youth academy players (n = 96) from U10–U16 age groups undertook anthropometric profiling (height, mass and somatic maturation) and fitness assessments (10 m, 20 m & 30 m linear sprints, 505-agility test, countermovement and squat jumps). Each players lead coach (n = 4) subjectively graded players utilising a red, amber and green (RAG) rating system on a weekly (current performance) and quarterly (perceived potential) basis, across 25 weeks. A MANCOVA, controlling for maturation, was applied to determine differences in (de)selection by physical performance. Mann Whitney-U tests were used to distinguish difference in (de)selection by subjective grading (weekly and quarterly). The key finding was that quarterly subjective gradings established a higher cumulative score of green ratings in selected players and a low cumulative score of red ratings, and vice versa for deselected players (P ≤ 0.001 to 0.03). However, whilst these findings suggest that quarterly subjective grades of potential were able to provide the best predictors for player (de)selection, the findings should be viewed with caution due to high potential for confirmatory bias.
Introduction
In an attempt to develop home-grown professional soccer players, the English Premier League implemented the 'Elite Player Performance Plan' (EPPP) within all English professional soccer academies [35]. The EPPP was created to provide a long-term model of development, encompassing the holistic development of players (technical/tactical, physical, psychological and social) [35]. Within the EPPP, national benchmarking of sports science and medical assessments are a mandatory criterion of data collection [35], ensuring that each player's physical profiles are monitored and measured throughout their academy journey. Additional variables include anthropometric measures in order to determine somatic maturation [21,26,27]. Physical assessments typically measure components of fitness; speed, power, stamina, to list a few [13,41,45].
Previous research [10,13,14,23,36] in academy soccer investigating the discriminative effects of fitness (speed, power, endurance) and anthropometrics on player status (elite to non-elite, academy to non-academy, and selected to deselected players) has provided inconsistent and conflicting findings. An example of such conflicting findings has been observed with reports from le Gall et al. [23] acknowledging anthropometric differences in playing status yet no differences in speed performances. In contrast, Deprez et al. [10] established speed as a defining factor for playing status, yet anthropometry to be an insignificant variable. Therefore, the use of physical attributes to define (de)selection status remains somewhat inconclusive.
Within the EPPP, coaches are required to provide subjective feedback and reflections on player performance, providing a timeline of evidence for player development [35]. The collation of subjective feedback can later be used to inform coaches of a players performance developments, especially when considering player (de)selection. This is particularly important when considering a player's 'potential', which may be predictive of their future abilities. Typically, subjective feedback is attained through coach intuition, defined as a coach's subjective beliefs based on experience and acquired knowledge, perceived contextual performance and self-belief in a coach's capacity to develop the individual [38]. Whilst research has demonstrated confidence in using coach intuition [22,37,42], concerns are raised in a coaches ability to distinguish performance within homogenous groups. Dugdale et al. [13] established that coaches could not determine performance differences within the top or bottom-performing groups of players. Likewise, Jokuschies et al. [19] reported that coaches vary in perceived areas of importance, reducing the objectivity within coach intuition. Consequently, these findings highlight limitations in the talent development process, based solely on coach intuition. However, when coach intuition is supplemented with objective assessment data, greater accuracy is attained within the player selection processes [12,13,42]. Therefore, in line with previous reports [13,42], holistic assessments employing both subjective (i.e., coach intuition) and objective outputs (i.e., components of fitness, maturation, etc.) are likely to provide greater precision in player development and (de) selection outcomes.
Maturation has been established to be a confounding variable within player selection processes. In the absence of informed data, coaches have reportedly demonstrated an unconscious bias in perceived levels of potential, with latematuring players deemed as having low-potential to succeed and a bias towards early-maturing players [7,28,30]. Maturational variations will likely influence selection outcomes, particularly around the years of the peak height velocity (PHV) [4,11]. Given that players of the same chronological age may vary in maturation by two years or potentially greater [29], physical performance variations are likely to be present. Moreover, several reports have demonstrated that players exhibiting advanced maturational timing are more likely to possess superior performances when testing components of fitness [8,11,24,47]. However, beyond PHV, when all players have transitioned into early adulthood, physical advantage from early maturational timing is often attenuated and/or reversed [8,9]. Moreover, the 'Underdog Hypothesis' has provided evidence that late-maturing players can possess a greater long-term potential in performance, due to higher reported levels of self-regulation evolving from extensive periods of overstimulation [9]. Therefore, informing coaches about each players' maturational status may reduce coach (sub)conscious selection bias, enhancing development opportunities and realising player potential. Likewise, caution should be taken when (de)selecting players based on physical markers alone, particularly within adolescent age bands [8,14,24].
The present study serves as a pragmatic exploration to observe the current processes utilised to measure and identify player standards and the protocols used for player (de)selection within a professional soccer academy. Moreover, an ethnographic research approach is employed to understand the suitability of the current processes applied within professional practice. Therefore, the data analysed emerged from the current methods and instruments used within a singular academy. Whilst this provides feedback unique to one academy, it is of the firm belief that similar processes are currently employed within other professional academies and clubs. Such assumptions are derived from previous research, whereby similar player grading instruments have been utilised [7,12]. Therefore, upon the completion of data collection, data interrogation will look at the difference in (de)selection and performance grading, with the hypothesis that i) subjective grading will align with (de) selection outcomes and ii) objective measures will highlight attributes aligned with (de)selection outcomes (whilst controlling for maturation). In answering these questions, further clarity is provided on academy decisions and processes applied, whilst further offering the research findings and practical applications for other academies' interpretations.
Objective Assessments
Objective assessments consisted of anthropometry and components of fitness tests and were measured using a singular time point in November (with additional planned assessment dates cancelled due to COVID-19 protocols). Given the ethnographical approach to this research, the assessments were already applied in practice. Two age groups undertook all of the assessments per day, with all age groups completing testing within one-week. Schedules were planned to ensure that each age group was provided a minimum of 48 h rest from previous training or games. Of the two age groups tested per day, a rotation was performed whereby one group would perform speed and agility tests first, whilst the other underwent anthropometric and jumping assessments. The players were familiar with the testing protocols and had undertaken the assessments previously (as part of the academy quarterly testing battery). A standardised warm-up preceded testing, consisting of a pulse raiser and muscle activation and mobility, to ensure players were suitably prepared.
Physical Profile
Anthropometric measures were taken using a stadiometer and scales (Seca, UK) with the removal of footwear. All measures were taken abiding by the guidelines provided by the International Society for the Advancement of Kinanthropometry (ISAK), taken by the same practitioner throughout. Somatic maturation was determined by calculating the percentage of adult height [21,26], which required a players decimal age, current height and mass, and mid-parent heights. Parental heights were attained before the start of the season, and where self-reports were used, necessary adjustments were applied to handle typical over-estimation [15].
Components of Fitness
Linear sprints and 505 change of direction (COD) tests were complete using single-beam light gates (Smartspeed, USA) on an indoor 3G pitch. Tests were initiated with a falling start, whereby the player starts 0.5 m before the first gate, with the feet in line. In both linear and COD assessments, players were informed to run beyond the final gate to prevent early deceleration. In the COD, light gates were placed at 10 m, with test markings produced at 15 m for players to change direction on. The assessed foot was required to be placed beyond the 15 m marking before returning through the light gate. Three trials were used for the linear sprints, and two trials per turning leg in the 505, using the best trials for further analysis..
Jumping tasks were performed using two force plates (Pasco, USA) set to 1000 Hz, and a compatible analysis software package (Capstone, USA). For both the countermovement-and squat jumps (CMJ and SQJ), players placed one foot on each force plate and were asked to remain stationary to capture bodyweight. In both jumps, players used a selfprescribed jump depth with the arms on the hips until the completion of each jump and were asked to jump 'as high as you can'. During the SQJ, players were asked to hold the dipped position for at least 2 s before initiating the jump. Players were required to land back on the plates, whilst absorbing the landing forces. Three trials of each jump were collected, with the best trial being used for further analysis.
Subjective Assessments
Coach subjective assessments were used to measure the technical and tactical abilities and overall potential of players, based on coach beliefs and perceptions. Subjective measures were taken on a weekly basis, identifying current performance, and a quarterly basis identifying a players perceived future potential. The academy employed a Red, Amber and Green (RAG) rating system for all subjective gradings as standard practice. The RAG rating system is commonly used within academy infrastructures, whereby common definitions state that red is 'performing below the expected standard', amber is 'performing at the expected standard' and green is 'performing above the expected standard' (or similar). This system is further integrated into the EPPP online audit system, the PMA, arguably explaining the original use of this method. Tallies of both weekly subjective performance grading and the quarterly subjective potential grading were used for further statistical analysis. Previous studies have looked at the test-retest reliability of coach subjective grading of players [19] establishing partly acceptable and partly unacceptable (− 0.57 ≤ r ≤ − 0.81) reliability outcomes.
Weekly Subjective Grading
Weekly subjective player grading of technical and tactical abilities were determined by lead coaches of their respective age groups. Coaches would provide a weekly score (red, amber or green) per player. For the entirety of the 2020-2021 season, performance RAG ratings were collected and tallied into a total quantity of red, amber and green scores. However, due to COVID-19 restrictions during the season, only 25-weeks (of the traditional ~ 36-weeks) across all ages (excluding U16s) were recorded for assessment. Due to the U16 (de)selection process being earlier than other ages, due to scholarship transitions, only 16-weeks (of the traditional ~ 26-weeks) of coach gradings were collected before selection decisions.
Quarterly Subjective Grading
Further assessments included perceived measures of potential. Coaches underwent quarterly (Q1 = September, Q2 = December, Q3 = April) subjective assessments of player future potential, further utilising the RAG rating. Coaches would assign each player either a red, amber or green score during each quarter of the season. Therefore, following the end of the season, each player would have three RAG scores for potential. The quarterly subjective potential grading is determined across a player's holistic ability (psychology, technical, tactical, social, physical), using available objective outcomes (provided by the sports science and medical department), and the coach's belief in the continual rate of progression in development and performance.
Player Selection
The process of player selection was undertaken as per normal academy procedures. Coaches were provided with all objective data prior to meetings between the lead coach, academy manager and head of coaching to discuss player selections. The U16 age group was the only age group where more staff was present within the selection process, including sports science and medical staff, U18 coaching staff, head of recruitment and head of education. Consequently, the outcome of this process results in 29 deselected players and 67 selected players (Table 1).
Statistical Analysis
Data distribution was repeatedly assessed by age group, using the Shapiro-Wilk test of normality. The U11 and U12 age groups combined (21 players) featured only one deselected player, and therefore was ineligible for further analysis and removed from the dataset. A multivariate analysis of covariance (MANCOVA) was applied for the objective assessments, utilising maturation as a covariant and linear speed (10, 20 m and 30 m sprints), COD (505 left and right), jumping tasks (CMJ and SQJ) and anthropometry (height and mass) as dependent variables, to investigate differences between age groups and selection status (independent variables). Maturation was controlled for given its potentially confounding influence within the analysis of physical performance [25,28,33]. An alpha level of < 0.05 was applied and follow-up univariate analysis (with Bonferroni adjustments) [1] were used where appropriate.
Where violations of normal distributions were observed, such as within the tally of RAG ratings, non-parametric tests were used. A Mann Whitney U-test was applied to identify the difference in player (de)selection and RAG tallies. Tallies were determined using cumulative RAG scores awarded by coaches across a season, per player. Therefore, a player can only be awarded either a red, amber or green per week or quarter. Tallies of each red, amber and green were then compared for differences by (de)selection outcomes. Due to the multiple comparisons of data, a Bonferroni correction was applied to reduce type 1 error [1]. A Bonferroni correction level was determined as the quantity of independent variables multiplied by the quantity of dependant variables, with the outcome providing the division of alpha set at < 0.05. Therefore, a new alpha was set at < 0.017. For the Mann Whitney U tests, r values were determined from Z-scores [16,17], with outcomes ≥ 0.1-0.29 = small, 0.3-0.49 = medium and ≥ 0.5 = large effect size. Subsequently, Eta-squared was calculated from r value outcomes. Eta-squared effect sizes were interpreted as > 0.01 = small effect, > 0.06 = medium effect and > 0.14 = large effect. Outcomes from the Mann Whitney U test were reported as medians and interquartile range (IQR).
Given the high subjectivity of the selection process, further statistics were applied to explore the potential for subjective bias. Recent player performance may influence decisions on selection outcome (over the utility of the full season report of performance). Time course influence was investigated by comparing differences in group performances by selection status between weeks 1-20 and weeks 21-25, using a MANOVA with observations of status by weeks interaction. Additionally, a Cramer's V was used to interrogate the associations of within quarterly subjective potential gradings, and between quarterly subjective potential gradings and selection outcomes. Effect size for Cramer's V were interpreted based upon degrees of freedom (Table 2), as outlined by Cohen [5]. All data were analysed using SPSS Statistics for Windows, Version 26.0 (Armonk, NY: IBM Corp.). The outcomes of the weekly subjective grade tallies identified significantly (with applied Bonferroni correction) higher tallies of green ratings in selected players (U = 0.0, P < 0.001, r = −0.83, η 2 = 0.68) and red ratings in deselected players (U = 4.5, P = 0.01, r = −0.82, η 2 = 0.67), within the U14 age group only (Table 3). Within the quarterly grading tallies, a consistent finding was observed with the low quantity of red ratings for selected players, and high quantity for deselected players in the U13, U14, U15 and U16 age groups (Table 3). Only in the U15 age group were significant differences identified in the tally of green ratings (U = 2.5, P < 0.001, r = − 0.74, η 2 = 0.55) between the selected (median = 2, IQR = 2-3) and deselected (median = 0, IQR = 0-0) players.
In considerations of subjective bias at different time courses, the results from the MANOVA found no significant (P > 0.05) differences in all age groups when considering interactions between selection status (selected vs. deselected) and weeks (weeks 1-20 vs. weeks [21][22][23][24][25]. When investigating the associations of quarterly potential gradings, Cramer's V observed various associations between different quarters and selection outcome (Table 4). In comparison of gradings across quarters, whilst all age groups across each quarter demonstrated large effect sizes, only few significant associations were reported within Q1 and Q2, and Q1 and Q3. Likewise, when looking at associations with quarterly grading and selection outcome, only a few large significant findings were established in Q1 and Q2, whereas Q3 demonstrate large to perfect significant associations across all age groups.
Discussion
This study looked to report on the current processes employed within a professional soccer academy undertaking player (de)selection, via an ethnographical approach, with a focus on identifying the difference in performance between (de)selection status. The key findings of this study were that coach subjective measures of player abilities were inconclusive in determining selection status. Whilst coach perceptions of player potential were capable of determining selection status in the final quarter only, this was potentially indicative of confirmative bias given the close proximity to when selection decisions are finalised. Additionally, weekly subjective gradings (current performance) were unable to consistently distinguish selection status. The outcomes of quarterly RAG gradings reported a greater number of red grades associated with deselection, with observations of higher green frequencies associated to selection. This implies that coaches are capable in distinguishing players at either extremity of performance (i.e., top-or bottom-performing players). Such findings align with previous reports [13,42] that identify coach intuition as a capable tool in determining (de)selected players. However, these findings also demonstrated that coaches may display indecision concerning the players of a moderate standard (amber), similar to previous research [13]. Whilst this middle ground is expected to cater to players with greater uncertainty towards their future, the range of abilities is vast and features players that were opted for both selection and deselection. Therefore, to reduce this range of abilities, it may be of greater benefit to further sub-divide grades to offer greater clarity in perceptions of player abilities and potential, similar to the nine by nine grid of performance and potential proposed by Baker et al. [2].
Conversely, the quarterly subjective grade results fail to control for confirmation bias, given that the same coaches who graded the players were also the coaches who (de) selected them. Considering the process of deselection observed evolved from weekly performance feedback with managerial staff, it is fair to elude that the lead phase coaches may (sub)consciously inform and influence their viewpoints on player performances. Therefore, it is impossible to dismiss the potential for confirmation bias. Additionally, this study established large-to-perfect associations across all quarters of potential grading and selection outcomes, with Table 4 The associations between quarters of subjective gradings of potential, and the associations of quarterly gradings and selection outcomes Q1 = quarter 1, Q2 = quarter 2, Q3 = quarter 3 * = significant outcome < 0.017 (with Bonferroni correction) Given the proximity to selection decisions, it is probable that the Q3 potential scores are moreso representative of selection decisions. Furthermore, the gradings are gathered by the lead coach alone. Therefore, to both mitigate selection bias and improve clarity for player developments and selection, further objectivity towards measures of potential should be explored, potentially incorporating the use of multiple inputs from additional coaching staff. Weekly subjective grades demonstrated a low utility in identifying selected from deselected players, with only the U14 age group reporting significant findings of red and green grades. Whilst the weekly subjective grades found low ability in identifying (de)selection outcomes, one concern of this instrument surrounded the bias of more recent performance influencing selection decisions, over the full season collection of grades. However, when assessing associations between the final five weeks of performance to the complete season, fair to very strong associations were observed. Therefore, providing greater reassurance that the weekly subjective grades provided a fair reflection of performance across the season. Furthermore, associations between weekly and quarterly subjective grades established a consistent correlation within green scores primarily. This suggests that top performing players maintain a more consistent playing performance than mid to lower performing players.
One explanation as to why weekly subjective grades failed to distinguish player selection outcomes is due to coaches providing grades to their individual age groups only, with no additional inputs. Each coach will maintain an expected performance standard based upon experience, knowledge, and beliefs, as outlined within the definition of coach intuition [22,37]. This will undoubtedly vary between coaches. Therefore, whilst one coach may perceive a player to be a 'amber' performance grade, another may perceive them to be a 'green'. This lack of uniformity in player grading highlights the need for academies to provide comprehensive anchor points within the grading tool, so to enhance clarity in player grading, like in the earlier proposed instrument by Baker et al. [2]. Likewise, academies may benefit from the adoption of a validated and reliable instrument for the subjective measure of player performance.
In consideration of the differing outcomes between weekly and quarterly subjective gradings, it may be postulated that the focus of each instrument (weekly vs. quarterly) provides the difference in selection outcome. Whilst weekly subjective grading focuses on current performance, quarterly subjective grading emphasizes perceived future potential. It is therefore plausible that coaches may measure current performance against a player's perceived future potential, i.e., a player who is believed to have high potential may be graded as currently underperforming, based on the coach's perceptions of the players' future abilities yet to be realised.
Likewise, a player with low potential may be scored higher in their current performance, given that the coach perceives them as playing to the highest standard expected of them to achieve (or in some cases, beyond this). Therefore, current performance ratings may instead provide some context in the development and attainment of (perceived) potential and may be best employed by utilising frameworks that account for both 'current performance' vs. 'perceived potential' [2,44].
In the present study, physical performance was not capable of distinguishing selected from deselected players. Given that the study undertook an ethnographic approach, whereby procedures tested are standard practice, and that objective assessments failed to distinguish selection outcome, potentially highlights the need for wider metrics to be applied to aid player selection. For example, a previous report has demonstrated the high interaction of acceleration within the 505 COD task design, resulting in only 31% of time spent changing direction [32]. Therefore, those with greater linear speed abilities are more likely to demonstrate greater overall COD outcomes [31,32]. Nimphius et al. [31] suggested using a change of direction deficit (COD def ) calculation to enhance the measurement of turning ability, consequently mitigating linear speed bias and providing a more reflective measure of task assessment. It may be suggested that the use of more task-specific and holistic (physical, psychological, social, technical and tactical) assessments, or contrary, assessments with higher ecological validity, may better serve the player (de)selection process.
A consideration in the findings for physical performance is the control for maturational variation. The current study found maturation to be a significant factor within selection. Therefore, academies must be mindful towards the control for maturational influence within physical performance data. Previous research has highlighted the influence of maturation on physical performance [25,28,33], whereby players exhibiting an advanced maturation status will possess greater physical abilities to their biologically younger peers. As a consequence, players of an advance maturation status are selected due to being perceived as beholding a higher potential for success [7]. Conversely, players of a late maturation status are released, due to a perception of low potential. However, research has highlighted that early maturation players are no more likely to achieve senior professional success to their biologically younger peers [20]. Without the affordance of time to achieve adult stature (therefore 'catching up' with their early and average maturing peers), late maturing players will not realise their full potential. Furthermore, late maturing players may only progress through academy selection processes by 'survival' [34], whereby they can tolerate over-stimulation without enduring burnout, injury or demotivation [34,39,46]. In summary, given the high variation of maturation identified within youths, controlling for maturation in physical performance is essential for consideration within the player (de)selection process.
This study is not without its limitations. Whilst this was an observation of a single academy's practice, further research should include several academies to identify (de) selection trends. However, the outcomes of this study can inform academies utilising similar selection processes of the stated shortcomings and considerations for enhancement. Likewise, this study highlighted areas that are not currently accounted for during player selection process. Whilst there is an open understanding for holistic abilities, the lack of assessments for psychological abilities and social measures potentially implicates player developments and (de)selection. Further limitations of this study may consider the lack of reliability assessments undertaken through the use of the RAG rating method and coach scoring. Whilst previous studies have been undertaken that outline test-retest reliability in coach subjective scoring [19], further work should be undertaken assess the use of RAG ratings.
Additionally, uncontrollable challenges were associated with the present study; during the season of 2020-2021, the COVID-19 international pandemic brought closure to academy soccer within the UK at various time points e.g., delayed start and closures in December and January. Academy closure (therefore a cessation of training) and imposed restrictions negatively impacted training provisions and physical performance assessment collection, traditionally undergone quarterly (June, October, January and April). This resulted in a singular time point provided for the 2020-2021 season. Moreover, the outcome reflects the 2020-2021 academy season, where the best operating procedure was implemented to mitigate risks and superseded the need for further data collection.
Practical Implications
Based on the findings of this study, coach subjective perception of player future potential were inconclusive in determining player deselection. Coaches were only able to identify players perceived as high potential in the final quarter, and given the close proximity of selection decisions being made, suggests decisions are likely decided upon these outcomes. Moreover, this indicates a level of confirmative bias in the subjective process. Therefore, it may be suggested that further objectivity is required within the process for grading potential in order to enhance player development and selection processes.
When considering weekly subjective gradings, it has been postulated that performance grades may provide context and further inform the coach on the present status of the players' predicted journey. Moreover, the continual collection of weekly and quarterly subjective grading is logical, with further research looking to better understand what further distinguishes player abilities. However, caution should be raised around the consistent pressure placed on players to perform. Given that measures of potential were unable to predict subsequent selection status, players remain under consistent pressure to perform throughout the season to retain selection status. Such issues are pertinent for players of moderate player abilities. Given it was clear that coaches were able to differentiate players at each extremity of performance, uncertainty remained around moderate ability players. Further research needs to be undertaken to investigate a coach's ability to perceive the development needs of these players, to ensure they are provided the appropriate and optimal provisions.
Given the confounding effect of maturation, bio-banding interventions may be convenient to ensure optimal developments are maintained. Banding players by biological age entails grouping players by growth status, regardless of chronological age. Such provisions will reduce any dependence on physical prowess and the higher demand for technical and tactical ability, typically exhibited by biologically advanced players. Likewise, such provisions can also be offered to biologically younger players, moving down and age group, which should afford the time to develop confidence and develop leadership skills; opportunities less likely to be presented within their own age groups.
Conclusion
The present study identified that a season of coach subjective perceptions (current performance and perceived potential) were inconclusive of determining subsequent selection status. Furthermore, the results indicate a high chance of confirmation bias associated with the current selection process. Additionally, it was apparent that whilst coaches were capable of distinguishing the extremities of player performance, coaches remain uncertain around the grading for moderate ability players. Further work is required to provide more objective assessments to mitigate the potential for bias within assessments, and enhance the accuracy in the player selection process.
Data availability
The datasets generated and analysed during the current study are not publicly available due to maintaining confidentiality of the participants and the associated club, but are available from the corresponding author upon reasonable request.
Conflict of interest
The authors report no conflict of interest.
Consent to participate Consent to participation was obtained from all participants. | 2023-03-12T05:14:56.530Z | 2023-03-08T00:00:00.000 | {
"year": 2023,
"sha1": "4a4e940daa3d15f4d0f1c394ac8ea43d79a665bb",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "4a4e940daa3d15f4d0f1c394ac8ea43d79a665bb",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Medicine"
]
} |
30065545 | pes2o/s2orc | v3-fos-license | Ab initio calculations of BaTiO3 and PbTiO3 (001) and (011) surface structure
We present and discuss the results of calculations of surface relaxations and rumplings for the (001) and (011) surfaces of BaTiO3 and PbTiO3, using a hybrid B3PW description of exchange and correlation. On the (001) surfaces, we consider both AO (A = Ba or Pb) and TiO2 terminations. In the former case, the surface AO layer is found to relax inward for both materials, while outward relaxations of all atoms in the second layer are found at both kinds of (001) terminations and for both materials. The surface relaxation energies of BaO and TiO2 terminations on BaTiO3 (001) are found to be comparable, as are those of PbO and TiO2 on PbTiO3 (001), although in both cases the relaxation energy is slightly larger for the TiO2 termination. As for the (011) surfaces, we consider three types of surfaces, terminating on a TiO layer, a Ba or Pb layer, or an O layer. Here, the relaxation energies are much larger for the TiO-terminated than for the Ba or Pb-terminated surfaces. The relaxed surface energy for the O-terminated surface is about the same as the corresponding average of the TiO and Pb-terminated surfaces on PbTiO3, but much less than the average of the TiO and Ba-terminated surfaces on BaTiO3. We predict a considerable increase of the Ti-O chemical bond covalency near the BaTiO3 and PbTiO3 (011) surface as compared to both the bulk and the (001) surface.
I. INTRODUCTION
Thin films of ABO 3 perovskite ferroelectrics play an important role in numerous microelectronic, catalytic, and other high-technology applications, and are frequently used as substrates for growth of other materials such as cuprate superconductors. 1,2 Therefore, it is not surprising that a large number of ab initio quantum mechanical calculations, 3,4,5,6,7,8,9,10,11,12,13 as well as several classical shell-model (SM) studies, 14,15,16 have dealt with the atomic and electronic structure of the (001) surface of BaTiO 3 , PbTiO 3 , and SrTiO 3 crystals. In order to study the dependence of the surface relaxation properties on the exchange-correlation functionals and the type of basis (localized vs. plane-wave) used in the calculations, a detailed comparative study of SrTiO 3 (001) surfaces based on ten different quantum-mechanical techniques 17,18 was recently performed.
Due to intensive development and progressive miniaturization of electronic devices, the surface structure as well as the electronic properties of the ABO 3 perovskite thin films have been extensively studied experimentally during the last years. The SrTiO 3 (001) surface structure has been analyzed by means of low-energy electron diffraction (LEED), 19 reflection high-energy electron diffraction (RHEED), 20 X-ray photoelectron spectroscopy (XPS), ultraviolet electron spectroscopy (UPS), medium-energy ion scattering (MEIS), 21 and surface Xray diffraction (SXRD). 22 Nevertheless it is important to note that the LEED 19 and RHEED 20 experiments contradict each other in the sign (contraction or expansion) of the interplanar distance between top metal atom and the second crystal layer for the SrO-terminated SrTiO 3 (001) surface. The most recent experimental studies on the SrTiO 3 surfaces include a combination of XPS, LEED, and time-of-flight scattering and recoil spectrometry (TOF-SARS), 23 as well as metastable impact electron spectroscopy (MIES). 24 In these recent studies, wellresolved 1×1 LEED patterns were obtained for the TiO 2terminated SrTiO 3 (001) surface. Simulations of the TOF-SARS azimuthal scans indicate that the O atoms are situated 0.1Å above the Ti layer (surface plane) in the case of the TiO 2 -terminated SrTiO 3 (001) surface.
While the (001) surfaces of SrTiO 3 , BaTiO 3 and PbTiO 3 have been extensively studied, much less is known about the (011) surfaces. The scarcity of information about these surfaces is likely due to the polar character of the (011) orientation. (011) terminations of SrTiO 3 have frequently been observed, but efforts towards the precise characterization of their atomic-scale structure and corresponding electronic properties has only begun in the last decade, specifically using atomic-force microscopy, 25 scanning tunneling microscopy (STM), Auger spectroscopy, and low-energy electron-diffraction (LEED) 26 methods.
To the best of our knowledge, very few ab initio studies of perovskite (011) surfaces exist. The first ab initio study of the electronic and atomic structures of several (1×1) terminations of the (011) polar orientation of the SrTiO 3 surface was performed by Bottin et al. 27 One year later Heifets et al. 28 performed very comprehensive ab initio Hartree-Fock calculations for four possible terminations (TiO, Sr, and two kinds of O terminations) of the SrTiO 3 (011) surface. Recently Heifets et al. 29 performed ab initio density-functional calculations of the atomic structure and charge redistribution for different terminations of the BaZrO 3 (011) surfaces. However, despite the high technological potential of BaTiO 3 and PbTiO 3 , we are unaware of any previous ab initio cal-culations performed for the BaTiO 3 and PbTiO 3 (011) surfaces. In this study, therefore, we have investigated the (011) as well as the (001) surfaces of BaTiO 3 and PbTiO 3 , with an emphasis on the effect of the surface relaxation and rumpling, surface energies, and the charge redistributions and changes in bond strength that occur at the surface.
A. Computational method
We carry out first-principles calculations in the framework of density-functional theory (DFT) using the CRYSTAL computer code. 30 Unlike the plane-wave codes employed in many previous studies, 31,32 CRYSTAL uses localized Gaussian-type basis sets. In our calculations, we adopted the basis sets developed for BaTiO 3 and PbTiO 3 in Ref. [33]. Our calculations were performed using the hybrid exchange-correlation B3PW functional involving a mixture of non-local Fock exact exchange, LDA exchange, and Becke's gradient corrected exchange functional, 34 combined with the non-local gradient corrected correlation potential of Perdew and Wang. 35,36,37 We chose the hybrid B3PW functional for our current study because it yields excellent results for the SrTiO 3 , BaTiO 3 , and PbTiO 3 bulk lattice constant and bulk modulus. 17,33 The reciprocal-space integration was performed by sampling the Brillouin zone with an 8×8×8 Pack-Monkhorst mesh, 38 which provides a balanced summation in direct and reciprocal spaces. To achieve high accuracy, large enough tolerances of 7, 8, 7, 7, and 14 were chosen for the dimensionless Coulomb overlap, Coulomb penetration, exchange overlap, first exchange pseudo-overlap, and second exchange pseudo-overlap parameters, respectively. 30 An advantage of the CRYSTAL code is that it treats isolated 2D slabs, without any artificial periodicity in the z direction perpendicular to the surface, as commonly employed in most previous surface band-structure calculations (e.g., Ref. [8]). In the present ab initio investigation, we have studied several isolated periodic twodimensional slabs of cubic BaTiO 3 and PbTiO 3 crystals containing 7 planes of atoms.
B. Surface geometries
The BaTiO 3 and PbTiO 3 (001) surfaces were modeled using symmetric (with respect to the mirror plane) slabs consisting of seven alternating TiO 2 and BaO or PbO layers, respectively. One of these slabs was terminated by BaO planes for the BaTiO 3 crystal (or PbO planes for PbTiO 3 ) and consisted of a supercell containing 17 atoms. The second slab was terminated by TiO 2 planes for both materials and consisted of a supercell containing Unlike the (001) cleavage of BaTiO 3 or PbTiO 3 , which naturally gives rise to non-polar BaO (or PbO) and TiO 2 terminations, a naive cleavage of BaTiO 3 or PbTiO 3 to create (011) surfaces leads to the formation of polar surfaces. For example, the stacking of the BaTiO 3 crystal along the (011) direction consists of alternating planes of O 2 and BaTiO units having nominal charges of −4e and +4e respectively, assuming O 2− , Ti 4+ , and Ba 2+ constituents. (Henceforth we shall use BaTiO 3 for presentation purposes, but everything that is said will apply equally to the PbTiO 3 case.) Thus, a simple cleavage leads to O 2 -terminated and BaTiO-terminated (011) surfaces that are polar and have nominal surface charges of −2 e and +2 e per surface cell respectively. These are shown as the top and bottom surfaces in Fig. 2(a) respectively. If uncompensated, the surface charge would lead to an infinite electrostatic cleavage energy. In reality, the polar surfaces would probably become metallic in order to remain neutral, but in view of the large electronic gaps in the perovskites, such metallic surfaces would presumably be unfavorable. Thus, we may expect rather generally that such polar crystal terminations are relatively unstable in this class of materials. 3 On the other hand, if the cleavage occurs in such a way as to leave a half layer of O 2 units on each surface, we obtain the non-polar surface structure shown in Fig. 2(b). Every other surface O atom has been removed, and the remaining O atoms occupy the same sites as in the bulk structure. We shall refer to this as the "O-terminated" (011) surface, in distinction to the "O 2 -terminated" polar surface already discussed in Fig. 2(a). The non-polar nature of the O-terminated surface can be confirmed by noting that the 7-layer 15-atom Ba 3 Ti 3 O 9 slab shown in Fig. 2(b), which has two O-terminated surfaces, is neutral. It is also possible to make non-polar TiO-terminated and Ba-terminated surfaces, as shown in Figs. 2(c) and In the present calculations of the BaTiO 3 and PbTiO 3 (001) surface atomic structure, we allowed the atoms located in the two outermost surface layers to relax along the z-axis (the forces along the x and y-axes are zero by symmetry). Here we use the term "layer" to refer to a BaO, PbO, or TiO 2 plane, so that there are two layers per stacked unit cell. For example, on the BaO or PbOterminated surfaces, the top layer is BaO or PbO and the second layer is TiO 2 ; displacements of the third-layer atoms were found to be negligibly small in our calculations and thus were neglected.
The calculated atomic displacements for the TiO 2 and BaO-terminated (001) surfaces of BaTiO 3 , and for the TiO 2 and PbO-terminated (001) surfaces of PbTiO 3 , are presented in Table I approach. 16 Similarly, for PbTiO 3 (001), Table I shows comparisons with the plane-wave LDA calculations of Meyer and Vanderbilt. 5 The relaxation of the surface metal atoms in both the BaTiO 3 and PbTiO 3 surfaces is much larger than that of the oxygen ions, leading to a considerable surface rumpling, which we quantify via a parameter s defined as the relative displacement of the oxygen with respect to the metal atom in a given layer. The surface rumpling and relative displacements of three near-surface planes are presented in Table II. According to our calculations, atoms of the first surface layer relax inwards (i.e., towards the bulk) for BaO and PbO terminations of both materials. Our calculations are in a qualitative agreement with the ab initio calculations performed by Padilla and Vanderbilt 4 for BaTiO 3 , and by Meyer and Vanderbilt 5 for PbTiO 3 . However, the predictions of the SM calculation disagree with the firstprinciples calculations; the SM predicts that the firstlayer oxygen ions relax outward on the BaO-terminated BaTiO 3 (001) surface, 16 rather than inwards. However, the magnitudes of the displacements are relatively small (−0.63% of the lattice constant a 0 in this study and 1.00% of a 0 in the SM calculations) 16 which may be close to the error bar of the classical shell model. Outward relaxations of all atoms in the second layer are found at both (001) terminations of the BaTiO 3 and PbTiO 3 surfaces. From Table I, we can conclude that the magnitudes of the atomic displacements calculated using different ab initio methods and using the classical shell model are in a reasonable agreement.
In order to compare the calculated surface structures further with experimental results, the surface rumpling s and the changes in interlayer distances ∆d 12 and ∆d 23 , as defined in Fig. 1, are presented in Table II. Our calculations of the interlayer distances are based on the positions of relaxed metal ions, which are known to be much stronger electron scatters that the oxygen ions. 19 For BaTiO 3 (001), the rumpling of TiO 2 -terminated surface is predicted to exceed that of BaO-terminated surface by a factor of two. This finding is in line with the values of surface rumpling reported by Padilla and Vanderbilt. 4 In contrast, PbTiO 3 demonstrates practically the same rumpling for both terminations. From Table II one can see that qualitative agreement between all theoretical methods is obtained. In particular, the relaxed (001) surface structure shows a reduction of interlayer distance ∆d 12 and an expansion of ∆d 23 according to all ab initio and shell-model results.
As for experimental confirmation of these results, we are unfortunately unaware of experimental measurements of ∆d 12 and ∆d 23 for the BaTiO 3 and PbTiO 3 (001) surfaces. Moreover, for the case of the SrOterminated SrTiO 3 (001) surface, existing LEED 19 and RHEED 20 experiments actually contradict each other regarding the sign of ∆d 12 . In view of the absence of clear experimental determinations of these parameters, therefore, the first-principles calculations are a particularly important tool for understanding the surface properties. To our knowledge, we have performed the first ab initio calculations for BaTiO 3 and PbTiO 3 (011) surfaces. We have studied the TiO 2 -terminated, BaO or PbO-terminated, and O-terminated surfaces illustrated in Fig. 2(c), (d), and (b), respectively. The computed surface atomic relaxations are reported in Table III.
Focusing first on the BaTiO 3 surfaces, we find that the Ti ions in the outermost layer of the TiO-terminated surface move inwards (towards the bulk) by 0.0786 a 0 , whereas the O ions in the outermost layer move outwards by a 0.0261 a 0 . The Ba atoms in the top layer of the Ba-terminated surface of Fig. 2(d) and the O atoms in the outermost layer of O-terminated surface of Fig. 2(b) move inwards by 0.0867 a 0 and 0.0540 a 0 , respectively. The agreement between our ab initio B3PW and the classical SM calculations is satisfactory for all three of these surface terminations. In particular, the directions of the displacements of first and second-layer atoms coincide for all three terminations. This indicates that classical SM calculations with a proper parameterization can serve as a useful initial approximation for modeling the atomic structure in perovskite thin films. Turning now to our results for the PbTiO 3 (011) surfaces, we find that all metal atoms in the outermost layer move inwards irrespective of the termination. Surface oxygen atoms are displaced outwards for the TiOterminated surface, while oxygen atoms move inwards in the O-terminated surface. The displacement patterns of atoms in the outermost surface layers are similar to those of the BaTiO 3 (011) surfaces, as well as classical shell model results for BaTiO 3 . 16 For example, the atomic displacement magnitudes of Ti and oxygen atoms in the TiO-terminated PbTiO 3 (011) surface are −0.0813 a 0 and 0.033 a 0 respectively. The Pb atom is displaced inwards by 0.1194 a 0 for the Pb-terminated surface, similar to the corresponding BaTiO 3 case. Overall, Table III shows similar displacement patterns for the BaTiO 3 and PbTiO 3 (011) surfaces, as well as qualitatively similar results for both ab initio and classical shell-model descriptions.
C. BaTiO3 and PbTiO3 (001) and (011) surface energies
In the present work, we define the unrelaxed surface energy of a given surface termination X to be one half of the energy needed to cleave the crystal rigidly into an unrelaxed surface X and an unrelaxed surface with the complementary termination X ′ . For BaTiO 3 , for example, the unrelaxed surface energies of the complementary BaO and TiO 2 -terminated (001) surfaces are equal, as are those of the TiO and Ba-terminated (011) surfaces (and similarly for PbTiO 3 ). The relaxed surface energy is defined to be the energy of the unrelaxed surface plus the (negative) surface relaxation energy. These definitions are chosen for consistency with Refs. [17,28]. Unlike the authors of Refs. [27,29], we have made no effort to introduce chemical potentials here, so the results must be used with caution when addressing questions of the where X = BaO or TiO 2 specifies the termination, E unr slab (BaO) and E unr slab (TiO 2 ) are the unrelaxed BaO and TiO 2 -terminated slab energies, E bulk is energy per bulk BaTiO 3 unit cell, and the factor of four comes from the fact that four surfaces are created by the two cleavages needed to make the two slabs. The relaxation energy for each termination can be computed from the corresponding slab alone using where E slab (X) is a slab energy after relaxation. The relaxed surface energy is then Similarly, for the BaTiO 3 (011) case, a cleavage on a bulk BaTiO plane gives rise to the complementary TiO and Ba-terminated surfaces shown in Fig. 2 where the energy is the same for X = TiO or Ba, E unrel slab (Ba) and E unrel slab (TiO) are energies of the unrelaxed slabs. Relaxation energies can again be computed independently for each slab in a manner similar to Eq. (2).
Finally, the (011) surface can also be cleaved to give two identical self-complementary O-terminated surfaces of the kind shown in Fig. 2(b). In this case the 7-layer slab has the stoichiometry of three bulk unit cells, so the relaxed surface energy of the O-terminated (011) surface is where E slab (O) is the relaxed energy of the slab having two O-terminated surfaces. Everything said here about BaTiO 3 surfaces applies in exactly the same way to the corresponding PbTiO 3 surfaces. The calculated surface energies of the relaxed BaTiO 3 (001) and (011) surfaces are presented in Table IV. In BaTiO 3 , the relaxation energies of the TiO 2 and BaO- The corresponding results are also given for the (001) and (011) surfaces of PbTiO 3 in Table IV. The results for the (001) surfaces are similar to those for BaTiO 3 , although the relaxed surface energies are somewhat lower. For the case of the (011) surfaces, however, we find a different pattern than for BaTiO 3 . We find a very large relaxation energy of −1.75 eV for the TiO-terminated surface, compared with −1.08 eV for the Pb-terminated surface and −1.12 eV for the O-terminated surface. The average energy of the TiO and Pb-terminated surfaces is now 1.69 eV, to be compared with 1.72 eV for the Oterminated surface, indicating that the cleavage on a Pb-TiO or an O 2 plane has almost exactly the same energy cost.
D. BaTiO3 and PbTiO3 (001) and (011) surface charge distribution and chemical bonding
To characterize the chemical bonding and covalency effects, we used a standard Mulliken population analysis for the effective static atomic charges Q and other local properties of the electronic structure as described, for example, in Ref. [39,40]. The results are presented in Table V. Our calculated effective charges for bulk PbTiO 3 are +1.354 e for the Pb atom, +2.341 e for the Ti atom, and −1.232 e for the O atom. The bond population describing the chemical bonding is +98 me between Ti and O atoms, +16 me between Pb and O atoms, and +2 me between Pb and Ti atoms. Our calculated effective charges for bulk BaTiO 3 are +1.797 e for the Ba atom, +2.367 e for the Ti atom, and −1.388 e for the O atom indicate a high degree of BaTiO 3 chemical bond covalency. The bond population between Ti and O atoms in BaTiO 3 bulk is exactly the same as in PbTiO 3 , while that between Ba and Ti is slightly negative, suggesting a repulsive interaction between these atoms in the bulk of the BaTiO 3 crystal.
For the TiO 2 -terminated BaTiO 3 and PbTiO 3 (001) surfaces, the major effect observed here is a strengthening of the Ti-O chemical bond near the BaTiO 3 and PbTiO 3 (001) surfaces, which was already pronounced for the both materials in the bulk. Note that the Ti and O effective charges for bulk BaTiO 3 and PbTiO 3 are much smaller than those expected in an ionic model (+4 e, and −2 e), and that the Ti-O chemical bonds in bulk BaTiO 3 and PbTiO 3 are fairly heavily populated for both materials. The Ti-O bond population for the TiO 2 -terminated BaTiO 3 and PbTiO 3 (001) surfaces are +126 me and +114 me respectively, which is about 20% larger than the relevant value in the bulk. In contrast, the Pb-O bond population of +54 me) is small for the PbO-terminated PbTiO 3 (001) surface, and the Ba-O bond population of −30 me is even negative for the BaO-terminated BaTiO 3 (001) surface, indicating a repulsive character. The effect of the difference in the chemical bonding is also well seen from the Pb and Ba effective charges in the first surface layer, which are close to the formal ionic charge of +2 e only in the case of the BaTiO 3 crystal.
The interatomic bond populations for three possible In Table VII we present the calculated Mulliken effective charges Q, and their changes ∆Q with respect to the bulk values, near the surface. We analyzed the charge redistribution between different layers in slabs with all three BaTiO 3 and PbTiO 3 (011) surface terminations. The charge of the surface Ti atoms in the TiO-terminated BaTiO 3 and PbTiO 3 (001) surface is reduced by 0.151 e and 0.129 e, respectively. Metal atoms in the third layer lose much less charge. Except in the central layer (and, in the case of PbTiO 3 , in the subsurface layer), the O ions also reduce their charges, becoming less negative. The largest charge change is observed for BaTiO 3 and PbTiO 3 subsurface O atoms (+0.233 e and +0.175 e, respectively). This gives a large positive change of +0.466 e and +0.350 e in the charge for each BaTiO 3 and PbTiO 3 subsurface layer.
On the Ba-terminated and Pb-terminated BaTiO 3 and PbTiO 3 (011) surface, negative changes in the charge are observed for all atoms except for Ba and Ti in the BaTiO 3 third layer, Ti atom in the PbTiO 3 third layer, and subsurface oxygen atom in PbTiO 3 . The largest charge changes are at the surface Ba and Pb ions. It is interesting to notice that, due to the tiny difference in the chemical bonding between BaTiO 3 and PbTiO 3 perovskites, the charge change for the BaTiO 3 subsurface O ion (−0.095 e) and PbTiO 3 subsurface O ion (+0.092 e) have practically the same magnitude, but opposite signs.
For the O-terminated BaTiO 3 and PbTiO 3 (011) surfaces, the largest calculated changes in the charge are observed for the BaTiO 3 and PbTiO 3 surface O atom (+0.230 e and +0.221 e, respectively). The change of the total charge in the second layer is negative and almost equal for both materials. For the BaTiO 3 crystal, this reduction by 0.249 e comes mostly from Ti atom (−0.154 e). In the PbTiO 3 crystal, the reduction by 0.230 e appears mostly due to a decrease of the Ti atom charge by 0.104 e, as well as a reduction of the Pb atom charge by 0.097 e.
IV. CONCLUSIONS
In summary, motivated by the scarcity of experimental investigations of the BaTiO 3 and PbTiO 3 surfaces and the contradictory experimental results obtained for the related SrTiO 3 surface, 19,20 we have carried out predictive electronic structure calculation to investigate the surface atomic and electronic structure of the BaTiO 3 and PbTiO 3 (001) and (011) surfaces. Using a hybrid B3PW approach, we have calculated the surface relax-ation of the two possible terminations (TiO 2 and BaO or PbO) of the BaTiO 3 and PbTiO 3 (001) surfaces, and three possible terminations (TiO, Ba or Pb, and O) of the BaTiO 3 and PbTiO 3 (011) surfaces. The data obtained for the surface structures are in a good agreement with previous LDA calculations of Padilla and Vanderbilt,4 the LDA plane-wave calculations of Meyer et al., 5 and in fair agreement with the shell-model calculations of Heifets et al. 16 According to our calculations, atoms of the first surface layer relax inwards for BaO and PbO terminated (001) surfaces of both materials. Outward relaxations of all atoms in the second layer are found at both terminations of BaTiO 3 and PbTiO 3 (001) surfaces. In BaTiO 3 , the rumpling of the TiO 2 -terminated (001) surface is predicted to exceed that of the BaO-terminated (001) surface by a factor of two. In contrast, PbTiO 3 exhibits practically the same rumplings for both (TiO 2 and PbO) terminations. Our calculated surface energies show that the TiO 2 -terminated (001) surface is slightly more stable for both materials than the BaO or PbOterminated (001) surface. The O-terminated BaTiO 3 and TiO-terminated PbTiO 3 (011) surfaces have surface energies close to that of the (001) surface. Our calculations suggest that the most unfavorable (011) surfaces are the Ba or Pb-terminated ones for both the BaTiO 3 and PbTiO 3 cases. We found that relaxation of the BaTiO 3 and PbTiO 3 surfaces for is considerably stronger for all three (011) terminations than for the (001) surfaces. The atomic displacements in the third plane from the surface for the three terminations of BaTiO 3 and PbTiO 3 (011) surfaces are still large. Finally, our ab initio calculations indicate a considerable increase of Ti-O bond covalency near the BaTiO 3 and PbTiO 3 (011) surface relative to BaTiO 3 and PbTiO 3 bulk, much larger than for the (001) surface. | 2007-10-10T20:42:07.000Z | 2007-10-10T00:00:00.000 | {
"year": 2007,
"sha1": "9497ac21aee40316eefbc1aa987e2a5aebed9190",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0710.2112",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "9497ac21aee40316eefbc1aa987e2a5aebed9190",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
14460490 | pes2o/s2orc | v3-fos-license | Displacement fields of point defects in two-dimensional colloidal crystals
Point defects such as interstitials, vacancies, and impurities in otherwise perfect crystals induce complex displacement fields that are of long-range nature. In the present paper we study numerically the response of a two-dimensional colloidal crystal on a triangular lattice to the introduction of an interstitial particle. While far from the defect position the resulting displacement field is accurately described by linear elasticity theory, lattice effects dominate in the vicinity of the defect. In comparing the results of particle based simulations with continuum theory, it is crucial to employ corresponding boundary conditions in both cases. For the periodic boundary condition used here, the equations of elasticity theory can be solved in a consistent way with the technique of Ewald summation familiar from the electrostatics of periodically replicated systems of charges and dipoles. Very good agreement of the displacement fields calculated in this way with those determined in particle simulations is observed for distances of more than about 10 lattice constants. Closer to the interstitial, strongly anisotropic displacement fields with exponential behavior can occur for certain defect configurations. Here we rationalize this behavior with a simple bead-spring that relates the exponential decay constant to the elastic constants of the crystal.
Introduction
The properties of crystalline substances often crucially depend on the structure and dynamics of imperfections of the crystal lattice. In particular, point defects such as interstitials and vacancies play a pivotal role in determining the stability, transport properties, growth characteristics, and mechanical behavior of materials. Recent impressive experimental advances, such as optical tweezers and confocal microscopy [1,2], now permit to study the fundamental properties of point defects in condensed matter systems with "atomistic" space and time resolution.
Recently, a number of experimental studies have focused on the structure and dynamics of point defects in two-dimensional assemblies of micrometer sized colloidal particles [3,4,5] and, in particular, on their effective interactions [6,7,8]. In studying such defect interactions the question arises to which degree they can be rationalized in terms of continuum elastic theory. As a first step towards answering this question, in this article we investigate numerically the disturbances caused by isolated interstitial particles and compare the results with the predictions of continuum theory. In carrying out such a comparison, it proves crucial that in solving the equations of elasticity theory boundary conditions are used that match those of the simulations. For the periodic boundary conditions usually applied in computer simulations, the displacement fields of single defects can be determined using the technique of Ewald summation familiar from electrostatics [9,10]. While elasticity theory properly describes the disturbances and interactions created by lattice imperfections on a larger scale, discrete lattice effects dominate on spatial scales of the order of few lattice constants.
The remainder of this paper is organizes as follows. In Sec. 2 we define the model and describe the numerical methods. The treatment of point defects in a two-dimensional elastic continuum is discussed in Sec. 3 and comparison with the numerical results is discussed in Sec. 4. For certain defect configurations one observes an exponential rather than algebraic decay of the displacement fields. This behavior can be understood in terms of a simple bead-spring model introduced in Sec. 5 with parameters related to the elastic constants of the material. Some concluding remarks are provided in Sec. 6.
Simulations
In this paper we study a two-dimensional crystal of soft particle interacting via the Gaussian potential [11,12,13] where r is the inter-particle distance and ǫ and σ set the energy and length scales, respectively. In the following, energies are measured in units of ǫ and distances in units of σ. This so-called Gaussian core model, used here as a generic model for a system of soft spheres, is a realistic description for the short-ranged effective interactions between polymer coils in solution [14]. In three dimensions, the Gaussian core model can exist as a fluid, a bcc-and an fcc-solid depending on temperature and density [12]. In two dimensions, the perfect triangular lattice is the lowest energy structure of Gaussian core particles at all densities [15]. Computer simulations indicate that also in this system of purely repulsive particles point defects such as interstitials, vacancies or impurity particles of different size display attractive (as well as repulsive) interactions both in two and three dimensions [16].
To study the displacement field of a single interstitial numerically, we prepare a configuration of particles arranged on the sites of a perfect lattice configuration and insert an extra particle of the same species. After insertion, the system is relaxed to a new minimum energy configuration by steepest descent minimization, i.e., we study the defect structure at T = 0. Typically, 70.000 steepest descent steps are carried out. The system we study here consists of N = 199.680 Gaussian core particles (without the extra particle) at a number density of ρ = 0.6σ −2 corresponding to a lattice constant a = (2/ √ 3ρ) 1/2 = 1.3872σ. Periodic boundary conditions apply to the simulation box of length L x = 416a and height L y = ( √ 3/2)480 a = 415.692a. The aspect ratio of the almost square simulation box is L y /L x = 0.99926.
We quantify the perturbation caused by the defect in terms of the displacement field [17] Here, r ′ i and r i denote the position of particle i with and without the defect, respectively. As we will see in the following sections, simple point defects generate remarkably intricate displacement patterns that can be understood in terms of elasticity theory only on large length scales.
At T = 0, the elastic constants describing the macroscopic response of the system to perturbations can be calculated as a function of density from simple lattice sums. For a density of ρ = 0.6σ 2 , the Lamé coefficients (see Sec. 3) of the perfect triangular lattice have values λ = 1.1487 ǫσ −2 and µ = 0.06018 ǫσ −2 . At this density, the pressure is p = 0.5442 ǫσ and the energy density is e = 0.2691 ǫσ −2 corresponding to an energy per particle of E/N = 0.4485ǫ. The bulk modulus, which in two dimensions is related to the the Lamé coefficients by K = λ + µ, has a value of K = 1.2089ǫσ −2 .
Elasticity Theory
While close to a point defect the displacement field is highly anisotropic and strongly dependent on the atomistic details of the interactions, for large distances elasticity theory is expected to be valid. The differential equations describing the equilibrium of an elastic continuum are usually expressed in terms of the strain tensor [17] ǫ ij (r) = 1 2 where u i denotes the i-component of the displacement u and r i the i-th component of the position r. For a given external volume force f (r) with components f i acting on an isotropic system such as a crystal on a triangular lattice, Hook's law leads to the equilibrium condition for the strain: Here, λ and µ are the so-called Lamé coefficients and summation over repeated indices is implied. Solving this equation for a singular force yields the Green's function from which the response of the elastic continuum to an arbitrary force can be obtained by integration.
To model the displacement field caused by the introduction of point defects using linear continuum elasticity theory, we determine the displacement field caused by two pairs of opposing forces, one pair acting along the x-axis and the other one along the y-axis [18,19,20]. This idealized model of a defect is equivalent to inserting a small circular inclusion into a hole of different size [19]. Each force of this pair is of equal magnitude F but with opposite sign acting on two points separated by a small distance h. Such a force pair exerts a zero net force on the material. In the limit h → 0 where the force F → ∞ in a way such that F h remains constant, the equilibrium condition for the displacement can be written as Assuming that the displacement can be written as the derivative of a potential, This equation is the Poisson equation of electrostatics with a singular disturbance.
Since, as noted above, K(r) = − ln(r)/2π is a solution of ∆K = −δ(r) (see, for instance, Ref. [21]), we obtain the Green's function from which the displacement field u(r) follows by differentiation according to Equ. , In comparing the results of particle simulations with those of elasticity theory it is important to realize that the displacement fields predicted by continuum theory are of a long-range nature. Therefore, it is crucial that corresponding boundary conditions are used in both cases. All simulations discussed in this paper are done with periodic boundary conditions in order to minimize surface effects and preserve the translational invariance of the perfect lattice. Hence, also the continuum calculations need to be carried out with periodic boundary conditions.
Since the defect fields for the infinite material are long-ranged, the displacement field in the periodic system cannot be obtained by simply summing up the contributions of the periodic images. In fact, such a naive summation of the contribution of all image defects diverges. A more appropriate treatment that avoids this problem consists in determining the Green's function of the Poisson equation (7) for periodic boundary conditions. In this case, the solution of this equation in two dimensions, known from electrostatics [9,10], can be written as Ewald sum of a logarithmic potential embedded in a neutralizing background, Here, E i (x) = x −∞ (e t /t) dt is the exponential integral. The first sum is over all lattice vectors l in real space and the second sum is over all reciprocal vectors k in Fourier space. The adjustable parameter η, set to a value of η = 6/L x here, determines the rate of convergence of the two sums and A is the area of the rectangular simulation cell. The Fourier space sum can be evaluated accurately using about 2,500 reciprocal space vectors. From Equ. (10) for the scalar function φ(r) the displacement field of a point defect in a system with periodic boundary conditions is found by differentiation, For the systems considered in this paper, the real space sum may be truncated after the first term. Since the value of F h/2π(λ + 2µ) is undetermined, the parameter γ ≡ F h/2π(λ + 2µ) is treated as a fit parameter in the following. The Ewald sums of the above equations describe the effects of "image defects" at the center of the periodically replicated domains.
Results
First, we study the displacement field of a single interstitial. To generate such a defect, we insert an extra particle of the same species into a perfect 2d-crystal on a triangular lattice. After insertion, the system is relaxed to a new minimum energy configuration by steepest descent minimization, i.e., we study the defect structure at T = 0. Typically, 70.000 steepest descent steps are carried out. In each step each particle is moved in the direction of the force acting on the particle where the absolute value of the displacement in chosen to be small enough to ensure that the energy of the system decreases in each step. The extra particle can deform the crystal in different ways [4] and produces displacement fields of different symmetries (see Fig. 1). In one configuration, called I 2 interstitial or crowdion and shown in Fig. 1a, the additional particle pushes one particular particle of the crystal out of its equilibrium position. Both the original particle and the additional particle arrange themselves at equal distance around the lattice position of the original particle. The displacement pattern arising for this type of interstitial has two-fold symmetry and, of course, occurs in all three low-index lattice directions with equal probability. One may suspect that this defect configuration, with a symmetry that differs from the symmetry of the underlying triangular lattice, is caused by the rectangular periodic boundary conditions that are applied to the system. To rule out this possibility, we have repeated the calculation with hexagonal periodic boundary conditions obtaining the same result.
Another low-energy defect configuration is the I 3 interstitial with three-fold symmetry (see Fig. 1b). In this case the interstitial particle is located at the center of a basic lattice triangle and pushes its neighbors outward from their original positions. A third important interstitial configuration is the I d interstitial or dumbbell interstitial shown in Fig. 1c. In the 2d Gaussian-core model under the conditions studied here the I 2 pattern has a slightly lower energy than the I 3 interstitial and the I d interstitial.
The energy difference between a I 2 and a I 3 interstitial is 0.000674ǫ and the difference between I 2 and I d is 0.000665ǫ. All three displacement patterns are important for the diffusion of interstitials. An I 2 interstitial is very mobile in the direction of its main axis. The I 3 and I d forms are visited as intermediate configurations when the I 2 interstitial changes the orientation of its main axis and hence its direction of motion [22]. Next, we compare the displacement fields determined numerically with the predictions of elasticity theory. In particular, we verify to which extent the 1/rbehavior modulated by the periodic boundary conditions and embodied in Equ. (11) is realized in the particle system. The complex displacement patterns of the various interstitial configurations shown in Fig. 1 obviously differ from this expectation, at least near the defect, and indicate that continuum theory is not applicable in this region. Far away from the defect, however, the perturbation caused by the defect is small and the response of the material should be described accurately by linear elasticity theory.
The magnitude |u(r)| of the displacement vector u(r) is shown as a function of the distance from the interstitial in Fig. 2 for the I 2 defect configuration. Each point in the figure corresponds to one individual particle. For short distances, the displacement magnitude is not a unique function of the distance r reflecting the anisotropic nature of the defect. For larger distances, however, the displacement magnitude is mostly determined by the distance r. Eventually, however, the periodic boundary conditions lead to a spread of the displacement magnitude for even larger distances and a splitting into two branches corresponding to the xand y-directions and the directions along the diagonals, respectively. In the regime where u(r) behaves isotropically, the . Displacement magnitude |u(r)| as a function of distance from the defect r for the I 2 interstitial. Each red dot corresponds to one particle. The solid line represents the γ/r behavior. Here, γ = 0.2291σ 2 was used as this value yields the best fit of the results obtained vie Ewald summation to the results of the particle simulations in the far field. Inset: angle θ between the displacement vector u the position vectors r as a function of the distance r from the defect site. displacement follows the approximately 1/r-form predicted by elasticity theory for a point defect in an infinitely extended medium. The orientation of the displacement vector u(r), depicted in the inset of Fig. 2, behaves in an analogous way. The angle θ between u(r) and the position vector r, shown as a function of the distance r from the defect, is not a unique function of r near the defect. For larger r, θ vanishes indicating that in this distance regime the displacement vector points straight away from the defect. At even larger distances, the periodic boundary conditions imposed on the system eventually cause the angle θ to spread again.
The displacement fields calculated according to Equ. (11) and numerically for an interstitial in the I 2 configuration are compared in Fig.3. In this figure, the displacement components u x and u y are depicted as a function of the distance from the defect along the x-axis and y-axis, respectively. The prediction of continuum theory, calculated using the Ewald summation of Equ. (11), agrees well with the displacement field of the particle system for distances larger than about 10 lattice constants.
Harmonic Model
Near the defect, the predictions of continuum theory differ from the simulation results. The deviation is particularly pronounced in the direction of the main axis of distortion of the I 2 defect, in which the displacement appears to decay exponentially up to a distance of about ≈ 10 a. This unexpected exponential behavior can be understood in terms of a simple model with harmonic interactions. This model consists of a onedimensional chain of particles in which each particle is connected to its two neighbors with springs of force constant k 1 (except the first and last particle, which are coupled only to their neighbors on the right and left, respectively). In addition, each particle is attached to a fixed lattice position with another spring of force constant k 2 . The Figure 3. Displacement components ux and uy of the I 2 interstitial as s function the of the distance along the x-axis and y-axis, respectively (solid lines). Here, the direction of largest displacement of the I 2 defect is oriented in x-direction. Also plotted is the displacement computed from continuum theory according to Equ. (11) (dashed line), simple 1/r-behavior (dotted line) and the displacement obtained for the simple mechanical model described in the main text (dash-dotted line). The inset shows the region close to the defect location. As in Fig. 2 a defect strength of γ = 0.2291σ 2 was used for the evaluation of the displacement from elasticity theory.
Hamiltonian of this system is
where N + 2 is the number of particles, x j is the position of particle j and b is the equilibrium distance between two neighboring particles. In the minimum energy configuration of this chain, the particles are arranged such that x j = jb. We now imagine that particle 0 is pushed to the right by a distance of u 0 while particle N + 1 is kept fixed at x N +1 = (N + 1)b. If the system is then relaxed to a new energy minimum, all other particles will be displaced from their original positions too. For this simple model, the response of the system to the displacement of the first particle can be calculated analytically by direct matrix inversion (see Appendix A). In the large N limit, one finds that the displacement of the particles from their original position decays exponentially with their position, where u j is the displacement of particle j due to the forced displacement u 0 of the first particle. The decay constant α is related to the force constants of the model by To compare the prediction of this simple model with the simulation results we have to determine the force constants k 1 and k 2 felt by the particles in the main axis of the defect. While the force constant k 1 arises from interactions within this main axis, the force constant k 2 is related to interactions of the particles in the main axis with those from adjacent rows. Accordingly, we determine k 1 by calculating numerically the energy change caused by slightly displacing one single particle in a one-dimensional row of otherwise fixed Gaussian core particles without the presence of the neighboring rows. The distance of the particles in the row is chosen to be equal to the lattice constant at the density ρ = 0.6σ −2 considered throughout the paper. From the energy as a function of the displacement one obtains a force constant of k 1 = 0.015ǫ/σ 2 . To determine the force constant k 2 we calculate the energy change caused by translating a whole row of particles in the perfect crystal. The particles in the row are fixed with respect to each other and the remaining particles are kept at their lattice positions. From the energy change per moved particle a force constant of k 2 = 0.0013ǫ/σ 2 follows. The decay constant of α = 0.29 calculated according to Equ. (14) with these force constants is in perfect agreement with the computer simulation results shown in Fig. 3.
For a system in which only nearest neighbor interactions are important, the force constants k 1 and k 2 can be simply related to the bulk modulus K and the shear modulus µ. Then, the force constant k 1 is given by where v(a) is the pair potential at distance a. Since in this case the elastic moduli are given by and one obtains To the extent that the response of the system to shear is determined by the interaction of neighboring parallel rows of particles, the energy density caused by shifting a whole row of atoms between two fixed ones is the same as that of a shear of appropriate magnitude. Accordingly, the force constant k 2 is related to the shear modulus by This expression remains also valid if interactions beyond nearest neighbors are included between adjacent rows of particles. In terms of the elastic constants, the constant α describing the exponential decay of the displacement field along the principal axis can be expressed as or, in terms of the Poisson ratio ν, For a density of ρ = 0.6σ −2 , inserting the Poisson ratio of ν = 0.905151 determined from a simple lattice sum yields α ≈ 0.25, only slightly different from the correct value α ≈ 0.29. This deviation occurs, because in the Gaussian core model at the density ρ = 0.6σ −2 interactions between non-nearest neighbor particles are important in determining the elastic constants (in fact, considering only nearest neighbors would produce a negative shearing modulus µ in this case). For systems, in which only nearest neighbor interactions are relevant, the above expression is expected to hold accurately.
Conclusion
Point defects in two-dimensional crystals, such as interstitials and vacancies, can assume configurations with symmetries that vary from the symmetry of the underlying triangular lattice. While close to the defect the displacement field is highly anisotropic and strongly dependent on the atomistic details of the interactions, for large distances elasticity theory, which predicts isotropic behavior, is valid. For the particular I 2 interstitial configuration, the displacement decreases exponentially with distance along the main defect axis. The decay constant is simply related to the material properties via the Poisson ratio, which measures the ratio between transversal and axial strain upon stretching. In comparing the displacement fields computed from particles simulations with those obtained with continuum elasticity theory it is crucial to use equivalent boundary conditions in both cases. Since particle simulations are usually carried out with periodic boundary conditions, also the differential equations of elasticity theory need to be solved for a periodic system. We have shown here that Ewald summation, a technique routinely used in computer simulations to determine the electrostatic interactions of charges and dipoles, can be used for this purpose. In this method the sum over all interactions with periodic image defects is split into two sums in real space and reciprocal space, respectively. This particular treatment of the long-ranged nature of displacement fields effectively introduces a neutralizing background that leads to convergent sums. Note that exactly the same expression apply also to a system that is enclosed in a rigid container. Outside a core region near the defect, displacement patterns determined using such Ewald summation agree perfectly with those calculated in particle simulations.
Appendix A.
In this appendix we calculate the response at T = 0 of the one-dimensional beadspring model of Sec. 5 to a forced displacement of the first particle in the chain. The potential energy of the N + 2 particles, located at positions x j , is given by where b is the equilibrium distance and k 1 and k 2 are force constants. The vector x = x 0 , x 1 , · · · , x N +1 includes the positions of all particles. Minimizing the potential energy with respect to the particle positions x j by requiring that for all j, one finds that at the potential energy minimum the particle positions are x j = bj. We now displace particle 0 by an amount u 0 from its original position x 0 = 0 and keep particle N + 1 fixed at position (N + 1)b. If we hold particle 0 at this new position while minimizing the potential energy, all particles from 1 to N will move to new equilibrium positions. Thus, the minimum energy configuration of the system is a function of the displacement u 0 of particle 0, which may be viewed as a parameter that is controlled externally and perturbs the system. To make this distinction between the displacement of particle 0 and that of all other particles more explicit, we denote u 0 with an extra symbol, ξ = u 0 . The displacement u j of the particles j = 1, · · · , N is then a function of ξ, where x j (ξ) and x j (0) denote the particle position in the minimum energy configuration with and without perturbation, respectively. In the following, we will calculate the displacements u j (ξ) as a function of the perturbation strength ξ.
Since condition (A.2) defines the position of the energy minimum as a function of the perturbation strength ξ, its derivative with respect to ξ must vanish, Once z j = ∂x j (ξ)/∂ξ is known, x j (ξ) can be obtained by integration.
For the bead-spring model considered here, the first and second derivatives of the potential energy with respect to the particle coordinates are given by and To solve Equ. (A.11) we have to invert the symmetric tridiagonal matrix H ij . For this particular matrix, the inverse matrix is known analytically [23], Integration with respect to ξ then yields where the integration constant is given by C i = x i (0). Thus, the displacement of particle j is proportional to the displacement of particle 0 and decays exponentially with the distance from the origin, with a decay constant α that depends on the force constants k 1 and k 2 only. | 2008-05-20T14:49:34.000Z | 2008-05-20T00:00:00.000 | {
"year": 2008,
"sha1": "b2a2746d2920bcc96ff8ee2c68f272b670b85c66",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/0805.3094",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "b2a2746d2920bcc96ff8ee2c68f272b670b85c66",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Physics"
]
} |
246067090 | pes2o/s2orc | v3-fos-license | Structural and functional deficits and couplings in the cortico-striato-thalamo-cerebellar circuitry in social anxiety disorder
Although functional and structural abnormalities in brain regions involved in the neurobiology of fear and anxiety have been observed in patients with social anxiety disorder (SAD), the findings have been heterogeneous due to small sample sizes, demographic confounders, and methodological differences. Besides, multimodal neuroimaging studies on structural-functional deficits and couplings are rather scarce. Herein, we aimed to explore functional network anomalies in brain regions with structural deficits and the effects of structure-function couplings on the SAD diagnosis. High-resolution structural magnetic resonance imaging (MRI) and resting-state functional MRI images were obtained from 49 non-comorbid patients with SAD and 53 demography-matched healthy controls. Whole-brain voxel-based morphometry analysis was conducted to investigate structural alterations, which were subsequently used as seeds for the resting-state functional connectivity analysis. In addition, correlation and mediation analyses were performed to probe the potential roles of structural-functional deficits in SAD diagnosis. SAD patients had significant gray matter volume reductions in the bilateral putamen, right thalamus, and left parahippocampus. Besides, patients with SAD demonstrated widespread resting-state dysconnectivity in cortico-striato-thalamo-cerebellar circuitry. Moreover, dysconnectivity of the putamen with the cerebellum and the right thalamus with the middle temporal gyrus/supplementary motor area partially mediated the effects of putamen/thalamus atrophy on the SAD diagnosis. Our findings provide preliminary evidence for the involvement of structural and functional deficits in cortico-striato-thalamo-cerebellar circuitry in SAD, and may contribute to clarifying the underlying mechanisms of structure-function couplings for SAD. Therefore, they could offer insights into the neurobiological substrates of SAD.
INTRODUCTION
Social anxiety disorder (SAD) is a prevalent and disabling psychiatric disorder characterized by notable and persistent fear or anxiety in social situations [1]. People with SAD are intensely afraid of possible scrutiny and negative evaluation by others and gradually avoid participating in social activities, resulting in emotional, cognitive, and behavioral disabilities, as well as severe social function impairments [2]. Approximately 7.1-12.1% of people are estimated to suffer from SAD in their lifetime [1], and approximately 90% of SAD patients have at least one comorbid disorder [3]. Given the severe functional impairments of SAD, it is of great importance to understand its neuropathology and identify potential neural biomarkers, which may be crucial for achieving early diagnosis and timely intervention.
Over the last two decades, a large body of neuroimaging (particularly magnetic resonance imaging (MRI)) research has begun to explore the structural and functional abnormalities in SAD, but the results are heterogeneous and in need of validation and replication [4,5]. On the one hand, evidence from structural MRI (sMRI) studies regarding SAD has indicated a widespread pattern of gray matter (GM) differences in a majority of cortical and subcortical regions, as well as the cerebellum [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22]. Notably, those findings showed much heterogeneity, to which many confounding issues, such as demographic and methodological discrepancies, may contribute. Specifically, high comorbidity in SAD may significantly complicate the clinical course and diagnosis and make it challenging to study the pure and specific neuropathology of SAD. Nevertheless, different proportions of SAD patients comorbid with different psychiatric disorders have been included, but the effects of comorbidity have not been well handled in previous studies. Besides, previous sMRI studies have typically involved a small sample size of participants with SAD, with the majority of studies including fewer than 30 patients, and few studies justified the sample size or conducted a power calculation. It is well established that studies with small sample sizes are highly susceptible to inflated risks of false positives and negatives [23], and a sample size of 20-30 participants is not sufficient to detect reproducible relationships between the brain and behavior measures regardless of analytic methods [24]. Furthermore, many previous studies on SAD were based on region-of-interest (ROI) analyses, although it has been reported that predefined cerebral ROIs are not isolated due to biological factors, which makes it difficult to correct for multiple comparisons and dramatically increases the risk of type II errors [25]. Hence, to probe the pure neurobiological underpinnings of SAD, it is indispensable and beneficial to explore structural deficits via a whole-brain voxel-based morphometry (VBM) analysis using a sufficient sample size of non-comorbid SAD patients.
On the other hand, functional MRI (fMRI) studies have been performed to determine the functional anomalies in SAD in the context of various emotional, social, and cognitive, as well as other nonspecific tasks [26,27]. The most consistent findings from these studies were alterations in SAD patients compared to healthy controls (HCs) in the frontolimbic circuitry termed the fear circuitry, which includes hyperactivity in the dorsolateral prefrontal cortex (dlPFC), ventromedial prefrontal cortex (vmPFC), anterior cingulate cortex (ACC), insula, amygdala, and hippocampus/parahippocampus (ParaHIP) [28]. This model posits that dysfunctional top-down modulation is pivotal in the emotional hyperactivity and diminished cognitive processing observed in patients with SAD [29]. Moreover, increasing evidence from some recent studies has revealed other SAD-related functional alterations beyond the conventional fear circuitry [26,30,31]. In 2014, a systematic review and meta-analysis conducted by Brühl et al. updated the neurofunctional model of SAD; abnormal fear circuitry in SAD was confirmed, and new findings of hyperactivated medial parietal and occipital regions in response to SADrelated stimuli, as well as hypoconnectivity of parietal, limbic, and executive network regions, were added [26]. Nevertheless, the majority of those functional results were based on task-fMRI using various types of tasks, while it remains to be evaluated whether an analogous pattern of functional anomalies occurs in resting-state brain physiology (i.e., resting-state fMRI (rs-fMRI)), which could offer distinct insights into the intrinsic neurobiology of SAD without the potential confounding effects of task performance [32]. Indeed, the results from the most recent systematic review based on resting-state neuroimaging studies specifically for SAD indicated that the neurobiological substrates of SAD may be, to some degree, different from those classic models (i.e., summarized by Etkin et al.) that primarily originated from task-based studies [33], as this review showed that aberrant (hyper or hypo) connectivity between the amygdala and parietal, temporal, and frontal areas, and abnormal (hypo-or hyper) activity in frontal regions were the most consistently implicated in patients with SAD, as shown by a range of neuroimaging analyses. Additionally, this review suggested that findings from rs-fMRI were confounded by technological and methodological factors and sample characteristics and that further studies with larger samples and consistent analysis methods are warranted.
Furthermore, despite the widespread structural and functional deficits identified in previous studies, multimodal neuroimaging studies on structure-function coupling are relatively scarce in SAD; that is, existing neuroimaging studies on SAD have mainly been performed using a single MRI modality, and few multimodal analyses have been conducted to probe structural and functional deficits and their relationships with the diagnosis of SAD [34][35][36]. Indeed, some evidence has demonstrated that functional alterations in brain regions are accompanied by structural changes in the corresponding areas [37,38], and that functional connectivity (FC) and networks could be predicted by structural substrates [39]. A previous study reported that rs-fMRI parameters were variable and may exist where there were no direct structural connections, but their persistence, strength, and spatial statistics were confined by the underlying anatomical structure of the human brain [40]. In other words, structural connections shape and place constraints on FC across the brain network at various spatial scales [41]. As a result, normal structure-function coupling is vital for the brain, while the disrupted coupling of structure and function can be found in many neurological and psychiatric disorders [42][43][44]. Based on this evidence, a "brain structure-function behavioral coupling" psychoradiological hypothesis indicates that structural alterations in the brain may give rise to clinical syndromes via an impact on disrupted FC [45,46]. Therefore, a study combining sMRI and fMRI is likely to offer more information on the underlying relationships among brain structure, function, and SAD diagnosis.
Taking these issues into consideration, the present study aimed to determine resting-state FC (rs-FC) alterations in brain areas with structural deficits in non-comorbid adult patients with SAD and to explore potential mechanisms of structural-functional couplings for SAD. To achieve these goals, sMRI and rs-fMRI analyses were conducted in the current study. For structural analysis, an optimized and standardized VBM-Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra (DARTEL) procedure was conducted to measure the GM volume (GMV) [47], a well-validated and widely used index to investigate the neurostructural signatures of GM, which may represent the numbers and sizes of glial cells, unmyelinated neurons, and the volume of the synapses [48]. Turning now to rs-fMRI analysis, a seed-based rs-FC metric that reflects temporal correlations or couplings (i.e., synchronous and coherent fluctuation) of neuronal activity patterns between specific regions (i.e., seeds) with other spatially segregated areas was investigated [49,50]. As we intended to explore structural-functional couplings, seed-based rs-FC analysis is suitable for the current study, that is, regions with structural deficits could be a priori seeds. In view of existing findings, we hypothesized that a cohort of patients with SAD, compared to a group of HCs, may show altered GMV mainly in subcortical nuclei such as the striatum [16,21] and abnormalities in large-scale cortical-subcortical circuitry. Considering the sparse and mixed results (i.e., increased vs. decrease) on SAD-related GM change [26,33], we do not intend to hypothesize the alteration direction. In addition, it could be speculated that the intrinsic functional networks at rest may mediate the correlations between structural deficits and SAD diagnosis.
SUBJECTS AND METHODS Participants and procedures
All procedures in the present study adhered to the ethical standards of the Declaration of Helsinki and the ethical principles in the Belmont Report. This study was approved by the Medical Research Ethics Committee of West China Hospital at Sichuan University. Prior to the experiment, written informed consent was obtained from all participants after they were given a full explanation of the procedures.
This study included 49 right-handed adult SAD patients without any comorbid psychiatric disorders. Patients were recruited from the Mental Health Center of the West China Hospital at Sichuan University. The diagnosis of SAD was confirmed by the consensus of two experienced clinical psychiatrists in accordance with the criteria of Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) through the Structured Clinical Interview for DSM Disorders (SCID) [51]. As the power analysis using G Power software [52] indicated that we needed a sample of at least 102 participants to detect a medium-sized effect (Cohen's d = 0.5, α = 0.05, 1-β = 0.8) to conduct a two-sample t-test, 53 HCs were recruited from the local community through advertisements and were matched to the patients in terms of sex, age, and handedness for comparison analysis, and the SCID-Non-Patient Version was conducted to confirm the lifetime absence of psychiatric and neurological illness. The exclusion criteria for all participants were as follows: 1) comorbidity with other axis I psychiatric disorders; 2) axis II antisocial or borderline personality disorders (verified by the SCID); 3) a history of substance dependence or abuse; 4) learning or developmental disorders; 5) a history of head injury; 6) the presence of major neurological or physical diseases; 7) family history of mental disorders; and 8) current pregnancy or claustrophobia and other contraindications to MRI examination. Individuals were also excluded if they were aged under 18 or over 60 years to minimize age-related effects.
X. Zhang et al.
Illness onset was determined as the period between the first reported/ observed alterations in psychological/behavior state to the development of disease when the patients participated in the study [53], with the information provided by patients, their family members, and medical records. To evaluate and compare the levels of social anxiety between SAD patients and HCs, the self-administered Liebowitz Social Anxiety Scale (LSAS) [54] was administered to assess all participants. As the most commonly used clinical scale in SAD studies, the 24-item LSAS provides scores for both fear factor (LSASF) and social avoidance factor (LSASA), and the total score (LSAST) is their sum, which has shown good validity and reliability in Chinese populations [55].
Image acquisition and preprocessing
Image acquisition. Whole-brain structural and functional MRI images were acquired on a 3.0 T MR scanner (Siemens Trio, Erlangen, Germany) with a 12-channel head coil. During the scans, the subjects were instructed to keep their eyes closed, relax but not to sleep, and lie as still as possible. Earplugs were used to reduce scanner noise, and foam pads were used to restrict head motion as much as possible. High-resolution three-dimensional T1-weighted images were acquired using a spoiled gradient-recalled sequence: repetition time (TR)/echo time (TE) = 1900 ms/2.26 ms, flip angle = 9°, 176 sagittal slices, slice thickness = 1 mm, field of view (FOV) = 240 × 240 mm 2 , data matrix = 256 × 256, voxel size = 1 × 1×1 mm 3 , and inplane resolution = 0.94 × 0.94 mm 2 . The rs-fMRI data were obtained with the gradient echo-planar imaging sequence: TR/TE = 2000 ms/30 ms; flip angle = 90°; acquisition matrix = 64 × 64; FOV = 240 × 240 mm 2 ; thickness = 5.0 mm, without gap; voxel size = 3.75 × 3.75 × 5 mm 3 ; and 205 volumes. Each scan was inspected by an experienced neuroradiologist to rule out visible movement artefacts and gross structural abnormalities before image preprocessing.
Image preprocessing. Preprocessing of structural images was performed using Statistical Parametric Mapping software (SPM12; Welcome Department of Cognitive Neurology, London, UK; http://www.fil.ion.ucl.ac.uk/ spm/) [56]. First, all MRI images were manually reoriented on the anteriorposterior commissure line for better registration. Second, the highresolution T1-weighted images were segmented into GM, white matter (WM), and cerebrospinal fluid (CSF) via the new segmentation tool in SPM12. Third, the GM data were aligned, resampled to 2 × 2 × 2 mm 3 , normalized to Montreal Neurological Institute (MNI) space, modulated for the preservation of GMV, and smoothed with an 8-mm full-width at halfmaximum (FWHM) Gaussian kernel using DARTEL in SPM12 [57].
The rs-fMRI data were preprocessed using the Data Processing Assistant for Resting-State fMRI (DPARSF 4.3, http://rfmri.org/DPARSF), which is based on SPM (http://www.fil.ion.ucl.ac.uk/spm/) and the toolbox for Data Processing & Analysis of Brain Imaging (DPABI 4.3, http://rfmri.org/DPABI) [58] and includes the following steps: 1) removal of the first 10 volumes and slice timing correction; 2) realignment and correction for head motion (three SAD patients and one HC with excessive head motion above 2.5 mm or 2.5°in any direction were excluded), in which the framewise displacement (FD) was calculated for subsequent analysis; 3) spatial normalization to MNI space including the new segmentation and DARTEL; 4) regressing out linear trends, Friston's 24 head motion parameters [59], WM signal, CSF signal, and global signal; 5) resampling into 3 × 3 × 3 mm 3 and spatial smoothing with a 6-mm FWHM Gaussian kernel; 6) temporal bandpass filtering (0.01-0.08 Hz); and 7) calculating rs-FC based on the clusters showing significant group differences in the VBM analysis as the seed areas. The time courses were extracted from the seed areas, and the correlation coefficients between those time courses and all remaining brain voxels were computed. Finally, the correlation maps were z-normalized using Fisher's r-to-z transformation to improve the normality of the partial correlation coefficients.
Statistical analysis
Demographic and clinical data analyses. The differences in the demographic and clinical data between two groups were conducted through a chi-square test for discrete variables (i.e., sex) and two-sample t-tests for continuous variables (i.e., age, LSAST, LSASA, LSASF, and mean FD) using IBM SPSS Statistics 22.0.
VBM and FC analyses. Whole-brain voxel-wise comparisons of GMV between groups were performed using two-sample t-tests, with age, sex, and total intracranial volume (TIV) as covariates of no interest in SPM12. Group differences in rs-FC were conducted in DPABI software, in which two-sample t-tests were used to compare rs-FC between SAD patients and HCs with age, sex, and mean FD as covariates of no interest. The Gaussian random field (GRF) theory [60,61] was performed to control for multiple comparisons with a significance threshold of a voxel-wise value of P < 0.001 and cluster probability of P < 0.05 [62]. In addition, we applied the false discovery rate (FDR) to correct for four seeds in the FC analyses.
Correlation analyses. To identify the associations between the structural/ functional changes and clinical characteristics, the average GMV and rs-FC values in the significant clusters with between-group differences were extracted respectively, and then we performed a partial correlation analysis between the aforementioned mean values and clinical features (i.e., LSAST, LSASA, LSASF, and disease duration) using sex, age, and TIV/mean FD as covariates in the SAD group via IBM SPSS Statistics 22.0.
Mediation analysis. To explore the potential mediating roles of functional deficits in the relationship between the structural abnormalities and SAD diagnosis, a mediation analysis was performed with the SPSS macro PROCESS that included the bootstrapping approach [63,64]. To this end, GMV of the brain regions with significant between-group differences was considered the independent variable (X); corresponding FC of identified regions was the mediator variable (M); SAD diagnosis served as the dependent variable (Y); and age, sex, TIV, and mean FD were regarded as covariates. Then, mediation analysis was conducted to investigate the direct (i.e., path c' representing the relationship between X and Y after controlling for M) and indirect relationships between structural deficits and SAD diagnosis. The indirect effect represented the product of path a (i.e., the relationship between X and M) and path b (i.e., the relationship between M and Y after adjusting for X). The estimation of the indirect effect was considered significant if zero was not included in the bootstrapped 95% confidence intervals (CIs) (number of samplings= 5000). FD framewise displacement, HCs healthy controls, LSAST, LSASF, and LSASA total score and fear and avoidance factor scores on the Liebowitz Social Anxiety Scale (LSAS), SAD social anxiety disorder. Data are presented as the means ± standard deviations (minimum−maximum). *P value obtained using a chi-square test. **P value obtained using a two-sample t-test.
Demographic and clinical characteristics
One hundred and two participants (49 SAD vs. 53 HC) were included in the VBM analysis, while 98 subjects (46 SAD vs. 52 HC) were included in the rs-FC-related analysis because of the removal of 4 participants (3 SAD vs. 1 HC) due to head motion. No significant group differences appeared in terms of sex composition and age; the patients with SAD had significantly higher LSAS scores than the HCs (Table 1). Besides, there were no significant differences in mean FD between the two groups [t (98) = 0.519, P = 0.605].
Group differences in GMV
The whole-brain voxel-wise analysis demonstrated that the SAD patients, compared to the HCs, had significantly decreased GMV in the right thalamus, bilateral putamen, and left ParaHIP. No areas showed larger GMV in the SAD group ( Fig. 1 and Table 2).
Group differences in FC
Compared to the HCs, patients with SAD had increased rs-FC between the left putamen and left middle temporal gyrus (MTG)/ superior temporal gyrus (STG), and decreased rs-FC between the left putamen and left cerebellum; increased rs-FC between the right putamen and left STG, and decreased rs-FC between the right putamen and right cerebellum; increased rs-FC between the right thalamus and bilateral MTG/STG and right inferior temporal gyrus (ITG), and decreased rs-FC between the right thalamus and limbic lobe/ACC, supplementary motor area (SMA)/superior frontal gyrus (SFG), bilateral cerebellum, and left thalamus ( Fig. 2 and Table 2). When the seed area was located in the left ParaHIP, there were no regions with significantly different rs-FC.
Correlations between structural/functional deficits and clinical characteristics
After controlling for the confounders of sex, age, TIV, and mean FD, the partial correlation analysis showed that decreased GMV in the bilateral putamen was significantly inversely related to SAD duration (left: r = −0.34, P = 0.021; right: r = −0.407, P = 0.005), while decreased rs-FC between the right thalamus and limbic lobe/ACC (r = 0.355, P = 0.020) and decreased rs-FC between the right thalamus and cerebellum (r = 0.321, P = 0.036) were positively correlated with SAD duration (see Supplementary Materials). There was no significant association between the structural/functional alterations and LSAS scores.
Mediation analyses
In the mediation analysis, we found a All clusters survived correction for multiple comparisons using Gaussian random field theory with a significance threshold of a voxel-wise value of P < 0.001 and cluster probability of P < 0.05.
DISCUSSION
In the present study, we demonstrated that SAD patients had structural and functional deficits in the cortico-striato-thalamocerebellar circuitry and uncovered significant mediating effects of functional anomalies on the links between structural deficits and SAD diagnosis. To our knowledge, the current study was the first to combine VBM and rs-FC to reveal structural and functional deficits and couplings in relation to SAD, which may be integral to the neuropathology of SAD and to some degree contribute to the future early diagnosis and targeted therapy in SAD.
Cortico-striato-cerebellar circuitry in SAD First, the current study pointed to the dysfunctional corticostriato-cerebellar circuitry in SAD. The VBM analysis revealed decreased GMV in the bilateral putamen in SAD patients compared to HCs, while a recent multicentre mega-analysis showed that patients with SAD had larger GMV in the putamen [16]. Without regard to the alteration directions, those findings at least indicated the involvement of putamen in SAD [65]. Robust evidence has accumulated that the putamen (i.e., a part of the dorsal striatum) is involved in social learning, motor, and cognitive control, reward processing, and cognitive and emotional regulation [66,67]. It has been documented that SAD patients lack a processing preponderance in the putamen for social rewards compared to social punishments [68], so structural alterations in the putamen may be responsible for its involvement in the imbalance of the neural approach-avoidance motivation system underlying SAD. Meanwhile, increased FC between the putamen and the MTG/STG and decreased FC between the putamen and cerebellum posterior lobe were also presented in the current analysis. The STG/MTG is a crucial component of the perceptual system involved in facial emotion processing, social threat evaluation [69], analysis of the dispositions and intentions of others' actions [70,71], visual perception and mental imagery [72], and integration of interoceptive information with information about the current environmental situation [73], all of which may be related to SAD characteristics such as an excessive focus on others' intentions and facial expressions, excessive fear for negative evaluation and scrutiny by others [74], and increased saliency of the social situations when SAD patients envision themselves in hypothetical scenes [75]. Interestingly, a recent study also reported increased intrinsic rs-FC in the left MTG in families genetically enriched for SAD, indicating the crucial roles of the MTG as a network hub in the socially anxious brain [76].
Combined with the findings that striatal dysfunction was closely related to the information processing biases in SAD [77], increased FC between the putamen and the MTG/STG may reflect enhanced input of undue focus and speculation on social and individual stimuli into the striatum and subsequent cognitive and emotional dysregulation. Another interesting finding was decreased FC between the putamen and cerebellum posterior lobe in SAD patients. Not surprisingly, cerebellar structural and functional anomalies have been implicated in the emotional dysregulation associated with various psychiatric disorders [78][79][80], especially anxiety-related symptoms (e.g., hyperarousal) and psychosis [11,[81][82][83], and abnormal resting-state cerebellar activity and cerebellum-based FC were observed in patients with SAD [84,85]. Indeed, the cerebellum has traditionally been considered a region exclusively involved in motor control and coordination [86], but recently, its involvement in nonmotor domains, such as emotion regulation [87], cognitive processing (i.e., visual-spatial, executive, and working memory) [88], and reward-related learning [89], has drawn much attention. Convergent evidence indicates that dysconnectivity between the putamen and cerebellar lobules is implicated in various cognitive functions and is interconnected with the default mode and frontoparietal networks [90,91]. Therefore, we hypothesized that diminished putamen-cerebellar connectivity may lead to dysfunctional cognitive and emotional regulation in patients with SAD.
Cortico-thalamo-cerebellar circuitry in SAD Current research has also pointed to structural/functional deficits in cortico-thalamo-cerebellar circuitry. In agreement with a recent meta-analysis of VBM studies [92,93], we found that patients with SAD had lower volumes in the right thalamus and left ParaHIP, crucial components of limbic structures whose structural and functional abnormalities were found in SAD studies [8,[94][95][96]. It has been documented that the thalamus and ParaHIP are implicated in emotion regulation, emotional salience, and cognitive/executive networks [97]; dysfunction in the thalamus, as a part of the arousal system, may be related to hypersensitivity and hypervigilance to social stimuli and emotional dysregulation [98], while ParaHIP deficits in SAD may reflect disrupted contextual fear conditioning and failure to assign accurate saliency value to stimuli [99]. Hence, it is speculated that emotional and cognitive/ executive dysfunction may be linked to progressive atrophy of the thalamus and ParaHIP in SAD.
From a network perspective, the cerebellum is closely connected with both motor and nonmotor (cognitive and affective) cortical regions via feedback projections of the cerebellum [88], both motor and nonmotor thalamic nuclei receive outputs from the cerebellum [90], and dysfunction in the cortical-thalamic-cerebellar circuit could damage the efficiency of receiving input and producing output [100]. Furtherly, evidence indicated that cortical-(para)limbic imbalance is one of the core pathophysiologies of SAD in which the PFC/ACC fails to adequately mediate the limbic regions, which then demonstrate dysfunctional activity [26,[101][102][103], while current study also found the decreased connectivity between right thalamus and PFC/limbic lobe/ACC. In this sense, this evidence aligns well with current findings of decreased connectivity within the corticalthalamic-cerebellar circuitry. In addition, the SAD patients showed decreased connectivity within the SMA-thalamic-cerebellar circuitry in this study. A study in healthy participants reported that the early binding of gaze, gestures, and emotions is achieved in the motor system (e.g., SMA and premotor cortex), which may prompt the preparation of an adaptive response to another person's immediate intention [104], while gaze avoidance towards emotional stimuli is one of the important characteristics in SAD patients [105].
Taken together, evidence is accumulating for the interconnection of the cortex, basal ganglia, and thalamus in large-scale loops (i.e., cortico-striato-thalamo-cortical circuitry; CSTCort circuitry) and for their involvement in vital cerebral function [106,107], which is considered a prevailing model regarding the neural and pathophysiological underpinnings of obsessive-compulsive disorder (another important anxiety-related psychosis in DSM-IV) [108,109]. The current study not only confirmed the involvement of CSTCort circuitry in SAD but also highlighted the crucial roles of the cerebellum in SAD, thus indicating that aberrant corticostriato-thalamo-cerebellar (CSTCere) circuitry may contribute to the psychopathological and pathophysiological basis of SAD; that is, dysfunctional CSTCere circuitry may contribute to the undue appraisal of external stimuli such as facial emotion and potential social threat, defective cognitive, emotional, and social motor processing, and consequent excessive fear and avoidance of social interactions, which contribute to the occurrence and development of SAD.
Couplings of structural and functional deficits in the prediction of SAD diagnosis Furthermore, the subsequent analysis identified that functional deficits may partially mediate the influence of structural abnormalities on SAD diagnosis. Previous studies have reported widespread structural and functional abnormalities [26,27,31,33], and a considerable number of regions with structural anomalies were compatible with functional deficits, some of which have shown great accuracy for clarifying SAD patients and HCs [31,110], suggesting that structural and functional deficits may be the pathophysiological bases and could serve as potential biomarkers for SAD. Furtherly, as a relatively more stable variable, structural features (e.g., GMV) are generally deemed the basis of functional parameters (e.g., rs-FC), and functional characteristics may be outward manifestations resulting from structural changes [40]. As previous studies demonstrated, between-group differences in functional activation in certain regions showed much overlap with the structural alterations [35], and a wide range of areas could conform to the principle of "the greater the GM concentration, the greater the task-related activation change from baseline" [38]. Besides, spatial statistics, strength, and persistence of rs-FC, are determined by the large-scale cerebral structural backbone, indicating a close interrelation of brain structure and function [40,111]. In agreement with this, our mediation analysis confirmed that dysconnectivity based on the bilateral putamen and right thalamus seeds could partially mediate the relationships between Fig. 3 Mediating role of rs-FC deficits on the effects of GMV abnormalities on SAD diagnosis. Unstandardized regression coefficients are displayed (*P < 0.05, **P < 0.01, ***P < 0.001). Age, sex, total intracranial volume, and mean framewise displacement were controlled for in the model. Abbreviations: CI, confidence interval; GMV, gray matter volume; HCs, healthy controls; L_Cere, left cerebellum; L_Put, left putamen; R_Cere, right cerebellum; R_MTG, right middle temporal gyrus; R_Put, right putamen; rs-FC, resting-state functional connectivity; R-Tha, right thalamus; SAD, social anxiety disorder; SMA, supplementary motor area.
atrophy of subcortical nuclei and SAD diagnosis. In other words, we speculated that atrophy of the bilateral putamen and right thalamus may cause aberrant functional synchronization and then give rise to defective function in some vital areas, eventually leading to the occurrence of SAD.
As a result, current findings may provide novel insights into the pathophysiological and neurobiological substrates underlying SAD; that is, the corresponding functional deficits may be a potential intrinsic mechanism linking GMV alterations to SAD occurrence. In this sense, our results may support the view that those identified structural and functional alterations were more closely related to the category of disorder than the psychopathological dimension [20]. To some degree, this speculation may be further confirmed by our exploratory correlation analysis, as we failed to detect significant correlations between those neuroanatomical differences and symptom severity. Nevertheless, it should be mentioned that some previous studies indeed observed relationships between neuroanatomical alterations and SAD symptom severity [6,9,10,16], indicating that further studies need to be conducted to investigate this exact relationship. Instead, we observed significantly negative correlations between decreased GMV in the bilateral putamen and SAD duration and positive correlations between dysfunctional rs-FC of the right thalamus with limbic lobe/ACC or cerebellum and SAD duration. Our results indicated that as the disease course was extended, SAD patients experienced disrupted bottom-up processes and top-down control in response to external stimuli evoking social anxiety, thus suffering from more severe damage to the subcortical nuclei and cortical-subcortical functional circuitry that appeared as resultant subcortical atrophy and large-scale circuit dysconnectivity.
Comparison between our results and previous findings and potential implications Along with the rapid growth of neuroimaging studies of SAD, increasing attention has been given to the validation and replication of findings that are vital for clinical transformation. Indeed, the neuroimaging results on SAD have been of limited consistency. Previous sMRI and rs-fMRI studies, which have been summarized in [21,33], have reported a widespread but variable pattern of brain regions with GM structural and functional alterations, while the current study detected GMV alterations mainly in subcortical regions and functional deficits in large-scale CSTCere circuitry. The differences from our results may be attributed to several factors. From a methodological perspective, first, we used automated software SPM with optimized and standardized DARTEL procedure, a validated VBM method with good test-retest reliability, for preprocessing and statistical analysis, while manual segmentation or less accurate methods were performed in some previous studies [19,112,113]. Second, we used whole-brain voxel-wise analysis, different from the ROI analysis adopted in some other studies, which may significantly increase the risk of Type II error [25], and could be more sensitive to detect alterations especially for small structures such as the amygdala due to multiple comparison corrections [114]. Third, GMV is a complicated parameter that is different from other surface-based indices, such as cortical thickness, surface area, and cortical folding (gyrification index) [115]. Another reasonable interpretation for the discrepancies is that no changed GMV may be the result of cortical thinning with concurrent surface expansion or vice versa [115], as our recent study demonstrated dissociations in cortical thickness and surface area (i.e., decreased cortical thickness and increased cortical surface area) in SAD patients [22,116]. Fourth, differences in parameter settings on preprocessing and statistical analysis may also contribute to the heterogeneity. For instance, the definition of seeds (e.g., choosing a prior mask of brain areas with structural/functional abnormalities vs. a combination of coordinates and radius) may also cause different results. From the perspective of samples, demographic variations, such as disease severity, illness durations, and comorbidities can confound results [117,118]. Another nonnegligible reason for inconsistent results is the sample size, which is elaborated earlier (Introduction section). In summary, many aspects of the methodological, medical, and sociodemographic domains are associated with (or perhaps cause) neurostructural/ functional alterations in SAD.
Neuroscience plays a crucial role in a translational approach to inform the improvement and development of diagnostic and therapeutic strategies. In the future, along with the identification and confirmation of the close (even causal) relation between the GM structural/functional alterations and occurrence/progression of SAD, those identified regions could serve neural biomarkers for early diagnosis of SAD, as well as reliable and noninvasive tools for disorder prognosis and treatment efficacy assessment [31,116,119]. Furthermore, regions with GM abnormalities are potential therapeutic targets. These findings may not only contribute to the selection, optimal use, refinement, or development of targeted drugs, but also direct modulation of SAD-specific areas via nonpharmacological neurobiological interventions such as deep brain stimulation [120], repetitive transcranial magnetic stimulation [121,122], and real-time functional MRI neurofeedback [123], which may be promising choices in the future. For instance, based on accumulated evidence from neuroimaging studies highlighting the crucial roles of the lateral-medial PFC in SAD, a recent randomized, double-blind, and parallel-group study demonstrated that transcranial direct current stimulation over the dlPFC and mPFC can significantly alleviate SAD core symptoms (i.e., fear and avoidance), reduce attention bias to threat-related stimuli, and improve therapy-related variables (i.e., emotion regulation, depressive state, worries, and quality of life) [124]. In summary, neuroimaging studies could offer further insights into the neurobiological mechanisms of SAD, which is of vital importance for guiding effective diagnosis and therapy to improve the quality of life of SAD patients as much as possible [31]; this is also the aim of psychoradiology [125,126]. Notwithstanding many important efforts, we should also recognize that there is still a long way from bench to bed. It remains to be seen to what extent those neuroimaging results based on the symptombased diagnostic categories for psychiatric disorders could reflect the specific pathophysiological mechanisms and their relations to clinical symptoms. Furthermore, future research can benefit much from investigating social anxiety dimension-based measures in combination with longitudinal studies of interventional effects.
Limitations
When interpreting the current findings, some limitations deserve mentioning. First, it needs to be clarified whether our results could have been influenced by examination-related anxiety during the MRI scans. To do this, we would need to assess the psychological reactions and psychophysiological responses of participants before/ during/after the MRI examination [127]. Second, as a cross-sectional design cannot explicitly elucidate the causal relationships between structural/functional abnormalities and disease state, future longitudinal and developmental studies involving follow-up evaluations and studying people with high innate vulnerability to developing SAD (e.g., based on the genotypes and endophenotypes [128]) will be much more beneficial for providing further insights into the neurobiological and psychopathological underpinning and progression course of SAD. Third, it would have been desirable to measure and, if feasible, to match the patients with HCs on intelligence quotient measures, which showed a positive correlation with brain volume [129]. Nevertheless, as the TIV was included as a covariate in the analysis and there is no definite proof supporting the notion that SAD patients are subject to intellectual impairment, the effects of intelligence on the current findings are quite limited. Fourth, given that SAD typically evolves during late childhood and early adolescence, it remains to be seen whether the current findings based on adult SAD patients could be generalized to adolescent populations. To do this, future researches need to investigate the neuroanatomical alterations in a cohort of child and adolescent patients suffering from SAD. Fifth, although power analysis was adopted to guarantee medium-sized effects (to the best of our knowledge, the current study is thus far the relatively large singlecenter study investigating whole-brain structural and functional deficits in non-comorbid SAD patients), our sample size is not very large compared to recent studies exploring other psychiatric disorders. This is partly due to strict inclusion criteria of adult SAD patients without any comorbid disorders, with the hope of investigating the pure and specific neurobiology of SAD, which may also to some degree limit the generalizability of our findings. In the future studies, researchers could benefit from exploring the effects of confounded factors on structural/functional deficits, and our results need further replication via a larger sample. Sixth, the cross-sectional design makes it difficult to identify causal associations regarding brain structural/functional alterations and SAD diagnosis. In particular, the exact correlations between cerebral structure and function remain to be investigated. Consequently, we also conducted another mediation design (i.e., X = FC, M = GMV, Y = SAD diagnosis; age, sex, TIV, and mean FD were regarded as covariates), and we observed similar results to the main analysis, i.e., a significant mediating effect of decreased GMV in the bilateral putamen on the association between decreased rs-FC of bilateral putamen with cerebellum and SAD diagnosis respectively .923], P < 0.05]) and SAD diagnosis. Therefore, other potential mediating associations may exist among GMV, FC, and SAD diagnosis, although our main analysis attempted to investigate brain structure-function-behavior couplings to disclose the potential neurobiological mechanisms underlying SAD. In summary, considering that current data from the cross-sectional design were not temporally discernable, the mediating analyses conducted in this study were of a statistical or theoretical nature, and the main results may be just a possible mechanism linking GMV, FC, and SAD. To identify the causal relationships of these variables, longitudinal and interventional (e.g., therapeutic trial) designs are needed in future works. Finally, as the results of FC analyses based on the seeds with structural deficits may be partly biased by the seeds selection, future works need to establish structural and functional networks in whole-brain regions to investigate the effects of structural-functional couplings on SAD.
Conclusions
The current study complemented and extended prior SAD-related neuroimaging studies by identifying the involvement of the CSTCere circuitry in non-comorbid adult patients with SAD and revealing the potential coupling mechanisms of structural and functional deficits in the prediction of SAD diagnosis, which together indicate that the aberrant CSTCere circuitry may contribute to the neurobiological basis of SAD. The current findings might provide insights into understanding the neurobiological substrates of SAD and initial evidence for further identification of candidate neuroanatomical biomarkers, which may advance the early diagnosis, targeted treatment, and therapeutic evaluation for SAD.
DATA AVAILABILITY
The data and code that support the findings of present study are available from the corresponding author through reasonable request. The data and code sharing adopted by the authors comply with the requirements of the funding institute and with institutional ethics approval.
FUNDING AND DISCLOSURE
This study was supported by the National Natural Science Foundation of China (Grant Nos. 81621003, 81761128023, 81820108018, 82027808, and 31700964) and NIH/NIMH R01MH112189-01. The funding sources had no involvement in the study design, data collection and analysis, results interpretation, or writing of the paper.
AUTHOR CONTRIBUTIONS
QYG and SW designed the study and supervised the conduct of the study. XZ, XY, WHK, SW, HL, NFP, and MH contributed to the data collection. HL, NFP, MH, and QYL provided methodological advice. XZ and XLS performed the data analysis and results interpretation. XZ, XLS, and SW drafted the manuscript, which all authors reviewed and approved for publication. | 2022-01-21T14:46:32.883Z | 2022-01-21T00:00:00.000 | {
"year": 2022,
"sha1": "6b40256c573f57c0638986954ab2f10fb4a48a34",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41398-022-01791-7.pdf",
"oa_status": "GOLD",
"pdf_src": "Springer",
"pdf_hash": "6b40256c573f57c0638986954ab2f10fb4a48a34",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
11773057 | pes2o/s2orc | v3-fos-license | Statistical Basis for Predicting Technological Progress
Forecasting technological progress is of great interest to engineers, policy makers, and private investors. Several models have been proposed for predicting technological improvement, but how well do these models perform? An early hypothesis made by Theodore Wright in 1936 is that cost decreases as a power law of cumulative production. An alternative hypothesis is Moore's law, which can be generalized to say that technologies improve exponentially with time. Other alternatives were proposed by Goddard, Sinclair et al., and Nordhaus. These hypotheses have not previously been rigorously tested. Using a new database on the cost and production of 62 different technologies, which is the most expansive of its kind, we test the ability of six different postulated laws to predict future costs. Our approach involves hindcasting and developing a statistical model to rank the performance of the postulated laws. Wright's law produces the best forecasts, but Moore's law is not far behind. We discover a previously unobserved regularity that production tends to increase exponentially. A combination of an exponential decrease in cost and an exponential increase in production would make Moore's law and Wright's law indistinguishable, as originally pointed out by Sahal. We show for the first time that these regularities are observed in data to such a degree that the performance of these two laws is nearly the same. Our results show that technological progress is forecastable, with the square root of the logarithmic error growing linearly with the forecasting horizon at a typical rate of 2.5% per year. These results have implications for theories of technological change, and assessments of candidate technologies and policies for climate change mitigation.
Introduction
Innovation is by definition new and unexpected, and might therefore seem inherently unpredictable. But if there is a degree of predictability in technological innovation, understanding it could have profound implications. Such knowledge could result in better theories of economic growth, and enable more effective strategies for engineering design, public policy design, and private investment. In the area of climate change mitigation, the estimated cost of achieving a given greenhouse gas concentration stabilization target is highly sensitive to assumptions about future technological progress [1].
There are many hypotheses about technological progress, but are they any good? Which, if any, hypothesis provides good forecasts? In this paper, we present the first statistically rigorous comparison of competing proposals.
When we think about progress in technologies, the first product that comes to mind for many is a computer, or more generally, an information technology. The following quote by Bill Gates captures a commonly held view: ''Exponential improvementthat is rare -we've all been spoiled and deeply confused by the IT model'' [2]. But as we demonstrate here, information technologies are not special in terms of the functional form that describes their improvement over time. Information technologies show rapid rates of improvement, but many technologies show exponential improvement. In fact, all the technologies we study here behave roughly similarly: Information technologies closely follow patterns of improvement originally postulated by Wright for airplanes [3][4][5][6][7][8], and technologies such as beer production or offshore gas pipelines follow Moore's law [9,10], but with a slower rate of improvement [8,[11][12][13][14][15].
It is not possible to quantify the performance of a technology with a single number [16]. A computer, for example, is characterized by speed, storage capacity, size and cost, as well as other intangible characteristics such as aesthetics. One automobile may be faster, while another is less expensive. For this study, we focus on one common measure of performance: the inflation-adjusted cost of one ''unit''. This metric is suitable in that it can be used to describe many different technologies. However, the nature of a unit may change over time. For example, a transistor in a modern integrated circuit today may have quite different performance characteristics than its discrete counterpart in the past. Furthermore, the degree to which cost is emphasized over other performance measures may change with time [17]. We nonetheless use the changes in the unit cost as our measure of progress, in order to compare competing models using a sizable dataset. The crudeness of this approach only increases the difficulty of forecasting and makes it particularly surprising that we nonetheless observe common trends.
Analysis
We test six different hypotheses that have appeared in the literature [3,9,[18][19][20], corresponding to the following six functional forms: Moore log y t~a tzbzn(t) Goddard log y t~a log q t zbzn(t) SKC log y t~a log q t zc log (x t {q t )zbzn(t) Nordhaus log y t~a tzc log x t zbzn(t): The dependent variable y t is the unit cost of the technology measured in inflation-adjusted dollars. The independent variables are the time t (measured in years), the annual production q t , and the cumulative production x t~P t i~1 q i . The noise term n(t), the constants a, b and c, and the predictor variables differ for each hypothesis.
Moore's law here refers to the generalized statement that the cost y of a given technology decreases exponentially with time: where mw0 and Bw0 are constants [9,12]. (We assume throughout that tw0, and we have renamed a~{m and b~log B in Eq. (1)). Moore's law postulates that technological progress is inexorable, i.e. it depends on time rather than controllable factors such as research and development.
Wright's law, in contrast, postulates that cost decreases at a rate that depends on cumulative production: where ww0 and Bw0 are constants, and we have renamed a~{w and b~log B in Eq. (1). Wright's law is often interpreted to imply ''learning by doing'' [5,21]. The basic idea is that cumulative production is a proxy for the level of effort invested, so that the more we make the more we learn, and knowledge accumulates without loss. Another hypothesis is due to Goddard [18], who argues that progress is driven purely by economies of scale, and postulates that: where sw0 and Bw0 are constants, and we have renamed a~{s and b~log B in Eq. (1). We also consider the three multi-variable hypotheses in Eq. (1): Nordhaus [20] combines Wright's law and Moore's law, and Sinclair, Klepper, and Cohen (SKC) [19] combine Wright's law and Goddard's law. For completeness, we also test Wright's law lagged by one year. Note that these methods forecast different things: Moore's law forecasts the cost at a given time, Wright's law at a given cumulative production, and Goddard's law at a given annual production.
We test these hypotheses on historical data consisting of 62 different technologies that can be broadly grouped into four categories: Chemical, Hardware, Energy, and Other. All data can be found in the online Performance Curve Database at pcdb.santafe.edu. The data are sampled at annual intervals with timespans ranging from 10 to 39 years. The choice of these particular technologies was driven by availability -we included all available data, with minimal constraints applied, to assemble the largest database of its kind.
The data was collected from research articles, government reports, market research publications, and other published sources. Data on technological improvement was used in the analysis if it satisfied the following constraints: it retained a functional unit over the time period sampled, and it included both performance metric (price or cost per unit of production) and production data for a period of at least 10 years, with no missing years in between. This inclusive approach to data gathering was required to construct a large dataset, which was necessary to obtain statistically significant results. The resulting 62 datasets are described in detail in File S1.
These datasets almost certainly contain significant measurement and estimation errors, which cannot be directly quantified and are likely to increase the error in forecasts. Including many independent data sets helps to ensure that any biases in the database as a whole are random rather than systematic, minimizing their effects on the results of our analysis of the pooled data.
To compare the performance of each hypothesis we use hindcasting, which is a form of cross-validation. We pretend to be at time i and make a forecastŷ y (f ,d,i) j for time j using hypothesis (functional form) f and data set d, where jwi. The parameters for each functional form are fitted using ordinary least squares based on all data prior to time i, and forecasts are made based on the resulting regression. We score the quality of forecasts based on the logarithmic forecasting error: The quality of forecasts is examined for all datasets and all hypotheses (and visualized as a three-dimensional error mountain, as shown in File S1). For Wright's law, an illustration of the growth of forecasting errors as a function of the forecasting horizon is given in Fig. 1.
An alternative to our approach is to adjust the intercepts to match the last point. For example, for Moore's law this corresponds to using a log random walk of the form log y tz1~l og y t {mzn(t), where n(t) is an IID noise term (see File S1). We have not done this here to be consistent with the way these hypotheses have been presented historically. The method we have used also results in more stable errors. Developing a statistical model to compare the competing hypotheses is complicated by the fact that errors observed at longer horizons tend to be larger than those at shorter horizons, and errors are correlated across time and across functional forms. After comparing many different possibilities (as discussed in detail in File S1), we settled on the following approach. Based on a search of the family of power transformations, which is known for its ability to accommodate a range of variance structures, we take as a response the square root transformation of the logarithmic error. This response was chosen to maximize likelihood when modeled as a linear function of the hindcasting horizon~target { origin~j{i, using a linear mixed effects model.
Specifically, we use the following functional form to model the response: where r fdij is the expected root error. The parameters a f and b f depend on the functional form and are called fixed effects because they are the same for all datasets. a f is the intercept and b f is the slope parameter. The parameters a d and b d depend on the dataset, and are called random effects because they are not fitted independently but are instead treated as dataset-specific random fluctuations from the pooled data. The quantities a d and b d are additive adjustments to the average intercept and slope parameters a f and b f , respectively, to take into account the peculiarities of each dataset d.
In Finally, we add an E fdij random field term to take into account the deviations from the trend. This is assumed to be a Gaussian Var We also define an exponential correlation structure within each error mountain (corresponding to each combination of dataset and hypothesis, see File S1), as a function of the differences of the two time coordinates with a positive range parameter r and another small positive nugget parameter g quantifying the extent of these correlations: where the two Kronecker d functions ensure that each error mountain is treated as a separate entity. Equations (7) and (8) were chosen to deal with the observed heteroscedasticity (increasing variance with increasing logarithmic forecasting error) and the serial correlations along the time coordinates i (hindcasting origin) and j (hindcasting target). Based on the likelihood, an exponential correlation function provided the best fit. Note that instead of a Euclidean distance (root sum of the squares of differences), the Manhattan measure was used (the sum of the absolute differences), because it provided a better fit in terms of the likelihood. Using this statistical model, we compared five different hypotheses. (We removed the Nordhaus model from the sample because of poor forecasting performance [20]. This model gave good in-sample fits but generated large and inconsistent errors when predicting out-of-sample, a signature of over-fitting. This points to the difficulty in separating learning from exogenous sources of change [20].) Rather than the 62|5|2~620 parameters needed to fit each of the 62 datasets separately for each of the five functional forms, there are only 16 free parameters: 5|2 = 10 parameters a f and b f , three parameters for the covariance matrix of the bivariate random vector (a d ,b d ), and three parameters for the variance and autocorrelation of the residuals E fdij .
Results and Discussion
We fit the error model to the 37,745 different r fdij data points using the method of maximum likelihood. In Fig. 2 we plot the expected root error r fij~af zb f (j{i) for the five hypotheses as a function of the hindcasting horizon. While there are differences in the performance of these five hypotheses, they are not dramatic. The intercept is tightly clustered in a range 0:16va f v0:19 and the slope 0:024vb f v0:028. Thus all the hypotheses show a large initial error, followed by a growth in the root error of roughly 2:5% per year. This is a central tendency for the pooled data.
The error model allows us to compare each hypothesis pairwise to determine whether it is possible to reject one in favor of another at statistically significant levels. The comparisons are based on the intercept and slope of the error model of Eq. (6). The parameter estimates are listed in Tables S1 and S3 in File S1 and the corresponding p-values in Tables S2 and S4 in File S1. For example, at the 5% level, the intercept of Goddard is significantly higher than any of the others and the slope of SKC is significantly greater than that of Wright, lagged Wright and Goddard. With respect to slope, Moore is at the boundary of being rejected in favor of Wright. Fig. 2 makes the basic pattern clear: Goddard does a poorer job of forecasting at short times, whereas SKC, and to a lesser extent Moore, do a poorer job at long times.
We thus have the surprising result that most of the methods are quite similar in their performance. Although the difference is not large, the fact that we can eliminate Goddard for short term forecasts indicates that there is information in the cumulative production not contained in the annual production, and suggests that there is a learning effect in addition to economies of scale. But the fact that Goddard is not that much worse indicates that much of the predictability comes from annual production, suggesting that economies of scale are important. (In our database, technologies rarely decrease significantly in annual production; examples of this would provide a better test of Goddard's theory. ) We believe the SKC model performs worse at long times because it has an extra parameter, making it prone to overfitting. Although Moore performs slightly worse than Wright, given the clear difference in their economic interpretation, it is surprising that their performance is so similar. A simple explanation for Wright's law in terms of Moore's law was originally put forward by Sahal [22]. He noted that if cumulative production grows exponentially: then eliminating t between Eqs. (2) and (9) results in Wright's law, Eq. (3), with w~m=g. Indeed, when we look at production vs. time we find that in almost every case the cumulative production increases roughly exponentially with time. (Note that if production grows exponentially, cumulative production also grows exponentially with the same exponent.) This is illustrated in Fig. 3, where we show three representative examples for production and cost plotted as a function of time. Fig. 3 also shows histograms of R 2 values for fitting g and m for the 62 datasets. The agreement with exponential behavior ranges from very good to rather poor, but of course these are short time series and some of them are very noisy. We test this in Fig. 4 by plotting the measured value of w d against the derived valueŵ w d~m =g for each data set d. The values cluster tightly along the identity line, indicating that Sahal's conjecture is correct.
The differences in the data sets can be visualized by plotting a d and b d as shown in Fig. 5. All but one data set is inside the 95% confidence ellipsoid, indicating that the estimated distribution of (a d ,b d ) is consistent with the bivariate normal assumption. The intercepts vary in a range roughly {0:10va d v0:17 and the slopes {0:018vb d v0:015. Thus the variation in the corresponding logarithmic forecasting error for the different datasets is comparable to the average error for all datasets (Fig. 5) and about an order of magnitude larger than the difference between the hypothesized laws (Fig. 2).
To illustrate the practical usefulness of our approach we make a forecast of the cost of electricity for residential scale photovoltaic solar systems (PV). Fig. 6 shows the best forecast (solid line) as well as the expected error (dashed lines). These are not confidence limits, but rather projected absolute log deviations from the best forecast, calculated from Eq. (6) using a Moore , b Moore , a Photovoltaics2 , and b Photovoltaics2 . The sharp drop in the one year forecast relative to the last observed data point comes from the fact that forecasts are based on the average trend line, and because this data series is particularly long. PV costs rose in recent years due to increased material costs and other effects, but industry experts expect this to be a short-lived aberration from the long-term cost trend.
The expected PV cost in 2020, shown in Fig. 6, is 6 cents/kWh with a range (3,12). In 2030 the cost is 2 cents/kWh, with a range (0.4, 11). This does not include the additional cost of energy storage technologies. The current cost of the cheapest alternative, coal-fired electricity, is roughly 5 cents/kWh. This is the wholesale cost at the plant (busbar), which may be most directly comparable to industrial scale PV (rather than the residential scale shown in Fig. 6). Industrial scale PV is typically about two-thirds the cost of electricity from the residential scale systems. In contrast to PV, coal-fired electricity is not expected to decrease in cost, and will likely increase if there are future penalties for CO 2 emissions [23].
The costs of other technologies can be forecasted in a similar way, using historical data on the cost evolution to project future performance. The expected error in this forecast is calculated using our error model (Eq. (6)). The error is determined for each future year j from the present year i based on parameters specific to the technology of interest, as well as insight gained from examining data on many technologies. This approach allows us to forecast both the expected error and the expected cost. The method outlined is suited to Moore's functional form. Forecasting future performance based on production levels requires an additional step of forecasting future production over time.
Our primary goal in this paper is to compare the performance of proposed models in the literature for describing the cost evolution of technologies. Our objective is not to construct the best possible forecasting model. Nonetheless we outline above the steps one would take in making a forecast in order to demonstrate the utility of the general approach we develop, which centers on analyzing a large, pooled database, and estimating the expected, time horizon-dependent error associated with a given forecasting model. This approach can be applied to other forecasting models in the future.
The key postulate that we have made in this paper is that the processes generating the costs of technologies through time are generic except for technology-specific differences in parameters. This hypothesis is powerful in allowing us to view any given technology as being drawn from an ensemble. This means that we can pool data from different technologies to make better forecasts, and most importantly, make error estimates. This is particularly useful for studying technology trends, where available data is limited. Of course we must add the usual caveats about making forecasts -as Niels Bohr reputedly said, prediction is very difficult, especially of the future. Our analysis reveals that decreasing costs and increasing production are closely related, and that the hypotheses of Wright and Moore are more similar than they might appear. We should stress, though, that they are not the same. For example, consider a scenario in which the exponential rate of growth of PV production suddenly increased, which would decrease the current production doubling time of roughly 3 years. In this case, Wright predicts that the rate at which costs fall would increase, whereas Moore predicts that it would be unaffected. Distinguishing between the two hypotheses requires a sufficient number of examples where production does not increase exponentially, which our current database does not contain. The historical data shows a strong tendency, across different types of technologies, toward constant exponential growth rates. Recent work, however, has demonstrated super-exponential improvement for information technologies over long time spans [24], suggesting that Moore's law is a reasonable approximation only over short time spans. This evidence from information technologies [24], and the results presented here, suggest that Moore may perform significantly worse than Wright over longer time horizons.
Supporting Information
File S1 Supporting Information (PDF) | 2012-07-05T14:12:19.000Z | 2012-07-01T00:00:00.000 | {
"year": 2013,
"sha1": "986f2b05e2c4c3f98256100d90bebf401a5b0c1d",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0052669&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c33efc61bf435bb4fc3856d7661df36b0d775aa4",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Economics",
"Physics",
"Mathematics",
"Medicine"
]
} |
118216538 | pes2o/s2orc | v3-fos-license | Simulations of strongly phase-separated liquid-gas systems
Lattice Boltzmann simulations of liquid-gas systems are believed to be restricted to modest density ratios of less than 10. In this article we show that reducing the speed of sound and, just as importantly, the interfacial contributions to the pressure allows lattice Boltzmann simulations to achieve high density ratios of 1000 or more. We also present explicit expressions for the limits of the parameter region in which the method gives accurate results. There are two separate limiting phenomena. The first is the stability of the bulk liquid phase. This consideration is specific to lattice Boltzmann methods. The second is a general argument for the interface discretization that applies to any diffuse interface method.
Simulations of liquid gas systems with lattice Boltzmann have been restricted to small density ratios in the past. Those restrictions have lead to the development of hybrid lattice Boltzmann schemes to be able to simulate systems with high density ratios of about 100-1000 by Inamuro et al. [1]. In this article we explain how large density ratios can also be achieved with standard lattice Boltzmann methods. Furthermore we derive the conditions which limit the ability of the method to obtain stable, accurate, and unique results for the phase diagram. We present a new general argument for the minimum interface width required to accurately simulate a system at a given reduced temperature. This important argument is not restricted to lattice Boltzmann methods but only relies on the relation of the discretized interface to the expression for the pressure. It therefore applies to all diffuse interface methods.
The lattice Boltzmann method can be viewed as a discretization of the Boltzmann equation. The hydrodynamic limit of the Boltzmann equation gives the continuity and Navier-Stokes equations and the discretization of the lattice version is chosen such that it preserves this limit. The basic variables of the lattice Boltzmann equation are a set of densities f i (x, t) associated with a velocity set v i . The evolution equation for the f i is then given by [2] (1) The f 0 i are the equilibrium distribution corresponding to the ideal gas. Non-ideal contributions are included through the bulk forcing term F i or pressure term A i , following [2]. The fluid density is defined as ρ = i f i , and the momentum is ρu = i f i v i (although the total momentum contains additional contributions from the force). The moments of the equilibrium distribution are Here Q is a correction term that should be zero. Most velocity sets for lattice Boltzmann are limited to v ix ∈ {−1, 0, 1} so that v 3 ix = v ix . This restricts the third moment in (2) to θ = 1/3 and Q αβγ = −ρu α u β u γ .
The non-ideal contributions from the A i need to conserve mass and momentum and the moments are given by The forcing term F i has the moments A standard expansion of (1) gives the continuity equation ∂ t ρ + ∇(ρû) = 0 (5) and the Navier Stokes equation . Here the Newtonian stress tensor is given by and unphysical terms have been collected in the remainder tensor The kinematic viscosity is given by ν = (τ − 1 2 )θ. Note that most of the unphysical terms in (8) violate Galilean invariance [3,4].
We see from (6) that for A = 0 and F = 0 the lattice Boltzmann method enforces an ideal gas equation of state with p(ρ, θ) = ρθ = ρ/3. To simulate a fluid with a nonideal equation of state P (ρ, θ) = ρθ + P nid (ρ, θ) we can now choose which we will refer to as the forcing method [2]. The careful reader may have recognized that the Ψ term does not appear in the Navier-Stokes equation (6), but a higher order analysis shows that these terms are necessary to recover the correct equilibrium behavior [2]. An alternative choice for the moments is which we will call the pressure method [2,3]. For either approach we recover the Navier Stokes equation for a non-ideal gas In equilibrium both approaches lead to a constant pressure P and therefore to the same density profiles [2]. Most previous lattice Boltzmann simulations approached the simulation of non-ideal systems by using the ideal gas equation of state p = ρθ = ρ/3, as a starting point. Interactions are then included to allow the simulation of non-ideal systems. The speed of sound c s = ∂ ρ p will then recover the ideal gas value of 1/3 in the dilute limit. For a van der Waals gas with a critical density of 1 and a temperature of θ = 1/3 and an interfacial free energy of κ/2 (∇ρ) 2 the pressure tensor is given by Previous approaches matched the ideal gas equation of state in the dilute limit, leading to p 0 = 1. For the van der Waals gas the speed of sound increases rapidly for high densities. A problem arises when the speed of sound becomes larger then the lattice velocity |v i |, because information can not be passed on at speeds larger than the lattice velocity. When the speed of sound is increased above 1 the simulation becomes unstable. This clearly limits the range of critical temperatures for which we can obtain stable solutions in lattice Boltzmann, as shown in Figure 1. The stability analysis is slightly complicated by the fact that we have additional gradient terms in the pressure tensor. These terms further decrease the stability, as shown in a previous analysis of the pressure method by C. Pooley for one, two, and three dimensional lattice Boltzmann methods [5]. In the notation of this letter the linear stability condition is for a homogeneous system with density ρ. This suggests that, at least as far as the stability of the bulk phase is concerned, the most stable solutions should be found for κ = 0.
To lower the speed of sound in the liquid phase we now reduce the value of p 0 in (12). This decreases the speed of sound in the liquid by a factor p 0 . This also increases the range of stability for κ in (13). We now expect that lowering the speed of sound by a sufficient factor will reduce the speed of sound sufficiently to simulate systems with arbitrarily low temperature ratios θ/θ c .
To test this idea we performed simulations with near equilibrium profiles using a one dimensional three veloc- where ρ l and ρ g are the equilibrium gas and liquid densities and N x is the number of lattice sites. The interface width is given by This profile is not the exact analytical solution to the differential equation ∇P = 0, but it is very close to it. By initializing the simulation with this profile we can test the linear stability of the method around an equilibrium profile to good accuracy. The shape of a stable interfacial profile is independent of p 0 for both the pressure and the forcing method.
In Figure 2 we see that by lowering p 0 the method is now able so simulate very small values of the reduced temperature θ/θ c for interface width w > 1, but that significantly larger width are required to recover an accurate phase diagram for deep quenches. For values of θ/θ c between 0.9 and 1 we also find non unique solutions for small values of κ which is discussed in more detail in a previous paper [2].
To understand when the method fails to obtain accurate results note that, in equilibrium, equation (12) requires that where p b is the bulk pressure corresponding to the density ρ. For small values of κ the interface becomes sharp in the continuous limit so that the derivatives become arbitrarily large. But in the numerical implementation the derivatives are discrete derivatives. The discrete values are limited by the lattice spacing. In the one dimensional case we choose ∇ρ(x) = 0.5(ρ(x + 1) − ρ(x − 1)) and ∇ 2 ρ(x) = ρ(x + 1) − 2ρ(x) + ρ(x − 1). For higher dimensional stencils with the same stability limits for the bulk phase see C. Pooley's thesis [5]. The methods always lead to a constant pressure, even across an interface [2]. We can now perform a simple estimate of the minimum value κ m that allows this pressure to be the equilibrium pressure. For any point with density ρ s we can consider two neighboring points, one with a smaller density ρ − and one with a larger density ρ + . We can now find a lower limit for the smallest value κ m by varying the values of ρ + and ρ − κ m = max ρg <ρs<ρ l min ρ l <ρ−<ρs where ρ l is the liquid density and ρ g is the gas density. We performed a scan of the parameter space w and θ/θ c initializing the simulation with a near equilibrium profile for different values of p 0 . We will accept simulations that are stable, accurate and unique. We choose as the criterion of accuracy that log 10 (ρ min ) − log 10 (ρ g ) < 0.1. As can be seen in Figure 2, the results are not very sensitive to the exact value of the cutoff. For values of the interface width w < 1.5 we also test the uniqueness of the simulation by using initial profiles with bulk densities corresponding to the pressure at the spinodal points [2]. Our criterion for uniqueness is then that all simulations lead to the same minimum density to within ∆ρ < 0.01. The comparing (17), shown as a dashed line in Figure 3, and the numerical results for stable, accurate and unique solutions shows excellent agreement. The bulk stability of eqn. (13) gives the second limit for the acceptable parameter range for the pressure method. The forcing method leads to a slightly larger range of bulk stability. The underlying reason is that calculating derivatives in (9) leads to an additional information exchange allowing for speeds of sound slightly larger than 1. But the dependence of the stability on p 0 and w is very similar to the one for the pressure method. Note that previous lattice Boltzmann simulations correspond to p 0 = 0 which corresponds to a small acceptable parameter range.
The interface constraint (17) is remarkably successful at predicting the acceptable simulation parameters. It predicts how thin is too thin for an interface. It thereby detects when non-unique solutions occur and when solutions for deep quenches fail to deliver accurate results. The criterion presented so far is entirely numerical but because of its importance we want to examine two limiting cases for shallow and deep quenches for which we can obtain analytical results.
We first examine for which values of ρ s the minimum value of κ is reached in eqn. (17). The dashed line in the inset of Figure 4 shows how this density ρ crit varies as a function of temperature.
Near to the critical temperature, the orange line in the inset in Figure 4 lies close to the high density spinodal curve. This is because P − p b has its highest magnitude here, therefore helping to maximize κ m within this region. An analytical estimate for κ m can be obtained by expanding the pressure around the critical density, giving Within this regime, P −p b is large and negative, therefore ρ − and ρ + must be chosen to make the denominator in (17) as negative as possible. A suitable choice is ρ − = ρ g and ρ + = ρ s . We assume that the critical value of ρ s lies on the spinodal curve ρ spin = 1 + 2 √ θ c − θ. This allows us to obtain κ m (ρ crit ), and substituting this expression into (15) gives a minimum interface width of As the temperature is decreased in the inset of Figure 4 the critical density ρ crit makes a discontinuous jump to a regime in which it lies close to the gas density, ρ g . The minimum interface width can, in this case, be analytically obtained by expanding densities around ρ g . We define ρ s = ρ g + δρ and ρ + = ρ − + ∆ρ. Since P − p b is a positive quantity, a suitable choice for ρ − is ρ − = ρ s . Substituting these expressions into equation (17) gives Minimizing this with respect to ∆ρ leads to ρ s = ∆ρ/4. Re-substituting this result back into equation (20), and maximizing with respect to δ, we finally obtain ρ crit = 2ρ g . Using this we can calculate the minimum interface width, as shown by the triangles in Figure 4. This closely follows the numerical result at low temperatures. We have therefore shown how to simulate deep quenches with lattice Boltzmann and we were able to predict which simulation parameters will lead to accurate results. | 2019-04-14T02:12:15.692Z | 2006-08-22T00:00:00.000 | {
"year": 2006,
"sha1": "468ad15716b5c83e7c84a383975fabde190b564b",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "af53b9909228105d2e9ecdc5c2cb826b726a85fa",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
252217370 | pes2o/s2orc | v3-fos-license | Long-term postural control in elite athletes following mild traumatic brain injury
Background Traumas to the head and neck are common in sports and often affects otherwise healthy young individuals. Sports-related concussions (SRC), defined as a mild traumatic brain injury (mTBI), may inflict persistent neck and shoulder pain, and headache, but also more complex symptoms, such as imbalance, dizziness, and visual disturbances. These more complex symptoms are difficult to identify with standard health care diagnostic procedures. Objective To investigate postural control in a group of former elite athletes with persistent post-concussive symptoms (PPCS) at least 6 months after the incident. Method Postural control was examined using posturography during quiet stance and randomized balance perturbations with eyes open and eyes closed. Randomized balance perturbations were used to examine motor learning through sensorimotor adaptation. Force platform recordings were converted to reflect the energy used to maintain balance and spectrally categorized into total energy used, energy used for smooth corrective changes of posture (i.e., <0.1 Hz), and energy used for fast corrective movements to maintain balance (i.e., >0.1 Hz). Results The mTBI group included 20 (13 males, mean age 26.6 years) elite athletes with PPCS and the control group included 12 athletes (9 males, mean age 26.4 years) with no history of SRC. The mTBI group used significantly more energy during balance perturbations than controls: +143% total energy, p = 0.004; +122% low frequency energy, p = 0.007; and +162% high frequency energy, p = 0.004. The mTBI subjects also adapted less to the balance perturbations than controls in total (18% mTBI vs. 37% controls, p = 0.042), low frequency (24% mTBI vs. 42% controls, p = 0.046), and high frequency (6% mTBI vs. 28% controls, p = 0.040). The mTBI subjects used significantly more energy during quiet stance than controls: +128% total energy, p = 0.034; +136% low-frequency energy, p = 0.048; and +109% high-frequency energy, p = 0.015. Conclusion Athletes with previous mTBI and PPCS used more energy to stand compared to controls during balance perturbations and quiet stance and had diminished sensorimotor adaptation. Sports-related concussions are able to affect postural control and motor learning.
Introduction
Participation in high-velocity sports, such as football, ice hockey, and rugby, may lead to a risk of sustaining neck and head trauma from a collision with another player, the playing surface, or an advertising board (1). At impact, the athlete may sustain a concussion [termed a sports-related concussion (SRC)], that per definition is a mild traumatic brain injury (mTBI)(2) (2). High-velocity sports carry an increased risk of mTBI (3). The SRC rates are highest in men, but in gendercomparable sports, women have higher concussion rates (3). The reported number of SRCs has increased in the last 20 years, possibly due to increased monitoring and recognition. However, the true incidence is still likely underestimated (4)(5)(6). mTBI may cause persistent neck and shoulder pain, and headache. Typically, these symptoms subside within 14 days of the incident (2,(7)(8)(9). However, if the incident is severe, the individual can also experience imbalance, dizziness, and visual disturbances, and symptoms may take months or years to subside (10)(11)(12) or remain throughout life (8,13), termed persistent post-concussive symptoms (PPCS) (14). Recently, we have shown that PPCS athletes suffering from vestibular symptoms commonly have dysfunction of the inferior vestibular nerve (15). As the inferior vestibular nerve conveys information from the posterior semicircular canal and sacculus to the vestibular nucleus, PPCS is likely to diminish the perception of left-right tilt and angular head movements in an anteriorposterior direction. Given the contribution of head position to upright standing, it seems reasonable to assume that PPCS can affect postural control, and particularly when the balance is perturbed.
Imbalance following mTBI is not routinely assessed and individuals can overlook this symptom in a background of pain and headache. When postural control is assessed, it is typically measured from a quiet stance. However, a quiet stance is less sensitive to balance defects than a postural control submitted to balance perturbations. When patients are subjected to increased postural challenges, the likelihood to detect defects is increased and the level of reliance on each sensory system can be assessed. For example, during balance perturbations, the level of reliance on the visual system can be assessed by comparing postural responses with eyes open to eyes closed (16)(17)(18). Following an SRC, the results of postural control tests serve a role as supplementary to neurophysiological functioning tests (9,19,20), particularly when balance perturbations are used. When balance is perturbed repeatedly, the central nervous system acts to reduce the imbalance. This is a motor learning process called postural adaptation that depends on the integrity of the sensory and motor systems (17,18). The postural challenge from balance perturbations also increases stress, which may exacerbate balance defects (9,21). One method of perturbing balance is to apply bilateral mechanical vibration over the gastrocnemii muscles. The vibration produces an increased activation of muscle spindles, and subsequently, an illusion of muscle lengthening, which induces a stretch reflex that creates a posterior displacement. When the vibration ends, corrective mechanisms produce an anterior displacement. When balance perturbations are applied repeatedly, an adaptation in postural control normally occurs, which over time diminishes the amplitude of anterior and posterior displacement (22)(23)(24).
The study aimed to determine whether mTBI affects postural stability with eyes open and eyes closed, and the degree of motor learning through sensorimotor adaptation. The reliance on the visual system for postural control was also determined by comparing postural responses with eyes closed to eyes open.
Ethical approval
The investigations were performed in accordance with the latest version of the Helsinki declaration and all subjects gave written informed consent before any assessments. The study was approved by the Ethics Review Board (Dnr 2017/1049), Lund, Sweden.
Participants
The eligible mTBI group consisted of elite athletes, before their injury active in ice hockey (6 participants), soccer (4), karate (4), handball (2), floorball (2), wrestling (1), and riding (1). They all had to terminate their sports career due to neck and head injuries (SRCs/mTBIs) obtained while executing their sports, which were still causing them various degrees of postconcussive symptoms. They suffered their latest injury more than 6 months before the investigations and reported persistent symptoms for more than 6 months. PPCS is defined as persistent post-concussive symptoms beyond normal clinical recovery, typically more than 3-6 weeks post-injury. All participants in our cohort had symptoms persisting for more than 6 months. The most common complaints included fatigue/low energy (experienced by 100%), headache (in 95%), "pressure in head" (in 95%), difficulty concentrating (in 95%), difficulties in remembering (in 95%), and irritability (in 95%). After medical examinations, including oculomotor and vestibular tests using an Interacoustics TM
Procedure
The participants performed four posturography tests, identically executed by all subjects. Two tests included recording the stability during 120 s of quiet stance while standing with eyes closed (EC) or eyes open (EO) as instructed. Two additional tests, also performed with eyes closed and open, included a 30-s quiet stance period successively followed by a 200 s period with balance perturbations. The balance perturbations were randomized and induced by vibrators strapped over the gastrocnemii (calf) muscles. The vibrators (6 cm long and 1 cm in diameter) produced an 85 Hz and 1.0 mm amplitude vibration. The custom-made vibrators (Section of medical engineering, Skåne University Hospital, Lund, Sweden). The vibration was produced by revolving a 3.5 g weight placed 1.0 mm eccentric on the rotation axis. The vibrators were custom-made for its purpose by the Section of medical engineering, Skåne University Hospital, Lund, Sweden, and the design included using DC motors from Escap, La Chaux-de-Fonds, Switzerland. The vibrations were applied as a sequence of individual balance perturbations by turning ON and OFF the vibrators. The duration of both the ON and OFF state ranged randomly from 0.8 to 6.4 s, according to a pseudorandom binary sequence (PRBS) schedule (25). An identical stimulation sequence was used for all participants and in all balance perturbation tests.
The stability was assessed with a custom-built force platform (Department of Automatic Control, Lund University, Sweden), which recorded both torques and shear forces with six degrees of freedom (d.f.) using force transducers with an accuracy of 0.5 N. A custom-built software (Postcon TM , Department of Clinical Sciences, Lund University, Sweden) sampled the force platform data at 50 Hz and produced the balance perturbation sequences by using a 16-bit AD-board (PCI-6036E, National Instruments).
At each posturography test, the subject was instructed to stand in an erect and relaxed posture with their arms folded across the chest. The subject stood barefoot on the force platform and using guidelines, their heels were placed 3 cm apart and their feet placed deviating about 30 • open to the front. When performing tests with eyes open, the subject was instructed to focus on a 4 × 6 cm image placed at eye level on a wall 1.5 m in front of them. None of the test subjects had any prior experience with the posturography tests used in the study, and they received no information about how the balance perturbations would affect them. During the posturography tests, the subjects listened to calm classical music using headphones to avoid extraneous sound distractions. The subjects were allowed to rest for 5 min between each of the four posturography tests.
Analysis
Only the properties of the movements in the anteroposterior direction were analyzed since balance perturbations from calf muscle vibration primarily induce forward and backward movements (24, 26). The variance of the torque values recorded by the force platform was calculated because this parameter corresponds to the energy used toward the support surface to preserve stability (27), i.e., the parameter reflects the efficiency of the central nervous systems' (CNS) control of the standing (28). For a more detailed explanation of the relationship between recorded torque and standing postural control, see Johansson et al. (27).
The force platform recordings were divided into three spectral categories: (a) total torque variance (total energy); (b) torque variance below 0.1 Hz (low-frequency energy); and (c) torque variance above 0.1 Hz (high-frequency energy) using a fifth-order digital Finite Duration Impulse Response (FIR) filter. The filter components were selected to avoid aliasing. This spectral categorization was done to obtain information also about the energy used for corrective changes of posture (i.e., <0.1 Hz) and for fast corrective movements to maintain balance (i.e., >0.1 Hz) (29). Typically, the fast corrective movements (>0.1 Hz) are increased by decreased visual information (29) . /fneur. . or by factors like being overweight or fatigued (30). The lowfrequency movements (<0.1 Hz) are commonly increased by unstable surface conditions like standing on foam (31,32). We analyzed the stability during the two quiet stance tests as one continuous time period (0-120 s). The stability during the two balance perturbation tests was analyzed and segmented into five time periods, a quiet stance period (0-30 s), and during the vibratory stimulation as four consecutive 50-s time periods; Period 1 between 30 and 80 s; Period 2 between 80 and 130 s; Period 3 between 130 and 180 s; and Period 4 between 180 and 230 s. During all the four stimulation periods P1 to P4 analyzed, the vibration stimulus had a similar effective bandwidth in the region of 0.1-2.5 Hz.
Statistical analysis
Before the statistical analyses were performed, the force platform recordings during the four posturography tests (i.e., the two quiet stance tests and the two balance perturbation tests) were first separated into three spectral bandwidths (total: <0.1 Hz and >0.1 Hz). The torque variance values for these three spectral datasets were thereafter calculated and normalized using the subjects' height and weight to account for anthropometric differences, as these individual factors influence recorded torque values (27). Finally, the normalized torque variances were log-transformed (using natural log) in preparation for the statistical analyses.
As the initial step, statistical analyses were performed with repeated measures of General Linear Model (GLM) Analysis of Variance (ANOVA). The statistical method was used after ensuring that all dataset combinations analyzed in the study fulfilled the three criteria: (1) appropriate independency; (2) produced acceptable sphericity according to the Greenhouse-Geisser evaluation; and (3) that the model residuals had normal or close to a normal distribution, thus validating the method's appropriateness (33-35).
The main factor combinations analyzed for their effects on stability within the three spectral bandwidths (total; <0.1 Hz and >0.1 Hz) during the two balance perturbation tests were: The main factor combinations analyzed for their effects on the stability within the three spectral bandwidths (total; < 0.1Hz and > 0.1Hz) during the two quiet stance tests were: 1) Main factors Group (Controls vs. mTBI, df 1) and Vision (Eyes Open vs. Eyes Closed, df 1). The model parameter Group is a Between-Subjects factor, and the model parameter Vision is a Within-Subjects variable.
In all analyses, p-values < 0.05 were considered significant. Procedures were utilized to address potential Type I and Type II errors. The need to use Bonferroni correction was considered but regarded as not required as no dataset was included in the Within-subject or the Between-groups tests more than once. The Shapiro-Wilk test revealed that some datasets were not normally distributed, and that normal distribution could not be obtained by log transformation. Thus, non-parametric statistical methods were used in all post-hoc statistical evaluations (33).
In post-hoc analyses, Within-Subjects paired comparisons were performed to study the effects of Vision and the adaptive changes from Repetition, i.e., the cumulative changes from perturbation period 1 to period 4 (36) using the Wilcoxon matched-pairs signed-rank test (Exact sig. 2-tailed). Betweengroup comparisons of mTBI vs. controls were performed with Mann-Whitney U Tests (Exact sig. 2 tailed) (33).
Stability during the balance perturbation tests
Repeated measures GLM ANOVA of the model (Group, Vision, Repetition) showed by the main factor Group that subjects who suffered from mTBI used significantly more energy compared with controls during the balance perturbation within all three spectral categories; total (143% more energy, p = 0.004), low frequency (122% more, p = 0.007), and high frequency (162% more, p = 0.004), see Table 1 and Figure 1. The results for the main factor Vision revealed that significantly more energy was used with eyes closed compared with eyes open in the categories: total (51% more, p < 0.001) and high frequency (99% more, p < 0.001). Analysis of the main factor repetition revealed that across the vibration periods the subjects of both groups used less energy to handle the balance perturbations within all three spectral categories: total (27% less, p = 0.001), low frequency (33% less, p = 0.012), and high frequency (17% less, p = 0.002).
Repeated measures GLM ANOVA analyses of the model (Vision, Repetition) on group level revealed for the main factor Vision that mTBI subjects used significantly more energy with eyes closed compared with eyes open in the spectral categories: total (63% more, p < 0.001) and high frequency (93% more, . /fneur. . p < 0.001), see Table 2 and Figure 1. Analysis of the main factor Repetition revealed that across the vibration periods mTBI subjects used less energy to handle the balance perturbations within all three spectral categories: total (18% less energy, p < 0.001), low frequency (24% less, p < 0.001), and high frequency (6% less, p = 0.009). Repeated measures GLM ANOVA analyses of the model (Vision, Repetition) on group level revealed for the main factor Vision that control subjects used significantly more energy with eyes closed compared with eyes open in the spectral category; high frequency (116% more, p < 0.001), see Table 2 and Figure 1. Analysis of the main factor Repetition revealed that across the vibration periods control subjects used less energy to handle the balance perturbations within all three spectral categories: total (37% less, p = 0.024), low frequency (42% less, p = 0.047), and high frequency (28% less, p = 0.002).
Post hoc analysis of group
With eyes closed, during all balance perturbation periods from period 1 to period 4, mTBI subjects used significantly more energy compared with controls in the spectral categories: total (mean 171% more, p ≤ 0.035) and high frequency (mean 153% more, p ≤ 0.044), see Figure 1. Moreover, during balance perturbation periods 1 and 3, mTBI subjects used significantly more energy than controls in the spectral category: low frequency (mean 210% more, p ≤ 0.036).
With eyes open, during balance perturbation periods 2, 3, and 4, the mTBI subject used significantly more energy than controls in the spectral category: total (mean 137% more, p ≤ 0.044). During all balance perturbation periods from period 1 to period 4, the mTBI subject used significantly more energy than controls in the spectral category; high frequency (mean 183% more, p ≤ 0.047). Moreover, during balance perturbation period 3, the mTBI subject used significantly more energy than controls in the spectral category; low frequency (178% more, p < 0.001).
Post hoc analysis of vision
The mTBI subjects used significantly more energy with eyes closed compared with eyes open in the spectral category: totalduring the initial quiet stance period (103% more, p = 0.008) and during all balance perturbation periods from period 1 to period 4 (mean 63% more, p ≤ 0.008), see Table 3. The mTBI subjects also used significantly more energy with eyes closed compared with eyes open in the spectral category: high frequency-during the initial quiet stance period (75% more, p = 0.019) and during all balance perturbation periods from period 1 to period 4 (mean 93% more, p < 0.001). Moreover, mTBI subjects used significantly more energy with eyes closed compared with eyes open in the spectral category: low frequency-during the initial quiet stance period (116% more, p = 0.015).
The control subjects used significantly more energy with eyes closed compared with eyes open in the spectral category: total-during the initial quiet stance period (98% more energy, p = 0.009), see Table 3. The control subjects also used significantly more energy with eyes closed compared with eyes open in the spectral category: high frequency-during the initial quiet stance period (54% more, p = 0.003) and during all balance perturbation periods from period 1 to period 4 (mean 116% more, p ≤ 0.005).
Post hoc analysis of adaptation to balance perturbation
In mTBI, the balance perturbations caused a cumulative adaptation of about 34% with eyes closed in the spectral category: low frequency (p = 0.048), see Table 4.
In controls, the balance perturbations caused a cumulative adaptation of about 34% on average with eyes closed in the spectral category: total (p = 0.016) and a cumulative adaptation of about 31% in the spectral category: high frequency (p = 0.002), see Table 4.
Stability during the quiet stance tests
Repeated measures GLM ANOVA of the model (Group, Vision) showed that mTBI subjects used significantly more energy than controls during quiet stance within all three spectral categories: total (128% more, p = 0.034), low frequency (136% more, p = 0.048) and high frequency (109% more, p = 0.015), see Table 5 and Figure 2. The significant results for the main factor Vision revealed that less energy was used with eyes closed compared with eyes open in the spectral category: low frequency (18% less, p = 0.048). Moreover, significantly more energy was . /fneur. .
Post hoc analysis of group
With eyes closed, the mTBI subject used significantly more energy compared with controls in all three spectral categories: total (222% more, p = 0.003), low frequency (290% more, p = 0.014), and high frequency (124% more, p = 0.002), see Figure 2.
Discussion
The mTBI group used more energy to stand compared to the control group during balance perturbations and quiet stance. They also had poorer sensorimotor adaptation compared to the control group as evidenced by more energy use for repeated balance perturbations. These differences tended to be accentuated when standing with eyes closed and manifested through the higher contribution of fast corrective movements to maintain balance. The findings of this study support the premise that a sports-related concussion might reduce postural control and motor learning through sensorimotor adaptation. An impaired adaptive capacity has both diagnostic and therapeutic implications and might add to the explanation of symptoms perceived in some sufferers of mTBI. A reduced adaptive capacity may equate to prolonged symptoms and reduced effectiveness of rehabilitation. Moreover, our results also indicate the value of assessing postural control in individuals with mTBI.
General e ects of mTBI on stability
The mTBI group used significantly more energy to maintain stability compared with controls during the balance perturbation, both with eyes open and eyes closed, and within all three spectral categories. The spectral categorization was done to obtain detailed information about the energy used for corrective changes of posture and used for fast corrective movements to maintain balance (29). Typically, the fast corrective movements (>0.1 Hz) are increased by decreased visual information (29) or by factors like being overweight or fatigued (30). The slow movements (<0.1 Hz) are commonly increased by unstable . /fneur. .
surface conditions like standing on foam (31,32). Intriguingly, the largest differences found between groups were that mTBI subjects used significantly more fast corrective movements than controls. Changes in the fast subconscious movement control, imply that central sensorimotor processes are not operating as efficiently as they should. The difference in postural control between the mTBI group and the control group was greater during balance perturbations compared to quiet stance. Moreover, with balance perturbations, the mTBI group required more energy to stand with eyes closed and eyes open, whereas quiet stance tests revealed differences between the groups with eyes closed only (Figures 1, 2). Thus, the use of balance perturbations to examine postural control following mTBI adds important information compared to the quiet stance (9,21).
The predominant cause for persisting symptoms of mTBI is suggested to be central nervous system white matter pathologies, produced by the rotational forces during the trauma (37). However, it is also deemed that all relevant central white matter lesions or disturbances may not be large enough to appear as structural pathologies in conventional neuroimaging (38,39). In line with this notion, in mTBI patients examined on a median of 22 days after a trauma, a decrease in fractional anisotropy and an increase in mean diffusivity in the cerebellum correlated to imbalance or dizziness symptoms (40). However, Gard et al.
(15) recently reported that most of the patients included in this study had impaired saccular and posterior canal functions, in line with a lesion to the inferior vestibular nerves. This was observed without detecting obvious posterior fossa changes on 7 T MRI. Thus, in some mTBI patients, a vestibular dysfunction may contribute to the balance disturbances.
Adaptation
With repeated balance perturbations, the level of energy used to maintain upright standing diminished in the control group as expected, showing sensorimotor adaptation. However, the degree of sensorimotor adaptation was diminished in the mTBI group within all three spectral categories. When balance is perturbed, the brain detects the imbalance and makes changes to body posture, and generates predictive muscle responses to reduce the imbalance when the perturbation is repeated (24, 41-43). This adaptive response is mediated initially by cortical mechanisms and thereafter by subcortical mechanisms (44,45). The reduced ability to adapt to repeated perturbances suggests a reduction of such a function. Sensorimotor adaptation is crucial to sporting performance and therefore, a reduced ability in this respect will affect competitiveness at the highest levels. However, sensorimotor adaptation is also crucial to everyday motor activities including postural and locomotor control.
Another possibility to be contemplated is that mTBI subjects might distrust their postural ability and react to this with poor adaptation in the same way as patients with a fear of falling or persistent postural perceptual dizziness (PPPD) do (46-48). However, our mTBI subjects were elite athletes, and a fear of falling seemed less likely as the mere mechanism of the observations. Moreover, a common reaction to a psychologically instigated fear of falling is increased recorded activity within the >0.1 Hz frequency range but normal or lower than normal activity in the <0.1 Hz frequency range, due to increased rigidity caused by co-contraction of the anti-gravity muscles (49). The mTBI subjects did not display such characteristics but presented significantly higher activity than controls both in the <0.1 Hz and in the >0.1 Hz frequency ranges both during quiet stance and balance perturbations.
Role of vision
The differences between mTBI and controls were accentuated when standing with eyes closed. With visual feedback, the mTBI subjects presented smaller differences from controls in both adaptations and posture control measurements. That is, the mTBI subjects had less need for fast corrective movements and less need for major changes of posture. Arguably, the frame of reference provided by vision helps to stabilize posture and reduce the effects of perturbations. An alternative explanation is that the subjects with mTBI become more dependent on visual information for balance control. The latter is in line with the finding that most of these subjects showed signs of vestibular dysfunction (15). Symptoms that might be of vestibular origin following SRC are commonly reported. In one study, 68% of athletes reported dizziness and 36% imbalance following SRC (50). Most of the symptoms resolved within 2-7 days (67.1 %), but in a small group (7%) it persisted for more than a month (50). Of note, the mTBI subjects assessed in the current study suffered their injury more than 6 months before the study and reported persistent symptoms for more than 6 months. Thus, at the time of this study, all mTBI subjects should have recovered from the typical SRC symptoms of head and neck pain and of headache. Moreover, any vestibular impairment from the mTBI incident should have been compensated for by upweighting visual and somatosensory information and suppressing less accurate vestibular information.
When changing from standing with eyes open to standing with eyes closed, there is a shift from using vision as the predominant information source for stability control to instead using the mechanoreceptive and proprioceptive systems as the predominant information source. Vision provides a frame of reference for the head movements, and thus, for the movements of the top segment in our biomechanically multisegmented body. This promotes using a segmental up-down postural control strategy. The predominant frame of reference for the mechanoreceptive and proprioceptive systems is the lowest body .
/fneur. . segment's contact with the ground, which promotes the use of a segmental bottom-up postural control strategy. This change in control mode and available reliable sensory information is reflected by increased use of fast corrective movements, as the mechanoreceptive and proprioceptive sensory systems are able to support such feedback control processes (29). Both the mTBI athletes and controls changed in a similar way to predominantly using a mechanoreceptive/proprioceptive control mode. However, the mTBI subjects used significantly more energy with their eyes closed, not only within the highfrequency range but also within the total frequency range. Maintaining postural control is dependent upon accurate sensory cues. The sensory information from vision, vestibular, and somatosensory receptors are conveyed to the central nervous system to be integrated into an aggregative perception of the present body position and movements. Thus, there can be several potential reasons why stability is poorer with eyes closed. This includes the possibility of diminished somatosensory information, reduced integration of somatosensory information, and an impaired exaction of postural responses in a mechanoreceptive/proprioceptive control mode. Postural instability may follow a vestibular nerve lesion following an mTBI, particularly acutely, and with eyes closed. To compensate for this, the central nervous system may become increasingly dependent upon visual information (51). However, postural control was significantly poorer in mTBI subjects than in controls with eyes open as well, evidenced by the use of more energy to stand, and this difference tended to be accentuated during the latter periods of the balance perturbation. Noteworthy, the level of fast corrective movements in vision control mode was as high in mTBI subjects as it was in controls in mechanoreceptive/proprioceptive control mode. Hence, mTBI subjects were able to utilize vision to enhance their stability, but access to visual information was not alone able to compensate for the deficits in postural control in mTBI subjects.
Limitations
We evaluated a group of mTBI athletes under a specific set of criteria, e.g., that they suffered from significant and long-term (>6 months) post-concussive/mTBI symptoms. However, our selected cohort is likely not representative of all SRC athletes, and the findings of our study may not be observed in athletes with a more limited duration of symptoms. The results are also based on a relatively small sample size, and the study may only apply to mTBI subjects with a similar medical history. Another . /fneur. .
limitation of the study is that we included no control group of asymptomatic SRC athletes. We are also aware that additional relevant information might have been obtained if the subjects had been assessed earlier after subjects sustaining their injury and if the subjects had been assessed repeatedly to monitor recovery. The included mTBI subjects were young adults who exercised more than the average population. The controls were matched in terms of age and physical activity but were not elite level.
Conclusion
The mTBI subjects had significantly poorer stability than controls during balance perturbations and quiet stance. Moreover, controls had better adaptation to balance perturbations than mTBI subjects. These differences tended to be accentuated when standing with eyes closed and were manifested by an increased contribution of fast corrective movements to maintain balance. Hence, the findings of this study support the premise that sports-related concussions might affect postural control and sensorimotor adaptation. Furthermore, the findings suggest that posturography could be included in the battery of diagnostic tests to assess mTBI. Results from posturography may predict the length of expected recovery and the outcome of rehabilitation since sensorimotor adaptation is key to both.
Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Ethics statement
The studies involving human participants were reviewed and approved by Ethics Review Board (Dnr 2017/1049), Lund University, Sweden. The patients/participants provided their written informed consent to participate in this study.
Author contributions
MM, NM, FT, and YT contributed to the conception and design of the study. AA-H and AG carried out the data collection. P-AF performed the statistical analysis. MM, P-AF, FT, AA-H, and AG interpreted the results. AA-H, AG, P-AF, and MM wrote the first draft of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.
Funding
This work was funded by the Swedish Research Council for Sport Science CIF 2021-0105, Swedish Brain Foundation 2020, Swedish Research Council VR 2018-02500 (all to NM), and hospital ALF funds (to NM and MM). | 2022-09-14T18:07:19.265Z | 2022-09-12T00:00:00.000 | {
"year": 2022,
"sha1": "36878507bf20674c8d02d1ca7234e0bf7ee997ba",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "36878507bf20674c8d02d1ca7234e0bf7ee997ba",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
158525757 | pes2o/s2orc | v3-fos-license | Facilitating Spatial Negotiation : a pragmatic approach to understanding public space
The workshop ‘Facilitating Spatial Negotiation’, which took place as part of the ‘Past, Present and Future of Public Space’ International conference on Art, Architecture and Urban Design that took place in Bologna (2014), promoted by City Space Architecture, demonstrates a pragmatic approach to understanding how public space can be realised. The method of collaborative painting is employed within a participatory practice that adopts tactics from spatial agency and critical spatial practice. First, this paper provides a descriptive and visual insight into the discussion between six participants on the topic of the street as a public space, in light of the Social Street movement. Then, it sets out how the session can be understood, through analogy, as a creative exercise in performing a common space. By reflecting upon this event through the framework of participatory practice, the focus is on how conflict is revealed and negotiated within the group. Two instants of conflict are discussed, which raise the critical question whether people are, in fact, interested in working together towards the production and use of common space. It is suggested that the implications of this workshop are twofold. First, a truly public space cannot be realised if the principles of common space are not adopted within the process of its negotiation. Secondly, the finding of a common language in the process of negotiating public space is crucial to this process. The painterly approach offers a shared visual forum, but ultimately the use of any facilitating medium depends on people’s responsibility to participate.
Introduction
How to realise a truly public space?Instead of using abstract theories, this alternative contribution in the form of a workshop offers a more pragmatic approach to answering this question.The workshop 'Facilitating Spatial Negotiation' took place as part of the 'Past, Present and Future of Public Space' International conference on Art, Architecture and Urban Design, held in Bologna, promoted by City Space Architecture, on the 26 th of June 2014, and sought to provide the opportunity to actively engage participants in the discussion of public space.The provision of a platform for an interactive sharing of experience and opinions is in contrast to the conventional conference model which is based on a unidirectional knowledge exchange from speaker to audience with scarce time for a negotiation.The workshop's underlying idea is that prior to producing a physical public space, the interests of space must be negotiated to elucidate the public good.By realising such a different kind of event, the hope is to contribute towards a better understanding of the nature and characteristics of public space.So, how can an analogy be drawn between the workshop and public space, in order to arrive at a better understanding of how a truly public space can be realised?Such an analogy can be found on two levels: the first is a question of how public space was discussed, and the second of how public space was performed within the session.First, this paper sets out the methodology and theoretical context, followed by a summary of discussions held within the session and accompanied by visual documentation of the workshop.Then, the paper proceeds to provide a critical reflection on the modes of communication contained in the session, and finishes by stating the implications of this workshop.The conference workshop was organised in collaboration with the representatives of the Social Street movement.Social Street started in Bologna, with the aim to create a social network between residents in the same street, in order to make sharing of expertise and knowledge possible and to pursue collective projects of common interest.The goal is to reap all the benefits of greater social interaction between neighbours.At the early stages of Social Street, however, there were still questions as to the realisation and impact of this social movement.For this reason, the topic 'your street, your choice' became the point of departure for the workshop.The invitation to participate in the one and a half hour session was taken up by five conference attendees.As an artist-researcher, my role in this workshop was primarily as a facilitator, with certain objectives: to give participants equal time and space to contribute; to acknowledge differing and co-existing positions and to provide moments of reflection on the process.However, just like the other participants, I was also an active participant in the conversation, contributing situated knowledge as a citizen.
Methodology and theoretical context
The workshop was undertaken as part of an interdisciplinary practice-led doctoral research project, which explores how contemporary painting practice can become an agency-based strategy in the architectural design process.The workshop employs the method of painting and adopts certain tactics from spatial agency, as explained in more depth in 'Painting Architecture: Towards a Practice-Led Research Methodology' (Mlicka, 2014).In this research project, painting is more than just a medium or mode of (re)presenting: it is a critical and engaged practice which has the potential to have a transformative effect.The method of collaborative painting is employed for the particular advantages it offers above other tools and mediums concerning facilitating collaborative thinking.First of all, this method has been developed to improve accessibility and participation.The large dimensions of the canvas sheet, laid horizontally on a table, enable all participants to contribute to the conversation visually, thereby literally giving form to their ideas.This visual 'forum' functions as common ground on which to accumulate ideas, juxtapose arguments and construct shared goals.The relatively low-skill method of painting enables participants to communicate without jargon and is more accessible than the exclusive high-tech tools being developed today.Secondly, the medium-specific qualities of painting make it possible to reflect a diverse range of perspectives, for example through the rich choice of colours.More importantly, the medium makes it possible to create layers, so that ideas can be built upon and changed.As a flexible working method, it uses a variety of tools and techniques, making different forms of expression possible.Finally, the method of painting enables participants to focus on the process of sense-making, rather than the production of a physical outcome.No attention is paid to aesthetic decisions or creating a finished artwork or design.Instead, the act of painting slows down the conversation, giving people the opportunity to consider and listen to others.Taken together, the simultaneous use of a visual and a verbal language can provide a platform for a more democratic mode of communication.The approach to the methodology which informed the workshop is built upon two types of practices: spatial agency and critical spatial practice.These practices share many characteristics despite originating from the disciplines of architecture and the arts respectively.Spatial agency indicates a shift away from a focus on the architectural product towards a situated and embedded praxis which is conscious of, and works with, its social, economic and political context (Awan, Schneider, & Till, 2011).This different approach to architecture is based on the fundamental idea that architecture is dependent upon others at every stage of its development (Till, 2009).Thus, instead of spatial thinking, I employ the concept of spatial negotiation that necessarily takes place between people.A critical spatial practice is situated between the disciplines of art and architecture, investigating their modes of operation while drawing attention to the wider social and political problems (Rendell, 2006).Such investigation can use creative means, such as painting in this case, to facilitate and reflect upon how things are done.This research project harnesses the process of painting to engage people in the negotiation of space during informal meetings.Critical, agency-based spatial practices employ various methods to transform space and to engage and empower people, while remaining critical of their own approach and aware of the entire context of existing relationships, networks and processes.One such tactic is participatory practice, which provides a relevant framework for the analysis of this workshop.As a participatory event, the workshop's goal is to take the differing perspectives about public space to a new level by creating a common space in which (situated) knowledge can be discussed.Based on previous sessions, I have defined such a process through four stages: sense-making, confrontation, negotiation and collaboration.The essence of this process is to reveal conflicts, which is an opportunity for self-critical reflection that potentially sets in motion a process of transformation.Transformation is understood here as going through a certain process to arrive at a point that is fundamentally different from the starting point.This might be manifested inchanges in social relationships or a change in the way that space is produced.The approach of instigating a conflictual space builds upon the larger debate in which participatory practice is criticised for pursuing a consensus-based culture.Consensual participation, it is argued, results in stasis and perpetuation of the status quo (Miessen, 2010).A transformative © Queensland University of Technology participation, on the other hand, is developed from within the context of the given situation instead of through the application of abstract professional knowledge from the outside (Till, 2005).Therefore, the underlying question raised in this paper is to what extent the process of participatory painting can excavate conflicts existing in a specific situation and in addition, whether it can set in motion a transformative effect.
The Discussion
The first issue brought up in the group was related to the difficulties citizens face when asking for help, when help-seeking isseen as a weakness in our culture.It raised a further question as to whether people who have questions or are in need of help turn to their neighbour or rather tend to use the internet.We discussed how finding local solutions results in certain patterns of relationships (fig.1), and the role of a Social Street therein.A schism appeared between the arguments based on an academic, if not purely scientific approach, and the more grounded approach stemming from citizens' own experiences and observations of urban relationships.This gap was further widened by the tendency of participants to talk at, rather than with other participants.In this respect, the invitation of participants to paint their argument made it possible to replace long monologues with shorter exchanges.Nevertheless, the incompatible levels on which the question of social relationships was addressed resulted in a rupture of the discussion into two parallel conversations.The discussion shifted to public space as a field of conflicted interests, adding a second layer to the painting (fig.2).On the one hand, public space was described in terms the interest investors have in it based on its financial value.On the other hand, it was addressed as physical space that can be used by citizens.The presence of Social Street was invoked to incite an active use of public spaces to prevent the prioritisation of the financial value.Another participant argued that the problem is the external control of space.This division between citizen control and administrative control over public space is visualised through two opposite fields: the municipality and the public.The setting out of this polar opposition as the third layer of the painting proved beneficial to the continuation of the discussion, because it was pointed to when clarifying or positioning arguments.It also revealed problems in the verbal communication related to linguistic misinterpretation, since the session was conducted in both English and Italian.Whereas 'public' was understood as physical public space by the Italian speaking participants, it signified the public understood as people for the English speakers.Accordingly, 'public' was replaced with 'people'.One participant also asked to replace the written word 'comune' (meaning municipality) with 'public administration' (fig.3).Having clarified these terms, it became possible to continue thinking on the same level about possible bridges between the two fields.While discussing approaches like urban acupuncture and micro interventions, two new questions emerged: how can we make people open their doors again to connect to others and what can the public administration do to improve the social impact on the street?The latter question resulted in a disagreement as to where Social Street's responsibilities lie and its opportunities are, and to what extent such a movement needs money, space and time.On the one hand, it was argued that things can be achieved without the public administration but on the other hand it was argued that a close relationship and collaboration between the people and the public administration is necessary.The conservative structure of Italian bureaucracy was criticised and the idea was brought up that a third actor could be useful to mediate between these two bodies.This is marked with the yellow circle in the centre of the painting (fig.4).More crucially, the problem of passivity is brought up: people appear disinterested in engaging with others.While there are many followers on Social Street's Facebook page, very limited numbers actually attend the events.We discussed whether a Social Street could be more effective through the use of incentives, such as shop discounts for local residents, incentives with the potential to change the mentality of staying indoors.Incentives could encourage the public to socialise more, in order to create a sense of community.This could fundamentally change Social Street itself, from merely a virtual platform to a movement in public space.Alternatively, people could be nudged to socialise in public spaces by offering some form of play.While this was an interesting proposition, it was not developed further in the conversation.
Finally, the discussion returned to the question of whether Social Street's representatives could become mediators between street residents and the public administration (fig.5).
Although the opinions were divided, there was a clear tendency to keep Social Street as a virtual platform for the locals.
The Performance
The second approach to reflecting on the workshop is one that considers how the discussion took place.Through an analogy between the workshop and public space, some insights can emerge as to how such a common space can be realised.If the painting (fig.6) is understood as a site on which relations are produced and played out (Donszelmann, 2009), it can provide certain clues as to how the session developed.In particular, it can function as an indicator, illustrating if, and how conflict played out between participants.Two types of conflict can be identified in the session.The first conflict was the incompatibility between approaches to discussing social relationships.Whereas some of these attitudes were visualised through small diagrams, they remained isolated points of view.There was no visual negotiating taking place, reflecting the lack of mutual questioning and listening within the conversation.Visual negotiation can be discerned when, for example, participants work together on one image by layering ideas and using contrasting colours.In the first iterations of this painting, the visual concepts were neither in direct confrontation nor in constructive contribution to other ideas, even if they might have been responses to previously mentioned concepts.They existed on a different level of engagement with the topic of public space.Within the verbal communication, the participants talked past each other and parallel conversations were conducted to avoid confrontation.This problem of parallel conversations revealed even more systematic problems concerning how disciplinary boundaries create barricades against interdisciplinary collaboration.As has been observed in the discipline of architecture at large (Till, 2009), there was also a tendency in this workshop to maintain a hierarchical relationship between the professional as 'the expert' and the citizen as 'layperson'.Instead, a more productive common space could be achieved if participants acknowledged that they are simultaneously expert citizens as well as citizen experts (Till, 2005).Without this attitude, negotiation is evaded rather than invited.
© Queensland University of Technology The second conflict was revealed during discussions in relation to the question as to where Social Street should be situated between the people and the public administration.
The low density of the painting at this stage reveals that participants were disengaged from the process, leaving the table for periods of time or turning their attention elsewhere.This disrupted the discussion and made constructive negotiation unsustainable.
A number of images were painted out of time with the conversation, for example when participants returned to the table and started painting without joining in the discussion.On other occasions, whilst one participant was engaged in the act of painting another participant would attempt to take over the conversation, instead of allowing the discussion to slow down and in doing so make time for listening.As a result, it was difficult to bring the discussion through the four stages of sense-making, confrontation, negotiation and collaboration.In particular, what was lacking was a more critical selfconscious reflection on the participatory process.The conflicts were touched upon, but could not be played out without the full involvement of the participants.
A shared concern arises by comparing the way that public space was discussed and, through analogy, performed in this workshop: are people actually interested in socialising in or through a public space?Within the conversation, this question emerged as a fundamental problem to both Social Street and the existence of public space in cities.In regard to the process of creating a shared space through collaborative painting, there was both limited contribution and a limited capacity to negotiate.At the end of the session, there was no sense of having reached a higher or different level of understanding.If public space means connecting people with each other, then the way in which public space is negotiated must reveal the willingness to connect by way of listening, asking and talking.
There might indeed be a need, as noted in the session, for incentives or nudges to bring people together.In this workshop, collaborative painting was utilized as such an enticement to work together.While this might not be a preferred or familiar mode of creativity for all involved, participants need to take responsibility to participate (an active form as opposed to merely being a participant) if they want to realise a truly public space.After all, a public space without a public remains a blank canvas.
Conclusion
What are the implications of this pragmatic approach to understanding public space?There are two suggestions that I would like to make here.First, a truly public space cannot be realised if the principles of common space are not adopted within the process of thinking about public space.Without the willingness or ability to reach out to others and communicate in a constructive way there is no opportunity for a common space.This problem has been theoretically approached through three modes of proactive participation: attitude, relevance and responsibility (Miessen, 2010;Till, 2011).Such an ideal framework for participation, however, leaves unsolved the question of how to accomplish this in practice.This leads to the second suggestion, that to achieve a mode of communication which enables people to work together towards a common space, a shared language must be found.This means a shared language not only in the linguistic sense of arriving at the same interpretation of words, but also in terms of disciplinary language by discarding jargon, and in terms of how participants might address each other without a haughty attitude.If participants do not make an effort to meet each other on the same level, there is a limited likelihood of achieving a common space.
The workshop was a creative exercise in performing such a common space, with the painterly approach offering a common language.Firstly, this is because painting as a visual medium remains isolated from the problems inherent to verbal communication, such as jargon or linguistic issues, providing a secondary visual layer to develop ideas.Secondly, it is an unfamiliar medium to most participants so there is no schism between experts and laypersons, opening up ground for an equal platform for participation.The quest for a common language does not, however, mean that consensus is sought.Rather, it means that participants can: commence upon making sense of their differing perspectives; confront these differences and reveal the points of conflict; negotiate solutions and alternatives and finally, arrive at a point of a collaborative effort towards the new goal.Nevertheless, as the Chinese saying goes, if the wrong man uses the right means, the right means work in the wrong way.The way that painting is used as a means for facilitating spatial negotiation ultimately depends on the intentions of all participants.The workshop as analogy for creating public space implies that it is a challenge for the local community to use the street proactively, especially at a time when it is easier to communicate through Facebook than face-to-face.
Figure 2 .
Figure 2. Video still from conference workshop on 26.06.2014 after 20:50 minutes.Source: Video documentation by Agnieszka Mlicka.
Figure 3 .
Figure 3. Video still from conference workshop on 26.06.2014 after 37:24 minutes.Source: Video documentation by Agnieszka Mlicka.
Figure 4 .
Figure 4. Video still from conference workshop on 26.06.2014 after 55:04 minutes.Source: Video documentation by Agnieszka Mlicka.
Figure 6 .
Figure 6.Photograph of the collaborative painting from the conference workshop on 26.06.2014.Source: Photo documentation by Agnieszka Mlicka. | 2019-05-20T13:03:21.884Z | 2017-10-11T00:00:00.000 | {
"year": 2017,
"sha1": "190e2de4b719dd673638213fd7c420d9b731be55",
"oa_license": "CCBYNC",
"oa_url": "https://www.journalpublicspace.org/index.php/jps/article/download/257/256",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "190e2de4b719dd673638213fd7c420d9b731be55",
"s2fieldsofstudy": [
"Art"
],
"extfieldsofstudy": [
"Political Science"
]
} |
234096172 | pes2o/s2orc | v3-fos-license | THE CHALLENGES OF TEACHING EFL FOR ADULT LEARNERS: ONLINE LEARNING DURING THE COVID-19 PANDEMIC
This study was conducted to determine the challenges of teaching EFL for adult learners online learning in UIN Mataram during the covid-19 pandemic. This study used a qualitative method. Data collection instruments are questionnaires and interviews. The findings of this study show that there are challenges for lecturers in teaching EFL for adult learners. In conclusion, this paper has shown so many responses about the challenges experienced by the lecturer, and students while studying online, such as; the positive and negative impact of online learning.
INTRODUCTION
Policy on the implementation of online education conducted by schools and also universities in Indonesia are responding to the Covid-19 pandemic that has hit almost the whole world. The Ministry of Education and Culture of the Republic of Indonesia encourages the implementation of the learning process online. This is by the Circular Letter of the Minister of Education of the Republic of Indonesia number 3 of 2020 on the Prevention of Coronavirus Disease in the Education Unit, and the Letter of the Secretary-General of the Ministry of Education number 35492/A. A5/HK/2020 dated basis because they do not have a laptop, the fact the cost is too high, internet access is insufficient and internet quota Insufficient. The need now is to replace it with online mastering. This reminds researchers that even though we live in the same country, we now do not share the same resources. Online learning is an open and spread learning system using pedagogy tools, made possible through the internet and network-based technologies to facilitate the formation of learning and knowledge processes through meaningful action and interaction (Dabbagh, 2005). As stated by Inchiparamban & Pingle (2016) that online learning makes it possible for learners to take up a course without attending an educational institution. Learners get the benefit of taking up a course from their home or from any place they are comfortable with.
Regarding the implementation at UIN Mataram of online classes, the use of technology in supporting online learning depends on three pedagogical factors. The pedagogical approach which is the first factor includes student-centered learning, the role of the teacher as a facilitator, and the integration of knowledge Carrillo & Flores (2020: 2). The main objective of this study was to determine the challenges faced for adult learners in learning English during the Covid-19 pandemic. Even though, students can develop English language skills during difficult times with online learning. Based on observations, It was seen at the beginning of the meeting, WFH's online learning policy made students have to do activities that could still be done at home. This makes many students hone their creativity in English skills during online lectures. Students have many ideas and ways to fill their spare time. Proving that the WFH policy can have a positive impact or change for adult students in learning English, among others; 1) students become more diligent in reading English literature, English journals, or other references that support learning; 2) students can learn and hone soft skills such as making direct interaction communicating English with lecturers who teach English courses and trade online; 3) foster the courage to speak directly; 4) create independence, express ideas, to answer questions in group discussions. Sepulveda-Escobar, P., & Morrison, A. (2020) suggests that there are several challenging factors in teaching EFL online, such as the lack of direct lecturers 'interaction with students and sudden changes in settings are among those that most strongly influence the participants' learning process. Regardless of the challenges presented, the student-teacher suggested that this unique experience would contribute positively, at least to a specified extent.
Challenges, Online learning, EFL, Adult learner
Related to the implementation of online classes, (Carrillo, C., & Flores, 2020), explained the use of technology in supporting online learning depends on three factors of pedagogy. The first pedagogical factor includes student-centered learning, the role of teachers as facilitators, and the integration of knowledge. The second factor is the design of learning that includes the flexibility of learning, learning that suits the individual needs of each student, according to the context, social, learning process, and the use of appropriate tools and technologies. The third factor is facilitation which includes clear expectations, appropriate questions, understanding and sensitivity to cultural issues, timely feedback; constructively; and detail, as well as the high attitude and commitment of the students. The above exposure can be seen in the preliminary study titled "is the online learning good amid the Covid-19 Pandemic? The case of EFL learners" by (Allo, 2020). The purpose of this study was to investigate learners' perceptions of EFL learning online during the COVID-19 pandemic. This research applies to the qualitative method. The online learning system (in a network) is a learning system without face-toface learning between teachers and students but is done online using the internet network.
Lecturers must ensure that teaching and learning activities continue, even though students are at home. The solution, lecturers are required to be able to design learning media as an innovation by utilizing online media. This is by the Minister of Education and Culture of the Republic of Indonesia regarding Circular Number 4 of 2020 concerning Implementation of Education Policies in the Emergency of the Spread of Corona Virus Disease (COVID-19). The learning system is implemented through a personal computer (PC) or laptop connected to an internet network connection. Lecturers can learn together at the same time using groups on social media such as WhatsApp (WA), telegram, Instagram, zoom applications, or other media as learning media. Thus, lecturers can ensure students take part in learning at the same time, even though they are in different places.
All sectors are feeling the impact of the corona. One of them is the world of education. Judging from the surrounding events that are happening, both lecturers and students who do not have cellphones to support online learning activities feel confused. Some students who do not have cellphones are learning in groups, so they do learning activities together. Starting to learn through video calls connected with the lecturer concerned, being asked questions one by one, and taking attendance via Voice Notes Challenges, Online learning, EFL, Adult learner 013 available on WhatsApp. The materials are also given in the form of a video which is less than 2 minutes long. The problems that occur are not only in the learning media system but the availability of quotas that require quite high costs for students and lecturers to facilitate online learning needs. It is clear, the challenge of lecturers as teachers are to find the easiest solution faced by students to remain active in participating in EFL learning as a foreign language in higher education.
Online learning cannot be separated from the internet network. An Internet network connection is one of the obstacles faced by students whose living quarters find it difficult to access the internet, especially since these students live in rural, remote, and disadvantaged areas. Things become a special concern to find a solution. Even if someone uses a cellular network, sometimes the network is unstable, because the geographical location is still far from the range of cellular signals. This is also a problem that often occurs in students who take online learning so that the implementation is not optimal. It should be realized that the unpreparedness of educators and students for online learning is also a problem. The transfer of conventional learning systems to online systems was very sudden, without proper preparation. But all of this must be carried out so that the learning process can run smoothly and students actively follow even in the conditions of the Covid-19 pandemic.
In line with these challenges, the main challenges in online learning during a pandemic. First, lecturers have very limited time in preparing and/or adapting offline learning material to online. Second, the lack of or limited opportunities for teachers and students to interact directly and freely during online learning results in disruption of the learning process. Third, the use of an effective pedagogical approach requires more effort in motivating and activating students in online learning (Huang et al, 2020: 2). The challenges of online learning do appear to be visible before us, not just one or two schools but comprehensive in several regions in Indonesia. The very important components of the online learning process need to be improved and improved. First and foremost is a stable internet network, an application with a user-friendly platform, and online socialization that is efficient, effective, continuous, and integrative to all education stakeholders.
The solution to this problem is that the government has made efforts for teaching and learning quotas that have been distributed free of charge, however, it must provide a policy by opening free online application services in collaboration with internet providers Challenges, Online learning, EFL, Adult learner and applications to help this online learning process. The government must also prepare an online learning curriculum and syllabus and disseminating information about the procedures for implementing online learning, about their roles and duties. To make it easier and online learning can run smoothly as expected. Thus, online learning is an effective solution for learning at home to break the chain of Covid-19 spread, physical distancing is also a consideration for choosing this learning. Good cooperation between lecturers, students, parents, and universities is a determining factor in making online learning more effective.
Research design
This research uses a qualitative approach, a type of case study which has the main objective of obtaining in-depth descriptive data, case study research is a research design that is comprehensive, intense, detailed, and in-depth and is more directed as an effort to analyze problems or problems a contemporary phenomenon (Herdiansyah, 2015). Thus, the main objective of this study is to identify and describe the challenges in learning English online faced by adult learners who are just starting with online learning methods. Challenges are identified, analyzed, and described in depth.
Data Collection
The data was obtained through observations on the English learning process that took place in the period September-December 2020 through the WhatsApp Group of MD A. UIN Mataram Faculty of Da'wah and Communication-Indonesia. Meanwhile, in interviews, researchers conducted interviews with lecturers that covered the challenges faced by lecturers teaching EFL during online lectures. Interviews were also conducted with students to find out the challenges of learning online and the impact of their application on their English skills.
Data analysis
From the amount of data obtained, then the researchers interpret all the data obtained, including a summary of all findings, comparing the findings with the theory, Challenges, Online learning, EFL, Adult learner draw conclusions, and propose limits and expectations of further research (Creswell & Creswell, 2018).
Findings
The implementation of physical distancing policy which then becomes the basis for the implementation of online lectures, by utilizing information technology that applies suddenly, makes lecturers and students because they are not ready. Some lecturers were surprised because they had to change the learning system that initially learned face-toface now all learning is done online. In detail, the findings obtained in this study are described as follows. Based on observations, documentation, and interviews, lecturers and students have challenges about class and technical management during online lectures.
Challenges faced by lecturers, and students. In detail, the findings obtained in this study are described in the table as follows. Based on questionnaire data obtained information about online lectures during the Covid-19 pandemic.
No
The Interview Section Aspects of the challenges 1 What are the challenges that you face for teaching online English subject in UIN Maatarm Da'wah Mangement Program?
The first is students are difficult to access the internet, so the students are difficult to get the materials, understanding materials given by the lecturer, and difficult to send the assignment to the lecturer. 2 Is that all your students difficult to access the internet?
No, it is just several students 3 What kind of challenges in teaching online English subjects in UIN Mataram?
Several students don't have a smartphone, while the students that have it are difficult to access the signal and the internet data packages. 4 Is it hard to know the capability of your students to understand your material?
Of course, yes. The lecturer does not much know the students' understanding of the materials that I have given to them since they respond are not directly as good as in online learning. 5 What's your solution to the challenges that you have to deal with?
My solutions are giving additional time to the students to send their assignment. Besides, if the students cannot use the handphone or access the internet, the students are allowed to send their Challenges, Online learning, EFL, Adult learner assignments to my house. The teacher also should be able to find interesting materials for the students are enjoy and not boring while learning the materials and do the assignment.
Based on data taken from interviews and online questionnaires, the challenges faced by lecturers, namely, some students face limited internet access because the signal area is not well covered. Then, it is also found that some students do not have smartphones, more detailed information such as lecturers prefer to teach offline classes.
The lecturer cannot deliver the material directly and do not the students understand the material or do assignments. Students are not active in responding to material provided by lecturers, especially male students and tend to be lazy when learning online. According to observations and interviews with lecturers, clear in However, how to make the use of google sites optimal? Therefore, this article will discuss how to make the use of Google sites to be optimal in distance learning.
b. Online learning or e-learning is learning that is done without face-to-face or learning using internet media, a learning process that utilizes information and communication technology (ICT) systematically by integrating all components of cross-space and time learning. E-learning is a learning system that is used as a means of teaching and learning that is carried out without having to face-to-face with the educator and the student (Setiawan, 2020).
c. Zoom is a free HD meeting app with videos and screen sharing for up to 100 people.
Zoom is a communication application using video. The application can be used in a variety of mobile devices, desktops, phones, and space systems. e. Google Classroom is a mixed learning foyer for educational scopes that makes it easier for teachers to create, share and classify each paperless assignment.
Google classroom is an application created by Google that aims to help lecturers and students if they are unable to, organize classes and communicate with students without having to be tied to the schedule of lectures in the classroom. In addition, lecturers can give assignments and directly provide value to students. Delivery of learning with e-learning is learning by utilizing internet technology to improve the learning environment with rich content with a wide scope. E-learning is the utilization of learning media using the internet, to deliver a series of solutions that can improve knowledge and skills. Each learning method must contain the formulation of organizing lesson materials, delivery strategies, and managing activities by paying attention to learning objectives, learning barriers, characteristics of learners, to obtain effectiveness, efficiency, and learning appeal (Miarso, 2004).
It can be concluded that the above exposure from some digital mode platforms mentioned above, WhatsApp is the most familiar among students and lecturers of UIN Mataram, because, before the Covid-19 pandemic, students and lecturers have also intensively communicated and interacted with learning through WhatsApp groups. There is no denying this, where the process aims to avoid the boredom of students in learning and absorb information related to teaching materials, in addition, to avoid monotonous patterns carried out by a large number of lecturers. Based on the table above data taken from the interview and online questionnaire, students state that they do not have laptop facilities and students in online lectures use mobile phones as a medium of online lectures. This data shows that there are still many students who have not been able to attend online lectures to the maximum. Based on the factor of time affordability, and different places and far from internet signals.
The paradigm shifts of conventional learning to online, although not difficult, but takes a long time. Because it is related to the paradigm change in academic culture.
As the results of the survey of students in the field in the description, there are still many shortcomings that must be addressed and equipped with readiness in learning English online. Based on the survey data above, shows that students are not ready to carry out this online learning, because it concerns academic culture, which includes values, attitudes, knowledge, and skills, as well as the readiness of facilities and infrastructure related to information literacy among students.
Teaching EFL Online Learning
Concerning the implementation of online learning, students who are claimed to have received challenges from online-learning. Data shows that online-learning offers flexibility, provides up-to-date information, provided rich, unlimited resources, encourages students to read, helps fewer active students become more active, and is faster and simpler. It is easy and makes students more independent in learning, asking, and answering online, through zoom meetings and short messages what app application.
According to Rosenberg, online-learning is a form of distance learning, but not all distance learning can be categorized as e-learning. Specifically refers to the use of Internet technology to convey and improve knowledge and skills based on 3 criteria, namely: (1) online-learning is a network capable of making changes, receiving, distributing, and sharing instructions or information quickly; (2) delivered to learners through computers using standard Internet technology; (3) focus on a broad view of a learning solution that is outside the traditional paradigm (Rosenberg, 2001)
Challenges, Online learning, EFL, Adult learner
Lecturer challenges are also clearly and clearly visible with the availability of internet facilities by providing teaching materials that are delivered to students can facilitate with all series of learning, even though distance learning from home to home, learning from the house, has been the path taken by each lecturer in applying and ingesting English teaching materials. Another challenges for lecturers and students is with the availability of quotas and time constraints each lecturer can distribute the material automatically quickly, and accurately and effective time and place, and easily precisely does not make students become bored and makes students easy to access when, where, and at a certain time, it depends on the activity of the learner at all times.
Online-learning offers flexibility. From interviews, flexibility becomes the main online-learning opportunity offered. The flexibility refers to the ease of access to which students can access anytime and from anywhere. Some of the students who have been interviewed included: Nurhidayati, Ahmad Fatoni, Sasmita, Marni, kilan and admitted that online-learning offers flexibility in terms of time and place. In interviews they stated: One of the opportunities is flexibility where we can access online-learning from anywhere, anytime, and under any situation. We just need to have gadgets and an internet connection. Online-learning is not limited to space and time. So, more flexible (Nurhidayati). Online-learning opportunities are more flexible and efficient in terms of time and place, as can be done anytime and anywhere (Ahmad Fatoni). The onlinelearning opportunity is that we can learn from anywhere not just from the classroom. As long as we have connections, we can learn (Sasmita). Online-learning is more flexible and easy. We don't have to come to class. Also, we can do it everywhere (Marni). Onlinelearning opportunities are flexible and accessible. It can be done anytime and anywhere as long as there is internet access. It's flexible in time and place (Kilan).
The participants made several key points regarding flexibility in online learning.
First, online learning is not limited to space and time. Second, by using online-learning, students can learn from any place, not just from the classroom. Lastly, to be able to access materials with online-learning, an internet connection is essential. Online learning provides flexibility in learning supported by (Smart, K. L., & Cappel, 2006) which argues that online-learning brings flexibility and convenience because online-learning allows students to access lessons anytime and anywhere, and students can complete units of all teaching materials delivered by educators quickly.
Teaching EFL Online Learning Challenges
Results of Observations, documentation, and interviews show that lecturers and students have challenges and opportunities concerning how to operate online classes.
Challenges faced by lecturers and learners such as limitations in explaining learning materials related to the variety of learning methods carried out. In the previous conventional class, lecturers had a creation, flexibility, and diversity of teaching methods that were filled with materials, characteristics of assignments, characteristics of learners, situations and learning environments, and so on. However, lecturers have limited ability to explain because learning is done online. It also challenges lecturers and students in carrying out learning activities. Activities are limited to sharing learning materials, videos, assignments, voice mail, and related information that has a limited interaction pattern, namely from lecturer to student and from student to the lecturer. Interaction patterns between students are quite limited. Group or pair work activities have not been seen in learning, so activities that require students to discuss and group are still limited. Studentcentered learning cannot be done because of all materials, activities, questions, tasks, and information centered or sourced from lecturers.
Lecturers and students are also constrained by learning time which makes the development, explanation, strengthening, enrichment of learning materials, and clarification of materials difficult to do. For example, some students who do not understand the material in a predetermined learning period should ask the lecturer directly through the lecturer's personal WA. Lecturers cannot explore widely both in terms of techniques, strategies, and methods to give lectures online due to limited space and time. time constraints and methods or techniques in providing satisfactory explanations to students are more than difficult enough to provide online. The same thing happens to students. Providing feedback, from all responses there are signal constraints as obstacles and cannot be connected properly through the internet network. These limitations can be caused by residences and domiciles that are far from the internet network that is sufficient to be connected.
The limited financial and economic condition of families for disadvantaged and remote villages as a concern: especially during the Covid-19 pandemic brought 22 heads of family to have difficulty in providing internet quota for their children during online classes that have a very high level of need during online. Thankfully, when the data Challenges, Online learning, EFL, Adult learner collection of this research ended, the government had provided free quota assistance for students since the 3rd week and 4th of September 2020. However, it is not having vouchers and internet pulses but another challenge that arises is some areas where students live difficult to reach signals. The village that the internet network is very weak inactivate.
Some students report that they are experiencing disruption or difficulty signaling in the area where they live.
We are well aware of that, that the importance of technology and information as a development that is highly superior to online learning. The solution is very appropriate for the government, lecturers, and other educators. So that lecturers give thought to create a WA group for each class collectively. This is a special concern for educators and lecturers to be used as challenges to be free and can be a common idea in finding the right solution. Challenges for lecturers and students of UIN Mataram Indonesia to adapt to English learning as an EFL foreign language online.
Challenges of lecturers and students in the utilization of information technology
in Online Learning. The use of smartphones and laptops in online learning can increase what will happen to learners (Anggrawan, 2019). That poly over-use of information and communication technology in online learning software, among others, means not tied to space and when (Pangondian, R. A., Santosa, P. I., & Nugroho, 2019). Research has been conducted that examines the use of gadgets such as smartphones and laptops in learning.
The ability of smartphones and laptops to access the internet helps students to follow online learning (Kay, R. H., & Lauricella, 2011) ; (Gikas, J., & Grant, 2013); (Chan, N. N., Walker, C., & Gleaves, 2015); (Gökçearslan, Ş., Mumcu, F. K., Haşlaman, T., & Çevik, n.d.). The use of online learning using zoom cloud meeting has the advantage of being able to interact exclusively between students and lecturers as well as teaching materials but has the disadvantage of being extravagant and less effective when more than 20 learners (Naserly, 2020).
Furthermore, the challenge of online learning is the availability of internet services. Some students access the internet using cellular services, and some use WiFi services. When the online learning policy was implemented at the University of Jambi, students returned home. They have difficulty cellular frequency when in their respective regions if there is a frequency produced that is very weak. Online learning has weaknesses when internet services are weak, and the instruction of lecturers who are poorly Challenges, Online learning, EFL, Adult learner understood by the student (Astuti, P., & Febrian, 2019). in (Sadikin, A., & Hamidah, 2020b).
Positive and Negative Impacts of Online Learning for Adult Learners
Students state that the opportunity to access subject matter from any location at any time is the most positive aspect of online learning. Besides, lecturers can integrate application or web-based resources into learning materials without difficulty.
This gives students individual and unique learning opportunities. However, some students argues that online learning has positive and negative impacts: Table 3. The impact of Online Learning
No
Those are online learning has positive and negative impacts A Online learning has positive impacts 1 The positive impact is that we can maximize the use of technology that has developed today. 2 We can get the material easily and in accordance with the way we learn can be in accordance with our desires for example when lying down, while eating, or when chatting with other people. 3 We can also be free to study without the standard time that is usually determined for each course on campus. B Online learning has negative impacts 1 The negative impact online learning process is only happen in one direction, making it difficult for students to consult with material that is felt to need a deeper explanation or understanding. 2 Teaching and learning activities which should be replaced by online lectures but only replaced by the accumulation of assignments that make me dizzy. 3 Online learning are seen in the technicality of the actual use of it. These impacts include how technology is not always efficient, it is harder for students to grasp concepts being taught, online learning can cause social isolation, and can cause students to not develop needed communication skills.
Researchers concluded that challenges of lecturers and in online learning at UIN Mataram, in the teaching of language lecturers and students with the challenge of learning from the house and work from house for these lecturers, is a new history in learning this century, teaching materials can be controlled with technology media in tying materials to students. The main challenge is that lecturers receive English language material every time they lecture online, lecturers cannot face-to-face and cannot control students well in online learning activities, but the positive impact for lecturers are very large because each
CONCLUSION
Based on the findings and discussion, the researcher concludes that the challenges of teaching online are real. The students or teachers dealing the challenges, difficult to get good internet access, and even several students do not have a smartphone.
Besides, since the pandemic situation the normal class at school cannot be used, so conduct online classes is a better way. However, the normal activity in the classroom is giving another atmosphere to the teaching-learning processes. Furthermore, the results also show that the problems caused by internet access and the teaching ways between online and normal classes are the main problem. The lecturer and students surely want to Challenges, Online learning, EFL, Adult learner back to the normal class, although the internet easier the teaching-learning process, however teaching online may not effectively be applied if the teacher and students do not ready yet. The solution are providing internet data packages for the students and the teacher, additional time for the students to send their assignment.
From the above exposure, it can be concluded that since the outbreak of the Covid-19 pandemic, all learning processes in different parts of the world have changed from face-to-face mode to online mode. The application of this online mode always provides challenges and convenience to all parties involved in it. Challenges arise not only in developing countries, but also in developed countries such as Germany, South Korea, and China. The challenges are related to class management and technical learning system which includes the preparation, implementation, and evaluation of online learning that still requires contributing advice from all stakeholders in the world of education including observers of the world of education and education policymakers.
Finally, even though online-learning has many potentials and advantages until the challenges of e-learning are considered, the full potential and advantages cannot be fully obtained by the student. The implication for the Institute, online -learning can be an alternative method to prevent the spread of covid-19 and learning continues to run normally and lancer. Online learning for lecturers and students provides free movement of space and time to provide materials and students can access freely and at any time both materials and tasks submitted. For other researchers, it is expected to be able to find simple methods and easier for future learners. | 2021-04-18T03:38:54.337Z | 2021-01-31T00:00:00.000 | {
"year": 2021,
"sha1": "f8dadbbe1498df8dd34cb729fd06b6db64d597b3",
"oa_license": "CCBYSA",
"oa_url": "http://journal.stbapontianak.ac.id/index.php/spectral/article/download/64/pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "f8dadbbe1498df8dd34cb729fd06b6db64d597b3",
"s2fieldsofstudy": [
"Education",
"Computer Science"
],
"extfieldsofstudy": [
"Psychology"
]
} |
259409852 | pes2o/s2orc | v3-fos-license | Cough as a Clue: Tracheal Endocarcinoma Unveiled
Tracheal adenocarcinoma (TAC) is a rare malignancy often characterized by significant delays in diagnosis, often attributed to the non-specific nature of symptoms, leading to subsequent challenges in management. The prognosis remains poor, highlighting the need for early detection and multidisciplinary treatment strategies. Surgical resection is recommended for eligible patients, followed by postsurgical irradiation. However, further research is required to give a better perspective on therapeutic interventions and enhance patient outcomes. This paper reports the case of a 50-year-old male, who presented with dyspnea, hemoptysis, and cough. The computed tomography (CT) revealed an intratracheal tissue mass. The cytological examination and immunocytochemistry confirmed the diagnosis of primary adenocarcinoma in the trachea. The treatment involved silicone tracheobronchial Y-stent followed by adjuvant chemotherapy with carboplatin and paclitaxel, and radiotherapy (60 Gray) with good clinical improvement.
Introduction
Tracheal adenocarcinoma (TAC) is an extremely uncommon and poorly reported tumor accounting for less than 1% of all malignancies [1]. The diagnosis is often made at an advanced stage, primarily attributed to the delayed onset of nonspecific symptoms such as hemoptysis, dyspnea, cough, and stridor [1]. Optimal management strategies hinge upon a careful assessment of the tumor stage and the feasibility of the surgical intervention, while radiation therapy and chemotherapy serve as viable alternative options [1,2]. Due to the limited clinical data, a multidisciplinary approach is indispensable for comprehensive patient care. It is imperative to conduct further investigations to enhance our understanding and to devise more efficacious therapeutic modalities for this condition.
Case Presentation
A 50-year-old male with a smoking history of 35 pack-years presented with an acute worsening of the frequency and the severity of cough superimposing a chronic dyspnea that persisted for eight months along with hemoptysis, asthenia, and significant weight loss. In the early stages of his presentation, the patient's symptoms of persistent cough and wheezing dyspnea were initially attributed to asthma and managed accordingly.
Physical examination revealed wheezing and stridor. The patient exhibited an oxygen saturation of 88% on room air. Chest radiography showed no abnormalities, but a subsequent chest computed tomography (CT) scan revealed a 24*29 mm intratracheal tissue mass accompanied by bronchial dilatation in the middle lobe ( Figure 1). CT scan reveals a luminal narrowing due to intratracheal tissue mass (arrows). Figure A shows the coronal view with a parenchymal window, Figure B displays the axial view with a mediastinal window, and Figure C presents the axial view with a parenchymal window.
During flexible bronchoscopy, a white tumor with a propensity to bleed was observed ( Figure 2). The tumor obstructed and infiltrated the tracheal wall in the lower 1/3 of the trachea, occupying more than 70% of the tracheal lumen. Rigid bronchoscopy confirmed the presence of a tumoral mass located proximal to the carina obstructing the bilateral bronchial tree and around 70-80% of the tracheal lumen.
FIGURE 2: Bronchoscopic view of the tumor
The tumor is causing tracheal obstruction and infiltration, with significant luminal occupation. Rigid bronchoscopy confirms tumor presence proximal to the carina (arrows), obstructing the main bronchi.
Histopathological evaluation of the biopsy sample revealed a poorly differentiated invasive squamous cell carcinoma, with immunocytochemistry favoring an invasive solid adenocarcinoma ( Figure 3). Further staging investigations revealed the presence of a locally advanced subcarinal mediastinal tumor, hilar lymph node involvement, and a pulmonary nodule in the right upper lobe. The tumor was classified as T4N1M1.
FIGURE 3: Histopathological images
Histopathological images demonstrating poorly differentiated invasive squamous cell carcinoma with features suggestive of invasive solid adenocarcinoma.
Following a multidisciplinary consultation, the decision was made to proceed with four sessions of chemotherapy with carboplatin and paclitaxel and radiotherapy (60 Gray) after placing a silicone tracheobronchial Y-stent (limb of 16 mm diameter and 4 cm length) to alleviate tracheal obstruction and dilatation after removing part of the tumor (Figure 4). The patient was subsequently transferred to the thoracic surgery team, which successfully relieved the tracheal obstruction and dilatation.
FIGURE 4: Endobronchial imaging
Endobronchial imaging showing the placement of the silicone tracheobronchial Y-stent: Relief of tracheal obstruction visualized Significant clinical improvement was noted immediately afterward, with the resolution of respiratory symptoms. A silicone tracheobronchial Y-stent placement with the same dimensions was scheduled for future management. Unfortunately, five months later, the patient died from confirmed coronavirus disease 2019 (COVID-19) pneumonia.
Discussion
Tracheal cancers are extremely rare, with an estimated incidence of approximately 0.1 per 100,000 individuals annually [3]. To date, the largest analysis of primary tracheal tumors was conducted using the comprehensive Surveillance, Epidemiology, and End Results (SEER) database, which included a total of 578 cases [3]. Within this report, 55% were male, and the prevailing histological type identified was squamous cell carcinoma, accounting for 45% of the cases [3]. Other histological types observed encompassed adenoid cystic carcinoma, small cell carcinoma, large cell carcinoma, sarcoma, adenocarcinoma, and unspecified or undifferentiated carcinoma. Notably, while primary tracheal tumors in adults are predominantly malignant, representing nearly 90% of cases, the rate of malignancy in children is significantly lower, ranging from 10 to 30% of cases [3].
TAC rarity and intricate management give rise to notable diagnostic and therapeutic complexities [1]. First, tracheal tumors often mimic asthma during the initial stages and the diagnosis is frequently delayed due to the late emergence of nonspecific symptoms such as hemoptysis, dyspnea, coughing, and stridor. While rare, the involvement of adjacent structures in TAC can lead to additional symptoms such as dysphagia and hoarseness [4]. The misdiagnosis is further supported by the presentation of a normal radiograph, adding to the challenge of accurate identification [1]. In our particular case, the patient manifested similar presentations associated with weight loss and asthenia and followed by a normal chest radiograph which are typical of the clinical features reported in TAC.
Chest radiographs rarely detect changes indicative of tracheal neoplasms [4]. This is primarily due to the superimposition of the thoracic spine and mediastinal structures on the trachea in the posteroanterior view. Therefore, the lateral view of a chest X-ray is often more helpful in aiding in diagnosing tracheal neoplasms [4]. However, physicians should be aware that chest radiographs may initially appear normal in most patients with tracheal neoplasms.
The diagnosis of TAC in our patient was confirmed through evaluation using flexible and rigid bronchoscopy, along with histopathological examination. The presence of a white growing tumor observed during flexible bronchoscopy, followed by a biopsy, revealed a poorly differentiated invasive squamous cell carcinoma. Immunocytochemistry results further supported the presence of invasive solid adenocarcinoma, highlighting the heterogeneity commonly observed in tracheal tumors [5].
The prognosis for patients with malignant tumors of the trachea is generally poor, and the long-term median survival of patients with TAC undergoing postoperative irradiation therapy is still poorly explored [2]. However, the prognosis for patients who are eligible for surgical resection is generally more favorable compared to those who are not [2]. Therefore, surgical resection is often recommended as the primary treatment approach for most cases of primary tracheal tumors, regardless of tumor burden, margin status, histology, or nodal status. All patients who undergo surgical resection for tracheal neoplasms require postsurgical irradiation as part of their treatment protocol [6].
Overall, early recognition, accurate diagnosis, and comprehensive treatment planning involving multidisciplinary approaches are crucial for optimizing patient outcomes. Further research is needed to better understand the biology of TAC and develop targeted therapies to improve patient prognosis.
Conclusions
TAC represents a rare and challenging form of cancer. The limited prevalence of this malignancy underscores the difficulties in its diagnosis and treatment. A multidisciplinary approach is essential to address complex management effectively. Furthermore, ongoing research endeavors are crucial to expand our knowledge, refine diagnostic techniques, and establish optimal therapeutic approaches for this rare condition.
Additional Information Disclosures
Human subjects: Consent was obtained or waived by all participants in this study. Conflicts of interest: In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no other relationships or activities that could appear to have influenced the submitted work. | 2023-07-11T02:32:17.840Z | 2023-06-01T00:00:00.000 | {
"year": 2023,
"sha1": "552f026660283904144964f1db84e41965fb3d15",
"oa_license": "CCBY",
"oa_url": "https://assets.cureus.com/uploads/case_report/pdf/162233/20230613-26017-1ylnzrw.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3b9684eac50f5f9240d8966a950cbc0dd588051c",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
207930182 | pes2o/s2orc | v3-fos-license | New DoS approaches to finite density lattice QCD
We present two new suggestions for density of states (DoS) approaches to finite density lattice QCD. Both proposals are based on the recently developed and successfully tested DoS FFA technique, which is a DoS approach for bosonic systems with a complex action problem. The two different implementations of DoS FFA we suggest for QCD make use of different representations of finite density lattice QCD in terms of suitable pseudo-fermion path integrals. The first proposal is based on a pseudo-fermion representation of the grand canonical QCD partition sum, while the second is a formulation for the canonical ensemble. We work out the details of the two proposals and discuss the results of exploratory 2-d test studies for free fermions at finite density, where exact reference data allow one to verify the final results and intermediate steps.
I. INTRODUCTION
Finding a suitable approach for Monte Carlo simulations of finite density QCD that extends the accessible part of the QCD phase diagram towards larger values of the chemical potential is currently one of the great challenges for lattice field theory. The problem is that at finite chemical potential the fermion determinant is complex and cannot be used as a probability in a Monte Carlo process. Among the approaches that have been explored to solve this so-called complex action problem are density of states (DoS) techniques, introduced to lattice field theory in [1,2]. The key challenge for a DoS approach is to compute the density with sufficiently high accuracy, such that it can be integrated over with fluctuating integrands that appear when evaluating observables at finite density. A naive determination with, e.g., simple histogram techniques turned out to be useful only for very low densities [3][4][5][6].
Inspired by the Wang-Landau [7] approach an interesting new development was presented by Langfeld, Lucini and Rago [8][9][10][11][12][13][14]. The idea is to use a parameterization of the density as the exponential ρ(x) = exp(−L(x)) of a piecewise linear and continuous function L(x). Vacuum expectation values restricted to the intervals where L(x) is linear are used to determine ρ(x) with very high precision. A variant of the Langfeld-Lucini-Rago method is the DoS Functional Fit Approach (FFA) [15][16][17][18][19] and in recent years both techniques were used to obtain interesting results for several bosonic lattice field theories at finite density (see, e.g., the review [20]).
However, no modern DoS formulation for systems with fermions was presented so far, and thus no clear path towards precise DoS calculations for finite density QCD has been outlined yet. The challenge is to formulate the DoS approach such that it is compatible with conventional pseudo-fermion Monte Carlo techniques that may be applied to a real and positive fermion determinant. * Member of NAWI Graz.
In this paper we discuss two proposals how to implement the DoS FFA for finite density lattice QCD. The first of the two is based on using a suitable pseudofermion representation of the QCD grand canonical partition sum. The imaginary part of the action is identified and the density is considered as a function of that imaginary part. DoS FFA is used to determine the corresponding density and observables are then obtained as integrals of the density.
The second proposal implements DoS FFA in a canonical setting. The canonical partition functions at fixed net quark number are written as the Fourier moments with respect to imaginary chemical potential µ = iθ/β. Considering θ as an additional degree of freedom in the path integral allows one to implement the DoS FFA and compute the density as a function of θ. Observables at fixed net quark number are then again obtained as integrals of the density.
In both formulations only Monte Carlo simulations without sign problem are needed, which furthermore can be implemented using well established techniques of standard lattice QCD simulations. We work out the details of the two new approaches and present the results of small exploratory 2-d studies of the free case where exact results can be used to assess the results and intermediate steps of the new proposals.
II. GENERAL FORMULATION OF THE DOS FFA APPROACH
Before we can discuss our two new DoS approaches to finite density QCD we first need to discuss the details of the DoS formulation we use, the Functional Fit Approach. The FFA [15][16][17] is here presented for a general bosonic theory with a complex action problem and we will later show that with suitably chosen pseudo-fermion representations finite density lattice QCD can be brought into the general form introduced in this section.
A. Partition sum and density of states
The vacuum expectation values O for some observable O that we consider here can be written as bosonic path integrals, where Φ denotes an arbitrary set of general bosonic lattice fields that can be based on sites or links, and D [Φ] is the corresponding product measure. We have already separated the exponent of the Boltzmann factor into two terms, the real part S R [Φ] of the action and the imaginary part αX [Φ]. We have allowed for a real-valued coupling α ∈ R multiplying the imaginary part, which is useful in some of the applications we have in mind. (2) where J [Φ] is an arbitrary real and positive functional of the fields. Below we will identify J [Φ] with some observable which in general can be decomposed into pieces that obey these requirements. Note that different choices of J [Φ] result in different densities ρ (J ) (x) and we use a superscript J to indicate which density we refer to.
With the densities ρ where in the second line we have explicitly listed also the particularly simple case where the observable is some The range of integration for the integrals dx in (3) depends on the properties of the imaginary part X [Φ]. If X[Φ] is bounded by some number x max , so is the integration range. We will see that this is the case for the canonical DoS formulation discussed in Section 4. Furthermore, usually one can identify symmetries to show that the densities ρ (J ) (x) are even or odd (depending on J ), such that the integration interval where we need to determine the densities is [0, x max ].
In case X[Φ] is unbounded the integration runs up to x = ∞, which is the case we will encounter in the direct DoS approach discussed in Section 3. Again symmetries can be used to show that ρ (J ) (x) is even (or odd) such that the actual integral is ∞ 0 dx. Furthermore we will see that the densities ρ (J ) (x) quickly decrease with x such that the range of integration can be truncated such that also in this second case we need to determine the densities in an interval x ∈ [0, x max ].
Having defined the densities ρ (J ) (x) and expressed observables as integrals over these densities we now have to address the problem of finding a suitable representation of the densities and how to determine the parameters used in the chosen representation.
B. Parametrization of the density
The densities ρ (J ) (x) are functions of the parameter x and we are interested in the densities in some finite interval [0, x max ]. For parameterizing the densities, we divide the interval [0, x max ] into N subintervals as follows: The densities ρ where the L (0) = 0 we can completely determine the constants a n as functions of the slopes k n . A simple calculation shows that L (J ) (x) can be written in the following closed form, from which we obtain the explicit form of the density ρ Thus our parameterized density ρ (J ) (x) depends only on the set of slopes k (J ) n , one for each of the intervals I n . We point out that the parametrization allows one to work with intervals I n that have different sizes ∆ n . In particular in regions where the density ρ (J ) (x) varies quickly one should use smaller intervals, while in regions of slow variation larger ∆ n can be used to reduce the computational cost. For a coarse scan of the density ρ (J ) (x) with the goal of determining the regions of quick variation, one can do a first numerically cheaper determination with large ∆ n which subsequently is refined with finer intervals. These techniques are referred to as preconditioning and are discussed in detail in [15][16][17].
C. Evaluation of the density parameters with FFA
To determine the density, we need to compute the slopes k with the corresponding restricted partition sums Z (J ) n (λ) given by where we have introduced the support functions Θ n (x) = 1 for x ∈ I n , 0 for x / ∈ I n .
In the restricted expectation values X (J ) n (λ) and the partition sums Z (J ) n (λ) we have introduced a free real parameter λ which couples to the imaginary part X [Φ] and enters in exponential form. Varying this parameter allows one to properly explore the x-dependence of the density in the whole interval I n . The expectation values X In the first step we have rewritten the restricted partition sum as the integral of the density ρ (J ) (x) over the interval [x n , x n+1 ]. In the second step the parameterized form (9) was inserted for that particular interval, which gives rise to a simple integral of an exponential that can be evaluated in the closed form on the right-hand side.
Comparing (10) and (11) it is obvious that the restricted vacuum expectation value X (J ) n (λ) can be computed as the derivative X (J ) n (λ) = d ln Z (J ) n (λ)/dλ, such that we find the closed expression, After multiplicative and additive normalization we can express the result for X (J ) n (λ) (which in its normalized form we denote as V (J ) n (λ)) in terms of a function h(s), and has the properties The strategy for determining the slope k (J ) n for an interval I n now is as follows: Using a standard Monte Carlo simulation without sign problem we compute the restricted vacuum expectation value X
III. DIRECT DOS APPROACH FOR LATTICE QCD WITH A CHEMICAL POTENTIAL
In this section we discuss the first of our two implementations of the new DoS approach to finite density QCD.
Here we use a suitable pseudo-fermion representation of the grand canonical partition sum and separate the part with the complex action problem. For this factor we set up the DoS FFA formulation, discuss its properties and present results of a first exploratory test in the free case.
A. Grand canonical partition sum and pseudo-fermion representation
We consider lattice QCD with N f mass-degenerate flavors of Wilson fermions. After integrating out the fermions the corresponding grand canonical partition sum with quark chemical potential µ is given by We consider the theory in d = 2 and d = 4 dimensions using lattices of size V = N d−1 s × N t . The SU(3)valued gauge variables U ν (x) live on the links (x, ν) of the lattice and obey periodic boundary conditions. Their path-integral measure is the product of Haar measures is the Wilson gauge action (we dropped the constant additive term), β g is the inverse gauge coupling and P [U ] the sum over the real parts of the traced plaquettes. By D[U, µ] we denote the Wilson Dirac operator with chemical potential µ in the background of a gauge field configuration U . We write the Dirac operator in the form, with the matrix elements of the hopping terms given by By γ ν we denote the Euclidean γ-matrices in d = 2 or d = 4 dimensions, and κ is the hopping parameter κ = 1/(2d + 2m) with m the bare quark mass. To be specific, we use a representation of the Euclidean γ-matrices where γ d is symmetric, which in d = 4 is, e.g., the chiral representation and in d = 2 the choice γ 1 = σ 2 , γ 2 = σ 1 with γ 5 = σ 3 . In (21) we use matrix/vector notation for the d Dirac indices of the γ-matrices and the 3 color indices of the link variables U ν (x). The chemical potential gives different weight for hopping in forward and backward temporal direction, i.e., the ν = d direction. The fermions obey periodic boundary conditions in the spatial direction(s) and anti-periodic boundary conditions in time, i.e., the terms in (21) that connect sites with x d = N t − 1 and x d = 0 have an additional minus sign.
In order to introduce a pseudo-fermion representation that is suitable for the DoS FFA we write the fermion determinant as In the second step we have written 1/ det D[U, µ] † as a bosonic integral over a complex-valued scalar field Φ(x) with 3d components for Dirac and color degrees of freedom, and in the exponent we use vector/matrix notation for all indices. The constant C is given by In the third step we have organized the exponent into real and imaginary parts such that the pseudo-fermion integral matches the general form introduced in (1), where here we set α = 1. The corresponding real and imaginary parts are given by where we also write the gauge field U as an argument in These two matrices are defined as A straightforward evaluation gives the matrix elements, (23) are real. Thus the pseudo-fermion integral in (22) has the form that allows one to use DoS FFA to evaluate that integral. This will be discussed in more detail in the next section.
Let us add a few comments on the first factor in (22), i.e., the determinant det D[U, µ] † D[U, µ] . Using the well known generalized γ 5 -hermiticity property corresponds to the fermion determinant of two mass-degenerate quark flavors with an isospin chemical potential which is free of complex action problems. We stress, however, that this isospin determinant is of course only a part of the weight and its coupling to the pseudo-fermion factor in (22) generates the full dynamics (see also the comments below).
The matrix D[U, µ] † D[U, µ] is obviously hermitian and has real and non-negative spectrum, such that it is directly accessible with pseudo-fermion methods. Possible approaches are a direct pseudo-fermion representation (below χ and χ j denote bosonic complex-valued pseudofermion fields), or an order-n Chebychev multi-boson representation [21,22] of the form where u j = e i2πj/(n+1) are the coefficients for the Chebychev factorization and we have used (20) can be obtained as follows: Using the triangle inequality one finds Using the definition (21) (32) Thus we find that there is a finite range of µ where the factor det D[U, µ] † D[U, µ] which is free of the complex action problem can be treated with conventional pseudofermion techniques. We stress again, that the estimate (32) is only a crude non-exhaustive bound that essentially reflects the situation of the free case, where condensation sets in at µ = m. For a dynamical background gauge configuration U the spectrum of H[U, µ] is known to contract such that values of µ that exceed the bare quark mass parameter m become accessible. To precisely delimit the range where the pseudo-fermion treatment of det D[U, µ] † D[U, µ] is possible beyond the bound (32) obviously is a dynamical question that has to take into account the emerging finite density physics as well as possible numerical instabilities of the HMC algorithm that can only be assessed in a full QCD simulation, which clearly goes beyond the scope of this presentation. However, already with the simple bound (32) we have established an interesting minimal region where the direct DoS approach is applicable in principle.
B. Implementation of the DoS FFA
To set up the density of states approach and to define the corresponding densities as outlined in the general presentation in Section 2 we need to write the grand canonical definition with the pseudo-fermion representation. Since we consider the general case of N f flavors, we need N f copies of the pseudo-fermion fields, Φ j , j = 1, ... N f , where by {Φ} we denote the set of all these fields. Based on the discussion of the previous section we thus write the grand canonical partition sum of QCD in the form that matches Eq. (1) with α = 1 (irrelevant overall constants were dropped), i.e., where the real and imaginary parts of the pseudo-fermion action, as well as the path integral measure were gener-alized to N f flavors, As we have outlined in the previous section the term (22) can be treated with conventional pseudo-fermion techniques and we combined the corresponding factor for N f flavors together with the gauge field action Following the general DoS FFA strategy outlined in Section 2.1 we now define the densities as where we allow for general observables J [{Φ}, U ] that can be functionals of both, the set {Φ} of pseudo-fermion fields Φ j , as well as the gauge fields U .
In the general outline of the method in Section 2 we have already announced that symmetries can be used to establish that the densities ρ (J ) (x) are either even or odd functions, depending on the observables J , which we assume themselves to be even or odd (general J may be decomposed into even and odd pieces). As an example we briefly discuss the simplest case of J = 1 and show that ρ (1) (x) is even. The symmetry transformation we consider is charge conjugation that for the gauge links and the pseudo-fermion fields is implemented as where * denotes complex conjugation and T transposition. It is straightforward to show that Equally straightforward is to show that the gauge action S g [U ] defined in (19) is invariant under the charge conjugation transformation (37), i.e., S g [U ] = S g [U ]. The invariance of the factor det D[U, µ] † D[U, µ] can be shown using the representation (27) and charge conjugation: Denote by C the charge conjugation matrix that where in the first step we used (27) Thus we have established that ρ (1) (x) is an even function and in a similar way one may analyze the symmetry properties for the general densities ρ (J ) (x) that contain the insertion of some observable J .
As the final step for the implementation of the DoS FFA we need to identify the restricted expectation values as defined in the general description of the method in Section 2. Comparing with the general form (2), (10) we can read off from the densities (36) the necessary restricted expectation values for full QCD, where in the second step we have written the combina- We stress that the ensemble considered in the restricted vacuum expectation values (41) is not simply QCD with isospin chemical potential reweighted to quark chemical potential, where a serious overlap problem would emerge. Instead the exponent of the Boltzmann factor in (41) is given by of the pseudo-fermion terms, which contribute the dynamics of the quark chemical potential.
C. First tests for the free case
For a first exploratory study of the new DoS approach we analyze the free case in two dimensions. Note that due to the restricted expectation values that need to be evaluated, this analysis already requires Monte Carlo simulations also for the free case and indeed provides a nontrivial test of the method. Insight about suitable sizes ∆ n for the intervals, the numerical cost, the accuracy that is needed for the density et cetera, can be obtained. Furthermore, the free case allows for a systematical comparison of the final results and the intermediate steps against analytical results that may be computed with Fourier transformation.
For the free case the density ρ (1) (x) defined in (36) with the help of the pseudo-fermion representation simplifies to (we consider the case of N f = 1 flavor) The gauge field integration has been dropped for the free case and also the Boltzmann factor (35) for the effective action since it is independent of x, such that it only would affect the overall normalization of the density which is set by requiring ρ (1) (0) = 1. Following the steps of the implementation of DoS FFA in the previous section, for determining the parameters k (1) j we need to evaluate the restricted expectation values defined in (41) which for the free case reduce to (44) The imaginary part X[Φ] can be obtained from (23) and (25) (drop the link variables U ν (x) there) as, and the kernel M[λ] in the Boltzmann factor of (44) is given by (see (42)), This matrix is obviously hermitian such that its eigenvalues are real. However, for the existence of the path integral needed for the evaluation of the restricted vacuum expectation values (44), the eigenvalues also have to be positive, and we now address this issue that has been neglected previously. It is easy to see that eigenvalues can become negative for large values of the parameter λ. The result for such an analysis is shown in Fig. 1, where we plot the value λ max as a function of µ/m and compare our results for different masses m and lattice sizes L × L. The value of µ/m where λ max becomes zero signals the breakdown of the method. We remark that the polygon-like behavior of the curves for the smaller volumes reflects the fact that for small volumes the momenta populate the interval [−π, π] with only a few values such that also a "sparse" spectrum emerges, and the different sections of the "polygon" correspond to a different eigenvalue becoming negative.
The data in Fig. 1 is organized in groups where in each group we consider a sequence of values L → ∞ and m → 0 at a fixed value of mL, i.e., we study the fixed volume continuum limit of the free theory. The dotted curves are for mL = 0.64, the dashed curves for mL = 1.28 and the full curves correspond to mL = 2.56. Note that the curves for different mL cluster according to the respective values of m. The figure shows that with increasing µ/m the values for λ max decrease and at a critical value of µ/m the boundary λ max becomes zero, signaling the breakdown of the method. We observe that for all three values of mL we study, the critical value of µ/m converges from above to a critical value of µ/m = 1, which is the value of the chemical potential where condensation sets in. Thus we expect that we can use DoS FFA all the way to the condensation point.
For the dynamical case one expects a similar behavior: For non-trivial gauge links the spectrum of the Dirac operator is known to contract, giving rise to an additive renormalization of the mass and a critical κ that is larger than the free value κ = 1/2d. Qualitatively one finds that for a larger critical κ a larger value of µ is accessible, and one may expect that also for the full case the critical value of µ coincides with the point where condensation sets in. We stress, however, that obviously this is only a very qualitative discussion of the situation in the fully dynamical case and future explicit Monte Carlo calculations will be necessary for a detailed analysis.
Having identified non-vanishing windows of λ where we can safely evaluate the restricted vacuum expectation values X (1) n (λ) defined in (44), we show some of these results for illustration in Fig. 2. We plot the restricted vacuum expectations X (1) n (λ) normalized to the form V (1) n (λ) defined in (15) as a function of λ. The symbols represent the data that we determined in a small Monte Carlo simulation on a 16 × 16 lattice using m = 0.1 and µ = 0.05. For x we use intervals of length ∆ n = 1 ∀n, such that the intervals are given by I n = [n, n + 1]. The symbols shown in Fig. 2 are the data for the intervals I n with n = 0, n = 10, n = 20, n = 50, n = 80 and n = 120. The lines are the fits of V (1) n (λ) with h(∆ n [λ−k (1) n ]) where h(s) is defined in (17).
From the fits of the restricted vacuum expectation value data with h(∆ n [λ−k (1) n ]) we can determine all slopes k (1) n , and from those compute the density ρ (1) (x) using the closed expressions (9). In Fig. 3 we show our results for ln ρ (1) (x) as a function of x, again using the DoS FFA data for V = 16 × 16, m = 0.1 and µ = 0.05. Note that this now is a quantity where for the free case we can compute analytical reference results. These are also shown in Fig. 3 and we find excellent agreement between the DoS FFA data and the analytic results, and stress at this point that we use a logarithmic scale on the vertical axis in Fig. 3.
We point out that further smoothening of the density with suitable fits will be part of a final strategy for DoS techniques -see, e.g., the recent systematic comparison of such techniques in [14].
We conclude this section with commenting on how the analytic reference results shown in Fig. 3 were obtained: Starting from the definition (43) of the density ρ (1) (x) we may use the integral representation of the Dirac delta and find
IV. DOS FFA FOR THE CANONICAL FORMULATION OF LATTICE QCD
In this section we present the second new DoS approach to finite density lattice QCD, now working with the canonical ensemble. The canonical partition sums at different net-quark numbers are expressed as Fourier moments of the grand canonical partition sum at imaginary chemical potential µ = iθ/β and then θ is considered as an additional degree of freedom in the path integral. In this form we may implement the DoS FFA and compute the density ρ(θ).
A. Canonical ensemble and density of states
The setting is as in the previous section, i.e., we study lattice QCD in d dimensions with N f degenerate flavors of quarks, and the grand canonical partition sum Z(µ) is defined in (18) - (21). The canonical partition sums Z N at a fixed net quark number N can be obtained as Fourier integrals over an imaginary chemical potential µ = i θ /β, where β is the inverse temperature in lattice units, i.e., β = N d , with N d being the number of lattice points in time direction (= d-direction), The corresponding free energy density at fixed N is de- Simple bulk observables can be obtained as derivatives of f N with respect to couplings of the theory. An example is the vacuum expectation value of the scalar fermion bilinear, The derivative generates the insertion of Tr D −1 [U, µ], i.e., the traced inverse Dirac operator (quark propagator) as an additional factor in the path integral. Note that also in the quark propagator the chemical potential µ appears and is set to the complex value µ = iθ /β, used for projecting to fixed net quark number N . General vacuum expectation values at fixed N have the form The expressions for the observables at fixed net quark number N can be rewritten with the help of densities ρ (51) J [U, µ] is an arbitrary functional of the gauge fields, which, if it contains the quark propagator, may also depend on the chemical potential µ. Note that again different choices of J [U, µ] result in different densities ρ (J ) (θ) and as before we use a superscript J to make clear which density we refer to.
With the densities ρ (J ) (θ) vacuum expectation values O N at fixed net quark number can be expressed as It is important to note that the densities ρ (J ) (θ) have symmetries that should be identified, because this allows one to reduce the range of θ that one needs to integrate over. Thus also the ρ (J ) (θ) need to be determined only in the reduced range of θ which lowers the numerical cost. In the previous section we have used charge conjugation symmetry to show that the density ρ (1) (x) for the imaginary part x ≡ X[U, Φ] is even in x. Also here it is straightforward to establish that ρ (1) (θ) is even. As before In a similar way as in (53) one can show that also the general densities ρ (J ) (θ) are either even or odd functions, depending on the symmetry of the insertion J [U, µ] (after decomposition into C-even and C-odd parts if necessary). Thus the integrals (52) for evaluating observables only run from 0 to π and exploring charge conjugation symmetry cuts the numerical cost in half.
We conclude this subsection with discussing another interesting symmetry property of the density, which not necessarily can be used to reduce the numerical cost, but reflects an important aspect of the underlying physics: If QCD is in a purely hadronic phase this is equivalent to ρ (1) (θ) being 2π/3 periodic. This property corresponds to the Roberge-Weiss symmetry and can be seen as follows: The statement that QCD is in a purely hadronic phase means that Z N = 0 for all net quark numbers N that are not multiples of 3. We first assume that ρ (1) (θ) is 2π/3-periodic. Then we find which shows that a 2π/3-periodic density ρ (1) (θ) implies that only Z N where N is a multiple of 3 are nonvanishing.
For the inverse statement we can use the completeness and orthogonality of the Fourier factors e iθN and sum over N the Z N in the form (52) with factors e iθN and find where in the second step we used that ρ (1) (θ) is even which in turn leads to Z N = Z −N . The relation (55) implies that if the Z N vanish for values of N which are not multiples of 3 the density ρ (1) (θ) is 2π/3-periodic. We remark, that the representation (55) of course holds in both, the hadronic and a possible non-hadronic phase, and in our small numerical test below we will use the form (55) to determine the canonical partition sums Z N from a fit of the density according to (55).
B. Implementation of DoS FFA
Having discussed the densities ρ (J ) (θ) and their symmetries we can now start the implementation of DoS FFA. For convenience we introduce the notation det D[U, θ ] ≡ det D[U, µ] | µ = iθ /β . For imaginary chemical potential γ 5 -hermiticity guarantees that det D[U, θ ] is real, such that the factor det D[U, θ ] N f is real and positive for even N f (or sufficiently large mass in case N f is odd), and we may write [24]. Using (56) we may write the canonical partition sum as It is interesting to note that the gauge fields U ν (x) and the phase variable θ enter the path integral in the same way, i.e., both are integrated over in the path integral and appear in the exponent of the Boltzmann factor. Thus one may view θ as one more degree of freedom in the path integral and compare (57) with the generic form (1) used in the general discussion of the DoS FFA in Section 2. The exponent in the integral is the action S[U, θ ] for all dofs. and we have already identified the real part of the action as S R [U, θ ]. The imaginary part, which only depends on θ , may be identified as X[θ ] = θ , and the parameter α in (1) is identified with the negative of the net quark number, i.e., α = −N .
Having found a form of the problem that matches the generic form discussed in Section 2, we may identify the restricted vacuum expectation values needed for the determination of the densities ρ (J ) (θ). They are given by where, as mentioned before, the fermion determinant may be represented using pseudo-fermions. The restricted vacuum expectation values (58) do not have a complex action problem and may be computed with standard Monte Carlo techniques. Note that the imaginary chemical potential θ is an additional degree of freedom that is restricted to the interval I n = [θ n , θ n+1 ] and needs to be updated as well.
After evaluating the restricted vacuum expectation values θ (J ) n (λ) they need to be brought into the normalized form V (θ) are computed using (8), (9), and finally observables in the canonical picture at fixed net quark number N are obtained from the densities via (52).
C. Tests of the canonical DoS FFA in the free case
Again we use 2-d free fermions at finite density for a first test also in the canonical formulation of the DoS FFA. For the free case the density (51) for the choice J = 1 reduces to the particularly simple expression (we where D[µ] denotes the Wilson Dirac operator (20), (21) in d = 2 with all link variables set to U µ (x) = 1. It is straightforward to evaluate this quantity using Fourier transformation and the reference data used in Fig. 5 below for verification were computed in this way. The restricted vacuum expectation values θ (1) n (λ) defined in (58) reduce to θ (1) Although we could use the symmetry of the density ρ (1) (θ) and restrict the determination of ρ (1) (θ) to the interval θ ∈ [0, π], we here determine the density for the full range θ ∈ [−π, π]. The symmetry of ρ (1) (θ) should emerge and serves as a consistency check for the calculation. The interval [−π, π] was divided into 100 equal size intervals I n of length ∆ n = 2π/100 ∀n. The Monte Carlo simulation for sampling the restricted θ-integral in each interval I n uses a statistics of 10 6 sweeps of local Metropolis updates separated by 20 sweeps for decorelation and 10 5 sweeps for initial equilibration. The determinant in the acceptance step was computed with Fourier transforma- (1) (θ) as a function of θ. We compare the DoS FFA result (thin blue curve) with the exact result (thick magenta curve). Note that we did not use the fact that the density is known to be an even function and for evaluation purposes numerically determined ρ (1) (θ) in the full range θ ∈ [−π, π]. tion und we typically use 10 values of λ for the evaluation of the restricted vacuum expectation values θ (1) n (λ). In Fig. 4 we show the results for the restricted vacuum expectation values θ (1) n (λ) already in their normalized form V (1) n (λ) according to (15). The symbols represent the data from the Monte Carlo simulation and the full curves are the fits with h(∆ n [λ − k (1) n ]) according to (17). The values of λ where the curves cross 0 are the slopes k (1) n . These crossing points start near 0 for the smallest n (i.e., intervals I n near −π) become negative then, revert back to 0, move to positive values and finally revert again back to 0 for intervals I n near +π. This full oscillation of the corresponding slopes k (1) n reflects the 2π-periodicity of the density ρ (1) (θ) (compare Fig. 5). From the slopes k (1) n obtained with the fits of the restricted vacuum expectation values we determined the density ρ (1) (θ) using (8) and (9). In Fig. 5 we compare the density determined in this way with the analytic result from Fourier transformation. The analytic result is represented by the thick magenta curve on top of which we plot the DoS result (thin blue curve). We stress again that the density ρ (1) (θ) was determined with DoS FFA for the full range θ ∈ [−π, π] and the fact that ρ (1) (θ) indeed comes out as an even function is a consistency check of the method. Anyway, the much more stringent test is the comparison with the analytic result where the plot shows that the DoS FFA curve perfectly falls on top of the exact curve determined as discussed above.
We complete our first test of the canonical DoS formulation with FFA by evaluating the canonical partition sums Z N from the density ρ (1) (θ) via the integrals (52) and comparing these Monte Carlo based results to the exact calculation based on a direct evaluation of (48) with Fourier transformation techniques.
In Fig. 6 we show the corresponding results for Z N normalized with Z 0 as a function of N . The blue diamonds represent the exact results and the red dots the DoS FFA data obtained with the integrals (52). The distribution resembles a Gaussian, rapidly decreasing with increasing |N | (which is of course a volume dependent statement). We find that the DoS FFA data based on (52) match the exact results very well.
We have already pointed out in the discussion of the direct DoS FFA approach that fitting the density with a suitable function will be an important part of future DoS strategies. Usually a large polynomial would be used for such a fit (see, e.g., [14,16,17] for related discussions), but for the canonical DoS approach the representation (55) of the density suggests another option for a fit, namely using a superposition of cosines (sines for odd densities). For the particular case of the density ρ (1) (θ) the fit parameters are the canonical partition sum Z N . In order to test this possibility, we determined the Z N also from a fit of ρ (1) (θ) with (55). The corresponding results are shown as black circles in Fig. 6 and again we find a very good agreement with the analytical results. This demonstrates that smoothening techniques based on periodic representations of the type (55) should be an interesting option to be explored in future development of canonical DoS techniques.
Also for the CanDos we would like to stress that the tests presented here constitute merely a very first assessment of the new approach and only the implementation in a full QCD simulation will show how well the numerical challenges can be brought under control in a calculation that includes the full gauge field dynamics.
V. SUMMARY, DISCUSSION AND OUTLOOK
In this article we have discussed two proposals for a modern DoS approach to finite density lattice QCD based on representations of the theory with pseudo-fermions. In the direct grand canonical approach the fermion determinant is represented with pseudo-fermions and subsequently their effective action is separated into real and imaginary parts such that the latter can then directly be treated with DoS FFA. We worked out the details of the formulation and provided some bounds on the involved kernels of the pseudo-fermion bilinears, showing that the method is applicable in an interesting range of values of the chemical potential µ. We presented very preliminary tests in the free case where a comparison to exact results allows one to assess the new approach. The direct DoS formulation in the grand canonical picture is rather straightforward, but has the disadvantage that also the densities depend on the chemical potential µ. As a consequence the densities have to be re-calcuated when changing µ. Whether this approach can beat our second suggestion, the canonical version of DoS FFA, has to be seen in future more detailed tests.
In the canonical formulation observables at a fixed net quark number N are obtained as the Fourier moments of the partition sum at imaginary chemical potential µ = iθ/β. In this setting we promote the angle θ to a new dynamical variable and interpret the exponent of the Fourier factors e −iθN as the imaginary part of the action. Again we treat this imaginary part with the DoS FFA approach and compute the density ρ(θ) as a function of θ. Observables at different net particle numbers N are then obtained by integrating the same density ρ(θ) with different Fourier factors e −iθN . Obviously here the re-sulting density ρ(θ) can be used for different net particle numbers N , but of course the accuracy of the determination of ρ(θ) has to be higher for larger N . Also here further tests that go beyond the first numerical checks we have presented here will be necessary to assess whether this formulation will be able to compete with other approaches to finite density QCD.
Both formulations we have suggested here, for the first time implement DoS techniques directly in a pseudofermion representation. This has the advantage that these well established techniques can be used in the framework of a modern DoS setting (here the DoS FFA is used but it is also straightforward to implement the ideas proposed here in the LLR framework). Obviously the simple exploratory numerical tests we have presented in this paper only serve to check the plausibility of the two new formulations and a much more detailed assessment will be necessary to explore their potential. Such further numerical tests are currently in preparation.
We conclude with remarking that the techniques developed here go beyond applications to finite density QCD.
The two approaches are general and can be applied to any lattice field theory with fermions where the interaction can be written with the help of a bosonic field such that the fermion action has a bilinear form and a fermion determinant emerges when integrating out the fermions. The bosonic fields do not have to be gauge fields, but also auxiliary fields of a Hubbard-Stratonovich transformation of quartic fermion interactions are a suitable option. We have begun to explore also these possible applications of the newly proposed DoS FFA techniques. Finally we remark that very recently [23] we presented a first test of the new approaches, now for the case of lattice QCD formulated with staggered fermions. | 2019-11-13T06:49:42.000Z | 2019-11-13T00:00:00.000 | {
"year": 2019,
"sha1": "14e7b0c110d2a8fd31e7c19be65f900b41923065",
"oa_license": "CCBY",
"oa_url": "http://link.aps.org/pdf/10.1103/PhysRevD.100.114517",
"oa_status": "HYBRID",
"pdf_src": "Arxiv",
"pdf_hash": "4fd5fadc3ee71cb720f4a061e82a451340c8c79e",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
53708822 | pes2o/s2orc | v3-fos-license | A Mass-Dependent Yield Origin of Neutron-Capture Element Abundance Distributions in Ultra-Faint Dwarfs
One way to constrain the nature of the high-redshift progenitors of the Milky Way is to look at the low-metallicity stellar populations of the different Galactic components today. For example, high-resolution spectroscopy of very metal poor (VMP) stars demonstrates remarkable agreement between the distribution of [Ti/Fe] in the stellar populations of the Milky Way halo (MW) and ultra-faint dwarf (UFD) galaxies. In contrast, for the neutron capture (nc) abundance ratio distributions [(Sr,Ba)/Fe], the peak of the small UFD sample (6 stars) exhibits a signicant under-abundance relative to the VMP stars in the larger MW halo sample (~ 300 stars). We present a simple scenario that can simultaneously explain these similarities and differences by assuming: (i) that the MW VMP stars were predominately enriched by a prior generation of stars which possessed a higher total mass than the prior generation of stars that enriched the UFD VMP stars; and (ii) a much stronger mass-dependent yield (MDY) for nc-elements than for the (known) MDY for Ti. Simple statistical tests demonstrate that conditions (i) and (ii) are consistent with the observed abundance distributions, albeit without strong constraints on model parameters. A comparison of the broad constraints for these nc-MDY with those derived in the literature seems to rule out Ba production from low-mass SNs and affirms models that primarily generate yields from high-mass SN. Our scenario can be confirmed by a relatively modest (factor of ~ 3-4) increase in the number of high-resolution spectra of VMP stars in UFDs.
INTRODUCTION
Our understanding of galaxies forming in a hierarchical universe suggests that a large fraction -and possibly the majority -of stars now in the halo of the Milky Way (MW) originally formed in smaller separate systems that were subsequently accreted and disrupted by our Galaxy (as originally proposed by Searle & Zinn 1978), with the remainder formed in situ within the main Galactic progenitor (Eggen et al. 1962;Abadi et al. 2003a,b;Zolotov et al. 2010;McCarthy et al. 2012;Tissera et al. 2012). While the relative contributions of accreted and in situ populations remain uncertain, simulations in which the stellar halo is assumed to be formed entirely by accretion Cooper et al. 2010) have been shown to have levels of substructure in space, velocities and stellar populations that are broadly consistent with observations (Bell et al. 2008;Schlaufman et al. 2009;Xue et al. 2011). This raises the following question: to what extent can the small systems that survive today (e.g. the satellite galaxies of the MW) be exploited to understand the properties of the small systems that fell in long ago (i.e. the primordial progenitors of the MW halo)?
One approach to this question is to compare and contrast the chemical abundance patterns of the stars in the stellar halo with those in satellite galaxies. For example, at metallicities [Fe/H] −2, stars in the low-mass classical dwarf spheroidals generally have lower α-element abundances than halo stars (as seen in compilations by Venn et al. 2004;Geisler et al. 2007). The observed differences can be explained, in general, by the low star formation rates and efficiencies detected in low mass dwarf spheroidals versus the likely progenitors of most halo stars (see, e.g., review by Tolstoy et al. 2009). Assuming a continuous star formation history, it is true that for all galaxies there exists an epoch for which no appreciable contributions from Type Ia supernovas (which predominantly produce the decline in [α/Fe]) are seen. This means that cosmological and astrophysical effects, which can prematurely quench star formation in galaxies such as reionization (Hoeft et al. 2006) and ram pressure stripping (Mayer et al. 2006), may determine whether low α-abundance ratios appear in systems that are accreted early-on. Therefore, these differences can also be explained within the hierarchical picture of structure formation as a result of star formation histories of the surviving satellites being much more extended than those of the progenitors of the bulk of the Halo (Robertson et al. 2005;Font et al. 2006). However, this statement pertaining to late-time evolution still begs the question: to what extent are the progenitors of the stellar halo similar to the progenitors of the MW's satellite galaxies? This can be addressed by comparing the abundance patterns of stars found in the "Very Metal Poor" (VMP) tail of the metallicity distribution (specifically those VMP stars with [Fe/H] < −2.5) which, because of their low metallicities, are supposed to have formed early on in the history of the Universe. The open black triangles in the top panel of Figure 1 demonstrate that the stellar halo has an average Titanium-to-Iron abundance ratio ([Ti/Fe]) that is roughly constant at all metallicities (measured by [Fe/H]), with a small dispersion that widens in the VMP population. This dispersion can arise when the stochastic nature of star formation is convolved with chemical yields that depend on the masses of the enriching stars (Audouze & Silk 1995;Ryan et al. 1996;McWilliam 1997McWilliam , 1998Norris et al. 2000). For example, Karlsson & Gustafsson (2005) point out that some VMP stars inherit their chemical compositions from gas enriched by just one or a few supernovae (SNe) and have the potential to reflect the full range of abundance ratios implied by the yields from stars of different masses (see also the discussion in Karlsson 2005;Tumlinson 2006;Carigi & Hernandez 2008;Koch 2009;Bland-Hawthorn et al. 2010). Indeed, the range in [Ti/Fe] exhibited in the stellar halo data at low [Fe/H] is consistent with the predictions for the range in individual yields of Ti from models of exploding stars of different masses (Nomoto et al. 2006;Heger & Woosley 2010). In contrast, most stars found with higher metallicities must have been enriched by many SNe, so all their abundances are closer to the average yield for the combined population which can be estimated by integrating the mass-dependent yields (MDY) of the individual stars over the initial mass function (IMF) of enriching stars.
The black open triangles in the lower panels of Figure 1 reveal a much wider spread in abundance ratios for the neutron-capture (nc) elements (here, Barium (Ba) and Strontium (Sr)) for VMP stars found in the stellar halo compared to Ti in the upper panel. At these metallicities, Roederer et al. (2010) suggest (see their Figure 13) that the same massive stars that produce Ti (and the αelements it emulates) also produce nc-elements (thought to originate from core-collapse SN explosions via the rprocess and perhaps from AGB/pre-SN winds via the s-process) but the forms of the MDYs for Sr and Ba are essentially unknown. 5 Hence, one viable explanation of the observed difference in the abundance ratio range between Ti and nc-elements for VMP stars is to again appeal to the stochastic nature of metal enrichment, but now assume a much stronger MDY for nc-elements than for Ti.
The green upside-down triangles in Figure 1 show abundance ratio measurements in stars in the ultra-faint dwarf satellites (UFD) of the MW Norris et al. 2010;Simon et al. 2010;Frebel 2010, and references therein). The Ti distributions for the VMP stars in the UFDs (green upside-down triangles) are very similar to the stellar halo (black open triangles), while the nc distributions show a significant difference, with a clear offset between the medians of the two populations that exceeds the spread due to systematic and observational errors (Frebel 2010). Several types of "differences" can be invoked to explain the origin of the galaxydependence of these abundance ratio distributions: 1. Differences in the mixing of Ti versus nc-elements due to differences in the formation site and process for each element, and, as a consequence, differences in the resultant properties of the enriched ejecta. Assuming that MW progenitors are predominantly larger in size, gas content, and dark matter mass than UFD progenitors, the strength of this effect is mediated by two environmental factors: (i) the depth of the gravitational potential dictates to what extent the different products can be blown out of their respective galaxies by corecollapse SNe; and (ii) the size and dynamics of local gas reservoirs influences how far the products can be evenly mixed in their respective galaxies.
2. Differences in the IMF or MDY of enriching stars due to preferential enrichment of UFDs from primordial populations of hypernovae (Nomoto et al. 2006) and/or pair-instability SN (PISN) ejecta from Pop III stars (see Frebel & Bromm 2012, and their "Case B" for a discussion of these scenarios).
3. Differences in the total masses of stars enriching the VMP populations in the MW halo and UFD (hereafter the "stochastic argument," similar to Case A of Frebel & Bromm 2012).
Note that all of the explanations above implicitly assume that the UFD progenitors are chemically isolated from MW halo progenitors, which has recently been demonstrated to be a plausible supposition in an analysis of N-body simulations by Corlies et al. (2013).
In this paper, we restrict our attention to the last of these "differences", which we consider the simplest model possible. We extend the discussion of dispersions and skews already in the literature to look at how stochastic chemical enrichment can influence the full shape of chemical abundance ratio distributions. Our aim is to isolate the influence of this one effect alone. Specifically, we examine to what extent the current abundance ratio distributions of VMP stars in the MW halo and UFDs can be explained without appealing to differences in mixing, varying the IMF or adopting unique yields. In §2, we outline and describe the assumptions made in our models. In §3, we present the general trends in the shapes of abundance ratio distributions produced by our models due to stochastic enrichment. In §4, we determine the likelihood of drawing the observed distributions of abundance ratios (found in the MW halo and the UFDs) from our simple models. In §5, we discuss the implications and limitations of our results in connection with expectations from other related studies. Finally, in §6, we summarize our results and discuss a possible test of the scenario with near-future observations.
GENERAL APPROACH
Our aim is to determine whether a simple model can simultaneously explain both the similarities in the distribution of [Ti/Fe] and the differences between the distributions of [(Sr,Ba)/Fe] seen for the two systems (the MW halo and the UFDs) represented in Figure 1. In our model, we assume that: i) the abundance ratios in each observed star represents enrichment from a previous enriching stellar generation (ESG); (ii) the stars within each ESG are sampled from a "normal" (Salpeter) IMF and produce enrichment with a power-law MDY; and iii) the stellar abundance ratio distributions for each system are the signature of enrichment from an ensemble of ESGs of a characteristic mass, M ESG . Note that our simple model assumes that enrichment from Pop III, metalfree stars with peculiar yields does not have a significant effect on abundance ratio patterns at the metallicities observed in UFDs.
Enriching stellar generation
Each ESG represents the combined enrichment by stars of total mass M ESG that could be formed in one or many different star clusters. Each ESG realization results from a Monte-Carlo sampling of a Salpeter (1955) and α = 2.35. We assume that the lower and upper stellar mass limit for the IMF are m low = 0.08M ⊙ and m upp = 40, respectively. (In Appendix A, a range of upper stellar mass limits, m upp = 30 − 80M ⊙ , are explored.) The lower threshold for stars contributing to chemical enrichment is taken to be m enrich,low = 8M ⊙ . The number of draws from the IMF is determined by the total and, subsequently, the remaining mass available to form a ESG of ∼ M ESG . Since this sequence of draws terminates when the total mass drawn exceeds M ESG , the actual ESG created only approximates the designated mass.
2.2. Stellar enrichment Each ESG realization produces a total mass yield for each element X by summation over all individual yields m X generated from stars of masses m ≥ 8M ⊙ . These yields are determined by a power law of index κ X and normalization β X : In our models, we are assuming that the sources of enrichment are the same in both UFD and MW halo stars. Currently, our models only take into account stellar enrichment from massive, short-lived stars which are thought to be the dominant source of enrichment for the VMP populations in both systems. Enrichment by longlived, low mass stars (excluding binaries) is assumed to become important only at higher metallicities. Although Ba is an archetypical s-process element at higher metallicities, the trace amounts of Ba observed in the VMP stars we are modeling are produced in core-collapse SNe by the r-process. There is also a large number of stars with measurable Sr abundances for the same VMP population even though Sr is primarily an s-process element thought to originate from the AGB phase in low-mass stars (known as the main s-process). Therefore, we anticipate that Sr-enrichment in the VMP population comes from a short-lived, but intense, pre-SN/super-AGB phase from massive stars, contributing weak s-process elements to the ISM prior to the SN phase (Herwig 2005); or, perhaps, is simply indicative of r-process at low metallicities (Roederer et al. 2010). Recent evidence pointing to fast rotating, massive stars as a viable source for s-process elements like Sr can be found in Chiappini et al. (2011), Frischknecht et al. (2012, and references therein. Hence, m X in our models represents a combined effective yield from both the pre-SN and SN phases of a star of mass m≥ 8M ⊙ . To construct abundance ratios, we first need to account for the common denominator -Fe abundance. The theoretical yield for Fe tabulated in Nomoto et al. (2006) varies only slightly over the range of enriching stellar masses examined -indeed, some previous studies using theoretical Fe yields of 0.05M ⊙ have assumed invariant Fe yields for SN ejecta. For consistency with other yields adopted in our models, we set yield parameters for Fe by fitting a power law to the Nomoto et al. (2006) predictions to find β Fe = 0.0607 M ⊙ and κ Fe = 0.072. Figure 2 shows our fits to the Nomoto et al. (2006) theoretical yields at Z = 0.001 (≃ Z ⊙ /18) for Fe along with fits to archetypical α-element MDYs. Also shown is our fit for Ti, which we chose as our known theoretical MDY because it exhibits the lowest scatter around a power-law fit and has a weak MDY. 6 For Ti, the power law fit yields an index of κ T i = 0.937. The yield normalization β T i is adjusted to maintain agreement between the average abundance ratio calculated for our assumed IMF, Ti Fe IMF (βT i, κT i), and average observed abundance ratio, Ti Fe OBS , calculated for our VMP ([Fe/H] < -2.5) MW halo sample (see Appendix B). This adjustment is made to compensate for the failure of the Nomoto models to get the amount of "fallback" for Ti correct in their SN explosions (for an explanation, see Figure 12 and §8.2 in their paper).
For Sr and Ba, which have no firm yield predictions, we examine a range in κ X (−20 ≤ κ X ≤ 20) which is wide enough to reveal the relative effects of stochastic sampling in ESGs of different mass (see §4.3-4.4) and allows for a comparison to some proposed yields in the literature (see §5.1). For each κ X a β X is derived by again requiring a match to the observed average in the MW halo sample, assumed to arise from the fully-sampled IMF.
Finally, it should be noted that while we do track the production of Fe in each ESG realization, we do not explicitly follow evolution in [Fe/H] since the latter is not critical to the scope of this project and would require more detailed assumptions regarding star formation efficiency, mixing, infall, and blowout.
Parent distributions and synthetic "Child" samples
Following the prescription given in §2.2, each ESG realization from §2.1 produces a chemical abundance ratio for [X/Fe] (where X represents Ti, Ba or Sr), which is supposed to represent a possible enrichment pattern for a subset of the total population of stars that exist in the observed systems. Thus, each ESG produces one was arbitrary. However, the difference in MDYs derived for Z = 0.001 (κ Fe = 0.072 and κ T i = 0.937) versus Z = 0 (κ Fe = 0.086 and κ T i = 1.130) are not significant to this study and use of either set of yields would lead to the same overall results. enrichment pattern from which many stars can sample. However, the numbers are proportional to how common that enrichment pattern is (as determined by the distribution of patterns from the ESGs generated). For a given set of parameters (M ESG , κ X ) we construct twodimensional "parent distributions" in the [Sr/Fe]-[Ti/Fe] and [Ba/Fe]-[Ti/Fe] planes from ensembles of enrichment by 1000 ESGs. Each parent represents a model for the intrinsic stellar distribution from which we can draw random synthetic samples ("children") to compare to the MW halo and UFD observed data distributions. Each child contains the same number of synthetic stars as the number of observed stars and their stellar abundance ratios are scattered by observational errors which are taken to be 0.15 dex (as a conservative lower bound).
RESULTS I: GENERAL EFFECTS
In this section we develop some intuition by examining the effect of varying parameters (M ESG , κ X ) on the shape of the abundance ratio distribution in [X/Fe] in one dimension. Figure 3 illustrates schematically the trends we expect to see in our distributions resulting from the combination of the IMF, the MDY(κ X ), and the number of enriching stars, n ⋆ , generated in a ESG (which is proportional, on average, to M ESG ).
Phenomenological expectations
In panel A, the Salpeter IMF is shown, illustrating that many more lower mass stars are produced for a given number of high mass stars in any ESG. This property is generic to all proposed IMFs in nearby galactic environments investigated in the literature (Kroupa 2002;Chabrier 2003;Elmegreen & Scalo 2005;Elmegreen 2006Elmegreen , 2007. In panel B, the MDY for various κ X are shown: an approximately constant mass yield across all stellar masses (κ X ≃ 0), a small/weak change in mass yield (low κ X values), and a large/strong change in mass yield (high κ X values). It should be noted that these power law fits are a rough 1 st -order approximation to the non-monotonic functions for MDYs anticipated in nucleosynthetic yield models (e.g. Nomoto et al. 2006;Heger & Woosley 2010) for both Ti and nc-elements. The detailed shape of these functions will be another key factor which contributes to the range and shape of observed abundance ratios, but is not considered in this paper to keep our models as simple as possible (and because the mass-dependence of stellar yields for most elements is not well understood at present).
In panel C, trends in the distribution of yields from an ensemble of enriching ESGs as a result of combining the IMF with MDY (IMF⊗MDY) are shown for different numbers of enriching stars per ESG, n ⋆ .
In the limit of n ⋆ = ∞ (right hand plot of panel C) complete sampling of the IMF is achieved, resulting in a single mean value for all realizations.
In the opposite limit of n ⋆ =1 (left-hand plot of panel C), we expect to directly sample the full range of yields contributed from individual stars, with frequencies dictated by the IMF. Hence a strong MDY (high κ X ; solid line/light-shaded area) will produce a wide distribution while a weak MDY (low κ X ; dotted line/dark-shaded area) will produce a narrow one. For positive κ X , the skew of these distributions will be positive or rightskewed, meaning that their extended tails are found to the right of the median and peaks are found to the left. In the case of negative κ X (not shown), the skew of the distributions will become negative, with the extended tail to the left of the median. A wide range of distributions can be observed between these two limits. For an element X with large, positive κ X (solid lines and light-gray areas in Figure 3), various distributions can be exhibited depending on the value of n ⋆ .
For example, with n ⋆ ="few", the convolution of yields with the IMF from a few enrichers can generate negatively-skewed (left-skewed) distributions. 7 Although massive enrichers are found less frequently than their lower mass counterparts, their individual chemical yields can dwarf those contributed by lower mass stars. Hence, the orientation of the tail of the distribution can flip compared to the n ⋆ = 1 case due to the weighted contribution of the "few" high mass enrichers with large absolute yields.
For n ⋆ ="many", the average number of n ⋆ realized in each ESG is high enough to start altering the distribution from a poisson-like distribution to a gaussian-like distribution via the law of large numbers. This effect arises from a counter-balance between the plentiful, although low impact, low-mass enrichers and the sparse, yet high impact, high-mass enrichers which leads to an "erosion" of possible abundance ratios at the margins of the distribution (homogenization), thus narrowing the distribution in accordance with the central limit theorem.
Model distributions
We can assess the validity of our phenomenological expectations, given in §3.1, by examining ensembles of many ESGs realized with identical parameters, to create chemical abundance ratio probability distributions. The features of interest are systematic changes in the: 1) variance (dispersion), 2) skewness (lopsidedness), and 3) kurtosis (peakedness) of the distribution. As noted in §2.2, the "means" of our distributions are set by the observed average abundance ratio but these higher moments emerge from the parameters specified for κ X and M ESG . Figures 4 and 5 illustrate the general trends found for various parameters (κ X , M ESG ). Figure 4 shows a number of features in these distributions that are similar to both our schematic framework and the observed distributions. Each panel corresponds to a different decade in M ESG (= 10 2 , 10 3 , 10 4 M ⊙ respectively) realized 1000 times to create distributions with average number of enriching stars given by <n ⋆ > ≃ 1, ≃ 7, and ≃ 65, analogous to the one, "few", and "many" enrichers in the schematic in Figure 3. Comparing the different colored histograms within each panel, increasing the value of κ X = 3 (red), 6 (green) and 9 (blue) leads to a broadening of the distribution. The black vertical dotted line shows the average for the distributions which correspond to yields from all ESGs generated with κ X = 0. Com- paring the same colored histograms across the panels, shows stochastic effects producing right-skewed distributions for n ⋆ ∼ 1 (left panel), left-skewed distributions for n ⋆ ∼ "few" (middle panel) and gaussian-like distributions for n ⋆ ∼ "many" (right panel). Figure 5 shows the general trends for negative MDYs. A comparison between Figure 4 and Figure 5 demonstrates that distributions derived with positive MDYs for X cause negatively or positively skewed distributions while negative MDYs only lead to negatively skewed distributions.
Most importantly, if we compare Figure 4 to the observations ("by eye") we see that the simple stochastic picture can be supported if the following criteria are met: • Abundance ratios in MW halo stars reflect enrichment by ESG with masses sufficient to produce "few"-to-"many" enrichers, while abundance ratios in UFD stars reflect enrichment by ESG with masses that would produce roughly "one" enricher; • Ti is well approximated by low κ X , resulting in a similar distribution for any n ⋆ (or M ESG ); • nc-element yields are well approximated by high |κ X |, resulting in noticeably different distributions for low versus high n ⋆ (or M ESG ).
MODEL PARAMETERS
Having demonstrated in principle that skewed abundance ratio distributions can be obtained when incomplete sampling of the IMF is coupled with strong MDYs we now assess whether this explanation is sufficient to explain the current observed samples.
Selecting a comparison sample
As stated earlier, our models were designed to track stellar abundance ratios that originate from the evolution of high-mass stars that are not in binary systems (i.e., the combined yields from a super-AGB/pre-SN phase and/or post-SN wind). Hence we select our sample from the Frebel (2010) compilation to exclude stars whose abundance ratios are likely to include enrichment from other sources.
Specifically, abundance ratio contributions from lowmass stars (e.g. AGB winds or Type Ia SNe) are limited by looking at VMP stars (with [Fe/H] < -2.5 in our case): because of their low-metallicity, VMP stars are assumed to have formed before long-lived low-mass stars had a chance to contribute significantly to chemical enrichment Vargas et al. (2013).
In addition, we use Figure 6 to exclude stars whose abundance patterns could reflect enrichment during binary evolution by identifying those stars that fall within the abundance ratio boundaries of [Ba/Fe] > 1.0 dex and [Ba/Eu] > 0.5 dex (indicated by the grey rectangular region, from the diagnostic prescription listed in the review by Beers & Christlieb 2005).
These "Barium stars" are thought to be produced during binary evolution from s-process-enhanced Barium enrichment in the common envelope (Smith & Lambert 1990;McClure & Woodsworth 1990) or wind accretion (Boffin & Jorissen 1988) phases. The validity of this simple diagnostic is confirmed by the locations of the stars highlighted in red and blue which indicate where Frebel (2010), using a more detailed abundance ratio analysis, designated stars as enriched by both r+s-process (in red) or by r-process (either class I or II) alone (in blue). For those stars with no Eu detection (vertical grey stripe) we also exclude those stars with [Ba/Fe] > 1.0 dex since a non-detection for Eu ensures that [Ba/Eu] >> 0.5 dex.
4.2.
Comparing data and models with a "paternal likelihood test" To directly compare our models to observations, we construct a test to determine the likelihood that the observed stellar abundance ratio samples for the MW halo or UFDs, shown in Figures 7 and 8, could be drawn from the 2-d parent distributions generated by a particular parameter set (see §2.3). Our "paternallikelihood test," is built around the comparison of our samples to each parent using the D-statistic derived from Black triangles are MW halo stars without given r-or s-process abundance ratio designations. Blue squares refer to stars with r-process abundance ratios (either class I or II) and red diamonds refer to r+s-process stars as designated by Frebel (2010). The pure r-process upper-limit, designated Barium stars (exclusion) region, and stars with nondetections of Europium and identified by a dashed grey line, grey rectangular region, and dark grey stripe, respectively (see text for explanation). the two-dimensional Kolmogorov-Smirnov (2dKS) test (Press 1992). The D-statistic represents the maximum difference (supremum) between two cumulative distribution functions (CDFs) -a smaller supremum indicates a higher likelihood that both CDFs are drawn from the same population.
While the values of the D-statistic can be used to rank our parameter sets given the observed data, the 2dKS test alone is insufficient for our purposes. The multitude of possible data orderings used to create CDFs in multi-dimensional samples (Peacock 1983;Fasano & Franceschini 1987) means that the D-statistic cannot be simply converted to a likelihood in a modelindependent manner. This problem is particularly challenging given the small number of stars (6) used in the UFD samples where large differences in D-statistics be- tween parameter sets may not actually represent significantly different likelihoods. Our paternal-likelihood test addresses this limitation by generating child-parent distances (D cp ) for a large number of synthetic child samples (with sample sizes equaling the observed data size) drawn (bootstrapped) from the parent. The distribution of D cp can then be used to assess the likelihood of observing the distance D dp between the collected data samples and the parent.
Specifically, we generate n children = 100 from each parent (defined by parameters M ESG , κ X , m upp ). Each child is comprised of n randomly-sampled stellar abundance ratios from the parent distribution where n equals the number of observed stars from the observed comparison data sample. Figure 9, for example, shows a distribution of D-statistic ranks calculated for the [Ti/Fe]-[Ba/Fe]plane using children drawn from one of our parent distributions to assess parental likelihood for the MW halo (upper panel) and UFDs (lower panel), respectively. The spreads in the distributions are influenced by both the observational/systematic errors and the sample size. As expected, a larger sample of stellar abundance ratios increases our certainty about the likely parent of the observed distribution.
We assess the significance of the comparison rankings between the observational data and the parent, D dp (indicated by vertical dashed line in Figure 9) by calculating a p-value -i.e. the fraction of children that are ranked as more different from the parent than the observed data (shown as the fraction of the histogram that lies to the right of the vertical line in Figure 9): The higher the p-value, the more likely the observed abundance ratios are a potential "offspring" of the parent. The color of the plot indicates the likelihood (i.e. the pvalue) of the observations being drawn from a parent of particular M ESG and κ X , and for a fixed M upp = 40M ⊙ 8 . From the upper panel it is immediately apparent that models with M ESG 10 3 M ⊙ are preferred in generating MW halo-like distributions. Furthermore, these models are consistent with a wide range of |κ X | 2 values due to the degeneracy between stochastic sampling of the IMF (governed by M ESG ) and the effect of varying the strength of the MDY: the IMF is more completely sampled as M ESG gets larger which will tend to homogenize the stellar abundance ratios, but this effect can be compensated for with a higher MDY strength in order to maintain a sufficient width to match the MW halo distribution.
Differences between the location and width of the trends apparent in the upper panel for ±κ Sr can be attributed to the relative weighting of low/high-mass enrichers in each case. Since there are significantly more low-mass enrichers than high-mass enrichers generated for M ESG a few hundred solar masses, homogenization is reached sooner for negative κ X (i.e. at a lower ESG mass) than for ESGs with a positive κ X . Also, the 8 We find that some spurious likelihoods can arise from models that have sample dispersions of ∼0.3 dex or less (i.e., on par with the observational or systematic errors). These artifacts are caused by a limitation in the way the 2dKS test handles models with a relative dearth of data sampled in the wings of its distribution (see Babu & Feigelson (2006); Babu & Rao (2004); Stephens (1974) for an explanation). Models with intrinsic dispersions of ≃ 0 are emblematic of this limitation. Fortunately, such models can be trivially identified (by their aforementioned dispersions) to be incompatible with the observed data and are therefore recorded with likelihoods of less than 5%. smaller width of the probability distribution for κ X < 0 reflects the diminished contributions of high mass stars because they are (in this scenario) both rare and have yields that are small relative to their less massive counterparts, thus shrinking the range of M ESG capable of producing the observed MW halo distribution.
The lower panel displays the results of the same analysis for the six stars in the UFD sample. The two regions of significant likelihood are analogs to the negative and positive κ X trends found for the MW halo, but the paucity of observed stars in the UFD sample means that a much broader set of models are compatible with the observed chemical distributions. Therefore, we see models with substantial likelihoods (p-values) across more than two decades in M ESG for a variety of κ X values. Despite the breadth of possible solutions found in each panel, they demonstrate (as a whole) that our simple model of stochastic enrichment is sufficient to explain the Ti and Sr abundance ratio distributions in the MW halo and UFDs simultaneously, provided that: (i) the UFD systems were enriched by a lower ESG mass than the progenitors to the MW halo stars; and (ii) Sr yields can be characterized by a power law with a relatively larger |κ X | when compared to Ti yields. plane: that the same simple model of stochastic enrichment with the same masses for MW halo and UFD enrichers preferred can also explain the distributions in this plane. The UFD results here suggest a slightly lower κ X for the MDY of Ba compared to Sr. Also, in the case of Ba, a negative MDY seems highly unlikely from our analysis. This result can be explained by comparing the UFD distributions from Figures 7 and 8 to the M ESG = 10 2 M ⊙ models from (Figures 7 and 8) reveals that [Sr/Fe] values are significantly more similar to the negative κ X for M ESG = 10 2 M ⊙ models than the [Ba/Fe] values. However, it should be noted that we rule out the existence of a negative κ Ba based on the MW halo data as the current UFD data are inconclusive on their own. In the next section, our results for "allowed" MDY strengths are compared with the most recent yields found the literature.
DISCUSSION
In this section, we evaluate how our model-derived MDY strengths compare to others found in the literature. We also examine how our selection of data affects our reported results.
Comparison to Other MDY Estimates
In Table 1 we compare our derived MDY strengths to the latest predictions given in the literature. In particular, we compare our values to those extracted from ab initio yields (i.e. yields derived from simulations) for Sr given in Frischknecht (2012, PhD Thesis) and from inferred values from the ab initio-and empirically-derived yields (i.e. chosen to match observations) for Sr and Ba applied in Cescutti & Chiappini (2013).
• Empirical Yields for Sr and Ba (8 − 10M ⊙ production site) In Cescutti & Chiappini's work, their homogenous stochastic models are chosen to fit the general distribution of halo stars without binary enrichment. These models, which they refer to as empirical models, are employed by the authors to examine the distributions produced by applying both their empirically-determined MDYs for the standard r-(and extended r-)process sites and the newly derived ab initio yields from Frischknecht's thesis work. To generate MDY strengths for their empirical yields, we consult the -3 ∼ -15 or -18 ∼ 4.5/6.6 ∼ 7.4/-( −10), ( 7) -5 -∼ 3.6/3.6 Barium (Ba) -3 ∼ -15 -∼ 3.9/-∼ (6 − 12) a Derived from empirical yields given in Cescutti (2012). b Derived from Figure 4.14 of Frischknecht (2012, PhD Thesis) for non-rotating (nr)/rotating stars (rs). Yields for Ba were not given. c Derived from Cescutti & Chiappini (2013) for rotating stars (rs) [their as-models]/spinstars (ss) [their fs-models]. † Chieffi & Limongi (2004) and Limongi & Chieffi (2012) provide another set of theoretical MDYs for Sr. From Chieffi & Limongi (2004) we find that the estimated MDYs for Sr given for progenitors with z > 0 to z ≃ z ⊙ results in strengths that are 1 κ Sr 4. The MDY for Sr for zero metallicity stars is κ Sr ≃ 8 -compatible with our work. However, more recent work by the same authors (Limongi & Chieffi 2012) produces a κ Sr 5 for zero metallicity stars. This result is only marginally compatible with our findings.
figure of Sr and Ba yields given in Cescutti (2012) which are reported to be similar to the yields used in Cescutti & Chiappini (2013).
• ab initio Yields for Sr (15 − 40M ⊙ production site) In Frischknecht's work, he conducts a suite of simulations that produce various chemical yields from massive stars as a function of the stars' metallicity and rotation. From his work, we approximate ab initio strengths (κ ab initio ) for 88 Sr 9 by examining Figure 4.14 of Frischknecht (2012, PhD Thesis). Unfortunately, we are unable to make a direct comparison to MDYs strengths for Ba (which are also evaluated by Frischknecht) because they are not available in his published work.
• Inferred Yields for Sr and Ba (15−40M ⊙ production site) We also generate an estimate of the MDYs for Sr, and more importantly, for Ba (unreported) from Frischknecht's unpublished results.
To do this, we input the various inferred ∆[ X F e ], displayed in Figure 1 of Cescutti & Chiappini (2013), along with their progenitor stellar mass range into the difference between logarithmic values of Eqn. 2. If we assume that Fe-yields for these stars are weakly mass dependent, we get: The estimates for the inferred MDYs strengths derived from Eqn. 5 are also listed in Table 1.
The final column of Table 1 gives our preferred MDY strengths, which are chosen by identifying ranges of κ X that could be simultaneously compatible for BOTH the MW and UFD's (i.e. looking at both upper and lower panels) As seen in Figure 10, both positive and negative MDY strengths for Sr are allowed. In particular, both a κ Sr 7, consistent with Frischknecht's 15 − 40M ⊙ ab initio yields and a κ Sr −14, consistent with Cescutti's 8 − 10M ⊙ empirically-derived (standard r) yields, are favored for Sr. Additionally, the inferred κ Sr from a combination of such yields should, in fact, be intrinsic to our analysis -however, inferences about combined yields are beyond the scope of this investigation and shall be addressed in future work. Figure 11 shows us results for Ba yields. Positive MDYs with κ Ba ∼ 6 − 12 are preferred and may be related to Frischknecht's spinstar yields. However, the extremely low likelihoods for negative κ Ba when compared to positive κ Ba , supports the notion that such yields are improbable. This strongly suggests a lack of Ba production from an ∼ 8 − 10M ⊙ production site which is consistent with more recent hydrodynamic simulations (e.g., Fischer et al. 2010;Wanajo et al. 2011) but contrary to other expectations for nc-yields found in the literature (see, e.g., Cescutti & Chiappini 2013;Cescutti 2012;Qian & Wasserburg 2008;Wanajo et al. 2003;Ishimaru & Wanajo 1999;Wheeler et al. 1998;Mathews et al. 1992).
These preliminary results illustrate the advantage of using statistical techniques that address the full density of the observed distributions and not only the average of their spreads as implemented in Cescutti & Chiappini (2013) and other previous studies. Further development of this technique may provide the best chance to uncover the "galactic genealogy" of the MW and its closest companions in the Local Group.
Data Compilations
One concern about using a compilation of data such as Frebel (2010) is that the dispersion in abundances may be artificially inflated by differences between the data sets. Frebel states that systematic difference between data sets are likely to inflate the dispersion by no more than 0.3 dex (for both the UFDs and MW halo). In particular, the dispersion between different measurements of Ti abundance (i.e., via Ti i/Ti ii or a combination thereof) is typically, ∼0.1-0.15 dex (e.g., Shetrone et al. 2003;Aoki et al. 2007;Frebel et al. 2010) which is pre-cisely on par with observational errors. In contrast, both the offset (under-abundance) and scatter (dispersion) of nc-elements are a factor of ∼3-5 and ∼10 bigger than these systematic uncertainties, respectively. Thus, we conclude that the differences between surveys cannot significantly alter our current results. Moreover, artificially inflated dispersions for UFD and MW halo distributions would serve to decrease their expected M ESG while leaving a significant ∆M ESG between the distributions intact. Hence our result of lower M ESG for UFDs verses significantly higher M ESG for MW halo progenitors is insensitive to these systematic differences.
Ignoring Data with Upper Limits
Our parental likelihood test is not strictly applicable to samples containing upper limits. However, as a check, we apply the test to the Frebel data compilation, including upper limits, to determine the possible effects, if any, of leaving data with upper limits out of our analysis. Including the upper limits also increases the scatter of our MW halo samples, which, again, effectively decreases the inferred M ESG slightly while, in this case, increasing the inferred MDYs. These values are not significantly different from the values we report. The similarity of the results from the two samples is compatible with the fact all stars with only upper limits for Sr and Ba are consistent with having [Sr/Fe] and [Ba/Fe] abundance ratios above those stars with the lowest known levels of nc-elements Roederer (2013); which is to say that stars with upper limits would actually be detected if higher signal-to-noise spectra were available. Hence, star with upper limits are consistent with residing in, not below, the distributions of detected stars. The insensitivity (or compatibility) of the models to the exclusion or inclusion of data with upper limits proves that our work is sufficient for the purposes of broadly testing whether our simple scenario for chemical enrichment of UFDs in comparison to MW progenitors is plausible. Once the observed data sets for UFDs are larger a more rigorous statistical approach will be required to actually place strong limits on -for example -the detailed nature of MDY for nc-elements.
CONCLUSION
While the distribution of [Ti/Fe] is similar in both the MW halo and UFDs, the means/medians of nc abundance ratios for VMP stars found in these two systems are significantly offset. Although the current UFD sample is still small, this discrepancy motivates questions concerning the nature of hierarchical merging in the construction of the MW halo. In particular, discrepant abundance ratios suggest that past accreted dwarfs galaxies (i.e. progenitors of the stellar halo) may have been quite unlike the progenitors of the current MW satellites. Possible solutions include appealing to inhomogeneous chemical mixing, differential blowout of metals from SN winds, differences in primordial abundance ratios due to population III stars or differences in the IMF of stars within the progenitor systems.
In this paper, we explore an entirely different possibility for these discrepant abundance ratios: that progenitors of MW halo were enriched by a larger prior generation of stars when compared to UFD progenitors (as could be the case if, for example, UFD progenitors were more isolated than the MW progenitors, as suggested in Corlies et al. (2013). We demonstrate that this simple hypothesis can qualitatively and quantitively explain both the similarities of Ti distributions and differences between the nc-distributions for the current observed samples provided that the nc-elements have much stronger MDYs (currently unknown) than the (known) MDYs for Ti. Specifically, a viable model that simultaneously fits the distributions of [Ti/Fe], [Sr/Fe] and [Ba/Fe] is one in which MW progenitors were enriched by prior stellar generations of mass M ESG 10 3 M ⊙ and UFD progenitors were enriched by M ESG 10 2 M ⊙ . The most likely MDY strengths (given the data used and the simplicity of our models) are characterized by a power law index of |κ Sr | ∼ 7 − 14 for Sr and κ Ba ∼ 6 − 12 for Ba with lowest plausible values of |κ Sr,Ba | 4 (compared to κ Ti ∼ 1). These numbers were derived from enriching stars sampled from for a Salpeter IMF with an upper limit of 40 M ⊙ (We show in Appendix A that a different m upp leads to a similar explanation, though with different numbers for M ESG and κ X ). 10 In this study we have demonstrated that our simple approach can explain the current data. However, it is known that many other effects can influence abundance ratio distributions in these systems, and, that ultimately, the relative importance of each effect needs to be assessed by building a more complete model. We see the current work as a foundation for more complete models in the future.
Despite the simplicity of our models, there are a number of interesting implications from our results. First, a relatively modest increase in the number of highresolution spectra in UFDs could be used to test the specifics of our model -if our interpretation is correct (barring other effects), then we should find UFD members with abundance ratios skewed above the bulk of the MW distribution as well as below. Figure 12 illustrates the likelihood of finding at least one UFD star with positive values of either [Sr/Fe] (upper panel) or [Ba/Fe] (lower panel) for different sample sizes. The gray region indicates the full range of probabilities for all parameter sets for which we found p-values ≥ 0.05 when compared to the current data sets in Sections 4.3 and 4.4. These probabilities were calculated from the parent distribution for each qualifying parameter set by finding the fraction of realizations, f , that had positive abundance ratios, and then adopting f in the binomial theorem to estimate the probability of drawing at least one such star for sample size N obs . The solid and dotted line indicates the median and 25 th /75 th -percentiles for all qualifying parameters sets at a given N obs . Overall, the figure indicates that, if our hypothesis of nc-abundance ratio distributions being skewed by strongly mass-dependent, power-law-like yields is a predominant effect, then sample sizes of ∼ 15 -25 VMP stars 11 in UFDs should start to contain some nc-rich ([Ba,Sr/Fe] 0) counterparts to the nc-poor ([Ba,Sr/Fe] < 0) populations observed so far. Efforts made to extend stellar abundance ratio sam-ples into the main sequence of UFDs (e.g. Vargas et al. 2013) should eventually provided samples large enough to determine whether stochastic sampling plays a predominant role in observed abundance ratio distributions. Figure 12. Gray region indicates the range in the probability that an observed UFD sample of size N obs could contain one star with [Sr/Fe] > 0 (top panel) or [Ba/Fe] > 0 (bottom panel) for parameter sets that had p-values greater than 0.05 (see lower panels of Figures 10 and 11). The solid and dotted lines indicated the median and 25 th /75 th -percentiles for these parameter sets, respectively.
Second, it should also be noted that as sample sizes increase, the likelihood distributions in our parameter space will become more concentrated, providing stronger constraints on the form of MDY for nc-elements, and, by extension, their origin. A preliminary comparison of our current results with predictions for MDY in literature already suggests that while production of Sr from 8−10M ⊙ stars is quite possible, production of Ba from these stars is highly unlikely. Our results also support the viability of recent ab initio yields for 15 − 40M ⊙ stars.
In conclusion, our results indicate that abundance ratio distributions in nearby systems contain intriguing signatures of their early isolation (or conversely, contamination): more/less isolated systems should be enriched by smaller/larger prior enriching generations (i.e. to have lower/higher M ESG ). These signatures could potentially be exploited to probe the progress of metal enrichment on MW scales in the early Universe -a local window on a regime that cannot be seen directly.
KVJ thanks the Observatories of the Carnegie Institution of Washington for their hospitality during a visit which provided the inspiration for this work, and Andrew McWilliam and Ian Thompson for conversations on this topic in particular. DL would also like to acknowledge the encouraging and insightful conversations he had with Ian Roederer, Andrew McWilliam, Alan Dressler, Chris Sneden, George Preston, Luis Vargas, Tim Beers, Anna Frebel, and Volker Bromm. Finally, we would like to thank the referee for comments which helped to broaden the scope of our work to compare our mass-dependent yields with those found in recent literature. DL was supported by NSF grants AST-0806558 and AST-1107373. Figure A1. Distributions of abundance ratios produced from 1000 realizations of an ESG, with M ESG = 10 2 M ⊙ (top row), 10 3 M ⊙ (middle row), and 10 4 M ⊙ (bottom row). Each column represents models generated with different mupp for the IMF: 40M ⊙ (first column), 60M ⊙ (second column), and 80M ⊙ (third column). Colors are the same as found in Figure 4 of the paper. | 2013-07-10T19:59:58.000Z | 2013-07-10T00:00:00.000 | {
"year": 2013,
"sha1": "5d4abd8d575c812ca941b8b388824f451ab33850",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1307.2894",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "5d4abd8d575c812ca941b8b388824f451ab33850",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
206818567 | pes2o/s2orc | v3-fos-license | Follow-up tricuspid annular plane systolic excursion predicts survival in pulmonary arterial hypertension
Few studies have examined the utility of serial echocardiography in the evaluation, management, and prognosis of patients with pulmonary arterial hypertension (PAH). Therefore, we sought to evaluate the prognostic significance of follow-up tricuspid annular plane systolic excursion (TAPSE) in PAH. We prospectively studied 70 consecutive patients with PAH who underwent baseline right heart catheterization (RHC) and transthoracic echocardiogram, who survived to follow-up echocardiogram after initiation of PAH therapy. Baseline TAPSE was 1.6 ± 0.5 cm which increased to 2.0 ± 0.4 cm on follow-up (P < 0.0001). The cohort was dichotomized by TAPSE at one-year follow-up: Group 1 (n = 37): follow-up TAPSE ≥ 2 cm; Group 2 (n = 33): follow-up TAPSE < 2 cm. Group 1 participants were significantly more likely to reach WHO functional class I–II status and achieve a higher six-minute walk distance on follow-up. Of the 68 patients who survived more than one year, 18 died (26.5%) over a median follow-up of 941 days (range, 3–2311 days), with significantly higher mortality in Group 2 versus Group 1 (41.9% vs. 13.5%; P = 0.003). While baseline TAPSE stratified at 2 cm did not predict survival in this cohort, TAPSE ≥ 2 cm at follow-up strongly predicted survival in bivariable models (hazard ratio, 0.21; 95% confidence interval, 0.08–0.60). In conclusion, follow-up TAPSE ≥ 2 cm is a prognostic marker and potential treatment target in a PAH population.
Hemodynamics
All patients underwent standard hemodynamic assessment by RHC including mean (Endexpiratory) right atrial pressure (RAP), mean pulmonary arterial pressure (mPAP), pulmonary arterial wedge pressure (PAWP), as well as cardiac output (CO) via the Fick method and expressed as L/min. Mixed venous oxygen saturation (MVO2) was determined while patients were breathing room air. Heart rate and noninvasive blood pressure were recorded during the procedure. Trans-pulmonary gradient (TPG) was calculated as mPAP-PAWP and pulmonary vascular resistance (PVR) was calculated from these measurements as TPG/CO and expressed in Wood Units (WU). Stroke volume (SV; CO/HR), cardiac index (CI; CO/body surface area [BSA]), stroke volume index (SVI; CI/HR) were calculated. Repeat hemodynamic data were available in 35 subjects.
Echocardiography
All echocardiograms were performed using either the Philips IE33 (Philips Healthcare, Andover, MA), or the GE Vingmed Vivid 7 Ultrasound (GE, Vingmed Ultrasound, Horten, Norway) ultrasound platforms. The systolic eccentricity index was obtained from the parasternal short axis view, and calculated as the ratio of the minor axis parallel to the interventricular septum (D1) to the minor axis perpendicular to the interventricular septum (D2) at end-systole (D1/D2). Right ventricular outflow tract velocity time integral (RVOT VTI), acceleration time (AcT) and systolic notching pattern was obtained from the RVOT pulse wave Doppler profile in the parasternal short or long axis views, as previously described. [4][5][6] The maximal trans-tricuspid flow velocity was obtained in the usual manner and used to quantify the RV-PA pressure gradient using the modified Bernoulli equation (4v 2 ). 7 Tricuspid regurgitation (TR) severity was assessed semi-quantitatively (graded 0-3). Diastolic function, was assessed by trans-mitral Doppler velocity and tissue Doppler using standard techniques, as previously described. 4,8 All studies were analyzed off-line using ProSolv CardioVascular Client software (Fujifilm Medical Systems, Stamford, CT). Analyses were performed by two experienced cardiologists trained in the echo-Doppler assessment described above, and blinded to the study subjects' clinical, hemodynamic and outcome status.
Data Analysis
The cohort was dichotomized by serial TAPSE value as follows: Group 1: follow-up TAPSE ≥ 2 cm and Group 2: follow-up TAPSE < 2 cm. This cutpoint was chosen based on normative data of TAPSE from prior studies 9,10 and as proposed by the current European Society of Cardiology treatment guidelines for PH. 11 Univariable and bivariable Cox proportional hazards models were constructed using TAPSE as a continuous or dichotomous variable and included variables found to be significant in univariable analyses (p value < 0.20) and variables previously shown to have prognostic significance in order to adjust for potential confounding factors. A landmark analysis was also performed to assess survival by TAPSE threshold (TAPSE ≥ 2 cm), with entry at 1 year from initial cohort enrollment, and survival time assessed from landmark entry. 12 The echocardiogram obtained closest to this landmark was used as the follow up study that was compared to the baseline echocardiogram. Variables that were found to be collinear by variance inflation factor testing in multiple linear regression were excluded from Cox multivariable analyses. The proportional hazards assumption was examined for all covariates using a continuous timevarying predictor and generalized linear regression of scaled Schoenfeld residuals on functions of time.
Intra-and inter-observer agreement was expressed by using intraclass correlation coefficients (ICCs) with 95% confidence intervals. Analyses were performed using Stata version 13.1 (StataCorp., College Station, TX), as well as GRAPHPAD Prism version 6.07 (GraphPad Software Inc; La Jolla, CA).
Supplemental figure 1A demonstrates a scatter plot analysis of follow-up 6MWD over the range of TAPSE measurements. The combination of a low follow-up 6MWD (< 400 m) and low follow-up TAPSE (< 2 cm) was associated with increased mortality whereas a normal follow-up TAPSE (≥ 2 cm) despite a low follow-up 6MWD was associated with lower mortality. More specifically, of the 27 patients with follow-up 6MWD < 400 m and TAPSE < 2 cm, 13 died (48%), compared to 3 of 18 patients (17%) with 6MWD < 400 m but a follow-up TAPSE ≥ 2 cm, yielding an odds ratio of 4.64 (95% CI 1.08-19.8) and a difference that trended toward significance (p=0.055), highlighting the importance of preserved TAPSE with regard to survival even in the setting of reduced 6MWD.
We also assessed the prognostic importance of follow-up RV size, as assessed by RV:LV ratio in addition to RV function as assessed by TAPSE. As seen in supplemental figure 1B, 14 of the 45 patients with follow-up RV:LV > 1 cm (31%) died while 5 of the 24 patients with RV:LV ≤ 1 cm died (21%; p=ns). However, when assessing those with follow-up RV:LV > 1 cm by TAPSE group, more patients died who had a follow-up TAPSE < 2 cm as compared to those with a TAPSE ≥ 2 cm (11/25 patients (44%) vs. 3/20 (15%), respectively, p=0.05). Thus, the presence of an enlarged and dysfunctional RV on repeat assessment was associated with higher mortality, as compared to those with an enlarged RV and normal function, with the lowest mortality (2 out of 17 subjects, 12%) noted in those with both relatively normal RV size and normal RV function. | 2018-04-03T04:17:03.080Z | 2017-03-01T00:00:00.000 | {
"year": 2017,
"sha1": "8a2a1db3010a8301ae44eaeca304a4fbebcd1026",
"oa_license": "CCBYNC",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/2045893217694175",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "420181efbdf21501d08afae0650f3439fa442aa0",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
248801652 | pes2o/s2orc | v3-fos-license | Investigating the effect of Mindfulness-Based Stress Reduction on stress level and brain activity of college students
Financial constraints usually hinder students, especially those in low-middle income countries (LMICs), from seeking mental health interventions. Hence, it is necessary to identify effective, affordable and sustainable counter-stress measures for college students in the LMICs context. This study examines the sustained effects of mindfulness practice on the psychological outcomes and brain activity of students, especially when they are exposed to stressful situations. Here, we combined psychological and electrophysiological methods (EEG) to investigate the sustained effects of an 8-week-long standardized Mindfulness-Based Stress Reduction (MBSR) intervention on the brain activity of college students. We found that the Test group showed a decrease in negative emotional states after the intervention, compared to the no statistically significant result of the Control group, as indicated by the Perceived Stress Scale (PSS) (33% reduction in the negative score) and Depression, Anxiety, Stress Scale (DASS-42) scores (nearly 40% reduction of three subscale scores). Spectral analysis of EEG data showed that this intervention is longitudinally associated with increased frontal and occipital lobe alpha band power. Additionally, the increase in alpha power is more prevalent when the Test group was being stress-induced by cognitive tasks, suggesting that practicing MBSR might enhance the practitioners’ tolerance of negative emotional states. In conclusion, MBSR intervention led to a sustained reduction of negative emotional states as measured by both psychological and electrophysiological metrics, which supports the adoption of MBSR as an effective and sustainable stress-countering approach for students in LMICs.
Introduction
In low-middle income countries (LMICs), several studies reported the high prevalence of mental health issues among university students, up to 55% (Dessauvagie et al., 2021;Pham et al., 2019;Pham Tien et al., 2020). With the intensive impact of COVID-19, nearly 44.6% of students reported having experienced adverse psychological effects during the pandemic (Chinna et al., 2021). Consequently, these stressful experiences could influence brain structures, cognition, and behavior, which profoundly affect working motivation and emotional regulation Lupien et al., 2009;McEwen, 2000). Therefore, considering a coping stress strategy with high effectiveness and affordability is essential for LMICs students.
One potential stress alleviation strategy is the Mindfulness-Based Stress Reduction (MBSR), a psychological therapy that offers intensive meditation practice to guide participants to cope with acute stress and decrease distraction (Kabat-Zinn, 2003;Virgili, 2015). With the availability of its curriculum and online adaptation by the Palouse Mindfulness, MBSR represents a low-cost stress alleviation therapy (Knight et al., 2015;Miller et al., 1995). This therapy has demonstrated positive efficacy in alleviating psychological and physical conditions in different populations (Anh et al., 2020;Chi et al., 2018). For students of diverse ages, mindfulness-based therapies were shown to be reliable and effective in reducing negative emotional states (Zenner et al., 2014). None of such psychological effects have been thoroughly validated in the LMICs context, where mental health issues rapidly increase amidst the negative economic and social impacts of the COVID-19 pandemic while facing a lack of intervention programs. Hence, it is uncertain whether the adaptation of MBSR might be feasible and effective considering the social and cultural differences between developed and developing countries. Therefore, it is important to validate if MBSR could potentially serve as a low-cost and effective stress-reducing strategy for students in LMICs.
Previous studies have not fully addressed whether MBSR-associated psychological effects could be sustained and what might be the neural correlates of these effects. Concomitantly, it is currently unclear whether the overall psychological effects of MBSR could link with the practitioner's ability to control their stress and anxiety levels when exposed to immediate stress stimuli. To address these inquiries, combining psychological questionnaires with real-time brain activity monitoring techniques could allow us to further understand the effect of MBSR. Electroencephalogram (EEG) is one such technique that has been widely applied to monitor the neural oscillation of distinct brainwaves such as alpha wave. An increase in alpha band power in the prefrontal and frontal lobe during mediation or eyes closed states has supported the effectiveness of the MBSR program on stress reduction and selfawareness enhancement (Aftanas and Golocheikine, 2001;Cahn and Polich, 2006;Gao et al., 2016;Morais et al., 2021;Moynihan et al., 2013). Although most studies successfully demonstrated the immediate and short-term effect of the MBSR program, the sustained impact of this intervention on neural activity has not been shown. Most longitudinal studies reported that questionnaires and brain activity outcomes return to the baseline (pre-MBSR) values (Bennett and Dorjee, 2016;Gouda et al., 2016;McIndoo et al., 2016;Moynihan et al., 2013). Recently, utilizing electrophysiological measures (i.e., ECG, EEG, and EDA), Morais et al. has found that MBSR is associated with an increase of alpha power in the prefrontal cortex during and after the intervention, but no significant similar increase was observed at other brain areas and especially, at two months post the training course (Morais et al., 2021). Additionally, these results are hindered by the lack of a control group, thus giving less power to the conclusion that the change in prefrontal alpha power might be associated with mindfulness training. In light of these findings, it is necessary to design the EEG study to clarify the short-term and sustained effects of MBSR on brain activity of students, especially when exposed to immediate stress conditions. These understandings will certainly shed light on the mechanism of MBSR's effectiveness.
The present study aims to validate the feasibility and explore the potential sustained effect of the MBSR program in reducing the negative emotional states of a college student population in a low-middle income country (herein: Vietnam). A Control group is included to separate MBSR-associated effects from those due to subjects' increasing familiarity with the experimental design across the course of the study. We hypothesized that MBSR intervention could effectively lower subjective stress perception and induce changes in brain activity right after the intervention and in a 2-month follow-up. Furthermore, by integrating stress-induced tasks during EEG acquisition, we examine whether the MBSR program helps students retain higher alpha power, which represents stress alleviation and relaxation states, during short-term stress conditions.
Participant recruitment
Forty-nine students with an age range from 18 to 22 years old were initially recruited (35 females; mean age = 19.9 ± 0.75 years and 14 males; mean age = 20.0 ± 0.6 years). All participants had normal or corrected-to-normal vision and were neurologically healthy (Association, 2013). Additionally, the participants who scored from severe to extremely high score on the Depression Anxiety and Stress Scale (Depression > 21, Anxiety > 15, and Stress > 26) were excluded. After signing the consent form, subjects were randomly assigned into two groups: Test grouppractice MBSR (N = 25) and Control group -do not practice MBSR (N = 24). In the final measurement session (Sustained measurement), the total number of remaining subjects was 41 subjects (N = 20 for the Test group and N = 21 for the Control group). This study was approved by the Ethics Committee of the School of Biomedical Engineering, International University, Ho Chi Minh City Vietnam National University.
The intervention: MBSR program
The MBSR program includes 8-week practices strictly following the Palouse Mindfulness course (Palouse). This course was derived from the Jon Kabat-Zinn program (Kabat-Zinn, 2003) under the guidance of certified mindfulness coaches. The curriculum and training materials of the course were translated into Vietnamese to optimize the effectiveness of the training process. The course objective is to raise self-awareness of well-being and resilience for all students with negative emotional states. Participants were required to attend at least six out of eight lessons and one meditation retreat day. They attended weekly 1.5 h-long classes and a 7-h retreat day, all under instructions from the Palouse Mindfulness certified coaches.
The overall design of data collection
The experiment includes three main data collection sessions. The first data collection, called the pre-MBSR measurement, occurred right after recruitment (Fig. 1A). The post-MBSR measurement was conducted immediately after the MBSR intervention (Fig. 1A). The last one, called Sustained measurement, happened two months after the previous MBSR session (Fig. 1A). The structure of these measurements is similar for all subjects. At each session (Fig. 1B), subjects were asked to complete two questionnaire packages (See Section 2.4.) and perform two Stressinduced tasks (See Section 2.5.) while their EEG signals (See Section 2.6.) were collected. All questionnaires and stress-inducing tasks were delivered via the Omniscience platform (EMOTIV Inc., Australia). A survey was conducted at the Sustained measurement to track the continuation of meditation practice of the Test group.
Perceived Stress Scale (PSS)
The PSS includes 10 questions about the unpredictable, uncontrollable, and overloaded situations respondents encountered during the previous month. Four positively phrased items are referred to as "Perceived Coping" or "Perceived Self-Efficacy," while six negatively phrased items are referred to as "Perceived Distress." The PSS score is calculated by summing the reverse scores of the positive items and the score of negative items. Higher scores imply higher levels of perceived stress in a range of 0-40 total scores (Maroufizadeh et al., 2018;Roberti et al., 2006).
Depression, Anxiety and Stress Scale (DASS-42)
The DASS-42 measures the negative emotional states, namely depression, anxiety, and stress. With 42 items, three subscales such as "Stress," "Anxiety," and "Depression" are defined with 14 numbered questions in the form. The score for each category is summed up to be classified from 0 to 42. Five severity labels -"normal," "mild," "moderate," "severe" and "extremely severe" are used to describe the meaning of each subscales' scores (Crawford and Henry, 2003;Lovibond and Lovibond, 1996).
Both PSS and DASS-42 questionnaires were translated into the native language of the subjects, revised and finalized by a Vietnamese neuroscientist and a Vietnamese psychologist before being used. Considering that the main scope of our study is measuring the changes in negative emotions after MBSR training, the contents and length of these two questionnaires were most suitable, as both DASS-42 and PSS were confirmed to be valid and reliable in both clinical and non-clinical samples including the Vietnamese population, and have been widely used in previous MBSR studies (Crawford and Henry, 2003;Jovanović and Gavrilov-Jerković, 2015).
Cognitive tasks 2.5.1. Mental arithmetic (MA) task
Mental arithmetic tasks act as artificial stressors to trigger acute mental stress (Li-Mei Liao and Carey, 2015;Ushiyama et al., 1991). The tasks include one trial stage and three main stages (i.e., Stages A-C). Each stage includes three rounds with different difficulty levels (Fig. 2). The calculations in this task include addition, subtraction, and multiplication. The trial stage is set at the easiest level, which consists of adding or subtracting one digit, for participants to get used to the task and the keyboard. The difficulty level is designed in the order of hard-medium-easy for each stage. The calculation in the hard level applies all the mentioned operations, while the medium and easy levels require from one to two operations of addition or subtraction in one digit. After finishing one round, subjects were asked to evaluate the difficulty of the tasks and their stress level while doing the calculations.
Color Stroop task
Stroop color task is utilized as an artificial stressor (Li-Mei Liao and Carey, 2015;Svetlak et al., 2010). The task consists of three rounds; each round has a time limit of approximately 60 s (Fig. 3). There will be four different colors (yellow, red, purple, green) in this task, which were translated into Vietnamese. Each color corresponds to a single key on the keyboard, which are I-L, respectively. The participants are required to remember the color related to the key. When a word appears on the screen, the participants need to press the key corresponding to the color of that word but not the meaning of the word. The speed of displaying each word is increased after each round; therefore, the time response for one displayed word is reduced.
Electroencephalogram (EEG) recording and analysis 2.6.1. EEG and EOG signal acquisition
The Alice 5 Polysomnography recording system was utilized in conjunction with Alice Sleepware software or Sleepware G3 software (Philips, Respironics Inc., Pittsburgh, PA) to collect raw EEG signals. All recording parameters are kept the same for three measurement sessions except for the sampling rate (i.e., which is set to be 500 Hz for Premeasurement, and 200 Hz for Post-and Sustained measurements). It is worth noting that this difference in sampling rate between Pre-and Post-, Sustained measurement sessions would not have an effect on EEG analysis as all raw EEG data in Pre-measurement was down-sampled to 200 Hz prior to the analysis. We selected two channels located at the prefrontal (Fp1/Fp2) and four at the frontal (F3/F4 and F7/F8) cortices, as these areas are associated with stress-related states . Also, two temporal channels (T5/T6) and two occipital channels (O1/O2) were selected since the included stress tasks in our study were expected to mainly affect temporal and occipital cortices (Wang and Sourina, 2013). Thus, the overall EEG setup comprised 10 Ag/AgCl electrodes with the ground electrode Fpz, and these were re-referenced to mastoid electrodes (M1/M2). EOG signal was recorded from two electrodes located at the tail of the eyes to detect horizontal eye movement. The impedance of all electrodes was kept below 5 kΩ.
EEG preprocessing
The continuous raw EEG data of pre-measurement was firstly downsampled to the sampling rate of 200 Hz to ensure consistency among all three recording sessions (i.e., Pre-, Post-, and Sustained measurements). The down-sampling process was assisted with the MNE-Python package's resampling function (i.e., mne.filter.resample), which uses the Fourier method with the improvement of edge padding and frequencydomain windowing (Gramfort et al., 2013;Virtanen et al., 2020). Next, the raw EEG data were subjected to an empirical mode decomposition (EMD) based preprocessing pipeline to remove high-frequency A. An et al. noises (e.g., electromyography (EMG) induced and powerline interference noises) (Zhang et al., 2008). EMD is a data-driven signal decomposition method, of which the underlying notion of instantaneous frequency provides insights into the time-frequency feature of the signal (Huang et al., 1998). EMD decomposes the signal into intrinsic mode functions (IMFs). In this study, an enhanced version of EMD, referred to as masking EMD (Deering and Kaiser, 2005), was used to resolve the inherent mode-mixing problem in the original EMD (Deering and Kaiser, 2005;Nguyen et al., 2019;Tsai et al., 2018).
The procedure for EEG signal preprocessing using an open-source Python EMD package is described as follows (Quinn et al., 2021). Firstly, the EEG signal was decomposed sequentially by EMD with an initial masking signal of frequency at 50 Hz to a maximum of ten (N = 10) intrinsic mode functions (IMF) ( Supplementary Fig. S1A, B). Secondly, the Hilbert transform was applied to each IMF to obtain the frequency distribution of each IMF ( Supplementary Fig. S1C). Each IMF can be distinctively characterized by its frequency range and power. Thirdly, the first mode characterized by high-frequency components (which usually contain power interference and EMG noise) and the last three modes dominated by low-frequency components were removed. The signal was then reconstructed from the remaining IMFs.
EEG Power Spectral Density (PSD)
Power Spectral Density (PSD) is a widely used feature for EEG signal analysis as it can provide insights into brain activation. Since the previous studies suggested that increasing alpha and/or theta power, estimated from PSD, are often observed during meditation compared with control conditions (Aftanas and Golocheikine, 2001;Fan et al., 2014a;Takahashi et al., 2005;Tang et al., 2009). Hence, including PSD is relevant in the context of this study to evaluate the sustained effect of MBSR. The procedure for PSD calculations can be described as follows. The clean data underwent a Fast Fourier transform (FFT) with 4-s Hanning windows and 50% overlap. For each channel, spectral power (µV 2 ) for all task and rest conditions was computed for the following bands: theta (4-7 Hz), alpha (8-13 Hz), and beta (15-30 Hz). To minimize the inter-individual differences in absolute power, which can be potentially caused by technical variability of each measurement session, we normalized the spectral power of three bands by the baseline spectra similar to previous EEG studies (Bian et al., 2014;Papagiannopoulou and Lagopoulos, 2016;Sammler et al., 2007;Zhao et al., 2018). Specifically, the relative power (RP) for each channel and band was then obtained by dividing the spectral power of that band by the total spectral power between 4 and 40 Hz as in Eq. (1): where f 1 (lower bound) and f 2 (upper bound) represent the cut-off frequencies for each band of interest. For example, for the alpha band, the f 1 and f 2 are 8 and 13 Hz, respectively.
Statistical analysis
Statistical analyses were only conducted on participants who completed all data collections (final total subjects n = 41). We performed the statistical results in GraphPad Prism 8.0 software to analyze statistics in three stages in two groups. In the analyzes described below, the significant results were demonstrated when the p-value was < 0.05. We specified the star representing the p-value reported in APA style, according to: .033 (*), .002 (**), < 0.001 (***) and < 0.0001 (****) for all figures. The statistical results were plotted as the column graphs.
Two-way ANOVA with Š idák correction for multiple comparisons was applied for the psychological self-evaluation and performance results of the two stress-inducing cognitive tasks.
For EEG band power, the relative power value that differs more than two folds of the standard deviation from the group mean in each electrode in multiple frequency bands was first excluded from the analysis. Given this outliers exclusion and missing data, the linear mixed-effects model with Sidak correction was used to examine group differences (Bates et al., 2015). Linear mixed-effect models are appropriate to process longitudinal data with missing values (Magezi, 2015;Overall and Tonidandel, 2006). The Greenhouse-Geisser correction was used for violations of sphericity. It is important to note that interactions not involving groups are not reported.
We also applied linear regression in scatter plots to understand the correlation between two variables, which are Stress vs. Difficulty and PSD features vs. questionnaires. Then, the two-sample t-test is utilized to determine the differences between slopes in two subject groups.
MBSR intervention reduces stress, anxiety, and depression scores
In the previous studies, the MBSR was shown to effectively reduce the stress levels in several subject groups such as Social Anxiety Disorder patients (Faucher et al., 2016), and university students (Galante et al., 2018), and employees (Janssen et al., 2018). However, it is unclear if a student group in the LMICs context could also benefit from MBSR. Here we utilized the PSS and the Stress score in the DASS-42 scale to assess the subjective stress perception of student participants. Our result demonstrated that right after the intervention (post-MBSR measurement), only the Test group's DASS stress score was found to decrease (p = 0.002, 30%, Fig. 5A), while no change in PSS scores was noticed in post-MBSR measurement (Fig. 4). Interestingly, the Test group's stress score for both measures substantially decreased when comparing between pre-MBSR measurement score and the Sustained measurement score (p = 0.002, 33%, Fig. 4A and p < 0.001, 50%, Fig. 5A for PSS negative score and DASS stress score, respectively). The combined PSS sum score of the Test group was also reduced by 25% (p < 0.001) in the Sustained measurement (Fig. 4C). On the contrary, the DASS stress score and PSS negative score of the Control group remained unchanged over time ( Fig. 5A and Fig. 4A). The PSS's positive score (i.e., the "Perceived Coping" items) also did not show any change in both groups (Fig. 4B). Note that the p-values of all multiple comparisons in PSS and DASS scores between three measurement sessions of both Control and Test group can be found in Supplementary Table 1. To understand whether MBSR could influence other mental health issues, we analyzed the Depression and Anxiety scores from the DASS-42 scales. In general, the Control group had a stable subscale score over time. Interestingly, in the Sustained measurement, the Anxiety and Depression subscales of the Test group were reduced by almost 40% from the initial scores (p = 0.002, Fig. 5B and 5C), though no change in this score is observed between the pre-MBSR and post-MBSR measurement ( Fig. 5B and 5C). It remains unclear whether the slow effect of MBSR on anxiety and depression levels (Fig. 5) should be attributed to the training course or the continuation of mediation practice after course completion (Supplementary Fig. S2).
Fig. 3. Experimental design of the Stroop color task.
A. An et al.
Validating the stress-inducing effect of the Mental Arithmetic (MA) and Stroop tasks
Since one of the main goals of MBSR training is to improve the participants' ability to cope with their negative emotions, we hypothesize that MBSR could influence brain activity in response to stress (Carroll et al., 2021;Johns et al., 2016;Katahira et al., 2018). To validate this hypothesis, we first designed a set of stress-induced tasks, including the Mental Arithmetic (MA) tasks and the Stroop tasks. We arranged the tasks in the order of hard-medium-easy repetitively during three-stage MA tasks performance, while the Stroop tasks speed up the reaction after each round Jun and Smitha, 2016) Comparing different difficulty levels of the MA tasks, we found that the difficulty and stress rating of the subjects is aligned with our experimental design (p < 0.001, Supplementary Fig. S3). Specifically, these difficulty and stress ratings reduced significantly from timed rounds to untimed rounds in both groups at all three measurements ( Supplementary Fig. S3).
The Control group's difficulty and stress perception scores toward the MA tasks were reduced at several task rounds between pre-MBSR and Sustained measurement, specifically on hard and medium rounds (Fig. 6A ii and Fig. 6B). In contrast, the perception of the Test group toward these two metrics remained largely stable over time. On the other hand, at the easy level, both groups did not report any considerable difference except for the reduction in stress perception of the Control group in Round 6 (p = 0.033, Fig. 6B vi). Note that the p-values of difficulty and stress level in each round are summarized in Supplementary Table 2. These findings could be attributed to the increasing familiarity with the tasks of the Control group, which resulted in the reduction of mean reaction time (p = 0.033, Supplementary Fig. S4A) and higher accuracy (Supplementary Fig. S5A).
Overall, increasing difficulty perception of the tasks is positively correlated with higher stress sentiment among the subjects of both experimental groups across all three measurements (p < 0.001, R 2 ~0.372-0.667, Fig. 7). Additionally, when comparing the linear Fig. 6. Self-reported evaluation of difficulty (A) and stress (B) in different task levels in three data measurements for both Control group (n = 21) and Test group (n = 20). Rounds 1, 4, 7; rounds 2, 5, 8; and rounds 3, 6, 9 are defined as difficult, medium, and easy levels, respectively. Each bar represents Mean ± SEM.
A. An et al. regression slopes of the two groups, the self-reported evaluations did not show differences between the stress level and difficulty perceptions of all subjects (p > 0.05) (Fig. 7). Therefore, both groups reported equivalent perceptions about difficulty and stress levels corresponding to the tasks. In conclusion, these findings confirm that the stress-induced tasks successfully generated stress sensation, especially at higher difficulty levels for both the Control and Test groups.
Regarding the Stroop color tasks, the mean reaction time of Round 1 tended to reduce in all subjects of both groups, while similar metrics of the last two rounds were stable over time. Meanwhile, the Control group increased accuracy (p = 0.002) at all three measurements in the Stroop tasks, indicating that they got accustomed to the tasks ( Supplementary Fig. S4B, S5B).
Power Spectral Density (PSD)
As the previous EEG studies on mindfulness have shown that meditation was associated with increased alpha power and theta power in both healthy individuals and in patient groups (Fan et al., 2014a(Fan et al., , 2014bLomas et al., 2015;Takahashi et al., 2005;Tang et al., 2009), we hypothesized that the Sustained effect of participating in MBSR training would be characterized by the increase in power of alpha and theta oscillations. To test this hypothesis, we performed the PSD calculation based on preprocessed EEG data, then compared the relative power of oscillations (e.g., alpha, beta, and theta) during task performing and resting phases between the Control and Test groups. The initial analysis of relative power in 22 individual phases (i.e., 9 rounds of mental arithmetic tasks, 3 rounds of Stroop tasks, and 10 rests) indicated no difference (p > 0.05) between the two groups. We then averaged the relative power of all rounds of mental arithmetic tasks, all rounds of Stroop tasks, and all rest sections related to each task to perform statistical analysis. Thus, the results of statistical analysis are reported for the following conditions: Mental Arithmetic Tasks, Mental Arithmetic Rest (i.e., averaged value of six resting phases in-between each round of mental arithmetic tasks), Stroop Tasks, and Stroop Rest (i.e., averaged four resting phases in-between each round of Stroop tasks). The p-values of all multiple comparisons between PSD values of three measurement sessions of both Control and Test group are summarized in Supplementary Tables 3 and 4, respectively.
MBSR intervention is associated with a sustained increase in alpha band power
Test group's relative alpha power was distinctively characterized by significant increases in the frontal, temporal and occipital areas right after and 8-week after the intervention (i.e., post-MBSR measurement and Sustained measurement) (Fig. 8). No change was observed for the Control group. There were also no noticeable differences in alpha power between two groups over time when applied One-way ANOVA. The increase in the relative alpha power of the Test group can be revealed spatially in the topographic map (Fig. 8A). Specifically, in the frontal area, relative alpha power for the Test group showed robust increases at both post-MBSR (p = 0.033, Fig. 8B i, v, at Fp1 and F7 channels) and Sustained measurement (p = 0.002, Fig. 8B ii, at F3 channel) when compared with pre-measurement. A similar rise in alpha band power was also found in the temporal cortex, which was present in both post-MBSR measurement (p = 0.033, Fig. 8B iii and p < 0.001, Fig. 8B vi) and Sustained measurement (p = 0.002, Fig. 8B iii and p = 0.033, Fig. 8B vi). At the occipital area, the alpha power after intervention for the Test group was also higher at both post-MBSR and Sustained measurement for the O1 electrode ( Fig. 8B iv, vii, ix) or only post-MBSR measurement for the O2 electrode ( Fig. 8B viii and xi). Overall, the MBSR training led to a global strengthening in alpha band power at both rest and task stages. To estimate the correlation between changes in alpha band power and changes in subjective stress perception, we applied linear regression in the delta values (i.e., postpre) of alpha power at each channel and delta values of each questionnaire subscale. We found that the Test group has a significant correlation between the delta values of alpha power of temporal and occipital lobes with the delta values of the DASS results, especially in the Anxiety and Depression subscales after completing the MBSR course (Supplementary Table 5, p < 0.05 and R 2 = 0.305), implying that the change in alpha power at these brain regions might be a neural representation of the psychological improvement.
Lower beta power was observed after the intervention
Theta and beta activities were also lower for the Test group in postand Sustained measurements (Fig. 9A), but these changes were not prevalent and distinctive as those in alpha power. In particular, the relative beta power of the Test group in Mental Arithmetic Rest and Stroop Tasks at the O1 position was reduced eight weeks after intervention (Sustained measurement, p = 0.002, p = 0.033, respectively) ( Fig. 9B i and iii). Surprisingly, both the Control and Test groups had lower beta relative power during Stroop Tasks at the T5 position (p = 0.002, Fig. 9B ii), implying that this reduction might be independent of the MBSR training. With regards to relative theta power, the statistical analysis only revealed one decrease at the T5 position for the Test group during Mental Arithmetic Rest right after intervention (postmeasurement) (p = 0.033), but then followed by an increase in Sustained measurement (p = 0.033) (Fig. 9B iv), indicating that this effect of MBSR on theta band power is not long-lasting. Overall, Supplementary Fig. S6 summarizes the effect of MBSR practice on brain activity during rest and task stages. Here, it was evident that MBSR practice is associated with sustained stronger alpha power at rest stages (Supplementary Fig. S6A, B) and during mental arithmetic task performance but not during Stroop task (Supplementary Fig. S6B). Compared to the increase in relative alpha power, the decrease in beta and theta is less pervasive and not particularly distinctive for MBSR intervention. A. An et al. Fig. 8. The effect of MBSR practice on brain activity during Resting phase and during Stress-induced tasks is reported by the relative power of alpha band. A) Topo maps indicate the changes in brain activity in Test group. B) The column plots summarize specific channels having statistically significant changes in alpha band power. Each bar represents Mean ± SEM.
MBSR training led to a sustained reduction in negative emotion states
The overall negative emotional states were steeply reduced in the Test group post-intervention in the DASS-42 Stress subscale (Fig. 5A). These results are consistent with previous studies that demonstrated stress reduction in either university students or other populations (cancer patients and the elderly) via PSS or DASS-42 scales (Aftanas and Golocheikine, 2001;Fan et al., 2014b;Matousek et al., 2011;McIndoo et al., 2016;Morais et al., 2021;Moynihan et al., 2013;Song and Lindquist, 2015). Despite the disruption caused by the COVID-19 outbreak during the last two weeks of the MBSR program, the MBSR program still brings out the robust effect on stress reduction.
This study was also one of the first longitudinal studies that showed the sustained effects of MBSR on stress, depression, and anxiety reduction ( Fig. 4 and Fig. 5). In previous works, the PSS or DASS scores mostly returned to baseline in the follow-up measurement (McIndoo et al., 2016;Morais et al., 2021;Moynihan et al., 2013), which might occur due to the difference in subject groups or practice environment. It was rather intriguing to notice that most of these MBSR-induced positive psychological impacts were more evident in the 2-month follow-up rather than immediately after the intervention, implying such a program would be more profoundly beneficial if subjects continue to implement the skills and knowledge taught during the course. The majority of Test group subjects in our study maintained their meditation sessions ( Supplementary Fig. S2), resulting in the sustained psychological effect of MBSR and highlighting the high feasibility of this intervention.
Mindfulness practice primarily induced higher alpha band power
We found a significant increase in alpha power for the Test group in frontal, temporal, and occipital cortices in both hemispheres postintervention (Fig. 8). This finding is in line with several previous studies which reported the increase in alpha synchronization associated with mindfulness compared with pre-intervention (Morais et al., 2021). The alpha band power is indicative of the state of stress reduction and Fig. 9. The effect of MBSR practice on brain activity during Resting phase and during Stress-induced tasks is reported by the power of beta and theta frequency bands. A) Topo maps indicate the changes in the beta band (i-vi) and theta band (vii-viii). B) The column plots summarize specific channels having significant changes in the relative power of beta band and theta band. Each bar represents Mean ± SEM. relaxation (Seo and Lee, 2010). This increase in alpha power post-intervention of the Test group, especially in the prefrontal and frontal area, might reflect the improvement of internal processing, suggesting the enhanced ability to direct attention internally, with a decrease in thought dispersion (Cooper et al., 2006).
As shown in Fig. 9A and Supplementary Fig. S6, changes in beta power are not typically distinctive for the Test group, as there was the T5 electrode of the Control group showed a significant main effect or interaction. These changes in beta power might be misinterpreted as sustained effects of MBSR training if the Control group was not included, as there were in similar MBSR studies (Morais et al., 2021). Longitudinal reduction of beta-band power has been associated with previously reported mindfulness training (Saggar et al., 2012). However, there was a discrepancy among EEG studies on mindfulness regarding changes in beta oscillations, and the interpretation of the functional significance of beta is also mixed (Lomas et al., 2015). We suspected that changes in both Test and Control groups might reflect the participants' familiarity with included cognitive tasks, since beta oscillations are typically associated with the sensorimotor processing (Brovelli et al., 2004). While both the brain response and the negative emotion scales exhibited similar trends after MBSR practices for only the Test group, a concomitantly significant correlation between these measurements was found in both subject groups (Supplementary Table 1). No difference in the degree of correlation was detected. Therefore, MBSR potentially did not affect the intrinsic relationship between psychological measurements and their corresponding neural representative.
Limitations
While the current study shows several promising results, few limitations are presented. Firstly, the relatively small-scale study (n = 49, in one university) could potentially limit the statistical analysis. Additionally, the MBSR session requires 8-week consecutive practice; therefore, participants, such as full-time employees, might drop out due to a limited time budget. One potential strategy to maintain the participants' commitment is to utilize shortened versions of MBSR, which were successfully implemented on subject groups such as pharmacy students and demonstrated clear psychological improvement (Hindman et al., 2015;McIndoo et al., 2016;O'Driscoll et al., 2019). An official and validated version of a shortened MBSR program is, however, not available; hence, their implementation is rather challenging.
Conclusion
Academic pressure tends to affect the mental health of college students negatively. This study aims to clarify an effective MBSR program for college students by assessing the sustained efficacy of the intervention on stress alleviation and brain activity when responding to shortterm stress conditions. Our results indicated that MBSR effectively reduced stress scores in the Test group after completing the MBSR training, whereas the similar measure of the Control group remained stable (Fig. 5A). Even two months after the training, subjects in the Test group exhibited lower negative emotional states, especially in the Anxiety and Depression scores ( Fig. 4 and Fig. 5). The stress reduction effect in the Test group was accompanied by an increase in frontal, temporal and occipital alpha power (Fig. 8) when subjects were exposed to short-term stress stimulations at both post and Sustained measurements. In conclusion, the sustained effects of MBSR were successfully demonstrated based on the reduction of subjective stress perception score and the ability to maintain high alpha power while coping with stress-stimuli. | 2022-05-16T15:03:55.532Z | 2022-05-01T00:00:00.000 | {
"year": 2022,
"sha1": "e1a0d6dedd26e5d99afe954a667df1fad626d81d",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1016/j.ibneur.2022.05.004",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "2a353932b3abc1a8ea078c6784ea54f607daa81e",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
122700494 | pes2o/s2orc | v3-fos-license | Studies of CP Violation and Mixing in the D Mesons decays from BABAR
After an introduction on the mixing and CP violation (CPV) in the neutral charmed mesons system, we present the current status of the measurements. We discuss some recent analyses performed on the data sample collected by the BABAR detector at the PEPII e+e− asymmetric collider.
Introduction
Mixing occurs in neutral mesons systems in which the flavour eigenstates differ from the mass eigenstates. The D 0 − D 0 system is one of the four systems in which this phenomenon is predicted by the Standard Model (SM) and observed experimentally, the other systems are B 0 − B 0 , K 0 − K 0 and B s − B s . The oscillation rate is different in each system, the D system is the one that shows the smallest mixing, this is the reason why it's the last one in which evidence of mixing has been found experimentally.
Theory and Formalism
The following is a brief overview on mixing and CP violation, for more details see PDG [1].
The time evolution in the D 0 -D 0 system is described by a Schrödinger equation: where M and Γ are two Hermitian matrices. C, P and T conservations impose constraints on the elements of M and Γ. In the following we will assume CP T conservation, which implies The two D 0 mass eigenstates are obtained diagonalizing the equation above and can be represented as: The normalization condition is |p| 2 + |q| 2 = 1. The mass eigenstate |D i has a mass M i and widthΓ i . Assuming CP |D 0 = +|D 0 , in case of CP conservation the |D 1 is the CP -even state.
CP Violation
CP Violation is an important ingredient when neutral meson mixing is involved. The only source of CP Violation (CP V ) in the Standard Model is the complex phase in the CKM matrix. There are three possible manifestations of CP V : (i) in the decay (direct CP V ); (ii) in mixing (indirect CP V ); (iii) in the interference between mixing and decay (indirect CP V ).
There is direct CP V when the probability of a meson D to decay into a final state f differs from the probability of the anti-meson to decay into the CP conjugate statef . In other words when: H d is the Hamiltonian responsible of the decay. It's important to underline that the moduli |Āf | and |A f | can be different only if the decay proceeds through the sum of two different contributions, with different phases. If the phase is the same for the two contributions, then the asymmetry will be zero. This kind of CP V is independent of mixing and it is expected to be negligible in the decays of D mesons [2].
When CP V in mixing occurs then the probability of the transition D 0 → D 0 is different from the probability of the inverted process D 0 → D 0 . The signature for CP V in mixing is: The CP V in the interference between mixing and decay can occur when both the D 0 and the D 0 have access to a common final state f . In this case f can be reached in two possible ways: (i) with a direct decay: D 0 → f ; (ii) with a decay after the mixing: CP V occurs in the interference between these two paths and it is possible also when CP is conserved in the decay and in the mixing separately. In this case we define λ f : where δ f and φ f are, respectively, the strong and the weak phase differences between the two paths. There is CP V when the weak phase difference is different from zero: 3. Mixing and CP V in the charm sector The mixing in the D system is described by two real parameters, the mixing parameters that are proportional to the difference of masses and widths of the mass eigenstates: withΓ ≡ (Γ 1 + Γ 2 )/2. If either x D or y D is non-zero, mixing will occur. If D 1 and D 2 have different lifetimes then y D = 0. The mixing parameters defined above are the ones directly related to the physical quantities that drive mixing. Other parameters not directly sensitive to these quantities may be defined. The parameter y CP D [3] measures how much the lifetime of the CP -even eigenstate (τ CP + ) differs from the D 0 lifetime: In case of no CP V then y CP D = y D . The CP V is also quantified in the parameter∆ Y : In case of no mixing∆ Y = 0. In the limit of CP conservation y CP
Standard Model Predictions and Measurements
Standard Model predictions for x D e y D are of the order of 10 −2 or smaller for both, as showed in Fig. 1. The suppression of mixing in the D system with respect to the other systems has been understood within the Standard Model. Since the mixing process changes the flavour of two units, the transition must proceed through an intermediate state that can be either real or virtual. One possible transition proceeding through a virtual intermediate state is represented in Fig. 2 (left). The contribution of these diagrams is small [5] due to the CKM factors associated to the vertices: the contribution of the b quark is doubly Cabibbo suppressed (DCS) while the contributions of the d and s quark are GIM suppressed. The biggest uncertainty related to the computation of these diagrams is due to the poor knowledge of the hadronization processes. It's important to notice that the D system is the only one in which the virtual quarks are down-type since the top quark is too heavy to form long-living bound states and the π 0 coincides with its anti-particle. From this point of view the D system is unique.
Some examples of real diagrams are reported in Fig. 2 (right), the intermediate states are all the possible final states common to the D 0 and the D 0 decays, such as K + K − and π + π − . This second class of diagrams is expected to be dominant in the mixing process though the uncertainty associated to their computation is even larger [5]. The first evidence of D 0 mixing was found by the Belle experiment [6] and the BABAR experiment [7] in 2007, and it was confirmed a year latex by the CDF experiment [8]. The measurements are all in agreement with the SM predictions and show no evidence of CP V . Fig. 3 shows the mean world values for x D e y D , extracted from all measurements, computed by the HFAG group [9].
Impact of the measurement
Historically mixing measurements have played an important role as SM tests and in building New Physics (NP) theories. The D system is the last one in which mixing has been observed, still not yet with a precision that can bring information on a possible or impossible NP structure and on CP V .
Measuring mixing in the D system completes the picture of neutral meson mixing within the SM bringing new information since this is the only system in which the virtual quark are down-type. Moreover, since the mixing parameters values are quite small, NP could manifest itself with a higher relative value (v. Fig. 1). Unfortunately, except for big values of x D and y D (hypothesis currently not favored by the experimental results), the interpretation of the results is not straightforward, it is limited by the uncertainties in the SM predictions.
The search for CP V is probably the most effective probe for NP: an evidence of CP V with the current experimental sensitivity would be a clear sign of NP.
The BABAR Experiment
The BABAR detector [10] has been running at the B-Factory PEP II, the e + e − asymmetric collider located at the SLAC National Accelerator Laboratory. The data taking started in 1999 and ended in April 2008, resulting in a total integrated luminosity of 530 fb −1 . The collider operated mostly at a center of mass energy of 10.58 GeV, corresponding to the mass of the Υ(4S) resonance (Run1 to Run6). During the last period of data taking it also operated at lower energies, corresponding to the masses of the Υ(3S) and Υ(2S) (Run7). The asymmetry in the beam energies led to a boost βγ = 0.56. The internal part of the detector consists of the Silicon Vertex Tracker (SVT), the drift chamber (DCH), the Cherenkov light detector (DIRC) and the electromagnetic calorimeter (EMC). Outside there is the instrumented flux return (IFR) and a superconductive solenoid which produces a uniform magnetic field of 1.5 T parallel to the direction of the electron beam. The B Factory produced around 690 millions of e + e − → cc events, working also as a Charm Factory.
The D 0 mesons tagging
In CP V measurements it is of crucial importance to identify the flavour of the neutral D meson at production. In order to do so the decay chain D * + → D 0 π + is reconstructed and the charge of the pion tag the flavour of the D meson. Samples of D mesons reconstructed in this way are referred to as tagged samples. In mixing analyses the knowledge of the flavour of the D is not always needed and in this cases we talk of untagged samples.
The tagged samples contain a much lower level of background but are around four times less populated than the correspondent untagged samples.
A time-dependent analysis on the Dalitz Plot
measurement of the mixing parameters is possible through a full time-dependent analysis of the Dalitz plot (DP) for the three-body final states K 0 S h + h − (h = π, K) with the K 0 S reconstructed in the π + π − final state. The three-body D 0 decay is assumed to proceed through two-body intermediate states where one is a resonance, and it is described by two independent Dalitz variables, m 2 + and m 2 − which are the reconstructed invariant mass squared of K 0 S h + and K 0 S h − . In the following we assume CP conservation in the decay, i.e. A(m 2 − , m 2 + ) =Ā(m 2 + , m 2 − ). The total amplitude A(m 2 − , m 2 + ) can be written as the superimposition of the amplitudes of each resonance (r) plus a non-resonant (NR) term: where a r and φ r are the modulus and the phase of the amplitudes and A r (m 2 − , m 2 + ) reproduces the dependence on the Dalitz variables. According to [11] the modulus and the phases are extracted from data while a phenomenological model for each A r (m 2 − , m 2 + ) is assumed. After selecting a sample of D 0 of purity greater than 98% for both the D 0 decay modes, we discriminate signal from background events fitting the distribution in the (∆m = m D * − m D 0 , m D 0 ) plane for the signal and background categories. An extended maximum likelihood fit is then performed on the combined samples of D 0 → K 0 S π + π − and D 0 → K 0 S K + K − decays in order to extract the Dalitz model and the resolution function parameters, yielding [12]: y D = [5.7 ± 2.0(stat) ± 1.3(syst) ± 0.7(model)] × 10 −3 This measurement, done on 485 fb −1 , is the most precise direct measurement of x D and y D . The contour plot for the 1 to 3 standard deviations regions is reported in Fig. 4.
The most important systematic errors are related to the Dalitz model assumption (referred to as "model" in the error), the fit bias and the detector misalignment. Figure 4. Fit results for K 0 S π + π − − K 0 S K + K − combined fit with the confidence-level (CL) contours for 1−CL = 0.317 (1σ), 4.55×10 −2 (2σ) and 2.70×10 −3 (3σ). Systematic uncertainties are included. The no-mixing point is shown as a plus sign (+).
6. Lifetime Ratio analyses: D 0 → K − π + , K + K − , π + π + The BABAR Collaboration has presented two analyses on the measurement of the ratio between the lifetimes of D 0 reconstructed in CP eigenstates and in CP mixed states, y CP D . One was performed on an untagged sample, the other on a statistically independent sample of tagged events. In both analyses τ D 0 is extracted from the proper time distribution of D 0 reconstructed in the final state K − π + assuming an exponential decay 1 . The same hypothesis is under the extraction of the value of τ CP + from K + K − final state and, in the tagged analysis only, the π + π − final state. The value of y CP with a significance for mixing of 4.1 standard deviations, resulting the most significant mixing measurement.
7. T violation with D 0 → K + K − π + π + Considering the four-body decay D 0 → K + K − π + π + , the T odd observable C T ≡ p K + · ( p π + × p π − ) built with the momenta of the tracks, is used to measure a T asymmetry: Since final state interactions can produce a non zero value of A T , the relevant parameter is , whereĀ T is extracted from a D 0 sample, and therefore a tagged sample of D 0 is 1 Since the mixing effect is smaller than 1%, the assumption is experimentally acceptable. needed. In order to extract the value of A T from a sample of 470 fb −1 a two-steps analysis is performed. First, the resolution function parameters for signal and background components are extracted from data in the (∆m,m D 0 ) plane. Then the sample is divided in four sub-samples (depending on the flavour of the D and the sign of C T ) and, after fixing the resolution function, the number of events in each sub-sample is extracted and A T is computed. The measurement is still statistically limited [15]: but is represents an improvement of one order of magnitude with respect to the previous FOCUS measurement. The most important systematic error is associated to the particle identification, other less important contributions come from the cut on the D 0 momentum inn the center of mass, the detector misalignment and the fit bias.
8. CP violation with D + → K 0 S π + A measurement of the direct CP asymmetry is possible studying the decay D + → K 0 S π + and and computing: A CP violation contribution from K 0 −K 0 mixing must also be taken into account. The expected value of A CP due to the CP violation in the K 0 system is A K CP = [−0.332 ± 0.006]% [14], any deviation from it would be evidence of CP violation.
The measured asymmetry A = N D + −N D − N D + +N D − will have two contributions besides the CP asymmetry: the forward-backward asymmetry and the detector-induced asymmetry in the reconstruction of the charged particles. The former is extracted directly from data exploiting its oddness in the center of mass polar angle. In order to remove the latter a new data-driven technique has been developed.
The track reconstruction efficiency
If the detector is not completely symmetric in charge reconstruction, the measurement of CP violation will be affected by a detector-induced asymmetry. In the signal channel the only track affected by this asymmetry is the charged pion from the D decay. In order to estimate this effect and subtract it from the final result a high populated control sample with no asymmetry from physics is selected: if no detector asymmetry is present then the number of reconstructed π + (N π + rec ) should be equal to the number of reconstructed π − (N π − rec ). Using the Υ(4S) decays 2 the ratio of the charged pion detection efficiency as a function of the D momentum was computed as: and this correction is applied to the negative D candidates. In Fig. 5 the value of the ratio of the detection efficiency and its error are reported: the plots shows that the correction is small and dominated by statistical fluctuations.
Results
After the subtraction of the forward-backward asymmetry (v. fig 6), the measured CP asymmetry is: it is consistent with A K CP and therefore with no CP violation in the charm sector, as predicted by the Standard Model at these levels of error. This measurement is the most precise CP asymmetry measurement at B-Factories. The most important systematic errors are relative to the charge asymmetry correction and the K 0 -K 0 regeneration in the material. | 2019-04-20T13:04:28.213Z | 2011-12-28T00:00:00.000 | {
"year": 2011,
"sha1": "2ca5f1c8ab69e203f13c14c6dce74049ca27b03d",
"oa_license": null,
"oa_url": "http://iopscience.iop.org/article/10.1088/1742-6596/335/1/012043/pdf",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "0295785b2d20d849d9b3bed3050f7c3bc25e4fb4",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
251556925 | pes2o/s2orc | v3-fos-license | Regulatory constraint and small business lending: do innovative peer-to-peer lenders have an advantage?
Introduction This paper investigates the regulatory advantage conferred on innovative Peer-to-Peer (P2P) lenders, in respect of lending to small businesses. It does this through the lens of the response to regulations imposed by the Frank-Dodd Act of both traditional banks and their online P2P competitors. The later are sometimes colloquially referred to as “FinTechs”, in reference to their use of financial technology. In fact, P2P lenders are a subset of the FinTech sector. As P2P lenders are not deposit takers, they are subject to less regulation than traditional banks. Small businesses1 are the backbone of the U.S. economy and the provision of credit is central to their functioning 2 Since 1995, small businesses have created two—thirds of Abstract
banks was hindered in the U.S. as a result of the Dodd-Frank Wall Street Reform 7 (commonly referred to as Dodd-Frank Act) and the Consumer Protection Act enactment on July 21, 2010.
According to academic studies (Bordo and Duca 2018;Acharya et al. 2018;Bouwman et al. 2018), the regulations of the Dodd-Frank Act 8 strained already high operational costs and increased capital constraints on banks, especially those with $10 billion or more in assets under the Federal Reserve's stress test requirements. The cumulative number of regulations are detailed in Fig. 1. Cortés et al. (2020) claim that such stress tests create a direct link from bank lending risk to capital and impose heavy capital requirements on small business loans. Therefore, the Dodd-Frank Act regulatory requirements cut down on the incentives for banks to make loans to serve businesses, especially small businesses, for which bank credit is one of the important sources of external financing (Mills and McCarthy 2014).
Under the Dodd-Frank Act, the average tier 1 risk-based ratio of U.S. banks increased by 22-27% between 2008 and 2015 (Buchak et al. 2018). In addition, banks with more than $10 billion in total consolidated assets are subject to an annual stress test which consists of dynamic capital requirements that impose risk-sensitive capital buffers on banks for expected deterioration in an adverse economic scenario (Bindal et al. 2020). In addition, Bindal et al. (2020) state that stress tests impose dramatically higher capital requirements on small business lending.
As mentioned, during the same period, the credit needs of small businesses started to be targeted by a new set of lenders that use innovative FinTech to disrupt the small business lending market (Mills 2018). These are collectively referred to as Peer-to-Peer lenders. Although being small relative to incumbents, these alternative lenders provide rapid turnaround and online accessibility for borrowers and use new data-rich credit score algorithms (Palladino 2018). According to Jagtiani and Lemieux (2016), these lenders are enabled by technology and have little (or indeed, are not subject to any) regulation. It could be argued this makes alternative lenders attractive to small business lenders in a post-crisis environment, and thus emerging of alternative P2P lenders had begun to alter the game for how small businesses access financing in the U.S. (Mills and McCarthy 2014). Alternative P2P lender total loan origination volume, loan application number and county number are presented in Table 1.
In order to provide causal evidence that the Dodd Frank Act impacted the provision of loans to small businesses, we use a quasi-natural experiment. This allows us to investigate how the new requirements affected treated banks with $10 billion total assets or more small sized business loans supply relative to untreated banks with less than $10 billion assets. It allows us to evaluate how the lack of the regulatory requirements gave FinTech lenders an advantage.
Firstly, to address the impact on the banks, we used small business bank and countylevel data. We replicate the method used by Tang (2019). After classifying treated and control banks (1), we investigate trends at the county level some counties have banks that were subject to the regulation, and others did not. It is suggested that those counties that had an impact from the Dodd-Frank Act saw less competition in banking, and therefore saw less of an impact. This follows the observations of Boot and Thakor (2000) regarding the development of relationship lending when there is less interbank competition.
We measure the banking competition intensity by (1) the concentration ratio of the "big three banks" (C3) and (2) the Herfindahl-Hirschman Index (HHI) using the banks' market share in terms of bank branches number in counties following Degryse and Ongena (2007) and Chong et al. (2013).
Treated counties are defined as counties if there is a bank with $10 billion assets or over which subject to the Dodd-Frank Act. We define treatment groups as counties with a high concentration of Dodd-Frank eligible banks. We further classify them as where there is a low banking competition at the 75th percentile of C3 and HHI. This means that where there is a bank asset that is below $10 billion, and there is a high competition at the 25th percentile of C3 and HHI, it is defined as a control county. In this way, our sample can be used to identify the impact of the Dodd-Frank Act impact on (1) aggregate county-level small business lending. Further, it can be used to identify (2) alternative P2P lender activity in treated and control counties. In this regard, according to the results in Table 5, we conclude that treated banks saw a decrease in the amount of small business lending. In addition, we note that county-level aggregate small business loan volume declined after the enactment of the Dodd-Frank Act. At the same time, when bank small business loan supply declines, demand for alternative P2P lending increases. Supportive of our findings in the concentrated counties, Hodula (2022) found evidence that FinTechs may act as substitutes in highly concentrated markets.
To the best of our knowledge, ours is the first paper to investigate the activities of both traditional banks and innovative alternative lenders in the small business market using the Dodd-Frank Act as an exogenous shock at the county level. In addition, our paper adds alternative P2P lenders to the debate in the literature on small business lending (e.g. Buchak et al. 2018;Tang 2019;Fuster et al. 2019;Hughes et al. 2022;De Roure et al. 2022). We note that Bordo and Duca (2018) and Zou (2019) also focus on small business lending and the global financial crisis. We, however, utilize the Dodd-Frank Act's impact on small business lending to identify the regulatory advantage of the P2P lenders.
After the credit crisis, regulation was focused on both capital and liquidity requirements by regulators, particularly in view of the fact that reserve requirements for U.S. banks. According to Thakor (2018), higher capital requirements can make it more challenging for banks to attract capital, and so they decreased lending in response to an anticipated rise in regulatory capital requirements after the financial crisis. There are several reasons why small business owners might turn to business loan alternatives. These include lower credit requirements, easier qualification and faster approval thanks to innovative technology (Milne and Parboteeah 2016). Akhigbe et al. (2016) present evidence that following the transition of the Dodd-Frank Act, banks discretionary risk-taking decreased due to the rising bank capital ratios and banks decreasing their non-performing loans levels. Andriosopoulos et al. (2017), meanwhile, investigate the impacts of key legislative events of the act and their conclusions support our view that there were changes to the competitive structure of the financial services industry. Allen et al. (2018) further investigate the market's response to the elimination of too-big-to-fail for large banks against the passage of the Dodd-Frank Act and suggest that act do not eliminate Too-Big-to-Fail banks. In their recent study, Calem et al. (2020) investigate banks stress test exercises impact on the supply of mortgage credit which is implemented under the Dodd-Frank Act Stress Testing (DFAST) regulatory programs and according to the paper that stress tests only alter originations of credit in the jumbo mortgage market. Additionally, Bindal et al. (2020) investigate the Dodd-Frank Act's size based regulatory requirements impact on banks merger and acquisitions and small business lending. Their results indicate that the size-related regulatory thresholds created by the Dodd-Frank Act has significant real effects on loans to small businesses but have indirect treatment effects on bank acquisitiveness.
In summary, our use of the Dodd-Frank Act as a natural experiment ties together separate strands of the literature relating to small business lending and the growing role of innovative alternative lenders.
Small businesses lending and the role of innovative sources of lending
Our working hypothesis is that the innovative P2P lenders benefit from a regulatory advantage. We therefore use two testable hypotheses related to small business lending. This ties the Dodd-Frank Act and the increasing role of alternative lenders together.
The distinctive features that distinguish small businesses from medium and large sized enterprises have long been the subject of research. Ang (1991) claims that the source of the structural and managerial differences could be traced to several features peculiar to small businesses. Out of this set, small firms are shown to make financial decisions in a different way than large companies. In this line of enquiry, several papers investigate small business lending from different perspectives such as bank consolidation, mergers and acquisitions or banking market size structure effects on small business lending, relationship lending, opaque small businesses, and economies of small business finance. We suggest the nature of small businesses makes them more amenable to the use of FinTech.
Consolidation of the banking sector is ruled out as an exogenous factor. Weston and Strahan (1996) and Takáts (2004) claim that consolidation does not adversely affect the credit availability to small businesses contrast with those of Berger et al. (1998) and Sapienza (2002), who find that the effects of consolidation reduce the small business lending activity of banks. Peek and Rosengren (1998) also indicate that while acquirer banks have a higher degree of specialization in small business lending than non-acquirer banks, similar to the mergers increase the consolidated bank small business loans. In another study, results show an external impact of consolidation in which the bank lending to small businesses can be reduced by mergers and acquisitions (Berger et al. 2004).
The size of financial institutions does matter. DeYoung et al. (1999) reveal that there is a negative relation between the size of the bank and its small business lending activity, and Berger and Udell (1995) claim that as banks become larger and more complex, they can reduce to provide loans to small firms. Regarding the market size structure of local market participants, Craig and Hardee (2007) investigate whether banking consolidation has affected small business lending by using the Small Business Finances Survey. They find that access to bank credit for small businesses is lower in markets dominated by the largest banks. Berger et al. (2007) also investigate market size structure affects the credit supply to small firms both in terms of prices and quantity and the point out that large banks compared to small banks tend to have lower loans to small businesses to assets, however, large banks take advantage of some transaction lending technologies to lend opaque small businesses.
Additionally, Mcnulty et al. (2013) indicate that the propensity to lend to small firms decreases as bank size rises. Further, that most loans to small businesses are made by small banks. In a recent study, Berger et al. (2015) show how local banks' market size structure impacts the loans received by small businesses and find that during normal times there is a greater market presence of small banks in more lending opaque and small firms, but this effect vanishes during the financial crisis.
Furthermore, Petersen and Rajan (1994) investigate the effect of the relation between a small firm and their creditors (banks) on the availability and funding costs of credits and they find that the close relationship between the firm and the bank has little impact on credit pricing. Berger and Udell (1995) claim that small business pays lower interest rates and less collateral if there is a longer banking relationship.
Moreover, Cole (1998) shows that lenders are more likely to expand credit to firms with which they have a constituted relation. Berger et al. (2001) examine the bank relation with internationally opaque businesses and find that some foreign-owned and large banks that are generated by mergers and acquisitions and foreign institutions may have problem to provide loans to opaque small businesses. Berger and Black (2011) analyse the comparative advantages of large and small banks in specific lending technologies and show that small banks have a comparative advantage in relationship lending for small firms.
The relationship between the larger bank and small business lending has also been investigated. Begley and Srinivasan (2021) looked at the effects of new regulations that banks are exposed to after the global crisis on mortgage lending. They argue that the share of especially four big banks in mortgage loans has decreased, and some of this gap is provided by FinTech lenders in parallel with our study. But Gallo (2021) argues that these online FinTech platforms are not fully efficient, and these platforms may suffer from misrepresentation. This makes it difficult to know lenders' credit history and lead to problems with collections.
We argue that the new regulations applied to the banks negatively affected those banks with a particularly large and high market share. As a result, we observe that loans to small businesses have decreased in the counties where these banks are located and there is low competition. This yields our first hypothesis: Hypothesis 1 Ceteris paribus, after the Dodd-Frank Act, aggregate small business lending declined in the counties where the banks affected by this legislation had a presence, and there was low competition to provide credit.
Apart from the small business lending studies, we further observe that small business loan origination occurs outside the traditional banking system with changing the regulatory structure of the banking system.
As mentioned, the FinTech phenomena began at the same time. There is now a growing literature on alternative P2P lenders (e.g., Cornaggia et al. 2018;Buchak et al. 2018;Tang 2019;Fuster et al. 2019;Allen et al. 2019;Hughes et al. 2022;De Roure et al. 2022). They all suggest P2P FinTechs' are becoming an alternative source of lending to traditional banks. This strand of the literature investigates these new type of lenders activities in the small business lending market. In this regard, Tang (2019) examines whether alternative P2P lending platforms act as substitutes for traditional financial intermediaries or instead as complements and find that alternative FinTech lending is a substitute for bank lending with regards to serving infra-marginal bank borrowers and complements for small loans.
Following a method similar to that used by Tang (2019), we observe that alternative P2P lenders can increase market share if the bank lending criteria are tightened and bank credit supply declines. Philippon (2016) evaluates the potential impact of FinTech on the finance industry and claims that it provides efficiency-enhancing benefits. In this respect, Fuster et al. (2019) point out that the FinTech lenders provide a rapid origination process that is less susceptible to demand fluctuations than traditional lenders and so P2P lenders adjust supply in a more flexible way. In this regard, they are better positioned to deal with the external mortgage demand shocks. Wang et al. (2021) claim that online P2P lending services give consumers and small firms a convenient and affordable loan option. Similarly, Havrylchyk et al. (2020) examine the drivers of P2P earnings growth. They produce evidence on both the role of the Internet and weak banking competition being responsible for the growth.
In a recent study, Balyuk (2016) investigates how FinTech innovation in the form of alternative P2P lending affects the credit provided by traditional intermediaries, for example, banks demonstrate that alternative lending impacts the principles in the consumer credit market by developing the information environment. According to Balyuk (2016), financial innovation can play a significant role in lowering shortcomings in the consumer credit market and FinTech innovations mitigate these shortcomings by creating information spill overs to traditional financial intermediaries. In addition, Li et al. (2021) maintain that banks may benefit from financial innovations in the clustering of financial data for a number of financial applications such as fraud detection, reject inference, and credit evaluation. On the other hand, Kou et al. (2021a, b) contend investments in FinTech can assist banks in decreasing their operating expenses and payment and transactional data enhance SME bankruptcy prediction.
In addition, recent papers focused on P2P lending suggest that alternative lending platforms compete with incumbents at a certain level. Cornaggia et al. (2018) set up a causal relationship between alternative lending infringement and commercial bank lending. They did this by using the differences in regulatory barriers to P2P lending on the borrower, and investor. They conclude that small banks' lending volume decline due to the activities of alternative lenders. Buchak et al. (2018) investigate the shadow banks' growth, particularly FinTech shadow banks, in the mortgage market. They show that both regulatory burdens and improved technology can explain the growth in FinTech shadow banking in the mortgage loan market. On the other hand, Jagtiani and Lemieux (2018) investigated whether alternative lenders' loans penetrated potentially underserved areas, where there are low-income borrowers, inadequate competition in banking services, and regions where bank branches have decreased more than others and regions with fewer bank branches per capita.
Finally, similar to Tang 2022) show that alternative P2P lenders are bottom fishing when unexpected financial regulations generate a competitive disadvantage for some incumbents. This is supportive of our findings. Havrylchyk et al. (2020) contend that alternative lending platforms have partly absorbed banks in some U.S. counties that were more affected by the financial crisis. Moreover, Tang (2019) and De Roure (2022) claim that the banks affected by the decrease in loan supply are not fully substituted by other banks serving in the same region.
We also posit the view that while banks' small business lending activity is slowing down, thanks to digital solutions such as digital tools for loan processing and credit underwriting, information asymmetry and searching cost is reduced. Consequently, alternative small firms have an advantage in respect of accessing funds easily. This allows them to increase their lending market share in the county where the large and high market share banks were affected negatively by Dodd-Frank. This yields our second hypothesis: Hypothesis 2 Ceteris paribus, after the Dodd-Frank Act, loans to small businesses are granted by P2P lenders increased in those counties where the banks that were affected by this legislation had a presence, and there was low credit competition.
Data
The main source of data is the Federal Financial Institutions Examination Council's (FFIEC) Consolidated Reports of Condition and Income (Call Reports) that are filed by U.S. banks. To address regulatory deficiencies identified during the last financial crisis,
Table 2 Descriptive statistics of bank characteristics
This table presents the summary statistics of bank-level statistics. The main dependant variable SBLvol is the log amount of small business loan volume. SBLnmbr is the logarithm of small business loan number. Size is the logarithm of banks total asset. TRBCapital is a total risk-based capital ratio. Core capital is a leverage (core capital) ratio. CoreTier1 is a Tier 1 risk-based capital ratio. Capital is the total bank capital to total assets. Deposits is the total deposits to total assets. ROE is the return on equity ratio. ROA is the return on assets ratio. NPL is the non-performing loans to total loans. SBLtoTLoan is the small business loan to total loans. SBLtoTA is the small business loan total assets. C&I Loans is the logarithm of commercial and industrial loans. Variables are winsorized at the 1 st and 99 th percentile banking regulators were directed to begin collecting annual data on lending to small businesses by the Federal Deposit Insurance Corporation's (FDIC) Improvement Act of 1991. Regulators provide information on loans to small businesses in the Call Report of June each year as required by this act. The Call Report data covers 2009-2012. Table 2 presents the summary statistics of bank-level data.
The second primary source of county small business data is the FFIEC's Community Reinvestment Act (CRA) database. In 1977, CRA was enacted by Congress and had been carried out by bank regulators. In regard to CRA, Congress aimed to stimulate each financial institution to meet the needs of each firm that doing business.
In part, regulations of CRA require that financial institutions report annual information on their lending to small businesses. Especially, it is necessary to report the amounts and numbers of business loans originated in amounts less than $100,000, more than $100,000 through $250,000 and more than $250,000 through $1 million. In addition, they must report the number and amount of loans originated to firms with less than $1 million in revenues. This paper is on a similar tack as ours, albeit looking at it the other way around from a size perspective. It covers annual CRA data covering the total amount and number of loans to small businesses between 2009 and 2012.
In addition to county small business loan data, county level macro variables are collected from the U.S. Census Bureau, St. Louis and New York FED database, Bureau of Economic Analysis (BEA) and FDIC, which displayed with county level data in Table 3.
Lastly, the P2P lender data is sourced with comprehensive information on funded loans and loan volume from Lending Club's website. We justify our use of lending club data following the extensive analysis of the publicly available databases by Teply and Polena (2020). As a U.S. based alternative lender, only Lending Club makes its data publicly. We note this as a limitation of our study but find comfort in the dominant market share position the company enjoyed at this time. This data covers the credit score of Table 3 Descriptive statistics of county characteristics This table shows the descriptive statistics of county-level variables. There are three main dependant variables. SBLoan is the log amount of loans to small businesses in each county. SBLoan1 is the log amount of loans to small businesses for businesses with gross revenues less than $1 million in each county. SBLoan2 is the log amount of loans to small businesses for businesses with gross revenues more than $1 million in each county. The sample also covers county level variables. Population is the county level population. DebtoIncome is the median household debt-to-income ratio by county. Income is the dollar amount of income per person by county. Unemployment is the ratio of jobless people by county. BRNUM is the number of branches per capita in the county. C3 is the share of deposits of the three largest banks in the county. HHI is the Herfindahl-Hirschman index and HHI ratio accounts for the market share of banks in the county. Domdep is the sum of the dollar amount of banks' branch domestic deposits by county. Except for DebtoIncome, Unemployment and C3, all variables are logarithmic and is taken logarithm before they are applied. Variables are winsorized at the 1 st and 99 th percentile borrowers, payment information of funded loans, status of loan and all loan application details from 2009 to 2012 is displayed in Table 4. After data is collected for the bank, county and alternative P2P lending variables, we merged the three datasets into one. To find treated bank and county and control bank and county unique 5-digit zip code is used. However, although bank and county small business data is provided with a 5-digit zip code level, the alternative P2P lender data is identified at the 3-digit zip code level. In order to evaluate alternative P2P lender activity in treated and control counties, county-level and alternative P2P lender data are merged according to this unique 3-digit zip code.
Method
We use a method that allows us to look at the impact of the regulation at a county level, following the approach taken by Tang (2019). We then apply a difference-in-differences (DiD) approach to obtain our empirical results.
We limited the research period so that the 2008 global financial crisis does not affect the data set exogenously. Our sample period starts after this date and due to using policy change in 2010 as an exogenous shock in our research method, we kept sample period limited to 4 years between 2009 and 2012 in order to mutually coincide the pre and post Table 4 Descriptive statistics for alternative P2P lender at county-level This table presents the summary statistics of alternative small business lender Lending Club. According to the Lending Club dictionary, the main dependant variable P2PSBL is the logarithm of amount of small business loan volume. Term is the payment numbers on loan. Int_rate is the interest rate on loan. Loan_status is a dummy variable and set to 1 if charged off, set to 2 for a fully paid loan. Annual_inc is the annual income provided by the borrower. Dti is the "ratio calculated using the borrower's total monthly debt payments on the total debt obligations, excluding mortgage and the requested Lending Club loan, divided by the borrower's self-reported monthly income". Fico is the credit score of borrowers. Term,Fico and Annual_inc are logarithmic. The county control variables are described in Table 3 periods. 9 This sample was analysed in with a similar empirical method in Tang's (2019) article where the period is 2009Q1-2012Q2. After the research period was limited to this period, we performed parallel trend analyses to test the robustness of the analyses results, and the results were confirmed. In order to isolate the regulatory impact, we apply a negative shock at county level to supply of bank loans that leads banks to tighten their lending criteria. In this regard, we consider an arguably exogenous shock to bank small business credit supply that was due to the implementation of the Dodd-Frank Act in June 2010 which is described as the beginning point of the post-shock term. Using small business loan data at bank and county level in regard to the Dodd-Frank Act, we follow Tang 10 (2019) and De Roure (2022) analyses who find that treated banks reduced lending.
In order to provide causal evidence, the Dodd-Frank Act is used as an exogenous shock. The DiD model compares the volume of small business lending one year before and two years after July 21, 2010 (the implementation date of the Dodd-Frank Act). The treatment group are banks that are affected by this regulation and control group are banks that are not affected.
We cannot completely exclude the possibility that time-varying, unobserved market variables, even with the "DiD" technique, simultaneously affect the development of Fintech loans and the position of traded banks before the shock. To alleviate this problem, we present in Fig. 2 findings that show a parallel trend of FinTech lending in both traded and non-traded markets before 2010Q2. We also show that the benefits of treatment began to take effect in the second quarter of 2010. Given the date of the Dodd-Frank, Figure 2 shows the trend of the annual mean values of small business loan volume of treated and control banks before and after the introduction of the Dodd-Frank Act. Data Source: FFIEC we also examine the impact of other additional regulations in the robustness section, it seems unlikely that other variables are responsible for this trend.
There are two cut-offs for financial institutions according to the Dodd-Frank regulations. The first one is for banks which are exceeding $10 billion in assets that subject to annual stress test and higher disclosure requirements. And the other is one for bank holding companies that are exceeding $50 billion in assets (called "systemically important banks") that subject to semi-annual stress tests and a far-reaching list of disclosure requirements. However, due to having limited data about bank holding companies, we could not include systemically important banks in the DiD model, which are exceeding $50 billion in assets; therefore, we only use $10 billion as a cut-off and therefore could not apply alternative method Regression Discontinuity.
Firstly, by using equation one, we test and analyse the qualification of existing research related to the Dodd-Frank Act impact on bank level small business lending activity.
where log(SBLoan) i,t is originated small business loans (origination volume $1 million or less) by bank i in year t. Treated i is a dummy variable that identifies the treatment group, one if the banks with assets over $10 billion threshold which are subject to the Dodd-Frank Act and zero for the banks with assets right below $10 billion threshold and exempted from Dodd-Frank Act. DFA t is the treatment dummy that takes the value one from Dodd-Frank Act enactment date (21 th July 2010), and zero prior for this date. C i,t is a vector of bank-level control variables are defined in Table 2. θ t is a variable for the county-year fixed effects and i is a variable for bank fixed effects, and both are used to help remove unobserved heterogeneity such as variation in local loan demand due to (county-specific) business conditions and for unobservable bank characteristics. ǫ i,t is an error term.
The four columns of Table 5 report the Dodd-Frank Act impact on bank small business loan volume. According to results, the coefficient of interaction term, Treated i x DFA t, is negative and highly in all estimations with bank, county and year fixed effects. The results show that small business lending volume in treated banks decreases.
In order to check traditional banks' responses to Dodd-Frank Act in the counties for evaluating small business loan applications, we use the following equation: where SBLoan t,c is originated loans to small businesses(loans origination volume $1 million or less) in county c in year t. Treated c is a dummy variable that identifies the treated counties and takes the value of 1, if there is a bank with $10 billion assets or over affected by the Dodd-Frank Act and there is low competition according to the C3 and HHI, which are in the top 75th.
If the county has a bank asset below $10 billion, and there is high competition in the bottom 25th, it is defined as a control county and takes 0. Counties other than the 75th and 25th percentile are not included in the model. DFA t is the treatment dummy that takes the value one from Dodd-Frank Act enactment date (21th July 2010), and zero prior for this date. C t,c is a vector of county-level control variables. δ c variable for the (1) county fixed effect, and γ t is a variable for time fixed effect. ǫ t,c is an error term. The county level variables are defined in Table 3. Table 6 reports the Dodd-Frank Act's effect on county small business lending activity. The first column shows the result for the aggregated small business loan activities county and columns 6 and 9 show the small business loan for businesses with gross revenues less than $1 million and for businesses with gross revenues of at least $1 million, respectively.
According to results, the coefficient of the interaction term, Treated c x DFA t is both negative and high in all predictions with county and time fixed effects. The results show that small business lending in treated counties decrease relative to control group counties after the Dodd-Frank Act in terms of aggregate small business loan and for businesses with gross revenues less than $1 million, respectively. There is no significant impact on for businesses with gross revenues of at least $1 million. Table 5 shows the difference-in-differences estimation results in Eq. (1). The dependant variable SBLoan is the bank level total loan volume. The variable Treated takes on the value 1 for the banks with assets over $10 billion and zero for the banks with assets right below $10 billion. DFA is the treatment dummy that takes the one from July 2010 onwards and zero prior to that date. Standard errors are clustered at the bank level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses In order to check if alternative lenders increased their lending in counties where small business lending decreased due to the credit supply shock's effect on small business loan applications, we use the following equation: where SBLoan P2P t,c is small business loan origination volume of alternative lenders loan in county c in year t. Treated c is a dummy variable that identifies the treated counties and takes the value of 1 if there is a bank with $10 billion assets or over and affected by Dodd-Frank Act and there is low competition according to the C3 and HHI, which are in the top 75th. If the county has a bank asset below $10 billion exempts from the Dodd-Frank Act and there is high competition in the bottom 25th, it is defined as a control county and takes the value of 0. Counties other than the 75th and 25th percentile are Table 6 Impact of Dodd-Frank Act on aggregate county-level small business lending Table 6 shows the difference-in-differences estimation results in Eq. (1). The variable Treated takes on the value 1 for the counties where there is a bank with $10 billion assets or over affected by the Dodd-Frank Act and there is low competition according to the concentration of the three largest banks (C3) and Herfindahl-Hirschman Index (HHI), which are in the top 75th. If the county has a bank asset below $10 billion, and there is high competition in the bottom 25th, it is defined as a control county and takes 0. Counties other than the 75th and 25th percentile are not included in the model. DFA is the treatment dummy that takes the one from July 2010 onwards and zero prior to that date. There are three dependant variables. SBLoan is the county level total small business loan volume. SBLoan1 is a total small business loan for businesses with gross revenues less than $1 million and SBLoan2 is a total small business loan for businesses with gross revenues of more than $1 million. Standard errors are clustered at the county level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses
Total small business loan volume Small business loan for businesses with gross revenues less than $1 million
Small business loan for businesses with gross revenues more than $1 million (1) (2) (3) (4) (5) (6) (7) (8) not included in the model. DFA t is the treatment dummy that other takes the value one from Dodd-Frank Act enactment date (21 th July 2010), and zero prior for this date. C t,c is a vector of county-level control variables. δ c variable for the county fixed effect, and γ t is a variable for time fixed effect. ǫ c,t is an error term. All variables are defined in Table 4 with loan-level variables. We acknowledge that it is not clear whether the DiD coefficient of this regression reports the effect of Dodd-Frank exposure (the main point of our paper) or the effect of bank concentration (unrelated to the paper). That said, we emphasize that the high concentrated counties with low competition are exposed to more regulatory impact and that this in turn should result in an advantage to P2P lenders in the less concentrated counties.
The main dependant variable measures lending volume of the alternative P2P lender data that we used the dollar amount of alternative P2P lender origination volumes from the loan book that is specified at the county level. Due to having limited county-level Table 7 Impact of Dodd-Frank Act on aggregate alternative P2P lending Table 7 shows the difference-in-differences estimation results in Eq. (3). The dependant variable P2PSBL is the small business loan origination volume of alternative lenders in counties. The variable Treated takes on the value 1 for the treated counties where there is a bank with $10 billion assets or over and affected by Dodd-Frank Act and there is low competition according to the C3 and HHI, which are in the top 75th. If the county has a bank asset below $10 billion exempts from the Dodd-Frank Act and there is high competition in the bottom 25th, it is defined as a control county and takes the value of 0. Counties other than the 75th and 25th percentile are not included in the model. DFA is the treatment dummy that takes the one from July 2010 onwards and zero prior to that date. Standard errors are clustered at the county level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses data, instead of using normalized 11 variables similar as in Tang (2019) paper, the logarithm value of the small business loan origination is used in the analysis. The results of Eq.
(3) are presented in Table 7. It is proved that in regard of control counties, loan origination volume of alternative P2P lender enhanced remarkably in treated counties after the Dodd-Frank Act became law in July 2010, in terms of the total loan amount. According to our results, there was a notable difference, between control and treated counties, in alternative P2P lender loan volume after the enactment of Dodd-Frank. The trend after the Dodd-Frank Act proves that the growth in demand for alternative credit between control and treated markets is unlikely to be urged by observable differences.
In accordance with Table 4, we find that treated counties experienced an increase in alternative P2P lender mall business loan applications compared to control counties.
This result is coherent with FinTechs' and banks being substitutes or complements with the findings of Tang (2019). However, this analysis is necessary for validating the Dodd-Frank Act as a negative shock to incumbents' small business loan supply. We emphasise the limitation to our approach is the restricted data available on alternative lenders. To sum up, the results on the volume of alternative P2P lender loans reveal that, when incumbents cut lending in the small business credit market, some borrowers tend to move from incumbents to alternative P2P lenders.
To check the parallel-trends assumption, we present Fig. 2, which shows lending by banks overtime for the treated and control group.
The Fig. 2 shows that in treated states, new small business loan volume is similar to that in control states before the Dodd-Frank Act. This indicates that the parallel-trends assumption is valid. After the Dodd-Frank Act, the new small business loan volume decreased both for treated and control banks, but it decreased more and faster in treated counties than in control counties which are presented in Fig. 3.
Similarly, we check the parallel-trends assumption with an alternative P2P lender. Figure 4 shows an alternative P2P lender credit provision in treated and control counties. It shows that the volumes of new alternative P2P lender loans to small businesses in control and treated counties displayed parallel trends prior to the Dodd-Frank Act. After the Dodd-Frank Act, P2P small business lending increased in treated counties.
Robustness and additional tests
As a robustness check, we also conducted the difference-in-differences analysis for a restricted 2009-2010 period. By reducing the research period, we compare the predicted treatment and whether the parallel trend assumption is violated. The results are even more significant. At both the county bank lending level and the individual bank level, we have an even bigger negative coefficient for the interaction term: treatedb _ EBAt, and this coefficient is always significant at the 1% level except county level analysis results for the small business loan for businesses with gross revenues more than $1 million. Detailed results are reported in Tables 8, 9 and 10.
We also conduct the main analysis conditioned on the bank-and county-year-fixed effects and various bank characteristics, with concurrent shocks that impose disparate effects on small business lending and the control banks. As part of this, two major potential coincident changes are examined. Collectively, these tests mitigate concerns regarding omitted concurrent shocks that drive the primary result.
Next, the Troubled Asset Relief Program (TARP) 12 was evaluated. TARP introduced by the U.S. government through the Emergency Stabilization Act (2008) to respond to the global financial crisis (Cornett et al. 2013). The TARP was planned to stabilize the Figure 4 shows the trend of the annual mean values of small business loan volume of Alternative P2P lenders in treated and control counties before and after the introduction of the Dodd-Frank Act. Data Source: Lending Club financial system by purchasing troubled assets from banks to inject liquidity into the financial system, and reactivate the credit markets (Harris et al. 2013).
According to Black et al. (2013), it was expected from the TARP to increase the lending of participating banks in the initial funding program. In this regard, Li (2013) finds evidence that TARP banks significantly increased bank loan supply. In addition, Berger et al. (2019) and Chu et al. (2019) document that banks increased credit supply to businesses by way of TARP capital injections. However, Cole and Damm (2020) find no evidence that the TARP program increased lending and claim that non-TARP banks reduced lending less than TARP recipient banks.
During the research period, we note that it is possible that control banks received more government aid from TARP after the financial crisis. They would therefore extend more credit to small businesses relative to treatment banks. To test the impact of this we used the period 2009-2012 (TARP participation it ). This variable is equal to one and zero otherwise is created as a new one and interact this variable with Table 8 Robustness results for bank level data Table 8 shows that by limiting the research period to one year before and after treatment, there is no change in banks' small lending activity and the effect of the Dodd-Frank Regulation is still significant. Standard errors are clustered at the county level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively. Year2010 + , and then added to the regression. The results are shown in Table 11. small business lending continues to load (two-tailed p-value < 0.01). We also reviewed a non-TARP program, the Small Business Lending Fund (SBLF), which was passed by U.S. Congress and signed into law in 2010 (Wilson 2013). The SBLF was created as part of the Small Business Jobs Act to encourage liquidity in the interbank lending market and intended to provide low-cost funding since, therefore, banks could lend to small businesses (Berger et al. 2020). Balla et al. (2017) claim that participants in the SBLF program were well-capitalized and healthier financially so that after two-quarters of the start of the SBLF program, SBLF participated banks experienced stronger aggregate growth in lending to small firms. In contrast, Basset et al. (2020) find evidence that there was not any difference between the loan growth of participated and non-participating banks in government financial aid program.
To test the impact of SBLF, we create an indicator equal to one if a bank is participated (SBLF participation it ), and zero otherwise, and interact this variable with Year2010 + . After adding this interaction to the regression, unlike TARP, we find a significant coefficient on small business lending continues to load (two-tailed p-value < 0.01) in the second column of Table 11. Table 9 Robustness results for county level data Table 9 shows that by limiting the research period to one year before and after treatment, there is no change on countylevel small business lending and the effect of Dodd-Frank Regulation is still significant except for small business loans businesses with gross revenues of more than $1 million. Standard errors are clustered at the county level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses A limitation of our approach is the limited-time sample. Parallel trends cannot be strongly verified if there is only one time period in the pre-period. Without strong evidence of parallel trends, it is difficult to assume that the treatment and control counties would have seen a similar credit growth after the regulation. Treatment counties had larger banks and a more concentrated banking environment. Such counties were also disproportionately exposed to the housing crisis since larger banks had higher MBS exposure. It is plausible that lower credit growth is an artifact of the damage caused by the crisis. A larger time sample would help address such concerns, but this was simply not available.
We observe that our results are consistent with Cortés et al. (2020) analysis of the way in which the Dodd-Frank Act acted on banks at the local level. They suggest that affected locals raise interest rates to compensate for the capital burden imposed by the stress test element. This gives an advantage to P2P lenders because banks reduce small business loans that are more like commodities as that leads borrowers to switch.
Table 10
Robustness results for alternative P2P lending Table 10 shows that by limiting the research period to one year before and after treatment, there is no change on alternative (P2P FinTech) lending for small businesses and the effect of Dodd-Frank Regulation is still significant. Standard errors are clustered at the county level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses (1) (2)
Conclusion
This paper investigated how innovative lending models gained a regulatory advantage over traditional banks, particularly in respect of loans to small businesses. We developed an empirical model for bank, county, and innovative P2P lending. We separately tested the impact of a negative regulatory shock on small business lending. We examined two main hypotheses. First, we investigated new regulations' impact on county-level small business loan origination at traditional banks. We found that in treated counties where there was a bank with $10 billion assets or over and affected by Dodd-Frank Act, and where there was low competition according to the C3 and HHI, there was a decrease in the small business loan volume according to control counties. We conclude that unexpected regulatory reform like the Dodd-Frank Act has led regulators to make changes that impact financial institutions, especially banks, and may cause them to reduce their lending to small businesses.
Second, we examined whether innovative P2P lenders increase their lending in counties where small business lending decrease due to the credit supply shock's effect on small business lending. The analysis shows that alternative P2P lender volume of loan origination rose considerably in treated counties after the Dodd-Frank Act became law. This shows that there was a regulatory advantage.
We conclude that policy makers should consider whether the regulatory advantage is equitable and/or desirable. Clearly, FinTech lenders can be regulated like traditional banks, but they would then lose this regulatory competitive advantage. Our contribution is in Table 11 Tests for successive shock Table 11 shows the result of additional tests for bank small business lending volume. In column 1, TARP is an indicator equal to one if a bank or its affiliated holding company participates in the TARP program and zero otherwise for years 2009-2012. In column 2, SBLF is an indicator equal to one if a bank or its affiliated holding company participates in the SBLF program and zero otherwise between 2009 and 2012. In column 3, all two potential successive shocks are controlled for. Standard errors are clustered at the bank level and shown in parentheses. Statistical significance at the 10%, 5% and 1% levels is denoted by * , ** and *** , respectively.t-statistics are presented in parentheses showing how the lack of regulation gives FinTech lenders a comparative advantage over traditional banks. The important implication of our paper's findings is that higher capital requirements and regulatory enforcement on banks may lead regulated banks to reduce their loans to small firms and thereby providing an opportunity for P2P lenders to grow market share. | 2022-08-15T14:05:05.844Z | 2022-08-15T00:00:00.000 | {
"year": 2022,
"sha1": "7cd41d23d3884404d568ec781a5e23acaaa1badd",
"oa_license": "CCBY",
"oa_url": "https://jfin-swufe.springeropen.com/track/pdf/10.1186/s40854-022-00377-y",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "a150c99b7f0614ad7e53ef59d95fae45dfe04b01",
"s2fieldsofstudy": [
"Economics",
"Business"
],
"extfieldsofstudy": []
} |
58360020 | pes2o/s2orc | v3-fos-license | Personality and Learning Styles of Final-Year Medical Students and the Impact of these Variables on Medical Specialty Choices
Objective: Medical student profile is changing on campuses today and there is a much greater variation in the range of personality type and learning style preferences to be considered. In this study it is aimed to determine the learning styles of medical school students at Gazi University and to find out whether there is any relationship between students’ personality types, learning style preferences and their medical specialty choices. Methods: The study was conducted on 170 final year students (96.6%) at the Gazi University School of Medicine in the 2011-2012 academic year. The authors used Myers-Briggs Type Indicator (MBTI) to determine the personality traits and Grasha Riechmann Student Learning Style Scale (GRSLSS) to establish the learning styles. Results: During the study, 91.8% of the students declared that they wanted to be a specialist and 4.1% wanted to be a general practitioner in the near future. Most preferred specialty appeared to be dermatology (11.2%) in whole group. Choices of female students who want to be a specialist were dermatology, ophthalmology and obstetrics and gynecology and their distributions were 22.8%, 13.0% and 9.8%, respectively (p<0.05). The most common personality type in all preferred specialty areas was Introverted Sensing Thinking Judging (ISTJ). The students with ISTJ personality type had a higher score on the collaborative and competitive learning style. Conclusion: Last-year medical students are characterized by a ISTJ personality type in most of the medical specialty preferences. Furthermore, these students have collaborative and competitive learning styles.
INTRODUCTION
Medical education is a difficult and life-long process both in undergraduate and postgraduate levels. The students are expected to demonstrate competency in areas of technical skills, team working and lifelong learning skills beyond medical education before and after graduation (1,2). In addition to these, the selection of medical specialty is as important as the educational process. The selection of medical specialty is based on many determinants such as personal (e.g., personal and learning styles), cultural, national and international values, academic achievements, finances, lifestyle and role models. Although many factors can influence this choice, personal features may play a stronger role in their specialty choices.
There are many kinds of personality measures used with medical students (3). One of the widely used ones is Myers-Briggs Type Indicator (MBTI). It has been used for assessing personality types of people for decades and hundreds of studies over the past 40 years have proven the instrument to be both valid and reliable. It has been also used for medical students in many studies (4)(5)(6)(7)(8)(9)(10)(11). The instrument addresses the two related goals in the development and application of the MBTI instrument: 1. The identification of basic preferences of each of the four dichotomies specified or implicit in Jung's theory. 2. The identification and description of the 16 distinctive personality types that result from the interactions among the preferences. The instrument evaluates the individual's favorite world, way of processing information, and way of making decisions and structuring the outside world. A four-letter personality type code results from how the questions are answered along four dichotomies (12).
The Grasha-Riechmann Student Learning Style Scales (GRSLSS) is designed specifically for use with senior high school and college/university students and it focuses on how students interact with the instructor, other students, and with learning in general (13). GRSLSS promotes an optimal teaching/learning environment by helping the faculty design courses and develops sensitivity to the students' needs (14).
Grasha-Riechmann student learning style model, which describes six dimensions of an individual's learning style, was developed in the early 1970s. The learning styles have been defined as personal qualities that influence a student's ability to acquire information, to interact with peers and the teacher, and otherwise participate in learning experiences (13).
Many studies have been conducted comparing specialty choices and the personality (4-11) but limited number of them compares the relationship between personality and learning styles as they impact students' choices.
In this study it is aimed to determine the categorization of the personalities and learning styles of last year medical students (interns) and the relationship of these factors with students' medical specialty choices.
Thus in this study researchers focused on the following questions: 1. What are the personality types of the last year (intern) students at Gazi University Medical School suggested by Myers -Briggs? 2-Is there a significant difference between personality types and sexes? 3-What are the learning style preferences of the last year (intern) students at Gazi University Medical School in terms of six dimensions suggested by Grasha-Riechmann? 4-Is there any significant relationship between learning styles and the medical specialty of choice?
Participants
In Turkey, upon completion of the medical degree, doctors may be qualified to train in one of the specialties of their choice for residency based on their performance on a Central National Residency Matching Examination.
The research was conducted at Gazi University School of Medicine in Ankara, the capital city of Turkey, in the academic year of 2011-2012. At the end of the 2012 semester a total of 176 students had graduated, and 170 of them (96.6%) (55.1% female, 44.9% male) have participated in the study. The forms were given to final year students and were filled under the supervision of the researchers.
Instruments
Research data form included MBTI and GRSLSS. The MBTI is an instrument that identifies a person's preferences for gathering data, processing information, and making decisions using four dichotomous scales (6). The first dichotomy identifies whether a person is energized by the outer world of people and things (Extraversion) or the inner world of ideas and experiences (Introversion). The second dichotomy identifies a person's preference for gathering information using their 5 concrete senses (Sensing) or by a sixth sense or "hunch" that allows them to recognize patterns and possibilities (Intuition).
The third dichotomy focuses on whether people prefer to make judgments and decisions based on logic and objective data (Thinking) or based on personal values and subjective data (Feeling). The fourth dichotomy identifies whether a person prefers to achieve closure and have things decided (Judging) or whether a person prefers to continue to consider options (Perceiving) rather than reaching a closure. The four MBTI dimensions analyzed are Extroversion-Introversion (E-I), Sensing-Intuition (S-N), Thinking-Feeling (T-F), and Judging-Perception (J-P). Based on the individual's responses to the questions, a four-letter personality type was generated which consists of four dimensions (eg, ESFP, INFJ). The validity and the reliability analyses of the Turkish version of MBTI had been conducted elsewhere (15).
The GRSLSS promotes understanding of learning styles that have six categories: independent students, dependent learners, competitive students, collaborative learners, avoidant learners, and participant learners. The GRSLSS was made up of 60 items that can be answered on a 5-point Likert Scale and has six sub-scales with 10 items on each scale. The participants are grouped into low, moderate and high on each of the subscales (Table 1). Turkish validity and reliability studies of the scale of GRSLSS had been conducted elsewhere (16). Table 1. Low, Moderate, and High score definitions based on the norms of each learning scale of GRSLSS (without dividing scores by 10) (4).
Low
Moderate Statistical analyses Data were analyzed using SPSS for Windows, version 16.0 (SPSS Inc, Chicago, Illinois). Chi-square analyses were used to statistically evaluate significant differences in the medical specialty choices, personality profiles of the students and gender. Kruskal-Wallis test was used to statistically evaluate significant differences in the personality profiles of the students and their learning styles. All tests were considered to be statistically significant when p<0.05.
Specialty, sex and personality
Students' choices in specialties were determined by asking them "Do you want to study as a general practitioner or a specialist? If a specialist, please indicate." While 44.9% of students were men, 55.1% were women and 91.8% of the students declared that they wanted to be a specialist and 4.1% wanted to be a general practitioner. The most preferred specialty expressed by students was dermatology (11.2%) ( Figure 1). Table 2 shows that 22.8% of female students will seek a residency in dermatology, 13.0% in ophthalmology and 9.8% in obstetrics and gynecology (p<0.05). 11,2 8,4 7,9 6,1 6,1 6,1 5,6 5,6 5,6 In examining MBTI personality types by sex, ISTJ was the most common type (41.9%), and no difference was detected between the sexes (Figure 2).
The most common personality type was ISTJ in all specialty areas. There were statistically significant differences in MBTI types between decided and undecided students and students whose choices are psychiatry (p<0.05). While 31.2% of the students who preferred psychiatry were in the category named "others", 23.1% of the undecided students had ISFJ personality type. In the detailed analysis, the students who preferred psychiatry had INFJ personality type (p=0.001). Among the students who prefer radiology, the 11.1% have an ENFP type (p>0.005) ( Table 3).
Specialty and learning style
The students who sought further training in ophthalmology displayed statistically significant lower scores on the avoidant learning style and higher scores on competitive learning style as compared to the other students who were not willing to undergo training in this area (p<0.05). On the other hand the medical students with a special interest in cardiology had statistically significant lower scores on avoidant learning style and higher scores on participant learning style (p<0.05). There were no statistically significant differences between learning styles in other specialty areas (Not presented in the table).
Personality and learning style
According to score definition presented in Table 1, the interns who have ISTJ personality type had high collaborative and competitive learning styles scores, moderate independent, avoidant, dependent and participant learning style scores. The interns who have other type of personality have statistically higher avoidant learning style scores than ISTJ, INFP, ENFP, ISTP, and ISFP personality types(p<0.05) ( Table 4).
DISCUSSION
Although this study is based on a single survey in the Faculty of Medicine, it has been informative in terms of revealing the personality and learning style factors affecting medical students' decision to choose a certain specialty in Turkey.
The specialties mostly preferred in previous years, such as plastic surgery, general surgery, cardiovascular surgery, and pediatrics are not listed in the first choices of students in recent years and were replaced by less-risky areas in terms of malpractice likelihood such as dermatology, radiology, and psychiatry. In our study, while male students do prefer radiology, ENT, and psychiatry, female students seek training in dermatology, and ophthalmology. There may be many factors influencing the decision making in choosing a specialty. We have focused on personality types and learning style factors. The medical malpractice law, which has been enforced for the last couple of years in Turkey, may be an influencing factor in deciding on less risky specialties.
Using MBTI, our last year medical students are characterized as Introverted-Sensing-Thinking-Judging (ISTJ) types. The most common MBTI styles for the students in our study correspond nicely to the most common preferences found in other studies conducted with medical students and medical residents (6)(7)(8)(9). Individuals who are ISTJ types are characterized as quiet and serious. They earn success by thoroughness and dependability. They are practical, matter-of-fact, realistic, and responsible. They decide logically what should be done and work toward it steadily, regardless of distractions. Also they take pleasure in making everything orderly and organized -their work, their home, and their life (4). Similar to our study, another study found that female physicians had significantly higher sensing components as compared to their male colleagues (10). In addition, a study that analyzed the changes in MBTI types and medical specialty choices over time reported that the proportion of feeling types was the highest and the most permanent among women (11).
Alltogether, specialties and MBTI types revealed that ISTJ was the most common personality type in all specialty areas. However, ISFJ type was significantly higher among emotionally unstable persons as compared to other groups. The difference is related to the feeling component of undecided students. Unstable behavior may be related to dominant feelings. The results of the study by Stilwell et al. (11) revealed that there is a shift towards judging type over the years among doctors due to an increase in technology and knowledge in all fields of medicine. This study also demonstrated that the physicians used perceiving skills more frequently in examination and diagnosis in the 1950s, but today doctors order tests and interpret the results, rather than relying on more inductive processes.
The majority of our study group consisted of ISTJ type students, and their competitive and collaborative learning style scores are in the high category. The ''competitive" learner is classified as a student who learns material in order to perform better than others in the class. They prefer teacher-centered instructional procedures (13). Indeed, the medical students often study the lecture notes for examination by spending more time on the important parts. People with introverted personality type learn with internal reflection and distill one's thoughts independently (9). Students with collaborative learning styles feel they can learn by sharing ideas or talents. They cooperate with the teacher and like to work with others (13). A study done at the same setting with a different student group revealed high scores in competitive and collaborative learning styles as in our study (17). The students with collaborative learning style are eager to learn and take responsibility for the process of learning. These students are very curious and hands-on (13). This is also a specific feature of thinker persons who are the majority of our study group. According to MBTI thinkers are logical, reasonable, questioning, critical, and tough (18).
Although almost all of the last year students in our institution participated in the study, we have compared the personality preferences and learning styles of students in a single institution. This is the most important limitation of this study.
CONCLUSION
Last year medical students are characterized as Introverted-Sensing-Thinking-Judging types in most of the medical specialty choices. The students having this type of personality have collaborative and competitive learning styles. Although the graduates of medical schools receive the Medical Doctor degree, professional counseling may be beneficial in their career planning. Mentoring for specialty choices of last year medical students including personality tests and learning styles may help them have a better near future. Therefore, the establishment of Career Counseling Centers in schools of medicine may be useful.
It is important to keep in mind that the results should not be used in isolation. Medical students and educators are cautioned against over-valuing personality types in the career selection process. To maximize learning, faculty should provide guidance in a manner that allows all students to use or express their individual preferences toward understanding, appreciating, and applying skills. | 2018-12-11T12:35:04.063Z | 2014-10-26T00:00:00.000 | {
"year": 2014,
"sha1": "80a0ff0fdd9902e8c389c25dcfd8b93ff2ea9195",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.12996/gmj.2014.43",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "80a0ff0fdd9902e8c389c25dcfd8b93ff2ea9195",
"s2fieldsofstudy": [
"Medicine",
"Education",
"Psychology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
149446175 | pes2o/s2orc | v3-fos-license | Near-Field Manipulation in a Scanning Tunneling Microscope Junction with Plasmonic Fabry-Pérot Tips
Near-field manipulation in plasmonic nanocavities can provide various applications in nanoscale science and technology. In particular, a gap plasmon in a scanning tunneling microscope (STM) junction is of key interest to nanoscale imaging and spectroscopy. Here we show that spectral features of a plasmonic STM junction can be manipulated by nanofabrication of Au tips using focused ion beam. An exemplary Fabry–Pérot type resonator of surface plasmons is demonstrated by producing the tip with a single groove on its shaft. Scanning tunneling luminescence spectra of the Fabry–Pérot tips exhibit spectral modulation resulting from interference between localized and propagating surface plasmon modes. In addition, the quality factor of the plasmonic Fabry–Pérot interference can be improved by optimizing the overall tip shape. Our approach paves the way for near-field imaging and spectroscopy with a high degree of accuracy.
FIB fabrication of the Au tips
The FIB fabrication of electrochemically-etched Au tips was performed an FEI Helios NanoLab G3 FIB-SEM DualBeam system and Ga ions were used. It provides Gallium ions with an acceleration voltage up to 30 kV and enables milling and deposition of structures with critical dimensions of less than 10 nm. The tips were mounted on a pre-tilted specimen holder and oriented towards the ion beam such that the axis of the tip is collinear with the ion beam.
We performed a multiple-step annular milling process with different parameters and a varying order, depending on the initial tip profile and the required sharpened length. Taking into account the re-deposition dynamics during ion milling and the fact, that the main re-deposition of removed material occurs underneath the ion beam incidence point at highly steep surfaces, we preferred an inner-to-outer-radius scan direction, i.e., from the apex downwards, in order to simultaneously remove re-deposited material.
Several tip fabrication procedures using FIB milling were already suggested before. 1,2 Here we introduce a similar procedure with a reverse milling sequence and different parameters.
We first start the tip fabrication with a low ion energy step at 5 kV in order to shape and smoothen the apex. Subsequently, we increase the inner and outer radii as well as the ion energy and beam current to obtain higher milling rates to only sharpen the shaft to the required length.
This step order aims to minimize damage and Gallium implantation into the apex and in close vicinity to it. Finally, we apply a low energy polishing step at 5 kV to the sharpened length of the shaft to reduce damage layer caused by 30 kV ions.
A key parameter in ion milling processes is the volume per dose value ΔV, i.e. the removed material volume per primary ion. It is material-specific and determines, at a certain ion energy, the milling time t for a required depth Z and a given beam current I: For the groove fabrication and in order to precisely control the groove depth Z and the milling time t, we set this value to 1.5 µm³/ nC and 0.42 µm³/ nC for Au and Ag, respectively, 3 and used, at 30 kV ion energy, the lowest ion beam current of ca. 7 pA to avoid excessive milling and thus damaging the apex. Figure S1 shows Multi-peak fitting analysis of the bias voltage-dependent STML spectra in Fig. 2c. It appears that a slight blue-shift is observed at long wavelengths (peak number 4 and 5). As shown below (Fig. S2), the peak positions remain almost identical in the currentdependent STML spectra where the tip-surface distance is also changed. Therefore, the blueshift in the voltage-dependent STML spectra may not be related to the gap distance. As seen in Fig. 1 (the voltage-dependent STML spectra for a non-groove tip), the applied voltage determines the quantum cutoff of the LSPR excitation and affects the overall spectral response.
Peak position analysis in the bias voltage-dependent STML spectra
The different spectral feature of the LSP may lead to different coupling to the propagating SPP modes.
Current dependence of the STML intensity for the grooved tip
The luminescence intensity at 615 nm for the 3-μm grooved tip (Fig. 2d) is plotted as a function of the tunneling current, indicating a linear dependence. Figure S2. Peak intensity at 614 nm of the STML spectra in Fig. 2d is plotted as a function of the tunneling current. The black line indicates the linear fitting result.
Numerical simulation of electromagnetic field distributions
Numerical simulations were performed to calculate the plasmonic response of the nanotip and STM-junction by solving the time-harmonic wave equation for the electric field within the RF-Module of COMSOL Multiphysics 5.3a. Simulations are performed in 3D for a groove distance of L = 3 µm (although the problem is fully radially symmetric, simulations needed to be performed in 3D as the point dipole excitation feature is not available in 2D radially symmetric models in COMSOL 5.3a). The structural parameters of the tip were taken from the SEM micrograph (Fig. 2b). The width of the simulation volume is 1.3 µm and the tip is truncated above the groove at 4 µm distance from the apex. Except for the 100 nm thick sample, which is chosen thick enough to not transmit any fields, the simulation volume is surrounded by perfectly matched layers to absorb all outgoing waves. | 2019-05-11T13:03:34.013Z | 2019-05-09T00:00:00.000 | {
"year": 2019,
"sha1": "f6fb1f140637e591400ac8ae932658dabca7fafb",
"oa_license": "CCBY",
"oa_url": "https://pubs.acs.org/doi/pdf/10.1021/acs.nanolett.9b00558",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "52556ad842cee4e9c05252a5c4708e73e811b0e2",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Medicine",
"Materials Science"
]
} |
250445244 | pes2o/s2orc | v3-fos-license | Analysis of Chloramphenicol in Shrimp Using Standard Addition Method Based on Diazotization
: This research aimed to determine the concentration of chloramphenicol in shrimp using the standard addition method based on the diazotization reaction using Zn powder as a reducing agent of chloramphenicol, followed by the use of N-(1-naphthyl) ethylenediamine dihydrochloride as a coupling agent and measured at 565 nm. Based on the test, the shrimp sample was found to contain 1964.91 g/kg of chloramphenicol and it exceeded the requirements set by the European Commission which was 0,15 g/kg. The limit of detection (LOD) value is 0.19 g/mL and the limit of quantitation (LOQ) value is 0.64 g/mL. The correlation coefficient (R 2 ) was 0.9991 for the concentration range of 0-50 ppm. The analysis of the results showed that the %recovery in shrimp analysis using this method was 87.41%-107.73% with an average of 109.38%.
INTRODUCTION
Due to Indonesia's status as one of the world's largest exporters of shrimp, shrimp is a valuable commodity in Indonesia's aquaculture. Production of farmed shrimp would reach 911.200.00 tons in 2020 [1]. Along with the development of shrimp farming, farmers are faced with several obstacles in producing errors such as diseases caused by Aeromonas hydophila which can be treated with antibiotics [2]. Chloramphenicol (CAP) is an antibiotic compound with strong antibiotic power and is active against both aerobic and anaerobic microbes. This antibiotic works by inhibiting the protein synthesis process of an organism, so it is widely used as a drug in aquacultures such as shrimp and fish to control the disease by increasing growth and production. The chloramphenicol structure is shown in Figure 1.
The use of CAP in aquaculture can cause accumulation and residue left in the tissue, organs, and flesh of these animals. Consumption of food products contaminated with chloramphenicol can cause allergic reactions for people with antibiotic allergies, decreased immunity, digestive tract disorders, and aplastic anemia. With several harmful effects of using chloramphenicol, the European Commission in 2019 set a minimum limit of chloramphenicol content in food of animal origin at 0,15 g/kg [4]. In addition, in Indonesia the use of CAP antibiotics is prohibited from being added as a food additive and is regulated in the Indonesia National Standard (SNI) 01-6366-2000 in 2000 [5]. Several studies shows that numerous techniques have been employed to determine chloramphenicol, such as Enzyme-Linked Immunisorbent Assay (ELISA), High-Performance Liquid Chromatography (HPLC), and Liquid Chromatography-Mass Spectrometry (LC-MS) [6]- [8]. Each of these methods has advantages, but these methods require a log time analysis and expensive, so there is a possibility of degradation, which causes a decrease in the concentration of CAP. So, an easy, fast, and accurate method of analyzing CAP is needed. This research was conducted based on diazotization process in which the colourless chloramphenicol has been change into colour substances, then was analyzed using spectrophotometer UV-Vis. This method was quite fast and easy as the colour can be seen directly after process. In the analysis of CAP in shrimp samples, it is necessary to do protein precipitation. Precipitation protein can be done by several methods, such as dialysis which removes the interfering substances but is not able to concentrate protein. In addition, organic reagents, Sodium Dodecyl Sulfate (SDS), and salts such as ammonium sulphate can also be used [9]. Ammonium sulphate has the ability to fractionate protein, so it is added salt to precipitate protein because it is able to maintain protein stability. However, the addition of this salt is required in high concentrations, so it is less effective. In addition, it can also be done with the addition of acids such as trichloroacetic acid (TCA). TCA is often used to precipitate proteins and purify protein from complex matrixes. [9]. The use of TCA was chosen because it is able to precipitate protein with the addition of TCA in small concentrations. Then it can be continued with the next stage of analysis. In this research, chloramphenicol in shrimp was analyzed using the standard addition method, which is more accessible, cheaper, and more accurate in detecting CAP content using UV-Vis spectrophotometric analysis based on diazotization reactions. The standard addition method was chosen because the sample is added to a certain quantity of standard solution whose concentration is known accurately so that the sample's matrix does not influence the analysis. The diazotization reaction was chosen because it is simpler, more accurate, and cheaper in detecting CAP by producing an intensely colored azo dye solution that produces a maximum absorption at a wavelength at 575 nm [10]. Ravisankar [11] analyzed azo dye at a wavelength at 550 nm. So in this research, the absorbance of azo dye solution was measured at a wavelength at 550-575 nm.
Research Methods Application of CAP in Shrimp
Application in shrimp begins with 250 g of shrimp washed and separated from the shell and mashed. In a beaker glass, 100 g of mashed shrimp are placed. 100 mL of TCA 15% was added to stand for one night. Separate the filtrate and the residue. 10 mL of the filtrate should be divided among three 50 mL volumetric flasks. Each volumetric flask was added with 1 mL, 2,5 mL, and 5 mL of a standard solution of CAP 500 ppm. Methanol was added up to the mark. Thus, concentration of 10 ppm, 25 ppm, and 50 ppm were attained. The same procedure was used to generate blanks without the addition of 500 ppm standard solution of CAP. Then, the filtrate is separated from the residue. Reduction of CAP 5 mL of sample filtrate was added with 2,5 mL of aquabidest, 2,5 mL of formic acid, and 0,3 g of Zn powder. Then it was stirred for 40 minutes with a rotation speed of 350 rpm. The solution was filtered. Formation an Azo Dye 7 mL of the reduced filtrate was pipetted. Added 1 mL HCl, 1 mL NaNO2 0,2%, and 1 mL ammonium sulfamate 0,5%. Each addition was vortexed and allows to be kept at a cold temperature. The solution was added with ethanol and 0,5 mL of 0,1% N-(1naphthyl) ethylenediamine dihydrochloride and kept at a cold temperature for 20 minutes to form an azo dye.
RESULT AND DISCUSSION Application of CAP in Shrimp
The first step in preparing shrimp samples was to precipitate protein from shrimp using a 15% TCA solution. This protein precipitation was carried out because CAP is relatively not bound to protein [12]. TCA was used to inactivate protease enzymes at low pH and modify the conformation of proteins so that proteins can precipitate and remove contaminants like salts and polyphenols [13]. As a result of the protein's isoelectric point, the addition of acid can result in changes in pH that change the protein's structure. Proteins become positively charged when acid is added because the amino groups on the protein capture protons and the pH decreases. Therefore, precipitation results from an imbalance in the protein structure [14]. The addition of strong acids such as TCA can trigger strong protein-protein interactions, as opposed to protein-solution, resulting in precipitation [15]. TCA 15% (w/v) is used because TCA concentration between 5%-45 % (w/v) is the most optimal for precipitating protein [14]. The photograph before and after addition of TCA shown in Figure 2. The obtained aliquots were then added to methanol. Methanol is used because CAP is highly lipid-soluble, and lipids are easily soluble in polar solvents such as methanol, ethanol, and chloroform. Additionally, methanol is a chloramphenicol solvent [12]. Methanol was employed to eliminate the remaining lipid interferences in the extracts by precipitation and it showed the highest lipid removal efficiency up to 97% [16]. Therefore, methanol is needed to draw CAP in aliquots and eliminate the lipid from the extracts.
Reduction of CAP
The purpose of the reduction procedure is to convert the nitro group into a primary amine group. The formic acid solution is used because it produces high sensitivity, which is optimal for reducing the NO2 group in CAP to N-H groups [17]. The mechanism of the chloramphenicol reduction reaction is shown in Figure 3.
Formation an Azo Dye
In the subsequent step, a solution of sodium nitrite (NaNO2) and hydrochloric acid (HCl). is added to the reduced CAP. When sodium nitrite and hydrochloric acid react, nitrosonium ions and water are produced. Furthermore, the addition of ammonium sulfamate was carried out and did not affect the color intensity of the coupling reaction can reduce the excess sodium nitrite required for the reaction [19]. The nitrosonium ion then will react with reduced chloramphenicol to form a diazonium salt. All additions were stored at a cold temperature. This is due to diazonium salt's poor thermal stability, which requires low temperatures at 0-5C [18]. The mechanism of diazonium salt formation is shown in Figure 4. . Reduction reaction mechanism [18] . Figure 4. The mechanism reaction for the formation of diazonium salt [18]. Diazonium cations react with N-(1-naphthyl) ethylenediamine dihydrochloride (NEDA) through electrophilic substitution, with diazonium salts as electrophiles to produce the purple azo dye shown in Figure 5. Purple-colored azo compounds can be identified by spectrophotometry at the maximum absorption wavelength of 565 nm. NEDA is used because it is highly sensitive, can be used in acidic solutions, and is widely used as a reagent for determining samples containing primary amine groups [11], [20]. The mechanism of formation of azo compounds is shown in Figure 6.
Figure 5. Azo dye from chloramphenicol
The results of the study on determining the content of CAP in blank shrimp samples obtained from the traditional market in Surabaya found the content of 1964.91μg/kg of CAP. The results of this study indicate that the sample of CAP in shrimp exceeds the European Union's minimum limit for CAP content which was 0.15 μg/kg. This indicates the sample does not meet the requirements for consumption. Based on the findings of this study, the agency in charge of food supervision should supervise the distribution of food more closely, and the consumer should be more selective in their food selections. Based on previous research, no CAP content was found in shrimp using the HPLC method. Table 1 shows a comparison of the analytical methods. Figure 6. The mechanism of azo compounds formation [18]. High accuracy is one of the requirements for a good analytical procedure. This study used %recovery to determine the accuracy of the research. The calculation of %recovery addition was derived from the concentration of CAP added to the sample and analyzed using a diazotization reaction with a standard CAP concentration of 500 ppm before the shrimp were added. The results are shown in Table 2. The percentage recovery produced is very good, with a range between 87.41 to 107.73%, with an average of 109.38%. This value falls within the 80-110% permitted by AOAC in 2016, so the validation results based on the %recovery value satisfy the requirements [21]. This indicates that the recovery test for CAP in shrimps using the standard addition method based on the diazotization reaction has good accuracy. In order to determine the result of an analysis, a detection limit (LOD) and a quantization limit (LOQ) are required. In this study, the LOD was 0.19 μg/mL, and the LOQ was 0.64 μg/mL. Linearity in the study gave good results with the correlation coefficient (R 2 = 0.991) shown in Figure 7. Wafi, et al., [22] described a spectrophotometric method with a diazotization reaction at room temperature to analyze shrimp with a LOD of 0.36 μg/mL and a LOQ of 1.19 μg/mL. Sharma, et al., [10] detected chloramphenicol residues in milk with a diazotization reaction at room temperature resulting in a %recovery of 98.66-101.12%. Hussein, et al., [23] analyzed using an oxidative coupling reaction to produce LOD, LOQ, and %recovery of 0.241 μg/mL, 0.804 μg/mL, and 99.77-100.5%, respectively. The present study yielded a higher percentage of recovery and a lower detection limit based on prior research. This indicates that the CAP sample analysis using the standard addition method based on diazotization has greater accuracy, linearity, lower LOD, and LOQ. This study demonstrates that the standard addition method by diazotization reaction is an excellent method for enrichment and detection of CAP in shrimp samples and a quantitative and qualitative CAP analysis method.
CONCLUSION
Based on the test results, the shrimp sample did not meet the standards set by the European Commission, which found content 1964.91 μg/kg of CAP. The spectrophotometric method for CAP analysis using the addition method based on the diazotization reaction has good accuracy shown by %recovery was 87.41%-107.73% with an average of 109.38% and it has good linearity as shown by the correlation coefficient (R2) was 0.9991. The limit of detection (LOD) value is 0.19 μg/mL and the limit of quantitation (LOQ) value is 0.64 μg/mL. | 2022-07-12T05:53:11.813Z | 2022-07-10T00:00:00.000 | {
"year": 2022,
"sha1": "113848e98fc90669afd781269658c21002c2f366",
"oa_license": "CCBY",
"oa_url": "https://ijcsrr.org/wp-content/uploads/2022/07/18-10-2022.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "eda096fff54377aab5cd3d14df8a7a513dc44ae8",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": []
} |
251589586 | pes2o/s2orc | v3-fos-license | Local Neumann semitransparent layers: resummation, pair production and duality
We consider local semitransparent Neumann boundary conditions for a quantum scalar field as imposed by a quadratic coupling to a source localized on a flat codimension-one surface. Upon a proper regularization to give meaning to the interaction, we interpret the effective action as a theory in a first-quantized phase space. We compute the relevant heat-kernel to all order in a homogeneous background and to quadratic order in perturbations, giving a closed expression for the corresponding effective action in $D=4$. In the dynamical case, we analyze the pair production caused by a harmonic perturbation and by a Sauter pulse. Notably, we prove the existence of a strong/weak duality that links this Neumann field theory to the analogue Dirichlet one.
Introduction
One of the major successes of Quantum Field Theory (QFT) has been the prediction of the Casimir effect [1], which builds a bridge between the world of macroscopic media and that of quantum effects [2][3][4][5][6]. In a first approximation, the bodies may be modelled as perfect conductors and thus implemented as boundary conditions on the electromagnetic field [1]. Of course, this is not enough to describe the always improving experimental results [7,8]. Moreover, this simplification is thought to be the root of some theoretical issues that include the divergence of the energy density at the boundaries [9,10] and a possible ill-definition of the vacuum self-energy [11].
Another possibility regards the substitution of perfect Dirichlet or Neumann boundary conditions by more general ones. In the Dirichlet case, one can mention for example semitransparent boundary conditions [24][25][26][27]. These can be seen as a special case of the ones described in the previous paragraph, where potentials get localized in a thin shell and allow the recovery of perfect boundary conditions as a special limit of the coupling.
Following these lines one can also consider local semitransparent conditions, in which the coupling (or background field living on the thin shell) becomes spacetime dependent [28,29]. This allows the study of interesting phenomena and cases, such as particle creation [30] and smooth nonplanar geometries (say gratings [31]) in the effective-medium approach [32].
A by far less studied case regards the analog for Neumann boundary conditions. It has been shown in Ref. [33] that Neumann semitransparent (or imperfect) boundary conditions for a scalar field can be modelled by the operator of quantum fluctuations where ζ ∈ R is the coupling constant. From a mathematical point of view, the problem at hand mixes several interesting ingredients. Indeed, it is contained in the so-called four-parameter family of point interactions [34], i.e. it arises in the study of self-adjoint extensions of the one-dimensional second-derivative operator acting on functions appropriately defined on R/{0}. In this context the interaction was baptized δ (x), leading to some misunderstandings in the literature. This point was already explained in the paper byŠeba [35], which showed that the correct interpretation of this self-adjoint extension is in terms of a ∂δ∂ potential with a renormalized potential, as also discussed in [33,36,37]. In particular, ∆ ζ is not related to the proposal in [38], where a true generalized potential is considered. One formal way to introduce the operator ∆ ζ is through the implementation of boundary conditions [34,39]. In this case, the action of ∆ ζ should be understood as implying the boundary conditions where 0 ± denotes the left/right limits to zero, and the derivative should be regularized for example as φ (0) = φ (0 + )+φ (0 − ) 2 , see [34,37]. This interpretation through boundary conditions has been pursued recently in [40,41], where δ-δ structures have been considered.
In the present paper we will focus on a scalar field living in a D-dimensional flat space. It will satisfy local Neumann semitransparent boundary conditions on a flat thin shell, i.e. we will upgrade ζ to a spacetime-dependent coupling (or field living on the shell). In particular, we will follow the idea that such a field may develop small fluctuations η around a mean (or vacuum expectation) value ζ, so that an exact treatment in ζ but a perturbative one in η should provide rich physical information. We thus begin our exposition in Sec. 2 with a description of the relevant QFT, consisting of a scalar field interacting with a classical background. In Sec. 3 we explain how to compute the heat-kernel of ∆ ζ to all order in the coupling ζ employing the Worldline Formalism, i.e. a path-integral approach. In doing so, we devise a tailor-made regularization which allows to perform in Sec. 4 a perturbative computation for spacetime-dependent contributions, which is also exact in the constant background.
Then, in Sec. 5 we analyze the effective action in a four-dimensional setting, showing that the renormalized theory is dual to the problem of local (Dirichlet) semitransparent boundary conditions. Afterwards, we study in Sec. 6 the effect of pair production in our system by considering time-dependent background fields, an analogue of the widely studied dynamical Casimir effect. In particular, as examples we consider a harmonic perturbation and a Sauter pulse. Finally, in Sec. 7 we state our conclusions.
We leave necessary technical information to the appendices, where we compute the Worldline generating function relevant to our current problem (App. A), compute integrals involving chained free heat-kernels (App. B; others containing also Hermite polynomials are included in App. C) and calculate in closed form a series of Hermite polynomials (App. D). We use Planck units so that and c are taken to be unity. It will prove convenient to split D-dimensional coordinates x into the coordinate perpendicular to the shell, x D , and those D − 1 parallel to it, x ; in Minkowski space, the set of spacelike coordinates of the latter is denoted by x . We definex := x − L.
The model
We begin by considering a scalar quantum field ϕ living on a D-dimensional Euclidean flat spacetime; it interacts with an external (classical) scalar field η that lives on a plate according to the following action: In this action m is the mass of the field ϕ and we have splitted the coordinates into the direction perpendicular to the plate (x D ) and those parallel to it (x ). Additionally, the plate is placed at x D = L and ζ may be understood as a mean value of the field η over the plate. From a physical point of view, the classical field describes the properties of the thin plate that are relevant in the interaction with the quantum field. Notice also that η has dimensions of length, independently of the dimension D of the spacetime. The interpretation of this action in Eq. (2.1) is more evident once we consider a homogeneous configuration for which η(x ) ≡ 0 [33]: in the limit ζ → ∞ one expects to obtain two semi-spaces, on whose boundaries ϕ satisfies Neumann boundary conditions, very much akin to the generation of Dirichlet boundary conditions through delta potentials 1 . In this sense, for finite ζ the action (2.1) can be interpreted as imposing Neumann semitransparent boundary conditions on the field ϕ; allowing for inhomogeneous field configurations η(x ), we will refer to local Neumann semitransparent boundary conditions. Following the usual way, upon a path-integral quantization one can obtain the generating functional Z for this system, (2.2) Afterwards, integrating out the scalar field ϕ, we can write the effective action Γ in terms of the operator of quantum fluctuations A; a direct computation gives where Γ one−loop denotes the quantum contributions to Γ. We will assume that η + ζ ≥ 0, such that A admits only a continuum spectrum. A peculiarity is that the operator A may be interpreted as a Schrödinger operator with a derivativedependent potential, different from the more diffused case of an only positiondependent potential 2 . As we will see in the next section, this admits an interpretation of boundary conditions in phase space.
On physical grounds, we expect two interesting regimes for our system: one in which η develops small fluctuations around a constant background, admitting thus an expansion in powers of the fluctuations, and one in which topology plays an important role, such that η should be considered to all orders. In the following we will consider the former, leaving the latter for future studies.
A path integral approach: the Worldline Formalism in phase space
One simple way to compute the quantum contributions to the effective action is to employ the well-known equivalence between Log Det and Tr Log, as well as Schwinger's propertime trick (or Frullani's representation for the logarithm of a quotient [42]). In this way we may recast where the heat-kernel K A (x, x ; T ) := x|e −T A |x has appeared in a natural way. From the formal side, the study of the spectral functions of operators with singular operators or generalized boundary conditions has attracted much attention in recent years [43][44][45][46][47][48]. One efficient way to perform this kind of computations is through the Worldline Formalism, in which one interprets the heat kernel in terms of path integrals in a first quantization procedure. Indeed, one may notice that the arguments of A are momentum and position operators in a first quantization, realized as (p,x) → (−i∂, x), so that e −T A can be understood as the evolution operator in imaginary time t = iT .
The Worldline Formalism has been successfully applied to several problems, see the book [49], the reviews [50,51] and references therein. In particular, it has recently proved useful in three situations that are relevant to the present computation: in phase space, where it has been applied to investigate noncommutative quantum field theories [52][53][54] and Berry phases [55], in the analysis of singular potentials and metrics [30,56,57] and in the study of boundaries [58][59][60] (see also [61][62][63] for related path integral approaches).
As a first step, we will perform an all-order computation with a constant background, setting η ≡ 0. It should be clear that under such assumption one can 2 Another way to look at the problem is to focus on the similarities that the operator A has with Laplace-Beltrami operator with a singular metric. We are not going to pursue this way in this article. disentangle the contribution in the Dth direction, yielding the remaining components just a free path integral. Therefore, for the rest of this section we may simply work in a one-dimensional setup. Additionally, we will consider the massless case, given that the mass may be directly included in the heat-kernel at the end of the computations, simply by adding a factor e −m 2 T .
Taking these considerations into account, we follow the Worldline Formalism approach to compute the transition amplitudes, so that the relevant heat-kernel may be written as In obtaining the master equation (3.2) one needs to Weyl-order the operator A, which involves employing the commutation relation between derivatives and coordinates to render the operator symmetric in terms of p and x, see [49,64]; this is the origin of the δ contribution and one of the reasons why previous attempts to compute the heat-kernel of the operator A may have failed [65]. As a matter of fact, one should keep in mind that this additional δ term does not enter in any way in the original operator; instead, it just plays a role in the path integral interpretation.
One further important point is that the path integral over the momentum variables has no boundary conditions. Since from the QFT point of view we are interested in the trace of the heat-kernel, we could have imposed periodic boundary conditions on the phase space path integral. However, such a choice would not allow the study of the heat-kernel out of the diagonal that we are going to undertake in the following section.
3.1. The heat-kernel K ζ . We will compute the heat-kernel in a perturbative fashion, expanding the result (3.2) in the coupling parameter ζ. In this way we obtain To proceed further we compute the momentum integrals, which are Gaussian upon introducing a source denoted by j: The factor N is a normalization constant whose only role is to be fixed when determining the value of the free path integral and therefore may be safely dismissed.
Having recasted every momentum as a variation in j, one can simplify the problem noting that δ (x − L) = ∂ 2 L δ(x − L); this enables us to use the Dirac deltas to impose constraints on the paths as following: where for any t-dependent quantity X(t) adding a subindex means X l := X(t l ) and the integral over the intermediate times has been ordered, such that Additionally, we have introduced one position variable (L i ) for each insertion of the potential, in order to avoid undesired mixing of the derivatives; as we will see shortly, this will bring its own benefits.
In this way the computation is reduced to a chain of partition functions (or generating functionals), which can be readily computed as in App. A. Using those results we get where we have introduced the natural notation for the intermediate displacements ∆x m := x m+1 − x m and intermediate periods ∆t m := t m+1 − t m ; correspondingly we define x 0 := x, x n+1 := x , t 0 := 0, t n+1 := T and for i = 1, · · · n we set x i := L i . Even if at first sight the computation of the variations and derivatives may seem a hard task, the implementation of a Hubbard-Stratonovich transformation to linearize the problem allows a direct computation: (3.8) Notice that had we set L i ≡ L at this point, the integrals in the intermediate times would have become divergent. Indeed, it has been shown in the past that the problem at hand requires a renormalization of the coupling [33,35] or, alternatively, an extension of the the potential to let it act on more general functions [66]. This feature is shared by some related problems, such as point interactions in three dimensions [34,67]. In our case, the regularization is already implemented by the separation of the intermediate points, which in physical terms corresponds to the idea of a plate with finite width suggested in [33]. The limit L m ≡ L will thus be left to the last stage of the computation.
As explained in App. B, one can use the results in [30] to perform the integrals in the intermediate times one by one. Considering the first cases, one realizes that the result for arbitrary n involves Hermite polynomials; using the general result of App. C, we obtain where the nth order coefficient in this expansion reads In this expression we have employed the sign function sign(·) and defined the displaced variables,x We can perform a resummation in Eq. (3.9) by using the method described in App. D; in this way we get a closed result in terms of the complementary error function erfc(·): The expression (3.12) coincides with the one obtained in [65] from a fermionic path integral. One further confirmation of the correctness of our result can be obtained from its trace. Integrating the heat-kernel (3.12) over the whole space we get where V denotes the volume (length) of the whole space. Using this formula as point of departure, we can analyze the small and large coupling regimes by using the following expansions: (3.14) In particular, it is immediate to see that we recover the free (ζ = 0) and the Neumann (ζ → ∞) cases [68]. As a final comment, recall that we are considering a positive ζ. The case with ζ < 0 is subtler, given that a bound state with energy E b = − 4 ζ 2 arises [65]. In order to obtain the heat-kernel trace for negative coupling one should then not only change the sign of ζ in Eq. (3.13) but also add a further contribution coming from the bound state, which reads Interestingly, the appearance of the bound state is already signaled as a nonperturbative factor in Eq. (3.13), which for ζ < 0 generates a strong divergence as ζ → 0. This reminds us of similar effects in resurgence theory, where information about the nonperturbative sector is stored in the perturbative results, see [69][70][71][72] and references therein. A more detailed analysis of this fact will be left to a future publication.
3.2.
Relation with the heat-kernel of a Dirac delta potential. One fundamental remark concerns the relation of this heat-kernel with a similar problem, viz. that involving a Dirac delta potential. If one defines the operator then its heat-kernel has been shown to be [30,73,74] This means that we may recast our result (3.12) as This relation is similar in nature to the Fermi-Bose duality introduced by Girardeau [75] and then further developed by Cheon and Shigehara [76] for a gas of interacting particles. However, our case differs from theirs, inasmuch as one can verify that K ζ possesses no defined symmetry underx → −x and consequently no statistics can be clearly assigned. Alternatively, projecting our heat-kernels K ζ and K delta respectively to the space of antisymmetric and symmetric functions around L, one can obtain the Fermi-Bose duality at the level of heat-kernels. Importantly, the similarity between the heat-kernels involved in (3.18) gets enhanced once we consider their diagonal, since then some partial cancellations occur and the sign functions simplify. As we will see in Secs. (4.1) and (5.1), the relation will be upgraded to a map between heat-kernels for the case of inhomogeneous backgrounds and will entail a duality at the level of QFTs, linking Neumann and Dirichlet local semitransparent boundary conditions.
An expansion for local Neumann semitransparent boundary conditions
Let us now turn our attention to the operator in Eq. (2.4). As precedently commented, we will split the classical background into a constant ζ plus a small perturbation η(x ) that may depend on the (D − 1) coordinates parallel to the plate. The idea is to perform an expansion in η(x ), keeping the full dependence on ζ. Following the lines in the preceding section one can obtain a Worldline formula for the corresponding heat-kernel: This expression can be readily expanded in powers of η. Moreover, we expect that in a large variety of physical situations η would be such that its average would vanish; as a consequence, we will neglect the first order contribution, so that the first new contribution will appear at second order in η. In formulae, we obtain ζ (x, y; T ; η] : = in which we have explicitly used the symmetry under exchange of the intermediate times s i=1,2 . Once more, the computation is more involved than the usual case, given that the potential on one side involves derivatives and on the other should be regularized. For this reason, it proves convenient to employ a series expansion in ζ, instead of trying to make direct use of the heat-kernel in Eq. (3.12). Calling If we now want to interpret the Dirac delta functions as fixing the path at given times, then there is an additional difficulty related to the fact that the intermediate times s i=1,2 are not ordered with respect to the other times, t i . However, whatever value s i=1,2 may take, they will of course fall into one of the intervals (0, t 1 ), (t 1 , t 2 ), · · · , (t n , T ). We can thus define ordered timest i (lm), such that (t 0 (lm),t 1 (lm), · · · ) := (0, t 1 , · · · t l , s 1 , t l+1 , · · · , t m , s 2 , t m+1 , · · · , T ), (4.4) and introduce an analogous definition for the 3L i andx i . Hiding the (lm) dependence for reasons of readability, we obtain Notice that we may also reorder the series to get the compact expression where the S 1 and S 2 functions are defined as follows: In the usual case, i.e when one considers potentials that do not involve derivatives, the S i functions would correspond to heat-kernels with constant coupling. In the current model, the S i play instead the role of regularized derivatives of the heatkernel with constant coupling ζ. Employing the results in Apps. C and D, we can perform a resummation of the power series in ζ and obtain them in closed form: 4πT . (4.9) At this point we can perform the following fast check. If we consider a constant η, then the result in Eq. (4.6) reduces, as expected, to a free heat-kernel in the parallel directions, multiplied by K ζ+η (x, y; T ) restricted to quadratic order in η.
Considering once more the expression (4.6) for the heat-kernel, the integrals in the parallel directions are trivial Replacing in Eq. (4.6) we obtain our final expression for the contribution to the heat-kernel of quadratic order in η: A more carefully comparison shows that the proportionality extends also to the heat-kernel expression at quadratic order in η, so that, up to a rescaling in the inhomogeneities, the weak-background problem in the delta case (or more precisely, local semi-transparent Dirichlet boundary conditions) is mapped to a strong-background regime in our current model and vice versa 4 : In particular, this means that we may borrow some results from [30]; as an example, the trace of the heat-kernel (which will be employed to compute the effective action) is simply given by (4.16) An even more detailed inspection shows that this is not an accidental relation valid only for the quadratic expansion in η. Indeed, one can repeat the computations performed in the previous section for an arbitrary order in η, to find that, dismissing the path integrals in the parallel directions, one obtains a chain which is exactly the one we would have obtained in the delta case. This provides an order by order proof of the existence of a map between the heat-kernels of these two different problems; more precisely, the map is given by .
(4.18)
This map depends strongly on the fact that η lives on the plate and will be broken if one introduces for example an additional potential with support outside from it. As we will see in Sec. 5, the mapping that we have discussed will automatically translate into a duality at the level of the renormalized semiclassical field theory.
4.2.
The purely inhomogeneous coupling. One interesting case is that in which the background field is small, such that the fluctuations may become larger than it 5 . In this regime we will be able to obtain a closed expression for the trace of the heat-kernel and the effective action in the massive case. Moreover, as we will see this will turn out to be an instructive limiting case. Let us then begin with expression (4.16). Taking its small-ζ limit involves expanding the heat-kernel of the delta potential for large coupling, which gives 4 We are defining K delta (x, y; T ; 4ζ −1 ; −4ζ −2 η] as the contribution to the heat-kernel of the operator A delta D := −∂ 2 + m 2 + η(x ) + ζ δ(x D − L) which is quadratic in η. 5 But keeping always ζ + η > 0 such to avoid instabilities triggered by possible bound states so that after setting x ≡ L the leading term cancels, rendering the expression more divergent as T → 0: (4.20) Setting also x ≡ L, the first contribution to K delta is of order ζ 2 , which is exactly the power needed to cancel the inverse powers of ζ in Eq. (4.16). In our computation the limit x, x → L should be taken at the end, since they act as regulators; the final expression once the mass contribution is reinstated is given by where g(·) is defined in terms of the modified Bessel functions of the first kind I α (·): To have an intuition of the result (4.21), one can study the behaviour of g(·) for large and small arguments: (4.23) Notice that the limit of a homogeneous configuration gives a vanishing contribution, in agreement with the previous results in Sec. 3.1. Moreover, the first terms in a small-propertime expansion are local. If instead the field η acquires modes with large momenta, then the expansion involves half-integer powers of k , which is tantamount of saying that nonlocal terms will play an important role. The result in (4.21) deserves one last comment, which will become important in the analysis of the effective action in the following section. A consequence of the small ζ limit is that the expression for the heat-kernel, in the expansion for small propertime, is more divergent than the finite ζ case by a factor T −1 . More generally, every time we increase the order in η by one, the expansion of the heatkernel for small T will be more divergent by a factor T −1/2 . This is reminiscent of the findings for a constant coupling, cf. (3.10), signaling that for a small coupling a resummation may be needed in order to see the real behaviour for small T .
The effective action
5.1. Duality for the Field Theory. Consider now the mapping in Eq. (4.18) at the level of the effective action adding a mass term. Employing the formula (3.1) for the effective action in terms of the heat-kernel's trace, we get the following relation between the quantum contributions to the effective action in our generalized Neumann case, Γ 1−loop , and those in the generalized Dirichlet case, Γ delta 1−loop : Inasmuch as we don't consider the interaction with gravity, the constant factor 4 −D/2 is irrelevant in the computation of physical quantities, since it will be absorbed in the renormalization of the cosmological constant. Therefore, we can see that there is a duality between both theories at the quantum level: if one desires to compute the large background expansion in one theory, one may simply study the small background of the other. These assertions are valid independently from the dimensions D of spacetime in which we choose to work.
5.2.
The inhomogeneous and massless case in D = 4. As an immediate consequence of the duality discussed in the previous paragraphs, we can study the massless case in D = 4 by borrowing results from the delta-potential case previously obtained by some of the authors of this work [30]. The explicit result for the effective action at quadratic order in η is where the form factor F has been split into three terms, one divergent as D tends to four, another which is finite and local (F L ), and the remaining which is finite and nonlocal (F N L ): The explicit expressions for these contributions, defining b 2 := 16(ζk) −2 and in terms of Lerch's transcendent function Φ(·, ·, ·), read In order to render the nonlocality of F N L more visible, one can perform expansions for large and small b, for which we get either log(k 2 ) contributions or half-integer powers of k 2 that preclude a so-called derivative expansion (see [77] for its application to a Casimir configuration which is similar in spirit to ours): At this point some comments are in order. First, the leading terms were to be expected from a simple dimensional analysis of the problem. Indeed, this is the reason why corrections proportional to k are so frequently encountered in the bibliography [78]. Second, the vanishing Neumann limit of Γ (2) seems to be well-motivated: indeed, a small variation around infinity should make no difference, at least as long as η is small, which was one of our hypothesis. The physical mechanism is similar to that in the Dirichlet case, where a larger coupling tends to repel the quantum field from the sheet, correspondingly attenuating the interaction with the background η. The only difference is that the coupling in the current situation involves derivatives of the quantum field.
Third, Eq. (5.3) signals that the theory needs to undergo renormalization. In dimensional regularization, the only term that needs a counterterm is the mass term for η. However, in other schemes one may obtain additional divergent terms, as discussed for general cases in [5,79]. Although this is not the case if ζ > 0, the ζ = 0 case is more subtle and will be discussed in Sec. 5.3.
Lastly and related to the previous point, in the limit of vanishing coupling the effective action in expression (5.2) is divergent, as can be seen from Eq. (5.3) together with the corresponding definitions and the expansion in Eq. (5.6). Such divergence simply indicates the fact that the expansion for small η and ζ do not commute. To gain insight into this point, consider the constant coupling case. A straightforward computation shows that where ψ(·) is the polygamma function and V is the volume over the plate. If we now consider homogeneous perturbations by replacing ζ → ζ +η, the expansion will be in powers of η/ζ, rendering clear our statement. The physical intuition of why the expansion is singular for ζ around zero is related to the instabilities generated by the bound state for ζ < 0. An alternative heuristic way to make sense of those divergences is to interpret them as a need of an additional renormalization. We will analyze this point further in the following section.
On the purely inhomogeneous and massive case.
To understand better the ζ → 0 limit of the previous expressions, let us set ζ ≡ 0 right from the beginning; the simpler formulae enable us to include the effects of a nonvanishing mass. One can then compute the effective action employing the heat-kernel's trace in Eq. (4.21); the result for a massive field in such case is 6 where the form factor is defined in terms of the hypergeometric function 2 F 1 (·, ·; ·; ·): The asymptotic expressions of this form factor in the limits of large and small mass can be obtained from the corresponding expansions of the hypergeometric function, to read (5.10) In particular, the massless limit corresponds to a nonlocal term proportional to a half-integer power of k 2 , namely (k 2 ) 5/2 . Notice that the result in Eq. (5.8) is automatically finite in dimensional regularization. However, as mentioned in the previous section the situation may change in other schemes. If instead of dimensional regularization a cutoff is introduced, then also the terms (k 2 ) i |η| 2 , i = 1, 2, should be renormalized. Indeed, computing 6 We are considering η(x ) > 0, so that effectively one may extract a mean value and perturbations around it. However, it proves convenient for the following discussion to keep η as one single entity. the effective action from (4.21), we introduce a UV cutoff Λ with dimensions of momentum, set D ≡ 4 and expand for small T to obtain where the dot points denote finite terms as Λ tends to infinity. As a consequence, the theory lacks predictivity for the terms depicted in (5.11). In particular, one may set all of them to zero, as in the substraction of large mass terms suggested in [80]. At this point one may understand the ζ → 0 limit of the results in the previous section as follows. The divergent contributions in Eq. (5.8) as ζ → 0 should be reabsorbed in a renormalization process; the explicit equivalence between both approaches can be seen by comparing them to Eq. (5.11). Taking this comment into account, one then sees that the ζ → 0 limit of Eq. (5.2) and the massless limit of expression (5.8) agree at the renormalized level.
One can also envisage what would happen once higher powers in η are considered. At the end of Sec. 4 we have mentioned that, in the purely inhomogeneous scenario, the expansion of the heat-kernel for small propertimes acquires one additional T −1/2 for every extra power of η. This implies that the number of terms to be renormalized will also correspondingly increase. However, this situation is reminiscent of the perturbative expansion of the heat-kernel for homogeneous coupling ζ, cf. Eq. (3.10), where arbitrary large negative powers of the propertime appear. Once the series is resummed, the result (3.12) is seen to have only a T −1/2 divergence for small propertime. We expect that a similar mechanism should be behind the need for renormalization of terms like those in (5.11) as long as η is strictly bigger than zero, since we have not seen them in the expansion studied in Sec. 5.2. In other words, the resummation in ζ performed in Sec. 4, if η/ζ < 1, is expected to be enough to avoid the singularities of the effective action's expansion for vanishing total coupling (ζ + η).
Dynamical Casimir effect and particle creation
Up to this point we have restricted ourselves to the consideration of the theory in Euclidean space. As customarily done, one can appeal to a Wick rotation in order to consider the problem in Minkowski space. This will allow us to mimic a situation of dynamical Casimir effect through time-dependent properties of the wall; for more information on this effect, see the reviews [81,82] and references therein.
To examine this dynamical scenario, we will assume that the argument of the delta function in Eq. (2.1) corresponds to a spatial coordinate, so that time may only be an argument of the external field η. Calling τ the Euclidean time and x 0 the Minkowski one, the rotation x 0 =: −iτ (accompanied by analogous rotations for every 0-component of a tensor) may be performed without encountering singularities in the form factors studied in this manuscript. One obtains then a master formula for the effective action Γ where we have introduced Feynman's prescription through an infinitesimal parameter , we have taken the branch cut of the square root to be in the negative real axis and the Fourier transform in Minkowski space is defined as 7 Of course the form factor F , depending on the situation under study, may be chosen among those in Eqs. (5.3) or (5.9). The expression for the effective action may be used to compute the creation of particles in a dynamical situation. Indeed, in the usual in-out formalism, the vacuum persistence's amplitude is given by the effective action as If the effective action develops an imaginary contribution, which may be possible by the appearence of branch cuts in (6.1) after the Wick rotation, then the vacuum becomes unstable through a process of pair creation, whose probability P is defined as In the most frequently studied situation, i.e. for weak pair-creation processes, we may approximate P ≈ 2 Im Γ M . Therefore, the quantity in which we are interested is Turning back to Eq. (5.9), the hypergeometric function 2 F 1 − 3 2 , 1 2 ; 2; x possesses a branch cut, which in the case of the principal branch runs from 1 to ∞ on the real x-axis. We may thus recast the probability of pair creation as where Θ(·) is the Heaviside function that signals the threshold of the pair-creation process: the external field η must provide at least the rest energy of two particles for the process to take place. In the case described in Eq. (5.2), such threshold is absent because particles are taken as massless.
6.1. Harmonic perturbations. A simple model that mimics the dynamical Casimir effect, introduced in [83] for the one-dimensional case and studied also in [30] for an inhomogeneous delta potential, is given by perturbations that are harmonic in time with frequency ω 0 ; for simplicity we will consider it independent of the spatial coordinates 8 : In Eq. (6.7), the exponential factor is employed to impose a boundary in time, since otherwise the number of pairs created becomes infinite. In the limit of large T , a straightforward computation shows that its Fourier transform satisfies The set of components that are parallel to the plates and space-like is denoted by k . 8 Following the discussion in Sec. 5.3, it should be clear that formally it is not enough to consider the amplitude η 0 small; one should also think that there is an additional background ζ (not necessarily much) bigger than η 0 for the expansion in η to be well-defined.
which, defining the threshold frequency ω c := ω 0 /(2m), thus leads to the following probability of pair creation rate per unit area of the plate in the purely inhomogeneous case (A denotes the area of the plate): To derive the last line we have employed the result in [84] for the jump across the branch cut of the hypergeometric function, which is proportional to the desired imaginary part. Notice that in the harmonic case, the pair production rate should be understood. The panel on the right is a density plot of the pair production probability per unit area for the Sauter pulse as a function of the frequency ω S and the mass m. In all the cases, the amplitudes η 0 are set to unity.
Analytically, we can compute the expansions for small and large ω c , which show that ω 5 0 sets the scale of the probability rate: (6.10) On one side, we see that in the massless limit we indeed recover (up to a rescaling) the infinite coupling result for the delta case [30], being the first corrections of order ω −2 c . Once more, on dimensional grounds this situation is reproduced in some analogous setups, such as a moving mirror [85].
On the other side, for large masses the result evidently vanishes as a consequence of the mass threshold, since the energy of the oscillations are not enough to provide the minimum energy of two particles at rest. For values of ω c slightly larger than the threshold, the particle creation rate behaves as a third power in the difference (ω c − 1). These behaviours can be confirmed from the plot in the left panel of Fig. 6.1, where the pair production rate per unit area is shown as a function of the frequency; the dashed blue line and the violet dashed-dotted line correspond respectively to m = 0.1 and m = 3.
As a generalization of this simple harmonic example, one can also consider a perturbation that resembles a plane wave over the plate, where k is its wavenumber, σ → 0 is a regulator and v the speed of the wave (recall that v = 1 in our units equals c, the speed of light in vacuum). After removing the regulator we get the following pair production rate per unit area: On one side we see that pair production is possible only if v ≥ 1. Notice that this situation should be understood not as a travelling wave with speed faster than light, but rather as an active fast modulation of a property over the plate, in a way analogous to that proposed for example in [86].
On the other side, defining an effective frequency ω 2 eff := (v 2 − 1)k 2 , the result (6.12) can be obtained from (6.9) by simply trading ω 0 → ω eff . In particular, the exclusively time-dependent case η H can be understood as its infinite speed limit (after an appropriate rescaling of k).
6.2. Sauter pulse. One widely diffused profile in the literature of QED in external backgrounds is the Sauter pulse [87], defined as , ω S ∈ R − {0}, η 0 > 0. (6.13) Replacing this profile in our general equation (6.6), we can compute the probability of pair creation P S . First, for the massive case (and ζ = 0), we show a density plot of P ζ=0 S per unit area in the right panel of Fig. 6.1 as a function of the frequency ω S and the mass m. On the one hand, as could be expected, as the mass increases the probability of pair creation diminishes, since the cost of creating the pair becomes higher. On the other, if the frequency ω S becomes larger, the distribution of η in Fourier space is widened, so that creation of pairs is more favoured. This is in agreement with the following analytical asymptotics where ζ R (·) is Riemann's zeta function. Although the exponential suppression for large masses may remind the one present in the Schwinger pair production for rather general electric fields [88], keep in mind that our process is perturbative in the background field amplitude and therefore intrinsecally different in nature. Additionally, in the left panel of Fig. 6.1 we show the behaviour of the pair production probability per unit area as a function of the frequency ω S ; the solid red line and the yellow small-dashed line correspond respectively to m = 0.1 and m = 3. Compared to the harmonic case, if frequencies are small we see that the exponential supression allows for a faster setting in of pair production; in other words, the Sauter pulse always embodies some frequency components above the mass threshold that enables the creation of pairs. On the contrary, for larger frequencies the trend will be reverted and the harmonic pair production will become greater, as dictated by Eqs. (6.10) and (6.14). 2. Probability of assisted particle production for a Sauter pulse per unit area, ζ 2 P ζ S /A. The left panel shows its density plot as a function of the frequency ω S and the background ζ. The right panel displays its behaviour as the frequency varies, considering different backgrounds. We have set the amplitude of the pulse (η 0 ) to unity.
We can also consider the massless case together with an arbitrary ζ. An expansion of the pair production probability for large and small frequencies thus gives In particular, for vanishing ζ we recover the expected massless limit of Eq. (6.14).
Increasing the value of ζ gives rise to an anti-assisted effect, contrary to the one found for electromagnetic backgrounds for Schwinger pair production [89,90]. Naively, since a constant potential alone does not contribute to pair creation in our scalar setup 9 , one could have expected that its addition would have had no effect on pair creation. Instead, the nonlocal feature of the generated form factor is nontrivial. The fact that it diminishes the pair production can be physically understood from the fact that a larger ζ tends to repel the quantum field, as explained in Sec. 5.2. Lastly, taking into account the discourse developed in Sec. 5.3, we can analyze the case in which we rescale the amplitude of the perturbation, η 0 =: ζη, such that η is taken to be small. At the practical level, this is analogous to saying that we multiply the pair production probability by ζ 2 . We have plotted in the left panel of Fig. 6.2 a density plot of the rescaled probability as a function of both ω S and ζ; one can see that for ω S 1.5, increasing the coupling ζ leads to larger probabilities. The right panel of Fig. 6.2 provides a clearer picture of this effect, showing the rescaled probability as a function of ω S for three distinct values of ζ ∼ O(1). For ω S ∼ 1.5 we see that the hierarchy between the curves is inverted.
Conclusions
In the present article we have studied the problem of a quantum scalar field theory with local semitransparent Neumann boundary conditions on a plate, which can also be understood as interactions with an external (classical) field confined to the latter.
First of all, we have shown that the quantum contributions to the effective action can be understood in terms of quantities in the phase space of a first quantization. Worldline techniques are by far much more developed in the case of configuration space [50], rendering the study of such path integrals per se an interesting problem.
Second, taking into account the well-known fact that the studied interaction requires regularization, we have devised a regularization appropriate to our pathintegral methods. We have shown that in this way we are able to rederive previously obtained results for the homogeneous case. Notice that this regularization provides also a physical interpretation in terms of an effective finite width of the plate, since we have to evaluate the intermediate heat-kernels at noncoincident points.
These technical developments allowed the computation of the heat-kernel at quadratic level on the perturbations η around a homogeneous background field ζ and the corresponding effective action in the D ≡ 4 case. Notably, these results are exact in ζ; they are (generally) given by a nonlocal operator acting on the perturbations η, what can be seen from the appearance of nonanalytic contributions.
Such nonanalytic contributions are responsible for the pair production that we have encountered in Minkowski spacetime, where the perturbations are used to model dynamical properties of the plates, very much akin to the situation in the Dynamical Casimir effect. In this scenario we have considered two possibilities, a harmonic excitation and a Sauter pulse. For large masses, the Sauter background displays an exponential cutoff, which is softer than the Heaviside present for the harmonic pulse. In the massless Sauter case, the introduction of a background ζ leads to a scenario of anti-assisted pair production. This effect can be inverted in a region of the parameters space if the amplitude of the pulse scales linearly with ζ, what is possible taking into account the discussion in Sec. 5.3 regarding the smallness of η and ζ.
A special comment deserves the finding of a duality between the field theory obeying local semitransparent Neumann boundary conditions and the equivalent Dirichlet one. Indeed, we have shown that there exists a mapping between the relevant heat-kernels, which becomes an exact strong/weak duality at the level of the corresponding renormalized quantum effective actions. Notice that this is a new duality, different from the Fermi-Bose duality discussed in condensed matter by means of the Girardeau mapping [75,76,91]. To clarify this point, first notice that the latter connects the Lieb-Liniger model [92] and the Cheon-Shigehara one, which are both one-dimensional models, while our results are valid in D-dimensions. Second, as discussed in Sec. 3.2, we impose no type of symmetry under the parity transformationx → −x and we always work with a scalar bosonic field, contrary to the change in statistics of the Girardeau mapping. Additionally, we don't introduce a self-interaction; instead, our field interacts with a spacetime-dependent background potential η. Nevertheless, taking into account the several experimental accomplishments based on the Fermi-Bose duality [93,94], it will be of interest to explore possible experimental roads of the newly devised duality.
Regarding possible future developments, it will be interesting to try to generalize our results to the problem where nonlocal boundary conditions are imposed. This may provide a way to analyze the appearance of topological effects.
One further peculiarity of the interaction considered in this article is that it can be thought as the first term in an effective field theory expansion, which has undergone a thin-shell limit. As such, it may find applications in trying to understand the nature of dark matter. Indeed, one open possibility is that it may behave as a field with unusual couplings [95,96].
Finally, another useful advance would be the development of a numerical code to tackle the problem discussed here. The situation is more involved than the usual cases, since a naive adaptation of the numerical Worldline techniques [97][98][99] to phase space suffer from the so-called sign problem.
where the required symmetric Green function is defined as G(s, t) := 4 (t 2 − t 1 ) (s − t 1 )(t 2 − t), t 1 < s < t < t 2 . One can further simplify the expression in the present case integrating by parts in the exponent; explicitly employing the boundary conditions satisfied by the Green function, we get y(t2)=0 y(t1)=0 Dy e − t 2 A direct computation shows that the second partial derivative involved in the computation is given by Adding all these results toghether we are led to our final expression Extending this analysis to any sign of x and y, and allowing also for a final time T different from one, we prove the following relation A way to obtain a closed expression for this series is to notice that, by using (C.2), we have where we have employed the formal expression of the geometric series. Solving it for T we arrive at the following differential equation T (x, β) = e −x 2 , (D.3) whose more general solution is given bỹ The "constant of integration" c 1 may be fixed by analyzing the behaviour of T for small β, from which we get .
(D. 5) This implies that the desired function is given by 2β .
(D. 6) 10 Notice that if x = 0 or y = 0, then the departing integral in Eq. (C.3) will be in general ill-defined. This is not an obstacle to the computations in the body of this article, given that we always work in a regularized framework, in which we have to consider the limiting cases x, y → 0. | 2022-08-17T01:16:23.558Z | 2022-08-15T00:00:00.000 | {
"year": 2022,
"sha1": "e134c510e42c48105e14033a4d31ce43a73a2d1e",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "e134c510e42c48105e14033a4d31ce43a73a2d1e",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Physics",
"Mathematics"
]
} |
235704602 | pes2o/s2orc | v3-fos-license | The creation of 3D building models using laser-scanning data for BIM modelling
The digital 3D models of the buildings based on laser-scanning data become a vital source of data and information repository to the Architecture, Construction, Engineering and Facilities Management (AEC & FM) sectors. A major advantage of the points cloud data captured by laser-scanning technology is its ability to the representation of the details of the three-dimensional models to exemplify the as-is conditions of buildings. However, the creation process of 3D models from the dense coloured 3D points provided by laser scanners has a significant impact on the quality of that produced models including building edges, walls, doors and windows. In particular, much uncertainty still exists about the compatibilities between the points cloud data formats and software extensions for the creation of as-is Building Information Modelling (BIM) models of the buildings. This paper presents a new framework for the creation of a 3D building model by transferring laser-scanning data into BIM software, such as Autodesk Revit. In our framework, an adopted link between software extensions was established in creating an accurate 3D building model. This framework is a road map of the required steps for investing the points cloud data relevant to BIM modelling. The promising results of the new approach illustrate to extract the 3D models of the buildings can reduce time-consuming, cost and efforts in dealing with or transferring laser-scanning data into Autodesk Revit for BIM models.
Introduction
Buildings are a key component in the different sectors, such as construction, building design and management as well as urban sustainability [1]. The dimensions of the buildings and the details of their outdoor, indoor details as well as their structures became crucial requirements to design, reconstruction, maintain and promote additional levels of building life cycle. Such required details can be exploited in the form of the two and three dimensions (2D and 3D). However, presenting a building in the form of 3D as a 3D model can aid a better understanding of building geometry which in turn can lead to more accurate specifications of the enterprise or scheme requirements [2]. The production of the 3D models of the buildings can also minimise the efforts, time and cost for the geomatics works in an efficient manner [1]. To create a precision 3D model of the building, very accurate data is required. A technology, the same as a laser scanner, provides 3D data that can depict the objects in a delicate way. Laser-scanning data is high-density coloured points which are known point clouds [3]. The high level of accuracy, cost-effective and high level details of point cloud data are essential for a wide range of engineering projects and applications. Because the point cloud data can represent the three dimensions of the objects, this feature facilitates the generation of 3D building models and their IOP Publishing doi: 10.1088/1757-899X/1105/1/012101 2 environments. The increasing attention of point cloud data resulted from it is being able to integrate with BIM tools, such as parametric build models in Autodesk Revit. Nevertheless, the main challenge faced by many users and researchers is to identify the most suitable harmonisations of data format and software (Autodesk Revit) extensions in terms of data processing and analysis. This study is devoted to proposing a general framework for dealing with the point cloud data in BIM tools, such as Autodesk Revit software. Therefore, in the following section, we mainly consider the previous studies aimed at employing laser-scanning data in generating 3D building models and analysis approaches for BIM modelling.
Previous work
More recent attention has focused on the provision of 3D models of the buildings for a number of applications, such as architectural design and visualisation, construction, city modelling and urban sustainability [2,4]. A number of authors have considered the use of laser-scanning data for generating 3D models of the buildings [5][6][7][8]. The advantages of the point cloud data captured by the laser scanner have also motivated researchers and users to develop new methods for BIM modelling. For instance, Hajian H and Becerik-Gerber B [5] reviewed the consecutive stages of generating asbuilt BIM using 3D laser scanning technology and determined the errors and inadequacies at every step. Their investigations concentrated on the accuracy of scanned point cloud registration and the required time for 3D modelling processes. The results of the study showed the creation of 3D models starting from the scanning to the BIM modelling processes is not semantic data and there is a need for identifying a framework as a workflow for managing building's semantic information. The required framework, as they recommended, can connect the collected data to as-built BIMs during maintenance and various operations. Another method was suggested by Mill T et al [6] presented the workflow of creating a BIM model for data management purposes. In their method, the BIM model of a building was generated based on the point cloud data collected from surveying observations including terrestrial laser scanning (TLS) combined with data collected by the total station instrument. Although the BIM model was able to outfit the interior and exterior features of the BIM model, the creation process of the BIM model from different survey data was problematic in processing in terms of deficiency of flexibility when merging different sets of point cloud data. In point cloud computing applications the merger process between tools of BIM software and data formats should be taken into account. As a result, further editing processes of data are going to consume plenty of time and effort. This, in turn, can impact on the quality and accuracy of the produced building model. With a different method, Bosché F et al [7] introduced an automated comparison method in earning value tracking and reducing the discrepancies between as-built and as-planned BIM models of pipes depending on combining two techniques: Hough transform and Scan-to-BIM. The authors claimed that their approach has the ability to improve the recognition of the objects in both the built and planned models as well as the pipe completeness according to a metric. Nonetheless, the proposed method has deficiencies in terms of the correct recognition and identification of pipes as well as the complex integration of two techniques. In addition, their method needs to further improvements that may cause increasing the processing time. Faltýnová M et al 2016 later reviewed many methods that can be possibly used for building documentation based on transferring spatial data (e.g. laser scanning data) into BIM software. Their review work was focused on laser scanning and Photogrammetry in addition to its merits and flaws to the documented building, condition of sites and level of the renovation of the covered methods. The findings showed that both photogrammetric and laser scanning methods are convenient for the construction, project and building documentation. However, in their study, authors have used both software and algorithm for generating the 3D building models or 3D construction elements so that may be very difficult to import into the Revit software for further BIM analysis. In another recent work, Cepurnaite J et al [9] proposed a workflow method for the renovation of the existing buildings based on utilising 3D BIM models in terms of improving the assessment of energy supply, setting the architectural and structural solutions, and devising the integrated solution of different tools for data management and analysis. Unfortunately, in their method, they only mentioned the use of 3D laser scanning data and no details on how was such data integrated into the BIM environment. Another recent research presented by Adán A et al [10], a 6D-based (3D model with 3 IOP Publishing doi:10.1088/1757-899X/1105/1/012101 3 image bands) method to process 3D dense coloured points in obtaining the small structural components in buildings, such as sockets, switches, signs, alarm devices and extinguishers on walls.. The recognition of elements of buildings was detected automatically based on the fusion of the images and geometric algorithms for providing automatic BIM modelling. The method has many limitations, for instance, the position of some objects on the semantic walls of the reconstructed as-is 3D BIM model of the building was not determined as they exist on the real wall. In a new study, Sanhudo L et al [11] introduced a laser framework for the rapid and factual acquisition of the real existing building geometric data that can be used in BIM modelling. The authors modelled a BIM representation of the as-is bus station depending on employing the point clouds which were collected from scanning the station. The point cloud quality and the survey's time and the scanning file size were tested and evaluated. However, in order to process the point cloud data (data registration and cleaning) then export it to the Autodesk Revit (as a BIM software) to create a 3D BIM model of the bus station, Leica Cyclone 9.1 was used for accomplishing that purpose. The problem is that such commercial software is not free in use for all researchers and users and this is a significant constrain to follow and adopt their work. Therefore, our current research helps to fill this knowledge gap in the scope of the study. To date, these few studies that have investigated the association between the utilisation of the point clouds in the BIM environment are still lacking applicability in how to deal with laser scanning data by BIM software in a direct way. Moreover, the creation of 3D building models with the detailed elements of the constructions and accurate structural components for BIM modelling is technically challenging. In particular, it is an irritating issue of any project when attempting to import the different formats of the point cloud data into BIM software. Additionally, further time for delivering the engineering projects in addition to extra effort and cost will be requested when using several software extensions for processing the point cloud data to attain accurate 3D models. Hence, to mitigate these inadequacies by considering all of this evidence, it seems that further studies are required. This paper proposes a new genuine framework to use the point cloud data in the BIM environment for creating a 3D building model. The specific aim of this work is to identify the key steps in importing the point clouds into the Autodesk Revit software (one of BIM environments) without using additional software tools or programme extensions or algorithms. There are three main parts of our contribution: (1) data collection, (2) data processing and cleaning, and (3) modelling and validating applied to the chosen case study. The remainder of this study is presented as follows. The next section (3) is to introduce our methodology. In Section 4, the results and discussion of our approach are presented. Finally, the conclusion and suggestions for potential future works are given in Section 5.
Proposed approach
A summing-up view of our developed framework is demonstrated in Figure 1. Laser scanning data were used in this work. To create a 3D building model based on employing the point clouds by exporting these dense coloured points to the Autodesk Revit software for further processing and structural analysis to the created 3D BIM model. Therefore, the following subsection presents the type of datasets that were utilised and then the following steps will be detailed in the other subsequent subsections.
Laser scanning datasets
Laser scanning is a voguish form of land surveying, capable of accurately measuring and collecting data from various surfaces, features, buildings and landscapes. Laser scanners father data about objects in the pattern of point cloud data, consisting of millions of 3D coordinates (X, Y and Z coordinates). Present-day laser scanners can collect comprehensive and precise point clouds and thereafter these collected data points can construct digital 3D models of the detection zone with point cloud processing software. Photo detectors, laser beams, receiver electronics, advanced sensors, Inertial Measurement Units (IMU), and Global Positioning Systems (GPS) are the components of laser scanners allowing this advanced technology to compute accurate coordinates of structures and their surfaces. Capturing point clouds by modern devices of laser scanning might be through Photogrammetry, such as Light Detection and Ranging (LiDAR) or terrestrial laser scanners (e.g. the devices that are placed on the Table 1 depending on the type of scanners and/or data processing software. In our current work, we used the point cloud data of a multi-level villa as free downloadable sample data from the online sampled data © Copyright -PointCab GmbH 2019. This dataset is captured by Figure 1. The general overview of the present study FARO Focus 3D and it is registered and presented as the coloured point clouds. However, the dataset did not clean yet and has unwanted points' presenting much noise to the building in addition to it is not ready to be processed as a BIM modelling. Therefore, for the aforementioned features, we have chosen it for applying our new framework. Our experiment to set up the framework procedures is demonstrated in the next subsection.
A proposed framework for creating a 3D model
Our proposed steps of the new framework are detailed as shown in Figure 2. First, after searching and collecting the point cloud dataset in a suitable and specific file format, the dataset was downloaded. As previously stated, the dataset represents the multi-level villa with its landscape. The selection of this specific point cloud dataset is to meet the need of two purposes: (1) examining our developed framework of the creation of a 3D BIM model and (2) working on more than one level of the building storeys. Free downloadable, a building geometry, the dataset file format and size were set and identified as criteria for choosing the laser scanning data. These criteria might change from work to another according to the goal of that work. Second, experimental work was conducted to explore and thereafter decide the most compatible, effective and convenient software with the BIM software to import the downloaded dataset. This step is also changeable and it depends on the type of engineering projects, enterprises and work demands. In this stage, Autodesk ReCap™ Pro © 2018 Autodesk Inc. was chosen after many attempts in which we examined some available software to accomplish the task of the generation 3D building model, Table 2. It is important to mention that the reasons behind we established the framework based on using Autodesk ReCap Pro software is because there is some evidence to show that it was possible to meet all aforementioned criteria in an efficient and feasible manner. For instance and most importantly, it is able to perfectly work with Autodesk Revit (BIM software); it deals with wide variety point clouds file formats; it is easy to install (as a software version for students); it has many tools, options and functions to process the point cloud data, and last but not least, Autodesk ReCap Pro is capable of exporting the point clouds as one united file presenting 3D building geometry with two common distinct file format extensions: (.rcp) and (.rcs), as shown in Table 3. In the second step, we managed to open, modify, merge, eliminate and finally extract the 3D model in the one combined file that can be exported to Autodesk Revit for further processing 3D BIM model. Third, the next step in the framework for generating the 3D model is to employ the 3D dense coloured points within the BIM environment. This final stage involves an essential step in terms of the integration of the generated 3D model into the BIM software and the validation of our framework. Autodesk Revit student version © 2019 Autodesk Inc. is considered one of the most powerful BIM software and its environments for producing organised, reliable and detailed model-based designs. Regarding this point, Revit® was used to obtain a 3D BIM model and conducted some fundamental structural computations and analysis for evaluating the proposed framework (in Section 4). We managed to model with accuracy and precision the point clouds sample that we chose to be a 3D building providing 3D geometry to all storeys structures and construction elements. As this paper is scheduled, the following section will present the results and discussion of the present work.
Results and discussion
Driving accuracy and efficiency works across the project lifecycle starting with fictional designs and ending with the real construction through visualising and analysing every single element is a core of building sustainability. The digital recording of the geometry and location IOP Publishing doi:10.1088/1757-899X/1105/1/012101 6 of buildings is an essential mechanism for long-term permanence [12]. Data that is remotely collected (e.g. the devices of remote sensing technology) most often aids to provide reliable information in addition to it is considered a highly vital data source that cannot replaceable [13]. In this vein, as a matter of fact, the outputs of this current research are that laser scanners (one of the remote sensing devices) provide high-quality data with a high level of details on any surveyed object. Table 2. The evaluation of software for processing the selected datasets including: Availabilityfree of use; Flexibilitythe ability to deal with wide range of data format; Independencythe ability to run and execute the processing without utilising other software; and lastly, Compatibilitythe ability to work with Autodesk Revit (BIM modelling) We were capable of employing the point clouds for the BIM modelling as a reference without problems during data processing in both Autodesk ReCap Pro and Revit. In fact, no other sources of the geometry extraction are being utilised for this purpose as same as the point clouds can do. The first step of the developed framework was to determine the suitable data for creating a 3D BIM model representing a building, as shown in Figure 3. However, the chosen point clouds contain multiple storeys of the building with its landscape. To attain that detached building, the landscape, which, however, hid the building by the surrounding trees, fence and the electric poles were removed Figure 4. Figure 4 also shows the step of data processing in Autodesk ReCap Pro. The pre-final merged model has been exported in both extensions (.rcp) and/or (.rcs), this, in turn, facilitated the merged 3D point clouds be processed in Autodesk Revit, as seen in Figure 5. Thereafter, in the Revit environment, the 3D BIM model was created in Figure 6. In this step of the proposed framework, the creation process of the 3D building model was based on a new technique in combining the merged dataset with the created BIM model for matching and marking the axes of the building, aligning the building boundaries and edges of the walls, and the floor heights, Figure 6 and 7. The assessment step of how much the 3D BIM model matches the 3D building model captured by the point clouds was also achieved within the former procedure. The façade surface was modelled entirely utilising cloud data from the laser scanning stage. Because Revit does not automatically detect the best fit for the façade surface position, the modeller manually selected the position. As a result, it can be very strenuous to manually pick the right position for the surface, particularly if the surface is irregular surface. The various structural and architectural parts of the building were built by applying Revit's commands, such as walls, stairs, doors, windows, roof and slabs, Figure 7. within the 3D point cloud model. Further, the final version of the 3D BIM model was having a geometry discrepancy in some parts of the building. One possible explanation can give is that the elimination of the additional point clouds which display greater details on the specific object may cause losing the accurate representation of its geometry.
Figure 8.
The final 3D BIM model after modelling all structural and architectural parts comparing to the 3D model of the building using the laser scanning data with a sample of the basic constructional calculations and structural analysis for the succeeding BIM applications
Modelling time
All processing work and modelling were performed on a laptop computer (Lenovo W520) with a CPU Intel Core i7-2720QM processer with 2.20 GHz and 4 GB RAM. We used NVIDIA Quadro 1000M graphic card. Although the point cloud file size was 1.1 GB, the estimated total performance time was eight hours.
Conclusion
This article presented the creation framework of a 3D model of a building using laser scanning point clouds for BIM modelling. The framework is considered as workflow steps of a prompt and accurate acquisition of the 3D geometric information of any building that can help users and researchers to be used in the BIM environment. Three key stages were included in this work. First, the point cloud data 10 was collected based on the purpose of the engineering project, file format and its availability. Second, Autodesk ReCap Pro was used to open, cleaning and processing data to obtain the 3D building model composed of the point clouds. Third, the final 3D model and the structural computations applied to the model were attained using Autodesk Revit functions. The method can mitigate the problem of dealing with the different data file format and extensions. However, there is still a need to do further work in determining the compatibilities of various software that deal with laser scanning data to avoid the complex processes and reduce the time, efforts and cost.
In the future, two goals are planned to achieve which are the use of different formats of the point cloud data that will be surveyed and captured by the laser scanner device for a particular building. Additionally, we are decided to experiment with a different design and shape of the building that prevails in the region. | 2021-07-02T20:02:26.625Z | 2021-06-01T00:00:00.000 | {
"year": 2021,
"sha1": "ac7a750d011e70ae9ee4aa88004d7a4a239c1a43",
"oa_license": null,
"oa_url": "https://doi.org/10.1088/1757-899x/1105/1/012101",
"oa_status": "GOLD",
"pdf_src": "IOP",
"pdf_hash": "ac7a750d011e70ae9ee4aa88004d7a4a239c1a43",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Physics"
]
} |
236966183 | pes2o/s2orc | v3-fos-license | Tumor growth suppression by implantation of an anti-CD25 antibody-immobilized material near the tumor via regulatory T cell capture
ABSTRACT In this study, we designed and synthesized an implantable anti-CD25 antibody-immobilized polyethylene (CD25-PE) mesh to suppress tumor growth by removing regulatory T cells (Tregs). The PE mesh was graft-polymerized with poly(acrylic acid), and the anti-mouse CD25 antibody was then immobilized using the 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide reaction. Immobilization of the antibody on the PE mesh was confirmed by immunostaining. The CD25-PE mesh could effectively and selectively capture CD25-positive cells through antigen-antibody interactions when the CD25-PE mesh was incubated with a suspension of mouse spleen cells, including CD25-positive cells. In addition, implantation of the CD25-PE mesh into mice subcutaneously demonstrated the Treg-capturing ability of the CD25-PE mesh with only a weak inflammatory reaction. In tumor-bearing mice, tumor growth was suppressed by subcutaneous implantation of the CD25-PE mesh near the tumor for 1 week. These results suggested that the anti-CD25 antibody-immobilized material could capture Tregs in vivo and inhibit tumor proliferation in a limited tumor-bearing mouse model. Further research is needed to facilitate cancer immunotherapy using implantable anti-CD25 antibody-immobilized material as a Treg-capturing device.
Introduction
Cancer treatment can be classified into surgical treatment, chemotherapy, radiotherapy, and immunotherapy. Cancer immunotherapy is a method for treating cancer using the immune system. To date, various cancer immunotherapies have been proposed, including vaccine therapy using autologous cancer vaccines [1], dendritic cell vaccines [2], and adoptive immunotherapy using natural killer (NK) cells and cytotoxic T cells [3]. Among these approaches, cancer immunotherapy related to regulatory T cells (Tregs) has recently become a major research focus. Tregs, i.e., CD4-, CD25-, and FoxP3-positive T cells, are key players in immune suppression [4] and function by controlling the activation of antigen-presenting cells via cytotoxic T lymphocyte antigen (CTLA)-4 and immunosuppressive cytokines (e.g., interleukin-10). In addition, Tregs play roles in suppressing the attack of T cells and other immune cells by modulating the production of transforming growth factor-β [5]. Furthermore, in the tumor microenvironment, which is formed by various components, including cancer cells, immune cells, and the extracellular matrix, Treg accumulation is induced by secretion of the chemokine C-C motif chemokine ligand 22 (CCL22) from cancer cells and tumor-infiltrating macrophages, resulting in an antitumor immune response [6,7]. Several treatments that inhibit immunosuppressive signal transduction by immune checkpoint inhibitors (e.g., anti-CTLA-4 and anti-programmed death-1 antibodies) and depletion of Tregs by administration of anti-C-C motif chemokine receptor 4 antibodies have been proposed as Treg-related cancer immunotherapies [8,9]. The development of selective Treg removal methods is also proposed [10,11]. Although the efficacies of these treatments have been demonstrated, treatment with immune checkpoint inhibitors can induce serious side effects owing to activation of T cells [12]. In addition, because Tregs are strongly related to autoimmunity, Treg-removing treatments may cause systemic autoimmune diseases. Therefore, the development of a method for local Treg removal at the tumor is essential.
In our previous reports, we developed an antibodyimmobilized material for the selective capture of immune cells, including Tregs [13][14][15]. The antibodyimmobilized material consisted of a grafted polymer and an antibody, in which the selective capture of target cells was achieved based on the nonadhesive properties of polymer grafting and the antigen/antibody interaction.
In this study, for the development of novel cancer immunotherapies related to Tregs, we designed and synthesized an implantable anti-CD25 antibodyimmobilized polyethylene mesh and investigated its properties, including selective Treg capture in vitro and in vivo and ability to suppress tumor growth.
Preparation of anti-CD25 antibody-immobilized mesh
Scheme 1 shows the preparation of an anti-CD25 antibody-immobilized mesh. After Soxhlet treatment with ethanol at 60°C for 8 h, corona discharge treatment was performed on both surfaces of the PE mesh (15 kV, 1 min). The mesh was immersed in 2% and 5% acrylic acid solutions. After degassing, the mesh was subjected to heat polymerization in a water bath at 60°C for 30 min. After washing the meshes with hot water, the poly(acrylic acid) (PAAc)-grafted PE meshes were immersed in a solution of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC, 10 mg/mL) and anti-mouse CD25 antibody (0.5, 5, 50 µg/mL) for 2 h at room temperature.
Evaluation of the antibody-immobilized PE mesh
The PAAc-grafted PE mesh was evaluated by methylene blue staining (0.02%) and weight measurements before and after polymerization. The anti-mouse CD25 antibody-immobilized PE mesh (CD25-PE) was stained using an immunostaining kit (POD Conjugate Anti Rat, for mouse tissue; TaKaRa) following the manufacturer's instructions. The amount of antibody immobilization was measured using ImageJ software. The obtained color image of the antibodyimmobilized PE mesh was converted to a monochrome image and inverted black and white. The gray values of the mesh were then measured.
Treg capture using the anti-CD25 antibody-immobilized PE mesh in vitro
All experiments involving mice were conducted using protocols approved by the Institutional Animal Care and Use Committee of Tokyo Medical and Dental University (approval number: A2019-313C). A spleen was harvested from a mouse (5 weeks of age) and placed into cell strainer (pore size: 70 µm), which was set on a dish of 6 cm diameter with 15 mL of red blood cell lysis buffer (BD Biosciences, San Jose, California, USA). The spleen was made some incisions and mashed through the cell strainer into the dish using a plunger end of a syringe. The spleen cells were transferred to a conical tube of 50 mL and centrifuged at 300xg, 4°C for 10 min. After removal of the supernatant, the cells were suspended with phosphate-buffered saline (PBS) including 2% bovine serum albumin (BSA) and were filtered with cell strainer (30 µm). A suspension of spleen cells (1.0 × 10 6 cells/mL; total volume: 0.5 mL) was prepared. The meshes were blocked with 2% BSA in PBS at room temperature for 30 min, and five meshes (diameter: 8 mm, area: 0.96 cm 2 ) were placed in a tube of 2 mL Spleen cells were suspended and incubated with gentle shaking at 4°C for 1 h. The number of cells that had not adhered to the mesh was measured, and the cell capture rate was calculated by dividing the number of adhered cells by the number of seeded cells.
Flow cytometric analysis of spleen cells
Mouse spleen cells were mixed with FITC-labeled antimouse CD25 antibodies and APC-labeled anti-mouse CD4 antibodies and incubated for 30 min in an ice bath. The cells were washed twice with PBS containing 0.2% fetal bovine serum and subjected to flow cytometry analysis (Novocyte, Agilent Technologies Japan Ltd., Hachioji, Japan).
Implantation of the anti-CD25 antibody-immobilized PE mesh into healthy mice
Mice (5 weeks old) were anesthetized, and an anti-CD25 antibody-immobilized PE mesh was subcutaneously implanted into each mouse. After 7 days, the mice were euthanized, and the mesh and skin were extracted. The experimental protocol was approved by the animal ethics committee of our institution (Animal experiment approval number: A2019-313 C).
Hematoxylin-eosin (H-E) staining and FoxP3 immunostaining of implanted anti-CD25 antibody-immobilized PE meshes
The extracted mesh and skin were fixed with 10% neutral formalin buffer, embedded in paraffin, and sliced to a thickness of 4 µm. The slices of the samples were stained with H-E. The number of cells within 30 µm of the mesh fiber was then counted, and the slices were subjected to antigen activation treatment with ethylenediaminetetraacetic acid. Samples were immersed in 3% hydrogen peroxide solution at room temperature for 10 min, blocked with PBS (3% BSA) at room temperature for 1 h, and then mixed with anti-FoxP3 antibodies at 4°C overnight. DAB staining was performed using a Simple Stain MAX-PO (Rat) and DAB substrate kit. The numbers of cells and Tregs within 30 µm of the fiber were measured, and the Treg ratio was calculated.
Statistical analysis
Results were expressed as the mean ± standard error of the mean, with each experiment performed at least three times. Tukey's multiple comparison test was used to test for statistical significance. A p-value < 0.05 was considered to be statistically significant.
Preparation of an anti-CD25 antibody-immobilized mesh
The PE mesh was subjected to corona discharge treatment and graft polymerization with acrylic acid. The PAAc-grafted mesh was evaluated by methylene blue staining and weight measurements before and after polymerization (Figure 1(a,b)). The blue-stained mesh was observed at an acrylic acid concentration of 2 wt% and became dark blue at an acrylic acid concentration of 5 wt%. In addition, the weights of the mesh before and after graft polymerization increased with increasing acrylic acid concentration. Subsequently, anti-CD25 antibody was immobilized on 2 and 5 wt% PAAc-PE using the EDC reaction. DAB-stained meshes are shown in Figure 2(a). For PE meshes without PAAc graft-polymerization, white meshes were observed at anti-CD25 antibody concentrations of 0, 0.5, and 5 µg/ mL, whereas brown-stained meshes were observed at an antibody concentration of 50 µg/mL, indicating the nonspecific absorption of the antibody on the PE mesh. For the 2 and 5 wt% PAAc-PE meshes, meshes became dark brown as the concentration of anti-CD25 antibody increased. Also, the anti-CD25 antibodyimmobilized PE mesh grafted with 5 wt% PAAc was darker than that grafted with 2 wt% PAAc. So, we used the anti-CD25 antibody-immobilized PE mesh grafted with 5 wt% PAAc for below experiments. The DABstained meshes containing 5 wt% acrylic acid were analyzed using Image J (Figure 2(b)). An increase in the immobilization amount of antibody was observed with increasing antibody concentrations.
In vitro cell capture of antibody-immobilized mesh
Mouse spleen cells were used to investigate the cell capture of anti-CD25 antibody-immobilized mesh. Mouse spleen cells were assessed by fluorescenceassisted cell sorting (FACS; Fig. S1). The percentages of CD25-positive and CD4-positive cells were approximately 8% and 18.5%, respectively. The anti-CD25 antibody-immobilized meshes (PAAc: 5 wt%, feed conc. of anti-CD25 antibody: 50 µg/mL) were set in suspensions of mouse spleen cells (1 × 10 6 cells/mL) and incubated with gentle shaking at 4°C for 1 h. Cells that had not adhered on the meshes were counted, and the cell adhesion rate was calculated ( Figure 3). For PE, 5% PAAc-PE, and anti-human CD25-PE meshes, there were no significant differences between cell capture rates, indicating that the cells adhered nonspecifically. In contrast, the number of adhered cells was increased for anti-mouse CD25-PE and anti-mouse CD4-PE compared with that of other meshes, indicating that cell capture on antibody-immobilized meshes involved an antigen-antibody interaction.
Implantation of the anti-CD25 antibody-immobilized mesh into healthy mice
To investigate the in vivo reaction of the anti-CD25 antibody-immobilized mesh, the mesh was subcutaneously implanted into healthy mice (Figure 4(a)). When PE and 5 wt% PAAc-PE meshes were implanted, red color, as a sign of the inflammatory reaction, was observed. In contrast, there were no significant differences between the sham operation and CD25-PE meshes when various concentrations of antibody were used. The implanted meshes were subjected to H-E staining (Figure 4(b)). Cells within 30 µm of the mesh fiber were observed and counted (Figure 4(c)). The highest cell accumulation was observed for the PE mesh. Compared with PE, the cell number decreased for the PAAc-PE and CD25-PE meshes. There were no differences between the CD25-PE meshes at various concentrations of antibody.
Subsequently, the implanted meshes were subjected to immunostaining with an antibody targeting FoxP3, a specific marker of Tregs (Figure 4(d)). The number of Tregs within 30 µm of the mesh fiber was counted (Figure 4(e)). Notably, Tregs were not observed around the PE fibers, but were observed for all CD25-PE meshes. The percentages of Tregs increased as the antibody concentration increased.
Suppression of tumor growth by implantation of the CD25 antibody-immobilized mesh into tumor-bearing mice
To examine the tumor-suppressive effects of the anti-CD25 antibody-immobilized mesh, CD25-PE mesh was implanted near the tumor in tumor-bearing mice ( Figure 5(a)). The implanted meshes were observed near the tumor after 7 days of implantation, although a slight shift in position was observed for the PE meshes. Changes in tumor volume over time are shown in Figure 5(b). For the sham operation, the tumor volume was increased to approximately 600 mm 3 after 7 days. For PE and PAAc-PE mesh implantation, similar tumor growth was observed until 4 days after implantation, and tumor growth was inhibited at 6 and 7 days after implantation compared with that of the sham operation. For the CD25-PE mesh, the greatest suppression of tumor growth was found after 7 days, and the tumor size rapidly increased between days 6 and 7. The implanted CD25-PE mesh was subjected to H-E staining and FoxP3 immunostaining ( Figure 5(c,e)). Although the CD25-PE mesh was slightly farther away from the tumor, Tregs were observed around the fibers of the mesh. The number of cells around the fiber of the mesh was counted, and the Treg ratio was calculated ( Figure 5(d, f)). Compared with the CD25-PE mesh implanted into healthy mice, the number of cells was increased for CD25-PE meshes implanted into tumor-bearing mice ( Figure 5(d)). There were no differences in Treg ratios between healthy and tumor-bearing mice ( Figure 5(f)).
Discussion
Cancer immunotherapy can involve enhancement of immune attack or release of immune tolerance [17]. Importantly, FoxP3 + CD25 + CD4 + Tregs have been shown to strongly affect tumor immune tolerance [4]. In the tumor microenvironment, Tregs infiltrate and accumulate through the secretion of chemokines, such as CCL22, from cancer cells and inhibit the attack of immune cells, such as NK cells and effector T cells, to cancer cells [4,5]. In this study, in order to remove Tregs from the tumor and release immune tolerance, we designed and prepared an antibody-immobilized mesh and subsequently investigated the effects of the mesh on tumor growth suppression following implantation in mice. The transcription factor FoxP3 is a specific marker of Tregs and is FoxP3 expressed inside the cells; thus, this specific marker could not be used in our system. CD25 is a surface marker of Tregs; therefore, we chose anti-CD25 antibodies in this study. A recent study showed that three types of Tregs, including naïve Tregs, effector Tregs, and non-Tregs, were present and exhibited variations in immunosuppression. Among these types of Tregs, effector Tregs, which show high immunosuppressive ability, are present as the CD25 ++ FoxP3 ++ cell population. Moreover, effector Tregs have been identified in proliferative tumors [18]. Therefore, in this study, we designed an anti-CD25 antibody-immobilized material to remove the effector Tregs from tumors.
The anti-CD25 antibody was immobilized on the surface of PAAc-PE mesh with acrylic acid concentrations of 2 and 5 wt% through the EDC reaction. The amount of immobilized antibody increased as the concentration of the antibody increased for both acrylic acid concentrations. However, the non-graftpolymerized PE was stained by immunostaining, suggesting nonspecific absorption of the antibody. Therefore, we used the anti-CD25 antibodyimmobilized PE mesh at an antibody concentration of 5 µg/mL for subsequent experiments. Analysis of the cell capture ability of the obtained CD25-PE mesh in mouse spleen cells (CD4 − CD25 − , CD4 + CD25 − , CD4 − CD25 + , and CD4 + CD25 + cells) showed cell capture ratios of approximately 14% and 20% for PE and 5 wt% PAAc-PE meshes, respectively. We assumed that the cells adhered to the meshes nonspecifically because the cells were observed at the crossover region of the mesh fibers (data not shown). Although the cell capture ratio of the anti-human CD25-PE antibody was still approximately 20%, an increase in the cell capture ratio was found for antimouse CD25-PE (37%) and anti-mouse CD4-PE (46%) meshes. These results indicated that cell capture occurred through antigen/antibody interactions on the antibody-immobilized PE meshes in addition to nonspecific cell capture. For FACS analysis, the rates of CD25-positive and CD4-positive cells were approximately 8% and 18%, respectively. Unexpected increases in the cell capture ratios for the anti-mouse CD25-PE and anti-mouse CD4-PE meshes were observed. Thus, further studies are required to clarify the mechanisms involved in these effects.
Analysis of the Treg capture ability of the CD25-PE mesh in vivo demonstrated inflammation and cell accumulation around the mesh fibers of the PE mesh owing to the foreign body reaction. In contrast, for the CD25-PE mesh, the inflammation reaction was not observed, and sham operation suppressed cell accumulation. In addition, although Tregs were not observed for PE meshes immunochemically stained with FoxP3, Treg accumulation around the CD25-PE mesh was also achieved. Surface modification of implant materials with biomolecules, such as peptides, proteins, and antibodies, can control cell behavior and function [19][20][21][22]. Thus, Tregs accumulated around the CD25-PE mesh may inhibit inflammation through immunosuppression.
Finally, we investigated the tumor-suppressive ability of the CD25-PE mesh in tumor-bearing mice. Our results showed that the PE and PAAc-PE meshes caused decreased tumor volumes only on days 6 and 7. It is assumed that the implantation of them induced changes of the tumor surroundings, such as micro-circulation state, capsulation, inflammation, and it is required to clear the mechanism in future. In contrast, CD25-PE resulted in high suppression of tumor growth until 7 days after implantation compared with the other meshes; the tumor volume decreased to 323 mm 3 , indicating that this mesh showed high tumor suppression ability. We assumed that the immune cells suppressed by Tregs in or around the tumor were activated by capturing Tregs on the CD25-PE mesh. This assumption was supported by the observation that the number of cells around the CD25-PE mesh of tumor-bearing mice was higher than that of healthy mice. In addition, spherical immune cells were observed for the CD25-PE mesh near the tumor, whereas the cells around CD25-PE meshes implanted into healthy mice were spindle-shaped cells, similar to fibroblasts. However, for FoxP3 immunostaining, the Treg ratio among the cells around the CD25-PE mesh was approximately 5%, similar to the CD25-PE ratio in healthy mice. Furthermore, although the tumor growth curves in mice in the CD25-PE mesh group gradually increased over 6 days after implantation, a remarkable increase in tumor volume was observed from days 6 to 7. These results suggest that it may be necessary to consider long-term application of CD25-PE mesh and that additional improvements to the CD25-PE mesh may be required.
Conclusion
In this study, we designed and synthesized an anti-CD25 antibody-immobilized PE mesh as an implantable Treg capture device to suppress tumor growth by releasing immune tolerance. The CD25-PE mesh effectively captured Tregs via antigen/antibody interactions in vitro and in vivo. Additionally, transplantation of the antibody-immobilized mesh into tumor-bearing mice suppressed tumor growth. These results suggest that such Treg control may have applications as a novel type of cancer treatment. Further research is needed to facilitate cancer immunotherapy using implantable anti-CD25 antibody-immobilized material as a Treg-capturing device. | 2021-08-11T05:25:10.932Z | 2021-06-22T00:00:00.000 | {
"year": 2021,
"sha1": "581af0965633f04db83c3609ead301d2b9429bb7",
"oa_license": "CCBYNC",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.1080/14686996.2021.1944782?needAccess=true",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "581af0965633f04db83c3609ead301d2b9429bb7",
"s2fieldsofstudy": [
"Medicine",
"Engineering"
],
"extfieldsofstudy": [
"Medicine"
]
} |
158532662 | pes2o/s2orc | v3-fos-license | The modernization of the educational system in France : the New Public Management between the affirmation of the State and the decentralized government a modernização do sistema educacional na França : a Nova Gestão
This text examines the process of modernization of the Educational System in France. It aims to understand the affirmation of the State and the decentralized government in the context of the discussion of the New Public Management. It examines the legacies, narratives, and policies of modernization, as well as the paradoxes of the new French public administration in education. It is questioned if the New Public Management was, in fact, implemented in the French educational system. It is stated that, in education, only administrative and financial responsibility entered the institutions without making many consequences on the work of teachers. The main teachers and supervisors are signing contracts and developing evaluations and audits, but the bureaucratic structure still prevails. It is understood, however, that a third way, between the State and the market, is being sought, especially regarding decentralization, with the emphasis on the most shared local responsibilities and the possible changes in the status of civil servants. _____________________________________________________________ Keyword: New Public Management; educational system in France; decentralized government.
INTRODUCTION
The French education administration is proud of its traditions dating back to the Enlightenment and stabilized by the Napoleonic Empire.A lot of educational plans were published during the second part of the 18th century affirming the following principle: education is the affair of the State; it does not concern families, communities and even less so religious congregations.This led to a great mistrust at the local level which was reinforced by State planning in the 1960s.The latter defined a school catchment area for the registration of pupils according to the location of their home.The same Statist concern inspired the definition of a curriculum focusing on academic disciplines and access to universalism.This French republican tradition is held by key professional bodies such as the Inspection Générale or embodied in the Agrégation (a special selectionbased exam to become high-ranking teacher in French secondary education).It has also penetrated the culture of the teaching profession and the trade-unions' countervailing power.All school modernization projects had to adjusted and adapted to this framework: it is the case for French comprehensive schools but also for the implementation of New Public Management.
The education system is also largely public.A private system, Catholic in its great majority, enrols about 17% of schoolchildren.Since the separation of the Church and the State (1905), this private education was no longer subsidised.However, at the beginning of the 5th Republic political regime, the Debré Act established a compromise (1959): a private and contracted education is subsidized but it has to respect State regulations by teaching the same curriculum, by being inspected and in giving the same training for teachers.
From the mid-70s to the 1980s, French comprehensive school system (college unique) was implemented by both right-wing and left-wing governments.This policy was a promise of democratization for many educators and parents but it did not fulfil their expectations and hopes (Derouet 1992).It led to a crisis of trust in the school system while some intellectuals from both the Left and the Right denounced the "false democratization" in secondary education which led, for some of them, to the "defeat of reflection".It was also challenged by claims for the recognition of ethnic and religious differences affirmed during the last years of the 20th century (Honneth 2000, Fraser 2013).The mistrust of the French republican tradition towards communities and multiculturalism remains an obstacle in intellectually grasping the issue.In a period in which the European Union is putting stock on the social inclusion of minorities, some parts of French society are tempted by a move back to more traditional definitions of the Republic and secularism.These legacies will be described in the first part of this chapter.
The pessimistic climate regarding the failure of the comprehensive school system facilitated the introduction of some recommendations related to accountability from international organizations (Éducation et Sociétés-29 2012).However, the Left and the Right mainly remain hostile to market ideas in education.They also have doubts about management and managerialism.Traditionally, the Left is attached to the civil service and mistrusts entrepreneurial and managerial discourses.However, there is another Left which is promoting some new ideas about governance, decentralization, local democracy and less State intervention.But it has not lead to French policy-makers converting to liberalism and free market policies: some of them are only reacting against the bureaucratic State and claims for more efficiency and quality.Overall, the French education system has included some principles of New Public Management in its bureaucratic tradition (Bezès 2009).
All these issues refer to different values and intertwined national and international political agendas.It is therefore difficult to characterize legacies and changes from the last fifty years.However, it is possible to provide the following analysis regarding the introduction of New Public Management into the French administration: a period of relative openness from the beginning of 1980s.It was marked by the general law on decentralization, voted in 1981, and by a new definition of justice imported from British examples (Derouet et Derouet-Besson 2008).This movement is at the root of the Education Priority Areas and school autonomy policies.The first Lisbon conference (2000) amplified the movement by instigating France to take into account the key European recommendations.The first measure linked to this new direction was the Institutional Act related to Finance Laws (Loi Organique relative aux Lois de Finances (LOLF), unanimously voted by Parliament in 2001, and which proposed a new organization of public services based on accountability.A second proposed direction is a basic skills framework which replaced the structuration of curriculum in disciplines (2005 Act).This conception is far from the French tradition and its real impact among practitioners can be questioned, but now basis skills are now part of the professional and political culture.Comparatively, the past few years have given the impression of a closure of national identity.The traditional conception of the Republic is threatened and the governing socialist party has returned to its fundamental principles: public service, centralization, secularism (Lawn & Normand 2014).
In France, there is no regulation by the market, no business, and no high-stake accountability system.Data are provided by the ministry of education to compare student outcomes but they are used to measure inequality of opportunities and not the performance of teachers and students.The idea of performance is mediated through a bureaucratic apparatus linked to the reform of the State beyond education.The words of "Management" and "managers" (Clarke & Newman 1997) do not fit the representation of executives who consider mainly they are civil servants respecting and applying regulations from the State.The LOLF proposes a general restructuring of public services but it has more impacted on accountancy procedures than on actors and schools.LOLF indicators are guiding the action of principals and inspectors but they have no influence on the conditions of teaching and learning which leave a great professional autonomy for teachers.A soft accountability is however emerging through the changing missions of the bodies of inspection who develop more audit and selfevaluation procedures but it remains on an experimental and non-statutory basis.The French education system is entered in a post-bureaucratic regime and has implemented its first standards in curriculum, literacy and numeracy.But the idea that schools could make difference is limited to issues about the school climate considered as a mean to fight against violence, drop-outs and social exclusion.There are no proposals about linking curriculum, assessment and performance.The French public management is a mix of modernization and conservative values inherited from the legacy of the Republic school system: neutrality of the State, equality of opportunity, common citizenship.It maintains its tradition of centralized standardization and it is blind to the recognition of differences and local particularities.It gives a powerful influence to professional bodies and trade unions at the summit of the State while New Public Management reform remains a top-down and loosely process.Even the reform of decentralization and the development of national assessments began in the 1980-90s has been slown down during the last decade.The NPM in France is characterized by a set of paradoxes which are explained throughout this chapter.It is a singular case in the European landscape of NPM reforms and it must be considered as so.It is also necessary to explain some legacies and narratives which characterize this particular situation.
I. LEGACIES, NARRATIVES AND POLICIES OF MODERNIzATION
The last decades saw an intensive legislative activity.The 1975 School Modernization Act created the comprehensive school system (collège unique).The notion of a "school development plan" was enshrined by the 1989 Act which remains the backbone of the new education system's regulation.This act is an umbrella law which fixes the key principles but gives certain autonomy at the local level.However this type of compromise, inspired by progressive education and "placing the pupil at the centre of the education system", has not really been understood and has even been refused by the great majority of the educative community for whom the transmission of knowledge, teaching and not learning, has to be the main concern of the school system.In the end, society at large was not only disappointed by the poor performances of the comprehensive schools in the reducing inequalities, it was also worried about the effect of extensive schooling: the school system had not brought about the social advancement that was expected.But there has also been disillusionment regarding were the achievement of pupils.The first publications of international surveys were not reassuring for the pessimistic.The republican link between school and society was broken.
Beyond these uncertainties of French society, the Lisbon Conference (2000) introduced some elements which were implemented into the objectives of the 2005 Act.The Right introduced a basic skills framework inspired by the European key competencies framework and defined by the European Commission for Lifelong Learning.When the Left came to power in 2012, it promulgated an Act for the Re-foundation of the School System.This title expresses the feeling of a loss of direction in French society regarding its education and the will to return to the neo-Kantian tradition of the Republican school's founders in the beginning of the 1880s.The secularist passion, which had faded with the decline of the Catholic Church, has regained power in the face of Islamic fundamentalism (Éducation et Sociétés-33 2014).
This legislative activity was supported by the creation of policy tools in charge of its implementation (Normand & Derouet 2016).
THE DECENTRALISATION AND THE AUTONOMY OF SCHOOLS
The attempt to decentralize followed a reflection after the 1968 movement about the possibility of schools becoming a school management unit while the centralized school system, with its million civil servants, was often compared to the Red Army.During the beginning of 1970s, a significant number of measures were experimented with but the 1975 Act ended this shift and France returned to the tradition of State planning.France put into place the comprehensive school system later than other OECD countries.The notion of school autonomy, which had been conceptualized from a pedagogical perspective, then took a managerial meaning.This new idea of a school development plan was introduced in 1982 as an experiment during the reform of junior schools.In 1984, a decree in the general Decentralization Act gave every secondary school the status of Public Local School with the possibility for the Board to define its school development plan.The Left added a social objective: adapting teaching methods to pupils' needs in order to prevent school inequalities.However, a certain managerial vision remains and it was inspired by the ideas of the sociologist Michel Crozier.The title from one of his books summarizes his thoughts: "Modern State, Earnest State" (1986).
While the notion of a school development plan was extended through the entire education system, the 1989 Act became the pillar of the new regulations.It proposed to establish a series of individual and moral contracts between the pupil, the school and his/her family without renouncing the concept of the school catchment area.However, the law introduced some possibilities of limited and framed school choice for families which did not accept the school-based project.It was recognition of the rights of families without moving towards a market-based system.Another limitation of this autonomy was the "untouchable" national curriculum.Autonomy was therefore quickly limited to a local and narrow management with no flexible means to achieve national objectives.
The 2005 Act attempted to revive the principle of school autonomy via a cautious liberal conception.The main measure was the creation of a "pedagogical board": the trade unions refused to allow issues on teaching to be discussed at the administrative board level as a lot of board members have no competency in this field.But the pedagogical board, which only includes teachers, could manage the national curriculum and local teaching conditions.However, its implementation has been long and difficult and has resulted in disappointing effects.The 2005 Act also took on board the recommendations of international organisations regarding the diversification of the school curriculum as a means to promote effectiveness and equity.Article 34 of the Act scheduled some possibilities for innovation by allowing schools to have more freedom outside of national regulations.This measure could help some schools create a specific identity but their choice and school development plans, with the notable exception of a few of them, were not really new or creative.According to the same logic, in 2007, the Minister announced more flexibility in the school catchment area policy and its abolition was scheduled in 2010.After a lot of heated debate, certain changes were made to this announcement.Local authorities, which are mostly against school choice, did not implement these instructions and it has continued to limit the possibilities of those families wishing to work outside of its scope.The Left, back in power, overturned the policy and has reinforced the catchment area policy.
THE DEVELOPMENT OF ASSESSMENT
The period is characterized by the implementation of an assessment system.It is the result of a long history.During the 1970s, the former system of administrative statistics evolved towards new missions and objectives developing an assessment system.In 1986, this administrative service became a ministerial directorate: the Directorate of Assessment and Forward Planning (Direction de l'Évaluation, de la Prospective et de la Performance (DEPP).It successive heads have shared the same thoughts about in-depth large-scale surveys and the culture of the State's statistics.The DEPP had important responsibilities.It reassures those who fear that school autonomy would lead to a loss of control in the steering of the education system; the law enabled the assessment system to prevent "some possible drifts".In response to society's concerns about the quality of learning, the DEPP also had the responsibility of implementing regular assessments of pupils' skills at different key-stages of the education system.All these missions were embedded in a certain conception of the education policy: the aim was to design tools for the State via indicators built from a national perspective which would take into account the diversity of local practices (Derouet & Normand 2010).
On behalf of its forecasting mission, the DEPP tendered several calls for educational research.The first one, in the late 1980s, regarded the return on investment in education and subtlety introduced the principles of French accountability.The return on investment was not only measured through performance but through the reduction of inequalities of opportunity.The second mission concerned the educational investment of families.This call was in the slipstream of the emergence of a movement of school consumers and new choices for private schools.The latter highlighted the reality and importance of school violence.All the research findings from the selected projects were presented in DEPP reports disseminated and summarized by the press and media.After this prosperous period, the DEPP's missions of were revised and reduced at the end of the 1990s: it had gained too much influence in comparison to other directorates and even the Minister himself or herself, and had to reintegrated into the rank and file.The debate was shifting: was it normal that assessment was led by a Ministry which designed and implemented education policies?Diverse reflections were inspired by Scandinavian examples where evaluative institutions are placed under the watch of Parliament.French policymakers are not entirely familiar with this concept.Even the word "agency" is considered by them to be too liberal and they prefer "high councils" which maintain a strong dependency on the State.A National Council for the Assessment of the School System (Conseil National d'Évaluation du Système Scolaire (CNESCO) was created in 2014.An academic was appointed President of the Council by the Minister, but all the resources are provided by the Ministry's departments.Moreover, this new council has not abolished the previous ones: the DEPP remains active and the Inspectorate is still in charge of assessing teachers and schools.
This situation can be considered as emblematic.The principles of New Public Management have been affirmed and this is not purely rhetorical.It has given place to an important legislative and regulative activity: France has progressively adopted European recommendations.But according to a strange mix, these principles have been included in the French administrative mindset which has reformulates the key issues.This is why some political scientists name this evolution "path dependency" or "hybridization of policies" in a national context.This process limits or even neutralizes the impact of international recommendations from the OECD and the European Commission.From this perspective, it is possible to illustrate the paradoxes of this modernization and to examine how New Public Management has been implemented in different areas, with some examples of policy borrowing from other countries and international organizations (Charlier, Croché & Leclercq 2012).
II. BEYOND LEGACIES AND REFORMISM: THE PARADOxES OF FRENCH NEW PUBLIC MANAGEMENT IN EDUCATION
If planning, through the action of the Planning Committee, was considered for long time as a lever to reconcile the objectives of equality of opportunity with economic development, the economic crisis and the failure of comprehensive schools forced the Educative State into a change of policy.At the beginning of the 80s, as described in the first part of this chapter, guidance remained a major concern for policymakers but assessment appeared as a new tool of governance for the education system.This explains the development of the first national assessments and the creation of the Directorate of Assessment and Forward Planning at the Ministry of National Education.
After the devolution acts, the French New Public Management (NPM) has corresponded to a education modernization project but, contrary to other countries, it has strongly resisted the market and privatization (Pollitt, 1990, Pollitt & Bouckaert 2004).As described in the first part of this chapter, the republican legacy is an initial explanation: the Republican school system was always eager to push back private interests while education was being progressively unified as a public service.So NPM reform is a compromise between tradition and modernization, and one which raises numerous paradoxes while, in the past decade, French education policy has become increasingly permeable to the effects of globalization and Europeanization (Hood 1991;Hood & Peters, 2004).
ASSESSMENT DEALING WITH A BUREAUCRATIC LOGIC
The creation of the Directorate of Assessment and Forward Planning or DEPP (Direction de l'Évaluation et de la Prospective) is a good example of this kind of French compromise.While it was inspired by the School Effectiveness Unit created at the UK Department of Education, it was first conceived as a planning instrument to forecast student enrolments after the socialist Minister Jean-Pierre Chevènement had announced the target of "80% of a same generation to achieve the baccalaureate in 2000".But this was also the result of an international expertise which France was involved, along with the USA and the OECD, in designing international indicators education.If the assessment logic has progressively penetrated the French education system, it was not to assess its quality and effectiveness, at least at the beginning.National assessments, as indicators for schools, were tools designed to measure the inequality of student outcomes and were presented as a mean to reduce these inequalities and to democratize access to education.The objectives of the Ministry was not to promote school choice and the market but to fight against the raw rankings published by the press which impeded a fair assessment of the social characteristics and merit of each school.Today, tests are still formative and not summative: they serve teachers in improving their teaching practices but they are not used for selecting students.
It was only during the 1990s that assessment began to be thought of as a tool for measuring the education system's effectiveness and quality.In the meantime a new paradigm was emerging.Claude Thélot, who played an important role as the Head of the DEPP, was the driver of this transformation (Thélot 1993).
Assessment espoused the principles of New Public Management (Economy, Efficiency, Effectiveness) while a High Council for the Assessment of Education (Haut Conseil de l'Évaluation de l'École) was created.It quickly became a think tank for experts and policymakers.This High Council published reports which claimed to align the French assessment system with the international surveys led by the OECD, particularly the PISA survey (Henry et alii, 2001).The High Council has also promoted the idea of a basic skills framework after a widespread national enquiry entitled "the Great Debate on Schools" based on data and questions prepared by the DEPP with the support of a consultancy firm .France was later joined by the European Standing Group on Indicators and Benchmarks to participate in building the indicators of the Open Method of Coordination.The PISA survey has progressively become a benchmark for policymakers, and Finland an example of a successful reform in education.
In France, education is a public service and a State administration (Derouet, 2000).It is therefore directly subjected to reforms enacted by the State.As we have seen, assessment has become a major component in the action of the State via the promulgation of the Institutional Act related to Finance Laws in 2001 (Loi d'Organisation des Lois de Finance or LOLF).This Act institutionalized new regulations for public expenditure through national programs and objectives which have to be assessed .Therefore, each administration and department of the State has to be accountable.But accountability in education remains very administrative and financial and, even if it includes pupil exams' results in its indicators, it does not put any pressure on schools regarding performance contrary to England (Mahony & Hextall, 2000;Gleeson & Husbands, 2001).Indeed, no system of information or digital assessment tool has been developed to make teachers more accountable.The LOLF remained very bureaucratic and has mainly served to justify the decision-making process for the reduction of budgets and cost-cutting processes with raw instruments even from a managerial point of view.In education, management does not share the same values as managers: they more often use the word "monitoring" to avoid a managerial vocabulary they often qualify as "neoliberal" .A lot of them do not make clear distinctions between "control" and "evaluation" even if audit practices in schools are currently being developed by the inspection bodies (Power 1997).
A LIMITED DECENTRALIzATION IN TERMS OF TRANSFER OF RESPONSIBILITIES
Decentralization is limited in its extent.Certainly, the first acts of decentralization delegated important powers to local authorities in building schools, renovation and equipment.The latter used these new responsibilities to receive significant investment and some prestigious operations for electoral purposes.The aim was to prove that local authorities can do a better job in a context of reduced investment from the State.But, in education, decentralization was stopped in the beginning of the 1980s.It was only in 1995 that a new act transferred the responsibilities of the youth vocational training from State to Regional Authorities.However, this decentralization was partial: the State continues to manage vocational schools and apprenticeships even if the regions are in charge of regulating the provision of vocational training via five-year plans.
Education remains narrowly statist and centralized.The State is in charge of defining the curriculum and the volume of teaching hours, the selection, recruitment and careers of teachers and other staff, initial and further training, controlling and inspecting schools, the guidance and professional inclusion of schoolchildren, the diplomas, certifications and recognition of qualifications.If devolution was implemented into the education system, by giving more autonomy to chancellors (recteurs), they remain very dependent on the decisions taken by the Ministry of Education.In the regions, relations between the State and local authorities can be tense due to conflicts regarding the sharing of jurisdiction or ideological opposition.Indeed, the primary and secondary education sectors are loosely coupled from a cultural and institutional perspective, and this does not facilitate cooperation and shared governance.Objective-based contracts define relations between the State and Local Authorities, but also between Local Education Authorities (Rectorats) and schools.Some networks of schools are emerging in particular to overcome the big divide between the primary and the secondary education sectors and to develop cooperation around the implementation of the basic skills framework (see below).
A source of heated debate is the transfer of civil servants to local authorities.It has been carried out for technical and maintenance staff in schools.Some similar attempts were made for the school guidance councillors.But this failed due to large-scale protests by the professional body of School Guidance Councillors and Psychologists (Conseils d'Orientation Psychologues) which was ideologically opposed to a concept of counselling defended by the local authorities via a strong cooperation with regional businesses and services involved in the assessment of skills or professional integration.Counselling also has powerful influence within the Ministry of Education and acts as a kind of internal lobbyist.Experts and policymakers are currently thinking of the creation of a regional public counselling service but nothing concrete has been yet proposed by the ministry.
THE RETENTION OF A CULTURAL TRADITION DESPITE A BASIC SKILLS POLICY
The Basic Skills and Knowledge Framework (Socle Commun de Connaissances et de Compétences) is the masterpiece of the 2005 School Act voted under the ministry of François Fillon.It gave rise to a whole of set of narratives (we could even say storytelling) which described it as the legacy of successive education plans from the foundation of the Republican School System.But, this framework is a translation, as we have seen, with some minor changes, of the European key competencies framework designed in 2004 by the European Commission while France has been involved for several years in the implementation of the Lisbon Strategy.It only resumed, after more than two decades, the basic skills travelling policy implemented in the USA and in the UK in the beginning of the 1980s (Ozga & Jones 2006).However, the French Basic Skills Framework is completely disconnected from issues of assessment and learning.It has led a curriculum war in France through ideological and strongly mediatized battles (Shor 1986).In terms of curriculum, modernizers are opposed to traditionalists.The former want to adapt the teaching of school disciplines to student needs and claim a stronger link between contents to be transmit by teachers and skills to be acquired by pupils.The latter wish to maintain a high level of contents requirement and criticize an instrumental conception of curriculum which distorts the culture transmitted to pupils.That is why the current socialist government has added "common culture" to the "basic skills and knowledge" framework to satisfy the claims of the main teachers' trade union.However, this divide goes beyond the traditional opposition between the Left and the Right.
From this perspective, the action of the State is torn between several contradictory requirements.It wants the basic skills framework to be a tool of pedagogical diversification to support the individualized counselling and achievement of pupils.But at the same time, it remains attached to an objective of equal teaching conditions for all pupils, and it defends a standardized conception of the curriculum.In addition to this paradox between standardization and diversification, there is a strong tension between assessment and curriculum (Revue Française de Pédagogie 2011).
Each teacher is considered autonomous in his/her classroom and on behalf of his/her "pedagogical freedom" is recognized and reaffirmed in the Code of Education.At the same time, as civil servants, they have to apply official instructions enacted by the Ministry for the implementation of the curriculum (Normand 2012).However, they have a discretionary power to assess students generally through marking.The lack of link between curriculum and assessments stops teachers from taking into account the issue of student skills while they do not feel concerned by student learning but only by teaching content.It explains why the High Council of Curriculum (Conseil Supérieur des Programmes has had to adapt the curriculum to the Basic Skills Framework, and recently proposed to implement an assessment without marks, to graduate students in accordance with their levels of learning difficulties, as has already been done in other European countries.But, up to now, the High Council's recommendations have not had much impact on policymaking.
SCHOOL CHOICE WITHOUT THE DEVELOPMENT OF THE MARKET
The ideology of the market served the policy of deregulation of catchment areas.The Right, under the Sarkozy government, sought to raise the issue of school choice while the Left was strongly opposed defending a social mix in schools.However, contrary to England, this deregulation was linked to a certain number of requirements which limited its extent (Ball 2008).Firstly, there was the issue of limited places in the best schools.Secondly, the selection and enrolment of pupils had to respect strict criteria (siblings, scholarship, special needs, etc…) which restricted the number of cases examined via bureaucratic regulations which restricted the voice of parents and their mobilisation.Head teachers, along with some local managers, were also reluctant to implement this policy.However, as has been observed elsewhere, the result was an increase in social segregation with the challenge of schools losing their best students, and this policy did not compensate the dominance of middle-class and upper families in the school choice strategies.It did not succeed either in developing a market for schools and strengthening competition between schools, as it is the case in the UK (Tomlinson 2005, Walford 2006).This policy was abandoned by the Left when it came to power in 2012.
Simultaneously, the very strong attachment to the equality of opportunities has led to the conception and implementation of some systems mixing school choice with principles of meritocratic selection against deprived pupils.That is why some higher education institutions, following the example of Sciences-Po Paris, have developed mentoring procedures in schools with difficulties, while preparatory classes to higher education institutions have opened their doors to deserving pupils on behalf of positive discrimination.As some sociological research findings demonstrate, this action has allowed higher education institutions to display a policy of openness to silence criticism of their excessive elitism while maintaining a strong selection in their entrance examinations.The other system invented by the Right was the Internats d'Excellence boarding schools copying the US Charter Schools.These schools for deprived students propose better support in teaching and learning while they isolate pupils from their family and social context to offer better studying conditions.However, management appears extremely heterogeneous from one school to another depending on the involvement of local authorities, the mobilisation of teaching teams, the recruitment procedures, the degree of autonomy of the pedagogical structure, etc.These schools have contributed to claims of imaginary meritocracy whilst serve as propaganda tools in the media to promote a positive discrimination with limited effects in the end.
THE FAILURES OF THE CONSERVATIVE REFORMISM OF THE LEFT
Since the socialists came to power in 2012, this policy of diversification and school choice was stopped as they claimed the will to reduce inequality of opportunities and to strengthen the school-mix.School autonomy, which the Right wanted to promote by giving more responsibility to Head Teachers was also stopped while a legitimistic conception has given power back to the General Inspection Body.Vincent Peillon, the Minister of National Education, brought together all the high-level managers of districts in Paris and told them that "management" and "governance" did not belong in his vocabulary.Instead, a rhetoric on a new foundation of the school system was disseminated while the principles of the Republican School System were reaffirmed particularly through the implementation of the teaching of "secular morality" in schools.In fact, the minister has a background ‡ in philosophy and he has remained very attached to Republican values and principles and has been inspired by the founders of the Republican School System (Kahn 2015).Sticking to its republican values, this left-wing government is promoting the Basic Skills Framework as a mean of democratisation and reduction of inequalities of opportunities.
The Commission for the New Foundation of the School System, created by the Minister to implement a new Act, despite its numerous working groups and its media coverage, has not lead to a substantial reform.The idea to focus the efforts of the education system on the primary education sector has only taken on board some recommendations of international organizations.The development of a national plan for digital technologies corresponds to similar aims without profoundly engaging the Ministry while the equipment is mainly depended on local authorities.Furthermore, the reaction of local authorities explains the failure of the reform of school timetables while it was presented as a key program of the Act.Succumbing to the lobbying of physicians close to the Academy of Sciences, the Minister decided to implement a national plan for the restructuring of timetables in primary schools after it was accepted by the trade-unions.Once the reform had been announced, it did not take long for the trade unions to disavow the Minister while teachers, local authorities and parents expressed their dissatisfaction to a badly-prepared, poorly-negotiated and under-funded reform.It was the same for the reform of graduate schools in education (ESPE : Écoles supérieures du Professorat de l'Éducation) and of the initial training of teachers.It was very quickly embedded in a bureaucratic maelstrom and a resurgence of conflicts of interest.Meanwhile, the Minister attempted to put the reform of the teaching profession on the political agenda, he did not have the time to implement it and his followers did not give him enough backing.They preferred to focus their action on restructuring the national curriculum according to the basic skills framework and to try to promote the school mix by transforming school provision, particularly by diminishing some elitist options like German, Latin or some bilingual courses which has led to a lot of protests from disciplinary-based interest groups and trade unions.
CONCLUSION
A lack of restructuring of the teaching profession, a limited autonomy for schools, school choice on the margins, a school market with restricted consequences, a managerial ideology with significant opposition, an unsuccessful decentralization: in these conditions, it is difficult to say that New Public Management has been implemented in the French education system.It contrasts strongly with the health sector where performance management, quality procedures, flexibility and mobility, contracts and agencies have created a new configuration of public service in hospitals.In education, only administrative and financial accountability has penetrated institutions in the long-term without having had much consequence on the work of teachers.Head Teachers and inspectors are developing assessment and audits, entering into contracts, but the bureaucratic structure predominates.However, current reflections among experts and policymakers, from both Left and Right, lead us to think that they are searching for a kind of Third Way between the State and the market (Normand 2016).A third step in decentralization, following the creation of new regional entities, leading to a restructuring of how local responsibilities are shared could be the main objective of the next reform in education.Another issue is related to the reform of the status of civil servants which could have consequences on the National Education public service.However, the confrontation between the Left and the Right regarding this project remains very decisive, and the trade unions are ready to fiercely defend their rights. | 2018-12-11T16:37:48.744Z | 2017-12-31T00:00:00.000 | {
"year": 2017,
"sha1": "7ef6d9dd74bb225b4f5bd9f16a141e4e5a846a16",
"oa_license": "CCBYNC",
"oa_url": "https://seer.ufrgs.br/rbpae/article/download/79294/46230",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "7ef6d9dd74bb225b4f5bd9f16a141e4e5a846a16",
"s2fieldsofstudy": [
"Political Science",
"Education"
],
"extfieldsofstudy": [
"Political Science"
]
} |
259098772 | pes2o/s2orc | v3-fos-license | Serum proteome profiles in cats with chronic enteropathies
Abstract Background Serum protein biomarkers are used to diagnose, monitor treatment response, and to differentiate various forms of chronic enteropathies (CE) in humans. The utility of liquid biopsy proteomic approaches has not been examined in cats. Hypothesis/Objectives To explore the serum proteome in cats to identify markers differentiating healthy cats from cats with CE. Animals Ten cats with CE with signs of gastrointestinal disease of at least 3 weeks duration, and biopsy‐confirmed diagnoses, with or without treatment and 19 healthy cats were included. Methods Cross‐sectional, multicenter, exploratory study with cases recruited from 3 veterinary hospitals between May 2019 and November 2020. Serum samples were analyzed and evaluated using mass spectrometry‐based proteomic techniques. Results Twenty‐six proteins were significantly (P < .02, ≥5‐fold change in abundance) differentially expressed between cats with CE and controls. Thrombospondin‐1 (THBS1) was identified with >50‐fold increase in abundance in cats with CE (P < 0.001) compared to healthy cats. Conclusions and Clinical Importance Damage to the gut lining released marker proteins of chronic inflammation that were detectable in serum samples of cats. This early‐stage exploratory study strongly supports THBS1 as a candidate biomarker for chronic inflammatory enteropathy in cats.
obtained via gastrointestinal endoscopy or surgery for histopathologic assessment are the current standard required for definitive diagnosis. 1,4 When diagnoses are not reached with histopathological evaluation, additional diagnostic tests such as immunohistochemistry and clonality testing might be required in ambiguous cases. 1 However, even with immunohistochemistry and clonality testing, definitive diagnoses are not always achieved. 1,5 There is a high rate of false positive samples with a specificity of 33% with clonality testing in cat samples. 5,6 Biomarkers have been investigated as potential diagnostic tests for CE with minimally invasive diagnostic methods. 7 The most commonly used biomarkers in cats in clinical practice are serum cobalamin and folate which can be misleading for diagnosis of CE in cats if comorbidities are present. 1 In humans, serum protein biomarkers are used to diagnose, monitor treatment response, and to differentiate between different forms of CE and therefore represent a novel, noninvasive, diagnostic, and monitoring tool for human IBD. [8][9][10][11][12][13] In cats, proteomics has been studied in several areas including chronic kidney disease, mammary carcinoma, pancreatic disease, osteoarthritis, and cat semen. [14][15][16][17][18][19] Given the extensive use of serum protein profiles in IBD in human medicine and studies in other areas of small animal medicine, serum proteomics might provide useful biomarkers in cats with CE. 8,14 The discovery of new proteomic biomarkers is a multiphase process which involves screening for potential biomarkers using an untargeted approach with a small sample size at the discovery phase followed by verification of selected proteins using targeted approaches with a larger sample size at the validation phase. 20,21 A recent study investigated the intestinal mucosal proteome in cats with 9 proteins found to be differentially expressed between healthy cats and cats with IBD and LGAL. 22 However, western blot analysis did not confirm significant differential protein expression. 22 The utility of serum proteomics in diagnosis of CE in cats has not been explored.
The aim of this exploratory study was to investigate changes in the serum proteomic profiles of cats with CE. This study provides useful preliminary data on the serum protein markers differentiating healthy cats from cats with CE. The ability to use a minimally invasive diagnostic method to assist in diagnosis of CE might reduce the need for more invasive diagnostics and anesthesia for pets with comorbidities and lead to reduced costs for pet owners.
| Animals and sample collection
Cases were recruited from 3 veterinary hospitals, including referral and primary care practices, between May 2019 and November 2020. Cats presented for evaluation of CE with clinical signs (vomiting, diarrhea, weight loss, or a combination of these signs) of at least 3 weeks' duration, and that were eventually diagnosed with CE/LGAL, were prospectively recruited for this study. Cat information including signalment, diagnostic tests, diet, previous and current treatment, and comorbidities was recorded. Cats with comorbidities such as pancreatitis, cholangiohepatopathy, urinary tract disease, and endocrinopathies, and cats without intestinal biopsies were excluded from further assessment. Diagnoses of CE were made on the basis of clinical signs, histopathology and exclusion of other causes of gastrointestinal manifestations, including metabolic disease, infection, parasitic disease, hepatic, and renal disease. The following diagnostic tests were performed: a complete blood count, a serum biochemistry profile, serum total T4 concentration, serum concentrations of cobalamin, abdominal ultrasonography and fecal flotation and PCR (feline coronavirus, Tritrichomonas foetus, Cryptosporidium species, panleukopenia virus, Clostriudium perfringens, Giardia species, Salmonella species, Toxoplasma gondii, Campylobacter jejuni, Campylobacter coli).
Serum feline pancreatic lipase immunoreactivity (fPLI) and feline trypsin like immunoreactivity (fTLI) were measured in some cats. All the cats diagnosed with CE using endoscopic biopsies were scored using the feline chronic enteropathy activity index (FCEAI) which was calculated based on their clinical signs, clinicopathological, and endoscopic findings. 3 Cats diagnosed via surgical biopsy were not FCEAI scored. All endoscopies were performed by a single board-certified veterinary internist (LB). Eight out of 10 abdominal ultrasounds were performed by board-certified veterinary radiologists from a single referral hospital while 2 ultrasound examinations were performed by general practitioners (cases recruited from primary care practices). All histopathological examinations of biopsied tissue samples were performed by boardcertified anatomic pathologists from different institutions and 7 cases were retrospectively reviewed following the WSAVA standards (MK).
Controls were healthy cats that did not show clinical signs of gastrointestinal disease or weight loss confirmed with history collection and unremarkable physical examination findings. Some control cats were staff cats while others were enrolled during annual wellness health check at primary care practices. Blood samples were collected via jugular venipuncture, followed by centrifugation for serum separation. The serum was then collected into a serum tube and stored at À20 C before analysis.
| Protein sample preparation
All serum samples were assessed for total protein concentration using the 2D Quant kit (Cytiva, Massachusetts, USA) as per manufacturer's instruction. Samples of 100 μg total protein were mixed in 50 μL AMBIC buffer (50 mM Ammoniumbicarbonate, 10 mM DTT, 2 M urea at pH 8) and trypsin digested at 25 C for 16 hours in a 1:100 enzyme-to-protein ratio based on the calculated serum protein concentration. 11 Digestion was halted by acidification. Each sample was then dried to remove the AMBIC, reconstituted in 50 μL 0.1% formic acid and desalted and nonpeptide contaminants removed using C18 stage tips (Thermo Scientific, Illinois, USA) according to the manufacturer's recommendations except that the elution buffer consisted of 80% CH 3 CN, 0.1% Formic acid.
| Mass spectrometry
Digested peptides were reconstituted in 10 μL 0.1% formic acid and separated by nano-LC using an Ultimate 3000 HPLC and autosampler (Dionex, Amsterdam, Netherlands) and followed methods described previously. 23
| Protein characterization
Protein dataset-peak lists were generated from raw files using Mascot Identifications were accepted if they could establish less than 5% false discovery rate (FDR) and contained at least 2 identified peptides per protein. 24
| Statistical analysis
Descriptive statistics of cat information including age, sex, weight, FCEAI and cobalamin concentrations were analyzed using R software.
Continuous data with non-normal distribution including age, weight and protein concentrations was compared and analyzed using the Mann-Whitney U test. Sex status was analyzed using Chi-square test.
Data were Log 2 transformed using NCSS (version 9) software and normal distribution confirmed by D'Agostino's K-squared test.
Changes in expression of protein (defined as fold change) were analyzed using Fisher's exact test with Benjamini-Hochberg adjusted P values. Differences in fold change were considered statistically significant at P < .02, with a minimum fold change of ≥5. [25][26][27] Fold change ratios and significance were calculated for CE/Control. For bioinformatics analysis, the Gene Ontology concept was used to investigate molecular function, biological processes, and cellular components. The identified biological module groups were evaluated by their enrichment score and the significance of the module's enrichment was determined by AmiGo Panther analysis. 28 Data were also analyzed with STRING version 11.5 database (https://string-db.org). 29 Proteins identified with significant (and fold change >2) changes in expression were uploaded into STRING to map corresponding protein-protein interactions. Graphical networks for these proteins were constructed based on their connectivity algorithms. mesenteric lymph node liver, or a combination of these tissues. The median number of biopsy specimens examined per site was 4 (range, [1][2][3][4][5][6][7][8][9][10][11]. Six cases were classified as lymphoplasmacytic enteritis (4 mod between the CE and clinically healthy cats. There was also no T A B L E 2 Consolidated and significant differential proteins between cats with CE and controls (CE/C) were found in 26 proteins with fold change ≥5 (P < .02). Using log2 transformed fold change, the significantly changing proteins between control and cats with CE based on total spectral counts, 2 peptide identification and FDR of 5%, P ≤ .05 is shown in
| DISCUSSION
Proteomic analysis allows large scale detection and rapid identification of proteins of pathological significance. 30 nisolone and with Cushing's syndrome. 41,42 In our study, no correlation was found between THBS1 abundance and cats treated with prednisolone in all groups.
A further finding in the study reported here was the involvement of coagulation factors in enteropathies. Coagulation factor V showed a significant difference with >100-fold increase in relative abundance (P < .001) in cats with CE compared to controls in our analysis. Coagulation factor V is a protein of the coagulation system and acts as a non-enzymatic cofactor for activated factor X to form the prothrombinase complex which contributes to generation of thrombin. 43 The relationship between hemostasis and CE is complex. Increased risk of thromboembolism and platelet activation have been observed in IBD in humans. 44 Further, increased generation of thrombin has been demonstrated in human IBD patients compared to controls and was thought to be associated with a partial loss of function of the natural anticoagulant pathways. 44,45 The imbalance between procoagulant and anticoagulant pathways with the changes in the fibrinolytic system are thought to contribute to thromboembolism in human IBD. 44 A hypercoagulable state has also been detected by thromboelastography in dogs with normoalbuminemic and hypoalbuminemic chronic inflammatory enteropathy. 46 The relationship between proteins in our study was mapped using the STRING algorithm (Figure 3), and the findings were dominated by the involvement of ECM proteins. ECM remodeling is a hallmark of IBD in humans. 51 Recent studies support the role of the ECM as an active component in promoting inflammation in the pathogenesis of IBD. 51 The alteration of ECM in IBD is characterized by inflammatory mediators and activated matrix metalloproteinase-9 (MMP9), which increases intestinal permeability, epithelial apoptosis and loss of goblet cells in colitis in humans. 51 In dogs, upregulation of mucosal active MMP2 and MMP9 was found in the intestines of dogs with CE compared to healthy dogs. 52 In addition, intestinal stricture formation and fibrosis have been linked to increased production in specific components of the ECM including fibronectin and collagens in IBD. 51 In our study, many ECM components such as fibronectin, inter-alpha-trypsin inhibitor heavy chains, profilin-1 and actin were upregulated in cats with CE compared to healthy cats.
Further, as shown in the data reported here (Figure 3), the interrelationship between inflammation, ECM and platelets contributes to the upregulation of protein clusters involving ECM and altered coagulation. The dysregulation of vascular ECM in human IBD patients has been found to promote adhesion and extravasation of leukocytes and platelets in the endothelium. 51 The cell adhesion molecules such as E-selectin interact with components of the ECM (fibronectin, collagens, laminin) to regulate the recruitment of circulating leukocytes and control endothelial permeability. 51 An apparent change in lipid metabolism, with upregulation of apolipoproteins A-I, B-100, C-I, C-III, and E, was identified in cats with CE compared to control cats in our study. Of these proteins apolipoproteins B-100, C-I, C-III, and E showed fold change ≥5. In cats, the distribution of serum lipoproteins and apolipoproteins is unlike that observed in humans. Lipoproteins in the cats are larger and richer in triglycerides compared to humans. 53 In the study reported here, cats with CE had increased levels of other known acute phase reactants such as complement proteins C1 and C3. The complement system comprises of many plasma proteins that function as receptors or regulators of complement activation through 3 activation pathways: the classical pathway, lectin pathway, and the alternative pathway. 58 Complement activation leads to the formation of C3 convertase which is the major and most abundant component in the complement cascade. 58 Hyperactivation of complement has been reported in chronic inflammatory diseases and immunological diseases in people. 58,59 Furthermore, in people with Crohn's disease, alterations in complement cascade and innate immune response have been described. 60 Serum C3 and C4 complement components were increased in human patients with Crohn's disease, with higher levels found in human patients with active disease compared to inactive disease. 61 The increased fold change in complement proteins in cats with CE in our study could be related to the immunological basis of IBD or inflammation. Nevertheless, further studies are needed to establish the link between the complement system and CE in cats.
There are several limitations in our study. First, the small sample size was a limiting factor. Although differences in protein profile were observed between cats with CE and LGAL, we did not perform further analysis comparing the 2 groups due to the small sample size of cats with LGAL and therefore a definitive conclusion of difference in protein profiles between CE and LGAL cannot be drawn. Second, as the management of each case was at the discretion of the F I G U R E 4 A dot plot showing the relative abundance based on spectral count of THBS1 in individual control and cats with chronic enteropathies (CE). The median is illustrated by the black line, median for cats with CE is 4.5 (range, 0-21) and median for controls is 0 (range, 0-1).
attending veterinarian, not all diagnostic tests were performed in each cat and treatment was not standardized. Specifically, fecal PCR, flotation, fPLI and fTLI results were only available for some cats.
Although our analysis did not show any correlation in THBS1 abundance in cats treated with prednisolone, we cannot completely | 2023-06-08T06:16:41.724Z | 2023-06-06T00:00:00.000 | {
"year": 2023,
"sha1": "151ec9dae5cfa4e30734bec93a494ba8b0746f42",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "f4ae59eb28544d82db163f54654b7e4cdf0533b1",
"s2fieldsofstudy": [
"Medicine",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
84004366 | pes2o/s2orc | v3-fos-license | Is ornithogenic fertilisation important for collembolan communities in Arctic terrestrial ecosystems
In the Arctic, areas close to seabird colonies are often characterized by exceptionally rich vegetation communities linked with the high nutrient subsidies transported by seabirds from the marine environment to the land. These areas also support soil invertebrate communities of which springtails (Collembola) often represent the most abundant and diverse group. Our study focused on springtail community composition in the vicinity of seabird (little auk, great skua and glaucous gull) nesting areas in different parts of Svalbard (Magdalenefjorden, Isfjorden and Bjørnøya), and on their comparison with adjacent areas not impacted by seabirds. Out of a total of 35 springtail species recorded, seven were found only within the ornithogenically influenced sites. Although geographical location was the strongest factor differentiating these springtail communities, ornithogenic influence was also significant regardless of the location. When each location was considered separately, seabirds were responsible for a relatively small but strongly significant proportion (8.6, 5.2 and 3.9%, respectively, for each site) of total springtail community variability. Species whose occurrence was positively correlated with seabird presence were Folsomia coeruleogrisea, Friesea quinquespinosa, Lepidocyrtus lignorum and Oligaphorura groenlandica near Magdalenefjorden, Arrhopalites principalis, Folsomia bisetosella and Protaphorura macfadyeni in Isfjorden, and Folsomia quadrioculata on Bjørnøya.
Abstract
In the Arctic, areas close to seabird colonies are often characterized by exceptionally rich vegetation communities linked with the high nutrient subsidies transported by seabirds from the marine environment to the land. These areas also support soil invertebrate communities of which springtails (Collembola) often represent the most abundant and diverse group. Our study focused on springtail community composition in the vicinity of seabird (little auk, great skua and glaucous gull) nesting areas in different parts of Svalbard (Magdalenefjorden, Isfjorden and Bjørnøya), and on their comparison with adjacent areas not impacted by seabirds. Out of a total of 35 springtail species recorded, seven were found only within the ornithogenically influenced sites. Although geographical location was the strongest factor differentiating these springtail communities, ornithogenic influence was also significant regardless of the location. When each location was considered separately, seabirds were responsible for a relatively small but strongly significant proportion (8.6, 5.2 and 3.9%, respectively, for each site) of total springtail community variability. Species whose occurrence was positively correlated with seabird presence were Folsomia coeruleogrisea, Friesea quinquespinosa, Lepidocyrtus lignorum and Oligaphorura groenlandica near Magdalenefjorden, Arrhopalites principalis, Folsomia bisetosella and Protaphorura macfadyeni in Isfjorden, and Folsomia quadrioculata on Bjørnøya.
To access the supplementary material for this article, please see Supplementary files under Article Tools online.
Arctic terrestrial ecosystems are generally regarded as being relatively simple, species-poor and characterized by short food chains. Very strong seasonality, a short, cold growing season, nutrient deficiency, permafrost, scant liquid water and regular freezeÁthaw cycles strongly restrict primary and secondary production (Ims & Ehrich 2013). Low energy and limited snow-and ice-free land, the relatively young age of contemporary Arctic terrestrial ecosystems, and spatial isolation contribute to generally low species diversity, especially in the case of higher plants and vertebrate herbivores and predators (Payer et al. 2013). However, spatial heterogeneity in, for instance, temperature, precipitation, wind exposure, hydrology, geomorphology (elevation), proximity to coastlines and soil chemistry create environmental gradients and complex mosaics of habitats that may support considerable diversification of communities of smaller organisms such as invertebrates (Hertzberg et al. 2000;Sinclair & Sjursen 2001;Ims & Ehrich 2013). Even very small patches of suitable favourable habitats surrounded by a hostile environment, such as individual Carex tussocks embedded in cyanobacteria-covered ground (Hertzberg et al. 1994;Hertzberg et al. 2000;Ims et al. 2004), Azorella cushion plants on sub-Antarctic Marion Island (Hugo et al. 2004 cryoconite holes (De Smet & Van Rompu 1994) and glacier mice (Coulson & Midgley 2012), may provide viable habitats for these animals.
Most of the Arctic invertebrate fauna inhabit soil and soil surface environments, and these organisms can develop much higher abundance, species diversity and food web complexity in suitable habitats than any other non-microbial eukaryotes on the land (Hodkinson & Coulson 2004;Hodkinson 2013;Coulson et al. 2014). Springtails (Collembola) are often the most abundant and diverse group (Bengston et al. 1974;Birkemoe & Leinaas 2000). Although it is widely assumed that springtails play essential roles in many key polar ecosystem processes, such as decomposition, energy flow and nutrient cycling, mainly through grazing on microorganisms and physical alteration of soil and litter (Hopkin 1997;Rusek 1998;Bardgett & Chan 1999;Filser 2002), detailed knowledge about the distribution and autecology of most species is lacking (Hogg et al. 2006;Hodkinson 2013).
Studies of polar (both Arctic and Antarctic) collembolan and other invertebrate assemblages over multiple spatial scales have revealed strong heterogeneity in their distribution and abundance. Between different geographical regions this may be driven by environmental conditions such as temperature and moisture associated with climate (Babenko 2000; Bokhorst et al. 2008), as well as historical dispersal and colonization processes (Á vila-Jimé nez & Coulson 2011). In the typical mosaic of High Arctic terrestrial habitats, microtopography and habitat moisture, temperature and habitat quality parameters such as physico-chemical properties of substrate, decomposition rate and food availability affect the invertebrate communities at the scale of meters (Usher & Booth 1986;Dollery et al. 2006;Zmudczyń ska et al. 2012;Hodkinson 2013). The size and isolation of habitat patches and especially their plant species composition may further contribute to the variability of springtail assemblages (Hertzberg et al. 1994;Hertzberg et al. 2000;Ims et al. 2004). At the scale of centimetres, invertebrates may be associated with particular plant species and/or local environmental conditions formed beneath them (Coulson et al. 1993;Block & Convey 1995;Coulson et al. 2003;. Finally, even within relatively homogeneous habitat, clustering of invertebrates may be also explained by inter-species relationships (Usher & Booth 1986;Caruso et al. 2013), pheromone-induced aggregation (Leinaas 1983;Usher & Booth 1984, 1986Benoit et al. 2009), and past stochastic events causing uneven mortality or other demographic phenomena Chown & Convey 2007).
Within the patchy High Arctic terrestrial ecosystem, particularly favourable habitats for many organisms are found in the vicinity of seabird nesting sites, especially the larger bird colonies which may consist of several hundred thousand individuals (Lindeboom 1984;Odasz 1994;Stempniewicz et al. 2007). These areas are fertilized by nutrients transported by the birds from the marine environment and deposited on land in the form of guano, feathers, egg shells and carcasses (Bokhorst et al. 2007;Zwolicki et al. 2013). The ornithogenically subsidized areas support exceptionally lush vegetation (Zmudczyń ska et al. 2008;Zmudczyń ska-Skarbek et al. 2013) including specific plant communities (Eurola & Hakala 1977;Elvebakk 1994;Zmudczyń ska et al. 2009), and populations of herbivores, predators and scavengers (Croll et al. 2005;Jakubas et al. 2008;Kolb et al. 2011).
Collembola feeding on fresh and dead organic matter may specifically graze on or accidentally ingest fungi, algae, bacteria and other microbiota (Hopkin 1997;Rusek 1998;Worland & Lukešová 2000) and may also attain very high population densities in ornithogenic substrates around seabird colonies (Coulson et al. 2014). While the contribution of collembolans to decomposition in the Arctic terrestrial ecosystem is thought to be vital, relatively little attention has been given to their role in ornithogenically influenced habitats (see Mulder et al. 2011). Byzova et al. (1995) reported extremely high density and biomass of springtails in the vicinity of a little auk (Alle alle) colony in Hornsund (south-west Spitsbergen), reaching more than 10 5 ind. and 5 g m (2 , these values being higher than reported either in most non-influenced ecosystems or in nutrient-enriched manure heaps. We have also reported similar values both associated with little auks and below a nearby cliff-nesting colony of Brunnich's guillemots (Uria lomvia) and kittiwakes (Rissa tridactyla) (Zmudczyń ska et al. 2012). High springtail density has also been noted from seabird-influenced sites in another west Spitsbergen fjord, Kongsfjorden (Bengston et al. 1974;Sømme & Birkemoe 1999), and on Nordaustlandet, the northern-most island of the Svalbard Archipelago (Fjellberg 1997). Sømme & Birkemoe (1999) and Zmudczyń ska et al. (2012) described gradual changes in collembolan community composition with distance from seabird colonies. Specific communities have also been described in areas experiencing the most intensive ornithogenic manuring, with few species strongly represented, including Megaphorura (formerly Onychiurus) arctica, Hypogastrura viatica, Folsomia quadrioculata and Xenylla humicola (Hodkinson et al. 1994;Fjellberg 1997;Sømme & Birkemoe 1999;Zmudczyń ska et al. 2012).
Most studies to date have focused on single seabird colonies and locations, and addressed basic parameters of springtail community description, such as overall species richness, density and biomass of Collembola as a whole, or only focusing on the most abundant species. In a previous study of two seabird colonies at Hornsund, we identified significant correlations between springtail density and soil physical and chemical properties, and vegetation biomass (Zmudczyń ska et al. 2012). Factors significantly influencing the springtail community in that study included the cover of green nitrophilous alga Prasiola crispa (explaining 11% of the total springtail variability), total plant biomass (9%) and soil conductivity (6%). These factors were all clearly associated with distance from the seabird colonies and guano deposition (see also Zwolicki et al. 2013).
In the current study, we used multivariate analytical techniques to generate quantitative estimates of, first, the proportion of total variability in the Collembola community explained by seabird influence. To our knowledge, this question has not previously been tested, and multivariate (ordination) approaches have rarely been applied in the study of ornithogenic impact on invertebrate assemblages worldwide (but see Orgeas et al. 2003;Towns et al. 2009;Kolb et al. 2012;Zmudczyń ska et al. 2012). Second, this approach will permit the identification of springtail species that are specifically linked with seabirdmediated changes in the environment. We analysed collembolan communities collected over a much wider geographical scale than has been achieved previously, ranging from the relatively small island of Bjørnøya (748N), which is the southernmost island of Svalbard, to the second largest Spitsbergen fjord Isfjorden located in the centre and mildest part of the island (788N), to the northern-most west Spitsbergen coast close to Magdalenefjorden (798N). We obtained samples from six seabird nesting sites and respective control areas. This coverage also enabled us to assess the role of geographical location within the archipelago in explaining variability between the springtail communities found. Finally, as other studies have related the occurrence of Collembola and vegetation communities/species (Hertzberg et al. 1994;Coulson et al. 2003;Ims et al. 2004), we evaluated the strength of this relationship in the vicinity of seabird colonies.
Study area
The study was conducted at three locations within the Svalbard Archipelago, one on Bjørnøya (Bear Island) and two on Spitsbergen (Fig. 1). Within all sampling plots we identified vascular plant species and visually (using a plot-sized quadrat subdivided into 20)20 cm units) estimated the individual species and total moss percentage contributions to vegetation cover.
Bjørnøya (Bjo). This is the southernmost island (176 km 2 ) of the archipelago, midway between the Norwegian mainland and Spitsbergen. Three sites in the vicinity of different seabird species nesting areas, together with respective control sites, were sampled. Site B-A (74838?N 19803?E) was close to a relatively large colony of the planktivorous little auk situated on a gentle slope on Alfredfjellet, exposed to the north and descending to Lake Ellasjøen. The upper part of the site consisted of vegetation-covered rock debris, while the lower part approaching the lake shore was flat and waterlogged. Vegetation cover was complete in both parts of the area. Vegetation consisted of vascular plants, mainly Salix sp. and Saxifragaceae (and Equisetum arvense in the waterlogged area), interspersed with compact moss carpets and clumps. Site B-Ac was the control site for B-A, located on Alfredfjellet and parallel to B-A but ca. 500 m from the colony and separated from it by a seasonal stream. Similar plant species to site B-A were present, but the total cover of vegetation was around 20% (see above for method of estimation). Site B-L (74847?N 18878?E) was located in the north-west, flat part of the island, close to the cliff edge and to a concentration of nests of the predatory (feeding on fish, large pelagic invertebrates and other seabirds) glaucous gull (Larus hyperboreus), with patches of dense vegetation surrounding each nest. Vascular plants were usually underlain by a dense moss layer (80% moss cover on average), and were dominated by Festuca cf. rubra subsp. arctica (up to 100%), with less than 10% admixture of Oxyria digyna, Saxifraga caespitosa and Draba sp. Site B-S (74847?N 18876?E) was located inland from B-L, in close proximity to nests of the predatory (feeding on fish and other seabirds) great skua (Stercorarius skua). Vegetation was similar to that of B-L, but F. cf. rubra subsp. arctica was less abundant (on average 70% cover), with a generally more species rich herb/shrub flora, and more abundant mosses (90%) present. In addition to the species listed above, Salix sp., Cerastium sp. and other Saxifragaceae were present (up to 80, 20, and B1% cover, respectively). Magdalenefjorden (Mag). The sites M-A1 and M-A2 (79852?N 10870?E) were situated on the talus slope of Aasefjellet, exposed to the west and descending to the open sea, adjacent to very large little auk colonies. The vascular plant layer was underlain by dense moss, giving up to 95% cover, and mostly consisted of C. arcticum, P. alpina var. vivipara and C. groenlandica. Site M-Ac, the control site for both M-A1 and M-A2, was located on Aasefjellet (ca. 700 m and 500 m from M-A1 and M-A2, respectively), facing north and descending to Hamburgbukta. Boulders that were not overgrown with vegetation composed a significant proportion of this area (up to 60% in some plots), with the remaining area predominantly covered by mosses.
Sampling protocol
The study was conducted in the summer months of July and August, during expeditions to Bjørnøya (2008) . Due to practical logistic limitations, on transects B-A and B-Ac every second plot was sampled (plots 1, 3, 5, 7 and 9). We collected three soil cores (together with vegetation cover, see below) from three sites along the same diagonal of each sampling plot (from the centre and the two corners of each square) (Table 1). In the cases of great skua and glaucous gull nesting sites located on flat ground, three plots (100)100 cm each) were sampled in the vicinity of each nest: (plot 1) with a nest situated in the centre, (2) adjoining plot 1, still within the patch of dense vegetation surrounding the nest (both plots included in sites B-L and B-S), and (3) beyond the boundary of the compact vegetation patches, 3 m on average from the nest (sites B-Lc and B-Sc, respectively). We collected one soil core from the centre of each plot, except in plot 1 where the sample was taken adjacent to the nest (Table 1). Nests containing eggs or chicks were not sampled. Sampled nests had been recently occupied as evidenced by the presence of down and other nest material, and food scraps, but we cannot be fully certain that they had been occupied in that year's summer season.
Samples were taken with a cylindrical probe (diameter 6 cm) from the soil surface (mainly organic) layer, and included the vegetation covering the area and the underlying soil to a depth of ca. 5 cm. Each sample was sealed in a plastic container and, within a few hours, returned to the laboratory where it was subsequently placed for 48 h in a modified Tullgren apparatus illuminated with 60W bulbs (Barton 1995). Extracted springtails were preserved in 96% ethanol and identified to species level following Fjellberg (1998Fjellberg ( , 2007. We calculated frequencies of occurrence (%) and densities of particular species (number of individuals per m 2 ). For cores obtained from the 160 )160 cm plots, total springtail species counts for each plot (i.e., the sum of the three samples obtained) were analysed.
Statistical analyses
To test for differences in the springtail species richness between the study areas the non-parametric Mann-Whitney U-test was used on account of the non-normal distributions of data and a relatively low number of sampling plots per group tested. Data were processed using STATISTICA 10.0 (StatSoft, Inc. 2011).
Numerical ordination methods were used to describe total (qualitative and quantitative) variability of springtail communities and vegetation: (1) independently of any environmental influence (unimodal indirect gradient analysis: detrended correspondence analysis [DCA]), to describe the general pattern of variability in the studied community; and (2) in relation to environmental variables (unimodal direct gradient analysis: canonical correspondence analysis [CCA]). Two nominal environmental factors were tested: AREA, determining the geographical location (Bjo, Isf or Mag), and SEABIRD, representing the presence (Seabird) or absence (Control) of a seabird colony in the vicinity of sampling sites. All species data were log-transformed to normalize their distributions. After CCA, a Monte Carlo permutation test was performed (with 499 permutations) to identify which of the factors significantly influenced the model. To calculate the factors' unique contribution to explaining variability in the springtail species composition we used variation partitioning test (ter Braak & Š milauer 2012). To provide more accurate estimation of variation explained with canonical (CCA) analyses, we adjusted the variation value using the number of degrees of freedom as suggested by Peres-Neto et al. (2006). Each time the results of constrained ordination were compared with those of unconstrained ordination (% variability explained by an environmental factor was divided by % variability explained by one (in the case of SEABIRD) or two (AREA) axes of the unconstrained analysis). Thus we obtained the efficiency of the environmental factor(s) (%) in explaining the total variability present in the data (ter Braak & Š milauer 2012). In order to relate ordinations based on springtail community composition with those of vegetation composition, we used co-correspondence analysis (CoCA; ter Braak & Š milauer 2012). To explore significant relationships between individual springtail species and the environmental factor SEABIRD we employed t-value biplots (Van Dobben circles) that approximated the t-values of the regression coefficients of a weighted multiple regression (ter Braak & Š milauer 2012). Data were processed using CANOCO 5.0 software (ter Braak & Š milauer 2012).
Results
Within the total of 140 plots (246 samples) studied, we recorded 35 springtail species: 17 on Bjørnøya, 21 in Isfjorden and 30 in Magdalenefjorden (Table 2, Supplementary Table S1). The average numbers of species were similar across SEABIRD and CONTROL sites in all plots taken together and in each location, except for Magdalenefjorden where it was higher in M-A1 than in M-Ac (Mann-Whitney test, U 03.00, p 00.003). However, in total, more species occurred within SEABIRD sites versus their respective CONTROL sites in the case of three little auk colonies: I-A (20) versus I-Ac (18) Table S1). Two species were recorded only on CONTROL sites: Agrenia bidenticulata (M-Ac) and Thalassaphorura duplopunctata (B-Ac) (Supplementary Table S1). Almost 20% of variability in springtail species composition was explained by two hypothetical gradients (unconstrained ordination axes), with 12.1% explained by axis 1 and 7.7% by axis 2 (DCA, gradient length04.00 SD; Table 3, Fig. 2a). CCA (constrained ordination) revealed that the AREA factor (Bjo, Isf, Mag) was responsible for 13.4% of the total collembolan variability, while the SEABIRD factor (Seabird, Control) accounted for 2.1% (Fig. 3). These factors therefore contributed 67.7% (AREA) and 17.4% (SEABIRD) of the variation explained by the model. The AREA and SEABIRD factors were independent of each other and did not share the explained variation (variation partitioning test, F 08.8, p 00.002).
The areas studied were distinct with respect to their vegetation composition. The plots from different areas formed distinct groups in the unconstrained ordination (DCA) space based on plant species and moss abundances recorded (Fig. 2b). The AREA factor explained 19.3% of the total vegetation variability, equating to 83.0% of the variation described with DCA (axis 1: 15.3%, axis 2: 8.0%, gradient length 02.58 SD). The variabilities of springtail and vegetation communities were not significantly correlated (CoCA, p !0.05). Because of the clear differentiation of the three localities (both with respect to collembolan and vegetation diversity, as well as in terms of geographical distance), in subsequent analyses each area was treated separately. Neither springtail nor vegetation variability were significantly co-correlated within any of the individual areas studied (CoCA, p !0.05).
On Bjørnøya, the SEABIRD factor was responsible for 3.9% of the total springtail variability, constituting 30.7% of the variability identified by the theoretical unconstrained analysis (Table 3). Of the collembolan species best fitted to the first CCA axis (equivalent to the SEABIRD factor; the species shown in Fig. 4a), Folsomia quadrioculata was significantly positively associated with this explanatory variable while four other species*Desoria tshernovi, Hypogastrura viatica, Tetracanthella arctica and Thalassaphorura duplopunctata*were associated negatively (species selected using Van Dobben circles; Table 4). In Isfjorden, the proportion of variation in collembolan community composition that was influenced by seabird presence was 5.2%, equating to 24.6% of the available variation (Table 3). Three species were positively related to seabird impact*Arrhopalites principalis, Folsomia bisetosella and Protaphorura macfadyeni*and two species related to it negatively Lepidocyrtus lignorum and Parisotoma notabilis: (Fig. 4b, Table 4). In Magdalenefjorden, seabird influence accounted for 8.6% of the total springtail variability, or 44.8% of the total available (Table 3). Here, four species reacted positively to seabird impact: Folsomia coeruleogrisea, Friesea quinquespinosa, Oligaphorura groenlandica and L. lignorum (which responded negatively in Isfjorden). Among six species that were negatively associated with seabirds in Magdalenefjorden was F. quadrioculata determined as positive in Bjørnøya (Fig. 4c, Table 4). The relationships with seabird influence were still present for nine species (five positive and four negative) when all the areas were analysed together (Table 4).
Discussion
Although the key role of colonial seabirds in the enrichment of the otherwise poor Arctic terrestrial ecosystem is relatively well recognized, data on soil invertebrate assemblages inhabiting ornithogenically modified areas are still sporadic. The six previous studies that we are aware of listed species occurring close to rich seabird nesting sites along with, at most, data on overall Collembola density and biomass (Bengston et al. 1974;Hodkinson et al. 1994;Byzova et al. 1995;Fjellberg 1997;Sømme & Birkemoe 1999). Our previous studies at Hornsund have identified some factors significantly correlated with springtail abundance and community composition, including the amount of guano deposited and soil and vegetation parameters that are closely associated with bird colony impact (Zmudczyń ska et al. 2012;Zwolicki et al. 2013). The present study is the first to attempt to estimate quantitatively the proportion of variation in Arctic Collembola communities explained by seabird influence, and one of very few that has addressed this question worldwide (Kolb et al. 2011). Almost 20% of total variability of springtail species composition in our data set (comprising 35 species recorded in 140 sampling plots from different habitats and geographical regions of the Svalbard Archipelago) was explained by two hypothetical environmental gradients (unconstrained DCA axes). The most important factor accounting for this variation (68%) appeared to be related to geographical location. We sampled three widely separated parts of the archipelago. At this regionally large spatial scale the sampling areas differed in exposure to the marine environment and, in particular, to the influence of ocean currents and temperatures of the water masses predominating at each (Loeng & Drinkwater 2007;Saskaug et al. 2009). For instance, with a mean July temperature of 5.98C in 1961Á1990 (eKlima 2014), Isfjorden is the warmest part of Svalbard, mostly due to the large inflow of warm Atlantic water. Bjørnøya, surrounded by cold Arctic water but also in close proximity to Atlantic water masses, experiences large amounts of fog and high winds (Summerhayes & Elton 1923;Saskaug et al. 2009), higher precipitation (30 mm in July compared to 18 mm in Isfjorden) and lower summer temperature (4.48C in July), while being the mildest throughout winter ( (8.18C in January). The key role of geographical factors underlying Collembola distribution has been emphasized by Caruso (2007) and Babenko (2009) in studies sampling different locations across latitudinal gradients in northern Victoria Land (Antarctica) and on the western Taimyr Peninsula (Russian Arctic), respectively. Although these studies did not assess the total species richness in the study areas, both noted that some species did not occur at all localities along each transect. demonstrated that variability in springtail species richness between their study sites was the same whether considering distance scales of 10 km or 100 m. DCA ordination of plots in our study revealed a similar pattern, with some samples from Bjørnøya being more similar to those from other Svalbard areas than others within the island itself (Fig. 2a). Furthermore, although the southernmost and also the most extensively sampled location (117 samples, in comparison with 54 from Isfjorden and 75 from Magdalenefjorden), Bjørnøya hosted the lowest number of springtail species (17 as compared with 21 and 30, respectively). We therefore conclude that the inter-area distinctions identified from our data do not represent a latitudinal gradient but, rather, large-scale variability resulting from a range of factors, including local climate, historical dispersal and colonization processes (Babenko 2000;Bokhorst et al. 2008;Hodkinson 2013). Similarly, geographical factors may underlie the distinctions identified between the study areas in terms of vegetation community composition, accounting for 83% of the variation described by DCA. In this case, the length of the gradient representing beta-diversity (ter Braak & Š milauer 2012) was lower (2.6 SD) than that calculated for Collembola (4 SD) while, as noted above, there was no correlation apparent between the vegetation and collembolan communities. However, the survey scales also differed between these two groups, with springtails being extracted from cores of 6 cm diameter, and vegetation composition documented in complete 160 )160 cm or 100)100 cm plots, areas that are likely to host multiple Is ornithogenic fertilization important for collembolan communities?
invertebrate microhabitats (Usher & Booth 1986;Hodkinson 2013). Irrespective of the scale considered and the geographical influence, our analyses demonstrate that seabirds exerted significant influence on springtail communities in Svalbard. Seven of the 35 springtail species recorded were present only in the seabird-influenced sites. Collembolan community variability was explained by the SEABIRD factor to a greater extent when each area was considered separately than when data from all the areas were combined (Table 3), probably a consequence of the community composition differing in detail between the areas. Hence, the springtail assemblage recorded within any one area contained one to four species that were positively correlated with the seabird influence, without the same relationship being identified in the other study areas. For instance, Folsomia quadrioculata is recognized as a widely distributed species in the Arctic and has been considered to have no clear habitat preferences (Fjellberg 1994), although it has also been recorded as notably abundant below bird cliffs (Fjellberg 1997;Sømme & Birkemoe 1999;Zmudczyń ska et al. 2012). In the current study, this species' presence was significantly correlated with seabird influence on Bjørnøya, but negatively correlated in Magdalenefjorden. Other species positively associated with seabird influence in analyses of the remaining areas are known for their occurrence in rich ornithogenic sites, and include both species characteristic of wet (e.g., Oligaphorura groenlandica) and dry (e.g., Lepidocyrtus lignorum) habitats (Fjellberg 1994). Those species for which significant responses were not necessarily found in ordinations (likely due to small sample sizes), but were found exclusively within seabird-influenced sites, were also typical of eutrophic and usually wet or moist habitats. Two such Fig. 4 Canonical correspondence analysis (CCA) ordinations of 10 best-fitted species with respect to the SEABIRD factor (axis 1) in each area. Pie slices based on species percentage occurrence within SEABIRD (black) and CONTROL (white) sites. Boldface indicates species that significantly and positively react to the SEABIRD explanatory variable (on the basis of t-value biplot). (Fjellberg 1994) and Russebukta/Edgeøya (www.artsobservasjoner.no), and the latter from Jan Mayen (Fjellberg 1994), Kinnvika/ Nordaustlandet and Barentsburg/ Isfjorden (Coulson et al. 2013). Notably, some species responding negatively in the current study to the seabird factor (e.g., Parisotoma notabilis, Tetracanthella arctica) or totally absent from the seabird areas (Thalassaphorura duplopunctata) have previously been noted in rich (also ornithogenic), both wet and dry sites (Fjellberg 1998(Fjellberg , 2007. This emphasizes the importance of conducting analyses of the entire springtail community inhabiting an area rather than focusing on individual species that may react differently to a given factor when additional environmental conditions are considered. A priori, the most obvious potential link by which seabird influence might interact with collembolan communities would appear to be via the soil and vegetation developing on rich ornithogenic substrates around nesting sites, both of which clearly differ from those of nonfertilized areas (Eurola & Hakala 1977;Zwolicki et al. 2013). We have shown previously that the density of all springtails as well as that of the locally predominating species (F. quadrioculata and H. viatica) were significantly though moderately correlated (correlation coefficients ranging from 0.2 to 0.5) with individual chemical and physical soil properties (Zmudczyń ska et al. 2012). Each of these factors was strongly correlated to seabird guano deposition (coefficients of 0.6 to 0.9, Zwolicki et al. 2013), but only soil conductivity significantly influenced the Collembola composition. Such evidence supports springtails not being directly dependent on a single factor but rather influenced by a range of complex environmental factors. As some studies have identified significantly distinct microarthropod assemblages associated with particular plant species/communities and/or the specific properties of soil forming beneath them (e.g. Coulson et al. 2003), we hypothesized that a strong relationship would exist between springtails and vegetation composition in the studied areas. However, no such correlation was apparent in our data, even when each study area was analysed separately. Plants are potentially available for springtails as food (typically during decay rather than through active grazing), as well as contributing to the modification of habitat properties such as soil moisture and temperature. However, seabirds may alter the quality and quantity of several different food resources, including algae, fungi and other microorganisms (Matuła et al. 2007;Wright et al. 2010). For instance, in Hornsund a green nitrophilous alga species, Prasiola crispa, growing abundantly below the seabird colonies was significantly associated with the collembolan community (Zmudczyń ska et al. 2012) and might therefore be an important diet component there. Moreover, seabirds may affect distribution and abundance of different invertebrate groups, such as mites, beetles or dipteran larvae that may compete for the resources and/or prey on collembolans (Bengston et al. 1974;Caruso et al. 2013;Basset et al. 2014). Unfortunately, as is often the case in studies of polar terrestrial ecosystems (Worland & Lukešová 2000;Hogg et al. 2006), detailed autecological information (including diet studies) of most of the springtail species reported here is lacking.
The multivariate analyses applied here provided clear evidence of the significant role that seabird influence plays for these soil invertebrate communities in the Arctic. While the explanatory power of the seabird enrichment was relatively low, the ornithogenic effect was significant both at the scale of the entire Svalbard Archipelago and within each specific geographical location. Other factors, such as small-scale habitat conditions (enhanced by the typical patchiness of tundra habitats [e.g., Hertzberg et al. 1994;Ims et al. 2004]), population density fluctuations from year to year (Sømme & Birkemoe 1999), and natural tendency for aggregation (Leinaas 1983;Usher & Booth 1984, 1986Benoit et al. 2009) contribute additional variation to the Collembola community. Given that springtails provide key ecosystem services contributing to organic matter decomposition, energy flow and nutrient Table 4 Significant positive (') and negative ( () response of Collembola species for the SEABIRD factor, chosen on the basis of t-value biplots made separately for each area: Bjørnøya (Bjo), Isfjorden (Isf), Magdalenefjorden (Mag) and all the areas together (ALL). cycling, and considering that these processes, together with the invertebrate communities and colonial seabirds, are expected to be strongly influenced by predicted climate warming, especially in the polar regions (Callaghan et al. 2005;Hodkinson 2013;Ims & Ehrich 2013), further studies of Collembola distribution, abundance and functional ecology are required, with these being planned over appropriate spatial and timescales. | 2019-03-20T13:05:31.689Z | 2015-01-01T00:00:00.000 | {
"year": 2015,
"sha1": "b1deedc9f05ffb6b21fd71dd3f2e9d977d4a129c",
"oa_license": "CCBYNC",
"oa_url": "https://doi.org/10.3402/polar.v34.25629",
"oa_status": "GOLD",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "a2cdc31a2cf3131b49b6d8a97f9ccace57a3cea4",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Biology"
]
} |
15280655 | pes2o/s2orc | v3-fos-license | The CD45 tyrosine phosphatase regulates phosphotyrosine homeostasis and its loss reveals a novel pattern of late T cell receptor-induced Ca2+ oscillations.
CD45 is a transmembrane tyrosine phosphatase implicated in T cell antigen receptor (TCR)-mediated activation. In T cell variants expressing progressively lower levels of CD45 (from normal to undetectable), CD45 expression was inversely related to spontaneous tyrosine phosphorylation of multiple proteins, including the TCR zeta chain, and was directly correlated with TCR-driven phosphoinositide hydrolysis. The Ca2+ response in these cells was altered in an unexpected fashion. Unlike wild-type cells, stimulated CD45- cell populations did not manifest an early increase in intracellular Ca2+, but did exhibit a delayed and gradual increase in mean intracellular Ca2+. Computer-aided fluorescence imaging of individual cells revealed that CD45- cells experienced late Ca2+ oscillations that were not blocked by removal of extracellular Ca2+. CD45 revertants had the signaling properties of wild-type cells. Thus, CD45 has a profound influence on both TCR-mediated signaling and phosphotyrosine homeostasis, and its loss reveals a novel role for this tyrosine phosphatase in Ca2+ regulation.
S timulation of T cells via the TCR invokes both early and late activation events. Classical early events include phosphoinositide (PI) 1 hydrolysis, increases in intracellular calcium ([Ca2+]i), activation of protein kinase C, and induction of protein tyrosine kinase (PTK) activity. The complex relationships among these events have recently been demonstrated in a number of ways. PTK activity is detectable before either PI hydrolysis or increases in [Ca2+]i (1), raising the possibility that tyrosine phosphorylation of intracellular substrates may be the primary signaling event initiated via the TCR. Furthermore, two drugs that antagonize PTK activity also inhibit activation-induced increases in inositol phosphates and [Ca2+]i (2,3). More support for a direct role of PTKs in T cell activation comes from studies in which v-src was introduced into T cell hybridomas (4). Cells expressing the active pp60 v-~ PTK spontaneously produced IL-2, even after depletion of protein kinase C. Moreover, their [Ca a +]i 1Abbreviations used in this paper: GAPDH, glyceraldehyde phosphate dehydrogenase; PI, phosphoinositide; PTK, protein tyrosine kinase. was constitutively higher than in wild-type cells, and increased to much higher levels upon TCR crosslinking (5). Together, these data suggest that tyrosine phosphorylation has a fundamental role in regulating T cell [Ca 2+ ]i and cellular activation.
The extent to which intracellular substrates are phosphorylated on tyrosine residues is also determined by tyrosine phosphatases. By removing phosphate groups from tyrosines, phosphatases can specifically counter the actions of trfKs. Furthermore, tyrosine phosphatases can have the opposite effect, actually enhancing PTK activity by removing phosphates from negative regulatory tyrosines (6)(7)(8)(9)(10). The best characterized lymphocyte tyrosine phosphatase is CD45 (also known as leukocyte common antigen, T200, B220, and . This molecule is present on most hematopoietic cells and exists as multiple isoforms. The isoforms have primary amino acid sequence differences (due to alternate splicing of at least three 5' exons) and different N-and O-linked glycosylation patterns (11). In contrast to the extraceUular portion, the in-traceUular piece of CD45 is invariant. It consists of two homologous domains that are similar in sequence to a placental tyrosine phosphatase, and CD45 has intrinsic tyrosine phosphatase activity (12). Recent work has shown that CD45deficient T cells are aberrant in their ability to proliferate or, in the case of cytotoxic T cells, to kill target cells when stimulated with antigen (13,14). Furthermore, decreases have been noted in TCR-mediated PI hydrolysis, Ca 2 § flux, and PTK activation (15,16).
In the present report we examine a series of murine T cell CD45 loss variants. The analyses indicate a critical role for CD45 in normal tyrosine phosphorylation homeostasis. Furthermore, these variants reveal the existence of both CD45dependent and CD45-independent mechanisms of mobilizing intraceUular Ca 2 +. The latter appears to be a novel pathway that predominantly involves late osdllations of Ca 2+ derived from intracellular stores.
Materials and Methods
Cells The YAC-1 wild-type T lymphoma cell (designated WT) was derived from an A/Sn mouse that had been inoculated with Moloney leukemia virus (17), and was obtained from the American Type Culture Collection (Rockville, MD). To generate CD45deficient variants, the YAC-1 buLk population was treated with ethyl methanesulfonate (500/~g/ml) overnight. The cells were washed and cultured in medium alone for 5 d. After this time the cells were stained with the anti-CD45 antibody M1/9.3 (Boehringer Mannheim, Indianapolis, IN) plus FITC-goat anti-mouse F(ab')2, and the dullest 2% of the population were sterilely sorted by flow cytometry (FACS IV| Becton Dickinson & Co., Mountain View, CA). A total of six such sterile sorts were performed, each occurring 3-7 d after the previous sorting, and the dullest 2% of the cells were selected each time. After the sixth sorting the cells were subcloned at 0.3 ceUs/well; the N1 and N2 subclones were isolated from different wells. To obtain cells expressing intermediate amounts of CD45, the CD45 l~ population obtained after the sixth sorting was sorted once again, but this time the brightest 2% of the cells were sdected, and then subdoned at 0.3 cells/well in 96-well microtiter plates (3596; Costar, Cambridge, MA); the D17 and D4 cells were isolated from different wells. After -2 mo of continuous culture it was noted by flow cytometry that a small fraction (1-2%) of the CD45-N2 cells once again expressed wild-type levels of CD45. The CD45 § cells in this population were incubated with anti-CD45 and selected with magnetic beads (Dynabeads M-450; Dynal, Inc., Great Neck, NY) coated with sheep anti-rat IgG, and subcloned at 0.3 cells/well. Approximately one-third of the subdones expressed high levels of CD45. Three such CD45 revertant subdones, N2.R23, N2.R33, and N2.R47, were selected for further study. LK 35.2 cells are Fc receptor-bearing B cell hybridomas (18). All ceUs were grown in medium consisting of RPMI 1640 plus 10% FCS, 5 x 10 s 2-ME, 2 mM glutamine, and antibiotics (complete medium).
Flow Cytometry. Flow cytometric analysis was performed with a FACScan | (Becton Dickinson & Co.). Approximately 10 + cells were incubated with the indicated antibodies on ice. After 30 min the cells were washed and stained with either FITC-goat anti-mouse F(ab')2, FITC-goat anti-hamster F(ab')2, or PE-avidin (Jackson Im-munoResearch). Fluorescence data were collected using logarithmic amplification.
Northern Blot Analysis. Total RNA was extracted by guanidinium thiocyanate and prepared as described (25). After ethanol precipitation, the water-solubilized RNA was quantified by spectrophotometry, and 15/~g of total RNA from each sample was separated by electrophoresis (0.9% agarose gel with 6.5% formaldehyde) for 5 h at 60 V (26). The ILNAs were transferred to Nytran membranes (Schleicher & Schuell, Inc., Keene, NH) in 20 x SSC (27). After immobilization of nucleic acids by baking for 2 h at 80~ the falter was prehybridized and then hybridized with specific probes labeled by random hexamer priming (27), A CD45-specific 2-kb HindlII fragment was prepared from the p70ZI.C eDNA clone (generously provided by Dr. Matthew Thomas, St. Louis, MO). The same falter was subsequently hybridized with a glyceraldehyde phosphate dehydrogenase (GAPDH) probe as an internal control. After room temperature washes in 2x SSC, 0.1% SDS, and washing at 55~ in 0.1x SSC, 0.1% SDS, the filter was air dried and autoradiography was performed.
Quantitation of TotallnositolPfiosphates. Cells were loaded with
[3H]myoinositol (14 Ci/mmol; New England Nuclear, Boston, MA) for 3 h at 37~ as described (28). The cells were thoroughly washed and divided into duplicate groups of 106 cells each in tubes containing 106 LK 35.2 cells in medium containing 10 mM LiC12. 2Cll (1:30 dilution of culture supernatant, final concentration) was added and 60 rain later the cells were lysed, lipid extracted, and the total water-soluble labeled products of phosphoinositide hydrolysis measured by anion exchange chromatography (28). The mean cpm achieved for each experimental point was divided by the total cpm incorporated by the cells (.yielding percent labeled phospholipid), and then the percent labeled phospholipid generated in the absence of 2Cll (background) was subtracted to obtain percent change labeled phospholipid.
Measurement of Tyrosine Pfiospfiatase Activity. Tyrosine phosphatase activity was assayed by measuring the release of 32p from a tyrosine-containing substrate, [ValS]-angiotensin II (Sigma Chemical Co., St. Louis, MO) as described (29). Briefly, 107 cells were lysed in 100/~1 of buffer containing 0.5% Triton X-100 in Tris/NaC1 buffer, pH 7.4, leupeptin, aprotinin, and PMSF, for 30 rain on ice. Either 10 ~1 of lysate was analyzed direcd)', or 50/~1 of lysate was immunoprecipitated with M1/89 plus protein G-Sepharose beads (Gibco-BRL, Gaithersburg, MD). 10/~1 of labeled substrate was added to the lysate or the beads and the reaction was allowed to proceed at 30~ for 60 s and then terminated by the addition of 5% activated charcoal suspended in 20 mM Hepes, pH 7.4. Supernatants were counted in a B scintillation counter.
Pfiospfiotyrosine Iraraunoblotting. 5 x 107 cells were resuspended either in 1 ml of complete medium (5-6 x 107 cells/tube) and kept on ice for 30 min (unstimulated) or in I ral of 2Cll culture supernatant and incubated for 30 rain at 37~ with frequent shaking (stimulated). The cells were then washed in ice-cold PBS with phosphatase inhibitors (0.4 mM sodium orthovanadate, 0.4 mM EIYrA, 10 mM sodium fluoride, and 10 mM sodium pyrophosphate; pH 7.6), resuspended in lysis buffer (0.5% Triton X-100, 50 mM Tris, and 150 mM NaC1) with phosphatase inhibitors for 30 min on ice. After centrifugation, the postnuclear fractions were immunoprecipitated with protein A-agarose beads precoated with either 2Cll or 4G10 (antiphosphotyrosine mAb; Upstate Biotechnology, Inc., Lake Placid, NY) for 2 h on ice with frequent shaking. The beads were boiled in reducing sample buffer (50 mM TKIS, pH 6.8; 1% SDS, 10% glycerol, 3% 2-ME, plus bromophenol blue and phenol red) for 5 rain, pelleted, and the supematants separated on SDSoPAGE. The proteins were transferred to nitrocellulose, immunoblotted (using BSA as a blocking agent) with the 4G10 antibody followed by 12SI-protein A-agarose (ICN, Costa Mesa, CA), and autoradiography was performed.
Measurement of[Ca~ § Determination of [Ca2+]i by flow cytometry was modified from the procedure of Rabinovitch and June (30). Briefly, the indicated cells were loaded with 2.5 #M indo-1 (Molecular Probes, Junction City, OK) in HBSS for 25 min at 31~ After establishing baseline values, purified 2Cll (20 #g/ml final dilution) was added as indicated. Cells were then analyzed for violet/blue fluorescence emission ratio (395 nm/500 rim). A fluorescence digital image processing system was used to detect changes in [Ca 2+ ]i in individual ceils by a procedure modified from Poenie et al. (31). Briefly, cells were loaded in 3 #M fura-2 AM (Molecular Probes) at 37~ for 30 rain in the presence of 0.0175% F-127. After settling on a glass cover slip, the cells were analyzed in a chamber maintained at 36.5 ~ + 0.25~ containing 2 ml of medium consisting of HBSS (containing 1 mM CaCh), 0.1% FCS, 10 -5 M 2-ME, and 20 ram Hepes. A baseline 350/380 ratio was established, after which purified 2Cll (20 #g/ml final) was added while images were acquired and stored digitally at 15-s intervals. The hardware consists of an image processor (FD5000; Gould, Fremont, CA), an Axiovert microscope (Carl Zeiss, Inc., Thomwood, NY), and a filter changer (Ludl Electronic Products, Scarsdale, NY) with excitation filters centered at 350 _+ 10 and 380 _+ 10 nm. The image processor and filter changer are interfaced to a Microvax host computer (Digital Equipment Corporation, Maynard, MA). Images (16 at each wavelength) are acquired at 30/s through a CCD camera (Cohu, San Diego, CA) and image intensifier (Videoscope, Washington, DC) and processed according to code kindly provided by R. Y. Tsien.
[CaZ+]i values were calculated from fura-2 ratio by the equation: [Ca2+]i = g(K-K~)(Km~-R); where R~ and tL~ are the 350 nm/380 nm ratios obtained in Ca 2+ absence or saturation, respectively. K is the product KD(F0/F,), where KD is the effective dissociation constant of fura-2 with respect to Ca 2 § (224 riM), F0 is the 380-nm excitation efficiency in the absence of Ca 2+, and F, is the 380-rim excitation efficiency at saturating Ca 2 § concentrations (31)(32)(33). The parameters are determined in the bath chamber from solutions containing fura-2-free acid.
Generation of YAC-1 CD45 Loss Variants. YAC-1 T cells
were analyzed by flow cytometry for cell surface expression of molecules involved in T ceU activation ( Fig. 1 A). The wild-type YAC-1 cells (WT) express both CD45 and the TCR. These cells were chemically mutagenized and CD45-deficient subdones sdected by repetitive rounds of cell sorting, as detailed in Materials and Methods. To allow a quantitative evaluation of the relationship between CD45 expression and biological behavior, four YAC-1 subdones that expressed progressively lower levels of CD45 were selected for study. As shown in Fig. 1 A, the expression of CD45 (relative to WT cells as judged by mean fluorescence inten-sity) was: D17, 25%, D4, 3.5%, N1 and N2, undetectable. The TCR was expressed equally by all of the cells. Similar analyses revealed that all of the cells expressed equivalent levels of Thy-1 and LFA-1; none of the cells expressed detectable CD4, CD8, or Ly-6 (data not shown). The CD45 and CD3 cell surface phenotype of two representative CD45 + revertants, N2.R23 and N2.R47, are shown in Fig. 1 B. Similar results were obtained with the N2.R33 subdone (data not shown).
To determine why cell surface CD45 was not expressed, RNA from WT cells and its variants was subjected to Northern blot analysis with a CD45-specific eDNA probe. WT cells expressed easily detectable CD45 mRNA (Fig. 2). The hierarchy of CD45 mRNA expression paralleled the cell surface expression of CD45: WT > D17 > D4 > N1, N2. The two cell surface CD45-variants, N1 and N2, expressed barely detectable CD45 mRNA. Thus, the loss of CD45 from the surface of the YAC-1 variants is due to a defect in transcription or, possibly, in posttranscriptional stabilization of CD45 mRNA.
CD45-specific and total cellular tyrosine phosphatase activity was determined (Fig. 3). WT cells had abundant anti-CD45-precipitable tyrosine phosphatase activity. The amount of activity precipitated from the CD45 variants correlated well with the relative abundance of this molecule as detected by flow cytometry and Northern blot analysis. In particular, no CD45-associated tyrosine phosphatase activity was detected in the N1 and N2 cell lines. Comparison of enzymatic activity in whole cell lysates suggests that *40% of the total tyrosine phosphatase activity in the YAC-1 cells can be attributed to CD45. Thus, WT cells and the CD45 variants provide a series of cells that have a graded range of cell surface CD45 protein, CD45 mRNA, and CD45-associated tyrosine phosphatase activity.
Altered Intracellular Steady-State Tyrosine Phospkorylation as a Function of CD45. TCR-driven and kinase-mediated tyrosine phosphorylation of the TCR ~" chain is an early event in T call activation (34). Given the dose physical proximity of CD45 and the TCR (29), we asked if the loss of the CD45 tyrosine phosphatase influenced the level of ~" phosphorylation in the resting and the activated state. YAC-1 calls, unstimulated or stimulated with 2Cll (anti-CD3), were lysed, and anti-CD3 immunoprecipitates were analyzed for phosphotyrosine-containing species by immunoblotting. YAC-1 cells exhibited two phosphorylated ~" species (Fig. 4). As described (34), tyrosine phosphorylated ~" chains (which have six intracellular tyrosine residues potentially available for phosphorylation [35]) migrate aberrantly on SDS-PAGE, having an apparent relative molecular mass of 21,000 rather than 16,000. The two bands in YAC-1 cells (pp21 and pp23) presumably represent the usual phospho-~" and a hyperphosphorylated form (the unphosphorylated ~" chain in YAC-1 cells has a mass of 16 kD, data not shown). As shown in and pp23 phospho-~'. Both forms increased substantially with activation. D4 had more resting phospho-~" than either WT or D17, and this again increased with activation. In contrast, the N1 and N2 cells had maximally phospho-~" in the unstimulated state. If anything, stimulation with 2Cll actually decreased the amount of phospho-~" in these cells to a small degree. In another experiment (Fig. 4 b), unstimulated WT ceils expressed a small amount of phospho-~" that increased in intensity after exposure to 2Cll. The CD45-N2 cells had high levels of phospho-~" in the absence of stimulation. Two different CD45 revertant ceils, N2.R23 and N2.R47, had phospho-~" levels that were indistinguishable from the WT ceils. These results demonstrate that CD45 expression is inversely related to spontaneous TCR ~" phosphorylation.
There are tyrosine phosphorylated substrates other than ~" in activated T ceils (36,16). To determine if the presence or absence of CD45 influenced these substrates as weU, 1ysates of YAC-1 cells were immunoprecipitated with antiphosphotyrosine antibodies, resolved by SDS-PAGE, and immunoblotted with an antiphosphotyrosine antibody (Fig. 5). Fig. 5, A and No bands were detected when phosphotyrosine was included while the CD45-cells were being immunoblotted with the 4G10 antiphosphotyrosine antibody, while the addition of phosphoserine or phosphothreonine had no effect on the intensity of the bands, confirming this reagent's high degree of specificity (data not shown). Thus, loss of CD45 results in a global alteration of tyrosine phosphorylation status.
TCR-mediated PI Hydrolysis and Ca 2 § Flux Vary as a Function of CD45 Expression.
Other TCR-coupled signaling pathways were also examined. YAC-1 cells were loaded with [3H]myoinositol and stimulated with 2Cll or A1F4- (Fig. 6). Both the WT and the D17 cells responded well to TCR crosslinking. D4 cells responded poorly. Neither N1 nor N2 ceils gave any detectable response. In contrast, all cells responded equivalently to A1F4-, a reagent that stimulates GTP-binding protein-dependent T cell PI h3atrolysis in a TCRindependent fashion (37). The fact that D17 ceils were as good as WT cells in their ability to hydrolyze PI (and to mobilize 838 CD45 Regulates TCR-mediated Signaling and Phosphotyrosine Homeostasis [Ca 2+ ]i; see below) indicates that for YAC-1 cells a four-to fivefold decrease in CD45 expression is still sufficient to support normal TCR-mediated responses. Only when the CD45 level falls dramatically (to that of D4 cells, <5% of normal) is a signaling defect revealed. Ca 2+ flux is another measure of TCIk-mediated cell activation. In preliminary experiments it was found that the Fc receptor-negative (based upon staining with the anti-Fc receptor 2.4G2 antibody [38], data not shown) YAC-1 cells responded well to 2Cll in the absence of accessory cells (i.e., no external crosslinking of 2Cll was required). Therefore, YAC-1 cells were loaded with the Ca2+-sensitive dye indo-1 and analyzed by flow cytometry (Fig. 7 A). Whereas W T and D17 cells responded to 2Cll with a rapid and vigorous increase in [Ca2+]i, D4 cells responded poorly. However, the shape of the D4 response curve was similar to those of the clones that expressed higher levels of CD45 (i.e., there was a rapid and short-lived rise). N1 and N2 cells displayed a very different Ca 2+ response pattern. First, no 2Cll-induced early (within I min) response was detected. However, during the m6 min of monitoring, there was a slow and steady increase in mean [Ca2+]i (baseline/maximal [Ca2+]i levels: N1, 134 nM/236 nM; N2, 134 nM/265 nM). Monitoring the fraction of responding cells was also revealing (Fig. 7 B). The large majority of W T (84070) and D17 (94070) cells responded at the peak; only about half of the D4 cells responded. Neither N1 nor N2 cells responded by 1 rain after activation. However, by 3 min after stimulation, it was evident that a significant number of cells were responding with elevated WT cells or the CD45-deficient variants were cultured in medium alone or in the presence of 2Cll for 30 min at 37~ At that time the TCR chains were immunoprecipitated, separated by SDS-PAGE, and immunoblotted with antiphosphotyrosine antibodies. The YAC-l-phosphorylated TCR ~" chain appears as a doublet. (b) Phospho-~" was analyzed as in a. Cells were cultured in medium alone (-) or with 2Cll for 30 min at 37~ (+). Because the intensity of the bands varied greatly, two different times of exposure are shown: the lane in group A represents the immunoprecipirate from N2 cells exposed for 10 h; the lanes in group B represent immunoprecipitates from WT and N2 cells, as well as two CD45 revertant cells, exposed for 60 h.
[] N1
[] N2 k 2Cll ~r~ Figure 6. Phosphoinositide hydrolysis as a function of CD45 expression. 106 WT calls or its CD45-deficient variants plus 106 LK 35.2 cells were cultured in medium containing t0 mM LiCh. Anti-CD3 (2Cll) or AIF4-was added at time 0, and total inositol phosphate generation was measured 60 or 30 rain later, respectively. The data are presented as the fraction of" total incorporated label that was duted from the ion exchange columns.
Changes in [Ca2+]i were also measured in CD45 revertants. As shown in Fig. 7, C and D, three independent revertant cell lines, N2.R23, N2.R33, and N2.K47, responded to activation with a Ca 2 + response that was indistinguishable from the WT cells. In contrast, the N2 cells (from which the revertants were derived) displayed a delayed and rather small increase in [Ca 2+ ]i. Five other CD45 revertants were analyzed, and all were found to express the wild-type pattern of [Ca 2+ ]i rise; two CD45-subclones derived from N2 at the same time as the CD45 + revertants had delayed Ca 2+ responses of the type manifested by N2 (data not shown). Thus, there are two distinct Ca z+ response phenotypes that depend upon the presence or absence of CD45.
[Ca2+]i in Single Cells. Flow cytometric analysis is
limited in that it can only provide a static picture of responding cells. Therefore, we employed computer-aided fluorescence imaging, which allows one to monitor the [Ca2+]i of individual cells over time, in order to analyze the atypical Ca 2 + response of the CD45-deficient cells. WT and the two CD45-subclones were labeled with fura-2 and monitored for changes in [Ca 2+ ]i after stimulation with 2Cll (Fig. 8). I I I I I I 2 3 4 5 6 7 800 A 700 6oo 2i cillations, although they were small in amplitude and only manifested by a minor fraction of the cells. Surprisingly, both N1 and N2 cells gave a qualitatively different response. First, 2Cll did not induce a rapid increase in [CaZ+]i. In fact, cells generally did not begin to respond until several minutes after stimulation. Second, unlike the concerted response of the WT cells, the response of the CD45-cells was unsynchronized. Third, whereas the WT cells generally manifested a large early Ca 2+ rise with little in the way of oscillations, the N1 and N2 cells exhibited only oscillations, which were much larger in amplitude and more frequent. The fraction of CD45-cells that responded over the course of the observation period varied from experiment to experiment, but was always >50%. In this experiment the fraction of N1 cells that responded was 75% and the fraction of N2 cells that responded was 88%. The responses of seven representative individual cells (WT vs. N2) are shown in Fig. 9. N1 cells had profiles that were indistinguishable from N2 cells (data not shown). EGTA was also used to chelate extracellular Ca 2+. This maneuver greatly blunted the early response of the WT cells, and completely abrogated any late response. In contrast, the Ca 2+ oscillations by the N2 cells (and N1 cells; data not shown) were hardly affected by the chelation of extracellular Ca 2+. The response of the YAC-1 cells to 2Cll was specific; no changes in [Ca2+]i were detected when Thy-1 was bound by the 30H12 antibody (Fig. 8 B). Furthermore, both CD45 + and CD45-cells labeled equivalently with fura-2 and responded to a TCR-independent stimulus, thapsigargin, which directly elevates [Ca2+]i by inhibiting the sarcoplasmic or EP, ATP-dependent Ca 2+ pumps responsible for sequestering [Ca2+]i (39). Together, these results demonstrate that the absence of CD45 allows T cells to manifest a specific large and periodic TCR-driven release of Ca 2 + from intracellular stores.
Discussion
The data in this study provide a view of the profound way in which the tyrosine phosphatase CD45 influences T cell signal transduction pathways. Three major and commonly studied signaling pathways were aberrant in CD45-cells. The defects, which consisted of dysregulated cellular homeostatic mechanisms and responses to TCR-mediated activation, were not merely of a quantitative nature. Rather, CD45cells had qualitative changes in both tyrosine kinase/tyrosine phosphorylation and TCR-mediated Ca 2+ mobilization. The most obvious consequence of CD45 loss was the constitutive phosphorylation of multiple proteins on tyrosine residues. Because of the substantial amount of tyrosine phosphatase activity remaining in the YAC-1 CD45-deficient cells, we were surprised at the extent of the hyperphosphorylation. This observation underscores the high degree of tyrosine phosphatase specificity in vivo, and proves that for at least a cohort of FFK substrates (including the TCR ~" chain), other tyrosine phosphatases will not substitute to serve CD45's regulatory function. There are a number of possible explanations for the hyperphosphorylation. The simplest possibility is that constitutive CD45 phosphatase activity normally counters the activity of constitutively active PTKs. Although the tyrosine-phosphorylated species detected in N1 and N2 cells include those induced in WT cells by TCR ligation, many new bands also appear, implying that either CD45 regulates phosphorylation initiated by some PTKs that are associated with T cell activation and some that are not, or that T cell activation somehow limits the variety of substrates available for the FFKs, perhaps by regulating their subcellular localization. Another possible explanation for the enhanced spontaneous phosphorylation is that the absence of CD45 activates one or more PTKs. Although there is evidence that CD45 activates the p56 ~k PTK by removing a regulatory (inhibitory) phosphate from Tyr-505 (40)(41)(42), phosphorylation of some regulatory tyrosines can also enhance enzymatic activity. Tyr-394 may play such a role in p56 ~k, since site-directed mutagenesis of this residue prevents the enhancement of enzyme activity associated with mutation of Tyr-505 (43). In the case of c-src, mutation of Tyr-416 (a tyrosine kinase target) to a phenylalanine suppresses the transforming ability of this kinase, and diminishes the PTK activity induced by Tyr-527 removal (8). Thus, lack of CD45 could potentially increase PTK activity by allowing unopposed phosphorylation of regulatory (enhancing) tyrosine residues, or by some more complex intermediary set of interactions. Interestingly, the doublet seen at 60 kD in activated WT, which is very prominent in unactivated N1 cells, is in the molecular mass range that is commonly associated with autophosphoryhted src family members (44). Efforts are now underway to identify which PTKs are active in the CD45-YAC-1 subclones.
The extreme degree of spontaneous hyperphosphorylation contrasts with previous results reported with CD45-deficient Jurkat cells, in which baseline tyrosine phosphorylation was normal, although tyrosine kinase activity was not induced by TCR crosslinking (16). It is possible that this discrepancy is due to the degree to which the YAC-1 cells and the Jurkat variant T cells are "CD45 deficient." Whereas N1 and N2 cells had no detectable cell surface CD45 and no enzymatic activity could be precipitated by anti-CD45, the CD45deficient Jurkat cells still expressed "~8% of their usual cell surface CD45 (16). It may be that the small amount of CD45 remaining in the CD45 I~ Jurkat cells was sufficient to maintain tyrosine phosphorylation homeostasis in unactivated cel/s, although insu~dent to aUow normal activation-induced increases in PTK activity. This could not account for another case, in which antiphosphotyrosine immunoblotting of whole call lysates from several CD45-cell lines found them to have a basdine tyrosine phosphorylation pattern similar to that of their CD45 + counterparts, except for hyperphosphorylation of p56 ~k (41). We also had difficulty demonstrating differences in tyrosine phosphorylation between CD45 + and CD45-YAC-1 cells when whole cell lysates were used (data not shown); only when antiphosphotyrosiue immunoprecipitation was first performed was a clear-cut difference seen upon subsequent immunoblotting. This is not dissimilar from the results of Ostergaard et al. (41), in which p56 ~k hyperphosphorylation was inconsistently seen in whole call lysates from three independent CD45-cell lines, although it was quite marked when p56 kk was specifically immunoprecipitated. We are currently using whole cell lysates or antiphosphotyrosine immunoprecipitates to evaluate the tyrosine phosphorylation status of other CD45-deficient cell lines.
An unexpected finding was the novel late Ca 2+ oscillations induced by TCR crosslinking. Data obtained with CD45-deficient human HPB-ALL andJurkat cells suggested that TCR-mediated increases in [Ca 2+ ]i were simply either absent or markedly blunted (15,16). It should be emphasized that the view one obtains of the aberrant Ca z+ response is largely a function of the technique employed (fluorimetry, flow cytometry, or computer-aided fluorescence imaging). Flow cytometric analysis of activated YAC-1 CD45-cells demonstrated a relatively small (*40 nM; n = 7) increase in mean [Ca2+]i, and responding cells were detected only minutes after activation, rather in the usual time of seconds. The finding that [Ca2+]i did change in response to a TCKmediated stimulus in the YAC-1 CD45-deficient cells was convincingly obtained only with the sensitive technique of computer-aided fluorescence imaging. Furthermore, by allowing one to monitor individual cells over time, asynchronous oscillations that do little to raise the mean [Ca2+]i of the population can be detected. The finding that this was the case with the CD45-YAC-1 derivatives is consistent with the small but reproducible finding with flow cytometry that, on average, from 15 to 20% of the cells were responding at any given moment I rain or more after activation.
The loss of the early rise in [CaZ+]i and the novel pattern of TCK-mediated fluctuations in [Ca2+]i is intriguing. Experiments using EGTA to chelate extracellular Ca 2 + indicate that both intracellular and extracellular Ca 2 + contribute to the early response in YAC-1 cells, while intracellular stores provide most of the Ca 2 + for the late response. Although not prominent in WT cells, the CD45-cells manifested late Ca 2+ oscillations that were almost as great in magnitude as the initial response in WT cells. These results suggest that either CD45 normally tends to prevent late oscillations or that loss of CD45 (with the attendant dysregulation of FFK activity) allows atypical late Ca 2+ oscillations to occur. We have recently shown that tyrosine kinase activity, by itsdf, can alter intracelluhr Ca 2+ regulation both in the absence or the presence of TCR-mediated signaling (5). One attractive possibility is that an altered level of tyrosine phosphorylation in CD45-cells, perhaps due to a change in tyrosine kinase activity, is the cause of the aberrant Ca 2+ response phenotype. Alternatively, CD45 may serve to couple the TCR to Ca 2 + responses by directly dephosphorylating critical but uncharacterized substrates. Whatever the exact mechanisms, together with previous studies, these data demonstrate that CD45 profoundly influences all of the well-characterized signaling pathways in T cells. The answer to how a tyrosine phosphatase regulates PI hydrolysis, [Ca2+]i mobilization, and PTKs will ultimately explain a great deal about how the TCR couples to biological function.
We are grateful to Douglas S. Smoot for performing the flow cytometric analyses, Paul A. Fischer for cell sorting, and Allan Weissman for helpful discussions and critical review of the manuscript. | 2016-05-04T20:20:58.661Z | 1992-09-01T00:00:00.000 | {
"year": 1992,
"sha1": "00ede83834f9ec4c2aea162100ab2f7a54aef01e",
"oa_license": "CCBYNCSA",
"oa_url": "http://jem.rupress.org/content/176/3/835.full.pdf",
"oa_status": "BRONZE",
"pdf_src": "PubMedCentral",
"pdf_hash": "bf6d9910faa0ebe52a53a8ed30cd9db21b01738c",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
235638040 | pes2o/s2orc | v3-fos-license | Enhancing student nurses’ clinical education in aged care homes: a qualitative study of challenges perceived by faculty staff
Background Ageing populations are increasing the demand for geriatric care services. As nursing schools respond to this demand, more high-quality clinical placements are required, and aged care homes offer suitable placement sites. Although an aged care experience for students is beneficial, the basis for effective implementation of these placements is yet to be fully established. The aim of this study was to explore faculty staff perspectives on the challenges associated with providing effective clinical education in aged care homes for first-year student nurses. Methods An exploratory qualitative study was performed. Fifteen in-depth interviews were conducted with program leaders of nursing degree programs (n = 4), course leaders (n = 6) and practice coordinators (n = 5) in three Norwegian universities. Data were analysed using thematic analysis. The findings were reported using the Standards for Reporting Qualitative Research (SRQR). Results Five themes were identified regarding the perceived challenges to implementing effective clinical education in aged care homes: (1) low staffing levels of registered nurses limit the capacity to effectively host students; (2) prevalence of part-time teachers can compromise the quality of students’ learning experiences; (3) tensions about the required qualifications and competencies of nurse teachers; (4) variation in learning assessments; and (5) lack of quality assurance. Conclusions These challenges signal key areas to be addressed in quality assurance for effective aged care placements. Further research into the minimum staffing levels required to support student learning in the aged care setting is required. Methods for developing shared practices to facilitate learning in aged care homes need to address the prevalence of part-time teaching appointments. Further research into the levels of qualification and competence required to support student learning in aged care facilities can assist with setting standards for this sector. Finally, academic-practice institutions must engage with government officials and national nursing bodies to develop national standards for clinical education in aged care homes. Supplementary Information The online version contains supplementary material available at 10.1186/s12912-021-00632-0.
Background
Populations are ageing globally, with one in six people expected to be over 65 years of age by 2050. In 2018, for the first time in history, people aged 65 years and older outnumbered children under five years of age [1]. Further, the number of people aged 80 years and older is expected to triple by 2050 [1], making the ageing of the population a momentous demographic transformation in the 21st century.
Nurses constitute a significant percentage of the health and social workforce, so university nursing programs are essential in preparing a workforce to meet future healthcare needs, including those associated with an ageing population [2]. However, a scoping review found little evidence of geriatric or gerontologic theory components in nursing education [3]. Instead, the evidence suggests that nursing students mostly learn about ageing and aged care through placements in aged care homes [4].
The renewed interest in aged care homes as clinical placement settings for nursing students has been prompted by the growing health care needs of an ageing population, saturated clinical placements in health services, and regulatory requirements in countries such as Norway [4]. Aged care home placements are often provided for first-year students to foster learning about the essence of nursing [5] and instil positive attitudes about older people [6].
A review of nursing students' experiences of aged care home placements found that students valued the chance to build relationships with older people, improve their communication skills, and assist the elderly with their daily living activities, but noted that students had to overcome several challenges to achieve learning [2]. For example, when assigned to aged care homes for clinical placement, nursing students did not value working with personal care assistants and often reported that the practice sites were unprepared for them to learn effectively [2].
A recent scoping review of teaching strategies and activities to enhance students' clinical placement in residential aged care facilities found that the most common strategy was to use care staff to provide facility orientation and to serve as student mentors [7]. Whilst this review highlighted a range of strategies, the effectiveness of these strategies was not addressed. The researchers identified timing and delivery of teaching strategies and the need for a multi-faceted and collaborative approaches as critical to student placements in residential aged care facilities [7]. However, methods to deliver strategies remains problematic in this setting.
A review of nurse staffing standards in aged care homes across six countries found a reliance on personal care workers rather than professional or registered nursing staff [8]. In Norwegian nursing homes, 20 % of nursing positions are reported to be vacant, with staff members often holding part-time positions, and around 25 % of the staff were not gerontology trained [9]. Immigrant nurses, often unfamiliar with the native country's culture and history, can comprise up to 43 % of the staff [9]. Collectively, these factors may contribute to the marginalization of aged care homes as appropriate clinical placements for nursing students [9]. Yet, older people living in aged care homes have complex, but stable, nursing care needs, thereby rendering these settings as potentially ideal learning environments for firstyear nursing students.
The predominant clinical placement model for nursing education in Norway and elsewhere in Europe is the preceptorship model [7.] In this model, students work with staff and receive mentoring from a registered nurse (RN) and follow up by a university-based nurse teacher [7,10]. The teacher focuses on the relationship between the RN mentor and the student, supporting the integration of theoretical and practical learning. The teacher is responsible for coordinating the students' learning, acting as a liaison rather than getting involved with hands-on patient care [10]. Hence, contact with and support from these teachers is important for first-year students. This support includes opportunities for critical reflection with peers to provide deeper learning about practice [11,12].
Given the limited numbers of professional nurses or RNs working in aged care homes, the role of the nurse teacher in student learning is significant. For clinical placements, the qualities of the nurse teacher should enhance student learning [13,14]. Nurse teachers should, therefore, be able to form positive relationships with students [15], meet them regularly [10], and be familiar with the nursing care in the placement setting [16]. In one study, differences in learning in an aged care home were attributed to the facilitation styles of nurse teachers [17]. How nurse teachers perceive their work may assist with further developing the nurse teacher role in the aged care home setting.
Placement in aged care homes is further complicated by limited access to experienced gerontology nurse teachers. Nurse teachers with backgrounds in acute care may be less interested in the care of older people and can bring negative attitudes into the aged care home setting [14]. Within universities, first-year student education is often relegated to the most junior academic staff, who may have limited teaching qualifications and may experience stress and role confusion, compromising their ability to effectively support students [18,19]. The limited number of RNs to serve as mentors in aged care homes, combined with the lack of experienced gerontology nurse teachers generally, poses significant challenges to quality learning in aged care home placements.
Globalization and migration have enhanced the internationalization of labour and education, leading to an increase in culturally and linguistically diverse students in internationally orientated nursing degree programs [20]. Over the last decades a sharp increase in the international mobility of students has been seen in Norway as well as globally [21]. Cultural diversity in nursing programs is perceived as an asset providing enriching experiences [20]. However, the greatest challenge for culturally and linguistically diverse students has been associated with clinical education [22], where students' language proficiency influences the successful completion of clinical placements [23]. In a systematic review, one identified barrier to learning in clinical placement by culturally and linguistically diverse healthcare students was reported as experiences with university support and instructions [22].
A survey of Norwegian nursing students on placement in nursing homes found that students viewed the clinical learning environment more negatively than hospital placements on nearly all dimensions [24] as assessed by the Norwegian version of the Clinical Learning Environment, Supervision and Nurse Teacher evaluation scale [25]. The negatively reviewed dimensions included pedagogical atmosphere, nursing care and learning situations, content of the supervisory relationship and leadership style on the ward.
Despite an emerging body of research into how students learn and experience learning in aged care home placements, few investigations have examined faculty perceptions regarding implementing student placements in aged care homes. The reliance on aged care staff mentors with limited gerontological expertise, the complexity of teaching strategies within the preceptorship clinical education model, and the need for collaboration between mentors, students and nurse teachers suggest that further investigation is required. The aim of this study was to explore faculty staff perspectives on the challenges associated with providing effective clinical education in aged care homes to first-year student nurses.
Design
This study adopted an exploratory constructivist approach to qualitative interpretive inquiry. A constructivist approach assumes a relativistic ontology, acknowledging multiple realities, and a subjectivist epistemology, in which understanding is co-created by the participant and the researcher [26]. In-depth interviews [27] were used to explore faculty staff perceptions of the challenges associated with providing clinical education in aged care homes. In-depth interviewing is a research technique comprising intensive individual interviews with a small number of respondents in order to explore a new issue in detail [27]. The Standards for Reporting Qualitative Research (SRQR) guidelines were used [28].
Setting and participants
The study setting comprised six nursing programs across three universities in Norway. These universities were purposively selected and predefined prior to data collection to provide variation [29] in terms of geographical area (rural/metropolitan), institutional size of the nursing programs (from 100 to 300 or more students), and institutional ranking. Institutional ranking is a national process used by the Norwegian higher education sector to monitor student experiences [30]. Both low-and high-ranking universities were included.
Recruitment was based on purposive criterion-based sampling [31]. Nurse program leaders, course leaders and practice coordinators were invited to participate. Nurse program leaders are responsible for the quality of bachelor's degree programs in nursing. Course leaders are responsible for the content, administration, and quality of the courses, including aged care home placements. Practice coordinators are educational administrators responsible for placing students and coordinating with municipalities and placement sites.
Prior to data collection, approval was obtained from each of the three participating universities. Invitations to participate were emailed to eligible staff with information about the study.
Data collection
A semi-structured interview guide was developed based on the integration of the literature and input from a panel of student nurses (n = 3), nurse teachers (n = 2) and nurse mentors (n = 2) (see Supplementary File 1). This panel of end-users ensured that the data collection tool was validated and anchored in the experiences of these key stakeholders. The guide addressed topics such as collaboration between education and practice, learning environment, mentors, nurse teachers, quality assurance work, barriers, and improvement measures. The guide allowed flexibility in responses depending on the participants' interests and experiences. The first and last authors [KAL, IA], experienced qualitative researchers with backgrounds in nursing and education, conducted the interviews. The interviews were conducted in the participants' workplaces and lasted approximately 60 min. Each interview began by explaining the study, confidentiality and informed consent procedure. The data were collected between November 2018 and February 2019.
Data analysis
All interviews were recorded using a digital recorder and transcribed verbatim by three of the co-authors [KAL, CF, IA]. The analytical approach followed Braun and Clarke's [32] framework for thematic analyses, a flexible approach to analysing qualitative data that searches for themes or patterns. Each theme captures data that informs the research question and can assist with identifying a response pattern or meaning within the data set [32]. Hence, the thematic analysis was guided by the research aim and followed the six phases described by Braun and Clarke [32]: (1) becoming familiar with the data; (2) generating initial codes; (3) searching for themes; (4) reviewing themes; (5) defining and naming themes; and (6) producing the report.
Four of the authors [KAL, CF, KA, IA] independently read the interviews to become familiar with the transcripts. They summarized and shared their impressions with the research team and explored different perspectives on the data. The transcripts were initially coded by the first and last authors, who highlighted segments relevant to the research questions to delineate patterns. The coded data were then sorted into potential recurring themes and patterns by looking for perceived challenges. Four of the co-authors [KAL, CF, KA, IA] discussed and achieved consensus by reviewing, modifying and refining the themes and potential sub-themes of the pre-defined analytic concept (e.g., challenges to clinical education). Five themes were identified. Examples of the themes were compared across the data items (e.g., interviews) and the most illustrative were selected for each theme. To ensure confidentiality and mask institutional affiliation, number identifiers were randomly assigned to each participant, including practice coordinators (e.g., PC1-5), program leaders (e.g., PG1-4) and course leaders (e.g., CL1-6).
Ethical considerations
The study was approved by the Norwegian Centre for Research Data (2018/61,309 and 489,776) and exempted from ethical approval by the Norwegian Regional Committees for Medical and Health Research Ethics since no health information or patient data was collected. Participation was based on informed, voluntary, written consent. To protect the anonymity of the participants and the educational institutions, details on university demographics, institutional and participant characteristics are not included in the paper.
Results
Of the 17 potential respondents invited to participate, 15 consented. The participants were nurse program leaders (n = 4), course leaders (n = 6) and practice coordinators (n = 5) across the three universities. The analyses identified five themes related to perceived challenges to clinical education in aged care home placements for first-year student nurses: (1) low staffing levels of registered nurses limit the capacity to effectively host students; (2) prevalence of part-time teachers can compromise the quality of student experiences; (3) tensions about required qualifications and competence of nurse teachers; (4) variation in learning assessment; and (5) lack of quality assurance. These themes are presented and discussed below.
Theme 1: Low staffing levels of registered nurses limit the capacity to effectively host students All participants mentioned that aged care homes were generally welcoming of students on placement. However, placement capacity was limited by the low numbers of RNs available as mentors: The main challenge, as I see it, is capacity for placement, as the entire class is in aged care home placement simultaneously. There is also lower nurse coverage in aged care homes than in other placement sites (PC1).
Participants from one educational institution tried to resolve this problem by dividing the class into two groups with different clinical placement periods. At another institution, students were placed in home healthcare due to the lack of capacity in aged care homes. Many participants emphasized that regulations governing students' clinical placement within primary care and aged care homes were less sufficient compared to placement regulations in specialized health care: The municipalities are not obligated to take students on placements as the hospitals are by regulation. So, it is a bit unpredictable. Suddenly, just weeks before placement, we get notice from an aged care home that cannot take students on placement after all, due to sick leaves, shortages of nurses and so forth. And, what do we do then? (PL2) All three educational institutions reported having formalized agreements with the municipalities with stated capacity for students on aged care home placements. However, some participants indicated discrepancies between the number of placement positions offered and the number of students stipulated in the agreements. Participants described constant changes and unpredictability regarding the number of placement positions that complicated the placement planning process. Reasons for the reduction in placement positions offered by aged care homes included shortage of RNs, lack of supervisory competence, lack of time for mentoring, and understaffing. Moreover, the shortage of RNs could lead to suboptimal placements: There is lower nursing coverage in aged care homes than in other placement sites. We want to have RNs to supervise the students. But due to the shortage of RNs or sick leaves, students might be followed up by experienced auxiliary nurses for parts of their placement period. Students are not always happy about that (CL2).
A few participants questioned whether limited placement capacity would compromise placement quality: Clinical education capacity in aged care home placements is a real challenge. It is difficult to talk about quality, because we are in a state of debt of gratitude towards the municipalities and the aged care home placement sites as they accept students for placements. So, we need to be constantly thankful towards the nursing home placement sites. It also is difficult to make demands and talk about quality. Talking about placement quality becomes a bit secondary (PL3).
Although most participants claimed they had to find alternative ways to maintain positive collaborations within the clinical setting, these solutions sometimes created new problems. For example, one practice coordinator emphasized that some aged care homes had the capacity and the willingness to accept students and provide effective clinical placements. However, these facilities were in rural and remote areas that were difficult for students and the nurse teacher to access: A lot of rural municipalities would be pleased to have and take students on placement in aged care homes. So, I do wish there was a way to supervise the students remotely. We are not using all the available placement sites in aged care homes because they are too far away for the nurse teachers to follow up with the students during placement (PC3).
Each nurse teacher was responsible for 4-24 students, which placed a strain on the available pedagogical support and feedback provided to students: It is challenging for the nurse teachers when they have a lot of students to follow up during the placement period. When a teacher is responsible for following up 24 students, it becomes challenging to keep track and provide individual supervision and feedback (CL3).
Participants did not identify an upper limit of students per nurse teacher. It was emphasized that a teacher's availability, work plan and wishes determined the nurse teacher to student ratio. A course coordinator indicated that a nurse teacher could be responsible for students in several aged care homes, and travel should, therefore, be a factor in student allocations.
Theme 2: Prevalence of part-time teachers can compromise the quality of students' learning experiences Most participants reported that a large number of nurse teachers providing clinical education for firstyear student nurses in aged care home placements were part-time staff. Participants reported that clinical RNs were hired to assume an educational role and act as nurse teachers. Aged care home placements for first-year student nurses emerged as the placement with the highest proportion of part-time nurse teachers: One of the biggest challenges we have concerning first-year students' placement in aged care homes is the lack of continuity among the nurse teachers. We agree that it is important to give students a good placement experience in their first-year placement. However, at our institution, almost half of the teachers we use in aged care home placement are external part-time staff (PL1).
Most of the participants reported that management did not prioritize clinical education nor question the high usage of external nurse teachers: Management knows we need to hire nurse teachers externally to carry out students' placements. It is the management that does not secure enough resources [to prioritize] clinical education. Management tells us to call, call, call somebody you know that can be hired to oversee clinical education. Management is fully aware of the situation. But nothing happens or changes. It has been like this for years (CL4).
Moreover, several participants reported that parttime nurse teachers were often recruited based on faculty employee acquaintances. There appeared to be few formal competence requirements for the nurse teachers to provide clinical education aside from being a RN. For example: We ask our colleagues if there is someone they know that could act as teachers in clinical education in aged care home placements. So, it is a bit random. We sure want them [the hired teachers] to have a master's degree. But a lot of them only have a bachelor's degree. We do not have any specific educational requirements concerning the teachers we hire for overseeing clinical placements (PL4).
Most participants stated they preferred to use internal nurse teachers because first-year students often are more vulnerable and in need of support. However, several participants stated that because of shortages of nursing faculty it was difficult to avoid hiring part-time nurse teachers: It is cheaper to hire a teacher to conduct clinical education than one with higher qualifications. However, [achieving a high-quality education] is difficult when we have a large number of externally hired nurse teachers to carry out clinical practice education that is not part of our internal staff (PL1).
A few participants claimed that clinical education and placement follow-up were a lower priority for the nurse education institutions and that nurses who held doctorates were assigned mostly to advanced research and education. According to one practice coordinator, 'clinical education is given a lower priority among staff than other responsibilities' (PC5).
In addition, participants noted the lack of formal preparation and orientation of the hired nurse teachers prior to the placement period. The educational institutions varied in their hiring practices. One educational institution mentored its externally hired nurse teachers; participants at other institutions claimed this was an area in need of improvement at their institution. In summary, across the educational institutions, problems existed regarding the competence and continuity of the staff entrusted with the supervision of students on placements.
Theme 3: Tensions about required qualifications and competencies of nurse teachers
Participants offered diverse views on the competence level of the nurse teachers. One participant considered clinical experience and expertise far more important than their level of academic degree: You do not need to be an associate professor to carry out clinical education. Nursing is a practical profession. People who have spent time building competence within academia have not been practicing nursing for a long time. I think students would benefit more from having clinical nurses with hands-on knowledge and expertise from the clinical field as nurse teachers. I think they can do a really good job (CL5).
In this quotation, the clinical competence of the teacher is considered critical for student learning in the workplace. However, most participants agreed that the nurse teachers' pedagogical competence and the RN mentors' competencies in supervision and assessment were most salient.
Consequently, participants across educational institutions reported offering free courses to strengthen RN mentors' supervisory competencies, but found it difficult to get the RNs to participate in these courses due to lack of financial compensation and practicalities: The aged care homes do not have the resources to send the RN mentors on courses that the university offers. The mentors don't want to participate on their day off and if they don't get compensation time for the course by their employee. So, I believe that it is more practical challenges than the mentors' willingness to enhance their supervisory competencies (CL6).
Moreover, some participants proposed that students' learning in clinical education needed to be emphasized and that the RN mentors' pedagogical competencies needed to be improved beyond their supervisory skills: I wish we had a system where the RN mentors learned more about workplace learning, learning in general and supervision. There is insufficient focus on learning during the student's placements in aged care homes (PL2).
When talking about the RN mentors' competencies and supervisory skills, a program leader at one of the educational institutions noted: It is a requirement that you as a teacher have university teaching and pedagogical basis of competence. But that is not the same as competencies in supervision. We talk about that RN mentors should be required to have supervisory competencies. However, no one is taking about the same requirements concerning the nurse teachers following up the students on placement. And that is interesting (PL1).
Theme 4: Variation in assessment of learning
Across the educational institutions, assessment of students' performance and competence was inconsistent, with various assessment instruments being used. One institution used a pass or fail grading scale, another used a verbal scale and the third used a numerical scale. The participants reported a range of satisfaction with these assessment instruments. Some justified using a numerical grading scale, while others preferred the freedom of writing a narrative. Several participants, especially course and program leaders, claimed that valid and reliable assessment of students' competence was challenging due to the differences in tasks and student readiness. Reported challenges related to language difficulties, scoring/assessment of learning outcomes, assessment criteria, the RN mentor's competence in assessment, and interaction among the student, RN mentor and the nurse teacher during the assessment process. Assessment was negotiated between nurse teachers and mentors, with nurse teachers making the final decision about individual student performance. For example: Much is up to the nurse teachers; they have the last word. The nurse teachers take control, lead, and make final decisions concerning whether the students have achieved their learning outcomes. … The mentors are the ones that sees [sic] how the student performs in the care of patients and how they perform and behave in the clinical setting. (PL4).
Difficulties with the language used in the competence assessment documents were proposed as a potential barrier for interaction and involvement from both the student and the RN mentors during the assessment discussions. The participants reported that students and mentors reported difficulty understanding the concepts used to describe student learning outcomes and attributed this to linguistic challenges. For example: It is challenging when you have students with a foreign first language and Norwegian as a second language. The same goes for the mentors. Because some of the mentors can also be difficult to understand due to linguistics. If you have a student and a mentor that both have linguistic challenges paired together, than you can have a real challenge. (CL4).
At another institution, the course leader shared a similar opinion: It is difficult with students with linguistic challenges who have not mastered communication with patients or staff. Then, there can be a lot of misunderstandings. This is something I find worrying (CL5).
Theme 5: Lack of quality assurance
Few participants were familiar with goal-oriented efforts directed towards ensuring quality in clinical education in aged care home placements. Goal orientation existed at a faculty level, but not in terms of clinical education or placement quality: We do not have a specific strategy when it comes to improving or ensuring placement qualitywe do not (PL1).
A course leader at another educational institution concurred: When it comes to quality in clinical education and aged care home placements, we don't do much except provide supervisory courses to enhance the mentor's competencies. Except for that we don't do very much (CL3).
One participant reported that their institution monitored its own performance in clinical education and placement quality through self-reports from students, teachers, or the practice field (e.g., the RN mentors or placement sites). Some educational institutions collected students' course evaluations, which included clinical education based on standardized course evaluations. However, several participants emphasized that evaluations were only randomly followed up. The responsibility for monitoring placement quality was seen as relying on the individual nurse teacher and his or her initiative to conduct evaluations after the placement period: It is up to each individual teacher. Or, frankly, it is written in the instruction contract that the teacher should conduct an evaluation meeting with the stakeholders in the aftermath of placement. However, I must admit that I do not have full control over it, if they [evaluation meetings] are conducted and if so, who participates in these meetings, if it is with the RN mentors or with the management team at the aged care home. No, I strictly don't know (PL1).
Moreover, several participants reported that if the nurse teachers conducted evaluations on their own initiative, these data were not followed up by the educational institutions and used for systematic quality improvements. For example: If we gather information based on placement experiences from the various stakeholders, it will generate a huge amount of data. Evaluations commit. If we gather all these placement experiences and evaluations, we need to do something about them, follow them up in some way. We get struck with a lot of information without having a plan to proceed with it. A lot of times I feel that we gather a lot of information without knowing what to do with it and deal with it (CL2).
Most participants described the effort to assure the quality of clinical education as insufficient: We do not conduct evaluation meetings with practice or with students in the aftermath of placement. We don't. But it is something we probably should do and improve (PL3).
Discussion
The findings from this study suggest at least two key challenges in providing effective first-year student clinical placements in aged care homes: (1) limited access for student placements; and (2) lack of qualified nurse teachers and available RN mentors. Other challenges pertain to the reliability and validity of assessment practices and the lack of a process for improving the quality of nursing student education from within the educational institutions.
Because the clinical placement model in aged care homes requires students to work with qualified staff [10], the capacity to host students is limited by the low number of RNs, understaffing and reliance on agency staff. Some placements were cancelled at short notice. Thus, the findings suggest that due to limited capacity, aged care homes may be less than optimal clinical placements.
Norway has more than 950 aged care homes [9], which should ensure an adequate number of placement opportunities. A case study of selected Norwegian nursing homes found variations in nurse competence and staffing [33], with consequences for students' access to RNs. In the current regulatory environment, this variability complicates the formation of collaborative legal agreements between the higher education institution and each aged care home. These agreements may also differ when accounting for the unique capacities of each municipality and their geriatric facilities. Making and monitoring these arrangements is labour-intensive, which may create an impediment to aged care home placements for nursing students.
In Australia, guidelines for collaborative agreements have been developed [34]. Creating such guidelines as part of a broader policy program around quality clinical supervision for students in aged care homes and other health services may enhance access to aged care homes for student placements. However, these agreements must specify minimum staffing levels to ensure adequate student learning. Further research into the minimum levels of staffing required to support student learning in the aged care home is required. In Norway, as in most EU countries, no specific educational requirements or national standards guide the quality of clinical practice placements and mentoring practices [35]. However, in the UK, national standards for clinical education have existed for many years [36].
The lack of qualified teachers in aged care homes is manifested as a high proportion of part-time nurse teachers and staff with limited geriatric qualifications and competence. Moreover, there was strong agreement among participants that part-time nurse teachers did not provide consistency for first-year students, as they are not involved in theoretical teaching on campus prior to placement. Further, given the shortage of RNs available to mentor and support student nurses' learning in aged care homes, support from nurse teachers may become even more important. Indeed, a study by Skaalvik et al. [37] found that students evaluated the role of the nurse teacher as more important in their first year than in their final year of study. Furthermore, the use of parttime nurse teachers in aged care homes required a high resource investment to prepare teachers for their roles and often lacked a systematic approach, leading to high variability in quality. Methods for preparing teachers to work in the aged care setting require further development, including online options possibly supported by a community of practice model for peer support and development of shared practices.
Other approaches to clinical education in aged care homes require further investigation. Due to budget cuts and increasing constraints, educational leaders may commit to long-term partnerships with one or a small group of aged care homes [38]. Having nurse teachers consistently assigned to the same aged care facility is essential for developing and strengthening their collaborative relationship [38]. In emerging models, nurse teachers can be jointly funded by the facility and the institution of higher education partners. Between student placements, these teachers could offer education and training in student supervision for RNs and collaborate with staff to identify and develop learning activities that are appropriate for students' learning objectives. This model has been used in the Student Nurse Led Ward model of clinical education in aged care homes with some success [39].
A limited amount of teachers in aged care homes possess geriatric qualifications and competencies. In the United States, a group of gerontology nurses has developed national competency standards for gerontological nurse educators, which are endorsed by the National Hartford Center of Gerontological Nursing Excellence [40]. Further research into the applicability of these competencies to the nurse teacher role is aged care homes is warranted. For sustainable and high-quality clinical education, future research should critically explore and extend our understanding of the relationship between nurse teachers' clinical expertise, competence, pedagogical skills set and students' learning outcomes.
Lack of pedagogical competence and supervisory skills among RN mentors (e.g., clinical nurses) who supervise students in aged care homes reflects a frequently mentioned challenge related to supervision of students in aged care facilities [24]. Developing educational interventions such as web-based educational support and peer networks for RN mentors may also provide a method for improving their pedagogical knowledge and mentorship practices [41].
Another notable challenge in providing clinical education in aged care homes pertained to consistent, reliable, and valid assessment practices. This challenge is not unique to aged care homes or to Norway. Research into clinical nurse and academic perspectives on student clinical assessment in Singapore suggests that the lack of a valid and reliable assessment tool, limited mentor competence in assessment, and limited academic-clinical collaboration were common issues [42]. A systematic review of student competence assessment found that the use of a valid and reliable assessment tool, with clearly identified criteria, and continued education and support for mentors are critical to quality learning [43]. Moreover, study findings indicate that nurse teachers play a dominant role during the assessment process due to a lack of engagement from RN mentors. Given the homogeneity in aged care home populations, there is an opportunity to develop an international collaborative into learning and assessment in aged care homes, possibly starting with establishing standards for first-year nursing students.
However, language difficulties associated with an internationally diverse workforce and student population appeared to have a profound effect on students' assessment in the aged care home context, contributing to reduced clarity about performance expectations. In this study linguistic challenges among students and RN mentors were reported to constrain student competence assessment. The importance of workforce diversity within healthcare systems to reduce health care disparities has been emphasized [44]. However, the effects of linguistically diverse RN mentors on mentorship and assessment practices, as well as on student learning in aged care placement, are not well established. Further research into the emerging educational challenges associated with international diversity is required.
The final challenge was the lack of a systematic process for improving the quality of nursing students' clinical education, especially in aged care homes. Some institutions collected information from clinical placements, but the volume of data produced was difficult to review due to the academic staff's limited resources. For those collecting student experience data, the great variation in placement experiences is consistent with other research conducted in Norway [45]. For a richer and more cohesive understanding of the aged care home as a clinical placement, evaluations and feedback from RN mentors, the placement host, nurse teachers and student nurses are recommended. Again, an international collaborative is required to develop standards for the quality of educational experiences in the aged care home setting.
Clinical placements in aged care homes may not be sufficient preparation for graduate nurses who will be working with an older population. Lane and Hirst [4] recommend, at a minimum, a carefully constructed curriculum incorporating gerontological theories and placement experience in aged care homes. McSharry and colleagues [46] suggest that lecturers have a responsibility to bring student experiences from aged care homes into the classroom for analysis and further learning. As curricula are designed with more geriatric content, comprehensive studies are required to investigate those educational strategies that improve student competence in geriatric care [3].
Study limitations
This study has several limitations that merit consideration when interpreting the findings. First, this study was based on a relatively small sample conducted within a Norwegian context, which restricts the transferability of the findings. Nevertheless, the findings and issues raised are relevant in a national and international context. Sample size and data saturation in qualitative research has been subject to enduring discussions due to a variety of conceptual understandings [47]. In this study, sample size was defined prior to data collection according to the purposeful criterion-based sampling strategy [31] and the predefined study settings. The sample was highly specific for the aim of the study and the interview dialogue was strong, which enhances information power [48]. The concept of information power is related to saturation, and indicates that high sufficiency of information within a sample deemed relevant for the study aim requires a lower number of participants [48]. Furthermore, a high degree of consensus emerged during data analysis, in which themes were replicated across the data set and deemed sufficient to satisfy the exploratory nature of this in-depth study [49].
Furthermore, potential biases should be acknowledged since the data collection and analysis were conducted by researchers with a background in nurse education, which entails a prior understanding of the context. To control for research bias, we applied triangulation during data analysis. Four of the authors participated in data analysis and reflected upon the findings, which provided a basis for checking interpretations [50]. The analysis was not reviewed by the participants (i.e., member checking of transcripts and interpretations), which could have been conducted to verify the findings [50].
The study implications on practice
While student placements in aged care homes are important for student learning, the variability of these homes in terms of staffing, combined with the limited preparation of nurse teachers to work in this setting, renders high quality learning serendipitous. A regulatory approach to including geriatric content and experience in undergraduate nursing curricula is long overdue. While a first-year placement offers opportunities to learn about the essence of nursing, opportunities for second-year students to learn more about the clinical presentations of complex comorbidities in this setting have yet to be explored. As the population over 80 years of age continues to grow, the need for nurses in aged care homes will continue to rise. To increase the number of qualified and competent nurses in aged care homes, further research into the benefits of placements for second-and third-year students is required.
Overall, there is an urgent need for national, and possibly international, standards for gerontological nursing education in undergraduate nursing programs, including standards and guidelines to support students' placements. As national governments examine and develop policies to address the burgeoning health issues associated with an ageing population, investment in geriatric nursing knowledge as a generalist competency through educational standards is recommended.
Conclusions
The world population is ageing, and as people live longer, many will experience chronic multiple morbidities. A nursing workforce prepared to work with older people is critical. Aged care homes offer a unique opportunity for first-year undergraduate nursing students to learn more about older people and chronic diseases and to develop skills to communicate and assist them with daily living activities. However, the sector's lack of maturity in providing educational experiences, particularly in relation to the quality of staff, is a barrier to effective placements. While the health services have a history of hosting students and are widely accepted as suitable environments for student learning, optimal learning environments are not yet fully developed in aged care homes. Providing such an environment will require industry regulation around staffing for student supervision and collaboration between the aged care sector and universities to set standards for nurse teachers and training.
This study investigated challenges to clinical education in aged care homes as perceived and experienced by faculty staff. The research provides insight into the many impediments to students' clinical education in these settings, which limit the quality of placement experiences as well as the maximization of learning potential. Hence, we propose that targeted efforts are warranted to enhance student nurses' clinical education in aged care homes.
In conclusion, we propose that nurse education institutions, in partnership with aged care homes, must invest in clinical education. Collaboratively, the academicpractice institutions must lead changes to improve the quality of students experience in aged care homes by evaluating their practices and contributions in clinical education and by engaging with government officials in health and aging as well as national nursing bodies.
Evaluations should be based on indicators appropriate for the aged care home setting that integrate a multiple stakeholder perspective. Moreover, the findings call attention to the need for national regulations and incentives and international research and development collaboratives that support quality of learning in aged care home placements. Taken together, these measures may enhance placements that stimulate and maintain students' interest in the care of older people. | 2020-11-26T09:02:05.750Z | 2020-11-25T00:00:00.000 | {
"year": 2021,
"sha1": "f258dc9a71dba00a0464225b4e733822ba1f0d2f",
"oa_license": "CCBY",
"oa_url": "https://bmcnurs.biomedcentral.com/track/pdf/10.1186/s12912-021-00632-0",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "717928f4883f27260e4fb8e647601c43c17f5592",
"s2fieldsofstudy": [
"Education",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
246625710 | pes2o/s2orc | v3-fos-license | Financial Soundness of Insurtech Companies in India – An Analysis
Insurtech is the latest buzz word that is shaking up the insurance world globally. In the simplest terms, insurtech can be defined as insurance coupled with technology. Though insurtech is still believed to be in nascent stages in India, the insurtech start-ups in India are changing the market dynamics to a great extent. The present study aims to analyse the financial soundness of insurtech companies in India and also to compare the financial soundness of these two companies. Two leading insurtech companies namely, Go digit and Acko are taken as sample of the study. The financial data pertaining to these two companies for a period of three years from 2017-18 to 2019-20 was used for the analysis. The financial indicators of CARAMELS model were used for analysing the financial soundness of these companies. Six parameters of the CARAMELS model namely C - Capital adequacy, A - Asset quality, RA - Reinsurance and Actuarial issues, M - Management soundness, E - Earnings and profitability and L – Liquidity were used for the purpose of analysis. Through the results of the study, it was found that the financial soundness of Go digit was better than Acko as in four out of six parameters i.e. capital adequacy, reinsurance and actuarial issues, management soundness and earnings & profitability, it has exhibited better performance. Further only in terms of Asset quality and liquidity Acko was performing better. The current study has analysed the financial soundness of insurtech companies in India. Six parameters of CARAMELS model excluding sensitivity to market risk have been assessed through various financial indicators. The data pertaining to two insurtech companies for a period of three years from 2017-18 to 2019-20 was considered for the study. The results of the data analysis have shown that in terms of capital adequacy, Go digit has exhibited better performance than Acko. The assets quality of Acko has improved from 2017-18 to 2019-20 whereas in case of Go digit the asset quality has decreased over the three years. An assessment of reinsurance and actuarial issues has shown that Go digit has higher risk bearing capacity with higher retention levels compared to Acko. Even in terms of management soundness Go digit has exhibited better management control in terms of operating expenses than Acko. An assessment of earnings and profitability indicators shows that Go digit has performed well in terms of underwriting profitability and overall profitability. Investment income performance was same in the case of both the companies. In terms of liquidity Go digit has registered lower ratios than Acko. Finally, it can be concluded that the financial soundness of Go digit was better than Acko as in four out of six parameters i.e. capital adequacy, reinsurance and actuarial issues, management soundness and earnings and profitability, it has exhibited better performance than Acko. Only in terms of Asset quality and liquidity Acko was performing better. However, besides the relative performance, both the companies should focus on improving their underwriting efficiency particularly in terms of loss ratio. Partnering of insurtech companies and traditional insurance companies would be the recommended model which would be a win - win situation for both in future.
INTRODUCTION:
Over the last decade, there was a sea through change in the landscape of insurance across the world. Insurtech is the latest buzz word that is shaking up the insurance world globally. In the simplest terms insurtech can be defined as insurance coupled with technology. By adding technologies to insurance, insurers are able to change the way they sold their products and served their customers. Leveraging the latest technologies like artificial intelligence, big data, chat bots, smart phone apps etc. the insurtech companies are able to provide simpler and faster products and there by leaving a seamless experience to their buyers. Though insurtech is still believed to be in nascent stages in India, in the last three years, there was a gradual growth in the number of digital players in both life and non-life insurance sectors. Many fully digital insurance intermediaries have also sprung up in the recent times in India. The emergence of insurtech start-up companies like Acko, Toffee insurance, Go digit have started changing the market dynamics in the Indian insurance industry in a big way. The monthly business statements of insurance companies in India which are published by IRDAI [1] reveal an interesting fact about two budding insurtech start-ups in the non-life insurance sector namely Go digit and Acko. According to these published statements, both these companies which have a market share of less than 1% have registered a premium growth of 82% and 164% respectively by the end of FY2019-20. This is relatively a very high growth compared to that of large traditional insurance players. Against this backdrop, it is interesting to understand the sustainability of these companies in the long run and also to know whether the business operations of these companies are resulting in positive earnings or not. Hence, the current study has been taken up to conduct a detailed analysis of the financial soundness of these two companies.
RELATED STUDIES:
A Review of literature related to financial performance of insurance companies and insurtech in general has been presented below: Joo (2013) [2] has studied the solvency position of non-life insurers in India by using ISI predictors. Further a multiple regression analysis was conducted to identify various determinants of solvency of Indian non-life insurers. He found that firm size and claims ratio have a significant impact on solvency position of non-life insurers. Dey, Adhikari and Bardhan (2015) [3] have examined the various determinants of financial performance of life insurers in India. They have found that size and underwriting risk have a significant positive relationship with ROE and leverage and volume of capital have a significant negative relationship with ROE of Indian life insurers. By taking the data of eight non-life insurance companies in India, Daare (2016) [4] has studied the factors determining the profitability of non-life insurance companies in India. He found that liquidity, company size and inflation have significant impact on profitability of these selected companies. Chellasamy & Valarmathi (2017) [5] have analysed the financial soundness of top five non-life insurance companies in India. They also tried to examine the relationship between the various parameters of CARAMEL model. They have found that among the select companies, New India Assurance has exhibited a relatively satisfactory performance in terms of all financial indicators. Pal, Bhattacharjee and Pal (2017) [6] have assessed the future growth potential of non-life insurers in India. They conducted a detailed analysis of underwriting performance of Indian non-life insurers by taking a fifteen-year data from 2000-01 to 2014-15. They have found that there is an underwriting cycle patterns present in the performance of these companies. Stoeckli, Dremel and Uebemickel (2018) [7] have explored the transformational capabilities and characteristics of insurtech innovations in order to understand the value creation of insurance in the digital world. They have developed a model consisting of 14 transformational capabilities and 52 characteristics which explains the firm level value creation as well as role of digital intermediaries in insurance market. From the review of literature, it was observed that though there are many studies that have used CARAMEL model to analyse the financial performance of Indian insurers, none of them have focused on insurtech companies. The empirical studies related to insurtech have not focused on the financial performance of these companies. Therefore, understanding the growing importance of insurtech across the world and particularly in India, the following study has been taken up.
OBJECTIVES:
1. To analyse the financial soundness of insurtech companies in India 2. To compare the financial soundness of insurtech companies in India 4. METHODOLOGY:
Sources of Data
The study is based on secondary data collected from the published annual reports and public disclosures of insurtech companies in India.
Period of Study
The period of study was 3 years from 2017-18 to 2019-20.
Sample
Two insurtech companies currently operating in the non-life insurance sector of India namely, Go Digit General Insurance Company Ltd and Acko General Insurance Company Ltd. were taken as the sample of the study.
Tools and Techniques
In order to analyse the financial soundness of insurtech companies, following the methodology of Chellasamy & Valarmathi (2017) [5], Ghimire & Kumar (2014) [8], Simpson & Damoah (2008) [9] CARAMELS model was used. Six parameters of the CARAMELS model namely C-Capital adequacy, A -Asset quality, RA-Reinsurance and Actuarial issues, M-Management soundness, E -Earnings and profitability and L -Liquidity were used for the purpose of analysis. S-Sensitivity to market risk was excluded due to non-availability of data. The various ratios used to analyse these parameters and their measurement is given in table 1. The results and analysis of various parameters of CARAMELS model through various financial indicators has been presented below.
Capital Adequacy
Capital adequacy of a company indicates whether it has sufficient capital to meet its claims. In this study solvency ratio has been chosen to assess the capital adequacy position of the insurtech companies. Solvency ratio assesses the capacity of an insurtech company to meet its short term and long-term debt obligations. Solvency ratio is measured as the ratio of available solvency margin to required solvency margin. In India IRDAI has stipulated that all the insurance companies should maintain a minimum solvency requirement of 150%. A higher solvency ratio indicates better credit worthiness of an insurtech company and vice versa. The solvency ratios of insurtech companies in India (Table 2) indicate that both Go digit and Acko were meeting the minimum solvency requirement of IRDAI. While Go Digit's solvency ratio was higher in 2017-18 and 2018-19, Acko maintained a higher solvency ratio in 2019-20. [10] and Acko [11]
Asset Quality
This parameter assesses the quality of assets of am insurtech company. As in Jansirani and Muthusamy (2019) [12], equities to total assets ratio was chosen as a proxy to measure asset quality. This ratio explains the proportion of total assets of a company that are funded by its equity capital. Higher this ratio, the better is the asset quality of the company which thereby indicates a sound financial position of the company. On the contrary, a low asset quality is riskier for the company. The equity to total assets ratios of the insurtech companies (Table 3) shows that in case of go digit in the years 2017-18 and 2018-19, the equities are more than total assets but in the year 2019-20 the ratio has fallen to 0.73. This shows that the equity share capital of the company has decreased. In the case of Acko, the ratio is more than 1 in all the years and more over the ratio has doubled from 2017-18 to 2018-19 with a slight decrease in 2019-20.
Reinsurance and Actuarial Issues
In order to evaluate this parameter, the risk retention ratio is chosen for analysis. Risk retention ratio indicates the proportion of risk that has been retained by the insurtech companies without transferring to the reinsurers. This ratio is measured as the ratio of net premium to gross premium. A higher ratio indicates higher retention by the company and a lower ratio indicates lower retention of risk by the company. In a way this ratio indicates risk bearing capacity of an insurtech company. The risk retention ratio of insurtech companies shows that go digit has a higher ratio than Acko in all the three years of sample period indicating a higher risk bearing capacity of go digit than Acko.
Management Soundness
The operating expenses to gross premium ratio have been considered to evaluate the management soundness of insurtech companies. This ratio reflects the operational efficiency of management of the company. A lower operating expense to gross premium ratio is preferred as it indicates lower operational expenses incurred to write the gross premium and vice versa. The operating expenses to gross premium ratio of insurtech companies (table 5) shows that the ratio has decreased from 90.74 in the year 2017-18 to 42% in 2019-20 in case of go digit which is nearly 48% reduction. In case of Acko the ratio was very high in the year 2017 -18 (1521.68%) but gradually the company was able to bring it down to 81.11% by the year 2019-20.
Earnings and Profitability
The earnings and profitability analysis has been conducted through five different ratios namely, loss ratio, expense ratio, combined ratio, investment income to net premium earned ratio and return on equity. The loss ratio indicates the amount of claims incurred in relation to premium earned. The expenses ratio indicates the amount of commission and operating expenses incurred in relation to premium earned. The combined ratio indicates the underwriting profitability of the company. A lower loss ratio, expenses ratio and combined ratio is always preferred as it exhibits the underwriting efficiency of the company. The loss ratio of the insurtech companies (Table 6) shows that both the companies had lower claims than premium earned. In all the three years the proportion of claims incurred to net earned premium of Acko was lesser than that of Go digit. Coming to expense ratio (Table 7), in case of Go digit, it has decreased from 112.75% in the year 2017-18 to 47.76% in the year 2019-20. In case of Acko the expense ratio was 1640.48% in 2017-18 and has come down to 159.60%. However, Go digit has better expense management compared to Acko. The combined ratio of insurtech companies (Table 8) shows that in case of go digit, the ratio was lesser than that of Acko in all the three years. The investment income to net premium earned ratio indicates the extent of investment income generated by an insurtech companies apart from its premium income. It explains the contribution of investment income in the profitability of a company. A higher ratio indicates a higher proportion of investment income in total earned premiums. The investment income ratio of insurtech companies (Table 9) shows that in the year 2017-18 both the companies had negligible or no investment income. In 2018-19 Go digit's ratio was 5.21% whereas Acko's ratio was a little higher with 7.46%. However, by 2019-20 both the companies had a ratio of 10% and 10.09% respectively. Return on equity is the most popular parameter used to measure the overall profitability of a company. As in Bawa and Verma (2017) [13], it was calculated as the ratio of net profits after taxes to net worth. It explains the efficiency of the company's management in utilising the shareholders' funds i.e., how much profit it is generating out of these funds. A higher ROE tells us that the management was able to generate more return on shareholder's funds and vice versa. The return on equity of insurtech companies (Table 10) shows that both the companies were experiencing losses in all the three years of study period. The year 2018-19 marked highest losses for both the companies. However, by 2019-20 go digit was able to reduce its proportion of losses by 72% while Acko could reduce it by 41%. [10] and Acko [11] [10] and Acko [11] [10] and Acko [11] [10] and Acko [11] [10] and Acko [11] 5.6 Liquidity The liquidity analysis was conducted through liquidity ratio. The liquidity ratio explains the capacity of a company to meet its short-term debt obligations. A liquidity ratio of 1:1 is considered to be ideal for any company. The liquidity ratios of insurtech companies (Table 11) shows that the liquidity position of go digit was always lower than that of Acko. The liquidity ratio of go digit has fallen from 0.56 to 0.20 in 2019-10 whereas the liquidity ratio of Acko has fallen from 0.85 to 0.40 in 2019-20. [10] and Acko [11] 6. DISCUSSION AND FINDINGS: It was found that both the insurtech companies has met the minimum solvency requirement of 150% set by the regulator. The insurtech companies have maintained highly adequate capital as they maintain a much higher solvency margin in all the three years than the minimum requirement. The asset quality of go digit was found to have fallen in the year 2019-20. Further the asset quality of Acko was better than Go digit in all the three years. It was found that the average risk retention ratio of Go digit was 76% while that of Acko was 53%. This clearly indicates a better risk bearing capacity of Go digit compared to Acko. During all the three years from 2017-18 to 2019-20 it was found that Go digit's management soundness was effective when compared to Acko. Through the analysis of earnings and profitability it was found that in spite of relatively higher claims incurred, Go digit had a lower combined ratio in all the years. This was mainly possible due to the lower commission and operating expenses of go digit. An observable difference was not found between both the companies in terms of investment income ratio.
By the year 2019-20 both the companies had an investment income ratio of around 10%. It was found that the return on investment of Acko was much lower than that of Go digit in the years 2018-19 and 2019-20. However, both the companies registered negative returns in all the three years. It was found that the average liquidity ratio of Go digit was 0.43 and that of Acko was 0.74. This shows that the liquidity position of Go digit was less than half of Acko.
CONCLUSION :
The current study has analysed the financial soundness of insurtech companies in India. Six parameters of CARAMELS model excluding sensitivity to market risk have been assessed through various financial indicators. The data pertaining to two insurtech companies for a period of three years from 2017-18 to 2019-20 was considered for the study. The results of the data analysis have shown that in terms of capital adequacy, Go digit has exhibited better performance than Acko. The assets quality of Acko has improved from 2017-18 to 2019-20 whereas in case of Go digit the asset quality has decreased over the three years. An assessment of reinsurance and actuarial issues has shown that Go digit has higher risk bearing capacity with higher retention levels compared to Acko. Even in terms of management soundness Go digit has exhibited better management control in terms of operating expenses than Acko. An assessment of earnings and profitability indicators shows that Go digit has performed well in terms of underwriting profitability and overall profitability. Investment income performance was same in the case of both the companies. In terms of liquidity Go digit has registered lower ratios than Acko. Finally, it can be concluded that the financial soundness of Go digit was better than Acko as in four out of six parameters i.e. capital adequacy, reinsurance and actuarial issues, management soundness and earnings and profitability, it has exhibited better performance than Acko. Only in terms of Asset quality and liquidity Acko was performing better. However, besides the relative performance, both the companies should focus on improving their underwriting efficiency particularly in terms of loss ratio. Partnering of insurtech companies and traditional insurance companies would be the recommended model which would be a win -win situation for both in future. | 2022-02-07T16:19:03.558Z | 2020-11-03T00:00:00.000 | {
"year": 2020,
"sha1": "4466db4985c99ad9bc0c0a5d38b550e8b1f65a82",
"oa_license": null,
"oa_url": "https://doi.org/10.47992/ijcsbe.2581.6942.0090",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "43b317435297e6ce691d7399816db60752506e4e",
"s2fieldsofstudy": [
"Business",
"Economics"
],
"extfieldsofstudy": []
} |
73446113 | pes2o/s2orc | v3-fos-license | Treating Patients with Type 2 Diabetes Mellitus Uncontrolled on Basal Insulin in the Czech Republic: Cost-Effectiveness of IDegLira Versus iGlarLixi
Introduction Few patients with type 2 diabetes mellitus (T2DM) achieve recommended glycemic control targets in the Czech Republic. Novel therapies, such as fixed-ratio combinations of basal insulin plus glucagon-like peptide-1 receptor agonists, may contribute to better glycemic control. In the analysis presented here, the present analysis assessed the long-term cost-effectiveness of two fixed-ratio combinations, IDegLira (insulin degludec/liraglutide) and iGlarLixi (insulin glargine/lixisenatide), for the treatment of patients with T2DM inadequately controlled with basal insulin from a healthcare payer perspective in the Czech Republic. Methods A cost-effectiveness analysis was performed over patient lifetimes using the IQVIA CORE Diabetes Model. Treatment effects were obtained from an indirect treatment comparison as no head-to-head data for IDegLira versus iGlarLixi are currently available. IDegLira was compared with two iGlarLixi pens (100 U/mL insulin glargine + 33 μg/mL and 50 μg/mL of lixisenatide, respectively). Direct medical costs associated with pharmaceutical interventions, screening and diabetes-related complications were captured. Deterministic and probabilistic sensitivity analyses were performed. Results IDegLira was associated with gains in life expectancy of 0.11 years and in quality-adjusted life expectancy of 0.14 quality-adjusted life-years (QALYs) versus iGlarLixi, due to a lower cumulative incidence and delayed onset of diabetes-related complications. IDegLira was also associated with higher projected costs due to higher acquisition costs; however, these were partially offset by cost savings from avoided complications. IDegLira was associated with incremental cost-effectiveness ratios of Czech Koruna (CZK) 695,998 and CZK 348,323 per QALY gained versus iGlarLixi pens containing 33 and 50 μg/mL of lixisenatide, respectively. These ratios were below the commonly used willingness-to-pay threshold of CZK 1,200,000 per QALY gained. Conclusion The present analysis indicated that IDegLira was associated with clinical benefits relative to iGlarLixi over patient lifetimes and was likely to be cost-effective in the treatment of patients with T2DM uncontrolled on basal insulin in the Czech Republic. Funding Novo Nordisk. Plain Language Summary Plain language summary is available for this article. Electronic Supplementary Material The online version of this article (10.1007/s13300-019-0569-7) contains supplementary material, which is available to authorized users.
was likely to be cost-effective in the treatment of patients with T2DM uncontrolled on basal insulin in the Czech Republic. Funding: Novo Nordisk. Plain Language Summary: Plain language summary is available for this article.
PLAIN LANGUAGE SUMMARY
• Patients with type 2 diabetes mellitus (T2DM) benefit from reductions in blood sugar levels and body weight, which lower the risk of long-term diabetes-related complications. Novel treatments, such as fixedratio combinations (FRCs) of insulin plus glucagon-like peptide-1 receptor agonists, can help patients to achieve these treatment targets, at low risk of hypoglycemia. • In the Czech Republic, too few patients with T2DM achieve treatment targets, and diabetes imposes a substantial cost burden on the healthcare system. Modern antidiabetic treatments, such as FRCs (e.g. IDegLira and iGlarLixi), provide the means to improve diabetes treatment and reduce diabetes-related costs. As healthcare budgets are not limitless, healthcare payers need to choose cost-effective treatments to achieve the best possible use of budgets. The present study evaluated the long-term cost-effectiveness of IDegLira versus iGlarLixi in Czech patients with T2DM poorly controlled on basal insulin. • A recent network meta-analysis (NMA) comparing IDegLira and iGlarLixi reported reductions in blood sugar levels (measured as glycated hemoglobin) and body weight, as well as lower hypoglycemia rates, for IDegLira relative to iGlarLixi. As no headto-head studies comparing the two FRCs are available, the NMA is the best source to inform long-term modeling. • Relative to iGlarLixi, IDegLira was associated with higher life expectancy and qualityadjusted life expectancy. Over patient lifetimes, the costs of diabetes-related complications were lower in patients treated with IDegLira and partly offset the higher acquisition costs of IDegLira. • In the Czech Republic, IDegLira is a costeffective alternative to iGlarLixi for the treatment of patients with T2DM poorly controlled on basal insulin.
INTRODUCTION
Diabetes is considered a ''global pandemic'' of the twenty-first century and associated with a substantial clinical and economic burden on patients and healthcare systems [1]. In 2017, it was estimated that 863,106 people in the Czech Republic were living with diagnosed diabetes (patients with impaired glucose tolerance not included), of whom 84% had type 2 diabetes mellitus (T2DM) [2]. Good glycemic control is crucial to reduce the incidence of diabetes-related complications and, consequently, the clinical and economic burden associated with diabetes [3,4]. In the Czech Republic, lifestyle changes and metformin therapy are recommended as first-line therapy for patients with T2DM [5]. If glycated hemoglobin (HbA1c) levels do not fall below 7.0% within 6 months while on this therapeutic regimen, then intensification to dual therapy with other non-insulin antidiabetic medications (including glucagon-like peptide-1 [GLP-1] receptor agonists) or insulin is recommended. If the HbA1c target is then not reached within another 6 months, treatment with an intensive insulin regimen or combination therapy (of non-insulin antidiabetic medications) is recommended to achieve a target HbA1c level of 7.0% [5]. In the Czech Republic, however, only about onethird of patients with T2DM achieve an HbA1c target of 7.0%, as recently demonstrated in the DIAINFORM study [6]. No measurable improvements in glycemic control were identified over the 3 years prior to the study, and clinical inertia was considered by the authors to be a likely cause for the lack of progress, leading them to call for the use of novel antidiabetic therapies [6].
Fixed-ratio combinations of GLP-1 receptor agonists plus basal insulin represent such novel treatments for patients with T2DM who fail to achieve adequate glycemic control. These combinations could be attractive treatment options as they combine the complementary effects of their components [7,8]. While basal insulin provides a stable, long-acting reduction in HbA1c levels, GLP-1 receptor agonists stimulate insulin secretion and reduce hepatic glucose production in a glucose-dependent manner, thereby improving glucose control, particularly of post-prandial glucose levels, with a low risk of hypoglycemia [9]. In addition, compared with basal insulins which often lead to gains in body weight, fixed-ratio combinations are generally associated with reduced body weight in populations poorly controlled on basal insulin or oral antidiabetic therapy [7,8].
The first fixed-ratio combination to be introduced to the European market was IDe-gLira (XultophyÒ; Novo Nordisk, Bagsvaerd, Denmark), which is a combination of insulin degludec (IDeg) and liraglutide [7,9]. IDegLira was demonstrated in the Dual Action of Liraglutide and Insulin Degludec in Type 2 Diabetes (DUAL) trial program to be associated with improved glycemic control in a wide range of patients, including patients treated with pioglitazone, sulfonylureas, oral antidiabetics (OAD), GLP-1 receptor agonists and basal insulin [10][11][12][13][14]. Importantly, reductions in HbA1c were associated with reduced glycemic variability and achieved without weight gain and at low risk of hypoglycemia [10,[13][14][15]. More recently, iGlarLixi (SuliquaÒ; Sanofi S.A., Paris, France), a fixed-ratio combination of insulin glargine (IGlar) plus lixisenatide, became available in the Czech Republic. In the LixiLan trial program, iGlarLixi was shown to be associated with larger HbA1c reductions than IGlar in patients with T2DM inadequately controlled on OAD therapy and basal insulin plus metformin, with fewer gastrointestinal side effects than with lixisenatide alone [16,17].
Both IDegLira and iGlarLixi are reimbursed in the Czech Republic for adult patients with T2DM inadequately controlled on basal insulin. Given the differences in the clinical and cost profiles of the two drugs, a cost-effectiveness analysis was conducted to inform resource allocation within budget constraints of healthcare systems [18,19]. Cost-effectiveness analysis and budget impact analysis represent the two key assessments required for a drug to be reimbursed in the Czech Republic. Therefore, the aim of the analysis reported here was to assess the long-term cost-effectiveness of IDegLira versus iGlarLixi for the treatment of patients with T2DM inadequately controlled on basal insulin, from a healthcare payer perspective in the Czech Republic.
Choice of Comparator
Before iGlarLixi became available in the Czech Republic, comparators for IDegLira were limited to basal-bolus insulin regimens and combination therapies, such as GLP-1 receptor agonists plus basal insulin, which have been included in an earlier economic evaluation versus IDegLira [5,20]. Since receiving its marketing authorization in 2017, iGlarLixi has been considered the most relevant comparator for IDegLira. As the two therapies have not previously been assessed with regard to their relative cost-effectiveness in the Czech Republic, iGlarLixi was chosen as the comparator in the present analysis.
In the Czech Republic, iGlarLixi is available in pre-filled pens in two strengths: a pen combining 100 U/mL of IGlar with 33 lg/mL of lixisenatide (providing daily doses of 30-60 dose-steps) and a pen combining 100 U/mL of IGlar with 50 lg/mL of lixisenatide (providing daily doses of 10-40 dose-steps). As acquisition costs differ between pens, both pens were included in the present cost-effectiveness analysis.
Description of the Modeling Approach
The cost-effectiveness of IDegLira versus iGlar-Lixi was evaluated by projecting long-term cost and health outcomes for both treatments. Costs were expressed in monetary units (Czech Koruna [CZK]) and health outcomes in qualityadjusted life-years (QALYs). Cost-effectiveness was expressed as an incremental cost-effectiveness ratio (ICER), i.e. the difference in projected costs divided by the difference in projected health outcomes. The ICER, which was reported as CZK per QALY gained, was compared to a willingness-to-pay (WTP) threshold to assess if IDegLira could be considered to provide good value for money. In the Czech Republic, the commonly accepted WTP threshold is the gross domestic product per capita multiplied by three, as suggested by the World Health Organization [20][21][22][23]. In 2018, this value was CZK 1,200,000 (per QALY gained), which was used in this analysis as the WTP threshold.
Long-term estimates of cost and health outcomes associated with each treatment were obtained from simulations using the IQVIA CORE Diabetes Model (CDM; IQVIA, Basel, Switzerland). The CDM is a web-based, nonproduct-specific, interactive computer model developed to project the long-term health and economic outcomes associated with antidiabetic treatment interventions [24,25]. The model and its validations against real-life data have been described in detail elsewhere [24][25][26]. Briefly, a series of interdependent sub-models is used to simulate background mortality and diabetes-related complications and assess their impact on quality of life and costs over time. The model can, therefore, be used to extrapolate short-term results, such as those from a clinical trial, to long-term outcomes regarding life expectancy, quality-adjusted life expectancy (QALE), cumulative incidence and time to onset of diabetes-related complications, as well as direct medical costs.
Time Horizon, Treatment Duration and Discounting
The base case analysis was performed over patients' lifetime. A long time horizon is recommended to fully capture long-term diabetesrelated complications and their impact on life expectancy, quality of life and costs. Shorter time horizons were explored in sensitivity analyses [27]. The model accounted not only for complication-related mortality but also for Czech Republic-specific background mortality [28].
Patients were assumed to receive IDegLira or iGlarLixi for the first 5 years of the analysis before treatment was intensified to basal-bolus insulin for the remainder of their lifetimes. These assumptions reflect the need for further intensification, required for most patients with T2DM to maintain good long-term glycemic control, and are in line with previous cost-effectiveness analyses of IDegLira in the Czech Republic [20].
As cost and benefits occurring in the future are generally valued differently from cost and benefits occurring in the present, discounting future outcomes is recommended by Czech guidelines for health economic analyses with a time horizon of more than 1 year [21,29]. As specified by guidelines, a discount rate of 3% per annum was applied in the base case to both clinical and cost outcomes. In sensitivity analyses, discount rates of 0 and 5% per annum were used.
Clinical Data: Baseline Characteristics, Treatment Effects and Progression of Physiological Parameters
Baseline characteristics of the simulated patient cohort were sourced from the IDegLira arm of the 26-week, phase 3 DUAL II trial, in line with previous health economic analyses of IDegLira for the Czech Republic (Table 1) [13,20]. The average number of cigarettes smoked per day and mean weekly alcohol intake, which were not reported in the DUAL II trial, were obtained from Czech Republic-specific data, assuming that the pattern observed for the general Czech population was applicable to the simulated cohort [30,31].
Treatment effects for IDegLira were taken from the DUAL II trial (Table 2) [13]. Treatment effects for iGlarLixi were calculated by applying the between-treatment differences (in HbA1c, body weight and daily doses) and rate ratios (in hypoglycemic event rates) obtained from a previously published indirect treatment comparison (ITC) to the treatments effects for IDegLira [32]. For example, an HbA1c reduction of -1.92% was applied for IDegLira based on results from DUAL II. In the ITC, iGlarLixi was shown to be associated with a mean HbA1c that was 0.44% higher than that for IDegLira. Consequently, the HbA1c treatment effect for iGlarLixi was calculated as -1.92% ? 0.44%, yielding an HbA1c reduction of -1.48% applied for iGlarLixi in the present analysis (see Table 2 for detailed calculations on how iGlar-Lixi treatment effects were obtained). The difference in body weight was used to calculate the difference in body mass index (BMI) based on the mean height of patients in the IDegLira arm of DUAL II [13]. This approach to calculate treatment effects was considered the most appropriate approach as no head-to-head study comparing IDegLira and iGlarLixi has yet been conducted. The ITC therefore represents the best currently available evidence for the relative efficacy of the two treatments [33,34].
Treatment effects for HbA1c and BMI were applied in the first year of the analysis, and differences in HbA1c and BMI were maintained during fixed-ratio combination treatment in the first 5 years of the analysis. Upon treatment intensification after 5 years, an HbA1c of 7.0% was assumed in both arms for the remainder of patient lifetimes, in line with glycemic control targets in the Czech Republic [5]. On treatment intensification, BMI was assumed to return to the baseline value. Differences in daily insulin dose or in rates of hypoglycemia were maintained over the first 5 years of the analysis and abolished upon treatment intensification after 5 years. With this approach, clinical differences were maintained only when there was a difference in costs, i.e. during treatment with IDegLira or iGlarLixi.
Resource Use and Cost Data
Costs were estimated from the perspective of the healthcare payer in the Czech Republic and expressed in 2018 CZK. Direct costs were included in the current analysis, capturing pharmacy costs, costs of diabetes-related complications and concomitant patient management costs.
Pharmacy costs were calculated based on resource use data obtained from DUAL II data for IDegLira (45.0 dose-steps per day) [13]. The mean difference in end-of-trial daily insulin doses between IDegLira and iGlarLixi reported by the ITC was applied to the IDegLira dose to obtain the daily iGlarLixi dose (48.6 dose-steps per day) [32]. Daily basal and bolus insulin doses were assumed to be equal following intensification after 5 years of fixed-ratio combination treatment in both arms, and were based on the DUAL VII study [14]. Throughout the analysis, patients were assumed to receive a daily concomitant metformin dose of 2000 mg. During treatment with IDegLira or iGlarLixi, patients were assumed to require one needle Table S1 for cost values and detailed references).
Health-State Utility Data
Diabetes-related complications have been shown to be associated with reductions in health-related quality of life, and these reductions were captured in the analyses. Utilities were sourced from the published literature, including a systematic review of utilities for economic modeling in T2DM and a survey of quality-of-life loss associated with hypoglycemic events (see ESM Table S2) [36][37][38][39].
Sensitivity Analyses
Long-term projections of clinical and cost outcomes, based on short-term data, are associated with uncertainty [40]. Health economic guidance for the Czech Republic recommends addressing this uncertainty through the use of deterministic, one-way sensitivity analyses as well as probabilistic sensitivity analysis (PSA) [21].
In line with this guidance, a range of sensitivity analyses was conducted. The impact of the choice of time horizon was explored by Values are presented as the mean with the SD in parenthesis for (1) and (3), and as the mean with the 95% confidence interval in parenthesis for (2) ITC Indirect treatment comparison a Differences in body weight were converted to differences in BMI based on the mean height (168 cm) of patients receiving IDegLira in DUAL II using shorter time horizons (10 and 20 years). Importantly, as not all modeled patients died over these shorter time horizons, not all complications and costs were captured. The effect of discount rates on projected outcomes was investigated using lower (0%) and higher (5%) discount rates, as suggested by Czech guidance [21]. The influence of treatment effects was explored in several sensitivity analyses, including analyses in which the only between-treatment difference was assumed to be in HbA1c (with all other treatment effects, hypoglycemia rates and daily fixed-ratio combination doses equal to those in the iGlarLixi arm), in which HbA1c progression was assumed to follow the UK Prospective Diabetes Study (UKPDS) HbA1c progression built into the CDM, and in which the upper and lower 95% confidence interval bounds reported by the ITC for the estimated between-treatment difference in HbA1c were used (Table 2) [24,32]. Similarly, the influence of BMI was assessed by running analyses in which the BMI difference between treatments was abolished (with all other treatment effects, hypoglycemia rates and daily doses applied as in the base case), in which the BMI difference between treatments was maintained over patient lifetimes, and in which the upper and lower 95% confidence interval bounds reported by the ITC for estimated between-treatment difference in BMI were used (TableI2) [32]. In addition, sensitivity analyses were conducted in which the difference in hypoglycemic event rates between treatments was set to zero (with all other treatment effects, hypoglycemia rates and daily doses applied as in the base case), and in which only statistically significant differences in HbA1c, BMI and severe hypoglycemic events (SHE) rates were applied while daily insulin doses and non-severe hypoglycemic event rates were set to those of the iGlarLixi arm in both arms. Alternative treatment switching patterns were also explored by assuming earlier (after 3 years) and later (7 years) switches to basalbolus therapy than in the base case (5 years). The impact of the choice of risk equation was investigated by conducting a sensitivity analysis using the UKPDS 82 equation built into the CDM. Use of this risk equation is recommended by the CDM proprietors for sensitivity analysis [26].
The effect of over-or under-estimation of direct costs of diabetes-related complication was explored in two sensitivity analyses. In the first, the cost of treating complications was increased by 10%; in the second, the cost was decreased by 10%. Disutilities were also varied for hypoglycemic events and BMI [41,42]. In addition, a sensitivity analysis was conducted assuming diminishing disutilities for non-severe hypoglycemic events [43].
The PSA was performed by using the functionality provided by the CDM, which samples complication costs, treatment effects and cohort characteristics from distributions and feeds the sampled values to second-order Monte Carlo simulations.
Statement of Ethics Compliance
This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.
Base Case Analysis
Long-term projections suggested that in comparison to iGlarLixi, IDegLira was associated with clinical benefits (Table 3). Compared to iGlarLixi, IDegLira was associated with gains in discounted life expectancy of 0.11 years and in discounted QALE of 0.14 QALYs. The clinical benefits of treatment with IDegLira relative to iGlarLixi resulted from a reduced incidence and delayed onset of diabetes-related complications (Fig. 1).
The clinical benefits associated with IDegLira came at an increased cost. Relative to the iGlarLixi pen containing 33 lg/mL of lixisenatide, lifetime direct medical costs for patients treated with IDegLira were approximately CZK 94,029 higher per patient ( Table 3). The corresponding incremental cost relative to the iGlarLixi pen containing 50 lg/mL of lixisenatide was CZK 47,058. The higher incremental costs for IDegLira were due to higher IDegLira acquisition costs over the first 5 years of the analysis, which, however, were partially offset by reduced costs for the treatment of diabetes-related complications. In particular, Estimation of the long-term clinical outcomes indicated that both life expectancy and QALE were improved with IDegLira treatment compared with iGlarLixi treatment, at an increased cost from a healthcare payer perspective. IDegLira was associated with an ICER of CZK 695,998 per QALY gained versus the iGlarLixi pen containing 33 lg/mL of lixisenatide and of CZK 348,323 per QALY gained versus the iGlarLixi pen containing 50 lg/mL of lixisenatide ( Table 3). As these ICERs fell below the commonly used WTP threshold, IDegLira is likely to be considered cost-effective versus iGlarLixi.
Deterministic Sensitivity Analyses
Deterministic sensitivity analyses demonstrated that the projected cost-effectiveness of IDegLira relative to iGlarLixi was generally not sensitive to changes in input data or assumptions ( Fig. 2; ESM Tables S3 and S4).
Shorter simulated time horizons were associated with increases in ICER relative to the base case. When a shorter time horizon was used, the incremental clinical benefit of IDegLira was smaller as long-term benefits were not captured in full, indicating that a lifetime perspective is indeed appropriate, as recommended by health economic guidance. Still, most ICERs calculated for shortened time horizons fell below the WTP threshold, suggesting that IDegLira was costeffective versus iGlarLixi even when some of the clinical benefits of IDegLira were not accounted for. When no discounting was applied, the ICER decreased, while the opposite effect was observed for a discount rate of 5% per annum.
A difference in cost-effectiveness outcomes was observed when treatments were assumed to differ in HbA1c only, i.e. if the benefits of IDe-gLira on BMI, hypoglycemia rate and daily doses were not accounted for. In this analysis, ICERs increased relative to the base case, suggesting that the non-HbA1c benefits of IDegLira are also important drivers of clinical and cost outcomes.
Abolishing the difference in BMI between treatments increased the ICER while maintaining the difference over patient lifetimes decreased the ICER. Using the lower 95% confidence interval bound for the treatment effect on weight reduced the clinical benefit associated with IDegLira over iGlarLixi, thereby increasing the ICER relative to the base case as incremental costs remained unchanged (although the ICER still fell below the WTP threshold). Conversely, using the upper bound of the 95% confidence interval for the treatment effect on weight increased the clinical benefit associated with IDegLira and, given unchanged costs, reduced the ICER relative to the base case.
Similarly, when the difference in hypoglycemic event rates was abolished, the ICER increased relative to the base case due to the reduced incremental clinical benefit of IDegLira and slight increase in incremental costs. The same pattern was observed when only statistically significant differences in treatment effects were considered. Assumptions regarding the timing of intensification to basal-bolus insulin also did not affect conclusions regarding costeffectiveness. Treatment switching after 3 years led to a small increase in the ICER as the reduction in incremental clinical benefits associated with IDegLira was slightly larger than the reduction in incremental acquisition costs. When intensification occurred after 7 years of fixed-ratio combination treatment, a small increase in the ICER was also observed due to the fact that the increased clinical gains associated with IDegLira were offset by higher acquisition costs.
Higher complication costs were associated with a reduced ICER relative to the base case, while the converse was observed for lower complication costs. Use of the UKPDS 82 risk equations for the prediction of cardiovascular events was associated with an increase in ICERs, but IDegLira was still considered to be cost-effective versus both iGlarLixi pens. The use of an alternative disutility associated with BMI and hypoglycemia, as well as assuming a diminishing disutilities approach for hypoglycemic events, did not have a large impact on the ICER, indicating that cost-effectiveness outcomes were not driven by these assumptions.
Probabilistic Sensitivity Analysis
The PSA conducted with sampling around cohort characteristics, treatment effects, complication costs and utilities produced mean results similar to those for the base case. For the comparison of IDegLira with the iGlarLixi pen containing 33 lg/mL of lixisenatide, the mean improvement in QALE gained with IDegLira versus iGlarLixi was 0.12 QALYs, at mean incremental costs of CZK 99,703, giving an ICER of CZK 843,898 per QALY gained. The corresponding values for the comparison with the iGlarLixi pen containing 50 lg/mL of lixisenatide were 0.12 QALYs at incremental costs of CZK 50,716, yielding an ICER of CZK 440,739 per QALY gained.
Cost-effectiveness scatterplots, which were based on 1000 simulated cohorts of 1000 patients, showed that most of the sampled ICER fell in the upper right quadrant (74 and 66% for the comparison of IDegLira with the iGlarLixi pen containing 33 and 50 lg/mL of lixisenatide, respectively), indicating that IDegLira was associated with increased effectiveness and costs (Fig. 3). At a WTP threshold of CZK 1,200,000 per QALY gained, there was a 59% probability that IDegLira was cost-effective versus the iGlarLixi pen containing 33 lg/mL of lixisenatide, and a 68% probability that IDe-gLira was cost-effective versus the iGlarLixi pen containing 50 lg/mL of lixisenatide.
DISCUSSION
The present analysis assessed the cost-effectiveness of IDegLira versus iGlarLixi in the Czech Republic for patients with T2DM who were inadequately controlled on basal insulin. Based on clinical data from the DUAL II trial and an ITC comparing IDegLira with iGlarLixi, outcomes were projected over patient lifetimes using a validated health economic model [13,24,32]. At a WTP threshold of CZK 1,200,000 per QALY gained, IDegLira was found to be cost-effective versus both iGlarLixi pens currently available (containing 33 and 50 lg/ mL of lixisenatide, respectively). Improvements in HbA1c and BMI in patients treated with IDegLira led to reductions in cumulative incidence and delayed onset of diabetes-related complications, resulting in increased life expectancy and QALE. IDegLira was associated with higher treatment costs than iGlarLixi, a reflection of the former's higher acquisition costs; however, these were partially offset by lower costs of treating diabetes-related complications, due to complications avoided or delayed in patients treated with IDegLira. Over patient lifetimes, IDegLira was associated with a gain of 0.14 QALYs and incremental costs of CZK 94,029 and CZK 47,058 versus iGlarLixi pens containing 33 and 50 lg/mL of lixisenatide, respectively, leading to ICER of CZK 695,998 and CZK 348,323 per QALY gained. The sensitivity analyses suggested that the estimated treatment difference for HbA1c as well as differences in hypoglycemia rates were key drivers of the results. In addition, the sensitivity analyses demonstrated that IDegLira was associated with benefits to patients over their entire lifetime, so long-term analyses were required to fully capture the impact of IDegLira.
While no previous health economic evaluation of IDegLira versus iGlarLixi is available, the results of the present analysis analysis are broadly aligned with those obtained in earlier studies that showed IDegLira to be cost-effective versus its competitors in a range of settings. In a recent CEA conducted for the Czech setting, IDegLira was compared with insulin intensification regimens in patients with T2DM inadequately controlled on basal insulin [20]. The authors of this study considered IDegLira to be cost-effective versus both basal-bolus therapy and basal insulin plus GLP-1 receptor agonist therapy from the perspective of the healthcare payer in the Czech Republic. Similar long-term analyses have been conducted in the UK, the Netherlands, Sweden and Slovakia, all of which showed IDegLira to be cost-effective versus basal-bolus insulin and/or basal insulin plus GLP-1 receptor agonists from the perspective of the respective healthcare payer [44][45][46][47]. For the UK, a recent short-term CEA also showed that IDegLira was likely to be cost-effective even over 1 year relative to treatment with basal-bolus insulin therapy [48].
To date, no head-to-head clinical trials comparing IDegLira with iGlarLixi have been conducted. Therefore, data from a previously published ITC were used to compare the two fixed-ratio combinations, which may be considered a limitation of the present analysis [37]. Evidence synthesis methods, such as indirect comparison or meta-analysis, have become widely used and accepted in health technology assessments and health economic modeling, particularly to provide healthcare payers and decision-makers with robust evidence even if no head-to-head clinical data are available [49,50]. In addition, the clinical data used in the present analysis were sourced from a consistent source, namely the DUAL clinical trial program investigating IDegLira. An additional limitation of the analysis was the reliance on relatively shortterm clinical trial data (pooled in the ITC) as the base for long-term projections of clinical and economic outcomes, which were therefore associated with uncertainty. Despite this limitation, which is common to many health economic analyses, projections of outcomes over patient lifetimes are recommended in guidelines for the economic evaluation of interventions for patients with diabetes as a means to inform decision-making even if long-term clinical data are not available [5,27]. In the present analysis, every effort was made to reduce the uncertainty surrounding the long-term outcomes by using a diabetes model which has been extensively published and validated against real-life data [24][25][26]. In addition, extensive sensitivity analyses were conducted, and these consistently showed IDegLira to be cost-effective relative to iGlarLixi.
Clinical trials have consistently demonstrated improved short-term clinical outcomes associated with IDegLira in patients with T2DM who failed to achieve glycemic control on basal insulin alone [13,14]. The present analysis indicated that shortterm outcomes translated into long-term benefits. Importantly, outcomes from clinical trials have been confirmed by real-world studies. In a chart review of more than 600 European patients with T2DM who initiated IDegLira at least 6 months before the review was performed, IDe-gLira was associated with a 0.9% reduction in HbA1c at 6 months after treatment initiation relative to baseline [51]. Similarly, reductions have been observed for weight (0.7 kg at 6 months vs. baseline) and for use of concomitant antidiabetic medications. These findings were confirmed by an observational study of patients with T2DM in Switzerland, in whom reductions in HbA1c and BMI were observed over a 6-month period following initiation of IDegLira therapy; in addition, required insulin doses were reduced, without SHE over the course of the 6 months [52]. The benefits associated with IDegLira are reflected in physicians' perceptions, as reported by a survey of 235 primary and secondary care physicians from Europe [53]. Based on their experience in clinical practice, physicians reported being more satisfied with prescribing IDegLira than basal-bolus insulin to achieve important treatment targets, including avoidance of weight gain and hypoglycemia, while also considering IDegLira to be simpler to administer [53]. Importantly, the simplicity of IDegLira therapy, which is matched by reports of higher patient satisfaction and easier training of patients, has been suggested as an important component of approaches to reduce treatment intensification inertia [54].
CONCLUSION
From the perspective of the Czech healthcare payer, IDegLira is likely to be cost-effective versus iGlarLixi in the treatment of patients with T2DM who are uncontrolled on basal insulin. IDegLira was associated with a low risk of hypoglycemia and reduced cumulative incidence as well as delayed onset of diabetes-related complications, which improved quality of life and reduced the costs of treating diabetesrelated complications. IDegLira is therefore a valuable option for intensification of antidiabetic therapy in the Czech Republic.
ACKNOWLEDGEMENTS
Funding. This study and any processing charges related to the submission of the manuscript were funded by Novo Nordisk. All authors had full access to all of the data in this study and take complete responsibility for the integrity of the data and accuracy of the data analysis.
Authorship. All named authors meet the International Committee of Medical Journal Editors (ICMJE) criteria for authorship for this article, take responsibility for the integrity of the work as a whole, and have given their approval for this version to be published.
Disclosures. Johannes Pöhlmann is an employee of Ossian Health Economics and Communications, which received consulting fees from Novo Nordisk for conducting the analysis and preparing the present article. Monika Russel-Szymczyk is an employee of Novo Nordisk Pharma Sp. z.o.o. Pavel Holik is an employee of Novo Nordisk s.r.o. Karel Rychna is an employee of Novo Nordisk s.r.o. Barnaby Hunt is an employee of Ossian Health Economics and Communications, which received consulting fees from Novo Nordisk for conducting the analysis and preparing the present article.
Compliance with Ethics Guidelines. This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.
Data Availability. The datasets used and/or analyzed during the current study, in addition to those made available in the Electronic Supplementary Material, are available from the corresponding author on reasonable request.
Open Access. This article is distributed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/ by-nc/4.0/), which permits any noncommercial use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. | 2019-03-08T14:03:49.259Z | 2019-01-31T00:00:00.000 | {
"year": 2019,
"sha1": "99ed555d6fe09220ce92f42483539f66ce53caa2",
"oa_license": "CCBYNC",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s13300-019-0569-7.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "99ed555d6fe09220ce92f42483539f66ce53caa2",
"s2fieldsofstudy": [
"Medicine",
"Economics"
],
"extfieldsofstudy": [
"Medicine"
]
} |
7080924 | pes2o/s2orc | v3-fos-license | Structural Characterization of the Degradation Products of a Minor Natural Sweet Diterpene Glycoside Rebaudioside M under Acidic Conditions
Degradation of rebaudioside M, a minor sweet component of Stevia rebaudiana Bertoni, under conditions that simulated extreme pH and temperature conditions has been studied. Thus, rebaudioside M was treated with 0.1 M phosphoric acid solution (pH 2.0) and 80 °C temperature for 24 h. Experimental results indicated that rebaudioside M under low pH and higher temperature yielded three minor degradation compounds, whose structural characterization was performed on the basis of 1D (1H-, 13C-) & 2D (COSY, HSQC, HMBC) NMR, HRMS, MS/MS spectral data as well as enzymatic and acid hydrolysis studies.
Introduction
Recently, many soft drink manufacturers have driven their focus towards naturally occurring high-potency sweeteners to reduce calories by introduction of non-caloric sweeteners into their OPEN ACCESS beverage systems. Stevia rebaudiana (Bertoni), a perennial shrub of the Asteraceae (Compositae) family native to certain regions of South America (Paraguay and Brazil) [1,2] is one such example in recent years which resulted in the isolation of several potently sweet diterpenoid glycosides namely rebaudiosides A and D, stevioside, and dulcoside A; also known as stevia sweeteners of which stevioside and rebaudioside-A are the major compounds. These compounds are all glycosides of the diterpene ent-13-hydroxykaur-16-en-19-oic acid known as steviol [3,4]. Recently we have reported the isolation and sensory evaluation of rebaudioside M (1, also known as rebaudioside X), a minor constituent from S. rebaudiana Bertoni, which is about 160-500 times sweeter than sucrose [5]. Due to continuing demand of minor steviol glycosides like rebaudioside D and M, there has been intense interest on S. rebaudiana and that is why it is grown commercially in a number of countries, particularly in Japan, Taiwan, Korea, Thailand, Indonesia and China.
As a part of our continuing research to discover natural sweeteners, we have recently isolated several novel diterpene glycosides from the commercial extracts of the leaves of S. rebaudiana obtained from various suppliers around the world [6][7][8][9]. Apart from isolating novel compounds from S. rebaudiana and utilizing them as possible natural sweeteners or sweetness enhancers, we are also engaged in synthesis of various novel steviol glycosides and understanding their physicochemical profiles as well as their stability in various systems of interest [10][11][12][13][14]. We are also engaged in studying the stability data of various major steviol glycosides like rebaudioside A and stevioside under fluorescent and acidic conditions by isolating and characterizing their degradation products using various spectroscopic and chemical studies [15][16][17]. In continuation of our stability studies, we are reporting the isolation and characterization of the major degradation products of the minor steviol glycoside from S. rebaudiana namely rebaudioside M under acid conditions at higher temperature.
Results and Discussion
The structures of the degradation compounds 2-4 ( Figure 1) identified during the course of this study were characterized on the basis of extensive spectroscopic data ( 1 H-& 13 C-NMR, COSY, HSQC, HMBC, MS, MS/MS) and hydrolysis studies.
Compound 2 was obtained as a white powder and its molecular formula was assigned as C 56 H 90 O 33 from its HRMS, which showed [M + H] + and [M + Na] + ions at m/z 1291.5439 and 1313.5254, respectively; this was supported by the 13 C-NMR spectral data. The 1 H-NMR spectrum of 1 showed the presence of three methyl singlets at δ 1.33, 1.35 and 1.89, eight methylene and two methine protons between δ 0.77-2.69, and a trisubstituted olefinic proton at δ 5.03, suggesting the presence of an ent-13-hydroxykaur-15-en-19-oic acid skeleton in its structure [7]. The presence of 15-ene ent-kaurane diterpenoid skeleton in 2 was supported by COSY (H-1/H-2; H-2/H-3; H-5/H-6; H-6/H-7; H-9/H-11; H-11/H-12) and HMBC (H-1/C-2, C-10; H-3/C-1, C-2, C-4, C-5, C-18, C-19; H-5/C-4, C-6, C-7, C-9, C-10, C-18, C-19, C-20; H-9/C-8, C-10, C-11, C-12; H-14/C-8, C-9, C-13, C-15, C-16, H-15/C-8, C-14, C-16, C-17, and H-17/C-13, C-15, C-16) correlations. In addition, the 1 H-NMR spectrum of 1 also showed the presence of six sugar units in its structure by exhibiting the anomeric protons at δ 5.33, 5.44, 5.45, 5.47, 5.81, and 6.33. The presence of six sugars was confirmed as hexoses by the fragment ions corresponding to the successive loss of six hexose moieties from its [M + H] + ion. Acid hydrolysis of 2 afforded D-glucose which was identified by preparing the corresponding thiocarbamoyl-thiazolidine carboxylate derivatives and in comparison of its retention times with the standard sugars as described in the literature [18]. The 1 H-and 13 C-NMR values for all the protons and carbons in 2 were assigned on the basis of COSY, HSQC and HMBC correlations and are given in Table 1. The molecular formula of compound 3 was determined to be C 56 H 92 O 34 by the HRMS data that showed (M + H) + and (M + Na) + ions at m/z 1309.5588 and 1331.5414, respectively. The 1 H-NMR spectrum of 3 showed the presence of three methyl singlets at δ 1.28, 1.31, 1.32, nine methylene and two methine protons between δ 0.78-2.67. In the absence of any unsaturated protons or carbons together with the appearance of a methyl group at δ 1.32 corresponds to a methyl group connected to a tertiary hydroxyl group suggested the structure of 3 should be similar to ent-13,16β-dihydroxykauran-19-oic acid [11]. The presence of tertiary hydroxyl at C-16 position in 3 was supported by the carbon signal appeared at δ C 77.1 in its 13 .3707, respectively in the positive ESI mass spectrum. This was also supported by the 13 C-NMR spectral data. The 1 H-NMR spectrum of 4 showed the presence of three methyl singlets at δ 0.81, 0.98 and 1.42; nine methylene and two methine protons. Enzymatic hydrolysis of 3 furnished an aglycon that was found to be identical to isosteviol on the basis of its NMR spectral data reported in the literature [19][20][21]. The presence of the isosteviol skeleton
Reagents and Chemicals
Ammonium hydroxide (NH 4 OH) was from Fluka (a part of Sigma-Aldrich, Bellefonte, PA, USA), and 85% phosphoric acid (H 3 PO 4 ) was from Fisher Scientific (Pittsburgh, PA, USA), all of which were reagent grade. HPLC grade acetonitrile (MeCN) was purchased from Burdick & Jackson (Muskegon, MI, USA). Water was purified using a Millipore system (Billerica, MA, USA).
HPLC Conditions
An Agilent (Wilmington, DE, USA) 1200 HPLC equipped with a quaternary pump, a temperature controlled column compartment with additional 6-port switching valve, an auto sampler and a UV absorbance detector, was used for the analysis. A Charged Aerosol Detector (CAD), ESA, Inc. (Chelmsford, MA, USA), was also used for the analysis. The scale on the CAD was 100 pA and the filter was set to medium. The switching valve diverted the first 5.5 min of each injection away from the CAD detector to prevent fouling of the detector. The system was controlled using Waters (Milford, MA, USA) Empower software. The separation HPLC column was maintained at a temperature of 25 °C with a flow rate of 5.0 mL/min. The RP-HPLC employed on a Gemini C 18 column (250 × 10 mm, 5 µm) (Torrance, CA, USA) with a Gemini Security guard C 18 cartridge. A binary solvent mobile phase as shown in Table 2 was used for detection and isolation of the three compounds 2-4. The injection volume of each sample was 150 µL, which were kept at ambient temperature while in the auto sampler. In all cases for UV detection, a 4 nm bandwidth was used with a reference wavelength of 210 nm (100 nm band width).
General Instrumentation
NMR spectra were acquired on Bruker Avance DRX 500 MHz instrument (Emory Univeristy, Atlanta, GA, USA) with a 5 mm inverse detection probe using standard pulse sequences. The NMR spectrum was referenced to the residual solvent signal (δ H 8.71, δ C 149.9 for pyridine-d 5 ), chemical shifts are given in δ (ppm), and coupling constants are reported in Hz. MS and MS/MS data were generated with a mass spectrometer (AMRI, Albany, NY, USA) made by Waters Premier Quadrupole Time-of-Flight (Q-Tof) equipped with an electrospray ionization source operated in the positive-ion mode and Thermo Fisher Discovery OrbiTrap in the positive Positive Mode Electrospray. Samples were diluted with water:acetonitrile (1:1) containing 0.1% formic acid and introduced via infusion using the onboard syringe pump.
Degradation of Rebaudioside M (1)
A 0.1 M phosphoric acid solution was made and adjusted to pH 2.0 with concentrated ammonium hydroxide. Ten milligram of 1 was added to 10 mL of phosphoric acid solution. The solution was placed on a heat block at 80 °C for 24 h.
Isolation of Degradation Compounds 2-4
HPLC purification was performed using the method described in Table 2 and the peaks eluting at retention times 7.86, 13.19, and 23.98 min were collected over several injections and dried by rotary evaporation under reduced pressure to yield compounds 2, 3 and 4 respectively. Samples of the compounds 1-4 are available from the authors.
Enzymatic Hydrolysis of 4
Compound 4 (250 μg) was dissolved in 2.5 mL of 0.1 M sodium acetate buffer, pH 4.5 and crude pectinase from Aspergillus niger (50 µL, Sigma-Aldrich, P2736) was added. The mixture was stirred at 50 °C for 48 h. The product precipitated out during the reaction and was filtered and then crystallized. The resulting product obtained from the hydrolysis was identified as isosteviol, characterized by comparison of its co-TLC with standard compound and 1 H-NMR spectral data [19][20][21].
Conclusions
The degradation of rebaudioside M (1) under acidic conditions has been studied at high temperature. The complete 1 H-and 13 C-NMR spectral data for the degradation compounds 2-4 are reported herewith for the first time based on COSY, HSQC, and HMBC spectroscopic data as well as enzymatic and acid hydrolysis studies. | 2016-03-01T03:19:46.873Z | 2014-01-01T00:00:00.000 | {
"year": 2014,
"sha1": "3dbf3fbdd21a78f9ad9612db83cbfea00b8ee853",
"oa_license": "CCBY",
"oa_url": "http://www.mdpi.com/1422-0067/15/1/1014/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3dbf3fbdd21a78f9ad9612db83cbfea00b8ee853",
"s2fieldsofstudy": [
"Agricultural And Food Sciences"
],
"extfieldsofstudy": [
"Chemistry",
"Medicine"
]
} |
1455390 | pes2o/s2orc | v3-fos-license | Monitoring the Turmeric Finger Disease and Growth Characteristics Using Sensor Based Embedded System — A Novel Method
In this cyber era, novelty plays a prime role in the field of agriculture that majorly depends on computer-based measurements and control. Herein before, it was totally controlled and performed by the agriculturists. One of the technological innovative methods to measure and monitor the turmeric finger growth characteristics is the embedded system that is based on sensor array module such as flex sensor, temperature sensor and pH sensor. The experimental work has been designed and tested with five set of nodes and the growth of turmeric finger is tenuously monitored by measuring the change in flex resistance. Out of five nodes, two nodes were diseased. Deliberately, one node was left as such and the other node was treated with natural pesticides (pseudomonas and viride) to restrict the rhizome rot disease attack. After cultivation, it was found that the rhizome rot disease attack on the node which was treated with pesticides was comparatively lesser than the other node. The five different nodes have been used in the experimental work with an average flex sensor resistance of 3.962 cm/kΩ. In a nutshell, this proposed method manifests the farmers to detect the rhizome rot disease at its earlier stage and to prevent it as well by screening the growth of the turmeric fingers when it is under the soil.
Introduction
Agriculture is acquiring more importance in the ongoing modern era with the incorporation of computer integrated technology and application of advanced control systems [1].Nowadays, the agribusiness increasingly banks on automation using computer based systems and usage of robotics for replacing the activities performed by man power, usually more superior in performance to human [2].There is a huge demand in the production of machines required for agricultural land by integrating computer, electromechanical and information systems due to the increasing complexity in performing farming [3].
The need for the new methodologies to meet the challenges of intricacy prevailing in the agricultural scenario has also led to the emergence of precision agriculture.Precision agriculture is very much useful in improving the efficiency of performance of agricultural land and also increases the quality and reliability in spite of the ruggedness in the environment [4].The effect over the growth of crops due to the presence of obstacles in the internode's communication is compared with the performance of sensor node in the greenhouse environment running with low power operation by Hyun-Joong Kang [5].The existing wireless sensor networks like ZigBee, Bluetooth and WiFi are operated with 2.4 GHz Industrial Specific and Medical (ISM) band which are applicable for modern precision agriculture [6] [7].All these wireless networks provide license free operation, huge spectrum allocation and worldwide compatibility [8].Miranda [9] evaluated the irrigation amount based on distributed soil water measurements using closed loop systems.
"Indian Saffron" as it is called, the turmeric with a long past of 4000 years of medicinal usage and also as an integral part of drugs, cosmetics and dye industries.Nitrogen, phosphorus, potassium forms the vital nutrient requirement whereas the calcium, magnesium and Sulphur are the follow-on nutrients.The root characteristic is estimated using various sensors like flex sensor, temperature and pH sensors.The fingers of turmeric get affected by fungus during 4 th month of its growth and the rhizome rot disease is unpredictable because it is underground.The above mentioned rhizome rot diseases is noticed when it spreads out as "Leaf Spot" [10].As this ruins the entire plant life, the flex sensor is used for continues monitoring and identified at earlier stage itself.After identification is done, then precautions can be made by applying pseudomonas and viride [11] [12] that serves as the best natural bio-pesticides.
In this research, an innovative method of investigating the turmeric finger growth with flex sensors using embedded system and the definite advantage of detection of rhizome rot disease in the earlier stage itself is proposed.The various sensors used are LM35 for temperature measurement and PHE-45P is used to measure soil pH range.The sensor array collects the data of n-number of nodes and transmits it individually using Zigbee module [13] that are collectively received and monitored using coordinating module and LCD display [14].The GSM (Global System for Mobile Communications) is used to transmit the data to a remote monitoring location.The Zigbee module operates with a voltage and frequency of 3.3 V and 2.4 GHz respectively covering 25 meters [15] [16] with an advantage of affordability and less power consumption than the other communication protocol (like RF, Bluetooth etc.) [15].Number of sensor nodes depends on area of the field.One node covers 20 m 2 area accordingly it is required to put number of nodes.To test this experimental setup 0.25 hectare of land was chosen.A module with five designed nodes had been put into practice without a specific location in the field to notify the readings from five nodes.Each field data measurement module (FDMM) comprised of a) temperature, b) pH and c) five flex sensors.The setup was kept at random in order to predict whether the turmeric finger was affected by the disease.Electrical signal outputs from temperature, pH and flex sensors were given to the signal conditioning unit so as to strengthen and sent to microcontroller's internal Analog to Digital Converter (ADC) module.The converted digital values were then processed into temperature value, pH value and flex resistance by the software in the microcontroller (LPC2148).The microcontroller was programmed to store values taken at three time intervals-8:00 am, 12:30 pm and 4:00 pm on a regular basis.To minimize the power consumption, microcontroller's internal real time clock was programmed to find out and store the readings for about 3 days and the data were transferred to the centre co-ordinating node via ZigBee transmitter.The obtained data were sequentially transferred in the form of message to computer which was at remote location via GSM technology [17].As and when the computer is ON, all the data which are transferred earlier will be updated in the system.All the nodes were operated with 9 V/15Ah battery and had been charged once in 40 days except computer.
Flex Sensor
The flex sensor was used to assess the angle displacement of the crop and it is capable of bending physically.Robotics, Gaming (Virtual Motion), Medical Devices, Computer Peripherals, Musical Instruments Physical Therapy, Simple Construction and Low Profile are some of the other real time application of flex sensors.The array of five flex sensors were fixed in the specified position around the plant for observing the finger growth.
The four sensors F1, F2, F3, and F4 had been fixed on four sides (around) of the plant.Whereas F5 was positioned at the tip of the plant to screen the descendent turmeric finger growth and it showed the maximum growth rate rather than the other sensors.The electrical resistance of flex sensor is 10 kΩ and the value varied pertinent to the force applied.The flex sensor has a power rating of about 0.5 Watts.
The flex sensor and its signal conditioning unit are shown in Figure 2 and Figure 3 respectively.The input impedance R in tends to infinity and the feedback impedance R f decreases to zero in case of non-inverting configuration.A feedback of 100% is given by directly connecting the output to the non-inverting terminal and V in is approximately equal to V 2 due to a fixed unity gain amplifier.The non-inverting input voltage V in and the amplifier gain is specified as follows (Equation (1) to Equation (3)): R 1 is the flex sensor resistance and varies accordingly to its bend.The error due to the source impedance of flex sensor acting as a voltage divider is reduced by decreasing the bias current and then the operational amplifier can be used in the basic flex sensor circuit as an impedance buffer (Q1).Output from the operational amplifier is fed to the microcontroller 10 bit ADC.
( ) ( )
Step size of the ADC 1 LSB 1024 where The variance of resistance measurement was between 10.1 kΩ and 13.8 kΩ.Hence output voltage from the impedance buffer was between 2.1008 V (for 13.8 kΩ) and 2.4876 V (for 10.1 kΩ) respecting the Equation ( 2).Due to the small variations in the voltage, it was required to increase the range of voltage by using differential amplifier.Voltage to the ADC circuit was computed by the Equation (4) ( ) where According to the Equation ( 4), output voltage from differential amplifier was between 0.7984 V and 0.0248 V. Since the flex sensor is more flexible [18], it was chosen to trace the growth of the turmeric finger.By the protrusion of the finger, the flex sensor had bent accordingly there was a change in resistance.To screen the finger growth, flex sensors were placed around the plant at the stage of planting itself.
Temperature Sensor
One of the precision integrated circuit temperature sensors are known as LM35 series.It is majorly used for remote applications and the output which is directly proportional to the centigrade [3].For its operation, the LM35 series can use either a single or dual power supply.From the above supply provided it draws only 60 μA and has low self-heating of 0.1˚C in still air.In the LM35 series, the rated temperature range of LM35 is −55˚C to 150˚C while LM35c provides improved accuracy of −10˚C with a range of −40˚C to +110˚C.
The temperature measurement circuit diagram is shown in Figure 4.According to the Equation (3) the ADC resolution was 0.977 mV.Temperature varies between 17.1˚C and 21.8˚C and the corresponding output voltage from the sensor module was between 171 mV and 218 mV.The temperature and pH sensors were used to determine the soil characteristics.In view of the fact that the temperature and the pH parameters vary during rainfall and fertilizer application, it's mandatory to examine the soil characteristics to enhance the plant growth.The performance of linear circuits connected with wires in an aggressive environment is affected due to intense electromagnetic radiation from transmitters, wires acting as receiving antennas and internal junctions forming rectifiers.The above mentioned problem can be eliminated using a bypass capacitor from V in to ground (C2-0.01μF) and a series R-C damper, such as 75 Ω (R1), in series with 1 μF/10V (C1) from output to ground.Out of the three (flex sensor, temperature sensor and pH sensor) measurements, flex sensor was given much importance than the other two whose measurement acts as an effective technique for detecting the finger growth rhizome rot disease in turmeric plants at an early stage.
pH Sensor
The PHE-45P sensor measures the pH value of aqueous solution used in the industrial and municipal process application.PHE-45P electrode is made of glass and the PHE-45P sensor's glass electrode must be ensured to be always wet for its proper functionality.These sensors are enabling immediate usage due to the fluid filled cap over it.In order to restore functionality, the electrode should be hydrated for 24 hours if it is dried and must be mounted vertically (electrode facing down) possible and also after purchasing the sensor.While mounting the angle of sensor must be at least 10˚ above horizontal.The life time of the sensor is maximised using a high volume dual junction salt-bridge and the chance of fouling is minimised using the annular junction which provides a large surface area.Further the contamination of the reference solution is diminished using the large electrolyte volume and dual reference junctions.The second glass pH electrode immersed in a reference solution acts as the reference element and this reference system importantly increases the range of sensor applications.
The integrated preamplifier is attached with the sensor.The low impedance signal is created from the amplifier to fetch the stable readings in noisy environment.The electrode breakage, loss of sensor seal integrity or integral temperature element failure warning can be given to the user using the system diagnostics in the sensor.The temperature sensor used in PHE-45P is called Pt1000 RTD which is used for obtaining highly linear and accurate output.
To test the performance of the system (mainly on flex sensor), an artificial method of turmeric finger rhizome rot disease was caused by stacking the water for more than two days to some of the nodes.
Centre Data Processing Module [CDPM]
The CDPM module consists of one central co-ordinating node that collects the information on the five different nodes via Zigbee communication.On interrupt basis (periodic updates based on RTC), out of the five nodes, the CDPM module receives data and that has been communicated to DLAM module via GSM.The product of Zigbee is XBEE module which is primarily used for fulfilling the IEEE 802.15.4 standards and requirements of the wireless sensor networks such as low cost, low power [19].The XBEE module is widely used for remote applications since the cost of the XBEE module is low [18].The module working within the ISM 2.4 GHz frequency band requires a minimal amount of power to exchange data between devices.In addition to the above, XBEE module provides pin-for-pin compatibility with each other.
Data Logging and Analysis Module [DLAM]
The DLAM module consists of one GSM receiver and one personal computer.GSM module receives the data and that has been converted to RS232 voltage (TTL logic to RS232 logic) by MAX232 IC.On both transmitter and receiver part SIM900 module has been used.After getting the data, data are being stored in the computer by using visual basic (VB) software for continuous monitoring of the parameter deviation.
Results and Discussion
A module has been designed with array sensors such as flex sensor, Temperature sensor and pH sensor which transmit data via the ZigBee module.Figure 5 and Figure 6 show the hardware and the implementation in the field.
Soil pH Readings
The moisture content of the soil was measured on watering and raining using pH readings as shown in the graph and the five different nodes 1 through node 5 were placed in different places under the soil.Figure 7(a) and Figure 7(b) graphs show the overall soil pH readings of five different sensor nodes.The pH value actually gets decreased after the occurrence of a rain or increased when fertilizers are applied.Both well and bore well are used to irrigate the field.The last two months (during the time of research) the water was highly scarce which made the farmers to use bore well increasing the pH value of the field and it is indicated using last ten set of reading.This led to the decrease in the chlorophyll content in large amount and the shade of the leaves turned to yellow.Chlorophyll is a green photosynthetic pigment playing a vital role in mounting the crop yield.The aforesaid fact had been discussed with the personnel of Coimbatore Agricultural University (Government of Ta-milNadu) and it was suggested that the supply of water during high pH value (above 6.3) can be considerably reduced and after watering once, the Nitrogen (N) fertilizer could be used in the range between 10 kg and 25 kg per hectare depending on the pH value to compensate the change in pH.
Soil Temperature Readings
During rain, the inference was low in temperature and the awareness about the basic growing conditions of turmeric, the crop becomes strong and vibrant based on soil, water and atmospheric temperature.The proposed system was implemented in the month of May because turmeric is cultivated between the month of May and June.The temperature remains very high and also increases the soil temperature.All the nodes were watered during evening (between 5 pm and 7 pm) except the node 2 and 5 which were watered during day time (between 11 am and 2 pm).From Figure 8 graphs, some of the readings directly reflected the high temperature because of water supplied at noon.Thus the perfect timing for watering is early mornings or late evenings where the temperature is maintained between 17˚C and 18˚C to increase the yield.
Flex Sensor Readings
In an array of flex sensors N1 through N5 represents the individual node that consists of 5 sensors ( 5
Node 1
Figure 10 graph shows the growth rate measurement based on readings taken at regular interval of the turmeric finger (node 1). Figure 11 shows the output image of the turmeric finger (node 1).
Growth Result Table 1 indicates the growth rate of the turmeric finger (cm) in accordance with the change in the resistance of the flex sensor-node 1 (after cultivation).The difference between minimum (initial) and maximum (final) resistance was used to obtain the following table at the end of the cultivation and finger length based on the actual physical measurement (cm scale) with respect to the centre of turmeric [20].The same procedure was followed for all the finger growth measurements.
Node average growth rate with respective to flex sensor resistance G avg is found using the Equation ( 5).
( ) (
) Figure 12 graph shows the relationship between the turmeric finger growth (cm) and the change in resistance (R) of the flex sensor-node 1.The measurement of finger growth is identified only after cultivation which is the normal practice that existed among farmers and they never get a chance to view the finger during cultivation.The finger growth of N1 and N4 was not much when compared to that of the other nodes.On the contrary, N2 had more growth than others.As turmeric is one of the row crops, the finger direction and the number of fingers depend on it [21].In node-1, an average flex sensor resistance relating to the finger growth had been calculated as 3.96 cm/kΩ.The number of primary and secondary finger branches was 7 and 16 respectively.The maximum length of the finger was 11.5 cm and the minimum length was 2.4 cm and the yield was 0.429 kg.After steam and dry process, the polished turmeric yield was 0.085 kg for sale.Node 1 had been kept in the middle of the row, there was minimum space between the crops with that of the standard distance [22] and so the yield was average.Naturally, there was no Rhizome rot disease attack in node 1.
Node 2
Figure 13 graph shows the growth rate measurement based on readings taken at regular interval of the turmeric finger (node 2).
Figure 14 shows the output image of the turmeric finger (node 2).
Growth Result
Table 2 indicates the growth rate of the turmeric finger (cm) in conforming to the change in the resistance of the flex sensor-node 2 (after cultivation).
Figure 15 graph shows the relationship between the turmeric finger growth (cm) and the variance in resistance (R) of the flex sensor-node 2. Node average growth rate appropriate with flex sensor resistance G avg was found using the Equation (7).
cm k
The average flex resistance with respect to the finger growth is 3.96 cm/kΩ in the node 2 with 7 and 12 numbers of primary and secondary branches.The maximum and minimum length of finger was 10.3 cm and 2.4 cm respectively with the yield of 0.273 kg.The polished turmeric yield was 0.055 kg for sale after the steam and dry process.The yield was average because the node 2 had kept in the middle of the row with an optimum distance between the crops.
Node 3
Figure 16 graph shows the growth rate measurement based on readings taken at regular interval of the turmeric finger (node 3).
Figure 17 shows the output image of the turmeric finger (node 3).
Growth Result
Table 3 indicates the growth rate of the turmeric finger (cm) in accordance with the change in the resistance of the flex sensor-node 3 (after cultivation) Figure 18 graph shows the relationship between the turmeric finger growth (cm) and the change in resistance (R) of the flex sensor-node 3. Node average growth rate with respect to flex sensor resistance G avg was found using the Equation (8).The average flex resistance with respect to the finger growth is 3.96 cm/kΩ in the node 3 with 7 and 1 numbers of primary and secondary branches.The maximum and minimum length of finger was 6 cm and 0.8 cm respectively with the yield of 0.127 kg.The polished turmeric yield was 0.025 kg for sale after the steam and dry process.An artificial method of turmeric finger Rhizome rot disease was caused by stacking water for more than two days.Hence the yield was very low.The natural pesticides like pseudomonas and viride were not intentionally applied on the affected turmeric crop (node 3) for testing the rhizome rot disease.
Node 4
Figure 19 graph shows the growth rate measurement based on readings taken at regular interval of the turmeric finger (node 4).
Figure 20 shows the output image of the turmeric finger (node 4).
Growth Result
Table 4 indicates the growth rate of the turmeric finger (cm) in accordance with the change in the resistance of the flex sensor-node 4 (after cultivation).
Figure 21 graph shows the relationship between the turmeric finger growth (cm) and the difference in resistance (R) of the flex sensor-node 4. Node average growth rate with respect to flex sensor resistance G avg was found using the Equation (9).
cm k
The average flex resistance with respect to the finger growth is 3.96 cm/kΩ in the node 4 with 12 and 25 numbers of primary and secondary branches.The maximum and minimum length of finger was 13.9 cm and 7.1 cm respectively with the yield of 1.396 kg.The polished turmeric yield was 0.28 kg for sale after the steam and dry process.The distance between the crops was optimum in node 4 which is kept in the middle of the row [22] and an artificial method of turmeric finger.Rhizome rot disease was caused by stacking water for more than two days.The natural pesticides like pseudomonas and viride were applied on the affected crop (node 4) for testing the rhizome rot disease.Initially, it was found that there was growth in all nodes (F1, F2, F3, F4 and F5 of node 4) and at the middle (F2, F3 and F5) there were no further improvements in the growth.Thereby, it was identified that the above said nodes were attacked by Rhizome rot disease.After identifying the Rhizome rot disease, natural pesticides (pseudomonas and viride) were applied which resulted in an improvement on those nodes.The yield was more as compared to that of the node 3 which was affected and not treated for rhizome rot disease.
Node 5
Figure 22 graph shows the growth rate measurement based on readings taken at regular interval of the turmeric finger (node 5).
Figure 23 shows the output image of the turmeric finger (node 5).
Growth Result
Table 5 indicates the growth rate of the turmeric finger (cm) in accordance with the change in the resistance of the flex sensor-node 5 (after cultivation).
Figure 24 graph shows the relationship between the turmeric finger growth (cm) and the deviation in resistance (R) of the flex sensor-node 5. Node average growth rate with respect to flex sensor resistance G avg was found using the Equation (10).The average flex resistance with respect to the finger growth is 3.96 cm/kΩ in the node 4 with 8 and 16 numbers of primary and secondary branches.The maximum and minimum length of finger was 11.9 cm and 1.6 cm respectively with the yield of 0.397 kg.The polished turmeric yield was 0.08 kg for sale after the steam and dry process.Naturally, there was no Rhizome rot disease attack in node 5.
The application of nitrogen (N) and potash (K 2 O) as well as two micronutrients Zinc (Zn) and boron (B) play an essential role on growth and yield [23].From the data of all nodes, it was obvious that only during the last four month; the turmeric finger had full-fledged growth.Therefore, monitoring the finger growth and applying the fertilizer at times, the farmers can get high yield in terms of weight and quality of the turmeric with high density which obtains more market prices.Modern precision agriculture requires adaption of lot more advanced technology [24] [25].
Conclusion
In the agricultural field at present the technology occupies a paramount position and it is important because agriculture forms the backbone of India occupying the major portion of economy by increasing its yield.The greatest requisites of this period are the amalgamation of computer technology in the agricultural field.By using embedded system, sensor array module with array of sensors such as flex sensor, temperature sensor and pH sensor were used to view the turmeric finger growth characteristics.The ZigBee and the GSM module had been used for communicating the data wirelessly and the overall growth rate had been estimated by examining the graph plotted on a daily analysis.The five different nodes have been used in the experimental work with an average flex sensor resistance of 3.962 cm/kΩ.Of the five nodes, the node 3 and 4 were diseased and 3 was left as such purposely while the node 4 was treated with pseudomonas and viride to restrict the Rhizome rot disease.It was finally observed that node 4 contained comparatively less rhizome disease than node 3. The definite advantage of this system is the early detection of the diseases and monitors the growth of the turmeric fingers when it is under the soil.By implementing this concept, farmers can be greatly guided to predict the final outcome by testing the growth rate and to take preventive measures in the early stage.
Figure 6 .
Figure 6.Designed module implemented in the field.
flex sensor (F1 to F5) × 5 nodes = 25 sensors were used).The positions of five individual flex sensors are shown in the following Figure 9. | 2017-12-04T23:14:12.198Z | 2016-06-02T00:00:00.000 | {
"year": 2016,
"sha1": "7b8e906946f7a439b4478d44a293ed807d051e15",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=67167",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "7b8e906946f7a439b4478d44a293ed807d051e15",
"s2fieldsofstudy": [
"Agricultural and Food Sciences",
"Engineering"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
268909520 | pes2o/s2orc | v3-fos-license | The Influence of Social Media on Perception of Body Image and Beauty Standards on Young People
. Through different sections, this paper discovers the complex relationship between social exposure and body image perceptions among young individuals. In the era of technological development, the significance of media exposure to young people cannot be underestimated. The Causes and Consequences section is divided into the previous studies and theories and the consequences of frequent media exposure. These two parts separately discussed the possible reasons for media influencing young people and the results, for instance, anxiety, lack of confidence, or dissatisfaction with their body image. It examines how these portrayals reinforce traditional gender roles, cultivate body dissatisfaction, affect interpersonal relationships, and shape societal norms and expectations. The third part introduces some mitigating factors and interventions. Including education initiatives and media literacy programs to support young people’s critical thinking and viewing skills when facing media exposure, it also includes policy interventions through government agencies and health services to improve young people’s mental health. This research fosters a healthier relationship between young people and media exposure.
Introduction
The evolution of technology has closely followed the widespread adoption of the internet, mainly through various forms of media.This influence began with television and movies and has now extended to include videos and streaming media.Over the past few decades, this rapid technological advancement has significantly impacted individuals, especially the younger generation, who have grown up in this media-centric era.These positive and negative effects are particularly evident in how unrealistic beauty standards have been shaped.
While society has always placed importance on beauty ideals, the rise of media and easy access to the internet has exponentially amplified this phenomenon in the digital age.Young people, in particular, are constantly exposed to manipulated images and narratives depicting distorted body ideals, which have, in turn, started to define beauty standards.Social media platforms play a crucial role in perpetuating these standards.
Media exposure is nearly ubiquitous among young people in today's digital landscape.According to a 2019 study by Common Sense Media, American youths aged 13 to 18 spend an average of over seven hours per day engaging with various media forms, including television, video games, social media, and online content [1].This research paper explores the intricate relationship between media exposure and young individuals and how this exposure influences their perceptions of body image.This topic holds substantial sociological significance, as it has far-reaching consequences on societal norms and values.The primary focus of this investigation is to analyze how media exposure shapes young people's perceptions of their bodies and to understand the sociological implications of these influences in perpetuating unattainable beauty standards.
Previous studies and theories
The internet's and technology's rapid development immediately demonstrates the significance of this social issue.The connection between media exposure and how people perceive their bodies has become more important with the rapid diffusion of information.Television, magazines, social media, movies, and advertising are just a few examples of the media that greatly impact how individuals perceive their bodies.In this literature review, we will focus on the social aspects of unrealistic beauty expectations to explore the key themes and conclusions about the effects of young people's exposure to body images.
Recent studies have revealed the significant role of media exposure in shaping body image perceptions among young individuals.One such study by Perloff [2] examined the interesting features of social media and its distinctive content.He found that it has a significant impact on body image issues through unfavorable social comparisons.According to this study, the exposure to "ideal" bodies on TikTok and Instagram frequently was linked to negative comments about body dissatisfaction and low self-esteem.
Another study by Fardouly et al. [3] looked at the impact of media exposure, including viewing digitally manipulated photos, on young women's body image dissatisfaction.Others who used social media said they felt more depressed than others who stayed on the control page.In addition, when browsing social media sites versus the control page, women who tended to compare their appearances reported higher differences in their facial features, hair, and skin.According to this study, photo altering on social media could make people feel less positive.
There are also sociological consequences of media-driven beauty standards.They are rooted in complex mechanisms that connect with societal norms and expectations.Tiggermann and Slater [4] examined, focusing on the internet, how exposure to media affects preteen girls' anxieties about their bodies.The internet was concluded to represent a potent sociocultural force among pre-teenage girls.The pressure caused girls to keep focusing on the internet's ideal body image information, at last resulting in body image concern.
Additionally, Groesz et al.'s study [5] examined how media exposure can cause people to internalize social standards for beauty and thinness, which eventually contributes to unfavorable body image evaluations.After viewing thin media images as opposed to photos of average-size models, plus-size models, or inanimate objects, participants' body image was much more unfavorable to themselves.Participants under the age of 19, who participated were more likely to have their thinness schemas activated showing this impact was more severe.
The social comparison theory is also applied to media exposure topics.This theory was initially proposed by social psychologist Leon Festinger in 1954, and it focused on the belief that people try to gain accurate self-evaluations by comparing with other people.According to a study, more women are making upward comparisons and evaluating themselves by the inflated expectations set by the media [6].Self-perceived resemblances to role models on social media can also impact men's and women's self-esteem.Self-esteem can be boosted by feeling more like a role model and lowered by feeling less like one [7].This theory implies that the rapid development of technology can negatively affect people's health.Through the years, more and more social platforms will be created, and the time young people are willing to put in is also increasing.Therefore, the chances of young people gaining negative influences are also increasing.The significance of this social issue is also growing.
Other than theories, there is a new trend of beauty filters and technological ways of improving our body shape.To pursue beauty ideas, beauty filters have evolved.There are pros and cons to the widespread use of beauty filters on social media.While new technologies offer creative and enjoyable ways to enhance one's appearance, they exacerbate issues that blur the line between reality and digital constructs, exacerbating body dissatisfaction and unrealistic beauty standards [8].This research indicates that the growth of digital media brings the consequences of illustrating romantic body images.This paper will continue to be based on this literature as we move forward to understand the cases and the consequences we obtained from this issue.
Media Portrayals of Unrealistic Beauty Standards
In this society, media significantly shaped people's beauty ideals and perpetuated unrealistic standards of attractiveness.In this section, the paper will discuss the different impacts of various medial portrayals of unrealistic beauty standards.
Advertisement and fashion have been the two essential components of perpetuating distorted body image.Advertising is crucial in projecting idealized beauty standards since it is a pervasive and compelling medium.Models and celebrities who follow particular physical standards are frequently used as brand ambassadors.Their representations allude to the consumers that their standards are easy to reach.Research like those by Grabe, Ward, and Hyde [9] has examined how viewers' exposure to images of idealized beauty in advertising can cause them to feel unsatisfied with their bodies.Such representations encourage a consumer culture that combines the pursuit of an idealized appearance with the use of products.The emphasis on thinness has also been an essential motif in fashion.These slender body standards, have been questioned for their propensity to accelerate body image problems [10].This emphasis on thinness upholds a limited definition of beauty by ignoring the diversity of body types and shapes.
Popular culture, including movies, television, and other entertainment media, maintains and promotes false beauty standards.These media greatly influence young people's perceptions of themselves and others, frequently presenting beauty as a crucial quality intimately connected to success, attractiveness, and desirability.Characters that meet traditional beauty standards commonly appear in mainstream movies and TV shows.Leading actors and actresses, whose appearances are frequently extremely managed, who become cultural icons representing beauty's predominant ideals.Children who watch these values on television often may acquire the impression that they are desirable and attainable.
The impact of these images is shown by research by Tiggemann and Slater [4], which shows that media representations of muscularity and thinness indicate more general societal norms related to our usual appearance and value.
The influence of entertainment media is influential throughout adolescence and young adulthood.Identity exploration, self-discovery, and an increased sensitivity to societal norms and expectations are characteristics of these life stages.Young people are more receptive to media messages that define beauty ideals and frequently compare their appearance to these arbitrary standards.
Social media and other digital channels have emerged as essential tools for spreading beauty standards in the modern era.Online personas are deliberately curated by social media influencers, celebrities, and users, who frequently present an idealized picture of themselves [2].Due to people's tendency to make upward social comparisons, creating well-maintained online personas impacts how people perceive beauty.
Consequences of Unrealistic Beauty Standards
Unrealistic beauty standards will cause various consequences, including the reinforcement of gender roles, the cultivation of distorted beauty ideas, and interpersonal relationships.This section will focus on the sociological consequences of these standards.
The media frequently reinforces traditional gender norms while also maintaining artificial beauty standards.These depictions have a disproportionately negative impact on women.Women are commonly portrayed in the media as objects of desire, with physical attractiveness taking advances over other qualities.People are expected to follow to cultural ideals of femininity and beauty, and women's agency and opportunities may be restricted by this reinforcement of traditional gender norms.Men are also not exempt from these effects since media depictions of muscularity and physical perfection are increasingly common.Promoting these values might cause problems with body image and harmful body behaviors in young males.
Teenagers' growing body dissatisfaction is one of the most critical societal effects of unattainable beauty standards.Numerous studies [3] show that media exposure and body dissatisfaction are directly related.People frequently believe they fall short of the idealized pictures in the media, which lowers their self-esteem and causes them to have unfavorable judgments of themselves.This body dissatisfaction can negatively impact relationships, mental health, and general well-being.Particularly in young people going through a sensitive time, it can contribute to the emergence of mental health conditions such as anxiety and depression.
The distorted standards may also influence interpersonal relationships, especially in romantic relationships.One person may be affected by unrealistic beauty standards through the media and ask for requirements or pressure the other half to achieve a perfect body.It will bring tension and insecurity into the relationship; after all, it can impact the dynamics and stability of a relationship.
The last societal consequence will be that distorted beauty ideas will be extended to other aspects of individuals' lives.Some social media users frequently criticize others' body shapes; however, these standards appear in other spaces where people may see beauty over other essential qualities, fostering a superficial view of beauty.The societal consequences usually extend to workplace dynamics, where appearance-based judgments can influence hiring decisions and working treatments.
In conclusion, there are numerous and widespread sociocultural effects of the exaggerated ideals of beauty promoted by the media.They strengthen cultural norms and expectations, foster body dissatisfaction, affect interpersonal interactions and promote established gender roles.These repercussions significantly affect young people's mental health and self-perceptions, underlining the need for critical analysis and suggesting ways to change media and cultural structures that promote beauty standards.
Education Initiatives and Media Literacy Programs
This section will examine the potential media literacy programs and educational initiatives.Those methods will help young people to set up healthy values and beauty standards.Other than concentrating on the impact of distorted body image caused by media, cultivating young people by promoting diverse media representations, empowering critical consumers, and providing educational initiatives can be helpful.
For young people, media literacy can be a tool for empowerment.The process of advancing media literacy skills is known as media literacy education, and its goals are to increase awareness of media influence.This can foster an engaged attitude toward both consuming and producing media.Young people who receive such education are more equipped to evaluate media content critically and challenge authorities.They give people the abilities to realize the photo manipulation, the business motives behind advertising, and the sociocultural settings that influence media portrayals [11].
At the same time, incorporating educational initiatives can help the influence of raising awareness of media impact.Formal education can mitigate the sociological influences.Schools can integrate media literacy classes as an interdisciplinary curriculum to discuss body image, self-esteem, and critical media analysis.These educational initiatives aim to raise awareness and foster resilience against media-driven beauty pressures [2].
There is another path to mitigate the consequence of media exposure.When schools and institutions can help young people establish a healthy beauty standard, film companies and other institutions can eliminate the negative consequences of unrealistic beauty ideas by promoting diverse media representations.Introducing more diverse body images can effectively challenge the social norm and contribute to a more inclusive and accepting media landscape.Collaboration with the media industry is also essential in this endeavor.Encouraging responsible media practices, such as the disclosure of digital retouching in advertising, can contribute to transparency and help young people differentiate between manipulated images and reality.
Media literacy and education initiatives aim to empower young individuals as critical consumers of all media products.Critical thinking can actively challenge unrealistic body standards when equipped with the tool.As long as enough people have a critical mind, media influence will substantially decrease.
Aside from school and the education system, policy implications and interventions could also be extremely useful.This section discusses potential policy implications and interventions that can mitigate the sociological consequences of unrealistic beauty ideals among young people.
Policy Implications and Interventions
Aside from school and the education system, policy implications and interventions could also be extremely useful.This section will discuss the potential policy implications and interventions that can mitigate the sociological consequences of unrealistic beauty ideals among young people.
Media Regulation and Accountability can be improved by government agencies.Government agencies can be vital in regulating the media sector, especially advertisements.Consumers can make informed decisions if policies are implemented to ensure that advertising discloses the use of digital retouching and photo editing.Regulating excessively digitally altered images in advertising can also encourage openness [2].Also, mental health services should include the mental health implications of unrealistic beauty standards.Policymakers should allocate resources to support mental health services, particularly for young individuals.Accessible and affordable mental health resources should be available to those struggling with body image issues and related psychological challenges, including counseling and therapy.
In conclusion, policy implications and interventions are vital tools in addressing the sociological consequences of unrealistic beauty standards perpetuated by the media.Through a concerted effort involving policymakers, media organizations, educators, and advocacy groups, it is possible to create a media landscape that fosters a more inclusive, responsible, and empowering portrayal of beauty ideals, benefiting the well-being and self-esteem of young individuals.
Conclusion
This research paper has reviewed the intricate relationship between media exposure and body image perceptions, which has a sociological focus on the reasons, consequences, and actions we can take to prevent more severe consequences from happening.
Advertising, entertainment, and social media are just a few places where media reinforce conventional gender roles, promote body dissatisfaction, affect interpersonal interactions, and influence societal norms and expectations.
This essay has also emphasized potential protective suggestions enabling young people to interact with media more critically.They will be taught under media literacy initiatives and educational programs to question irrational notions of beauty and understand more critical content behind the media.Furthermore, the collaborative role of governments, media companies, schools, and advocacy groups in creating a more responsible and inclusive media environment is underscored by policy implications and interventions, ranging from media regulation to diversity programs.
The connection between media exposure and young people's conceptions of their bodies continues to be a dynamic and changing phenomenon as we continue in the digital age.It requires ongoing investigation, discussion, and action.We want to create a future where young people interact with media content with assurance, resiliency, and a genuine sense of self-worth by identifying the complex social effects of media-driven beauty standards and implementing educated regulations and solutions.By doing this, we support a society that honors variety, inclusivity, and the appreciation of beauty. | 2024-04-05T17:45:24.458Z | 2024-04-01T00:00:00.000 | {
"year": 2024,
"sha1": "3e0ae9d66af6040fde39f434103a848c069c75bd",
"oa_license": "CCBYNC",
"oa_url": "https://wepub.org/index.php/TSSEHR/article/download/1075/1070",
"oa_status": "HYBRID",
"pdf_src": "Anansi",
"pdf_hash": "71314e7a526abbb7977e924f84821911ae710bf1",
"s2fieldsofstudy": [
"Sociology"
],
"extfieldsofstudy": []
} |
6253796 | pes2o/s2orc | v3-fos-license | Tay-Sachs carrier screening in the genomics age : Gene sequencing versus enzyme analysis in non-Jewish individuals
Purpose: To compare the sensitivity of Hexosaminidase A (HexA) enzyme-based testing to gene sequencing for carrier detection in non-Jewish individuals. Methods: Blood samples were obtained from parents and relatives of affected patients at an annual TaySachs and Allied Diseases Foundation meeting. A family history was taken for each individual. Samples were analyzed for leukocyte HexA activity, serum HexA activity and subjected to extensive gene sequencing. The results from these analyses were combined with our previously published data describing 34 obligate Tay-Sachs disease (TSD) carriers. Results: Twelve additional TSD carriers were detected in this study. Gene sequencing successfully identified all 12 carriers whereas enzyme analysis identified 11 of 12 carriers. This individual is a carrier of the B1 variant that is known to cause false negative results with enzyme testing. Combined data from 46 non-Jewish TSD carriers revealed that gene sequencing had a higher sensitivity rate than HexA enzyme-based testing (94% versus 87%) in non-Jewish TSD carriers. In our series, approximately 4% of non-Jewish TSD carriers have this mutation. Conclusions: HexA gene sequencing provides a higher sensitivity for TSD carrier detection than HexA based enzyme analysis in non-Jewish patients primarily due to the presence of individuals with the B1 variant.
INTRODUCTION
As automation and technical improvements continue to drive down the costs of DNA sequencing, it has now become feasible to perform Hex A gene sequencing as a primary screening test for Tay-Sach's Disease (TSD) carrier detection.This study enlarges upon an earlier study that demonstrated that sequencing and enzyme analysis had identical sensitivities in identifying TSD carriers to determine whether DNA sequencing would be an appropriate initial screen for non-Ashkenazi Jewish (AJ) individuals.
Tay-Sachs disease, or GM2 Gangliosidosis, is an autosomal recessive neurodegenerative disorder caused by a deficiency of beta-hexosaminidase A (HexA), resulting in lysosomal accumulation of GM2 ganglioside.There are several forms of TSD with the most common being the classic infantile form that leads to death between 2 and 5 years of age.Infrequently Tay-Sachs disease can first manifest symptoms in adulthood and is called late onset TSD or LOTS.However, even in LOTS there is a progressive of symptoms are progression eventually resulting in profound disability [1].
Certain ethnic groups have an increased carrier frequency for Tay-Sachs disease with the highest being Ashkenazi Jews (AJ) who have a carrier frequency of approximately 1:30.French Canadians, Cajuns and Pennsylvania Dutch also have increased carrier frequencies relative to other ethnic groups.Low risk populations have a carrier frequency of approximately 1:300 [1].
Throughout most of the developed world, high risk populations are offered carrier screening for Tay-Sachs Disease.In the United States and Canada carrier screening was begun in the 1970's using an enzymatic approximation of HexA activity [1,2].Enzyme based screening has resulted in an approximate 90% reduction in new cases of TSD in these countries.[1] Because at risk couples are usually screened, most new cases of Tay-Sachs disease in the US and Canada are born to parents where at least one parent belongs to non-high risk group.The American College of Obstetricians and Gynecologists recommends carrier screening for couples of Ashkenazi Jewish, French-Canadian, and Cajun descent as well as for couples in whom only one member of the couple belongs to a high risk ethnic group [3].
Enzyme based HexA carrier screening is performed using an artificial fluorogenic substrate.Unfortunately there is a second enzyme, Hexosaminidase B (HexB) that acts on the same substrate.Discrimination between HexA and HexB activity heat inactivation; takes advantage of the observation that HexA is relatively more heat labile than HexB.The enzyme assay is run twice, both before and after a period of heat inactivation.The percentage of HexA activity is calculated and this is used to determine if an individual is a carrier or noncarrier [2].The assay can be performed on serum, white blood cells and platelets.The serum assay cannot be used during pregnancy because of the presence of the HexP isozyme which is heat stable and artificially reduces the apparent percentage of HexA.Therefore during pregnancy, only white cells and platelet assays are reliable [4,5].Regardless of sample type there is a significant overlap between noncarrier and carrier ranges for percentage HexA with between as many as 10% of patients results falling in the indeterminate or "gray zone".Often even after repeated analyses are performed on multiple samples, certain individuals will be unable to be classified in terms of their TSD carrier status [6].
After population based screening was instituted, it was observed that certain unaffected individuals who presented for carrier detection actually had HexA levels in the affected range.This finding eventually led to the discovery that there are specific DNA mutations, now called pseudodeficiency alleles that cause the enzyme to be deficient in the metabolism of the artificial substrate in vitro, but have normal in vivo activity against the native substrate.Therefore all individuals who are apparent carriers by enzyme analysis require molecular testing to rule out the presence of a pseudodeficiency allele.Approximately 2% of AJ and 3% of non-AJ individuals carry a pseudodeficiency allele [7,8].Thus all individuals regardless of ethnic background who are apparent TSD carriers by enzyme testing require followup with molecular testing to investigate the presence of a pseudofediciency allele.
There are also at least 5 known mutations that cause a phenotype called the B1 variant where Hex A shows normal or near normal activity against the artificial substrate but is deficient in vivo against the naturally occurring substrate [9][10][11].Consequently, individuals heterozygous for this genotype will have false negative or indeterminate enzymatic carrier tests but are still carriers for TSD.Since these individuals may have false negative enzyme testing they would remain undetected without molecular testing.
Although reasonably accurate and cost effective, the enzyme assay suffers from some inherent limitations.As noted above, there is significant overlap between the ranges of non-carrier and carriers, the presence of a pseudodeficiency allele causes false positive results, and the presence of a B1 allele can cause false negative results.Enzyme analysis can be affected by medications including oral contraceptives which can cause false positive results [12,13].
The enzyme assay also cannot differentiate between carriers of the severe infantile form of TSD from the milder forms.Finally, HexA enzyme analysis requires special logistics because the samples must be received in the laboratory in a timely fashion at appropriate temperatures in order to obtain reasonably accurate results.Once the samples are received in the laboratory they must be processed rapidly leading some laboratories to only accept specimens on certain days and times.
In Ashkenazi Jews, 3 mutations account for 92% -98% of all carriers of TSD.Many authors have suggested that for AJ individuals, a panel consisting of these 3 mutations and the 2 pseudo deficiency alleles is preferable to enzyme testing for a carrier screening program [6,14].However, for non Jewish individuals, this panel would detect only about 50% of TSD carriers [6,14,15].
Previously we collected samples from 34 obligate or self reported non-AJ TSD carriers at a National Tay-Sachs and Allied Diseases meeting.We compared 4 modalities of testing for sensitivity for detection of TSD: enzyme analysis of serum, enzyme analysis of leucocytes, common mutation analysis, and HexA gene sequencing.As expected, the common AJ mutation panel had a sensitivity of only 52% in this population.The enzyme analysis and DNA sequencing analysis had equal sensitivities of 91%.Specificities could not be determined since these individuals were known TSD carriers.The current study represents and extension of the original study [15].In addition we examined our data from 64 samples submitted for clinical HexA sequencing to determine the prevalence of novel mutations and variants of unknown clinical significance (VUCS).
Clinical Samples
Volunteers were requested at the 2009 Annual Meeting of the Tay-Sachs and Allied Diseases Association.
Inclusion Criteria: Any individual affected with TSD and any individual who was a first or second degree relative of a patient with TSD, with Tay-Sachs disease were offered enrollment in the study.
Exclusion Criteria: Patients and individuals related to patients with diseases other than TSD were excluded from the study.
Patients were consented in a protocol approved by the Western Institutional Review Board.For this study, only samples obtained from 23 individuals who were self identified as being of non-Jewish ancestry and were related to patients affected with infantile TSD were used.Each individual was tested for serum Hex A activity, white cell Hex A activity, and complete gene sequencing.
DNA Extraction
DNA from most of the samples were extracted using Gentra Autopure LS (Minneapolis, MN) exactly as described by the manufacturer, and some samples were extracted using Qiagen Biorobot 9604 platform using MegAttract ® DNA blood M96 kit (Valencia, CA).Whole blood was used for DNA extraction in most cases, and in some cases, leukocytes were used.We observed no differences in sequencing quality between different extraction platforms and different starting materials.
HEXA Gene Sequencing
DNA sequencing and PCR were performed essentially as described by Huang et al. [16] except that DNA purification post-PCR was done using exonuclease (USB Corporation, Cleveland, OH)/calf intestinal alkaline phosphatase (Promega, Medison, WI) digestion, and post dye terminator reaction DNA was purified by ethanol precipitation and resuspended in Hi-Di formamide (ABI, Foster City, CA).Fourteen amplicons, each amplifying one of 14 exons, were generated through individual PCR reactions.Amplification efficiency and correctness of sizes were verified by agarose gel electrophoresis.Sequencing covered exons 1 through 14 of chromosome 15 and at least 20 bp flanking sequences.Samples were resolved on an ABI 3730 automated DNA analyzer and data were analyzed using SeqScape software (both from Applied Biosystems).
HEXA DNA Common Mutation Test
We purchased xTAG, Ashkenazi Jewish Reagents from Tm Biosciences (Toronto, Canada) and used them according to manufacturer's instructions as described previously [17].Mutation analysis was performed for 5 common mutations and 2 pseudoalleles: 1278insTATC; ).This is a multiplex assay that analyzes 31 mutations in various genes using multiplex PCR followed by allele specific primer extension, Luminex bead conjugation, and analysis on a Luninex 200 instrument (Luninex Corporation, Houston, TX).
HEXA Enzyme Assay
Serum collected from subjects was separated from blood clots by centrifugation and kept frozen at −20˚C until analysis.Leukocytes were extracted by the method of Snyder and Brady [18].with some minor modifications and protein concentration of the leukocyte cell pellets was determined by the method of Lowry [19].
The Hex-B enzyme is heat stable, but Hex-A is heat labile.The amount of Hex-B was determined by measuring total hexosaminidase activity, both before and after heat inactivation.The percent HEX-A is determined by the equation: (HEX (total) -HEXB)/HEX (total).Our established reference intervals are non-carriers: percent HEXA > 57% and <80%, TSD carriers percent HEXA >= 25% and <52%, with indeterminate (gray zone) percent HEXA 52% -57%.
Statistical Analysis
Chi square analysis was performed using the software program Epistat.
RESULTS
Of the 23 tested individuals, 11 were determined to be non-carriers by both enzymatic and molecular techniques.Table 1 tabulates the results for the 12 carriers in this series.The most illuminating case is an individual with a B1 phenotype and the genotype of c.533G > A (p.Arg178His).This individual is a TSD carrier but has indeterminate results on leucocyte testing and a false negative non-carrier result on serum testing.In the previous series we had detected another individual with the same mutation who had false negative non-carrier by both the leukocyte and serum enzyme assay.Combining the data from this study and the previous study [11].demonstrates that for the individuals tested in these 2 studies, extensive gene sequencing has a higher sensitivity than enzyme analysis for TSD carrier detection.DNA sequencing had a sensitivity of 94% versus 87% for enzyme analysis in 40 TSD carriers.Although the numbers in this study are too small to reach statistical significance, it is clear that DNA sequencing has at least an equal, if not superior sensitivity to enzyme analysis.
Leukocyte testing performed slightly better than serum testing due to an increased rate of indeterminate results in the serum (25%) versus leukocyte series (8%).
OPEN ACCESS
One of the complications of performing gene sequencing for TSD carrier detection is the possibility of detecting variants of unknown clinical significance (VUCS).In the current series, 4 of the 12 carriers had novel variants defined by the fact we could not find any publications or databases containing these mutations.Of these 4, none would be considered a VUCS.The mutation c.1214-1215delAAinsG causes a frame shift mutation and therefore almost certainly deleterious, as is the c.759-774dup.The nonsense mutation c.1292G > A (p.W431X) is also clearly pathogenic.The c.253 + 1G > A affects a consensus splice donor sight and therefore would also be predicted to be pathogenic.
In order to investigate the frequency of VUCS that might be encountered should HexA gene sequencing be used as a first line TSD carrier screen, we reviewed the data from our initial 64 clinical sequencing assays.There were 10 patients with novel mutations and are listed in Table 2.The mutation p.Met459Val was seen in 3 unrelated patients.One of these patients is affected with TSD and is a compound heterozygote for a classic TSD mutation.A second individual was a heterozygote was identified as a carrier for TSD by enzyme analysis thus confirming that this allele is almost certainly a deficiency allele.There were 2 individuals in the same TSD family with the novel variant p.Ala479Thr that is predicted to probably damaging by the algorithm of Poly-Phen (http://genetics.bwh.harvard.edu/pph2/)[21].The allele c.316 + 1G > T alters a consensus splice site and other mutations that alter this site have been demonstrated to be causative mutations [22].suggesting that this is also almost certainly a TSD allele.The variant c.1A > C is also almost certainly a disease causing allele since it alters the origin of transcription and other mutations altering this site are demonstrated to be TSD alleles.The variant p.Gly520Asp is predicted to be benign by the Poly-Phen site.An individual with indeterminant enzyme results had a novel VCUS of p.Glu162Lys (c.484G > A).Therefore this individual's carrier status could not be resolved.
A second VCUS result was an individual with the variant p.Phe434Leu (c.1302C > G).We were unable to determine if this patient had prior enzyme testing.Thus, 2 of 64 (3%) cases were reported as VCUS.This is comparable or lower than the number of individuals who fall within the gray zone for HexA enzyme testing results in most laboratories.
DISCUSSION
Although population based carrier detection using enzyme analysis has reduced the births of affected children in high risk ethnic groups drastically, infants with TSD continue to be born.Since almost all new cases of TSD have at least one parent from a non high risk group, it is important to optimize the sensitivity of carrier detection in those populations.It is clear that a panel consisting of the common AJ and French Canadian-Cajun alleles will only detect approximately 51% of all carriers from other ethnicities.This study has demonstrated that the 4% prevalence of the B1 phenotype, in addition to the 8% indeterminant rate for leukocyte HexA activity limits the sensitivity of the enzyme assay to below 90% in the patients tested in this study.Our observed sensitivity for the enzyme based assay for the 40 confirmed carriers in our combined studies was 87%.Gene sequencing performed better with a sensitivity of 94% although the numbers were to low to reach statistical significance.Since there are so few births of affected children, getting more obligate carriers to enlarge this study would be difficult.
In terms of specificity, the low rate of VCUS makes sequencing equal or better than enzyme analysis.Individuals with VCUS can have confirmatory enzyme testing, similar to the current practice of all TSD carriers on enzyme analysis being reflexed to common mutation analysis to rule out a pseudodeficiency allele.In this manner almost all VCUS will be classified into benign variants and disease causing mutations.As the database expands, progressively fewer VCUS will be encountered.It is possible that for a small number of individuals who have a VCUS, a negative family history and negative or indeterminant enzyme analysis could have an, as yet, undescribed B1 variant and be a carrier for TSD.It is impossible to predict how often this might occur, but the possibility could be discussed in a genetic counseling session.
Using DNA sequencing as the primary screening test for non AJ patients would have several advantages over HexA enzyme analysis.In addition to advantages listed in the introduction, prenatal diagnosis by mutation analysis is immediately available for couples.Pseudodeficiency and LOTS alleles will be identified and the false negative results for individuals with the B1 phenotype will be avoided.Even dried filter paper spots can be used for DNA sequencing analysis allowing samples to submitted from remote areas.Saliva testing and buccal swab testing would also be possible.
Gene sequencing is not 100% sensitive either.In the initial study 3 individuals had negative sequencing results and positive enzyme results.In addition DNA sequencing will not detect the non-Jewsish TSD allele 7.6 kb deletion (7.6 kb delex1).This deletion is detected in our common mutation panel.If DNA sequencing is introduced as a primary screen for non-Jewish individuals, a separate assay would need to be performed to detect this mutation.
Since neither enzyme based testing nor molecular based testing is capable of detecting all TSD carriers, the combination of both modalities will provide the highest sensitivity.Although not appropriate for all patients, this should certainly be considered for a spouse of a known carrier in order to optimize the negative predictive value of the testing.
However, for routine carrier testing, the data suggests that for an AJ, a common mutation analysis is sufficient.For a non-AJ, DNA HexA sequencing with a 7.6 kb del assay could replace the enzyme based assay and yield equal, if not higher sensitivity and specificity for carrier detection.Therefore, now that DNA sequencing can be performed at a comparable cost and turn around time to HexA enzyme analysis it may be time to consider molecular testing as first line approach to TSD carrier screening in all populations.
Table 2 .
Novel variants discovered in HexA sequencing assay. | 2017-10-02T05:37:31.502Z | 2013-03-26T00:00:00.000 | {
"year": 2013,
"sha1": "370d87803199d86da3bd4bd392a60075386cbae7",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=29367",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "370d87803199d86da3bd4bd392a60075386cbae7",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Biology"
]
} |
252031644 | pes2o/s2orc | v3-fos-license | Evaluation of Ecuadorian genotypes of Capsicum spp. against infestations of Bemisia tabaci
The genus Capsicum, native to tropical and subtropical America, belongs to the Solanaceae family, which includes commercially important vegetables such as chilies and green peppers. The silverleaf whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae), causes losses to vegetables including Capsicum species. Among the al - ternatives of pest control, an effective, economical, and environmentally compatible method is the resistance of the host plant. Infestation by B . tabaci was evaluated in 73 Capsicum genotypes
Introduction
The genus Capsicum belongs to the Solanaceae family and includes chili peppers and peppers; within this botanical family, Capsicum, stands out among the 90 genera of commercially important vegetables worldwide (Tripodi & Kumar 2019). This genus is native to tropical and subtropical America in an extensive region from Mexico to the southern part of the Andes, in which archaeological evidence suggests its use from the year 6000 BC (Tripodi & Kumar 2019). Capsicum comprises 42 species, including Capsicum annuum L., C. baccatum L., C. frutescens L., C. pubescens (Ruiz and Pavon) and C. chinense Jacq., (Qin et al. 2014, Vallejo-Gutiérrez et al. 2019; the first three are the most cultivated in the world (Anjos et al. 2018). Chili peppers and peppers have a wide variety of shapes, sizes, and colours of fruits, in which the hot ones are used as spices, and the sweet ones as vegetables, in addition to having decorative, medicinal, and cosmetologically purposes (Gálvez et al. 2021).
The whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) causes losses in vegetables in tropical and subtropical areas of the world (Latournerie- Moreno et al. 2015, Li et al. 2021, due to the direct damage caused by the sap sucking involving chlorosis, leaf deformation and plant weakening, as well as indirect damage resulting from the presence of a fungus (Capnodium spp.) that produces sooty mould and interferes with the photosynthetic process (Ortega et al. 2019). However, the transmission of viral diseases is the most important damage caused by B. tabaci (Lorenzo et al. 2016, Guo et al. 2020).
The commonly used control measures against insect pests in horticultural crops are based on the use of organo-synthetic pesticides; however, these products are mostly toxic to the environment and to non-target species and favour the development of resistant populations (Nombela & Muñiz 2010). An effective, economical, and environmentally compatible method for the control of insect pests is host plant resistance (Ballina-Gomez et al. 2013).
Painter (1951) defined resistance as the genotypic condition of a plant that allows it to be less damaged than another of the same species under similar environmental conditions. It includes three categories: antixenosis, antibiosis, and tolerance. In antixenosis, the plant may not be preferred for oviposition, shelter, or feeding because it has certain qualities that make it a poor host (Painter 1951, Kogan & Ortman 1978, Jeevanandham et al. 2018, while antibiosis describes the adverse effects of the host plant on the biology of the insect, including death of the early stages, abnormal growth rates, failure to pupate, under size of adults (Painter 1951). Finally, resistant plants may be tolerant if they survive below levels of infestation that could kill or severely injure susceptible ones.
Studies have shown that Bemisia tabaci is affected by the physical characteristics of the leaf surface, such as villi, glandular trichomes, and leaf shape (Ballina-Gomez et al. 2013, Al-Aloosi et al. 2020, Sripontan et al. 2022. It is also known that pepper varieties and cultivars have different chemical composition, antioxidant and allelochemical compounds that influence resistance against phytophagous insects (Sripontan et al. 2022). Particularly, domesticated plants and wild relatives of Capsicum can be important sources of resistance to reduce damage by phytophagous (Ballina-Gomez et al. 2013, Tripodi & Kumar 2019. Therefore, the local germplasm of domesticated chillies and peppers could be sources of resistance especially in Ecuador, which is a country of high diversity of Capsicum species. Consequently, the objective of this study was to evaluate the response of Ecuadorian Capsicum genotypes to B. tabaci infestations.
Material and methods
The experiment was conducted from February until May 2018 inside a shade house of 1625 m 2 (25 m x 60 m) built with high-density polyethylene shade cloths and a transparent polyurethane roof, at the Experimental Station "La Teodomira", Faculty of Agricultural Engineering, Universidad Técnica de Manabí (UTM), Lodana, Manabí province, Ecuador (coordinates: 01° 09' 51" S and 80° 23' 24" W), altitude 60 meters above sea level). Inside the shade house, the average temperature was 26.25°C and the relative humidity was 79.75%. The life zone corresponds to a tropical dry forest.
Capsicum genotypes native to Ecuador (73 accessions) were evaluated, corresponding to C. annuum, C. baccatum, C. frutescens, C. pubescens and C. chinense (Syn. C. sinense Murray), from INIAP genebank originally collected from different provinces of Ecuador (Monteros-Altamirano et al. 2018). The accessions are being characterized morphologically and molecularly, whose genebank codification (numbers), respective species classifications as well as the provinces where the accessions were collected are presented in Table 1.
Seeds of each genotype were sown in 50-well germination trays (one seed per well) with peat moss where they remained for 30 days. Subsequently, plantlets were transplanted in the shade house at 1.30 m between rows and 0.70 m between plants (10 plants per genotype) in a completely randomized design. Irrigation was carried out twice a week for 15 minutes at the beginning of the cycle (first 35 days) and later for 30 minutes through a drip system with 0.02 mm tapes, located every 20 cm and a capacity of 3 L. hour -1 . Following previously established methodologies to induce populations of whiteflies through insecticide applications in tomato crops , all plants by Capsicum genotypes were sprayed with insecticides; four applications were made every 15 days starting with the transplant. The first two sprays were conducted with imidacloprid (15 days) and thiamethoxam (30 days) at doses of 2 cc.L -1 . For the other two applications chlorpyrifos (2.5 cc.L -1 ) and Thiocyclam hydrogen oxalate (1 g.L -1 ) were used, following local recommendations.
After insecticide spraying, observations were made on leaves to corroborate the establishment of B. tabaci individuals in all genotypes, which occurred approximately one month after the last spraying. Since then, the populations were monitored weekly on the leaves in each genotype until 90 days. The populations of B. tabaci were counted on four random leaves in four plants per genotype, two leaves in the upper layer and two in the middle layer of the plant. The leaves were kept in trans-parent plastic bags, labelled for each genotype, and taken to the Entomology Laboratory of the Faculty of Agricultural Engineering, Universidad Técnica de Manabí. There were observed under a 10 -40X magnification stereoscope Carl-Zeiss®️ brand, counting per leaf the number of: eggs, nymphs, and adults, of B. tabaci.
The number of eggs, nymphs and adults per leaf was analysed by ANOVA (p < 0.05). The comparison of means by genotype were evaluated with the Scott-Knott test (p < 0.05) and by species with the LSD Fischer test (p < 0.05). A dendrogram including the average of the total number of individuals of B. tabaci per leaf by genotype was plotted to establish similarity relationships based on infestations, using the unweighted arithmetic average method and Euclidean distance. The analyses were performed using the statistical software InfoStat professional version 2019 (Di Rienzo et al. 2019).
Results
In relation to number of adults, the Scott-Knott test determined three groups according to media differentiation (a, b, and c) (Fig. 1, p < 0.05).
Cp-119 91
Cb -11 993 Cf -11 99 4A C f-1 1 9 9 4 B C f-1 1 9 9 5 C s -1 1 9 9 6 C b -1 2 8 3 1 C b -1 2 8 3 3 C f -1 2 8 3 8 The first group (a) identified high number of adults of B. tabaci in two genotypes of C. baccatum (2231 and 2233 codes). The second group included 15 accessions of C. baccatum and one accession of C. annuum. The third group with the least number of adults included 25 accessions of C. sinense, 10 of C. baccatum 16 of C. frutescens and 3 of C. annuum. Regarding the higher number of eggs (range or group "a") (Fig. 2, p < 0.05) were found in 4 accessions of C. baccatum (12859,12847,12845,12842) one accession of C. pubescens (11991) and one C. baccatum (2269). The second group includes 14 accessions of C. baccatum and one accession of C. sinense. Finally, the third group with the lesser number of eggs includes 24 accessions of C. sinense, 4 of C. baccatum, 17 of C. frutescens, 3 of C. annuum and 2 of C. pubescens. Average of the number of nymphs are also separated in three groups according to the Scott-Knott test (Fig. 3, p < 0.05).
Cs -91 26 C s -9 1 1 9 Cs-91 29 The first group with the higher number of nymphs are 3 accessions of C. baccatum (12853, 12845 and 2232). The second group includes other 3 C. baccatum accessions (12847,12842,12831) and the third group with the lesser number of nymphs includes 17 accessions of C. baccatum, 25 of C. sinense, 17 of C. frutescens, 4 of C. annuum and 3 of C. pubescens (Fig. 3). When comparing the variables by species according to the differences in the degrees of significance, the number of adults, eggs, and nymphs of B. tabaci were significantly higher in C. baccatum and lower in C. frutescens and C. sinense ( Table 2).
The dendrogram constructed by genotype including average of density per leaf of all individuals of B. tabaci shows three groups with similar genotypes depending on the level of pest infestation (Fig. 4). A first group (right to left) made up of 14 genotypes that showed a high infestation, of which 13 belong to C. baccatum, one to of C. annuum. The second group is made up of 6 genotypes of C. baccatum. A third grouping includes those genotypes of Capsicum spp. with lower infestations including all genotypes of C. frutescens and C. sinense, as well as three of the four genotypes of C. annuum, two of C. pubescens and three of C. baccatum.
Discussion
Resistance of C. annuum genotypes to B. tabaci infestations has been proven in field and laboratory research carried out in different geographical areas e.g. Al-Aloosi et al. (2020) in field conditions at Iraq, evaluated a local variety and two commercial varieties of C. annuum "Anaheim chili" and "Aleppo"; the last one showed the lowest infestations by B. tabaci, whose resistance is attributed to the presence of secondary metabolites and other factors, such as the high density of trichomes, thickness and colour of the leaves. Jeevanandham et al. (2018) in a study conducted under shade house conditions, detected fewer adults and eggs of B. tabaci in 4 genotypes of C. annuum suggesting their strong antixenotic and antibiotic effects. Free-choice tests conducted inside entomological boxes evaluated the resistance of C. annuum genotypes collected in south-eastern Mexico, to B. tabaci (Ballina-Gomez et al. 2013, Chan et al. 2014. Of the twelve genotypes evaluated by Ballina-Gomez et al. (2013), three (Blanco, Bolita and Pico Paloma) showed low egg hatching and little or no survival of B. tabaci nymphs, mentioning that resistance could be associated with antibiosis due to low nutritional quality or toxic secondary metabolites. Chan et al. (2014) assessed 14 genotypes, in which the one collected in a wild habitat (Maax ik), showed the least attraction of adults and low preference for oviposition.
Besides C. annuum, resistance to B. tabaci has also been examined in genotypes of the other species such as C. frutescens, C. pubescens, and C. sinense. Pantoja et al.
(2018) in a B. tabaci oviposition preference test found a small number of adults and eggs on some accessions of C. annuum, C. frutescens, C. pubescens, and C. sinense, conferring resistance to the infection due to antixenosis. Sripontan et al. (2022) evaluated in entomological boxes resistance to B. tabaci of C. annuum, C. frutescens and C. sinense cultivars; small number of adults of B. tabaci were found in cultivars of C. annuum while those of C. frutescens and C. sinense presented a higher number of both adults and eggs. Firdaus et al. (2011) evaluated the resistance of forty-four genotypes of C. annuum, C. baccatum, C. frutescens and C. sinense species, finding a high negative correlation between the number of adults and eggs of B. tabaci and two characteristics of the leaf (the density of glandular trichomes and the thickness of the cuticle). Additionally, C. annuum was the species that developed the lowest whitefly populations. Kumar et al.
(2020) evaluated the resistance of 125 Capsicum genotypes and associated the non-preference of B. tabaci to the presence of glandular and non-glandular trichomes, as well as the flavonoids contained in the plants.
The results presented allow to identify three situations among the native Capsicum of Ecuador evaluated to the infestation of B. tabaci: One, in which a high population density of B. tabaci was observed in 76% of the C. baccatum genotypes; two, in which most of the evaluated genotypes of C. annuum and C. pubescens showed low populations; and third, all genotypes of C. frutescens and C. sinense showed low whitefly infestations.
Genetic variability has been detected in Capsicum accessions and hybrids, which present distinct morphological and molecular characteristics (Costa et al. 2016, Cardoso et al. 2018, Tripodi & Kumar 2019). This could explain the different whitefly infestation rates in genotypes of C. baccatum and C. annuum. Low infestation in C. pubescens could be associated with the pubescence of the leaves; however, it is not ruled out that the presence of volatile compounds, high concentrations of capsaicinoids in this species as well as C. frutescens and C. sinense have affected the colonization of B. tabaci populations on the evaluated accessions.
The high attraction of B. tabaci for C. baccatum genotypes, as well as the low infestation in all the genotypes of C. frutescens and C. sinense is demonstrated in this study. The non-preference of adults and the scarce oviposition of B. tabaci on genotypes of C. frutescens and C. sinense suggest resistance due to antixenosis that could later have affected the lower number of nymphs. Nevertheless, antibiotic effects of the genotypes on B. tabaci nymphs cannot be ruled out. The term antixenosis was proposed by Kogan and Ortman (1978) and defined as the resistance mechanism employed by the plant to deter colonization by an insect and may include morphological changes, such as subtle variations in plant surface colour, waxy or hairy leaves, and flavour, as well as defensive exudations of gums or resins. Tripodi and Kumar (2019) reported that an important number of accessions of cultivated and wild species are stored in genebanks around the world, which represent a valuable resource for breeding to transfer traits related to resistance to selected cultivars. Certainly, these results can be used as a basis for genetic improvement for the resistance of Capsicum species to B. tabaci.
This research shows the high attraction of the whitefly B. tabaci, towards evaluated genotypes of C. baccatum. Likewise, results indicate the non-preference of adults and the scarce oviposition of B. tabaci over genotypes of C. sinense and C. frutescens, suggesting resistance due to antixenosis. These results could guide breeding programs for the resistance to B. tabaci infestations of Capsicum species. | 2022-09-03T15:13:17.819Z | 2022-08-28T00:00:00.000 | {
"year": 2022,
"sha1": "005ae1e4a419c7f8b28c36738d8168555cae0d1d",
"oa_license": "CCBYNCSA",
"oa_url": "https://revistasinvestigacion.unmsm.edu.pe/index.php/rpb/article/download/22729/18552",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "aba6c55f4c096e2c6157d5af3b21bfd988f85db9",
"s2fieldsofstudy": [
"Biology",
"Agricultural And Food Sciences"
],
"extfieldsofstudy": []
} |
257091145 | pes2o/s2orc | v3-fos-license | An innovative green synthesis approach of chitosan nanoparticles and their inhibitory activity against phytopathogenic Botrytis cinerea on strawberry leaves
Green synthesis is a newly emerging field of nanobiotechnology that offers economic and environmental advantages over traditional chemical and physical protocols. Nontoxic, eco-friendly, and biosafe materials are used to implement sustainable processes. The current work proposes a new biological-based strategy for the biosynthesis of chitosan nanoparticles (CNPs) using Pelargonium graveolens leaves extract. The bioconversion process of CNPs was maximized using the response surface methodology. The best combination of the tested parameters that maximized the biosynthesis process was the incubation of plant extract with 1.08% chitosan at 50.38 °C for 57.53 min., yielding 9.82 ± 3 mg CNPs/mL. Investigation of CNPs by SEM, TEM, EDXS, zeta potential, FTIR, XRD, TGA, and DSC proved the bioconversion process's success. Furthermore, the antifungal activity of the biosynthesized CNPs was screened against a severe isolate of the phytopathogenic Botrytis cinerea. CNPs exerted efficient activity against the fungal growth. On strawberry leaves, 25 mg CNPs/mL reduced the symptoms of gray mold severity down to 3%. The higher concentration of CNPs (50 mg/mL) was found to have a reverse effect on the infected area compared with those of lower concentrations (12.5 and 25 mg CNPs/mL). Therefore, additional work is encouraged to reduce the harmful side effects of elevated CNPs concentrations.
. Analysis of variance for chitosan nanoparticle biosynthesis using Pelargonium graveolens leaf extract as affected by initial pH level, the incubation period (min), chitosan concentration (%), and temperature (°C). * Significant values, F: Fisher's function, P: Level of significance. www.nature.com/scientificreports/ Statistical and multiple regression analysis. To discover the significance of the BBD and its factors, both were exposed to the analysis of variance (ANOVA) and multiple regression analysis ( Table 3). The overall model generated significant performance with a P-value < 0.05. The linear, interaction, and quadratic effects followed the same significant trend. On the other hand, the lack-of-fit error recorded a higher P-value > 0.05 that did not allow it to reach the significance threshold. Another, the values of the coefficient of variation (C.V. = 1.60) and adequate precision (58.08) are other evaluating parameters that showed the good performance of the model. All these are indicators of the significance of the model and its involved factors. The general mean and standard deviation values of the investigational runs are 94.95 and 1.53, respectively. Exploring ANOVA shows the aptness of the various individuals, interactions, and quadratic effects. The three types of R 2 are adequate to suggest a high significance of both the model and its components. However, the linear, mutual interactions, or quadratic coefficient estimate showed various positive or negative significant responses, being negative for the incubation period and quadratic terms, whereas positive for the rest of the other terms (temperature and initial chitosan concentration) and all the interactions.
Source of variance
Model adequacy checking. The model adequacy was further checked to approve the appropriateness of the model. Mathematical statics were checked and described. The normal probability plot of the externally studentized residuals ( Fig. 2A) displays that data points are concerted thoroughly along the straight line and follow the normal distribution without linearity. In addition, the depiction of the Box-Cox of the power transformation ( Fig. 2B) shows that the current lambda (λ) value is equal to one. However, the best λ value (green line) was located between the confidence intervals (two red lines).
Likewise, the values of externally studentized residuals vis predicted values by the model were drawn. The plot (Fig. 2C) shows an equal scatter of the residual data above and below the 0-axis. This pattern is ideal enough to prove the suitability of the BBD model. Similarly, the predicted values were drawn against the trial values to (Fig. 2D). Again, the linear regression analysis displays better fitting of the model forecast points that lie nearer to the line of perfect prediction.
The three-dimensional (3D) surface plot. To explore the mutual influence of each pair of the tested parameters, 3D-surface plots of the simultaneous effect of the three independent factors on CNP biosynthesis using Pelargonium graveolens leaf extract were constructed. Fig. 3A-C illustrate two independent factors (at their central points) on the X-and Y-axes against the Z-axis (CNP biosynthesis). The maximum CNP biosynthesis was situated around the central points of temperature, incubation period, and chitosan concentration; out of these ranges, a marked decline in CNP biosynthesis was noticed.
Experimental validation of the model. The prime aim of the investigational strategy is to find the bestpredicted situations for obtaining the maximum CNPs. To find the best-predicted conditions for an extreme response, the desirability function was used, and the best combination of conditions was calculated based on the following prediction equation: CNP biosynthesis = 9.66 − 0.36 × incubation time + 0.10 × temperature + 0.69 × chitosan concentration + 0.94 incubation time × temperature + 0.35 incubation time × chitosan concentration + 0.54 temperature × chitosan concentration − 1.60 incubation time 2 − 1.43 temperature 2 − 2.14 chitosan concentration 2 .
According to the model's equation, this predicted value was achieved at the tested variables of 57.53 min (incubation period), 50.38 °C, and 1.08% (chitosan concentration). The highest theoretical (predicted) value of CNP biosynthesis was estimated to be 9.73 mg/mL (Fig. 3D). Such optimum points of the tested levels recorded a satisfactory desirability value (1.0).
To verify the green synthesis of CNPs by using P. graveolens leaf extract under the optimal predicted conditions, a triplicate set of experiments was carried out, and the experimental results were judged to the predicted values. The actual laboratory value of the green synthesis of CNPs by using P. graveolens leaf extract was 9.82 ± 3 mg/mL, verifying a high degree of model precision and confirming the validation of the model under the design matrix.
Electron microscopy investigation. The surface morphological structure was examined using scanning electron microscopy (SEM). The size, morphology, and structure were inspected at various magnification scales (Fig. 4A,B). The SEM image of the morphological construction of CNPs shows spherical-like particles, and their mild agglomerated state revealed a highly porous surface. Furthermore, the size of the nanoparticles also showed uniformity and homogeneity. www.nature.com/scientificreports/ Another efficient tool for discovering the morphological structure of CNPs is transmission electron microscopy (TEM), which was performed at various magnification powers (Fig. 4C,D). TEM derives additional details, such as a particle's aggregation and agglomeration. The two-dimensional (2D) TEM image introduced a wider conception than SEM, especially regarding the non-aggregation and lower-agglomeration status of CNPs with a highly porous surface. Furthermore, the CNP size analysis showed a measurement from 6.02 to 10.87 nm. It is of particular importance to note that both SEM and TEM were found to be complementary to the characterization of CNPs.
Energy-dispersive X-ray spectroscopy (EDXS) analysis. For further characterization, CNPs were explored for their elemental compositions using EDXS (Fig. 5A), which can quickly generate information about the kinds of elements, distributions, and concentrations. The EDXS spectra of CNPs confirmed the presence of the four elements that chitosan is composed of hydrogen, carbon, nitrogen, and oxygen. Such a composition represents the main elemental component of chitosan.
Zeta (ζ) potential analysis. The ζ-potential analysis of the synthesized CNPs was investigated; nevertheless, ζ-potential analysis is often the only available route for the description of double-layer properties. The depicted ζ-potential (Fig. 5B) shows decent stability of the positively charged CNPs with a ζ-potential of + 32.6 ± 5.26 mV at 25 °C. Furthermore, the ζ-potential potential distribution has a single peak, indicating excellent uniformity of CNPs.
Fourier transform infrared (FTIR) investigation. CNPs were analyzed using FTIR (Fig. 6A) to explore the biomolecules for the possible occurrence of various functional groups that bind with CNPs due to reduction and stabilization. The detected intense bands were compared with standard values to classify the functional groups. The adsorbent spectra were measured in the range between 4500 and 500 cm −1 . The characteristics of CNPs were shown by a broad absorption band. The FTIR spectra show absorption bands at 3736, 3442, 2350, 1572, 1413, 1072, 914, and 645 cm −1 .
X-ray diffraction (XRD) analysis. The crystallographic structure of CNPs was investigated using XRD ( Fig. 6B). After irradiating CNPs with incident X-rays, the intensities and scattering angles of the X-rays that left the CNPs were measured. XRD of CNPs showed three peaks at 2-theta of 13, 19, and 35°. , drying as a function of temperature can easily be seen as a quick initial mass dropping (− 23.118%). Then, the weight loss of CNPs showed multistage decomposition with increasing temperature but with lower weight losses. The lowest weight loss was recorded at a temperature range of 483.38-613.34 °C, which was 2.461%. Although CNPs have reasonable heat constancy, at the last temperature stage (704.92-799.91 °C), the nanoparticles kept a reasonable amount, i.e., ̴ 20%, of their mass. Investigation of the thermo analytical technique of differential scanning calorimetry (DSC) was applied to CNPs at different heating rates to determine the variation (positive or negative) in the amount of heat flow of CNPs as a function of temperature in the presence of a solvent reference. Both the CNPs and reference were sustained at the same temperature throughout the experiment. Thermo analytical information was gathered to create a phase diagram (Fig. 7B). As the thermodynamic system is changed, the CNPs undergo phase transitions, showing two definite broad endothermic peaks. The first broad endothermic peak was detected between 90 and 143 °C at 122 °C, requiring a heat amount of − 358.87 J/g CNPs. The other broad endothermic peak appeared at 242 °C, requiring − 64.18 J/g CNPs. A single glass transition temperature of the CNPs was found at 180 °C.
Isolation and identification of B. cinerea. Following the isolation and purification trials, the most severe isolate was selected from several isolates based on the pathogenicity test. The colony characteristics of B. cinerea SIB-1 were observed on PDA plates, on which the fungus developed and sporulated on the culture medium to cover the majority of the 9-cm plate during six days. The growth pattern on the plates showed the colonies as warty, fluffy, and appressed, grayish-white to light gray or dark gray. The conidia produced on the surface of the medium, in abundance all over the plates, can also be seen as concentric rings.
Data extracted from the SEM investigation ( Fig. 8) show the conidial clusters carried on conidiophores and the branched mycelium, with conidiophores arising directly from the mycelium. They were more or less straight, monopodial branched toward the apex. The conidia were solitary and arranged in clusters or masses. The forms of the conidia observed were ellipsoidal and globose, and they were smooth, often with a slightly protuberant hilum and unicellular. The conidia measured 13.97 ± 2.04 μm in length and 9.53 ± 1.45 in width. S1) The amplicon was sequenced and compared to the NCBI public genome map using the BLAST program. Phylogenetic analysis using the neighbor-joining method was performed ( Fig. 9B) to determine the relatedness between the sequenced gene and its homologs in other organisms. The created phylogeny is totally annotated and displays a tight correlation with those of comparable strains. Multiple nucleotide sequence alignment showed that the B. cinerea SIB-1 isolate was closely related to B. cinerea strain RGM 2565 (K934584). Accordingly, the current strain was banked in the NCBI GenBank with the accession number MZ570270.
Antifungal activity of CNPs. The in vitro antifungal activity of CNPs and chitosan ( Fig. 10 and Table 4) revealed an obvious reduction in the growth of B. cinerea SIB-1 on potato dextrose agar plates, but the greatest fungal inhibition was the result of CNPs treatments. CNPs inhibited fungal growth by 73.58% at 1.0 mg/mL, while chitosan showed only 31.16% growth inhibition under the same condition. In comparison, the recommended dose of Ridomil Gold (positive control) caused about 50% inhibition of the fungal growth.
To investigate the protective roles of CNPs against gray mold disease developed by B. cinerea SIB-1, different concentrations of the CNPs were tested on detached strawberry leaves (Fig. 11). The average percent of leaf area infected was determined for each concentration. In general, after five days, leaves treated with the CNPs had a lower disease severity compared with the control, which had a mean infected area of 85% with a severity class value of 4. The lowest disease severity (class 0) was observed at the concentration of 25 mg/mL, which only showed a 3% infected area. However, CNPs at 12.5 and 50 mg/L recorded leaf area infections of 27 and 23% with severity class values of 2 and 1, respectively. It was found that the disease dramatically progressed over time from one to five days on both the adaxial and abaxial sides.
Discussion
Nanotechnology has had a significant impact on several high-tech businesses in recent years, and it has been demonstrated that it has an impact on many microbial species, as well 1 .
There are several advantages of nanoparticles over the bulk form. Biologically, owing to the tiny size, nanoparticles straightforwardly penetrate and are easily taken up by the cell, which permits proficient accretion at the target site in the organism. Moreover, the retention of the nanoparticles at the target site has a longer clearance www.nature.com/scientificreports/ time, leading to an increase in therapeutic stability, bioavailability, and efficiency compared to the same dosage of the non-nanoparticulate form 7,18 . Nanoparticles can be generated utilizing an assortment of strategies, including physical (ball milling, ultrathin films, spray pyrolysis, thermal evaporation, plasma arcing, lithographic procedures, pulsed laser desorption, layer-by-layer growth, sputter deposition, molecular beam epitasis, and diffusion flame synthesis of nanoparticles) and chemical (chemical solution deposition, electrodeposition, chemical vapor deposition, sol-gel process, soft chemical method, catalytic route, wet chemical procedure, hydrolysis coprecipitation, and Langmuir-Blodgett) methods, as well as hybrid techniques. These methods use high radiation and highly concentrated reductants and stabilizing agents that are destructive to human health and the environment as well 7,[18][19][20] .
Alternatively, the biological-based procedure or green manufacture of nanoparticles is a bioreduction process with lower energy requirements. The technique is environmentally friendly and nontoxic, with greater stability, and nanoparticles are biosynthesized by applying a single-step process 18 . Moreover, to the best of the authors' knowledge, all phytofabrication of nanoparticles is mostly restricted to metal ions, and no previous reports on CNPs phytofabrication.Many examinations have demonstrated that plant extracts act as potential precursors for the biosynthesis of nanoparticles in nondangerous manners. Therefore, plants are utilized effectively and economically in the biosynthesis of several metal-nanoparticles 7,10 .
The universal procedure of metallic nanoparticle biosynthesis employs the plant as a bioreducing agent and metallic salt as a precursor, resulting in biocompatible and stable nanoparticles. This promising route of nanoparticle production using a biological system utilizes three main approaches, i.e., 1) the selection of solvent intermediate, 2) the choice of an ecological, benign reducing agent, and the choice of a nontoxic material as a www.nature.com/scientificreports/ capping agent to stabilize the biosynthesized nanomaterials 21,22 . Additionally, twelve well-known green chemistry principles have now become a reference guide for developing less hazardous chemical products 23 .
All previous approaches were found to be achieved in the selected plant (Pelargonium graveolens) and the transformed polymer (chitosan), additionally, the 12-green chemistry principles were strongly applied in the current work. To the best of the authors' knowledge, no previous work has reported on the synthesis of chitosan polymers into nanoforms using Pelargonium graveolens plants. Accordingly, the current paper describes a novel protocol for the green biosynthesis of CNPs, employing the leaf extract of Pelargonium graveolens. This procedure offers several merits over ordinary fabrication procedures.
Before proceeding to maximize the biosynthesis of CNPs, a primary characterization test based on UV/ visible spectra was applied to ensure the development of the nanoparticles. The current absorption peak wavelength was detected at 295 nm; this result is in harmony with that previously reported at 285 nm 24 and 320 nm in the UV region 2 . Compared with CNPs, the UV/visible spectrum of chitosan showed a wider absorption band intensity; therefore, the sharp intensity level of the CNP biopolymer indicates the success of the phytofabrication of CNPs 2,24 .
Exploring the experimental data of CNPs of the design matrix shows the highest level of CNPs located at the middle levels of the three tested variables, indicating the accuracy of both selected independent variables and their tested levels. It was obvious that the predicted values of CNPs were very adjacent to those of the trial CNPs; consequently, the residuals or error values were at their minimum, signifying another proven accuracy of the investigated parameters and their levels.
For the selection of the most appropriate model, the effect of each model term was screened at the P level of 0.05. The tested terms (factors) that had lower P values were considered significant and reliable for the modeling process. Regarding the model selection, R 2 is considered a very important selection criterion; if R 2 is higher than 0.9, the regression model is defined as very significant, and the model is adequate; however, the R 2 value should not be lower than 0.75 25 . The current quadratic model had high R 2 , adjusted R 2 , and predicted R 2 values, which were very close to one. Consequently, the quadratic model was the best-fitted model. R 2 is defined as the amount of change in the observed (experimental) response (CNPs) that is described by the three tested factors. Generally, R 2 can help choose the best-fitted model. All types of R 2 range from zero to 1. The closer to 1, the better the modeling of the experimental data. Interestingly, increasing the number of factors (predictors) leads to a continuous increase in the R 2 value, irrespective of the significance of the factors. Therefore, the adjusted R 2 is an improved R 2 that considers the number of factors (variables) in the model. Contrary to R 2 , the adjusted R 2 may be reduced with the addition of extra terms (factors) to the model. Therefore, the adjusted www.nature.com/scientificreports/ R 2 is a better indicator than R 2 to judge how well the model fits the data. The predicted R 2 is used to determine the degree of the predictive capability of the model, e.g., to predict the value of the CNP response at new levels of the tested factors. Moreover, it is more beneficial than the adjusted R 2 for comparing and selecting the models. Therefore, the quadratic model was selected as a modeling base in CNP bioprocessing. The BBD data were subjected to ANOVA. The model exhibited a high F-value and low P-value; additionally, the lack-of-fit was not nonsignificant, indicating the significance of the proposed overall model. Moreover, the ratio of adequate precision is higher than 4, which is a suitable indicator that this model can be successfully employed to work within the tested range of the various tested factors along with the design space to maximize CNP biosynthesis. Another precision and trustiness of the experimental design can be noted by the lower value of the C.V. and greater value of adequate precision, which are desirable for the reliability of the model.
The weight of every individual factor was diagnosed, and the P-value was again utilized. However, the values of P indicate that the model terms are significant (< 0.05), indicating that they are important phytofabrication parameters of CNPs. This also suggests that the variables and their established levels, as well as the investigational design, are well defined and attain the peak performance of CNP phytofabrication. Therefore, the projection model was created based on such proven terms.
Data of ANOVA displays that the predicted R 2 and adjusted R 2 are close to each other. The values of both kinds of R 2 should be less than 20% of each other to be in decent agreement 25 . In the current investigation, the predicted R 2 was in harmony with the adjusted R 2 value, indicating high compatibility between the predicted and experimental values of CNP biosynthesis and indicating the satisfactory predictive ability of the model within the examined range. Some of the model terms showed a negative coefficient estimate value, which indicates that such a variable has an antagonistic effect on CNP biosynthesis by P. graveolens at higher concentrations. The positive coefficient value, on the other hand, indicates a cooperative effect, and the variable(s) increase CNP biosynthesis with the continuous increment of the level of the investigated factor within the region of the experiment.
Model adequacy was checked. The normal plot of residuals was plotted to check the externally studentized residuals versus normal probability (%). Values show an equal distribution of the residual data, indicating that the variance of CNP biosynthesis was independent of the biosynthesis process, thus supporting the adequacy of the model. Moreover, residuals were found to be very low at all tested points. This implies that the model can fit the actual experimental data faithfully. Additionally, the model prediction points vis the actual points lie much nearer to the line of perfect prediction. Thus, the model has a significant generalization capacity for CNP biosynthesis. Moreover, the Box-Cox plot of the model transformation of chitosan nanoparticle biosynthesis using P. graveolens leaf extract confirms the suitability of the design and data. The value of l concluded no recommendation for data transformation in this model. Consequently, these two adequacy tests authorize the aptness of the design and obtained data. Model adequacy was checked by plotting the normal probability of the externally studentized residuals. Most of the data points aggregated thoroughly around the straight line and were, thus, considered normally distributed without linearity. No value is located away from the general mean since the The pattern also displays a normal distribution, supporting the adequacy of the model. Analysis of the 3D plots demonstrates that all the pair tested factors generated a peak of CNP biosynthesis around the center point of the design space, meaning that the tested ranges of the three factors were carefully selected, and the model best fit the design.
To find out the best-predicted combinations that maximize CNPs, the desirability function was used, whose value ranges from undesirable (zero) to desirable (one). As the response approaches the goal, the desirability value becomes closer to 1. The desirability value is generally estimated as a mathematical evaluation of the optimization process before experimental validation 26 . Accordingly, the optimum levels of the incubation period, temperature, and initial chitosan concentration that maximize CNP biosynthesis by P. graveolens were estimated by solving the prediction equation. Among several options, the best solution was selected based on the desirability value. The current desirability value was sufficiently high since it reached the peak that validate the optimization process. Then, the predicted amount of CNPs was estimated, which was found to have a high intensity of agreement between the experimental values, suggesting that the desirability function effectively ascertains the best-predicted situations for the green synthesis of CNPs by using P. graveolens leaf extract.
The theoretical estimation of the polynomial model is an estimate based on a reasonably studied area of the tested independent variables; therefore, the guarantee of the real prediction effectiveness of the equation under real conditions is critical. However, the theoretical value of CNP biosynthesis was valid, since it showed close similarity to that of the experimental one. That is, in turn, produces strong evidence for the fitness of the design and the modeling process, utilizing the tested ranges of the studied variables.
Following the optimization conditions of the biofabrication process of CNPs from chitosan by P. graveolens leaf extract, the surface morphological structure of the obtained CNPs was monitored by SEM, which are widely considered the main accepted procedures for the characterization of nanoparticles. These techniques are erroneously used interchangeably, but in reality, they vary substantially. However, both techniques provide some similarities but a distinct analysis, which is why the accurate interpretation of their images is essential. Collectively, SEM and TEM offer powerful tools for the investigation of size, shape, surface area, crystal structure, and morphological structure. Although TEM systems can bring much greater 2D resolution for size analysis, SEM provides accurate information about the 3D surface and shape features 27,28 .
The 3D SEM image of CNPs exhibits a good dispersion of the nanoparticles, which are entangled to form a larger exposed surface area, making the CNPs very appropriate for adsorption 29 . Like the current phytofabricated CNPs, most of the CNPs prepared from chitosan were spherical in shape 30 , and few had oval pleated 31 or rod-shaped structure 29,32 .
The 2D TEM image undoubtedly indicates that the CNPs show a highly porous surface owing to low agglomeration attributes. These porous and agglomerated CNPs have been considered key phenomena for the synthesis of novel CNPs, hence maximizing their usefulness as antibiological phytosynthesized nanomaterials in biomedical and agricultural applications, where the porous nature can effectively adsorb harmful chemicals and antagonize the pathogens 2,33 . In contrast to bulk materials, which have lower porosity, nanoparticles with high porosity have a greater specific exterior area and high reaction activity 1 .
The terms agglomeration and aggregation are repeatedly used interchangeably, but they definitely differ, where agglomeration indicates more weakly bonded particles and aggregation indicates strongly bonded or fused particles. In our case, the CNPs showed low agglomeration without aggregation, and the low agglomeration phenomenon is accepted since many nanoparticle types have high ionic strength and agglomerate in aqueous matrices, such as in phosphate-buffered saline and cell culture medium 34,35 .
EDXS study was used together with electron microscopy investigation to analyze the component elements of CNPs. When the electron beam of SEM hits the inner shell of an element atom, its inner-shell electron is relocated by another electron from an outer shell to fill the vacancy, and the process is accompanied by the release of an energy difference in the form of an X-ray that is unique to the specific element. Moreover, the intensity of the specific X-ray is directly related to the concentration of the element in the particles 36 . However, the test confirms the presence of the various elemental compositions of native chitosan, confirming the uniformity and stability of CNPs during the biotransformation process.
The ζ-potential is an indicator of the stability of colloidal dispersions. The weight of the ζ-potential specifies the degree of electrostatic repulsion among similarly charged adjacent particles 37 . For tiny particles and molecules, a high ζ-potential confers stability and resists aggregation of nanoparticles in the solution or dispersion. In contrast, at small ζ-potentials, attractive forces may exceed, leading to flocculation owing to the breakdown of the dispersion. Therefore, colloids with high negative or positive ζ-potentials are more electrically stabilized than those with low ζ-potentials, which tend to coagulate or flocculate 38 . The current ζ-potential value of CNPs suggests nanoparticles with good stability. The CNPs were positively charged. From an antimicrobial point of view, when ζ-potential is positive, particles can easily interact with the negatively charged cell membrane and/ or DNA of a biological system and can then be released simply into the cytoplasm of the cell 39 .
Regarding FTIR, the presence of various intense bands indicates the presence of a capping agent, which acts as a stabilizer that inhibits the overgrowth of nanoparticles and prevents their aggregation and/or coagulation in colloidal synthesis. Therefore, the observed intense bands were matched with standard values to classify the functional groups. The first range of bands that appeared in the spectra was due to stretching vibrations of OH groups at wavenumbers ranging from 3736 to 3442 cm −1 , indicating the presence of alcohols and phenols. The stretching vibration of methylene (C=H) was at 2350 cm −1 . It is also already known that the band at 2350 cm −1 generally arises from the background CO 2 40 and has no corresponding group associated with the chitosan structure. Absorption at wavenumber 1572 was correlated to the vibrations of carbonyl bonds (C=O stretching) of the amide group CONHR or protonated amine (NH 2 . Bending vibrations of the methyl group (C-H bending, alkane) of CNPs were visible at 1413 cm −1 . Absorption in the wavenumber range 1072 and 914 cm −1 is generated www.nature.com/scientificreports/ from the stretch vibration of CO groups (COH and COC) in the oxygen bridge, emerging from the deacetylation of chitosan. At the end of the FTIR spectra, the small peak at 645 cm −1 corresponds to the wagging of the saccharide structure of chitosan 41 . FTIR analysis strongly emphasizes the structural stability of chitosan during phytoconversion into CNPs. XRD analysis is a fast practice, primarily used in materials science for the phase identification of a crystalline nature and can deliver information on unit cell dimensions; thus, the XRD pattern is considered the fingerprint of periodic atomic arrangements in a given material 32 . The XRD of the current CNPs showed three peaks at 2-theta of 13, 19, and 35°. It is conventionally accepted that chitosan stretches two characteristic peaks at 2-theta of 10 and 20°, and the current shift that occurred to CNPs from the normal chitosan peaks indicates the amplified amorphous nature, thus lessening the crystal structure of chitosan, which comes in line with studies that focused on decreasing the crystallinity for improving the sorption properties of the materials 32,42 .
Both TGA and DSC investigations were performed, and both are measures of the thermo analytical features used to describe the analysis of nanoparticles that take part in chemical reactions over a controlled temperature range. TGA measures the differential thermal analysis in terms of the change in mass of the sample in relation to temperature changes or as a response of time with constant temperature and/or constant mass loss. DSC, on the other hand, measures the heat flow released or required against the temperature change at a particular time. The main dissimilarity between TGA and DSC is the method of measuring the changes in samples that are triggered by heat 29,31,32,43 .
At the beginning of TGA, common drying as a function of temperature can easily cause a quick initial drop in CNP mass due to the loss of residual water bound to the two polar groups in CNPs, which is not known to correspond to any chemical reactions 32 . Another reason for the drop in weight at the beginning step may be due to the dehydration of the saccharide rings, depolymerization, and decomposition of volatile products 29 . Next, successive weight losses in CNPs with increasing temperature may be due to evaporation and/or sublimation; however, multistage decomposition shown as a step-like pattern is due to the thermal degradation of CNPs 43 .
The loss of weight in the next stages of thermal analysis may be due to the decomposition of the polymer matrix; however, the CNPs did not fully decompose at the high temperature (800 °C) and conversely showed some stability in the polysaccharide structure. This result indicates that CNPs are thermally stable over a temperature up to 800 °C, which may be due to the high crosslinking of the CNPs that forms a stronger and stiffer hydrogel network 29,32,43 .
However, the thermal analysis steps were not blended during dynamic TGA; however, there is still a possibility of hidden interference of the decomposition steps, necessitating either far slower heating rates or stepwise TGA methods. That is why TGA itself may not be sufficient to identify the decomposition products, therefore chemical testing such as DSC is often, required alongside TGA, to ascertain the identities of suspected decomposition products 31 .
DSC of CNPs was performed to highlight phase transitions and clarify every single step of the thermal degradation mechanism. Usually, the temperature program for a DSC analysis is designed such that the sample holder temperature increases linearly as a function of time. That, in turn, can offer pieces of evidence about physical phenomena, such as glass transition, thermal stability, and purity 31 . Two definite broad endothermic peaks were generated, the first at a lower temperature that was due to the removal of absorbed water. The second endothermic peak that appeared at 242 °C is generally associated with the breakage of cross-linkage of CNPs. Additionally, the higher value of the glass transition temperature is due to the presence of a crosslinking agent and high thermal stability. Only a single glass transition in the DSC heating curves indicates the uniformity of the CNPs under high temperatures 29 . The transformation of chitosan into CNPs decreases the crystallinity due to changes in the solid-state structure of chitosan due to crosslinking, and thus, the decomposition of CNPs occurs above 300°C 32 .
Interestingly, in the present study, all proceeding investigations came in harmony with each other to provide an accurate perspective for the characterization of CNPs. Moreover, the currently proposed phytofabrication method for CNP preparation is considered ideal for generating high-quality CNPs.
The developed phytosynthesized CNPs were monitored regarding their antifungal properties. Depending on the molecular weight, concentration, degree of substitution, and the type of functional groups on chitosan, the free chitosan polymer exhibits various antifungal activities against a wide array of fungi. Derivatives of the polymer can be formed to target specific pathogens. Chitosan shows natural antifungal capability without the necessity for any chemical alterations 2 , which is why the generated CNPs were tested against the isolated B. cinerea SIB-1.
Koch's postulates were applied to the isolated fungi to confirm the pathogenic ability and to select the most aggressive isolate as well. The colony characteristics of the isolated fungus and SEM investigation showed the typical features of the already known phytopathogenic B. cinerea SIB-1. These features are in line with those previously described 17 .
A severe phytopathogen (Botrytis cinerea SIB-1) was used in this study as a model for the evaluation of CNPs as an anti-biological agent. The main reason for selecting such phytopathogen is the wide host range since it can infect more than 200 host plants, and it can infect several parts of the plants, including the upper parts such as seeds, leaves, bulbs, and other propagation material at pre-and postharvest stages. Moreover, Botrytis spp. infect the host plant in all climate areas of the world and under great humidity in the presence or absence of water films. The fungus can generate high numbers of conidia that pose a long-lasting threat to susceptible hosts; in addition, the genotypic and phenotypic variation of the fungus is another broad-spectrum thread for the plant production sector 16,17 . Importantly, the fast alterations in populations and resistance in response to exposure to xenobiotics, e.g., fungicides, are quite widespread in the genus 44 , urging the discovery of alternative commercial approaches of considerable disease suppression to be integrated into crop management protocols 15 .
The selected fungus was identified as B. cinerea SIB-1 on a molecular basis, which is sensitive and specific for the rapid recognition of filamentous fungi at different systematic levels. The ITS region is usually used and www.nature.com/scientificreports/ can be adequate for fungal identification at the species level. The ITS region is also contemplated among the markers with the fastest and uppermost probability of precise identifications for a very broad group of fungi 45 . Interestingly, the culture morphology, SEM investigation, and molecular identification computably confirmed the fungus to be B. cinerea SIB-1. Gray mold caused by the phytopathogenic Botrytis cinerea, is a serious disease that affects all strawberry growing regions and is the main cause of concern most years. The gray mold disease is a problem not only in the field, but also during storage, transportation, and marketing of strawberries as a result of severe rot as the fruits begin to ripen. Leafs, fruit caps, flower stalks, petals, and crowns are among the other parts infected by the fungus. CNPs showed strong inhibition against B. cinerea SIB-1. Possible protocols include inhibition of mycelial growth and sporogenesis 3 .
There are three mechanisms proposed as the inhibition mode of chitosan. In the first mechanism, the cell membrane of fungi is the main target of chitosan. The inhibitory effect of chitosan may be related to its interaction with the cell membrane of the fungal cell and alteration of membrane permeability 46 . The positive charge of chitosan allows it to interact with phospholipid components of the fungal cell membrane that are negatively charged. This increases membrane permeability, allowing cellular contents to leak out, ultimately resulting in cell death. The second mechanism involves chitosan acting as a chelating agent, binding to trace elements and rendering them unavailable to fungi for normal growth. Finally, the third mechanism proposed that chitosan could pass through fungi's cell walls and bind to their DNA or proteins. This will stop the production of essential proteins and enzymes by inhibiting the synthesis of mRNA 3 . Chitosan inhibited mycelia growth, sporulation, and spore germination. It induces morphological changes characterized by excessive branching, mycelial swelling, agglomeration of hyphae, abnormal shapes, dissolution of protoplasm, large vesicles, cytoplasm aggregation, or empty cells devoid of cytoplasm in the mycelium 47 .
Ridomil Gold is a systematic fungicide that inhibits fungal development by interfering with the biosynthesis of sterols in the cell membrane. Thus, provides excellent disease control, ensures double protection to the target plants due to systemic activity of Mefenoxam fungicide and contact protective activity of mancozeb fungicide. Mefenoxam penetrates rapidly into the plant through the leaves and stems and is distributed upwards with the flow of sap. This way, new growth is protected as well. Mancozeb provides a protective film on the surface of the plant and inhibits germination of the spores and controls leaf and tuber blight as well as leaf spot disease.
The current experimental results on strawberry leaves show that treatment with CNPs reduced infection lesions. A key factor for a pathogenic fungus to be able to magnificently infect plants is to secrete a category of effector proteins into plant cells, which makes plants more susceptible to diseases 48 . These effector proteins decreased after treatment with chitosan 5 ; moreover, chitosan can stimulate defense-related enzymes and augment the accumulation of antimicrobial ingredients in the infected plant, mainly diminishing the success rate of the fungal infection and inducing plant resistance 49 .
It is of essential importance to note that leaf treatment with a high concentration of CNPs (50 mg/mL) was found to have a reverse effect on the infected area compared with those of lower concentrations. These results suggested that increasing the concentration of nanoparticles might result in the crowding of these particles on the leaf pores that limit their penetration into the inner tissues. Furthermore, several findings concluded the potential phytotoxicity of these nanoparticles at high concentrations. In this respect, the application of CNPs at a concentration higher than the optimum causes a reduction in the mineral and nitrogen contents of the coffee leaves 50 . Similarly, high concentrations of chitosan nanoparticles markedly reduce the growth and development of Capsicum annuum, while lower concentrations have a growth-promoting effect 51 . These results suggest additional investigation on the optimum concentration of CNPs and further indicate the need for caution when using CNPs to reduce the harmful side effects of elevated concentrations.
Materials and methods
Formulation of plant extract. Fresh leaves of Pelargonium graveolens were collected from the El-Natrun area (30° 22′ 39″ N latitude and 30° 21′ 1.08″ E longitude), Northern West Nile Delta, 62 miles from Cairo, Egypt. The leaves were washed thoroughly three times with tap water and then with distilled water to eliminate any impurities. Finally, 25 g of thoroughly washed and finely cut leaves were dipped into a conical flask containing 100 mL of distilled water, mixed, and boiled for 10 min. After boiling, the solution was then filtered through filter paper (Whatman No. 1). The filtered extract of Pelargonium graveolens was collected and used for the biosynthesis of CNPs.
Phyto-synthesis procedure of CNPs. Chitosan, obtained from Bio Basic Inc., Toronto, Canada, was liquefied at 1% (w/v) with acetic acid (1%, v/v), and the pH was raised to 5 with 1 N NaOH and kept under magnetic stirring for 24 h. Equal amounts (10 mL) from each of the plant extracts and chitosan solution were blended and incubated at 50 °C under shaking for 30 min. After incubation, the turbidity was centrifuged at 10,000×g for 10 min. To remove the unreacted chitosan from the produced nanoparticles, it was washed several times with acetic acid solution. The precipitate was then obtained by centrifugation of the mixture at 10,000×g for 10 min. The resultant CNPs were then freeze-dried. The resultant CNPs were redissolved again in 1% acetic acid, and the UV/visible absorbance spectrum of the prepared CNPs was determined by a double beam spectrophotometer at 295 nm. A standard calibration curve prepared from known concentrations of CNPs was prepared in 1% acetic acid at various serial dilutions and was used to estimate the final concentration of CNPs (mg/mL).
UV-Visible spectrum. The generated CNPs were monitored by detecting the peak absorbance at an array between 200-400 nm utilizing an Optizen Pop-UV/Vis spectrophotometer. www.nature.com/scientificreports/ Optimization of CNPs biosynthesis by BBD. The matrix of BBD was generated to establish the best levels of the chosen process variables affecting chitosan nanoparticle biosynthesis using P. graveolens leaf extract and to study the individual and interaction effects between these variables. The factors were incubation period (X 1 , 30-90 min), temperature (X 2 , 40-60 °C), and chitosan concentration (X 3 , 0.5-1.5%). Each process variable ranges at three levels (− 1, 0, + 1), with five central points, resulting in a total of seventeen trials ( Table 1). The linear, interaction, and quadratic influences of the three selected process variables affecting chitosan nanoparticle biosynthesis were determined to explore the relationship between chitosan nanoparticle biosynthesis (Y) and the optimum level of each of the tested variables. The next second-order polynomial function was applied: where Y is CNP biosynthesis using P. graveolens leaf extract, X i is the process variable at the coded levels, β 0 is the regression coefficient, β i is the linear coefficient, β ij is the interaction coefficient and β ii is the quadratic coefficient. The investigational design and statistical examination were conducted using the Windows edition of Design-Expert software (version 12, Stat-Ease, Minneapolis, USA). The STATISTICA version 8 program was used to draw 3D surface plots. FTIR spectroscopy analysis. FTIR spectroscopy assessment was conducted to analyze the surface properties of the CNPs, which were ground with KBr pellets used for FTIR measurements. The FTIR spectrum of CNPS was measured using a Shimadzu FTIR-8400 S spectrophotometer in the range of 4500-500 cm −1 at a resolution of 1 cm −1 .
Characterization of
XRD pattern. XRD is a crucial technique for determining the structural properties of nanoparticles. The XRD patterns were recorded using a diffractometer (Bruker D2 Phaser 2nd Gen). The X-ray source was operated with a Cu anode with a voltage of 30 kV and a current of 10 mA. Diffraction intensity was measured at 25.7 °C and a scanning rate of 2°/min for 2θ = 10-50 53 .
TGA of CNPs. The CNP sample was withered at 60 °C for 1 h and mounted in a platinum sample pan. TGA was accomplished using a thermoanalyzer of type 50-H. TGA was operated in the range from 25 to 800 °C, with a 20 mL min -1 flow rate, under a nitrogen atmosphere at an increment of 10 °C min −1 . The chart was drawn as temperature versus weight loss percentage. DSC analysis. DSC (60-A) was used to assess the CNP pyrolysis pattern. The CNP sample was dried at 60 °C for 1 h and mounted in an aluminum sample pan. The examination was operated under nitrogen atmosphere conditions with a heating rate of 10 °C min −1 and a flow rate of 30 mL min -1 . The thermogram was generated between 25 to 500 °C. The initial breakdown temperature for the CNPs, as presented by thermogravimetric analysis, was chosen for the DSC upper limit. The graph was mapped as heat flow versus temperature.
Isolation and identification of the fungal pathogen. The model fungus used in this study was B. cinerea, a causative agent of gray mold. The fungus was isolated from blighted fruits of strawberry (Fragaria virginiana, var. Fortuna) collected from open fields in the Menoufia government (30°18′ 02″ N latitude 31° 00′ 30″ E longitude) during March 2021. The collected samples were kept at 25 °C and 90% relative humidity for 4 days in a glass box to facilitate fungal sporulation. The purified culture was obtained by a single-spore technique on PDA medium, purified, and then stored (4 °C) until further use. Primary morphological identification was carried out with the aid of the Miclea et al. 17 method. One pathogenic isolate of B. cinerea SIB-1 was selected based on the pathogenicity test. The most severe isolate was selected, and further confirmational identification methods were used. The morphology of the Botrytis cinerea SIB-1 was examined on potato dextrose agar plates after 7 days at 30 °C. The gold-coated dehydrated specimen was examined at different magnifications with Analytical Scanning Elec- Genomic DNA of B. cinerea SIB-1 was utilized as a template for PCR amplification of the ITS regions using primers ITS1 (5'-TCT GTA GGT GAA CCT GCG G-3') and ITS4 (5'-TCC TCC GCT TAT TGA TAT GC-3') described by White et al. 55 . The specific primers were obtained from Clinilab (Clinilab Analysis Co., Egypt). The reaction mixture (25 µl) contained 5 µl of Taq red buffer, 20 ng of template DNA, 10 pmol of each primer, 0.25 U of Taq polymerase (Bioline, UK). PCR cycling conditions were as follows: 95 °C for 2 min, 30 cycles of 95 °C for 30 s; 50 °C for 30 min; 72 °C for 1 min; and 1 cycle of 72 °C for 5 min. The PCR product was checked on a 1.5% agarose gel stained with 0.05 μg/mL ethidium bromide, and the target band was purified with a GF-1 PCR clean-up kit according to the manufacturer's guidelines.
The ITS sequence was computationally evaluated using the BLASTn program (http:// www. ncbi. nlm. nih. gov/ BLAST). Sequences were aligned using Align Sequences Nucleotide BLAST. The obtained sequence was banked in GenBank to attain strictly related sequences; then, the accession number of the fungal strain was established. The evolutionary relationship was inferred, and the phylogenetic tree was generated using MEGA 10 software.
Antifungal activity of CNPs. Two tests were performed to investigate the in vitro antifungal activity of CNPs against the gray mold pathogen B. cinerea SIB-1. The first test was performed on agar plates containing PDA medium. The pH of the culture medium was adjusted prior to autoclaving using 1 M NaOH to be 6.5. So that, after sterilization and addition of CNPs or chitosan, the pH falls within the specified range (found to range from 5 to 5.5).
Known amounts of CNPs were dissolved in acetic acid solution (1%) with stirring overnight under aseptic conditions. The nanomaterial was added to PDA to ultimate concentrations of 0.5 and 1.0 mg/mL. To prepare the fungal inoculum, B. cinerea SIB-1 mycelia were taken from a single colony and transferred to potato dextrose agar plates. The inoculated plates were incubated at 28 °C for 7 days. A fungal disk (5 mm) was taken from the active growing margin and aseptically transferred to the center of the plates. The inoculated plates were checked daily until the control (zero CNPs) treatment touched the edge of the plate.
The in vitro antifungal activity of chitosan against B. cinerea SIB-1 was also investigated. On the other hand, the antifungal activity has been compared with the commercial Ridomil Gold fungicide on the mycelium growth of B. cinerea SIB-1 under the same conditions. Ridomil Gold is a leading fungicide recommended for controlling many important diseases on many crops. It is a mixture of both systemic (Mefenoxam) and contact (Mancozeb) fungicides to control certain diseases.
The second test was carried out to evaluate the protective effect of CNPs against gray mold disease. CNPs were prepared by centrifugation at 10,000 rpm for 10 min and washed three times to remove any residue of acetic acid. Then, the washed CNPs were resuspended in sterile distilled water at concentrations ranging from 0 to 50 mg/ mL. Healthy detached strawberry leaves were collected, and surface sterilized with 2% sodium hypochlorite for 3 min and 70% ethanol for 1 min, rinsed with sterile water, and finally dried on filter paper. The collection of plant material has complied with relevant institutional, national, and international guidelines and legislation. The surface-sterilized leaves were then immersed in the nanoparticle solutions (0, 12.5, 25, and 50 mg/mL) for 5 min and placed on moist paper towels in 9 cm Petri dishes with their adaxial surface up. Two mycelial plugs (5 mm) were acquired from the margin of a 5-day-old culture of B. cinerea SIB-1 and adhered to the treated leaves, with five leaves for each treatment. The dishes were sealed with plastic film and incubated at 25 °C in the dark for up to 5 days. The average percent of leaf area infected was determined for each concentration. Disease severity was evaluated in comparison to a disease severity scale of 0 (healthy leaf), 1 (< 25% infection); 2 = (25-50% infection) 3 = (51-75% infection), and 4 (> 75% infection) 56 .
Conclusions and challenges
The current study introduced a novel biological-based approach for the green synthesis of chitosan nanoparticles using Pelargonium graveolens. Optimization of the bioprocess parameters was studied and maximized. All characterization tests on CNPs confirmed the suitability and efficacy of the plant as the bio-converter agent. The antifungal activity of the generated CNPs against a model phytopathogen with a wide host range (B. cinerea SIB-1) proved the ability of the CNPs to replace or minimize the extensive use of pesticides and to be applied in various technological and medical fields.
Although the data of the current study are promising, other several challenges are still recommended to be elucidated, such as the exact mode of action of the generated CNPs, the suitable inhibitory doses on a wider range of fungi, and the cytotoxicity of CNPs. However, the large-scale production of CNPs is another challenge before being approved as a commercial product. | 2023-02-23T14:08:19.693Z | 2022-03-03T00:00:00.000 | {
"year": 2022,
"sha1": "f2245689fc077fb65fef48517fd9bcf75ca7297e",
"oa_license": "CCBY",
"oa_url": "https://www.nature.com/articles/s41598-022-07073-y.pdf",
"oa_status": "GOLD",
"pdf_src": "SpringerNature",
"pdf_hash": "f2245689fc077fb65fef48517fd9bcf75ca7297e",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": []
} |
10921025 | pes2o/s2orc | v3-fos-license | Heterogeneous multimodal biomarkers analysis for Alzheimer’s disease via Bayesian network
By 2050, it is estimated that the number of worldwide Alzheimer’s disease (AD) patients will quadruple from the current number of 36 million, while no proven disease-modifying treatments are available. At present, the underlying disease mechanisms remain under investigation, and recent studies suggest that the disease involves multiple etiological pathways. To better understand the disease and develop treatment strategies, a number of ongoing studies including the Alzheimer’s Disease Neuroimaging Initiative (ADNI) enroll many study participants and acquire a large number of biomarkers from various modalities including demographic, genotyping, fluid biomarkers, neuroimaging, neuropsychometric test, and clinical assessments. However, a systematic approach that can integrate all the collected data is lacking. The overarching goal of our study is to use machine learning techniques to understand the relationships among different biomarkers and to establish a system-level model that can better describe the interactions among biomarkers and provide superior diagnostic and prognostic information. In this pilot study, we use Bayesian network (BN) to analyze multimodal data from ADNI, including demographics, volumetric MRI, PET, genotypes, and neuropsychometric measurements and demonstrate our approach to have superior prediction accuracy.
Introduction
Alzheimer's disease (AD) is a highly prevalent neurodegenrative disease and is widely recognized as a major, escalating epidemic and a worldwide challenge to global health care systems [1]. Considerable research efforts have been devoted to establish a disease model of AD that could lead to greater understanding of the events that occur in AD. One major development is the Aβ hypothesis that assumes AD begins with abnormal processing of transmembrane Aβ precursor protein (APP). Such a malfunction of the APP metabolism will in turn trigger a series of pathological events, resulting in the toxic betaamyloid plaque in the human brain which is one defining characteristic of AD.
This disease model has been articulated in Jack et al. [2] who presented a hypothetical model for biomarker dynamics in AD pathogenesis. The model begins with *Correspondence: shuaih@uw.edu 1 Industrial Engineering Department, University of Washington, Seattle, WA, USA Full list of author information is available at the end of the article the abnormal deposition of Aβ fibrils, as evidenced by a corresponding drop in the levels of soluble Aβ42 in cerebrospinal fluid (CSF) and increased retention of the amyloid positron emission tomography (PET) radioactive tracers in the cortex. Subsequently, neurodegeneration and synaptic dysfunction follows, indicated by increased levels of CSF tau protein, brain atrophy, and decreased glucose metabolism measured using [ 18 F]fluorodeoxyglucose (FDG) PET. As neuronal degeneration progresses, atrophy in certain areas typical of AD such as the hippocampus regions becomes detectable by magnetic resonance imaging (MRI). So far, Jack's model has been widely studied, confirmed, refined, and enriched. While many details in the disease model are still unknown, investigators from academia and the pharmaceutical industry have been actively developing biomarkers to gain better and more accurate knowledge of the mechanisms of AD pathogenesis and progression to facilitate a range of clinical tasks such as early diagnosis, treatment efficacy evaluation, treatment planning, better clinical trial design, and drug developments.
While most of the existing efforts mentioned above focus on single modality of biomarkers, recently, there have been a few studies that proposed to study many biomarkers of heterogeneous nature jointly. For instance, Ye et al. [3] integrated multiple complementary data and initiated the work to use the multiple kernel learning method for multimodal integration for AD research. Zhang et al. did a sequence of work on multimodal classification [4] and regression [5] based on multimodality data and achieved better prediction accuracy than those models with a single biomarker. However, most of these works focus on prediction. Less effort has been devoted to study the interactions of these multimodal biomakers for better understanding of the disease as a whole.
Thus, in our study, we take a systematical perspective to study patterns of disease progression. We take into consideration multimodal biomarkers such as APOE (apolipoprotein E) genotypes, SNP variants, demographics, FDG-PET, amyloid PET, MRI, and neuropsychological assessment. We adopt a powerful machine learning model, the Bayesian Network (BN), as the major tool for studying the influential relationships among the variables. A main premise of using BN model for multimodal biomarker integration is that it could provide more details regarding the potential mechanism of the disease progression than those black-box prediction models [3][4][5]. Specifically, while the existing black-box prediction models throw in all the multimodal biomarkers as predictors parallel in the prediction equation regardless of their heterogeneous clinical nature, their clinical roles are not revealed since each biomarker is assigned with a quantitative weight in the prediction equation that only determines whether or not the biomarker is important. Moreover, this weight is not an absolute presentation of evidence, as it is essentially a multivariate concept that depends on the existence of other biomarkers in the equation. This results in the risk of excluding important biomarkers which hold significant clinical value but not significant statistical prediction value due to redundancy with other biomarkers. Also, from these black-box prediction models, there is no indication of how the biomarkers influence each other, whether or not some biomarkers mediate the effects from other biomarkers to disease outcomes. Presumably, the relationships between the multimodal biomarkers could be very complex, and our study is motivated by the lack of capacity of existing multimodal biomarker integration methods to discover and model these relationships. On the other hand, although not a causal model, BN models have been found very effective in a range of applications to study the "layers" of influence among variables. It could lead to very useful knowledge regarding the "chain reaction" of a sequence of events captured by the biomarkers' measurements. BN is a powerful data-driven model that seeks the best mechanistic model that is consistent with a set of measurements from a cohort of patients. Thus, it translates naturally into a semantic description of the disease similar to a clinician's intuitive description of its progression.
The remainder of the paper is structured as follows: In Section 2, we will provide description of the dataset that will be used in this study and the BN, particularly the mixed type Bayesian network due to the heterogeneous nature of the biomarkers. In Section 3, we will present the learning results and validation efforts. We then conclude our study in Section 4.
Data
The data used in this paper were obtained from ADNI database adni.loni.usc.edu. The primary goal of ADNI has been to test whether the serial MRI, PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early AD. Determination of sensitive and specific markers of very early AD progression is intended to aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well as lessen the time and cost of clinical trials.
ADNI is the result of efforts of many co-investigators from academic institutions and private corporations, subjects have been recruited from over 50 sites across the USA and Canada. The initial goal of ADNI was to recruit 800 adults, aged 55 to 90, to participate in the research with approximately 200 cognitively normal older individuals followed up for 3 years, 400 people with MCI followed up for 3 years, and 200 people with early AD followed up for 2 years.
Subjects
The ADNI general eligibility criteria are described at www.adni-info.org. Briefly, subjects are between 55 and 90 years of age, having a study partner able to provide an independent evaluation of functioning. Specific psychoactive medications will be excluded. The general inclusion/exclusion criteria are as follows: (1) healthy subjects: mini-mental state examination (MMSE) scores between 24 and 30, a Clinical Dementia Rating (CDR) of 0, non-depressed, non-MCI, and non-demented; (2) MCI subjects: MMSE scores between 24 and 30, a memory complaint, having objective memory loss measured by education adjusted scores on Wechsler Memory Scale Logical Memory II, a CDR of 0.5, absence of significant levels of impairment in other cognitive domains, essentially preserved activities of daily living, and an absence of dementia; and (3) mild AD: MMSE scores between 20 and 26, CDR of 0.5 or 1.0, and meets the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS/ADRDA) criteria for probable AD.
Our study includes the baseline measurements of 517 ADNI subjects. The cohort contains 114 AD patients, 283 MCI patients, and 120 healthy controls. Table 1 lists the demographics of these subjects.
Biomarkers
The description about biomarkers to be analyzed is listed in Table 2. These biomarkers are heterogeneous in terms of both clinical nature and statistical characteristics. While this list is still limited, it provides a good presentation of the genetic, demographic, neuroimaging, and clinical aspects of the disease. Among these markers, some are categorical biomarkers, such as sex (male or female) and SNPs (carrier or non-carrier), while some are numeric biomarkers such as some clinical measurements. Note that we also include some SNPs variants which are the top genetic risk factors for AD reported at http://www. alzgene.org/TopResults.asp.
Bayesian network
A BN is a graphical model that characterizes the influential relationships among variables is a finite set of nodes and E is a finite set of directed edges between the nodes. The DAG defines the structure of the BN. Each node v ∈ V in the graph corresponds to a random variable X v , i.e., in our study, a biomarker is a variable. In the DAG, the relationship between each variable X v with its parent variables denoted as pa(v) can be characterized as a conditional probability distribution, p(x v |x pa(v) ). Then, the joint probability distribution of a BN could be deduced as For this reason, the set of conditional probability distributions for all variables in the network, denoted as P, is called the parameter of the BN. A Bayesian network for a set of random variables X is then the pair (D, P).
Mixed type Bayesian network
In this paper, we adopt the mixed type Bayesian network model that handles both discrete and continuous variables, which is developed in [6]. For mixed type BNs, the set of nodes V can be further specified as V = ∪ T, where and T are the sets of discrete and continuous nodes, respectively. The set of variables X can then be denoted as , τ ∈ T}, where I and Y are the sets of discrete and continuous variables, respectively. For a discrete variable δ, we let I σ denote the set of levels.
It has been a challenge to model the mixed type Bayesian network. As mentioned earlier, a BN consists of the structure D and the parameter P. The central challenge for modeling mixed type Bayesian network is the development of appropriate models for characterizing P. In our study, we follow the seminar work in [6] that models the joint probability distribution by factorizing it into a discrete part and a mixed part, so where the first part of products of conditional probabilities is for discrete nodes and the second part is for continuous nodes.
For discrete nodes, conditional probabilities are parameterized as where θ σ |i pa(σ ) = (θ i σ |i pa(σ ) ) i σ ∈I σ . The parameters are subject to the constraints that i σ ∈I σ θ i σ |i pa(σ ) = 1 and 0 ≤ θ i σ |i pa(σ ) ≤ 1. For continuous nodes, the local probability distributions are Gaussian linear regressions on the continuous parents with parameters depending on the configuration of the discrete parents, as shown in below: so that
Learning of mixed type BN from data
With the BN model specified for mixed type variables, the next task is to identify a structure learning algorithm that can find the optimal DAG structure. The basic formulation of this problem, according to the score-based method, starts with a dataset T and a scoring function φ. Then, the task is to find a Bayesian network B ∈ B n that maximizes the values φ(B, T). The standard methodology is to use search algorithms, such as heuristic search, greedy hillclimbing, genetic algorithms, and tabu search, conducted over eligible search space B n to search the DAG structure that maximizes the score. In this study, we use the score function developed in [7] for mixed type BN, which can be readily implemented in the R package "bnlearn" [7]. After having identified the optimal DAG structure, parameter estimation could be conducted via maximum likelihood estimation according to (2). We refer interested readers to [8][9][10] for more details of the learning algorithms for mixed type BN.
Results
We apply the mixted type BN on the heterogeneous biomarkers of the ADNI cohort we have collected. In order to identify a stable DAG structure, first, we use a bootstrap method to generate 100 new training sets by sampling the original data set with replacement, then, learn the optimal DAG structure on each bootstrapped dataset. We then derive the final DAG structure by keeping those arcs which appear at least in half of these DAG structures learned from bootstrapped datasets. This strategy has been suggested in previous works for BN applications [11] that has been found effective to robustify the learning result. Note that, here, we also utilize the prior knowledge in the learning of the DAG structure, i.e., the genetic factors could be parents of other factors not the other way around, while the disease outcome variables such as ADAS-cog and MMSE score could only be in the bottom of the BN model. This prior knowledge is used in the BN learning and greatly reduces the search space of the eligible DAG structures. Note that, to impute missing values, the median is used for continuous variables while the mode is used for discrete variable. The final BN model is shown in Fig. 1. Note that some variables in Table 2 are not shown in Fig. 1. This indicates that the algorithm was not able to detect significant and robust relationship among these variables with others. We use green to represent categorical variables while using blue to represent numerical variables. The probability tables of categorical variables and the parameters of the conditional Gaussian distribution w, b for continuous variables are shown along the DAG structure as well. For example, node HippoNV in Fig. 1 has five parents: sex is binary when the other four are numerical. The relationship between the HippoNV with other variables such as AGE, EDU, AV45, and FDG is characterized as a regression model, while parameters of this regression model vary according to the categorical variable SEX.
Overall, this network structure is consistent with the existing knowledge in AD literature. As expected [12][13][14][15][16], the APOE e4 was associated with higher amyloid burden (as measured by AV45 PET imaging) and lower cerebral glucose metabolism (as measured by FDG-PET). A direct impact of e4 to MMSE score was also identified in our results in agreement with previous reports [17,18], although its underlying mechanism warrants further investigation. An association of the SNP rs11136000 with amyloid burden was also identified, in agreement with the potential role of clusterin (CLU, the gene that SNP rs11136000 is associated with) in Aβ clearance [19,20]. Based on this study, it is also identified that there were direct relationships between amyloid burden and cognitive performance which may reflect the direct neurotoxic effect of Aβ and its derivatives or indirect impact through pathways that were not represented in the biomarkers we included in this study [21][22][23]. The direct interaction between cerebral glucose metabolism and cognitive function as identified in this study was also in agreement with prior knowledge [24][25][26][27]. The identified relationship between years of education and the cognitive performance might be a cognitive reserve effect as reported by a number of studies [28][29][30][31]. In summary, using Bayesian network, we identified inter-biomarker relationships that are in good agreement with the existing knowledge about AD. Fig. 1 Learn mixed type Bayesian network using heterogeneous multimodality data at baseline
Evaluation of the prediction accuracy with BN
Besides comparing our results with AD literature, we further pursue numerical validation. Specifically, as "MMSE" and "ADAS-cog" are two important clinical outcomes, it is of interest to see if the learned BN owns significant prediction capability of the two outcomes. Thus, in this section, we compare the prediction capability of BN with three common regression techniques (implemented in R environment), such as linear regression (lm()), decision tree (rpart()), and random forest (randomForest()). The target metric we would like to measure and compare is mean square error (MSE), which serves as the goodness of fit in a regression problem. We use 10-fold cross validation to obtain unbiased estimates of MSE. To set up cross validation procedure, we randomly divide the original dataset into 10 subsamples. In each round, a single subsample is retained for testing the model while the remaining nine subsamples are used as training set. Table 3 lists the mean and standard deviation of MSE of the models. In terms of the average of the MSE, the BN achieves a better accuracy than the linear regression and decision tree in both MMSE and ADAS-cog prediction, while its performance is close to the random forest which has been known to be a very powerful prediction model despite its black-box nature. Similar observation could also be made in terms of the variance of the MSE.
Validation of the identified BN via the covariance patterns
We also analyze the covariance patterns to help validate the learned BN model. The covariance patterns essentially characterize the undirected associations among variables.
Thus, a BN model that aims to explain the influential relationships between the variables is expected to be able to explain the associations that are observed in data. Specifically, to derive the associations among variables, we use Pearson correlation for continuous variables, polychoric correlation for categorical variables, and polyserial correlation for a categorical variable and a continuous variable. The heterogeneous correlation matrix is computed using R package "polycor". Figure 2 shows the associations we have observed from the biomarkers. Each row/column represents one biomarker. The color intensity shows the strength of an association. Note here that we only present the magnitude of the associations to focus the purpose on validation with the BN model. Overall, the association patterns revealed in Fig. 2 is consistent with our learned BN model. For instance, from Fig. 2, it is clear that the ADAS-cog is strongly associated with the variables FDG, AV45, HippoNV, and APOE4. While this is consistent with the BN as shown in Fig. 1, we also notice that in Fig. 2, we could not detect that the association between APOE4 with ADAS-cog could be mediated by the variable FDG. Thus, by learning the BN model, we could identify more layers in the relationships between the variables and could shed light to useful discoveries of the underlying mechanism of the disease progression.
Validation of the identified BN via RuleFit
In order to validate the structure of the learned BN, another approach we propose to use is the RuleFit [32] method. RuleFit is a powerful method to discover complex interactions among variables. Again, it is a predictive model, so it lacks the capability of the BN to provide possible explanations of the relationships among the variables. But in the same spirit as the use of the association patterns to validate the BN model, we hope to see consistence between the BN structure with the interaction patterns the RuleFit could identify.
Thus, we apply the Rulefit on our data to identify the interactions among the biomarkers that can predict the two outcomes, MMSE and ADAS-cog. Table 4 lists the five rules we have identified. Column 1 gives the scaled importance for each rule. Column 2 (support) refers to the fraction of the samples in the dataset to which the rule applies. Apparently, it seems that there is great consistence between the two methods. For example, to predict MMSE, both the BN and RuleFit identified that HippoNV, FDG, EDU, and APOE4 are important. And to predcit ADAS-cog, both the BN and RuleFit identified that FDG, HippoNV, and AV45 are important. There are some interesting differences as well, e.g., RuleFit identified that the interaction between AGE and HippoNV is important to predict MMSE; however, it is revealed in the BN model that HippoNV actually mediates the effect from AGE to MMSE. Thus, given the consistency in the results, we could conclude that the BN model can provide more details of the underlying relationships among the variables.
Conclusions
In this paper, we propose to use the mixed type Bayesian network to model the interactions among heterogeneous multimodal biomarkers. We conduct this study using ADNI baseline dataset and find that the learned BN model provides findings that are consistent with the AD literature. We further validate the learned BN structure via the prediction accuracy of clinical outcomes, capability to explain association patterns among variables, and comparison with powerful feature selection method. In future work, we would like to investigate the use of dynamic BN models to incorporate the temporal data that is available in the ADNI dataset. Critical changes of the biomarkers that may indicate disease progression may be discovered, and how these significant clinical events could be synthesized to be a systematical disease model is a very interesting and exciting research direction. Also, note that mixed type BN has been a challenging topic which is worthy of further methodological study. For example, different from the approach we used in this study, it is suggested to convert the continuous variables into discrete variables to enable application of discrete BN learning algorithms. Examples of these methods could be found in [8,10]. Another approach is to directly model the mixed type variables, as the one we have used in our study. Examples of these methods could be found in [9]. The approach we used, although it has been a benchmark method and well evaluated in some applications, limits its applications to where only discrete variables could be parents of continuous variables. Both methods could lead to discovery of different types of relationships among variables, and different applications may require different approaches for optimal performance. Thus, we believe it will also be a future research direction. | 2017-08-11T00:35:22.260Z | 2016-08-19T00:00:00.000 | {
"year": 2016,
"sha1": "ed667ab13d3e67194e371b0fdedf1af8bf0c2302",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1186/s13637-016-0046-9",
"oa_status": "HYBRID",
"pdf_src": "PubMedCentral",
"pdf_hash": "ed667ab13d3e67194e371b0fdedf1af8bf0c2302",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Medicine",
"Computer Science"
]
} |
220470348 | pes2o/s2orc | v3-fos-license | Photoacoustic topography through an ergodic relay for functional imaging and biometric application in vivo
Abstract. Significance: Photoacoustic (PA) tomography has demonstrated versatile biomedical applications. However, an array-based PA computed tomography (PACT) system is complex and expensive, whereas a single-element detector-based scanning PA system is too slow to detect some fast biological dynamics in vivo. New PA imaging methods are sought after. Aim: To overcome these limitations, we developed photoacoustic topography through an ergodic relay (PATER), a novel high-speed imaging system with a single-element detector. Approach: PATER images widefield PA signals encoded by the acoustic ergodic relay with a single-laser shot. Results: We applied PATER in vivo to monitor changes in oxygen saturation in a mouse brain and also to demonstrate high-speed matching of vascular patterns for biometric authentication. Conclusions: PATER has achieved a high-speed temporal resolution over a large field of view. Our results suggest that PATER is a promising and economical alternative to PACT for fast imaging.
Photoacoustic (PA) imaging provides functional and molecular information by sensing optical absorption, which supports a wide range of biomedical applications. [1][2][3][4] PA computed tomography (PACT) has successfully imaged structural and dynamic features in animals and humans. [5][6][7] Using an array of ultrasonic transducers, a PACT system can detect signals from a large field of view (FOV) in parallel, but the multichannel detection and acquisition system is complex and expensive. 8,9 Moreover, PACT systems are often bulky. On the other hand, conventional PA microscopy systems scan a single-element ultrasonic transducer to form images, with reduced imaging throughput. 4,10 As an alternative, we developed photoacoustic topography through an ergodic relay (PATER), a novel high-speed imaging technology with a single-element ultrasonic transducer. 11 An acoustic ergodic relay (ER) is an acoustic waveguide that encodes sound waves from the input points to an output point with distinct acoustic reverberant characteristics. 12 We had previously shown that a right-angle prism works as an ER for PATER. 11 Using only a single-element ultrasonic transducer, PATER simultaneously detects widefield PA signals encoded by the ER and then mathematically decodes the received signal to form a widefield image. 11,13,14 Consequently, PATER can be used to study dynamic activities with a submillisecond temporal resolution over a large FOV. Applying PATER in vivo, we monitored changes in oxygen saturation in a mouse brain and demonstrated high-speed recognition of vascular patterns for biometric authentication. Our results have demonstrated that PATER is a promising and economical alternative to PACT for a broad range of biomedical applications. Figure 1(a) shows a schematic of the PATER system. A 532-nm pulsed laser beam (INNOSAB IS8II-E, Edgewave GmbH, 5-ns pulse width, and 1-kHz pulse repetition rate) passes through an optical-element wheel (LTFW6, Thorlabs, Inc.), which switches the active optical elements (a lens and an engineered diffuser) in and out of the light path according to the acquisition mode. The laser beam then passes through the ER and illuminates the object on the ER's imaging plane. PA waves are encoded inside the ER and finally detected by a single-element ultrasonic transducer (VP-0.5-20 MHz, CTS Electronics, Inc.).
PATER requires two acquisition steps. In the first step-a point-by-point scanning called calibration mode [ Fig. 1(b)], a laser beam is focused by a plano-convex lens (LA1433, Thorlabs, Inc.; 150-mm focal length) to a small spot (∼30 μm) on the input surface of the ER that interfaces with the object to be imaged. Because the pulse width of the laser (∼5 ns) is much shorter than the central period of the ultrasonic transducer (50 ns, correspoinding to 20 MHz) and the focused beam spot (∼30 μm) is much smaller than the central acoustic wavelength (∼300 μm inside the ER), each PA wave input to the ER can be approximated as a spatiotemporal delta function. 11,15 Therefore, each calibration measurement quantifies the impulse response of the system at one scanning position. The ER is driven by a customized two-axis motorized stage for raster scanning along the x and y axes, so impulse responses over the entire FOV can be calibrated. The second step, referred to as widefield imaging mode, uses a broad laser beam for illumination [ Fig. 1(c)]. The laser beam passes through an engineered diffuser (EDC-5-A-1r, RPC Photonics, Inc.; 5.5-deg divergence angle) that homogenizes the beam for uniform illumination. PATER's system setup and reconstruction method were reported in Ref. 11. Each widefield measurement can be expressed as a linear combination of the impulse responses from all pixels: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 1 ; 1 1 6 ; 5 9 2 where s denotes the detected widefield PA signal, t is the time, i is the pixel index, N p is the total number of pixels, k i is the normalized impulse response from the calibration, and P i is the rootmean-squared (RMS) PA amplitude. 11 The RMS value of the raw calibration signalk i ðtÞ for the i'th pixel was calculated as E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 2 ; 1 1 6 ; 4 9 0 where N t denotes the number of sampled time points and t is the time. A 2-D density plot of RMS i over all pixels is a calibration image. To construct the system matrix K, the normalized impulse response was computed for each time point through k i ðt j Þ ¼k i ðt j Þ∕RMS i . Eq. (1) can be recast to matrix form by discretizing time t: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 3 ; 1 1 6 ; 3 8 3 where K ¼ ½k 1 ; k 2 ; : : : ; k N p is the system matrix. Pixels with RMS values lower than twice (6 dB) the noise amplitude were considered as the background that was too dark to calibrate for; therefore, the impulse responses of these pixels were excluded from the system matrix K. The widefield image P is reconstructed by solving the inverse problem of Eq. (3) as a minimizer of the objective function, adopting a two-step iterative shrinkage/thresholding algorithm: 16 E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 4 ; 1 1 6 ; 2 9 0P ¼ arg min P ks − KPk 2 þ 2λΦ TV ðPÞ: Here Φ TV ðPÞ is the total variation regularization term and λ is the regularization parameter. 16 We tested the linearity of the PATER system by measuring concentrations of the Evans Blue (EB) dye (E2129, Sigma-Aldrich, Inc.) in two tubes with 532-nm light illumination. Two silicone tubes with a 0.65-mm inner diameter were placed on the ER surface in parallel, separated by ∼3 mm. Ultrasonic gel was applied between the tubes and the ER to facilitate acoustic coupling. An EB solution with a 0.6% concentration by mass was injected into the two tubes for calibration. The concentration of EB in one tube was kept unchanged as a control, whereas the concentration of EB in the other tube was varied from 0% to 0.9% [ Fig. 2(a)]. The measured concentrations, calculated based on the widefield images, agreed well with the preset concentrations [Figs. 2(b) and 2(c)], which proved the linearity of the PATER system's widefield measurement.
For in vivo studies, we used female ND4 Swiss Webster mice (Envigo; 18 to 20 g, 6 to 8 weeks). All the laboratory animal protocols were approved by the Animal Studies Committee of Washington University in St. Louis and the Institutional Animal Care and Use Committee of California Institute of Technology. The mouse was anesthetized in a small chamber with 5% vaporized isoflurane mixed with air for anesthesia induction and then transferred to a customized animal mount where it was kept anesthetized with a continuous supply of 1.5% vaporized isoflurane. The animal mount consisted of a stereotaxic frame that fixed the mouse's head and a heating pad that maintained the mouse's body temperature at ∼38°C. The hair on the mouse's head was razor trimmed, and the scalp was surgically removed, but the skull was left intact. The scalp was removed to enable direct contact between the skull and ER to facilitate acoustic coupling. Bloodstains on the skull were carefully cleaned off with phosphate buffered saline solution, and ultrasound gel was applied on the skull for acoustic coupling. Then the animal mount was raised until the mouse's skull was in contact with the imaging surface of the ER. An adequate amount of pressure was maintained between the mounted animal and the ER to prevent the mouse's head from moving, but not so much pressure as to interrupt the blood supply in the brain.
We first imaged the in vivo dynamic change in blood oxygen saturation (sO 2 ) in a mouse brain using a deoxy-hemoglobin-dominated absorption wavelength of light at 620 nm. Oxygen challenges were performed to stimulate changes in the sO 2 level in the mouse brain by manipulating the oxygen concentration of the mouse's inhaled gas. In this study, a mixture of 95% oxygen and 5% nitrogen was initially used, with gaseous isoflurane for anesthesia. The mouse brain vasculature was first imaged through the intact skull in calibration mode. For the oxygen challenge, the mixture was changed to 5% oxygen and 95% nitrogen for 3 min; it was then changed back to the initial concentration to end the challenge.
To estimate the change in sO 2 in a mouse brain using a single wavelength of light, a few assumptions are required. First, absorption in blood mainly comes from oxy-and deoxyhemoglobin. Thus the absorption coefficient μ a of blood can be calculated as E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 5 ; 1 1 6 ; 2 2 5 μ a ¼ lnð10Þðε HbO 2 C HbO 2 þ ε Hb C Hb Þ; where ε is the molar absorption coefficient (M −1 cm −1 ), C is the concentration (M), and the subscripts HbO 2 and Hb denote oxy-and deoxy-hemoglobin, respectively. The oxygen saturation in the blood is calculated as E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 6 ; 1 1 6 ; 1 5 7 where the total hemoglobin concentration (C HBT ) is given by C HBT ¼ C HbO 2 þ C Hb . Therefore, the change in the blood oxygen saturation can be calculated as Second, if we assume that the change in the total hemoglobin concentration in blood is insignificant, then a change in blood oxygen saturation signifies that ΔC Hb ¼ −ΔC HbO 2 . At a deoxy-hemoglobin dominant absorption wavelength, such as 620 nm, the ratio of ε Hb ∕ε HbO 2 is ∼7∶1, thus a change in absorption is mainly due to a change in the concentration of deoxyhemoglobin. 17,18 Therefore, we can assume that Δμ a ≈ lnð10Þε Hb ΔC Hb and that the change in PA signal amplitude at 620 nm is proportional to the change in blood oxygen saturation.
To monitor the oxygen challenge, a tunable dye laser (CBR-D, Sirah GmbH), using DCM (SDL-550, Sirah GmbH) dissolved in ethanol as the gain medium, was pumped by the 532-nm pulsed laser (INNOSAB IS8II-E, Edgewave GmbH, 5-ns pulse width, 1-kHz pulse repetition rate) to generate laser light at 620 nm. The calibration (200 × 200 pixels, 30-min acquisition time) was performed at both 532 and 620 nm wavelengths, with an FOV of 3 × 3 mm 2 . We recorded the same FOV in widefield imaging mode at 620 nm during the oxygen challenge (Video S1) and calculated the signal differences pixel by pixel from the widefield images after temporal running averaging. Two oxygen challenge cycles were performed and analyzed [ Fig. 3(a)]. The rate of signal change during the challenge was smaller than that during recovery from hypoxia [ Fig. 3(b)], which is consistent with the results reported previously. 19,20 To provide dual-wavelength measurements for sO 2 calculation, widefield measurements at 532 nm were taken before and at 3 min into the oxygen challenge and reconstructed with the 532-nm calibration data. A vessel-segmentation and sO 2 quantification algorithm was used to identify vessels and compute the sO 2 within those vessels. 21,22 The sO 2 in the brain was found to have dropped significantly during the challenge [ Fig. 3(c)].
In our second in vivo study, we demonstrated PATER's ability to differentiate blood vessel patterns for potential biometric authentication applications. Biometric authentication utilizes unique biological characteristics of individuals to verify their identities. Internal characteristics such as vascular patterns can more securely identify an individual than external characteristics such as fingerprints 23 because the internal characteristics are less exposed and contain in vivo physiological features-such as blood flow, arterial oxygenation, and venous oxygenationthat cannot be readily duplicated by others. Security applications based on internal biometric characteristics have great potential, but they require high processing speed and accuracy to be reliable. First, one mouse was fixed in a stereotaxic frame and a region of the cortical vasculature was recorded in calibration mode [ Fig. 4(a)]. Then, the same FOV was imaged in widefield imaging mode. During the widefield recording, we detached the mouse from the ER and then reattached it to the same position using a linear translational stage (PT1, Thorlabs, Inc.). Only noise was recorded while the mouse was detached, and signals were observed again when the mouse was reattached. This process was repeated for the second mouse. We then tried to reconstruct the widefield images of the first mouse's vasculature using each of the two recorded calibration data sets. As a result, the widefield image reconstructed from the matched calibration data (the first mouse's) revealed the original vasculature, whereas the widefield image reconstructed from the mismatched calibration data (the second mouse's) could not [ Fig. 4(b) and Video S2]. The correlation coefficients between the widefield reconstruction images and the calibration images were quantified [ Fig. 4(c)]. The plot indicates that the widefield images reconstructed from the matched calibration data have a much higher correlation than those reconstructed from the mismatched calibration data. Also the vasculature is again recognizable after being detached and reattached to the ER. Several aspects of this proof-of-concept experiment still need to be addressed to make the technology more applicable. First, the region where the object is reattached to the ER needs to be the same as the calibrated region, requiring an effective repositioning method. Second, the deformation of soft tissue should be minimized for the current system, as it could change the boundary conditions. Third, the present PATER system requires calibration for each object; a universal calibration method is being explored, which will promise more biomedical applications in the future.
In summary, we have demonstrated PATER's ability to quantify in vivo functional processes such as oxygen saturation changes in a mouse brain and also to identify vasculature patterns based on PATER's unique detection method. PATER's single-channel ultrasonic detection system can be a feasible alternative to a PACT's multichannel ultrasound detection system. Compared to PACT, PATER has greatly reduced cost and system complexity, making it more affordable for portable applications, such as a wearable device to monitor vital signs in patients. Furthermore, since it can both identify vessel patterns and quantify functional processes, PATER can potentially provide comprehensive, secure, and robust biometric authentication.
Disclosures L. V. W. and K. M. have financial interests in Microphotoacoustics, Inc., CalPACT, LLC, and Union Photoacoustic Technologies, Ltd., which did not support this work. | 2020-07-11T13:02:15.438Z | 2020-07-01T00:00:00.000 | {
"year": 2020,
"sha1": "13ba072c8c4c429e2a56a95bacd70ecc6627d507",
"oa_license": "CCBY",
"oa_url": "https://www.spiedigitallibrary.org/journals/journal-of-biomedical-optics/volume-25/issue-7/070501/Photoacoustic-topography-through-an-ergodic-relay-for-functional-imaging-and/10.1117/1.JBO.25.7.070501.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "970737a26c1723ba02d6b3bcf584aef7b0117e6b",
"s2fieldsofstudy": [
"Engineering",
"Medicine"
],
"extfieldsofstudy": [
"Computer Science",
"Medicine"
]
} |
174812897 | pes2o/s2orc | v3-fos-license | How Does the Regulatory Genome Work?
The regulatory genome controls genome activity throughout the life of an organism. This requires that complex information processing functions are encoded in, and operated by, the regulatory genome. Although much remains to be learned about how the regulatory genome works, we here discuss two cases where regulatory functions have been experimentally dissected in great detail and at the systems level, and formalized by computational logic models. Both examples derive from the sea urchin embryo, but assess two distinct organizational levels of genomic information processing. The first example shows how the regulatory system of a single gene, endo16 , executes logic operations through individual transcription factor binding sites and cis -regulatory modules that control the expression of this gene. The second example shows information processing at the gene regulatory network (GRN) level. The GRN controlling development of the sea urchin endomesoderm has been experimentally explored at an almost complete level. A Boolean logic model of this GRN suggests that the modular logic functions encoded at the single-gene level show compositionality and suffice to account for integrated function at the network level. We discuss these examples both from a biological-experimental point of view and from a computer science-informational point of view, as both illuminate prin-ciples of how the regulatory genome works.
INTRODUCTION
''There exists today a very elaborate system of formal logic, and specifically, of logic as applied to mathematics. This is a discipline with many good sides, but also with certain serious weaknesses. .Everybody who has worked in formal logic will confirm that it is one of the technically most refractory parts of mathematics. The reason for this is that it deals with rigid, all-or-none concepts, and has very little contact with the continuous concept of the real or of complex number, that is, with mathematical analysis. Yet analysis is the technically most successful and best-elaborated part of mathematics. Thus formal logic is, by the nature of its approach, cut off from the best cultivated portions of mathematics, and forced onto the most difficult part of mathematical terrain, into combinatorics.''-John von Neumann M echanisms to annotate genomic sequences encoding RNAs and proteins are well established, but the term ''regulatory genome'' refers to parts of the genome that provide information not for the structure of molecules but for when and where molecules are produced within an organism (Davidson, 2006;Peter and Davidson, 2015). What will it take to annotate the regulatory genome? Which structural and functional definitions will be adequate to describe the regulatory genome? Typically, regulatory DNA encodes binding sites for transcription factors that in turn control gene expression. At the sequence level, however, it is so far not possible to distinguish regulatory from nonregulatory sequences, since transcription factor binding sites are small and occur throughout the genome in regulatory as well as nonregulatory DNA. Thus at the structural level, the rules by which regulatory DNA encodes gene expression patterns are not clearly understood, and the current definition of regulatory sequences relies on the observed function in gene regulation.
From a computational point of view, the function of the regulatory genome is to execute highly complex information processing functions at several levels of organization. At a basic conceptual level, the function of regulatory DNA is to control the expression of individual genes. Even at this level, regulatory systems associated with individual genes display a complex modular form, with clusters of transcription factor binding sites encoded in multiple cis-regulatory modules that all contribute to the correct gene expression output. With the discovery of gene regulatory networks (GRNs), it became clear, however, that the function of the regulatory genome goes beyond the control of individual genes. Thus, from a systems-level perspective, the regulatory genome also provides the information system for the development of the animal body plan. GRNs are networks of regulatory genes encoding transcription factors and signaling molecules, and of regulatory sequences encoding their interactions. Cis-regulatory sequences controlling the expression of transcription factors and signaling molecules affect not just the activity of single genes, but they also affect all other genes expressed downstream of these regulators. Regulatory sequences that control the expression of transcription factors contribute directly to the interpretation of the regulatory genome, since they determine the combination of expressed transcription factors, the regulatory state (Peter, 2017). At this level, the regulatory genome encodes information for genome activity in all different developmental and physiological contexts, throughout the life of an organism.
How are the different levels of informational organization encoded in the regulatory genome? Important insights into how the regulatory genome works have been generated in the sea urchin embryo by detailed system-level dissection of regulatory systems at both the single-gene and GRN levels. One of the first and best understood cis-regulatory control systems for an individual gene controls the expression of endo16, a gene expressed in the midgut of sea urchin embryos (Soltysik-Espanola et al., 1994;. The expression of endo16 is controlled by several cis-regulatory modules, as is typical for any gene, and each module includes binding sites for several transcription factors. The impact of the individual modules, and even individual transcription factor binding sites within these modules, on the gene expression output of endo16 has been analyzed experimentally and shows a complex code for information processing even at the single-gene level. At the GRN level, one of the most complete experimental analyses of a network has also been conducted in the context of endomesoderm development in the sea urchin embryo (Davidson et al., 2002a;Oliveri et al., 2008;Croce and McClay, 2010;Peter andDavidson, 2010, 2011;Sethi et al., 2012;Materna et al., 2013;Cui et al., 2014). This GRN consists of *50 regulatory factors that control gene expression and the specification of several distinct endomesodermal cell fates during 30 hours of development.
Curiously, the computational function of the regulatory genome, at both the single-gene and GRN levels, has been made accessible through computational logic models. The observation that the function of the regulatory genome can be approximated by computational logic formulas, not unlike the logic gates used in computer science, indicates that formal logic approaches successfully capture the information processing functions of the regulatory genome at different scales. We discuss the insights that have been generated by experiments and computational models, and that illuminate the functional properties of the regulatory genome.
686
ISTRAIL AND PETER
CONTROL OF A SINGLE GENE: THE ENDO16 REGULATORY SYSTEM IN EXPERIMENT AND MODEL
The cis-regulatory system controlling the expression of the sea urchin endo16 gene was one of the first to be experimentally dissected in great detail, and as of today is probably still one of the best understood regulatory control systems (Soltysik-Espanola et al., 1994;Yuh et al., 1994Kirchhamer et al., 1996;. The regulatory function of this sequence was first experimentally discovered, and then captured in an elegant computational model (Yuh et al., 1998). But to be able to appreciate this model, we will first revisit some of the experimental work that served as its foundation.
The sequence fragment that encodes sufficient information to recapitulate the expression pattern of endo16 during sea urchin endoderm development was identified by reporter assays (Yuh et al., 1994. A DNA fragment of 2300 bp just upstream of the transcription start site of endo16, when placed upstream of a reporter gene and injected into sea urchin embryos, drives expression at first in endodermal progenitors and in the midgut during later development. Similar experiments have been performed with many cis-regulatory sequences, demonstrating that regulatory DNA encodes information for particular gene expression patterns. In the case of endo16, the individual regulatory functions encoded within the cis-regulatory system have been carefully dissected. First, the 2300-bp fragment was separated into seven modules (modules A-G; Fig. 1A) by restriction enzymes. Each module contains clusters of binding sites that are recognized and bound by transcription factors, which regulate gene expression (Yuh et al., 1994). All modules together encode a total of >30 binding sites for 13 transcription factors. The seven modules were then tested for transcriptional activity, either alone or in various combinations . These experiments revealed that of the seven modules, three contributed to the activation of gene expression in the endoderm (modules A, B, and G) and four contributed to the repression of endo16 in the ectoderm (modules E, F) and skeletogenic mesoderm (modules C, D).
The key to understanding the regulatory logic of the endo16 cis-regulatory system is that neither the modules nor the transcription factor binding sites operate by simple linear addition or subtraction of individual regulatory functions. Instead, individual transcription factor inputs and individual modules operate by nonlinear combinatorial synergism (Yuh et al., 1998). Thus, the complete construct with all seven modules produces specific expression in the endoderm during development of the sea urchin embryo . Furthermore, modules A, B, and G are each capable of driving endoderm expression on their own, although they also show some expression in the ectoderm and skeletogenic mesoderm that is not observed with the full construct . Similarly, a construct with all three modules A, B, and G produces correct endodermal expression in addition to ectopic expression in ectoderm and skeletogenic mesoderm. Of the three activating modules, module G shows only weak activity, while module A drives endodermal expression predominantly during early development and module B drives expression predominantly during later development in the midgut. Adding either module E or module F to the construct with all three activating modules suppresses the ectopic expression in the ectoderm, meaning that both modules encode binding sites that are sufficient to repress expression of endo16 in the ectoderm. Modules C and D on the contrary are both required simultaneously to repress ectopic expression of the ABG construct in the skeletogenic mesoderm .
The interesting feature of this regulatory system that is perhaps common to many transcriptional control systems is that the modules, when present together, do not operate independently despite the fact that each module also shows activity when tested individually. Thus, a construct containing the two activating modules A and B shows more transcriptional activity, then the sum of the activity of the two constructs carrying modules A and B alone. The data indicate that the contribution of module A can be described as a linear amplification function by a factor of 4 of the output of module B . Similarly, adding module A to a construct including both modules B and G will lead to an amplification of the output of construct BG by a factor of 3. Thus, module A, although active on its own, functions as a modulator of activity when placed in combination with other modules. Even more remarkable, module A is also required to mediate the repressive activity of modules CD, E, and F. Thus, when the repressive modules are combined with the GBA activator modules or with module A alone, they turn off gene expression in nonendodermal cell fates. However, if modules CD, E, and F are placed in combination with modules G and B, without module A, the repressive modules have no effect on gene expression. This means that module A contributes to both activation and repression of endo16 in response to alternative regulatory modules, in a Janus-like function, while activating gene expression by itself. So how does an activating module mediate the function of repressive modules?
The model for endo16 regulation shown in Figure 1B summarizes the logic operations that are executed by module A (Yuh et al., 1998). This model captures the function of individual transcription factor binding sites within module A that perform the computation of gene activity. Thus, binding sites CG1 and P are both required for the interaction of module A with module B. The mutation of either CG1 or P will lead to gene expression comparable with module A alone, even when module B is present, and to a reduction of gene activity by a factor of 2. When by itself, module A drives gene expression in the endoderm of early sea urchin embryos. This expression pattern is mediated by the binding site for Otx, and Otx is absolutely required for endodermal expression. Without a functional Otx binding site, the endodermal activity of module A is abolished (Yuh et al., 1998). However, even though mutation of the Otx binding site abolishes module A function in the early endoderm, it does not interfere with the ability of module A to interact with module B through binding sites CG1 and P. Furthermore, module A interacts with modules F, E, and DC through binding site Z. The interaction of module A with the basal transcription apparatus (BTA) is mediated by binding sites CG2, CG3, and CG4, and mutation of these sites leads to a reduction in gene expression by a factor of 2. Thus, interactions between modules A and B and between module A and the BTA are equally contributing to the fourfold increase of module B activity in the presence of module A. If we keep in mind that each site is just a few nucleotides long, then this entire complex operation is encoded in just a few short sequence elements with no particular apparent organization other than being part of module A.
The computation performed by module A in the control of endo16 expression can be approximated by the logic model shown in Figure 1C (Yuh et al., 1998). Interestingly, this computational model is a hybrid model including both discrete logic functions and the response to continuous inputs. The overall gene output function in this model is shown as a continuous function in time that is modulated by discrete
ISTRAIL AND PETER
repression or amplification functions. Each term represented by Greek letters in Figure 1B and C captures the regulatory impact of either transcription factor binding sites or cis-regulatory modules based on experimental observations. For some inputs, this impact is modeled as strictly Boolean, such as for the repressor modules F, E, and DC, where presence of any one of the repressing transcription factors dominantly turns off gene expression (e.g., if the integrated repressor function a = 1, then Z(t) = 0, and thus the final output Y(t) = g * Z(t) = 0). In a Boolean logic statement, this would correspond to a dominant NOT logic function. On the contrary, for other inputs, the impact is represented by an amplifier function, where presence of an activating transcription factor amplifies the gene output by a factor of 2 (e.g., if P = 1 AND CG = 1, b = 2, else b = 0). A few inputs are represented in this model as time-dependent continuous regulators determining the kinetics of the gene output.
There are many lessons to be learned from the work on the endo16 regulatory system. One is that even a relatively simple expression pattern requires complex computation of regulatory inputs. These regulatory inputs occupy their respective binding sites within regulatory DNA wherever they are expressed in the embryo, but whether this interaction leads to a gene expression output depends on the computation of the overall output based on the information encoded in the proximal cis-regulatory module A. Module A integrates the response to all other modules and the function of regulatory inputs binding to these modules. The developmental functions of this regulatory system, that is, activation in endoderm or repression in ectoderm, are mediated by separate DNA sequence modules, which makes it necessary to determine the final output through a proximal element responsive to all other modules. In addition, the endo16 model suggests a more profound truth, which is that the function of cis-regulatory sequences can be thought of in the context of a logic framework, each module contributing a unique function.
This function is mediated by transcription factors that bind to regulatory modules, and can be described by a combination of discrete or continuous functions. In the endo16 case, a few binding sites were identified, which determined the dynamic change in expression levels while the repressive modules were better approximated as Boolean ON/OFF switches. But regardless of the qualitative contribution of each input, the overall gene expression output is computed by the integration of individual regulatory functions according to strict logic rules. This idea was further explored by Istrail and Davidson (2005). This work found that by comparing the function of many different cis-regulatory systems, several logic operations could be defined, which were commonly executed by regulatory sequences. Thus, the regulatory systems controlling expression of single genes can be described as a repertoire of logic gates that is valid and applied across organisms, and a fundamental feature of the regulatory genome that is also used in the following to describe higher level GRN functions.
LINKING REGULATORY SYSTEMS: THE OPERATION OF REGULATORY CIRCUITS
The endo16 example shows a view of the regulatory genome that we are perhaps most familiar with, which is the function of cis-regulatory systems to control gene expression in response to transcription factor inputs. But the information contributed by the regulatory genome goes beyond the regulation of single genes in response to a regulatory state. The regulatory genome is also responsible for generating the regulatory states, thereby controlling the activity of the genome. The information for the control of genome activity is stored in the genome in the form of GRNs (Peter and Davidson, 2015). GRNs consist of genes encoding regulatory factors and of cis-regulatory sequences controlling gene expression. GRNs control the expression of transcription factors and signaling molecules that in turn regulate the expression of all other genes.
Interestingly, the regulatory systems controlling expression of transcriptional regulators are in principle no different from the regulatory systems controlling expression of any other gene, although perhaps slightly more complicated in design. However, what is substantially different is that cis-regulatory systems controlling expression of transcriptional regulators are connected with one another through regulatory interactions, whereby the transcription factor expressed as the output of a regulatory gene will serve as an input into other cis-regulatory systems. As a result, the regulatory circuits that are formed by multiple regulatory genes and their regulatory interactions have properties that go beyond regulating the expression of individual genes in response to transcription factor inputs. By connecting multiple regulatory systems, these circuits are able to execute more complex multigene logic functions.
HOW DOES THE REGULATORY GENOME WORK?
An example of a small regulatory circuit is shown in Figure 2A. Here, the three genes gcm, gatae, and six1/2 are connected by a positive feedback circuit that is active in the sea urchin mesoderm Davidson, 2006, 2012;Peter and Davidson, 2017). In this circuit, Gcm activates the expression of gatae, and Gatae activates the expression of six1/2. The two regulatory feedbacks are provided by Gcm, activating its own expression, and by Six1/2 activating gcm expression. The cis-regulatory system of gcm encodes the response to both Gcm and Six1/2, and both transcription factors are required to ensure late expression of gcm (Ransick and Davidson, 2012). The initial activation of this positive feedback circuit comes from Delta/Notch signaling, which activates the expression of gcm (Ransick and Davidson, 2006;Croce and McClay, 2010). In the presence of Delta/Notch signaling, gcm expression is activated despite the absence of Gcm and Six1/2. And vice versa, Delta/Notch signaling is only present during early development and then turns off. At this point, Gcm and Six1/2 regulate gcm expression even in the absence of Delta/Notch signaling (Ransick and Davidson, 2012). Thus, Delta/Notch operates in OR logic to the other two inputs, while Gcm and Six1/2 regulate gcm expression in AND logic ( Fig. 2A). This regulatory logic is reflected in the cis-regulatory system of gcm. The initial input, Delta/Notch signaling, activates a cis-regulatory module (CRM) that is independent of the module responding to Gcm and Six1/2. The two CRMs function independently, and therefore constitute an OR logic gate, while the two inputs regulating the second CRM have to be both present to activate the AND logic gate.
The function of this regulatory circuit can be captured in a Boolean logic model (Fig. 2). In this model, we assume that a gene is either expressed (1) or not expressed (0), and that expression of the gene will lead to the production of functional levels of the corresponding transcription factor. Furthermore, we take into account that there is a time delay between expression of a regulatory gene and activation of its target gene, which is defined by the time it takes to produce sufficient amounts of transcription factor product to activate
690
ISTRAIL AND PETER target gene expression (Bolouri and Davidson, 2003). If we assume that the delay time in this example is 1 hour, and we assume the regulatory logic for the three genes as given in Figure 2B, then all three regulatory genes are expressed within 2 hours after turning on the initial input Delta/Notch signaling (Fig. 2C). Moreover, the positive feedback into gcm is active after 3 hours (three regulatory steps at 1 hour each). If Delta/Notch signaling is turned off after 4 hours, all three genes remain being expressed because of the operation of the positive feedback circuit. However, if the Delta/Notch input lasts only for 1 hour, this does not provide enough time to activate the positive feedback circuit, and the three regulatory genes are only expressed transiently for as long as their inputs are present (Fig. 2C). Similarly, if the positive feedback circuit in this model is removed, expression of the three regulatory genes depends exclusively on the Delta/Notch signaling input, and gene expression turns off once the input is no longer available (Fig. 2C). This example shows how the regulatory inputs controlling expression of gcm do not just operate in isolation but are part of a regulatory circuit with a function beyond the control of gcm expression. Activating gene expression is a function of Delta/Notch signaling, but this signal lasts only for a few hours in the sea urchin embryo, and the positive feedback circuit is required for continued gene expression. The function of this positive feedback circuit is not to turn on expression of the three regulatory genes, but to maintain their expression once the initial input is no longer available. Similar positive feedback circuit configurations have been discovered in many GRNs that control very different developmental processes (Narula et al., 2010;Peter and Davidson, 2015). Very often, positive feedback circuits occur downstream of transient developmental signaling inputs, implying that they function in a way similar to the example discussed here. Since in each developmental context these positive feedback circuits are composed of cellfate-specific sets of transcription factors, the similarity in circuit function must be caused by the similarity in regulatory circuit architecture and not because of the specific molecular properties of the transcription factors involved. Since the architecture of regulatory circuits and GRNs is encoded in the regulatory genome, this means that important developmental functions are encoded in the regulatory genome in addition to protein coding sequences. We will now turn to the function of the regulatory genome at the GRN level, which is responsible for the control of entire developmental processes.
REGULATORY LOGIC AT THE LEVEL OF GENE REGULATORY NETWORKS: THE ENDOMESODERM GENE REGULATORY NETWORK
GRNs have been experimentally studied in many developmental contexts (Peter and Davidson, 2015). One of the most extensively characterized GRNs controls endomesoderm development in pregastrular sea urchin embryos (Davidson et al., 2002a(Davidson et al., , 2002bOliveri et al., 2008;Peter andDavidson, 2010, 2011;Materna et al., 2013). About 50 transcription factors and signaling ligands/receptors are involved in the specification of endoderm and mesoderm during the first 30 hours of sea urchin development. These regulatory factors have been identified based on a systematic analysis to be expressed in either endodermal or mesodermal cell fates. The regulatory interactions between these transcription factors were analyzed by systematically perturbing the expression of each transcriptional regulator and by monitoring the effect on the expression of all other regulatory genes in the system. The results of gene expression analyses and perturbation experiments were used to reconstruct the GRN that connects these regulators into a functional program for endomesoderm development.
The function of the endomesoderm GRN is to determine that skeletogenic cells will form at the vegetal pole in every embryo of this species, and that these cells will be surrounded by other mesodermal cell fates and the endoderm that gives rise to the gut. The developmental organization of these cell fates within the embryo is an important function of the GRN. Thus, the GRN ensures that the set of transcription factors associated with each endodermal and mesodermal cell fate are expressed in the correct position within the embryo. The GRN also controls which downstream differentiation genes are expressed in each cell fate. For example, proteins involved in synthesizing a skeleton are expressed in skeletogenic cells (Rafiq et al., 2014), whereas proteins with digestive enzymatic functions are expressed in the gut. The purpose of experimentally dissecting GRNs is therefore to obtain a causal understanding on how the genome controls the developmental organization of an embryo (Peter and Davidson, 2015).
The endomesoderm GRN model shown in Figure 3A shows how the expression of transcription factors is regulated in each endomesodermal cell fate (Davidson et al., 2002a;Longabaugh et al., 2005;Oliveri et al., 2008;Peter and Davidson, 2011;Longabaugh, 2012;Materna et al., 2013). Each of the colored boxes represents a distinct cell fate, and the genes shown in each box together compose the regulatory state of the corresponding fate. For each gene, the regulatory inputs that regulate its expression are shown as linkages into the associated cis-regulatory system, while the regulatory functions of the transcription factor outputs are shown as linkages into the regulatory systems of target genes. In a GRN, individual cis-regulatory systems are therefore connected through the transcription factors with which they are associated. We have seen in the example of endo16 how individual cis-regulatory binding sites and modules are computed to control expression of a single gene. But if we extrapolate this to the level of a GRN, how does the logic that is encoded in the regulatory systems of different genes operate when connected within a network? Do these systems operate independently, or is there an intrinsic logic to combining several genes into a network circuit? Is there a degree of freedom to connect cis-regulatory systems, or are there specific rules for the compositionality when combining the logic of cis-regulatory sequences?
A computational model of the endomesoderm GRN provides perhaps some answers to these questions. Thus, in an attempt to capture the dynamic behavior of a system of interconnecting regulatory genes, a Boolean logic model was mathematically defined based on the experimental observation of the endomesoderm GRN (Peter et al., 2012). The purpose of this computational model was to test whether the GRN that was reconstructed as shown in the topological model in Figure 3A would suffice to explain the developmental control of gene expression and the specification of different endomesodermal cell fates in the sea urchin embryo. The basic components of this model are (1) the regulatory logic controlling each gene in the GRN, identified based on the effect of experimental perturbation of its transcription factor inputs; (2) a temporal delay function associated with each regulatory interaction; and (3) maternal inputs that initiate the activation of the zygotic developmental program.
The regulatory logic of each gene in this model was captured by Boolean logic statements that were formulated based on the regulatory inputs controlling expression of each gene and based on the logic operation computed by the cis-regulatory system (Peter et al., 2012). For instance, if perturbation experiments indicate that two transcription factors A and B regulate the expression of gene C, and that perturbation of either A or B leads to a strong reduction of expression of gene C, this would indicate that the presence of both A and B is required for activation of gene expression. The Boolean logic statement that captures the regulatory logic for gene C would therefore be C = A AND B. The time function in this model derives from the time it takes from starting transcription of an upstream regulatory gene to producing levels of transcription factor sufficient to regulate target gene expression. This time was calculated based on RNA and protein synthesis rates to be *3 hours in sea urchins developing at 15C (Bolouri and Davidson, 2003). Surprisingly, using a temporal step function of 3 hours for almost all regulatory interactions was a correct assumption to reproduce the temporal and spatial expression of almost all genes in the system. And finally, the maternal inputs are transcription factors that are present in the egg and in the model these factors are turned ON by default for the first few hours of development. This Boolean computational model was used to compute expression or absence of expression for each gene in the endomesoderm GRN based on a Boolean logic statement and based on the presence or absence of its inputs. Except for the maternal factors, these inputs are present only if the corresponding regulatory genes are computed as expressed based on their own associated logic statements. The computed expression for all regulatory genes in the GRN model is shown in Figure 3B for the endoderm domain during 30 hours of development. A comparison of the gene expression patterns computed by the Boolean logic GRN and the gene expression experimentally observed in the sea urchin embryo demonstrates that the information captured in the GRN is sufficiently complete to recapitulate the embryonic gene expression patterns. Thus, this system behaves as an automaton, where early maternal inputs initialize a program that is self-sufficient to operate without any further inputs from outside the system. This analysis shows that it is possible to obtain a complete explanation for developmental gene activity based on the experimental analysis of a GRN. In addition, these results suggest that the cis-regulatory systems at each network node can be combined without further instructions to reproduce the correct system-level output of an entire GRN both in terms of gene expression and in terms of differential cell fate specification.
The observation that a system of interconnected cis-regulatory modules is sufficient to capture an entire developmental program provides a powerful confirmation of the information processing function of the regulatory genome. Of course, although the examples here derive from the sea urchin embryo, the computation of logic functions by regulatory sequences must represent a general property of the regulatory genome that applies to sea urchins as well as to any other organism, in development and beyond. It demonstrates that the regulatory logic controlling individual genes can be viewed as a system of logic gates that compute developmental gene expression. Returning to von Neumann's quote, although mathematical analysis is applicable to many areas of biology, the system-level information processing functions of the regulatory genome might be better approximated by formal logic. Thus, the logic encoded in regulatory DNA provides a unifying concept that defines the function of the regulatory genome, from the modular regulatory systems controlling individual genes to the networks controlling genome activity throughout biological processes.
HOW DOES THE REGULATORY GENOME WORK? 693 | 2019-06-07T20:32:29.372Z | 2019-07-01T00:00:00.000 | {
"year": 2019,
"sha1": "716ec0be65f46affc75bec946334100acd5686f6",
"oa_license": "CCBYNC",
"oa_url": "https://www.liebertpub.com/doi/pdf/10.1089/cmb.2019.0097",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "07cbb1baca4aaec6454f81d3e915e7a3a23725fa",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine",
"Computer Science"
]
} |
233859935 | pes2o/s2orc | v3-fos-license | Overview of water quality modeling
Abstract Due to population growth, urbanization, and industrialization, water demands have increased, and the quality of water is degraded. Water quality modeling is a significant tool that aids managers and policymakers in multiscale integrated water resources and environmental management. However, water quality modeling is challenging due to several constraints. The modern application of modeling is essentially utilized by the need to comply with rules and regulations. In view of this, water quality modeling requires the standardization of models, identification of common features of models, the hotspots of pollution, and the current state of policy-relevant models. This review presents an overview of water quality modeling and major models frequently applied for water quality assessment at the catchment and at waterbody scales. This review is intended to highlight the applicability of certain water quality models, and the state of water quality modeling, model classification, and uncertainties. Water quality models are described and selected based on their applicability, strengths, weaknesses, and intended use. Some models are applicable for the specific waterbodies, simulate selected water quality parameters, have uncertainties, are not commercially available, require skilled model users, and have huge data requirements. When selecting suitable models, it is recommended to consider the availability of data, model complexity, and type of waterbody, and intended objectives to be modeled.
PUBLIC INTEREST STATEMENT
Water quality modeling helps decision and policy makers to decide the best and practically resilient solutions for water quality management. Tangible results from water quality modeling are significant to make successful decisions and directions with the root causes of the environment and waterbody. Some scholars have been done on water quality modeling; however, some water quality models are improved, new models are developed and the gap of the model is identified. Hence, as the need for recent water quality model characteristics knowledge, this review on water quality modeling is essential. The principles, application, gaps, advantages, and types of water quality models have been reviewed. The review will contribute to the selection of an appropriate model for varied water quality problems and different waterbodies, illustrates available models and model uncertainties.
Introduction
Water is one of the key elements of the environment that determines the survival of life and restricts the socio-economic growth of the people (Stolarska & Skrzypski, 2012). Overseas and inland surface and sub-surface water systems play an incredible role in everyday life activities mainly for drinking, agricultural, industrial, recreational, and other public uses. Our everyday lives depend on the availability and quality of water. Accessibility of suitable water quality for different purposes (EPA (Environmental Protection Agency), 2003;FAO, 1975;FAO, 1979a andJonnalagadda et al., 1991) is becoming difficult due to rapid population growth and expansion of agro-industries. Some industrial, agricultural, and human activities have a serious effect on ecological diversity. In addition, surface water quality depends on natural phenomena; the quality of water in lakes and dams is suffering from incessant degradation due to natural processes resulting from eutrophication and anthropogenic causes (Stolarska & Skrzypski, 2012).
Water resources in the world are under pressure; particularly eutrophication is a major environmental problem. Eutrophication is caused by high loading of dissolved and particulate organic matter and inorganic nutrients (Goshu & Aynalem, 2017) due to lack of proper soil and water conservation practices in the watershed (Teshale et al., 2002;Yitaferu, 2007), inappropriate wastewater disposal, and the lack of wastewater treatment technologies. The continuous urbanization and industrialization with increased human activities have a negative impact on water quality and adversely affect the aquatic ecosystem. Human activities are one of the major factors of pollution in environments (EIC (Environmental International Consultant), 2009). Urbanization, industrialization, and agricultural practices have a substantial contribution to water pollution, and they discharge a huge quantity of organic and inorganic pollutants into the waterbodies. Most of the industries mainly in developing countries like Ethiopia discharge the wastewater into the environment without proper treatment. Mainly the cities have the highest pollution rate with inadequate waste management systems that have resulted in excessive pollution of receiving waterbodies. Sedimentation is also a big problem for the lakes, rivers, dams, and coastal areas. According to EEPA (Ethiopia Environmental Protection Agency (2003), some industries located in Addis Ababa have discharged 90 percent of their waste into the nearby waterbodies and open spaces without treatment. The discharges from agricultural inputs (fertilizers and pesticides), domestic and industrial wastes pollute freshwater systems and endanger the socio-economic and ecological values.
Contaminations of ecosystems commonly are organic (pesticides and hydrocarbons) and inorganic (phosphates, nitrates, and metals). The increased concentrations of nutrients cause eutrophication that is a manner of increased algae growth and growth of other forms of plants in the aquatic system (Hussein et al., 2015). Eutrophication can result in the reduction of the dissolved oxygen in the waterbody (X. E Yang et al., 2008;Conley et al., 2009). Watercourses are essential areas of the ecosystem, and are a source of water for human activities and provide habitat for aquatic animals. However, the discharge of pollutants from different sources results in the scarcity of reliable water quality for the definite functioning of ecosystems (L. Yang et al., 2013;Kumarasamy, 2015). For most of the rivers and lakes in urban areas of developing countries, the wastewater effluents at the downstream section are discharged from industries. Industrial sectors are liable for dumping 300-400 million tons of heavy metals, solvents, toxic sludge, and other waste into water sources globally each year (UNEP (United Nations Environment Programme), 2012). The highest ecological pressure in the environment is the one due to river water pollution. For instance, in Lake Tana, Ethiopia, the quality of water has extensively deteriorated over the years. Recent studies have shown a serious decline in fish stocks due to the spread of aquatic weed, water hyacinth around fish spawning grounds in Lake Tana (Solomon, 2017).
The forgoing problems mainly arise from imbalances of development interventions and environmental protection activities (Teshale et al., 2002). Major indicators of this imbalance include: accumulation of persistent pollutants in fish specifies located in reservoirs, decreased fish production, and extended reservoir eutrophication periods, growth of water hyacinths on the reservoir surface, increased costs of operating water treatment plants, increased reservoir sedimentation, and irrigation water quality reduction. Accordingly, the effects of the development of projects and other linked activities on the environment have to be considered before implementation in order to achieve the successful management of the environment and reducing the effects of wastes on the ecosystem.
Water resources and environmental management require continual monitoring in terms of quality and quantity. Proper assessment of the degree of water pollution is used as the benchmark for management and well-adjusted utilization of water resources. One of the basic approaches, which are required to solve the water pollution problem is the modeling of water quality changes. In the past years, the development of mathematical modeling has been rapid (Stolarska & Skrzypski, 2012) and many models have been utilized to date so far. Mathematical models have been applied to assess water quality changes as a result of wastewater discharge vagaries (World Bank Group, 1998). Various mathematical water quality models have been developed and applied by some researchers to study the quality of water in many countries, for example, in Poland (Stolarska & Skrzypski, 2012), the USA in the Shenandoah River watershed (Mbongowo et al., 2019), andFlorida Bay (Carl et al., 2000), and South Korea in Ara artificial canal (Zhenhao & Dongil, 2013). Water quality models have been applied for water quality assessment in various waterbodies, onedimensional (1D) and two-dimensional (2D) coupled hydrodynamic and water quality models were used in the Ca Mau Peninsula (Tri et al., 2018), development of surface water model (Q. Wang et al., 2013;Kayode & Muthukrishna, 2018), mathematical models have been applied for reservoirs (Stolarska & Skrzypski, 2012), the Rio Chone estuary (Stram et al., 2005), the East Johor and Singapore Straits (Sundarambal & Pavel, 2014) and Kas bay (Kagan & Lale, 2015). These studies are important for the planning and management of waterbodies. A summary review of water quality models has also been done (Bai et al., 2011;Q. G. Wang et al., 2009). One of the purposes of a water quality study is to determine its suitability for the intended use. For water quality modeling, the primary and secondary water level and water quality data have been used (Tri et al., 2018).
Investigative water quality models are both mathematical expressions and expert scientific judgment, which comprise process-based (mechanistic) and data-based (statistical) models. They are effective resources for assessing and predicting pollutant transport (Q. G. Wang et al., 2009;Bai et al., 2011,;Huang et al., 2012), identifying pollution, fate, and behavior of pollutants (Q. G. Wang et al., 2009), simulating and forecasting complex processes in water ecosystems (Liu, 2018), recognizing the spatial and temporal distribution of pollutants in the water, and intensifying decisions regarding how water quality will be altered (Q. Wang et al., 2013). Water quality models are also needed for investigating the environmental situation of different waterbodies, evaluating the variation in water quality when initial or boundary conditions are changed (Kayode & Muthukrishna, 2018), predicting long-term surface water quality, and performing environmental impact assessment with different pollution scenarios (Q. Wang et al., 2013). The models play on an important role in reducing the cost of labor, materials, and time for a large number of pollution mitigation scenario experiments to some degree (Q. Wang et al., 2013), for watershed and watercourses in the ecosystem. Many types of water quality models have been used to simulate the quality of water in several sorts of water systems comprising rivers, streams, lakes, reservoirs, estuaries, coastal waters, and oceans (Loucks & Van Beek, 2017). Since then, several water quality models have been established with different model algorithms (Liou et al., 2003;Q. Wang et al., 2013) by various organizations and researchers. However, due to different theories and algorithms used in the models, the modeling outputs of different models have big differences, as a result, the use of different models may provide different environmental management decisions when the outputs cannot be referenced to or associated with each other (Obropta et al., 2008). Assessment of pollutants through monitoring is a challenging task, which requires a continuous update of existing models and the development of new water quality models. The first study on water quality modeling for simulating BOD and DO in a river system was done by Streeter and Phelps in 1925 (Cox, 2003 andChapra, 2008).
Water quality modeling applies in the estimation and prediction of water pollution using mathematical simulation techniques. An illustrative water quality model consists of a collection of formulations, representing physical mechanisms that determine the position and movement of pollutants in a waterbody (Victoria, 2012). Mathematical water quality modeling is considered as one of the best approaches to estimate the existing pollutant load, pollutant transfer, and upcoming cause-effect relation between pollutant sources and water quality (Nair & Bhatia, 2017). Water quality modeling allows decision and policy makers to choose better, more technically strong solutions among alternative possibilities for water quality management. The models are required to determine better alternatives for solving sustainable water quality problems in the long term. In addition, models are essential to provide a basis for economic analysis, and then decision-makers can use the output to assess the environmental implications of a project and the cost-benefit ratio. The quality of water has been evaluated and modeled with its physical, chemical, and biological characteristics. The relations between the processes related to these characteristics are necessarily multifaceted, and water system managers must pursue to develop a worthy understanding of the main factors and processes that affect the water quality of each local water resource they are responsible for if they are to make correct or improved management decisions (Liu, 2018). The concentrations and distribution of contaminants are influenced by the inactivation of contaminants and some dynamic processes, including diffusion, dispersion, and advection. These processes are closely related to the water flow characteristics, influent and effluent entering and leaving, respectively, the waterbody, wind stress, and temperature stratification (Liu, 2018). Hence, for a better understanding of how the environment and water systems are polluted and to make fruitful decisions and directions, concrete knowledge derived from water quality modeling is significant.
Integrated water quality assessment such as physical observation, and model-based simulation can be used by agencies, resource managers, planners, scientists, engineers, project implementers helping for achieving basin-wide and at large and small-scale level load reduction goals. The water system normally is studied with various approaches including theoretical analyses, mathematical modeling, laboratory tests, and field observations. Laboratory analysis and field observations are the most reliable ways to acquire tangible information for a specific system, which will provide a reliable basis for analysis and modeling. Meanwhile, observed or measured data are typically rare and not sufficient to indicate or predict a complete picture of the real scenario in the large and complex waterbody. Moreover, the available data may not necessarily be very reliable and low-quality data with high errors that might lead researchers to compose a wrong or misleading idea of what is actually happening. Thus, mathematical water quality modeling coupled with observations for verification and calibration is essential in such cases. Integrated models such as hydrodynamic and water quality models have been extensively developed and used (Liu, 2018) for assessing lakes, rivers, reservoirs, ponds, estuaries, and coastal water in many aspects.
Due to the need of water quality models, several authors have rewritten review papers on water quality modeling (Beck, 1987;Tsakiris & Alexakis, 2012;Loucks & Van Beek, 2017;Rauch et al., 1998), developed models (Ambrose et al., 2009) and modeled with different water quality models (Q. Wang et al., 2013;Sundarambal & Pavel, 2014;Whitehead, 2016;Yuceer & Coskun, 2016). Therefore, in the view of up-to-date modeling concepts, improved model development, simulation, and prediction of water quality in a changing environment, it is essential to review water quality modeling regarding the principles, application, analysis, and development of water quality models by various scholars and organizations. The current review will support the selection of a suitable model for diverse water quality problems (Kayode & Muthukrishna, 2018) and different waterbodies highlight available models and model uncertainties.
Significance of water quality modeling
Water quality management is an essential component of overall integrated water resources management (UNESCO, 2005). The output of the model for different pollution scenarios with water quality models is an imperative component of environmental impact assessment (Q. Wang et al., 2013). Sound water quality is very limited in the world and more care to water quality modeling is inevitable (Davies & Simonovic, 2011). Water quality models have been utilized to study and determine existing circumstances for assessing potential impacts because of human activities (Hicks & Peacock, 2005). Water quality models are decision support tools for simulating the fate of pollutants in water and assessing their related hazards (Chapra, 2008;Q. Wang et al., 2013). Water quality modeling is essential to develop a perfect conceptual model based on the existing information, understand the transport regime of pollutants, test hypotheses, quantify the dominant controlling processes, and certify with governing principles and observations. Also, Water quality modeling is vital to understand the history of pollutant transport and to determine time ranges in which a pollution incident might have started or contaminants have reached a targeted concentration in areas of concern (Zheng and Gordon, 1995).
The purpose of modeling is to solve problems of surface water pollution and to track water quality changes (Chapra, 1997;Holnicki et al., 2000;Stolarska & Skrzypski, 2012). Water quality models are applicable to analyze the existing phenomena, predict and compute effects of changes in the aquatic environment, set limits for pollutant discharge or load, identify the location of sources of pollution and causes of water quality deterioration on a given segment of the stream, and selecting an optimal approach for sustainable development (Holnicki et al., 2000;Chapra, 1997). Various water quality modeling methods with diverse commercial software packages have been used in different studies; for instance, QUAL2K and HEC-RAS were used for the Keelung River in Taiwan (Fan et al., 2009) and QUAL2EU was used in the Yamuna River, India (Hussein et al., 2015) to evaluate the quality of water.
Water quality modeling problems and model standardization
Water quality modeling is challenging due to some constraints including lack of experience of a model user , sufficient representative site selection and sample gaps, and lack of calibration, errors in data reporting (Chapman, 1996). In some cases, a number of models are location and parameter specific; they depend on waterbody type, do not conform to certain dimensional analysis, and may not have the ability to model point and nonpoint source pollution together. The uncertainty in water quality modeling commonly comes from numerous sources of errors: measurements of input and response uncertainty (Rode & Suhr, 2007), parametric uncertainty, and structural error due to the incapability of a specified model structure to reproduce the physical mechanisms (Montanari, 2004). In most developing countries, a uniform model standardization system has not been recognized (J. Q. Wang et al., 2004;Cao & Zhang, 2006) that limits extensive utilization of those models for ecological and water management as a result of the lack of benchmarks and comparisons between different modeling outcomes (Q. Wang et al., 2013). Several serious problems in catchment-scale water quality modeling have spatial variability, which commonly takes over the catchment behavior, proper selection of representative sites, and integration of nonlinear biogeochemistry (Rode et al., 2010). On the other hand, the model complexity, lack of data, and poor data quality are other limiting factors for water quality modeling.
To fruitfully apply respectable model system regulation, it is very essential for most developing countries to develop their model standardization. Model standardization favors a sound understanding of the accessibility, accuracies, methods of computation, calibration, and development of various water quality models (Cao & Zhang, 2006;Politano et al., 2008). The modeling outputs are important and variable, and thus water quality models have to be more standardized, accessible, and consistent when they are applied to support programs to meet water quality standards and legal documents (Q. Wang et al., 2013). Water quality models could be structured and standardized through procedures described in recognized published research articles, workshops, or by setting up a resident workgroup, the formation of national model assessment indicators, and an authentication system (EPA (Environmental Protection Agency), 2003). Some developing countries need to standardize some widely utilized water quality models for effective environmental impact assessment. Standardization of water quality models will support environmental management agencies' guarantee in the uniform application of water quality models for regulatory uses (Q. Wang et al., 2013).
Model classification and selection
Numerous commercial and open-source models have been applied to simulate complex water quality processes in diverse environmental conditions. Some of these models including MIKE21 (Chapman, 1996), HEC-RAS RAS (US Army Corps of Engineers, 2014, 1998), QUAL2K (Fang et al., 2008), WASP 6 (Artioli et al., 2005), QUASAR (Lees & Sincock, 2002;Whitehead et al., 1997), and SWAT (Grizzetti et al., 2003). Because of data, the existence of captures the core of the problem, the simplest reliable model is always chosen over complex models. An excessively complex model will have increases the computational time, cost, and leads to extra uncertainties if detailed data are not available (Zheng and Gordon, 1995). Kayode and Muthukrishna (2018) have assessed the AQUATOX, QUAL2E, WASP, CEQUAL-RIV1, MIKE11, SWAT, and SIMCAT models and have defined their capabilities and applications for different waterbodies. MIKE 11 and QUAL2E do not consider the denitrification process for the period of its operation. Besides, QUAL2E and SIMCAT cannot model variable flow conditions since the flow rate is expected to be steady state. SWAT has the capacity of simulating in-stream fate and transport of a wide variety of pollutants, and it can be coupled with an in-stream model to provide a good outcome.
Model selection
Water quality models can be selected based on different criteria such as model complexity, availability of data, type of waterbody, water quality simulation capabilities, easy accessibility of the program code source, and the existence of good certification of the model (Kayode & Muthukrishna, 2018). Also, the model applicability, cost, familiarity, and support are criteria used for selecting suitable water quality models. Many water quality models are not widely used and are no longer being updated to comprise the latest developments. For deciding the most suitable models, it is essential to assess the existing water quality models (Smith Warner International, 2005). Different water quality models are widely used around the world, which has advantages and limitations. Nevertheless, the model must be calibrated and validated for an acceptable outcome. The U.S. Environmental protection Agency (Grimsrud et al., 1976) has described the model selection process and criteria with four levels of selection phases for models. The selection process is intended to provide users guidance to select some levels of features they might require for the problem to be solved. The phases of the model selection processes are Phase I (model applicability test), Phase II (cost constraint test), Phase III (performance index rating; simplified), and Phase IV (performance index rating; advanced). The first two phases are the elimination of unsuitable models and the last two phases are the ranking of the remaining models. The exclusion of available models in phase I is essential to condense the assessment of models in the next phase.
In the first phase, the elimination of models is done based on the suitability of the model to the problem at hand (the type of waterbody, time variability, discretization, special features, constituents modeled, model input data, driving forces, and boundary factors). As per the constraints, those candidate models that do not meet the users' requirements can be rejected. However, if the assessment shows that the model is possibly applicable, it might be preserved for further consideration in the next phase. If all the available nominated models are excluded in the first phase, then the user must search either for a new model; reexamine the applicability constraints, or unrestraint effort requirements to use a water quality model for the design requirements. Phase II also is an eliminatory phase, in which the model is selected based on cost (data acquisition, mechanism model and workers costs, acquisition, and equipment requirements). The third phase (Phase III) is required to rank models based on weights related to the criteria from phases II and I. Phase IV is the last phase of the model selection process, and it needed for advanced ranking of models based on relevant processes including accuracy (numerical stability, model representation, and dispersion), the competence of available model certification, ease of modification, data input design, and output form and content (Grimsrud et al., 1976). Similarly, models can be selected with simple approaches (Marcos et al., 2018). In general, before performing water quality simulation, suitable models should be selected. In order to select the type of water quality models required for different waterbodies, various factors should be considered, in particular inspecting the type of pollutant problem affecting the water system, determining the cause of water pollution, identify the best management practice solutions, the modeling objectives, and the available resources are essential. Also, identifying the project goal is essential when developing a water quality modeling tool through discussions with the stakeholders, regulating agencies, and technical personnel involved in the development (Kayode & Muthukrishna, 2018).
Model classification
Water quality models can be classified as physical (laboratory) and mathematical (analytical) models (Holnicki et al., 2000;UNESCO, 2005). Besides, they can be categorized according to the complexity of computer simulation (1D, 2D, and 3D models), data requirements (extensive databases and minimum data requirements models), type of approach (physically based, conceptual and empirical), pollutant type (nutrients, sediment, and salts, etc.), area of application (catchment, groundwater, river system, lake, coastal waters, integrated), nature (deterministic or Stochastic), state analyzed (steady state or dynamic simulation), and spatial analysis (lumped, distributed) (Tsakiris & Alexakis, 2012). The advantages, disadvantages, applicability, and assumptions for 1D, 2D, and 3D models are summarized in Table 1. On the other hand, based on the extent and spatial scales, some models are considered as operational, tactical, strategic, and directional models (Stolarska & Skrzypski, 2012). The SKM (2011) classification scheme has described categories for three types of water quality models such as catchment models (derive flows from the rainfallrunoff process and simulate related pollutant loads), in-stream models (simulate hydrodynamic behavior of flows and in-stream water quality processes), and ecological response models (simulate the ecosystem response to stressors, such as flow and water quality). Water quality models have been classified as steady state models (planning models: intended for long-term trends and routine monitoring), dynamic/stochastic models (design: short-term dynamics continuous and long-term trends and routine monitoring tasks), and dynamic models (operational: which is required for short-term dynamics and continuous monitoring for operational management) (Whitehead, 2016). Water quality models can also be classified as a simulation model and optimization model (Chapra, 2008;;Sharma & Kansal, 2013). The simulation model defines and represents changes in water quality in some mathematical form. However, optimization models are commonly applied to find the smallest number of alternative data before doing the model simulation. The models are typically classified with respect to model complexity, type of receiving water, and water quality parameters to which the model can be predicted. When the model is more complex, it is highly difficult and expensive for application to a given condition because of the data requirements (World Bank Group, 1998).
The major principle governing model preparation is the law of conservation of energy, momentum, and conservation of mass (Chapra, 2008). There are different formulas that can be followed to develop a water quality model and each application depends on the different types of parameters to be modeled (Kayode & Muthukrishna, 2018). Several water quality models (i.e.
Parameters and data for water quality modeling
Water quality modeling needs information and data to predict existing and future water quality situations in the ecosystem as a function of the baseline conditions and pollutant loads (Grimsrud et al., 1976). To model water quality, several parameters and data are required as input ( Table 2). The input data needed to investigate water quality characteristics can be found from the literature, organizations, measured directly in the field, or determined through model calibration (Liu, 2018). The availability and accuracy of data is a great concern in the development and use of models for water quality analysis and management (Loucks & Van Beek, 2017). The amount and quality of available input data will govern the complexity of the model to be applied for simulating water quality parameters (Kayode & Muthukrishna, 2018). The necessities of data for water quality models intensify with model complexity and range of applications (Rode et al., 2010). Existing input data and information that would be relevant for water quality modeling include initial and boundary concentrations, source of pollutants, baseline conditions, flow characteristics, the geometry of the modeled waterbodies (river, lake, coastal, ponds, reservoirs, etc.). Moreover, the initial Table 1. Features of 1D, 2D, and 3D water quality models (Balcerzak, 2000)
Model Advantages Limitations Applicability & assumptions
One-dimensional (1D) models • Used quickly for lake and reservoir water without pre-calibration and with a small available database of measurements • The simplest and most commonly used models in the analysis of river water quality.
• Do not describe the complex chemical, physical and biological reactions in water • Are not designed to estimate the variation of concentration with time.
• Assume significant changes in determining the water quality parameters arising only along with the longitudinal profile of the watercourse, • Valid in long creeks, rivers, streams, and narrow channels (Lubo, 2018).
• Typically applicable to rivers, and for estuaries and lakes with large length-width ratios.
Two-dimensional (2D) models
• Can be significant to examine water quality at various depths, • The study of individual parameters can be made at different time intervals (hour, day, week, month, and year).
• The models need more data and more skilled analytical users than one-dimensional models, • Requires careful calibration and is sensitive to changes in many parameters of water quality.
• Assumes significant changes in water quality occur both in along and the longitudinal profile of the watercourse • Applicable to simulate water quality mostly in reservoirs, deep rivers, and lakes.
Three-dimensional (3D) models
• Required to estimate the spatial distribution of concentrations of simulated water quality parameters • Need vast amounts of data and more skilled analytical experts • Due to the high complexity of the examined issues, the models are rarely used.
• Applicable to examine changes in water quality in reservoirs, dams, deep rivers, lakes, estuaries, and sea bays. Emphasizes on oxygen balance, and firstorder decay of BOD (Q. Wang et al., 2013).
•
The model is based on the assumptions that a single BOD input is distributed evenly at a cross-section of a stream/river and it moves as plug flow without mixing in the river (Lin & Lee, 2001), • One DO sink (carbonaceous BOD, CBOD) and one DO source (reaeration) only are considered in the standard (Schnoor, 1986). These overviews will increase errors in the model.
QUAL
1D steady state/dynamic model. Suitable to simulate contaminants in well-mixed streams and rivers (Brown & Barnwell, 1987). Usually applied to analyze the effect of point source discharge changes on water quality, including the impacts of nutrients on algal concentration and DO (World Bank Group, 1998). Vital to the analysis of the spatial and temporal variations of nutrients, T, BOD and DO concentrations in the water column (Kannel et al., 2011).
• Because of its steady-state assumption, it is not capable to simulate a river, for which temporal flow variation affects key water quality conditions • Not considered suspended sediment movement, macrophytes, and denitrification processes.
• In model development, the reaches and computational elements must not be more than 25 and 20 per reach or a total of 250, respectively.
•
The headwater and junction elements would have a maximum value of seven (Kayode & Muthukrishna, 2018).
• Forecast the direct and indirect effects on the resident organisms, • Analyze the transfer of biomass and chemicals from one section of the ecosystem to another.
• Aids to recognize the cause and effect dealings between the chemical water quality, physical environment, and aquatic life.
•
Is not capable to model metals and impossible to couple with hydrodynamic models, • While simulating the change in nutrients, chemicals, and sediment concentrations, it supposed a unit volume of water in the waterbody, • The internal nutrients are not represented in algal bioenergetics (Kayode & Muthukrishna, 2018).
(Continued) Has inadequate eutrophication kinetics in its processes and requires extensive experiment by users.
• Not suitable to simulate sediment transport processes in the river (Kayode & Muthukrishna, 2018).
•
The equations are written in the conservative form via Boussinesq and hydrostatic estimates.
• Vertical momentum is not considered and might give rise to imprecise outcomes, where substantial acceleration is present, • Its application is a complex and timeconsuming task (Cole & Wells, 2013).
• Well-mixed in the lateral direction and hydrostatic hypothesis for the vertical momentum equation
•
The main equations are laterally and layer averaged, • Due to the complications of the model, data to drive the model can be a prime restraint, • Computational and storage burden on a computer when making continuing simulations, • Accessibility of input data is the limiting issue for the application or misuse of the model (Cole & Wells, 2018).
(Continued) • Can universally to model the transport and contaminant fate in surface waterbodies.
• Vital to the analysis of some water quality characteristics including conventional pollutants (N, P, bacterial contamination, T, BOD, sediment oxygen demand), nutrients, metals, and toxic chemical movement (Stolarska & Skrzypski, 2012).
• Model calibration and application to predict water quality parameters necessitate is time-consuming for the user.
• Needs far-reaching training for users due to its complexity.
• Cannot predict the results of control structure (Nasser et al., 2017).
• Cannot run in batch mode, • Does not handle variable processes (i.e. mixing zone processes, non-aqueous phase liquids, segment drying, and metals speciation), • Large external hydrodynamic file (USEPA, 2005).
• Requires a large amount of data for calibration and verification, • Does not handle mixing zones or near field effects, sinkable/floatable materials, • Cannot successfully simulate suspended solids loading in the river (Kannel et al., 2011).
(Continued) • Required for recognizing the degree of salinity intrusion effects at varied river flow rates (Liangliang & Daoliang, 2014).
flow condition of the water system, inflow water quality concentrations, waterbody type, time, kinetic parameters, calibration, bathymetric, meteorological and flow data (Smith Warner Cole & Wells, 2018;International, 2005;Kayode & Muthukrishna, 2018), the water level at sampling, major aquifer type, aquifer map, flow duration statistics, lake and piezometric levels between sampling are required (Chapman, 1996). Nevertheless, the required input data used for water quality modeling depends on the expected objectives, waterbodies to be investigated, models applied for simulation, and data availability and quality.
The water quality of lakes, rivers, estuarine, streams, ponds, reservoirs, etc. can be characterized by various physical, chemical, and biological parameters including temperature, PH, the concentration of dissolved minerals, turbidity, salinity/total dissolved solids (TDS), Na, Ca, Mg, K, bicarbonates, suspended solids, nitrate, BOD, DO, chloride, the concentration of nutrients, bacteria/ coliforms, suspended sediment, nutrients, algae concentration, sulfates, heavy metals, CBOD and other constituents. These constituents can be analyzed with models; however, according to Liu (2018), the modeling outcomes of several parameter values shall be compared with the observed data. The model parameter values that realize the best match between measured and predicted constituents concentrations can be selected for advanced modeling runs.
Conclusions
Water pollution is one of the worldwide challenges facing both developed and developing countries. The cities mainly have the highest pollution rates because of inadequate waste management systems and urban runoff pollution. Water pollution problems are usually due to economic growth and have an impact on both the environment and human health. The causes of water contamination include soil erosion, deforestation, habitat destruction, improper waste management, overgrazing, lack of awareness in management, shortage of decision support tools, and an inadequate organized database system. In the current situation, the ecosystems require a sustainable management solution for improved socioeconomic development. For a better understanding of how the environment and water systems are polluted and to make fruitful decisions and directions, concrete knowledge of water quality modeling is needed. This review described an overview of water quality modeling emphasizing modeling application, commonly used water quality models, and model selection, application, and limitations for different waterbodies. Water quality modeling is a significant tool that helps to water managers and policymakers applying for unified water and environmental management. Scholars, policymakers, and designers through rules and regulations significantly require the practice and application of water quality modeling. Models have been applied to simulate various water quality characteristics and evaluate water quality changes as a result of wastewater discharge to the ecosystem. Various water quality models have been developed and applied in some countries to study the quality of water in various waterbodies with 1D, 2D, and 3D simulations. Different types of water quality models including some extensions of model software have been developed for predicting in different topography, waterbodies, and pollutants at different space and time scales. Water quality model studies are very important for providing solutions and directions towards sustainable planning and management of waterbodies.
One of the requirements for water quality modeling is to determine its suitability for the intended use. However, water quality modeling has several limitations. Some models are applicable for specific waterbodies, simulate selected water quality parameters, have uncertainties, require skilled model users, are not commercially available, and require a huge amount of data. On the other hand, various water quality models can be integrated with other hydrodynamic and hydrological models. Every model has its particular sole purpose and simulation features. Many countries are working to develop guidelines on water and environmental quality investigation and management, providing regulated models for water quality prediction. Consequently, it is advisable to standardize water quality models for all countries. When developing water quality modeling and selecting suitable models for a waterbody, it is vital to make the selection through discussions with stakeholders, based on modeling objectives, time, and available resources (data, cost, etc.). However, to meet the accuracy of the study objectives, the user should know the assumptions and model uncertainties. | 2021-05-07T00:02:58.680Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "b6d4b7c3a92c61b88d584d372b9052cd72ab9280",
"oa_license": "CCBY",
"oa_url": "https://www.tandfonline.com/doi/pdf/10.1080/23311916.2021.1891711?needAccess=true",
"oa_status": "GOLD",
"pdf_src": "TaylorAndFrancis",
"pdf_hash": "de8bb6410ae1d2a440c14af2c73d649ad3d948db",
"s2fieldsofstudy": [
"Environmental Science"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
261759399 | pes2o/s2orc | v3-fos-license | Time‐limited balanced truncation within incremental four‐dimensional variational data assimilation
Four‐dimensional variational data assimilation (4D‐Var) is a data assimilation method often used in weather forecasting. Based on a numerical model and observations of a system, it predicts the system state beyond the last time of measurement. This requires the minimisation of a functional. At each step of the optimisation algorithm, a full nonlinear model evaluation and its adjoint is required. This quickly becomes very costly, especially in high dimensions. For this reason, a surrogate model is needed that approximates the full model well, but requires significantly less computational effort. In this paper, we propose time‐limited balanced truncation to build such a reduced‐order model. Our approach is able to deal with unstable system matrices. We demonstrate its performance in experiments and compare it with α‐bounded balanced truncation, which is an another reduction approach for unstable systems.
derived [6].The quantity of interest is the maximum a posteriori estimate * 0 = arg min 0 ( 0 ), where the cost functional that needs to be minimised is given by subject to the forward model dynamics (1a) for , the state at time , and ‖ ‖ 2 ∶= T .This setting is called strong constraint 4D-Var and has its origins in weather forecasting [1], where the name 4D-Var refers to the three spatial dimensions and the additional evolution in time.The minimisation of the functional ( 0 ) requires to run the possibly nonlinear model (1) and its adjoint multiple times and is thus very costly to compute.In data assimilation, the incremental 4D-Var approach [3] is often used.It uses successive linearised versions of the operators and then minimises a quadratic functional, just like the inexact Gauss-Newton Method [5].The linearisation of the operators and around leads to the Jacobians and in the so-called tangent linear model: where denotes the state perturbation and ∶= − ( ) is the distance between the measurement and the output of the full model at time .Problems where 4D-Var is applied especially appear in weather forecasting [1,7] where one typically has to deal with unstable matrices after linearisation.This will be addressed by our model order reduction approach later.The quadratic cost functional derived from (3) that has to be minimised for 0 is given by subject to the forward model dynamics (3a).The minimisation of ( 2) is now done as an inner-outer iteration: In the outer loop the current guess for the initial condition is updated by ] and its expected observations for the computation of (4).The idea of obtaining * 0 incrementally is promising, but the minimisation of ( 4) is very costly for high dimensions, even in the linear setting.Instead of using the full model, we want to project the system (3) or its system matrices and onto a lower-dimensional subspace.We need a projection operator T ∈ ℝ × projecting the increment ∈ ℝ onto The state variables are reduced in dimension by this projection, hence x ∶= T for all time and iteration indices.The projected version of the prior covariance matrix is thus given as Γ Γ Γ ∶= T Γ Γ Γ pr .The projection does not effect the observations and the corresponding error covariance .The projected functional (4) reads: and is now less costly to compute because the dimension was reduced from to the much smaller .Suitable projection matrices and must be found so that the reduced system still approximates well the behaviour of the full system.The remainder of this paper discusses the choice of projection operators and demonstrates our approach in numerical examples.
Model reduction by balanced truncation
Model reduction for dynamical systems often relies on projection-based methods [8,9], and one such method that has been used in data assimilation [2,10,11] is balanced truncation (BT) [12,13].A more detailed explanation of BT can be found in the book [14] which we based this subsection on.The original BT method considers linear time invariant (LTI) systems with a control input ().For this work, we focus on the setting in discrete time ( ∈ ℤ + ): ( + 1) = () + (), () = (), (0) = 0 .
The main idea of BT is to only keep the subspace of states which are simultaneously easy to reach and easy to observe.The computations rely on the so-called reachability and observability Gramians, which are associated with the reachability and the observability energy of a state.The Gramians are functions of time, and defined as follows: T ( T ) , the time-limited reachability Gramian, and (8a) The time-limited Gramians describe the amount of energy necessary for a state to be reached or observed up to a fixed time .The system behaviour in the entire time span from zero to infinity gives the energy needed to reach or observe a state at any point of time.This justifies the definition of the limits of the finite Gramians (8a) and (8b).These limits only exist for systems (7) which are stable, meaning the eigenvalues of have to lie inside the unit circle such that is bounded for → ∞ and consequently and are bounded for → ∞.The limits ∞ = lim →∞ and ∞ = lim →∞ for a stable LTI system (7) are: T ( T ) , the infinite reachability Gramian, and (9a) ( T ) T , the infinite observability Gramian. ( Infinite Gramians have the advantage that they can be computed efficiently as the solutions of discrete-time Lyapunov (or Stein) equations.For a stable discrete-time system (7) the infinite Gramians (9a) and (9b) are unique, positive-definite solutions to the discrete-time Lyapunov equations ∞ T + T = ∞ and T ∞ + T = ∞ .The downside of using infinite Gramians is the limitation of BT to stable systems.In data assimilation applications unstable systems are very common.For this reason we propose the time-limited BT approach for unstable systems and rather use the time-limited Gramians evaluated at the last time of observation, namely ( ) and ( ).From now on, we denote the reachability Gramian by and the observability Gramian with , no matter whether they are time-limited or infinite.The Gramians are necessary to find a low-dimensional subspace that contains states with low reachability energy and high observability energy so that the states in this reduced subspace are both easily reachable and easily observable.Reachability and observability must be made comparable by choosing a basis in which both concepts are equivalent.This is done by making sure that the directions obtained by BT maximise the Rayleigh quotient (‖ ‖ ∕‖ ‖ −1 ) 2 .
This holds for the dominant eigendirections of the matrix pencil ( , −1 ) associated with the dominant eigenvalues 2 .They satisfy = 2 −1 and the eigenvalue square roots 1 > 2 > … > are called Hankel singular values.The matrix of the dominant eigendirections = [ 1 , .… , ] is used to build rank- projection operators and with T = , that project the state variables and matrices of the system (7) onto a suitable low-dimensional subspace, just as in (5).The algorithm for BT of systems of the form ( 7) is based on the singular value decomposition.Details can be found in the book [14].
Remark 2.1.BT can be applied without further assumptions on the system (7), as long as suitable Gramians can be defined.The standard application are stable systems with infinite Gramians (9) which can be obtained by solving Lyapunov equa-tions.Time-limited BT works with the harder to compute time-limited Gramians (8) but makes the approach applicable to unstable systems.For both choices of Gramians output error bounds are known [14,15].Remark 2.2.Another approach for unstable systems is -bounded balanced truncation [2] which we will use for comparison.In the discrete setting, all eigenvalues of the finite matrix in (7) lie inside a circle of radius ( ) (the maximal absolute value that an eigenvalue of has) around the origin.Once > ( ) is chosen, the system can be shifted by ∶= ∕, ∶= ∕ √ and ∶= ∕ √ .BT is now performed on the shifted system, which is stable.
BT within strong constraint incremental 4D-Var
The idea of model reduction by BT for 4D-Var first appeared in the work of Lawless et al. [4] and was subsequently applied to strong-constraint 4D-Var [2,16] and to the weak constraint setting (including model errors, which we ignore here) [10].
Within this work, we demonstrate the use of time-limited balanced truncation (TLBT) for the tangent linear model (3), as we proposed in our work [17].We assume that = and = for all to have a LTI system.Up to now we have ignored noise in our explanation of 4D-Var but want to take into account the prior and measurement uncertainty in Gramian computations.The tangent linear model including prior uncertainty is then given by: 0 = 0 ∼ (0 0 0, Γ Γ Γ pr ) and +1 = for ≥ 0 = , (10) so that the covariance of the initial increment 0 corresponds to the prior covariance.We additionally assume timeinvariance of the observation error, that is, = Γ Γ Γ for all .
We now apply BT which is defined for systems of type (7) to the inner loop LTI system (10) and thus need to define suitable Gramians.For ≥ 1 the systems are the same with the input port = 0 in (10).For = 0 we have to take into account the prior distribution in (10) and hence a different input operator at the initial time.This is resolved by using a different summand at = 0 in the computations of the Gramians (8a) and (9a), namely T ( T ) = 0 0 0( T ) = 0 0 0 for ≥ 1 and 0 Γ Γ Γ pr ( T ) 0 = Γ Γ Γ pr at time = 0 (see [10]).The Gramians for system (10) are given by: 0 0 0( T ) = Γ Γ Γ pr , the time-limited empirical reachability Gramian, and (11a) ( T ) T Γ Γ Γ −1 , the time-limited empirical observability Gramian.
( 1 1 b ) Infinite Gramians are obtained for → ∞ and it always holds 4D-Var = Γ Γ Γ pr .These empirical Gramians are found in our work [17] and similarly in a paper by Qian et al. [11] for the continuous setting.The observation error of (1b) was included in the observability Gramian as proposed by literature [11,16].This choice is reasonable because of an argument by Bernstein et al. [18]: The Γ Γ Γ −1 -weighted expected distance error between the observations of the full system (10) and the observations of its reduced projected version with operators (5), is minimised by the reduced model obtained using (11b).This is explained in more detail in our paper [17].
Time-limited balanced truncation can now be applied to system (10) using the Gramians (11).Projection matrices and T are built by TLBT as described in Section 2.1 and used to reduce (10) and obtain the corresponding reduced cost functional (6) of the inner loop.The minimisation in the inner loop and update of the outer loop can then proceed as described in Section 1. TLBT within the inner loop of incremental 4D-Var is particularly effective if the projectors and T do not change at each call of the inner loop but can be computed once for the entire incremental 4D-Var algorithm.This is the case if it can be assumed that the linearised system matrices of (3) do not change during the iterations of the outer loop.The non-linear model trajectories are not affected by the projection and need to be re-evaluated at each outer loop iteration.
2.3
Posterior precision as constant Hessian in a Quasi-Newton method The typical framework of incremental 4D-Var is the linearisation around the current iterate initial state () 0 and then finding the minimum of (4).This minimisation requires the computation of the gradient ∇ 0 J( () 0 ) and the Hessian ∇∇ 0 J( ).The authors of [5] demonstrate that for an exact tangent linear model (3) this corresponds to a Gauß-Newton step.We want to make this computationally tractable and use an approximate and reduced model.We thus propose a further simplification, namely to determine linearisations = ∈ ℝ × and = ∈ ℝ × in (3) primarily for all possible (initial) states, that is we chose one linearisation for the whole model and do not linearise at each step of the outer loop.We also assume time-invariance of the observation error, that is = Γ Γ Γ for all .Then, a system of type (1) with 0 = 0 0 0 reads: = + , ∼ (0 0 0, Γ Γ Γ ) for = 1, … , .
We apply the Bayesian statistical approach with the Gaussian prior distribution for 0 and the derived Gaussian likelihood where . It is well known that the posterior distribution of 0 is again Gaussian (see e.g., [19]): The Fisher matrix ∈ ℝ × is defined as Note that these summands also appear in the empirical observability Gramian (11b).It can be verified that Γ Γ Γ pos corresponds to the inverse Hessian for the minimisation of (4) for constant , and .Using = 0 we obtain: ), that is the gradient shifted by the weighted distance to the non-zero background error.We can use the reduced quantities (5) to compute reduced posterior quantities (12) and thus the (inverse) Hessian and gradient of the of the reduced cost functional (6) for a given initial state (and the corresponding trajectory).Qian et al. [11] have explained how BT is used for the efficient computation of a good approximation of (12) and we generalised the approach to arbitrary priors and unstable systems using TLBT [17].Standard methods [4,5] use the adjoint model to find the increment () 0 for the i-th step of the outer loop.Our approach is to use the described TLBT-based computations for building a low-resolution model in the inner loop and compute an approximation of () 0 as the minimiser of (6) effectively.
The assumption = and = for the whole model is particularly effective as we only need to build the reduced model by TLBT once and can then perform computations in the lower-dimensional setting at each iterate of the outer loop.Additionally, it can be seen that the computation of Γ Γ Γ pos in (12) does not depend on the current trajectory but only on the (reduced) system matrices.With our assumptions, they are constant all throughout the optimisation.Consequently, we only need to compute the reduced prior precision as approximate Hessian once in the beginning and can use it at each update step.Only the gradient computation needs to be performed for each iterate to determine the descent direction.Hence, this results in a Quasi-Newton method [20] and saves computational effort.
F I G U R E 1 RMS errors for 4D-Var as described above, with Γ Γ Γ pr Gaussian exponential matrix and all variables observed.RMS, root-mean-square.
In this section, we have proposed TLBT for model-reduction within the inner loop of incremental 4D-Var.With fixed linear system matrices and as approximate system and observation operator, we only need to compute an approximate Hessian once.This leads to a significant speed up in the computations for solving the initial value problem (1), at least approximately.The speed-up with our approach and the quality of the approximate solution will now be demonstrated in experiments.
NUMERICAL EXPERIMENTS WITH THE LORENZ-95 SYSTEM
The Lorenz-95 system was introduced by Edward Lorenz in 1995 [21] and is a nonlinear system of ODEs that is known for its behaviour similar to numerical weather models.It is characterised by its dimension and the forcing : d d = − −2 −1 + −1 +1 − + for = 1, … , and cyclic boundary conditions (13) For our experiments with the Lorenz-95 model, we have chosen the parameters = 500 and = 8.We want to consider an assimilation window of 50 steps with incoming measurements and then compute a forecast of 100 steps.The step size is set to ℎ = 0.01, leading to = 0.5 the end-time for the last measurement.The chaotic and unstable nature of the system only allows for short-term predictions.We provide the MATLAB code for reproducing the presented results at https://github.com/joskoUP/TLBT4DVar.For our setup, it can be chosen between = for the constant observation matrix or if only partial observations every of dimensions are made.The prior covariance can be chosen to be Γ Γ Γ pr = 0.1 or to be a Gaussian exponential, i.e. (Γ Γ Γ pr ) , = 0.1 exp( ) ∀, .The observation error covariance is Γ Γ Γ = 0.01 .
We create a true initial condition 0 = [ 1 (0), … , (0)] T by sampling from a standard normal distribution and then compute the background state 0 (which is the starting value for the outer loop in our optimisation) by disturbing 0 with (0 0 0, Γ Γ Γ pr )-distributed additive noise.All trajectories of the full model are computed from a given initial condition with the system equations ( 13) by a fourth order Runge-Kutta method.The trajectory starting from 0 is considered the truth, whereas the trajectory starting from 0 is the first guess before assimilation.We use incremental 4D-Var to infer the initial condition * 0 as best guess of 0 .For the full model, we linearise at each iteration of the outer loop to obtain the timevarying tangent linear model (3).In the reduced setups (TLBT and -BT), we assume = constant over all iterations and construct once in the beginning via linearisation in the middle of the assimilation window, which is 25 = 0.25.We compare the trajectory starting from * 0 (obtained after assimilation) to the truth in the root mean square error.This is plotted in Figure 1 with the Gaussian exponential prior and full observations and in Figure 2 for the case with Γ Γ Γ pr = 0.1 F I G U R E 2 RMS errors after assimilation for observed and unobserved variables in 4D-Var as described above, with Γ Γ Γ pr = 0.1 and observations every 9 points.RMS, root-mean-square.
TA B L E 1
Runtimes and root-mean-square error over the whole assimilation window for the full model compared to the reduced model with time-limited balanced truncation and -balanced truncation in the inner loop.and partial observations (only every 9th spatial point).It can be observed that TLBT does perform well compared to the full model and that the results for -BT depend strongly on the chosen setting.
𝚪 𝚪 𝚪
In Table 1 the runtimes and root-mean-square (RMS) errors for several settings are given.We can observe a significant speed-up even though we have only reduced the dimension from = 500 to = 375.This is mainly due to the fact that we use a constant Hessian and can precompute ( T ) T Γ Γ Γ −1 for all for the gradients.Building the projection matrices and the reduced model operators (5) requires a lot of effort.The Quasi-Newton steps are comparably cheap so that we can allow our approximate model to need more steps than the full model and still be faster.Model reduction by -BT is cheaper than by TLBT, because we can compute the Gramians as the solutions of Stein equations instead of the direct summation approach that is necessary with unstable system matrices.The big advantage of TLBT compared to -BT is that with TLBT, there is no need to solve an eigenvalue problem and determine a heuristic value for the parameter .With = ( ) + 10 −5 we have made a good choice for most of our settings.TLBT and -BT lead to assimilation errors in the same order of magnitude and -BT can outperform TLBT.In the setting with the Gaussian exponential prior and partial minimum * 0 =∶ () 0 of (4) is used in this update and the minimisation of (4) is called the inner loop.The nonlinear model is run from the current step initial condition () 0 to obtain the trajectory () = [ obtain the new increment (+1) 0 as the solution of ∇∇ 0 J( T =∶ x ∈ ℝ , with ≪ .The operator such that T = gives the reduced model and observation operators as: M ∶= T and Ĥ ∶= . | 2023-09-14T15:08:46.838Z | 2023-09-12T00:00:00.000 | {
"year": 2023,
"sha1": "0d540da328a0950f0fb1f7dd52aa8dc0191b20e2",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/pamm.202300019",
"oa_status": "HYBRID",
"pdf_src": "Wiley",
"pdf_hash": "1c98521fbce6502b3619dc380c99274b0bf09bb4",
"s2fieldsofstudy": [
"Environmental Science",
"Engineering",
"Mathematics"
],
"extfieldsofstudy": []
} |
55782735 | pes2o/s2orc | v3-fos-license | An Analysis of Time-delay for Remote Piloted Vehicle
Time-delay is one key factor affecting handling qualities of remote piloted vehicle. In this paper, the composition elements of time-delay were analyzed, in order to determine the relationship between latency and handling qualities. Then, comparing with the definition of time-delay in manned aircraft operation, the expression of time-delay was proposed for remote piloted vehicle and the measuring method of time-delay in piloted airplane was used as a reference. Moreover, considering different types of time-delay in RPV operation, typical openloop and closed-loop experiments were designed and conducted based on ground testing environment. According to these results, the relation between time-delay and handling qualities was obtained. Finally, validity of this relationship was demonstrated through simulated approach and landing during flight testing. Time-delay is a critical element of RPV handling qualities, the research of which will be helpful and meaningful in unmanned aerial vehicle design and flight testing.
Introduction
For remote piloted unmanned aerial vehicle, pilot in ground control station could not get real time feedback as their counterpart, manned aircraft pilot, such as status of aircraft, visual weather and vibration of fuselage.Consequently, vehicle could not receive real time control order via uplink from ground control station.Therefore, time-delay in unlink and downlink has been becoming one primary issue affecting operator who control UAV to complete assigned mission.Nowadays, a large amount of research are carried out to decrease influence of time-delay on handling qualities in remote piloted vehicle and many methods are applied on RPV operation to alleviate operator's workload and to improve handling qualities.However, there are still two unsolved questions: how to define time-delay in unmanned aerial vehicle clearly and how to determine relation between time-delay and handling qualities.In this paper, study both in theoretical analysis and flight testing is started to solve these two questions mentioned above.
Time-delay in manned aircraft
In the standard of manned aircraft flying qualities, there are two kinds of time-delay definition: equivalent time-delay and effective time-delay [1][2].Equivalent time-delay is mainly aiming at low level equivalent system.The effective time-delay refers to the time that equals to number of the maximum rate of slope intersecting with time axis, which input of control force/displacement order produce.
Table 1 shows requirements in handling qualities of piloted aircraft.Time-delay in handling qualities of manned aircraft is measured by interval from input of con to response of control surface in either equivalent or effective time-delay.
Composition of UAV time-delay
Time-delay of unmanned aircraft system consists of many items, not only the latency from input of manipulator system to response of control surface, but also latency of data link, digital processing in computers, and so on [3][4][5].Comparing with latency of aircraft response, magnitude of data link timedelay is greater.Figure 1 is the composition diagram of time-delay in remote piloted vehicle.The total time delay is the sum of time-delay in ground control station, time delay of data link and time delay in UAV.The time delay on ground contains pilot input, computer calculation in ground control station and interaction between ground control station and data link, while time delay in UAV covers calculation in flight control computer, response of actuator and servo, interaction between aircraft and data link.And time delay of data link includes uplink latency and down link latency.It can be seen that the compositions of UAV time delay is rather complex.There are so many factors affecting the analysis of time delay of UAV.
Definition of Unmanned Aerial Aircraft Time-delay
Like definition of manned aircraft time-delay, time-delay of RPV reflects the relation in timedifference between input of order sent by pilot from GCS and display in screen of GCS down link from aircraft [6].For the remote control, this expression reflexes pilot's feeling of latency in aircraft motion after manipulation, which has approximation for large time-delay RPV control as well as less accurate.It is difficult to measure equivalent time-delay and the main method for measurement is depends on equivalent matching algorithm to calculating latency.For the RPV with large magnitude of time-delay, adaptive condition happens easily.Hence, a pre-knowledge hypothesis is proposed for time-delay of data link.
Taking longitudinal movement of manned aircraft as an example, pilot got feedback of pitch angular acceleration firstly during manipulation, but it is difficult to measure acceleration in flight testing.However, as the integral of angular acceleration, angular velocity refers to variation of angular acceleration directly, and angular speed is easy to measure and has high accuracy.Therefore, that is why using angular speed as main parameter to evaluate time-delay.
For remote piloted vehicle, pilot in GCS could not get feedback on change of angular speed, but get information about attitude instead.Hence, it is suggested to choose attitude as main evaluation indicator for time-delay.Meanwhile, from perspective of measurement and calculation, the most part of latency is led by data link, so it is not to emphasize precision of latency in aircraft.Considering all points mentioned above, the definition of time-delay in RPV is proposed as follows.The time-delay in RPV refers to time span from sending input of manipulating order by remote pilot in GCS to display of status of system in GCS via down linking after response of aircraft to input.
Measurement of UAV time-delay
There are two kinds of measurement for time-delay in manned aircraft.Equivalent time-delay is the equivalent matching of angular speed, while effective time-delay is calculated on basis of maximum slope of angular speed and starting time of motion.The time-delay of RPV is to measure time-relation between sending input of manipulation order and receiving status of aircraft shown in screen of GCS.Thus, variation of attitude angle is becoming the main parameter in time-delay measurement.Operator in GCS could monitor changing speed of RPV attitude.The method of equivalent time-delay in manned aircraft could be used in RPV time-delay measurement to determine slope of attitude angle.As shown in figure 1, Step input starts in first second.The time-delay of manned aircraft is the difference value between starting time of order and crossover point of maximum slope and time axis, which is Δ t1.As shown in figure 1, according to measurement of manned aircraft, time-delay of RPV is the interval between start time of step input and the time, which value equals crossover point of maximum slope on attitude angle and time axis.
Testing environment for time-delay
Closed-loop ground test and flight test are conducted in small unmanned aerial aircraft in order to verify the relationship between time-delay and handling qualities.
Closed-loop ground testing introduction
Closed-loop ground testing system is built using simulator and aircraft, which is shown in figure 2. Firstly, input of manipulation input was sent to dual-redundancy flight control computer through data link.Then, Flight control computer will send orders to actuator to deflect control surface.On the other side, analog computer collects deflection of surface and calculates dynamic response of aircraft, which
Flight testing description
The experiment is executed in one remote piloted aircraft that was refitted from trainer aircraft.In order to achieve remote control, fly-by-wire system was added on the basis of mechanical manipulation system in front of cockpit, so that pilot in front could control vehicle through fly-bywire.The pilot in rear of cockpit still uses mechanical manipulation system as safe guarantee.In the ground control station, side stick was installed and layout of human-machine interface was optimized to make sure operation safety and effectiveness.On the other hand, data transmission system with high-performance was employed in order to decrease latency in data link, which means input of order and status of system could interact in short interval.
During testing, remote pilot control the RPV in GCS to measure the period from sending input of control to receiving status on screen.
Results analysis 4.1 Analysis of closed-loop test results
There are 5 sophisticated test pilots having more than 3000 flight hours in fixed wing manned-aircraft, who participate in testing.In the beginning phase of testing, pilots did not adapt to large time-delay manipulation.They can complete open-loop mission with easy input like impulse and step, while it is hard to accomplish closed-loop task using simple input.With much practice, pilots understand the characteristics of RPV time-delay and influence on aircraft motion from different magnitude of latency, and they could finish tracking task in closed-loop test.
The comments from pilots are shown in table 2. The open-loop task includes impulse and step input maneuvers, while closed-loop mission refers to approaching with inspecting disturbance and correcting.Since time-delay is measured on the basis of attitude, the magnitude of time-delay in RPV is greater than manned aircraft.From comments, it is seen that well-trained pilot could accomplish remote control in open-loop task and there is higher probability of successful landing with simulated correcting ability when magnitude of time-delay is around 300 milliseconds.When magnitude locates in 410-460 milliseconds, the operative difficulty becomes greater and coupling oscillation occurs when correcting during landing.Over 510 milliseconds, it is hard to execute normal landing, but there is less probability of occurring coupling oscillation.
With analysis of testing data, time-delay for 400-500 milliseconds is a sensitive range.The magnitude of time-delay coincides with responding time, which may produce coupling phenomenon.
Analysis of flight test results
In order to explore further relationship between time-delay and handling qualities, remote pilot control vehicle carried out simulated approach at 80m height, and then go around, shown in figure 3.During testing, remote pilot adjust heading to targeting at runway.It is found that heading is difficult for remote pilot.And it is hard to touchdown.Therefore, Go-around is the only choice for remote pilot due to safety consideration.
Comparison between tests
During approaching phase, step input of heading operation was repeated for several times to determine magnitude of time-delay.The results illustrate that time-delay in heading direction is around 400 milliseconds, as shown in figure 4.This magnitude of latency is almost the same with comments from remote pilots.Hence, the scope of remote operation time-delay locates between 400~500 milliseconds, and manipulation, especially closed-loop operation is hard to implement in this range, which would cause piloted induced oscillation .Finally, it is seen that the results between closed-loop ground test and flight test coincide badly based on analysis.From definition and results mentioned above, it is hard to control remote piloted vehicle if magnitude of time-delay is greater than 400 milliseconds.
Conclusions
As development of UAV and expanding in applications, latency in operation has been becoming one of primary issues in operation.From perspective of handling qualities, definition and composition of remote piloted vehicle are presented at first.To demonstrate relationship between time-delay and handling qualities, closed-loop ground test and flight test are executed.Results show that scope of latency is found and suggestions are proposed, which is helpful for study of remote piloted vehicle handling qualities in future.
Figure 1 .
Figure 1.The method of measuring time-delay
F1Figure 3 .
Figure 3. Simulated approach and go around in remote piloted control mode
Figure 4 .
Figure 4. Statistics of heading time-delay
Table 1 .
Time-delay requirement in handling qualities
Table 2 .
Comments about closed-loop test from pilots | 2018-12-08T19:44:39.277Z | 2017-01-01T00:00:00.000 | {
"year": 2017,
"sha1": "2eb64d84e28d1482abe0edc5c640e9ef62cf12e1",
"oa_license": "CCBY",
"oa_url": "https://www.matec-conferences.org/articles/matecconf/pdf/2017/28/matecconf_2mae2017_04012.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "2eb64d84e28d1482abe0edc5c640e9ef62cf12e1",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Engineering"
]
} |
260543554 | pes2o/s2orc | v3-fos-license | Effect of Itraconazole on the Pharmacokinetics of Diclofenac in Beagle Dogs
The objective of this study was to investigate the potential effect of itraconazole on the pharmacokinetics of diclofenac potassium in beagle dogs after oral coadministration. Five male beagle dogs received a single oral 50 mg dose of diclofenac potassium alone in phase I, and along with a single oral 100 mg dose of itraconazole in phase II. Blood samples obtained for 8.0 hours post dose were analysed for diclofenac concentration using a validated high performance liquid chromatography (HPLC) assay method. The area under plasma concentration-time curve (AUC0–∞), maximum plasma concentration (Cmax), time to reach Cmax (Tmax) and elimination half-life (t1/2), were calculated for diclofenac before and after itraconazole administration. The coadministration of itraconazole with diclofenac potassium has resulted in a significant reduction in AUC0–∞ and Cmax of diclofenac, which was about 31 and 42%; respectively. No statistically significant differences were observed for Tmax and t1/2 of diclofenac between the two phases. Therefore, it could be concluded that oral coadministration of itraconazole may have the potential to affect the absorption of diclofenac as indicated by the significant reduction in its AUC and Cmax in beagle dogs.
Introduction
Diclofenac is a nonsteroidal anti-inflammatory drug (NSAID) with analgesic and antipyretic properties. It is widely used in management of mild to moderate pain particularly when inflammation is also present as in cases of rheumatoid arthritis, osteoarthritis, musculoskeletal injuries and some postoperative conditions [1][2][3]. Its pharmacological effects are believed to be due to blocking the conversion of arachidonic acid to prostaglandins by inhibiting cyclo-oxygenase enzymes [4].
Diclofenac is almost completely absorbed after oral administration. However, due to its first-pass hepatic metabolism, only about 50% of the absorbed dose is systematically available [5][6][7][8]. The major metabolite of diclofenac in human is 4'-hydroxydiclofenac, which is mainly formed by cytochrome P4502C9 (CYP2C9) enzyme [9][10]. The minor diclofenac metabolites are formed by several enzymes including CYP3A4 [11]. About 99% of the drug is bound to human plasma proteins, mainly albumin [12,13]. The potassium salt of diclofenac was found to be particularly useful for quick pain relief compared to the sodium salt because of its higher solubility in the stomach acidic medium [14].
Itraconazole is a triazole antifungal agent that is used for a number of indications including systemic and superficial mycosis [15]. It has been shown to interact with many drugs mainly by inhibiting their CYP3A4-mediated metabolism and/or multidrug resistance protein 1 (MRP1)-mediated transport [16][17][18][19]. In addition, itraconazole is highly bound to plasma proteins, primarily albumin and therefore, may interact with other albumin highly bound drugs [15,20].
A literature search revealed that pharmacokinetic interactions between itraconazole and diclofenac have never been investigated. In fact, there is a very limited published information about the pharmacokinetic interactions between triazole antifungals and NSAIDs in general [21]. Coadministration of itraconazole with diclofenac may be considered for the treatment of certain fungal infections in cases of rheumatoid arthritis, musculoskeletal injuries and some postoperative conditions. Therefore, it is important to investigate the potential interaction between the two drugs. The aim of this study was to investigate the effect of itraconazole on the pharmacokinetics of diclofenac in vivo using beagle dogs as an animal model.
Materials
Diclofenac potassium 50 mg tablets (Cataflam ® , Novartis Pharma, Egypt) were purchased from the market while itraconazole 100 mg capsules (Sporanox ® , Janssen-Cilag, Beerse, Belgium) were obtained from the pharmacy of King Khalid University Hospital. Diclofenac and flufenamic acid (internal standard) analytical powders were purchased from Sigma (St. Louis, MO, USA). Acetonitrile was obtained from BDH, England, UK and glacial acetic acid from Polysciences Inc. Warrington, PA, USA. All other reagents were of analytical grade.
Animal study
The protocol of the in vivo study in beagle dogs was approved by the Experimental Animals Care Centre of College of Pharmacy, King Saud University, Riyadh, Saudi Arabia.
Five healthy male beagle dogs weighing 10.2-13.6 kg were used. The study was conducted in two phases with one week wash-out period. The dogs were fasted for 12 hrs before drug administration and continued fasting for 2 hrs post dose but allowed free access to water. Each dog was administered one tablet of 50 mg diclofenac potassium alone (phase I) or with a 100 mg itraconazole capsule (phase II). No other medications were taken during the study period. Venous blood samples (3.0 ml) were taken from the femoral vein into heparinized tubes before drug administration (to serve as a blank) and at 0.25, 0.50, 0.75, 1.00, 1.50, 2.00, 3.00, 4.00, 6.00 and 8.00 hr after drug administration. Samples were centrifuged immediately at 5000 rpm for 10 min. and the separated plasma samples were kept at −20 °C for analysis.
Assay of diclofenac in plasma
The plasma concentration of diclofenac was determined by a modified high-performance liquid chromatography (HPLC) assay method [22] The mixture was shaken again on a vortex mixer for 1 min., and centrifuged for 5 min. at 10000 rpm. The supernatant was transferred to an autosampler vial for injection in HPLC.
Calculation of pharmacokinetic parameters
Maximum plasma concentration (C max ) and the time to reach it (t max ) were obtained directly from plasma data. Elimination half-life (t 1/2 ) was calculated as 0.693/K el , where K el is the elimination rate constant obtained from the slope of the terminal exponential phase. The total area under plasma concentration time curve (AUC 0-∞ ) was calculated as the sum of AUC 0-8hr and AUC 8hr-∞ , where AUC 0-8hr was determined by the trapezoidal rule method and AUC 8hr-∞ as the last plasma concentration divided by K el .
Statistical analysis
The significance of the differences between plasma concentrations of diclofenac at each sampling time and the pharmacokinetic parameters of treatment group versus control were evaluated using Student's paired t-test. P value ≤ 0.05 was taken as the criterion for statistically significant difference. Figure 1 shows the plasma concentration of diclofenac in beagle dogs following oral administration of 50 mg diclofenac potassium tablet alone (control) or with 100 mg itraconazole capsule. It was noticed that the plasma concentration of diclofenac was significantly affected by the presence of itraconazole upon oral coadministration to beagle dogs. This finding was supported by the significant reduction in the C max and AUC 0-∞ after itraconazole administration (about 42% and 31%; respectively, Table 1). No statistically significant differences (P > 0.05) were observed for the values of t max and t 1/2 after itraconazole treatment compared to control, which indicated similar times to reach maximum concentration and similar elimination rate constants.
Results and Discussion
Itraconazole is a known inhibitor of CYP3A subfamily of enzymes and most of the reported pharmacokinetic interactions that are caused by this drug are believed to be mainly due to inhibition of CYP3A. However, the observed interaction in this study can not be explained by inhibition of drug metabolism simply because such inhibition would result in an increase in the AUC and plasma concentration of the affected drug rather than a decrease as observed with diclofenac in this study. In fact, CYP3A subfamily is not a major contributor to diclofenac elimination both in human and in beagle dog [5,6,9,10]. In addition, alteration of diclofenac elimination, in general, does not seem to play a major role in the observed interaction as strongly indicated by the lack of significant itraconazole effect on the t 1/2 of diclofenac [ Table 1].
Displacement of diclofenac from its plasma binding sites by itraconazole is unlikely to be a major cause for this observed interaction. This is supported by the finding of Hynninen and colleagues who have shown that itraconazole could significantly reduce the C max and AUC 0-∞ of meloxicam in human subjects without affecting the unbound plasma meloxicam concentration [21]. Mean plasma concentration (± SE) of diclofenac in beagle dogs following oral administration of 50 mg diclofenac potassium tablet alone (control) or with 100 mg itraconazole capsule (n=5). *P < 0.05.
The results obtained from this study suggest that itraconazole reduces the plasma concentration of diclofenac by interfering with its gastrointestinal absorption. Impairment of absorption by itraconazole has been suggested for the first time by Hynninen and coauthors to explain itraconazole effect on the plasma concentration of another NSAID; meloxicam [21]. However, the exact mechanism of such interaction is not clear. Appropriate mechanistic studies that involve suitable intestinal absorption models may help to clarify the mechanism(s) behind this interaction.
In conclusion, itraconazole was shown in this study to significantly reduce the C max and AUC 0-∞ of diclofenac in beagle dogs suggesting that it has the potential to decrease the intensity of diclofenac pharmacological effect. The exact mechanism of this interaction is not clear but the results suggest an alteration in diclofenac absorption by itraconazole. In addition, results obtained from this study warrant further investigation in human subjects to evaluate the clinical relevance of this interaction. Table 1. Pharmacokinetic parameters of diclofenac in beagle dogs (mean ± SD) following oral administration of 50 mg diclofenac potassium tablet alone (control) or with 100 mg of itraconazole capsule (n=5). *P < 0.05. | 2019-04-06T13:04:18.948Z | 0001-01-01T00:00:00.000 | {
"year": 2010,
"sha1": "2c4984981a81f95c979636ca47e936fb5fd2f60f",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2218-0532/78/3/465/pdf?version=1475149726",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "93f6e9ef3a42227d64804eb4dbdd0125ae8a472f",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
6197766 | pes2o/s2orc | v3-fos-license | Bilateral Field Advantage in Visual Enumeration
A number of recent studies have demonstrated superior visual processing when the information is distributed across the left and right visual fields than if the information is presented in a single hemifield (the bilateral field advantage). This effect is thought to reflect independent attentional resources in the two hemifields and the capacity of the neural responses to the left and right hemifields to process visual information in parallel. Here, we examined whether a bilateral field advantage can also be observed in a high-level visual task that requires the information from both hemifields to be combined. To this end, we used a visual enumeration task—a task that requires the assimilation of separate visual items into a single quantity—where the to-be-enumerated items were either presented in one hemifield or distributed between the two visual fields. We found that enumerating large number (>4 items), but not small number (<4 items), exhibited the bilateral field advantage: enumeration was more accurate when the visual items were split between the left and right hemifields than when they were all presented within the same hemifield. Control experiments further showed that this effect could not be attributed to a horizontal alignment advantage of the items in the visual field, or to a retinal stimulation difference between the unilateral and bilateral displays. These results suggest that a bilateral field advantage can arise when the visual task involves inter-hemispheric integration. This is in line with previous research and theory indicating that, when the visual task is attentionally demanding, parallel processing by the neural responses to the left and right hemifields can expand the capacity of visual information processing.
Introduction
Enumerating visual objects from a visual scene is a task the human brain must continuously perform. When individuals are asked to determine the numbers of briefly presented visual items, differences in the slope accuracy are typically found between small versus large numbers [1,2]. Enumeration remains fairly accurate up to four items but deteriorates rather dramatically with larger numbers. The process of enumerating small numbers is known as subitizing [3], while enumerating large numbers is thought to either reflect counting (when sufficient presentation time is given) or to engage the approximate number system (when the items are briefly presented) [4].
In a typical visual enumeration task, the items are randomly presented on a computer screen, with some items inevitably falling in the left visual field and others in the right visual field. Given that the information from the left visual field generates initially a greater neural response in the right visual cortex and the information from the right visual field in the left visual cortex, the brain needs to integrate all that information across the hemispheres to represent a single quantity. The corpus callosum allows processing occurring in one hemisphere to be transmitted to and integrated with processing occurring in the other hemisphere [5]. The effect of inter-hemispheric integration on visual enumeration, however, is currently unknown and two predictions can be made.
On the one hand, recent findings have suggested temporal and qualitative differences between the integration of visual information within and across the hemifields, with within-hemifield integration preceding [6,7] and being more efficient [8,9] than across-hemifield integration. For instance, Large and colleagues [6] found that the same regions in the lateral occipital cortex (LO) respond both to the upper and lower visual fields, but with a clear contralateral preference. Using the technique of fMRI adaptation, the authors found a greater adaptation to vertical translations of faces within the same hemifield than across-hemifield translations, suggesting that the upper and lower visual representations are combined in the contralateral LO prior to the integration of the left and right representations. Consistent with this, the completion of illusory contours [9] and processes of perceptual grouping [8] are stronger when the stimuli appear within the same hemifield than when they cross hemifields. Accordingly, as across-hemifield integration is required in an enumeration task when the items are split between the two hemifields, we may predict a unilateral field advantage in visual enumeration, namely better performance when the items are unilaterally presented as when they are bilaterally displayed.
On the other hand, another line of research has suggested that there exist independent attentional resources for the left and right hemifields [10,11] and that parallel processing by the neural responses to the left and right hemifields can expand the capacity of visual information processing. This has been reported in a number of attentional demanding visual tasks, such as object tacking [12], short-term memory for spatial locations [13], item identification [14] and orientation discrimination and detection [15] among other visual tasks. These tasks are better performed when the items are distributed across the left and right visual fields as when they were all displayed within a single hemifield. Contrary to visual enumeration, the tasks employed in those studies do not require the information from the left and right visual fields to be combined. Rather, visual representations from both fields could be processed independently by each of the hemispheres and still support task performance. However, if such a bilateral advantage is a general feature of selective attention, as previous findings seem to suggest [12][13][14][15], we may predict a bilateral field advantage in visual enumeration only when the task requires a certain amount of attentional resources. More precisely, a bilateral field advantage may be observed beyond the subitizing range, that is to say when at least four items have to be enumerated. In the present study, the effect of distributing visual items across the two hemifields on visual enumeration was directly tested by pitting the unilateral and bilateral field advantage hypotheses against each other.
Experiment 1
In this experiment two to eight dots were quickly presented on a computer screen and fell either all in the same visual field (unilateral condition) or in the two visual fields (bilateral condition). Participants were asked to keep their eyes on the centre of the screen and to enumerate the dots as accurately as possible. The crucial question was whether visual enumeration would benefit, or alternatively suffer from the bilateral presentation.
Method
Participants. A total of 20 volunteers (13 women), aged between 19 and 37 years (mean = 24) took part in the experiment. In all experiments, the participants had self-reported normal or corrected-to-normal vision, and they were naïve to the experimental aims. They provided written and informed consent before experiments, and all procedures were approved by the ethic committee of the University of Leeds. They were offered £6 in exchange for their time.
Stimuli and procedure. The stimuli were presented on the 17-in. monitor of a Pentium-based computer running E-Prime 1.1 software (Psychology Software Tools, Inc. www.pstnet.com/ eprime). Participants viewed the computer screen at eye level from a distance of approximately 60 cm. Eye movements were not monitored, but participants were encouraged to keep their eyes focused on the centre of the screen throughout the experiment.
The resolution of the screen used was 10246768 pixels and the screen background was black. The displays consisted of four invisible quadrants (subtending 4.43u64.43u each) placed around a central fixation point and separated vertically and horizontally by 2.32u. Green dots (0, 130, and 0 on red, green, and blue phosphors respectively) with a diameter of 10 pixels (0.34u) were used as stimuli. The dots were placed randomly within two of the four quadrants with a minimum centre-to-centre spacing between dots of 38 pixels (1.3u). The number of dots within one quadrant ranged from 1 to 4. The total number of dots on the screen varied therefore between 2 and 8, with all possible combinations evenly used. There were two conditions: unilateral and bilateral (see Figure 1). In the unilateral condition, the dots appeared in two quadrants from the same visual field (upper-left/lower-left or upper-right/lower-right). In the bilateral condition, they appeared in two horizontally symmetrical quadrants from different hemifields (upper-left/upper-right or lower-left/lower-right).
The experiment was conducted in a quiet and dimly illuminated room. A single trial started with a blank screen for 1000 ms, followed by a central fixation point (a small white cross subtending 0.61u60.61u) for 500 ms. The stimulus display was then presented for 150 ms, followed by a blank screen that endured until a response was made. Participants responded by pressing the space bar and simultaneously speaking their response (for a similar procedure, see [16][17][18]). Participants were then prompted to encode their response by pressing a number key on the computer key pad. Twenty practice trials were completed followed by 8 blocks of 56 test trials (2 conditions67 numbers of dots632 trials). All conditions were randomized within blocks. After each block, participants were given the opportunity to take a break during which they were shown their correct response rate and mean response latency. They were politely warned if their accuracy was lower than 60%. Participants then pressed the space key to continue.
Results and Discussion
To avoid eye movements, the stimulus displays were presented only for a very short time (150 ms). Therefore, we used accuracy (percentage of correct responses) as a measure of performance as well as the coefficient of variation (CV) (the ratio of the standard Figure 1. Samples of displays used in Experiment 1. In the unilateral condition, the items appeared in two quadrants from the same hemifield (either the left or right hemifield). In the bilateral condition, the items appeared in two horizontally symmetrical quadrants from different hemifields. Note that dotted lines that delimitate the four quadrants are shown for illustration purpose only. The quadrants were invisible in the experiment. doi:10.1371/journal.pone.0017743.g001 deviation and the mean response) as a measure of response precision. In all analyses, Greenhouse-Geisser corrections for nonsphericity were applied where appropriate. The results are plotted in Figure 2.
Possible left-right visual field asymmetries were also examined. Accuracy in the left visual field was significantly higher than in the right visual field, t(19) = 2.36, MSE = 0.013, p,.05. Previous investigations on visual field asymmetries for enumeration processes have provided controversial results. Some studies have revealed a left visual field advantage for enumeration [19][20][21], whereas other neuroimaging [22,23] and neuropsychological [24] studies have found equivalent enumeration performance in both hemifields. More research is needed to clarify this issue and this will not be further discussed.
Finally, we asked whether the variability in responses also varied between the unilateral and bilateral conditions. Past research has shown that when more than 3 or 4 items are briefly presented, the approximate number system is engaged and both the mean and the standard deviation of responses increase linearly and in direct proportion as a function of numerosity, resulting in constant coefficients of variation (CVs) [25][26][27]. A 2 (condition) 67 (numerosity) ANOVA (repeated measures) on CVs revealed no effect of condition (p..24), a main effect of numerosity, F(3.17, 60.24) = 46.25, MSE = 0.002, p,.001, and a significant condition 6 numerosity interaction, F(6, 114) = 2.20, MSE = 0.000, p,.05. Figure 2b shows constant CVs of about 0.08 for numerosities greater than four in both conditions (p..8), yielding evidence that the approximate number system was engaged in the task [25][26][27]. Furthermore, whereas the mean CVs for small numerosities (2)(3)(4) did not differ between the bilateral and unilateral conditions (p..15), the mean CVs for larger numerosities (5)(6)(7)(8) were significantly smaller in the bilateral condition (CV = 0.083) than in the unilateral condition (CV = 0.094), F(1, 19) = 11.92, MSE = 0.000, p,.003. This indicates less variability, thus higher precision, in responses when the to-be-enumerated items were split across the left and right visual fields.
Those results show that visual enumeration is more accurate and more precise when the items are displayed in the two visual fields relative to when they appear within the same hemifield. This bilateral field advantage was observed when more than four objects had to be enumerated, suggesting that this bilateral effect occurs when sufficient attentional resources are requisite. This is consistent with the notion of independent resources in the left and right hemifields [10,11] and with the findings that parallel processing by the neural responses to the left and right hemifields can expand the capacity of visual information processing [12][13][14][15]. Furthermore, the present work extends those findings by showing that when the information needs to be integrated across the two hemifields, the initial parallel processing still benefits the task. In Experiment 1, the dots were vertically aligned in the unilateral condition, whereas they were horizontally aligned in the bilateral condition. Therefore, the possibility that the bilateral field advantage in visual enumeration actually reflects a horizontal advantage in number processing remains. To control for this, the same two spatial arrangements of dots (i.e., vertical versus horizontal) were used in Experiment 2, but were always presented within a single hemifield (see Figure 3, and [12] for a similar procedure). If the bilateral field advantage can be explained by the horizontal alignment of the dots, then we would expect better performance when the dots are horizontally aligned than when they are vertically aligned.
Method
Participants. Seventeen new volunteers (11 women), aged between 21 and 39 years (mean = 26) took part in the experiment.
Stimuli and procedure. This experiment replicated Experiment 1 except that the four quadrants shifted 6.58u to the left or right of fixation so that all four quadrants fell within a single hemifield ( Figure 3). In the vertical condition, the dots appeared in two vertically aligned quadrants either on the far left, near left, near right, or far right. In the horizontal condition, the dots appeared in two horizontally aligned quadrants either on the top left, bottom left, top right, or bottom right. Twenty practice trials were followed by 16 blocks of 56 trials (2 conditions 62 hemifields 67 numbers of dots 632 trials). Figure 4a). Surprisingly, it was the vertical condition that yielded better performance (72% and 69% in the vertical and horizontal conditions, respectively). We conducted a 2 (condition) 62 (hemifield) ANOVA (repeated measures) to see whether the vertical advantage was present in both hemifields. The results revealed a main effect of condition, F(1, 16) = 10.34, MSE = 0.001, p,.005, no effect of hemifield (p..3), and a significant condition 6 hemifield interaction, F(1, 16) = 14.75, MSE = 0.000, p,.001, indicating a vertical advantage in the left hemifield only (see Figure 4b). To further explore this effect, we carried out separate analyses for each of the four vertical positions (i.e., far left, near left, near right, and far right) and the results showed a vertical advantage only when the dots were displayed in the far left quadrants, t(16) = 28.66, p,.001 (see Figure 4c). This finding suggests that the vertical advantage observed in this experiment may simply reflect a left hemifield bias in enumeration [19][20][21]. Finally, the 2 (condition) 67 (numerosity) ANOVA (repeated measures) on CVs revealed no effect of condition (p..05), a main effect of numerosity, F(3.20, 51.27) = 51.72, MSE = 0.001, p,.001, and no significant condition 6 numerosity interaction (p..54). Figure 4d shows constant CVs of about 0.10 for numerosities greater than four in both conditions (p..15). The mean CVs did not differ between the horizontal and vertical conditions for both small numerosities (p..08) and large numerosities (p..28). The critical finding in Experiment 2 was the absence of a horizontal advantage in visual enumeration. Thus, the bilateral field advantage observed in Experiment 1 cannot be explained by the horizontal alignment of the dots. Rather, the effect must have been caused by the separate placement of the dots in the left and right visual fields.
Experiment 3
The bilateral advantage observed in Experiment 1 for numerosities greater than four seems to reflect an advantage of dividing attention between the left and right hemifields as compared to within the same hemifield, as greater numerosities may require greater attention. However, a stimulus-based (''bottom-up'') explanation remains plausible. In particular, retinal stimulation in the bilateral condition may differ from that in the unilateral condition and that could potentially account for the bilateral advantage observed in Experiment 1. For example, previous research has shown that the classical receptive field (CRF) of a visual neuron is surrounded by the non-classical receptive field (nCRF), where stimuli can modulate the responses to CRF [28]. Even if the bilateral and unilateral displays shared identical stimulation in one quadrant, the unique stimulation from the nonshared quadrant could differentially drive the nCRFs of neurons responding to the shared quadrant. Such a stimulus-driven explanation would indeed be more parsimonious than an attention-based (''top-down'') explanation. Experiment 3 was designed to rule out stimulus-driven explanations such as the one above. In order to match the retinal stimulation between the bilateral and unilateral conditions, our approach was to present dots in all four quadrants on all trials. Prior to the dots, a spatial cue indicated which two quadrants to select for dots enumeration (and which two quadrants to ignore). This experimental design eliminated stimulus-driven differences between the bilateral and unilateral conditions and tested more directly genuine attentional ability.
Method
Participants. Nineteen new volunteers (16 women), aged between 18 and 38 years (mean = 23.6) took part in the experiment.
Stimuli and procedure. This experiment replicated Experiment 1 with the following changes: (i) prior to the presentation of the dots, a spatial cue (i.e., a small white arrow of 1.3u61.3u of visual angle) was centrally presented (50 ms) and indicated which two quadrants to select for dots enumeration; (ii) the cue was followed by a 50 ms blank interval before the presentation of the dots; (iii) dots were then presented in the four quadrants on all trials (see Figure 5). Participants were instructed to enumerate the dots from the two cued quadrants and to ignore those from the two other quadrants. In the unilateral condition, the cue pointed to left or right, whereas in the bilateral condition, the cue pointed to the upper or lower visual field. To equalize retinal stimulations between the two conditions, the number of dots presented in the uncued quadrants always matched the number of dots in the cued quadrants. For example, if three dots were presented in the upperleft quadrant and two dots in the lower-left quadrant, then two dots were presented in the upper-right quadrant and three dots in the lower-right quadrant. Twenty practice trials were followed by 8 blocks of 56 trials (2 conditions 67 numbers of dots 632 trials). Figure 6b shows constant CVs of about 0.12 for numerosities greater than three (with the exception of eight) in both conditions (p..20). Importantly, although the mean CVs for small numerosities (2)(3)(4) did not differ between the bilateral and unilateral conditions (p..69), the mean CVs for larger numerosities (5)(6)(7)(8) were significantly smaller in the bilateral condition (CV = 0.11) than in the unilateral condition (CV = 0.12), F(1, 18) = 6.80, MSE = 0.001, p,.018. Similarly to Experiment 1, this indicates less variability, thus higher precision, in responses when the to-beenumerated items were split across the two hemifields.
General Discussion
The primary aim of the present study was to examine the effect of dividing visual stimuli across the two hemifields on visual enumeration. Enumeration requires the integration of information into a single quantity and since across-hemifield integration has been found to be more efficient [8,9] and to occur temporally after the completion of within-hemifield integration [6,7] one might expect enumeration to be more efficient when the items are presented in one hemifield only. Against this, the present study reveals that visual enumeration is actually more accurate and more precise when the visual items are distributed between the left and right visual fields as when they are all presented within a single hemifield. This bilateral field advantage, however, was only observed when four or more items have to be enumerated, thus when the task was sufficiently attentionally demanding. The study also shows that neither the horizontal alignment of the dots per se (Experiment 2) nor the retinal stimulation differences between the unilateral and bilateral displays (Experiment 3) can account for the observed bilateral field advantage. Rather, this finding is consistent with the notion of independent attentional resources in the left and right hemifields [10,11] and with recent data that have shown that parallel processing by the neural responses to the left and right hemifields allows more information to be processed [12][13][14][15].
There are several possibilities how the existence of independent resources in the two hemifields can facilitate the enumeration process. One possibility is that when the number of to-beenumerated items exceeds the subitizing range, one lateralized quantity is enumerated first while the other is held in short-term memory before being processed for subsequent enumeration and integration with the first value. Recent findings have provided evidence for independent short-term memory representations in the two hemifields [13], which would be necessary for this account.
Another possibility is the existence of two independent enumeration processes, or two pools of attentional resources, one in each hemifield and working in parallel. According to this proposal, the quantities from each visual field are processed independently and simultaneously, predominantly in the contralateral hemisphere. The two resulting quantities are then integrated together via the corpus callosum to provide the final response. Although such a process ultimately requires two quantities to be combined and added together, these costs could be minimal compared to the gain of splitting a large number into two smaller ones in each hemifield. Previous data have indeed suggested that magnitude information is represented in both hemispheres [29][30][31][32] and that numerical information can be rapidly transferred from one hemisphere to the other during a number comparison task [33,34]. If the transfer of numerical information across the corpus callosum is fast, the integration of two quantities from different hemifields may also be rapid and efficient. Furthermore, this account would be also consistent with the extensive research by Banich and colleagues [35][36][37][38] that shows that dividing processing across the hemifields is beneficial when the task is demanding because the subcomponents of the task can be divided between the hemifields and processed in parallel. This account fits well with the present findings, where the bilateral field advantage was observed only when more than four items had to be enumerated, thus when the task was rather demanding in terms of attentional resources. With numerosities less than four, however, there might be sufficient resources to subitize efficiently within a single hemifield removing any splithemifield differences. Further research is needed to test those non-exhaustive possibilities. However, whichever account is proposed, the present data strongly demonstrate that the bilateral presentation of information in the visual field can benefit a high-level task that requires inter-hemispheric integration such as visual enumeration. | 2014-10-01T00:00:00.000Z | 2011-03-23T00:00:00.000 | {
"year": 2011,
"sha1": "a887e9da509a0d89de355c58ee2a7a6a425042e7",
"oa_license": "CCBY",
"oa_url": "https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0017743&type=printable",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "c8ab6eb31e9175e2dcff3a9c7821240d3bb2f84a",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
17985667 | pes2o/s2orc | v3-fos-license | The Spontaneous Breakdown of the Vacuum
We discuss the spontaneous breakdown of the vacuum by a strong electromagnetic field as observed in SLAC experiment E-144. We show that the data follow the Schwinger non-perturbative result obtained for a static field.
Pair Production in a Strong em Field
In his 1951 paper, J. Schwinger 2 predicted that an intense static electric field will break down the vacuum to produce e + e − pairs. This occurs when the field-strength approaches the critical value E c = m 2 c 3 /eh = 1.3 × 10 16 V/cm. We argue that this effect has been observed in the scattering of high energy electrons from the focus of an intense laser pulse 1 .
The experimental conditions 1 differ from the original Schwinger premise in two respects: (a) The field strength in the laboratory is only E ∼ 3 × 10 10 V/cm but reaches near-critical value in the rest frame of the 46.6 GeV incident electrons (E * = 2γE ≃ 1.8 × 10 5 E).
(b) The field in the laboratory is not static but a well-defined coherent wave field. However Brezin and Itzykson 3 have shown that spontaneous e + e − pair production will also occur in a time-dependent field. They derive the probability for pair production in both the perturbative (low field, E * < E c ) and non-perturbative (E * > ∼ E c ) regimes. It is convenient to introduce the normalized vector potential of the field where ω 0 is the angular frequency of the field which is assumed to be monochromatic and sinusoidal. The probabilities per unit time -unit volume derived 3 are Eqs. (2,3) have an immediate interpretation in physical terms. When η ≪ 1 we are in the perturbative regime and n = 2mc 2 /hω 0 is the number of photons required to produce the pair. Thus the probability is proportional to the n th power of the square of the normalized vector potential: (η 2n ). When η ≫ 1 the probability depends on the electric field strength through the singular expression exp (−πE c /E). In the static case this behavior can be interpreted 4,5 as quantum-mechanical tunneling through a potential V 0 ∼ 2mc 2 .
In the intermediate case, η ∼ 1, where the function smoothly interpolates between the two regimes. In the non-perturbative regime it is customary to introduce the dimensionless parameter If an electron moves through the electric field with 4-momentum p (γ = p 0 /mc) then in the electron's rest frame the parameter Υ takes the value where F µν is the field tensor. Eq. (6 ′ ) can also be used when a high energy photon of 4-momentum k ν traverses the field. For head-on collisions we can write where γ was defined previously and ω 0 is the frequency of the em field. A few comments are in order: (1) Υ is a relativistic invariant describing the interaction of a particle with the electric field (2) Υ is well defined for a static field (3) It is a measure of the cm energy in the collision of the incoming particle with one laser photon (in units of the electron mass) multiplied by the normalized potential η.
In the experiment reported in ref. [1] electron-positron pairs were produced when 46.6 GeV/c electrons crossed the focus of a laser pulse of wavelength λ = 527 nm. This observation can be interpreted as a two-step process in which first a photon backscatters off an electron to become a high energy γray (ω γ ∼ 29 GeV) and subsequently the γ-ray scatters from at least four laser photons to produce the pair. The photon density in the focus is adequately high (n ω ∼ 2.5 × 10 26 cm −3 ) so that multiphoton processes up to n = 5 could be observed over the course of the experiment. Support for this interpretation comes from plotting the positron yield as a function of η and observing that it varies as η 2n with n = 5.1 ± 0.2 as expected from Eq. (2) if one replaces ω 0 by 2γω 0 . In fact the data agree with an exact calculation of the multiphoton Breit-Wheeler equation 1 as shown in Fig. 1.
To examine the alternative interpretation in terms of the spontaneous breakdown of the vacuum we wish to test Eq. (3). Note that in the experiment, η ∼ 0.3 namely one is between the two regimes. We plot the data as a function of 1/Υ as shown in Fig. 2 where we use the form of Eq. (6 ′ ) for Υ; here we have also included data obtained at 49.1 GeV. A fit to the Υ dependence of Eq. (4) yields for the factor in the exponent πg(η) = 2.01 ± 0.12 ± 0.4 the first error being statistical and the second systematic. The prediction of Eqs. (4,5) for η = 0.25 is g(η) = 0.58 and thus πg(η) = 1.82. However, the result of Eq. (7) must be corrected for two factors. In Fig. 2 we used the rms value of the electric field to define Υ and η, whereas in ref. [3] the peak values are used. Secondly the frame of reference in which the high energy gamma and one photon collide should be used in the definition of γ entering Eq. (6 ′ ); namely Υ = Υ rms √ 2(29.2/46.6). We then find that Υ has to be reduced by a factor of 1.14, while η must be increased by √ 2. This leads to πg(η) = 1.77 ± 0.35 observed πg(η) = 2.12 predicted (7 ′ ) We see that the dependence of the pair production rate on the field strength agrees with the predictions of refs. [1,3].
To estimate the positron yield predicted by Eq. (3) we must integrate the probability over volume and time. We associate a volume equal to λ -3 c (λc is the Compton wavelength of the electron) for each electron crossing the focus and use ∆t = (ℓ/c)(1/γ) for the time of interaction in the electron rest-frame; here ℓ is the length of the focus, ℓ = 2d/ sin θ ∼ 20 µm. We then obtain For Υ = 0.24 and η = 0.35 we find for the probability per incident high energy γ-ray However per laser shot only ∼ 10 6 γ-rays cross the laser focus and we must account for the fraction of γ-rays of sufficient energy to produce a pair (∼ 10 −2 of the total spectrum). These qualitative arguments predict a pair production rate of ∼ 4/laser shot as compared to the observed rate of 0.1/laser shot. We make two additional remarks. First, that eventhough the electric field seen in the electron rest-frame is time-dependent, the period is longer than the formation time of the pair by a factor of fifteen. In the rest-frame whereas the quantum-mechanical uncertainty time associated with an energy fluctuation of the order of an electron mass is Thus one can treat the fields seen in the electron rest-frame as static as considered in ref. [2]. Note however that this assumption is not needed in deriving Eqs. (2,3). In the experiment of ref. [1] the energy of the electron and positron in the pair is provided by the incident high energy γ-ray. The presence of an incident particle resolves the issue of energy-momentum balance since it is known that a plane wave (for which E 2 −B 2 = 0) cannot produce pairs. On the other hand in a focused wave, there are regions near the focus 6 where E 2 − B 2 > 0. The value of the invariant is approximately 1 2 It would be of considerable interest to observe the breakdown of the vacuum without the participation of an incident particle. This would require an intensity of the em flux I = 5 × 10 29 W/cm 2 to reach Υ = 1. Some future FEL's are planned to operate in the 1Å region and one could consider extremely tight focussing (say to 10Å 2 ) to reach high intensity. The peak power will be at best P = 10 11 W so that I = 10 26 W/cm 2 which is still short of producing critical field in the laboratory frame.
The condition for the breakdown of the vacuum is that the electric field E c be such that an electron gains energy equal to its rest mass in one Compton wavelength eE c λc = mc 2 (10) Eq. (10) leads to the definition of the critical field E c introduced previously The above is a statement on the interaction of the electron with the field. However for the pair to be produced there must also be sufficient energy in the field in a volume of order V = λ -3 c . We therefore obtain a necessary condition on the field energy Combining (9) and (10) leads to an upper limit on the value of the fine structure constant This inequality is satisfied in nature and appears to be a necessary condition for the spontaneous breakdown of the vacuum. Fig. 1 Dependence of the positron rate on the laser field-strength parameter η. The rate is normalized to the number of Compton scatters inferred from the EC37 monitor. The solid line is the prediction based on the numerical integration of the two-step process of laser backscattering followed by multiphoton Breit-Wheeler pair production. From ref. [1], 46.6 GeV data. | 2014-10-01T00:00:00.000Z | 1998-05-01T00:00:00.000 | {
"year": 1998,
"sha1": "dd6d588a31a14639591adc57e28f37f8222d6055",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "169a6b1053342bb637c628db59a30c1420782990",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
237289710 | pes2o/s2orc | v3-fos-license | Natural Language Processing Accurately Categorizes Indications, Findings and Pathology Reports from Multicenter Colonoscopy
Colonoscopy is used for colorectal cancer (CRC) screening. Extracting details of the colonoscopy findings from free text in electronic health records (EHRs) can be used to determine patient risk for CRC and colorectal screening strategies. We developed and evaluated the accuracy of a deep learning model framework to extract information for the clinical decision support system to interpret relevant free-text reports, including indications, pathology, and findings notes. The Bio-Bi-LSTM-CRF framework was developed using Bidirectional Long Short-term Memory (Bi-LSTM) and Conditional Random Fields (CRF) to extract several clinical features from these free-text reports including indications for the colonoscopy, findings during the colonoscopy, and pathology of resected material. We trained the Bio-Bi-LSTM-CRF and existing Bi-LSTM-CRF models on 80% of 4,000 manually annotated notes from 3,867 patients. These clinical notes were from a group of patients over 40 years of age enrolled in four Veterans Affairs Medical Centers. A total of 10% of the remaining annotated notes were used to train hyperparameter and the remaining 10% were used to evaluate the accuracy of our model Bio-Bi-LSTM-CRF and compare to Bi-LSTM-CRF.
Introduction
The amount of complex and varied information (e.g., structured, semi-structured, and unstructured) has grown dramatically during the past century in numerous sectors, including education [1], media [2], power [3], and healthcare [4]. In healthcare, electronic health records (EHRs) contain information in both unstructured and structured formats on patient health, disease status, and care received, which can be useful for epidemiologic and clinical research [5]. In EHRs, unstructured data is documented without standard content specifications, often recorded as free text, and structured data is generally entered into discrete data fields with standardized responses or parameters (e.g., age, weight) [6]. Compared to the structured data within the EHR, free-text (unstructured data) data often conveys more granular and contextual information of clinical events and enhances communication between clinical teams [7]. As such, extracting information from these free-text document resources has considerable potential value in supporting a multitude of aims, from clinical decision support and quality care improvement to the secondary use of clinical data for research, public health, and pharmacovigilance activities [8]. Natural language processing (NLP) can be an efficient way of automatically extracting and arranging this information, making NLP a potentially pivotal technology for enabling quality measurement from EHR data.
Background
Colonoscopy is a well-established procedure used to detect and prevent colorectal cancer (CRC), the third most common cancer in men and women, which accounts for almost 10% of all cancerrelated deaths in the United States [9]. There is evidence that colonoscopy use has risen significantly in recent years, primarily because of increased colonoscopy screening rates [10]. Understanding the dynamics of the changes in colorectal polyps and cancer involving sizes, shapes, and the number of polyps is critical because these are associated with the efficacy of colorectal screening strategies. Manually extracting the detailed findings of a colonoscopy (e.g., number of polyps, polyp sizes, and polyp locations) from a colonoscopy procedure report and connecting with the pathology report (e.g., the histology of polyps found during colonoscopy) is quite time-consuming [14]. Extracting information entity identification from the clinical free text requires scalable techniques such as NLP, machine learning, and neural networks to convert clinical free text into a structured variable for text mining and further information extraction [15][16][17]. Colonoscopy named entity identification has drawn a considerable amount of research in recent years, and various techniques for identifying entities have been suggested [22]. Conventional methods such as Decision Tree, Support Vector Machines (SVM), Hidden Markov Model (HMM), and Conditional Random Field (CRF) are either rule-based or dictionary-based, which usually need to be manually formulated and for the current datasets only. When the dataset is updated or replaced, updating the rules contributes to high machine costs. This will cause a low recall if the original rules or dictionary are used. The latest advanced techniques are data-driven, including the methods of machine learning and deep learning. In particular, the models of Bi-LSTM-CRF (Bi-Long Short-Term Memory and Conditional Random Field) have been used successfully in the biomedical text to achieve better results and represent some of the methods that are most commonly used [23]. Existing work regarding NLP in connection with CRC and colonoscopy is limited in scope and quantity. Adapting a clinical information retrieval system, Named entity recognition (NER), identified pathology reports consistent with CRC from an electronic medical record system [26]. An application was proposed to automatically identify patients in need of CRC screening by detecting the timing and status of colorectal screening tests mentioned in the electronic clinical narrative documentation [2,14,26,27]. Neither of these studies assesses the colonoscopy procedure's detailed findings, such as the number, shape, and size of polys, due to the problem of entity boundary extraction with different entity lengths and named entity recognition. Understanding the dynamics of the changes in colorectal polyps and cancer, involving sizes, shapes, and the number of polyps, is critical because these are associated with the efficacy of colorectal screening strategies [28].
Named entity recognition (NER), also known as entity identification, entity chunking, and entity extraction, is a process of information extraction that helps to locate and classify named entities mentioned in unstructured and structured text into pre-defined categories such as medical codes, person names, etc. NER in the medical domain is more complicated for two key reasons [24]. First, because of non-standard abbreviations, acronyms, terminology, and several iterations of the same entities, several entities seldom or sometimes fail to appear in the training dataset. Second, text in clinical notes is noisy due to shorter and incomplete sentences and grammatical and typographical errors. Additionally, clinical NER for multiple facilities is more difficult due to potential interfacility variation compared with single facility data. On the one side, clinical reports involve a paragraph border, incomplete sentences, rendering it impossible to identify complex characters, and sentence boundary detection [25].
A dependency of NLP is the sentence boundary detection (SBD) task is to identify the sentence segments within a text [19]. Good segmentation of sentences will not only improve translation quality but also reduce latency in colonoscopy reports. Most of the existing such as SBD systems SATZ -An Adaptive Sentence Segmentation System [20], proposed by Palmer and Hearst (1997), Disambiguation of Punctuation Marks (DPM), and Automated Speech Recognition (ASR) system are highly accurate on standard and high-quality text, but extracting sentences or detecting the boundary of sentences from a clinical text in colonoscopy reports is a challenging task because clinical language is characterized, in addition to peculiarities like misspellings, punctuation errors and incomplete sentences, by an abundance of acronyms and abbreviations [21].
Neural embedding methods such as word2vec [29] and GloVE [30] have been widely explored over the past five years to construct low-dimensional vector representations of words and passages of text and entity boundary extraction. Neural embeddings are a common group of low-dimensional vector representation methods, phrases, or text (50-500 dimensions) whose values are assigned by a neural network functioning as a black box. However, it isn't easy to view these dimensions in a meaningful way in clinical notes. It takes extensive training and tuning of several parameters and hyperparameters to construct neural embeddings. This paper describes and evaluates a method for identifying clinical entities and entity boundary extraction from the CRC clinical report. Given a sequence of words, our model represents each word, phrase, or text using a concatenated low-dimensional vector of its corresponding word and character sequence embedding. A near-comprehensive vector representation was built of words and selected bigrams, trigrams, and abbreviations. The word vector and dictionary features are then projected into dense representations of the vectors. These are then fed into the Bidirectional Long Short-Term Memory (Bi-LSTM) to capture contextual features. Lastly, a Conditional Random Field (CRF) is used to capture dependencies between adjacent tags. With the ability of Bio-Bi-LSTM-CRF, we capture detailed colonoscopy-related information within colonoscopy records like indicators, the number of polyps, the size of polyps, and the polyp's location, and the procedure and polyp histology like adenoma and tubulovillous and adenocarcinoma.
Section Methods explains the proposed Bio-Bi-LSTM-CRF in depth. Section Model Evaluation explains about strategies for model evaluation Section Results presents the experiments and results. Section Discussion discusses several important issues of the proposed model. Section Strengths and Limitations explains about strengths and limitations of the model and f inally, Section conclusion concludes the current paper and points out potential future work.
Dataset and Annotation
A total of 44,405 de-identified colonoscopy reports, which represent all colonoscopies between 2002 and 2012 on 36,537 subjects, were extracted from four Veterans Affair Medical Centers ( Figure 1). These colonoscopy reports contain sections such as indications and findings that contain text input by a colonoscopist. The exact sections can vary between facilities. Only sections that contain text with information on the indication of polyps found during the colonoscopy were retained for analysis. Pathology reports include sections such as impressions and findings with text input by a pathologist. Only sections that contain text with information on polyp pathology were retained for analysis.
Manual annotation of colonoscopy reports provided a gold standard to benchmark the automated methods. A total of 4,000 colonoscopy reports were randomly selected (1,000 from each facility) to annotate sentences. We used 80 percent of the reports as training data, 10 percent as data validation set for hyperparameter tuning, and the remaining 10 percent as test data. The annotation processes are described in detail [31]. Briefly, two annotators manually marked the sentences devised a list of lexical units (trigger phrases) and corresponding elements (attributes) for each sentence. Based on the existing literature, three common phenotyping tasks were selected. The final list is shown in Table 1. Table 1. List of the variables of interest from colonoscopy reports
Lexical Units Description
The main components of the entity extraction pipeline are depicted in Figure 2. The pipeline includes (1) data cleaning and pre-processing, (2) sentence boundary detection, (3) vector representation, and (4) the entity extraction model. We used python 3.7 and the open-source packages NLKT, sci-kit-learn, and Tensor Flow that provide various NLP methods, machine learning methods, and neural network modules as described in depth in the sections below.
Data cleaning and Preprocessing
We pre-processed the text reports to remove all special characters (e.g., commas in a list). Misspellings in our data set included typographical errors (e.g., "cacel" instead of "cecal"), phonetic errors that could be associated with lack of familiarity with medical terms (e.g., byopsi and methastasis), and everyday language errors (e.g., hoooooootsnare). The correction of spelling errors from clinical entries using the MetaMap and auto spell checker by Google's query suggestion service.
Sentence Boundary Detection
Sentence segmentation is not a simple task in clinical data. Automatically segmenting clinical free text is an essential first step to information extraction. There are many sentence boundary detection (SBD) packages that mainly relied on punctuation marks and wide spaces. In the most text, sentence boundaries are indicated by a period, but colonoscopy reports frequently contain intersentence periods (e.g., "Dr. Patil" or "size of 3.5 cm" or "No. of polyps:") contrasting with typical biomedical text that usually had concepts from standard terminologies. Theoretically, boundary failures can result from standard medical terminologies. To solve this problem, we applied a machine learning approach to identify the sentence boundaries.
To perform sentence boundary detection, it is essential to represent the input sentences with suitable tags. S.F. (S=beginning of a sentence, F= ending of the sentence) format is used tagging in sentence boundary detection. We used the features and keywords listed in Table4to identify and tag the sentence boundaries. We used a decision tree for sentence boundary detection classification. This approach is constructed by applying the C4.5 algorithm. The example is shown in Textbox 1.
Textbox 1. Example of clinical note.
The clinical note extracted sentences containing essential lexical units. To avoid duplicate sentences and improve sentence diversity, the sentences were sorted by TF-IDF cosine distance. The sentences were manually de-identified and checked by automatic sentence boundary detection.
Vector Representation
Its complexity characterizes information on clinical notes in vocabulary and morphological richness. Therefore, using the recognition of a named entity to capture morphological information from complex medical terms can avoid vocabulary problems and compensate for traditional embedding words. The name entity recognition of clinical colonoscopy notes is usually regarded as a sequence labeling task. Given a sequence of words in a clinical notes sentence, we label each word in the sentence with a BIEOS (Begin, Inside, End, Outside, Single) tag scheme.
For example, in Textbox 2 below, "The mass was circumferential" and "Estimated blood loss was minimal "sentences are not significant. The widely used annotation modes for named entity recognition training data are BIO, BIEO, and BIESO, where B is the begin ning of an entity, I reflect the center of an entity, E is the end of an entity, S is a single entity , and O is not an entity.
Textbox 2. Example output after sequence labeling task.
In this study, we create a new vector representation by integrating dictionary function vector due to the uncertainty in the boundary of medical vocabulary terms and the complexity of representing uncommon and unknown entities. Each word's vector representation consists of the word vector and the dictionary feature vector as = + Where + is the concatenation operator, as shown by the diagram, the word vector is constructed as a combination of word embedding and character embedding. To represent character embedding, we use bidirectional LSTM to extract character features and implement certain character features.
Dictionary feature to vector
We use the n-gram models to divide the original sentence into text segments based on the word's context. After that, combined with the domain dictionary (colonoscopy) C, we produce a binary value dependent on whether or not the text segments are in C. Dictionary feature vectors are built in various dimensions, based on the number of entity categories in the dictionary. Bi-LSTM converts the final dictionary function vector from the dictionary element.
Entity Extraction Model
The entity extraction utilizing Bi-LSTM-CRF [32] used word embedding as inputs; we employ embedding character and word embedding. It allows the model to get trained with more feature information. The combined character words and dictionary feature vectors are feed into the model. This step has the following advantages:(1) more details about the features; and (2) better identification of the entity boundary problem. After that, we feed this into a Bi-LSTM model. Instead of splitting the two feature vectors as inputs, three separate Bi-LSTM versions of the proposed model are seen in Figure 3. To obtain a vector of the final character representation, we implemented a Bi-LSTM neural network and added the forward and backward sequences' output vectors. The key aim of utilizing Bi-LSTM is that it functions best to take on long-term dependencies in a phrase, and the bidirectional sequential design provides further advantages by remembering both a word's previous and potential meaning. CRF Layer: The colonoscopy reports for a sentence, the measurement of the from BI-LSTM neural network obtains a word representation, ignoring neighboring marker dependencies. For example, the tag "I" (Inside) cannot be followed with the tag "B" (Begin) or "E" (End). We use conditional random field (CRF) to determine labels from a sequence of background representations, rather than using it independently for tagging decisions can produce greater tagging accuracy in general.
Rules for Specific Task
As we are dealing with unstructured colonoscopy clinical notes from multiple facilities, we use manual rules to fix this problem. The aim of defining rules is to identify sentences that contain certain entities from the text. These rules will be dynamically updated for each facility. We explained the main four rules in Table 2. Keywords (e.g., small, medium, large, etc.), number preceeding "CM", or number preceeding "MM" Location of polyps Keywords (e.g., rectal, sigmoid, etc.) or distance from rectum given in "CM". Word processing Abbreviations (e.g., "SIG" for sigmoid, "REC" for rectal, etc.) and special word processing (e.g., "Recto-sigmoid" is a combination of "rectal" and "sigmoid", "Tubulovillous" is a combination of "Tubular" and "Villous", etc.) Number of polyps Keywords (polyps, number, numerics, and sessile)
Dataset and Experimental Settings
As previously mentioned, there are 4,000 manually annotated colonoscopy reports and five kinds of medical named entities of interest: a number of polyps, size of the polyp, location of the polyp, the procedure of removal of polyps, and status of removal as shown in Table 3. There are several sentences to every instance. These sentences are further separated into clauses by using "S" as the beginning of the sentence and "F" as mentioned in section 4.3. We calculate micro-averaged precision, recall, and F1-score for exact matching words and classify sentences accordingly to conceptual meaning from colonoscopy reports. We compare our model Bi-LSTM-CRF with improved NER and sentence boundary detection with the Bi-LSTM-CRF. We used Tensor Flow to complete the CRF module. We first run a bidirectional LSTM-CRF model forward pass for each batch, including the forward pass for both the forward and backward states of LSTM. We divide the learning rate by five if the validation accuracy decreases and used mini-batches of size 100, and training is stopped when the learning rate goes under the threshold of 10 -8 . For classifiers, we set hidden layers size to 300; because neural network architectures are usually overparameterized and vulnerable to overfitting to solve it, we used regularization strategies that have been implemented: Elastic Net regularization has been added to each LSTM hidden layer. The parameters predominantly include tag indices and batch size, where tag indices indicate the number of actual tags, and the batch size reflects the number of batch samples. The parameter settings for the proposed model are shown in Table 4.
Model evaluation
The proposed Bio-Bi-LSTM-CRF method and the state-of-the-art model Bi-LSTM-CRF were trained on the training subset, and the validation subset was used for tuning the hyper-parameters. Ultimately, the trained models were tested on the test subset, and the following outcomes were reported as the models' performance. We used precision, recall, accuracy, and F1 score to evaluate the Bio-Bi-LSTM-CRF model and Bi-LSTM-CRF model's performance compared with the overall accuracy.
Evaluations of precision, recall, accuracy, and F1 score were performed for each report. For example, all of the polyp numbers, sizes, and locations in a findings report must be correct for it to be considered accurate regardless of the number of polyps. Also, the precision, recall, accuracy, and F1 score were performed at the entity-level (e.g., each polyp size, number, and location) to find the models' performance for each facility. An exact match with the gold standard of manually annotated records is required for each entity versus the record-level where each record is considered separately.
Results
The proposed Bio-Bi-LSTM-CRF outperformed Bi-LSTM-CRF and achieved higher accuracy by 5.6%, 6.6%, and 19.1% percentage points in the indication, findings, and pathology reports, respectively, as shown in Table 5. The Bio-Bi-LSTM-CRF model's absolute accuracy was relatively high for all three reports: 88.0% for findings reports, 93.5% for indication reports, and 96.5% for pathology reports. The proposed Bio-Bi-LSTM-CRF thus reports a 15.3% increase in precision, a 14.8% increase in recall, and a 10.3% increase in F1over the state-of-the-art model Bi-LSTM-CRF. The hierarchical influence structure of Bio-Bi-LSTM-CRF aided in reducing noisy text from clinical colonoscopy notes. Both the Bi-LSTM-CRF and Bio-Bi-LSTM-CRF models performed better in entity identification indications reports, and pathology reports achieved an average of 84.75% and 95% accuracy between facilities, as shown in Table 6. However, they struggled in organizing the findings reports that mentioned characteristics of number polyps and locations of polyp in body and procedure of removal of a polyp, e.g., two diminutive mass polyp, and reports where the status of the location of the polyp was defined, e.g., A single small 3-5mm polyp away from rectum and sigmoid colon' by long-range dependencies and words. In indication reports and pathology reports, there are no long dependency sentences and multiple sentences. Whereas findings have multiples sentences representin g the information of polyps. To understand polyps' characteristics, the model needed to look further back in the sentence and relate " polyp" or "sessile." On the other hand, Bio-Bi-LSTM-CRF models entity identification on findings reports correctly. Bi-LSTM-CRF models achieved an average of 92.6% recall, as shown in Table 8. The Bio-Bi-LSTM-CRF model only made a few misclassifications of entities where the polyp location was represented in the form of distance from the rectum (130 cm). The number of polyps was described in an atypical fashion, including uncommon terms such as "semi-mass of large 10mm polyp," which only appeared in one and a few notes respectively in the dataset.
The Bio-Bi-LSTM-CRF model's overall accuracy was between 8.0% and 13.0% points higher than the Bi-LSTM-CRF method in entity identification. The accuracy of the Bio-Bi-LSTM-CRF methods between facilities ranged from 92.0% to 94.0%, as shown in Table 7. The F1-scores of Albany facility, Ann arbor facility, Detr oit facility, and Indianapolis facility are 93.7%, 96.3%, and 87.1%; as shown in Table 8The Bio-Bi-LSTM model achieved the highest overall results for all three facilities. The proposed Bio-Bi-LSTM-CRF thus reports a 12.8% increase in the F1 score. Likely, our Bio-Bi-LSTM-CRF model adapted well to other physicians' reporting style and language to achieve comparable performance in other facilities. For the polyp removal procedure, the performance of the two models of about the same level (at around 90.0%).
In comparison, the Bio-Bi-LSTM-CRF model shows an improvement of 20.9% F1 score over the Bi-LSTM-CRF in a type of CRC information from pathology reports. Bi-LSTM-CRF is struggling to extract entities from long dependences terms and multiple sentences. The weakest performance area was for location information of polyp. There is a large variety of location representation in the dataset. These are also relatively specific and unlikely to occur in the other facility data.
Discussion
This study proposed a Bio-Bi-LSTM-CRF model for efficient and accurate entity extraction from colonoscopy report notes and free-text medical narratives. The novel Bio-Bi-LSTM-CRF accurately analyzed a large sample of colonoscopy reports. Our findings demonstrate a clear improvement in accuracy, precision, and recall within a highly regarded academic healthcare system. Bio-LSTM-CRF model consistently obtained better entity extraction accuracy than a Bi-LSTM-CRF model with identical feature sets. Over experiments, it was found that the addition of sentence boundary detection, rules, and domain dictionary in the tasks could significantly boost the accuracy, thus proving the efficacy of merging practices and domain dictionary in entity extraction tasks.
Our results highlight several advantages of using Bio-Bi-LSTM-CRF in routine quality measurement using data in EHRs. The critical advantage of Bio-Bi-LSTM-CRF is that it is economically feasible. It would be expensive and time-consuming to review tens of thousands of colonoscopies reports manually. Another advantage is that Bio-Bi-LSTM-CRF allows providers to continue to use natural narrative when describing patient care. There has been criticism that structured note systems in current EHRs force providers to create unnatural and overly structured notes, which take extra time to develop and impede communication because these notes are difficult to read [33].
Compared to the direct and implicit term metrics, the neural embeddings were observed to have the inferior performance on the biomedical term similarity/relatedness benchmarks, as assessed both by our tests (against several different implementations of word2vec that other groups have developed for similar tasks) and those reported by others (see Results). However, we acknowledge that that word2vec performance can be optimized for specific tasks by adjusting different choices of parameters and hyperparameters, different neural network architectures, and other ways of handling multi-word phrases [34,35], and we did not test all possible implementations. Thus, we do not claim that our novel metrics necessarily perform better than neural embeddings in general. We decided to include only the most essential, informative words and phrases in our vocabulary in contrast to the word2vec based embeddings that encompass a much larger vocabulary . We feel that the relatively prevalent words and phrases are likely to be the most valuable for representing biomedical articles. However, it is worth exploring the effects of different vocabulary restrictions, especially when applying our vectors to other corpora, domains, and tasks.
Strengths and Limitations
In many aspects, our research is novel. First, to our knowledge, this is only the research to examine the validity of NLP for multiple center data to investigate indications, findings, and pathology laboratory results of dictated consultation notes for this purpose [36,37]. Secondly, NLP research in CRC clinical notes concentrates exclusively on a single clinical condition, such as tumor presence [26]. Our analysis is significantly more comprehensive than other more widely reported studies, looking at four different medical centers. There are multiple drawbacks to our research. In our initial training dataset (n=4,000), our choice to randomly sample resulted in CRC cases' under-sampling. As a result, the model was never trained on this feature and subsequently performed poorly during the final test set for this feature, possibly underestimating a properly trained model's effectiveness. This suggests that a real-world application of this technology may require a more purposive sampling strategy than our random sampling approach. Although the inclusion of many bigrams, trigrams, and abbreviations should reduce the problem of term ambiguity to some extent, the vector representations of words represent an overall average across word instances and different word senses. The implicit term metrics have relatively few parameters, which we have attempted to set at near-optimal values. However, the choice of 300-dimensions for the vector representations, and the particular weighting schemes for the similarity of vectors, may be regarded as best guesses and might be tuned further for optimal performance in specific tasks. However, this study's goal was to establish the feasibility of using a deep learning model to extract clinical features from dictated consult notes and inform the approach to more extensive future studies.
Conclusion
Our proposed Bio-BI-LSTM-CRF improved the perception of unknown entities, which means that the model should have a better capability to deal with emerging medical concepts and multicenter data without extra training resources. The pre-processing of the word-level input of the model with external word embeddings allowable to improve performance further and achieve state -of-the-art for the colonoscopy clinical notes entity extraction. This idea could not only be applied in medical entity extraction but also other medical named entity recognition applications such as drug names and adverse drug reactions, as well as named entity recognition tasks in different fields. | 2021-08-26T01:16:10.106Z | 2021-08-17T00:00:00.000 | {
"year": 2021,
"sha1": "a5d00ea8fd260c49669ead141de2b67d7ee8557c",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "a5d00ea8fd260c49669ead141de2b67d7ee8557c",
"s2fieldsofstudy": [
"Medicine",
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
269121707 | pes2o/s2orc | v3-fos-license | Recycling Polyethylene/Polyamide Multilayer Films with Poly(isoprene-g-Maleic Anhydride) Compatibilizer
Polymers generally form incompatible mixtures that make the process of recycling difficult, especially the mechanical recycling of mixed plastic waste. One of the most commonly used films in the packaging industry is multilayer films, mainly composed of polyethylene (PE) and polyamide (PA). Recycling these materials with such different molecular structures requires the use of compatibilizers to minimize phase separation and obtain more useful recycled materials. In this work, commercial polyisoprene–graft–maleic anhydride (PI-g-MA) was tested as a compatibilizer for a blend of PE and PA derived from the mechanical recycling of PE/PA multilayer films. Different amounts of PI-g-MA were tested, and the films made with 1.5% PI-g-MA showed the best results in terms of mechanical properties and dart impact. The films were also characterized thermally via thermogravimetric analysis (TG) and differential scanning calorimetry (DSC), using Fourier-transform infrared spectroscopy (FTIR), and morphologically using a scanning electron microscope (SEM). Other parameters, such as tearing and perforation, were analyzed.
Introduction
Packaging is an inevitability of commercial activity and the global transaction of goods.Hence, recycling plastic packaging is becoming increasingly important in preserving limited resources and protecting the environment from plastic waste [1].In the case of food packaging, multilayer films are used to maintain the integrity of food and protect it against degradation processes.The justification for the different layers films is due to the different functions carried out by the different polymer films [2].For example, lowdensity polyethylene (PE) or ethylene vinyl acetate copolymer (EVA) is commonly used to provide low-temperature heat seal ability, polyethylene terephthalate as the outside layer to improve printability and abrasion resistance, and polyamide 6 (PA) or poly(vinyl alcoholco-ethylene) as a gas barrier layer [3].So far, multilayer packaging is mostly incinerated or landfilled because it is very hard to recycle a material that can be used for the same function [4], and the functionality of the materials is greatly lost mainly due to the general incompatibility between polymers.The European Union (EU) strategy stipulates that by 2030, all plastic packaging placed on the EU market must be reusable or economically (mechanically) recyclable.
The ideal recycling process corresponds to separating the different components of multilayer films and reprocessing each layer separately.The mechanical recycling of multilayer films is an easy and attractive operation due to its simplicity and no need for specialized equipment.However, it is a big challenge since the plastic layers are generally not compatible with each other.Consequently, the mechanical recycling of multilayer films results in a single material with distinguishable phases and domains, poor mechanical properties, and barrier effects far below those of the components [3][4][5].The simpler way to improve the quality of this material is to use compatibilizers to reduce phase segregation in the final product.Compatibility is usually promoted by adding block or graft copolymers with structural affinity for the different polymers to reduce the interfacial tension between the phases, improve dispersion, and promote adhesion [3][4][5][6].
Much work has been performed to improve the compatibility of PE/PA blends using maleic anhydride groups.Moreno et al. [1] evaluated the effect of the concentration of PEg-MA as a compatibilizer with blends of LDPE/PA6, confirming that the compatibilizer causes changes in the morphology and rheology of the blend and improves the mechanical performance.Anjos et al. [20] evaluate the effect of maleic anhydride-grafted linear low-density polyethylene (LLDPE-g-MA) as a compatibilizer on the rheological, thermal, mechanical, and morphological properties of PA6/LLDPE blends and find that the addition of the compatibilizer agent increases the impact strength of the 50/50 (PA6/LLDPE) blend by 160%.Moreno et al. [21] studied the properties of blends of LDPE and PA6 generated from mechanical recycling of multilayer films, using polyethylene-graft-maleic-anhydride (PE-g-MA) as compatibilizer and different amounts of virgin PA6.The blends were immiscible at all compositions, and with a high content of PA6, a minor effect of the compatibilizer was observed.Czarnecka-Komorowsk et al. [23] investigated the influence of maleic anhydride grafted polyethylene (PE-g-MAH) on the morphology and mechanical properties of the recycled PE/PA blends.It was found that the addition of PE-g-MAH was beneficial to the blend, and it was proved that it was possible to produce granules for industrial applications.Silva et al. [24] studied the effects of the poly(ethylene-alt-maleic anhydride) (HDPE-alt-MAH) compatibilizer on the mechanical properties of HDPE/PA12 blends.It was found that the addition of 2 wt% of compatibilizer was sufficient to produce blends with significantly better mechanical properties.
There are publications where some polybutadienes are used in polyethylene mixtures [25] or modified polyamide properties [26][27][28][29][30]. Like natural rubber, polyisoprene compounds exhibit good building tack, high tensile strength, good hysteresis, and good hot tensile and hot tear strength.The presence of an anhydride group in polyisoprene could originate a compatibilizer for the PE/PA blend.In this work, we developed a process for recycling films of PE/PA made up of seven layers into a single material with improved properties using commercial polyisoprene-graft-maleic anhydride as a compatibilizer.The PE and PA constitute 70% of the total amount of polymers; others include poly(vinyl alcohol) (PVOH) and adhesive layers.The introduction of small amounts of polyisoprene-g-maleic anhydride improves the final film performance, particularly in terms of mechanical properties.
Extruders
The (PE/PA)rec pellets were mixed with the polyisoprene-graft-maleic anhydride compatibilizer (1.5-6%) and 3% of tixosil to improve the mixing quality and extruded.The production of pellets was carried out in a double screw extruder in which the mixture of raw materials was dosed in a chamber with two screws.
Tables 1 and 2 show the temperature profile and the extrusion conditions to obtain the pellets.
Production of the Films
The next step was extruding the above pellets in a balloon extruder to obtain the films.Table 3 shows the equipment parameters for producing the films via a balloon extruder.The obtained films have 100 µm of thickness.
Preparation of the Samples for Characterization
Solvent fractionation experiments were performed to investigate the composition of the binary PE/PA blends.The treatment of small samples with 85% formic acid (a solvent of the PA phase) was carried out at room temperature for 62 h.After this procedure, the samples were washed with ethanol and dried in an oven at 40 • C.
Characterization Techniques
• Thermogravimetric analysis (TGA) and Differential scanning calorimetry (DSC) The thermal stability of films was studied using thermogravimetric analysis (TGA) that was conducted using NETZSCH TG 209F1 (Netzsch, Germany).Samples were heated in a temperature range of 30-600 • C at a heating rate of 10 K•min −1 under nitrogen purge flow.Also, thermal behavior was evaluated using differential scanning calorimetry (DSC) made in a NETZSCH DSC 204 F1 Phoenix model (Netzsch, Germany).All samples were analyzed in an aluminum pan with an ordinarily closed aluminum lid.The samples were heated from room temperature to 300 • C, then cooled to −50 • C, and followed a heating cycle to 300 • C. A heating/cooling/heating rate of 10 • C•min −1 was used.A dry nitrogen environment with a purge flow was applied.
•
Fourier-transform infrared spectroscopy (FTIR) Films were characterized using FTIR in ATR mode using an Agilent Technologies Carey 630 spectrometer equipped with a Golden Gate Single Reflection Diamond ATR in the 4000-600 cm −1 range at room temperature.Spectra were collected with 4 cm −1 spectral resolution and 64 scans.OMNIC software(version 8.2.0.387) was used to analyze spectra.
•
Scanning electron microscopy To investigate the compatibilization of the polymeric blend, the fracture surfaces were analyzed via scanning electron microscopy (SEM).The specimens were frozen in liquid nitrogen prior to fracture to diminish the risk of plastic deformation.The fracture surfaces were coated with gold and analyzed with 1 kV of acceleration voltage in a field emission scanning electron microscope (FESEM), ZEISS MERLIN Compact/VPCompact, Gemini II.
• Tensile testing
Tensile tests were performed on an INSPEKT solo 2.5 mechanical tester equipped with a 500 N load cell.The film rectangular-shaped specimens were presented to tension at a rate of 50 mm•min −1 until failure.The thickness of films was measured with a digital micrometer screw gauge (precision 1 µm), and measurement was taken at three different locations on each film, and the mean value was used in the calculus of the mechanical test results.The present values are an average of five valid tests.
•
Dart Impact test and Tear test Dart impact test (dart drop test) and tear test are often used when having blown and cast films manufactured from LDPE.The analyses of dart impact test were performed on CAST and Tear test on an Instron mechanical tester equipped with a 1 kN load cell.
Film Preparation
The films of (PE/PA)rec with the polyisoprene-graft-maleic anhydride compatibilizer were produced using tubular/balloon extrusion.As polyisoprene-graft-maleic anhydride is a viscous compound, to facilitate the processing, it was necessary to add a small amount of commercial silica (3%), since without the introduction of silica, the (PE/PA)rec pellets stick to the extruder's feeding chamber, making it impossible to obtain homogeneous material.The silica impregnated with the desired amount of PI-g-MA disperses much better in the (PE/PA)rec pellets, making the final film more homogeneous.The small amount of silica also has the advantage of being a drying agent.
Thermogravimetric Analysis
Figure 1 shows the TGA thermograms of the samples.Figure 1a exhibits the graph of weight vs. temperature, and Figure 1b shows the weight derivate vs. temperature of obtained films with different amounts of PI-g-MA.
The profiles of the different samples are very similar, indicating that the presence of the compatibilizer does not change the thermal stability of the blends (Table 4).The profiles present two major mass losses, one between 360 and 400 • C, with losses between 15 and 17%, and the second between 440 and 500 • C, with losses between 70 and 80%, where most of the material decomposes.The first one is probably due to the small amount of PVOH in the original multilayer film [31].The second one corresponds to the decomposition of the PE/PA blend, which happens as a single event due to the proximity of the temperature decomposition of these polymers [7,20,32].These events are much clearer on the DTGA curve, where a shoulder at lower temperatures precedes the main degradation step.The presence of the compatibilizer does not bring any significant changes in the degradation profile of the mixtures, except for a small shift in the maximum degradation temperatures.clearer on the DTGA curve, where a shoulder at lower temperatures precedes the main degradation step.The presence of the compatibilizer does not bring any significant changes in the degradation profile of the mixtures, except for a small shift in the maximum degradation temperatures.
Differential Scanning Calorimetry
DSC analysis was performed to evaluate the changes in thermal behavior caused by the introduction of the PI-g-MA compatibilizer, Figures 2, 3, and S1 and Tables 5 and S1.
As can be seen in Figures 2 and 3, all blends exhibit two zones characteristic of endothermic melting processes and two crystallization zones corresponding to the presence of immiscible polymers.The first zone between 109 °C and 120 °C corresponds to the polyethylene melting with broad signals and the melting of polyamide-6 as a single signal around 220 °C.The crystallization zones are between 50 °C and 115 °C for polyethylene and 160 °C to 190 °C for PA.The presence of a broad melting zone for polyethylene indicates the presence of different low-density polyethylenes on the recycled material derived from wasted multilayer films.Compared with the literature, the broad melting signal for LLDPE corresponds to the presence of crystals with different lamella thickness due to the heterogeneous distribution of short chain branches, which is a characteristic of LLDPE prepared using a Ziegler-Natta type catalysis.The peak corresponds to the crystallization of longer linear chains with less short-chain branch content, and the shoulders correspond to the crystallization of shorter linear chains with higher short-chain branch content [33][34][35].Thermal events for the first heating are presented in Figure S1 and Table S1.Crystallization events (Figure 2 and Table 2) also
Differential Scanning Calorimetry
DSC analysis was performed to evaluate the changes in thermal behavior caused by the introduction of the PI-g-MA compatibilizer, Figures 2, 3 and S1 and Tables 5 and S1.
As can be seen in Figures 2 and 3, all blends exhibit two zones characteristic of endothermic melting processes and two crystallization zones corresponding to the presence of immiscible polymers.The first zone between 109 • C and 120 • C corresponds to the polyethylene melting with broad signals and the melting of polyamide-6 as a single signal around 220 • C. The crystallization zones are between 50 • C and 115 • C for polyethylene and 160 • C to 190 • C for PA.The presence of a broad melting zone for polyethylene indicates the presence of different low-density polyethylenes on the recycled material derived from wasted multilayer films.Compared with the literature, the broad melting signal for LLDPE corresponds to the presence of crystals with different lamella thickness due to the heterogeneous distribution of short chain branches, which is a characteristic of LLDPE prepared using a Ziegler-Natta type catalysis.The peak corresponds to the crystallization of longer linear chains with less short-chain branch content, and the shoulders correspond to the crystallization of shorter linear chains with higher short-chain branch content [33][34][35].Thermal events for the first heating are presented in Figure S1 and Table S1.Crystallization events (Figure 2 and Table 2) also show two major zones corresponding to the crystallization of the two polymers on the blend.A small crystallization event occurs at 62 • C for all the samples.This event is not currently seen [23] and is related to the nature of the LDPE.This event is explained by the melt topology and entanglements of PE chains during the crystallization process [35].
to the ∆Hc value, which shows greater differences between the two polymers in the presence of the compatibilizer.With respect to PE, the compatibilizer increases the ∆Hc [38], and in the case of PA, the presence of compatibilizer decreases the ∆Hc, as observed by others [8,39].This decrease in ∆Hc could be explained by fractional crystallization, in which the compatibilizer influences the size of the PA particles, causing small particles to crystallize at lower temperatures [8,39].With the introduction of the compatibilizers small changes are observed in the DSC profiles for PE melting and crystallization.In relation to PA, the Tm is similar for all blends, but in relation to Tc, there are some differences.The addition of compatibilizers reduces the temperature of crystallization.This effect in PA is reported as a signal for the interaction between PA and the compatibilizer [36,37].The effect of compatibilization via PI-g-MA can be better seen by analyzing the ∆H values for melting and crystallization of the blends [12].For PE and PA, the ∆Hm value generally decreases only slightly compared to the ∆Hc value, which shows greater differences between the two polymers in the presence of the compatibilizer.With respect to PE, the compatibilizer increases the ∆Hc [38], and in the case of PA, the presence of compatibilizer decreases the ∆Hc, as observed by others [8,39].This decrease in ∆Hc could be explained by fractional crystallization, in which the compatibilizer influences the size of the PA particles, causing small particles to crystallize at lower temperatures [8,39].
Infrared Spectroscopy (FTIR)
The ATR-FTIR spectra of the pellets of (PA/PE)rec blended with different amounts of PI-g-MA are shown in Figure 4.
The PE part is identified by the stretching vibrations of the -CH 2 groups at 2912 cm −1 (asymmetric stretching) and 2842 cm −1 (symmetric stretching).The band at 1462 cm −1 is identified as C-H bending deformation [40,41].The absorption bands at 728 cm −1 and 717 cm −1 correspond to the rocking deformations of the CH 2 group [40].
The PA present in the blend shows the band corresponding to the amide group (N-H) at 3292 cm −1 , while the C-H stretching bands of the methylene segments were observed at 2912 cm −1 and 2842 cm −1 [42,43].The adsorption bands at 1636 cm −1 and 1557 cm −1 are associated with amide groups, the first with amide-I and the second with the amide-II forms [44].The other band at 1462 cm −1 could be assigned to the CH 2 bending of the amide group [41- 43,45].
In the FTIR spectrum of PI-g-MA, the two absorption bands at 1791 cm −1 and 1712 cm −1 can be attributed to the anhydride ring, and the one weak absorption peak at 1656 cm −1 can be attributed to the C=C stretching vibration of polyisoprene [46].
The absence of bands corresponding to the carboxylic anhydride (two signals were assigned at 1856 cm−1 and 1780 cm−1) of the maleic anhydride ring (especially in the sample with 6% PI-g-MA) could suggest that the cyclic anhydride was reacted during processing.
To verify whether maleic anhydride reacted during processing, the FTIR spectrum of the sample in the form of pellets with 6% PI-g-MA (one processing) was compared with the spectrum of the same material in the form of film (two processings), Figure 5. Analysis of the spectra shows that, when the material is in pellets, there are small bands at 1705 cm −1 and 1739 cm −1 corresponding to the maleic anhydride band.These bands are no longer noticeable after the second processing of the material into a film.The absence of bands corresponding to the carboxylic anhydride (two signals w assigned at 1856 cm −1 and 1780 cm −1 ) of the maleic anhydride ring (especially in the sam with 6% PI-g-MA) could suggest that the cyclic anhydride was reacted during processi To verify whether maleic anhydride reacted during processing, the FTIR spectrum the sample in the form of pellets with 6% PI-g-MA (one processing) was compared w the spectrum of the same material in the form of film (two processings), Figure 5. Analy of the spectra shows that, when the material is in pellets, there are small bands at 1 cm −1 and 1739 cm −1 corresponding to the maleic anhydride band.These bands are longer noticeable after the second processing of the material into a film.Wavenumber(cm -1 ) 0% PI-g-MA 6% PI-g-MA 100% PI-g-MA Figure 5. ATR-IR spectra of pellets of (PE/PA)rec with different percentages of PI-g-MA and insert ATR-IR spectra of the film (two processings) and pellets (one processing) of a sample with 6%PI-g-MA.
SEM
Figure 6 shows the SEM images of sections of cryo-fractured samples of the films made from de (PE/PA)rec with different amounts of PI-g-MA before and after etching the Polymers 2024, 16, 1079 9 of 14 3.5.SEM Figure 6 shows the SEM images of sections of cryo-fractured samples of the films made from de (PE/PA)rec with different amounts of PI-g-MA before and after etching the samples with formic acid to remove the PA fraction.Considering the incompatibility between PE and PA, it was expected that the morphology of the samples would show the typical distribution of one continuous phase of PE with PA droplets of different sizes across the continuous phase, as observed by others [7,37].In our case, this pattern is not seen.The PE/PA sample shows the material in an elongated shape, like fibers that are not connected to each other.This different morphology could be because the films are subjected to stretching when blowing, and the materials, therefore, feel stretching forces and take on this shape [47].In other works, the samples commonly resulted from injection or pressing processes, which implies compression forces and, therefore, a different morphology.The presence of the compatibilizer changes the structure into a lamellar and more compact one.With the removal of the PA part with formic acid (etched samples), a smoother surface appears, and a series of small craters are visible, corresponding to the PA phase dispersed in the PE.It is possible that some dissolved PA in the formic acid covers the surface, contributing to a more homogeneous surface.
Mechanical Performance
The mechanical performance of the samples was evaluated using the tensile-strain tests.The results are presented in Figure 7.The results show that, in general, when adding PI-g-MA, there is an increase in elongation at break compared with the sample without the PI-g-MA.Also, in relation to tensile strength, there is a small gain in the compatibilized samples, with 1.5% and 3% compatibilizers.These facts are indicative of the compatibilizing effect of the PI-g-MA polymer.The material with the highest elongation at break and the highest tensile strength is the sample with 1.5% PI-g-MA with gains relative to the 0% PI-g-MA of 16% and 44%, respectively.The presence of the reactive anhydride groups reacting with the PA portion and the isoprene part interacting with the PE part contributes to this improvement, probably due to the creation of positive interfacial interactions between PE and the end groups of polyamide-6.Similar results were presented by other authors [48,49], who noted that incorporating HDPE-g-MAH into blends improved hardness as a result of increased interfacial adhesion of phases in the blends.
interfacial interactions between PE and the end groups of polyamide-6.Similar results were presented by other authors [48,49], who noted that incorporating HDPE-g-MAH into blends improved hardness as a result of increased interfacial adhesion of phases in the blends.
Tear Resistance, Perforation, and Impact Fall Dart
In the industrial environment, important properties related to film applications were studied.Table 6 shows the results of the tearing, drilling, and dart drop tests.The tear resistance measures the resistance of the film to a tear propagation event.Films containing the compatibilizer (1.5 and 6%) show better properties in the longitudinal direction than in the transverse direction compared with samples without compatibilizer.Regarding puncture resistance, there is no difference between compatible and non-compatibilized samples.With respect to the impact of the fall dart, which is a measure of the impact strength of a film, there is a clear advantage of the compatibilized films (1.5% PI-g-MA obtained 75% of the gain and 6% PI-g-MA obtained 55% of gain) compared with the noncompatibilized sample.This gain is visible in the greater resistance that the film has in relation to its breakage.
Conclusions
This work transforms a mixture of PE/PA from residues of a multilayer packaging film into a monolayer film with PI-g-MA as a compatibilizer.The resulting films were extensively characterized in terms of their morphology, thermal behavior, and physical properties.FTIR analysis confirmed that the cyclic anhydride from MA reacted during processing.SEM analyses showed differences between the non-compatibilized sample 0% PI-g-MA with a fibrous structure and the compatibilized samples with a more homogeneous and rougher structure.The compatibilized films showed an increase in elongation at break and a small gain in terms of tensile strength.Films with 1.5% and 6% compatibilizer show better properties in the longitudinal direction and 75% gain and 55% gain in the impact of the fall dart, respectively, compared with samples without a compatibilizer.These facts are indicative of the compatibilizing effect of the PI-g-MA polymer.
Figure 2 .
Figure 2. DSC thermogram of first cooling flux of films with different percentages of PI-g-MA.Figure 2. DSC thermogram of first cooling flux of films with different percentages of PI-g-MA.
Figure 2 .Figure 3 .
Figure 2. DSC thermogram of first cooling flux of films with different percentages of PI-g-MA.Figure 2. DSC thermogram of first cooling flux of films with different percentages of PI-g-MA.Polymers 2024, 16, x FOR PEER REVIEW 7
Figure 3 .
Figure 3. DSC thermogram of second heat flux of films with different percentages of PI-g-MA.
Figure 4 .
Figure 4. ATR-IR spectra of films of (PE/PA)rec with different percentages of PI-g-MA and only g-MA.
Figure 4 .
Figure 4. ATR-IR spectra of films of (PE/PA)rec with different percentages of PI-g-MA and only PI-g-MA.Polymers 2024, 16, x FOR PEER REVIEW 9 of 15
Figure 5 .
Figure5.ATR-IR spectra of pellets of (PE/PA)rec with different percentages of PI-g-MA and insert ATR-IR spectra of the film (two processings) and pellets (one processing) of a sample with 6%PI-g-MA.
Figure 6 .
Figure 6.SEM micrographs of the cryofracture surface of the (PE/PA)rec with different percentages of PI-g-MA.Figure 6. SEM micrographs of the cryofracture surface of the (PE/PA)rec with different percentages of PI-g-MA.
Figure 7 .
Figure 7. Elongation at break and tensile strength of the films.
Figure 7 .
Figure 7. Elongation at break and tensile strength of the films.
Table 4 .
Result obtained using TGA analysis.
Table 4 .
Result obtained using TGA analysis.
Table 5 .
DSC data of second heat flux of the films with different percentages of PI-g-MA.
* in parenthesis, the values for the observed shoulders.
Table 6 .
Results of tear resistance, puncture resistance, and resistance to the impact of the fall of the dart. | 2024-04-14T15:08:55.084Z | 2024-04-01T00:00:00.000 | {
"year": 2024,
"sha1": "e3a759ad148e835d366168e4fa0aeba2d07a03e9",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4360/16/8/1079/pdf?version=1712922175",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "81c2c5d3225d7f843da455c3e7b811faf376c8e8",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": []
} |
6642366 | pes2o/s2orc | v3-fos-license | Identification and characterization of the highly polymorphic locus D14S739 in the Han Chinese population
Aim To systemically select and evaluate short tandem repeats (STRs) on the chromosome 14 and obtain new STR loci as expanded genotyping markers for forensic application. Methods STRs on the chromosome 14 were filtered from Tandem Repeats Database and further selected based on their positions on the chromosome, repeat patterns of the core sequences, sequence homology of the flanking regions, and suitability of flanking regions in primer design. The STR locus with the highest heterozygosity and polymorphism information content (PIC) was selected for further analysis of genetic polymorphism, forensic parameters, and the core sequence. Results Among 26 STR loci selected as candidates, D14S739 had the highest heterozygosity (0.8691) and PIC (0.8432), and showed no deviation from the Hardy-Weinberg equilibrium. 14 alleles were observed, ranging in size from 21 to 34 tetranucleotide units in the core region of (GATA)9-18 (GACA)7-12 GACG (GACA)2 GATA. Paternity testing showed no mutations. Conclusion D14S739 is a highly informative STR locus and could be a suitable genetic marker for forensic applications in the Han Chinese population.
Aim To systemically select and evaluate short tandem repeats (STRs) on the chromosome 14 and obtain new STR loci as expanded genotyping markers for forensic application.
Methods STRs on the chromosome 14 were filtered from Tandem Repeats Database and further selected based on their positions on the chromosome, repeat patterns of the core sequences, sequence homology of the flanking regions, and suitability of flanking regions in primer design. The STR locus with the highest heterozygosity and polymorphism information content (PIC) was selected for further analysis of genetic polymorphism, forensic parameters, and the core sequence.
Conclusion D14S739 is a highly informative STR locus and could be a suitable genetic marker for forensic applications in the Han Chinese population.
Short tandem repeats (STRs) comprise the repeat units of 2 base pairs (bp) to 7 bp in length (1). Due to a high degree of length polymorphism as a result of variation in the number of repeat units and a short size of amplification products, they have become the most popular genetic markers for the identification of individuals and paternity testing (2). However, only a small number of STRs with high degree of length polymorphism is suitable for use as genotyping markers. Multiplex assays commonly include non-coding tetranucleotide and pentanucleotide repeats, which enables high combined power of discrimination (CPD) and combined power of exclusion (CPE) in a single test. Currently, commercial kits, such as PowerPlex ® Fusion System (Promega, Madison, WI, USA) and GlobalFiler ® Express Kit (Thermofisher Scientific Inc., Waltham, MA, USA) allow simultaneous amplification of more than 20 autosomal STR loci, which simplifies forensic DNA profiling (3,4).
STRs are prone to mutation in meiosis, which might result in a false maternal or paternal exclusion due to gain or loss of repeat units. Therefore, additional genetic information is required to increase the combined paternity index (CPI), which allows the detection of true parental relationships in a pedigree and reduces the chances of false exclusion. Currently, commercially available kits include some STR loci with a low power of discrimination (PD) and low power of exclusion (PE), such as TPOX. Furthermore, STRs included in Combined DNA Index System (CODIS) and European Standard Set (ESS) belong to only 18 of the 22 autosomal chromosomes (5). Therefore, some new multiplex STR typing systems were developed to provide additional information for paternity testing, such as 26plex STR assay (6). However, most STR loci used in the expanded assays, such as D14S1434, also have low PD and PE (7).
The development of six dyes permits a simultaneous detection of more STR loci in a multiplex STR typing system (4). CPD and CPE can be increased if an STR locus with low PD and PE in a multiplex STR typing system is replaced by a new STR locus with high PD and PE from the same chromosome, or if such a locus is added to the multiplex STR typing system. This is especially important for new STR loci with high PD and PE from the chromosomes that are not included in multiplex STR typing systems. The addition of these may help to avoid linkage potential between STR loci. Therefore, it is necessary to systemically select and evaluate new STR loci as genotyping markers for forensic application (8). For this purpose we intended to identify STR loci with high degree of polymorphism on chromosome 14. In fact, no STR locus on chromosome 14 has been included in common multiplex STR typing systems, even the latest PowerPlex ® Fusion System and GlobalFiler ® Express Kit. Although several STR loci on chromosome 14 have been used as expanded genotyping markers, including D14S1434 and D14S608, the use of these loci has several disadvantages. D14S1434 has been reported to have low PD and PE (9) and while D14S608 has relatively high PD, its allele frequency does not show normal distribution in all tested populations (10)(11)(12)(13)(14). D14S608 was also observed to have significant deviation from Hardy-Weinberg equilibrium (HWE) in German population (11).
In this study, STR loci on chromosome 14 were filtered from the Tandem Repeats Database (TRDB) (15) and their core and flanking sequences were further evaluated. D14S739 was shown to be highly polymorphic in a small sample size and was further characterized in the Han Chinese population.
Selection of STr loci
A total of 386 repeats on chromosome 14 were preliminarily filtered from TRDB using the following rules: 'Pattern Size' was equal to 4; 'Copy Number' was ≥8 and ≤30; 'the content of GC' was 20%-55%, '%Indels' was equal to 0, and '%Matches' was ≥90%. A set of 26 STR loci was selected based on the positions on the chromosome, repeat patterns of core sequences, sequence homology of flanking regions, and suitability of flanking regions in primer design.
Primer design, amplification, and electrophoresis
Primers were designed by using Primer v5.0 (Premier Biosoft Interpairs, Palo Alto, CA, USA). The amplification of STR loci was performed by polymerase chain reaction (PCR) including 2.5 μL 10 × PCR buffer (with MgCl 2 ), 2.0 μL deoxynucleotide mixture (2.5 mM), 1.0 μL FAM TM -labeled or unlabelled primer set (100 μM, Sangon Biotech., Shanghai, China), 1.0 μL rTaq DNA polymerase (5U/μL), and 1.0 μL sample DNA in a 25 μL final reaction volume. After an initial denaturation at 94°C for 3 minutes, PCR was carried out for 31 cycles under the following conditions: denaturation at 94°C for 30 seconds, annealing at 58°C for 35 seconds, extension at 72°C for 30 seconds, and a final extension at 72°C for 25 minutes. PCR products were separated by agarose gel electrophoresis or capillary electrophoresis in ABI PRISM 3130xL Genetic Analyzer (Thermofisher Scientific Inc.). Croat Med J. 2015;56:482-9 www.cmj.hr naming of the alleles and allelic ladder The pilot investigation of genetic polymorphism was performed with 35 individual DNA samples. The number of alleles of each STR locus was determined and the forensic parameters were evaluated. The PCR products of each allele were cloned in plasmid vectors and sequenced by 3130xL Genetic Analyzer. The alleles were named according to the sequencing results and the recommendations of the DNA Commission of the International Society of Forensic Genetics (ISFG) (16). The alleles were amplified, and then the products were diluted, mixed together, analyzed, and balanced to produce the allelic ladder (17). Panel and bin files for GeneMapper ID software v3.2 were programmed by using fixed size of allelic ladder.
Population investigation and data analysis
The bloodstains were collected from 511 unrelated individuals after informed consent had been obtained and the DNA samples were prepared by 10% Chelex-100 solution (Bio-Rad Laboratories, Hercules, CA, USA) and proteinase K (18). The allelic ladder, panel, and bins were updated when new alleles were observed. The values for allele frequencies, observed heterozygosity (Ho), expected heterozygosity (He), polymorphism information content (PIC), PD, PE were calculated, and the exact test of HWE was performed using the PowerStats v1.2 software (19) and PowerMarker software v3.25 (20). The study was approved by the ethics committee of Shanghai Medical College, Fudan University.
Selection of STr loci on chromosome 14
From a total of 27 552 loci in TRDB, we obtained 386 STR loci. The sequence homology of flanking regions was evaluated by the Blat tool (http://genome.ucsc.edu/cgi-bin/ hgBlat) and the suitability of flanking regions in primer design was assessed by Oligo v. 7.0 software (Molecular Biology Insights, West Cascade, CO, USA). A set of 26 STR loci with a spacing of about 3 Mb from each other was selected for further investigation ( Figure 1 and Table 1).
Pilot investigation of genetic polymorphism
The specificity of primer sets for the 26 STR loci was tested by PCR amplification and agarose gel electrophoresis (Table 1 and Figure 2) and further evaluated by capillary electrophoresis. Pilot investigation of genetic polymorphism showed that the locus with the highest heterozygosity, PIC, PD, and PE locus No. 20 with 9 alleles. University of Cal- ifornia Santa Cruz (UCSC) Genome browser analysis (http:// genome.ucsc.edu) showed that the locus No. 20 had an identical location on chromosome 14 as D14S739. Therefore, D14S739 was further analyzed.
The population analysis of d14S739
The allelic ladder with 9 alleles of D14S739 was prepared and the genetic diversity of D14S739 in the Han population was investigated. In all tested samples we observed 14 alleles. Insertion-deletion polymorphisms (Indels), which result in microvariants, were not observed. The forensic parameters of D14S739 including allele frequencies, Ho, He, PIC, PD, and PE were calculated and no deviation from HWE was observed (Table 2). Compared with the polymorphism and forensic parameters of CODIS STRs obtained from our laboratory (21), D14S739 was comparable to the FGA locus and superior to other loci.
The core sequence analysis of d14S739 We next analyzed the sequence of D14S739 in the human genome version 19 (Hg19). In its core region, there are two repeat motifs GTCT and ATCT. However, D14S739 was originally cloned with the oligonucleotide probe of GATA repeats (22). According to the nomenclature for STR alleles, the repeat motifs of D14S739 should be defined as GATA motif and GACA motif (16). To further determine the nucleotide sequences of all 14 alleles, the representative samples containing the alleles of D14S739 were used to amplify the target region and PCR products were cloned into pMD TM Because of the combination of two repeat motifs in the core region, alleles with the same size had different repeat patterns ( Figure 3A and B). The single nucleotide variation in alleles was also observed. The transition of cytosine to thymine in the GACA motif led to the appearance of GATA motif ( Figure 3C). Other alleles might have a similar pattern although we did not sequence all the alleles in the population. In fact, the single nucleotide polymorphism in the core region of D14S739 was confirmed by the UCSC Genome Browser Database (http://genome.ucsc.edu).
detection of d14S739 in paternity testing
The allelic ladder with 14 alleles of D14S739 was prepared and the performance of D14S739 in paternity testing was investigated in 200 trio paternity tests using PowerPlex ® 21 System (Promega, Madison, WI, USA). The transmission of alleles from parents to their offspring conformed to Mendelian laws and no mutation was observed. The representative genotypes of one trio paternity test together with the allelic ladder are shown in Figure 4.
diSCuSSion
In this study, we performed a comprehensive screening of STR loci on chromosome 14 and identified D14S739 as a highly polymorphic STR locus in the Han Chinese population. Generally, the degree of polymorphism for a genetic locus can be measured by two distinct parameters -heterozygosity and PIC (24). Our results showed that D14S739 had higher heterozygosity (0.8691) and PIC (0.8432) in the Han Chinese population than D14S1434 (0.682 and 0.645, respectively) (9) and D14S608 (0.8110 and 0.8399, respectively) (14). Similarly, D14S739 had higher PD (0.9615) and PE (0.7328) than D14S1434 (0.863 and 0.378, respectively) (9) and D14S608 (0.9504 and 0.6659, respectively) (14). Therefore, the inclusion of D14S739 in multiplex STR typing systems could help to achieve high CPD and CPE.
The addition of independent STR loci with high degree of polymorphism in multiplex STR typing systems could mini- mize adventitious matches in forensic casework. However, not all STR loci are suitable for the forensic purposes, so thorough evaluation is needed to filter out unsuitable ones (8). In this study, 26 STR loci with a spacing of 3 Mb on the chromosome were used as candidates. It is possible that we missed some of the suitable loci. In fact, it is difficult to evaluate a large number of STRs by manual screening. Therefore, a primary alignment using results from the whole genome sequencing against the reference genome could provide an overall view of the variation of all STR loci in a population, which can decrease the chance of missing polymorphic STR loci during the screening.
D14S739, also known as GATA65G10, was first cloned with the oligonucleotide probe of GATA repeats and subse-quently used for the construction of human genetic maps (22,25). The transition of cytosine to thymine creates GATA motif in the repetitive GACA region having as a consequence that alleles with the same size have different DNA sequences. The sequence variants make it difficult to accurately determine the DNA sequence of alleles with the same size. In these cases sequencing of a large number of alleles should be performed. The sequence polymorphism in the repeat motif was also observed in other STR loci, such as vWA (26). The internal allele variation might not be an important consideration in forensic casework since STR variation is primarily size-based and alleles of several STR loci with the same size, such as D21S11 and FGA, contain variable repeat blocks in the core region (26). In this study, the alleles of D14S739 were named Figure 4. electrophoretograms of d14S739 in a trio paternity test. The serial numbers 015a, 015b, and 015C represent father, son, and mother, respectively.
based on the size of the core region. The size-based variation of D14S739 leads to a high degree of polymorphism, and therefore has enough discriminating power for forensic purposes. Besides the fragment length, the sequence variation of D14S739 can provide additional information for the application of next-generation sequencing in forensic practice.
During meiosis an STR locus might lose or gain one or more repeat units, which affects the interpretation of paternity testing results. Previous studies showed the highest mutation rate for FGA and D21S11, which can be derived from the relatively large number of repeat units (27). However, FGA and D21S11 alleles with incomplete repeat units were widely observed (28). We did not observe incomplete repeat units at D14S739 locus, although D14S739 has 21-34 repeat units. We also did not observe mutation of D14S739 in 200 trio paternity tests. Therefore, D14S739 might have a relatively low mutation rate during meiosis. Since this study had a relatively small sample size, studies with larger sample sizes are needed to further determine the mutation rate of D14S739.
acknowledgments The authors thank Prof. Ziqin Zhao, Dr Huaigu Zhou, Prof. Chengtao Li, and Prof. Shilin Li for their valuable help with experiment design and manuscript preparation.
Funding This work was supported by the National Natural Science Fund (81571853 and 31270862).
ethical approval This study was conducted in accordance with the ethical guidelines of the Declaration of Helsinki 2008 and was approved by the ethics committee of Shanghai Medical College, Fudan University.
authorship declaration CS designed the experiments, performed the experiments, and analyzed the data. YZ, YZ, WZ, HX, and ZL provided technical expertise necessary for completion of this study. QT designed the experiments and reviewed the manuscript. YS and JX designed the experiments, analyzed the data, and wrote the manuscript.
Competing interests All authors have completed the Unified Competing
Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work. | 2018-04-03T04:15:51.920Z | 2015-10-01T00:00:00.000 | {
"year": 2015,
"sha1": "aac297dd040aea607fd079983b60f85ede6d07c0",
"oa_license": "CCBY",
"oa_url": "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4655933/pdf/CroatMedJ_56_0482.pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "aac297dd040aea607fd079983b60f85ede6d07c0",
"s2fieldsofstudy": [
"Biology"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
119269858 | pes2o/s2orc | v3-fos-license | All-order asymptotics of hyperbolic knot invariants from non-perturbative topological recursion of A-polynomials
We propose a conjecture to compute the all-order asymptotic expansion of the colored Jones polynomial of the complement of a hyperbolic knot, J_N(q = exp(2u/N)) when N goes to infinity. Our conjecture claims that the asymptotic expansion of the colored Jones polynomial is a the formal wave function of an integrable system whose semiclassical spectral curve S would be the SL_2(C) character variety of the knot (the A-polynomial), and is formulated in the framework of the topological recursion. It takes as starting point the proposal made recently by Dijkgraaf, Fuji and Manabe (who kept only the perturbative part of the wave function, and found some discrepancies), but it also contains the non-perturbative parts, and solves the discrepancy problem. These non-perturbative corrections are derivatives of Theta functions associated to S, but the expansion is still in powers of 1/N due to the special properties of A-polynomials. We provide a detailed check for the figure-eight knot and the once-punctured torus bundle L^2R. We also present a heuristic argument inspired from the case of torus knots, for which knot invariants can be computed from a matrix model.
Introduction
The asymptotic expansion of the colored Jones polynomial J N (K, q) of a knot K when N → ∞, and more generally of invariants of 3-manifolds, has received much attention recently. The terms of such an asymptotic expansion are also invariants of 3-manifolds, which are interesting for themselves. They are generically called "perturbative invariants". Many intriguing properties of these expansions have been observed, first in relation with hyperbolic geometry and the volume conjecture [57], [72], then concerning arithmeticity [31], modularity [61] or quantum modularity [88, Examples 4 and 5].
Solutions of the A-hat recursion relation
Garoufalidis and Lê have shown that the Jones polynomial of a knot K ⊆ S 3 , denote J N (K, q), is q-holonomic [47]: if we denote by ∆ the shift N → N + 1, there exists an operator A K (q N/2 , ∆, q) which is polynomial in its three variables, so that A K · J N (K, q) = 0. More generally, one may consider the space of solutions J (u) of the difference equation The Wilson loops in the representation of dimension N in the SL 2 (C) Chern-Simons theory (viewed as a perturbative quantum field theory after expansion around a flat connection with meridian holonomy u, with coupling constant = iπ/integer) provide such solutions, let us call them J CS,(α) (u). For some examples of hyperbolic 3-manifolds and a choice of triangulation, it has been observed [31] that the asymptotics of Hikami integral 3 J H,(α) (u) (which depend on a choice of integration contour γ α ⊆ C) coincides with J CS,(α) (u). In other words, J H,(α) are also solutions (in those examples) of (1).
We would like to propose a third method which we conjecture to provide formal solutions of 1 for any hyperbolic 3-manifold M, and relies only on algebraic geometry on the SL 2 (C) character variety of M (Conjecture 5.5). The latter is a complex curve obtained as the zero locus of a polynomial A(m, l) ∈ Z[m, l], called the A-polynomial of M. The AJ conjecture [45] states that: lim →0 A (m, l, e 2 ) ∝ A(m, l), where ∝ means up to an (irrelevant) polynomial in m. Eqn. 3 has been checked in numerous examples (it holds for instance for the figure-eight knot) and has been proved recently for a infinite class of knots [62]. In the light of the AJ conjecture, we can summarize our work by saying that we propose an algorithm to construct formal solutions of 1 starting only from the classical limit of the operator A. For the figure-eight knot, we have checked that it gives a correct result for the first few terms. The final goal would be to identify our series with the genuine all-order asymptotics of invariants in 3-dimensional defined in the realm of quantum topology, like the colored Jones polynomial. This step is subtle because of wild behavior when q is a root of unity, and non trivial Stokes phenomena, as one can already observe already in the case of the figure-
Short presentation
Let us give a flavor of our construction (all terms will be defined in the body of the article). The geometric component A(m, l) = 0 of the A-polynomial of K has a smooth model which is a compact Riemann surface C 0 of genus g. It is endowed with a point p c corresponding to the complete hyperbolic metric on S 3 \ K, and a neighborhoodŨ ⊆ C 0 of p c in bijection with a neighborhood U ⊆ C of iπ which parametrizes deformations of the hyperbolic metric of S 3 \ K. Let p u the unique point inŨ such that m(p u ) = e u . We denote ι the involution of C 0 sending (m, l) to (1/m, 1/l). In particular we have p −u = ι(p u ). In the following, p denotes a point of the curve, and in the comparison with the asymptotics of the colored Jones near u = iπ, one wishes to specialize at p = p u . Let (A, B) be a symplectic basis of homology on C 0 . We construct a formal asymptotic series with leading coefficient: and for χ ≥ 1, The notation • is used for ι(o) for some basepoint o, ω h n are the differentials forms computed by the topological recursion for the spectral curve (C 0 , ln m, ln l) with a Bergman kernel normalized on A-cycles, and U ,d,• is (0, |d|) tensor which is a sum of terms of the form: where J 1 , . . . , J s denote a partition of {1, . . . , } is s subsets. ϑ µ ν denotes the theta function with characteristics µ, ν ∈ C g associated to the matrix of periods of C 0 for the chosen basis of homology. The notations ϑ • µ ν means that we specialize its argument to w = w • ≡ • da + ζ , where da is the vector of holomorphic differentials dual to the A-cycles, ζ ∈ C g is a constant defined in Eqn. 98, and ∇ the gradient acting on the argument w. Besides, the undotted version of theta means that we specialize to w = ζ .
Formula 5 depends on a choice of basepoint o and of a characteristics µ, ν ∈ C g . If we change the homology basis (A, B), we merely obtain the same quantities for a different characteristics µ, ν. The dependence of −1 (u) of the choice of branches for the logarithms will be discussed later. For instance, the first coefficient is given by: In general, ζ depend on in a non trivial way, so our series is not a priori a power series in . It is however a well-defined formal asymptotic series: as we will see, j χ (u) is actually a function of , which does not have a power series expansion in powers of . But Apolynomials of 3-manifolds are very special polynomials: for K-theoretical reasons, ζ is constant along sequences = iπ/k where k ranges over the integers. Hence, j χ (u) specialized to such subsequences of , is indeed a function of u only. We also point out that another huge simplification occurs for a certain class of knots (containing the figure-eight knot). Let us denote ι * , the linear involution induced by ι on the homology of C 0 . When ι * = −id, we have actually ζ = 0. Thus, our series is always a power series in (without restriction) in this case. The proposal of Dijkgraaf, Fuji and Manabe [28] is tantamount to setting U ,d,• ≡ δ ,1 δ d,0 , and thus miss the theta functions. For the figure-eight knot, as we indicated, U ,d,• contributes as constants in j χ for χ ≥ 1, and their value explain the renormalizations observed by these authors. To summarize, they are due the fact that the geometric component of the character variety is not simply connected.
The leading coefficient −1 (p u ) is known to be related to the complexified volume of M for a family of uncomplete hyperbolic metrics parametrized by u. Within our conjecture, the other coefficients χ (u) also acquires a geometric meaning, as primitives of certain meromorphic 1-forms on the SL 2 (C) character variety. The computation of the coefficients with our method is less efficient than making an ansatz like Eqn. 2, pluging into the A-hat recursion relation and solving for the coefficients [31], [90]. However, it underlines the relevance of the geometry of the character variety itself for asymptotics of knot invariants, and also suggests unexpected links between knot theory and other topics in mathematical physics (Virasoro constraints, integrable systems, intersection theory on the moduli space, non-perturbative effects, etc.), via the topological recursion. It also provides a natural framework to discuss the arithmetic properties of perturbative knot invariants, at least when ι * = −id.
Outline
We first review the notions of geometry of the character variety needed to present our construction (Section 2), and the axiomatics of the topological recursion with the definition of the correlators, the partition function and the kernels (Section 3 and 4). We state precisely our conjecture concerning the asymptotic expansion of the Jones polynomial in Section 5, and check it to first orders for the figure-eight knot and the manifold L 2 R. Our intuition comes from two other aspects of the topological recursion, namely its relation to integrable systems [13] and to matrix integrals [3], [35], [21]. We give some heuristic motivations in Section 7, by examining the relation of our approach with computation of torus knots invariants from the topological recursion presented in [17]. This section is however independent of the remaining of the text. In appendix A, we propose a diagrammatic way to write χ , which may help reading the formulas, but requires more notations.
up to a global conjugation. When M is a knot complement in S 3 , the choice of l and m is canonically the longitude and the meridian around the knot. In general, we continue to call the (arbitrarily chosen) (m, l) meridian and longitude.
The locus of possible eigenvalues (m, l) ∈ C * × C * , also called character variety, has been studied in detail in [23]. It is the union of points curves. In particular, the union of the 1-dimensional components is non empty and coincide with the zero locus of a polynomial with integer coefficients: A(m, l) = 0. The latter is uniquely defined up to normalization and is called the A-polynomial of M. The A-polynomial is topological invariant of 3-manifolds endowed with a choice of basis of π 1 (∂M ), and it contains a lot of geometric information about M.
The A-polynomial has many properties, and we shall highlight those we need along the way. The first one is that, since (m, l) and (1/m, 1/l) describe the same representation up to conjugation, the A-polynomial is quasi-reciprocal: there exists integers a, b and a sign ε such that ε m a l b A(1/m, 1/l) = A(l, m). To simplify, we assume throughout the paper that M is a knot complement in a homology sphere, although most of the ideas can be extended to arbitrary 3-manifolds. In particular, this assumption implies [23] that the A-polynomial is actually even in m. We take this property into account by defining x = m 2 and y = l. The 1-form: φ = ln l d ln m is related to a notion of volume and will play an important role. The A-polynomial might not be irreducible. We denote generically A(x, y) one of its irreducible factor, which is a polynomial with integer coefficients considered in the variables x and y. We use the generic name component to refer to the subvariety C 0 defined by A(x, y) = 0 in C × × C × . There always exists reducible representations with l = 1 and m = 0, so one of the irreducible factor is (l − 1), and it defines the abelian component. The A-polynomial of the unknot is precisely (l − 1), but in general there are several non-abelian components. In a given non-abelian component C 0 , there always exists points corresponding to reducible representations, i.e. Z 0 = C 0 ∩ {(x = 1, y = 1), (x = 1, y = −1)} = ∅. Z 0 is actually the set of singular points of C 0 . After a birational transformation with integer coefficients and poles at the singular values of x, we can always find a smooth algebraic curve C which models C 0 . x and y are then meromorphic functions of C. We refer to the triple (C, x, y) as the spectral curve of the component we considered. The examples treated in Section 6 illustrate the method to arrive unambiguously to the spectral curve.
Properties of the A-polynomial
As a polynomial, the A-polynomial of a 3-manifold is very special: it satisfies the Boutroux condition and a quantization condition. These two properties hold for any 3-manifold (and any component of its A-polynomial). They come from a property in K-theory, which is proved in [23, p59], and were clarified in [63] Before coming to the K-theoretic point of view, let us describe these properties.
Boutroux condition
We have a Boutroux property: for any closed cycle Γ ⊆ C \ Z 0 , For hyperbolic 3-manifolds, this is related to the existence of a function giving the hyperbolic volume. The Boutroux condition has been underlined in [56] for plane curves of the form Pol(x, y) = 0 endowed with the 1-form φ = y dx. It appears naturally in the asymptotic study of matrix integrals, (bi)orthogonal polynomials and Painlevé transcendants, and is related to a choice of steepest descent integration contours to apply a saddle-point analysis [9], [8]. Actually, Hikami observed [53] that the A-polynomial can be obtained as the saddle-point condition in integrals of product of quantum dilogarithm constructed from triangulations and related to knot invariants. So, it is not surprising to meet a Boutroux property here.
Quantization condition
The real periods of φ are quantized: there exists a positive integer ς such that, for any closed cycle Γ ⊆ C \ Z 0 with base point p 0 such that ln m(p 0 ) = 0, iπ, This condition has first been pointed out by Gukov [50] in his formulation of the generalized volume conjecture, as a necessary condition for the SL 2 (C)-Chern-Simons theory to be quantizable. In our framework also, Eqn. 11 implies the existence of a expansion in powers of for certain quantities. We explain the mechanism in Section 2.10.
Triangulations and hyperbolic structures on 3-manifolds
The A-polynomial of a hyperbolic 3-manifold M is closely related to the deformations of the hyperbolic structure on M. Since all our examples are taken from hyperbolic manifolds, we review this relation and follow the foundational work of W. Thurston [83] and Neumann and Zagier [77]. By definition, a 3-oriented manifold M is hyperbolic if it can be endowed with a smooth, complete hyperbolic metrics with finite volume. There exists an infinite number of hyperbolic knots, i.e. knots whose complement in the ambient space is hyperbolic. Mostow rigidity theorem then states that the metrics in the definition above is unique. M is either compact, or has c cusps. Thurston explains that, modulo Dehn surgery on the cusps, M can be decomposed in a set of ideal tetrahedra glued face to face. Ideal means that all vertices of the triangulations are on the cusps. We imagine that the Dehn surgery has already been performed and start with a triangulable manifold M. It is then the interior of an oriented, bordered compact 3-manifold, whose boundary consists in c tori. So, its Euler characteristics is 0, and counting reveals that the number of tetrahedra N T equals the number of distinct edges in the triangulation. And, by construction, the number of vertices is c, the number of cusps.
In a ideal tetrahedron T, let us choose an oriented edge e pointing towards a vertex •. If we intersect T by a horosphere centered at •, we obtain a triangle whose sum of angles is π. It is thus similar to some euclidean triangle T (z e ) with vertices 0, 1 and z e . We choose a representative for which the image of z e in the tetrahedron belongs to e, and such that Im z e > 0. z e is called a shape parameter, and we may define in a unique way logarithmic shape parameters ζ e = ln z e , which are more natural to express geometric conditions. For a given vertex with incident edges e 1 , e 2 , e 3 in cyclic order, the shape parameters are z e1 , 1 − z −1 e1 and (1 − z e1 ) −1 . As a manifestation of the angle sum condition around a Euclidean condition, we have: ζ e1 + ζ e2 + ζ e3 = iπ (12) and in particular, the product of the shape parameters around a vertex is always −1. The shape parameter of an edge in opposite orientation is z −e = −z −1 e . Opposite edges in the tetrahedron have the same shape parameter. Thus, the triangulation depends a priori on N T shape parameters.
An oriented edge in the ideal triangulation of M correspond to the identification of a collection of distinct oriented edges (e j ) j of the tetrahedra. Since M is smooth along edges, we have N T gluing conditions, which are in general redundant: The data of N T shape parameters satisfying Eqn. 13 fixes a hyperbolic metrics with finite volume for M, which in general becomes singular at the vertices of the triangulation. The completion M z of M with respect to this metrics is a topological space, which may differs from M by addition of set of points at the cusps. M z happens to be a genuine hyperbolic manifold iff for any vertex • ∈ {1, . . . , c} in the triangulation: where {e • } is the set of oriented edges of tetrahedra whose image in the triangulation points towards •. It is shown in [83], [77], that the set of solutions of 13-14 is discrete. Moreover, in the neighborhood of a solution (i.e. of a manifold M z0 ), the cusp anomalies α • are local coordinates for the set of solutions of 13. In a triangulated hyperbolic 3-manifold M z0 , there is a natural PSL 2 (C) representation, namely the holonomy representation. If we assume only c = 1 cusp, let us choose two closed paths γ m , γ l ⊆ N which are representatives of a meridian and a longitude. Then, the holonomy eigenvalues (m, l) arise such that m 2 (resp. l 2 ) is the product of the shape parameters of the oriented edges crossed by γ m (resp. γ l ). The holonomy representation can be lifted [23, p71] to a SL 2 (C) representation. The lift is not unique because of a choice of square root, but we always have l c = −1 [19], and we can choose m c = −1. Now, the hyperbolic structure of M has a 1-parameter deformation in the neighborhood of M z0 , and (m, l) can be defined in a unique way as continuous (in fact, holomorphic) functions along the deformation. Also, the locus U of (m, l) achieved by the deformation is included in some 1-dimensional component of the SL 2 (C)-character variety: the geometric component C geom 0 . In other words, the deformation selects an irreducible factor of the A-polynomial, as well as a point p c on the spectral curve C geom . p c is uniquely defined by the initial value (m c , l c ) = (−1, −1) and the infinitesimal deformation around. Logarithmic variables on C geom are also very useful. We define (u, v) as analytic functions on C assuming the initial value (iπ, iπ) at p c , and such that (m, l) = (e u , e v ). The branches of the logarithm in Eqn. 9 can be unambiguously chosen as: For a countable set of points p in C geom , the space M z is not as wild as in the generic case. Indeed, if there exists coprime integers q, q such that qu + q v = iπ, M z is a manifold that differ from M by adjonction a geodesic circle at the cusp, which is obtained by performing a (q, q ) Dehn surgery on M.
Volume and Chern-Simons invariant
By a standard computation, the volume of an ideal tetrahedron with shape parameter z, endowed with its complete hyperbolic metrics, is given by the Bloch-Wigner dilogarithm D(z), which is a continuous function defined on C \ {0, 1} as: The hyperbolic volume of M z is thus: and the functional relations satisfied by the dilogarithm ensure that it does not depend on the triangulation. Another invariant of hyperbolic 3-manifolds is the Chern-Simons invariant. For compact manifolds, it was introduced in [22] and belongs to R/(2π 2 Z). For manifolds M z obtained by Dehn surgery on a hyperbolic manifold M, this definition was generalized in [67] and the invariant belongs to R/π 2 Z. Its definition in terms of a triangulation involves e Re Li 2 (z e ) plus a tricky part described in [73]. Notice that, a priori, the Chern-Simons invariant CS(M z ) only make sense when M z is a true manifold.
From a differential geometry standpoint, [77] and independently [87,Theorem 2] proved that both invariants can be extracted from the function on C geom : For a point p ∈Ũ ⊆ C geom , the volume of M z(p) is directly related to the imaginary part of Φ(p): and thanks to the Boutroux condition, it does not depend of the path from p c to p. If we assume that M z(p) is a manifold obtained by (q, q ) Dehn surgery, the real part is related to the Chern-Simons invariant. The formula involves the conjugate integers (r, r ) such that qr − q r = 1: and thanks to the quantization condition, it does not depend modulo 2π 2 Z/ς of the choice of path from p c to p. In this article, we call Vol a the analytic volume, and CS a the analytic Chern-Simons term.
Remark 2.1 Even if M is not hyperbolic, the primitive of the 1-form ln l d ln m defined over (one of the component of ) the SL 2 (C) character varieties defines a notion of complexified volume, whose imaginary part is closely related to the notion of volume of a representation.
It is enlightening to understand the volume, the Chern-Simons invariant and the properties raised in § 2.2 from the point of view of K-theory. This is the matter of the next two paragraphs.
Bloch group and hyperbolic geometry
Let K be a number field or a function field. To fix notations, K × is the multiplicative group of invertible elements of K, and K + is just K considered as an additive group. For an abelian group G, the exterior product Λ 2 Z G is the Z-module generated by the antisymmetric elements x ∧ y for x, y ∈ G, modulo the relations of compatibility with the group law (n · x) ∧ y = n(x ∧ y). When S is a set, Z · S is the free Z-module with basis the elements of S.
The pre-Bloch group [11] P(K) is the quotient of Z·(K × \{1}) by the relations [z]+[1−z] = 0 and [z] + [1/z] = 0 for any z ∈ K × \ {0}, and the five term relations: for any z, z ∈ K × \ {1} such that zz = 1. Those combinations appear precisely in the functional relations of the function D(z) of 16. Indeed, D induces a well-defined function D : For a hyperbolic manifold M with a triangulation, a point p in C geom determines shape parameters z(p) = (z e (p)) e for the triangulation. We can apply the above construction to a field K where the functions z live. It is in general an extension of the field C(C geom ), of finite degree that we denote d. The element is actually independent of the triangulation. Also, the volume is a well-defined function on C geom , given by D(ξ z ).
Up to now, the introduction of the pre-Bloch group has served merely as a rephrasing of § 2.4. Neumann and Yang [76] took a step further to reach the Chern-Simons invariant. We introduce the Rogers dilogarithm, which is a multivalued analytic function on C × \ {1}: Some computations shows that the diagram below 5 is well-defined and commutative: The Bloch group of C by definition B(C) = ker µ, and we have ρ(B(C)) ⊆ Ker e. As a matter of fact, ζ → 1 ∧ ζ ∈ Λ 2 Z C + induces an isomorphism between C/Q and Ker e. Thus, there is a mapρ from the Bloch group to C/Q. Coming back to hyperbolic geometry: since two edges carry the same shape parameter in each tetrahedron, the element ξ z defined in 22 actually sits in B(C) ⊆ P(C). When M z is a manifold, it was proved in [75] thatρ gives the irrational part of the Chern-Simons invariant:
K-theory viewpoint
We now review the interpretation of the Boutroux and quantization condition in the context of K-theory, and its relations to hyperbolic geometry.
Symbols
After a classical result of Matsumoto [68, §11], the second K-group K 2 (K) of a field K is isomorphic to Λ 2 Z K × modulo the relations z ∧(1−z) = 0. In other words, K 2 (K) = coker µ/2, where µ is the morphism introduced in § 2.5. The elements of K 2 (K) are usually called symbols, and denoted {z 1 , z 2 }. When C is a component of an A-polynomial of a 3-manifold, a theorem [23, p. 61] shows the existence of a integer ς, that we choose minimal, such that 2ς · {m, l} = 0 ∈ K 2 (C(C)).
Regulators
If z 1 , z 2 ∈ C(C), and Z denotes the set of zeroes and poles of z 1 , z 2 , the regulator map is defined as: o is a basepoint in γ and given a choice of branch of ln z 1 and ln z 2 at o, the logarithms are analytically continued starting from o along γ. One can show that this definition does not depend on o, on the initial choice of branches for the logarithm, and of the representative z 1 , z 2 of the symbol {z 1 , z 2 }. Hence, there exists a map: If {z 1 , z 2 } is 2ς-torsion (as in Eqn. 25), we see that r[z 1 , z 2 ](γ) is a 2ς th -root of unity for all closed cycles γ. We deduce that, for any closed cycle γ with basepoint o such that ln z 1 (o) = 0, iπ and the integral is well-defined, This line of reasoning has been written explicitly in [63]. This can be applied to {m, l} for a component of an A-polynomial, and justifies the Boutroux and the quantization condition of § 2.2.
Tame symbol and Boutroux condition
Given an algebraic curve C with two functions z 1 , z 2 defined on it, it might not be easy to check if {z 1 , z 2 } is torsion. However, it is elementary to check if there is a local obstruction to being torsion, i.e. if Eqn. 28 holds for all contractible, closed cycles γ in C. We focus in this paragraph only on the imaginary part of Eqn. 28, which gives rise to the Boutroux condition, and discuss its relation with the tame condition. The reason is that the Boutroux condition already has interesting consequences for the Baker-Akhiezer kernel ( § 2.10) and thus the construction of Section 4. This is formalized as follows. To any z 1 , z 2 ∈ K × , we can associate the regulator form, which is the 1-form: For any point p ∈ C 0 , let T p : K 2 (K) → C × be the map defined by: This expression is indeed independent of the representative of {z 1 , z 2 }. It is also independent of the branches of the logarithms and of the basepoint to define the integral over a small circle around p. A computation shows: so T p is closely related to the regulator map r[z 1 , z 2 ] evaluated on a small circle around p. For a given curve, z 1 and z 2 only have a finite number of zeroes and poles, so T p ({z 1 , z 2 }) = 1 except at a finite number of points. Notice that the Riemann bilinear identity applied to the meromorphic 1-forms dz1 z1 and dz2 z2 implies p∈C0 T p ({z 1 , z 2 }) = 1. We say that {z 1 , z 2 } is a weakly tame symbol if |T p ({z 1 , z 2 })| = 1 for all p, i.e. we define the subgroup: It is very easy to check if an element of K 2 (C(C)) is weakly tame of not, given Eqn. 30, and this provides a local obstruction for the Boutroux condition, and a fortiori for being torsion. Moreover, if there exists an integer ς 0 such that T p ({z 1 , z 2 }) is a 2ς th 0 -root of unity for all p ∈ C, and if {z 1 , z 2 } is torsion, ς 0 must divide the order of torsion. The tame group itself is defined as: η[z 1 , z 2 ] is always closed, since η[z 1 , z 2 ] = Im d ln z 1 ∧ d ln z 2 ) = 0. It is in general not exact, but η[z, 1 − z] = dD(z). So, we can illustrate this discussion in the context of hyperbolic 3-manifolds. The shape parameters (z e ) e sit in an extension K of C(C geom ) of some degree ς, we have at our disposal the element ξ ∈ B(K) (see Eqn. 22) and the symbol {m 2 , l} = e {z e , 1−z e } is by construction zero in K 2 (K). By coming back to K 2 (C(C geom )), one only obtains that ς 0 · {m 2 , l}, so {m, l} is 2ς-torsion. Hence {m, l} is weakly tame in a trivial way.
Arithmetics and cusp field
We now come to aspects of the A-polynomial which are relevant to the arithmeticity properties of the perturbative invariants of 3-manifolds. We aim at preparing for a clarification of the arithmetic nature of the invariants defined from the topological recursion in Section 3.1, especially when applied to A-polynomials. Unless precised otherwise, we work in the remaining of this section with any of the irreducible factor of the A-polynomial which is not of the form (lm a ± 1), and the corresponding spectral curve (C, u, v) is endowed with a marked point p c ∈ C such that m 2 (p c ) = l 2 (p c ) = 1.
We already stated that m(p c ), l(p c ) is a singular point for A. More precisely, A(m, l) ∝ (l − l(p c )) a C l−l(pc) m−m(pc) in the neighborhood of this singularity. C is a polynomial with integer coefficients, called the cusp polynomial. Although it contains less information than the A-polynomial, it retains some geometric significance and is closely related to the Cpolynomial studied by Zhang [89]. We also introduce the cusp field F, which is the splitting field of the cusp polynomial. In particular, at the vicinity of p c in C, we have (l+1) ∼ γ(m+1) where γ is a root of C, thus an element of the cusp field.
There are several notions of fields associated to a hyperbolic 3-manifold M. In presence of an ideal triangulation of M, the tetrahedron field is the field generated by the shape parameters of the tetrahedra. From another point of view, M can be realized as quotients H 3 /Γ where Γ is a discrete subgroup of PSL 2 (C) of finite covolume. One can define the invariant trace field, which is the field generated by the trace of squares of elements of Γ. It is clear that: cusp field ⊆ invariant trace field ⊆ tetrahedron field.
There are examples of hyperbolic knot complements where the cusp field is strictly smaller than the tetrahedron field [74]. The numbers produced from the topological recursion will naturally live in the cusp field F.
Remark 2.2
When C 0 is the geometric component of an A-polynomial of a triangulated 3manifold, [20] ensures that the shape parameters are rational functions of l and m. Hence, the order of torsion of {m, l} is 2 (i.e. ς = 1), and the invariant trace field coincides with the tetrahedron field.
Definition of A-spectral curves
Since we will often use this setting, we give the name A-spectral curve (over a field K) to the data of: • a curve C defined by an equation of the form Pol(m, l) = 0 (with coefficients in A), such that {m, l} is 2ς-torsion in K 2 (C(C)) for some minimal integer ς. We assume that Pol(m, l) is irreducible and not proportional to lm b ± 1 for some integer b.
• a compact Riemann surface C 0 which is a smooth model for C, and a marked point p c ∈ C 0 such that l(p c ) 2 = m(p c ) 2 = 1.
• two functions u = ln m and v = ln l on C 0 , and the differential form φ = v du.
• we add the technical assumption that the zeroes of v are simple.
One may wonder if all A-spectral curves over K arise as components of the A-polynomial of some 3-manifold. The answer does not seem to be known. K tame 2 (C) (and a fortiori K w.tame 2 (C)) for a compact Riemann surface C 0 of genus g ≥ 1 defined over Q is in general not trivial. Part of a conjecture of Beilinson predicts that a certain subgroup of K tame 2 (C) has rank g. Yet, non zero tame symbols are not easy to exhibit, see for instance [32] where elements in the tame group of some hyperelliptic curves are constructed.
Algebraic geometry on the spectral curve
We now come to the study of algebraic geometry on the spectral curve (C, u, v) Topology, cycles, and holomorphic 1-forms The curve C 0 defines a compact Riemann surface of a certain genus g, which does not depends on the smooth model C 0 for C. Actually, the genus can be computed from the polynomial A(m, l) as the dimension g of the space of holomorphic forms, i.e. rational expressions h(m 2 , l)dm which are nowhere singular. Let (A j , B j ) j be a symplectic basis of homology cycles: For the moment, we choose an arbitrary basis, and we will have to consider later how objects depend on the basis, i.e to describe the action of the modular group Sp 2g (Z). There is a notion of dual basis of holomorphic forms (da j ) j , characterized by: Then, the period matrix is defined as: and a classical result states that it is symmetric with positive definite imaginary part. We choose an arbitrary base point o, for example o = p c , and introduce the Abel map: When g = 1, C 0 is an elliptic curve and a is an isomorphism. When g ≥ 2, this is only an immersion.
Theta functions and characteristics
For any g × g matrix τ which is symmetric with positive definite imaginary part, we can define the theta function: Where there is no confusion, we omit to write the dependance in τ . θ is an even, quasi periodic function with respect to the lattice Z g ⊕ τ Z g : We define a gradient ∇ acting implicitly on the variable w, and a gradient D acting on the variable τ : The theta function is solution to the heat equation: Throughout the article, we are going to use tensor notations, and indicate with a · the contraction of indices. We consider ∇θ and Dθ, and more generally ∇ ⊗l θ (resp. D ⊗l θ) as a l-linear form (resp. a 2l-linear form), i.e. a [0, l] tensor. For example, if T is a [l, 0] tensor, we may write: ∂ l θ ∂w j1 · · · ∂w j l T j1,...,j l .
A half-characteristics is a vector c = 1 2 (n + τ · m) where n, m ∈ Z g . It is said odd or even depending on the parity of the scalar product n · m. Eqn. 40 implies that θ(c|τ ) and its even-order derivatives vanish at odd half-characteristics, while the odd-order derivatives of θ(w)e iπµ·w vanish at even half-characteristics of the form 1 2 (n + τ · m). There is a notation for theta functions whose argument is shifted by a half-characteristics c = ν + τ · µ, Notice that we still have:
Bergman kernel
For us, a Bergman kernel is a symmetric (1, 1) form B(p 1 , p 2 ) on C 0 × C 0 which has no residues and has no singularities except for a double pole with leading coefficient 1 on the diagonal, i.e. in a local coordinate λ: If we pick up a symplectic basis of homology (A, B), there is a unique Bergman kernel B(p 1 , p 2 ) which is normalized on the A-cycles: ∀j ∈ {1, . . . , g}, Moreover, B is symmetric in p 1 and p 2 and the basis of holomorphic form is retrieved by: Any other Bergman kernel takes the form: where κ is a symmetric g × g matrix of complex numbers and t denotes the transposition.
As a matter of fact, B κ satisfies the relations 47 and 48 if we replace (A, B) by a symplectic basis of generalized cycles (A κ , B κ ) defined by: In this formula, A and B should be interpreted as column vectors with g rows. The Bergman kernel normalized of the A-cycles can always be expressed in terms of theta functions: where c is any non singular odd half-characteristics. Non-singular means that the right hand does not vanish identically when p 1 , p 2 ∈ C 0 , and such characteristics exist [69]. Yet, this formula is not very useful for computations when g ≥ 2. In practice, one may start from the equation A(m 2 , l) = 0 defining C and C 0 , and find "by hand" a Bergman kernel and a basis of holomorphic forms expressed as rational expressions in m 2 and l with rational coefficients. Both methods are illustrated for genus 1 curves in Section 6.2.
Prime form
Let c be a non singular odd half-characteristics. We introduce a holomorphic 1-form: It is such that its 2g − 2 zeroes are all double. Then, the prime form It is antisymmetric in p 1 and p 2 , it has has a zero iff p 1 = p 2 in C 0 , and in a local coordinate λ: The prime form appears in this article through the formulas:
Modular transformations
The group Sp 2g (Z) acts on those objects by transformation of the symplectic basis of homology cycles. Let γ be an element of Sp 2g (Z).
• The cycles A and B, interpreted as column vectors with g rows, transform by definition as: where a, b, c, d are g × g integer matrices. The new basis ( γ A, γ B) is symplectic (see Eqn. 35) iff t b d and t c a are symmetric and t a d − t c b = 1. These are indeed the condition under which M γ belongs to Sp 2g (Z).
• The dual basis of holomorphic forms, interpreted as a row vector with g columns, transforms as a modular weigth −1 vector: • The matrix of periods transforms as: Using the relations defining Sp 2g , one can check: so that γ τ is indeed symmetric. We have denoted M t , the transposed of a matrix M .
• The Bergman kernel B κ ; τ defined from the chosen basis of cycles (we have stressed the dependance in τ ), transforms as: • The generalized cycles (A κ , B κ ) on which B κ is normalized are modular expression of weight 1: We have used the relation • The theta function transforms as: where ∆ γ is the half-characteristics ∆ γ = 1 2 (diag(ab t ) + diag(cd t )τ ) and Ξ γ a eighth root of unity.
Baker-Akhiezer spinors
Given a 1-form ω on C 0 , a complex number H ∈ C × , and vectors µ, ν ∈ C g /Z g , we set: with For a vector w ∈ C g , we have denoted frac[w] the vector of [0, 1[ g which is equal to w modulo Z g . ψ BA is called a Baker-Akhiezer spinor, it is a (1/2, 1/2)-form defined a priori on the universal cover of C 0 × C 0 , since we have: It is regular apart from a simple pole when p 1 = p 2 : and has an essential singularity when p 1 or p 2 reach a singularity of ω, of the form: Baker-Akhiezer functions have been introduced in [59] to write down some explicit solutions of the KP hierarchy. They can be obtained from the Baker-Akhiezer spinor when ω is a meromorphic 1-form, and by sending p 2 to a pole of ω with an appropriate regularization (see for instance [13]). Modular transformations act on ψ BA only by a change of the vectors µ, ν. We have introduced a normalization constant H ∈ C × , to be adjusted later. In general, the ratio involving ϑ µ ν does not have a limit, neither has a power series expansion when H → 0.
But we can say more if we assume the Boutroux and the quantization condition, i.e. that there exists ς ∈ N * such that, for all closed cycles Γ: Let us denote s A and s B integer vectors such that: It is then natural to consider values of H −1 belonging to arithmetic subsequences on the imaginary axis: Indeed, we find: so that the argument of the theta functions only depend on r = k mod ς. We have: and the Boutroux condition also ensure that Im p1 p2 ω does not depend on the path of integration between p 1 and p 2 . For a hyperbolic 3-manifold, if we choose ω = v du, the right hand side is Vol a (p 1 ) − Vol a (p 2 ) and this asymptotics is exactly the one involved in the generalized the volume conjecture (see § 5.2)
Branchpoints and local involution
In this article, we reserve the name ramification points to points in C 0 which are zeroes of du = d ln m. The value of m at a ramification point is called a branchpoint. We use generically the letter a to denote a ramification point. Since C is defined by a polynomial equation A(e u , e v ) = 0, we must have m(a) = 0, ∞. When a is a simple zero of d ln m, we call it a simple ramification point, and we can define at least in a neighborhood U a ⊆ C 0 of a the local involution p → p: Since A is quasi-reciprocal and has real coefficients, the involution ι : (l, m) → (1/m, 1/l) and the complex conjugation * act on the set of coordinates (m(a), l(a)) of the ramification points, and decompose it into orbits with 2 elements (for an a such that (l(a), m(a)) is real or unitary) or 4 elements (in general). Amphichiral knot complements admit an orientation reversing automorphism, so that ι m (m, l) = (1/m, l) by ι l (m, l) = (m, 1/l) are separately symmetries of their A-polynomial. Then at the level of spectral curves, the set of ramification points can be decomposed further into orbits of 2, 4 or 8 elements.
Topological recursion
The topological recursion associates, to any spectral curve (C 0 , u, v), a family of symmetric (1, . . . , 1) forms ω h n (p 1 , . . . , p n ) on C n 0 (n ∈ N * , h ∈ N) and a family of numbers F h (h ∈ N). These objects have many properties, we shall only mention those we use without proofs. We refer to [41] for a detailed review of the topological recursion. The fact that, here or in topological strings, one encounters spectral curves of the form Pol(e u , e v ) = 0 rather than Pol(u, v) = 0, does not make a big difference in the formalism.
We assume that all ramification points are simple. This is satisfied for most of the A-polynomials we have studied (see Figs. B.3-B.3). The topological recursion can also be defined when some ramification points are not simple [80], [14], but we do not address this issue here.
Recursion kernel
We introduce the recursion kernel: K(p 0 , p) is a 1-form with respect to p 0 globally defined on C 0 , and a (−1)-form with respect to p which is defined locally around each ramification point.
Differential forms
We define: and recursively: In the left hand side, I = {1, . . . , n−1} and p I in a (n−1)-uple of points of C 0 . For any J ⊆ I, p J is the uple of points indexed by the subset J. In the right hand side, we take the residues at all ramification points, and the in the right hand side ranges over h ∈ {0, . . . , h} and all splitting of variables J ⊆ I, excluding (J, h ) = (∅, 0) and (I, h). The formula above is a recursion on the level χ = 2h − 2 + n. ω h n has a diagrammatic interpretation ( Fig. 1), it can be written as a sum over graphs with n external legs, h handles, and thus Euler characteristics −χ. However, the weights of the graphs are non local, they involve stacks of 2g + 2 − n residues where the ordering matters.
Although Eqn. 75 seems to give a special role to the variable p 0 , one can prove (e.g. from the diagrammatic representation) that ω h n (p 0 , . . . , p n−1 ) is symmetric in p 0 , . . . , p n−1 . Except maybe ω 0 1 , the ω h n are meromorphic (1, . . . , 1)-forms on C n 0 , which have no residues and have poles only at the ramification points. We illustrate the computation at level 1. To write down the residues it is convenient to choose a local coordinate at each ramification point, for instance λ a (p) = m(p) − m(a) = √ e u(p) − e u(a) , which has the advantage that λ a (p) = −λ a (p). If r is a function or R is a 1-form, we denote: = + Figure 1: Diagrammatic representation of the topological recursion which defines ω h n . Each ω h n is represented as a "surface" with h handles and n punctures, i.e. with Euler characteristics χ = 2 − 2h − n. The diagrammatic representation of the topological recursion, is that one can compute ω h n with Euler characteristics χ in terms of ω h n with χ = 2 − 2h − n > χ, by "removing a pair of pants" from the corresponding surface.
Then, we find: To write down ω 1 1 , we need to expand: Then, we have:
Stable free energies
We have already met the abelian function: For h ≥ 2, we define: Since ω h 1 has no residues, F h does not depend on the basepoint o. The numbers F h are called the stable free energies of the spectral curve. We are not going to give an explicit definition of the unstable free energies F 0 and F 1 . Actually, for the computation of the BA kernels and later the asymptotics of the colored Jones polynomial, it is not necessary to know how to compute the free energies, we only need one of their key property called special geometry (see Eqn. 84). So, we just state that there exists F 0 and F 1 satisfying Eqn. 84, it is in fact a way to define them.
Deformation of spectral curves
By abuse of notations, we write F h = ω h n=0 , i.e. we consider ω h n for all n, h ∈ N. Unless specified, the properties mentioned below also hold for the unstable free energies. Special geometry expresses the variation of ω h n when φ = v du is deformed by addition of a meromorphic 1-form Ω. By form-cycle duality on C 0 , to any meromorphic 1-form Ω we can associate a cycle Ω * and a germ of holomorphic function on Ω * denoted Λ Ω , such that: Then, for a smooth family of spectral curves S α = (C 0 , u α , v α ) such that: we have: Notice that from the expression of ω 0 3 in Eqn. 77, one retrieves as a special case the analog of Rauch variational formula [81] for the variation of the Bergman kernel ω 0 2 = B along any meromorphic deformation.
In this article, deformations by holomorphic 1-forms and by 1-forms with simple poles will play a special role.
Variations of filling fractions
The filling fractions are defined by: Performing a variation of filling fractions amounts to add to v du a holomorphic 1-form, i.e. use the deformation: We denote ω h,(l) n the [0, l]-tensor of l th derivatives of ω h n with respect to the filling fractions, and according to Eqn. 84: In particular, the tensor of second derivatives of F 0 = ω 0 0 is the matrix of periods:
Deformation by simple poles
Given a couple of distinct points (p 1 , p 2 ), we denote: This 1-form is characterized by a simple pole at p = p 1 (resp. p = p 2 ) with residue 1 (resp. −1), no other singularities, and vanishing A-cycle integrals. If we perform an infinitesimal deformation with Ω(p) = dS p2,p1 (p), we obtain according to Eqn. 84:
Symplectic invariance
The topological recursion also has nice properties under global transformations of the spectral curve (C 0 , u, v). To simplify, we consider in this paragraph (n, h) = (0, 0), (0, 1), (1, 0), and just mention that the properties below are slightly modified for those cases. It is very easy to prove from the definitions: ) with f at least a germ of holomorphic function in the neighborhood of the values u(a), ω h n are unchanged. According to the first property, replacing m = e u and l = e v by some of their powers i.e. use (±m a , ±l b ) instead of (m, l), only affect the ω h n by a scaling factor. The second property tells us that the ω h n are the same if we change the signs of m and l, or even replace 6 l by lm a for some power a. There is conjecturally a third property concerning the exchange of u and v: , the F h are unchanged, and for n ≥ 1, the cohomology class of ω h n is multiplied by the sign (−1) n . This has only been proved [40] when u and v are meromorphic function on the curve C 0 , that is for spectral curves defined by an equation Pol(u, v) = 0. This invariance of the free energies under this exchange has meaningful consequences in random matrix theory and enumerative geometry (see [39, §10.4.1] for an example). Here and in topological strings, we rather have to consider spectral curves of the form Pol(e u , e v ) = 0. We believe that Property 3.3 survives in this context with a few extra assumptions, although this has not been established yet. For example, within "remodeling the B-model", it implies the framing independence of the closed topological string sector.
In other words, if Property 3.3 holds, the F g , and cohomology classes of the ω h n up to a sign, are invariant under all the transformations which preserve the symbol du ∧ dv. This suggests to consider the F h and the ω h n up to a sign as "symplectic invariants" of the function field K = C(C). We have seen in § 2.5 that the real part of the primitive of ω 0 1 essentially coincide with the Bloch regulator of the symbol {m, l}. It would be interesting to investigate the possible meaning of the topological recursion in terms of K-theory of K.
Deformation of the Bergman kernel
Instead of B(p 1 , p 2 ), we could have used in the definitions 73 and 75 another Bergman kernel: We denote ω h n|κ the corresponding objects. They are polynomials of degree 3h − 3 + n in κ, and it is not difficult to prove: The special geometry (Eqn. 84) for meromorphic deformations normalized on the A-cycles still holds for ω h n|κ at any fixed κ. However, variations of κ and filling fractions are mixed, since the holomorphic forms da j in Eqn. 86 are defined from B = B κ=0 and not B κ . The appropriate formula can be found in [39], it is closely related to "holomorphic anomaly equations" [7], but it will not be used in this article.
Effect of an involution
The A-polynomial comes with an involution ι : (m, l) = (1/m, 1/l). It induces an involutive linear map ι * on the space of holomorphic 1-forms on C 0 . The g eigenvalues of ι * are thus ±1. By integration, it induces an involutive isomorphism of the Jacobian of the curve, that we denote ι * . The number of +1 eigenvalues is the genus of the quotient curve C 0 /ι.
The case ι * = ε id is of particular interest. When ε = 1, ι * is a translation by a halfperiod, and when ε = −1, ι * is a central symmetry. In these two situations, all admissible Bergman kernels: are invariant under ι, and so is the recursion kernel K κ (z 0 , z). Since the set of ramification points is stable under ι, we can recast the residue formula by choosing a representative a in each pair {a, ι(a)} of ramification points: By recursion on 2g−2+n ≤ 0, we infer that E h n|κ (ι(p), p I ) = E h n|κ (p, ι(p I )) and ω n|κ (p 0 , p I ) = ω h n|κ (ι(p 0 ), ι(p I )). This result has an interesting corollary when ι * = −id: by duality, ι * B κ = −B κ , hence and in the case (n, h) = (1, 0), since ln d ln m is always invariant under ι, we have: As one can see in Figs. B.3 and B.3, ι * = −id is neither rare nor the rule for complement of hyperbolic knots. We observe however that the genus of the quotient C 0 /ι is low compared to the genus of C 0 : the "simplest" knot we found for which the quotient has not genus 0 is 8 21 . The geometrical significance of these observations from the point of view of knot theory is unclear.
Non-perturbative topological recursion
The perturbative partition function is usually defined as: where F h are the free energies. However, the genuine partition function of a quantum field theory (like the Chern-Simons theory or topological string theory) should have properties that Z pert,H does not satisfy. For instance, it should be independent of the classical solution chosen to quantize the theory (background independence), and it should have modular properties (e.g. S-duality) whenever this makes sense. From the topological recursion applied to a spectral curve (C 0 , u, v), and theta functions, we are going to define a non-perturbative partition function T H which implements such properties. Modular transformations correspond here to change of symplectic basis of cycles on C 0 . Then, one can define non-perturbative "wave functions". To keep a precise vocabulary, we shall introduce quantities that we call n|n-kernels, which depend on 2n points on the curve. In particular, the leading order of ψ [1|1] H (p 1 , p 2 ) when H → 0 will be given by the Baker-Akhiezer spinor. We prefer to use a new letter for the formal parameter. We shall find later that in the application we consider, it must be identified to defined in terms of the parameter q = e 2 in which the colored Jones polynomial is a Laurent polynomial, but this identification might be different when considering other problems.
Definitions
We use the notations of § 2.9. We take as data a spectral curve (C 0 , u, v) endowed with a basis of cycles, we choose two vectors µ, ν ∈ C g and we set: We give the definitions, which we comment in § 4.2.
Partition function
The non-perturbative partition function is by definition: We may isolate its leading behavior by writing where now lim H→0TH = 1. We consider this expression as a formal asymptotic series with parameter H → 0. The coefficient of H χ in general depend on H, but does not have a power series expansion in H. Thus, it is meaningful to speak of the χ th -order term in the expansion, keeping in mind that this coefficient may also depend on H.
(1|1)-Kernel
In integrable systems, the Sato formula expresses the wave function as Schlesinger transforms of the tau function, which in our language correspond to adding a 1-form with simple poles to φ = v du. Actually, we prefer to work with the kernel ψ H (p 1 , p 2 ) which is a function on C 0 × C 0 , defined as: where dS was defined in Eqn. 89. We introduce shortcut notations: ψ H (p 1 , p 2 ) can be computed thanks to special geometry: In the second line of Eqn. 103, all the n j variables in ω hj ,(dj ) nj are integrated over p1 p2 , and recall from special geometry that: Again, Eqn. 103 should be understood as a formal asymptotic series with parameter H → 0. It can be shown [13] that ψ H (p 1 , p 2 ) does not change when p 1 or p 2 goes around an A or a B cycle. Since ψ is the ratio of two partition function, the exponential involving the free energies F h in the numerator of the first line of Eqn. 103 cancels with the same factor present in the denominator. As we claimed earlier, only the expression of ω h n is needed to compute the kernel, not the expression of the free energies. We may isolate its leading behavior: where now lim H→0ψH (p 1 , p 2 ) = 1.
n|n-kernels
If we perform n successive Schlesinger transformations, we are led to define the n|n-kernels: which are functions on C 2n 0 . Eqn. 103 has a straightforward generalization: In this context, • = n i=1 pi oi and ϑ • stands for:
Diagrammatic representation
In Appendix A, we explain that the formulae for the non perturbative partition function (Eqn. 99) and the n|n kernels (Eqn. 107) can be represented as a sum over (maybe disconnected) diagrams. To a given order in H, there is only a finite sum of allowed diagrams. With this formalism, it is easy to reexponentiate the series above, i.e. to compute the asymptotic series for ln T H or lnψ H : they can be written as sum over connected diagrams.
Special properties for A-spectral curves
When the symbol {e u , e v } is 2ς-torsion in K 2 (C(C)), the spectral curve satisfies the Boutroux condition and the quantization condition. So, when k is an integer going to infinity along arithmetic subsequences of step ς and:
Remarks
Eqn. 99 for the non-perturbative partition function was first derived in [36] as a heuristic formula to compute the asymptotics of matrix integrals, N = H −1 playing the role of the matrix size. In [38] it was proved that it has order by order in powers of H −1 a property of background indepedence, and that it transform like a theta function of characteristics [µ, ν] under modular transformation. Actually, Eqn. 99 is the result of summing all perturbative partition functions over filling fractions shifted by integer multiples of H. This operation looks very much like the Whitham averaging in integrable systems, and we conjectured (and checked to the first non trivial order) in [13] that T H is indeed a formal tau function of an integrable system whose times are moduli of the spectral curves. In that article, we also introduced a spinor version of the kernel ψ H (p 1 , p 2 ) (Eqn. 101), in order to build a wave function in the language of integrable systems. "Wave function" is a generic name for any complex-valued solution of a linear ODE's or difference equation. Ψ α (m) ≡ ψ H (p α (m), p 0 ) should be considered as the asymptotic series of a wave function, where p 0 is a point hold fixed in C 0 . Different branches p α (m) give rise to wave functions with dominant asymptotic behavior in different sectors. Typically, a wave function is a linear combination of Ψ α (m), and thus its asymptotics is subject to the Stokes phenomenon when one goes from one sector to the other. The advantage of introducing the kernels is that the Stokes phenomenon is described by a single object ψ H (p, p 0 ) with p ∈ C 0 , through the branching structure of the covering m : C 0 → C.
One can in principle derive the difference equation satisfied by ψ H (p, p 0 ) order by order in H, and we expect it to have an expansion in powers of H, no matter if the spectral curve satisfy the Boutroux and the quantization condition. However, a general expression for the resummated difference operator annihilating the wave function (and its n|n counterpart) just from the data of the spectral curve is not available. Recently, Gukov and Su lkowski [52] have pointed out that, adding some assumption on the form of the answer, allows to reconstruct the full ODE or difference operator from the knowledge of the first orders. In particular in the context of hyperbolic geometry, the A-hat polynomial [45] is expected to appear as one of those operators (cf. § 5.3). The A-hat polynomial is known in closed form for many knots, and Dimofte [29] has discussed a procedure to construct theÂ-polynomial from the A-polynomial. Those observations might give hints towards a general theory for the reconstruction of an exact integrable system whose tau function has precisely an asymptotic given to all order by Eqn. 99 in the limit H → 0.
Rewriting in terms of modular quantities
It was proved in [38] that T H has modular properties. Since the Bergman kernel B 0 is not modular invariant, the ω h n are not either modular. Similarly, although the theta function is modular, its derivative are not. So, the modular properties in the expression 99 are not manifest.
From Eqn. 60, one sees that the deformed Bergman kernel B κ is modular invariant if we have chosen κ = κ(τ ) as a function of τ which is quasimodular of weight 2, namely: If this is the case, it is straightforward to deduce from the topological recursions formula that ω h n|κ(τ ) is modular invariant when 2h − 2 + n ≥ 0. It is often easier to compute modular objects than non-modular ones, so imagine that we have computed the ω h n|κ(τ ) . We would like to write T H only in terms of ω h n|κ(τ ) . This can be done using Eqn. 92 to express ω h n ≡ ω h n|κ=0 in terms of ω h n|κ . The result (valid for any κ) is: where: F and: Certain linear combinations of derivatives of theta are modular, and the T d|κ(τ ) precisely provide such combinations. In fact, the proof that T H is modular given in [38] amounts to prove that T d|κ(τ ) are modular. In the context of elliptic curves, we shall see in § 6.1 that it is natural to choose κ(τ ) proportional to E 2 (τ ), and for this choice, T 2d|κ(τ ) is related to the d th -order Serre derivative of theta functions.
Effect of an involution
When the genus of the quotient C 0 /ι is zero, only the terms with even d j remain in the partition function (Eqn. 99) and the kernel (Eqn. 103). In this paragraph, we assume it is the case. The conclusion of § 3.5 was that only even order derivatives of theta functions appear in the formulas, since the odd order derivatives are contracted with zero. Then, we may trade ∇ ⊗2 for a derivative with respect to the period matrix: Besides, from Property 3.5 we learn that ζ H = 0 for any H. Then, the non-perturbative partition function and the non-perturbative n|n kernels happen to be formal power series in H. And, in order to compute them, we only have to compute derivatives of Thetanullwerten with respect to the matrix of periods. For instance, the partition function reads: On the other hand, if we compute the perturbative partition function with the Bergman kernel B κ , we find with help of Eqn. 92: This expression is very similar to the non-perturbative partition function computed with the Bergman kernel B 0 . More precisely: The analogy carries at the level of the kernels. For instance, the perturbative kernel computed with B κ is defined as: and we observe that: Examples of knots for which ι * = −id can be read off Figs. B.3-B.3. For instance, it happens for the figure eight-knot and the manifold L 2 R. These two examples have be studied in [28], where it was proposed that asymptotics of the colored Jones polynomial could be computed from ψ pert,H|κ , at the price of ad hoc renormalizations of κ ⊗d to all orders. This phenomenon is explained by Eqns. 117 and 118, and this explanation is verified on examples in Section 6.
Application to knot invariants
Our main conjecture is formulated in § 5.4. We first explain the background of Chern-Simons theory and facts about volume conjectures, which allow a better understanding of the identification of parameters and of the complicated Stokes phenomenon when we consider functions in the variable u.
Generalities on Chern-Simons theory
With compact gauge group The partition function of Chern-Simons theory of compact gauge group G (and corresponding Lie algebra g) in a closed 3-manifold M is formally the path integral over g-connections A on M, of the Chern-Simons action: It depends on the Planck constant . A way to define properly this integral is to choose a saddle point A cl of the action S CS , and perform an expansion around A cl as usual in perturbative quantum field theory. By construction of Chern-Simons theory, the saddle points (also called "classical solutions") are flat connections on M, i.e. those satisfying dA cl + A cl ∧ A cl = 0. However, there are in general many equivalence classes of flat connections, and one wishes the genuine partition function to be a sum over all classes of the perturbative partition functions, with some coefficients α cl : This sum is finite when assumes a value 7 of the form: where K is an integer called level and H ∨ is the dual Coxeter number of G. There is actually a rigorous definition of the Wilson lines for these values [84], [82].
With complex gauge group
The complexification comes in two steps. We shall be sketchy here and refer to [34] for details. Firstly, one considers a Chern-Simons theory with complex gauge group G C , whose Lie algebra is obtained from g by Weyl's unitary trick, and with the new action S CS [A] + S CS [A * ]. The partition function then admits a decomposition in perturbative blocks, defined by expansion around a g C -valued flat connexion: The partition function is real when * and A * are the complex conjugates of and A. Secondly, at the level of the perturbative blocks, one consider a complexified version of the theory by assuming A and A * independent g C -valued connections. The blocks then have a factorization Z cl The Φ cl G are called holomorphic blocks, and they will play an important rôle in the following. By construction, they have an expansion in power series of , whose coefficients can be computed as well-defined sums over Feynman diagrams.
Wilson loops and colored Jones polynomial
The most important observables in Chern-Simons theory are the Wilson loops: given an oriented loop K in M, and a representation R of G, they are defined as where P is the ordering operation along the loop. L can be considered as a knot drawn in M, and in fact the Wilson loop is a partition function for the knot complement M \ K, where the classical solutions are now flat connections on M\K with a meridian holonomy prescribed by R. To be precise, if ρ = 1 2 α>0 α is the vector of Weyl's constants, Λ R = (λ j ) j is the highest weight associated to R, we must identify the holonomy eigenvalues to e (ρj +λj ) = e uj .
A foundational result is that the Wilson loops define knot invariants. When G = SU(2) and R is the spin N −2 2 representation, which has dimension N and is represented by the Young diagram the Wilson loop is related to the colored Jones polynomial J N (K, q), with identifications: The denominator accounts for the normalization of the Jones polynomial, which is 1 for the unknot in S 3 , denoted . The Wilson loop of the unknot is itself given by:
The volume conjectures
Initially, the Jones polynomial J 2 (K, q) has been defined in [54] and its colored version J N (K, q) in [84], [82], in the context of quantum groups. J N is a Laurent polynomial in q with integer coefficients. The number J N (K, q = e 2iπ/N ) is usually called the Kashaev invariant, and the original volume conjecture is: It was later enhanced by Gukov [50] to include hyperbolic deformations of S 3 \ K, and subleading terms: • δ (α) is an integer computed from cohomology, and (α) 0 (u) is related to the Ray-Singer torsion.
• χ (u) for χ ≥ 1 are the coefficients in the -expansion of a certain holomorphic block Φ (α) (u) for SL 2 (C) Chern-Simons theory on M with boundary condition specified by u.
The statement about the leading order is called the generalized volume conjecture (GVC). The range of validity in u in not obvious, because of resonances and Stokes phenomena, that we attempt to describe in the next paragraph. The Kashaev invariant is retrieved for u = iπ. We first recall two rigorous results about the leading order of the GVC. The first one is due to Garoufalidis and Lê: i.e. the GVC holds with the choice of the abelian component.
The second is due to Murakami, who studied the figure-eight knot starting from a closed formula available in this case.
• When u = iπ or u ∈ 5iπ 6 , 7iπ 6 / ∈ iπQ, the lim is given by the GVC for the geometric component.
• When u = iπP Q with P, Q coprime integers and P/Q ∈ 5/6, 7/6[, the lim is 0 when N → ∞ along multiples of Q, whereas the lim is given by the GVC for the geometric component if N → ∞ avoiding multiples of Q.
We recall that the A-polynomial of the figure-eight knot has two components, one abelian (l−1) and one geometric, which intersect at m 2 = −1 and m 2 = 3± √ 5 2 . One recognizes in the latter a value of u at which a transition between components occur for the GVC to be valid according to Theorem 5.4. The example of the figure-eight is special in two ways. Firstly, its branchpoints are located at m 2 = e ±2iπ/3 and m 2 = 3± √ 5 2 , so two of them coincide with the intersection points. So, we do not see a change of branch within a single component at 3 + √ 52 but actually a transition to the abelian component, and the behavior around the other branchpoints is beyond the range of validity of Theorem 5.4. Secondly, CS a (p) vanishes along the path from p c to p in the geometric component such that: so that Theorem 5.4 is only sensitive to the volume, not to the Chern-Simons part.
5.3Â-polynomial, AJ conjecture and Stokes phenomenon
Garoufalidis and Lê [46] showed that J N (K, q) always satisfy some recurrence relation on N . At the level of the analytic continuation, this turns into the existence of an operator A K ∈ Z[e 2 ∂u , e u , e ] so that: The AJ conjecture [45] states that the limit → 0 of A coincides withwith the A-polynomial of K up to a factor which is a polynomial in e u , i.e. the A-polynomial is the semiclassical spectral curve associated to the difference equation Eqn. 131. It has been proved recently in [62] for hyperbolic knots satisfying some technical assumptions and for which the Apolynomial has only a single irreducible factor apart from (l − 1). If we treated Eqn. 131 like an ODE, the leading asymptotic of the colored Jones when → 0 would be given naively by a WKB analysis, namely: where l and m satisfies lim →0Â (m, l) = 0, and p u is a point on this curve such that ln m(p u ) = u. At the heuristic level, it explains the appearance of the complexified volume in the leading asymptotics of the colored Jones polynomial, by combining the AJ conjecture and Neumann-Zagier results reviewed in § 2.4. Going a step further, we could imagine to introduce a infinite set of times and embed Eqn. 131 (at least perturbatively in the new times) in a system of compatible ODE's, for which we know how to associate quantities satisfying loop equations [6], [12], [5]. Those loop equations have many solutions, and the non-perturbative topological recursion applied to the semiclassical spectral curve provide distinguished solutions as formal asymptotic series in [13]. This naive approach can be seen as a vague intuition why it is sensible to compare objects computed from the topological recursion to the asymptotics of solutions of the A recursion relation, which we attempt to do in § 5.4. Difference equation are of discrete nature, and if we treat it like an ODE we may miss resonance phenomena, which here occur when q is a root of unity. On top of that, we have to take into account the usual Stokes phenomenon, hidden in the specification of the point p u on the semiclassical spectral curve which projects to ln m(p u ) = u. This choice comes in three part: • to which component C (α) of the A-polynomial should p u belong ?
• in which sheet of the covering m 2 : C (α) → C should p u belong ?
• which determination of the logarithms in −1 (u) should be chosen ?
We call the data of such a triple T (α) = (C (α) , p (α) , ln) a determination. Although we can consider the RHS intrinsically as a function of a point p in the the universal covering of the SL 2 (C) character variety (defined component by component), it is a non trivial issue to predict for which determination it can be matched to the asymptotics of the LHS which is a function of u. The transition between different determinations occurs across Stokes curves in the u-complex plane. Although the values of u at which several components intersect, and branchcut structures of the coverings m 2 : C → C represented in the u-plane obviously play a role, we do not know of a unambiguous algorithm which would give, for any knot, the determination corresponding to each domain and the correct pattern of Stokes curves which separate them. For second order differential equation (i.e. for semiclassical spectral curves having a single component, the form y 2 = Pol(x)), the algorithm yielding the Stokes curves is known [8], but it is not obvious to generalize this construction to curves of the form Pol(e x , e y ) = 0 and having several components. The only reliable facts are that, for hyperbolic knots, one has to choose: • for u close to iπ: the determination corresponding to the geometric branch of the geometric component (see § 2.3). We call it the geometric determination.
• and for u nonnegative and close to 0, the determination corresponding to the abelian component, so that −1 (u) ≡ 0.
Main conjectures
Let M is a hyperbolic 3-manifold with 1-cusp, and let us consider an A-spectral curve (C 0 , u, v) coming from an irreducible component of the A-polynomial of M. We would like to consider the asymptotics series constructed from the 2|2-kernel introduced in § 4.1: depending on a choice of basepoint o and a characteristics µ, ν ∈ C g . We identified the formal parameter H to . We recall that ι is the involution (m, l) → (1/m, 1/l) defined on C 0 . Let us recapitulate its properties: • J n.p.TR (p) is defined as a formal asymptotic series: • The leading order is the complexified volume up to a constant: • For any χ ≥ 1, χ (p) is a meromorphic function of p ∈ C 0 , which is either independent of , or is a function of which does not have a power series expansion when → 0. We give in § 5.5 its expression up to χ = 3.
• If 2ς is the order of torsion of the symbol {m, l} in K 2 (C(C 0 )), for any χ ≥ 1, χ (p), seen as a function of , assumes a constant value on the subsequences = iπ k where k is a integer with fixed congruence modulo ς.
Conjecture 5.5
There exists a choice of o and µ, ν such that J n.p.TR (p) is annihilated by the A-operator.
We also attempt to formulate a stronger version of the conjecture to identify this series with the all-order asymptotics of the colored Jones polynomial:
Conjecture 5.6
If M is the complement of a prime 8 hyperbolic knot, with a choice of determination as in the GVC (Conjecture 5.2) and keeping the same notations, we have the all-order asymptotic expansion: for a constant C independent of u, and a prefactor B(u) independent of . In other words, for any χ ≥ 1, the (α) χ (u) of Eqn. 128 coincide with χ (p (α) (u)) up to a constant independent of and u.
First few terms
In the comparison to the colored Jones polynomial, there is always an issue of normalization, which is reflected in the prefactors C and B(u) that we do not attempt to predict. Thus, the definition of 0 is irrelevant here, and we refer to [27] for some discussion on the computation of the constant term 0 (u) in the GVC in terms of algebraic geometry on the A-polynomial.
Comments
We check that For the once punctured torus bundle L 2 R (a knot complement in lens space), we check in § 6.5 up to o( 2 ) that: where J H (u) is a Hikami-type integral associated to L 2 R, and for the RHS, o is chosen at a branchpoint, [µ, ν] is the unique half-integer characteristics with reality properties, and we choose the geometric determination. However, the normalization is now C = 1+ 2 32 +o( 2 ). The free parameters in our conjecture are the basepoint o for computing primitives, and the characteristics [µ, ν] of the theta functions. Notice that different choices of o affects χ (p) in a non trivial way, since it contains products of primitives. We have not found a general rule to specify neither o, nor µ, ν, and the choices might be also subjected to Stokes jumps regarding the identification to asymptotics of the Jones polynomial. In the examples treated in Section 6, the curve has g = 1 and we find natural choices for them. In general, we think that it must chosen among even half-integer characteristics, so 2g+1 g possibilities are left. Recall that, for hyperelliptic curves, they are in bijection with partitions of the 2g + 2 Weierstraß points in two sets of g + 1 elements. For A-spectral curves listed in Appendix B.1 that we found to be hyperelliptic 9 , it turned out that they can be represented after birational transformations with rational coefficients (m, l) → (X, Y ), in the form: where S 1 and S 2 are polynomials with integer coefficients and of the same degree g +1, hence providing a canonical choice of even half-characteristics, for which ϑ µ ν (0) 8 computed by Thomae formula is an integer. This suggests that a deeper study of the SL 2 (C) character variety could entirely fix the appropriate choice of µ, ν.
In such a conjecture, it is natural to identify the Planck constant of Chern-Simons theory with the parameter H of the non-perturbative partition functions of Section 4, since special properties arise on each side when H and assume values of the form iπ/integer. In the framework of Chern-Simons theory, the Wilson line can be thought as a wave function, hence it is natural to compare them to kernels. The 2-kernel ψ [2|2] H (p 1 , p 1 ; p 2 , p 2 ) is symmetric by exchange of (p 1 , p 1 ) with (p 2 , p 2 ), so the right hand side is invariant under the involution ι, which is also a property of the holomorphic blocks. We attempt to motivate 10 further the precise form of the conjecture in Section 7. We shall see that, for torus knots, ψ [2|2] without the power 1/2 appears heuristically in the computation of the colored Jones polynomial. For torus knots, it is known [50, Appendix B] that the Chern-Simons partition function Z SL2(C) coincide with Z SU(2) up to a simple factor. For hyperbolic knots, we rather have Eqn. 122, which incite to identify the holomorphic block with the analytic continuation of Z SL2(C) . This may account for the power 1/2 in Conjecture 5.6.
Examples
From the point of view adopted in this article, the complexity of hyperbolic 3-manifolds with 1-cusp is measured by the complexity of the algebraic curve defined by the geometric component C geom 0 of its A-polynomial: to compute J n.p.TR (p), we need to compute explicitly meromorphic forms (and their primitives) on the curve, as well as values of theta functions and their derivatives. From the tables of A-polynomials of Culler [26], [25], we collected the genus of the A-polynomial components of various knots in Fig. B.1.
The simplest non trivial class of manifolds correspond to those for which C geom 0 is a genus 1 curve, i.e. an elliptic curve. This happens for the geometric components of the figure 8knot and the manifold L 2 R. The theta and theta derivatives values can be computed in a simple and efficient way thanks to the theory of modular forms (Section 6.1).
The next simplest class corresponds to manifolds for which C geom 0 is hyperelliptic. In this case there are uniform expressions for a Bergman kernel in terms of the coordinates m 2 and l, and the theta values are well-known in terms of the coordinates of Weierstrass points. For curves of genus g ≥ 2, in principle, the values of theta derivatives can be related to the theta values via the theory of Siegel modular forms and the work of [10]. The 5 2 knot and the Pretzel (-2,3,7) give rise to A-polynomial with a single component, of genus 2 thus hyperelliptic. We leave to a future work explicit computations for A-spectral curves of genus 2 and comparison to the perturbative invariants obtained by other methods.
We observe many times that some components of the A-polynomial of different knots are either the same, or birationally equivalent. For instance, the A-polynomial of the 5 2 and the Pretzel(−2, 3, 7) are birationally equivalent, and one of the two factors of the A-polynomial of the 7 4 coincide with the A-polynomial of the 4 1 . This remark has some interest because values of theta derivatives, which provide the corrective terms to be added to the topological recursion for comparison with the asymptotics of the colored Jones polynomial, only depend on the isomorphism class of C 0 as a Riemann surface, i.e. only depend on A(m, l) up to birational equivalence.
Since C is a singular curve, we do not expect a naive inequality between the degree of A or of the invariant trace field (which contains the cusp field), and the genus g. We observe that: • g looks experimentally much lower than the genus of a generic smooth curve of the same degree as the A polynomial.
• the genus of the quotient C/ι also drops compared to the genus of C.
It would be interesting to have a interpretation of g as well as those two observations from the point of view of representation theory. In the same vein, we can ask other open questions, here focused on geometric components: Problem 6.1 Describe the set of elliptic curves over Q which are obtained as geometric components of hyperbolic 3-manifolds. Do all elliptic curves arise in that way ?
Problem 6.2 Characterize the hyperbolic 3-manifolds so that the quotient C geom /ι has genus 0. Problem 6.3 For a given genus g, do an infinite number of non-isomorphic curves of genus g arise as geometric components of a hyperbolic 3-manifold ?
These problems are already interesting in one replaces "geometric component" by the class of "A-spectral curves" defined in Eqn 2.8.
Thetanullwerten for elliptic curves
In this section, we give a self-contained presentation to compute the theta functions and their derivatives appearing in Section 4 and the computation of J n.p.TR (p) for a genus 1 spectral curve. For more details about elliptic modular forms, the reader may consult the recent textbook [18, Chapter 1].
Modular forms and their derivatives
Elliptic curves are characterized by the orbit of their period τ in the upper-half plane H under the modular group SL 2 (Z). A modular form of weight k for a subgroup Γ ⊆ SL 2 (Z) is a by definition a holomorphic function f : H → C such that f (τ ) = O(1) when q = e 2iπτ → 0, and satisfying: When the subgroup is not precised, it is understood that Eqn. 144 holds for the full modular group. Obviously, modular forms are 1-periodic functions, so have a Fourier expansion: where only nonnegative indices appear owing to the growth condition when q → 0. The Eisenstein series provide important examples of modular forms of weight 2l when l ≥ 2. The zeta value in the denominator enforces the normalization E 2l (τ ) = 1 + O(q) when q → 0. We find convenient to absorb a factor of π per unit weight, and introduce non standard notations E 2l = π 2l E 2l . It is well-known that the ring of modular forms is generated byẼ 4 andẼ 6 . Thus, identities between modular forms of a given weight can be proved by checking that only a finite number of their Fourier coefficients match.Ẽ 2 fails to be modular, indeed one can show: Let us define a differentation operator with an accurate normalization for our purposes: Obviously, derivatives of modular forms are not modular forms. If f is a modular form of weight k for some subgroup Γ, we rather have: This behavior is captured by the notion of "quasi-modular forms" and its relation with "non-holomorphic modular forms" [18, Chapter 1]. We adopt however a more pedestrian way. It is easy to check that the combination: is modular of weight k + 2. d k f is called the Serre derivative of f . Consequently, the differential closure of the ring of modular forms is generated by E 2 , E 4 and E 6 . The basic relations in the new ring are: Since the vector spaces of modular forms of weight 4, 6 and 8 are 1-dimensional, these relations can be proved by checking from Eqn. 147 that d 1Ẽ2 is modular of weight 4, hence of the form c 4Ẽ4 , and similarly d 4Ẽ4 = c 4Ẽ6 and d 6Ẽ6 = c 6Ẽ6 . Then, one finds c 2l by matching the constant Fourier coefficients of the two sides.
Theta functions and their derivatives
In genus 1 there are 3 even characteristics 1 2 , 0 and τ 2 . The corresponding theta values are: and they satisfy the relation: ϑ 4 2 + ϑ 4 4 = ϑ 4 3 . The ϑ i are modular forms of weight 1/2, but only for a congruence subgroup Γ(2) of SL 2 (Z) (this is related to the shift of argument and the eight root of unity in Eqn. 61). Their fourth powers build a vector modular form of weight 2: It is possible to build out of ϑ 4 i expressions which are modular forms, resulting in relations to Eisenstein series upon checking a few Fourier coefficients. As before, we prefer to work withθ i = π 1/2 ϑ i , and we obtain: Combining Eqns. 156-157 to Eqns. 151-153, we obtain after some algebra the basic relations in the differential ring generated by theθ i : From there follows the computation of the d-th derivative ofθ i to all orders, and we observe especially that : where P d is a polynomial with integer coefficients. By Nesterenko's theorem,Ẽ 2 cannot be an algebraic number. So it might seem hopeless to obtain any explicit number to compute the kernels, e.g. Eqn. 103. But we explained in § 4.3 how the kernels could be computed in terms of combinations of derivatives which were modular. We immediately see that the appropriate combination must be equal to 3 −d P d (θ 4 2 ,θ 4 4 , 0). This coincides the definition of the d th order Serre derivative 11 : In the following, we focus on the computation of the T 2d;i .
Application to elliptic curves
In this paragraph we consider a curve C defined by an equation Pol(m 2 , l) = 0 with integer coefficients, whose smooth model C 0 is a Riemann surface of genus 1. Alternatively, there exists x, y ∈ Q(m 2 , l) such that the defining equation of C 0 can be brought in Weierstraß form: and g 2 , g 3 are called elliptic invariants. Up to a multiplicative constant, the unique holomorphic 1-form on C 0 is dz = dx y . We assume we have chosen A and B cycles on the curve, it is not necessary to be precise about this choice as we will see in a moment. If we denote 2 A = A dz and 2 B = B dz, the holomorphic 1-form normalized on the A-cycle is da = dz 2 A and the period is τ = B A . The curve C 0 is isomorphic to C/(Z ⊕ τ Z), and we can uniformize Eqn. 163 by where ℘ is the Weierstraß function: (165) 11 More precisely, d k is −8π 2 times the Serre derivative in the notations of [18].
Let us recall the main properties of ℘(w|τ ). It is an even periodic function with periods 1 and τ , which has a double pole with coefficient 1 and first subleading order O(w 2 ). Its full asymptotic expansion when w → 0 is: where B 2j are the Bernoulli numbers. The values of the Eisenstein series for C 0 can be expressed in terms of g 2 and g 3 , by a comparison of the expansion of the left and right hand side of Eqn. 163 when z → 0: The equations Eqns. 156-157 allow in principle the determination of: Yet To summarize, any solution of Eqns. 156-157 will give us the fourth powers of the theta values, maybe in desorder and with the wrong sign. But the sign does not matter to compute D dθ i /θ i from Eqns. 158-160, so the choice of another solution just results in a permutation of i = 2, 3, 4, i.e. of the label of the even characteristics.
Arithmetic aspects
The modular discriminant ∆(τ ) = e 2iπτ ∞ n=1 (1 − e 2iπnτ ) 24 is another important modular form, of weight 12. In terms of Eisenstein series: Equivalently, we find its value from Eqns. 156-157 or 167: If we assume g 2 and g 3 rational (and this is so when C 0 comes from an A-polynomial), we learn from Eqns. 156-157 thatθ 4 i /(2 A ) 2 for i = 2, 3, 4 are algebraic number. Even more, the reality of g 2 and g 3 imply that the complex conjugates (t * 2 , t * 3 , t * 4 ) must be in the list of solutions 169, and looking case by case we infer that one of the numbersθ 4 i /(2 A ) 4 is real (if ∆ > 0) or pure imaginary (if ∆ < 0), while the two others are complex conjugates up to a sign. When ∆ < 0, we also have a privileged choice of even characteristics, namely the one for whichθ 4 i0 /2 2 A is purely imaginary. This would remain true if τ was slightly changed by addition of an imaginary part (theẼ 2j (τ ) would remain real). We deduce that the d th order Serre derivatives T 2d;i0 /(2 A ) 2d are real, algebraic numbers. This last statement is also true for all T 2d;i /(2 A ) 2d when ∆ > 0, because the three numbersθ A set of elliptic invariant is g 2 = 8 3 and g 3 = − 1 27 , and the discriminant is ∆ = −19. We find that T 2d;i0 ∈ Q(α) with: 11A3: A set of elliptic invariants is g 2 = − 4 3 and g 3 = 19 27 , and the discriminant is ∆ = −11. We find that T 2d;i0 ∈ Q(α) with: 43A1: y 2 + y = x 3 + x 2 (10 139 ) A set of elliptic invariants is g 2 = − 4 3 and g 3 = 35 27 , and the discriminant is ∆ = −43. We find that T 2d;i0 ∈ Q(α), with: In the two first examples (figure-eight knot and L 2 R, the fact that T 2d;i0 are rational numbers imply that the coefficients χ (p) (for the choice of the characteristics associated to i 0 ) for χ ≥ 1 sit in the same function field as the amplitudes of the topological recursion. On the contrary, for the three last examples, they will sit a priori in an extension by the element α of the function field where the amplitudes of the topological recursion live.
Remark 6.4 The two first elliptic curves do not have complex multiplication, whereas the three last do. We thank Farshid Hajir for pointing us this property.
Bergman kernel for elliptic curves
In genus 1, there is only one odd characteristics, and the corresponding theta function is: It satisfies: The Bergman kernel normalized on the A-cycle is: On the other hand, the Weierstraß function provides another natural Bergman kernel: By "natural", we mean that it can be written in terms of the coordinates x and y (see Eqn. 163) thanks to the addition relation for ℘. The result is: There is a well-known relation between ℘ and ϑ 1 : In other words, Eqn. 179 is the expression for a Bergman kernel normalized on the A κ(τ )cycle, with the value:
Application to some degree 2, elliptic A-spectral curves
We assume in this paragraph that the spectral curve (C, ln l, ln m) has a defining equation: where P 1 , P 2 , R are polynomials and S is a polynomial of degree 4 with simple roots and leading coefficient 1. Also, it admits a smooth model C 0 of equation: We also assume that {m, l} is 2-torsion (i.e. ς = 1) and that ι * = −id. The spectral curve for the figure-eight knot and L 2 R takes this form. Then, many simplifications occur.
Writing the n-forms
First, the ramification points a i of the spectral curve coincide with the Weierstraß points of C 0 , i.e. with the roots of S. Besides, the local involution correspond to changing the sign of the squareroot S(X) → − S(X), and is in fact defined globally on C 0 . We define a variable z by integration of the holomorphic 1-form: Afterwards, it is straightforward to check the formula: In particular, we find: which intervenes in the computation of ω 1 1 . We also compute: To avoid cumbersome notations, we use the same letter a i to denote the image of the ramification point a i in the Jacobian of C 0 (on the left hand side) and the value of the X-coordinate at the same ramification point. The second term in Eqn. 188 is odd when we change the sign of the squareroot, so disappear when we integrate from z to z: Besides, one can check by differentiating the two sides of the equality: where C i only depends on the basepoint o. This can be checked by differentiation and the fact that both sides of the equality are invariant under ι. If one denotes {a i , ι(a i ), a i , ι(a i )} the set of ramification points, one finds that C i = 0 when the basepoint o is chosen as a i or ι(a i ). These formulas allow to complete the computation of the ω h n|κ(τ ) , i.e. the topological recursion with the Bergman kernel B κ(τ ) defined in Eqn. 179. We just need to compute expansion of the quantities in Eqns. 189 at the branchpoints X → a j , and then take residues. The coordinate z is convenient for these computations, because Taylor expanding then amounts to differentiating the Weierstraß function with respect its argument. This method yields ω h n|κ(τ ) as a linear combination with rational coefficients of elementary n-forms of the type: where p j are even integers. It is easy to integrate ω h n|κ(τ ) over cycles with such a representation (we have to use 189 for terms with some p j = 0). In particular, integrating z j over 1 2iπ B κ(τ ) in Eqn. 191 gives 0 if p j > 0, and replaces the j th -factor by (2 A ) −1 if p j = 0. ι * has a single eigenvalued, which we assumed to be −1. The discussion of § 4.4 applies and can be explicitly checked: we find that the amplitudes: vanish if d is odd. Then, the computation of T 2d;i for d integer as detailed in § 6.1 is all we need to evaluate J n.p.TR (p).
Arithmetic aspects
Since S is palindromic, we can write the results in a more compact form with a variable w = m 2 +m −2 2 . We also denote: Notice that w is a uniformization variable for the quotient C 0 /ι which has genus 0 by assumption. The amplitudes G dzj 2 A appears in ω h n (z 1 , . . . , z n ). At level 1, we have: It is easy to construct recursively a set Q h n ⊇ P h n from the residue formula Eqn. 75. At level one, we define Q 1 1 = P 1 1 and Q 0 3 = P 0 3 , and if we know all Q h n at level χ, we use the following rules to define Q h n at level χ + 1: • if (p, p , p 2 , . . . , p n ) ∈ Q h−1 n+1 , then (p + p + 2, p 2 , . . . , p n ) ∈ Q h n . • for all (n , h ) = (1, 0), (n+1, h) such that 0 ≤ h ≤ h and 0 ≤ n ≤ n, if (p, p 2 , . . . , p n +1 ) ∈ Q h n +1 and (p , p n +2 , . . . , p n ) ∈ Q h−h n−n , then (p + p + 2, p 2 , . . . , p n ) ∈ Q h n .
We already know r 0,(0) 3 = 3, r 0,(2) 3 = 1, r 1,(0) 1 = 3, and by recursion one can show: Thus, our construction naturally entails: At the points such that m 2 = 1 (in particular the reference point where the hyperbolic metric is complete), we have w = 1. From the definition, we see that the cusp field is F = Q[ σ(1)], and our construction naturally entails: In the examples below, we need C i ≡ 0, so one can forget about C i in Eqn. 198 and Eqn. 199. In this case, our Conjecture 5.6 predicts that the coefficients of the asymptotic expansion of the Kashaev invariant belong to the cusp field F.
4 1 (figure-eight knot)
Apart from the abelian factor (l − 1), the A-polynomial has a unique factor (necessarily the geometric one): It defines a curve C 0 of genus 1. The symbol {m, l} is 2-torsion, and ι * = −id. The spectral curve can be put in the form of Eqn. 182: so the results of § 6.3 can be applied, and we introduce: The curve has 4 ramification points, of coordinates The local involution z → z is defined globally on C 0 , and it corresponds to (m, l) → (m, 1/l). Incidentally, it coincides with the amphichiral symmetry. The cusp field is Q[ √ −3]. It is known [65] that the hyperbolic volume of M u = S 3 \ 4 1 with cusp angle 2 Im u and Re u = 0, is: where β(u) = arccos ch(2u) − 1/2 and Λ is the Lobachevsky function: In particular, it vanishes when u = ±2iπ/3, and this value coincide with the u-projection of two of the four branchpoints. Hence, it we denote by a 0 any of these points, we find that Im
Amplitudes
We now derive the three first terms of J n.p.TR . We choose to compute the primitive with C i ≡ 0. We computed the ω h n up to level 3 (i.e. 2h − 2 + n ≤ 3). We just present here the expression for the non-vanishing amplitudes G h,(d) for level 1 and 2 coincide 12 with the amplitudes computed in [28, § 3.3].
(207) 12 More precisely, since [28] uses the spectral curve 2v du (instead of v du here), we retrieve their amplitudes F (h,n) by multiplying our results by (1/2) First orders of J n.p.TR /comparison to colored Jones We choose the even-half characteristics [µ, ν] leading to real-valued theta derivatives (the last column of the table for the curve labeled 15A8 in § 6.1.5 is selected). The reader may recognize in those values the ad hoc renormalizations of the constants G k found by the authors of [28]. We use the general expressions given in § 5.5 to compute: . These authors as well as [28] also find the same coefficients in the asymptotic expansion of a Hikami-type integral J H (u) associated to the figure-eight knot. This is believed to be the correct asymptotic expansion for the colored Jones polynomial in the GVC near iπ.
Specialization to u = iπ/comparison to Kashaev invariant
We recall that the complete hyperbolic point correspond to w = 1. The coefficients χ (w = 1) belong to Q[ √ −3]. We find for the first orders: This is in agreement with the asymptotic expansion of the Kashaev invariant, proved with help of numerics [31] to be: This manifold is hyperbolic [79], [49]. Apart from the abelian factor (l−1), the A-polynomial has a unique factor (necessarily the geometric one): A(m, l) = l 2 m 4 + l(−m 6 + 2m 4 + 2m 2 − 1) + m 2 .
It defines a curve of genus 1. The symbol {m, l} is 2-torsion, and ι * = −id. The spectral curve can be put in the form Eqn. 182: so the results of § 6.3 can be applied, and we introduce: The curve has 4 ramification points, of coordinates: The local involution z → z is defined globally of C 0 . The cusp field is Q[ √ −7].
Amplitudes
We again choose to compute all primitives with C i ≡ 0.
First orders of J n.p.TR /comparison to colored Jones We choose the even-half characteristics [µ, ν] leading to real-valued theta derivatives (the last column of the table for the curve labeled 14A4 in § 6.1.5 is selected).
This is in agreement with the results of [28], and we recognize again their ad hoc renormalizations in the values of theta derivatives. These authors have computed the asymptotic expansion of a Hikami type integral J H (u) associated to L 2 R: where ± indicates the dependence of the integration contour. We have:
Heuristics imported from torus knots
Let (P, Q) be coprime integers. The A-polynomial of the torus knot K P,Q contains a nonabelian component of the form A(m, l) = lm P Q + 1. Since the corresponding spectral curve does not have branchpoints, its partition function and kernels are ill-defined, so our conjecture for the Jones polynomial cannot be correct as such for torus knots for the nonabelian branch. Nevertheless, we shall see heuristically how the shape of our conjecture for any Wilson line arises in the case of torus knots. It is only at the end of this derivation, when we specialize to the Jones polynomial, that one discovers that the A-polynomial should be replaced by a blow-up of one of its deformation for the conjecture to be meaningful.
General case
Thanks to toric symmetry, the Wilson loops of K P,Q can be computed by localization [85], [60], [4], [55], and the sum over flat connections on S 3 \ K P,Q can be written as a matrix-like integral: where α > 0 are the positive roots of G, χ R is the character of the representation R, and the normalization constant Z P,Q is: This can be written even more explicitly, using Weyl's formula for the characters: where ρ = 1 2 α>0 α is the vector of Weyl's constants, Λ R is the highest weight associated to R.
SU(n) case
For SU (n), the positive roots are α i,j = e i −e j with i < j and where e i = (0, . . . , 0, 1, 0, . . . , 0) with 1 in the i th position, and ρ = n i=1 ( n+1 2 −i)e i . The Weyl group is the symmetric group S n . Irreducible representations R are in correspondence with Young diagrams with n rows, or partitions λ = (λ 1 ≥ λ 2 ≥ · · · ≥ λ n ≥ 0), and we have Λ R = (λ 1 , . . . , λ n ). The character associated to the representation indexed by λ is the Schur polynomial: where H i = λ i − i + c, c is an arbitrary constant. From Harish-Chandra formula, we also have: where dU is the Haar measure on U(n) with total mass 1, and ∆(X) = 1≤i<j≤n (X i − X j ) is the Vandermonde determinant. Thus, Eqn. 222 gives: W SU(n),R (K P,Q , ) = ∆(H) (P Q) n(n−1)/2 Z P,Q R n ×U(n) dX dU ∆(X) where we have defined the potential V(X) as: Notice that V(X) is invariant under translation of X by a matrix proportional to the identity matrix 1 n . We may include the factors i = j since they are equal to 1, and rewrite: where B l the l th Bernoulli number. To get rid of the linear term in the exponential in Eqn. 228, we shift: The contour of integration R n is shifted to C n R,H accordingly: We define the normal matrix M = U XU † , and the invariant measure on the space H n (C ,R ) of normal matrices with eigenvalues on the contour C R, is: Therefore: The Vandermonde of the H's is related to the dimension of the representation: and the trace of H is related to the number of boxes in λ: So far, the constant c was arbitrary, in particular it can depend on λ. The choice: allows to have Tr H = 0, and we now stick to it. The normalization constant Z P,Q is computed with the same steps for the trivial representation R ∅ of SU(n). Thus: where the multiplicative constant is given by: The integral in the numerator is similar to that in the denominator, except for the external field H (resp. H ∅ ) encoding the highest weight associated to R (resp. to the trivial representation). We also shifted the contour from R to C R, = R + 2 P Q n |λ|. Since the discussion to come remains at a formal level, we shall move it back to R, and assume the range of integration to be the space of hermitian matrices H n (R) in Eqn. 239.
Principle
For matrix integrals (with or without external potential) of the form Eqn. 234, we have [39] where S n, = (C, x, y) is the spectral curve of the matrix integral, which in general depend on and the size of the matrix n, and T is the non-perturbative partition function defined in Section 4. We have define somewhat arbitrarily 13 Eqn. 241 means that the asymptotic expansion of the left hand side is given by the right hand side (which we defined as a formal asymptotic series). Adding an external field in the form e Tr H M amounts (see for instance [78]) to modify the spectral curve by addition of simple poles p 1 , . . . , p n ∈ C to x with residue H with respect to dy, and such that y(p j ) = H j , and some other simple poles o 1 , . . . , o n ∈ C with residue −H.
Similarly, we denote p ∅ j the poles associated to the external field H ∅ , i.e. we have y(p ∅ j ) = − i + n+1 2 . We thus find: Now, the simple poles are added to the second function in the spectral curve, and we recognize the n|n-kernel described in § 4.1, for the spectral curveŜ n, = (C, y, x) for some basepoint o:
SU(2) case: Jones polynomial
For SU(n = 2) and the representation R associated to (λ 1 , λ 2 ) = (N − 1, 0), we retrieve the colored Jones polynomial (see Eqn. 125). It is thus computed from the 2-kernel with points p 1 and p 2 of projections: and for the trivial representation: This leads to: .
and we insist that the kernels are computed for the spectral curve of the matrix integral after exchange of x and y.
Spectral curve for the torus knots
Let P , Q be integers such that P Q − Q P = 1. The spectral curve S n, = (C, x, y) of the matrix integral in Eqn. 223 was derived in [17], in the regime when n is of order 1: S : e −(P +Q)y − e −nH e −(P x+Qy) − e −(Q x+P y) + e nH e −(P +Q )x = 0.
This curve can be uniformized with a variable z ∈ C ≡ C: hence its genus is 0, and there is no theta function in the definition of its partition function and kernels. It was argued in [17] that the topological recursion for this curve reproduces the torus knots invariants.
Here, we are interested in the regime where is small and n is fixed (and in particular n = 2). If we keep x and y of order 1, the curve is trivial: in the sense that it does not have ramification points. So, the partition function T of this spectral curve is ill-defined. Actually, when the limit spectral curve is ill-defined, the information about the unstable terms (i.e. the terms decaying with → 0) are actually obtained contained in the blow-up of S n, at its basepoint o. It is realized by setting x = √ 2n x and y = √ 2n ỹ withx andỹ of order 1, and retaining the first non trivial order in Eqn. 249 when n → 0, we find: The formulae 241 and 245 are expected to be correct if applied to the spectral curve of Eqn. 252 at least for the terms of order o(1) when → 0. The non decaying terms are rather given by the limit spectral curve itself, and thus are trivial. This is in agreement with the fact [58] that there is no exponential growth of the Jones polynomial of torus knots (they are not hyperbolic), in other words −1 ≡ 0.
General mechanism
We expect that the mechanism described for torus knots complements is general (see for instance the conjecture in [60, Section 6]), and our proposal should essentially compute Wilson lines for more general 3-manifold M, provided the spectral curve is well-chosen. A scenario would be that: where V consists of eigenvalues of the holonomy operator P exp ∂M A, and S(V) the effective action for V after integrating out the other degrees of freedom against the Chern-Simons action. If a formula like 253 holds, the steps of § 7.1.2 can be repeated to find: where M = U X U † , U ∈ SU(n) and H i = λ i −i+c for some constant c. Then, if the integral dM e S(M ) is a Tau function with spectral curve S n,H = (C, x, y), we would find again: where the n|n-kernel is computed for the spectral curveŜ n, = (C, y, x) after the exchange (x ↔ y). For hyperbolic manifolds, one expects to find as spectral curve a deformation of the A-polynomial (or at least of subcomponents of it), which reduces to the A-polynomial in the SU(n = 2) case, i.e. to m ∝ e 2y , l ∝ e x and A(m, l) = 0. This is plausible because it is known that the A-polynomial can be obtained as the saddle point equation obtained by elimination from the Neumann-Zagier potential [53]. In our argument, we see then from Eqn. 246 that the points p 1 and p 2 needed to compute the Jones polynomial have m-projection: ln m(p 1 ) = N + cte ln m(p 2 ) = −N + cte, and one recognizes 15 (up to constant shift) the identification between the hyperbolic structure parameter m = e u and the quantum group parameter q = e 2 and N appearing in Eqn. 125.
Perspectives
We have constructed a formal asymptotic series J n.p.TR (p) depending on a point on the SL 2 (C) character variety of a hyperbolic 3-manifolds with 1-cusp, which has interesting properties per se. It depends on a choice of characteristics µ, ν ∈ C g (which might be restricted to even-half characteristics) and basepoint o for the computation of iterated primitives. Provided a accurate choice is made for those data, we have conjectured that it computes the asymptotic expansion of the colored Jones polynomial, discarding roots of unity. We have made a non-trivial check to first orders for the figure-eight knot. A weaker conjecture is that our series is a formal solution of the A-hat recursion relation. We made a closely related check to first orders for the L 2 R. We think that working on the A-hat recursion relation satisfied by the colored Jones polynomial is a good approach in an attempt to prove our conjecture (or a slight modification of it). The intuition behind our construction comes from the theory of integrable systems and its relations to loop equations. J n.p.TR (p) was defined formally by introducing an infinite number of infinitesimal deformations of the A-polynomial curve, and one may wonder if this can be interpreted as an integrable perturbation of the Wess-Zumino-Witten CFT.
The main interest of our proposal is rather structural than computational. Although we do have an algorithm, it requires the use of a basis of meromorphic forms which behaves well under integration, and the computation of theta functions and derivatives, so is less efficient than other methods. Yet, the expression in terms of the topological recursion suggest possible connections between knot theory and respectively integrable systems [41], [13], intersection theory on the moduli space [37] and enumerative geometry, which deserve further investigation.
This conjecture gives also a framework for the study of the arithmetic properties of perturbative knot invariants. It would particularly interesting to compare our predictions for the expansion of the Kashaev invariant (i.e. at the complete hyperbolic point) to those of [30] obtained by a gluing procedure.
When the quotient of (a component of) the SL 2 (C) character variety by the involution ι(m, l) = (1/m, 1/l) is a genus g ι = 0 curve, J n.p.T R is a formal power series, although it takes into account non-perturbative effects. When this property does not hold, it is rather an asymptotic series which contains fast oscillations to all orders when → 0. However, the K-theoretical properties of the A-polynomial imply that, if we specialize to sequences iπ/k with k an integer of fixed congruence, we retrieve a formal power series. It would be interesting to know if this wild behavior can be seen in the asymptotics of the colored Jones polynomial for knots such that g ι = 0. The simplest examples of this kind we know are 8 21 and the Pretzel(−2, 3,9), and are currently under investigation. Actually, earlier experiments on the asymptotics of the colored Jones have been performed to our knowledge only for knots with g ι = 0. If Conjecture 5.6 contains some part of truth, new phenomena may be discovered. Else, one would need to understand how it should be modify to preserve the matching for 4 1 and L 2 R.
Our conjectures could be generalized in several directions.
• One may wonder if the n|n kernels can be identified to asymptotics of other relevant knot invariants. Notice also that Hirota equations imply determinantal formulas [13]: where ∝ means equality up to a factor involving prime forms. A naive guess, inspired by Section 7, would be to compare ψ [n|n] to Wilson lines in a Chern-Simons theory with SU(n|n) gauge group in the limit of large representations. The rescaled size of the representation would be in correspondence (see § 5.1 in the case of SU(n) with points p 1 , . . . , p 2n on the character variety.
• One may wish to study the asymptotics of the colored Jones when q → ζ d (a root of unity, instead of q → 1 here). It is natural to propose a conjecture similar to 5.6, with the curve of equation lim q→ζ d (e u , e v , q) replacing the A-polynomial.
• One may wish to study the a-(or Q-) deformation of the knot invariants, considered recently in [2]. In the regime when q → 1 but keeping a finite, this amounts to study asymptotics for gauge groups of large rank. We guess that the non-topological recursion for the a-deformed A-polynomial will come into play.
• And, at the top of the hierarchy, one may consider the categorified knot invariants, which results from another deformation with a variable t [33]. These knot invariants can be seen as generating series of BPS invariants. They are conjectured to be annhilated by an operator A(e u , e ∂u , a, q, t), called the super-A-hat polynomial, which is explicitly known in a few examples [43], [44]. In this case, although a t-deformed spectral curve can be defined, we think that a "deformed topological recursion" should be used in order to compute something meaningful about their asymptotics. This intuition is based on the fact that Schur polynomial have to be replaced by Macdonald polynomials under this deformation, and ongoing work suggests that the analysis of the matrix model of Eqn. 222 requires a deformation of the topological recursion.
Although the identification for the colored Jones polynomial (N = ln m(p u )) was rather simple, appropriate and non-trivial "mirror maps" (like in topological strings [1], [15]) could be necessary to make any of those generalizations effective. In yet another direction, the (generalized) volume conjecture can also be formulated for links with L components. The SL n (C) character variety has local complex dimension (n − 1)L at a generic point [66]. It is a challenging problem to reduce -if only possible -the asymptotics of 3-manifolds with L cusps to algebraic geometry on this variety.
A Diagrammatic representation for the non-perturbative topological recursion
A.1 Non-perturbative partition function
The non-perturbative Tau function T H was defined in Eqn. 99. We had: with:T We also recall that, owing to special geometry, the k th derivative of ω h n with respect to filling fractions is: For 2 − 2h − n < 0, we represent ω h n by a surface with h handles and n legs, and we represent ∇ ⊗k ϑ/ϑ by a black vertex with k legs.
Then, with those diagrammatic notations, Eqn. 259 is represented as a sum of graphs. Each graph has exactly one black vertex, whose legs are attached to the legs of a product of ω h n 's, such that all legs are paired.
where χ Euler is sum of the Euler characteristics of all punctured surfaces of the graph (each of them having a negative Euler characteristics), and # Aut ∈ N * is the symmetry factor of the graph.
A.2 Logarithm of the non-perturbative partition function
Notice that the generating function for the derivatives of ln ϑ, is related to the generating function for the derivatives of ϑ, by keeping the cumulants. If we represent ∇ ⊗k ln ϑ by a white vertex with k legs, we have that the black vertex is the sum of all possible products of white vertices having the same legs.
and thus
A.4 Perturbative knot invariants to first orders
The central object in our conjecture concerning the asymptotics of the colored Jones was: and from Eqn. 271 they acquire a diagrammatic representation. When ι * = −id, ω h,(d) n vanish whenever d is odd, so the only graphs with non-zero weight are those where each surface is contracted with an even number of legs incident to a white vertex. We give below the two first orders in diagrams in this case.
B.3 Properties
We present properties of spectral curves for various knots. Each block collect equivalent curves modulo birational transformations. Notice that l → C l ± m a implies ω h n → (±) n ω h n . The column g gives the genus of the curve, and the column g ι gives the genus of the quotient curve C/ι, i.e. the number of +1 eigenvalues of ι * . In the column H, we indicate if ι coincide or not with the hyperelliptic involution. If this is the case, we necessarily have ι * = −id. |{a}| indicates the number of ramification points. They are all simple, except when we indicate with a superscript +1 the presence of one extra ramification point of order 3 at (m, l) = (−1, 1). We then indicate the minimal positive integer ς such that 2ς · {m, l} = 0 ∈ in K 2 (C). When the knot is amphichiral and when the component is stable under α(m, l) = (1/m, l), we indicate the number of +1 eigenvalues of the induced map α * in homology. We put question marks when we could not obtain the answer in a reasonable time with maple. | 2012-09-13T13:25:57.000Z | 2012-05-10T00:00:00.000 | {
"year": 2012,
"sha1": "06d1e927cce14057166f999d9b028019762ab4ec",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1205.2261",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "06d1e927cce14057166f999d9b028019762ab4ec",
"s2fieldsofstudy": [
"Mathematics"
],
"extfieldsofstudy": [
"Mathematics",
"Physics"
]
} |
254929149 | pes2o/s2orc | v3-fos-license | Assessing the therapeutic impact of resveratrol in ALS SOD1-G93A mice with electrical impedance myography
To aid in the identification of new treatments for amyotrophic lateral sclerosis (ALS), convenient biomarkers are needed to effectively and uniformly measure drug efficacy. To this end, we assessed the effects of the nutraceutical resveratrol (RSV) on disease onset and overall survival in SOD1-G93A (ALS) mice and compared several standard biomarkers including body mass, motor score (MS), paw grip endurance (PGE), and compound motor action potential (CMAP) amplitude, with the technique of electrical impedance myography (EIM) to follow disease progression. Eighteen ALS mice (nine females, nine males) received RSV in the chow (dose: 120 mg/kg/day) starting at 8 weeks of age; 19 ALS mice (nine females, 10 males) received normal chow; 10 wild type (WT) littermates (five females, five males) fed standard chow served as controls. Biomarker assessments were performed weekly beginning at 8 weeks. No differences in either disease onset or overall survival were found between RSV-treated and untreated ALS mice of either sex; moreover, all biomarkers failed to identify any beneficial effect of RSV when administered at this dose. Therefore, for the comparative evaluation of the ability of the various biomarkers to detect the earliest symptoms of disease, data from all animals (i.e., RSV-treated and untreated ALS mice of both sexes) were combined. Of the biomarkers tested, EIM impedance values, i.e., surface EIM longitudinal phase at 50 kHz (LP 50 kHz), and CMAP amplitude showed the earliest significant changes from baseline. LP 50 kHz values showed a rate of decline equivalent to that of CMAP amplitude and correlated with both PGE and CMAP amplitude [Spearman rho = 0.806 (p = 0.004) and 0.627 (p = 0.044), respectively]. Consistent with previous work, these findings indicate that surface EIM can serve as an effective non-invasive biomarker for preclinical drug testing in rodent models of ALS.
To aid in the identification of new treatments for amyotrophic lateral sclerosis (ALS), convenient biomarkers are needed to e ectively and uniformly measure drug e cacy. To this end, we assessed the e ects of the nutraceutical resveratrol (RSV) on disease onset and overall survival in SOD -G A (ALS) mice and compared several standard biomarkers including body mass, motor score (MS), paw grip endurance (PGE), and compound motor action potential (CMAP) amplitude, with the technique of electrical impedance myography (EIM) to follow disease progression. Eighteen ALS mice (nine females, nine males) received RSV in the chow (dose: mg/kg/day) starting at weeks of age; ALS mice (nine females, males) received normal chow; wild type (WT) littermates (five females, five males) fed standard chow served as controls. Biomarker assessments were performed weekly beginning at weeks. No di erences in either disease onset or overall survival were found between RSV-treated and untreated ALS mice of either sex; moreover, all biomarkers failed to identify any beneficial e ect of RSV when administered at this dose. Therefore, for the comparative evaluation of the ability of the various biomarkers to detect the earliest symptoms of disease, data from all animals (i.e., RSV-treated and untreated ALS mice of both sexes) were combined. Of the biomarkers tested, EIM impedance values, i.e., surface EIM longitudinal phase at kHz (LP kHz), and CMAP amplitude showed the earliest significant changes from baseline. LP kHz values showed a rate of decline equivalent to that of CMAP amplitude and correlated with both PGE and CMAP amplitude [Spearman rho = .
(p = . ), respectively]. Consistent with previous work, these findings indicate that surface EIM can serve as an e ective non-invasive biomarker for preclinical drug testing in rodent models of ALS. KEYWORDS electrical impedance myography, resveratrol, amyotrophic lateral sclerosis, mouse, therapy Introduction Amyotrophic lateral sclerosis (ALS), is a progressive neurodegenerative disease that affects nerve cells in the brain and the spinal cord leading to loss of upper and lower motor neurons, muscle denervation, loss of voluntary muscle control, muscle atrophy, and eventual death. In order to assess new potential drug therapies, most preclinical studies using ALS rodent models rely on a variety of measures such as body mass, survival, muscle girth, electrophysiological testing, and assorted behavioral parameters including paw grip endurance test (PGE), rotarod, treadmill, open field activity, and/or motor score (MS) (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). To aid in the identification of new treatments for ALS, additional biomarkers are needed to more effectively and uniformly measure disease onset and progression.
Electrical impedance myography (EIM) has shown promise as a biomarker in both human and animal studies of ALS (12,13). In surface EIM, a rigid 4-electrode array is placed on the skin overlying a muscle of interest, and a weak, multi-frequency (1 kHz−1 MHz) electrical current is passed across the two outer electrodes and into the muscle. The two inner electrodes record the impedance to current flow by the muscle fibers (14-17). Alterations in the measured impedance values (i.e., resistance, reactance) and the derived phase values provide data on muscle condition such as myofiber size and the development of fat and connective tissue as occurs in longer standing disease (18)(19)(20).
Human studies have shown that surface EIM longitudinal phase at 50 kHz (LP 50 kHz) is sensitive to decline in ALS patients and the rate of deterioration correlates to length of survival, suggesting that surface EIM could potentially serve as a surrogate outcome measure (12). Moreover, when surface EIM was used to track a single rapidly deteriorating muscle, excellent sensitivity to overall ALS progression was achieved, and 81% fewer subjects would be required to determine clinical benefit as compared to standard measures such as ALSFRS-R (21). Additional advantages of surface EIM include that it is non-invasive, painless, rapid to apply, requires minimal training, is highly reproducible, and can be used to study virtually any superficial muscle (22,23).
In preclinical studies, surface EIM applied to the SOD1-G93A rat model of ALS demonstrated sensitivity to disease progression and a strong relationship between the rate of decline in LP 50 kHz values and length of survival of the animals (13). In a study designed to evaluate drug efficacy, although SOD1-G93A ALS mice treated with riluzole showed no increase in time of disease onset or overall survival (24), there was good correlation between LP 50 kHz values and other biomarkers including paw grip endurance (PGE), and Compound Muscle Action Potential (CMAP) amplitude, supporting the conclusion that changes in surface EIM values reflect motor neuron loss and declining motor function.
Hoping to further assess EIM's sensitivity to drug effect, we undertook an additional study, focusing on the potential therapeutic effect of the nutraceutical resveratrol (3,4 ′ ,5trihydroxystilbene) (RSV). RSV, a polyphenol compound found in the skin of red grapes, has been shown to mimic effects of caloric restriction, exert anti-inflammatory and antioxidative effects, and affect the initiation and progression of different diseases through a variety of mechanisms (25). We recently demonstrated the ability of RSV to mitigate muscle deconditioning in a rat model of reduced mechanical loading (26). RSV has also received considerable attention for its potential neuroprotective effects in some neurodegenerative disorders (27), but there is conflicting information about its value in the treatment of ALS (28). One study reported that a single dose of RSV at 25 mg/kg in the SOD1-G93A ALS mouse model did not improve motor abilities or extend survival in ALS mice (29). A second study used intraperitoneal injections of RSV at a dose of 20 mg/kg/twice a week, which were reported to improve survival and delay disease onset in this model (30). A similar positive effect of RSV on survival and motor function was observed with a higher dose of RSV (160 mg/kg/day) administered orally in ALS mice (31). Based on these previously published papers, we evaluated longitudinally a group of SOD1-G93A ALS mice treated with RSV vs. a group of untreated animals with two goals in mind: (1) to assess the effects of RSV treatment on disease progression and overall survival; and (2) to compare the effectiveness of several biomarkers [i.e., CMAP amplitude, paw grip endurance (PGE), motor score (MS), and surface EIM] to detect disease onset and progression, as well as the potential drug efficacy of RSV in this model. litters, but did not have significant numbers to litter match the treatment vs. the non-treatment groups. In addition, WT littermates (five females, five males) served as controls. Animals were housed in micro-isolator cages in groups of up to five mice/cage in the animal facility equipped with a 12:12 light/dark standard lighting cycle, and were monitored on a daily basis by research staff, evaluating conditions of fur, grooming, the presence of porphyrin staining, mobility and gait, and ability to feed.
Treatment with resveratrol
HPLC-purified trans-resveratrol (RSV), (obtained from Megaresveratrol.net, Candlewood Stars, Inc, Danbury, CT, United States) was incorporated into Purina Formulab Diet 5008 at a concentration of 0.1% by Envigo Teklad Diets, Madison WI. A cohort of untreated ALS mice was fed normal chow (NC), while another cohort of ALS mice was fed RSV incorporated in their chow. Starting at 8 weeks of age, 18 ALS animals (nine females, nine males) were fed with RSV chow ad lib; 19 ALS mice (nine females, 10 males) received Purina Formulab Diet 5008. WT littermates (five females, five males) were fed standard chow and served as controls. All animals in the study were switched to Gel-packs (DietGelH 76A, PharmaSer, Framingham, MA) when the first ALS mice in either treatment group became too weak to reach their chow. These Diet Gel-packs did not contain Resveratrol. All animals were then followed out to the anticipated time of death (∼120-140 days).
Dosage of resveratrol
RSV-treated ALS mice were fed Purina Formulab Diet 5008 chow that had been incorporated with RSV at an initial concentration of at a concentration of 0.1%. An analysis of the chow immediately prior to initiation of the study (after a delay of 6 months due to lab shutdown caused by the COVID-19 pandemic) yielded an RSV concentration of 0.075%. Therefore, based on an average daily food intake of 3-5 g per 25 g body mass, RSV was administered to the ALS mice at an average dose of ∼120 mg/kg/day (i.e., 25% less than our originally planned target of 160 mg/kg/day).
Behavioral measurements
ALS mice were monitored daily to assess feeding and movement throughout the course of disease. Since the BIDMC IACUC would not approve the commonly used endpoint in ALS studies, i.e., a prolonged (>30 s) righting reflex (7), an alternative approach was approved and used for all animals. A standard motor score (MS) assessment (graded on a 0-4 scale) was assigned to all animals (4,24). A MS of four was given for animals with no sign of motor dysfunction; three for animals with detectable tremors when suspended by the tail; two when the animals had mild difficulty ambulating; one when mice were dragging at least one of their hind limbs; zero when both hind limbs were fully paralyzed. Disease onset was determined when an animal received a MS equal to 3. When their MS reached a value of 0, mice were deemed moribund and euthanized by inhalation of carbon dioxide (CO 2 ) gas delivered from a compressed gas canister.
The paw grip endurance test (PGE) (4,24), also known as the hanging wire or inverted screen test (1)(2)(3)(4), was performed bi-weekly beginning at 8 weeks of age to assess muscle strength (32). Briefly, each mouse was placed in the center of a 30 × 42 cm wire rack with 1 × 1 cm square openings and 1 mm diameter wire. The wire rack was inverted quickly and placed on a 25 × 37 cm rectangular Plexiglas support 25 cm above a padded surface. The time until the mouse let go with all four limbs and dropped onto the padded surface below was noted. Each mouse was given three attempts to hold onto the inverted lid for an arbitrary maximum time of 120 s and the longest time achieved by each mouse was recorded.
Compound muscle action potential amplitude
Compound muscle action potential (CMAP) amplitudes were measured bi-weekly using the Natus UltraPro S100 EMG system with Synergy software (Natus Neuro, Middleton WI) as previously described (33).
Electrical impedance myography
All electrical impedance myography (EIM) measurements were made bi-weekly with the mView impedance spectroscopy system (Myolex Inc., Boston, MA) using 41 logarithmically spaced frequencies from 1 kHz to 10 MHz. Surface EIM measurements were performed as described (34) with the animals under 1% isoflurane anesthesia with a heating pad underneath the animal to maintain consistent body temperature at 37 • C. After the fur was clipped, a depilatory agent was applied for 1 min and the skin cleaned with 0.9% saline solution. This process was repeated a total of three times to ensure complete fur removal. The leg was taped to the measuring surface at ∼45 • angle extending out from the body. A fixed rigid fourelectrode array was applied over the gastrocnemius (GA) muscle to obtain longitudinal measurements (35). Measurements were repeated twice and averaged. The array was rotated 90 degrees, and measurements repeated to obtain transverse values. Longitudinal and transverse resistance (R) and reactance (X) values were collected across the entire frequency range and
Statistical analyses
Biomarker data are reported as either the mean ± standard error across the groups. Kaplan-Meier curves were constructed to assess the effect of RSV on the time of disease onset (determined as the time of the first recorded MS value = 3) and length of overall survival (determined as the time of euthanasia required when the MS value = 0). The average time of disease onset and average survival times were reported as mean ± SEM. The median time of disease onset and the median survival time were reported together with their respective Hazard ratios (Log-rank) and their 95% confidence intervals. Basic statistical analyses of the physiological and impedance values were performed using GraphPad Prism V8 (GraphPad Software, Inc. La Jolla, CA). Unpaired t-tests were used to compare the means between two groups; multiple group comparisons were performed by one-way ANOVA with Tukey's multiple comparisons test. The rates of decline of the various biomarkers were determined by simple linear regression. The linearity of the rate of decline of the various biomarkers was quantified as described (24) by performing a least-squares fit based on the average data for each week across the entire measurement period. Residuals for each of the weeks were calculated; the absolute value of these residuals were taken and averaged for each measure and then divided by the average value of that parameter across all measurement points, thus providing a gauge of how data from each week deviated from the regression line. Spearman's correlation coefficient, rho, was calculated to assess the relationship between survival and rate of decline in various biomarker values, as well as for correlations between LP 50 kHz and PGE and CMAP amplitude. For all analyses, p < 0.05, two-tailed was considered as significant. Figures 1, 2 there was no significant difference in the average time of disease onset or average length of survival between the RSV-treated and the untreated ALS animals for either sex when studied individually or when the sexes were combined. As shown in Table 1, neither the median time of disease onset nor the median survival time differed between the RSV-treated and the untreated ALS animals for either sex when studied individually or when the sexes were combined.
E ect of resveratrol on time of disease onset and animal survival
As shown is Figure 3 for both female and male ALS mice, the rates of decline for body mass, CMAP amplitude, LP 50 kHz, PGE, and MS, were similar between untreated and RSV-treated animals. Body mass did not show a significant change from WT until 112 days for females and slightly earlier at 105 days for males. Similarly, the decline in MS became significant at 119 days for both females and males, while the change in CMAP amplitude became significant at 84 days for females and 105 days for males. The decline in PGE became significant at 112 days in females and at 105 days in males. The decline in LP 50 kHz occurred between 84 and 98 days of age for both female and male ALS mice and was significantly different from the WT in females at day 112. Table 2 shows the individual slopes of decline over time for each biomarker. The p-values indicate that there were no significant differences in the rates of decline of any biomarker between RSV-treated and untreated ALS mice for either sex.
Alterations in biomarker values over time
Since neither female nor male RSV-treated animals showed a difference from their untreated counterparts in any of the physiological or electrophysiological biomarkers examined, the data from all animals i.e., the RSV-treated and untreated ALS animals of both sexes, were combined to increase statistical power. A comparison of CMAP amplitude, LP 50 kHz, PGE, and MS values is presented in Figure 4. The arrows in each figure mark the time of significant change from their baseline values (p < 0.05) for two consecutive weeks or longer; this information is summarized in Table 3. When all of the experimental data was combined, the LP 50 kHz values showed a similar change from baseline as that of CMAP amplitude, i.e., at 10 weeks (i.e., 70 days), and slightly better than PGE, i.e., at 12 weeks (i.e., 84 days), whereas the MS did not show a significant change until 15 weeks (i.e., 105 days). The rates of decline, and the linearity of the decline, for each biomarker are summarized in Table 3. As can be seen in Table 3, LP 50 kHz had the smallest mean residual, indicating that the surface EIM impedance data would track very closely along a fitted line. Compared to CMAP amplitude, which also became significantly different from baseline at 10 weeks of age, and which has the next lowest average residual value, the LP 50 kHz residuals were significantly lower (p < 0.0001). In comparison, PGE showed a significant change from baseline at 12 weeks and MS showed a significant change at 16 weeks and these two biomarkers had mean residuals of 0.167 and 0.201 respectfully. Correlations between survival and rate of deterioration of LP kHz, PGE, CMAP amplitude, and motor score /fneur. . Figure 6 shows the correlations between LP 50 kHz and PGE and CMAP amplitude. For this analysis, the data from each week was averaged across all the animals by sex. LP 50 kHz correlates with both PGE and CMAP amplitude (Spearman rho: 0.806 and 0.627, respectively; p-values: 0.004 and 0.044, respectively) suggesting that these biomarkers could be used interchangeably to follow disease progression in this model of ALS.
Correlations between LP kHz and PGE and CMAP
Comparison of disease onset using motor score vs. LP kHz Finally, we compared the median time of disease determined using the MS with that obtained using LP 50 kHz as the biomarker. Disease onset was determined either as the time of the first recorded MS = 3 or as the time of a 20% reduction in LP 50 Hz value from baseline. Figure 7 shows the resultant Kaplan-Meier curves. The median time of disease onset was 112 days for both RSV-treated and untreated animals using MS and 98 days for both RSV-treated and untreated animals using LP 50 kHz. Although no RSV treatment benefit was detected using either strategy, the median time of disease onset was 14 days earlier using the electrical impedance parameter LP 50 kHz as compared to the MS value.
Discussion
The results of the present study are 2-fold: (1) Despite evaluating a number of different biomarkers to assess drug efficacy, RSV at the dose and route of administration used here (i.e., 120 mg/kg/day; orally in the chow) showed no beneficial effect in SOD1-G93A ALS mice. As determined by MS analysis, RSV did not delay disease onset or increase overall survival (2). The various biomarkers evaluated showed different abilities to detect disease onset and track disease progression. EIM had the ability to detect disease onset sooner than two other biomarkers evaluated here, i.e., PGE and MS, and was equivalent to the ability of CMAP to detect deviations from baseline values, at ∼10 weeks of age.
The lack of efficacy of RSV treatment was unexpected since an earlier study reported a modest therapeutic effect of RSV in this mouse ALS model, albeit that study was conducted using a slightly larger dose of RSV (31). Although our target dose was 160 mg/kg/day, due to pandemic-related delays of more than 6 months, the actual dose of RSV used in the present study was ∼25% less than that used previously (31) and could account for the differences seen here. Moreover, whereas the mice in the previous study received RSV in their chow up until the time of their death, in the present study, all ALS mice were switched to Gel-packs, i.e., ending RSV treatment, as soon as the first ALS mice in either group showed signs of paralysis in at least one hindlimb (MS = 1) at ∼15-16 weeks of age. This switch was required because mice were not housed individually and once a Gel-pack was provided in the cage for one animal, all other animals in that cage would have access to the Gel-pack as well and could favor it over the RSV-incorporated chow. Of note, some ALS animals then went on to live another 2-3 weeks without receiving any RSV in their chow. This could have had a detrimental impact on their survival, particularly if treatment with RSV is critical at these very late stages of the disease. Another difference between these two studies was the method used to determine the endpoint. A righting reflex (i.e., the time it takes for an animal to right itself after being placed on its side) of >30 s was used in the previous study (31) but could not be used here due to restrictions by our IACUC that were designed to decrease suffering in the mice. Rather, we were . /fneur. . required to euthanize mice when their MS reached a value of 0 (i.e., when both hindlimbs were paralyzed thus prohibiting their access to the food in the Gel-packs). Therefore, in the absence of a common endpoint (7), it is difficult to compare the effectiveness of RSV treatment as reported in these two studies.
Nevertheless, we were able to compare the different biomarkers evaluated here for their ability to detect disease onset and track disease progression. As shown in Figure 3, all biomarkers that we examined were able to follow disease progression in this ALS model, although body mass and MS appear especially insensitive to disease status until relatively late in disease progression. Of the remaining biomarkers, LP 50 kHz and CMAP amplitude appear to be the strongest indicators of disease-induced deviation from baseline for several reasons. First, when all data from all ALS mice (i.e., RSV-treated and untreated from both sexes) were combined, LP 50 kHz and CMAP amplitude were able to identify a deviation from their respective baseline values at just 10 weeks of age, earlier than . /fneur. . any of the other biomarkers examined. Second, the subsequent decline in LP 50 kHz appeared linear, thus allowing us to determine a rate of disease progression. The results obtained in the present study using EIM as a biomarker are very similar to those that we described previously in this same ALS model when studying the potential therapeutic effects of the drug riluzole (24). Although there was no drug benefit found in either study, EIM was able to track disease onset and progression as well as or better than several other biomarkers in both studies. Finally, there is a good correlation between LP 50 kHz and PGE and CMAP amplitude, indicating that alterations LP 50 kHz values reflect muscle fiber atrophy, decreased muscle strength and deteriorating motor function (18) and could be used interchangeably to follow disease progression. Critically, EIM appeared to perform similarly to CMAP in these analyses. CMAP is a relatively familiar concept to .
/fneur. . the clinical neurophysiology community and can be easily obtained with standard nerve conduction equipment, as was the case here. However, it is important to point out that EIM has three valuable characteristics that makes its application in humans especially appealing. First, it does not require the stimulation of nerve. This allows EIM to be performed easily on many different muscles, including even paraspinal muscles, for which a CMAP would be virtually impossible to obtain given the inaccessibility of the nerves. This is especially important in ALS, where tracking muscle deterioration may benefit from not being limited to only the most distal muscles in the upper or lower extremities. Indeed, it has even been used on the tongue successfully (36). Second, it is entirely painless. Third, it is exceedingly fast to perform. These last two features are especially important for human application since it allows application of the technique to be tailored to the region(s) most rapidly progressing in any patient, which can be quite varied in human disease. It also means that it is possible to even perform measurements at home for disease tracking purposes (37).Thus, even if EIM performs similarly to CMAP, in human application, the technique is simply more versatile as tool to monitor deterioration and potential response to therapy.
In addition to the unfortunate reduction in dose of RSV due to the pandemic, limitations to this study include the fact that it was not blinded to the presence or absence of RSV in the chow (since the chow was color-coded); however, this issue would probably only be of concern if a treatment effect had been identified. Second, only a single dose of RSV was studied. Ideally, it would have been preferable to study several other doses of RSV, both higher and lower. Future studies could deliver RSV in drinking water as noted previously (38); however, the limited aqueous solubility of RSV may preclude this approach especially when high daily doses of RSV are required. Alternatively, lower doses of RSV, possibly used in conjunction with other drugs (39) or treatment with other polyphenolic compounds altogether (28) might provide detectable therapeutic benefit in this ALS model. A third limitation of the present study is that we chose to use the PGE test rather than rotarod performance as a measure of muscle health. Whereas, the PGE test, or hanging wire test, is a basic, inexpensive motor test which requires only balance and grip strength and no significant training, the rotarod test is a more complex test requiring balance and motor coordination as well as muscle strength .
to perform and necessitates special equipment and training of the animals (40). Nevertheless, since rotarod is often used to follow disease progression in ALS mice (1)(2)(3)31), it will be important in future studies to evaluate EIM and rotarod in a side-by-side comparison to determine their respective ability to detect disease onset, track progression and evaluate drug efficacy. A fourth limitation of the present study is that we did not examine any potential targets of RSV in the spinal cord, e.g., sirtuin 1, as described in the previous study (31) and therefore we cannot draw any conclusions regarding the level of RSV in this target tissue. A fifth limitation is that interpretation of the validity of biomarkers using this ALS model is challenging since it is difficult to show biomarker correlation to survival because the mice die over a relatively short period of time, especially when faced with strict mandatory euthanasia criteria. A final limitation of the study is the lack of any standardized commercial surface EIM measurement tool for mice that could be used to replicate our findings. However, to address this issue, we have recently published information about how to create a surface array with an identical footprint to the one used here, as well as specific step-by-step instructions describing and demonstrating how to perform EIM in rodents (34) with the hope of expanding the use of EIM in preclinical studies. In addition to evaluating the therapeutic efficacy of RSV in model of ALS, another major goal of this study was to determine if surface EIM could be sensitive to a treatment effect in ALS mice. Indeed, although EIM has been studied in a variety of neuromuscular disorders and has been shown to be sensitive to deterioration, there remains a dearth of data establishing its potential capability to detect a beneficial effect of therapy in either humans or animals. In non-motor neuron conditions, a therapeutic effect was identified in boys with Duchenne muscular dystrophy (41) who began corticosteroids and also in humans recovering from disuse atrophy (42). The only example of EIM's ability to detect a potential therapeutic benefit in ALS was in a single patient in a study of intraspinal stem cell implantation who also experienced a clinical improvement (43). Since RSV at the dose given did not produce any clinical effect here, it is at least reassuring that EIM also did not identify one.
In summary, whether RSV has a potential therapeutic effect, remains uncertain and additional study of this nutraceutical at doses of 160 mg/kg/day and higher in mice, in these authors' opinion, remains worthy of study. Unfortunately, it is challenging to engage clinical trialists to pursue a study of RSV since it is a readily available nutraceutical and would be easy for trial participants to supplement on their own. Moreover, its nutraceutical status limits the financial incentive to encourage its use since there is no available patent protection. The second motivation for this study was to assess the technique of EIM, and the data collected here continue to support the technology's use in the assessment of motor neuron disease progression. Further development of a standardized animal system and also a regulatory approved system for human use is currently underway. Once available, EIM can be added to the collection of available tools to assess disease progression in preclinical as well as clinical studies of ALS.
Data availability statement
The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author. | 2022-12-22T14:32:56.285Z | 2022-12-22T00:00:00.000 | {
"year": 2022,
"sha1": "635a44757cf78190a63b34a3f4e3734e970c23d8",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "635a44757cf78190a63b34a3f4e3734e970c23d8",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
254976108 | pes2o/s2orc | v3-fos-license | Cerebrotendinous Xanthomatosis: A practice review of pathophysiology, diagnosis, and treatment
Cerebrotendinous Xanthomatosis represents a rare and underdiagnosed inherited neurometabolic disorder due to homozygous or compound heterozygous variants involving the CYP27A1 gene. This bile acid metabolism disorder represents a key potentially treatable neurogenetic condition due to the wide spectrum of neurological presentations in which it most commonly occurs. Cerebellar ataxia, peripheral neuropathy, spastic paraparesis, epilepsy, parkinsonism, cognitive decline, intellectual disability, and neuropsychiatric disturbances represent some of the most common neurological signs observed in this condition. Despite representing key features to increase diagnostic index suspicion, multisystemic involvement does not represent an obligatory feature and can also be under evaluated during diagnostic work-up. Chenodeoxycholic acid represents a well-known successful therapy for this inherited metabolic disease, however its unavailability in several contexts, high costs and common use in patients at late stages of disease course limit more favorable neurological outcomes for most individuals. This review article aims to discuss and highlight the most recent and updated knowledge regarding clinical, pathophysiological, neuroimaging, genetic and therapeutic aspects related to Cerebrotendinous Xanthomatosis.
. Introduction
Cerebrotendinous Xanthomatosis (CTX) or Cerebral Cholesterinosis (MIM #213700) is a rare autosomal recessive inherited metabolic lipid-storage disorder related to bile acid biosynthesis pathways (1). CTX is caused by bi-allelic pathogenic variants in CYP27A1 (2q35), which codes sterol 27-hydroxylase, a mitochondrial enzyme of cytochrome P450 oxidase system. Reduction of the activity of this enzyme leads to increased formation and storage of abnormal lipid content in several tissues, especially tendons, lenses and peripheral, and central nervous system (2).
Several hundreds of cases have been reported since the first description in 1937 by Van Bogaert (3). Current data suggest that the disease appears to be substantially underdiagnosed: the incidence in the United States is between 1:72,000 and 1:150,000, and the disease frequency among Sephardim Jews of Morocco has been estimated to be 6 per 70,000 (4, 5). More than 400 individuals with CTX have been reported worldwide (6), with larger groups of affected individuals being reported in the medical literature from Italy, the Netherlands, Germany, Japan, China, Turkey, Israel, and Spain. CTX incidence ranging from 1:134.970 to 1:461.358 in Europeans, 1:263.222 to 1:468.624 in Africans, 1:71,677 to 1:148,914 in Americans, 1:64,267 to 1:64,712 in East Asians and 1:36,072 to 1:75,601 in South Asians (7).
The aim of the present article is to present current evidence on the main clinical, biochemical, radiologic and treatment aspects of CTX.
. Pathophysiology of CTX CTX is caused by pathogenic variants in CYP27A1, which leads to of deficiency in sterol 27-hydroxylase, a mitochondrial enzyme with a key role in cholesterol metabolism and bile acid synthesis pathways. Multiple variants associated with CTX have been identified, including missense, insertion/deletions, splice-site and nonsense variants, and there is no known clear genotype-phenotype correlation (1,7).
Bile acid synthesis occurs in two main metabolic pathways. The classical pathway initiates with 7α-hydroxylation of cholesterol, in which the enzyme cholesterol 7α-hydroxylase acts. The alternative pathway's first step is 27-hydroxylation of cholesterol, catalyzed by sterol 27-hydroxylase, leading to oxidation of side chains of different sterol intermediates (8). In CTX, the impaired activity of CYP27A1 compromises the formation of chenodeoxycholic acid (CDCA) and cholic acid, to a lesser extent. The loss of the negative feedback effect of CDCA on cholesterol 7α-hydroxylase results in increased levels of 7α-hydroxy-4-cholesten-3-one and its metabolites in the classical pathway ( Figure 1). Elevated serum levels of cholesterol and urine bile acids as glucuronides are found (11). The increased cholesterol metabolites adhere to tissues. Furthermore, there are raised levels of other abnormal pathological intermediates, such as cholestanol. Their accumulation mainly in the brain, eye lenses and tendons cause progressive neurologic dysfunction, cataract and xanthomas, respectively, which are some of the classic clinical manifestations of the disease. However, there is a wide number of phenotypes with diverse systemic and neuropsychiatric symptoms (12).
Accumulation of cholestanol in the brain is still not fully understood, since it does not efficiently cross the blood-brain barrier (BBB) (7,12). Impairment or increased permeability of the BBB has been suggested, possibility endorsed by the high levels of cholestanol and apolipoprotein B found in the cerebrospinal fluid (CSF) in patients with CTX. This change of the BBB may be an effect of circulating bile alcohol glucuronides (1). Nevertheless, some studies have shown an intact BBB in CTX patients, indicating that the increased cholestanol may result from insufficient removal or from synthesis of cholestanol in the brain from cholesterol or another precursor. Furthermore, the bile acid precursor 7α-hydroxy-4-cholesten-3-one crosses the BBB and can be converted to cholestanol by neurons, astrocytes, microglia, and human monocyte-derived macrophages (1, 7) ( Figure 1).
The lesions in CTX present significant deposits of cholesterol as well as cholestanol, although serum cholesterol levels are usually normal. However, elevated serum levels of lathosterol and phytosterol are found, indicating increased de novo synthesis and incremented intestinal absorption of cholesterol, respectively (11). Also, regarding the metabolism of cholesterol, patients with CTX develop early atherosclerosis and xanthomas, which might be related to diminished transport of peripheral cholesterol to the liver, since the levels of 27-hydroxycholesterol, the product of the 27-hydroxylase activity that passes more efficiently the cell membranes, are significantly reduced (1,13). Increased cytoplasmic Nεcarboxymethyl-lysine related to oxidative stress dysfunction has been also evidence in foamy histiocytes from the dentate nucleus (14).
. Clinical overview and diagnosis of CTX
The diagnosis of CTX is mainly based on clinical, neuroradiological, genetic, and biochemical findings. The clinical presentation of the disease is highly heterogeneous, which can lead to significant diagnosis delay. In young people, CTX-related findings are primarily bilateral juvenile cataracts (82%), chronic diarrhea (31%), and intellectual disability (48-74%) (1,15). In adults, these factors are added to the appearance of tendon xanthomas (76%), as well as psychiatric disturbances (11.4%) and neurological disorders, such as peripheral neuropathy (45%), cerebellar .
/fneur. . ataxia (36-83%), movement disorders (parkinsonism, dystonia, myoclonus, postural tremor), cognitive decline (87%), and spastic paraparesis and other pyramidal signs (64-92%) (1, 6, 15, 16) ( Figure 2). In childhood, the earliest manifestations may be infantile diarrhea and neonatal cholestatic jaundice. Typically, jaundice is self-limited and transient, without severe complications, and associated with elevated conjugated hyperbilirubinemia, liver transaminases and alkaline phosphatase. Gamma glutamyl transferase serum levels are generally normal or slightly elevated. The marked reduction of CDCA content does not stimulate the activation and expression of farnesoid X receptor, leading to reduced bile salt exportation and transportation in bile canaliculi (1,19). There are, however, rare descriptions of severe neonatal cholestasis in patients with CTX, leading to very early lethal progression or evolving with the need of liver transplantation (19). Chronic unexplained infancyonset diarrhea is the most common gastrointestinal scenario (76% of patients) and occurs due to the presence of bile alcohol in the intraluminal region and relative absence of CDCA. Steatorrhea, cholestanol, cholesterol and fatty acids are absent in the stool content, as well as malabsorption and failure to thrive are not usually observed (1). CDCA represents a highly effective therapy for symptomatic diarrhea remission. Bilateral juvenile cataract is also a common finding, being described in 85% of patients (20). During juvenile and adulthood periods, optic neuropathy and premature retinal vessel atherosclerosis may represent additional neuroophthalmological complications (1). Neurological signs and tendinous xanthomas often develop after cataracts appear. Xanthomas may be present in 71% of patients with CTX, appearing in the 1st to 3rd decade of life, being more common in late adolescence and early adulthood. They are represented by large amounts of foamy macrophages full filled with complex lipid crystal cleft structures (1,7,12). They appear more frequently in the Achilles tendon but can also be identified in the tibial tuberosity, triceps, and fingers tendons (Figure 3). Tendon involvement in inherited metabolic disorders is not limited to xanthomata in CTX and may also occur in familial hypercholesterolaemia type 3 (PCSK9), sitosterolemia (ABCG8), ataxia with vitamin E deficiency (TTPA), hyperlipoproteinemia type III (APOE), primary hypoalphalipoproteinemia type 2 .
/fneur. . (APOA1), Alagille syndrome (JAG1), and rarely in congenital hypophosphatasia, ochronosis and galactosemia (21). Other typical examination findings in familial hypercholesterolemia, such as corneal arcus and eyelid xanthelasmata, are not observed in CTX. Neurological dysfunction is almost always present, with onset usually in late adolescence or early adulthood (12) (Figure 2). Psychiatric symptoms (behavioral disorders, depression, hallucinations, agitation), dementia, and intellectual disability can be present (15). Intellectual disability is commonly one of the most common neurological complications in CTX starting during the first decade of life (12). Pyramidal (spasticity and hyperreflexia) and cerebellar signs (progressive ataxia and dysarthria) are frequent. Although less common, movement disorders, such as parkinsonism, dystonia, myoclonus, and tremor have been reported. Dystonia is mostly multifocal, with reports of blepharospasm, oromandibular, cervical, and limb dystonia (22). Both positive and negative myoclonus have been reported as one of the earlier movement disorders features of CTX, mainly involving upper extremities, which can have a polyminimyoclonus pattern, resembling intention, or action tremor (23). Palatal myoclonus with pharyngeal, laryngeal, and lingual involvement may also be present (22). Seizures, peripheral neuropathy (that may be axonal, demyelinating or mixed), motor, or sensorimotor (24,25) and pes cavus are also possible features ( Figure 3). Later, with advancing age, other frequent findings can be observed, such as premature atherosclerosis, osteoporosis, and cardiovascular disease, which include ischemic heart disease, mitral valve insufficiency, abdominal aortic aneurysm, coronary artery dissection, and thickening of the interatrial septum due to lipomatous hypertrophy (1). There are also descriptions of cardiac autonomic dysfunction, ventricular tachycardia, and atrial fibrillation in the disease (7,26). Osteoporosis represents a challenging chronic complication of CTX and commonly leads to important morbidity, especially in lately diagnosed patients and generally with poor response to CDCA therapy (1, 2, 6, 7). Childhood and juvenile-onset osteoporosis may be also a possible early complication of CTX and probably underdiagnosed during the first decades of life (1,6).
Despite ataxia is usually considered the main gait disturbance presented by patients with CTX, pyramidal findings are more frequent than cerebellar signs (15) and cases of spinal xanthomatosis, characterized sometimes as pure forms of spastic paraparesis, have been reported in the literature (27)(28)(29). A recent literature review on spinal xanthomatosis, reviewed 34 cases, reporting a mean age of onset of the neurological symptoms of 24 years, with most cases presenting with complex hereditary spastic paraplegia (HSP) phenotype, presenting dementia, ataxia, polyneuropathy, seizures, and psychiatric disease as the complicating feature. Interestingly, 23.5% reported patients had spastic paraplegia as the sole neurological phenotype and only 31% of patients with spinal xanthomatosis presented xanthomas. On the other hand, cataracts and chronic diarrhea were frequent features being present in 78 and 65% of cases, respectively (30). Since the report by Burguez et al. (29) the center of one of the authors of the present manuscript (Saute JA) has been screening CYP27A1 in the investigation of patients with HSP suspicion. Among 115 screened families, CTX was diagnosed in six of them, representing 5% of this cohort in southern Brazil (Saute JA personal communication), confirming that HSP phenotype should lead to CTX suspicion with biochemical or genetic screening for the disease.
The presence of two of four clinical hallmarks (premature cataracts, diarrhea, progressive neurologic signs, tendon xanthomas) should trigger comprehensive biochemical testing for CTX (31). Ophthalmologists may notice unexplained bilateral cataracts, which are a common symptom, especially in children and teenagers (32). Given that these are among of the early indications and symptoms, the association of juvenile cataracts and chronic diarrhea is particularly significant (33)(34)(35). Additionally, children and adolescents with psychiatric disorders including autism spectrum disorder, attention deficit hyperactivity disorder (ADHD), irritability, aggressive outbursts, or oppositional-defiant disorder should undergo further testing, especially in the context of consanguinity or in the presence of cataracts or chronic diarrhea (36).
In clinical practice, the Mignarri index of suspicion can be used to calculate the CTX prediction score and guide the best diagnostic approach for each individual. This index assigns different scores to certain groups of findings (Table 1): (i) family history, (ii) systemic signs, and (iii) neurological involvement (37). The highest indicator score (100) is attributable to "very strong indicator" (A), including positive family history of a sibling with CTX (A1), and/or presence of tendon xanthomas (A2). "Strong indicator" (B) with an individual score of 50 is given to the presence of consanguineous parents (B1), and/or juvenile-onset cataracts (B2), childhood-onset chronic diarrhea (B3), prolonged unexplained neonatal jaundice (B4), and/or ataxia or spastic paraparesis (B5), dentate nuclei signal changes at brain MRI (B6), and/or intellectual disability or psychiatric disturbances (B7). "Moderate indicator" (C) is attributable to individual score of 25 and given to early osteoporosis (C1), and/or epilepsy (C2), parkinsonism (C3), and polyneuropathy (C4) (7,37,38). Plasma cholestanol level assessment is indicated in patients with Mignarri scores ≥100. With previous high levels of plasma cholestanol or a Mignarri score ≥200 (including at least 1 "very strong indicator" or 4 "strong indicator"), there is a formal indication for genetic analysis of CYP27A1 gene (2,37,39,40). The clinical use of the Mignarri score should not limit the early investigation of patients with clinical features highly suggestive of CTX diagnosis (i.e., juvenile cataracts, childhoodonset chronic diarrhea), even in the absence of score values higher than 100 or 200 points or other clinical signs.
As for the biochemical characteristics that provide subsidies to aid in the diagnosis, high serum concentrations of cholestanol are the main diagnostic marker of CTX (7). Elevation of other cholesterol precursors, such as 7-dehydrocholesterol and 8-dehydrocholesterol, is also commonly observed in plasma testing in CTX (41). High levels of bile alcohols, such as glucuronides, can be found in bile, plasma, and urine and are biomarkers for CTX. In tissues, cholesterol tends to be increased, while in plasma its concentration is normal or reduced. Other bile acid precursors in plasma and bile (such as lathosterol, lanosterol) are increased. Classical CTX form usually leads to significantly higher plasma levels of cholestanol than atypical forms and spinal CTX. Plasma levels .
/fneur. . of cholestanol and abnormal intermediates of bile acid synthesis may be elevated also in chronic cholestatic biliary tract diseases, such as Primary biliary cirrhosis and Progressive Familial Intrahepatic Cholestasis type 3 (ABCB4), and in inherited metabolic disorders, such as Niemann-Pick disease type C, sitosterolemia, familial hypercholesterolemia, and peroxisomal biogenesis disorders (1,6,42,43). Drugs which promote abnormal activity of bile acid metabolism, such as intravenous propofol during total anesthesia, may lead to similar metabolic profiles to primary bile acid synthesis disorders (44). There is also important variation in plasma cholestanol levels in different ethnic and age groups, mainly comparing neonatal, childhood and adult (45). Chronic steroid use may reduce plasma cholestanol levels, leading to potential false-negative results and normal values (46), while hypothyroidism may lead to increased levels (1). In CSF analysis, it is possible to find high levels of cholestanol, cholesterol, fragments of apolipoprotein B, apolipoprotein-A1, and albumin (1,7,11). Liver biopsy may demonstrate the presence of electrodense deposits dispersed in the cytoplasm and crystal formation (1, 6, 39), although they are not routinely indicated. The quantification of the bile acid precursor 7 alpha-hydroxy-4-cholesten-3-one is being proposed as a rapid and potentially alternative diagnostic test for CTX (7,12), as well as an optimal therapeutic biomarker during clinical follow-up (47). Sequencing of CYP27A1 should be performed in all patients with a suspected diagnosis of CTX. Some authors have suggested genetic testing in patients with high cholestanol levels or with very high clinical suspicion (34), but nowadays the accessibility of genetic testing is greater than dosing cholestanol for most centers. Bi-allelic pathogenic variants combined with typical clinical findings are diagnostic of CTX, but variants of unknown significance should always be confirmed with plasma cholestanol analysis (6). As genetic testing methods became more accessible and available for investigation in most centers specialized in Rare Diseases, organizing knowledge about potential clinicgenetic correlations in CTX has become a major challenge in clinical practice ( Table 2). The development of specific diagnostic criteria for CTX became an essential measure for diagnostic purposes and future clinical trials (Table 3).
Regarding differential diagnoses, sitosterolemia, familial hypercholesterolemia (both of which can also manifest with tendon xanthomas), Smith-Lemli-Opitz syndrome (characterized by elevated 7-dehydrocholesterol, which may also be present in some CTX patients), other inborn errors of bile acid metabolism (such as HSP type 5A), and non-specific liver disease are among the disorders with features similar to CTX (40,(50)(51)(52). Progressive neurologic symptoms, as well as cataracts and chronic diarrhea, can distinguish CTX from these disorders (34,50). Congenital diarrhea and Alagille syndrome are important differential diagnoses in childhood onset, as well as other causes of neonatal jaundice. In adult patients, differential diagnosis is made with other causes of progressive neurologic disease, such as HSP, hereditary cerebellar ataxias, multiple sclerosis, leukodystrophies, mitochondrial disease, histiocytosis, and other causes of acquired ataxia, and in these cases tendon xanthomas and cataracts are among the most important clues for CTX (53).
The considerable diagnostic delay before the correct diagnosis and proper treatment might be prevented if CTX was included in national newborn screening programs. Newborn screening with dried blood spots is considered by several groups a key step for early diagnosis and treatment (54). However, inclusion in newborn screening programs should be done cautiously and initially in a research context due to the likelihood of detecting mild variants that may remain asymptomatic for a longer period without treatment (6). There is not a definite metabolite or battery tier to perform metabolic or genetic neonatal screening for CTX. However, the most characteristic biomarker in CTX positive newborns after screening was 5-β-cholestane-3α,7α,12α,25-tetrol,3-O-β-D-glucuronide (GlcA-tetrol) (54, 55), and both GlcA-tetrol and the ratio of GlcA-tetrol to tauro-chenodeoxycholic acid .
. Neuroimaging findings
The typical neuroimaging finding of CTX is T2-weighted (T2W) hyperintensity in the dentate nucleus, with cerebellar hypointensity occasionally being seen in the late stage as a result of hemosiderin depositions and microhemorrhages following cerebellar vacuolation (57).
A review article of 38 patients has shown brain MRI abnormalities in 84% of patients, with supra and infratentorial cortical atrophy, subcortical and periventricular white matter abnormalities, brainstem lesions, cerebellar atrophy and other cerebellar parenchymal abnormalities involving the dentate nuclei and the surrounding white matter as the main findings (9). T2W and FLAIR brain MRI may show symmetric hyperintense lesions in the periventricular white matter, posterior limbs of internal capsules, globus pallidum, cerebral peduncles extending into the substantia nigra, anterior region of the pons, inferior olive, or in the cerebellar parenchyma, involving the dentate nuclei and the surrounding white matter, .
TABLE Diagnostic criteria and categories for CTX, based on Sekijima's ( ) and modified from Stelten's criteria ( ).
Diagnostic category: which were hypointense on T1W and diffusion-weighted images (DWI) (7, 40) (Figures 4, 5). The dentate nuclei may present with hypointensities on T2W/FLAIR and susceptibility-weighted (SW) images over time. T2W/FLAIR signal abnormalities in the dentate nuclei were the most common findings in patients with CTX (30). The reasons for preferential involvement of the dentate nucleus remain unclear (12,30). A significant clinical-imaging correlation was only found between the extent of dentate hyperintense lesions and disability expressed by the modified Rankin Scale (30).
The distribution of lesions along corticospinal tracts or in the cerebellum were consistent with the clinical presentation of pyramidal or cerebellar signs, whereas abnormalities of the . /fneur. .
Neuroimaging patterns observed in CTX. (A)
Sagittal T -weighted brain MRI showed hypointense signal in deep cerebellar white matter (white arrow). Axial brain MRI disclosed signal change in deep cerebellar white matter (white arrows) hypointense in T W (B) and hyperintense in T W (C) and FLAIR sequences (D). Axial brain MRI disclosed hyperintensity in deep cerebellar white matter in DWI (E) and ADC sequences (F). Coronal brain MRI showed hypointensity in the deep white matter in T W imaging (G), as well as corresponding hyperintensity (white arrows) in T W imaging (H).
substantia nigra may be associated with Parkinsonian features (30,40). Spinal xanthomatosis may present with non-enhancing long T2W hyperintense lesions predominantly involving the central and posterior cord ( Figure 6). One study has found it to have a relatively mild clinical course, compared with the classic form of the disease (30). This case series with 33 patients reported that patients usually presented pyramidal signs and 48% had dorsal column signs. One of the patients presented with late-diagnosed CTX and after treatment discontinuation had psychiatric symptoms and marked spinal xanthomatosis (rare), which manifested as spastic paraparesis in the absence of xanthomas. Spinal MRI revealed new linear hyperintensities of the corticospinal and gracile tracts (30,57).
Neuroimaging findings, despite being highly variable among patients, generally disclose features which are rarely reversible in multi-imaging modalities after treatment with CDCA, even in individuals with early diagnosis of CTX (13). There is not, however, a direct correlation between the severity of neurological compromise and the extension of dentate nuclei and white matter involvement, including patients with severe motor compromise and cognitive decline with unremarkable neuroimaging studies. Neuroimaging features are quite similar in both adult and childhood-onset cases, despite a more typical pattern being identified in adult patients (1,13,57).
In summary, brain MRI shows diffuse cerebellar atrophy; exvacuum dilatation of the IV ventricle; symmetric hyperintensity of the dentate nuclei and of the cerebellar white matter with some associated DWI hypointensities in the adjacent zone; a soft hyperintensity on T2W and FLAIR may be noted in the superior cerebellar peduncles and in the pyramidal tracts; initial signs of cerebral atrophy may also present be with enlargement of both the insular and frontal spaces (13).
Additionally, magnetic resonance spectroscopy (MRS) may reveal typical lipid peaks, increased choline, and decreased Nacetyl-aspartate peaks in the involved regions, which indicates extensive axonal damage and mitochondrial dysfunction (40). Functional dopaminergic studies showed presynaptic denervation, which is consistent with the mild improvement with levodopa in some patients (58, 59).
Usually tendon xanthomas appear as hypo-to isointense on T1W images and showed low to intermediate signal on T2W images. Bilateral Achilles tendons were most frequently involved. CT scans may show soft tissue enlargement with .
. Clinical management and therapeutic approaches
Generally, neurologic and neuropsychologic evaluations, plasma cholestanol concentration, brain MRI, echocardiography, and total-body bone density should be evaluated annually; more frequent surveillance may be indicated in newly diagnosed patients until biochemical indicators of disease stabilize. Of note, because cholestanol levels may require considerable time to return to a normal range (reduction of 91 µmol/l in an average follow-up of 34 months) after specific treatment initiation, other substrates in the cholestanol pathways, such as intermediate bile alcohols, could be used for short-term follow-up or as surrogate outcomes in clinical trials (7).
In CTX pathophysiology, there is markedly reduced production of bile acids, especially chenodeoxycholic acid (CDCA), and to a lesser extent cholic acid (CA). As CDCA and CA have a negative physiologically feedback on 7-α-hydroxylase, the rate-limiting enzyme of bile acids synthetic pathway, in CTX 7-α-hydroxylase activity is highly enhanced (60). This results in reduced synthesis of CDCA, high production of cholestanol and its subsequent accumulation in different tissues, as well as normal or low levels of cholestanol in plasma and bile alcohols in urine (61). Evidence that cholestanol may be neurotoxic is supported by the finding of cholestanol deposition and apoptosis in neuronal cells, most notably Purkinje cells, in the cerebellum of rats fed a 1% cholestanol diet (62). CDCA also blocks and antagonizes GABA A and NMDA receptors (63) (Figure 1).
Given the reduced synthesis of CDCA and high production of cholestanol, as well as the recent evidence that cholestanol may be neurotoxic, CDCA has become the standard of care for CTX patients. It prevents the accumulation of cholestanol by inhibiting bile acid synthesis through a negative feedback pathway (31). This drastically lowers plasma cholestanol concentrations in patients and its accumulation in tissues. While initial studies with CDCA reported clear short-term clinical benefit in most patients with CTX, long-term studies have rather reported stabilization in some patients (61). In 1984, one of the first studies was published that spoke in favor of long-term benefit from CDCA therapy in outcomes such as decreased serum cholestanol, improved neurological examination and electroencephalographic findings (16). Despite this, there have been no specific randomized placebocontrolled clinical trials of CDCA in patients with CTX to this date. CDCA treatment typically does not significantly reduce tendon xanthomas or improve cataracts, but can stabilize or improve neurologic manifestations, including cognitive deterioration, pyramidal tract signs, and cerebellar deficits (15). Therefore, given the natural course of CTX, the primary aim of treatment is stabilization or improvement of neurological signs and symptoms based on results of retrospective trials (31).
While most CTX patients do well in response to CDCA therapy, others continue to deteriorate neurologically, especially patients diagnosed over the age of 25 years who already have significant neurologic disease (64). When significant neurologic pathology has occurred, the effect of treatment seems to be limited (15). Early diagnosis and treatment in CTX are imperative to prevent potentially irreversible neurological damage and it changes the disease course in a positive way, alleviating both the neurologic and systemic symptoms of CTX. In a study with 43 CTX patients with a follow-up of 8 years, cognitive impairment (74%), premature cataracts (70%), tendon xanthomas (77%), and neurologic disease (81%) were the most frequent conditions, and treatment with CDCA improved symptoms in 57% of patients, despite of 20% continued to deteriorate (64).
In the largest retrospective cohort study of CDCA treatment with 56 patients, all patients diagnosed and treated before the age of 24 had complete resolution of previous neurologic symptoms and no new onset symptoms, while 61% of patients diagnosed and treated after the age of 24 had neurologic deterioration, with parkinsonism as the main treatment resistant feature (6). These findings suggest that CDCA treatment should be instituted as soon as possible, and that early diagnosis is paramount to good outcomes in this disease (65).
The currently recommended dosing for CDCA ranges from 5 to 15 mg/kg per day in children and 750 mg per day, in three divided doses, in adults. There is a formal recommendation of a slowly progressive dosing introduction with 500 mg per day, for 2 weeks, followed by a weekly increase of 250 mg per day, until the recommended dose is reached (6). If serum cholestanol or urine bile alcohols remain elevated after 3 months, CDCA may be raised up to 1,000 mg per day. For children and adolescents, it is recommended an initial dose of 5 mg/kg per day, in three divided doses. Few specific adverse events or safety concerns have been reported for CTX patients treated with CDCA, with most reports indicating no major adverse events. Discontinuation of therapy due to adverse events occurs in <5% of cases (31). Hepatotoxicity is considered the major concern with CDCA, however in most cases with minor serum aminotransferase elevations (66). Hepatotoxicity leads to the need of dosing adjustment (67), despite in most cases minor serum transaminase elevations (up to three times the upper limit of normality) represent a transient phenomenon with complete resolution in up to 6 months after drug discontinuation. Patients with serum aminotransferase levels over three times the upper limit of normality and evolving with recurrence of such laboratory changes after reintroduction of CDCA may discontinue therapy. CDCA restart is recommended generally at lower initial doses of 5 mg/kg per day and maintained at such dosing with no significant complications (31). If patients evolve with persistent diarrhea or severe gastrointestinal complaints, transient reduction of the recommended dosing is performed until improvement of symptoms, when effective dosage is restarted. CDCA therapy is contraindicated in patients with moderate to severe hepatocyte dysfunction, intrahepatic cholestasis, primary biliary cirrhosis, sclerosing cholangitis, biliary pancreatitis, biliary gastrointestinal fistula, acute cholecystitis or cholangitis, or biliary tract obstruction (31, 64). Patients with absolute contraindication to CDCA use or severe adverse events may potentially benefit from the alternative use of CA (68).
Early reports of combination therapy with low doses of CDCA with HMG-CoA reductase inhibitor pravastatin suggested that this combination reduced plasma levels of cholestanol and avoided the increase in triglyceride and low-density lipoprotein (LDL)-cholesterol with CDCA alone; however, the follow-up period was too short to detect relevant clinical changes (69). Other case reports also favored the combination of CDCA with HMG-CoA reductase inhibitors like simvastatin and atorvastatin, with biochemical response and reports of improvements of peripheral neuropathy and cognitive symptoms when statin was added to CDCA (70, 71). LDL-apheresis with CDCA and HMG-CoA reductase inhibitor is another possible approach, despite the consistent reduction of cholestanol to normal or even subnormal levels, a definite improvement of clinical symptoms was not noted with this aggressive cholestanol lowering therapy (72,73). CDCA alone presents stronger evidence on clinical outcomes than combined therapy with CDCA and HMG-CoA reductase inhibitor. In a recent consensus statement with Delphi method, the expert panelists considered CDCA alone the preferred first line therapy for CTX, but also considered that combination therapy with HMG-CoA reductase inhibitor improves/stabilizes the prognosis. The panel disagreed that LDL apheresis improves/stabilizes prognosis (6).
Previous studies have also evaluated the possible role of CA therapy in the management of patients with CTX with variable patterns of clinical response, especially regarding neurological involvement (49,74). Through the suppression of endogenous bile acid biosynthesis by negative feedback mechanisms, similar to CDCA, CA supplementation provided significant reduction in the urinary excretion and serum production of intermediate biomarkers of bile acid biosynthesis pathway (75). A retrospective Franco-Belgian multicentric study evaluated the safety and efficacy of CA in the treatment of both CDCA-naïve and non-naïve patients with CTX (60). More than 80% of individuals had clinical improvement or stabilization and marked reduction of plasma cholestanol levels during treatment period, disclosing a possible role of CA as a second-line or .
/fneur. . alternative therapy in patients who presented moderate to severe side effects with CDCA therapy (60). There is currently no consensus regarding the use of CA as a monotherapy in CTX (6). Other studies have evaluated the potential role of several other compounds, such as cholestyramine (76), hydrophilic ursodeoxycholic acid (76), ursodeoxycholic acid (76) and, in biochemical and clinical parameters of patients with CTX, however no significant responses or benefits were observed (76). Finally, preclinical studies showed that single intravenous administration of adeno-associated virus (AAV) expressing CYP27A1 directed to liver provided full metabolic restoration of the disease in a transgenic mice model of CTX in a greater extent than CDCA (77), being a promising therapeutic option that should be pursued by future studies.
. Prognosis
Age at definite diagnosis and treatment introduction as early as possible represents the most important prognostic factors related to treatment responses and outcomes, especially in asymptomatic patients or individuals without significant neuropsychiatric involvement (1,6,7). Progressive reduction of plasma cholestanol levels after treatment initiation leads to slowing clinical progression of CTX (6). Patients with typical brain MRI involvement of the deep cerebellar white matter with vacuolation seen as hypointense on T1-weighted and FLAIR sequences generally evolve with worse prognosis (60), as well as the absence of dentate nuclei signal changes is generally associated with better prognosis (6). Some CYP27A1 pathogenic variants have been also associated with more severe neurological and multisystemic involvement and then worse prognosis ( Table 2). As CDCA represents a specific drug therapy developed for continuous use, long-term treatment availability for initiation and maintenance represents a key issue for better neurological and systemic outcomes and increased life expectancy (6).
. Conclusion
CTX is a rare and potentially treatable genetic disease that results in multisystemic involvement. The main cause of disability results from neurologic manifestations including pyramidal signs, ataxia, and cognitive impairment. Neuroimaging with typical T2W hyperintensity in the dentate nucleus, bilateral juvenile cataracts and the presence of tendon xanthomas are important clues for diagnosis. Treatment with CDCA is safe and appears to be effective based on results of retrospective studies, especially if initiated early. Prompt diagnosis, possibly with neonatal screening, may significantly reduce the burden of this disease.
Author contributions
PN and AB: conception and design, acquisition of data, analysis and interpretation of data, writing of the first draft, review and critique, and final approval of the version. RR, SV, DA, VG, HF, CS, DD, AP, and WP: conception and design, analysis and interpretation of data, writing of the first draft, review and critique, and final approval of the version. JS, PS, and PB-N: conception and design, acquisition of data, writing of the first draft, review and critique, and final approval of the version. All authors contributed to the article and approved the submitted version. | 2022-12-23T14:29:18.433Z | 2022-12-23T00:00:00.000 | {
"year": 2022,
"sha1": "c796f4d61b121c64d793693e5c7fec900b2185e2",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "c796f4d61b121c64d793693e5c7fec900b2185e2",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
264532764 | pes2o/s2orc | v3-fos-license | To use or not to use? Understanding doctoral students’ acceptance of ChatGPT in writing through technology acceptance model
While artificial intelligence-based chatbots have demonstrated great potential for writing, little is known about whether and how doctoral students accept the use of ChatGPT in writing. Framed with Technology Acceptance Model, this study investigated doctoral students’ acceptance toward ChatGPT in writing and the factors that influence it. The questionnaire survey revealed a high intention to use ChatGPT in writing among doctoral students in China. The findings further indicated that attitude was a significant predictor of behavioural intention to use ChatGPT in writing and mediated the impacts of perceived usefulness and perceived ease of use on it. Perceived ease of ChatGPT use was in turn influenced by students’ past ChatGPT use experience. This study provides powerful evidence for the applicability of Technology Acceptance Model in the acceptance of ChatGPT in writing. The results have significant implications for leveraging ChatGPT for writing in higher education.
Introduction
Artificial intelligence (AI) technologies play a crucially important role in the increasingly digitalized world (Lee et al., 2022;Farrokhnia et al., 2023).As a generative AI chatbot, ChatGPT is a large language model that can autonomously learn from data and produce human-like texts (van Dis et al., 2023).It can converse on a wide range of topics and generate human-like responses after training huge quantities of text data (OpenAI, 2023).Ever since its release in November 2022, ChatGPT has sparked debates about its implications for education (Farrokhnia et al., 2023;Tlili et al., 2023;van Dis et al., 2023).While ChatGPT can potentially transform educational practices by providing a baseline knowledge of diverse topics (Tlili et al., 2023) and facilitating personalized, complex learning (Farrokhnia et al., 2023), it may supply incorrect texts, encourage cheating, and threaten academic integrity (Dwivedi et al., 2023;van Dis et al., 2023).The controversies have made ChatGPT "the most high-profile and controversial form of AI to hit education so far" (Williamson et al., 2023, p. 2).
Writing has been one of the most influenced domains in the ChatGPT era (Taecharungroj, 2023;Yan, 2023).While writing plays an important role in higher education (Kirkpatrick, 2019), it has been oftentimes considered challenging for language learners, especially for those who learn and use English as an additional language (Ma, 2021).Prior research has suggested that chatbots are effective in addressing this challenge, since they could supply meaningful guidance and substantive feedback to support language learners to write at their own pace in a less anxiety-inducing environment and improve writing quality (Guo et al., 2022;Zhang et al., 2023).As a chatbot powered by generative AI, ChatGPT has demonstrated improved abilities than earlier chatbots (e.g., ELIZA) to understand natural language, generate appropriate responses, and engage in free-flowing conversations throughout the writing process, hence opening a new avenue for writing practice (Barrot, 2023;Su et al., 2023).As succinctly summarized by Imran and Almusharraf (2023), ChatGPT is "a complete package from generation to final proofreading and editing of writing material" (p.2).Nevertheless, till now, scarce attention has been paid to the acceptance and usage of ChatGPT in English writing-a daunting but critical work facing doctoral students (Kirkpatrick, 2019).Little is known about whether and how doctoral students intend to use ChatGPT in writing and the key determined factors.Informed by Technology Acceptance Model (TAM; Davis, 1989), the present study seeks to fill the void by addressing the following two questions: (1) how is the doctoral students' acceptance intention to ChatGPT in writing?( 2) what factors may influence doctoral students' acceptance intention to ChatGPT in writing?Such information is important, as the individuals' intention to adopt and use AI technology is critical to improving teaching and learning of writing (Cheng, 2019;Yan, 2023).
Literature review 2.1. The use of ChatGPT in writing
Chatbots, computer programs or AI systems designed to simulate human conversations and interact with users via natural language, have gained considerable attention and increasingly applied in writing in the past decade (Zhang et al., 2023).Chatbots have demonstrated great potential as a writing assistant and learning partner in writing classrooms, as they can provide a broad array of language choices and feedback to students' writing process and make students feel less stressed about their writing performance in the learning process (Guo et al., 2022).ChatGPT was developed in 2022 as a novel chatbot rooted in Generative Pre-training Transformer architecture, and outperforms early chatbots in terms of the capability for understanding and producing human-like texts as well as providing feedback on long texts (Dwivedi et al., 2023;Farrokhnia et al., 2023;Su et al., 2023;Tlili et al., 2023).Such affordances make it a powerful writing assistant and writing tool (Barrot, 2023;Dergaa et al., 2023;Imran and Almusharraf, 2023).As shown in Taecharungroj's (2023) analysis of early reactions on Twitter, ChatGPT has been most frequently used for writing, such as essays and articles.
Given the close link between ChatGPT and writing, a growing body of research has been undertaken to investigate the benefits and threats associated with the use of ChatGPT in writing.Piloting ChatGPT for academic writing, Bishop's (2023) user experience demonstrated that ChatGPT is effective in explaining well-known concepts, translating between languages, giving timely and personalized feedback, adjusting the style and tone of texts to imitate different writers, and perfecting the mechanics of writing, thereby enhancing writing efficiency and promoting writing quality.Zooming into the use of ChatGPT in second language writing context, Barrot (2023) and Su et al. (2023) further unpacked the potential of collaborating with ChatGPT in writing classrooms.For them, ChatGPT has taken into consideration various writing constructs, such as pragmatics, coherence and syntax, and could support the structural, dialogical and linguistic aspects of quality writing by assisting students in topic generation, outline preparation, content revision, proofreading and post-writing reflection.Taking stock of the research on ChatGPT in academia, Dergaa et al. (2023) and Imran and Almusharraf (2023) highlights the need to leverage ChatGPT as a valuable writing assistant tool to support the writing process and enhance academic writing.
Notwithstanding the benefits, the use of ChatGPT in writing has also raised concern for inaccurate and unintelligent responses, academic integrity, learning loss and educational inequality (Dwivedi et al., 2023;Farrokhnia et al., 2023;Tlili et al., 2023).As noted by the developer itself (OpenAI, 2023), "ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers." Such incorrect and biased information can mislead students and be further incorporated into their writing, thereby harming knowledge practice and science progress (Tlili et al., 2023;van Dis et al., 2023).Another limitation of using ChatGPT in writing is associated with its unintelligent responses, typified by its frequent use of irrelevant statements, template rigidity of writing, and insufficiencies in emotional depth in writing (Barrot, 2023).Also, ChatGPT does not always reference sources appropriately and cannot be held accountable for their work, which raises pertinent issues concerning plagiarism and academic integrity (Dergaa et al., 2023;van Dis et al., 2023;Williamson et al., 2023;Yan, 2023).Additionally, the generative nature of ChatGPT allows students to complete writing assignments simply through unwitting copy-and-paste, and hence results in learning loss, especially when students become too reliant on the AI-powered chatbot for convenience (Barrot, 2023).Likewise, using ChatGPT in writing could lead to educational inequality (Dwivedi et al., 2023).Focusing on ChatGPT's text generation functionality, for example, Yan's (2023) research showed the undergraduates were much concerned with its impact on educational equity, given that writing teachers may not effectively distinguish texts produced by students from those produced by ChatGPT.
While the above user cases and scholarly discussions are helpful in unpacking the potentials and pitfalls of using ChatGPT in writing, the research into ChatGPT is still at its early stage (Barrot, 2023).Little empirical research has been conducted to examine the socio-technical aspects of using ChatGPT in writing.Since writing is essential to doctoral education (e.g., Kirkpatrick, 2019) and subject to the advances in AI technologies (Yan, 2023), it is necessary to explore and examine doctoral students' intention toward ChatGPT and the influencing factors.Such information could shed light on doctoral students' acceptance of ChatGPT in writing, and generate useful insights to leverage ChatGPT and other similar generative AI technologies for the teaching and learning of writing in higher education.
Technology acceptance model
User acceptance refers to the prospective users' predisposition toward using technology (Lee and Lehto, 2013).TAM, emerging from the theory of reasoned action, has become an influential 10.3389/fpsyg.2023.1259531Frontiers in Psychology 03 frontiersin.orgsocio-technical model that seeks to identify and explain the end-users' acceptance of technology (e.g., Cheng, 2019;Granić and Marangunić, 2019).In TAM, individuals' acceptance of a particular technology is operationalized as their behavioural intentions to use it (Lee and Lehto, 2013).TAM postulates that people's actual usage of technology is determined by their behavioural intentions.Behavioural intentions, in turn, are jointly determined by people's attitudes and perceived usefulness (Davis et al., 1989).Attitude towards technology underscores individuals' affective reactions to and evaluation of the use of the technology (Ajzen, 1991;Lee and Lehto, 2013) and it is closely related to one's intrinsic motivation (Davis et al., 1992).If people have a more favourable attitude toward the technology, they are more likely to form positive intentions to use it (Davis et al., 1989;Estriegana et al., 2019).Perceived usefulness is people's belief about the extent to which using the technology will improve their performance (Davis, 1989).It is a type of extrinsic motivation in determining technology acceptance and technology usage behaviour (Davis, 1989;Lee and Lehto, 2013).That is, if students believe that using the technology will improve their performance in writing, they tend to have a positive inclination to use it.The perceived usefulness is also hypothesized to have a positive influence on attitudes and thus affect behavioural intentions (Davis et al., 1989).If the technology is viewed as useful in enhancing writing performance, students are apt to appraise the technological means positively and inclined to use it (Estriegana et al., 2019).Therefore, this study proposes the following hypotheses.
Hypothesis 1: Attitude towards using ChatGPT in writing would significantly and positively influence students' behavioural intention to use ChatGPT in writing.
Hypothesis 2: Perceived usefulness of using ChatGPT would significantly and positively influence students' behavioural intention to use ChatGPT in writing.
Hypothesis 3: Perceived usefulness of using ChatGPT would significantly and positively influence students' attitude towards using ChatGPT in writing.
Hypothesis 4: Attitude towards using GPT would significantly mediate the effects of perceived usefulness on students' intention to use ChatGPT in writing.
Furthermore, TAM posits that attitude is jointly determined by perceived usefulness and perceived ease of use which refers to "the degree to which a person believes that using a particular system would be free of effort" (Davis, 1989, p.320).In TAM, perceived ease of use is assumed to have a significant effect on perceived usefulness and attitudes, resulting in increased behavioural intention (Davis et al., 1989;Alfadda and Mahdi, 2021).If the technological tool is perceived to be easy to use, students tend to consider it helpful and develop a favourable attitude, thereby demonstrating a strong inclination to use it in writing (Alfadda and Mahdi, 2021).Subsequently, the following hypotheses can be proposed.
Hypothesis 5: Perceived ease of use would significantly and positively influence students' perceived usefulness of ChatGPT in writing.
Hypothesis 6: Perceived ease of use would significantly and positively influence students' attitude towards using ChatGPT in writing.
Hypothesis 7: Attitude towards using GPT would significantly mediate the effects of perceived ease of use on students' intention to use ChatGPT in writing.
Meanwhile, a number of studies have revealed a strong and direct association between perceived ease of use and behavioural intention (Granić and Marangunić, 2019).In Yang and Wang's (2019) study, for instance, the perceived ease of use showed a significant and positive impact on students' behavioural intention to use machine translation.As argued by Shiau and Chau (2016), when people perceive that using a technological tool does not require much effort, they will be more intended to use it.Hence, the following hypothesis is proposed.
Hypothesis 8: Perceived ease of use would significantly and positively influence students' behavioural intention to use ChatGPT in writing.
According to Davis et al. (1989), perceived usefulness and perceived ease of use are influenced by a range of external variables, among which experience is one best studied external factor (Abdullah and Ward, 2016).The existing literature suggests that experience influences both learners' perceived usefulness (e.g., Chang et al., 2017;Yang and Wang, 2019) and perceived ease of use of educational technologies (e.g., Purnomo and Lee, 2013).For instance, Chang et al. (2017) found that students who have more experience in using computers tend to demonstrate more positive perceptions regarding the ease of use and usefulness of e-learning.Hence, this study assumes that students who have experience in using generative AI chatbots are more prone to understand usefulness of ChatGPT and become more proficient in using it in EFL writing.The following hypotheses are accordingly proposed.
Hypothesis 9: Past ChatGPT use experience would significantly and positively influence perceived usefulness of ChatGPT in writing.
Hypothesis 10: Past ChatGPT use experience would significantly and positively influence perceived ease of using ChatGPT in writing.
Taken together, and in line with the existing literature on TAM, a conceptual model is formulated in the present study (see Figure 1).
Measures
To determine doctoral students' acceptance of ChatGPT in writing and the factors influencing it, an online survey was administered in March 2023.The survey instrument consisted of two sections subsuming questions pertaining to demographic profiles (gender, major, and past ChatGPT use experience) and those concerning the constructs in TAM.The survey items in the second part were adapted from Davis (1989), Edmunds et al. (2012), Lee and Lehto (2013), and Rafique et al. (2020), and in light of the usage of ChatGPT in writing.In the second section, the respondents indicated their agreement level on every item by recording their response in a 6-point Likert scale, ranging from "1" (Strongly Disagree) to "6" (Strongly Agree).
Behavioural intention to use ChatGPT in writing
Behavioural intention to use ChatGPT in writing was measured on a five-item scale adapted from Lee and Lehto (2013) and Rafique et al. (2020).The five items (e.g., "I intend to use ChatGPT to improve my English writing ability in the future") showed high reliability (Cronbach's α = 0.871).According to Hu and Bentler (1999), the CFA results demonstrated good construct validity (χ 2 = 7.976, df = 5, RMSEA = 0.050, CFI = 0.995, TLI = 0.990), with factor loading ranging from 0.659 to 0.838.
Past ChatGPT use experience
In the present study, students' past ChatGPT use experience was operationalized as whether the students had used ChatGPT de facto at the time of data collection.It was measured via one item, i.e., "Have you ever used ChatGPT before?"The respondents indicated their past experience on a yes-no scale (Yes = 1, No = 0).
Data analysis
SPSS 24.0 and Mplus 7.4 Software were used for data analysis.First, the SPSS software was used to conduct descriptive analysis and correlation analysis.Then, the Mplus software was utilized to construct structural equation modelling (SEM), with a view to calculating relationships among focus variables and conduct mediation analysis.For mediation analysis, bias-corrected bootstrapping method with 2000 times of resampling was employed to calculate the point estimates of the confidence intervals regarding the mediating effects.
In light of Hu and Bentler's (1999) research, the fit of the model was evaluated by the following cut-off values: Root mean-square error of approximation (RMSEA) < 0.08; Tucker-Lewis index (TLI) > 0.90; and comparative fit index (CFI) > 0.90.Additionally, Harman's single factor test was conducted by SPSS software to exclude possible common variance bias.The results showed that less than 50% (46.80%) of the total variance of variables were explained after all the items were loaded into one factor, indicating no need to control common variance bias (Mat Roni, 2014).
Preliminary analysis
The descriptive statistics of all variables are presented in Table 1.Except for past ChatGPT use experience, the other four focus variables' score fall between 3.954 and 4.159, indicating mid-to-high levels on behavioural intentions, attitudes, perceived usefulness and perceived ease of use regarding ChatGPT.Particularly, the students reported the highest score on behavioural intention (M = 4.159), revealing doctoral students' high intention to use ChatGPT in writing in this study.
Structural equation modelling
SEM analysis was conducted to examine the relationships among focus variables with gender being controlled for all the structural relationships.As shown in Figure 2, the model had a high explanation for variance in students' behavioural intention to use ChatGPT in writing (80.1%), attitude towards using ChatGPT (70.2%), and perceived usefulness of ChatGPT (65.7%), respectively, and a low explanation for variance in perceived ease of ChatGPT use (2.4%).The model fit indices (χ 2 = 350.545,df = 198, RMSEA = 0.056, CFI = 0.951, TLI = 0.943) indicates a good SEM model fit.
Perceived attitude towards using ChatGPT in writing had significant and positive impacts on students' behavioural intention to use ChatGPT in writing (β = 0.850, p < 0.001), supporting H1.Perceived usefulness of using ChatGPT had significant total influences on students' behavioural intention to use ChatGPT (β = 0.577, p < 0.001), but did not have significant and direct influences on it (β = 0.117, p > 0.05), thus rejecting H2.However, perceived usefulness of ChatGPT had positive and significant influences on students' attitude towards using ChatGPT in writing (β = 0.541, p < 0.001), thus supporting H3.Besides, perceived ease of use had significant and positive effects on students' perceived usefulness of ChatGPT in writing (β = 0.817, p < 0.001), thus supporting H5.Perceived ease of ChatGPT use had positive and significant influences on students' attitude towards using ChatGPT in writing (β = 0.337, p < 0.001), thereby supporting H6.Perceived ease of use had significant total influences on students' behavioural intention to use ChatGPT (β = 0.689, p < 0.001) but had no significant and direct influence on it (β = −0.069,p > 0.05), rejecting H8.In addition, past ChatGPT use experience had significant and positive influences on students' perceived ease of using ChatGPT in writing (β = 140, p < 0.05) but had no significant influence on perceived usefulness of ChatGPT (β = −0.065,p > 0.05).Therefore, the results supported H10 but rejected H9.
Additionally, results of mediation analysis (Table 2) show that students' attitude towards using ChatGPT significantly mediated the effects of perceived usefulness of ChatGPT on their behavioural intention to use ChatGPT in writing (β = 0.460, p < 0.001, 95% CIs: 0.149 to 0.771), hence supporting H4.It also significantly mediated the influences of perceived ease of ChatGPT use on students' behavioural intention to use ChatGPT in writing (β = 0.287, p < 0.05, 95% CIs: 0.022 to 0.552).Thus, H7 was supported.
Discussion
While ChatGPT has ignited debates about its applications in education (e.g., Farrokhnia et al., 2023), it remains unknown whether students are willing to use it or not in writing.This research contributes to the existing literature by investigating Chinese doctoral students' acceptance toward ChatGPT in writing and its major influencing factors.Through the lens of TAM, the present study revealed a strong intention to use ChatGPT in writing among doctoral students, which was affected by their attitudes, perceived usefulness, and perceived ease of use.The findings provide a deeper understanding of doctoral students' acceptance inclination toward ChatGPT and other generative AI chatbots in writing in higher education.
Although ChatGPT remains new, the doctoral students demonstrated a strong intention to use it in writing.This corroborates Taecharungroj's (2023) finding that ChatGPT has been mainly used in the writing domain.Students' high behavioural intentions might be attributed to the affordances of ChatGPT for writing.As shown in prior research (e.g., Bishop, 2023;Yan, 2023), ChatGPT could help students to brainstorm ideas, obtain timely and personalized feedback, translate language items, and improve written drafts.This makes it a potential mediation tool for doctoral students to write more fluently and effectively in the publish-or-perish system (Kirkpatrick, 2019).
Consistent with our prediction, doctoral students' attitude towards using ChatGPT in writing was found to be a significant predictor of behavioural intention.While a number of prior studies have removed attitudes from TAM due to its weak role in mediating the effects of perceived usefulness and perceived ease of use on behavioural intention (e.g., Lee and Lehto, 2013;Yang and Wang, 2019), this study found that attitude not only directly influences behavioural intention but also mediates the impacts of perceived usefulness and perceived ease of use on it.The finding lends support to the original TAM (Davis et al., 1989).It also supports Ajzen's (1991) argument that personal attitude towards a behaviour functions as a major determinant of people's intentions to perform it.In other words, when doctoral students have more positive evaluation of using ChatGPT in writing, they are more willing to perform the behaviour.Also, as suggested by the expectancy-value model of attitudes (Ajzen, 1991;Ajzen and Fishbein, 2008), people's attitude is further determined by salient beliefs regarding the outcome of performing the behaviour and attributes associated with the behaviour, such as the cost and effort incurred by performing it.In this sense, positively valued outcomes and easier management of the technology could strengthen users' affective reactions towards the technology and boost their sense of efficacy, hence contributing to their favourable attitude towards it and the resultant increasing behavioural intention (Davis et al., 1989).As shown in this study, doctoral students' attitude towards using ChatGPT in writing, shaped by the perceived usefulness and ease of use, played an important role in mediating their effects on students' intention to use ChatGPT in writing.Furthermore, the results revealed that perceived usefulness and perceived ease of use had significant total influences on students' behavioural intention to use ChatGPT in writing.This echoes the central role of perceived usefulness and perceived ease of use in the adoption process of technology in prior research examining TAM (Cheng, 2019;Granić and Marangunić, 2019;Alfadda and Mahdi, 2021).Nevertheless, the study found no significant direct influence of them on doctoral students' behavioural intention.Instead, they only influenced behavioural intention through attitudes.This surprising finding is inconsistent with previous studies on people' acceptance of educational technology (e.g., Estriegana et al., 2019;Yang and Wang, 2019).This might be due to the fact that some researchers (Davis, 1989;Lee and Lehto, 2013;Chang et al., 2017;Yang and Wang, 2019) did not include the attitude variable in their models and consequently failed to explore its mediating effects.Another plausible explanation might be that ChatGPT remains new, and early adopters use ChatGPT mainly because it facilitates inherently enjoyable and interesting experience (Taecharungroj, 2023;Tlili et al., 2023).In other words, the use of ChatGPT at this stage is primarily intrinsically motivated (Davis et al., 1992).Accordingly, the expected outcome of using ChatGPT for enhancing writing performance at the extrinsic level and perceived ease of using ChatGPT at the technical level could be instrumental, when such beliefs catalyse intrinsic motivations and when using ChatGPT in writing appeals to individuals (Ryan and Deci, 2000).
Also, the study found that perceived ease of use was found to be significantly and positively influenced perceived usefulness of ChatGPT in writing.This is analogous to Rafique et al. 's (2020) study, in which users' perceived ease of using mobile library applications had a significant influence on perceived usefulness.By the same token, users' perceived ease of using ChatGPT in writing could greatly shape the perceived usefulness (Davis et al., 1989).If doctoral students consider it challenging to apply ChatGPT in writing, they are likely to hold that ChatGPT has little effect on their writing.When they perceive ChatGPT easy to use, they tend to regard it as useful and helpful for writing.
In addition, this study extends prior research on TAM by including experience as an external factor to enhance the model explanatory power.Doctoral students' past ChatGPT experience is proved to be a significant predictor for perceived ease of use.The more experienced the students are, the more positive they are about the ease of using ChatGPT in EFL writing.This is compatible with Purnomo and Lee's (2013) study, where prior computer experience had a positive influence on learners' perceived ease of use an e-learning system and such influence was stronger than that on perceived usefulness.The findings also support of argument Nelson's (1990) that the acceptance of technology relies upon not only the technology itself but also individuals' expertise in using it.Students with experience in using generative AI chatbots could employ the knowledge and skills obtained from prior experience to writing, develop a better personal control, and accordingly perceive it easier to use it in writing (e.g., Purnomo and Lee, 2013;Chang et al., 2017).
Conclusion
Despite the increasing interest in ChatGPT in educational settings, research on its acceptance is still scarce in education.Based on TAM, descriptive statistics, correlation analysis, and SEM were employed to gauge doctoral students' acceptance of ChatGPT in writing and explore the influencing factors.Data analysis revealed a high-level intention to use ChatGPT in writing, shaped by doctoral students' attitudes, perceived usefulness, and perceived ease of use.The present study could contribute to ChatGPT research in both theoretical and practical ways.Theoretically, the inclusion of experience in TAM helps to reveal the variables that could influence doctoral students' adoption of ChatGPT in EFL writing.As our model explained 80.1% of the variance in behavioural intention, this study overall supports and advances the applicability of TAM in ChatGPT, a new technology in writing education.
Practically, the results of the study could also generate useful implications for technology developers, policy-makers, writing teachers, and doctoral students to leverage ChatGPT for the teaching and learning of writing.Doctoral students' strong intention to use ChatGPT in writing suggest that ChatGPT may augment its function as an educational tool for writing in higher education.Considering the significant and strong effect of attitude on students' behavioural intentions to use ChatGPT in writing, it is of necessity for educational institutions, writing teachers, and technology developers to be aware of students' attitudes and increase their positive evaluation of and affective reactions towards using ChatGPT in writing.For instance, technology developer can make the usage of ChatGPT more innovative, enjoyable and interesting so as to create more positive attitudes and boost learners' intrinsic motivation to use ChatGPT in writing.Given the increasing concerns for information, ethical and learning risks associated with ChatGPT (e.g., Barrot, 2023;Dwivedi et al., 2023) and doctoral students' strong intention to use ChatGPT for writing, measures must be taken to mitigate such negative impacts of ChatGPT on doctoral students.For example, technology developers can strengthen the quality control of generated responses.Similarly, writing teachers need to provide trainings on effective, ethical and responsible use of ChatGPT in writing.Besides, perceived ease of use and perceived usefulness are found to have a significant influence on students' attitude, which could further exert an effect on students' intentions to use ChatGPT in writing.The sequential and circular influential relationship among the variables implies a need for technology developers to increase the usefulness and ease of using ChatGPT in writing to make it more functional and user-friendly.For Given the significant effect of past ChatGPT experience on perceived ease of ease, instructing doctoral students to increase their use of ChatGPT, and reflect upon and communicate the skills for utilizing ChatGPT to promote writing performance could be an effective way to develop their expertise in ChatGPT.Also, doctoral students can experiment with ChatGPT in a conscious manner, and record their hands-on experience to continuously improve the capability for effective and ethical use of ChatGPT for writing.Regardless of the contributions, there are several limitations that need to be taken into consideration in future research.Firstly, while the study revealed a high intention to use ChatGPT in writing among doctoral students, it was exploratory in nature and only used questionnaires to gauge students' acceptance of ChatGPT.Future research can thus employ case study research deign or mixed study research design and collect multiple sources of data (e.g., semistructured interviews, user reflections, and screenshots) to obtain an idiosyncratic and in-depth understanding of students' actual process and outcome of using ChatGPT in writing.Secondly, the present study was based on a sample of doctoral students from a science and technology university in China.The types of writing assignments they face and their needs for using ChatGPT to improve writing could be very different from other learner groups like undergraduates (Yan, 2023) and students in other countries, which limits the generalizability of this study.Therefore, future research can expand the sample scope to include students with varied educational levels and backgrounds to increase the generalizability and representativeness.It may also be interesting to conduct cross-section research to examine whether the level of use acceptance across different learner groups in the future.Thirdly, our data was collected from participants who interacted with ChatGPT shortly after the release of ChatGPT and who used ChatGPT primarily for its inherently enjoyable and interesting experience (Taecharungroj, 2023;Tlili et al., 2023).Given the increasing ethical, learning and information concerns concerning the use of ChatGPT in writing in academia (Barrot, 2023;Su et al., 2023) and students' growing experience, knowledge and skills regarding ChatGPT, their attitudes, perceptions and intentions of using ChatGPT in writing may alter over time.Longitudinal research can be conducted to trace the development of knowledge concerning the use of ChatGPT for writing among doctoral students, and how such knowledge influences their attitudes towards, as well as perceptions and intentions of using ChatGPT in writing.Considering the doctoral students' high intention to use ChatGPT for writing and the increasing concerns for information, ethical and learning risks associated with ChatGPT (e.g., Barrot, 2023;Dwivedi et al., 2023), it is also promising to explore effective ways to integrate ChatGPT in writing instruction and construct writing models to empower students to collaborate with ChatGPT in an effective, ethical and responsible manner.
total number of 242 doctoral students (151 males and 91 females) participated in the study through convenience samplings in one technological university in China.The students, ranging from 24 to 43 in age, were enrolled in the compulsory course entitled Writing for Academic Success taught by the first author.The course aims to empower doctoral students to improve English for academic writing 10.3389/fpsyg.2023.1259531 A skills.The participants were from different disciplinary backgrounds, such as computer science, mechanical engineering, materials science, economics, and education.
TABLE 1
Results of descriptive statistics and correlation analysis.
TABLE 2
Results of mediation analysis. | 2023-10-28T15:19:09.734Z | 2023-10-26T00:00:00.000 | {
"year": 2023,
"sha1": "07df9048aa2307a581eeb82843a7c133ef36c2e5",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "PubMedCentral",
"pdf_hash": "0eff1fe9342b442e908477264ca714bc1a3309c0",
"s2fieldsofstudy": [
"Education",
"Computer Science"
],
"extfieldsofstudy": [
"Medicine"
]
} |
53629374 | pes2o/s2orc | v3-fos-license | Non-Perturbative Quantum Mechanics from Non-Perturbative Strings
This work develops a new method to calculate non-perturbative corrections in one-dimensional Quantum Mechanics, based on trans-series solutions to the refined holomorphic anomaly equations of topological string theory. The method can be applied to traditional spectral problems governed by the Schr\"odinger equation, where it both reproduces and extends the results of well-established approaches, such as the exact WKB method. It can be also applied to spectral problems based on the quantization of mirror curves, where it leads to new results on the trans-series structure of the spectrum. Various examples are discussed, including the modified Mathieu equation, the double-well potential, and the quantum mirror curves of local $\mathbb{P}^2$ and local $\mathbb{F}_0$. In all these examples, it is verified in detail that the trans-series obtained with this new method correctly predict the large-order behavior of the corresponding perturbative sectors.
Introduction
It has been known for some time that certain string theories may be encoded in simple quantum mechanical models. This has led to a very fruitful interaction between string theory (and its close cousin, supersymmetric gauge theories) and Quantum Mechanics, as illustrated by [1,2]. It was recently pointed out that this connection is more general than previously thought. More precisely, it was argued in [3] that the all-orders WKB approximation in generic one-dimensional quantum systems is encoded in the so-called refined holomorphic anomaly equations of topological string theory [4][5][6]. This makes it possible to calculate the quantum periods (also known as Voros multipliers [7]) to all orders, by using both modularity and the direct integration of the holomorphic anomaly equations [8,9]. In the case of polynomial potentials, there is a physical reason for the connection between the WKB method and the holomorphic anomaly equations [10,11]: it turns out that these quantum-mechanical models can be engineered in terms of Seiberg-Witten (SW) theories with gauge group SU (N ) (where N corresponds to the degree of the potential), in the NS background [1], and in a particular scaling limit near the Argyres-Douglas point in moduli space. Since the holomorphic anomaly equation governs the quantum periods of supersymmetric gauge theories in this background [12][13][14], they are inherited by the quantum-mechanical model.
So far, the results in [3] are purely perturbative: the expansion in the WKB method is treated as a genus expansion in topological string theory, and obtained, order by order, by using the recursive structure of the holomorphic anomaly. It is now natural to ask whether the connection between quantum mechanical problems and the holomorphic anomaly goes beyond perturbation theory, and can be used to compute non-perturbative effects in Quantum Mechanics. In other words, is it possible to extend this connection to quantum mechanical trans-series 1 including exponentially small corrections? The key ingredient for such an extension would be a generalization of the refined holomorphic anomaly equations to the realm of trans-series.
The standard anomaly equation governing the conventional topological string has already been generalized non-perturbatively in [17,18]. This has led for example to a very precise determination of the large-order behavior of the genus expansion for topological string theory on toric Calabi-Yau (CY) threefolds. More recently, it has been used in [19] to provide a semiclassical decoding of a recent proposal for a non-perturbative topological string partition function [20]. Our first goal in this paper is to generalize the work of [17,18] to the refined topological string. This leads to a general trans-series solution for the refined topological string free energy. Since we want to make contact with simple quantum mechanical problems, we focus on the so-called Nekrasov-Shatashvili (NS) limit [1]. The trans-series solution of the refined holomorphic anomaly equations should correspond, in this limit, to the non-perturbative sector of Quantum Mechanics.
In order to test this idea, we look at two different types of problems. The first type consists of conventional quantum mechanical models in one dimension, involving the Schrödinger operator. Extensive work since the late 1970s has led to a good understanding of the trans-series structure in this type of examples, by using exact versions of the WKB method [7,[21][22][23], multi-instanton calculations in Quantum Mechanics [24][25][26], or the uniform WKB approximation [27][28][29]. We start our analysis by looking at the modified Mathieu equation, which is closely related to the NS limit of the Ω-deformed SW theory [1,30,31]. The study of the refined holomorphic anomaly in this model is best implemented by using modular forms, as first pointed out in [8]. Interestingly, the study of trans-series solutions requires an extension of this framework to deal with exponen-tially small corrections. We propose an extended modular ring which captures the properties of the trans-series, and we test our results against the large-order behavior of the non-holomorphic extension of the quantum periods. Based on this analysis, we find a universal form for the oneinstanton correction at all orders in . We also determine a higher instanton solution of the holomorphic anomaly equations which agrees with the result for the trans-series in Quantum Mechanics, in the holomorphic limit. The structure found for the modified Mathieu equation extends straightftorwardly to other canonical examples in Quantum Mechanics, like the double well and the cubic potential studied in [3]. These results indicate that the correspondence between quantum mechanical problems and the refined holomorphic anomaly equations found in [3] extends to the non-perturbative realm.
The power of a new method should also be measured by its ability to go beyond what is already known. Our new non-perturbative method in Quantum Mechanics shows its true strength in a second type of problems, namely, the spectral problems associated to quantum mirror curves. These spectral problems involve difference, rather than differential, equations, and they have been intensively studied in the last few years. In spite of this, very little is known about their trans-series structure. Our method gives a concrete tool to calculate non-perturbative trans-series for the all-orders WKB expansion in this problem. This leads again to predictions for the large-order behavior of the WKB series, which we test in detail in the case of local P 2 . In addition, we can deduce from our results the full one-instanton trans-series for the eigenvalues of the difference equation. The systematic, perturbative expansion of these eigenvalues has been recently studied in [32], and we use our results to predict and test the asymptotic behavior of this expansion in the case of local F 0 .
The all-orders WKB method produces an asymptotic expansion in powers of 2 , which is sometimes called the quantum volume. In the case of quantum mirror curves, the relation to topological strings has led to a plethora of exact results which in particular promote the quantum volume to a true function. It is then natural to ask what is the relation between the asymptotic expansion and the exact result. We perform this analysis in the case of local P 2 and we find that, as in previous related examples [19,33], the asymptotic expansion of the quantum volume is Borel summable but its Borel resummation does not agree with the exact result. This opens the way for a "semiclassical decoding" of the exact quantum volume in terms of a full trans-series, which we leave for future work. This paper is organized as follows. In Section 2 we review the connection between the all-orders WKB method and the refined holomorphic anomaly equations, and we study transseries solutions to these equations in the NS limit. Sections 3, 4, 5 and 6 are devoted to detailed discussions of examples. In Section 3 we consider the modified Mathieu equation. We present the trans-series solution to the corresponding holomorphic anomaly equations in terms of modular forms. We test this solution against the large-order behavior of the quantum free energies, and we also check that it reproduces known results in Quantum Mechanics. Section 4 does a similar analysis for another important example in Quantum Mechanics: the double-well potential. Sections 5 and 6 are devoted to examples based on quantum mirror curves: the case of local P 2 and the case of local F 0 , respectively. In the first case, we do a comparison between the Borel resummation of the quantum free energies and the exact answer obtained from topological string theory. In the second case, we use our results on trans-series to predict the large-order behavior of the perturbative series for the energy levels worked out systematically in [32]. We conclude in Section 7 and list some open problems. Two Appendices contain supplementary material on the topics discussed in the paper. In the first Appendix we present the formulation of the refined holomorphic anomaly equations of [6,34] in terms of a master equation, while in the second Appendix we present a large-order test of the trans-series obtained in Quantum Mechanics for the modified Mathieu equation.
The all-orders WKB method
In this paper we will be interested in spectral problems in one dimension. One general strategy to attack these problems is to use the all-orders WKB method [35] (see for example [36,37] for clear presentations). In this method, one defines a formal power series in 2 which we will call the quantum volume, whose classical limit is the volume in phase space defined by a maximal energy E. Let us review how this quantum volume is defined in the case of Schrödinger operators. We start with the Schrödinger equation for a one-dimensional particle in a potential V (x), where we have set the mass m = 1. Let us consider the standard WKB ansatz for the wavefunction, which leads to a Riccati equation for Q(x), We now solve for the function Q(x) as a power series in : If we split this formal power series into even and odd powers of , we find that Q odd (x) is a total derivative, and the wavefunction (2.2) can be written as Let us now consider the Riemann surface Σ defined by We will restrict ourselves to curves of genus one. The turning points of the WKB problem are the points where p 2 (x) = 0, and correspond to the branch points of the curve (2.9). For a curve of genus one there are only two independent one-cycles A and B encircling turning points. The A-cycle corresponds to an allowed region for the classical motion, while the B-cycle corresponds to a forbidden region. The period of the one-form y(x)dx along the A-cycle gives the volume of the classically allowed region in phase space, By using the formal power series in (2.6), we define the quantum volume as a formal power series in 2 , The all-orders WKB quantization condition is then One important question in Quantum Mechanics (asked e.g. in [38]) is whether there is a welldefined function of E and whose asymptotic expansion in , at fixed E, coincides with (2.11).
In other words, is there a non-perturbative definition of the quantum volume? Surprisingly, such a definition is not known in general. For the Schrödinger equation with polynomial potentials, Voros has constructed exact quantum volumes for states of definite parity, defined as fixed points of a functional recursion (see e.g. [39,40]). In the context of integrable systems, the quantum volume can be obtained from the so-called Yang-Yang function. The Yang-Yang function can be explicitly written down in some cases via the relation to supersymmetric gauge theory [1] or to topological string theory [20,[41][42][43][44]. In all those cases, the quantum volume is a well-defined function and its asymptotic expansion agrees with (2.11). We can use the integrals of the differential y(x)dx around the two cycles A and B of the curve (2.9) to define the classical periods, with appropriate choices of branch cuts for the function y(x). It is useful to introduce the prepotential or classical free energy F 0 (t) by the equation We should regard t as a flat coordinate parametrizing the complex structure of the curve (2.9).
If we now use the quantum-deformed differential P (x)dx, we promote the classical periods to "quantum" periods, where we have introduced the variable ν = t( )/ . Both quantum periods are defined by formal power series expansions in 2 , and the leading-order term of this expansion, which is obtained as → 0, gives back the classical periods. Note that the quantum A-period is proportional to the quantum volume, and the all-orders WKB quantization condition reads The quantum B-period defines the quantum free energy as a formal power series in 2 , where each coefficient is a function of the full quantum period ν, This quantity is the analogue of the NS free energy in N = 2 supersymmetric gauge theories and topological string theory. In writing the quantum free energy and the quantum A-period we have already made a choice of "frame", but as it is well-known from SW theory, there is an infinite number of frames related by duality transformations. We will see in the examples below that, in some cases, it is convenient to use different frames to perform the analysis. The other type of spectral problems that we want to address in this paper are obtained by quantization of mirror curves to toric CY threefolds. In the genus-one case, mirror curves can be writen as O(e x , e y ) + κ = 0, (2.18) where O(e x , e y ) is a polynomial in e x , e y . As explained in e.g. [20], we quantize the function O(e x , e y ) by promoting x, y to canonically conjugate Heisenberg operators x and y on L 2 (R), satisfying the commutation relation Ordering ambiguities are solved by Weyl's prescription. In this way we obtain a spectral problem of the form O(e x , e y )|ψ = −κ|ψ , (2.20) where |ψ belongs to the domain of the operator O(e x , e y ) inside L 2 (R). When working in the x representation for the wavefunctions, e y acts as a difference operator, and the spectral problem (2.20) can be written as a difference equation. It has been proved in [45,46] that, in many cases, the operators O(e x , e y ) have a discrete spectrum (more precisely, their inverses are trace class operators in L 2 (R)). On the other hand, difference equations can also be solved with the WKB method [47], by using an ansatz of the form (2.2). The leading order term is given by where y(x) is the (multivalued) function defined by (2.18). As shown in [12,48], this WKB analysis makes it possible to define quantum periods, similarly to what we have discussed in the context of the Schrödinger equation. One of the quantum periods defines the quantum volume, as in (2.11), and this in turn makes it possible to write down a perturbative quantization condition of the form (2.12). One can also define a quantum free energy, as in (2.15). It was argued in [12,48], and further tested in [49,50], that the quantum free energy obtained with the WKB method agrees with the NS limit of the refined topological string free energy for the corresponding CY manifold. The refined topological string free energy satisfies a generalized or refined set of holomorphic anomaly equations [5,6] which extend the original construction in [4]. In particular, in the case of the difference equations (2.20), the quantum free energy defined by the WKB method satisfies the NS limit of the equations in [5,6]. In [3], evidence was given that the quantum free energy of many one-dimensional Schrödinger problems also satisfies these equations, even when the spectral problem is not related to any known supersymmetric gauge theory or topological string theory. It was then conjectured in [3] that the connection between the WKB method and the holomorphic anomaly equations should be valid for general one-dimensional problems 2 . This conjectural connection, if true, provides a unified framework to study spectral problems coming from Schrödinger operators and from quantum mirror curves. We will now review in some detail the refined holomorphic anomaly equations and we will study their trans-series solution.
The refined holomorphic anomaly
We consider the B-model refined topological string on a local CY manifold described by a Riemann surface Σ. This could be of the form (2.9), as arising in ordinary Quantum Mechanics, or a mirror curve defined by an equation like (2.20). The moduli space of complex structures on the CY manifold is a special Kähler manifold with metric G km . In the case of a local CY manifold built upon a Riemann surface Σ, the metric is related to the period matrix τ of Σ by where β is a real normalization constant. We will denote the corresponding covariant derivatives by D i . The prepotential F 0 of this special Kähler manifold is precisely the function of the moduli defined by (2.14). The Yukawa couplings C ijk are defined by and we can fix β so that An important quantity entering into the formalism is The basic quantities in refined topological string theory are the perturbative free energies F (g 1 ,g 2 ) (t i ), with g 1 , g 2 ≥ 0 (see for example [5,31,52,53] for more details). The total free energy is a function of two parameters, 1,2 , and it is defined by the asymptotic expansion There are two important limits of this quantity. The first one corresponds to 1 = − 2 = g s . In this limit, only the perturbative free energies with g 1 = 0 contribute, and we recover the standard topological string with string coupling g s . The second limit is the so-called NS limit [1], in which we take 1 = 0. More precisely, we define the quantum or NS free energy by the limit It has the asymptotic expansion and we shall denote for simplicity The refined topological string energies can be computed with many different techniques: instanton calculus [31], the refined topological vertex [52,54], BPS invariants [53,55], and, in the case of the NS limit, the WKB method mentioned above [12,48]. Another powerful technique is based on the refined holomorphic anomaly equations. These equations exploit the fact that the refined free energies can be promoted to non-holomorphic functions of both the moduli t i and their complex conjugatest i , F (g 1 ,g 2 ) (t i ,t i ). The refined holomorphic anomaly equations govern the anti-holomorphic dependence of these functions, and they read These equations are valid for g 1 + g 2 ≥ 2. In the standard topological string limit, with g 1 = 0, one recovers the BCOV holomorphic anomaly equations [4]. In the NS limit, the first term in the r.h.s. of (2.30) drops out, and we obtain the simplified equations In this paper we will focus on the case in which Σ has genus one, so that the moduli space of complex structures has dimension one. It can be parametrized by the elliptic modulus τ , which is related to the prepotential by In this case, the anti-holomorphic dependence of the free energies can be encoded in a single function, usually called the propagator S, defined by ∂tS =C tt t . (2.33) Here, we have denoted the single index by t, which refers to the modulus of the CY. By using the propagator, we can write the refined holomorphic anomaly equations in the case of curves of genus one as [5,6], where ∆ is essentially the discriminant of the curve Σ (it can contain additional functions of the moduli). Using (2.36) as the initial condition, as well as the special geometry of the moduli space, the holomorphic anomaly equations (2.31) determine the functions F n recursively, up to a purely holomorphic dependence on the moduli which is usually called the holomorphic ambiguity. When using the refined holomorphic anomaly to compute the quantum free energies, it is important to take into account an important subtlety. The argument t of the functions F n (t) is, as it should, the full quantum period t( ) appearing in (2.15). The anomaly equations typically give the F n s as functions of the complex modulus of the curve-parametrized by the elliptic modulus τ or by the complex parameter appearing in the equation of the curve-which we will denote by z. However, the relation between z (or τ ) and t( ) is the one determined by the classical period [5,6].
As first noted in [8,9], in the case of curves of genus one, it is very useful, both conceptually and computationally, to re-express the anomaly equations in the language of modular forms. This leads to convenient parametrizations of the free energies and their holomorphic ambiguities, and to a fast symbolic (computational) implementation of the recursion. When the Riemann surface has genus one, there is a single Yukawa coupling, which we will denote by (2.37) To satisfy (2.33), the propagator S can be written as the non-holomorphic modular form [8] S = − β 12 E 2 (τ,τ ), (2.38) where E 2 (τ,τ ) is defined by and E 2 (τ ) is the weight-two Eisenstein series. It is also very useful to introduce the Maass derivative acting on (almost-holomorphic) modular forms of weight k, . (2.40) The refined holomorphic anomaly equations read in this case, in the NS limit, where the coefficients f n,r are holomorphic in τ . Since F n has modular weight zero for n ≥ 1, the coefficients f n,r have weight −2r. The coefficient f n,0 is the holomorphic ambiguity. As first explained in [8], one can determine the holomorphic ambiguity by first finding an appropriate parametrization in terms of modular forms, which depends on the particular curve under consideration. The holomorphic ambiguity is written in this way as an unknown linear combination of known modular forms. To determine the coefficients in this linear combination, one imposes boundary conditions at special loci in the moduli space of the curve. This method has been used in e.g. [9,13,56]. We will see concrete implementations of this procedure in the examples of this paper.
Trans-series solutions of the refined holomorphic anomaly equations
As set-up originally in [4], the holomorphic anomaly equations are inherently perturbative. Due to the recursive nature of these equations, it is natural to ask whether they can have trans-series solutions which capture non-perturbative effects. This was answered in the affirmative in [17], and further developed and exemplified in [18,19,57,58]. In this section we will extend some aspects of [17,18] to the refined case, focusing on the NS limit which is relevant for quantum mechanical problems. The first step in constructing the trans-series solution is to obtain a "master equation" for the total free energy (2.27). Such an equation is given by and It is easy to see that this reproduces the recursion (2.35). A more general master equation can be written away from the NS limit, which should provide the starting point for an analysis of general trans-series in the refined topological string. We present this master equation in Appendix A. We now postulate a trans-series ansatz for the solution of this master equation, rather than the perturbative series (2.28). The simplest ansatz is a one-parameter trans-series with an infinite number of exponentially small corrections, as the one already used in [17,18] for the standard anomaly equation of [4]. We will write (2.46) In this equation, A is the instanton action, σ is a trans-series parameter, and F (n) ( ) denotes the perturbative expansion in around the n-th instanton. All multi-instanton sectors are themselves asymptotic series (as the perturbative series (2.28) already was) where b (n) is a "characteristic exponent" or "starting genus". We now proceed as in [17], i.e. we insert the trans-series (2.46) into the master equation (2.43). One recovers the perturbative NS-limit holomorphic anomaly equation (2.35) for the perturbative coefficients in (2.28), alongside a new non-perturbative extension for the nonperturbative coefficients in (2.47). At first non-trivial non-perturbative order one obtains i.e. the instanton action is holomorphic (since the anti-holomorphic dependence is all contained in the propagator). This is in complete analogy with what happens in the standard topological string case [17]. Note, however, that this condition only occurs if the combination of starting genera is strictly positive 3 , B nm > 0 (this already occurred in the standard topological-string case; see [17] for further details). We shall proceed under this assumption. The holomorphicity condition (2.48) puts little restrictions on the actual value of the action. However, one expects that it is a period of the CY manifold, as it was postulated in [59] based on previous insights [38,60,61]. By explicitly using holomorphicity of the instanton action, the remaining terms give us the non-perturbative extension of the NS-limit holomorphic anomaly equations. We find, In these equations, the D (n) i are operators defined as follows: (2.51) Note that the operator in the first line is a multiplicative operator involving no derivatives. The equations (2.50) become very simple for the first instanton sector. We find, and For example, for k = 1, 2 we obtain, (2.54) 3 Otherwise there will be extra (non-trivial) constraints.
There is a slightly simpler master equation which we will also use in the following. Let us define Then, the holomorphic anomaly equation can be written as which can be also solved with a trans-series ansatz, as we will see in our examples.
One of the signposts of resurgence is that higher instanton corrections "resurge" in the largeorder behavior of lower instanton series (see e.g. [15,16]). In the case we are considering here, the perturbative series is the series of NS free energies (2.28). We expect the large-order behavior of this series to be controlled by the first instanton series F (1) . More precisely, we expect the leading double-factorial behavior A is the smallest action in absolute value, and the coefficients µ n are given by the loop corrections in the one-instanton sector, In this equation, Σ is a Stokes parameter which has to be determined in each problem.
In the rest of this paper, we will study the trans-series solution of the NS limit of the refined holomorphic anomaly in various examples.
Examples in Quantum Mechanics: the modified Mathieu equation
We now apply the formalism developed in the previous section to various examples. The first one is the modified Mathieu equation, which has been studied in detail in the recent literature (see e.g. [62][63][64][65][66][67]), due to its connection to SW theory and its quantum deformation [1,31].
WKB analysis and the refined holomorphic anomaly
The modified Mathieu equation describes a one-dimensional, quantum mechanical particle in a cosh(x) potential. The classical Hamiltonian is and the corresponding Schrödinger equation is The spectral problem associated to the modified Mathieu equation leads to an infinite number of discrete energy levels, labelled by an integer n = 0, 1, 2, · · · . These energy levels can be computed by using the all-orders WKB method. For a given energy E, the turning points of the motion are given by ±x + , where The classical volume of phase space is given by the integral, where K(m), E(m) denote the complete elliptic integrals of the first and the second kind, as a function of the squared modulus m = k 2 . The all-orders WKB method gives an asymptotic expansion for the quantum volume, of the form (2.11). It is possible to calculate the very first orders of this expansion by using conventional techniques. One finds, for example, at the nextto-leading order, As explained in Section 2.1, the WKB method makes it possible to define quantum periods, as well as a quantum free energy. The quantum volume gives the quantum A-period. There is a quantum B-period associated to a cycle which goes around the imaginary axis, and is given by The classical limit of this period will be denoted by a(E), and it is given by The classical prepotential is defined by, where a = a(E) is the classical limit of the B-period. Note that, in this problem, we define the prepotential and the quantum free energy in terms of the A-period. This is because we are choosing here a particular frame, which we will call the "electric" frame. There is a dual, "magnetic" frame, obtained by performing an S transformation which exchanges the A and the B periods, and which is more useful to analyze the spectral problem, as we shall see. The prepotential can be computed at large a as This is, up to a choice of normalization, the prepotential of SW theory [30], and the Riemann surface underlying the Hamiltonian (3.1) is equivalent to the SW curve where u, the complex modulus of the curve, is related to the energy in (3.2) by This relation can be regarded as a consequence of the connection between SW theory and classical integrable systems [68,69], since (3.1) is the only non-trivial Hamiltonian of the classical N = 2 Toda lattice. The classical prepotential F 0 (a) can be promoted to a full quantum free energy by the equation where a is now the full quantum period in (3.6). The resulting quantity, is the NS limit of the Nekrasov free energy for SU (2), N = 2 supersymmetric Yang-Mills theory [1,12] (or, more precisely, its asymptotic expansion in powers of 2 ). Let us now introduce the "magnetic" quantum period (3.14) The corresponding "magnetic" or dual quantum free energy Then, the all-orders WKB quantization condition can be written as From this quantization condition one can derive in particular the perturbative expansion of the energy E = E(ν) as a function of the quantum number ν, by simply expanding the quantum period a D ( ) around u = 1, order by order in , and solving for u as a function of and ν. One finds in this way, This is precisely what one obtains in the standard perturbative analysis of the modified Mathieu equation, by using for example the BenderWu package [70].
In this quantum-mechanical problem, the connection to topological string theory and its geometric engineering limit indicates that the F n (a) can be computed by using the NS limit of the refined holomorphic anomaly equation. As noted in (2.36), the first correction to the classical prepotential is given by In order to obtain efficiently the higher order corrections F n (a), with n ≥ 2, we rewrite the holomorphic anomaly equations in terms of modular forms, as we explained in Section 2.2. The relevant modular forms are the ones associated to the SW curve, and discussed e.g. in [8]. They form a ring with generators where ϑ i (q) are the Jacobi elliptic functions, and E 2 (τ ) is the second Eisenstein series. These generators have modular weight 2. Their argument is where τ is the elliptic modulus and is related to the prepotential by We note that the Maass derivative (2.40) acts on the above generators as (3.25) As usual in SW theory, we have to specify the electric-magnetic frame for the calculation of the quantum prepotential. We will reserve the notation F n for the quantum free energies in the electric frame. When n = 0, 1, the free energies in the electric frame are given in (3.8), (3.19). Expressions in the magnetic frame can be obtained, in the language of modular forms, by an S-duality transformation We also introduce It will be also useful to introduce the quasi-modular forms, (3.28) The first one corresponds to the electric period a(E), while the second one is related to the B-period by They have the following expansions in terms of the exponentiated modulus, The free energies in the magnetic frame F D,n , appearing in (3.15) can be obtained from the F n by an S-transformation, and depend on the variable a D . We recall that the holomorphic anomaly equations will give the F n s as functions of the elliptic modulus τ or its dual τ D . To obtain them as functions of the full quantum periods a( ), a D ( ), we should use the classical equations (3.28) relating τ and τ D to a, a D . In order to use the anomaly equations (2.41) we have to specify as well the Yukawa coupling, which is given by We also need an appropriate parametrization of the holomorphic ambiguity. In this case, since the relevant ring of modular forms is generated by (3.20), we parametrize the ambiguity by where the a n,i are constant numbers. They are fixed by the following boundary conditions for the dual free energies, where B 2n are the Bernouilli numbers. In this way we find, for example, Higher order free energies can be easily computed recursively.
The free energy trans-series
We now proceed to compute the free energy trans-series from the extended holomorphic anomaly equations. We first introduce the covariant derivative w.r.t. the modulus a as We look for a trans-series solution of the holomorphic anomaly equations, involving exponentially small quantities of the form e −A/ . In order to use the formalism of modular forms, we have to take into account that the exponents in these quantities are instanton actions, given by periods, and are not modular invariant. We will now introduce a formalism which makes it possible to exploit modularity in spite of this fact. First of all, we enlarge the ring of modular objects as follows. Since E 2 does not transform as a modular form under an S transformation, we introduce the quantity so that i.e. E 2 , E D 2 provide a vectorial representation of the S transformation 4 . In addition, we introduce a degree-counting constant which transforms with weight one under S, and which we denote by ω 1 . In any evaluation it should be taken to 1. This leads to an enlarged ring with additional generators E 2 , E D 2 and ω 1 . To define the action of the Maass derivative on these additional generators, we note that D τ can be written as and we assign a weight −2 to τ , so that We can then calculate in a straightforward way, In addition, we postulate the following transformation of ω 1 under S-duality, Both the Maass derivative of ω 1 and its S transformation lead to a formalism with very useful properties. Let us introduce the actions, In addition, the introduction of ω 1 , together with its S transformation rule in (3.42) makes it possible to work out the dual expansion of the actions. For example, after an S transformation, we find, which after setting ω 1 = 1 leads to the following dual expansion, Similarly, one finds Now that we have an appropriate, enlarged modular formalism, let us look for a trans-series solution of the NS holomorphic anomaly equations. We will use the master equation (2.56) for the modified quantum free energy (2.55), and we consider the trans-series ansatz Here, F (1) is the one-instanton correction i.e. G is the exponentiated instanton action together with all quantum corrections around the one-instanton configuration. As we shall see, it is possible to determine G in a single strike. By plugging the ansatz (3.48), (3.49) in (2.56), we find Our goal is now to solve this equation for G. We know that G = A + · · · . Let us then consider the ansatz where S A has weight two and is holomorphic Our first observation is that ∂ E 2 and D a (or D τ ) do not commute in general: by looking at the algebra of extended generators (3.25) and (3.41), we find that, when acting on an object of weight k, This can be also written as We deduce that ∂ S and D a commute when acting on objects of zero weight. We have introduced ω 1 precisely so that actions have zero weight, therefore since actions are holomorphic, therefore independent of the propagator, as established in (2.48) (in this context, holomorphy of A means that it does not involve the generator E 2 of the extended ring). ∂ S and D a also commute on the weight zero object F (0) . We then calculate and we conclude that This equation relates S A and A. In fact, when the action A is one of the actions in (3.43), we can solve for S A as follows. As a consequence of the algebra (3.41), one has We conclude that (3.60) Finally, upon using (3.44), we find: This gives the full one-instanton correction after exponentiation. Let us note that the functions G A,B have a very non-trivial expansion. For G A , one finds, (3.63) By expanding the exponential, we can immediately calculate all the coefficients F k , k ≥ 1. We should mention that trans-series for the quantum free energies associated to the Mathieu equation have been computed in [64] by using WKB methods. They find solutions involving exponentially small corrections in the periods, as we have found above.
An immediate application of the above calculation is the determination of the trans-series for the quantum volume function, which in this case is given by (3.12). We can write it as calculated in the electric frame. The one-instanton trans-series associated to this is of the form For instance, for the B-period action we find
Application: large-order behavior
We have now determined the full one-instanton correction to the quantum free energy, for arbitrary values of τ andτ (which can be taken to be independent variables). This correction depends on a choice of instanton action, which corresponds to the A or the B period. Some ingredients in the answer, like the exponent b, are still undetermined.
which are obtained from (3.43) once we set ω 1 = 1 . The relative dominance of these actions depends on where we are in moduli space, as explained in [59]. As shown in Fig. 1, the A-period dominates the asymptotics at large u, while the B-period dominates the asymptotics at small u, and there is an exchange of dominance as we move from the strong-coupling region of small u to the weak coupling region of large u. We then expect that the full large-order behavior of the F (0) n is controlled by the one-instanton amplitudes associated to the A and the B periods, which in turn can be obtained by exponentiating the functions appearing in (3.61), respectively. The Stokes parameter Σ appearing in (2.59) can be extracted from the boundary behavior in (3.34), and turns out to be given by Note that the non-holomorphic F n depends on q,q through the modular form and we can regard q,q as independent variables. The trans-series also depends on q,q, and should control the large-order behavior for arbitrary values of these two variables.
In order to test the predictions (2.59) for the coefficients µ m , we can proceed as follows. By using the values of F (0) n and the predicted values for µ , = 0, · · · , m − 1, we consider the sequence where (x) n = x(x + 1) · · · (x + n − 1) is the Pochhammer symbol. This sequence should converge to µ m as n → ∞. In addition, we can accelerate the convergence of this sequence by using Richardson transforms (in the present context, see for example [16,71]). We will denote by µ which gives u = 1.492.... This is in the strong-coupling region, and the relevant one-instanton amplitude is associated to the B-period. The holomorphic limit is obtained whenτ → i∞, or equivalently whenq → 0. In this limit, and from the explicit value of G B , we find the predictions: The numerical value obtained for µ 5 by using the sequence of the F n up n = 45 and with 10 matching 12 digits with the resurgent prediction of µ 5 . This is a strong test of this and the previous coefficients, which were used as input for the numerics. In Fig. 2 we show the sequence (3.70) up to n = 45, its first Richardson transform, and the predicted value for µ 5 from the trans-series. Similar tests can be done for more general values ofq, and in all cases we find perfect agreement between the prediction obtained from G B and the actual large-order behavior. An interesting case occurs whenτ = 0. This corresponds to the free energies in the magnetic frame, F D,n , and in the holomorphic limit. In this case, G B = A B and the asymptotics is trivial, in the sense that This is precisely what is found in a numerical study of the large-order behavior. We have tested the predictions obtained from our analytic instanton results also in the region of large u, where the relevant instanton action is associated to the A-period.
Higher-order instanton corrections
In the previous section we have studied the one-instanton trans-series, but we expect higher instanton corrections. In this section we find a solution to the holomorphic anomaly equations describing multi-instanton corrections. First of all, we introduce the following multi-instanton ansatz for the function F introduced in (2.55): where each F (m) has the following structure: We note that We have already solved for φ (1) in Section 3.2, namely where we used the value of b = 2 obtained from the large-order analysis. After plugging it into the master equation (2.56), and using (3.51), we obtain, for m ≥ 1, The left hand side of this equation involves the operator that annihilates G in (3.51), which we will denote by It has the following properties: In order to solve (3.79) we have to "integrate" with respect to W . Let us first consider the m = 2 instanton correction. By using (3.78), we obtain from (3.79) the equation We recall from (3.58) that D a T = 0, (3.84) and by using (3.53) and (3.56) we find that Now we apply W on the following weight zero object, where we have used (3.51) and (3.84). We conclude that 0 in the kernel of W . The simplest solution compatible with the large-order behavior of the quantum free energies is that φ It is now clear that T always accompanies the D a derivatives so that the full object keeps the correct weight. Define Suppose now that X 0 has weight zero, so that ∂ S and D a commute when acting on it. Let us use (3.83) to write, Then the action of W can be written in terms of G, We then have the following commutator, and we can build the recursion with these elements. Let us look for example at the third-order instanton, with m = 3. The equation determining φ (3) is The building blocks to solve this equation can be obtained by using (3.93) and WG = 0, and we find (3.96) Therefore, Proceeding in this way, it is possible to calculate the m-th multi-instanton correction in terms of a set of constants f (1) , · · · , f (m) . In principle these constants should be related, since we expect a single trans-series parameter. In the case of the trans-series controlling the large-order behavior of the F n s, for example, the values of these constants can be found, in principle, by using explicit large-order results and resurgence relations. In the next section we will fix the values of these constants by comparing to results obtained in Quantum Mechanics.
Comparison with previous results in Quantum Mechanics
In conventional Quantum Mechanics one can find exact quantization conditions for the spectrum by using the exact WKB method [7,21,23], instanton calculus [24][25][26], or the uniform WKB approximation [27][28][29]. These quantization conditions are in fact equations defining implicitly a trans-series for the quantum period ν in (2.15). This leads, by a formal trans-series expansion, to a trans-series for any function of ν, like for example the energy E = E(ν). The exact spectrum is then obtained by applying Borel-Écalle resummation to the resulting trans-series. The trans-series obtained from exact quantization conditions are usually based on instanton solutions to the Euclidean EOM. However, there are no real instanton solutions for the cosh(x) potential, and one needs complex instantons [38] coming from classical trajectories along the imaginary axis in the complex x plane [72,73], where we have a periodic potential. One way to find the appropriate trans-series for the modified Mathieu equation is to start with the cos(x) potential (i.e., the Mathieu equation). In the cos(x) potential the quantization condition was obtained in [24][25][26] by using instanton calculus and derived in [29] by using the uniform WKB method. It reads 1 + e ±2πiν = f SG (ν) + 2 cos θ f SG (ν).
(3.98)
Here, θ is the quasimomentum, and f SG (ν) can be written as where A SG (ν, ) is a certain regular function of ν, . The ± sign in (3.98) corresponds to the choice of lateral resummation. To make contact with the modified Mathieu equation, we change → − in the function A SG (ν, ), and we eliminate the dependence in θ by taking θ = π/2. We end up with the equation, The function f (ν) turns out to be related to the derivative of the dual quantum prepotential introduced in (3.16), as follows: where a D ( ) and ν are linked by (3.17). We recall from (3.34) that ∂F D (a D , )/∂a D ( ) has a singular part at a D = 0. After exponentiation, the singular part gets resummed into the prefactor of f (ν) involving the Gamma function. A(ν, ) is then the regular part of this derivative. It has the expansion, The relation (3.100) defines a trans-series for the quantum period ν. To compute it explicitly, we write, as in [27,28], where ∆ν (k) ∝ e kA(ν, )/ (3.105) is a k-instanton contribution. From the equation (3.100) one finds (we pick the + sign for simplicity) and so on. In the following, we will rewrite f (ν) as . The extra 2/ comes from the relation between a D and ν. The trans-series for u(ν, ) (or, equivalently, for the energy) can be obtained by promoting the perturbative relation u = u(ν) to a trans-series relation, as explained in e.g. [27,28]. We obtain, u(ν + ∆ν) = n≥0 u (n) (ν). (3.108) The first three instanton corrections for u read, where we have denoted u = u (0) , u = ∂ ν u. To verify that (3.100) gives the correct trans-series for the energy, we have checked in detail that (3.108), (3.109) lead to the appropriate large-order behavior of the perturbative energy series and of the first instanton series; see Appendix B.
Let us now show that (3.108), (3.109) are compatible with the results obtained from the holomorphic anomaly equation. In the context of the anomaly equation, we obtain a transseries for the quantum free energy, so the comparison is easier to made if there is a functional relation between u and F (or F D ) which can be promoted to a trans-series relation. In fact, such an equation exists and it is often referred to as "perturbative/non-perturbative" or PNP relation 5 . PNP relations in one-dimensional quantum systems were first noted by G.Álvarez and his collaborators in a series of papers [27,28,74,75], and they have been generalized to other models in [29]. In the case of the (modified) Mathieu equation, the PNP relation coincides with the extension of the Matone relation [76] to the quantum NS limit [77], see for example [63,66,78]. In our notation, the PNP relation reads: Let us now search for the appropriate trans-series of the quantum free energies which leads, through (3.110), to (3.108), (3.109). In view of (3.16), the "classical" action should be Now we need the G function corresponding to this action. Since f (ν) was written in terms of F D , we should write it in the magnetic frame. From (3.61), we find To go to the magnetic frame, we perform an S transformation. Since ω 1 has weight one, it also transforms, and we obtain Writing it in the form, (3.115) The holomorphic limit in the magnetic frame corresponds toτ D → i∞. Therefore, in this limit, The covariant derivative also gets an i factor in the magnetic frame (this is due to the fact that, under an S transformation, a goes to −ia D ). Therefore, in this frame, where we have used the relation (3.17). The G function becomes This is precisely what is needed. Let us now denote by F (k) D the instanton corrections obtained in Section 3.4 in the magnetic frame, and with the above choice (3.118) for the function G. According to (3.110), we have We have verified that this relation holds true for k = 1, · · · , 5 with the following choice of the constants in the holomorphic ambiguity: (3.120) As an example, let us consider the three-instanton free energy. From (3.97) and (3.119) we get
(3.121)
This reproduces precisely the last line in (3.109), once (3.120) is used. We conclude that our trans-series solution of the holomorphic anomaly equations not only leads to the correct largeorder behavior of the quantum free energies, but it also reproduces correctly the trans-series obtained from exact quantization conditions in Quantum Mechanics. In the next section we will see more examples in quantum-mechanical models.
The double-well potential
The double-well potential in Quantum Mechanics is given by We will set λ = 1. A detailed analysis of the all-orders WKB method in terms of the refined holomorphic anomaly was already presented in [3]. Here we summarize some of the results. The classical A-period t corresponds to the allowed region, while the classical B-period t D = ∂ t F 0 corresponds to the tunneling region between the wells. They define together a classical prepotential F 0 (t). The modulus τ of the corresponding elliptic curve satisfies: and the energy is related to τ by the relationship where K 2 , L 2 are the modular forms introduced in (3.20). From (4.2) we deduce that β = 1/2, therefore The direct integration of the resulting holomorphic anomaly equations, in the NS limit, makes it possible to calculate the functions F n systematically, as shown in [3]. We can now use the techniques developed in this paper to obtain the one-instanton transseries. Since quantum-mechanical problems associated to genus-one curves have the same structure, the one-instanton correction is still given by the general solution (3.49), where G is given in (3.52). Equivalently, we can write it as (3.83), where T is given by (3.82). The only ingredient that changes is the parameter β appearing in (2.38), which does not affect the derivation of (3.83).
Let us now find the trans-series responsible for the large-order behavior of the quantum free energies in the double well. There are two relevant instanton actions, given by the periods In Fig. 3 we plot their absolute value as a function of q 2 , where we clearly see their regions of dominance. By using the explicit formulae for A, we obtain from (3.82): (4.6) The normalization of the trans-series can be determined by the singular behaviour at the conifold points identified in [3], which correspond to the energies E = 0, E = 1/32. One finds, (4.7) By plugging now the above results in (3.49), we find, for the instanton correction associated to the A A action, while for the A B action, (4.9) We have tested the above expressions systematically, in different regions of dominance for the instanton actions, and for different values of q,q (not necessarily complex conjugate values). Let us give an example, corresponding to the values q = 1/2,q = 1/4. We show the convergence to this value in Fig. 4.
Comparison with the exact quantization condition
As in the case of the modified Mathieu equation, we can compare the trans-series obtained with the refined holomorphic anomaly equations, with the trans-series obtained from the exact quantization condition in Quantum Mechanics. The exact quantization condition for the doublewell potential was first obtained in [24] with instanton techniques, and then derived in [7] with the exact WKB techniques of Voros-Silverstone [21,22] 6 . For us, the most convenient form for this quantization condition is the one derived by G.Álvarez in [27] by using the uniform WKB method. It reads, 1 + e 2πit( )/ = i e −t D (t, )/2 . (4.13) In this equation, t( ) = ν is the quantum A-period, takes into account the parity of the states, and t D (t, ) is the quantum B-period, re-expressed in terms of the quantum A-period. From now on we will set = +1 for simplicity. Note that (4.13) is very similar to the equation (3.100) appearing in the context of the modified Mathieu equation, and it can be also used to define a trans-series for ν (or equivalently, the quantum A-period t), as we did in (3.104), (3.106). Any function of ν, like the energy, gets promoted to a trans-series as it happened in (3.108). We find, similarly to (3.109), (4.14) where we have denoted t D = ∂ t t D (t, ) and E = ∂ t E(t, ).
Let us now show how the result above can be reproduced by using the trans-series for the free energy. The only ingredient we need is the PNP relation for the double well obtained in [27], which we write in the form where ξ is an appropriate constant. Suppose now that we choose an instanton action proportional to the B-period, and its associated one-instanton trans-series (3.49). The function T is given by (3.82), and one obtains (see e.g. the second equation in (3.60)) In the electric frame, the non-holomorphic E 2 becomes simply E 2 , and with (2.32) T e = α, (4.18) while the G function becomes With the value α = 1/2 for the double-well problem, we find By using (3.89), this also means that By using these results, we can verify that the multi-instanton results obtained in Section 3.4 reproduce the results obtained from the exact quantization condition. Let us take for example the m = 2 instanton correction, (3.88). Following the PNP relation, the corresponding correction to the energy should be given by 4.22) and Using values of the constants similar to (3.120), we get precisely what appears in (4.14). We have verified the agreement up to m = 5 (the five instanton correction).
All the results obtained in this section for the double well can be extended to the cubic oscillator studied in [3]. In that case one has that β = 1, and the relevant instanton action is also of the form (4.16) with α = 1. The function G in the electric frame is also proportional to the quantum B-period. One can also check that the multi-instanton series obtained in Section 3.4 reproduce the trans-series obtained from the exact quantization condition obtained in e.g. [7,74], provided the constants f (m) take the value (4.26)
Examples of quantum mirror curves: local P 2
The examples analyzed so far involve Schrödinger operators from Quantum Mechanics. We have seen that the trans-series obtained from the refined holomorphic anomaly give us new results for the asymptotics of the quantum free energies. These results are compatible with the trans-series obtained with standard techniques in Quantum Mechanics. In this section we will consider the spectral problems associated to quantum mirror curves (see [2] for a review and references). In these problems there are conjectural exact quantization conditions [20,42] which determine the spectrum of these operators. This creates the opportunity to compare these exact results with the results obtained with approximation schemes: the all-orders WKB expansion and the standard perturbative expansions [32,34], as well as their trans-series extensions. In this and the next section, we will study the spectral problem for two different toric CY manifolds, local P 2 and local F 0 , in the all-orders WKB expansion, through the refined holomorphic anomaly equations. We will also calculate the corresponding trans-series. For these spectral problems, there are no trans-series results in Quantum Mechanics to compare with, so the holomorphic anomaly gives the only concrete approach to understand their resurgent structure.
Refined holomorphic anomaly, trans-series and large-order behavior
In the case of the local P 2 geometry, the corresponding quantum-mechanical operator is O P 2 = e x + e y + e −x−y . (5.1) It was conjectured in [20] and then proved in [45,46] that this operator has a discrete spectrum and its inverse ρ = O −1 is trace class. Although these operators are not of Schrödinger type, and they lead to difference equations instead of differential equations, one can still use the all-orders WKB approximation [47], as it has been done in [34,48,50,80]. The associated Riemann surface is just the mirror curve of local P 2 , which has the form e x + e y + e −x−y + κ = 0. (5. 2) The calculation of the classical volume reduces to the calculation of classical periods on this curve. We will parametrize the moduli space with the coordinate z, which is related to κ by The standard, classical periods in the large-radius frame (which is appropriate for the point z = 0) are given by −t = log(z) + 1 (z), After integration, we find the prepotential Then, a simple calculation shows that (see for example [5,20] for more details) where the F 0 symbol means, as in [20], that we changed e −t → −e −t in the exponentially small corrections appearing in the expansion (5.6). The relation between t and E is given by Explicitly, one finds [34] vol The most efficient way to calculate the higher-order corrections to the quantum volume is to use the refined holomorphic anomaly equations. To set up these equations, we proceed as in e.g. [5,13,18,81]. We will use as our global coordinate the modulus z introduced in (5.3). We also need some preliminary ingredients from special geometry. We introduce the discriminant of the curve (5.2), ∆ = 1 + 27z, (5.10) the Yukawa coupling, 11) and the standard topological string genus-one free energy, The holomorphic limit of the propagator S, in the large-radius frame, is given by the equation where the superscript indicates that S is calculated in the large-radius frame. One finds, explicitly [18], (5.14) With these ingredients one can already solve the refined holomorphic anomaly equations in the NS limit, (2.35). The initial condition for the recursion is the value of F 1 [5], The holomorphic ambiguity is fixed by imposing appropriate boundary conditions. As usual, the holomorphic quantum free energies F n can be computed in different frames, and when needed we will indicate such a frame by a superscript. In the conifold frame, and near the conifold singularity at z = −1/27, the quantum free energies satisfy the gap condition [5,8,13] where t c is the flat coordinate at the conifold, given by Using all this information, it is straightforward to calculate the F n at very high order in n. One finds, for example, In order to make contact with the quantum volume, we note that at higher orders in , the relationship between t and E given in (5.8) gets quantum corrections, and one needs the socalled quantum mirror map t(E, ) [48]. From the point of view of WKB theory, the quantum mirror map just encodes the quantum corrections to the A-period, t. In this case, the quantum mirror map has the form t(E, ) = 3E − 3 q 1/2 + q −1/2 e −3E + · · · , q = e i .
(5.19)
The all-orders perturbative quantum volume is then given by where the F LR n , with n ≥ 1, are obtained by taking the holomorphic large-radius limit of the F n , changing e −t → −e −t in the exponentially small corrections, and then relating t to E via the quantum mirror map (5.19).
The first question involving trans-series that we can ask is: what is the large-order asymptotics of the series of quantum free energies F n ? General resurgence results predict that the asymptotics should be of form (2.57), with the relation (2.59). Since z = −e −3E + O( 2 ) is naturally negative for this problem, we will focus on negative values of z. A similar problem, concerning the large-order asymptotics of the standard topological string free energies of local P 2 , was studied in detail in [18]. The instanton action that controls the asymptotic behavior depends on the point where we are in moduli space. We will perform the analysis in a region in between the conifold point and the large-radius point, i.e.
It turns out that, in this region, the asymptotics of the F n is controlled by the action where the conifold coordinate t c has been defined in (5.17). This is also the action controlling the asymptotics of the standard topological string free energies in this region, as found in [18]. Let us now determine the trans-series associated to this action. We will use the equations (2.50) to determine the trans-series at the one-instanton level, F n . We will parametrize the moduli space with the coordinate z. We will also need boundary conditions in order to fix the ambiguities. To do this, we proceed as in [18] and we note that, in the conifold frame, we have the behaviour (5.16). By using B 2n = (−1) n−1 2(2n)! (2π) 2n 1 + 4 −n + · · · (5.23) this determines b = 2 (5.24) and the large-order coefficients (2.57) in the conifold frame, Let us now analyze the equations for the trans-series. First of all, according to (2.52), F (1) 0 is holomorphic and has no propagator dependence. Therefore, this quantity does not depend on the frame and it can evaluated e.g. in the conifold one. By comparing to (5.25), and by using (2.59), we conclude that Σ 2πi F The next correction is non-trivial. By solving the first equation in (2.54), we find where f 1 (z) is a holomorphic ambiguity. We fix it again by going to the conifold frame, and by using that where S C is the propagator in the conifold frame. It has the explicit expression [18] where P ν (x), Q ν (x) are Legendre functions. Proceeding in the same way, we solve the second equation in (2.54), and we find The pattern we obtain in this way is very similar to what we obtained in the analysis of the Mathieu equation, see e.g. (3.52). This suggests that the full one-instanton amplitude is given by By expanding (5.32), we reproduce the results obtained above for F 1,2 , and we obtain very explicit expressions for all the coefficients F (1) k . One finds, for example, This leads, through (2.59), to predictions for all the coefficients µ k controlling the large-order behavior of the F n . Since we can generate the functions F n up to a large value of n, we can test the above expectations in great detail. This is done as follows: we fix a value of the propagator (typically corresponding to a choice of frame) and a value of z. We use the sequence F n (z), up to a given value of n, to extract numerical approximations for the action A and for the coefficients µ k , k = 0, · · · , 4, improved with Richardson extrapolation. The numerical results are then compared to the predictions coming from (5.22) and (5.32). We show tests of our predictions in Fig. 5, Fig. 6 and Fig. 7 for A and µ 0 , for µ 1,2 , and for µ 3,4 , respectively. In all cases, we consider the large-radius frame, and values of z of the form z = −2 −ξ . We indicate the number of matching digits between the numerical approximation and the prediction as a function of the total number of terms in the sequence. As we can see, in the region (5.21) the agreement is impressive. However, as we get closer to z = 0, the number of matching digits decreases. The reason for such a loss of precision was already clarified in [18]: near z = 0 there is another action, given by the large-radius period t, which is of the same order than t c , and an additional trans-series enters into the asymptotics. We have performed tests for complex values of z and for more general values of the propagator in the region of dominance of (5.22), and the agreement is again excellent.
Quantum free energy: exact versus all-orders WKB
As we have already remarked, the quantum volume is defined as an asymptotic expansion in , and does not always have a non-perturbative definition. The same thing happens with the quantum free energy (2.17), which is defined by an asymptotic series. It turns out that, in the special case of spectral problems associated to topological strings on toric CY manifolds, there is an exact, non-perturbative function whose asymptotic expansion gives back (5.20). Let us first review how the exact quantum volume is constructed. First of all, we note that the quantum free energies F LR n (t) have an expansion as t → ∞ of the form It turns out that the formal double sum F LR inst (t, ) = n≥0 k≥1 a n,k e −kt 2n (5.36) can be first resummed in in the form [52,82] 37) where N d j L ,j R are integer numbers, called BPS invariants, which generalize the Gopakumar-Vafa invariants of the CY [83]. This means in particular that the quantum volume can be also resummed in the form (5.38) This resummation does not lead to an appropriate quantization condition due to the poles which appear at ∈ 2πQ, as first noted in [80]. One needs to add corrections invisible in an expansion, which were determined in [20] as a consequence of a general correspondence between spectral theory (ST) and topological strings (TS), or ST/TS correspondence. The quantization condition was written later in a simpler form by using BPS invariants in the NS limit [42] (the equivalence between both formulations in many cases was derived in [84].) Let us denote by the last term in the r.h.s. of (5.38). Then, the non-perturbative volume is simply given by The total, exact quantum volume is then defined as vol ex (E, ) = vol p (E, ) + vol np (E, ). (5.41) There is strong evidence that the above expression defines a function of E and , for real and E sufficiently large, which gives the actual spectrum of the operator (5.1) through the exact quantization condition vol ex (E, ) = 2π ν. (5.42) In some cases, this exact volume function can be derived from a first principles, resummed WKB calculation [85]. It is clear (see e.g. [86]) that the above procedure also defines an exact function The asymptotic expansion of this function for small and fixed t is precisely the total quantum free energy (2.17) of local P 2 , in the large-radius frame: It is then natural to ask whether the asymptotic series in the r.h.s. is Borel summable, and in case it is, whether its Borel resummation agrees with the exact function in the l.h.s. Borel summability in the region of negative z in (5.21) follows from the large-order analysis above, since A 2 is negative. We have found numerically that the Borel resummation of the F LR n , which we denote by BF (0) , differs from the exact result (5.43), as shown in Fig. 8. For example, we obtain BF (0) z = −2 −6 , = π = 2.0571102 . . . , F ex z = −2 −6 , = π = 2.0565565 . . . .
(5.45)
Here, z is obtained from t by using the inverse of the classical mirror map given in the first line of (5.4). This is in contrast to what happens with the perturbative series in calculating the energy levels of the spectral problem of (5.1). This series is Borel summable and can be Borel-resummed to the exact values of the energies [32,86]. At the same time, the mismatch we find is not surprising, and it seems to be the default behavior for "stringy" series which diverge doubly-factorially, as it has been realized recently in related examples [19,33].
Our numerical results suggest that the mismatch between the exact result and the Borel resummation is an exponentially small effect, controlled by the same instanton action which was found in [19]. In view of this mismatch, one could ask whether the exact quantum free energy could be recovered by considering the non-trivial trans-series associated to this instanton action and performing Borel-Écalle resummation, as in [19]. In other words, can we "semiclassically decode" the exact function (5.43) in terms of its WKB expansion and an appropriate transseries? Without further input, this is a difficult problem, since we have to find the appropriate trans-series parameter, which could depend on both and the modulus z. We leave this issue for future work. 6 Examples of quantum mirror curves: local F 0
Refined holomorphic anomaly and all-orders WKB
Let us finally consider another important spectral problem, corresponding to the local F 0 geometry. The quantum-mechanical operator is It has been proved rigorously [45,46] that O −1 F 0 is trace class and positive when m F 0 > 0. In this paper we will focus on the special case m F 0 = 1, in which the theory simplifies considerably. The all-orders WKB method for this operator has been studied in [34,48,50,80]. The associated Riemann surface is the mirror curve of local F 0 , e x + e −x + e y + e −y + κ = 0. (6.2) As in the example of local P 2 , the calculation of the classical volume of phase space reduces to the calculation of classical periods on this curve. The appropriate global coordinate in the moduli space is The classical periods are determined by the equations (see e.g. [87]) where the integration is fixed by the leading-order behavior (6.5) One has in this case that [5,20] vol 0 (E) = 2 ∂F 0 ∂t − 2π 2 3 , (6.6) where t = 2E + O(e −2E ) is related to E by the classical mirror map, i.e. by the first equation for the period in (6.5), once we set z = e −2E . As in the case of local P 2 , the higher-order free energies can be easily computed with the NS limit of the refined holomorphic anomaly equations. In the special case we are considering with m F 0 = 1, one can formulate the problem in terms of modular forms, as it was done in [88] for the original anomaly equations of [4]. We introduce the modular forms, as well as E 2 . The argument q = e πiτ is related to the prepotential F 0 by This provides the map between the modular variables and the geometry of the curve (6.2). In particular, we can recover the modulus as For this model, the NS limit of the refined anomaly equations is of the form (2.41) with the β = −2 value. The initial condition for the recursion is F 1 , which is given by We now parametrize the holomorphic ambiguity in (2.42) as In order to fix the ambiguity, we have to introduce boundary conditions. This requires a discussion of the frames appropriate for different regions in moduli space. The large-radius frame, which is appropriate for the region near z = 0, is defined at the classical level by the standard large-radius periods (6.5). The quantum corrections in this frame are obtained simply by setting E 2 → E 2 (q) in the above formulae. As explained e.g. in [81,88,89], there are two other important frames. One is the conifold frame, appropriate near the conifold singularity z = 1/16 (equivalently, near κ = 4). The appropriate periods at this point are defined by where C is Catalan's constant. The second equation defines the conifold prepotential F C 0 (t c ). The conifold modulus q c = e πiτc is given by and it is related to τ by an S transformation, which can be implemented in the modular forms The quantum corrections to the free energy in the conifold frame can then be obtained from the non-holomorphic F n as and applying the transformation (6.15). The gap boundary condition at the conifold reads To obtain more boundary conditions, we consider the theory in the orbifold frame, appropriate near κ = 0. The corresponding periods are given by The orbifold modulus q o = e πiτo is given by 19) and the passage to the orbifold frame is implemented through the modular transformation The quantum corrections in the orbifold frame can then be obtained from the non-holomorphic F n as and applying the transformation (6.20). The gap boundary condition at the orbifold is These boundary conditions (together with the absence of constant terms in the expansion at large radius) fix the holomorphic ambiguity completely. One finds, for example, In order to obtain the quantum corrections to the quantum volume, we have to take into account that the relation between t and E is now given by the quantum mirror map t = t(E, ), which in this case reads [48], The all-orders WKB quantization condition is then given by
Trans-series and large-order behavior for the energy levels
We can now use the technology developed in this paper to solve an elementary problem in the Quantum Mechanics of the operator (6.1). Let us denote by κ(ν, ) = e E(ν, ) the eigenvalue of O F 0 , as a function of the shifted energy level ν = m + 1/2 and . One can use standard perturbation theory to find κ(ν, ) a perturbative series in , whose coefficients depend on ν. This was first addressed in [34,86], and studied more systematically in [32], who extended the BenderWu package of [70] to include difference equations associated to operators such as (5.1) and (6.1). By using the extended package of [32], one finds, for the very first orders, This result can be in principle derived from the all-orders WKB quantization condition.
In analogy with what happened in the Mathieu equation, the quantization condition (6.25) defines a quantum dual period which is the analogue of a D ( ) in our analysis of the modified Mathieu equation. We can now expand each term ∂ t F n around κ = 4 by using the quantum mirror map. This gives (6.28) After inverting this expansion, we obtain, which is a rearrangement of the perturbative expansion (6.26). We now ask the following question: what is the behavior of the coefficients κ (ν) appearing in the perturbative expansion (6.26), at large and fixed ν? (this question was asked in e.g. [32]). We expect a behavior of the form A derivation of this asymptotics directly from the difference equation, via uniform WKB or other techniques, is not available. However, we can answer this question by using the trans-series for the quantum free energies. By now it should be clear that the one-instanton correction for genus one curves is of the form where the action A is a period, and f (1) an overall constant. The value of the action can be determined from the large-order behavior of the sequence (6.26), and turns out to be As in the case of the modified Mathieu equation, we have to evaluate (6.31) in the magnetic frame. The same argument that led to (3.118) produces in here We now promote ν to a trans-series, as in the examples in standard Quantum Mechanics. By using the quantization condition (6.27), we obtain The exponent can be computed explicitly by using our results from the refined holomorphic anomaly, and going to the conifold frame. One finds, A very important property of this result is that the -independent part in the first line (which contains the singularities at ν = 0) can be exactly resummed in terms of a Gamma function (this happens in all the Quantum Mechanical models analyzed in [3] and also in the modified Mathieu equation, as we saw in (3.101)). More precisely, it is the large ν expansion of log √ 2π16 ν Γ 1 2 + ν . (6.37) Putting everything together, we find ∆ν (1) where the coefficients a k (ν) can be easily computed from the expansion of the functions ∂ tc F C n . One finds, for the very first coefficients, a 1 (ν) = 12ν 2 + 11 96 , a 2 (ν) = 144ν 4 − 160ν 3 + 264ν 2 − 392ν + 121 18432 , a 3 (ν) = 8640ν 6 − 28800ν 5 + 54000ν 4 − 96960ν 3 + 188100ν 2 − 64680ν + 22657 26542080 . (6.39) The trans-series for the energy can be calculated as in [27], κ(ν + ∆ν (1) + · · · ) = κ(ν) + κ (1) (ν) + · · · ; (6.40) and therefore κ (1) = ∂κ ∂ν It only remains to determine the coefficient f (1) , which can be fixed by the large-order behavior of the sequence κ (ν) and is given by We conclude that the action A and the coefficient b appearing in (6.30) are given by The first two coefficients µ 0 , µ 1 are given by µ 0 = 4 π 16 2ν Γ 2 1 2 + ν , µ 1 = 4 π 16 2ν Γ 2 1 2 + ν 12ν 2 + 24ν + 11 96 , (6.48) and further coefficients can be computed with the methods explained above. One can extend this trans-series to arbitrary ν (i.e. not necessarily a half-integer) by introducing a factor (−1) 2ν−1 in f (1) . We can test these predictions with an analysis of the sequence κ (ν) defined in (6.26). We extract a numerical prediction µ (k,l) i for the coefficients µ i by taking the first k terms in the sequence which is similar to (3.70), and performing in addition Richardson transforms. As an example, we quote the prediction for the value of µ 5 in the ground state ν = 1/2, agreeing on 20 digits. In Fig. 9 we plot the sequence (6.49) for i = 5, together with its first two Richardson transforms. The convergence to the predicted value (6.50) is clear. Figure 9. The sequence (6.49) for i = 5, denoted by blue circles, together with its two first Richardson transforms. The convergence to the predicted value (6.50) is manifest.
Conclusions and outlook
In this paper we have extended the correspondence between the refined holomorphic anomaly equations and the all orders WKB method into the realm of trans-series, providing in this way a new method to obtain non-perturbative results in Quantum Mechanics (in one dimension). We have constructed trans-series solutions to the NS limit of the refined anomaly equations which both recover and generalize known trans-series in standard quantum mechanical models. For spectral problems associated to quantum mirror curves, our trans-series solutions give information which cannot be obtained by straightforward generalizations of the current methods, as we have illustrated in the case of local P 2 and local F 0 . There are clearly many avenues for research open by our new methods. In this paper, our focus has been mostly on the one-instanton sector, for which we have in fact produced a universal expression (see e.g. (6.31)). We have also discussed the structure of the higher instanton sectors, and in particular, by using PNP relations, we have verified that the results of the holomorphic anomaly equations match existing results in Quantum Mechanics. In the case of spectral problems associated to quantum mirror curves, the higher instanton sector is less understood. One reason is that the PNP relationship in these examples is more problematic 7 . It would be important to calculate and test systematically the higher order instanton corrections in the case of quantum mirror curves, and compare them to the exact results of [20,41,42]. It has been noticed in [90][91][92] that quantum mirror curves turn out to be related to interesting spectral problems in condensed matter systems. It would be interesting to see if our methods lead to new non-perturbative results for this type of systems.
As a tool to analyze instanton trans-series, we have extended the ring of modular forms to take into account exponentially small corrections. This extension requires some unorthodox ingredients from the point of view of the traditional theory of modular forms, but it is very successful in producing correct predictions for the large-order behavior. There might be more natural versions of our formalism, and it would be very interesting to further clarify this new mathematical structure.
Our comparison of the Borel resummation of quantum free energies and the available exact results is clearly incomplete. First of all, this comparison can be done for other examples, such as the modified Mathieu equation, where the exact result for the quantum free energy is provided by instanton calculus [31]. The main question is whether all these exact results can be decoded in terms of the perturbative series, plus the trans-series found in this paper. This was achieved in [19] in a closely related example, but a deeper understanding of trans-series parameters is needed in order to have a systematic tool for such an analysis.
Finally, an important open problem is to extend our discussion (and the one of [17,18]) to the fully refined case, involving the two parameters 1,2 . This is a necessary step in order to unveil the trans-series structure of the refined topological string, and will hopefully open a new window on the non-perturbative structure of the topological string. (A.14) In the NS limit, we recover instead the results discussed in the main text.
B Large-order behavior in the modified Mathieu equation In this Appendix we give evidence that the trans-series equation for the modified Mathieu equation (3.100) leads to the correct large-order behavior of the perturbative and the one-instanton series. The trans-series for the energy reads, where f (ν) is given in (3.101). We will write, as in [27], where ξ(ν) = 2π Γ 2 ν + 1 2 32 2ν e 16 .
(B.4)
Let us now focus on the ground state ν = 1/2. We write, give the asymptotic behavior a k ∼ 2 π (−16) −k 1 − 5 2k − 13 8k 2 + · · · , k 1, (B.8) which is the result obtained in [29,72]. However, the trans-series should also give us the asymptotics of the coefficients of the first instanton series. To see how this goes, we note that the first instanton correction is formally purely imaginary, and it will get an exponentially small real piece related to the real part of the second instanton series. We expect Note the factor of 2 in the r.h.s., which is standard in resurgence (see e.g. equation (5.14) in [93] or else [94] for more details). If we write which can be checked in detail by using standard techniques 8 . | 2018-10-12T13:48:17.000Z | 2017-12-07T00:00:00.000 | {
"year": 2019,
"sha1": "99787c25d0be5e7265abfd1e52fc72c9eff309bc",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/1712.02603",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "99787c25d0be5e7265abfd1e52fc72c9eff309bc",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics",
"Mathematics"
]
} |
17415700 | pes2o/s2orc | v3-fos-license | Molecular Sciences Expression of Chemokines, Mip-1alpha and Rantes in Caprine Lentiviral Infection and Their Influence on Viral Replication
Chemokines belong to a super family of inducible and secreted, pro-inflammatory cytokines. They act primarily as chemoattractants and activators of specific types of leukocytes and are involved in a variety of immune and inflammatory responses. The status and role of chemokines, macrophage inflammatory protein (MIP-1α) and RANTES (Regulated upon Activation Normal T-cell Expressed and Secreted) in the immunopathogenesis by caprine arthritis-encephalitis virus (CAEV) are not fully elucidated. The objectives of this study were to, 1) determine the expression MIP-1α in goat synovial membrane cells (GSM cells) infected in vitro, and peripheral blood mononuclear cells (PBMC) of CAEV infected goats by RT-PCR, and 2) effect of exogenous MIP-1α and on replication of CAEV in GSM cells in vitro. RT-PCR results indicated higher expression of MIP-1α in PBMC of CAEV-infected goats as compared to controls. Similarly, higher expression of MIP-1α was observed in GSMC infected in vitro with CAEV as compared to that in uninfected cells. Exogenous MIP-1α (20 ng/ml) and RANTES (20 ng/ml) significantly inhibited CAEV replication in GSM cells by 75% and 65%, respectively as compared to the replication in GSM cells not treated with the chemokines. Results of this study suggest that CAEV infection may alter the expression of chemokines in goats, which may suppress the replication of the virus.
Introduction
Caprine arthritis encephalitis virus (CAEV) is a non-oncogenic retrovirus belonging to the subfamily lentivirinae.Exposure of goats to CAEV results in chronic mononuclear infiltration of various tissues [1].It leads to persistent infection and chronic arthritis in goats, which shows remarkable histopathological similarity to rheumatiod arthritis (RA) in humans.Feldman et al. [2] reported that the potential role of various cytokines and disruption in their normal function could lead to exacerbation of the inflammatory process in RA.
Chemokines are small inducible, and secreted (SIS) cytokines that are involved in a variety of immune and inflammatory processes mainly due to their chemotactic activity for specific types of leukocytes [3].RANTES and MIP-1α, two important pro-inflammatory chemokines belong to CCchemokine subfamily [4].RANTES is produced by circulating T-cells.Its synthesis is induced by TNF-α and IL-1α and inhibited by the stimulation of T-lymphocytes.RANTES is chemotactic for Tcells, eosinophils, and basophils [5].RANTES was also reported to be highly expressed in human synovial fibroblast of patients suffering from RA suggesting its role in the inflammatory process [6].
MIP-1α, produced by bacterial endotoxin stimulated macrophages, plays an important role in inflammation by inducing synthesis of other proinflammatory cytokines such as IL-1β, IL-6.It activates neutrophils, eosinophils and basophilic granulocytes and is involved in acute neutrophilic inflammation.MIP-1α is also known to induce the proliferation and activation of killer cells known as CHAK killer cells [4].
It has been shown that the altered expression and response of various chemokines may play a critical role in the immuno-and histopathology of infected goat [7].Since MIP-1α and RANTES are important pro-inflammatory cytokines, any alteration in their normal expression and function could subsequently affect immunopathology in the goat including the development of arthritis and replication of CAEV in infected goats.It has been shown that some chemokines were secreted as a result of human immunodeficiency virus (HIV) and simian immunodeficiency virus (SIV) infections and were implicated in suppressing viral activity [8][9].The effect of chemokines, MIP-1α and RANTES, on CAEV replication has not been investigated.Therefore, the present study was conducted, 1) to evaluate the effect of CAEV infection on the expression of MIP-1α by RT-PCR, and 2) to determine the effect of exogenous RANTES and MIP-1α on the replication of CAEV in vitro.
Inoculation of goats with caprine arthritis encephalitis virus
A total of six adult female mixed-breed goats were infected approximately 3 years ago with CAEV strain 75-G63 (1x10 6 TCID 50 /ml; ATCC#VR-905; ATCC, Rockville, MD) by injecting 1 ml intravenously and intraarticularly [10].Goats were confirmed seropositive to the major internal structural protein of CAEV by the agar gel immunodiffusion test (AGID; Veterinary diagnostic technology, Denver, CO) and by the polymerase chain reaction (PCR) technique [11].CAEV-infected goats were showing varying degrees of arthritis at the time of this study.Six age and sex-matched, uninfected goats, confirmed seronegative by the agar gel immunodiffusion test and PCR technique were used as controls.Tuskegee University animal care and use committee approved the protocol for use and experimental infection of goats with CAEV.
Isolation and preparation of peripheral blood mononuclear cells (PBMC)
Blood samples from control and CAEV-infected goats were collected into 10 ml vacutainers (Becton Dickinson, Franklin Lakes, NJ) containing sodium heparin by jugular venipuncture.PBM cells were isolated by density gradient centrifugation.Freshly collected blood was centrifuged at 275 x g for 10 min.Supernatant was discarded and the buffy coat diluted in 5 ml of RPMI-1640 medium was carefully layered on approximately 4 ml of Histopaque-1077 (Sigma, St. Louis, MO) and centrifuged at 800 x g for 20 min.The PBM cells that resolved into a white, buffy coat at the interface were collected and transferred into fresh 15 ml sterile centrifuge tubes.These cells were washed with 5ml PBS by centrifuging at 450 x g for 10 min, counted and 10 X 10 6 cells/ ml were used for the isolation of RNA.
Goat synovial membrane cell culture & CAEV cultivation
Goat synovial membrane cells, collected from 1 day-old goat kid, were grown under 5% CO 2 in minimal essential medium (MEM) containing 10% fetal bovine serum (FBS) and 10% streptomycin (Sigma, St. Louis, MO).The adherent cell layers were passaged at confluence.To infect, 1 X 10 6 TCID 50 /ml of CAEV strain 63 in MEM medium was added to GSM cells grown in 75cm 2 tissue culture flasks and incubated at 37°C for 24 h.The control uninfected and infected GSM cells were then used for RNA isolation.
RNA isolation & cDNA synthesis
Total RNA was isolated using an RNA isolation kit (Ambion Inc, Austin, TX).RNA was treated with DNase to eliminate any DNA contamination and confirmed by agarose gel electrophoresis.
Approximately 2 µg RNA was used to synthesize cDNA using the random primers included in the RETROscript TM kit (Ambion, Austin, TX). cDNA was used immediately for polymerase chain reaction (PCR) or stored at −20°C until needed.
Construction of goat MIP-1α cDNA clone
Human MIP-1α primers (BioSource International Inc, Camarillo, CA) were used initially to amplify a 279 bp product from total RNA isolated from CAEV-infected goats.The amplified product was cloned into pCR 2.1 vector, using original TA cloning kit (Invitrogen, Carlsden, CA) in order to sequence the PCR product using M13 forward and reverse primers.The PCR product was sequenced in the DNA Sequencing Laboratory (University of Arkansas for Medical Sciences, Little Rock, AR).
Goat specific primers, goat forward MIP-1α primer (GFMIP) and goat reverse MIP-1α primer (GRMIP) were then designed from the sequence to amplify a 324 bp product of goat MIP-1α.
RT-PCR for MIP-1α
Amplification of goat MIP-1α sequences from cDNA was performed using a Supertaq Plus kit (Ambion Inc, Austin, TX).The PCR reaction contained 1 X reaction buffer (10 mM Tris-HCl, pH 8.3, 50 mM KCl and 1.5 mM MgCl 2 ), 2.5 mM of each dNTP, 1µg of sense primer GFMIP (5' TCAGCAGATCTCTGCCATGAG3'), 1µg of antisense primer GRMIP (5'TGGTGTCGCTTAGGGACTTG3') and 1.5 units of Taq polymerase in a total volume of 50 µl.The PCR primers used were expected to amplify a 324 bp product of MIP-1α.After an initial denaturation at 94ºC for 5 min, the reaction mixture was subjected to 35 cycles of thermal cycling at 94ºC for 1min for denaturation, at 57.7ºC for 1 min for annealing and at 72ºC for 1 min for extension.Finally a 7 min extension at 72ºC was performed after the last cycle (Perkin Elmer GeneAmp PCR System 2400).
Attempts to amplify RANTES sequence in goat PBM cells and GSM cells using human RANTES primers have been unsuccessful.
Densitometric analysis of RT-PCR products
Polaroid pictures obtained of the RT-PCR electrophoresis gels were scanned into photoShop 6.0 TM .
Areas were delineated using the 'free hand lasso' PhotoShop 6.0 TM , and were computed using histogram measurements.
Effect of exogenous RANTES and MIP-1α on the replication of CAEV in GSMC
To determine the effect of exogenous RANTES and MIP-1α on the replication of CAE virus in vitro, GSM cells were grown in 6-well tissue culture plates.Upon confluence, either 20 ng/ml rHu (recombinant human) RANTES or 20 ng/ml rHu MIP-1α was added to the monolayers of GSM cells and incubated at 37ºC for 3 h. 1 X 10 6 TCID 50 /ml (100 µl) of CAEV strain 63 was then added to infect the GSM cells and incubated at 37ºC for 72 h under 5% CO 2 .In addition, RANTES and MIP-1α were also added to the cells at 24 h and 48 h post-infection.CAEV-infected GSM cells were observed daily for any signs of cytopathic effects.At the end of 72 h post-infection, supernatants were collected and frozen at -70°C until analyzed for reverse transcriptase activity.The GSM cells were then stained with 1% Giemsa's stain.The average number of syncytia (multinucleated giant cells with 3 or more nuclei) was counted from 20 fields of each well to determine viral replication.
Determination of reverse transcriptase (RT) activity
CAEV replication was also assessed by measuring the viral RT activity present in the infected GSMC culture supernatants 72 h after infection.RT activity was assayed using a colorimetric reverse transcriptase enzyme immunoassay kit (Roche Molecular Biochemicals, Nutley, NJ).
Statistical analysis
The data for area profiles of RT-PCR bands, RT activity and syncytia formation were presented as mean ± SE.Differences between means of control and infected groups were determined by student ttest and values of P≤0.05 were considered significant.
RT-PCR and gel electrophoresis analysis of MIP-1α expression in GSMC and PBMC
MIP-1α was constitutively expressed in both GSM cells and PBM cells.RT-PCR amplification of RNA isolated from GSM cells revealed a band of approximately 324 bp in both uninfected, control and CAEV-infected cells (Fig. 1).
However, expression of MIP-1α was found to be higher in CAEV-infected cells relative to control, uninfected cells.Similarly, comparison of PCR products revealed that the PBMC of CAEV-infected goats showed greater expression of MIP-1α compared to that of non-infected animals (Fig. 2).Results of area profiles of the RT-PCR bands also confirmed the differences of MIP-1α expression between CAEV-infected and non-infected goats (Fig. 3).Within GSM cells, areas of RT-PCR bands of MIP-1α varied significantly from non-infected cells to CAEV-infected cells (P≤0.0035).Significant differences were also observed between total area measurements of MIP-1α bands within PBMC (P≤0.036).GSM cells, and PBM cells in non-infected and CAEV-infected goats.Areas were delineated using the 'free hand lasso' PhotoShop 6.0 TM , and were computed using histogram measurements.Significant differences were observed between control and CAEV-infected RT-PCR bands of both GSM cells (P≤0.0035) and PBM cells (P≤0.036).Area measurements were expressed as pixels.*indicates significant difference from control.
Effect of exogenous chemokines on CAEV replication in vitro
Exogenous chemokines had significant effect on the RT activity of caprine lentivirus (P≤0.05).
Supernatants collected from the infected GSM cells at 72 h were assayed for RT activity using a sandwich ELISA method.Exogenous MIP-1α (20 ng/ml) reduced RT activity in GSM cells from 5.4 pg/ml to 0.7 pg/ml (P≤0.01),approximately an 8 fold reduction (Fig. 4).20 ng/ml exogenous RANTES also reduced RT activity in GSM cells by 2 fold (P≤0.01),from 5.4 pg/ml to 2.65 pg/ml (Fig. 5).Incubation of GSM cells with chemokines significantly reduced syncytia formation.Exogenous MIP-1α (20 ng/ml), and RANTES (20 ng/ml) reduced the number of syncytia by 75% and 65% respectively (P≤0.01) compared to the cells not treated with the chemokines (Fig. 6 and 7).Exogenous RANTES inhibits CAEV replication in GSMC in vitro.20 ng/ml exogenous RANTES was added 3 h prior to infection with 1 X 10 6 TCID 50 /ml CAEV, and at 24 h and 48 h postinfection.rHu RANTES reduced syncytia formation by 65%.After 72 h, supernatants were harvested for the determination of RT activity.*indicates significant difference from control (P≤0.01).
Discussion
The most significant biological functions of chemokines, leukocyte recruitment [12] and activation [4] are essential for a prompt and effective inflammatory response and in host defence against infections.RANTES and MIP-1α, which belong to the CC-chemokine family are critical in the establishment of chronic inflammation in various pathological conditions including rheumatoid arthritis [13][14].MIP-1α secreted locally by synovial fibroblasts and tissue macrophages was shown to recruit monocytes to the joint in RA [15] indicating the significant role played by this chemokine in the inflammatory process.Although the pathogenesis of chronic arthritis in CAEV-infected goats is not completely understood, changes in cytokine expression may be one of the factors involved [16,7].
In the present study, RT-PCR analysis showed that MIP-1α was constitutively expressed by goat synovial cells.However, CAEV infected GSM cells (in vitro) showed higher expression of MIP-1α compared to non-infected GSM cells.Importantly, increased expression of MIP-1α was also observed in PBM cells of CAEV-infected goats relative to that of non-infected goats.These results indicate that the CAEV infection may increase expression of MIP-1α, which may play an important role in the immounopathology observed in CAEV infection.It could be speculated that the increased activity of MIP-1α in caprine lentiviral infection may stimulate the synthesis of other proinflammatory cytokines such as IL-1, IL-6 and TNF in both fibroblasts and macrophages.It may also induce or increase proliferation and activation of CC-chemokine activated killer cells (CHAK) thereby increasing host's ability to fight infection.
Recently, both RANTES and MIP-1α were shown to have the ability to effectively inhibit lentiviral replication [8][9].Activated CD8+ cells in HIV-infected humans and SIV-infected non-human primates secrete soluble factors such as RANTES, MIP-1α and MIP-1β which act as antiviral factors.These antiviral factors were reported to decrease viral activity by decreasing its entry by competing with HIV-1 in binding to CCR5, the principal coreceptor for this virus [16,17,8].However, it is not known if these chemokines exert any pertinent effect on CAEV replication.It has been reported that some cytokines such as lentivirus-induced interferon (LV-IFN) were capable of adversely affecting CAEV replication in vitro [19].LV-IFN inhibits CAEV replication by suppressing monocyte proliferation and their maturation into macrophages and also by directly inhibiting CAEV gene expression in mature macrophages.Our results showed that exogenous RANTES and MIP-1α significantly inhibit CAEV replication in GSM cells in vitro.Analysis of RT activity and the syncytia formation clearly indicated significant reductions of CAEV replication in the presence of exogenous RANTES or MIP-1α.Like other chemokines, RANTES and MIP-1α mediate their effects on cells via G protein coupled receptors that contain seven transmembrane spanning regions [20].CCR1 and CCR5 act as receptors for both RANTES and MIP-1α [21][22], and can bind these chemokines with similar affinities.Incubation of GSM cells with both RANTES and MIP-1α did not result in any further decreases in CAEV activity (data not shown).This absence of combined effects of RANTES and MIP-1α could be attributed to the fact that RANTES and MIP-1α share same binding sites on CCR1 and CCR5 receptors.The exact mechanism by which these chemokines inhibit CAEV activity and replication is currently under investigation.
Based on the results of this study, it could be speculated that increased production of chemokines due to lentiviral infection play a dual role in the infected goat.They may help in reducing viral replication by directly affecting viral attachment to the macrophages or by inhibiting viral transcription [23].However, they may also contribute to the development of arthritis because of their chemotactic ability to attract T cells and macrophages, as well as activation of these cells leading to the production of other proinflammatory cytokines.Further investigations are needed to understand the precise role of these chemokines in the regulation of CAEV replication and in the immunopathology in the infected goat.
Figure 1 .
Figure 1.Effect of CAEV infection in vitro on the expression of a 324 bp fragment of MIP-1α mRNA in goat synovial membrane cells.Total RNA extracted from GSM cells was subjected to RT-PCR using goat specific primers.Samples were electrophoresed on a 2% agarose gel, stained with ethidium bromide and photographed utilizing ultraviolet transillumination.Stds: mol.size marker; Lanes 1-3: Control, uninfected GSM cells; Lanes 4-6: CAEV-infected GSM cells.
Figure 2 .Figure 3 .
Figure 2. Effect of CAEV infection in vitro on the expression of a 324 bp fragment of MIP-1α mRNA in peripheral blood mononuclear cells of goats.Total RNA extracted from PBM cells was subjected to RT-PCR using goat specific primers.Samples were electrophoresed on a 2% agarose gel, stained with ethidium bromide and photographed utilizing ultraviolet transillumination.Stds: mol.size marker; Lanes 1-4: control PBMC isolated from uninfected goats; Lanes 5-8: PBMC isolated from CAEVinfected goats.
Figure 4 .Figure 5 .
Figure 4. Exogenous rHu MIP-1α decreases RT activity of CAEV in vitro.CAEV replication was assessed by measuring RT activity present in the GSM cell culture supernatants.Supernatants were collected 72 h post-infection for RT analysis.Mean RT value of MIP-1α treated cell culture supernatant was 0.7 ng/ml compared to that of 5.4 ng/ml of control (no chemokine added) sample.*indicates significant difference from control (P≤0.01). | 2015-03-21T17:44:09.000Z | 2002-11-30T00:00:00.000 | {
"year": 2002,
"sha1": "fd1c910e336aafacf2a31be407c7802f84a4212e",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1422-0067/3/11/1177/pdf?version=1403128974",
"oa_status": "GOLD",
"pdf_src": "Crawler",
"pdf_hash": "fd1c910e336aafacf2a31be407c7802f84a4212e",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology"
]
} |
270406384 | pes2o/s2orc | v3-fos-license | Analysis of Cannabis and its Products in the Context Forensic Science
(hereinafter referred to as Δ9-THC ) and delta-8-tetrahydrocannabinol (hereinafter referred to as Δ8-THC ), as well as to identify other cannabinoids present. Solving such a large number of analytical problems requires reliable, reproducible and sensitive analytical methods. At the same time, the lack of reference materials for Δ9-THC , its isomers and other cannabinoids is hindering quality results of forensic laboratories.
Research Problem Formulation
Illegal circulation of narcotic drugs, psychotropic substances, their analogues and precursors is one of the leading social problems.Cannabis remains the most widely used illicit substance.Cultivation and production of hemp covers all regions of the world, the main emphasis is on achieving a high content of tetrahydrocannabinol (hereinafter referred to as THC) in plants.According to the report of the European Monitoring Centre for Drugs and Drug Addiction (EMCDDA)) for 2023, in the countries of the European Union in 2021, the average THC content in herbal cannabis was 9.5 %, in cannabis resin -20 % 1 .
Cannabis regulation is increasingly confronted with new forms and ways of its use.Currently, cannabis products are represented by substances with a high THC content or contain cannabidiol (hereinafter referred to as CBD) against the background of low levels of THC in oils or tinctures.These products are used in food production (such as pastries, candies, chocolate, marmalade, potato chips, pies, soups, drinks of various colors and flavors), as well as in electronic cigarettes and vaping liquids.Traditional cannabis resin is increasingly found in loose form, and there are known cases of cannabis resin being produced from plant material with a high CBD content.
The increase in the number of such materials and their various characteristics requires forensic experts to revise analysis methods of in order to prevent unreliable results.In addition, forensic experts (in particular, in EU countries) often have to not only detect, but quantify low levels of THC, differentiate its isomers, in particular delta-9-tetrahydrocannabinol (hereinafter referred to as Δ9-THC) and delta-8-tetrahydrocannabinol (hereinafter referred to as Δ8-THC), as well as to identify other cannabinoids present.Solving such a large number of analytical problems requires reliable, reproducible and sensitive analytical methods.At the same time, the lack of reference materials for Δ9-THC, its isomers and other cannabinoids is hindering quality results of forensic laboratories.
Plants of the Cannabis genus contain approximately 500 compounds, of which more than 144 are classified as phytocannabinoids (natural cannabinoids), while others are represented by flavonoids, quantification, profiling and age estimation of cannabis are considered.
Keywords: cannabis; phytocannabinoids; extraction; identification; quantitative definition; gas chromatography; liquid chromatography; nuclear magnetic resonance method; methods of vibrational spectroscopy.Theory and Practice of Forensic Science and Criminalistics.2024.Issue 1 (34) ISSN 1993-0917 e-ISSN 2708-5171 https://khrife-journal.org/index.php/journal94 fatty acids, and phenols.Among the chemical components, the most famous in plants of the genus Cannabis, THC and CBD are distinguished.Due to complexity of terpenophenolic composition of plants and the investigated matrix, for example, products containing cannabis extracts, their analysis and quantification require the use of multi-step sample preparation protocols with subsequent validation of each new method that will be implemented in the laboratory.
This topic is relevant for Ukraine, as cannabis and products containing cannabis are actively promoted on the market of recreational products.The speed of spread of these substances (against the fact that their psychoactivity is not always known, insufficient research and lack of control) gives them the status of high social danger and requires further research, which is already being conducted in EU countries.
Article Purpose
Review of the latest advances of scientific literature on isolation and analysis of cannabis and cannabis-containing products in a forensic context.Since the forensic expert must be well-versed in modern trends in the analysis of prohibited substances and use the latest data from the analytical and forensic literature in his work, such information will help them choose research methods, taking into account the actual resources and equipment of forensic laboratories.
Analysis of Essential Researches and Publications
Analysis of the special scientific literature of recent years indicates an active interest in the analysis of the complex matrix of cannabis plants.Many studies are aimed at studying the processes of phytocannabinoid extraction for further analysis.A study by G. Micalizzi and colleagues of the properties of solvents (methanol, ethanol, acetone, hexane) demonstrated that ethanol exhibits the best extraction properties against cannabinoid acids in their decarboxylated forms; on the other hand, the use of hexane reduces the extraction results of cannabinoids 2 .
N. Christinat et al. studied the properties of a mixture of water and acetonitrile as extraction reagents for the determination of various types of cannabinoids in samples of hempseed oil, milk, various beverages and food 3 .In other similar studies, isopropanol was recognized as the most effective extractant (compared to methanol and acetonitrile) during the extraction of the phytocannabinoid composition of hemp seeds 4 .
V. Brighenti and colleagues compared extraction methods, their parameters and solvents, studying the composition of cannabinoids in cannabis flowers; dynamic maceration in ethanol at room temperature demonstrated the highest efficiency for the further study of the total amount of cannabinoids, as well as their neutral and acidic forms 5 .
Investigating the effect of extraction conditions on the release of cannabinoids and cannabis flowers, E. М. M. А. ElSohly and C. Cuttler report a tendency to increase THC concentration in hemp plants from 10-14 % 9 to 20 % 10 .Quantification of cannabis is usually performed using gas chromatography (GC) with a flame ionization detector (GC-FID), as noted in G. R. Borges et al 11 .
V. Cardenia and his colleagues compared the methods of GC-FID and gas chromatography with a mass detector (hereinafter referred to as GC-MS), showed the interchangeability of detectors in the case of the study of cannabis and products containing it, and developed a GC-MS method for the quantitative determination of cannabinoids , approved for use in European countries, on a quadrupole GC-MS with electron impact ionization 12 .
C. Citti and colleagues noted that the GC-MS method for quantification requires the use of expensive deuterated standards, which are not available for all cannabinoids 13 .L. А. Ciolino and her colleagues recommend using N,Obis(trimethylsilyl)trifluoroacetamide for the derivatization process that occurs during the identification of phytocannabinoid acids 14 .G. R. Borges and his co-authors compared several chromatographic methods using Δ9-THC as an example and determined a significant amount of Δ8-THC and cannabinol (hereinafter referred to as CBN), which are known products of the breakdown of Δ9-THC and are present in the GC chromatograms, in contrast to the results obtained by the methods of highperformance Liquid chromatographymass spectrometry (hereinafter referred to as HPLC) 15 , nuclear magnetic resonance (hereinafter referred to as hereinafter referred to as) and thin-layer chromatography (hereinafter referred to as TLC).M. M. Delgado-Povedano, together with co-authors, described a method for research on extracts of plants of the genus Cannabis using quadrupole timeof-flight mass detectors in combination with GC for non-polar compounds or the method of liquid chromatography (hereinafter referred to as LC) for more polar cannabinoids and their acids 16 .
L. Calvi and colleagues evaluated the potential of HPLC with an Orbitrap detector for the study of cannabinoids in commercial cannabis extracts and investigated the degradation of these extracts under different storage conditions 17 .
N. Christinat et al. tested a variety of edible products sold in the EU for the presence and quantitative content of 97 cannabinoids using ultra-efficient liquid chromatography with a hybrid triple quadrupole-linear ion trap.Studies have shown that the concentration of CBD in the milk of cows fed with hemp reaches 10 mg/ kg.In their studies, the authors also noted the excess of permissible levels of toxicity and concentration of THC in some products containing cannabis extract 18 .
G. R. Borges and colleagues developed a fast (5 minutes) method of highperformance liquid chromatography in combination with a diode detector for the quantitative determination of THC without loss of resolution of the main cannabinoids 19 .
Application of liquid chromatography method with ultraviolet detection to classify cannabis samples in order to distinguish them into recreational and fibrous, was proposed by M. Mandrioli et al 20 .
M. Deville and her colleagues noted the existence of a significant metabolic difference between plants grown indoors and outdoors: outdoor cultivars had significantly higher concentrations of THC, CBD, and CBN 21 .G. Stefkov and his co-authors investigated phytocannabinoids from plants of the genus Cannabis using UV detection 22 .At the same time, a team of Italian researchers published a report on the difficulties of joint elution of cannabinoid peaks and their low sensitivity to UV radiation, which makes it difficult to quantify THC and CBD during one analytical cycle 23 .
C. Duchateau and colleagues described the difficulties of identifying and distinguishing between legal and illegal (legal and illegal) cannabis inflorescences encountered by law enforcement agencies in the field (outside laboratories), and developed an innovative approach to the classification of cannabis samples (in accordance with European and Swiss laws) using stationary and portable nearinfrared (hereinafter referred to as NIR) analyzers and chemometric methods for determining the concentration of THC on GC-FID 24 .
С. Sánchez-Carnerero Callado and coauthors reported obtaining acceptable prognostic results in the estimation of cannabinoid concentration using NIR and NIR with Fourier transformation and the use of mathematical and statistical models: in their opinion, this analytical method simplifies the estimation of the quantitative content of cannabinoids in plants of the genus Cannabis compared to traditional GC method 25 .
L. Sanchez and colleagues used Raman spectroscopy to investigate cannabis with high CBD content, achieving 100 % accuracy through the use of chemometric methods 26 .
J. A. de Leite and her colleagues noted difficulty of isolating cannabinoids from a complex plant matrix using the NMR method: they used preparative highpressure liquid chromatography to isolate and purify the samples 27 .
C. Citti 28 , G. Stefkov 29 and J. A. de Leite 30 in their research papers described the use of the NMR method for qualitative and quantitative research of cannabis components.
Chemometric analysis for estimating the age of herbal cannabis was used by K. C. Mariotti and co-authors 31 .G. R. Borges and his co-authors used chemometric methods to differentiate the fibrous and recreational types of cannabis 32 .A. Slosse processed data during cannabis profiling using chemometric methods 33 .
Main Content Presentation
Development of analytical methods for the identification and quantification of cannabinoids in various matrices is a special challenge.In nature, THC is found in plants of the genus Cannabis, it is considered the main psychoactive component of cannabis.The tetrahydrocannabinol term includes all stereochemical variants unless otherwise specified.For forensic investigations, the following components of cannabis are important: tetrahydrocannabinolic acid (hereinafter referred to as THCA), CBN, CBD, cannabigerol, cannabivarin, cannabichromene (hereinafter referred to as CBC).
THC, CBD and CBC are the main phytocannabinoid components of cannabis.In fresh biomass, approximately 95% of these components exist in the form of starting compounds: THCA, cannabidiolic acid (hereinafter referred to as CBDA) and cannabichromic acid.All of them are formed as a result of the enzymatic catalysis of cannabigerolic acid.The corresponding THC, CBD and CBC are formed by light and heat decarboxylation.
Decarboxylation rarely comes to an end, so both forms remain present in the matrix.CBN is a product of unnatural decomposition of THC, therefore, when processing cannabis and products containing it, it is necessary to take into account factors that affect the stability of cannabinoids and the consequences of sample storage conditions, laboratory analysis and interpretation of results (such as decarboxylation of THCA, oxidation of THC to CBN, CBD to THC isomers, isomerization of Δ9-THC in Δ8-THC).
Decarboxylation of THCA to form THC occurs during harvesting and drying of hemp, heating of the sample, for example, during smoking, under the influence of light, and during chemical research.Under similar conditions, THC can independently turn into CBN.Therefore, the correct storage of samples and products containing hemp, as well as the choice of appropriate methods for their chemical analysis, are of crucial importance.Important factor in determining total THC content is also the stability of cannabinoids in the studied object.Based on the THC and CBN content, the age of a particular cannabis sample can be determined, so comparative analysis is usually impractical three months after material extraction.
Cannabis resin is usually in the form of large, dense blocks; the degree of decarboxylation and degradation of cannabinoids, and therefore the profiles of cannabinoids, differ in different parts of such blocks due to different exposure to light and heat (cannabis resin is sensitive to heat and light during storage).
CBN and Δ8-THC have recently received increasing attention: Δ8-THC, which is a secondary component naturally occurring in the Cannabis plant, has been found as a major component in vaping liquids, chewing gum and tinctures.
In addition to Δ8-THC, THC isomers such as delta-6a-and delta-10atetrahydrocannabinol have been found in vaping liquids 34 .
For routine analysis (identification and quantification) of cannabis samples and products containing it, it is important to use representative material.General aspects of representative sampling of narcotic substances for analysis are presented in the European Network of Forensic Institutes (ENFSI) Narcotics Working Group Guidelines on Sampling of Illicit Drugs for Quantitative Analysis.
It is important to maintain the best storage conditions for the herb, including a dark place at -20 °С, as THC at this stage 100 is still sensitive to air and UV (light) that can cause THC to oxidize to CBN.Fresh plant material should be stored in paper bags, as polymer packaging can cause decomposition and the appearance of mold due to the high moisture content.
Samples must be dried.There are different approaches to drying Cannabis plants, for example, at temperatures below 70 °С to constant weight and a moisture content of 8% to 13% 35 , or at 40 °С in an oven for 12 hours 36 .The dried material has a heterogeneous composition: it should be crushed.Grinding helps to release the oily resin containing cannabinoids, terpenes and other secondary plant metabolites, increases the surface area and increases the extraction efficiency 37 .
Homogenization of plant cannabis is not required for qualitative chromatographic analysis when using plant parts with the highest THC levels.Drying cannabis resin is not necessary.However, due to the nature of the matrix, the grinding process is complicated, and freezing the samples can facilitate grinding with a mortar and pestle 38 .
Objects in the form of food, beverages, or supplements require preparation for a specific analytical method before analytical testing.For research on of liquid samples containing cannabinoids, they are mixed with a selected solvent, homogenized and (if necessary) decarboxylated 39 .This method of sample preparation is simple and straightforward, but can lead to overloading of the analyzer and unreliable results.
The extraction procedure is an important step in the preparation of samples containing cannabis.Extraction of samples should be simple, selective and reproducible 40 .Depending on the lipophilicity of the extracted cannabis components, solvents from polar to nonpolar are used 41 .Non-polar solvents (hexane, petroleum ether) extract neutral cannabinoids well, but acidic cannabinoids (for example, THCA) poorly.Therefore, the extracts obtained in this way do not satisfy the requirements of quantitative THCA analysis the sum of THC and THCA.To extract acidic cannabinoids, polar solvents are used (isopropyl alcohol, ethanol, methanol, mixtures of methanol with chloroform and acetonitrile) 42 .
Common methods are liquid-liquid extraction, which is used to extract bioactive substances from oils or other liquids containing CBD, and solid-liquid extraction, where an organic solvent is added to the plant material 43 .
In addition to the solvent type, total cannabinoid yield is affected by the number of consecutive extractions, particle size, and temperature.Extraction at elevated temperatures causes the decarboxylation of cannabinoid acids to neutral compounds, which distorts the results for each cannabinoid present, so extraction methods using ultrasound, microwave radiation, pressurized liquid extraction and supercritical liquid extraction are recommended 44 .
The choice of appropriate method for research on cannabis and products containing it depends on the goals of the analysis, the characteristics of the objects and the analytical requirements (qualitative and/or quantitative data, determination of low levels of Δ9-THC, differentiation of Δ9-THC isomers, detection of other cannabinoids present in the sample).
Thin-layer chromatography is often used for the initial qualitative screening of cannabinoids: it is a simple and affordable method for pre-screening for the presence of acidic and neutral cannabinoids.
GC is a rapid method with excellent resolution properties used for the analysis of cannabinoids in plant materials and biological matrices.However, due to the thermolability of acidic forms of phytocannabinoids, this method does not work for the identification of cannabinoid acids (THCA and CBDA), since their decarboxylation occurs in the hightemperature injector of gas chromatograph.
In default of specific legal requirements, it is established practice to determine the total THC content, as this best reflects the pharmacological activity of the material, but even then, elevated temperatures may lead to THC decomposition (to CBN).For non-smoking products containing THC, THC and THCA should be identified and quantified separately.THC, like cannabis, unlike THCA, is subject to international control.Total THC findings are common practice when the material is intended for smoking, as this process converts THCA to THC.In test materials that are not heated during use, THC does not convert to THC (e.g. in foodstuffs) and should therefore be identified and described separately.
In order to preserve the acid structure and evaluate all cannabis components separately, derivatization is carried out before starting the research, which helps to obtain a more detailed profile of cannabinoids (although this procedure lengthens sample preparation and increases analysis cost).There can be difficulty in achieving complete derivatization, leading to inaccurate results due to the possibility of thermal degradation of cannabinoids in the injection port and column.
Despite the above issues, GC method remains useful for the analysis of cannabinoids.The most common detectors for GC analysis of cannabinoids are mass spectrometric (MS) and flame ionization (FID) detectors.The gold standard for identification is the GC-MS method, which 102 uses electron ionization, which provides a high level of fragmentation of compounds together with the use of commercial and proprietary libraries for the qualification of cannabis products.There is no such possibility in GC-FID 45 , since the identification of substances depends on the retention time, and reference standards are used to identify components.At the same time, GC-FID offers a simple and relatively economical detection method with high resolution and a wider linear dynamic range, which ensures accurate quantification of cannabinoids 46 .
For accurate identification and quantification of the most common cannabinoids (Δ9-THC, Δ8-THC, CBD, CBC, cannabicyclol and their acid precursors that have the same molecular weight and identical mass fragments) during testing using mass spectrometric detectors, it is necessary to carefully choose an initial chromatographic separation method combined with sensitive detection methods.
Liquid chromatography (hereinafter referred to as LC) is an advanced method for separating complex cannabinoid samples.The main methods used for this are HPLC and ultra-efficient LC.
Advantages of LC compared to GC: reduced separation time and the ability to analyze all cannabinoids directly, including acid precursors, bypassing the derivatization process.The use of this method is not accompanied by heating of the samples, therefore decarboxylation of natural acids does not occur.This ensures the preservation of the authentic composition and allows detection of acids and their corresponding neutral forms, which contributes to a more complete profile 47 .
Diode array detectors and UV detectors are often used in conjunction with HPLC to quantify cannabinoids in complex extracts of Cannabis plants, simple and relatively inexpensive methods that provide accurate results.Diode array detectors have the advantage of simultaneously measuring a wide range of wavelengths (compared to UV scanning, which uses one fixed wavelength per cycle).Published HPLC analytical methods are fully 48 or partially 49 validated for various matrices 50 and meet the requirements of the international standard ISO/IEC 17025 51 , that stipulates general requirements for testing and calibration laboratories.However, the authors note different values of detection 103 levels, quantification, linear range, and repeatability due to matrix, instrument, and detector design properties.
Lack of selectivity of diode-array detectors when analyzing complex matrices of cannabinoid extracts, where other cannabinoid-absorbing compounds at the detection wavelength may co-elute with the same retention time and distort analytical results, has prompted the scientific community to investigate more sensitive and specific MS detectors.
LC-MS is the preferred method for cannabinoid analysis using ionization modes (electrospray ionization or atmospheric pressure chemical ionization).Cannabinoids can be analyzed in any type of mass detector because they contain ionized hydroxyl and carboxyl groups.Most of these mass detectors and their combined systems have been successfully used to determine cannabinoids in cannabis plant material 52 .
A time-of-flight mass detector (alone or in combination with a quadrupole) is used to identify and quantify cannabinoid compounds in complex plant mixtures and matrices: such as food products containing cannabis (e.g.cannabis chocolate) 53 .
Because the mass-to-charge ratio of cannabinoids and the nature of fragmentation are similar, experts use a dual approach to their quantification: chromatographic separation using a diode detector combined with quadrupole 52 Citti C., Russo F., Sgrò S., Gallo A., Zanotto A., Forni F., Vandelli M. A., Laganà A., quantitative analysis.Two separate LC-MS instruments can be used in research to effectively identify and quantify all cannabinoids in their extracts.
Triple quadrupole mass detector is particularly useful for the quantification of compounds with very low cannabinoid concentrations (e.g.hemp seed extracts and human body fluids) in positive ionization mode 54 .
Ion trap mass detectors (such as the Orbitrap) are often combined with a quadrupole mass detector to form a QTrap.Detectors of this type perform multistage mass spectroscopic analysis with high resolution and selectivity.Orbitrap type mass detectors, as well as LC with triple quadrupole mass spectrometer (LC-QqQ), are used to study products with low cannabinoid content (food and beverages) 55 .
In order to meet the high demand for the analysis of cannabis and products containing it, today special analytical tools are being developed to test such samples.For example, an HPLC analyzer from Shimazu is already available, designed exclusively for the quantification of cannabinoid content, which comes with a column, mobile phase and standard material and does not require laborious method development.
High-resolution mass spectrometry with direct real-time analysis (DART-HRMS) is used to screen products 104 containing cannabis (eg., leaves, stems, roots, powders, tinctures, capsules, and other plant products) for forensic purposes.This method makes it possible to quickly identify the required substance in complex matrices without pre-treatment of samples 56 .
The NMR method is a powerful tool for the analysis of chemical compounds, including cannabinoids.In forensic laboratories, the NMR method is used to identify and quantify cannabinoids in samples of narcotic drugs and psychotropic substances.NMR is based on the reaction of nuclei with a magnetic field and measures the emission of electromagnetic signals that provide information about the structure and amount of substances in the sample.Derivatization is a necessary step in the study of some samples to obtain reliable analysis results, especially for cannabidiol acids, which acquire a neutral form during decarboxylation.Direct measurement is possible for cannabis oils.One of the main advantages of the NMR method is that its use does not require the use of standard samples for calibration.This method makes it possible to quickly and reliably determine the composition of complex mixtures by measuring the interactions of nuclei with a magnetic field, which is especially important for the study of complex substances that contain many different compounds and in which it is difficult to determine the 56 Appley M. G., Chambers M. I., Musah R. A. Quantification of Hordenine in a Complex Plant Matrix by Direct Analysis in Real Time-High-Resolution Mass Spectrometry: Application to the "Plant of Concern" Sceletium Tortuosum.Drug Testing and Analysis.2021.Vol.14. Is. 4. Pp. 604-612.DOI: 10.1002/dta.3193(date accessed: 07.10.2023).57 Deidda R., Dispas A., De Bleye Ch., Hubert Ph., Ziemons É. Op. cit.DOI: 10.1016/j.aca.2021.339184(date accessed: 12.10.2023).58 Sanchez L., Filter C., Baltensperger D., Kurouski D. Op. cit.DOI: 10.1039/C9RA08225E (date accessed: 11.10.2023).59 Citti C., Russo F., Sgrò S., Gallo A., Zanotto A., Forni F., Vandelli M. A., Laganà A., Montone C. M., Gigli G., Cannazza G. Op. cit.DOI: 10.1007/s00216-020-02554-3 (date accessed: 12.10.2023).60 Deidda R., Dispas A., De Bleye Ch., Hubert Ph., Ziemons É. Op. cit.DOI: 10.1016/j.aca.2021.339184(date accessed: 07.10.2023).
standard concentrations of each individual component.
In recent years, methods of vibrational spectroscopy, in particular infrared and Raman, have been actively used to analyze cannabis.NIR method is considered the most common in the spectral range of infrared spectroscopy.The main advantages of these methods are ease of use, speed of measurements, as well as non-invasive and non-destructive analysis of plant material without the need for pretreatment of the sample 57 .
The availability of portable devices makes it possible to conduct testing outside the laboratory 58 .Despite the mentioned advantages, it is worth noting the low sensitivity of this method compared to LC and GC.The interpretation of such spectra requires a multidimensional analysis of the obtained data due to a significant amount of information 59 .It is worth noting that the results of Raman spectroscopy can be affected by molecules that cause fluorescence (for example, chlorophyll) 60 .
Analytical methods provide a large and complex volume of data that requires multidimensional analysis to obtain the necessary information.These data can be processed to select optimal measurement procedures, explore patterns, or predict certain properties.Chemometrics methods are used to obtain the most valuable information.Recently, for processing complex matrices (as in the case of 105 Cannabis), chemometric tools are used, which make it possible to obtain necessary information from sample data 61 .
Given that the results provided by forensic laboratories are part of law enforcement investigations and can be used as evidence in court, it is important to provide reliable and reproducible data.To confirm the technical competence of forensic laboratories, accredited ISO/ IEC 17025 methods should be used.The main parameter is the measurement uncertainty, especially for the quantitative determination of Δ9-THC and other prohibited cannabinoids to check their legality (0.08% is the maximum permissible rate of THC in vegetable raw materials in Ukraine).This parameter should be kept in mind as it reflects the variance of the results and has a significant impact on the interpretation of the results under applicable law.Without considering this parameter, there is a risk of misinterpretation and wrongful prosecution 62 .
Analytical standards are a cornerstone of quantitative cannabinoid analysis.attention to the legislation that regulates the purchase and import of such materials.Morphological and chemical analytical methods for the study of cannabis and products containing it are usually sufficient for its identification.However, in situations where the sample does not have distinct morphological characteristics of cannabis plant material or contains low levels of THC (for example, if it is highly fragmented material, young seedlings, seeds, roots or bare branches), the identification of cannabis based on DNA analysis is more efficient and effective for species-level identification 63 .
Conclusions
Highly professional forensic examination of narcotic drugs, psychotropic substances and precursors plays an important role for judicial system in particular and society as a whole.Most current research focuses on the three or four major cannabinoids and their precursor acids.With the expansion of knowledge about other cannabinoids, the demand for their research as additives to many consumer goods is likely to increase, which necessitates the development of identification protocols and quantification methods.
There are at least 16 to 20 different cannabinoids that require immediate development of procedures for their extraction and analysis due to the fact that interest in them as products for consumption is growing day by day.
106
The present review of cannabiscontaining object analysis methods shows that for the correct separation and identification of phytocannabinoids from the plant matrix and/or matrix of edible or other cannabis-derived products, various chromatographic methods offer different possibilities in terms of detection limits, linearity, reproducibility, and specificity.They depend on particular device parameters (type of detector, availability of ionization, capabilities of data collection and analysis software).Therefore, in order to choose the most appropriate analytical method, it is important to apply a balanced approach, which should be based on the volume of analysis and analytical capabilities of a particular laboratory.
While developing analysis procedures, it is necessary to take into account the storage conditions and the influence of various factors on composition of cannabis containing objects, as well as on the analytical research standards.This is especially important because cannabinoids by their nature can decompose or turn one into the other under the influence of light, heat, or oxidation.
With development of cannabinoid extract analysis, it is important to harmonize standardized research methods that will contribute to harmonization of testing methods and harmonization of analytical results.
Chemometrics is becoming an important and powerful forensic tool, especially when it comes to cannabis profiling or while working with spectroscopic data, where multivariate data analysis is commonly used.
Despite the rather complex modern approaches to cannabis research procedures and products containing it, they may not be needed for routine research.The choice of method of and decision on the need to apply additional methods remains at discretion of forensic expert and depends on availability of appropriate laboratory equipment and legal specifics.
This list of research methods for cannabis and cannabis containing products is not exhaustive.Within one article, it is impossible to cover all the analytical methods for research on plant material and products based on it used in forensic research: we have focused only on the leading ones.
Financing
This research did not receive any specific grant from funding institutions in the public, commercial or non-commercial sectors.Disclaimer Founders had no role in the research design, data collection and analysis, decision to publish or manuscript preparation.Participants Author has contributed solely to intellectual discussion underlying this document, case law research, writing and editing, and assumes responsibility for its content and interpretation.
Declaration of Competing Interest
Author declare no conflict of interest. | 2024-06-13T15:23:15.438Z | 2024-03-29T00:00:00.000 | {
"year": 2024,
"sha1": "7b77edd346ee6d1c311b6add4cec76596f4886eb",
"oa_license": null,
"oa_url": "https://doi.org/10.32353/khrife.1.2024.06",
"oa_status": "GOLD",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "626dce982925c93b348508bc6c97c840c4fadbb3",
"s2fieldsofstudy": [
"Chemistry",
"Medicine"
],
"extfieldsofstudy": []
} |
56165616 | pes2o/s2orc | v3-fos-license | Studies of the nucleon structure in back-to-back SIDIS
The Deep Inelastic Scattering (DIS) proved to be a great tool in testing of the theory of strong interactions, which was a major focus in last decades. Semi-Inclusive DIS (SIDIS), with detection of an additional hadron allowed first studies of 3D structure of the nucleon, moving the main focus from testing the QCD to understanding of strong interactions and quark gluon dynamics to address a number of puzzles accumulated in recent years. Detection of two hadrons in SIDIS, which is even more complicated, provides access to details of quark gluon interactions inaccessible in single-hadron SIDIS, providing a new avenue to study the complex nucleon structure. Large acceptance of the Electron Ion Collider, allowing detection of two hadrons, produced back-to-back in the current and target fragmentation regions, combined with clear separation of two regions, would provide a unique possibility to study the nucleon structure in target fragmentation region, and correlations of target and current fragmentation regions.
Introduction
The quark-gluon dynamics manifests itself in a set of non-perturbative functions describing all possible spin-spin and spin-orbit correlations.Many experiments worldwide are currently trying to pin down various effects related to the nucleon structure through semi-inclusive deep-inelastic scattering (HERMES at DESY [1][2][3][4], COMPASS at CERN [5], Jefferson Lab [6][7][8][9]) polarized proton-proton collisions (PHENIX, STAR and BRAHMS at RHIC) [10][11][12], and electron-positron annihilation (Belle and BaBar) [13][14][15][16].Azimuthal distributions of final state particles in in hard scattering processes, in particular, are sensitive to the orbital motion of quarks and play an important role in the study of transverse momentum distributions (TMDs) of quarks in the nucleon.Transverse momentum dependent distributions of partons, encoded in TMDs, and transverse space distributions of partons, encoded in Generalized Parton Distributions (GPDs), have been widely recognized as key objectives of the JLab 12 GeV upgrade and a driving force behind construction of the Electron Ion Collider (EIC).Correlations of the spin of the target or/and the momentum and the spin of quarks, combined with final state interactions define the azimuthal distributions of produced particles.During the last few years, the first results on transverse single-spin asymmetries (SSAs) have become available.HERMES and COMPASS [3,5,[17][18][19][20][21] measurements for the first time directly indicated significant azimuthal moments generated both by the Collins and the Sivers effects.Measurements of SSAs at JLab, performed with longitudinally polarized NH 3 [22] and transversely polarized 3 He a e-mail: avakian@jlab.org3 3 [23][24][25][26] indicate that spin orbit correlations may be significant for certain combinations of spins of quarks and nucleons and transverse momentum of scattered quarks.Large spin-azimuthal asymmetries have been observed at JLab for longitudinally polarized beam [6] and transversely polarized target [27], consistent with corresponding measurements at HERMES [28] and COMPASS [29], which have been interpreted in terms of higher twist contributions, related to quark-gluon correlations.Recent measurements of multiplicities and double spin asymmetries as a function of the final transverse momentum of pions in SIDIS at JLab [8,22] suggest that transverse momentum distributions may depend on the polarization of quarks and possibly also on their flavor.Calculations of transverse momentum dependence of TMDs in different models [30][31][32][33] and on lattice [34,35] indicate that the dependence of the transverse momentum distributions on the quark polarization and flavor may be very significant.
Combination of measurements in a large Q 2 range from HERMES, COMPASS and JLab, extended to EIC would allow studies of evolution effects and control possible higher twist contributions in the measurements of TMD observables in general, and of the Sivers asymmetry in particular.Wide acceptance of CLAS12 detector at Jefferson Lab (JLab) and EIC would allow also checks of Sivers effect in the target fragmentation region, where it is expected to change the sign [36].Much higher Q 2 range accessible at JLab12 with CLAS12 and EIC would allow for studies of Q 2 -dependence of different higher twist spin-azimuthal asymmetries, which, apart from providing important information on quark-gluon correlations are needed for understanding of possible corrections from higher twists to leading twist observables.An important process which can provide independent information on twist-3 TMDs is the di-hadron production in SIDIS described by interference functions [37][38][39][40][41][42][43].In fact, the measurement of SSAs with longitudinally polarized target or beam is sensitive in particular to the twist-3 chiral-odd distribution functions e and h L , in combination with the chiral-odd interference fragmentation function H < ) 1 [41].This effect survives after integration over quark transverse momenta and can be analyzed in the framework of collinear factorization.The dihadron production, thus, becomes a unique tool to study the higher twist effects appearing as sin φ modulations in target or beam spin dependent azimuthal moments of the SIDIS cross section [44,45].
EPJ Web of Conferences
The fracture function defining the conditional probability to produce the hadron h when a quark q is struck in a proton target(left) and the full list of leading twist FFs, where U,L,T stand for unpolarized, longitudinally polarized and transversely polarized nucleons (rows) and quarks (columns).
The interference fragmentation function H < ) 1 has been used to obtain information on the transversity parton distribution function [46].Although, dihadron production in SIDIS requires higher energies and Q 2 , than single hadron SIDIS, measurements of double-spin asymmetries at CLAS are already at 5.7GeV compatible with simple leading twist predictions for equality of double spin asymmetries in eX,eπ + π − X, and eπ 0 X, assuming the sea quark contributions are negligible at large x B and fragmentation functions sum of charged pions are flavor independent.CLAS measurement of the double spin asymmetry from inclusive DIS (also from HERMES and SLAC) are consistent with CLAS measurements of double spin asymmetry in charged pion pair production (see Fig. 1).Although, the Target Fragmentation Region (TFR) of DIS, when the hadrons are created from the target remnant, carries important information about the spin and flavor structure of the nucleon, it has not been studied systematically in experiments due to lack of theory fundamentals.The main physical question in the TFR is how the diquark-like remnant system after the DIS process dresses itself up to become a full-fledged hadron, i.e., by which mechanism the quark-antiquark pairs restoring color neutrality are produced, and how this process is correlated with the spin of the target or/and the produced particles.In the valence-quark region (x > 0.1) accessible at JLab with CLAS detector at 5.7 GeV the polarization transfer from the beam is expected to be significant.High luminosity and high polarization of the electron beam makes CLAS an ideal place for studies of correlations between target and current fragmentation regions.Recently the leading twist formalism for spin and transverse-momentum dependent fracture functions (FFs) has been developed [47].The production of two hadrons in polarized SIDIS, with one spinless hadron produced in the current fragmentation region (CFR) and another in the TFR, would provide access to the full set of leading twist FFs [48].In case of double hadron production process l( ) + N(P) → l( ) + h 1 (P 1 ) + h 2 (P 2 ) + X, at leading order, the cross-section includes all fracture functions, describing conditional probabilities to produce a hadron of certain type in a target fragmentation for a given flavor of struck quark [47,48]: dσ l( ,λ)+N(P,S )→l( )+h 1 (P 1 )+h 2 (P 2 )+X dx dy dz 1 dζ 2 d 2 P T 1 d 2 P T 2 dφ S = where D ll (y) = y(2 − y)/1 + (1 − y) 2 .The subscripts in the structure functions σ UT,UL,LT , specify the beam (first index) and target (second index) polarization (U, L, T for unpolarized, longitudinally and transversely polarized targets, and U, L for unpolarized and longitudinally polarized beam).
For an unpolarized target expressions there are two contributions σ UU and σ LU , involving convolutions of unpolarized, longitudinally and transversely polarized quark FFs (see Fig. 2) and fragmenta-tion functions of unpolarized, D 1 , and transversely polarized quarks, H ⊥ 1 , [47]: The structure functions F ... ... are specific convolutions [48] of fracture and fragmentation functions depending on x, z 1 , ζ 2 , P 2 1⊥ , P 2 2⊥ , P 1⊥ • P 2⊥ .The hadron 1 produced in the CFR (x F1 > 0) is described by standard scaled variable z 1 P•P 1 /P•q, and its transverse momentum P T 1 (with magnitude P T 1 and azimuthal angle φ 1 ) and the hadron 2, h 2 in the TFR (x F2 < 0) is described by similar variables, ζ 2 E 2 /E and P T 2 (P T 2 and φ 2 ), where x F is Feynman variable defining the fraction of the longitudinal momentum of the hadron in the virtual photon-proton center of mass (CM) frame.In electroproduction, the polarized lepton emits a virtual photon with non-zero longitudinal polarization, which in turn selects preferentially one polarization state of the struck quark.The opposite polarization of a remnant s s pair can again be transferred to the final-state Λ polarization, with the efficiency extracted from eN collisions.After removing a polarized scattered quark from an unpolarized nucleon, the remnant diquark may combine with an s quark, which could originate from the nucleon sea or from a color string between the diquark and the scattered quark to form a Λ hyperon.Significant polarization effects (∼ 15-20%) have been predicted in Intrinsic Strangeness Model (ISM) for Λ production in the TFR in deep-inelastic scattering [49].Longitudinal Λ polarization transfer coefficient as a function of x F has been already measured with 5.5 GeV electron beam at CLAS (e1f data set).In Fig. 3, the measured asymmetry is shown with the projected results for the future measurements using the CLAS12 detector and 11 GeV beam [50], compared with the ISM predictions [49].The spin transfer, is proportional to the beam polarization, P B , and the depolarization factor, D(y).The meson cloud model [51], due to the pseudoscalar nature of the NKΛ coupling, predicts the polarization of final-state Λ hyperons to be strongly anticorrelated to that of the nucleon, vanishing for an unpolarized target.The large acceptance of the EIC (see Fig. 4) would provide a unique possibility to study the nucleon structure in the target fragmentation region.
SSA in back-to-back hadron production
With hadrons detected in the TFR, the beam SSAs appear already at leading order.With two hadrons detected in the final state, structure functions may depend also on the relative azimuthal angle of the two hadrons, generating a long range correlation between hadrons produced in CFR and TFR.The kinematic plane for back-to-back hadron production is shown on Fig. 5. First measurements of single spin asymmetry defined by Eq.2 in semi-inclusive production of protons and charged pions in coincidence with the scattered electron in hard scattering kinematics (Q 2 > 1 GeV 2 , W 2 > 4 GeV 2 ) have been performed by the CLAS collaboration using 5.5 GeV and 5.7 GeV longitudinally polarized electron beams scattering off a 5-cm-long liquid-hydrogen target (CLAS e1f and e16 experiments).Target and current fragmentation regions were selected by cuts on the x F variables of protons ( x F < 0) and pions (x F > 0), in addition to standard data quality, vertex, acceptance,and fiducial cuts.Based on previous studies we have chosen 0.7 > z > 0.4 as the canonical cut on the final state π + to exclude contamination from exclusive events and decay pions from baryon resonances produced in the target fragmentation region.With a cut on the missing mass of the e π + X system, M X > 1.4 we have almost no data at z > 0.7, but we use this cut as it was determined in Hall C that strong deviations for the quark-parton model occur at high z [52].Additional cuts were imposed on the final state π + (0.7 > z > 0.4) to exclude contamination from exclusive events and decay pions from baryon resonances produced in the target fragmentation region.The contamination from target fragmentation, higher twist, or other effects are important for z < 0.3.Choosing as independent azimuthal angles Δφ = φ 2 − φ 1 and φ 2 , the beam spin asymmetry could be defined as sin(Δφ) The beam spin asymmetry, A sin Δφ LU , has been calculated as a sinusoidal modulation of the difference of azimuthal angles of proton and π + with respect to the lepton scattering plane, for different electron helicity states.The modulation was extracted for different bins in x, z of the pion and the product of transverse momenta of final state proton and pion with respect to the virtual photon in the CM frame.The P T -dependence of the SSA shows a trend for asymmetry to increase with increasing transverse momenta of pion and proton P T 1 ,P T 2 , (see Fig. 5), consistent with expectations from theory.The xdependence of the the A sin φ LU is consistent with asymmetry being large in the large-x B region, were the valence quark presence is very significant.Similar behavior has been observed also for events with neutral and negative pions in the final state, detected in coincidence with scattered electron and proton in the target fragmentation region.For quantitative comparison with ongoing measurements as well as projections for future measurements of different b2b processes using the CLAS12 and EIC one will need modeling of FFs, which can be modeled using different partonic models used to predict polarization of Λ hyperons in the target fragmentation region of DIS, such as the meson cloud model [51] or intrinsic strangeness model for Λ production in the target fragmentation region in deep-inelastic scattering [53,54].
Summary
In conclusion, kinematic dependences of single spin asymmetries are measured in a wide kinematic range at CLAS with polarized beam and unpolarized hydrogen target.Significant single-spin asymmetries have been observed in back-to-back pion and proton electroproduction for the first time.Measurements of single-spin asymmetries indicate that spin-orbit correlations may play an important role in description of the structure of nucleon in terms of elementary quarks and gluons going beyond the simple collinear partonic representation.Within the canonical ranges of x, Q 2 , z, P T , φ, M X , a LO pQCD model of SIDIS, observed asymmetries may be interpreted in a framework using FFs to describe the conditional probabilities of finding partons and hadronization functions describing the fragmentation of the parton to final state hadrons in the current fragmentation.A non-zero A LU in b2b SIDIS, measured for the first time, indicate that spin-orbit correlations between hadrons may be very significant opening a new avenue for studies of the complex nucleon structure in terms of quark and gluon degrees of freedom.The JLab 12-GeV upgrade will provide the unique combination of wide kinematic coverage, high beam intensity (luminosity), high energy, high polarization, and advanced detection capabilities necessary to study the transverse momentum and spin correlations in di-hadron production in doublepolarized semi-inclusive processes both in the target and current fragmentation regions.The large acceptance of the EIC combined with clear separation of target and current fragmentation regions would provide a qualitatively new tool to study the nucleon structure beyond the traditional current fragmentation.
DOI: 10
.1051/ C Owned by the authors, published by EDP Sciences,
Figure 1 .
Figure 1.Missing mass of the dihadron system showing the separation of exclusive process (left) and dihadron double spin asymmetry measured at 5.7 GeV in dihadron production, eπ + π − X (red triangles), compared with DIS in different experiments.
Figure 3 .
Figure 3. Dominant diagram for Λ production in the target fragmentation region due to scattering on a valence u quark (left) and Longitudinal Λ polarization transfer coefficient as a function of x F from CLAS 5.5 GeV e1f data set (right).
Figure 4 .
Figure 4. Angular distributions of e ΛK + X events for EIC configuration with 5 GeV electrons and 50 GeV protons in the Lab frame (left) and x F distributions for Λs and Kaons for different z fractions of Kaons (from left to wright the lower cut on z increasing from z > 0.1 to z > 0.5).
Figure 5 .
Figure 5. Kinematic plane for back-to-back (b2b) hadron production in SIDIS (left) and measured SSA in backto-back proton and π + production as a function of the product of their transverse momenta P T 1 P T 2 . | 2018-12-06T04:54:49.028Z | 2016-03-01T00:00:00.000 | {
"year": 2016,
"sha1": "543a8084da7f1982b586c3dec592391983aa1794",
"oa_license": "CCBY",
"oa_url": "https://www.epj-conferences.org/articles/epjconf/pdf/2016/07/epjconf_poetic2016_01003.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "543a8084da7f1982b586c3dec592391983aa1794",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
221479000 | pes2o/s2orc | v3-fos-license | Environmental analysis along the supply chain of dark, milk and white chocolate: a life cycle comparison
Environmental impact evaluation in the food sector is a key topic, due to both stricter legislations and higher consumer awareness towards sustainable choices. The case of chocolate is a remarkable example, owing to the increasing demand and the complex production process from cocoa beans to final bars. The present study aims at assessing the environmental impacts related to three chocolate types (dark, milk and white) through life cycle assessment (LCA) methodology. Consistent with food Product Category Rules (PCRs) and previous LCA literature, the study follows a cradle to grave approach. Among different raw material productions, it focuses above all on cocoa farming assuming three possible producer countries (i.e. Ghana, Ecuador and Indonesia), so that the influence of specific weather conditions and soil properties is underlined. Since the manufacturing step is supposed in the North Italian factory, different transport distances are also taken into account. Moreover, the work focuses on the possible use of several packaging materials and following disposal issues. In view of the open discussion about the most suitable functional unit in food sector, mass and energy amount approaches are compared. Along chocolate supply chain, different phases are evaluated according to LCA methodology. Among analyzed producer countries: Indonesia monoculture case results to be the most impacting situation, due to an intensive use of agrochemicals; pesticides give a wide contribution in Ecuador, whereas Ghana is penalized by the highest water consumption. The transport of beans to manufacturing plant influences mostly the GWP, owing to long travelled distances. Considering the whole production process, cocoa derivatives and milk powder are the main contributors to every impact category. From packaging point of view, the best solution is the use of a single polypropylene layer. A sensitivity analysis is performed to check the validity of different allocation procedures: both mass and energy content allocations lead to similar results. Through LCA methodology, the life cycle of dark, milk and white chocolate is compared. The study assesses different potential environmental impacts, assuming mass and energy content as possible functional units and references for allocation procedures. For all combinations of functional units and allocation rules, dark chocolate globally presents the best environmental performance, whereas the other two types have similar environmental impacts.
Introduction
Nowadays, the environmental sustainability is emerging as key-point in the agri-food sector because of its remarkable impacts. For instance, food sector causes more than 25% of global greenhouse gas (GHG) emissions (Solazzo et al. 2016). It needs high water consumption and it uses about half of icefree land area on the Earth for cropland and pasture, which provokes deforestation (Barona et al. 2010). Moreover, alimentary brands focus more and more on these issues, since consumers begin to be conscious of the global pollution and to consider not only the product quality but also its potential damages to the environment. Life cycle assessment (LCA) currently is the best standardized methodology to analyze the environmental aspects since it enables to highlight and study correlations between production systems and natural resource depletion, i.e. Water-Energy-Food Nexus (Del Borghi et al. 2020). Several studies aimed at evaluating environmental consequences due to food production chain: fruit and grain farm (Ingrao et al. 2015;Tricase et al. 2018), processed foods (Canellada et al. 2018;Del Borghi et al. 2014;Ingrao et al. 2018), seafood and meat (Hospido et al. 2006;López-Andrés et al. 2018). In this context, packaging is also considered to have an effective impact estimation (Del Borghi et al. 2018;Strazza et al. 2016). However, LCA still seeks to become a tool for combined analyses of economic value and eco-burden, creating new sustainable business models in view of the transition towards a circular economy (Scheepens et al. 2016).
Environmental concerns are critical in chocolate supply chain. Cocoa is only produced in tropical zones of America, Africa and Asia growing in specific humidity conditions, whereas transformation processes and principal markets are usually in Europe and North America. Indeed, there is a remarkable contrast between cocoa production and demand in different areas: for instance, Europe is the major consumer with an average annual request of 1812 ktons, followed by the USA characterized by a national consumption of 775 ktons, while only 146 ktons are eaten in Africa, which is the first worldwide farmer with 3185 ktons of produced cocoa (García-Herrero et al. 2019). Therefore, raw materials are transported for long distances. The manufacturing is quite complex because several co-products can be derived from cocoa beans and other ingredients are added to obtain the final product. The effects due to cocoa life cycle cannot be neglected anymore, in view of the increasing chocolate (ISO 2006a) for the chocolate sector. However, several studies propose the life cycle assessment of cocoa derivatives. The LCA is a standardized methodological tool that enables the assessment of the main environmental aspects associated to a specific product "from the cradle to the grave", through the evaluation of different input and output flows and their correlated potential environmental impacts according to ISO 14040 (ISO 2006b) andISO 14044 (ISO 2006c). In food fields, there are a lot of discussions concerning LCA approaches to be followed (McAuliffe et al. 2020). One of the main issues is the definition of proper functional units, which should provide a measurement of specific peculiarity of every product so as to enable the comparison among different systems. One kilogram of product is usually assumed (Roy et al. 2009), still this does not efficaciously represent the actual quantitative consumption of different food. A more realistic view is provided focusing on the nutritional value and the caloric intake. Here, the evaluation of potential environmental impacts is paired with health benefits. Some examples of this innovative perspective are the use of grams of proteins (Sonesson et al. 2017), energy amount (Nemecek et al. 2016), nutritional quality index or fullness factor (Chapa et al. 2020) as LCA functional unit.
In the case of chocolate, the common approach proposes the use of mass unit as reference for different inputs and outputs of the system. Its life cycle is usually divided into cocoa farming, transport, manufacturing, sale and end of life. There are some authors that analyze specifically the raw material cultivation in different areas: Ghana (Ntiamoah and Afrane 2008), Colombia (Ortiz et al. 2014) or Indonesia (Utomo et al. 2016). Others also consider the manufacturing (Büsser and Jungbluth 2009;Pérez Neira 2016), or the transport and the transformation neglecting the cultivation step (Vesce et al. 2016). Only packaging material can be also studied specifically (Allione et al. 2011), whereas few works take into account the whole life cycle (Miah et al. 2018;Recanati et al. 2018;Konstantas et al. 2018). Still, in all these cases, they focus on dark chocolate or chocolate derivatives, such as chocolate biscuits and wafers or moulded chocolate. So, the obtained results are not specific for chocolate life cycle, but Here, the present study aims at providing a more complete analysis through LCA methodology, based on a "cradle to grave" approach. Three common chocolate types (dark, milk and white, each one with a specific recipe) are evaluated and compared in order to detect which ingredients provoke the major potential environmental impacts. Since cocoa origin has more and more influenced consumer choice in last years (Torres-Moreno et al. 2012), a more detailed analysis of farming step is performed. The comparison among different producer countries highlights the relevance of this phase and how the results vary in function of the considered cocoa supply chain. Another relevant factor is due to packaging, so three different commercial solutions are compared. Firstly, the study follows the common literature approach of 1 kg of product as functional unit, consequently all the allocations are performed in terms of mass. Then, in view of chocolate calorie intake (Cooper et al. 2008), the analysis is also carried out considering 1 kcal as functional unit, to highlight if the previous identified trends are confirmed by this second approach.
Methods and data
The environmental potential impacts deriving from chocolate production are assessed according to LCA methodology as defined by ISO 14040-44 (ISO 2006b, c). Since no specific PCR is furnished by the International EPD® System for this product category, suggested approach for generic food products is followed (IES 2019).
Goal and scope of the study
The present study aims at detecting the potential environmental impacts due to the life cycle of three chocolate types: dark, milk and white ones. Two different approaches are followed: firstly, 1 kg of chocolate is assumed as functional unit according to the guidelines of the PCR Basic Module for food products (IES 2019); then, 1 kcal is defined as functional unit for a further comparison of the analyzed products. The study considers the life cycle "from cradle to grave", dividing it into raw material production (i.e. cocoa, milk powder, sugar and final product packaging), cocoa transport, chocolate manufacturing and packaging waste management (Fig. 1). The packaging material for cocoa bean transport, usually jute sacks, is excluded. The retail and storage steps are neglected since they The study is performed using the simulation software SimaPro 9 and the database Ecoinventv.3.5 (Wernet et al. 2016).
According to the PCR Basic Module for food products (IES 2019), the following indicators for environmental impacts and for resource use are considered: Acidification potential (AP) according to CML 2001 non-baseline-January 2016 (University of Leiden 2016); eutrophication potential (EP), global warming potential (GWP), abiotic depletionelements (ADP, el), abiotic depletionfossil fuels (ADP, ff) according to CML 2001 baseline-January 2016 (University of Leiden 2016); photochemical oxidant creation potential (POCP) according to ReCiPe 2008 (Goedkoop et al. 2009); net water use and cumulative energy demand (CED).
Life cycle inventory
In the inventory analysis, data about dark, milk and white chocolate supply chain are collected by secondary and tertiary sources: in particular, numerical data for cocoa co-products and milk powder production and for chocolate manufacturing are referred to existing recent literature, whereas data for the production of sugar and other auxiliary materials (e.g. fertilizers, pesticides) are retrieved from specific processes available in Ecoinvent v. 3.5 (Wernet et al. 2016). Then, all phases involved in the supply chain are modelled through SimaPro 9 software. The percentage of different ingredients is defined according to All. I, d.lgs. n. 178/2003 (European Directive 2000/36/CE) (ADICONSUM 2003). The lecithin, used as an emulsifier, is neglected due to low present amount. Table 1 reports proposed chocolate composition, with the reference ranges in European legislation.
In the following paragraphs, inventory data of the main steps of chocolate supply chain are presented in more details.
Raw material production and transport
The basic ingredient of chocolate is cocoa, which is cultivated in tropical regions. In the present LCA, three countries are taken into account as cocoa producers: Ecuador, Ghana and Indonesia. In a commercial stand, there is usually a density of 1100-1200 trees/ha. Since the yield is low in comparison with other tropical cultivars-for instance 150-450 kg cocoa/ha (ICCO 2020) instead of 500-1000 kg coffee/ha (Gebreselassie et al. 2017) and 30,000-50,000 kg banana/ha (Nayak et al. 2019)-the land occupation is very high. Moreover, cacao trees are not disease-resistant, so an intensive use of fertilizers and pesticides is required to increase the production. As shown in Table 2, input data for three considered areas change because of the climate, the humidity and the soil characteristics. In Ecuador case study, a traditional process, characterized by low fertilizer use per hectare and poor cocoa yield, is represented (Pérez Neira 2016). A traditional approach is also modelled for Ghana (Recanati et al. 2018), Fig. 2 Environmental impacts due to farm and transport of dried cocoa beans (Ecuador EC, Ghana GH, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) (2019) while both traditional monoculture and agroforestry systems are studied for Indonesia (Utomo et al. 2016). For the latter country, data are referred to cocoa pod: therefore, the request of 16 kg of cocoa pods to obtain 1 kg of cocoa dried beans is assumed. The emissions deriving from the application of fertilizers are evaluated according to the existing literature (Bouwman et al. 2002;EMEP/CORINAIR 2002;IPCC 2006) and are reported in Table 3. The emissions deriving from the use of diesel are instead evaluated according to specific Ecoinvent database processes. When cocoa fruits are harvested manually, the external husks are eliminated and usually left on fields as fertilizers. Then, there is the cocoa bean fermentation, which is a spontaneous process to improve aromas and reduce liquid content. The sun-drying for water and acidity elimination follows. Both phases do not request any specific energetic inputs (in few cases an artificial drying with hot air is used).
The dried cocoa beans are transported from tropical zones, where they are grown, to factories in North America and Europe. The itinerary is divided into three parts: & A first route from the cultivation site to the departure port by a lorry 3.5-7.5 t EURO3 & A trans-oceanic ship transport to the chocolate producer country & The final step from the Italian harbour to the transformation factory by a lorry 16-32 t EURO5 All transportation data are presented in Table 4. The first step is estimated from literature (Pérez Neira 2016; Ntiamoah and Afrane 2008; Recanati et al. 2018), whereas the others are calculated considering the distance between the specific port and the manufacturing factory located in Piedmont, Italy.
The input data of other ingredients are calculated considering the needed different amounts for every specific chocolate type (according to Table 1).
Chocolate manufacturing
The transformation step requests several unit operations. After the cleaning and the selection, cocoa dried beans are roasted at 120-180°C to develop aromas and to sterilize the product. Then, the milling transforms beans into cocoa liquor, which is partly fed at the pressing operation to divide the cocoa fat butter from the cocoa dried cake (its further grinding produces cocoa powder). After addition of all ingredients, the mixture is . The obtained liquid chocolate is tempered in order to cool it down slowly and then it is poured into specific moulds before wrapping into packaging. So, whereas water use is low, the whole process needs high amount of energy in the form of electricity, heat and cooling. A valuable opportunity is the addition of a trigeneration system to optimize process design, as described in Table 5 (Reverberi et al. 2011). As reported in Table 6, both manufacturing phases are modelled according to Recanati et al. (2018): it is assumed that the requested heat is provided by trigeneration and partially by supplementary natural gas, whereas electricity and cooling are derived only by the trigeneration process. Input data for the chocolate manufacturing are firstly mass allocated among different co-products-cocoa liquor (21.3%), cocoa butter (43.9%) and cocoa powder (34.8%)-and then through energy content approach (MP&F 2020).
Packaging materials and end of life
On market, chocolate bars are wrapped by several possible materials, considering the requested properties to guarantee intact aroma. As shown in Table 7, four different packaging solutions are evaluated. An option is a sole polypropylene (PP) packaging; another is an aluminium film packed with a fibre-based material: respectively a cardboard (Recanati et al. 2018) and a kraft paper.
The end of life of the different packaging materials is defined according to the Italian scenario in 2018, as reported in Table 8.
Results and discussion
The LCA provides practical key measures which allow an easy comparison among different possible conditions. Process hotspots are detected, so effective changes can be introduced for system optimization. In view of that, firstly, impacts due to cocoa farm and transport are presented to underline critical points of the main raw material in the chocolate supply chain. Then, the whole production process is considered to compare three chocolate types, owing to different requested ingredients. Finally, the production and the end-oflife treatment of packaging materials are also taken into account as addition causes common for every analyzed case.
Cocoa farm and transport
Cocoa farm is characterized by a relevantly high emission impact, in relation to other permanent fruit cultivars: low yield per hectare is the main reason in this regard. Indeed, the usable product is limited considering the elimination of husks, the weight lost during fermentation and sun-drying. Moreover, the increased demand in the last period has forced the production optimization through an intensive use of chemical substances. For every analyzed case, synthesis and usage of fertilizers are the main sources of environmental impacts. As Fig. 2 shows, Indonesia monoculture case represents the worst condition. The emission of NO into air and those of nitrate and phosphate into water, both due to N-and P-based component application, respectively contribute to more than 85% of AP (88.1%) and EP (88.5%). The direct and indirect emissions of N 2 O also cause 34.3% of the total GWP, whereas the fertilizer production adds another 38.9% to GWP and consumes 62.2% of the total energy (CED) requested by the cultivation phase. In the Ecuador case study, pesticides have a higher contribution: for instance, 23.3% of AP, 14.1% of EP and 35.9% of GWP in comparison with 2.5% of AP, 3% of EP and 12.3% of GWP in Indonesia monoculture system. Except for water consumption, Ghana shows the best performance in all the impact categories owing to the application of N-free fertilizers and the absence of diesel consumption in agricultural machinery. A possible optimization is the substitution of the agrochemicals with organic products. Since the cocoa production stage creates a large amount of solid waste due to husks (about the 67% of the fresh pod weight), these may become organic fertilizers. Moreover, cocoa residue could be also used for bioenergy production (Kamp and Østergård 2016).
As Fig. 2 shows, pollution due to the transport step is influenced by travelled distances, so Ghana scenario results to be the best solution. Among all the case studies, GWP, POCP and CED are the categories mainly affected by the transportation phase: from 10.8% for Indonesia monoculture system to 22.3% for Ecuador in the case of GWP; from 36.7% for Ghana to 52.3% for Indonesia agroforestry system in the case of POCP; and from 12.9% in Ghana to 24.1% in Indonesia agroforestry system in the case of CED.
As far as Abiotic Depletion is concerned, pesticide production results the most impacting process (above 75%) in terms of ADP, Fig. 7 ADP, elements impacts of 1 kg dark, milk and white chocolate according to mass allocation (Ecuador EC, Ghana GH, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) Fig. 8 ADP, fossil fuels impacts of 1 kg dark, milk and white chocolate according to mass allocation (Ecuador EC, Ghana GH, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) el, whereas fertilizer production contributes between 40.2% (Ghana) and 62.3% (Indonesia monoculture system) to ADP, ff.
Chocolate manufacturing
The environmental impacts caused by the production of 1 kg of chocolate are assessed and compared for dark, milk and white cases. The study evaluates the effects due to the production of ingredients (milk powder, sugar, cocoa liquor, powder and butter), energy and water consumption for final product refining. As in previous literature LCA studies (Konstantas et al. 2018;Vesce et al. 2016), Figs. 3, 4, 5, 6, 7, 8, 9 and 10 show that cocoa derivatives and milk powder provide the major contributions. The first ones are widely influenced by the producer countries. Indeed farming requests an intensive use of agrochemicals and the bean transforming phase needs a high energetic consumption (Ntiamoah and Afrane 2008). The milk powder manufacturing also has an intensive energetic usage because of evaporation and drying steps (Finnegan et al. 2017). For instance, in Ecuador case study, AP impacts are mainly due to cocoa derivatives (96%) in dark chocolate, cocoa derivatives (19%) and milk powder (63%) in milk chocolate and cocoa butter (27.6%) and milk powder (65.1%) in white chocolate. Similar percentages are obtained for EP: analyzing Ghana as farmer country cocoa derivatives contributes for 91% in dark chocolate, while 76.3% is due to milk powder in white one. In accordance with literature (Büsser and Jungbluth 2009), the milk and white chocolates have the most relevant GWP impact: considering an average value between proposed situations, about 4 kg CO 2 eq. are obtained in comparison with 2 kg CO 2 eq. due to dark chocolate production. POCP and ADP, ff have quite similar results, as the milk powder present in milk and white chocolate compensates for the major amount of cocoa co-products in dark chocolate. On the contrary, ADP, el impacts result higher for dark chocolate since the contribution (per mass unit) of milk powder is lower, due to the relevant impact of pesticides applied during cocoa cultivation. Except for Ghana case study where a considerable amount of water is used by cocoa farming, the milk powder production requests about 70% of net water consumption in chocolate supply chain (Fig. 9), whereas the water use for property chocolate refining step is very low (Vesce et al. 2016). Similar considerations are valid for needed energy: indeed, only the milk powder manufacturing spends 46 MJ (around 66%) as Fig. 10 shows. Fig. 9 Water use impacts of 1 kg dark, milk and white chocolate according to mass allocation (Ecuador EC, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) Fig. 10 CED impacts of 1 kg dark, milk and white chocolate according to mass allocation (Ecuador EC, Ghana GH, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) In general, as reported in Fig. 11, dark chocolate shows a better performance in the categories where impacts deriving from milk powder production are predominant (i.e. EP, GWP, POCP ADP, ff, CED), whereas it overtakes milk and white chocolate as milk powder contribution decreases. Water use represents a separate case since the comparison is strongly influenced by the water consumption for cocoa cultivation in Ghana (Fig. 2). Moreover, milk and white chocolate present similar results since they contain the same amount of milk powder and similar amounts of cocoa co-products (Table 1). Therefore, even though the comparison among different chocolate types varies according to the considered environmental impact category, still dark chocolate globally shows the best environmental performance, followed by white chocolate and then milk chocolate.
Packaging production and end-of-life treatments
In chocolate supply chain, the main causes of pollution are the used raw materials: above all dairy and cocoa derivatives. Certainly, the careful choice of products with lower environmental impacts, resulting from a better management of their cultivation and processing, could improve system performances. An alternative is the substitution of some ingredients; for instance, the use of soy milk, instead of cow milk, could reduce the impacts up to 70-90% (Miah et al. 2018). However, this solution is not always possible because the replacement changes the characteristics of the final product, such as taste, nutrition values and physical appearance. For this reason, an easier reduction of impacts can be obtained focusing on packaging materials. Figure 12 presents the environmental impacts generated by the packaging production to wrap 1 kg of chocolate. The polypropylene (PP) layer results to be the least impacting material in all chosen impact categories. Two different combinations of an aluminium foil with a fibre-based material result more impacting than the PP case, mainly because of aluminium-based material production. Consequently, the aluminium layer plus cardboard is the most impacting solution in all categories: respectively, 0.0021 kg SO 2 eq. for AP, 0.0008 kg PO 4 3− eq. for EP, 0.4228 kg CO 2 eq. for GWP, 0.0012 kg NMVOC eq. for POCP, 1.70 10 −6 kg Sb eq. for ADP, el, 4.1419 MJ for ADP, ff, 0.0035 m 3 for water use and 5.7136 MJ for energy consumption.
Sensitivity analysis
Mass allocation is usually suggested when allocation procedures cannot be avoided and no different physical relationships reflect the way in which the inputs and outputs are changed by quantitative variations in the products delivered by the system (IES 2019). Thus, mass allocation is applied in the first point to the cocoa co-products as defined in paragraph 2.2.2. Since different allocation choices could strongly affect the results and owing to the common use of chocolate as energy food, allocation rules based on the cocoa co-product energy content are proposed for the sensitivity analysis. The caloric intakes for cocoa liquor, cocoa butter and cocoa powder are respectively equal to 648.3 kcal/100 g, 899.05 kcal/ 100 g and 469.6 kcal/100 g (MP&F 2020). As reported in Table 9, this allocation choice leads to a higher allocation percentage for cocoa butter (56.7% instead of 43.9%) and to lower allocation percentages for cocoa liquor and cocoa powder (respectively 19.8% and 23.5% instead of 21.3% and 34.8%), proportionally affecting their environmental impacts. As shown in Figs. 13 and 14 for GWP category, the energy content allocation slightly rises the environmental impacts of both milk and white chocolate because of the increased impacts of cocoa butter. On the contrary, dark chocolate shows almost equal impacts as the presence of all three cocoa coproducts balances the result variation. The change linked to cocoa butter also leads white chocolate to become more impacting than milk chocolate, since cocoa butter-the only cocoa co-product contained in white chocolate-is strongly unfavoured by the energy content allocation. Except for different percentage changes, the same behaviour occurs for all considered impact categories and indicators as shown by the results presented in the Supplementary Material.
Possible variation in the results could also be caused by different proportions among the mass of cocoa co-products obtained in chocolate manufacturing, as cocoa liquor contains both cocoa powder and cocoa butter in roughly equal proportion. Therefore, according to the existing proportion between cocoa butter and powder and maintaining the same overall mass for cocoa co-products (Table 6), different percentage variations in the output of cocoa liquor are applied to the manufacturing phase in the case of energy content allocation. However, as shown in Fig. 15 for GWP, the variation of the results is substantially negligible for all chocolate types in the case of energy content allocation, whereas no change is present in the case of mass allocation.
Finally, a comparison between two allocation methods is evaluated considering a functional unit of 1 kcal. The conversion of functional unit is computed according to average energy content for three chocolate types (Verna 2013): 4950 kcal/kg of dark chocolate, 5150 kcal/kg of milk chocolate and 5400 kcal/kg for white chocolate. Thus, looking at chocolate for its primary function of energy food, the application of an energy-based functional unit turns back to favour-in Fig. 13 GWP impacts of 1 kg dark, milk and white chocolate according to energy content allocation (Ecuador EC, Ghana GH, Indonesia monoculture system IDm, Indonesia agroforestry system IDa) terms of GWP-white chocolate instead of milk chocolate in both the allocation rules applied (Fig. 16), as for the original case of 1 kg of product with mass allocation (Fig. 14). However, regardless of the functional unit and the allocation rules applied, the qualitative comparison among three chocolate types remains similar.
Conclusion
The environmental impact analysis of the food supply chain is becoming a relevant topic due to its considerable consequences and, at the same time, higher attention of consumers to more sustainable product choice. In this context and owing to the continuous increase of cocoa demand, the comparison among dark, milk and white chocolate life cycle is proposed through LCA methodology from cradle to grave. Several possible situations are analyzed, considering different cocoa producer countries. Indeed, each zone and farming technique (monoculture or agroforestry system) has specific environmental impacts depending on requested inputs. The analysis shows that Ghana case study has minor consequences, due to lower use of fertilizers and pesticides and travelled distance between cocoa fields and factory; yet a higher water value is consumed. The raw material production, specifically cocoa co-products and milk powder, has the major influence in all considered categories. In addition, packaging material comparison is proposed analyzing different possible choices. The best solution is a single PP layer, whereas the commonly used aluminium foil with an external fibre-based pack has higher environmental impacts. According to the performed sensitivity analysis, the comparison between two applied allocation procedures-mass and energy content-does not show a remarkable difference, highlighting their equal validity in the application to chocolate LCA studies. In both cases, dark chocolate globally presents the best environmental performance, whereas the other two types have similar environmental impacts. These results are also qualitatively confirmed in the case of calories as functional units.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. | 2020-09-04T14:11:24.692Z | 2020-09-04T00:00:00.000 | {
"year": 2020,
"sha1": "b84d9b1c338cf76239a9bff6cc3af533dee88022",
"oa_license": "CCBY",
"oa_url": "https://link.springer.com/content/pdf/10.1007/s11367-020-01817-6.pdf",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "f248174073ea741c12e19ad523065404b49c6e34",
"s2fieldsofstudy": [
"Agricultural And Food Sciences"
],
"extfieldsofstudy": [
"Business"
]
} |
84842978 | pes2o/s2orc | v3-fos-license | The VHE Gamma-Ray View of the FSRQ PKS~1510-089
The flat spectrum radio quasar PKS 1510-089 is a monitored target in many wavelength bands due to its high variability. It was detected as a very-high-energy (VHE) $\gamma$-ray emitter with H.E.S.S. in 2009, and has since been a regular target of VHE observations by the imaging Cherenkov observatories H.E.S.S. and MAGIC. In this paper, we summarize the current state of results focusing on the monitoring effort with H.E.S.S. and the discovery of a particularly strong VHE flare in 2016 with H.E.S.S. and MAGIC. While the source has now been established as a weak, but regular emitter at VHE, no correlation with other energy bands has been established. This is underlined by the 2016 VHE flare, where the detected optical and high-energy $\gamma$-ray counterparts evolve differently than the VHE flux.
Introduction
The correlations between blazar emissions in different energy bands are best probed with long-term monitoring, providing unbiased sampling. Especially for ground-based observatories this is hard to achieve for even a small number of sources. The Fermi satellite has transformed the monitoring of blazars in the high-energy (HE) γ-ray band (E > 100 MeV) through its continuous surveillance of the whole sky every three hours (although somewhat less uniform after its hardware failure in March 2018) as detailed in [1,2]. In the optical and radio bands many monitoring programs are run thanks to the large number of available telescopes. However, in other energy bands the monitoring capabilities are limited. In the X-ray band the Neil Gehrels Swift observatory runs a limited monitoring effort and can follow up on flares. MAXI on board the International Space Station provides all-sky capabilities within 1 orbit with limited sensitivity. In the very-high-energy (VHE) γ-ray band (E > 100 GeV) the monitoring effort is limited by sensitivity, e.g., for FACT and HAWC [3], or by time constraints due to competition with other objects. The latter strongly influences the monitoring efforts of the three large imaging atmospheric Cherenkov telescope (IACT) facilities H.E.S.S., MAGIC and VERITAS. Nonetheless, they have been running limited monitoring projects on a number of blazars.
Here, we report on the ongoing monitoring efforts by H.E.S.S. and MAGIC of the flat spectrum radio quasar (FSRQ) PKS 1510-089. It is located at a redshift z red = 0.361 and possesses a bright broad-line region (BLR), e.g., [4,5]. Hence, VHE photons produced within the boundaries of the BLR should be absorbed. As the emission region of γ-rays was thought to be close to the central black hole, VHE emission from FSRQs was considered unlikely by many. However, several detections of FSRQs [6][7][8][9][10][11][12] challenge this picture and suggest that jets are able to produce γ-rays also further downstream in the jet.
To verify that these are not simply one-time-only flaring events, but that FSRQs produce VHE emission on all time scales, monitoring programs have been initiated with H.E.S.S. and MAGIC on PKS 1510-089. While these are not unbiased, they have already provided important information. During a strong multiwavelength flaring event in 2015, variability on night-by-night scales at VHE γ-rays was observed for the first time from this source [13][14][15]. Furthermore, MAGIC observations integrated during low-states in the HE band revealed a significant VHE signal with an average, integrated fluxF(E > 150 GeV) = (4.3 ± 0.6) × 10 −12 cm −2 s −1 [16]. Hence, PKS 1510-089 is not only variable in VHE γ-rays but also a persistent source. This has a direct and very important consequence: the absorption of VHE photons through the BLR cannot be too severe, and the emission region must be at the edge or even outside of the BLR at all times. This, in turn, implies that the usual model for FSRQ γ-ray emission, namely inverse-Compton scattering of BLR photons, might not be correct.
This paper gives the status of the H.E.S.S. monitoring efforts on PKS 1510-089, and its early results. One of the important outcomes is the detection of an unprecedented VHE flare in 2016, which was also followed-up with MAGIC. Details of this flare are reported here. Additional multiwavelength data are gathered for comparison from Fermi-LAT in the HE γ-ray band, from Swift-XRT in the X-ray band and from ATOM [17] in the R-band.
Monitoring with H.E.S.S.
After the detection in 2009 [8], H.E.S.S. has continued observing PKS 1510-089 with low cadence. Since 2015 this effort has been significantly increased with several hours of observations each month during the visibility period (which typically lasts from February to July each year) resulting in observations almost every night without moon interference. The resulting nightwise lightcurve including all observations is shown in Figure 1, and a focus on the 2015 and 2016 season is shown in Figure 2. Note that nightwise bins do not guarantee a significant flux per night due to the limited sensitivity and the dimness of the source in the low state. In fact, about 50% of the nights shown in Figure 1 are compatible with zero. The bright VHE flare in 2016 clearly stands out with peak fluxes up to 10 times higher than the previous record holder in 2015. In order to reveal details of the other times, the inset shows the zoom in on the fluxes without the 2016 flare. The flare is further discussed in Section 3. The average, integrated flux for the whole time frame is F(E > 150 GeV) = (5.1 ± 0.3) × 10 −12 cm −2 s −1 , which is compatible within errors with the MAGIC low-flux level, but includes the bright states, as well. The average of the 2015-2016 time frame is compatible with the average of the whole time frame. In both cases, a constant flux is ruled out with very high significance. This is underlined by the fractional variability [18] where S 2 is the variance, σ 2 err is the mean square error, andF is the average flux of the source in the considered data set. For the whole data set F VHE var = 3.2 ± 0.1, and for the 2015-2016 time frame F VHE var = 3.3 ± 0.1. These large values are driven by the 2016 flare. Removing the two nights of that event give F VHE var = 0.8 ± 0.2, which still implies significant variability. Defining the variability time scale between two subsequent flux points as [19] the minimum variability time scale is t VHE var = (0.8 ± 0.06) h, which was exhibited during the major flare in 2016. The error on the variability time scale has been derived through error propagation. Unfortunately, many of the HE flares were not followed up with with H.E.S.S. due to observational constraints. Nonetheless, it is interesting to investigate whether there is any correlation between these two bands. Plotting the simultaneously recorded fluxes of the two bands against each other can reveal direct correlations. The scatterplot is shown for H.E.S.S. and Fermi-LAT fluxes in Figure 3a. The discrete cross-correlation function (DCCF) can uncover correlations with time-delays in non-simultaneous and unevenly spaced data [20]: where F a and F b are the fluxes of two lightcurves with meanF a andF b and variance S a and S b , respectively. The sum goes over all N pairs i, j in the time interval τ. The DCCF between the VHE and HE γ-ray fluxes is shown in Figure 3b for the full time frame and 2015-2016, respectively. The scatterplot does not contain any strong evidence for a direct correlation between VHE (here integrated above 150 GeV) and HE fluxes in the data. While on several occasions a high HE flux is accompanied with a significant VHE flux, this is not a general rule, as also VHE-flux levels compatible with zero are recorded for similar HE fluxes. On the other hand, similar VHE fluxes can occur at different HE-flux levels. The 2016 VHE flare again stands out for the relatively low simultaneous HE fluxes. The non-correlation of VHE and HE fluxes is also underlined by the flat DCCF. It should be noted that the different integration times (a few hours for H.E.S.S. and 24 h for Fermi-LAT) might influence the conclusions here given the relatively fast variability found in this source.
For the 2015-2016 time frame, data from Swift-XRT and ATOM have been analyzed, giving the X-ray and R-band lightcurves for these years. The X-ray average, integrated flux is F(2 keV < E < 10 keV) = (9.5 ± 0.1) × 10 −12 erg cm −2 s −1 . The flux is incompatible with a constant flux with high significance, and F X var = 0.19 ± 0.01. The fastest variability is t X var = 8 ± 4 h. 1 Given the low cadence in these observations as visible in the third panel of Figure 2, it is difficult to distinguish flares from a ground state. In the optical R-band, the average, integrated flux is F(R) = (8.605 ± 0.005) × 10 −12 erg cm −2 s −1 . The flux is highly variable with F R var = 0.679 ± 0.002, and t R var = 10.3 ± 0.5 h. The lightcurve, shown in the bottom panel of Figure 2, reveals a highly active state in 2015 and a mostly quiet state in 2016. In April and May 2015 ATOM recorded correlated activity in the optical band with the HE γ-ray band. The very bright optical flare in July 2015, which was the brightest flux state ever recorded with ATOM in PKS 1510-089, only had a mild counterpart in the HE band. Compared to the other optical flares in 2015, the July outburst was more than 2 times brighter.
Scatterplots have also been produced for VHE versus X-ray and VHE versus R-band fluxes, which are shown in Figure 4a,b, respectively. As the number of data points are low, no DCCFs have been calculated. No conclusions can be drawn from the VHE versus X-ray scatterplot at this point. Unfortunately, no X-ray coverage was obtained during the 2016 VHE flare. The VHE versus R-band scatterplot suggests that high optical fluxes (i.e., above 2 × 10 −11 erg/cm 2 /s) imply significant VHE 1 The large error implies that this value is not highly significant. Trials might reduce the significance further. Hence, this time scale should be regarded as a lower limit. fluxes (i.e., fluxes that deviate by more than 1σ from zero). The low number of data points makes this a weak conclusion. However, high VHE fluxes do not imply high optical fluxes (using the same threshold), as is demonstrated by the 2016 VHE flare.
The 2016 VHE Flare
The summary of the monitoring results in Sec. 2 has already hinted at the unprecedented nature of the flare in 2016. The nightly lightcurve around this event is shown in Figure 5. The flare lasted less than 3 days in the VHE band with a peak in the late hours of MJD 57538 (30 May 2016 -hereafter "maximum night"). In the HE band a flux rise seems to have happened. However, this is barely significant, as it is hovering around the long-term average and more than a factor 10 below previous flares. On the other hand, the spectral index clearly reduces compared to the average ∼2.4, peaking at ∼ 1.6. Hence, while the integrated flux in the HE band barely changed, the spectrum itself significantly hardened. The optical flux rises by a factor of 2 from the beginning of the event to its peak. However, this again is a much smaller flux than that exhibited in previous outbursts. Unfortunately, there is no strictly simultaneous coverage of this flare in any other band.
A detailed lightcurve of the maximum night is shown in Figure 6. The VHE flux exhibits a peak with a flux ∼80% of the Crab above an energy of 200 GeV and a subsequent decay. From the peak to the minimum the flux fell by almost an order of magnitude. As the low flux in the HE band coupled with the small effective area of Fermi-LAT inhibits short-time binning, individual photons recorded with Fermi-LAT with energies E > 1 GeV are shown in the second panel. Fermi-LAT recorded photons with energies up to E ∼ 25 GeV during the H.E.S.S. observation window, but only 2 photons with energies E > 1 GeV in the MAGIC observation window. This is indicative of a softening of the spectrum at that time. The optical R-band flux recorded with ATOM exhibits a double-peaked structure, which is different than the VHE γ-ray lightcurve. The optical flux only changes by ∼30%. The γ-ray spectra of the maximum night are shown in Figure 7 along with the HE and VHE γ-ray low-state spectra [16]. The VHE γ-ray spectra have been corrected for the absorption by the extragalactic background light (EBL) using the model of [21]. The resulting deabsorbed spectra of the flare are compatible with power-laws with indices Γ H.E.S.S. = 2.9 ± 0.2 stat and Γ MAGIC = 3.37 ± 0.09 stat , respectively. The HE γ-ray spectra in the two VHE observation windows are compatible within errors being Γ LAT = 1.4 ± 0.2 stat during the H.E.S.S. time frame and Γ LAT = 1.7 ± 0.2 stat during the MAGIC time frame, respectively. The spectral breaks between the HE and the VHE γ-ray band are, therefore, ∆Γ = 1.5 ± 0.3 stat during the H.E.S.S. time frame and ∆Γ = 1.7 ± 0.2 stat during the MAGIC time frame. The comparison with the low-state data clearly shows the significant shift of the peak energy from E ∼ 100 MeV to E ∼ 30 GeV during this flare. The spectral break can have several causes. The underlying particle distribution could exhibit a break due to the interplay of acceleration and cooling. The break could also be a sign of the Klein-Nishina reduction of the inverse-Compton cross-section at high energies. That the break remains roughly constant between the respective observation time windows while the spectra get softer, could be an indication of the Klein-Nishina reduction. A third possibility is that the spectral softening results from the absorption of VHE photons by external soft photon fields, such as those from the BLR. The resulting optical depth for different VHE emission region distances from the black hole is shown in Figure 8. Obviously, the distance of the emission region from the black hole has a strong influence on the optical depth τ γγ .
One can conservatively estimate an upper limit on the degree of absorption by assuming that the spectrum detected by Fermi-LAT represents also the intrinsic spectrum in the VHE domain. The degree of absorption τ can then be derived by where F extra is the extrapolated flux, and F obs is the observed flux. Without going into details [14], the calculation for the H.E.S.S. data set give a maximum value of τ for the highest energies of τ = 5.4 ± 0.9 stat , while the MAGIC data gives a maximum value of τ = 3.9 ± 1.4 stat . These estimates agree within errors. Assuming the absorption is due to the BLR, the absorption values can be translated into a minimum distance of the emission region from the black hole. The emission region could be located at roughly 2 × R Ly α ∼ 1.6 × 10 17 cm ∼ 0.05 pc from the black hole (c.f. Figure 8). Assuming that the distance of the Ly α line represents the radius of the BLR [22], the flaring region would at the very least be located on the outer edge of the BLR. This underlines the statement given above during the discussion of the monitoring data: the jet must be able to produce VHE γ-rays on distances on the order of a significant fraction of a parsec from the black hole.
Summary & Conclusions
FSRQs are by now established VHE γ-ray emitters. However, whether they are able to produce the VHE emission at all times or only during short bright flares has been an open question. This has led to the establishment of monitoring programs by the IACT experiments H.E.S.S. and MAGIC on the FSRQ PKS 1510-089, which is one of the closest of this type of blazars (redshift z red = 0.361). These programs are supplemented with multiwavelength data in the HE γ-ray, X-ray and optical regime. The first important result obtained by MAGIC is that PKS 1510-089 can be detected at VHE γ-rays during low states in the HE band [16]. Similarly, variability has also been established through MAGIC observations [13].
The H.E.S.S. monitoring described in detail here adds important features to these earlier results. Strong variability is detected in the VHE γ-ray domain, while the observations also hint to a persistent flux at other times. Interestingly, comparison of the VHE γ-ray lightcurve with other energy bands does not reveal any obvious correlation. This exemplifies the need for deep monitoring programs across the entire multiwavelength spectrum. Otherwise important effects-such as better sampling of correlation functions, variability time scales, etc.-might be missed for the interpretation of certain events.
The latter statement is further emphasized by the detection of an unprecedented VHE γ-ray flare in 2016 with H.E.S.S. that was followed up with MAGIC, as well. It was more than 10 times brighter than any flux seen at VHE γ-rays before (with a peak flux of 80% of the Crab) and lasted only 2 days. It was accompanied by a significant hardening of the HE γ-ray spectrum as observed with Fermi-LAT, while the HE fluxes remained rather low compared to other flares. The optical R-band observations with ATOM revealed a mild counterpart that was also much dimmer than previous flares, but exhibited a different flux evolution compared to the VHE band. Unfortunately, no other simultaneous data is available that could further constrain the spectrum.
All these observations reveal that the jet of PKS 1510-089 is able to accelerate particles to high energies to produce VHE γ-rays at all times. It also implies that these emission regions are probably located beyond the BLR, as otherwise the VHE emission should be strongly absorbed. This has been shown here specifically for the 2016 VHE flare. A simple estimate of the maximum absorption allowed for by the data results in a lower limit on the black hole distance, which indicates a location of the emission region on the edge of or beyond the BLR.
In conclusion, deep and, preferably, unbiased monitoring programs on FSRQs and blazars in general are important to reveal the general behavior of the sources, as well as to uncover new and unexpected features. | 2019-04-23T13:23:59.780Z | 2019-03-20T00:00:00.000 | {
"year": 2019,
"sha1": "c38fed9782eee8e94ae98f95d5864c7b459ed70a",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "a77e8b257110438d2a9ab0a31fe655555bf165e6",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
18185795 | pes2o/s2orc | v3-fos-license | Thoracic Duct Injury Following Cervical Spine Surgery: A Multicenter Retrospective Review
Study Design: Multicenter retrospective case series. Objective: To determine the rate of thoracic duct injury during cervical spine operations. Methods: A retrospective case series study was conducted among 21 high-volume surgical centers to identify instances of thoracic duct injury during anterior cervical spine surgery. Staff at each center abstracted data for each identified case into case report forms. All case report forms were collected by the AOSpine North America Clinical Research Network Methodological Core for data processing, cleaning, and analysis. Results: Of a total of 9591 patients reviewed that underwent cervical spine surgery, 2 (0.02%) incurred iatrogenic injury to the thoracic duct. Both patients underwent a left-sided anterior cervical discectomy and fusion. The interruption of the thoracic duct was addressed intraoperatively in one patient with no residual postoperative effects. The second individual developed a chylous fluid collection approximately 2 months after the operation that required drainage via needle aspiration. Conclusions: Damage to the thoracic duct during cervical spine surgery is a relatively rare occurrence. Rapid identification of the disruption of this lymphatic vessel is critical to minimize deleterious effects of this complication.
Introduction
The thoracic duct is the body's largest lymphatic vessel; its proximity to the vertebral bodies makes it an important anatomical consideration during low anterior spine surgery. Variability in the anatomic course of the vessel, as well as its small size, can result in potential injury. 1 Damage to the thoracic duct during spinal operations is a rare but potentially serious complication that manifests as a cervical chylous fistula, chylothorax, or chyloretroperitoneum. 2 Serious insult to the structure can result in significant nutritional deficiency, respiratory dysfunction, and considerable immunosuppression. 3,4 Conservative therapy initially consists of diet modification, electrolyte monitoring, and appropriate drain placement. 5 If needed, definitive treatment requires ligation of the thoracic duct, a procedure that can result in significant morbidity. 6 Current data regarding the occurrence of thoracic duct injury during spine surgery is primarily found as isolated case reports. [7][8][9][10][11][12][13][14] This study is the first to report the overall incidence of thoracic duct injury during cervical spine surgery based on a multicenter retrospective review.
Methods
We have conducted a retrospective multicenter case series study involving 21 high-volume surgical centers from the AOSpine North America Clinical Research Network, selected for their excellence in spine care and clinical research infrastructure and experience. Medical records for 17 625 patients who received cervical spine surgery (levels from C2 to C7) between January 1, 2005, and December 31, 2011, inclusive, were reviewed to identify occurrence of 21 predefined treatment complications. The complications included reintubation for the purpose of a hematoma evacuation, esophageal perforation, epidural hematoma, C5 palsy, recurrent laryngeal nerve palsy, superior laryngeal nerve palsy, hypoglossal or glossopharyngeal nerve palsy, dural tear, brachial plexopathy, blindness, graft extrusion, misplaced screws requiring reoperation, anterior cervical infection, carotid artery injury or cerebrovascular accident, vertebral artery injuries, Horner's syndrome, thoracic duct injury, quadriplegia, intraoperative death, revision of arthroplasty, and pseudomeningocele. Trained research staff at each site abstracted the data from medical records, surgical charts, radiology imaging, narratives, and other source documents for the patients who experienced one or more of the complications from the list. Data was transcribed into studyspecific paper case report forms. Copies of case report forms were transferred to the AOSpine North America Clinical Research Network Methodological Core for processing, cleaning, and data entry. Descriptive statistics were provided for baseline patient characteristics.
Results
Of the 17 625 total patients reviewed, 9591 individuals underwent surgery of the cervical spine using an anterior approach only. These cases were reviewed to identify cases of iatrogenic thoracic duct injury. There were 2 (0.02%) instances of damage to the thoracic duct. The highest incidence for any single institution was 0.079%, while 19 of 21 institutions reported zero instances of thoracic duct injury.
Case 1
The first occurrence of thoracic duct injury was in a previously healthy male that was involved in a motor vehicle accident. As a consequence, the patient developed radicular pain for which physical therapy did not provide adequate symptom relief. Therefore, the patient opted to undergo anterior cervical discectomy and fusion (ACDF) at the level of C5-C6. The operation was conducted using a left-sided approach. Intraoperatively, a chyle leak was noted and the otolaryngology team was consulted to address the disruption of the thoracic duct. Surgical clips were utilized to control the leakage. The patient was discharged home on postoperative day 1. There was one postoperative visit 80 days following discharge. No residual effects were noted from the intraoperative leak during the follow-up period.
Case 2
This patient underwent ACDF to address a herniated nucleus pulposus with associated radiculopathy at the level of C5-C6. The patient was female, had a body mass index of 21.9, and had no comorbid conditions at the time of the index surgery. A left-sided approach was used to gain access to the cervical spine. There were no issues intraoperatively and no chyle leak was noted. The patient was discharged home following the operation in stable condition. Approximately 2 months after surgery, the patient presented for outpatient follow-up during which a 2.5 Â 2.5 cm nontender, mobile mass was noted superior to the incision. There were no accompanying symptoms of fever or dysphagia; no signs of infection were noted around the incision. Ultrasound revealed a complex fluid collection sizing 2.3 cm in diameter. Needle aspiration of the lesion removed 3.5 mL of milky white fluid identified as chyle. The patient recovered uneventfully and no additional problems were noted during further follow-up. In all, the patient had 8 postoperative visits with the most recent occurring 3.5 years following the index operation.
Anatomy of the Thoracic Duct
The thoracic duct serves as the primary conduit for the return of lymph to the bloodstream from all lymphatic vessels except those found on the right side of the head, neck, thorax, and arm ( Figure 1). As such, it delivers three quarters of the lymph produced in the body to the venous circulation. 15 The length of the thoracic duct ranges from 36 to 45 cm in adults. Its origin is in the retroperitoneum at the cisterna chyli, located on the anterior surface of the first or second lumbar vertebra. After reaching the fifth thoracic vertebra, its course veers left and continues its ascent posterior to the aortic arch and the thoracic portion of the left subclavian artery. Traveling between the left side of the esophagus and pleura, it reaches the root of the neck and forms an arch rising approximately 3 to 4 cm above the clavicle. At the superior border of the clavicle, the thoracic duct is bordered by the left carotid sheath anteriorly, the omohyoid muscle laterally, the anterior scalene fascia posteriorly, and the esophagus medially. It then crosses anterior to the subclavian artery, vertebral artery and vein, and the thyrocervical trunk and terminates by opening into the junction of the left subclavian and left internal jugular veins. 16 Important variations in the course of the thoracic duct have been reported as a result of various cadaver studies. Gottlieb and Greenfield found one cadaver out of 75 with a thoracic duct that remained on the right side during its ascent and emptied into the right internal jugular vein. 17 The height of the arch of the thoracic duct in the root of the neck also varies, as it can be inferior to, at, or superior to the clavicle. 18 Hart et al described a case in which the arch was situated 7 to 8 cm above the clavicle. 13 Importantly, there is significant deviation in the termination pattern of the thoracic duct as it joins the systemic circulation. It may end as single or multiple outlets into the left internal jugular vein, the left subclavian vein, the left external jugular vein, the left brachiocephalic vein, the left transverse cervical vein, or the right internal jugular vein. [17][18][19][20] Greenfield and Gottlieb's study of 75 cadavers found 89.4% of thoracic ducts to have 1 termination, 6.6% to have 2 terminations, and 4% to have 3 terminations. 17
Chyle Leak During Spinal Surgery
The majority of reported cases of thoracic duct injury during spine operations have occurred during surgery of the thoracolumbar spine. [7][8][9][10][11][12] Colletta and Mayer describe a case report in which an intraoperative chyle leak is noted. 7 Despite attempted intraoperative repair, the patient developed a postoperative chylothorax, which was subsequently managed with a chest tube and low-fat diet until resolution was achieved. Nakai and Zielke reported 6 instances of chylothorax in an estimated 2000 cases of thoracolumbar spinal surgery; 8 all were treated conservatively with a combination of diet modification, needle aspiration, and chest tube placement. Propst-Proctor et al surveyed a total of 10 orthopedic spine surgeons who had combined to perform an estimated 1000 anterior thoracolumbar operations. 9 They reported a total of 3 cases of chylothorax that were managed conservatively with closed chest tube drainage and a restricted fat diet. Similarly, Bhat and Lowery reported 3 operations complicated by chyloretroperitoneum and/or chylothorax that resolved with a combination of diet modification, chest tube drainage, and/or continued maintenance of a retroperitoneal drain. 10 Su and Chen also described a patient that experienced retroperitoneal chyle leakage after undergoing lumbar fusion; 11 this patient had spontaneous cessation of drainage upon temporary clamping of his drain tube. Hussain et al described a series of 4 patients that developed chyloretroperitoneum following anterior lumbar surgery. 12 Three patients' symptoms resolved with fluoroscopic-guided percutaneous aspiration as well as drainage catheterization. The fourth patient had continued drainage of her abdominal lymphocele despite placement of a catheter and required reoperation 6 weeks after the initial surgery to create a peritoneal window to allow drainage of the lymphocele into the peritoneal space.
Thoracic duct injury during cervical spine surgery is sparsely described in the literature. Two case reports describing its occurrence in this setting were identified. Hart et al described a case of a patient undergoing a left-sided ACDF in which an intraoperative chyle leak was identified secondary to disruption of the thoracic duct in the supraclavicular region. 13 In this case, the arch of the thoracic duct extended 7 to 8 cm above the clavicle, making it significantly more superior than expected. Intraoperative suture ligation of the duct was conducted and the leakage was confirmed to have stopped; there were no postoperative complications. Warren et al reported a case of cervical lymphocele formation following right-sided ACDF during which the right lymphatic duct was penetrated. 14 Surgical closure of the lymphocele attempted 2 weeks postoperatively was unsuccessful; subsequent percutaneous drainage and sclerotherapy revealed a large branch of the thoracic duct communicating with the lymphocele. This branch was embolized via a percutaneous transcervical approach, resulting in immediate resolution of the patient's symptoms.
The rate of thoracic duct injury during anterior cervical spine surgery was 0% at the majority of institutions reviewed (19 of 21). The highest incidence at any single institution was found to be 0.079%. Estimates by Nakai and Zielke as well as Propst-Proctor et al suggest that the rate of thoracic duct injury during thoracolumbar surgery is approximately 0.3% 8,9 : Nakai and Zielke reported 6 cases of injury out of an estimated 2000 cases, and Propst-Proctor et al found 3 instances out of an estimated 1000 patients. The present study is the first to report an incidence of thoracic duct injury following cervical spine surgery, revealing a substantially lower incidence of this complication in cervical surgery versus thoracolumbar. This discrepancy is possibly due to the longer anatomic course of the thoracic duct in the thoracolumbar area as compared to the cervical spine.
Chyle Leak During Other Surgeries
The thoracic duct is subject to injury during any operation that involves the root of the neck. Such procedures are commonly conducted by thoracic surgeons and otolaryngologists in addition to spine surgeons. For instance, Marthaller et al describe a series of patients that incurred damage to the thoracic duct following esophagectomy; 21 this manifested through the formation of a chylothorax and/or a high-output chylous fistula. For such cases, the authors advocated for percutaneous embolization of the thoracic duct prior to open re-exploration. Similarly, Patel et al describe a case of thoracic duct injury following neck dissection that was treated by percutaneous embolization. 22 Thoracic duct damage leading to chylous leakage occurs in 1% to 2.5% of radical neck dissections. 23 In a retrospective review of 221 patients that underwent neck dissection, de Gier et al identified 11 individuals that suffered a chylous fistula. Dietary modifications alone were sufficient to stop the leak in 5 of these patients. The other 6 cases necessitated initiation of total parenteral nutrition. Two of these individuals required surgical re-intervention via a pectoralis major muscle flap transfer. Overall, there were no cases that resulted in permanent damage.
Identification and Management
If left untreated, injury to the thoracic duct can lead to chronic chyle loss and eventually result in significant weakness, dehydration, and emaciation. 24 Unfortunately, intraoperative diagnosis of lymphatic vessel injury can be difficult due to the usual fasting state of patients prior to surgery, which reduces lymphatic production and transport. Positive pressure application with the patient in Trendelenburg position may facilitate identification of a suspected leak. If a leak is observed intraoperatively, suture ligation of the corresponding lymphatic vessel should be carried out. Postoperative presentation of a thoracic duct injury is dependent on the location of the injured vessel, although the presence of a variably cloudy and white fluid (ie, chyle) from an operative drain or upon aspiration is often seen. Conservative management of a chylothorax in these circumstances consists of a low-fat diet or total parenteral nutrition, supportive care (eg, correction of electrolytes, rehydration), and, in severe cases, adequate drainage via insertion of chest tube. 5,25 Treatment with somatostatin or octreotide has also been shown to be effective in treating thoracic duct injury, as it decreases drainage output and results in earlier fistula closure. [26][27][28] Definitive management in refractory cases involves ligation of the thoracic duct; 6 however, conservative therapy for at least 1 to 2 weeks is recommended prior to taking this course.
Limitations
Limitations to consider when interpreting the results of this study include the consideration of biases (eg, selection bias, information bias) inherent to retrospective reviews. The multicenter, retrospective nature of the study prevented a standard methodology for identifying, documenting, and monitoring iatrogenic injury postoperatively between various institutions. This variation in follow-up protocols may have resulted in an underreporting of cases of thoracic duct injury occurring during the postoperative period. Additionally, there were likely other cases of thoracic duct injury that would have increased our understanding of the natural history and treatment of this complication that we did not encounter as they occurred on other services (eg, otolaryngology, thoracic surgery).
Conclusion
Complications involving the thoracic duct are rarely encountered during anterior cervical spine operations. Nevertheless, surgeons operating in this area need to remain cognizant of the presence of this critical lymphatic structure and should be cautious of anatomic variants that may increase the risk of injury. The manifestations resulting from thoracic duct disruption are dependent on the degree of injury as well as its location. In many cases, prompt identification of damage to the duct will allow for conservative management and resolution of symptoms without significant long-term effects.
Authors' Note
This study was ethically approved by the institutional ethics committees at all participating sites.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Adeeb Derakhshan reports grants from AOSpine North America during the conduct of the study; Michael P. Steinmetz reports grants from AOSpine North America during the conduct of the study; Gabriel A. Smith reports grants from AOSpine North America during the conduct of the study; Ziya Gokaslan reports grants from AOSpine North America, personal fees from AO Foundation, grants from AOSpine, outside the submitted work; Michael G. Fehlings reports grants from AOSpine North America during the conduct of the study; K. Daniel Riew reports personal fees from AOSpine International, other from Global Spine Journal, other from Spine Journal, other from Neurosurgery, personal fees from Multiple Entities for defense, plantiff, grants from AOSpine, grants from Cerapedics, grants from Medtronic, personal fees from AOSpine, personal fees from NASS, personal fees from Biomet, personal fees from Medtronic, nonfinancial support from Broadwater, outside the submitted work; Thomas E. | 2018-04-03T05:47:54.167Z | 2017-04-01T00:00:00.000 | {
"year": 2017,
"sha1": "a6f171b793259ebbe86900915aa33f37a2541d12",
"oa_license": "CCBYNCND",
"oa_url": "https://journals.sagepub.com/doi/pdf/10.1177/2192568216688194",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "a6f171b793259ebbe86900915aa33f37a2541d12",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
239022510 | pes2o/s2orc | v3-fos-license | The Outcome of Induction Only Versus Induction and Maintenance With A Low Dose Intravesical BCG Therapy For High-Grade NMIBC
Background: According to standard guidelines, high-risk NMIBC (Non-Muscle Invasive Bladder Cancer) is treated by TUR (Transurethral resection) followed by intravesical immunotherapy(BCG). Induction followed by maintenance is recommended for achieving maximum delay of tumor recurrence and progression. Objective: This study aimed to find the outcome of induction only vs induction and maintenance, considering recurrence and progression of the disease. Materials and Methods: This experimental study was conducted in BSMMU from June 2018 to December 2020 among the histologically proved high-risk NMIBC (Ta, T1, and/ Tis) patients. Patients were allocated in two groups. In one arm BCG induction only and another arm induction and maintenance were implemented. Patients were followed up upto 2 years period. Disease recurrence and progression along with different local and systemic adverse effects were recorded add analyzed. Results: Total 30 patients were allocated in 2 groups.14 patients in induction only arm and 16 in induction and maintenance arm. Upto 24 months follow upoverall disease recurrence was 23.3% and disease progression was found in 10 % of patients. 28.5% of the induction-only arm and 18.75% patients of induction and maintenance arm developed recurrence. Whereas 7.14% of the induction-only arm and 12.5% of other arm patients developed disease progression. Both were statically insignificant (p 0.198). Conclusion: For high-risk NMIBC inductiononly BCG therapy is not inferior to induction and maintenance therapy in terms of recurrence and progression, rather it has relatively fewer adverse effects.
INTRODUCTION
Bladder cancer is a common genitourinary malignancy. The Incidence of bladder tumor is rising sharply throughout the world in both males and females due to various causes. According to standard guideline NMIBC (Non-Muscle Invasive Bladder Cancer), those are classified as high-risk group -treated by TUR(Transurethral resection) followed by intravesical immunotherapy(BCG). Since it was discovered by Morales et al in 1976 [13], Till now it is commonly practiced treatment protocol. According to many standard randomized trials, BCG (Bacillus Calmette-Guerin) reduces disease recurrence and delay disease progression than TUR alone. SWOG (Southwest Oncology Group) BCG protocol is the commonly used regimen for BCG therapy. According to SWOG-21 doses of BCG were installed in the bladder in three years period. The BCG produces cumulative toxicity. So only 16% of patient of the SWOG trial could continue their treatment upto 3 years because of the toxicity [10]. The majority of current guidelines recommend maintenance BCG for 1-3 years. [5] Many other standard studies mentioned that maintenance is optional regarding efficacy and toxicity. [12] So BCG induction only therapy or induction and maintenance which is appropriate for high-risk NMIBC is still a controversial issue. Many studies are ongoing on this issue. This study aimed to find the outcome of induction only vs induction and maintenance, considering recurrence and progression of the disease.
MATERIALS AND METHODS
This Experimental study was conducted in BSMMU from June 2018 to December 2020. After TURBT high-risk NMIBC patients were included in the registry. High-grade Ta, T1, and/ Tis were considered as high-risk group. Patients were allocated in two groups. In one arm BCG induction only and another arm induction and maintenance were implemented. Patients were distributed in two groups according to Even and Odd serial numbers in the registry. Those having concomitant other malignancies, poor performance status (ECOG/WHO-3,4), unwilling to participate in the study were excluded. For induction only arm BCG(40 milligrams dissolved in 40 ml saline) was installed in the bladder -weekly for 6 weeks starting at least 14 days after TURBT(transurethral resection of bladder tumor). In another arm, induction and up to 3 years maintenance was planned according to SWOG guideline. For those patientsinstallation weekly for 6 weeks, then maintenance three weekly at 3,6,12,18,24,30 and 36 months was scheduled.
In this study follow up performed upto 2 years after TURBT. Patient scheduled for Induction only therapy if developed recurrence were treated by another 6 weeks of BCG instillation. Patients were followed up at 3,6, 9 months, and at1 year, thereafter 6 monthly with history, physical examination, cytology and cystoscopy. In this study complete response was defined as negative cytology and biopsy at 6 months follow up. During this period positive cytology or presence of tumor in biopsy specimen was considered as recurrence. On the other hand-muscle invention in the biopsy or metastasis was considered as disease progression. During this study period different local systemic adverse effects like-pain, frequency, haematuria, fever, malaise, were recorded and plotted accordingly. Data analysis performed using computer-based software.p vale <0.05 was considered as significant.
RESULTS
In this study total 30 patients were followed up for 2 years period. Out of them-14 patients were in induction only arm, whereas 16 patients were in another arm (Table-1). (Table 2) Induction only arm 4(28.50%) patients developed recurrence, out of them-1(7.14%) patient had disease progression another three patients had a recurrence but no progression found. Recurrence was re-treated with weekly 6 dose of intravesical BCG. 1 patient who developed disease progression died due to bladder cancer another (1) patient of this arm died due to CVS-related complications. In the maintenance arm, three patients(18.75%) developed recurrence among them two patients (12.5%) developed disease progression one patient diet later due to bladder cancer. Another patient developed upper tract TCC(transitional cell carcinoma) presented with lung metastasis and diet from TCC. 1 patient diet due to CVSrelated complications. (Table 3) Four patients developed recurrence of the induction-only arm. All of them developed in the second year. (two at 18 months, another two at 24 months). In the maintenance arm, 3 patients developed recurrence-one at 9 months, one at 18 months, another at 24 months. Regarding disease progression, one patient of the induction arm developed progression at 24 months. In another arm, two patients developed disease progression at 9 months and at 18 months. Both recurrence and progression were statically insignificant comparing between the two groups. (Fig.-1) Recurrence-free survival up to 2 years in induction only arm was 71.4% and 81.25% in another arm. (Fig.-2) Progressionfree survival upto two years was 92.8% in induction only arm but 87.5% in the opposite arm. -3).
DISCUSSION
In this study 30 patients of high-risk NMIBC-14 were in induction only arm and 16 were in induction and maintenance arm. Upto 24 months follow up ( 28.5%) of induction only and (18.75%) patients of another arm developed recurrence. 7.14% of the induction-only arm and 12.5% of patients of the opposite arm developed disease progression. Total 16.6% of patients died in this 2 year follow-up period.40% patient of them were in the 1 st group and 60% of patients were in the 2 nd group. The result of this study showed nearly similar recurrence and progression-free survival in comparison to some other standard randomized trials. 2-year recurrence-free survival rate was 72% on maintenance group and 10.4% of those patients had disease progression, which is similar to our data. [14] Standard review literature stated that 24 months recurrence-free survival rate varying between 54-89% by BCG therapy. [6] Another study result 2-year recurrence-free survival was 73% and progression-free survival was 11% in the induction only arm. [4] Local and systemic adverse effects by the standard dose of BCG induction (81 mg suspended in 50 ml saline) were relatively higher than this study. [10] In our study patient received nearly half of the standard dose BCG( 40 mg dissolved in 40 ml saline). this may be the cause of reduced adverse effects during this maintenance period without compromising the oncological outcome. It indicates BCG installation with a reduced dose can produce similar immunological efficacy. After all, results showed that the induction-only arm is not inferior to the induction and maintenance arm in terms of recurrence and progression upto 24 months follow-up period.
LIMITATIONS
It was a single-centered study where observation weres made on basis of 2 years follow-up period. Longer follow-up at least 5 years involving multiple centers, will be more effective regarding comments on decision making for adjuvant BCG therapy in high-risk NMIBC.
CONCLUSION
Maintenance BCG can reduce a few recurrences but the number is statically insignificant. Moreover maintenance therapy producers relatively more local and systemic adverse effects. In terms of recurrence and progression induction only is not inferior to induction and maintenance therapy rather it has relatively fewer adverse effects. So all the high-risk patient does not necessarily need maintenance therapy. However, in relapsing patients retreatment with induction therapy may be a logical salvage option. | 2021-08-27T16:41:48.690Z | 2021-01-01T00:00:00.000 | {
"year": 2021,
"sha1": "3bf0a2bfc434f0b6b5c065dfa91a7b69de692d4a",
"oa_license": null,
"oa_url": "https://doi.org/10.20431/2456-060x.060103",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "5fd6168cb54feeb63163f68aff0f5b2fee6f5de2",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
221614311 | pes2o/s2orc | v3-fos-license | Challenging the Erythropoiesis Paradigm in β-Thalassemia
Iron Research Program, Lindsley Kimball Research Institute, New York Blood Center, New York, NY, USA Department of Pathology and Laboratory Medicine, Weill Cornell Medicine, New York, NY, USA. The authors have indicated they have no potential conflicts of interest to disclose. Copyright © 2020 the Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the European Hematology Association. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 License, which allows others to remix, tweak, and build upon the work non-commercially, as long as the author is credited and the new creations are licensed under the identical terms. HemaSphere (2020) 4:5(e475). http:// dx.doi.org/10.1097/ HS9.0000000000000475. Received: 14 July 2020 / Accepted: 28 July 2020 -thalassemia is a genetic disorder due to defective -globin synthesis that results in severe anemia. A hallmark of the disease is the massive expansion of the erythroid b compartment associated with enhanced but ineffective erythropoiesis. Whether this result in remodeling and homeostatic changes of the bone marrow (BM) niche and what are the consequences on disease pathophysiology is unclear and remains understudied. For many years, the BM microenvironment was considered an inert scaffold, providing structural support for hematopoietic stem and pro genitor cell (HSPC) contained within. In the past decade, however, the BM microenvironment has been described as a vibrant and complex living tissue with crucial homeostatic roles in hematopoiesis. Along with cell-intrinsic factors, hematopoietic stem cell (HSC) self-renewal is maintained by extrinsic factors coming from the circulation and produced by the local microenvironment. Importantly, the BM microenvironment contributes to the maintenance of HSC homeostasis through a dynamic crosstalk with BM stromal cells (e.g. endothelial and mesenchymal stem cells, MSC). Although HSC transplant (HSCT) is an effective and currently the only curative approach for the correction of the erythropoietic defect, the risk of graft rejection in b-thalassemia is higher than in other indications for HSCT. A recent study by Aprile et al has brought the thalassemic BM niche into focus as potential explanation for limited HSCT success in this disease. By using the Hbb (th3) mouse model that recapitulates the b-thalassemia phenotype, the authors demonstrated that th3 mice have both a reduced number and more active HSCs. th3 HSCs show a loss of quiescence and a higher cycling rate (Fig. 1). The elevated DNA damage and reduced clonogenic potential highlight the increased replication stress and impaired self-renewal ability of HSCs in b-thalassemia. Through HSCT experiments, the authors demonstrated how an improved BM microenvironment can positively affect HSC behavior and ameliorate their functions.th3HSCs transplanted in a recipient th3 BM show a competitive disadvantage compared to their co-transplanted WT counterparts. On the contrary, transplantation of th3HSCs into a recipient wild-type BM restored HSC repopulating ability. Study of secondary transplants showed that th3 HSC transplant into wild-type recipients retained a long-term reconstitution ability, while their transplant into th3 recipients lead to progressive HSC exhaustion. To explain the disparity between transplant success, the authors next turned to the BM stromal niche. While there was no evident alteration in osteoblasts or osteoclasts, a decrease in bone network and mineral density, together with reduced type-I collagen levels were observed. Reduced serum alkaline phosphatase, osteopontin (OPN), and Notch-ligand Jagged 1 (JAG1) in th3 BM suggested altered osteoblast and MSC functions in b-thalassemia. Of note, OPN and JAG1 expression in osteoblasts and MSCs is dependent on the parathyroid hormone (PTH) and their reduction leads to a loss of HSC quiescence. The PTH axis is a critical hormonal system that controls bone remodeling and also provides signals for HSC maintenance via its specific receptor (PTHR) on BM stromal cells. In b-thalassemia reduced circulating PTH levels likely contributes to a defective and poorly supportive BM microenvironment. The administration of PTH in th3 mice restored bone density as well as OPN and JAG1 levels and improves the pool of quiescent HSCs (Fig. 1). In line with findings in the mouse model, patients with transfusion-dependent b-thalassemia (TDT) show reduced HSPC quiescence. Moreover, HSPC gene expression profile revealed upregulation of genes involved in cellular responses to oxidative stress and DNA damage as well altered BM collagen and JAG1 levels in TDT patients.
b -thalassemia is a genetic disorder due to defective -globin synthesis that results in severe anemia. A hallmark of the disease is the massive expansion of the erythroid compartment associated with enhanced but ineffective erythropoiesis. Whether this result in remodeling and homeostatic changes of the bone marrow (BM) niche and what are the consequences on disease pathophysiology is unclear and remains understudied. For many years, the BM microenvironment was considered an inert scaffold, providing structural support for hematopoietic stem and pro genitor cell (HSPC) contained within. In the past decade, however, the BM microenvironment has been described as a vibrant and complex living tissue with crucial homeostatic roles in hematopoiesis. 1,2 Along with cell-intrinsic factors, hematopoietic stem cell (HSC) self-renewal is maintained by extrinsic factors coming from the circulation and produced by the local microenvironment. Importantly, the BM microenvironment contributes to the maintenance of HSC homeostasis through a dynamic crosstalk with BM stromal cells (e.g. endothelial and mesenchymal stem cells, MSC). Although HSC transplant (HSCT) is an effective and currently the only curative approach for the correction of the erythropoietic defect, the risk of graft rejection in b-thalassemia is higher than in other indications for HSCT. 3 A recent study by Aprile et al has brought the thalassemic BM niche into focus as potential explanation for limited HSCT success in this disease. 1 By using the Hbb th3/+ (th3) mouse model that recapitulates the b-thalassemia phenotype, the authors demonstrated that th3 mice have both a reduced number and more active HSCs. th3 HSCs show a loss of quiescence and a higher cycling rate (Fig. 1). The elevated DNA damage and reduced clonogenic potential highlight the increased replication stress and impaired self-renewal ability of HSCs in b-thalassemia.
Through HSCT experiments, the authors demonstrated how an improved BM microenvironment can positively affect HSC behavior and ameliorate their functions. 1 th3 HSCs transplanted in a recipient th3 BM show a competitive disadvantage compared to their co-transplanted WT counterparts. On the contrary, transplantation of th3 HSCs into a recipient wild-type BM restored HSC repopulating ability. Study of secondary transplants showed that th3 HSC transplant into wild-type recipients retained a long-term reconstitution ability, while their transplant into th3 recipients lead to progressive HSC exhaustion.
To explain the disparity between transplant success, the authors next turned to the BM stromal niche. While there was no evident alteration in osteoblasts or osteoclasts, a decrease in bone network and mineral density, together with reduced type-I collagen levels were observed. Reduced serum alkaline phosphatase, osteopontin (OPN), and Notch-ligand Jagged 1 (JAG1) in th3 BM suggested altered osteoblast and MSC functions in b-thalassemia. Of note, OPN and JAG1 expression in osteoblasts and MSCs is dependent on the parathyroid hormone (PTH) and their reduction leads to a loss of HSC quiescence. 4,5 The PTH axis is a critical hormonal system that controls bone remodeling and also provides signals for HSC maintenance via its specific receptor (PTHR) on BM stromal cells. In b-thalassemia reduced circulating PTH levels likely contributes to a defective and poorly supportive BM microenvironment. The administration of PTH in th3 mice restored bone density as well as OPN and JAG1 levels and improves the pool of quiescent HSCs 1 (Fig. 1).
In line with findings in the mouse model, patients with transfusion-dependent b-thalassemia (TDT) show reduced HSPC quiescence. Moreover, HSPC gene expression profile revealed upregulation of genes involved in cellular responses to oxidative stress and DNA damage as well altered BM collagen and JAG1 levels in TDT patients.
HemaTopics
Overall, these findings support the concept that, while HSCT is an effective strategy to cure b-thalassemia, its successful accomplishment also depends upon the improvement of the BM microenvironment. 1,2 BM niche alterations results in impaired niche-HSC crosstalk, leading to HSC exit from quiescence and premature HSC exhaustion in b-thalassemia. This suggests that HSCs develop impaired functions and loose stemness upon prolonged persistence in an altered BM microenvironment. Recent evidence indicates that stress signals, including oxidative stress, inflammation, iron, and hypoxia derived from ineffective erythropoiesis and chronic transfusions, may alter the BM niche, leading to MSC pool pauperization and defective microenvironmental support for HSC. 2 Importantly, these observations indicate that the pathophysiology of b-thalassemia is not exclusively a dyserythropoiesis disorder, as the long-standing paradigm states, but a more complex disorder of hematopoiesis resulting from an impaired BM microenvironment and altered HSC functions. 1,2 These findings lay the groundwork for combined therapies which correct the erythropoietic defect and improve the BM microenvironment as well as HSC functions in pathologic conditions hallmarked by dyserythropoiesis. Targeting the BM microenvironment and correcting the HSCstromal niche crosstalk will also provide a valuable strategy to improve transplantation and gene therapy approaches in these diseases. | 2020-09-03T09:03:43.485Z | 2020-08-27T00:00:00.000 | {
"year": 2020,
"sha1": "9c4425df38b550604839e2e7604d3b27ff605f8e",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1097/hs9.0000000000000475",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "308f67d4ad342d93bd53fd4c88d621b3a92b6bc0",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
54702956 | pes2o/s2orc | v3-fos-license | High-Performance Computational Electromagnetic Methods Applied to the Design of Patch Antenna with EBG Structure
In this contribution High-Performance Computing electromagnetic methods are applied to the design of a patch antenna combined with EBG structure in order to obtain bandwidth enhancement. The electrical characteristics of the embedded structure (patch antenna surrounded by EBG unit cells) are evaluated by means of method of moment technique (MoM) whereas for designing the unit cell, the finite element method (FEM) together with the Bloch-Floquet theory is used. The manufactured prototypes are characterized in terms of return loss and radiation pattern in an anechoic chamber.
Introduction
Microstrip patch antennas, which are used for both defense and commercial applications, are replacing many conventional notprinted antennas.The most important properties of microstrip patch antennas are the lightweight, smallvolume and the mass production at low cost whereas intrinsic disadvantages that limit their applications are low gain, narrow bandwidth, and excitation of surface waves [1].In order to design antennas with better efficiency and gain and lower backlobe and sidelobe levels, EBG structures can be used [2][3][4][5][6][7].In previous works [8][9][10][11][12][13][14][15][16] several narrow band antennas have been mounted on EBG structures.In this contribution, for obtaining bandwidth enhancement in the 2.48 GHz band, the band-gap of the EBG lattice is designed to be adjacent to the frequency band of the patch antenna, so that when integrating together both structures their resonances couple each other and as a result a wider bandwidth will be generated.The final design is uniplanar and in addition there is no need for via holes.
The aim of this work is challenging because two resonant structures are involved and when integrated together their resonant behavior is mutually influenced.For this reason the precision in the simulation results is a key point to achieve bandwidth enhancement.At this point is where highperformance computational electromagnetics play a funda-mental roll in order to get a proper design in a reasonable time.
Microstrip Antenna
The antenna design has been carried out using Method of Moments (Momentum) electromagnetic simulator [17] and its geometry (Figure 1(a)) was optimized varying the values in a continuous range in order to obtain the frequency band of interest (2.48 GHz) and the minimum bandwidth requested (the process is an iterative one in which there will be a tradeoff between the variation of antenna's geometry and its effects on the performance).Using parameter sweep with MoM in a 2 core Intel Xeon X5560 48 GB RAM (equivalent to 16 execution threads) server, there can be obtained the solution for 81 different values of the swept parameter in just one minute.This is due to a mesh definition so that 20 cells per wavelength at 3 GHz are taken which leads to 249 rectangular cells and 171 triangular cells with a matrix size of 740.After applying mesh reduction a matrix size of 610 results.A dielectric substrate, ROGER3010 having 1.27 mm thickness, ε r = 10.2 relative dielectric permittivity, and loss tangent of 0.0023, has been used.The measured operating bandwidth of the patch antenna is 23 MHz (Figure 4).The difference in bandwidth between the simulated (20 MHz) and measured (23 MHz) results could be due to the fact that the commercial MoM software considers infinite extension for the dielectric substrate, or even more likely due to manufacturing tolerances.
EBG Characterization
The unit cell of an EBG lattice consists of metal pads and sometimes narrow lines that implement a distributed LC circuit having a resonant frequency [18].In [19] the same unit cell but in different frequency band is characterized from the point of view of an AMC, computing the phase of the reflection coefficient on its surface, from an uniform incident plane wave.To search for the frequency band in which the periodic structure shows the AMC behavior, the finite element method (FEM) together with the Bloch-Floquet theory is used.A single cell of the lattice (with periodic boundary conditions (PBCs) on its four sides) is simulated to model an infinite structure.The unit cell dimensions are W × W = 16.38 × 16.38 mm 2 .The simulation is carried out using a server with 2 processors Intel Xeon E5620 and 64 GB RAM in a configuration equivalent to 2 execution threats and 32 GB RAM.The air-box size is λ/2 at the lower frequency considered in the simulation (which is 1 GHz).The solution frequency is 3 GHz and a frequency sweep is carried out from 1 GHz to 3 GHz with a 0.01 GHz frequency step.The mesh is established to take at least 10 tetrahedra per wavelength at the solution frequency (3 GHz).Mixed-order basis functions and 30% lambda refinement are used.Maximum Delta S is fixed to 0.02 for the S-parameter calculations.All this leads to a mesh with 3642 tetrahedra (2889 for the air-box and 753 for the substrate), a matrix size of 16101, and a computational time of 8 minutes and 44 seconds for the aforementioned frequency sweep.However, for the intended application, the structure should show EBG behavior.Even though the finite element method (FEM) uses specific boundary conditions such as simulating just half the volume under study (using Perfect Magnetic Conductor (PMC) boundary condition in one of the volume walls, the one that would cut the volume in two identical parts, the size of the electromagnetic problem that needs to be simulated is reduced), for the current scenario the method of moments (MoM) could yield a faster computational time (using MoM only the conductive structure is meshed whereas using FEM, the whole volume around the structure including the air box is meshed).The periodic structure can be characterized as EBG using the suspended strip method [20,21] and so the transmission coefficient (S 21 ) of a suspended strip line over the periodic structure is simulated (Figure 2).The structure will block the transmission of power along the strip line for frequencies within the band-gap region and a noticeable reduction of S 21 can be observed at a certain frequency band.To accomplish this simulation the mesh has been defined so that 20 cells per wavelength at 3 GHz are taken leading to 2085 rectangular cells and 5008 triangular cells with a matrix size of 9706.After applying mesh reduction results a matrix size of 2352.The simulation time for 25 frequency steps is 50 minutes in a 2core Intel Xeon X5560 48 GB RAM (equivalent to 16 threads execution) server.
The band-gap of the EBG structure (45 MHz) is adjacent to the bandwidth of the patch antenna (see Figure 3); thus when combining the two structures together, bandwidth enhancement is obtained without disturbing other characteristics of the patch antenna, such as the radiation pattern.
Patch Antenna with EBG Structure
Once the antenna and the EBG structures have been designed, the next step is the integration of both resonant structures together in the same layer forming a uniplanar design [22].As it has been already mentioned in the introduction, the resonance frequency of both structures is mutually influenced, and so depending on the frequency difference between them, and how the unit-cell arrangement around the patch antenna is, the resulting resonance frequency will change.The design of the patch antenna surrounded by one row EBG lattice has been carried out (Figure 1(b)).In FEM, due to the air-box size when small frequencies are involved (as in this case), the electric size of the problem to be solved is rather big.A proper mesh should take at least 10 (generally 20) tetrahedra per wavelength.Depending on the prototype's physical size, this could make the matrix of the linear equation system to become dense, which is not desirable in FEM and generally leads to longer computational times and increased memory requirements.However the matrix in MoM is dense, so this is not a problem, and thus this could be a better choice in general.The disadvantage of MoM is related to dielectric managements as they are considered infinite sized.Using MoM the mesh has been defined so that 20 cells per wavelength at 3 GHz are taken which leads to 2460 rectangular cells and 6246 triangular cells with a matrix size of 12365.After applying mesh reduction a matrix size of 5832 results.The simulation time for 81 frequency steps is 63 minutes in a 2-core Intel Xeon X5560 48 GB RAM (equivalent to 16 threads execution) server.
A prototype of the Patch antenna-EBG has been manufactured using laser micromachining.The return losses of the manufactured prototype have been measured and compared to those of the microstrip patch antenna (Figure 4).The principle of achieving bandwidth enhancement is based on coupling the resonances of the structures involved.As the patch antenna resonates at adjacent frequency band compared to the band-gap of the EBG lattice, a significant bandwidth enhancement of the prototype combining the two structures (Patch antenna-EBG) is obtained.As shown in Figure 4 Radiation pattern cuts in the E and H planes of each manufactured prototype are plotted in Figure 5.Using an EBG structure to surround the patch antenna, the directivity increases due to the surface wave suppression (Table 1).
In the case of placing the patch antenna in a frequency range outside the band-gap of the EBG structure, MoM simulations show that the unit cells have no influence in the radiation properties or in the bandwidth.
Conclusions
Bandwidth enhancement of microstrip patch antenna by means of EBG structure for RFID SHF 2.48 GHz band has been presented.Using High-Performance computing electromagnetic methods the electrical characteristics of the resonant unit cell and the patch antenna have been evaluated both separated as well as combined in the same layer.The simulated and measured results are in good agreement due to precision of the methods (MoM and FEM) used in simulations.Both frequency domain techniques, MoM and FEM, can be used once the AMC is designed using FEM with PBCs.This is just an example of MoM and FEM techniques application to the design of antennas with metamaterials, but there are also other possible approaches, time domain based such as (Finite-difference time-domain) FDTD which could also be used.
There was reported a 50% increase of the initial bandwidth.The patch antenna-EBG prototype presented in this paper has several advantages such as planar feature, compact size, and low dielectric losses.Neither via holes nor multilayer substrates are required, simplifying practical implementation and reducing its cost.
Figure 2 :
Figure 2: Schematic of suspended line above EBG surface (top view).
Figure 3 :
Figure 3: Resonances to be coupled in order to achieve bandwidth enhancement.
Figure 4 :
Figure 4: Simulation and measurement comparison between the patch antenna and the patch antenna-EBG prototypes.
Table 1 :
Comparison between the two prototypes. | 2018-12-12T15:12:32.416Z | 2012-01-01T00:00:00.000 | {
"year": 2012,
"sha1": "dc67ba0799e2ae1f22bc3f89d6be4ea66b6f0caf",
"oa_license": "CCBY",
"oa_url": "https://downloads.hindawi.com/journals/ijap/2012/435890.pdf",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "dc67ba0799e2ae1f22bc3f89d6be4ea66b6f0caf",
"s2fieldsofstudy": [
"Business"
],
"extfieldsofstudy": [
"Engineering"
]
} |
251729137 | pes2o/s2orc | v3-fos-license | Therapeutic Window of Intravenous Immunoglobulin (IVIG) and its correlation with IVIG-resistant in Kawasaki Disease: a retrospective study
Objective To investigate the optimal timing of initial intravenous immunoglobulin (IVIG) treatment in Kawasaki disease (KD) patients. Methods KD patients were classified as the early group (day 1-4), conventional group (day 5-7), conventional group (day 8-10), and late group (after day 10). Differences among the groups were analyzed by ANOVA and Chi-square analysis. Predictors of IVIG resistance and the optimal cut-off value were determined by multiple logistic regression analyses and receiver operating characteristic (ROC) curve analysis. Results There were no significant differences in IVIG resistance among the 4 groups (p = 0.335). The sensitivity analysis also confirmed no difference in the IVIG resistance between those who started the initial IVIG ≤ day 7 of illness and those who received IVIG >day 7 of illness (p = 0.761). In addition, patients who received IVIG administration more than 7 days from the onset had a higher proportion of coronary artery abnormalities (p = 0.034) and longer length of hospitalization (p = 0.033) than those who started IVIG administration less than 7 days. The optimal cut-off value of initial IVIG administration time for predicting IVIG resistance was >7 days, with a sensitivity of 75.25% and specificity of 82.41%. Conclusions IVIG therapy within 7 days of illness is found to be more effective for reducing the risk of coronary artery abnormalities than those who received IVIG >day 7 of illness. IVIG treatment within the 7 days of illness seems to be the optimal therapeutic window of IVIG. However, further prospective studies with long-term follow-up are required.
Introduction
Kawasaki disease (KD) is an acute, self-limited disease leading to systemic vasculitis that predominantly affects children younger than five. 1 KD surpasses acute rheumatic fever as a leading cause of acquired heart disease in developed countries. 2 The most common complication of KD is coronary artery abnormalities (CAA), and its prevention is important to improve outcomes in patients with KD. 3 Intravenous immunoglobulin (IVIG) is the mainstay of treatment in KD and has been known to be effective in abolishing vascular inflammations leading to coronary artery lesions (CALs). 4 The fatality rates of KD have markedly decreased since the IVIG therapy was introduced in 1983. 5 However, the association between the earlier timing of IVIG administration of disease onset and risk for IVIG unresponsiveness remains debatable. 6 Moreover, the literature comparing the early and routine IVIG therapy regarding the efficacy in preventing cardiac sequelae is limited and controversial. [7][8][9][10][11][12] Furthermore, several latest guidelines on the optimal timing of IVIG administration and if IVIG can be given earlier remain inconclusive. 1,13,14 Given this background, the authors hypothesize that there is maybe a narrow therapeutic window of IVIG administration. Earlier or later IVIG treatment may affect the IVIG responsiveness and outcomes in KD patients. Thus, the authors performed a retrospective study to investigate the optimal timing of initial IVIG administration by comparing the outcomes of early and conventional IVIG treatment in KD patients.
Methods
The Chengdu Women's and Children's Central Hospital Ethics Committee approved the study protocol (Approval No.B202123) and waived informed consent requirements. All methods were carried out following the Declaration of Helsinki.
Study design and participants
All the patients with KD who were hospitalized at the Chengdu Women's and Children's Central Hospital between January 2018 and December 2020 were identified as participants in this retrospective study. The Chengdu Women's and Children's Central Hospital is a tertiary referral center that serves a catchment area of approximately 2 million people per year who live in the southwest region of China. Patients were diagnosed with KD according to the American Heart Association (AHA) guideline in 2017. 1 The inclusion criteria for study subjects were as follows: (1) patients were <18 years old; (2) patients of initial onset of KD; (3) patients received standard treatment with 2 g/kg of IVIG of single infusion during the acute phase of illness. Exclusion criteria were as follows: (1) recurrent KD; (2) patients had a severe infection, allergy, autoimmune diseases, or collagen disease; (3) patients who had received IVIG in the first three months of admission; (4) patients who did not receive IVIG or initial IVIG dose was <2 g/kg; (5) patients with missing clinical or laboratory information. The AHA guideline for KD recommends starting IVIG treatment within 10 days from the onset of symptoms and, if possible, within 7 days to prevent CAA because this is when vasculitis worsens. Several retrospective studies, however, reported that IVIG use at 4 days of illness had no benefit in preventing CAA, but was instead associated with increased IVIG resistance. 7,8 Therefore, the present study hypothesizes that the clinical outcomes differed among four categories (start of IVIG therapy up to 4 days, 5-7 days, 8-10 days, and more than 10 days). Thus, the remaining patients who meet the inclusion criteria were classified into four groups: those who received the initial IVIG day 4 of illness (the early treatment group), those who received IVIG on days 5-7 (the conventional treatment group), those who received IVIG on days 8-10 (the conventional treatment group), and those who were treated after day 10 (the late treatment group). The first day of fever was defined as day 1.
Data collection
The demographic, clinical outcomes, and laboratory data were extracted from the medical records. Blood samples were collected within 24 h pre-IVIG treatment. IVIG resistance was defined as recrudescent or persistent fever 36 h but not longer than 7 days after initial IVIG infusion. 1
Treatment protocol for KD patients
In the present study's hospital, all KD patients received the standard therapy with IVIG (2 g/kg) and aspirin (30-50 mg/ kg/d during the acute phase of illness) immediately after the diagnosis. The aspirin was lowered to 3-5 mg/kg/d 2-3 days after the patients were afebrile. Other therapies, including prednisolone, were not used in the initial treatment. Combined antiplatelet and anticoagulation therapy were recommended for patients with giant aneurysms. For IVIG-resistant patients, the 2nd IVIG of the same dosage was administrated. If fever persists 36 h after the 2nd IVIG infusion, intravenous methylprednisolone (30 mg/kg/dose) was performed for 3 consecutive days. No patients received additional treatment such as infliximab, plasma exchange, and cytotoxic agents.
Complete KD was diagnosed in any children with a fever and had four or more of the following five major symptoms: (Ⅰ) erythema and cracking of lips, strawberry tongue, and/or erythema of oral and pharyngeal mucosa, (Ⅱ) polymorphous exanthema, (Ⅲ) bilateral conjunctival congestion, (Ⅳ) changes of the peripheral extremities, and (Ⅴ) unilateral cervical lymphadenopathy. Incomplete KD was defined as a child with a fever with fewer than four major symptoms and compatible laboratory or echocardiographic findings. 1
Echocardiography
Echocardiography was used to detect CAA during hospitalization. Echocardiography was performed and supervised by an experienced echocardiographer and with appropriate transducers. The coronary artery abnormalities were defined as follows: (1) normal: Z score < 2; (2) only dilation: Z score 2 to < 2.5; or initial Z score < 2, Z score decline during follow-up (usually 6-12 months) 1; (3) small coronary aneurysm: Z score 2.5 to < 5; (4) medium coronary aneurysm: Z score 5 to <10, and absolute dimension < 8 mm; (5) large or giant coronary aneurysm: Z score 10, or absolute dimension 8 mm. 1 Patients with a Z score 2.0 were considered to be CAA. Coronary artery aneurysms were classified into small aneurysms, medium aneurysms, and large (or giant) aneurysms. 1 The Z-score was measured before IVIG administration using the formula by Dallaire and Dahdah. 15 The authors chose the maximum Z-scores of the left main coronary artery, left anterior descending coronary artery, left circumflex artery, and right coronary artery.
The primary outcomes were IVIG resistance and the development of CAA in the acute phase. The secondary outcome was the length of hospitalization.
Statistical analysis
The normality of the distribution of variables was checked using the Kolmogorov-Smirnov test. Continuous variables were expressed as mean § standard deviations or median and IQR (25th, 75th percentile) if non-normally distributed. Categorical variables were expressed by presenting the frequency and proportion in each category. To compare the difference of variables according to the timing of initial IVIG administration of disease onset ( day 4, days 5-7, days 8-10, and >day 10), analysis of variance (ANOVA) or Kruskal-Wallis H test was applied for continuous variables, and Chisquare analysis was applied for categorical variables. For significance found by ANOVA, Kruskal-Wallis H test, or Chisquare test, the further pairwise comparison was conducted through the Dunnett-t-test or the least-significant difference (LSD) test based on the homogeneity of data variance, or Bonferroni method for cross-table.
The multivariate regression analysis was conducted to explore the factors associated with the prevalence of IVIG resistance and coronary artery abnormalities, which were dependent variables. The independent variables for multivariate regression analysis contained both categorical and continuous variables. The continuous variables were first screened by univariate analysis, namely t-test or Mann-Whitney U test, based on the distribution normality of variables. The variables associated with dependent variables were included in the regression model. The area under the receiver operating characteristic (ROC) curve was analyzed to assess the predictive accuracy of initial IVIG administration time for IVIG-resistant and to identify the optimal cutoff point according to the Youden index. A subgroup analysis was performed by dividing CAA into dilation, small aneurysm, and medium to large (or giant) aneurysm according to the AHA guideline in 2017. 1 The authors also performed a sensitivity analysis to compare the primary and secondary outcomes between the patients who received the initial IVIG day 7 of illness and >day 7 of illness. A P value less than 0.05 was considered statistically significant. All statistical analyses were performed with the IBM SPSS Statistics 25.0 (SPSS, Chicago, IL, USA).
Characteristics of included KD Patient
A total of 965 KD patients who meet the inclusion criteria were included (Fig. 1), with a median age of 28.10 § 19.79 months, ranging from 1 to 132 months. The ratio of males to females was 1.53:1. All patients received initial IVIG treatment once the KD diagnosis was made. There were 883 patients diagnosed as complete KD, accounting for 91.50% of the total KD patients. Forty patients were classified into the early group ( 4 days), 726 into the conventional (5-7 days) group, 160 into the conventional (8-10 days) group, and 39 into the late group (> 10 days). There were no significant differences in age. A slight difference (p = 0.039) was observed in gender among the four groups. The proportions of complete KD in early and conventional (5-7 days) groups were significantly higher than patients in conventional (8-10 days) and late groups (p = 0.000). The detailed information on the characteristics of included KD patients can be seen in Table 1.
Laboratory findings
The neutrophils%, C-reactive protein (CRP), White Blood Cell (WBC) count, and Hemoglobin (HB) in early and conventional (5-7 days) groups were significantly higher than those in conventional (8-10 days) and late groups (p 0.05), suggesting inflammatory reaction is severe in early and conventional (5-7 days) groups. Platelet in early and conventional (5-7 days) groups were significantly lower than in conventional (8-10 days) and late groups (p 0.05). AST and ALT in the conventional (5-7 days) group were higher than in the conventional (8-10 days) group. The four groups showed no significant difference in albumin and erythrocyte sedimentation rate (ESR) (Appendix S1).
IVIG resistance, development of Coronary artery abnormalities, and the length of hospitalization A total of 101 patients had IVIG resistance, accounting for 10.47% of the included KD patients. The early group had a higher IVIG-resistance rate (17.5%) than the other three groups. The conventional (5-7 days) group had the lowest IVIG-resistance rate (9.92%) among the 4 groups. However, there were no significant differences in IVIG resistance rate among the 4 groups (p = 0.335) ( Table 2). The length of hospitalization of the late group was significantly longer than the two conventional groups (p = 0.000). There were 213 patients involved in CAA, accounting for 22.07%. The late group had the highest rate of CAA (38.46%). The rate of CAA in the late group was significantly higher than in the conventional (5-7 days) group (p < 0.05) ( Table 2).
In the subgroup analysis, there was no significance in the proportion of dilation between each group. The conventional (5-7 days) group had the lowest proportion of small aneurysms (11.71%) and medium to large aneurysms (1.93%). Moreover, the early group has a proportion of small aneurysms significantly higher than the conventional (5-7 days) group (p < 0.01), while the late group has a proportion of medium to large aneurysms significantly higher than conventional (5-7 days) group (p < 0.01) ( Table 3). The sensitivity analysis also confirmed no difference in the IVIG resistance between those who started the initial IVIG day 7 of illness and those who received IVIG >day 7 of illness. There was a significant difference in the higher proportion of CAA (OR = 0.680; 95%CI = 0.476-0.972, p = 0.034) and longer length of hospitalization (p = 0.033) for those who started IVIG administration more than 7 days from the onset (Appendix S2).
ROC curve regarding IVIG resistance and multivariate regression analysis
According to the ROC analysis, the optimal cut-off value of initial IVIG administration time for predicting IVIG resistance was >7 days, with a sensitivity of 75.25% and specificity of 82.41% (area under the curve was 0.785, 95% confidence interval was 0.758-0.811; p < 0.001) (Appendix S3). Results are presented in n (%), median (25th, 75th percentile).
Figure 1 Flowchart of included KD patients.
According to the multivariable logistic regression analysis, age and albumin were independent predictors of IVIG resistance in patients with KD (Appendix S4). Incomplete KD, Gender, CRP, and albumin were independent risk factors of CALs in KD patients (Appendix S5).
Discussion
This study aimed to investigate the optimal time option for initial IVIG treatment in KD patients. There were no significant differences in IVIG resistance among the 4 groups in the present study, although the early group had the highest IVIG resistance rate. The sensitivity analysis also confirmed no difference in the IVIG resistance between those who started the initial IVIG day 7 of illness and those who received IVIG >day 7 of illness. However, patients who received IVIG administration more than 7 days from the onset had a higher proportion of CAA and longer length of hospitalization than those who started IVIG administration less than 7 days. Thus, IVIG therapy within 7 days of illness is more effective for reducing the risk of coronary artery abnormalities than those who received IVIG >day 7 of illness.
The late and conventional (8-10 days) groups had higher proportions of incomplete KD than the early and conventional (5-7 days) groups in the present study. A higher proportion of incomplete KD may experience significant delays in diagnosis, 16 which leads to the late IVIG therapy. Furthermore, it is not surprising that patients treated earlier had a higher level of CRP, neutrophils%, WBC, and lower platelet counts. It was suggested that the patients in the early and conventional (5-7 days) groups might likely have more typical symptoms and more severe cases than conventional (8-10 days) and late groups, leading to an early initial administration of IVIG. 17 IVIG has been the mainstay of treatment in KD to reduce the prevalence of CAA. However, about 10À20% of patients show resistance to IVIG treatment and present a higher risk of coronary vasculitis. 18 Initial IVIG resistance is also associated with an increased risk of coronary artery aneurysms. 19,20 The criteria for when to provide IVIG are unclear and differ from previous studies. 7,12,[21][22][23] The 2004 AHA and 2018 Italian Society of Pediatrics guidelines stated that IVIG should be used within 10 days after the onset of disease and, if possible, within 7 days. In the latest 2017 AHA guideline, it is suggested that IVIG should be instituted as early as possible within the 10 days of illness onset of fever; However, there is no suggestion on the optimal timing of IVIG and if it can be given earlier. Several studies reported that IVIG use at 4 days of illness is associated with increased IVIG resistance. 7,21,22,24 These results must be interpreted with caution. Muta et al. 7 reported early treatment is likely to result in a greater requirement for additional IVIG. Still, they did not perform multivariate analysis because of limited data associated with the need for IVIG retreatment. The present results are similar to previous studies that showed earlier IVIG treatment within 4 days may not increase the higher incidence of IVIG resistance. 12,23 The early group showed a high rate of IVIG resistance because there might be more patients with severe inflammation or atypical clinical course, respectively.
The IVIG resistance may be associated with the dynamic activation status of macrophages and the level of serum cytokines during the acute phase of KD. 25 Inappropriate time options for IVIG treatment may affect the balance of pro-inflammatory and counter-inflammatory cytokines and subsequently the patient's clinical outcome, including additional treatment and development of CAA. Future studies should focus on the pharmacokinetics of IVIG and the influence of immunological changes during the acute phase of KD on treatment response. The ROC analysis showed the optimal cut-off value of initial IVIG administration time for predicting IVIG resistance was >7 days. Multivariable logistic regression analysis revealed that age and albumin were independent predictors of IVIG resistance, which is compatible with the previous studies. 10,26 The most common sequela of KD is CAA, which is speculated to be caused by acute systemic inflammation. 27 In the present study, the late group had the highest rate of CAA. It Results are presented in n (%), median (25th, 75th percentile). was significantly higher than the conventional (5-7 days) group, indicating the IVIG should be administrated within 10 days after the disease onset to prevent the CAA. The results of the subgroup analysis were similar -the late group has a rate of medium to large aneurysms significantly higher than the conventional (5-7 days) group. Although the early group has a higher proportion of small aneurysms than the conventional (5-7 days) group, one possible reason is that this was due to a bias caused by including severe cases with full clinical symptoms as early as day 4 after onset. The sensitivity analysis also showed that patients who began IVIG treatment more than 7 days of illness onset might develop CAA more frequently than those treated on days 7. To minimize cardiac sequelae, it is crucial to avoid any delay in IVIG treatment because earlier inflammatory suppression may contribute to avoiding developing CAAs. 28 Multivariate regression revealed that the incomplete KD, male gender, higher level of CRP, and lower level of albumin were independent risk factors of CALs, in line with other reported studies. 26,29,30 Limitations This study has several limitations. First, it was a retrospective observational study with potential selection and information bias. Second, all the KD patients came from Sichuan, southwest China, limiting the findings' generalization. Third, the study results could have been limited by incomplete data caused by patients being lost in follow-up after discharge from the hospital. A rigorous prospective study with long-term follow-up patients from other districts is needed.
Conclusion
IVIG therapy within 7 days of illness is found to be more effective for reducing the risk of coronary artery abnormalities than those who received IVIG >day 7 of illness. All patients meeting diagnostic criteria for KD should be treated as soon as the course of illness as the diagnosis can be established. IVIG treatment within the 7 days of illness seems to be the optimal therapeutic window of IVIG. However, further prospective studies with long-term follow-up are required.
Funding
This manuscript did not receive any funding.
Ethical approval
The Chengdu Women's and Children's Central Hospital Ethics Committee approved the study protocol (Approval No. B202123) and waived informed consent requirements. | 2022-08-23T15:11:54.954Z | 2022-08-01T00:00:00.000 | {
"year": 2022,
"sha1": "20a782e045e1bbb261cf3ee7feb8016c4f1b0366",
"oa_license": "CCBYNCND",
"oa_url": "https://doi.org/10.1016/j.jped.2022.07.003",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "1116a88d857963e02eeb70c9ecdbd5d40e17b669",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
249062582 | pes2o/s2orc | v3-fos-license | An Evolutionary Approach to Dynamic Introduction of Tasks in Large-scale Multitask Learning Systems
Multitask learning assumes that models capable of learning from multiple tasks can achieve better quality and efficiency via knowledge transfer, a key feature of human learning. Though, state of the art ML models rely on high customization for each task and leverage size and data scale rather than scaling the number of tasks. Also, continual learning, that adds the temporal aspect to multitask, is often focused to the study of common pitfalls such as catastrophic forgetting instead of being studied at a large scale as a critical component to build the next generation artificial intelligence.We propose an evolutionary method capable of generating large scale multitask models that support the dynamic addition of new tasks. The generated multitask models are sparsely activated and integrates a task-based routing that guarantees bounded compute cost and fewer added parameters per task as the model expands.The proposed method relies on a knowledge compartmentalization technique to achieve immunity against catastrophic forgetting and other common pitfalls such as gradient interference and negative transfer. We demonstrate empirically that the proposed method can jointly solve and achieve competitive results on 69public image classification tasks, for example improving the state of the art on a competitive benchmark such as cifar10 by achieving a 15% relative error reduction compared to the best model trained on public data.
INTRODUCTION
The success of machine learning continues to grow as it finds new applications in areas as diverse as language generation (Brown et al., 2020), visual art generation (Ramesh et al., 2021), chip design (Mirhoseini et al., 2020), protein folding (Senior et al., 2020) and competitive sports (Silver et al., 2016;Vinyals et al., 2019). The vast majority of machine learning models are designed and trained for a single task and specific data modality, and are often trained by starting with randomly initialized parameters, or with limited knowledge transfer from a pre-trained model. While this paradigm has shown great success, it uses a large amount of computational resources, and does not leverage knowledge transfer from many related tasks in order to achieve higher performance and efficiency.
The work presented in this paper is based on the intuition that significant advances can be enabled by dynamic, continual learning approaches capable of achieving knowledge transfer across a very large number of tasks. The method described in this paper can dynamically incorporate new tasks into a large running system, can leverage pieces of a sparse multitask ML model to achieve improved quality for new tasks, and can automatically share pieces of the model among related tasks. This method can enhance quality on each task, and also improve efficiency in terms of convergence time, amount of training examples, energy consumption and human engineering effort.
The ML problem framing proposed by this paper can be interpreted as a generalization and synthesis of the standard multitask and continual learning formalization, since an arbitrarily large set of tasks can be solved jointly. But also, over time, the set of tasks can be extended with a continuous stream of new tasks. Furthermore, it lifts the distinction between a pretraining task and a downstream task. As new tasks are incorporated, the system searches for how to combine the knowledge and representations already present in the system with new model capacity in order to achieve high quality for each new task. Knowledge acquired and representations learned while solving a new task are available for use by any future task or continued learning for existing tasks.
We refer to the proposed method as "mutant multitask network" or µ2Net. This method generates a large scale multitask network that jointly solves multiple tasks to achieve increased quality and efficiency for each. It can continuously expand the model by allowing the dynamic addition of new tasks. The more accumulated knowledge that is embedded into the system via learning on previous tasks, the higher quality the solutions are for subsequent tasks. Furthermore, new tasks can be solved with increasing efficiency in terms of reducing the newly-added parameters per task. The generated multitask model is sparsely activated as it integrates a task-based routing mechanism that guarantees bounded compute cost per task as the model expands. The knowledge learned from each task is compartmentalized in components that can be reused by multiple tasks. As demonstrated through experiments, this compartmentalization technique avoids the common problems of multitask and continual learning models, such as catastrophic forgetting, gradient interference and negative transfer. The exploration of the space of task routes and identification of the subset of prior knowledge most relevant for each task is guided by an evolutionary algorithm designed to dynamically adjust the exploration/exploitation balance without need of manual tuning of meta-parameters. The same evolutionary logic is employed to dynamically tune the hyperparameters multitask model components.
RELATED WORK
The main novelty of the presented work is to propose and demonstrate a method that jointly provides all of the following properties: 1) ability to continually learn from an unbounded stream of tasks, 2) automate the selection and reuse of prior knowledge and representations learned for previous tasks in the solving of new tasks, 3) search the space of possible model architectures allowing the system to dynamically extend its capacity and structure without requiring random initialization, 4) automatically tune the hyperparameters of both the generated models and the evolutionary method, including the ability to learn schedules for each hyperparameter, rather than just constant values, 5) ability to optimize for any reward function, also including non-differentiable factors, 6) immunity from catastrophic forgetting, negative transfer and gradient interference, 7) ability to extend any pre-existing pre-trained model, including extending its architecture and adapting the domain on which such model have been trained to other domains automatically, 8) introduction of a flexible access control list mechanism that allows expression of a variety of privacy policies, including allowing the use or influence of task-specific data to be constrained to just a single task or to a subset of tasks for which data or higher-level representation use should be permitted. Different lines of research have focused on distinct subsets of the many topics addressed by the proposed method. In this section we highlight a few cornerstone publications. Refer to Appendix A for an extended survey. Different methods have been proposed to achieve dynamic architecture extensions (Chen et al., 2016;Cai et al., 2018), some also focusing on an unbounded stream of tasks (Yoon et al., 2018), or achieving immunity from catastrophic forgetting (Rusu et al., 2016;Li & Hoiem, 2018;Rosenfeld & Tsotsos, 2020). Unlike our work, these techniques rely on static heuristics and patterns to define the the structural extensions, rather than a more open-ended learned search process. Neural architecture search (NAS) (Zoph & Le, 2017) methods aim to modularize the architectural components in search spaces whose exploration can be automated with reinforcement learning or evolutionary approaches (Real et al., 2019;Maziarz et al., 2018). More efficient (but structurally constrained) parameter sharing NAS techniques (Pham et al., 2018;Liu et al., 2019a) create a connection with routing methods (Fernando et al., 2017) and sparse activation techniques, that enable the decoupling of model size growth from compute cost growth Du et al., 2021). Evolutionary methods have also been applied with success for hyperparameter tuning (Jaderberg et al., 2017).
Cross-task knowledge transfer has gained popularity, especially through transfer learning from a model pre-trained on a large amount of data for one or a few general tasks, and then fine-tuned on a small amount of data for a related downstream task. This approach has been shown to be very effective in a wide variety of problems and modalities (Devlin et al., 2019;Dosovitskiy et al., 2021). Large scale models have recently achieved novel transfer capabilities such as few/zero shot learning (Brown et al., 2020). More complex forms of knowledge transfer such as multitask training or continual learning often lead to interesting problems such as catastrophic forgetting (McCloskey & Cohen, 1989;French, 1999), negative transfer (Rosenstein, 2005; Figure 1: Graphical representation of the two mutation types used by the proposed method: layer cloning mutation (left) and hyperparameter change (center). The graph on the right represents the model generated by the preliminary experiment described in Section 4. The bottom nodes display the task names, the top nodes display the validation accuracy, and internal nodes are represented with the color of the task that has last updated the parameters of the corresponding layer. (Chen et al., 2018;Yu et al., 2020). Research on these topics mostly focuses on approaches such as weighted combination methods (Liu et al., 2019b;Sun et al., 2020b) or gradient transformations (Sener & Koltun, 2018;Kendall et al., 2018), also methods automating knowledge selection at a layer level was proposed (Sun et al., 2020a). The method proposed in this paper can be considered large-scale and state-of-the-art focused progression from Gesmundo & Dean (2022).
EVOLUTIONARY METHOD
This section defines the proposed method capable of generating a dynamic multitask ML system. The multitask system is initialized with one root model. This model can be either pretrained or randomly initialized. During the evolutionary process, the proposed method searches for the best model for a single task at a time, referred to as the active task. During the active phase of a task, a population of models trained on the active task is evolved: the active population. The first time a task becomes active, its active population is empty. For subsequent iterations, the active population is initialized with all the models trained on the active task that have been retained from previous iterations. The active population is iteratively extended by: 1) sampling a parent model (Section 3.1), 2) applying to the parent model a sampled set of mutations (Section 3.2) to produce a child model, 3) performing cycles of training and validation in order to train and score the child model. Each trained model is assigned a score (Section 3.3). Early population pruning is performed by discarding the models that did not achieve a better score then their parent. An active phase is composed of multiple generations in which multiple batches of child models are sampled and trained in parallel. At the end of a task active phase, only its best scoring model is retained as part of the multitask system. A sequence of task iterations may be employed to generate solutions for a set or a stream of different tasks. A task can become active multiple times. Details of the method are reported below (and in Algorithm 1).
PARENT MODEL SAMPLING
The first attempt to sample a parent model for the active task is done over the active population of models for that task. The models in the active population are visited in decreasing order of score, starting with the highest scoring one. Each model, m, can be accepted as parent with probability: p parent (m|t) = 0.5 #selections(m,t) . Where #selections(m, t) denotes the number of times the candidate model, m, has been previously selected as parent to generate a child models for task t. If the current candidate parent is not selected, then iteratively the model with the next best score is considered to be selected as parent with probability p parent (·|t). If a full iteration on the active population is completed without a successful parent model selection, then the same method is applied to the randomly sorted list of all remaining models: all the models currently part of the multitask system that were trained on a task different from the current active task, t. This fallback list is randomly sorted since these models have not been scored for t. As a final fallback a parent is uniformly sampled among all the models currently in the system. This method prioritizes the exploitation of high scoring models that had few attempts at generating an improved model for the active task. But also, in combination with early pruning, it automatically transitions toward a more exploratory behavior in case the higher scoring models are unable to generate an improvement.
MUTATIONS
In this work we consider Deep Neural Networks (DNN) models. DNN are commonly defined by their architecture and hyperparameters. Architectures are composed of a sequence of neural network layers, each mapping an input vector into an output vector of variable dimensions. Hyperparameters specify the configuration details such as the optimizer or data preprocessing configurations. The presented method allows for two types of mutations ( Figure 1): Hyperparameter mutations can be applied to modify the configuration inherited from the parent. Each hyperparameter is associated with a sorted list of valid values (Table 1). If a hyperparameter is selected for mutation, then its new value is selected at random among the two neighbouring values in the sorted list ( Figure 1). This constraints hyperparameter mutations to only incremental changes. Notice that, every hyperparameter of a child model is set to a single value. However, considering that a child model continues its ancestors training with mutated hyperparameters, then the method can be regarded as capable of defining a piece-wise constant schedule over time for each hyperparameter.
Layer cloning mutations create a copy of any parent model layer that can be trained by the child model. If a layer of the parent model is not selected for cloning, then it is shared with the child model in a frozen state to guarantee immutability of pre-existing models. Child models can train only the cloned copies of the parent layers. The cloned layers are trained with a possibly modified version of the parent optimizer. The configuration of the child optimizer is defined by the mutated hyperparameters. If such optimizer is of a type that stores a state (i.e. momentum), then the state is also cloned from the state saved by the ancestor that has last trained the cloned layer. Notice that, a trainable layer may be followed by frozen layers. In this case the gradients for the trainable layer are propagated through the frozen layers and applied only to the parameters of the trainable layers while frozen parameters are left unchanged. The head layer is always cloned since it always needs to be trainable. If a child model is trained on a task different from the parent's task, then a new randomly initialized head layer is created with output shape matching the number of classes of the new task.
Each possible layer cloning or hyperparameter mutation is independently sampled to be applied with probability µ. µ is itself a hyperparameter that is mutated by the evolutionary process. Thus demonstrating that automatic tuning is not only applied to selecting the hyperparameters of the generated models, but can also be applied to self-tune the configuration of the evolutionary algorithm.
TRAINING AND SCORING
A newly sampled child model is trained on the active task for a given number of epochs. The model is evaluated on the validation set after each epoch. At each intermediate evaluation, the child model is assigned a score that the evolutionary algorithm aims to maximize. The score can be defined to optimize a mixture of factors such as validation quality, inference latency, training compute or model size, depending on the applications requirements. The presented experiments aim to compare against the state of the art for a large number of tasks without any size or compute constraint. Therefore, the validation accuracy is used directly as the score without additional factors. After training, only the parameters and optimizer state of the version of the child model achieving best score are retained.
DISCUSSION AND PROPERTIES
Notice that, none of the defined mutation actions or the evolutionary algorithm allow the creation of child models that can alter the parent model in any way. Once a model has been trained, the parameters storing its knowledge cannot be modified. This method guarantees immunity from catastrophic forgetting, since the knowledge of a trained model is always preserved. It also provides a solution to negative transfer, since it automates the selection of the knowledge that is most relevant for each new task. Furthermore, it also avoids gradient interference, that can arise when multiple gradients are synchronously applied to the same set of parameters. Nonetheless, models for new tasks can use knowledge and representations from prior tasks and even extend these to improve or specialize them.
The method compartmentalizes the knowledge of each task in a subset of components, allowing the implementation of different dataset privacy control policies. For example, we can introduce private tasks that can benefit from all the public knowledge embedded in the multitask system but are able to withhold the knowledge and representations derived from their private dataset from being used by other tasks. This is achieved by preventing other tasks from using or cloning components trained on private data. This also allows to remove the knowledge learned from the private dataset at any future date by simply removing its components. This private/public distinction can be generalized into an access-control-list mechanism. For example, a set of private tasks can share representations that are withheld from any other task. Privacy control capabilities are empirically demonstrated in Section 4.
EXPERIMENTAL SET UP
This section details the instantiation of the proposed method employed in the experimental analysis. The task type for the presented set of experiment is image classification. This choice allows us to define a large benchmark of publicly available datasets with standardized framing. It also allows us to build on top of state-of-the-art models whose architecture definition and checkpoints are public: the Visual Transformer (ViT) is used as root model (Dosovitskiy et al., 2021).
Architecture Layer cloning mutations can create a copy of any of ViT's layers: 1) Patch embedding: the first layer of the model maps the input image into a sequence of embedded tokens, each corresponding to a patch of the input image. 2) Class token: a classification token is prepended to the sequence. The final hidden state corresponding to this token is used as the aggregate sequence representation for classification tasks (Devlin et al., 2019). 3) Position embedding: the sequence representation is then augmented with an embedding that carries each patch positional information. 4) Transformer layers: the sequence representation generated by the input layers is iteratively transformed by a stack of transformer layers (Vaswani et al., 2017). 5) Model head: a final fully connected layer mapping the representation produced by the top-most transformer layer into the logits.
Parameters The parameters of the root model can be either randomly initialized or loaded from a checkpoint. The preliminary experiment demonstrates the evolution from random initialization (see Section 4), while the large scale experiment starts from a pretrained large ViT model (see Section 5).
Hyperparameters As default hyperparameters we use those resulting from the extensive study conducted by Steiner et al. (2021): SGD momentum optimizer, cosine decay schedule, no weight decay, gradient clipping at global norm 1 and 386×386 image resolution. The evolutionary method can change the hyperparameters of optimizer, image preprocessing and architecture (see Table 1).
PRELIMINARY EXPERIMENT
This section describes a small scale preliminary experiment that introduces details of the method application and illustrates the privacy control and initialization capabilities. This experiment demonstrates the ability to generate a multitask model from random initialization and a minimal architecture rather than evolving a pretrained state-of-the-art model. Therefore, a randomly initialized ViT Ti/16 architecture (Steiner et al., 2021) stripped of transformer layers is used as root model. To allow the method to build a capable architecture, we add an extra mutation action that can insert a new randomly initialized transformer layer just before the head layer.
Furthermore, the dataset privacy control technique (see Section 3) is demonstrated by adding a private task. Three numerals recognition tasks are used as benchmark: {bangla, devanagari, telugu}. Telugu is introduced as a private task, so that no other task can access the knowledge introduced into the system by its dataset. However, its models can leverage the knowledge provided by the public tasks.
This short experiment is configured to perform 2 active task iterations for each task. During each active task iteration 4 model generations are produced. In each generation 8 child models are sampled and trained in parallel on each of the 8 TPUv3 cores. The choice of small datasets, architecture and training budget is intended to facilitate reproducibility and fast experimental iterations. The experiment can be reproduced by using the published artifacts and completes in less than 5 minutes. Figure 1 (right) displays the resulting multitask model solving jointly the 3 tasks. We observe a high degree of cross-task knowledge and components sharing throughout the evolutionary process. Even though the root model has no transformer layers, multiple randomly initialized transformer layers are inserted and trained improving the score of each task. Note that, at any point during the evolution, the components trained on the private task (red) are only used by the private task.
LARGE SCALE CONTINUAL LEARNING EXPERIMENT
This section reports a single large scale continual learning experiment producing a multitask system jointly solving 69 visual tasks. A pretrained ViT L/16 is used as root model, which has been selected for its pretraining validation accuracy on the imagenet-21k dataset following Steiner et al. (2021).
VIT BENCHMARK
The first tasks introduced to the system are 3 tasks on which ViT was evaluated in Dosovitskiy et al. This configuration results in 8 epochs for imagenet, and 80 epochs for cifar. This is roughly equivalent to the fine-tuning setup of the baseline model we compare against (Dosovitskiy et al., 2021): 8 epochs for imagenet and 102.4 for cifar. The proposed method can be considered cheaper since: 1) ViT fine-tuning has been repeated multiple times for the hyperparameters tuning process and 2) setting µ = 0.2 results in cheaper training steps, since parameters updates can be skipped for frozen layers and gradient propagation can be skipped for frozen layers at the base of the model (preliminary experiments have shown a 2.5-4% training speed-up attributable to this). In order to provide a fair comparison, as a root model is used the same ViT L/16 architecture, same checkpoint pretrained on the i12k dataset, same 384×384 image resolution, and optimizer and prepossessing configuration. Table 2 reports the top 1 test accuracy achieved. µ2Net outperforms fine-tuning with comparable training steps per task. Extending the training with 5 additional tasks iterations leads to moderate gains on imagenet2012 and cifar10. Notice that, for cifar100 the accuracy decreases. This can happen since the best models are selected according the validation accuracy and, as the model gets close to convergence, a small validation accuracy gain may lead to a noisy perturbation of the test accuracy.
To quantify knowledge transfer, we consider the model produced for each task and examine the dataset on which each layer's ancestors were trained. On average, the layers composing the model generated for the imagenet2012 task have performed only 60.6% of the training steps on the imagenet2012 dataset, and have received 31.5% of the gradient updates from cifar100 and 7.9% from cifar10. The Table 3: Test accuracy achieved on the VTAB-full benchmark by: 1) fine-tuning with matching architecture and checkpoint (ViT L/16 i21k) reported by Steiner et al. (2021), 2) the Sup-rotation method (Gidaris et al., 2018) that achieved the best result in the VTAB-full leaderboard (Zhai et al., 2019b), 3-4) µ2Net results after 2 task iterations, 5) and after an additional iteration performed after the VDD benchmark introduction. Underlined models transfer knowledge from VDD tasks. layers comprising the cifar100 model have performed 42.3% of their training on imagenet2012 and 20.6% on cifar10. And layers comprising the cifar10 model performed 46.1% of training on imagenet2012 and 35.9% on cifar100. The tasks heterogeneity improves the representations produced by the different layers, and results in generally higher performance, as shown in Table 2.
The following sections 5.2 and 5.3 describe the extensions of the system performed by introducing additional benchmarks. After the introduction of each benchmark, we perform an additional iteration on imagenet and cifar tasks to analyze effects of further knowledge enrichment. As a representative example: the VDD benchmark (Section 5.3) includes a low resolution version of cifar100. The model that will be generated for vdd/cifar100 will be a mutation of the current full resolution cifar100 model. Afterward, the additional active task iteration on cifar100 will be performed, and the resulting improved cifar100 model will be a mutation of the low resolution vdd/cifar100 model.
After a final iteration, we note that 99.49 is the best cifar10 accuracy reported for a model trained only on public data: to the best of our knowledge, this constitutes a 15% relative error reduction compared to the 99.
VTAB-FULL BENCHMARK
Next, we introduce to the system the 19 VTAB-full tasks (Zhai et al., 2019a), plus 5 additional task variants that are not included in the standard evaluation set (Table 8). From this experiment onward, the infrastructure is scaled from 8 to 32 cores, as detailed in Appendix B. The number of task iterations is reduced from 10 to 2. These changes lead to a roughly similar exploratory and training budget per task. However, the increased parallelism results in faster task iterations. Table 3 reports the achieved results along with reference models that use limited knowledge transfer capabilities. Steiner et al. (2021) reports the quality achieved by fine-tuning a model equivalent to our root model. This outperforms µ2Net on only 2 tasks, even if it has been trained multiple times to perform hyperparameter tuning. Zhai et al. (2019b) reports the results of the best model identified with a large scale study. This state of the art model outperforms µ2Net on 4 tasks. Again, increasing number of task iterations and additional knowledge (VDD) in the system, seem to yield better quality.
VISUAL DOMAIN DECATHLON (VDD) BENCHMARK
The VDD benchmark (Bilen et al., 2017) is introduced next. The ML methodology proposed in this paper, can achieve higher efficiency by focusing the available compute on the training of a single multitask system. Though, the standard approach to measure variance relies on experiment repetitions. This section demonstrates how variance can be measured for any chosen segment of the training. In practice, 2 task iterations are performed to introduce the VDD tasks starting from the state achieved after the introduction of the last benchmark as usual. But, this experiment is run on 3 parallel copies of the system, allowing us to compute variance of the metrics for this set of task iterations.
The VDD benchmark is composed of 10 diverse tasks. This is also reflected in the diverse variance ranges measured (see Table 4). Variance is low for most of the tasks. However, for ucf101 and aircraft is significantly higher. The metrics that have highest correlation with standard deviation are error rate (linear proportionality in log scale) and number of training samples per class (inverse proportionality in log scale) (Figure 4). These can be considered metrics indicative of the complexity of the task. Furthermore, variance decreases with the second iteration: average standard deviation of 0.87 after 1 iteration and 0.61 after the second. These findings can support the intuitive hypothesis that tasks with higher complexity may benefit from more iterations to decrease variance and approach convergence. The next system extension continues from the state of one randomly selected replica.
MULTITASK CHARACTER CLASSIFICATION BENCHMARK
We continue extending the system by adding a set of 8 character classification tasks. Thus offering the opportunity to study knowledge transfer across tasks with high domain correlation. Table 5 reports the test accuracy achieved with 2 active tasks iterations. We observe that tasks with more training data (left) achieve convergence in the first iteration, this hypothesis is supported by lack of significant accuracy gains with the second iteration. While tasks with less training data (right) show a significant gain from a second training iteration. Smaller tasks use transferred in domain knowledge: bangla top model reuses components that embed knowledge introduced by emnist/letters, while devanagari transfers from omniglot and bangla, and telugu transfers from bangla and devanagari. Furthermore, the achieved quality is comparable to the state of the art published for each task. Table 6: Test accuracy achieved on the VTAB-1k benchmark by: 1) fine-tuning ViT L/16 i21k (matching root model) Dosovitskiy et al. (2021), 2) the Sup-rotation method (Gidaris et al., 2018) that achieved the best result in the VTAB-1k leaderboard (Zhai et al., 2019b). Underlined models have at least one ancestor trained on the corresponding full form task. Doubly underlined model inherit directly from the current best model for the matching full form task. Activated parameters per task Added parameters per task Figure 2: Activated and added parameters per task as percentage of the total number of parameters of the multitask system, measured through the course of the large scale experiment described in Section 5. Vertical lines highlight the start of the introduction for each of the considered benchmark.
VTAB-1K BENCHMARK
The system is further extended by adding the 1k-samples version of the VTAB tasks. Since the system contains already the knowledge learned from the full version of each task, this set allows to study how effective is the proposed method at retrieving knowledge that is already embedded in the system. Table 6 reports results along with reference models that use limited knowledge transfer capabilities. During the first iteration, the models generated for the short form tasks can retrieve the knowledge of the corresponding full form task also without directly mutating its model, but rather mutating a model having at least one ancestor trained on the full task. For example, the model generated for flowers102 1k mutates the dtd 1k model, that has 28 ancestors, of which only the 21 st was trained on oxford_flowers102 f ull . However, that is enough for flowers102 1k to achieve 99.2% test accuracy.
After only one task iteration, 5 tasks achieve better test accuracy than the reference models without reusing any knowledge introduced by the corresponding full form task. Particularly interesting is the case of kitty-dist 1k (a.k.a kitty/closest_vehicle_distance 1k ), that achieves a strong performance without transferring from the matching full form task but composing the knowledge of related tasks: kitty/closest_object_distance f ull (8 ancestors), kitty/count_vehicles f ull (3 ancestors) and kitty/count_vehicles 1k (3 ancestors). Thus learning to estimate distance of the closest vehicle by combining the knowledge of recognizing vehicles and estimating the distance of the closest object. Also, clevr-dist 1k achieves a strong performance by inheriting from the semantically equivalent task kitti/closest_object_distance f ull without reusing the knowledge introduced by clevr-dist f ull .
CONCLUSION
We introduced the µ2Net method, aimed at achieving state-of-the-art quality on a large task set, with the ability to dynamically introduce new tasks into the running system. The more tasks are learned the more knowledge is embedded in the system. A ViT-L architecture (307M parameters) was evolved into a multitask system with 13'087M parameters jointly solving 69 tasks. However, as the system grows, the sparsity in parameter activation keeps the amount of compute and the memory usage per task constant. The average added parameters per task decreases by 38% through the experiment, and the resulting multitask system activates only 2.3% of the total parameters per task (see Figure 2 and Table 7). The proposed method allows decoupling the growth of useful representations for solving new tasks from the growth of parameters/compute per task. Furthermore, experimenting with a large number of tasks allowed us to identify different patterns of positive knowledge transfer and composition, achieving higher efficacy on small datasets and across related tasks. The proposed approach to mutations allows to achieve immunity against common pitfalls of multitask systems such as catastrophic forgetting, negative transfer and gradient interference, and demonstrates the key data privacy properties we want to achieve in a continual learning system. Future work can continue to build toward systems that can acquire further capabilities and knowledge across multiple modalities.
All the experiments reported in this paper can be reproduced by using the public datasets and published µ2Net code, and can be extended by using the published checkpoints (see Appendix B).
A EXTENDED RELATED WORK SURVEY
The proposed method is designed to learn an unbounded number of tasks in a continual learning fashion. In such contexts it aims to learn each task with higher quality and efficiency by automating and optimizing the knowledge transfer among any subset of tasks that can provide useful knowledge to one another. The proposed model is designed to be immune from common multitask learning pitfalls: catastrophic forgetting, gradients interference, negative transfer. Cross-task transfer-learning has gained popularity, especially through transfer learning from a model pre-trained on a large amount of data for one or a few general tasks, and then fine-tuned on a small amount of data for a related downstream task. This approach has been shown to be very effective in a wide variety of problems across many modalities, including language (Devlin et al., 2019;Raffel et al., 2020) andvision (Dosovitskiy et al., 2021;He et al., 2016). The success of transfer-learning applications hinges on adequate prior knowledge selection to avoid typical negative transfer pitfalls (Rosenstein, 2005;. Common solutions rely on data or model selection techniques, often putting emphasis on the efficiency of the exploration (Zhang et al., 2020;Mensink et al., 2021), also method aiming to automate knowledge selection at a layer level have been proposed Sun et al. (2020a). Transfer learning capabilities are critical for multitask models. ML models trained jointly on multiple tasks can be affected by gradients interference if any subset of parameters receive gradients jointly from multiple sources (Chen et al., 2018;Yu et al., 2020), and by catastrophic forgetting of prior knowledge as new tasks are learned (McCloskey & Cohen, 1989;French, 1999).
These knowledge loss problems can be alleviated with weighted combination of tasks (Liu et al., 2019b;Sun et al., 2020b) and gradient transformation methods (Chen et al., 2018;Sener & Koltun, 2018;Kendall et al., 2018). Stronger guarantees are provided by methods that compartmentalize task specific knowledge in dedicated parameter subsets (Rebuffi et al., 2017;Houlsby et al., 2019;Rusu et al., 2016;Rosenfeld & Tsotsos, 2020). Addressing catastrophic forgetting and identifying what subset of parameters/knowledge that is beneficial to share with each task is also critical for continual learning or life long learning methods (McCloskey & Cohen, 1989;French, 1999;Ramesh & Chaudhari, 2022).
The proposed method relies on an evolutionary approach to jointly search the spaces of models architectures, hyperparameters, and prior knowledge selection while optimizing for an possibly multi-factor non-differetiable reward function. The automation of hyperparameter tuning has been commonly addressed with Bayesian optimization (Srinivas et al., 2010;Bergstra et al., 2011;Snoek et al., 2012), evolutionary methods have also been explored for this purpose (Jaderberg et al., 2017;Zhang et al., 2011). Hyperparameters tuning can be considered related to the neural architecture search (NAS), as architectures can be defined by the selection of a sequence of architectural hyperparameters. Initially, NAS methods have been based on reinforcement learning techniques (Zoph & Le, 2017) but sample efficient evolutionary approaches have also proposed (Real et al., 2019;Maziarz et al., 2018). Alternative NAS methods focusing on more efficient parameter-sharing (Pham et al., 2018;Liu et al., 2019a;Kokiopoulou et al., 2019) or optimization for multi-factor quality/cost trade-offs (Tan et al., 2019) have also been explored.
The proposed method is capable to dynamically extend the system, adding capacity or novel structures in an unconstrained fashion. A few methods have been proposed to achieve dynamic architecture extensions (Chen et al., 2016;Cai et al., 2018), some also focusing on an unbounded stream of tasks (Yoon et al., 2018;Yao et al., 2020), or achieving immunity from catastrophic forgetting (Rusu et al., 2016;Li & Hoiem, 2018;Li et al., 2019;Rosenfeld & Tsotsos, 2020).
The proposed method is sparsely activated, thus the unbounded growth of knowledge and parameters is decoupled from the growth of computational cost. The growth in capabilities of state of the art models often requires growth in terms of trainable parameters (Kaplan et al., 2020). Sparse activation techniques at sub-layer level Du et al., 2021) Figure 3: Graph representing the architecture of the multitask system solving jointly 69 image classification tasks generated by the large scale continual learning experiment described in Section 5. Each task is identified with a unique color. Bottom triangular nodes represent the data input of each task. Top rectangular nodes represent the head layer of each task. Each edges sequence of the same color connecting a task input to its head, a path, defines the layers sequence composing the model for each task. Each path traverses 27 round nodes representing ViT L/16 internal layers (see Section 3.5) in the following order from bottom to top: patch embedding, class token, position embedding and 24 transformer layers. Internal nodes are represented with the color of the task on which the parameters of the corresponding layer were trained last. Except for the gray nodes that have not received gradient updates from any of the 69 tasks and still carry the parameters of the root model that were loaded from a checkpoint of a ViT L/16 pretrained on the imagenet-21k dataset (see Section 5) (Video: youtu.be/Hf88Ge0eiQ8). Figure 4: Correlation in log scale with the standard deviation measured during the variance analysis conducted on Visual Domain Decathlon benchmark by running the training on 3 parallel replicas of the system. We display the 2 metrics that are most correlated with the standard deviation: error rate computed on the test set (left) and training samples per class (right). The red line is fitted to minimize the squared distance to the set of points. 1 of 2). For each dataset used in the experiments, this table reports: 1) dataset name indicative of the Tensorflow Datasets Catalogs identification string and linking to the corresponding catalog page ("visual_domain_decathlon" has been abbreviated as "vdd", and "diabetic_retinopathy_detection" as "drd"), 2) train, validation and test data splits, represented with the standard Tensorflow Datasets format ("validation" has been abbreviated as "val").
3) corresponding scientific publication reference. Datasets are listed in the order of introduction into the system. Notes: [1] The test split of the imagenet_v2 dataset is used as validation set for imagenet2012.
[2] The test split of the cifar10_1 dataset is used as validation set for cifar10.
[3] The VTAB-full benchmark also includes the cifar100 task. Cifar100 has been introduced to the µ2Net system as part of the initial benchmark. In the VTAB-full results tables we refer to the top 1 test accuracy achieved in the latest cifar100 training iteration without retraining it as part of the VTAB-full active training iteration.
[5] VTAB additional task, not included in the standard scoring set. These tasks were added to further scale the system and analyze transfer across related tasks. Table 9: Datasets splits and reference (part 2 of 2). | 2022-05-26T13:31:35.261Z | 2022-05-25T00:00:00.000 | {
"year": 2022,
"sha1": "f519131bd56270d07745d5c27314a39e50ac1544",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Arxiv",
"pdf_hash": "c9866e126cb70229e2017087f84b5db446f25a20",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
10872990 | pes2o/s2orc | v3-fos-license | Hrd1-mediated BLIMP-1 ubiquitination promotes dendritic cell MHCII expression for CD4 T cell priming during inflammation
Yang et al. demonstrate that Hrd1 plays an important role in DC induction of CD4 T cell immunity. The underlying mechanism involves the ability of Hrd1 to ubiquitinate and degrade BLIMP-1, thus releasing CIITA from transcriptional repression and promoting MHCII expression. As a consequence, Hrd1−/− DCs protect mice from MOG-induced experimental autoimmune encephalomyelitis.
MHC class I (MHC-I) and class II (MHC-II) are cell surface glycoproteins on APCs involved in the binding and presentation of peptide antigens to the TCRs of T lymphocytes. MHC-I presents antigen from endogenous sources to CD8 + T cells, whereas MHC-II presents peptide from exogenous sources to CD4 + T cells (Rudolph et al., 2006;Adams and Luoma, 2013;Merad et al., 2013). These MHC-peptide-TCR complexes are required to generate antigenspecific immune responses. MHC-II is constitutively expressed in APCs at a relatively low level under naive conditions but can be induced by IFN- (Collins et al., 1984;Koeffler et al., 1984). In addition, pathogen-associated molecule patterns, including LPS and bacterial DNA, are capable of inducing MHC-II gene expression through binding to TLRs (Reis e Sousa, 2004). TLR-induced MHC-II expression plays critical roles in host immunity against microbial infections and is associated with inflammatory autoimmune diseases (Friese et al., 2005).
The MHC class II transactivator (CIITA) is critical for both constitutive and inducible MHC-II gene transcription (Steimle et al., 1993). IFN- and TLR stimulation induce CIITA to activate MHC-II gene transcription during inflammation (Drozina et al., 2005).
CIITA-deficient mice do not express conventional MHC-II molecules on the surface of splenic B cells, macrophages, and DCs (Chang et al., 1994). It has been reported that B lymphocyteinduced maturation protein 1 (BLIMP1), a transcriptional repressor that is capable of triggering plasma cell differentiation, is a repressor of CIITA transcription. BLIMP1-mediated CIITA suppression turns off MHC-II-mediated antigen presentation to induce B cell differentiation into plasma cells (Piskurich et al., 2000). BLIMP1deficient DCs exhibit elevated expression of MHC-II and facilitate CD4 + T cell immunity (Piskurich et al., 2000;Tooze et al., 2006;Chen et al., 2007;Zhao et al., 2007;Kim et al., 2013).
The ubiquitin pathway has been shown to play a critical role in regulating MHC-II antigen presentation at multiple levels (Moffat et al., 2013). Ubiquitination of CIITA is required for its transcriptional activation of MHC-II gene expression (Greer et al., 2003). Ubiquitination also directly regulates MHC-II degradation,
Because Hrd1 has been identified as an anti-apoptotic molecule that protects cells from ER stress-induced apoptosis (Amano et al., 2003), we asked whether Hrd1 gene deletion affects CD11c + DC survival. Surprisingly, loss of Hrd1 function catalyzed by the membrane-associated RING (really interesting new gene)-CH (MARCH) family of E3 ubiquitin ligases (Ishido et al., 2009). Although several MARCH family members have been suggested as regulators of both innate and the adoptive immune responses, MARCH 1, which targets CD86 and MHC-II for ubiquitination-mediated degradation, is the most well characterized member (Matsuki et al., 2007;De Gassart et al., 2008;Young et al., 2008;Walseng et al., 2010;Tze et al., 2011). Given the critical roles of MHC-II in antigen presentation and the activation of the adaptive immune system, it is not surprising that a tight regulatory mechanism is necessary to ensure appropriate MHC-II antigen presentation. However, how the ubiquitin pathway controls MHC-II antigen presentation, in particular the specific E3 ubiquitin ligases that are required in this process, remains largely unidentified.
Hrd1, also known as Synoviolin, is a membrane-spanning protein on the endoplasmic reticulum (ER). It has a RING finger domain followed by a long proline-rich C terminus in its cytoplasmic portion, which is likely involved in recruiting cytoplasmic proteins for ubiquitination. Hrd1 was initially identified as a ubiquitin ligase involved in degrading misfolded proteins (Carvalho et al., 2006;Denic et al., 2006). Because Hrd1 expression is often up-regulated in synovial fibroblasts in patients with rheumatoid arthritis, it was renamed Synoviolin (Amano et al., 2003). We recently reported that proinflammatory cytokines, including TNF and IL-1, are responsible for inducing Hrd1 expression in synovial fibroblasts (Gao et al., 2006). We further observed that Hrd1 ubiquitinates IRE1 (inositol-requiring enzyme 1), a critical kinase in regulating the ER stress response . It has been shown that Hrd1 targets the misfolded MHC-I for degradation in the in vitro cultured cell lines (Burr et al., 2011;Huang et al., 2011). Although the ER stress functions of Hrd1 in misfolded protein degradation have been well studied, its physiological roles in immune regulation are not known.
Hrd1 promotes MHC-II expression by DCs
To study the physiological functions of Hrd1 in DCs, we generated Hrd1 floxed mice. The Hrd1 gene contains 16 exons ( Fig. 1 A), we floxed exons 8-11 that encode a large region of the Hrd1 protein from its fifth transmembrane domain (TM) to the proline-rich sequence leading to deletion of the functional RING finger (Fig. 1, B and C). To exclude the potential effects of the neomycin selection cassette on Hrd1 expression, this cassette was flanked by two flippase recognition target (FRT) sites, which can be deleted by FLP recombinase. This targeting vector was transfected into an embryonic stem cell line generated from C57/BL6 mice. Neomycin selects were screened by PCR. Seven clones were obtained and confirmed to be correct targeting by Southern blotting. Blastocyst injections resulted in several chimeric mice with the capacity for germline transmission. Breeding of heterozygous mice yielded Hrd1 wt/wt , Hrd1 wt/f , and Hrd1 f/f offspring without phenotypic abnormalities in expected Mendelian ratios ( Fig. 1 D and The ER membrane-anchoring protein Hrd1 carries 6 transmembrane (TM) domains, one RING finger domain, and a C terminus proline-rich domain. The deletion of floxed Hrd1 gene by Cre recombinase destroys Hrd1 protein expression. (D) Genotyping of Hrd1-floxed mice. Tail snips from a litter of Hrd1 flox/wt X Hrd1 flox/wt offspring were collected for DNA extraction and PCR analysis. The 302-bp PCR product is the WT allele and the 407-bp product is the mutant allele. (E and F) BM cells were isolated from WT and Hrd1 conditional KO (Hrd1 / ) mice and cultured under DC differentiation conditions. (E) Hrd1 protein expression was analyzed by immunoblotting (top) using -actin as a loading control (bottom). (F) Hrd1 mRNA levels were determined by real-time quantitative RT-PCR. Hrd1 levels in WT DCs increased with LPS treatment. (G) Cell surface expression of B220 and CD11c in total splenocytes from WT and Hrd1 / mice was analyzed by flow cytometry. (H) CD11c + B220 cells were gated and the expression of CD8 and CD11b was analyzed. (I and J) The mean percentages (I) and absolute numbers (J) of CD11c + cells in the spleens of 10 pairs of WT and Hrd1 / mice are shown (n = 10).
Hrd1 regulates MHC-II expression at the mRNA level in DCs
To delineate the mechanism by which Hrd1 regulates MHC-II expression in DCs, we analyzed MHC-II mRNA expression in WT and Hrd1-null BMDCs. Consistent with the observed reduction in MHC-II protein expression in Hrd1-null BMDCs by flow cytometry (Fig. 2), MHC-II mRNA expression was largely diminished in these cells (Fig. 4 A). Importantly, TLR stimulation for 2 h failed to enhance MHC-II transcription in Hrd1-null BMDCs. In contrast, the transcription of MHC-I was not affected by Hrd1 deficiency (Fig. 4 B). Diminished Hrd1 mRNA expression was confirmed in BMDCs from Hrd1 / mice (Fig. 4 C). In addition, a significant increase in DCs did not reduce survival; rather, it led to a slight increase in the percentage and a statistically significant increase in the total numbers of CD11c + DCs in the spleen. In addition, the percentages of CD11c + B220 conventional DCs and CD11c + B220 low plasmacytoid DCs were not altered in the spleens of Hrd1 / mice compared with WT mice (Fig. 1 G). Moreover, analysis of the gated CD11c + B220 DCs by their expression of CD11b or CD8 did not detect any changes in the percentages of CD11c + CD11b + CD8 B220 myeloid DCs and CD11c + CD11b CD8 + B220 lymphoid DCs with Hrd1 gene deletion (Fig. 1, G and H). In addition, a slight increase in the percentage (Fig. 1 I) and a statistical significant increase in the total numbers ( Fig. 1 J) of CD11c + cells were detected in the spleen of DC-specific Hrd1 knockout mice.
Notably, we detected a significant reduction in MHC-II expression on the surface of immature BM-derived DCs (BMDCs). Stimulation with LPS for 24 h led to a dramatic increase in MHC-II expression in WT DCs but failed to upregulate MHC-II expression in Hrd1-null DCs. In contrast, the expression levels of MHC-I, CD80, and CD86 were not altered by Hrd1 gene deletion (Fig. 2, A and B). A similar reduction in MHC-II expression in gated CD11c + cells from the spleens of Hrd1 / mice was further confirmed (Fig. 2, C and D). These results indicate that Hrd1 gene deletion in mice selectively impairs MHC-II expression by DCs.
Because MHC-II expression is critical for CD4 + T cell development and homeostatic proliferation, we asked whether the decrease in MHC-II expression in Hrd1-null DCs alters T cell development in mice. Neither the percentages nor the absolute numbers of CD4 and CD8 double-negative, doublepositive, and single-positive T cell populations were altered in the thymus of Hrd1 / mice (Fig. 3, A and B). In contrast, a 20-30% increase in total splenocytes in Hrd1 / mice was detected ( Fig. 3 D). This enlarged spleen size of Hrd1 / mice appeared to be due to a mean 30% increase in the percentage and total numbers of B220 + B cells (Fig. 3 E) and CD4 + cells (Fig. 3, F-H). Although there is a reduction in the percentage of CD8 + T cells, presumably due to the increase in CD4 + T cell populations in spleen, their absolute numbers were not affected (Fig. 3, F and H). Analysis of CD44 and CD62L expression on the surface of CD4 + T cells showed no changes in CD4 + T cell activation in DC-specific Hrd1-null mice (Fig. 3 I), excluding the possibility that chronic activation caused the observed increase in CD4 + T cells. In addition, there was an 30% reduction in the percentage of DC25 + FoxP3 + T reg cells in the spleens of Hrd1 / mice (Fig. 3, J and K), but absolute T reg cell numbers were not altered ( Fig. 3 L), suggesting that the reduction in T reg cell percentage was likely a consequence of the CD4 + T cell increase. To support this, FoxP3 + T reg cells were not changed in the thymus of DC-specific Hrd1-null mice (Fig. 3 C). Further characterization of the splenic B cells did not did not detect any changes in the percentages of follicular and marginal zone B cells by their expression of CD21 and CD23, and no changes in the cell surface IgM and IgD expression (Fig. 3 M). Similar to CD4 + T cells, no increase in chronic B cell activation was detected because their It has been reported that TLR stimulation suppresses MHC-II mRNA transcription in DCs 18-24 h after stimulation (Landmann et al., 2001;Pai et al., 2002;Yao et al., 2006), but with a transient increase during the early stage (0.5-3 h) of stimulation (Landmann et al., 2001;Casals et al., 2007). We dynamically analyzed the effects of Hrd1 deletion on TLRmediated MHC-II transcription. Indeed, consistent with these previous studies, after a transient increase at 2 h after stimulation, a 40-50% reduction in MHC-II transcription was detected in WT DCs 24 h after LPS stimulation. In contrast, this dynamic transcription of MHC-II in Hrd1-null DCs was dismissed ( Fig. 4 E). The transcription factor CIITA has been identified as a critical factor for TLR-induced MHC-II transcription in DCs (Steimle et al., 1993). Interestingly, we detected that CIITA mRNA expression was diminished in BMDCs and in the gated CD11c + DCs in the spleens of Hrd1 / mice ( Fig. 4 F), indicating that Hrd1 may regulate MHC-II expression by promoting CIITA gene transcription.
DC Hrd1 is a positive regulator for CD4 + T cell priming
The finding of reduced MHC-II expression on the surface of Hrd1-null DCs led us to test whether Hrd1 positively regulates antigen presentation by DCs. Alexia Fluor 647-conjugated chicken ovalbumin (Ax647-OVA) was developed as a convenient approach to measure antigen uptake and presentation as the processed peptides fluoresce (Mansour et al., 2006). We incubated WT BMDCs with Ax647-OVA and found that mean fluorescence intensity increased in a dose-dependent manner, indicating normal antigen presentation and processing. In contrast, mean fluorescence intensity was significantly reduced in Hrd1-null BMDCs incubated with Ax647-OVA ( Fig. 5 A). A similar reduction in fluorescence intensity was observed in freshly sorted CD11c + splenic DCs ( Fig. 5 B), indicating that Hrd1 functions are required for antigen presentation by DCs.
MHC-II expression by DCs is essential for antigen-specific CD4 + T cell priming by directly presenting peptide antigens to their TCRs. Because Hrd1 deficiency in DCs selectively impaired MHC-II expression, we tested whether Hrd1-null DCs were still able to prime CD4 + T cells. WT and Hrd1-null BMDCs were stimulated with LPS in the presence of OVA overnight, washed, and then incubated with naive CD4 + T cells from OT-II TCR transgenic mice. Naive CD8 + T cells from the OT-I mice were used as controls. Antigen-specific T cell proliferation was determined by 3 H-thymidine incorporation and CFSE dilution as previously described (Zhang et al., 2009). OT-II CD4 + T cells proliferated vigorously when co-cultured with WT BMDCs. Conversely, OT-II CD4 + T cell proliferation was diminished when co-cultured with Hrd1-null BMDCs and OVA protein as measured by 3 H-thymidine incorporation (Fig. 5 C) or CFSE dilution (Fig. 5 E). As expected, the proliferation of CD8 + OT-I T cells was comparable when co-cultured with either WT or Hrd1-null DCs (Fig. 5, D and E). As an E3 ubiquitin ligase that promotes protein degradation, the DC Hrd1 may regulate CD4 T cell activation by controlling the protein antigen processing onto MHC-II complex. To test this in Hrd1 mRNA and protein expression was detected in WT BMDCs upon TLR stimulation (Fig. 4, C and D), indicating that TLR signaling normally induces Hrd1 expression in DCs. Next, we studied the role of DC Hrd1 in regulating CD4 + T cell priming in vivo using an adoptive transfer approach. Naive CD45.2 + OT-II CD4 + T cells and control CD45.1 + CD4 + T cells were mixed at a 1:1 ratio, stained with CFSE, and adoptively transferred into lethally irradiated WT and DC-specific Hrd1 KO mice. 1 d after the adoptive transfer, recipients were immunized with OVA 323-339 /CFA. 5 d after immunization, vigorous proliferation of the CD45.2 + CD4 + OT-II T cells in the WT recipients was detected. In contrast, proliferation of CD45.2 + CD4 + OT-II T cells in the DC-specific Hrd1 / recipients was dramatically impaired (Fig. 5, I and J).
notion, we analyzed the T cell proliferation by co-cultivating with OVA peptides. Similar to the results from OVA protein co-culture experiments, loss of Hrd1 in DCs dramatically impaired their ability to prime CD4 + T cells in the presence of the MHC-II-presented OVA 323-339 peptide as measured by both 3 H-thymidine incorporation assay (Fig. 5 F) and CFSE analysis (Fig. 5 H). In contrast, Hrd1 gene deletion in DCs did not affect CD8+ T cell proliferation when co-cultured with the MHC-I-presented OVA 257-264 peptide (Fig. 5, G and H). Therefore, genetic deletion of Hrd1 in DCs impairs CD4 T cell priming is largely due to the reduced MHC-II expression.
(D) WT and
Hrd1 / BMDCs were stimulated with each indicated TLR agonist for 24 h. Hrd1 protein expression levels were determined by Western blotting with -actin as a loading control. (E) WT and Hrd1 / BMDCs were stimulated with 500 ng/ml LPS for 2 or 24 h. The expression levels of MHC-II were determined by real-time PCR. (F) CIITA expression in BMDCs and sorted CD11c + splenic DCs without TLR stimulation was quantified by real-time RT-PCR using -actin as a control. Data are reported as mean ± SD from five independent experiments (n = 5). Student's t test was used for statistical analysis. *, P < 0.05; **, P < 0.01; **, P < 0.01; ***, P < 0.005. (I and J) In vivo OVA-specific CD4 + T cell priming was analyzed as described in Materials and methods. (I and J) Representative images (I) and data reported as mean ± SD (J) from five pairs of WT and Hrd1 / recipients are shown (n = 5). Student's t test was used for statistical analysis. *, P < 0.05; **, P < 0.01; ***, P < 0.0001.
2472
Hrd1 programs DCs for CD4 T cell priming | Yang et al.
These results indicate that DC Hrd1 plays a critical role in antigen-specific priming of CD4 + T cells in mice.
Hrd1 interacts with BLIMP1 in DCs
The significant reduction in both MHC-II and CIITA mRNA levels in Hrd1-null DCs led us to ask whether Hrd1 regulates MHC-II expression by targeting upstream regulatory factors. The transcription factor BLIMP1 has been shown to suppress the expression of both CIITA and MHC-II (Piskurich et al., 2000). We speculated that Hrd1 controls MHC-II and CIITA expression through targeting BLIMP1 in DCs. To test this hypothesis, we first determined whether Hrd1 interacts with BLIMP1. Indeed, Hrd1 protein was detected in anti-BLIMP1 immunoprecipitates from HEK293 cells cotransfected with Hrd1 and BLIMP1, but not in cells transfected with Hrd1 alone (Fig. 6 A). Interaction between endogenous Hrd1 and BLIMP1 in mouse primary BMDCs was confirmed by coimmunoprecipitation and immunoblotting because Hrd1 was detected in immunoprecipitates from BMDC lysates using anti-BLIMP1 but not normal rabbit IgG (Fig. 6 B). In addition, the interaction between Hrd1 and BLIMP1 appears to be regulated by TLR signaling as LPS stimulation of BMDCs for 2 h enhanced their interaction in a dose-dependent manner (Fig. 6 C).
Hrd1 protein contains six transmembrane (TM) domains and its cytoplasmic tail carries an E3 ligase catalytic RING finger and a long proline-rich C terminus (Fig. 6 D). Analysis of BLIMP1 interaction with Hrd1 truncation mutants identified that the Hrd1 proline-rich region mediates its interaction with BLIMP1, as deletion of this region completely abolished the Hrd1-BLIMP1 interaction (Fig. 6 E). The BLIMP1 protein carries 2 acidic domains (AD), one positive regulatory (PR) domain, a proline-rich region, and 5 DNA-binding zinc finger (ZF) regions (Fig. 6 F). We generated a series of truncated BLIMP1 mutants and mapped the domains that interact with Hrd1. As indicated in Fig. 6 G, the BLIMP1 PR domain appears to be required for its interaction with Hrd1, as the N-terminal portion containing the PR domain interacted with Hrd1 and deletion of the PR domain-containing C terminus completely abolished the Hrd1 interaction.
Hrd1 promotes BLIMP1 protein degradation through ubiquitination in DCs E3 ligases often catalyze ubiquitin conjugation onto interacting proteins to modulate their function. We asked whether Hrd1 catalyzes BLIMP1 ubiquitination. Indeed, transient Hrd1 expression in HEK293 dramatically enhanced BLIMP1 ubiquitination. In contrast, only a low level of BLIMP1 ubiquitination was detected in cells without Hrd1 co-transfection, 1 and 2). Hrd1 interaction with BLIMP1 was determined by immunoprecipitation (IP) with anti-Flag antibody followed by immunoblotting with anti-Myc antibody (top, lanes 3 and 4). The same membrane was reprobed with anti-Flag antibody to confirm BLIMP1 expression (bottom). (B) Interaction between endogenous Hrd1 and BLIMP1 in mouse BMDCs was analyzed by IP with anti-BLIMP1 using normal rabbit IgG as a control, followed by immunoblotting with anti-Hrd1 (top). The same membrane was reprobed with anti-BLIMP1 (bottom). (C) BMDCs were stimulated with different doses of LPS as indicated for 2 h. The interaction between BLIMP1 and Hrd1 was determined as in B. (D and E) Truncation mutants of Hrd1 were generated (D) and their interactions with BLIMP1 in transiently transfected HEK293 cells were determined by IP and immunoblotting (E) as described in A. (F and G) Truncation mutants of BLIMP1 were generated (F) and their interactions with full-length Hrd1 in the transiently transfected HEK293 cells were determined by IP and immunoblotting (G) as described in A. TM, transmembrane; Pro, proline; AD, acidic domain; PR, positive regulatory; ZF, zinc finger.
presumably catalyzed by endogenous Hrd1 (Fig. 7 A). Mutation of a critical cysteine to alanine in the RING finger of Hrd1 (Hrd1/CA) that inactivates its E3 ligase catalytic activity completely abolished its ability to catalyze BLIMP1 ubiquitination (Fig. 7 A), although interaction between Hrd1 and BLIMP1 was not affected (Fig. 7 B). These results indicate that Hrd1 is an E3 ubiquitin ligase of BLIMP1 and its functional RING finger is required for Hrd1-mediated BLIMP1 ubiquitination. We then speculated that loss of Hrd1 function would impair or abolish BLIMP1 ubiquitination in DCs. Compared with Hrd1-null DCs, an anti-ubiquitin antibody detected bands with higher molecular weights in anti-BLIMP1 immunoprecipitates from WT BMDCs, suggesting that BLIMP1 protein ubiquitination in WT DCs is lost in Hrd1-null DCs (Fig. 7 C, lanes 1 and 3). Similar to our results that TLR signaling positively regulates Hrd1-BLIMP1 interaction ( Fig. 6 C), LPS stimulation further enhanced BLIMP1 ubiquitination in WT DCs (Fig. 7 C, lanes 1 and 2) but failed to enhance BLIMP1 ubiquitination in Hrd1-null BMDCs (Fig. 7 C, lanes 3 and 4). These results indicate that Hrd1 is required for BLIMP1 ubiquitination in mouse DCs and that TLR signaling enhances BLIMP1 ubiquitination.
We noticed a significant reduction in BLIMP1 protein expression in cells when WT Hrd1, but not the Hrd1/CA mutant, was coexpressed (Fig. 7, A and B). This prompted us to ask whether Hrd1 is involved in regulating BLIMP1 protein degradation. We sorted CD11c + splenic DCs from WT and Hrd1 / mice and analyzed the expression levels of both BLIMP1 protein and its mRNA. BLIMP1 protein expression was relatively low in the immature mouse splenic DCs, and stimulation with LPS for 24 h significantly induced BLIMP1 protein expression (Fig. 7 D, lane 1 vs. 3). Notably, compared with immature WT DCs, a dramatically higher level of BLIMP1 protein was detected in immature Hrd1-null DCs, which again was further enhanced by LPS stimulation (Fig. 7, D and E). Conversely, BLIMP1 mRNA levels were indistinguishable between WT and Hrd1-null DCs, before or after LPS stimulation (Fig. 7 F), indicating that Hrd1 suppresses BLIMP1 expression posttranscriptionally. To confirm this conclusion, we demonstrated that BLIMP1 protein expression level and half-life are significantly increased in Hrd1-null BMDCs (Fig. 7 G). Therefore, our results collectively indicate that Hrd1 is an E3 ubiquitin ligase of BLIMP1 that controls BLIMP1 protein stability in DCs.
Hrd1 regulates MHC-II expression through suppression of BLIMP1
Because BLIMP1 has been shown to suppress CIITA and MHC-II expression (Kim et al., 2013), and we demonstrated that protein expression of BLIMP1 is elevated in Hrd1-null DCs (Fig. 7 D), we hypothesized that Hrd1 promotes MHC-II expression by catalyzing BLIMP1 protein degradation. To test this hypothesis, we used an shRNA-mediated knockdown Hrd1 has been initially identified as an E3 ubiquitin ligase critical for misfolded protein degradation, raising the possibility that loss of Hrd1 expression attenuates MHC-II expression through unfolded protein response (UPR). However, analysis of the downstream UPR genes, including ERdj3, ERdj4, and Edem1 (ER-associated degradation), Pdi4 and Fkbp11 (protein folding), Ire1 (ER stress transducer), and Sec22L1 (ER-Golgi transport; Zhang et al., 2011), did not detect any significant increases in their transcription levels in Hrd1-null BMDCs, even after stimulation with the pharmacological UPR inducer tunicamycin (Fig. 8 G). Together with our observation that Hrd1 deficiency did not facilitate the ER stress-induced DC death (Fig. 2), our results largely exclude the possibility of misfolded protein responses caused by Hrd1 deficiency in DCs. In addition, we have previously reported that Hrd1 promotes the degradation of IRE1 . Indeed, we did detect that the IRE1 protein expression levels are increased in Hrd1-null DCs (unpublished data), raising a possibility that loss of Hrd1 expression may inhibits MHC-II expression due to the elevated IRE1 functions. However, the MHC-II expression levels, both its protein and mRNA, are not affected in IRE1-null DCs (Fig. 8 H), excluding the possible role of IRE1 in MHC-II expression in DCs. Collectively, our results indicate that Hrd1 regulates MHC-II expression through targeting BLIMP1 degradation.
Genetic Hrd1 deletion attenuates autoimmune response in mice
When primed by specific antigens, CD4 + T cells undergo clonal expansion and then differentiate into effector T helper approach and examined whether BLIMP1 suppression rescues MHC-II expression in Hrd1-null DCs. shRNA-mediated knockdown inhibited >95% of BLIMP1 protein expression in both WT and Hrd1-null DCs (Fig. 8 A). Suppression of BLIMP1 expression rescued MHC-II expression in Hrd1null BMDCs to levels comparable to those of WT BMDCs (Fig. 8, B and C). We also noticed that the MHC-II expression level in Hrd1-null BMDCs with BLIMP1 knockdown was slightly but statistically significantly lower than that in WT BMDCs with BLIMP1 knockdown; this may be due to incomplete BLIMP1 knockdown in Hrd1-null DCs. We also found that BLIMP1 knockdown rescued CIITA mRNA expression in Hrd1-null DCs (Fig. 8 D). Collectively, these data suggest that Hrd1 diminishes BLIMP1-mediated CIITA suppression to promote MHC-II expression by ubiquitination and degradation of BLIMP1 protein.
Next, we determined whether BLIMP1 knockdown could rescue the ability of Hrd1-null BMDCs to prime CD4 + T cells. As shown in Fig. 8 E, CD4 + T cells from OT-II mice failed to proliferate when co-cultured with Hrd1-null DCs and OVA 323-339 peptides, confirming that Hrd1 is required for DC function in priming CD4 + T cells. BLIMP1 knockdown in Hrd1-null BMDCs rescued CD4 + OT-II T cell proliferation (Fig. 8, E and F). BLIMP1 knockdown in WT BMDCs also resulted in a slight but statistically significant increase in CD4 + T cell proliferation (Fig. 8, E and F). Therefore, BLIMP1 protein accumulation in Hrd1-null BMDCs appears to interfere with the function of these cells in CD4 + T cell priming, supporting the model that DC Hrd1 regulates CD4 + T cell activation through BLIMP1 degradation. (E and F) BLIMP1 knockdown and control BMDCs were stimulated with LPS (200 ng/ml) and OVA 323-339 peptide overnight, washed, and co-cultured with CD4 + T cells from OT-II mice. Representative images of CD4 + T cell proliferation (E) and mean percentages of dividing CD4 + cells ± SD (F) from three independent experiments are shown. Student's t test was used for the statistical analysis of the data in (C, D, and F). *, P < 0.05; **, P < 0.01; ***, P < 0.005. (G) WT and Hrd1 / splenic DCs were stimulated with or without tunicamycin for 2 h and the mRNA expression levels of each indicated UPR genes were analyzed by real-time PRC. (H) The expression levels of MHC-II on the surface of WT and IRE1a / splenic DCs (top) and MHC-II mRNA were analyzed. Error bars represent data from five pairs of mice (n = 5).
presentation (Burr et al., 2011;Huang et al., 2011). However, we found that in mouse primary DCs, Hrd1 is not required for MHC-I protein and mRNA expression. Therefore, it is likely that Hrd1 selectively regulates MHC II gene transcription but degrades the misfolded MHC I molecules in DCs.
We identified Hrd1 as a specific E3 ubiquitin ligase that catalyzes BLIMP1 ubiquitination and degradation in mouse DCs. Loss of Hrd1 led to an accumulation of BLIMP1 in DCs, which reduced MHC-II expression; BLIMP1 knockdown rescued CIITA and MHC-II expression as well as CD4 + T cell priming. Therefore, Hrd1-mediated BLIMP1 degradation promotes the transcription of both CIITA and MHC-II genes. To our knowledge, Hrd1 is the first identified E3 ubiquitin ligase that targets BLIMP1. In addition to its important physiological functions in DCs, BLIMP1 has been shown as a critical regulator in B cell differentiation of plasma cells (Kikuchi et al., 1995;Simard et al., 2011;Maseda et al., 2012), follicular T helper cell differentiation (Crotty, 2011), CD8 + memory T cell development (Martins and Calame, 2008), and myeloid cell functions (Chang et al., 2000). It will be interesting to determine whether Hrd1-mediated BLIMP1 ubiquitination and degradation regulate the functions of these immune cell types. In addition to BLIMP1, this ER membrane-anchoring E3 ubiquitin ligase Hrd1 has been shown to target transcription factors including Nrf1 (Steffen et al., 2010;Tsuchiya et al., 2011) and p53 (Yamasaki et al., 2007). More recently, we have shown that Hrd1 catalyzes Nrf2 degradation to suppress Nrf2-mediated cellular protection during liver cirrhosis (Wu et al., 2014). As expected, similar to BLIMP1, Hrd1 recognizes these target proteins through its cytoplasmic C terminus domain. More importantly, the regulation of these transcription factors is unlikely triggered by misfolded protein response. Therefore, the accumulated evidences suggest that Hrd1 is involved in a variety of pathobiological functions by degrading cellular signaling molecules through its cytoplasmic domain.
MHC-II expression is critical for CD4 + T cell development. Mice with targeted gene deletion of MHC-II, or factors required for MHC-II gene transcription such as CIITA, lack CD4 + T cells (Chang et al., 1994;Madsen et al., 1999). Although the DC-specific Hrd1 knockout mice showed a significant reduction in MHC-II expression, the development of CD4 + and CD8 + T cells in their thymus was not affected. On the contrary, CD4 + T cells and B cells, but not CD8 + T cells, in the peripheral lymphoid organs were slightly increased in the DC-specific Hrd1 knockout mice compared with WT mice. The mechanisms underlying the increase in CD4 + T cells and B cells are not clear. It is possible that the relatively lower percentage of FoxP3 + T reg cells, which normally suppress the homeostatic proliferation of both T cells and B cells, permits an increase in the CD4 + T cells and B cell populations in the Hrd1 KO mice. However, T reg cell reduction cannot completely explain why the increase is specific to CD4 + T cells and B cells without affecting CD8 + T cells. The reduced percentage of T reg cells is probably not related to a developmental defect because their absolute numbers in the (Th) cells including Th1, Th2, Th17, and T reg cells. Both Th1 and Th17 are critical pathogenic factors in multiple sclerosis in humans and experimental autoimmune encephalomyelitis (EAE) in mice (Pierson et al., 2012). Because Synv-null DCs are incapable of priming CD4 + T cells, we reasoned that genetic deletion of Hrd1 in DCs would attenuate the CD4 + T cell-mediated autoimmune inflammatory response and disease progression. We used a myelin oligodendrocyte glycoprotein (MOG)induced EAE mouse model to test this hypothesis. Because both CD4 + T cells and B cells are increased in DC-specific Hrd1 / mice (Fig. 3), we generated Hrd1 f/f CD11c-Cre + /RAG1 / double KO (Hrd1 / /RAG1 / ) mice and used an adoptive transfer approach. CD11c + cell-depleted splenocytes from WT C57/B6 mice were adoptively transferred into Hrd1 / / RAG1 / double KO and Hrd1 +/+ /RAG1 / mice. EAE induction with MOG was initiated 1 d after adoptive transfer and the clinical symptoms were scored daily. As shown in Fig. 9 A, in contrast to the severe disease that developed in Hrd1 +/+ RAG1 / recipients after MOG immunization, only modest symptoms with a dramatic delay in onset were observed in Hrd1 / /RAG1 / mice, indicating that Hrd1 suppression in DCs protects mice from MOG-induced EAE. MOG-specific CD4 + T cell proliferation (Fig. 9 B) and IL-2 production ( Fig. 9 C) were largely diminished in the draining lymph nodes of Hrd1 / /RAG1 / recipients. As a consequence of impaired CD4 + T cell priming during disease progression in Hrd1-null mice, the differentiation of MOG-specific Th1 and Th17 cells was significantly inhibited (Fig. 9 D). In addition, we confirmed a significant reduction in MHC-II, but not MHC-I, expression on CD11c + DCs in the draining lymph nodes of Hrd1 / /RAG1 / mice (Fig. 9, E and F). Therefore, our studies collectively indicate that DC-specific Hrd1 suppression attenuates autoimmune EAE through down-regulation of MHC-II expression, which inhibits MOG-specific CD4 + T cell priming.
DISCUSSION
Our study shows that the ER membrane-spanning E3 ubiquitin ligase Hrd11 is required for DC-controlled CD4 + T cell priming in the autoimmune inflammatory response. This conclusion is supported by the following observations: targeted deletion of the Hrd1 gene in DCs impaired MHC-II expression at the transcriptional level; Hrd1-null DCs failed to prime CD4 + T cells without affecting CD8 + T cell activation; Hrd1 targets BLIMP1, a nuclear transcriptional repressor, for ubiquitinmediated degradation in DCs; Hrd1 promotes MHC-II gene transcription and CD4 + T cell priming through BLIMP1 degradation; and genetic deletion of Hrd1 gene in DCs partially protects mice from MOG-induced EAE.
Hrd1 appeared to specifically regulate MHC-II expression without affecting the expression of either MHC-I or costimulation molecules, including CD80 and CD86. As a consequence, Hrd1-null DCs failed to prime CD4 + T cells, but antigenspecific CD8 + T cell activation was not affected. Recent studies showed that Hrd1 catalyzes the degradation of misfolded MHC-I and is possibly involved in MHC-I-restricted antigen were increased. It has been shown that the ER stress responsive transcription factor XBP-1 (X-box binding protein 1) is required for the development and survival of CD11c + DCs (Iwakoshi et al., 2007). We have recently reported that Hrd1 can promote degradation of IRE1 , the only known enzyme that activates XBP-1 (Shen et al., 2001), and our unpublished data show that the IRE1 protein, but not its mRNA, is increased in Hrd1-null DCs. Together with our observations that Ire1 gene deletion did not alter MHC-II expression on DCs, these studies suggest that loss of Hrd1 may promote DC survival due to an elevation in IRE1mediated XBP-1 activation, but not for the impaired MHC-II expression. We are currently generating DC-specific IRE1 / / Hrd1 / double knockout mice to study the role of the Hrd1-IRE1 pathway in DC survival.
In summary, our studies reveal a previously unappreciated molecular mechanism that regulates MHC-II gene transcription, involving the ER membrane-spanning E3 ubiquitin ligase Hrd1. In this model, TLR signaling induces Hrd1 expression to promote BLIMP1 ubiquitination and degradation. Because BLIMP1 is a critical transcriptional suppressor of MHC-II expression, TLR-induced Hrd1 expression enhances MHC-II expression to facilitate CD4 + T cell response.
MATERIALS AND METHODS
Generation of Hrd1 floxed mice. The Hrd1-targeting vector was generated as in Fig. 1 A, and then transfected into an embryonic stem cell line generated from C57BL/6 mice. Neomycin selects were screened by PCR. Seven clones were obtained and confirmed by Southern blotting. Blastocyst injections resulted in several chimeric mice with the capacity for germline transmission. Breeding of heterozygous mice yielded Hrd1 wt/wt , Hrd1 wt/f , and Hrd1 f/f mice without phenotypic abnormalities in expected Mendelian ratios (Fig. 1 C). The DC-specific Hrd1-null mice were generated by breeding Hrd1 floxed mice with CD11c-Cre transgenic mice.
CD11c-Cre transgenic mice, OT-I and OT-II TCR transgenic mice, and RAG1 knockout mice, all of which are at the C57BL/6 genetic background, were purchased from The Jackson Laboratory. IRE1 floxed mice were used as previously described (Qiu et al., 2013). All mice used in this study were maintained and used at the Northwestern University mouse facility under pathogen-free conditions according to institutional guidelines. All animal study proposals have been approved by the institutional animal care and use committees (IACUC) at Northwestern University.
BMDC cultivation and activation. BMDCs were generated as previously described (Yang et al., 2013). BM cells were isolated from leg bones of 8-10wk-old WT and Hrd1 f/f CD11c-Cre + (Hrd1 / ) mice and were cultured in RPMI medium containing 10% FCS and GM-CSF (20 ng/ml; BioLegend). Cell cultures were fed on days 3, 6, and 8 and used on day 9 or 10. To select peripheral lymphoid organs of Hrd1-null mice were not different than in WT controls. Rather, it is likely that the reduction in T reg cell percentage is a consequence of the increase in CD4 + T cells. However, because MHC-II-restricted expression of self-antigen is critical for FoxP3 + T reg cell development (Sakaguchi, 2005), the possibility that Hrd1 regulates DC MHC-II expression to modulate T reg cell development cannot be fully excluded. Further, in addition to DCs, CD11c expression can be detected in other cell types (Miller et al., 2012). Therefore, further studies are needed to delineate how Hrd1 expression in DCs, as well as other types of CD11c + cells, affects CD4 + and B cell homeostatic proliferation in mice.
Hrd1 has been identified as an anti-apoptotic factor that protects cells from ER stress-induced cell death (Carvalho et al., 2006;Denic et al., 2006), raising the possibility that loss of Hrd1 might cause DC apoptosis. However, the percentages of CD11 + cells in the spleens of Hrd1 / mice were not altered and the absolute numbers of CD11 + cells in the spleen Figure 9. Genetic deletion of Hrd1 gene in DCs partially protects mice from MOG-induced EAE. (A) CD11c-depleted splenocytes from WT C57/B6 mice were adoptively transferred into 8-wk-old RAG1 / and Hrd1 / /RAG1 / double KO mice. 1 d after transfer, recipients were immunized with MOG 35-55 peptide (100 µg per mouse, emulsified with CFA). Mice were also given pertussis toxin (200 ng per mouse) on days 0 and 2 via tail vein injection. All mice were weighed and examined for clinical symptoms. Error bars represent data from six pairs of mice (mean ± SD; n = 6). **, P < 0.01. (B-F) Splenocytes from MOG-immunized mice were isolated, stained with CFSE, and co-cultured with MOG peptide for 3 d. (B) Proliferation of CD4 + T cells was analyzed by flow cytometry. (C) Percentages of IFN--producing Th1 and IL-17-producing Th17 cells were analyzed by intracellular staining. (D) IL-2 production in the culture supernatant was examined by ELISA. Data represent means ± SD from 6 pairs of mice. **, P < 0.01. (E and F) CD11c + conventional DCs in the draining lymph nodes from mice during disease were analyzed for their expression of MHC-II (E) and MHC-I (F) by flow cytometry. The mean fluorescence identity is indicated (mean + SD).
and assigned scores on a scale of 0-5 as follows: 0, no overt signs of disease; 1, limp tail; 2, limp tail and partial hindlimb paralysis; 3, complete hindlimb paralysis; 4, complete hindlimb and partial forelimb paralysis; 5, moribund state or death. Table S1 shows the sequence details of primers used in this study. Online supplemental material is available at http:// www.jem.org/cgi/content/full/jem.20140283/DC1. | 2016-05-12T22:15:10.714Z | 2014-11-17T00:00:00.000 | {
"year": 2014,
"sha1": "b76d09538df0da11a9a55a5e6e625bb3a7ee4918",
"oa_license": "CCBYNCSA",
"oa_url": "https://rupress.org/jem/article-pdf/211/12/2467/1011767/jem_20140283.pdf",
"oa_status": "BRONZE",
"pdf_src": "Adhoc",
"pdf_hash": "67e5a726f1abb241782d732c6c8f68a7a20d6cfb",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Medicine",
"Biology"
]
} |
244721 | pes2o/s2orc | v3-fos-license | Energy Harvesting Based Body Area Networks for Smart Health
Body area networks (BANs) are configured with a great number of ultra-low power consumption wearable devices, which constantly monitor physiological signals of the human body and thus realize intelligent monitoring. However, the collection and transfer of human body signals consume energy, and considering the comfort demand of wearable devices, both the size and the capacity of a wearable device’s battery are limited. Thus, minimizing the energy consumption of wearable devices and optimizing the BAN energy efficiency is still a challenging problem. Therefore, in this paper, we propose an energy harvesting-based BAN for smart health and discuss an optimal resource allocation scheme to improve BAN energy efficiency. Specifically, firstly, considering energy harvesting in a BAN and the time limits of human body signal transfer, we formulate the energy efficiency optimization problem of time division for wireless energy transfer and wireless information transfer. Secondly, we convert the optimization problem into a convex optimization problem under a linear constraint and propose a closed-form solution to the problem. Finally, simulation results proved that when the size of data acquired by the wearable devices is small, the proportion of energy consumed by the circuit and signal acquisition of the wearable devices is big, and when the size of data acquired by the wearable devices is big, the energy consumed by the signal transfer of the wearable device is decisive.
Introduction
Body area networks (BANs) are small wireless sensor networks (WSNs) which support a lot of medical applications and provide a solution for smart health monitoring [1]. For an exhaustive introduction to BANs, we refer the reader to [2,3]. BANs are configured with ultra-low power consumption wearable devices [4] and medical sensors [5] (such as digestible medical electronics). These sensors constantly monitor physiological signals and movement data of the human body; they transfer such signals and data to the cloud for analysis, thus realizing intelligent monitoring of the user's health [6,7].
However, data collection and transfer of BAN sensors consume energy. To increase the comfort of wearable devices, the battery of wearable devices is usually small, thus the battery's capacity is limited [8,9]. Old batteries need to be replaced frequently or recharged regularly. For medical sensors (such as digestible sensors), it is impossible to replace or recharge the battery. Thus, minimizing the energy consumption of wearable devices and optimizing BAN energy efficiency is still a challenging problem [10].
Nowadays, most research concerning BAN energy efficiency optimization focuses on the design of the routing algorithm [11,12], duty-cycle-based data collation [1], data reduction and compressed sending [13,14], and cross-layer design [15]. For example, some papers have proposed novel approaches to reduce energy consumption through an adaptive routing algorithm [16], dynamic programming for heterogeneous networks [17], and voltage/frequency scaling [18], and secure data transmission [19]. Other papers have dealt with the issue of data-generation uncertainty in the optimal design of BANs: [20,21] proposed a robust optimization model solved by fast mixed integer programming heuristics, based on the algorithm for robust capacitated network design proposed in [22]; Reference [23] has instead investigated the adoption of a min-max regret model. However, few works realize BAN energy efficiency optimization through energy harvesting. Generally, energy in the BAN sensor can be harvested in the following three ways: • Energy harvesting through the environment: some sensors harvest energy through renewable energy sources (including solar, wind, and luminous energy resources).
•
Energy harvesting through the human body: some sensors harvest energy from their own heat energy, bio-energy, body surface friction and body movement.
•
Energy harvesting through a wireless signal: some sensors harvest energy by acquiring wireless signals.
For energy harvesting through the environment, it is impossible for the user to be exposed to strong sunshine or strong wind for long time and this energy harvesting mode depends greatly on the weather and other conditions, which may lead to a longer delay [24,25]. Thus, it is not applicable to BANs. For energy harvesting through the human body, the bio-energy of the human body is unstable and may result in unreliable energy production; furthermore, wearing an additional energy-harvesting device may result in discomfort [26]. Thus, it is also not applicable to a BAN. For energy harvesting through a wireless signal or radio frequency (RF) energy harvesting [27,28], considering that wireless signals exist everywhere constantly and controllably, it is a feasible approach to provide reliable energy to the low power consumption sensors.
A RF energy harvesting based BAN includes two stages: wireless energy transfer (WET) and wireless information transfer (WIT) [29,30]. A challenging problem of the simultaneous wireless information and power transfer (SWIPT) is how to allocate resources between the WET and WIT, so as to minimize the energy consumed by the network. Some research has included primary exploration and discussion on BANs. For example, Abubaka et al. [5] proved in their research that WET to the sensor in the digestive tract could be realized through an antenna outside the human body and such energy was sufficient to keep the sensor working normally, including being able to monitor the environment and temperature of the digestive tract and the special nutrition cost.However, these studies failed to take into consideration the energy consumed by the circuit and data collection and processing of the sensor. In fact, the small size of the sensor in a BAN may result in a great proportion of energy being consumed by the circuit, data collection and processing. Thus, it is obviously impossible to neglect such power consumption.
In this paper, we propose the energy harvesting-based BAN, i.e., the sensor in a BAN can harvest energy from access points and transmit the collected data to access points. Furthermore, we study the resources allocation scheme that realizes minimized energy consumption in a BAN. To be specific, the main results and contributions of this paper include the following:
•
We introduce energy harvesting into a BAN to improve BAN energy efficiency. Compared with traditional BANs, the energy harvesting based BAN proposed in this paper can significantly improve the BAN energy efficiency.
•
We formulate the optimization problem concerning time allocation for the WET and WIT in a BAN, with the aim of minimizing energy consumption in the sensor when considering the WIT and WET time limits. Furthermore, we convert such a problem into a convex optimization problem under linear constraints.
•
We propose a closed-form solution to the optimization problem based on Karush-Kuhn-Tucker (KKT) conditions. Simulation results showed that when the size of data acquired by the wearable devices is small, the proportion of energy consumed by the circuit and information collection of the wearable devices is big, and when the size of data acquired by the wearable devices is big, the energy consumed by information transfer of the wearable device is decisive.
The remainder of this article is organized as follows. The system model is described in Section 2. We first formulate an optimization problem to minimize the energy consumption of the energy harvesting-based BAN. Then, we transform the problem to a convex optimization problem with linear constraints and propose a closed-form solution in Section 3. Our experimental results and discussions are provided in Section 4. Finally, Section 5 concludes this paper.
System Model
In this section, we introduce the network architecture of an energy harvesting-based BAN, as well as the WET and WIT model based on the time division multiple access (TDMA) protocol.
Energy Harvesting-Based Body Area Networks Model
We consider an energy harvesting BAN as shown in Figure 1; the ultra-low power consumption sensor configured to the BAN could collect the physiological signals of the human body, such as electrocardiography signals. The access point recharged the sensor through WET at a fixed interval and the sensor needed to deliver collected physiological signals of the human body to the access point. That is, the human body sensor harvested energy from the access point. When the sensor acquired energy, some of the harvested energy was used for signal transfer, while some of the harvested energy was consumed by the circuit and data acquisition of the sensor. In this paper, we assume that the access point has a stable energy supply and can provide sufficient energy to the sensor.
In this paper, the transfer protocol used is the TDMA protocol as shown in Figure 2. We assume that there are n sensors within the area covered by the access point and these sensors are denoted as S = {S 1 , S 2 , · · · , S n }. Let t 0 , t 1 , · · · , t n denote the time slot and T t 0 , T t 1 , · · · , T t n represent the duration of the time slot. Considering that the WET and WIT will be finished within the period T, T can be divided into n + 1 duration of the time slot, i.e., In the duration of the time slot t 0 , WET is performed and the access point transfers energy to n sensors by way of signal broadcasting. In the duration of the time slot t 1 , t 2 , · · · , t n , n sensors perform WIT and transfer information to the access point through the harvested energy.
Transmission Model
The transmission modes included WET and WIT. In WET, the sensor harvests energy through the access point. In WET, the sensor transfers acquired information to the access point. Detailed models are shown below.
Wireless energy transfer model: Let P b denote the access point transmission power; then, according to the the work of You et al. [31], the energy harvested by the sensor S i , denoted as E H i , is given as: where h DL i is the channel gains from the access point to the sensor S i , and η (0 < η < 1) is the energy converting efficiency.
Information transfer model: After the sensor harvested energy, it is necessary to transmit the collected information to the access point. The channel gain of the sensor S i is defined as h i , and the transfer power of S i is defined as p i . Then, according to the work of You et al. [31], the transfer rate of S i in the time slot t i , denoted as r i , is given as: where σ 2 is the variance of complex white Gaussian noise, B is the channel bandwidth from the sensor S i to the access point.
Energy Consumption Model
The energy consumption of the sensor S i in T t i can be divided into the following three parts:
1.
Energy consumed by the circuit of the sensor S i . Considering that the sensor works constantly, this energy is constant and we denote it as E c i .
2.
Energy consumed by signals (such as the perception, collection and storage of signals) processed by the sensor S i . This part of energy consumption is associated with the data size which is denoted as ω i to be processed. Let γ i denote the energy consumed by the processing of one bit of data. Then, this part of energy can be expressed as E proc i Energy consumed for information transfer by the sensor S i . This part of energy consumption is also related to the data size ω i and can be represented as E tran According to the work of You et al. [31], under a given time constraint, the most energy-efficient data transfer policy is fixed-rate transmission over the whole time slot. Thus, for the sensor S i in the duration of time slot t i , the lowest energy consumption transfer rate is fixed at r i = ω i /T t i . Based on the above discussion and Equation (5), we can rewrite the E tran i as follows: Thus, the total energy consumed by the sensor S i , denoted as E loc i , can be obtained as follows:
Problem Formulation and Solution
In this section, we introduce the optimization problem of BAN time allocation, with the aim of achieving rational allocation of resources, thus minimizing the energy consumption of the sensor.
Problem Formulation
In this paper, we assume that T t 0 is fixed. That is, the system gives the parameter T t 0 earlier, in which energy is transferred to sensors in the human body. T opt = [T t 1 , T t 2 , · · · , T t n ] with the aim of dividing the period T − T t 0 rationally, thus minimizing the energy consumption of the sensor. In view of the discussion described in Section II, the following optimization problems can be obtained: subject to: where the objective function (6) is the energy consumption of the minimized sensor. The constraint condition (7) meant that the WIT of n sensors was finished within the time duration T − T t 0 .The constraint condition (8) meant that the energy consumed by the sensor should not exceed the harvested energy. To solve the above optimization problem, we further adapt the problem and combine the constraint condition (8) and the objective function (6); then, we can obtain the energy savings ∆E i as follows: Thus, the following optimization problem, which is equivalent to P1, can be obtained: subject to: In the next subsection, we characterize the solution of the problem.
Closed-Form Solution
As for the optimization problem described above, it can be proven to be a convex optimization problem as follows. Theorem 1. Problem P2 is the convex optimization problem.
Proof of Theorem 1.
Define the function f (x) = 2 x B − 1; then, we can obtain the first and the second derivative of f (x) as follows: Here, it is clear that f (x) > 0, thus f (x) is convex. Considering the perspective function of is a convex function. Since the sum of convex functions is still convex, the objective function is a convex function. The restriction is a linear constraint. Thus, the optimization problem is convex [32].
As for problem P2, we can define its Lagrange function as follows: where λ is the Lagrange multiplier. It is assumed that T * t i and λ * are the optimal solutions of problem P2 and its dual problem, respectively. Based on the Karush-Kuhn-Tucker (KTT) conditions, the following conditions can be obtained: Based on these conditions, the optimal time allocation scheme can be derived as the following theorem. Theorem 2. The optimal time allocation scheme for Problem P2 is shown as follows.
where W(x) is the Lambert function and ∑ n i=1 Proof of Theorem 2. For the sake of simplicity, we define . Based on (16), it has: Using the properties of exponential functions, we can obtain: Based on the definition of the Lambert function, the solution of (19) can be obtained: Thus, the optimal resource allocation scheme will be given as follows: Furthermore, based on (15), we can obtain λ * = 0 or ∑ n i=1 T * t i + T t 0 − T = 0. (1) If λ * = 0, the optimal resource allocation scheme would be shown as below: since W(−e −1 ) = −1, the denominator of (22) is zero. Thus, λ * satisfy λ * > 0.
(2) If λ * > 0, the optimal resource allocation scheme would be obtained as follows: and it satisfies ∑ n i=1 T * t i = T − T t 0 .
Simulation Results
In this section, we conduct the simulation experiments. Firstly, we set the simulation parameter, then we evaluate the performance.
Parameter Setting
We assume that there are 6 sensors in BAN. As for the WET stage, assume that T = 1 s and the WET time slot is t 0 = 200 ms. The channels h i are modeled as independent Rayleigh fading with average power loss set as 10 −3 . We set the transmitted power of the access point as P b = 100 W. For the WIT stage, we set the bandwidth as W = 5 MHz and the variance of complex white Gaussian channel noise as σ = 10 −9 W. In this paper, for the purpose of convenience, we assume that the sensors have equal circuit consumption E c i and energy γ i consumed for processing one bit of data. Hence, in this paper, E i = 0.001 J and γ = 10 −4 J/bit. The data size to be transferred followed uniform distribution with the mean value ω = 1000 bit.
Energy Cost of Sensors
We give the effects of data size ω on the energy consumption of the sensor. The X-axis represents the size of data (Kbits), the Y-axis is the energy consumption of all the sensors and we used logarithmic coordinates to the axis Y, and the corresponding unit of measurement is log 10 Joule.
It can be seen from Figure 3 that along with the increase in data size ω, more energy will be consumed. It is because the increase of data size requires more energy for data transfer and information processing. It can be observed from Figure 3a that with the same size of data transfer, more energy consumed in processing each bit of data suggests more energy consumed by the system. It is clear in Figure 3b that with the same transferred data size, the bigger the energy E c consumed by the circuit of the sensor, the bigger the energy consumed by the sensor. Furthermore, it can be deduced based on these two figures that when the transfer data size is smaller, for instance, ω = 1000 bits, the difference between curves is bigger. It is because when ω is smaller, the energy consumed by the circuit of the sensor and the energy consumed by information processing take a greater proportion; when the transfer data size is bigger, such as ω = 1200 bits, the difference between curves decreases because when the transfer data size ω is big, the energy consumption of the sensor is mainly energy consumed for data transfer. Moreover, Figure 4 shows that more energy is required as the number of sensors n increases. The reason is that as the number of sensors increases, sensors can collect more data and deliver the data to the access point, which consumes more energy.
Time Duration Allocation of Sensors
In this subsection, we discuss the relationship between transmission data size ω and time duration allocation T t i . Figure 5a shows the effects of transmission data size ω on the time duration allocation. The X-axis represents the mean value of transmission data size that follows a uniform distribution. The Y-axis is the time duration allocation. From the figure, we can observe that the time duration allocated by sensors 1, 2 and 3 varies little with the increment of transmission data size. Compared with Figure 3, we can conclude that the size of data transmitted has a greater impact on energy cost than on the time duration allocation. Figure 5b shows how the system allocates time duration for each sensor when given a set of transmission data sizes. The X-axis represents the specific transmission data size and the Y-axis indicates the time duration allocation. From the figure, we can observe that the time duration allocation increases with the increment of given transmission data.
Conclusions
In this paper, we proposed an energy harvesting-based body sensor network, and a method based on time division multiple access (TDMA); we built upon the optimization problem of time division for wireless information transfer and proved that this optimization problem was actually a convex optimization problem; we gave a closed-form solution to the problem. The simulation results of the experiment indicated that when the size of data acquired by the sensor was small, the energy consumption of the sensor was mainly energy consumed by the sensor circuit and energy consumed for data acquisition; when the size of data acquired by the sensor was big, the energy consumed by the sensor for data transfer was decisive. | 2017-07-21T12:04:16.048Z | 2017-07-01T00:00:00.000 | {
"year": 2017,
"sha1": "7f7f8ee31ca5416d341085fcd19ebe3449923015",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1424-8220/17/7/1602/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "7f7f8ee31ca5416d341085fcd19ebe3449923015",
"s2fieldsofstudy": [
"Engineering",
"Medicine"
],
"extfieldsofstudy": [
"Computer Science",
"Engineering",
"Medicine"
]
} |
78088275 | pes2o/s2orc | v3-fos-license | Feasibility of Gastrointestinal Endoscopy Training in Surgery Residency at a Public Referal Hospital in Malawi
Objective: We set out to explore the impact of gastrointestinal (GI) endoscopy training on the surgery residency program at Kamuzu Central Hospital (KCH), Lilongwe, Malawi. Methods: We reviewed the hospital and published data regarding GI endoscopy and the surgical training at KCH from 2009 to 2015. The endoscopy database was reviewed determining endoscopic procedures done during the same period. Results: Since the onset of surgical residency program at KCH in 2009, 7 residents (5 now COSECSA Fellows) have been trained as independent GI endoscopists with support from the British Society of Gastroenterology andLiverpool Malawi Wellcome trust. Between them, with supervision, 1304 upper and 28 lower GI scopes have been done; 22% therapeutic. Endoscopy capacity has improved from an average of 12 patients per week to about 30 patients per week with a well-developed oesophageal variceal banding and tumour stenting capacity. Weekly endoscopy training lists by the fellows have also been established. During the same time period, 3 papers have been published from the endoscopy data with the residents as principle investigators or co-authors. Conclusion: With dedicated support endoscopy training during surgical residency is feasible in resource poor settings and improves diagnostic capacity for better patient care ultimately benefiting trainees in gaining endoscopy diagnostic and therapeutic skills, appreciating disease spectrum and management, and exposure to research skills.
hospitals in Malawi and the only one in the central region of the country.It has a bed capacity of 1200 and serving a population of approximately 6.3 million Malawians [1].In view of a high surgical disease burden at KCH and the country as Gastrointestinal (GI) conditions contribute a huge proportion of general surgical practice at KCH. Published data indicate that an average of 993 laparotomies per year due to GI related conditions were done from 2007 to 2010 at the institution [3].Additionally, looking at gastrointestinal endoscopy (GE) practice at KCH, Lindsay Wolf has described the spectrum of disease among patients with GI symptoms.Dysphagia (37%), haematemesis (21%), and epigastric pain (16%) are the most common indications for endoscopy whilst oesophageal cancer (27%) and oesophageal varices (17%) are the common diagnoses [4].This clearly means that the importance for GE at KCH does not need to be overemphasized.
As is the case with surgery, GE is a complex task that involves an interaction of cognitive and manual skills [5] Since the endoscopy training started at the institution, its impact on the STP and possible challenges have not been evaluated before hence our interest to carry out this study.
Place of Study
The study was conducted at KCH, Lilongwe, Malawi.
Data Collection and Analysis
We reviewed KCH GE and STP related published data and hospital records from 2009 to 2015.The data from published records was obtained from an exhaustive internet search for all articles relating to the training programs at KCH.The Hospital records reviewed included a Microsoft access GE database in the hospital's endoscopy unit and documents/records of trainee recruitment program in the department of Surgery.Published data in which residents were involved as authors was particularly noted.Data abstracted from hospital records included number of residents recruited in the STP and number of residents trained in GET.Here was no special criteria for inclusion of residents in the GE training as opportunity to train in GE was offered to all trainees in the STP Records regarding supportive and supervisory visits by members of the BSG and LMW were also reviewed.In addition, the KCH endoscopy unit data base was reviewed to determine the scope of endoscopic procedures done during the same period.
These procedures were categorized into diagnostic; lesion identification with/out biopsy and therapeutic procedures; with curative or palliative intent.
Ethical Consideration
In our study, no patient and surgical resident identifiers were used therefore ethical approval was not necessary.However, authorization to access hospital records including the endoscopy database was sought from the KCH hospital administration.
Results
Since the onset of the STP, 17 residents have been recruited in the program, with two residents training in orthopedics and 15 in general surgery.53% (n = 8) of the general surgery residents have been trained in GE (Table 1).
During the course of the endoscopy training program within the STP, supervisory and supportive visits by well qualified gastroenterologists were done (Table 2).These visits were supported by the British council, British society of gastroenterology (BSG), Mersey school of endoscopy (MSE) and the Malawi Liverpool welcome trust (MLW).
A total of 1304 upper and 28 lower GEs were done from 2012 to 2015 in which at least one surgery resident was involved.14 were colonoscopies and 14 sigmoidoscopies.The common findings are as shown in Figure 1.
Since onset of the STP and GET, three GE related papers have been published with residents as first authors or co-authors (Table 3).
Discussion
The future of surgical training in Africa depends on the availability of the leadership to recruit and attract trainees and trainers, adopt innovative educational technologies and growth of demand for quality surgical care [7].In Malawi, particularly KCH, the STP is young but has afforded to recruit and retain most of the trainees despite obvious resource capacity limitations (Table 1).Among other institutions in the country with STPs, GE training has been firmly established and incorporated in the STP at KCH only.Through links with the Mersey school of Endoscopy and the Liverpool Malawi Welcome trust and financial support from the British council and BSG, a number of supportive and supervisory visits to KCH endoscopy unit have been made since 2011.These visits included provision of basic skills course for those with no or very limited endoscopy experience, refresher/enhanced skills course for those with endoscopy experience (Table 2) [8].Milestones gained through the GE training are appreciable due to sustained skill impartation (Table 2), endoscopy equipment and supplies provision mainly from the STP sponsorship-University of Belgen, Norway, University of North Carolina (UNC), USA and BSG/Blackpool hospital in UK.
Surgical training programs elsewhere struggle with the integration of similar training into their curriculum given the constraints of time and trainee responsibilities and expectations [9].There is no consensus on the optimal way to teach endoscopy [5].Our model has seemed to produce conspicuous results that point towards feasibility of GE training incorporation in the training program (STP).Strictly speaking, the GE training is not mandatory at KCH nor it is in the COSECSA curriculum perhaps due to resource limitations.Given the benefits we have seen i.e. surgical trainees gaining diagnostic and therapeutic GE skills (Figure 1 and Figure 2), appreciating and managing endoscopically and probably surgically-those recommended for surgery, common GI pathologies seen in their practice, well designed GE trainings should be considered in every STP especially in Africa where demand for endoscopy is high and capacity for endoscopy is limited in many places [10].Resource constraints restrict widespread use of endoscopy both for diagnostic and therapeutic procedures in many African countries such that very few physicians in Africa have acquired skills in endoscopy [11].It is common knowledge now that COSECSA is producing more surgeons every year and could be a prime vehicle to promote GE training and service in the region and beyond.GE resource mobilization is crucial in that regard.3).This exemplifies the fact that given a well-established endoscopy practice in a surgery residency program, opportunities for research skills for trainees can be created.
Owing to the GE training program, the capacity for endoscopy at KCH has progressively improved (Figure 3).Obviously this eventually positively impacts on the surgical training and capacity at the institution.It is beyond the scope of this paper to ascertain accuracy in diagnostic capability among the trained residents and we recommend a well-designed prospective study to that effect.
a
whole and deficiency of well qualified surgeons, a surgery training program (STP) was established at the institution in July 2009 after accreditation by the College of Surgeons of East Southern and Central Africa (COSECSA) [2].Since its inception, 5 individuals have attained the fellowship qualification (4 general surgeons, 2 orthopaedic surgeons and 1 paediatric surgeon).
such that GE training may presumably be integrated in a STP without major challenges.Unfortunately in reality, time constraints during surgery residency hinder a smooth GE training program.In United Kingdom, high rates of dissatisfaction with endoscopy training have been observed among surgical trainees nationally.Reasons stated include no scheduled training lists, conflicting elective/emergency commitments and competition and absence of training lists [6].Before the residency program, one or two foreign surgeons were available for mainly diagnostic GE due to limited therapeutic resources.Two dedicated weekly endoscopy lists with a maximum of 12 patients per week were being done.Taking advantage of the STP, with support from the British Society of Gastroenterology (BSG) and Liverpool Malawi Well come Trust (LMW), a GE training program started in 2012 at KCH, with basic endoscopy skills training followed by skills enhancement and advanced (therapeutic) endoscopy skills at designated intervals.During all this period, surgery residents were under supervision from the experienced surgeons at KCH and regular visits by gastroenterologists from Black Pool and Glasgow in the UK, with personal resources and support from the BSG.
Figure 3 .
Figure 3. Progress of capacity for endoscopy at KCH.
With dedicated support gastrointestinal endoscopy training during surgical residency is feasible in resource poor settings and improves diagnostic capacity for better patient care ultimately benefiting trainees in gaining endoscopy diagnostic and therapeutic skills, appreciating disease spectrum and management, and exposure to research skills.
Table 1 .
STP and GE resident recruitment.
Table 2 .
Summary of GE training supportive visits | 2019-03-12T10:09:03.485Z | 2017-04-28T00:00:00.000 | {
"year": 2017,
"sha1": "557dbeefc47c3a159028990b9e839591fb93e59e",
"oa_license": "CCBY",
"oa_url": "http://www.scirp.org/journal/PaperDownload.aspx?paperID=75891",
"oa_status": "GOLD",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "557dbeefc47c3a159028990b9e839591fb93e59e",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
119393050 | pes2o/s2orc | v3-fos-license | X-ray observations of low-power radio galaxies from the B2 catalogue
We present an analysis of X-ray data, taken with ROSAT, for a well defined sample of low-power radio galaxies from the Bologna B2 catalogue. Where possible, the HRI has been used in order to take advantage of the higher spatial resolution provided by this instrument. A variety of models are fitted to radial profiles in order to separate the resolved and unresolved X-ray emission from the galaxies. We demonstrate a strong, approximately linear, correlation between the luminosities of the unresolved X-ray components and the 5-GHz luminosities of the radio cores in this sample. This suggests a physical relationship between the soft X-ray emission of radio galaxies and the jet-generated radio core emission. We infer a nuclear jet-related origin for at least some of the X-ray emission.
INTRODUCTION
It is widely accepted that the radio emission in BL Lac objects is dominated by synchrotron radiation from a relativistic jet pointed towards the observer, thus explaining, among other things, the superluminal velocities and rapid variability observed in several objects. Such a favourable orientation to the observer should not be common, implying the presence of a population of double radio sources whose twin jets lie in the plane of the sky. Low-power radio galaxies can be found which match BL Lac objects in extended radio power (Wardle, Moore & Angel 1984) and galaxy magnitude (Ulrich 1989). There is evidence that the total soft X-ray luminosity in such galaxies is correlated with the radio-core luminosity, implying a nuclear jet-related origin for at least some of the X-ray emission (Fabbiano et al. 1984). High spatial resolution X-ray measurements have further strengthened this argument by separating point-like emission from hot X-ray emitting atmospheres (Worrall & Birkinshaw 1994;Edge & Röttgering 1995;Feretti et al. 1995;Worrall 1997;Hardcastle & Worrall 1999). What has been lacking is highresolution X-ray observations of a large unbiased sample of low-power radio galaxies with which to investigate the association of unresolved X-ray emission with the nuclear radio jet.
The B2 bright sample of radio galaxies are Bologna Catalogue 408-MHz radio sources identified with elliptical galaxies brighter than m Zwicky =15.4 (Colla et al. 1975, Ulrich 1989. The radio survey occupies 0.84 steradian, and is complete at 408 MHz down to 0.2 Jy for (B1950) declinations between 29 and 34 degrees, 0.25 Jy for declinations between 24 and 29.5 degrees and between 34 and 40 degrees, and 0.5 Jy between 21.4 and 24 degrees (Colla et al. 1970(Colla et al. , 1975Ulrich 1989), although none of the bright sample radio galaxies lies in the last declination range. The sample comprises 50 galaxies, of which 47 are at z ≤ 0.065 (see Table 1). The high Galactic latitudes imply a relatively small Galactic neutral hydrogen column density and small resultant X-ray absorption. The B2 sample has been shown to be well matched with radio-selected BL Lac objects in their extended radio properties and galaxy magnitudes (Ulrich 1989). In this paper we present ROSAT X-ray measurements of 40 galaxies which constitute an unbiased subsample of the 47 galaxies at z ≤ 0.065. The data were taken from our pointed observations or from data in the ROSAT public archives.
Section 2 discusses the sample of galaxies observed with ROSAT and the general properties of the X-ray data. Details of our analysis and notes on some of the sources are in section 3. Section 4 describes X-ray -radio comparisons. Section 5 contains our conclusions.
A Friedmann cosmological model with H0 = 50 km s −1 Mpc −1 , q0 = 0 is used throughout this paper. c 1999 RAS 2 X-RAY DATA Table 1 lists the sample of 50 B2 radio sources associated with elliptical galaxies; we have not included B2 1101+38 and B2 1652+39 as these are the well known BL Lac objects Mkn 421 and Mkn 501. We tabulate the pointed observations taken with ROSAT using the instrument with the highest spatial resolution, the High Resolution Imager (HRI; David et al. 1997), or, if no HRI observations were made, the Position Sensitive Proportional Counter (PSPC;Trümper 1983;Pfeffermann et al. 1987). Where multiple observations exist, we present results for the longest on-axis viewings, concentrating where possible on HRI data. Most observations result in a detection. In some cases where good data exist from the HRI and PSPC, both were analysed and compared (this was the case for B2 0055+30, 0120+33, 0149+35, 1122+39; 1217+29 and 2335+26). Within the same ROR, some HRI observations were split into observing periods about 6 months or a year apart. Where this was the case, the data have been merged after the individual 'observations' were checked for anomalies. In many cases resolved X-ray emission (measured better with the PSPC) is seen in addition to point-like emission. Complementary work discussing the extended X-ray emission and X-ray spectra of B2 radio galaxies based on PSPC observations can be found in Worrall & Birkinshaw (in preparation). Our treatment of the resolved emission in this paper is restricted to modelling it sufficiently well to determine a best estimate for the contribution from central unresolved emission.
40 of the full sample of 50 B2 radio galaxies (and the 47 at z <0.065) listed in Table 1 were observed with ROSAT : the demise of the satellite in early 1999 prevented observation of the remaining 7 objects with z <0.065. No known bias was introduced into the set of 40 objects for which we have data by the process of prioritizing sources for ROSAT observation. This is illustrated in Fig. 5, which shows histograms of 1.4-GHz extended radio power (Pext), redshift, and absolute visual magnitude (Mv) for the 47 galaxies at z ≤ 0.065. Pext and Mv are indicators of isotropic unbeamed emission used to support the association of these galaxies with the hosts of BL Lac objects. A Kolmogorov-Smirnov test finds no significant (>90 per cent confidence) difference between the distributions of sources for which we have ROSAT data, and those for which we do not, in any of the quantities Pext, z , or Mv.
The ROSAT HRI has a roughly square field of view with 38 arcmin on a side. A functional form for the azimuthally averaged point spread function (PSF) can be found in David et al. (1997). The core of the HRI radial profile of a point-like source may be wider than the nominal PSF, and ellipsoidal images are sometimes seen, due to a blurring attributable to residual errors in the aspect correction. The major axes of such ellipsoidal images are not aligned with the satellite wobble direction (the wobble is employed to ensure that no sources are imaged only in hot pixels in the HRI or hidden behind the PSPC window-support structure) and depend unpredictably on the day of observation and therefore the satellite roll angle. The asymmetry is strongest between 5 arcsec and 10 arcsec from the centroid of the image. The PSF for the HRI begins to be noticeably influenced by the off-axis blur of the X-ray telescope at > ∼ 7 arcmin off-axis. All HRI observations discussed here are essentially on axis.
The ROSAT PSPC has a circular field of view, of diameter 2 degrees, and the PSF has a FWHM of about 30 arcsec at the center of the field, degrading only marginally out to off-axis angles of about 20 arcmin. At larger radii the mirror blur dominates the spatial resolution, and at 40 arcmin off axis the FWHM of the PSF is about 100 arcsec. Only one of the 40 sources discussed here (B2 0722+30) is significantly affected by mirror blur, as all others were observed in the central part of the field of view.
We used the Post Reduction Off-Line Software (PROS; Worrall et al. 1992) to generate radial profiles of the X-ray data. Background was taken from a source-centered annulus of radii given in Table 3. Where the radial profile has been used to probe the extended emission, the contribution from extended-emission models to the background region is taken into account (Worrall & Birkinshaw 1994).
We excluded confusing sources, defined as those separated from the target by > ∼ 15 arcsec (HRI) and > ∼ 30 arcsec (PSPC), showing up at ∼ 3σ above the background level and overlapping the on-source or background regions (or being slightly beyond, but still affecting the background due to the broad wings of the PSF). Optical images (e.g the Palomar Sky Survey plates, digitized by the Space Telescope Science Institute) and radio maps were overlaid on the Xray images, to check the identity of each target source and any neighbouring X-ray sources, and to help to classify and limit further the effects of confusing sources.
Our luminosity determination for unresolved emission assumes a power-law spectrum with an energy index α of 0.8 (fν ∝ ν −α ) modified by Galactic absorption. Variations in spectral form affect the luminosity, but this is normally a small error compared with statistical uncertainties.
RADIAL PROFILES AND NUCLEAR X-RAY EMISSION
We analyse the structure of an X-ray source by extracting a radial profile and fitting various models convolved with the energy-weighted PSF of the detector in question [the nominal PSF for the HRI (David et al. 1997) and the PSF for the PSPC (Belloni, Hasinger & Izzo 1994)].
As well as point-source models, we have fitted our radial profiles with β models (Sarazin 1986) Table 2. In most cases we find that for observations with enough counts a significantly better fit to the data is achieved by fitting the composite model (c) rather than either (a) or (b) individually (see Table 2).
Residual errors in aspect correction affect the HRI PSF and must be taken into account when evaluating the contribution from a point source [see Worrall et al. (1999), where it was found that a point source can appear more like a β model with a core radius of up to about 5 arcsec (dependent on β assumed) because of the aspect smearing]. Of the sources in this paper only 1833+32 has high enough count rate to perform the dewobbling procedure described by Harris et al. (1998); we find no substantial difference in the best-fit parameters after dewobbling. For observations where the data show extension on small scales, it is difficult to decide whether the source is really point like or indeed slightly extended. In some cases, the fit to the point+β model has a substantially lower χ 2 than the beta-model fit alone, and here we regard the point-like component as well measured. Where χ 2 remains unchanged for these two models, the total counts have been taken as an upper limit on the point-source counts. A literature search into the environments of the sources and a cross-correlation of the HRI and PSPC results, where possible, helps us further validate our HRI findings and place limits on the likelihood of the source being primarily point-like.
In the 7 cases where there are not enough counts to perform adequate radial-profile fitting , the total counts are taken as an upper limit to the contribution of point-like emission. For non-detections a 3σ upper limit, derived by applying Poisson statistics to a 5 by 5 arcsec detection cell for the HRI on-axis observations (3 cases), and a 120 by 120 arcsec detection cell for the PSPC observation of the off-axis source B2 0722+30, centred on the position of the radio and optical core, is taken as an upper limit on both the total and the unresolved X-ray emission. Table 2 gives results for the 16 sources with enough counts to allow radial-profile model fitting. Table 3 presents the net counts within a circle of specified radius and, for the sources with enough counts to allow radial profiling (consisting of a minimum of ∼70 counts over 3 data bins), the point-source contribution. There is a wide range in the ratio of unresolved to resolved counts. X-ray core flux and luminosity densities calculated from the unresolved count rates, and radio core flux and luminosity densities taken from the literature are given in Table 4.
Notes on individual sources
Where analysis of sample sources is present in the literature, comparisons have been made with the results presented here. Results for B2 0055+26 are taken from Worrall, Birkinshaw & Cameron (1995). B2 0326+39, 1040+31 and 1855+37 are discussed in detail in Worrall & Birkinshaw (in preparation). The HRI results for B2 2229+39 are consistent with the findings of Hardcastle, Worrall & Birkinshaw (1998), for a PSPC observation of the source.
B2 0120+33
Identified with the galaxy NGC 507, the source lies in Zwicky cluster 0107+3212 and is one of the brightest galaxies in a very dense region. Extended X-ray emission is seen out to a radius of at least 16 arcmin, and there is evidence for the presence of a cooling flow and possible undetected cooling clumps distributed at large radii (Kim & Fabbiano 1995). B2 0120+33 has a steep radio spectrum and weak core. It may be a source with particularly weak jets, or possibly a remnant of a radio galaxy whose nuclear engine is almost inactive and whose luminosity has decreased due to synchrotron or adiabatic losses ).
The HRI map shows the central region of this source to be asymmetrical, with a large extended emission region to the SW. Radial-profile fitting of the innermost parts of this galaxy with point, β, and point+β models, show that a good fit to the data is achieved by using a single β model with β=0.67, which gives a core radius of 4 ′′ . The β+point model gives a marginally better fit, but the additional component is not significant on an F-test at the 90% confidence level. Therefore, we have taken the total counts from the inner regions of this source as an upper limit on any unresolved emission present, keeping in mind that this may also include a cooling-flow contribution (this also holds for other sources with possible unresolved cooling flows such as B2 0149+35, 1346+26 and 1626+39). B2 0149+35 B2 0149+35 is identified with NGC 708 and is associated with the brightest galaxy in the cluster Abell 262. Braine & Dupraz (1994) suggest that it contains a cooling flow which may contribute excess central X-rays, and this may explain why B2 0149+35 has a higher point-like X-ray luminosity and flux than expected based on other sample members. It is not possible to separate spatially a cooling-flow contribution from unresolved X-ray emission using the PSPC observation, and the asymmetry of the source makes the extraction of a radial profile difficult. The HRI observation is split up into 2 OBIs (Observation Intervals), one of which shows a barred N-S structure. Each OBI was individually analysed by taking close-in source regions, and this gives results which are consistent with the PSPC data from the inner region of B2 0149+35. In the longer, and more reliable, of the two OBIs (12.7ks), 207 counts were detected. The point-like contribution to the net emission from this source however, is not significant at the 95 per cent level when an F-test is performed. The detected counts are therefore taken as an upper limit on the point model emission. The shorter OBI was not used. B2 0207+38 This source is described as being more similar to an S0 or to a spiral galaxy than an elliptical (Parma et al. 1986). It has also been called a post-eruptive Sa (Zwicky et al., 1968) or a distorted Sa (de Vaucouleurs, de Vaucouleurs & Corwin 1975). The radio structure is disc-like and there is no sign of either a radio core, or of jets or radio lobes (Parma et al. 1986). It is probably a starburst, like B2 1318+34. There are no ROSAT X-ray data for this source. B2 0836+29A This object (4C 29.30), this object has been often confused in the literature with the cD galaxy B2 0836+29 at z =0.079, which is the brightest galaxy in Abell 690. B2 0924+30 B2 0924+30 appears to be a remnant radio galaxy whose nuclear engine is inactive Cordey 1987;Giovannini et al. 1988). It is the brightest member in a Zwicky cluster (Ekers et al. 1981), and the X-ray data suggest extended X-ray emission, although the detection is of marginal significance. The relatively high X-ray emission for a source with no detectable core radio emission may therefore be due to the extended gas in the cluster. We have taken the detected emission as an upper limit on possible point source emission.
B2 1122+39
Analysis of this source, both in the PSPC and in the HRI, shows that ∼3 per cent of the total emission is contributed by an unresolved source. This is consistent with the findings of Massaglia et al. (1996) who find a contribution from a point source of <6 per cent. B2 1217+29 PSPC and HRI analysis of this source are consistent. A β+point model fits the data better than a β model alone (see Table 2).
B2 1254+27
There is a large discrepancy between the positions given for this object in the NASA Extragalactic Database (NED) and the SIMBAD Astronomical Database. This is because the radio source has, in some cases, been incorrectly associated with the galaxy NGC 4819 rather than the true host galaxy for the radio emission which is NGC 4839.
NGC 4839 is classified confusingly as morphological type S0 (Eskridge & Pooge 1991), E/S0 (Jorgensen, Franx & Kjaegaard 1992) or as a cD (Gonzalez-Serrano, Carballo & Perez-Fournon 1993;Fisher, Illingworth & Franx 1995). Andreon et al. (1996) also mention that the low average surface brightness suggests that this galaxy is dominated by an extended disc. The X-ray map of the source shows extended large scale emission to the SW [described by Dow & White (1995), as being in the process of interacting with the intracluster medium of the main (Coma) cluster]. This goes beyond the size of the optical galaxy and has been excluded here so as not to affect the background emission. About 88 per cent of the net counts arise from a point-like emission component.
B2 1257+28
The region of enhanced X-ray emission in B2 1257+28 in the Coma cluster is substantially smaller than the size of the optical galaxy. Small on-source (12 arcsec radius) and background (15-22.5 arcsec) source-centered circles were used in order to verify the contribution from unresolved X-ray emission given by our best-fit model. B2 1317+33 B2 1317+33 (NGC 5098A) has a companion galaxy (NGC 5098B) at a distance of ∼ 40 arcsec. We have checked that the X-ray and radio source come from NGC 5098A by overlaying the radio, optical and X-ray maps. B2 1318+34 B2 1318+34 is a classic merger-induced starburst, whose total radio flux can be attributed to starburst activity rather than an active nucleus (Condon, Huang & Yin 1991).
B2 1346+26
The source is a cD galaxy in Abell 1795, identified with 4C 26.42. It contains a central cooling-flow component, as discussed in Fabian et al. (1994). HST WFPC2 images of the core of this cooling flow are presented in Pinkney et al. (1996).
Our analysis of the HRI data detects a central point source which is significant on an F-test at the 95 per cent confidence level (see Table 2). About 1 per cent of the total counts lie in this point-like component (see Table 3). B2 1422+26 B2 1422+26 is not radially symmetric in the X-ray. An offaxis X-ray source in the same field of view gives a good fit to the nominal PSF, and so we can rule out the possibility that the X-ray extension seen in B2 1422+26 is due to the ROSAT aspect correction problem. The possible detection of a point-like component is not significant on an F-test at the 95 per cent level (although it passes at the 90 per cent level). We have nevertheless taken the point counts calculated from these model fits as our best estimate of the central emission, though the errors are large. B2 1615+35 HRI analysis is consistent with Feretti et al. (1995). The X-ray emission is largely point like (see Table 3), with ∼60 per cent of the net counts coming from the point source. B2 1621+38 B2 1621+38 was analysed by Feretti et al. (1995), who found a point-source contribution of ≤50 per cent of the total Xray flux. Our analysis is consistent with this result; we find the point-source contribution to be 18±4 per cent. B2 1626+39 B2 1626+39 lies in a cluster (A2199), with a prototypical cooling flow. Owen & Eilek (1998) conclude that the radio source is relatively young and has been disrupted by the surrounding gas. The ROSAT HRI data set for this source consists of 2 OBIs roughly 7 months apart. Only in the second observation does the source appear extended, with two adjacent peaks. We have taken this to be due to errors in the aspect correction or processing effects and therefore have used only the first OBI in our analysis. B2 1833+32 B2 1833+32 is an FRII radio galaxy (Laing, Riley & Longair 1983;Black et al. 1992) with broad emission lines (Osterbrock, Koski & Phillips 1975;Tadhunter, Perez & Fosbury 1986;Kaastra, Kunieda & Awaki 1991). Its higher than expected X-ray flux, as compared with the core radio strength, may arise from emission in the central accretion disc around the active nucleus, seen due to an advantageous viewing angle as indicated by the broad emission lines.
B2 2236+35
This source has a double symmetric radio jet embedded in a low surface-brightness region. The two extended lobes are similar in strength and size (Morganti et al. 1987). The X-ray emission at radii greater than about 20 arcsec seems aligned with the radio jets in this source. Model fitting shows ∼25 per cent of the total counts to be in the point source.
X-RAY -RADIO COMPARISONS
In Table 3 we list the net counts detected for each source within a specified radius and our best estimate of the contribution from unresolved emission derived as described above. 1-keV luminosity densities and broad-band soft X-ray luminosities calculated from the values for unresolved emission are given in Table 4, along with the radio core flux density and luminosity density. In Fig. 2 we show a logarithmic plot of radio core luminosity density against the unresolved core X-ray emission; the corresponding flux density-flux density plot appears in Fig. 3.
Both the logarithmic X-ray and radio flux densities and the corresponding luminosities are correlated at the > 99.99 per cent significance level on a modified Kendall's τtest which takes upper limits into account, as implemented in ASURV (Lavalley, Isobe & Feigelson 1992). The fluxflux correlation gives us confidence that the luminosityluminosity relationship is not an artificially-introduced redshift effect.
To determine the slope of the core flux-flux and luminosity-luminosity plots, a generalised version of the Theil-Sen estimator was used as presented in Akritas, Murphy & LaValley (1995). This takes into account the nature of the upper limits by assuming that the individual points are all part of the same parent population. For more details of this analysis and its advantages over the more commonly used survival-analysis method, see Hardcastle & Worrall (1999), where it is explained how using the bisector of two regression lines provides a more robust estimate of the slope. This method however does not give a value for the intercept of the best-fit line. To determine the best-fit line plotted on Fig. 2, a regression based on the bisector of slopes determined by the Schmitt algorithm as implemented in ASURV was used, because it does allow us to determine an intercept.
For the whole sample, the Theil-Sen luminosityluminosity slope is 1.05 with 90 per cent confidence limits 0.86-1.28. There are however, a few galaxies that should be removed. These are the starburst object B2 1318+34 (which is not member of a strict AGN sample), and the FRII B2 1833+32 (which is a broad-line galaxy). For comparison purposes we have retained these objects on Figs. 2 & 3. With the omission of these two sources, we find a Theil-Sen slope of 0.96 for the luminosity-luminosity relation, with a 90 per cent confidence range (derived from simulation) of 0.78 -1.21. The median logarithmic dispersion about the regression line is ∼ 0.2. Schmitt regression analysis of the slope gives consistent results, with a slope of 1.15 and a 90 per cent confidence limit of 0.92 -1.39. Normalisation of the Schmitt slope at a radio luminosity of 10 22 W Hz −1 sr −1 , gives the best-fit normalisation value of 15.55 (i.e. the predicted Xray flux at that radio luminosity is 10 15.55 W Hz −1 sr −1 . The 90 per cent confidence range on this is 15.2 to 15.9. From combining statistical methods, the best overall estimate and 90 per cent confidence uncertainty for the core X-ray-radio luminosity relation for low-power radio galaxies is given by: log(lx) = (0.96 +0.25 −0.18 ) log (lr/10 22 ) + (15.55 ± 0.35) A couple of sources lie significantly away from the regression line (B2 1254+27 and B2 1257+28). Both lie in the Coma cluster and may be contaminated by cluster emission.
B2 0120+33 has been classified as a possible remnant of a radio galaxy whose nuclear engine is inactive , and is shown as an upper limit above the expected value based on the X-ray correlation for other sample members. Its high total X-ray flux may be due to the cooling flow seen by Kim & Fabbiano (1995) (see Section 3 of this paper) and makes the extraction of the core flux difficult. B2 0924+30 is also a relic source. B2 0149+35 contains a cooling flow which may contribute excess central X-rays (Braine & Dupraz 1994).
The correlation shown in Figs. 2 & 3 suggests a physical relationship between the soft X-ray emission of radio galaxies and the jet-generated radio core emission. Correlations between the total X-ray emission and the radio core emission have been seen in the Einstein Observatory data (e.g Fabbiano et al. 1984), but did not have the spatial resolution necessary to separate point and extended components. Our ROSAT analysis and the decomposition of the X-ray emission into resolved and unresolved components, now shows that the nuclear X-ray emission is strongly correlated with the nuclear radio-core emission. This favours models which imply a nuclear jet-related origin for at least some of the X-ray emission.
CONCLUSIONS
Radial profiling and model fitting of ROSAT data, primarily from the HRI, have allowed us to separate point-like contributions from the overall X-ray emission in low power radio galaxies from a well defined, nearby sample. We find fluxflux and luminosity-luminosity core X-ray/radio correlations for such sources, with slopes that are consistent with unity. This suggests a physical relationship between the soft X-ray emission of radio galaxies and the jet-generated radio core emission, with the clear implication that at least some of the X-ray emission is related to the nuclear radio jet. In future work we will estimate X-ray beaming parameters under the assumption that radio galaxies are the parent population of BL Lac objects.
ACKNOWLEDGMENTS
This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. Support from NASA grant NAG 5-1882 is gratefully acknowledged. We thank the referee for useful comments. (Ulrich 1989). Solid shading shows the 40 sources for which we have ROSAT X-ray data. Table 1. Data on the 50 B2 sample galaxies. RA and DEC are given for the radio core where possible, and the optical identification otherwise (optical positions are labelled with a *). References are as follows: 1, Fanti et al. (1987), 2, Worrall et al. (1995); 3, Ekers (1978); 4, Fanti et al. (1982); 5, Bridle et al. (1991); 6, Measured from radio map in Laing (1996); 7, Johnston et al. (1995), 8, Measured from radio maps supplied by L. Rudnick; 9, Venturi et al. (1995); 10, Stocke & Burns (1987); 11, Fanti et al. (1973);12, Schillizzi et al. (1983), 13, Kim (1994); 14, Bridle et al. (1981); 15, Giovannini et al. (1988); 16, Estimated from map of Venturi et al. (1993). Galactic absorbing column density N H is taken from Stark et al. (1992). Redshifts are from Colla et al. (1975). Mv is the absolute visual magnitude of the host galaxy from Ulrich (1989). The ROR is the ROSAT observational request number, with 'rh' referring to HRI observations and 'rp' to PSPC observations. Offset refers to the off-axis angle, and is shown only for those sources with offsets ≥6 arcmin. A dash in the last three columns indicates no observation. Table 3. Net counts within the given source radius and the estimated contribution from unresolved emission (point counts). Sources marked with an asterisk lie in cluster environments, so the net counts are more dependent on the radii chosen for source and background regions than for field galaxies. Details of any foreign sources excluded from these regions are not shown. Consult Table 1 <10 Table 4. Radio and X-ray core flux densities and luminosities. For the radio core, references are as follows: 1, Giovannini et. al (1990); 2, Fanti et al. (1987); 3, Venturi et al. (1993); 4, Fomalont, private communication; 5, various, see text; 6, measured from maps supplied by R. Morganti. A source of type U is undetected in the X-ray, and the X-ray measurement is a 3σ upper limit. In a source of type P a compact source is detected in the X-ray, and all the X-ray counts are attributed to the X-ray core. In type R, a mixture of extended and compact X-ray emission is found. The X-ray core value is from the best-fit point-source contribution to a multi-component model. Sources of type L are upper limits corresponding to the total detected counts on a source, either because there are too few counts to perform an adequate radial profile fit or because the source has complicated structure which is not well characterised by our models. L 0.2−2.5keV point is the estimated luminosity of the X-ray point-like component between 0.2 and 2.5 keV. The X-ray values assume a power-law spectrum with energy index 0.8, and errors in the 1 keV luminosity density are statistical only.
B2
S Figure 2. X-ray core luminosity is plotted vs radio core luminosity. The line drawn is derived from the Schmitt regression [(excluding a broad-line radio galaxy and a starburst galaxy (see text)] and has a slope of 1.15 and an intercept of -9.82. Cluster sources are shown by squashed squares, an open circle denotes a relic source, a cross represents a starburst galaxy, and a closed star is everything else. The excluded broad line FRII source 3C382 has been displayed as a diamond; it is over-bright in X-rays for its core-radio strength, suggesting an additional X-ray emission component, consistent with unified models. The arrows show where upper limits have been taken and in all other cases 1σ error bars are shown. Figure 3. X-ray against radio flux density. Symbols as for Fig. 2. The correlation persists here, and is significant at the >99.99 level. | 2019-04-14T01:53:56.008Z | 1999-07-01T00:00:00.000 | {
"year": 1999,
"sha1": "7e40cad999934079aa673582b2f25d5237b1a644",
"oa_license": null,
"oa_url": "https://academic.oup.com/mnras/article-pdf/310/1/30/2946037/310-1-30.pdf",
"oa_status": "BRONZE",
"pdf_src": "Arxiv",
"pdf_hash": "e534f89e10eda8b29c91d6fec53ed6dbbb5f5639",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
8365651 | pes2o/s2orc | v3-fos-license | How can students contribute? A qualitative study of active student involvement in development of technological learning material for clinical skills training
Background Policy initiatives and an increasing amount of the literature within higher education both call for students to become more involved in creating their own learning. However, there is a lack of studies in undergraduate nursing education that actively involve students in developing such learning material with descriptions of the students’ roles in these interactive processes. Method Explorative qualitative study, using data from focus group interviews, field notes and student notes. The data has been subjected to qualitative content analysis. Results Active student involvement through an iterative process identified five different learning needs that are especially important to the students: clarification of learning expectations, help to recognize the bigger picture, stimulation of interaction, creation of structure, and receiving context- specific content. Conclusion The iterative process involvement of students during the development of new technological learning material will enhance the identification of important learning needs for students. The use of student and teacher knowledge through an adapted co-design process is the most optimal level of that involvement.
Background
Clinical skills training is a fundamental part of nursing education wherein students combine sensory, motor and cognitive learning processes and learn how to perceive and act in any situation presented to them [1]. This complexity of clinical skills acquisition demands a range of different learning approaches for nursing students to learn what they need to know [2]. A shift toward more learner active teaching strategies in higher education [3] and an expanding knowledge of information technology [4] has produced many changes in clinical skills training over the last few years. This change has produced multiple new learning strategies, such as simulation, serious games, online learning material, and personal digital assistants, which have emerged and become part of nursing student clinical skills training [5][6][7][8][9]. Nevertheless, the quest to determine the most optimal learning method within clinical skills is still being sought by many nurse educators [10].
Further, new policy initiatives and an increasing amount of the literature within higher education call for students, not only to be consulted during the development of learning strategies, but also become actually involved as co-designers, co-producers, and co-creators of their own learning [11,12]. The goal is to place student needs at the center of the design process [13] and thus view the student as a knowledgeable and critical partner in learning [14]. While the idea of user involvment already is an established best practice within health care services [15][16][17] nursing education has only to some extent actually embraced this student collaboration concept [18,19]. Student experiences have, however, been deemed valuable for future educational improvement [20] and student involvement has been used in some curriculum design [21,22]. There is also some comprehensive literature on student use, benefits, barriers, and their experiences with already developed programs and devices [23][24][25]. On the other hand, there is a shortage of literature on active involvement of nursing students in the actual development processes and especially a lack of descriptive studies that examine the actual personal role of the students when they are engaged in the creation of their own learning activities [26]. In Norway, undergraduate nursing education follows the Bologna requirements with 3 years of full-time study resulting in a bachlor degree [27]. Student involvement is ensured through law [28] where the minimum requirement is yearly student evaluation of the educational programme provided by the institution. The Ministry of Education also requires the educational institutions to gear their educational approach to the 'active, participating student', through a White Paper submited to the Norwegian Parliament [29]. While these official documents have ensured some participation, the room for individual interpretation of its execution often results in the use of representatives rather than participatory or prefigurative forms [30].
Aim of the study
The aim of this study was to explore and describe the actual process of student involvement when developing technological learning material for clinical skills training in a Norwegian nursing faculty. Two research questions were developed for this purpose: -How can nursing faculties actively involve their nursing students in the process of developing technological learning material? -How can both students' roles and contributions in the development process of such technological learning material be best described?
Design
The study was grounded in the idea of user involvement and the methodology of participatory design (PD). PD builds on the line of reasoning that key to finding the gaps that matters lies in involving the end users in development and design of services [31]. The process entails actively involving a group of people and bringing them to a consensus on what they want to do and how best to do it. In order to meet the actual needs of the users, their involvement must be incorporated into both design and development [31]. Through this process, PD has the potential of increasing the ease of implementation and of creating the benefits of credibility and legitimacy, while ensuring that the final design truly meets the precise needs of its users [19]. The approach has been specially suggested for use within educational settings due to its ability to take student perspectives into account [11]. While similar approaches such as experience-based co-design (EBCD) offers a series of stages to follow [32,33] PD does not entail an specific description of how to involve the end users in the development process, but rather focuses on the involvement itself. In this study, the methods of data collection therefore needed to both actively and creatively engage the students in the developmental process, while giving the researchers the opportunity to grasp the students' perspectives throughout the developmental process. An explorative qualitative approach was chosen as appropriate for arriving at an in-depth understanding of human behavior, by giving the participants room and opportunity to describe and explain their own experiences [34]. The development process was elaborated by the authors of the paper and divided into five phases; (1) initial phase, (2) Investigation phase, (3) revision phase, (4) exploratory test phase, and (5) finalization phase. The students contributed to different activities and to the collection of different data throughout the development process. An overview of activities and data collection is found in Table 1.
Contextual setting
The technological learning material was applied to the clinical skills course at a Norwegian nursing faculty to teach undergraduate nursing students the 13 clinical skills required to pass that course. The course the technological learning material was applied to was based on a combination of supervised and unsupervised practice sessions. There were nine different supervised training sessions wherein a teacher-led group of 10-12 students practiced the 13 different scenarios. In addition, the students were given unlimited access to the Clinical Skills Laboratory (CSL) at the campus where they were expected to administer their own unsupervised training sessions. At the end of the course, all students were tested in one of the 13, randomly chosen skills in practical oral examination. For details of the course and the CSL environment, see C Haraldseid, F Friberg and K Aase [35]. Portable Sim-Pad® tablets were used as technological mediators of the offered learning material. The main features of the tablet was; preprogramming correct actions that could be taken, feedback on actions taken, and to linking actions to responses. The user was thereby guided through a scenario, which could develop in multiple ways, as different actions might result in different outcomes. The software also gave the user a log of their actions at the end of each scenario and the programmer had the opportunity to add log comments, give the instructor instant messages, or set time limits for when actions needed to take place. By pre-programming the tablets, the students were able to run the required scenarios on their own. Prior to involving the student in the developmental process actively, four prototype scenarios were developed by a teacher team to exemplify for the students how the features of the tablet could be used. To demonstrate to the students what they were asked to do, all students (165) enrolled in the clinical skills course were given a 1-hour introductory instruction lecture on how to operate their tablets, including the possibility of testing the device in groups. The prototype scenarios were also made available for use during two compulsory, supervised training sessions where the students had the opportunity to access their tablets during unsupervised training sessions to test the scenarios and become comfortable with their use. After the introduction, the students were involved in different phases and in different activities as shown in Table 1.
Study participants
The study was undertaken at a Norwegian nursing faculty during Fall 2013 and Spring 2014 terms. In the Initial phase, all students enrolled in the clinical skills course were informed of the ongoing project and had the opportunity to familiarize themselves with the tablets and use of them as desired and when and how they wanted. The students participating in the Investigation and Exploratory Test phase were recruited during the initial phase through purposive sampling among all 165 students. This recruitment was done after the students completed their clinical skills course. Due to their participation in the course these students would have important experiences of their needs and the challenges that would present during clinical skills acquisition, together with first -hand user information on how the prototype of the learning material used in the course could be improved. The students were recruited by the first author through an open invitation in one of the faculty lecture classes. All students wishing to participate were encouraged to approach the first author personally or via an e-mail after class. There were no prerequisite for how much the students used the prototype of the learning material during the course, as those without excessive experience with the tablets could also contribute with important experiences leading to improvements. In total, 19 students contributed to four focus group interviews and two practical training sessions. In their own reporting, five of these 19 students stated they had used the tablet 'a little', six had used it 'some', and eight reported they had used it 'a lot'. During the focus groups in the Investigation Phase the 19 students were divided into two groups with eight students in Group A and 11 in Group B. The division into the groups were based on the students' schedules and their convenience. In the Exploratory Test Phase, 11 out of the original 19 students who participated did so based on availability with five from Group A and six from Group B. These 11 students were then divided into Groups C and D (see Table 1).
During the Revision and the Finalization phases, the first author organized meetings and conducted the process of making changes to the learning material. A clinical nurse specialist from the hospital contributed as a direct result of the students' feedback, and two faculty teachers were consulted to make sure the current alterations matched best practice guidelines and required course content. A senior interaction designer was consulted on how to integrate the students' feedback to the technological choices available on the tablet set-up.
Ethical considerations
During the Initial phase of the study the students were given oral information about the ongoing project, confirming that participation was voluntary, which is in line with the principles outlined in the Declaration of Helsinki (World Medical Association Declaration of Helsinki, 2005). The students who proceeded to participate in the Investigation Phase and the Exploratory Test Phase received both written and oral information on the background and goal of the study, including information about their right to withdraw from the study at any point during it. Written informed consent was collected prior to data collection in both the Investigation Phase and the Exploratory Test Phase. Since the current research study involved no medical interventions or collection of health related information, the approval authority is the Norwegian Social Science Data Services (NSD) who also assess the ethical aspects of recruitment and informed consent. Approval for the study was therefore obtained from the NSD (Reference Number. 36260) and from the head of the nursing faculty.
Data collection
The data in the study was collected through field notes and focus group interviews. Field notes were collected through informal meetings between the first author and the students, for example, during the delivery and return of the tablets or if any students approached the first author with comments about tablet use. In addition, notes were taken at all revision meetings, and the students took their own notes during the practical test session. All focus group interviews were conducted in a meeting room on campus. Interviews with Groups A and B were conducted 6 weeks after these students completed the clinical skills course, while the interviews with Groups C and D were conducted 9 weeks after their completion of the course. All interviews were moderated by the first author and assisted by the third author. Interaction between the students was encouraged with the moderator asking, prompting, and clarifying questions. Interviews were audio recorded while both the moderator and assistant moderator wrote field notes to complement the audiotaping. The Investigation Phase and the Finalization Phase had their own separate aims and interview guidelines, respectively.
In the Investigation Phase, the focus group interviews [36] lasted for 60-80 min. The goal was to explore the students' requirements during unsupervised training and how the technological learning material could contribute to fulfilling their learning needs. The interviews commenced with general questions about the students experiences in CSL training. Once the students seemed comfortable with the interviewers, the questions gradually turned to the theme of the study [36]. Those questions pertained to the issues the students enjoyed or found difficult in the CSL environment, their needs, and how their training could be improved.
The Exploratory Test Phase consisted of both practical training sessions and focus group interviews. The training session lasted for 45-60 min. In that session the students received a revised version of the technological learning material, based on the needs and feedback gathered during the Investigation Phase. They were given all the necessary equipment to complete the training session. The students were divided in groups of two or three and instructed to test the device as it suited them, but they had to complete the entire scenario. They were encouraged to take breaks in the scenario and discuss the process with each other, while taking notes of what they had experienced, felt and thought. Immediately after the practical training session, the groups were gathered for joint discussion in focus group interviews. The focus group lasted for approximately 30 min. It attended to different aspects of the learning material, in particular, the layout, the content, and areas that needed improvement and ways to undertake such improvement. In addition, the students handed in their personal notes from the practical test session for use as supplementary data material.
Data analysis
The main topic for analysis was the focus group interviews, while the field notes and student notes were used as supplementary data material. Qualitative content analysis was chosen as the method to analyze and categorize data [37]. All interviews were transcribed by the first author 1 or 2 days after the interviews. The transcripts and field notes were also analyzed and coded by the first author. In the first step, the data was read as openly as possible, trying to get an impression of both parts and the whole. In the second step, after reducing the number of words while still preserving the content, the meaning units were shortened and coded. This step compared the units and sorted the text into relevant themes [37]. As the authors reviewed and discussed these themes, it became clear that several themes were overlapping, so some themes were merged at a more abstract level in Step 3.
Step 4 consisted of reading the field notes and interview transcriptions again, making sure the final themes covered the whole picture. During this step it became clear that the themes represented five different learning needs that were especially important for the students: clarification of learning expectations, help to recognize the bigger picture, stimulation of interaction, creation of structure, and receiving context-specific content. To establish trustworthiness throughout the entire study, the first and third authors conducted the interviews and took all the field notes, while the second author formulated the critical questions needed to expand the understanding of the gathered data [37,38]. Different interpretations found during the analytical steps were repeatedly discussed and reinterpreted by all authors together.
Results
Through a process of actively involving nursing students in the development of technological learning material, their role evolved into being advocates for learning needs that are necessary for tailoring their learning material accordingly. While the nursing faculty staff may hold the key to what students should learn, the students described how their learning could be most constructively achieved. These learning needs were not initially explicitly described, but rather evolved over time as a result of the iterative involvement. By systematically collecting the students' experiences and using different data sources, their learning needs became both explicit and concise. These learning needs were subsequently used as the basis for identifying the practical implications and changes to be made to the technological learning material. The five themes evolved through the process of the material development and represent the students' different learning needs.
Clarification of learning expectations
The students undertook a range of different actions to prepare themselves for the final exam, among these were multiple choice questions, video films, assigned reading and correspondence with teachers through e-mails, online discussions forums, and personal meetings. While these different actions did serve different needs, the students' main goal was to understand what the faculty teachers actually expected of them in terms of learning. Their time and energy were often used to decipher the real or hidden meaning behind the information and questions they received from faculty teachers. This often led to uncertainty: 'If you don't have the answer, then we go back and forth. What do they mean? What do they think? How do you interpret it? Then you are left with three different answers…then this uncertainty appears (Interviews, Group B). These were all typical questions from the students. Their biggest fear was a failure to grasp what they needed to learn, which would result in their failing the exam. This fear left them uncertain and insecure, indeed more worried about what the faculty wanted them to know than about how they could learn better and understand the different aspects of the actual procedure. 'The students ask a lot of questions over and over again, and need detailed conformation and information about what to learn (Field notes). The students, therefore, needed better preparation and more information about their teachers' expectations. By clarifying expectations, important time and energy could be diverted toward achieving specific learning goals, instead of searching for them. When addressing this issue by integrating learning goals into the learning material, the students found it easier to grasp what was expected of them, as 'it stood there, in black and white: what is expected of you and what is the answer' (Interview, Group C).
Help to recognize the bigger picture
Another issue that claimed much of the students' attention was the variety of answers they could find for what they saw as being the same type of questions. In their struggle to find the 'right' answers, they often consulted different sources of information, resulting in them finding more discrepancies than clarification. For example, '…we ask the same question to different teachers and get different answers' (Interview, Group B). It seemed that the novelty of their profession led to an extreme attention to details, focusing more on the pieces of the puzzle than the big picture. They seemed to be self-aware of their own deficiency in recognizing the bigger picture while lacking the tools to do something about it '…there is probably many ways to Rome, and they are all right, but we cannot see all the possibilities. For us there is so much we need to keep in mind; it is this procedure and this procedure, we cannot see all the possibilities, we need it to be more specific; that's how it is. Maybe it sounds kind of square, but that's how it is! ' (Interview, Group B). While all these small variations were a source of frustration, their biggest issue was the differences between actual practice and what was taught at the faculty: 'I have practiced (on the procedures) the way I think the sensors would like me to solve the task at the exam, in order to pass. You need to know how it's supposed to be done when you come in there (to the school exam) because the reality in the CSL is not exactly the same as the reality we meet when we are on prac' (Interview, Group A).
The students therefore wanted answers that 'belonged' to every question and a recipe for how things were done and why. While the students searched for ways to simplify their quest for what they saw as 'right answers' the field notes also speculated that the real issue was understanding the bigger picture and indeed, 'recipes with belonging arguments of 'why' could help students think picture instead of pieces? (Field Notes). The original questions embedded in the learning material were therefore complemented with answers and arguments. This aimed to help the students better understand the whole scenario, seeing better connections between principles, actions and arguments: '(…) I think more now, I pay attention if the doctor (when in prac) does it correctly (…) Before I never had the knowledge to do that' (Interview, Group B).
Stimulation of interaction
Besides helping to recognize the whole picture and clarifying expectations, the students appeared to seek, and value every possibility for more interaction. Types of interactions varied between students and those between students and teachers. What all of the activities had in common, however, was that they gave the students' the ability to challenge their own knowledge, test their knowledge, and rate their knowledge to the knowledge of others and thus progress. While all forms of feedback were sought, teacher feedback was especially valued. The students saw this feedback as the safest source of information and information of the highest level to test their knowledge against, although it was often the least available option. The most used alternative was to practice, discuss, and receive precise feedback through group interaction with other students. The problem with this process, however, was uncertainty about the quality of the feedback coming from their peers: 'it is okay to ask each other, I might ask Mary, and then she answers and I think "hm…yes I'm satisfied with that answer", but sometimes I think "is Mary right?…is that the right answer?" And then you get hesitant, because we are not professionals any of us! So sometimes it would be great to have a teacher here!' (Interview, Group D).
The tablet, however, could be used to ask stimulating questions and give feedback that would trigger more interaction both between the students and the tablet and between the students who were practicing together. Critical questions created enthusiasm and engagement with the procedure, while also eliminating the uncertainty that could be raised between peers as in 'you know that what you learn is correct', 'it's a quality assurance' (Interview, Group B).
While the prototype scenarios entailed a limited number of questions, one of the later versions integrated questions into almost every answer to test how the student responded. As noted in the Field Notes, there was'A surprising enthusiasm about all the questions in scenario 4 (Field Notes)'. This mood seemed to be explained by the fact that the students saw the questions as a chance to be challenged about aspects of the procedure that they had not thought of, to 'get some aha-experiences for ourselves (Interview, Group A)' and also to receive tips for possible questions for the exam. All these characteristics, taken together, made the tablet interesting as a potential element for creating highly valued interactions among the students that helped them both prepare and learn.
Creation of structure
Training for the practical oral exam was seen by the students as a stressful event. While they valued all sorts of tools that could help them during training, it was important that these tools simplified, instead of complicating, their preparations. Simplicity, overview, and structure were thus keywords found in the students' feedback created through the layout and design of the content on the tablet. It was important that 'for someone that is doing this for the first time it should not feel so overwhelming', (Interview, Group C). Student feedback, therefore, led to scenarios that were structured chronologically, dividing the different tasks into separate sections to create a natural progression in the scenario. While this dissection could be seen as fragmenting the bigger picture, it accommodated the students' previous statements about needing a recipe to follow due to the novelty of their profession: 'in nursing there is so much (to know)…. But now it gets taken down a notch, and it gets easier to act accordingly' (Interview, Group B). Using the same basic structure in all scenarios created a sense of familiarity and predictability for the students, while giving them the structure they needed. Another important aspect for creating such a structure was enabling the students to follow it. The initial lack of attention to details often caused a gap between what teachers believed was communicated and what the students perceived as having been communicated. 'People are more amateurs than you think (…) I remember when we first started here (in the CSL) some of us had never measured a blood pressure before, and then you are presented with a film, and you see how they measure, but there is no sound. Yes, you blow up this and you put these in your ears, but you don't know how it is supposed to sound. It's like if I was to teach you how to bake a cake I could say: "then you take the flour…" but you would want to know how much flour to take wouldn't you?! (Interview, Group B).
This attention to detail often made the scenarios information rich and long, something that also claimed an opportunity to navigate back and forth in the scenarios and check information they were unsure of, while also to making it easier for them to repeat specific sections of the scenario while creating the structure. The students also pointed out where information needed to be elaborated on, what information could be misjudged or misunderstood, and how information should be phrased, thus keeping them truly on track to know what was important and avoid potential confusion.
Receive context-specific content
While creating a structure revolved around how information was given, the students' contributions were also concerned with what kind of information they needed. Multiple learning tools competed for their attention, and the trouble of their not knowing the best way to learn caused them to jump from one remedy to another. What made them favor the learning material on the tablet, however, was that the content could be specified to each context and situation. Disputes and frustration seemed to be more related to questions concerning context. Discrepancies in answers and information often were rooted in the fact that they were given for different contexts. By giving and explaining context specific information, more tailored to the scenarios, the process helped settle disputes rather than create more of them.
That the learning material was produced in collaboration with teachers and a clinical nurse specialist created a new coherence between what happened during practice, was written in the referenced literature, lectured about in class and the information stored on the tablet. Taken together, this process clarified several factors that had previously been seen as discrepancies by the students, and it helped them see that instruction could be done differently, depending on the context: (Student 1) 'then we actually get an answer….' (Student2) Instead of just us students discussing, because then we never get answers…(Interview, Group B)'. Training using the tablet also created an unexpected positive aspect that helped them prepare for the more psychological aspect of the exam: This is a very good way to work. You get kind of nervous, get some performance anxiety, because you know that she has something that resembles the exam (the tablet). You get to practice the exam situation in a systematic way (Interview, Group C)'. Making the instruction context specific also meant challenging students to think about the context. Asking for explanations and reasons for their actions in each specific setting, but also asking what would have changed if something in the context was changed: 'By using the Simpad, I got quite a few extra tips about the questions that might come, what the sensors could ask, it made me become more aware of the reasons behind things (Interview, Group A) …'.
Practical implications
In order to operationalize the findings for future development of technological learning material, the five different learning needs that evolved through the iterative student involvement process were linked to a set of practical implications. These practical implications can be seen as a checklist of important aspects to consider for future development of technological learning material. The implications are structured in a figure indicating the relationship between the iterative student involvement, the evolved student learning needs, and the practical implications (Fig. 1).
As Fig. 1 displays, each of the five identified learning needs can be operationalized through a set of different implications. It is important to remember that the iterative student involvement process entailed student validation of all implications in this study, and the findings may vary by context. In addition, several aspects related to students' involvement need to be considered, some of which are discussed in the following.
Discussion
This paper documents how nursing students can be actively involved in the development of their own learning materials and how their role indeed contributed to the identification of five different and important learning needs. In the following discussion we look at the involvement when using an iterative process and the level of student involvement for the best learning outcomes.
Using an iterative process for involvement
The students in this study were actively involved in several phases throughout the development process. The process was iterative and entailed identifying student needs and trying to meet them, before adjusting both the needs and the solutions. Without this repetitive process, the unveiling of the specific learning needs would have been more difficult. One of the most important catalysts that enables human beings to become proactive and engaged in activities according to RM Ryan and EL Deci [39] are the catalyzing factors in their environment. Among these are autonomy, which plays a vital role in human motivation [40]. Facilitating autonomy demands decreases external control, provision of individual choices, and acknowledgement of feelings [41]. We believe that the iterative process in contrast to a single mapping of students' experiences facilitates autonomy through acknowledging and integrating students' thoughts and feeling over time. Parallels can be drawn to Freire's [42] deliberative pedagogy where creativity and participation are taken into account. This choice again made the students in this study engaged and interested in the possibility of being able to influence their own learning material.
The process of iterative student involvement can be difficult to achieve due to limited time and resources. Teachers also often experience anxiety over reduced authority when they open up to students for feedback on their performance [43]. Further, students may feel insufficiently equipped to participate in the process [44]. On the other hand, engagement and student involvement, once undertaken, makes students more aware of their faculty's commitment to their own learning [45], thus enhancing knowledge of their own learning process [46], playing an important role in quality improvement [44] and increasing student satisfaction with the material provided them. While satisfaction should not be equalized with quality [47], dissatisfaction with teaching has negative effects on both motivation and engagement [48]. The results from this study indicate that iterative processes that do identify students' needs assumable can foster more motivation and engagement and have the possibility of ensuring the development of learning design that satisfies students' needs.
Level of student involvement
Although user involvement is deemed to be beneficial, there is ongoing debate concerning the extent of that involvement. C Bovill and CJ Bulley [12] adapted Arnstein's ladder of citizen participation [49] to revolve more around student involvement, and specifically distinguish between 'tutors in control' and 'students in control'. The highest level of participation is when students themselves control decision-making and have substantial influence, while the lowest level of participation is when there is no student participation [12]. The highest level of involvement removes the teacher from the equation, leaving the students absent from the influence of the tutor. While this active participation can bring to bear a high level of autonomy, as supported by EL Deci and RM Ryan [41], the removal of the tutor is still challenging in the higher education context due to quality assurance systems [12]. It could also be directly unwise sometimes, as the qualities of good teachers are still vital for the facilitation of learning according to J Hattie [50]. Striving for student participation at the highest rung also was contradicted by some of our findings. Our students clearly stated that the role of the teacher was important, as they needed clarification of learning expectations, along with questions, clues, and answers to help them see details they were not able to see for themselves. The teacher is, therefore, important when designing technological learning material and is supported by PA Kirschner [26]. Shared involvement in the overall process makes both students and teachers valuable, where the aim is not necessarily simply to strive to reach the highest rung of the ladder. Within other professions, user involvement and participatory approaches have gradually shifted toward similar approaches such as 'co-creation', 'co-design' or 'experience-based co-design' [51][52][53][54]. These methods reflect a more democratized approach where the different stakeholders are united in a partnership agreement that fosters a bottom-up approach [33]. The idea is to involve all parties in an ongoing creative process, giving endusers a larger role and the power to make decisions [51]. Education, as advocated by Paulo Freire should in itself be an empowering, participatory process [42]. Involving students through co-creation and co-design could therefore seem suitable for the educational setting since participation and empowerment are the direct consequences of this process. Although the literature on cocreation and co-design within education is somewhat scarce, the method has proven fruitful in areas like health care and service improvement [55][56][57]. Collaboration through combining experience, creativity, and engagement of both students and teachers in a co-design of technological learning material could therefore be beneficial for in many respects.
Although different learning styles are believed to suit different students, the focus of this study was not to match a specific style to a particular type of students but Fig. 1 From student involvement to practical implications rather to add to the body of learning materials in order to increase the chance that all students will find a type of learning material that suits their needs. An analysis of the effects of the learning material described here was beyond the scope of the reported study. Further research is needed to investigate how this learning material impacts students' learning processes. The active student involvement was limited to a group of student representatives. Their opinions might not correspond with other students in the faculty or other nursing faculties, and those differences should also be taken into consideration [58].
Conclusion
This study indicate that iterative involvement of students in the process of developing new technological learning material enhances student identification of important learning needs. Further, the use of students' and teachers' knowledge in an adapted co-design process appears to be the most optimal level of involvement for both students and instructors. Further studies is needed to optimize the approach for student involvement and adjust it to various settings and professions. | 2018-03-28T21:12:06.956Z | 2016-01-12T00:00:00.000 | {
"year": 2016,
"sha1": "440fe09087b568a76596f66bafc2551871d13691",
"oa_license": "CCBY",
"oa_url": "https://bmcnurs.biomedcentral.com/track/pdf/10.1186/s12912-016-0125-y",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "440fe09087b568a76596f66bafc2551871d13691",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Medicine"
]
} |
220835681 | pes2o/s2orc | v3-fos-license | Insulin: Trigger and Target of Renal Functions
Kidney function in metabolism is often underestimated. Although the word “clearance” is associated to “degradation”, at nephron level, proper balance between what is truly degraded and what is redirected to de novo utilization is crucial for the maintenance of electrolytic and acid–basic balance and energy conservation. Insulin is probably one of the best examples of how diverse and heterogeneous kidney response can be. Kidney has a primary role in the degradation of insulin released in the bloodstream, but it is also incredibly susceptible to insulin action throughout the nephron. Fluctuations in insulin levels during fast and fed state add another layer of complexity in the understanding of kidney fine-tuning. This review aims at revisiting renal insulin actions and clearance and to address the association of kidney dysmetabolism with hyperinsulinemia and insulin resistance, both highly prevalent phenomena in modern society.
INTRODUCTION
Insulin is a vital hormone with several functions among which its central role in energy and glucose homeostasis is fundamental (Plum, 2006). Beyond the traditional crosstalk between the most recognized insulin-responsive organs (liver, adipose tissue, and skeletal muscle), kidney insulin action and resistance have recently been suggested as critical components of metabolic and dysmetabolic states (Artunc et al., 2016). The important role of insulin in the kidney is further corroborated by the increased prevalence of chronic kidney disease in subjects with type 2 diabetes mellitus (T2DM) and non-alcoholic fatty liver disease (NAFLD; Musso et al., 2014;Kim et al., 2018;Mok et al., 2019;Kiapidou et al., 2020). Furthermore, the kidney can adjust the levels of essential players of nutrient homeostasis such as glucose and insulin (Krebs, 1963;Chamberlain and Stimmler, 1967).
First reports of insulin actions in the kidney were originated in the 1950s (Farber et al., 1951). Since then, the knowledge acquired has dramatically increased, nonetheless much remains unveiled. It is now known that insulin impacts on tubular glucose reabsorption . In fact, a whole therapeutic class based on this mechanism was recently developed. Additionally, hyperinsulinemia driven by pancreatic hypersecretion and/or impairments in hepatic insulin clearance could explain changes in glomerular filtration rates and increased renal gluconeogenesis, which in turn causes fatty liver deposition and, as a direct consequence, may lead to dysglycemia (Bril et al., 2014;Naderpoor et al., 2017;Jung et al., 2018;Bergman et al., 2019).
Kidney is, in fact, the most critical organ responsible for insulin clearance after the liver (Elgee and Williams, 1954;Narahara et al., 1958); still, its relevance in the maintenance of proper insulinemia and insulin sensitivity is underestimated. The control of insulin clearance/action exerted by the kidney is complex. Furthermore, the diversity of insulin actions is only possible thanks to the differentiation of specific kidney segments in which insulin regulates different pathways. In this review, we will briefly describe insulin production and its main actions in the liver, adipose tissue, and skeletal muscle. Moreover, we will focus on kidney and on renal insulin actions as well as insulin clearance/degradation. Overall, this work aims at highlighting the critical role of kidney-insulin interplay in the development of dysmetabolism.
OVERVIEW OF INSULIN BIOSYNTHESIS AND NON-RENAL ACTION AND METABOLISM
The insulin gene (INS) is represented by only one copy in the human genome, and its transcription is mainly regulated by the same enhancers of other glucose-related genes (Andrali et al., 2008;Gao et al., 2014). In pancreas β-cell, INS is first transcribed as preproinsulin and, after cleavage and folding in the rough endoplasmic reticulum (ER), proinsulin is formed (Hutton, 1994). The conversion into insulin occurs after the removal of the c-peptide in the Golgi apparatus, with a final maturation achieved by c-terminal amino acid elimination (Steiner, 2004). Mature insulin is composed of A and B chains linked by two disulfide bonds, with a third bond within A chain (Mayer et al., 2007). For storage, insulin is complexed in hexamers with zinc, and an actin network is involved in the organization of mature insulin granules for first and second insulin secretion phases (Kalwat and Thurmond, 2013). Insulin secretion is triggered by cytoplasmic increments in ATP levels, derived from glucose internalization and subsequent metabolism (Prentki and Matschinsky, 1987). This increase in ATP shuts down membrane ATP-sensitive K + channels, depolarizing plasma membrane, and activating voltage-gated calcium channels (Detimary et al., 1998). Indeed, the resultant Ca 2+ influx seems to be necessary for the fusion of insulin granules with the plasma membrane, ending up in its secretion (Hou et al., 2009). The first wave of insulin secretion is composed of primed granules that are in close proximity with plasma membrane (Olofsson et al., 2002). However, most of these granules remain more distant from the plasma membrane and are further boosted with newly synthesized insulin that will culminate in 75-95% of total insulin secretion in the second phase (Kou et al., 2014).
With the insulin pools ready to be released, insulin secretion does not occur in a constant manner, but in waves of 4-5 min interval (Song et al., 2000). The pulsatile secretion profile is considered primordial for proper insulin responsiveness, as constant release would result in down-regulation of insulin receptor (INSR) expression (Matveyenko et al., 2012). After binding and activation of its receptor, insulin, as an anabolic hormone, promotes increased biosynthesis of macromolecules further needed to maintain energetic balance during nutrient privation (Bedinger and Adams, 2015).
At cellular level, insulin mostly relies on the specific bind to its transmembrane receptor to perform its actions (Figure 1). Insulin receptor is a tyrosine kinase receptor with two extracellular α subunits and two transmembrane β subunits (Ye et al., 2017). Insulin binding to INSR produces an autophosphorylation that leads to the internalization of the complex and activation of downstream effectors. INSR substrate 1 (IRS1) and 2 (IRS2) are the main targets that intracellularly coordinate insulin action (Thirone et al., 2006;Kubota et al., 2016). In the liver, the activity of these proteins is differently controlled by insulin. Although both IRS1 and IRS2 are degraded after insulin activation, only IRS2 is transcriptionally repressed by insulin activation (Hirashima et al., 2003). Further transduction of insulin signaling occurs via phosphatidylinositol 3-kinase (PI3K)/AKT phosphorylation. AKT or Protein kinase B (PKB) are considered the main effectors of metabolic insulin actions (Taniguchi et al., 2006;Molinaro et al., 2019). On the other hand, AKT1 and 3 are the leading players of mitogenic effects of insulin by RAS/mitogen activated protein kinase (MAPK) cascade and crosstalk with growth hormones (Xu et al., 2006;Xu and Messina, 2009;Qiu et al., 2017).
Given the scope of this review, we will summarize insulin actions on non-renal target organs, but detailed information is available in recent reviews (Gancheva et al., 2018;Tokarz et al., 2018). The first organ to be reached by insulin is the liver, where 50-70% is removed at first pass (Duckworth et al., 1998). In hepatocytes, insulin signaling recruits forkhead box transcription factor 1 (FoxO1) to the cytoplasm, inhibiting the transcription of gluconeogenic enzymes (Ling et al., 2018). In such an abundance of glucose, insulin also stimulates the synthesis of glycogen by phosphorylation of glycogen synthase kinase 3 (GSK3), which leads to dephosphorylation and activation of glycogen synthase FIGURE 1 | Canonical insulin signaling cascade. Insulin (gray circle) binds to its transmembrane kinase receptor at cell membrane and triggers insulin signaling cascade. Insulin receptor present two isoforms with distinct affinity for insulin already associated to differential insulin internalization (INSR A and INSR B;Calzi et al., 1997). Binding of insulin auto-phosphorylates INSR at Tyr-960. Further recruitment and phosphorylation of insulin receptor substrate (IRS) 1 and 2 will mostly result in subsequent activation of phosphoinositide-dependent kinase 1 (PDK1) through phosphoinositide-3 kinase (PI3K; Yamada et al., 2002). PDK1 is responsible for propagation of insulin signal to one of the most important downstream effectors, Akt. Importantly, Akt has two distinct phosphorylation sites, Thr308 activated by PDK1 (Alessi et al., 1997) and Ser473 phosphorylated by mammalian target of rapamycin complex (mTORC) 2 protein (Bayascas and Alessi, 2005). Finally, fully activated Akt can interact with different proteins, eliciting different effects as stimulation of glucose uptake and glycogen synthesis by AS160 and glycogen synthase kinase 3 (GSK3), respectively (Ng et al., 2008). On the other hand, INRS activation also promote growth factor receptor-bound protein 2 (GRB2) interaction with Shc proteins and activation of mitogen activated protein kinases (MAPK; Skolnik et al., 1993;Xu et al., 2006). This as part of the insulin-mediated proliferative stimuli. PIP2, phosphatidylinositol 4,5 biphosphate; PIP3, phosphatidylinositol 3,4,5 triphosphate. (McManus et al., 2005). After the replenishment of liver glycogen stores in the fed state, glucose starts to be degraded to pyruvate, which is the precursor of lipid biosynthesis via de novo lipogenesis (DNL; Sanders and Griffin, 2016). Herein, insulin signaling is also crucial for upregulation of sterol regulatory element binding protein 1c (SREPB1c), a transcription factor associated to further transcription of DNL (Oh et al., 2003;Roder et al., 2007;Krycer et al., 2010).
Only a fraction of the insulin that reaches the liver will get to the periphery, to perform its action at skeletal muscle, adipose tissue, and other insulin-sensitive organs. Skeletal muscle is the greater insulin-dependent "kidnapper" of glucose, it is responsible for ∼60 to 80% of whole-body glucose disposal (DeFronzo et al., 1976;Fernandes et al., 2011). Insulin signals to myocytes through classical signaling, promoting glucose uptake by the dynamic translocation of glucose transporter 4 (GLUT4; Zorzano et al., 1996). Glucose is here mainly used for glycolysis or stored as glycogen as already described for the liver.
During the fed state, a clear inhibitory effect in lipolysis and consequent blunt of free fatty acid (FFA) release is mediated by insulin acting on adipocytes (Dimitriadis et al., 2011). The resultant decrease in FFA availability reduces its hepatic uptake and fatty acid oxidation, changing the disposal of substrates that are mostly used for gluconeogenesis (Galgani et al., 2008). Not exclusively for liver, the effect of insulin in reducing adipose tissue FFA release is important for other organs that, at fed state, change energy source to glucose and recover glycogen stores contributing to normoglycemia (Petersen and Shulman, 2018). In this context, glucose uptake is also stimulated by insulin in the fed state at adipocytes. Furthermore, translocation of glucose transporters to the membrane of adipocytes was very well characterized in 3T3-L1 adipocytes by activation of the canonical PI3K/AKT2 pathway (Ng et al., 2008).
As mentioned above, all these organs expressing INSR are responsible for insulin clearance and local degradation. However, a significant part of insulin degradation is played in the liver (50-70%) by its specialized machinery of internalization and degradation (Najjar and Perdomo, 2019). In the case of internalization, carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM-1) phosphorylation by insulin promotes rapidly receptor-mediated uptake (Poy et al., 2002). Carcinoembryonic antigen-related cell adhesion molecule 1 expression occurs preferentially in liver, and restoration of its function only at liver of null CEACAM-1 mice reverse the hyperinsulinemia and fatty liver deposition observed in global knockout CEACAM-1 model (Russo et al., 2017). Regarding degradation, insulin-degrading enzyme (IDE) is considered the most critical protease to cleave insulin intracellularly and is highly expressed in the liver (Kuo et al., 1993).
The amount of insulin that remains in circulation after liver first pass (30-50%) will be mostly degraded by the kidney (Elgee and Williams, 1954;Narahara et al., 1958). However, the kidney is not just an endpoint for insulin degradation, as insulin also acts throughout the entire nephron. Whereas hepatic insulin clearance is regaining attention as the very first alteration leading to hyperinsulinemia, compensatory mechanisms that could emerge from the kidney are not well known. Impairments in hepatic insulin clearance increase peripheral insulinemia, and very little is described regarding the exact effects of the increased amount of insulin that will further reach the kidney.
KIDNEY AND INSULIN ACTIONS
Kidney works together with the liver for maintenance of optimal insulin levels. In humans, kidney removes 6-8 U of insulin per day by two major routes (Figure 2), that apparently have distinct preferential physiological goals: post-glomerular secretion (insulin signaling) and glomerular filtration (nutrient conservation and homeostasis) (Ritz et al., 2013). In the kidney, insulin acts at multiple sites along the nephron, from the glomerulus (Coward et al., 2005;Welsh et al., 2010;Mima et al., 2011;Lay and Coward, 2014) to the renal tubule (Tiwari et al., 2008Li et al., 2013), to modulate different functions such as glomerular filtration, gluconeogenesis, renal sodium handling, among others.
Many alterations in insulin associated mechanisms in the kidney are driven by insulin resistance (IR), leading to diabetic nephropathy if not reestablished (Svensson and Eriksson, 2006). In the Asiatic population, it was described that the prevalence of diabetic nephropathy with impaired estimated glomerular filtration rate (eGFR) (with or without albuminuria) was 31.6%, and the prevalence of albuminuria (with or without impaired eGFR) was 16.9 and 22.0%, respectively (Mok et al., 2019). Patients may already show diabetic kidney disease (DKD) at the time of T2DM diagnosis. Nevertheless, 10 years after the diagnosis of T2DM, low-level albuminuria is present in 24.9% of the patients and 5.3% already present macroalbuminuria (Adler et al., 2003).
Glomerular Insulin Actions
Podocytes are the primary constituent cell of the glomerulus, with their long finger-like projections to the glomerular capillaries at the glomerular basement membrane (GBM). These cells have intercellular junctions that form filtration barriers to help maintaining normal renal function. When damaged, podocytes lose their arrangement resulting in a reduction in barrier function. Indeed, one of the DKD features is the podocyte loss with consequent albuminuria (Pagtalunan et al., 1997;Wolf et al., 2005). Lately, it has been unveiled the role of insulin in podocytes ( Figure 3); in other words, podocyte IR gives rise to the albuminuric feature.
Podocytes reveal to express proteins of the insulin signaling canonical pathway, namely the INSR, and both IRS1 and IRS2 (Coward et al., 2005;Santamaria et al., 2015), with IRS2 being the most prevalent one. Coward et al. (2005) depicted the effect of insulin in human podocytes, which results in glucose uptake not only through GLUT4 but also through glucose transporter 1 (GLUT1; Figure 3). As insulin promotes the translocation of GLUT4 to the membrane through the activation of PI3K-AKT2-PkB pathway, there is a remodeling of the cortical actin of the cytoskeleton with subsequent contraction (Welsh et al., 2010). In compliance, podocytes-specific deletion of INSR in mice revealed DKD features based on substantial albuminuria and histological features as podocyte foot structure loss and glomerulosclerosis (Welsh et al., 2010). AKT seems to play a central role as its phosphorylation appears to be severely affected in several models of both type 1 diabetes mellitus (T1DM; streptozotocin-induced T1DM) and T2DM (db/db mice, Zucker rats) (Tejada et al., 2008;Mima et al., 2011). Moreover, AKT isoform 2 deletion results in serious glomerular lesions in mice. This can lead to rapid disease progression, also associated with tubular dilatation and microalbuminuria (Canaud et al., 2013). Other relevant players that might contribute to podocyte IR is SH2-domain-containing inositol phosphatase 2 (SHIP2), a down regulator of the PI3K signaling pathway shown to be upregulated in the Zucker rats. Moreover, protein tyrosinephosphatase 1B (PTP1B), a negative regulator of the INSR activity, or phosphatase and tensin homolog when increased, appears to also compromise the insulin signaling pathway (Mima et al., 2011;Garner et al., 2018).
Podocytes also present an insulin-dependent alternative pathway, the cyclic guanosine monophosphate(cGMP)dependent protein kinase G (PKG), from which the PKG isoform I-alpha levels are increased in glomeruli of the hyperinsulinemic Zucker rats (Piwkowska et al., 2013). These high insulin levels increase glomerular barrier albumin permeability through a PKGI-reliant mechanism via the NAD(P)H-dependent generation of superoxide anion.
An important player in podocytes physiology is the protein nephrin, a podocyte-specific protein, which is responsible for the maintenance of the integrity of the filtration barrier. In fact, nephrin mutations are involved in severe nephrotic syndrome (Lenkkeri et al., 1999). Nephrin appears to play a most outstanding role in the trafficking of GLUT4 and GLUT1 by interacting with Vamp2 as well as by interacting with insulinstimulated actin remodeling (Coward et al., 2007;Lay et al., 2017). Of interest, nephrin can also induce phosphorylation of p70S6K in a PI3K-dependent manner independently of INSR/AKT2 activation (Villarreal et al., 2016 ; Figure 3).
FIGURE 2 | Continued
FIGURE 2 | Schematic representation of renal insulin handling. Insulin is depicted in gray circles at different portions of the nephron. As insulin is a small molecule, it will be fully filtered by the glomerular system until it reaches the proximal tubule. At proximal tubule cells, almost all the filtered insulin will be absorbed at the luminal membrane. In physiological conditions, only a small percentage will be excreted in urine. Beyond glomerular filtration, insulin also raise from the perivenous capillaries. In a closer look to the proximal tubule cells is represented both mechanisms of insulin clearance. The increased levels of insulin receptor (INSR) and consequent increased uptake of insulin at the basolateral membrane is also depicted.
The development of glomerular IR is triggered by several factors: high glucose and/or insulin levels, increased FFAs levels, or an inflammatory milieu. Moreover, insulin signaling appears to be relevant for the adaptative ER stress response in DKD; this is the case for mice with podocyte-specific heterozygous INSR deletion (Madhusudhan et al., 2015). In support of this view, stable overexpression of INSR or knockdown of PTP1B was protective against ER stress (Garner et al., 2018). Podocyte mitochondria play an essential role in cellular metabolism. When dysfunctional as a result of reactive oxygen species production, mitochondria triggers apoptosis, which can also be observed along with a compromised IR state (Susztak et al., 2006). Certainly, the preservation or reestablishment of podocyte integrity is essential in the prevention of the onset and development of DKD.
INSULIN TUBULAR ACTIONS
In the kidney tubule, insulin has several roles: metabolism, electrolyte and acid-base regulation and absorption of filtered substances. However, the exact mechanisms by which insulin performs these distinct roles is not fully understood. Nonetheless, it seems that, at least some of them, are mediated by INSR, and can be explained by the recruitment of specific IRS, as recently shown by Nakamura et al. (2020) specifically for gluconeogenesis and sodium reabsorption regulation. Still, there are overlapping mediators in downstream pathways. In the following paragraphs we will summarize the most relevant and well-known insulin actions in the tubular segment.
Insulin receptor is present throughout the entire nephron (Butlen et al., 1988;Stechi et al., 1994), however, insulin binding capacity of luminal and basolateral membrane tubule cells is different. There is evidence showing same affinity of INSR in both membrane sides of the cell, nonetheless its abundance is asymmetrical (Hammerman, 1985). In fact, the binding capacity of the contraluminal compared to luminal membrane seems to be several times greater due to higher expression of INSR (Talor et al., 1982). Figure 4 summarizes insulin signaling in proximal tubule (PT), regarding its actions in both gluconeogenesis and sodium reabsorption. Additionally, insulin actions through INSR are thought to be different in the proximal and distal nephron regions. Tiwari et al. (2013) demonstrated, in a mouse model with deletion of INSR in the kidney tubule cells, that depending on the segment targeted with INSR deletion, there were different phenotypes further described. In case of decreased INSR at PT, FIGURE 3 | Podocyte insulin signaling. Podocytes are the first cells to interact with insulin at the nephron and express several proteins of the canonical insulin signaling pathway. However, here the podocyte-specific protein nephrin is known to have a role in the trafficking of glucose transporters (GLUT1 or GLUT4) to podocyte membrane and consequently promote glucose uptake. The trafficking seems to involve Vamp2 and actin remodeling. On the other branch of insulin signaling, an effect in the large-conductance Ca 2+ -activated K + channels (BK Ca ) is also important for maintenance of podocyte integrity and proper glomerular filtration. GRB2, growth factor receptor-bound protein 2; GSV, GLUT storage vesicle; VAMP2, Vesicle-associated membrane protein 2.
animals had a mild diabetic phenotype, without increased IR when compared to control. These animals shown to have an higher activity of gluconeogenesis enzymes . On the other hand, in animals with the deletion of INSR targeted to distal parts of the tubule, elevated blood pressure and impaired sodium excretion was observed (Tiwari et al., 2008).
Glucose Reabsorption
Glucose is reabsorbed by the PT cells from the kidney tubule lumen to the bloodstream ( Figure 4A). In the kidney, GLUT2 is in the basolateral membrane and diffuses glucose out of the cell, contrary to the liver, where GLUT2 acts in glucose uptake. Sodium-glucose transporter proteins (SGLT) are responsible for glucose and sodium co-transport by the luminal membrane of kidney cells. Sodium-glucose co-transporter 2 (SGLT2) is a highcapacity/low affinity sodium-glucose cotransporter present in the apical membrane of proximal convoluted tubule cells (Vallon et al., 2011). These transporters are responsible for approximately 90% of filtered glucose reabsorption, and an important target for T2DM therapy (SGLT2 inhibitors;DeFronzo et al., 2012). The remaining 10% of glucose in the tubule is absorbed by SGLT1, a low capacity and high affinity sodium-glucose cotransporter (Wright et al., 2011). Kidney has a threshold for glucose excretion which, in healthy conditions, relates to a glycemic value around 180 mg/dl ( Figure 4B). This threshold, however, can be altered in diabetes (Rave et al., 2006). It is not clear if SGLT2 glucose transport is or not directly dependent on insulin signaling (Ferrannini et al., 2020). Nonetheless, SGLT2 expression was shown to be upregulated by insulin on human cultured PT cells, in a dose-dependent manner . Therefore, in hyperinsulinemic states, frequently observed in prediabetes and T2DM, an excessive glucose absorption can be observed. Of notice, glucose is still highly absorbed by SGLT2 in IR states, suggesting that this mechanism is not affected by IR, though it is upregulated by hyperinsulinemia. In this case, a vicious cycle can happen where increased insulin levels drive an increase in glucose SGLT2 overexpression increasing glycemic levels which in turn will increment the insulin secretion ( Figure 4C). Therefore, SGLT2 overexpression inhibition is a potential new target for highly prevalent hyperinsulinemia related conditions, namely some dysglycemic phenotypes and obesity. In this case, lowering insulin levels could be paramount to prevent SGLT2 overexpression. Thus, SGLT2 inhibitors might become a relevant therapeutic approach for hyperinsulinemia related conditions, other than T2DM, namely obesity and prediabetes.
Gluconeogenesis Regulation
Kidney has a major role in gluconeogenesis along with the liver and the intestine. Gluconeogenesis occurs mainly in PT cells essentially from lactate and glutamine ( Figure 4A). Moreover, PT cells do not use glucose, as they get their energy mostly from fatty acid oxidation (Gerich et al., 1963). Insulin regulates gluconeogenesis in PT cells to meet the fluctuating needs of the body. While in the fasting state the kidney contributes in 40% to overall gluconeogenesis, in the post-absorptive state its contribution drops to 20% (Gerich et al., 1963; Figure 4B). FIGURE 4 | Dynamics of proximal tubule cells at fasting, fed and insulin resistant states. Proximal tubule cells are subjected to distinct microenvironments (lumen and interstitium) and the regulation of absorption and reabsorption of molecules is complex. Although all the described processes occur in every cell of the proximal tubule simultaneously, each specific process is illustrated in a different cell. At fasting (A), low levels of insulin allow expression of gluconeogenic enzymes whereas sodium reabsorption is downregulated. Expression of glucose transporter 2 (GLUT2) at basolateral membrane is mostly associated to glucose output and not to its uptake. Moreover, albumin absorption is performed by megalin and cubilin at luminal membrane and transcytosis allow albumin to be rerouted back to the organism. At fed state (B), increased availability of insulin and glucose promote drastic changes in proximal tubule dynamics. In the case of insulin, luminal uptake is mostly associated to degradation and basolateral to signaling activation. Insulin receptor (INSR) activation downregulates gluconeogenesis and increases sodium reabsorption by different proteins as type 3 Na-H exchanger (NHE-3) and sodium-glucose transport protein 2 (SGLT2). Together with sodium, SGLT2 also co-transport glucose from the lumen. Finally, hyperinsulinemia is linked to perturbations of proximal tubule cells in many aspects (C). As in many other organs, insulin signaling desensitization is associated to inefficient inhibition of gluconeogenesis contributing to maintenance of increased levels of glucose. Derangements at podocyte level increases filtration of albumin and overloads luminal capacity of reabsorption. Such impairment in albumin reabsorption culminates with albuminuria, frequent observed in hyperinsulinemic states. Nakamura et al. (2020) demonstrated that insulin directly inhibits gluconeogenesis in isolated PT through IRS1/AKT2/mTORC1/2, and that mTORC1 positively regulates insulin signaling (Nakamura et al., 2020). Specifically, in the fasting state, suppressed insulin signaling increases FoxO1 activity, increasing the expression of gluconeogenic genes, such as PEPCK and glucose-6-phosphatase. In parallel, decreased glucose reabsorption via SGLT2 on the luminal membrane downregulates the NADH/NAD + ratio, activating sirtuin 1 and peroxisome proliferator-activated receptor-gamma coactivator 1α (PGC1α), a coactivator of FoxO1. Therefore, these two mechanisms lead to enhanced gluconeogenesis. On the contrary, in the fed state, the increased insulin levels and glucose reabsorption result in the suppression of gluconeogenesis through downregulation of the previously mentioned gluconeogenic genes Sasaki et al., 2017). Whereas glucose is reabsorbed by the luminal membrane, insulin interacts with the basolateral membrane, thus gluconeogenesis regulation results from the integration of signals from distinct cell microenvironment.
Sodium
Insulin impacts on the fine-tuning of several electrolytes by the kidney. Among them sodium handling is probably the best described (DeFronzo et al., 1975). However, the insulin role on sodium retention in normoglycemia is not completely established. Regulation of sodium absorption is paramount to maintain the extracellular volume in a physiological range. Kidney is the principal organ involved in sodium excretion, responding and adapting to sodium intake (Krekels et al., 2015). Additionally, sodium has a major role in driving electrochemical forces that support kidney primary role in finetuning body composition. Sodium reabsorption and excretion results from the integration of a complex network of sensors, neural-hormonal stimuli and hemodynamic and metabolic mechanisms (Frame and Wainford, 2017).
Contrary to glucose, sodium is absorbed along the nephron by distinct apical sodium transport proteins. Usually, approximately 65% of filtered sodium is reabsorbed in the PT along with water, mainly through type 3 Na-H exchanger (NHE-3) and SGLT proteins. In addition, the thick ascending limb is responsible for approximately 25% of reabsorption through sodium-potassium chloride cotransporter-2 (NKCC2). Finally, 5-10% of sodium is reabsorbed in the collecting duct by epithelial sodium channel (ENaC) and less than 10% is excreted in urine (Esteva-Font et al., 2012).
The association of insulin with sodium absorption was suggested almost a century ago (Atchley et al., 1933), nonetheless discrimination of insulin-and glucose-mediated effects has not been clarified (DeFronzo et al., 1976;Manhiani et al., 2011). It is still under debate if insulin has a causal effect on hypertension under normoglycemia. Nevertheless, it is known that insulin stimulates sodium absorption in all the tubule segments where it takes place (Kirchner, 1988;Friedberg et al., 1991;Ghezzi and Wright, 2012). In the PT, insulin regulates several sodium transporters, in both luminal (NHE-3, SGLT2) and basolateral membrane (Na/K-ATPase, NBCe1; Figure 4A; Gesek and Schoolwerth, 1991;Feraille et al., 1994;Ruiz et al., 1998). Recently, Nakamura et al. (2020) showed that, regarding sodium absorption in PT, insulin recruits an IRS (IRS2) different from the one orchestrating gluconeogenesis (IRS1). Insulin receptor substrate 2 acts through the AKT2/mTORC2 pathway (Nakamura et al., 2020; Figure 4A). In fact, mTORC2 is known to activate serum/glucocorticoid regulated kinase 1 (SGK1), that will then stimulate ENaC and NHE-3, increasing sodium reabsorption (Satoh et al., 2015). It has been suggested that, in healthy conditions, with rising insulin levels in fed state, IRS2 desensitize, suppressing sodium reabsorption at PT and increasing its delivery in the distal convoluted tubule ( Figure 4B). Moreover, with IR, the desensitizing mechanism is abolished and therefore sodium will not reach the distal tubule (Ecelbarger, 2020; Figure 4C). Of notice, by recruiting distinct IRS, kidney cells can somehow dissociate pathways performing distinct functions.
Finally, it must be kept in mind that insulin can interact with intrarenal and systemic renin-angiotensin-aldosterone system in several ways (Muscogiuri et al., 2008) and therefore, indirectly interfere with sodium reabsorption in different mechanism out of the scope of this review.
Albumin Absorption
The luminal membrane of PT cells is primarily responsible for the reabsorption of proteins that are freely filtered in the glomerulus by receptor-mediated endocytosis (Figure 4A; Christensen and Gburek, 2004). This is the case of albumin reabsorption that can have an important role in energy conservation. It has been suggested that albumin endocytosis is a regulated process, dependent on membrane receptors megalin and cubilin (Christensen and Birn, 2001). More recent evidence suggest that insulin might also have a role in the regulation of tubular albumin absorption (Kumari et al., 2019).
Albuminuria is of major clinical relevance in diagnosis and follow-up of kidney disease including subjects with diabetes. Insulin resistance was found to be associated with decreased INSR expression in tubular cells in rat models (Wang et al., 2005). In these observations, Kumari et al. (2019) analyzed urine samples from mice with targeted deletion of INSR from the renal PT. These mice had an impaired uptake of albumin, without any glomerulopathy. They also demonstrated that in healthy humans, albumin absorption capacity and excretion vary from the fast to the fed state. Moreover, IR was associated with microalbuminuria even in normoglycemia as described in the RISC study (Pilz et al., 2014) and thus can be present regardless of diabetes diagnosis. Altogether, these evidences suggest that albuminuria might be an important marker of kidney tubular dysfunction and might reflect tubule cells IR (Figure 4C). These reinforces the kidney contribution to diabetes development and highlights insulin and albumin dynamics prior and regardless of the development of diabetes.
KIDNEY INSULIN CLEARANCE
In the systemic circulation, besides insulin metabolization by the liver, the kidney is the major site of insulin clearance (around 25%) (Elgee and Williams, 1954;Narahara et al., 1958); its action might be required to limit excessive insulin levels. Evidences supporting this theory started to rise in the middle of the 20th century (Zubrod et al., 1951;Elgee and Williams, 1954;Narahara et al., 1958;Ricketts and Wildberger, 1962). In a study from 1966, Beck et al. (1966) found that when insulin was injected intravenously in mice, it concentrated in liver and kidney; however, with higher insulin doses, raised insulin levels were found in rat's kidney, while these levels were found to be reduced in liver. Nevertheless, this early study has some methodological limitations. Indeed, kidney insulin clearance remains constant in spite of insulinemia variations, but varies with creatinine clearance (Rubenstein et al., 1967). Globally, these evidences suggest that kidney insulin clearance is a non-saturable process, although dependent on glomerular filtration rate.
Insulin freely filtered in the glomerulus is absorbed by the lining cells of the PT (Figure 2). Upon entering the cell, insulin is transported through the luminal membrane into the PT cells and is degraded. Luminal insulin reabsorption limits urine insulin excretion, thus less than 2% of insulin reaches the urine in normal fasting conditions (Rubenstein and Spitz, 1968). Insulin is transported through the luminal membrane by a receptormediated endocytic mechanism (Rabkin et al., 1984). Endocytic internalization of insulin seems to be more related to insulin degradation than to insulin biological actions (Figure 2).
While glomerular filtration of insulin could not account for its total estimated renal extraction, a second mechanism was postulated. In this context, it was observed that a significant amount of insulin was cleared by the post-glomerular peritubular capillaries into the tubule cells (Chamberlain and Stimmler, 1967). In humans, this route represents around one-third of cleared insulin in the kidney, where it enters tubule cells not just by endocytosis, but also by INSR mediated uptake. In this case, through INSR binding, insulin signal to the kidneys' tubular apparatus is crucial to maintain central physiologic functions, similarly to what happens in extra-renal tissues, namely regarding glucose homeostasis and blood pressure.
Insulin degrading activity has been observed at cytosol, lysosomes and mitochondria in addition to the membrane, indicating that it occurs in distinct cell sites. Degradation at the membrane level, however, seems to represent less than 2% of total degrading activity (Rabkin et al., 1984). Insulin can be initially hydrolyzed by an insulin protease followed by the action of plasma-membrane-associated or lysosomal proteases. This pathway can degrade insulin entering through both luminal and contraluminal membrane. In another possible pathway, endocytic vesicles containing insulin fuse with lysosomes. This pathway comprehends glutathione-insulin transhydrogenase (GIT) action, followed by hydrolysis of intact A and B chains by lysosomal proteases, and seems to need insulin internalization. It may act primarily on insulin delivered by luminal uptake and it is most active when supraphysiological levels of insulin are present (Rabkin et al., 1984). Azizi et al. (2015) demonstrated that, in a culture of human adipose microvascular endothelial cells, insulin can go through microvascular endothelial cells by transcytosis (Azizi et al., 2015). Regarding insulin handling in the kidney tubule, Dahl et al. (1989) hypothesized that insulin molecules could also pass tubular cells by a retroendocytic pathway instead of being degraded. The authors demonstrated that cultured opossum kidney cells exhibited a retroendocytic pathway for insulin (Dahl et al., 1989). Using the same model, the authors later demonstrated that inhibition of insulin degradation diverted intact insulin from the degradative to the retroendocytic pathway (Dahl et al., 1990). Although captivating, especially regarding a potential contribution to hyperinsulinemic states, this hypothesis was not further explored.
Whether insulin clearance mechanisms attributed to other organs affects renal function and insulin clearance it is not clear. For example, the lack of CEACAM-1, a key protein enrolled in hepatic insulin clearance driving hyperinsulinemia, in the kidney leads to increased renin levels contributing to a potentiation of the RAS system and hypertension. These effects are exacerbated upon high fat diet exposure. Hence, the described CEACAM-1 renal effects can be due to the lack of its expression as well as the observed hyperinsulinemia (Huang et al., 2013;Li et al., 2015).
Despite early conflicting results, further studies showed that insulin is excreted in urine. However, in physiological conditions, it represents a minimal proportion of insulin filtered in the glomerulus. In health, a minor amount of insulin appears in the urine, as the majority is absorbed in PT. Tubule absorbing capacity of insulin does not saturate and thus the insulin fraction excreted in urine is constantly small, regardless of insulin levels. However, the amount of insulin excreted in urine varies physiologically (e.g., fasting and post-prandial) and in pathological conditions (obesity, diabetes) depending on the affected nephron region (e.g., glomerulopathy vs. tubulopathy) (Rubenstein and Spitz, 1968). Considering that insulin is internalized in the apical membrane by a receptor-mediated endocytic mechanism, the increased urinary insulin excretion might represent a tubular dysfunction. Subjects with tubulopathy show large amounts of insulin in urine approximating the amount that is filtered (Rabkin et al., 1984). Conversely, subjects with nephrotic syndrome show normal amounts of insulin in urine. When both glomerular and tubule lesion occur urine insulin excretion increase (Rabkin et al., 1984).
CLINICAL INSIGHTS ON INSULIN DYSREGULATION
Insulin resistance is a common feature in chronic kidney disease (CKD) patients, even in absence of diabetes (DeFronzo et al., 1981;Shinohara et al., 2002;Becker et al., 2005;Kobayashi et al., 2005;Landau et al., 2011), and it is a risk factor for CKD progression (Fox et al., 2004). Its prevalence in CKD ranges from 30 to 50%, and this mainly depends on the adopted method of measurement (Spoto et al., 2016). Insulin resistance can be detected at the very early stages, when eGFR is still within the normal range, suggesting a potential role in triggering CKD (Fliser et al., 1998). A large study based on the Atherosclerosis Risk in Communities (ARICs) cohort confirmed that CKD development increases in strict parallelism with the number of metabolic syndrome criteria measured in non-diabetic adults, and this relationship remains significant even after controlling for the development of diabetes and hypertension (Kurella et al., 2005). Insulin resistance has also been associated with prevalent CKD and rapid decline in renal function in elderly, non-diabetic, Asian individuals (Cheng et al., 2012), and with microalbuminuria in the general population (Mykkänen et al., 1998), and in patients with T1DM (Yip et al., 1993;Ekstrand et al., 1998) and T2DM (Groop et al., 1993), indicating that this relationship is independent of diabetes (Mykkänen et al., 1998;Chen et al., 2003Chen et al., , 2004. The proposed mechanism by which IR contributes to kidney damage involves the worsening of renal hemodynamics through activation of the sympathetic nervous system (Rowe et al., 1981), sodium retention, decreased Na + , K + -ATPase activity, and increased GFR (Gluba et al., 2013).
The etiology of IR in CKD is multifactorial, depending on classical and CKD-peculiar risk factors, such as physical inactivity, inflammation and oxidative stress, adipokine derangements, vitamin D deficiency, metabolic acidosis, anemia and microbial toxins (Spoto et al., 2016).
Long-term hemodialysis has a positive effect on IR (DeFronzo et al., 1978), but there is little clinical data regarding the effect of peritoneal dialysis.
In addition to being a risk factor for CKD onset and progression, IR is also involved in the increased cardiovascular (CV) risk in this population. IR may be responsible for high blood pressure via direct stimulation of renin-angiotensin-aldosterone system (Nickenig et al., 1998), activation of sympathetic system (Sowers et al., 2001) and downregulation of the natriuretic peptide system (Sarzani et al., 1999).
However, the association between IR and CV complications in CKD patients is still to be clarified, as well as the relationship between IR and all-cause and CV mortality.
A positive association between IR and all-cause mortality was found in smokers and physically inactive CKD patients and in a small cohort of 170 Japanese, non-diabetic, dialysis patients (Shinohara et al., 2002), whereas association with CV mortality was found in a cohort of peritoneal patients. However, no association was found in the ULSAM cohort, including 3-4 stage CKD patients (Jia et al., 2014) and in a small study performed in peritoneal dialysis patients (Sánchez-Villanueva et al., 2013).
Even though the prognostic value of IR for death and CV events need to be clarified, the association with CKD is well established. Even in this case, however, if IR is responsible of the onset and progression of CKD, or if CKD is responsible for IR is still to be clarified. A possible answer to this question could derive from clinical trials aiming at assessing the effect of drugs used for IR treatment on kidney function.
Thiazolidinediones (TZDs) are a class of oral diabetic medications that increase insulin sensitivity by acting on PPARγ (Yki-Järvinen, 2004). The effect of TZDs on kidney function has been previously described in mice models (Fujii et al., 1997). Treatment with TZDs has been demonstrated to improve insulin sensitivity in patients with T2DM after a 3-month treatment, and to reduce albuminuria, the last effect likely mediated by the concurrent increase in serum adiponectin concentration (Miyazaki et al., 2007). These results were confirmed in a metaanalysis of 15 double-blind, randomized, clinical trials involving diabetic patients (Sarafidis et al., 2010) and in a large study involving 4351 diabetic patients (Lachin et al., 2011). Even though a meta-analysis reported an increase in cardiovascular mortality linked to the use of TZDs in dialysis patients (Nissen and Wolski, 2007), no definitive proof are available on the risk related to this medication in this population.
Another interesting class of hypoglycemic drugs with positive kidney outcomes are the SGLT2 inhibitors (SGLT2i) which inhibit glucose and sodium reabsorption in the PT (Ferrannini, 2017). These drugs have a renoprotective effect in patients with T2DM (Perkovic et al., 2019) independently of glycemic control (Cherney et al., 2017). The renal protective effect can also be attributed to altered hemodynamics, reduced inflammation and fibrosis as well as controlled blood pressure and weight loss (Williams et al., 2020). In rats treated with SGLT2i the glycemic improvement was accompanied by a decrease in insulin and lipid levels (Huang et al., 2020). Moreover, the actions of SGLT2i are associated with increased insulin sensitivity and decreased albuminuria (Cherney et al., 2017). Interestingly, Jaikumkao et al. observed that in an animal model of diet induced obesity characterized by IR and impaired renal function dapagliflozin treatment resulted in improved IR, renal function and renal insulin signaling (Jaikumkao et al., 2018).
CONCLUSION
Insulin is a hormone which acts not only on the most recognized insulin-responsive organs (liver, adipose tissue, and skeletal muscle), but also on the kidney. Moreover, the kidney has a primordial role in insulin clearance and may impact on insulin plasma levels. Whereas its main action is mainly related to homeostasis of glucose, including modulation of gluconeogenesis and lipolysis, in kidney the effects of insulin and IR change according to whether the target is in glomeruli or tubules. More specifically, if in glomerular podocytes insulin promotes glucose uptake, with an involvement in barrier permeability, in tubules it contributes to glucose reabsorption and gluconeogenesis regulation, and plays an important role in sodium homeostasis.
Importantly, insulin intervenes in albumin reabsorption at tubular level. Moreover, IR has been associated with microalbuminuria even in normoglycemia, and thus can be present regardless of diabetes diagnosis. These findings reinforce the kidney contribution to diabetes development and highlights insulin and albumin dynamics prior and regardless of the development of diabetes.
Insulin is cleared in the PT of kidney by two major routes, either by absorption of filtered insulin, or by post-glomerular capillary secretion, and only a minor amount appears to be excreted in urine. A decreased renal insulin clearance might lead to higher insulin levels, in the organ or systemically, favoring IR. Nonetheless, the impact of renal insulin clearance affection in the kidney or in insulin plasma levels still needs to be further unveiled.
It is clear now that kidney is not a mere target of insulin action, but insulin, more precisely IR, is also able to trigger CKD even in absence of diabetes. IR has been associated with prevalent CKD, rapid decline in renal function and microalbuminuria in the general population and in diabetic patients. In addition to being a risk factor for CKD onset and progression, IR is also involved in the increased cardiovascular risk in this population. However, if IR is responsible for the onset and progression of CKD, or if CKD is responsible for IR is still to be clarified. Preliminary confirmations come from clinical trials aiming at exploring the effect of TZDs, a class of oral diabetic medications, on kidney function. However, more focused studies, aiming also at testing the safety of these medications in CKD patients, are needed to better understand if treatment of IR may improve renal function in this population.
AUTHOR CONTRIBUTIONS
AP and MPM conceptualized the study. All authors originally drafted the manuscript, reviewed, edited, and critically approved the final version of the manuscript. | 2020-07-29T13:06:53.461Z | 2020-07-29T00:00:00.000 | {
"year": 2020,
"sha1": "8e772193a536760981d7ed6893a8825520d63527",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fcell.2020.00519/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8e772193a536760981d7ed6893a8825520d63527",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
31905742 | pes2o/s2orc | v3-fos-license | Recurrent Haemorrhage Due to Cerebral Arterio-Venous Malformation in Successive Pregnancies: a Rare Presentation
Introduction: The rupture of an intracranial Arteriovenous Malformation (AVM) in pregnancy is a rare occurrence, but may have fatal maternal and fetal consequences. Association between AVM rupture and pregnancy has been proposed which may be caused by increased cardiac output or circulatory effects of the elevated estrogen levels. The presentation during the pregnancy is usually as a result of Intracerebral Haemorrhage (ICH) following their rupture. Prognosis depends on the site, size, angio-architecture of AVM and extent of bleed. The risk of rebleed in an untreated symptomatic AVM is increased and mortality rate associated with the second bleed is 13% with the rate increasing to 20% for each subsequent haemorrhage. Till date, only a few reports have been published with a rebleed in cerebral AVM in pregnancy. Case report: We report a case of a 29 year old female who presented with ICH following ruptured occipital AVM in both her pregnancies. She was managed conservatively in her pregnancy followed by Gamma knife surgery in the postpartum period. Despite a rebleed in AVM, she had good maternal and fetal outcomes. *Corresponding Author: Dr Tanuja Muthyala, Senior Resident, Department of Obstetrics and Gynecology, Postgraduate Institute of Medical Education and Research, Chandigarh. India, Email: drtanujambbs@gmail.com Citation: Muthyala, T., et al. Recurrent Haemorrhage Due to Cerebral Arterio-Venous Malformation in Successive Pregnancies: a Rare Presentation. (2017) J Gynecol Neonatal Biol 3(2): 4448. J Gynecol Neonatal Biol | Volume 3: Issue 2 www.ommegaonline.org
Introduction
Intracranial Arteriovenous Malformations are formed from a vascular plexus of direct arterio-venous connections that lack an intervening capillary bed [1] . During pregnancy AVMs usually present with intra cerebral haemorrhage (ICH), headache, seizures or focal neurological deficit [2] . Elevated cardiac output, blood pressure and blood volume may predispose pregnant patients with AVMs to intracranial haemorrhage. The management plan needs to be individualised for each patient. Surgical intervention for ruptured AVM during pregnancy could prevent re-bleed however there is a risk of complications related to the pregnancy, possibility of preterm labour and intrauterine foetal demise. Literature suggests that radiation exposure to the developing foetus secondary to stereotactic radiosurgery and endovascular procedures are below the risk threshold [3] . Nonetheless, these procedures should be performed during pregnancy only when medically necessary. Conservative management till term followed by a definitive treatment in the postpartum period is a viable option in these patients [4] . Here we present a case of repeated rupture of an intracranial AVM in a pregnant patient during successive pregnancies.
Case report:
A 29 year old prime gravid presented to our hospital with severe headache with vomiting at 35 weeks of gestation. At the time of this admission her blood pressure was normal, fundus examination revealed papilloedema. Magnetic Resonance Imaging (MRI) revealed a left occipital lobe AVM with feeders from Posterior Communicating Artery (PCA) and an inter-hemispheric subdural haemorrhage. Neurosurgery consultation was sought. She was managed conservatively on anti-epileptics till term followed by an elective cesarean section under General Anesthesia (GA). Her postoperative period was uneventful. She was planned for intervention in the postpartum period but was lost to follow up.
She presented again to neurosurgery emergency during her second pregnancy at 22 weeks of gestation with history of loss of consciousness for 4 hours. Her Glasgow coma scale was E3V5M6. She was a febrile with a pulse rate of 86 beats per minute. Her blood pressure was 110 systolic and 80 mm of Hg diastolic. Her neurological examination was normal except for right homonymous hemianopia. On obstetric examination, gravid uterus was relaxed and corresponding to 20 weeks period of gestation. MRI was performed which showed an abnormal bunch of vessels the left occipital region suggestive of an AVM nidus. A T1 isointense, T2 iso to hyperintense ICH was seen in the left occipital region, surrounding the nidus, abutting the occipital horn. The hemorrhage was also extending into the ventricular system via the left lateral ventricle ( Figure 1). Her clinical and neurological status remained stable thereafter and as the patient was remote from term, she was managed conservatively. A noncontrast Computed Tomography of the head was performed one week later which revealed a resolving hematoma with no evidence of hydrocephalus. At the time of discharge, 10 days after ictus, her vitals were stable but homonymous hemianopia was persistent. She was explained about the risk of repeat hemorrhage and further neurological deterioration. Digital subtraction angiography (DSA) with embolization were planned in the postpartum period. She was again lost to follow-up despite counselling and subsequently reported to the antenatal clinic at 36 weeks period of gestation. She underwent elective repeat caesarean section under GA. Her peri-operative period was uneventful. DSA was performed after the caesarean section, which revealed a 1.7 x 1.3 x 1.2 centimetre occipital lobe AVM with feeders from left PCA. Prominent venous channels draining into the Superior Sagittal sinus and the straight sinus were seen (Figure 2). She was given option of gamma-knife radiosurgery versus embolization. The patient elected to undergo gamma knife surgery postoperatively. At follow-up visits, 2 and 6 months thereafter, she was doing well. Another MRI was performed 2 years after radiotherapy. This revealed resolution of the left occipital haematoma and reduction in the size of the nidus ( Figure 3).
Discussion
The association between AVM rupture and pregnancy is controversial due to relatively low incidence of cerebral AVM. According to Liu et al the rate of ICH related to cerebrovascular malformations is similar in pregnant and non-pregnant women and that pregnancy does not increase the rate of first cerebral haemorrhage from an AVM [5] . However, other authors have shown an increased risk of AVM rupture during pregnancy [3] . There is a 2 -3% annual risk of ICH with AVMs. ICH may be seen in 8% to 71% of patients with brain AVMs. The risk of rupture increases to 10 -30 % in the first year after initial haemorrhage. The location, size, morphology, presence of associated aneurysms and drainage patterns of the lesion affects the risk of rupture [1] .
AVM treatment paradigms vary significantly, and are often individualized. Unruptured AVMs in pregnancy generally warrant conservative management due to the low risk of rupture. However, when they present with rupture, surgery needs to be considered either during pregnancy or in the postpartum period due to an increased risk of re-hemorrhage and associated mortality. The relative risks and benefits to the mother and her foetus must be carefully weighed. Endovascular embolization is widely accepted as an important component of contemporary, multimodal therapy for AVM. Although rarely curative, embolization can facilitate subsequent surgical resection or radio-surgery [6] . In patients with high operative risk, inoperable lesions or those with stable neurological status, conservative management can be adopted during the pregnancy. Definitive therapy may be performed after delivery in such cases. Precautions are recommended during delivery and factors that influence the type of delivery include mode of management of the lesion (conservative versus operative) and maturity of the pregnancy at the time of ictus. Amias et al showed a bias in favour of caesarean section in 30 cases when neurological management was conservative [7] . Caesarean section could be a better option as vaginal delivery could precipitate Valsalvamanauver thus precipitate rupture of AVM [2,3,5] . Close collaboration with a team of neurologists, neurosurgeons, obstetricians and anaesthesiologists is mandatory.
Perquin DA et al in 1999 reported 3 women with intracranial AVMs. One of them carried two pregnancies after diagnosis. Caesarean sections were performed in both pregnancies which were uneventful. Another woman was 32 weeks pregnant when she presented with a haemorrhage due to a ruptured AVM. A she recovered well and was delivered by caesarean section. The third woman was 15 weeks pregnant when her symptoms occurred. There was a large intracerebral haematoma with rupture into the ventricular system. This patient expired later [8] . Nagamine N et al reported a 38-week-pregnant woman who developed an intracerebral hematoma secondary to a ruptured arteriovenous malformation. She underwent a caesarean section followed by craniotomy under general anaesthesia. The anaesthetic course was uneventful and resulted in good foetal and maternal outcome [9] . Jermakowicz et al reported a 23-year-old woman who presented with headache and visual disturbance after the rupture of a left parieto-occipital arteriovenous malformation in the 22 nd week of her pregnancy. After taking precautions to shield the foetus from radiation, endovascular embolization followed by open surgical resection of AVM were performed [6] . Three important issues have emerged from this case for clinical implication firstly the chances of Haemorrhage due to Cerebral AVM is increased in pregnancy. Conservative approach should be the first line of care in pregnancy if the patient is stable. Caesarean section should be the mode of delivery in this case. Gamma knife surgery should preferably be avoided in pregnancy.
Conclusion
Management of cerebral AVM in pregnancy is complex and needs to be individualized considering obstetric and neurological status. Treatment options are similar to that of a nongravid uterus and ruptured AVM require definitive treatment due to high risk of rebleed and associated fatal consequences. However conservative management during pregnancy even after bleeding may be an option if patient remain stable. Though potentially safe levels of fetal radiation exposure may be achieved during endovascular and SRS procedures in pregnancy, no radiation exposure should be considered unless medically necessary. Mode of delivery should be by caesarean section. Multidisciplinary team management involving neurologists, neurosurgeons, obstetricians, ophthalmologists', intervention radiologists and anaesthesiologists in a tertiary centre is mandatory to facilitate good outcome in such high risk pregnancies. | 2019-01-15T14:15:40.179Z | 2017-11-21T00:00:00.000 | {
"year": 2017,
"sha1": "c8332e90e7b9d252b2c22fd470480f2a2031dea6",
"oa_license": "CCBY",
"oa_url": "https://www.ommegaonline.org/articles/publishimages/14971-Recurrent-Haemorrhage-Due-to-Cerebral-Arterio-Venous-Malformation-in-Successive-Pregnancies-a-Rare-Presentation.pdf",
"oa_status": "HYBRID",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "3c4722e39d81c8f3e4780fa6c3758221c7deb071",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
248187731 | pes2o/s2orc | v3-fos-license | Working Capital Management on Financially Constrained Firm
: Working capital management is a daily activity that will determine the availability of resources for the company. There are positive and negative effects in increasing the degree of working capital. There is an optimal degree as well. The degree of optimal working capital for each company is different, depends on financial conditions. This paper aims to examine the effect of non-linear working capital on firm performance and analyze the differences in the degree of optimal working capital between financially constrained and unconstrained firms. Data were obtained from OSIRIS with the observation period 2010–2019. Working capital is proxied by the net trade cycle. This study uses panel data regression models, i.e. fixed effect regression and random effect regression. The result of this study shows that working capital has a non-linear effect (U-shaped inverted) on firm performance when using ROA and ROE as a proxy for performance which means the company has an optimal degree of working capital. The result also shows that financially constrained firms grouped based on cash flow, interest coverage, and cost of external financing have lower optimal working capital which means that the benefits of working capital are more used by non-financially constrained firms.
INTRODUCTION
There are three decisions determined by financial managers to maximize the wealth of shareholders. There are investment decisions, funding decisions, and working capital decisions. One of the decisions that need to be considered by managers is making working capital decisions and ensuring the company's operational activities run smoothly (Hanafi, 2013). Working capital decisions are significant for the company because working capital decisions directly affect the company's daily operational activities. Most of the activities of a company's financial manager are allocated to manage current assets because there are certain types of companies, for example, manufacturing, most of their assets are current assets.
There are two perspectives regarding the effect of working capital investment on firm value. The first perspective explained that high working capital can increase sales and get discounts to increase company value (Deloof,2003). Meanwhile, the second perspective states that high working capital requires funding and increases the cost.so it can increase the risk of bankruptcy (Kieschink, et.al 2011). Based on these two perspectives, it means that there is a trade-off in determining working capital decisions. In this condition, a firm is required to be able to determine optimal working capital. Optimal working capital is the limit to the extent to which working capital investment can have a positive effect on the company and if working capital investment exceeds the optimal degree, then working capital investment will harm the company. Financial factors such as the availability of capital costs, internal funds, and access to the capital market will affect investments made by working capital management. Banos -Caballero et al. (2013) argue that optimal working capital for financially constrained firms will be lower than financially unconstrained firms. The imperfect market hypothesis explains that external funding is not the best substitute for internal funding because it increases costs. That makes financially constrained firms more sensitive to working capital investments.
When increasing inventory, financially constrained firms tend to rely on account payables because they will face difficulties to access external funding so financially constrained firms cannot use a discount for early payments if they buy inventory in cash to a supplier. That will be different from financially unconstrained firms. They can get a discount if they buy inventory in cash from a supplier because of having sufficient cash. They can easily access external funding, such as bank loans. The discounts provided by these suppliers will certainly be higher than the interest in order to attract firms to pay in cash. So, the research problem is a trade-off in the working capital decision. Empirical evidence is needed to show that influence. In addition, there are differences in the amount of optimal working capital between financially constrained and unconstrained firms. This paper aims to (1) examine the influence of working capital on firm performance and (2) analyze the differences in the level of optimal working capital between financially constrained and unconstrained firms.
Working Capital
According to Hanafi (2016), working capital consists of cash, accounts receivable, and inventory. Besides, working capital also consists of current debt, including trade debt and accrual debt. In contrast, according to Brigham (604, 2014), working capital symbolizes the company's assets that can be converted into cash to support the company's operations. According to Brigham (604, 2014), working capital management relates to two fundamental questions. First, what is the right amount of working capital both the specific number of accounts and the total amount. Second, how should working capital be funded because working capital is related to funding decisions, so a financial manager must decide on the company's cash amount and the amount of short-term funding for working capital.
Firm Performance
Firm performance is the result obtained by the firms through actions and policies taken (Neely, 68, 2004). Performance is a term, generally used for part or all of the actions and activities of an organization in a period with references to the number of standards such as past or projected costs, with the basis of efficiency, accountability or management accountability and so on. Santos and Brito (2012), company performance is a result of the effectiveness of the company in carrying out its activities, both operational and financial.
Financial Constraints
Firms often face financing constraints on investing. Kaplan and Zingales (1997) suggest that financial constraints occur due to unfavourable financial liquidity and the difficulty of accessing external funding because they face the difference between capital costs from internal funding and capital costs from external funding. According to Fazzari et al. (1988), the existence of asymmetric information results in external funding costs being more expensive than internal funding, resulting in financially constrained companies' investment decisions tends to be more sensitive to liquidity than nonfinancially constrained companies. This results when the company gets an investment opportunity, the company will fund its investment by using internal funding then external funding because internal funding raises costs that are cheaper than external funding.
The Influence of Working Capital on Firm Performance
Higher investment in inventories and accounts receivable and inventory can improve company performance. Companies with a more extensive inventory can reduce costs and price fluctuations and prevent disruption of business processes (Blinder and Maccini, 1991). That also allows the firm to provide the best service for customers (Schiff and Lieber, 1974). Besides, giving credit can increase a company's sales, companies can provide discounts on customers effectively, attract customers to make purchases when demand is low, and improve long-term relationships between suppliers and customers. Customers can ensure the quality of products and services before payment so that it can reduce asymmetry information between the seller and the buyer. Emery (1984) argues that trade credit provision will be more profitable than investing in shortterm securities and can be used as a liquidity reserve to prevent future cash shortfalls.
However, there are some negative impacts on investment in working capital when it has passed the optimal amount. First, higher inventory can increase costs such as warehouse rental fees, insurance, and security (Kim and Chung, 1990). Second, high working capital requires higher funding as well. That can increase the financial cost and opportunity cost. In companies that provide high trade credit will increase credit risk. It can also lead companies to financial distress and face bankruptcy (Kieschnick et al., 2011). The existence of positive and negative impacts on working capital reflects tradeoffs in working capital decisions. Each company must have an optimal amount of working capital to balance costs and benefits to company performance. This increase in performance will be in line with the increase in working capital until the optimal amount of working capital is achieved. Conversely, when the amount of working capital exceeds the optimal amount, the relationship between working capital and performance will be negative (Banos-Caballero et al. 2013). Based on the following explanation, this study composes the hypothesis as follows. H1: There is an inverted U-shaped impact of working capital on firm performance.
The Impact of Working Capital on Financially Constrained Firm
Suppose there is an inverse U parabolic effect of working capital on company performance. In that case, the optimal amount of investment in working capital will be different for companies with financial constraints and those without financial constraints. That is caused by working capital investment decisions that are strongly influenced by investment opportunities. Financially constrained firms will have difficult access to the capital market and have fewer investment opportunities so they can only rely on internal funding. Since market imperfections have asymmetry information and agency costs, means to get external funding will increase higher costs than internal funding (Greenwald, et al. 1984). External capital is not a perfect substitute for internal capital. In line with this, Fazzari et al. (1988) show that company investment depends on financial factors such as internal funding availability, access to capital markets, or funding costs. Companies with lively working capital require more funding, so the optimal amount of working capital will be lower for companies with financial constraints. When increasing the degree of inventory, financially constrained firms tend to rely on account payables because they will find difficulties to access external funding. That causes financially constrained firms cannot use a discount facility if they buy inventory in early payments from a supplier. On the one hand, a higher degree of account payable can reduce the working capital. That will be different for financially unconstrained firms. They can use discount facilities if buying supplies in cash to suppliers because they have sufficient cash. Even though it turns out that they do not have enough cash, they can easily access external funding, such as bank loans. The discounts provided by these suppliers will certainly be more significant than the bank interest in order to attract companies to do early payment financially unconstrained firms are relatively more extensive and more mature. Companies with a greater internal funding capacity and higher access to capital markets tend to have higher working capital Hill et al. (2010). Based on the following explanation, this study composes the second hypothesis as follows. H2: Financially constrained firms will have a lower degree of optimal working capital than financially unconstrained firms.
METHODOLOGY
The sample in this study is non-financial companies listed on the Indonesia Stock Ex-change during the period 2010-2019. Sources of data in the study came from OSIRIS. This research is quantitative. Based on the hypothesis and research analysis model, the variables in this study consist of dependent variables, independent variables, company grouping variables with financial constraints, and control variables. The dependent variable in this study is firm performance which is measured by accounting performance. Accounting performance is proxied by ROA and ROE.
The independent variable in this paper is the working capital of the company. The net trade cycle is a proxy for measuring working capital. According to Banos-Caballero et al. (2013) and Shin and Soenen (1998), the net
Measurement Proxy Dividend Pay-out Ratio
If the firms have interest coverage above the median then they are financially unconstrained firms, and vice versa.
Dividend Pay-out Ratio: Dividend paid i,t / Net Income i,t
Cash Flow
If the firms have cash flow above the median then they are financially unconstrained firms, and vice versa.
Cash Flow i,t = (earning before interest taxes i,t +depreciation i,t ) /total asset i,t Interest coverage If the firms have interest coverage above the median then they are financially unconstrained firms, and vice versa.
Interest coverage i,t = EBIT i,t / Interest expenses i,t
Cost of external financing
Companies that have a dividend pay-out ratio above the median are classified as companies with financial constraints. Conversely, if the company has a dividend pay-out ratio below the median, it is classified as a company with financial constraints.
Cost of external financing i,t = financial expenses i,t / total debt i,t Table 1 Variable Measurements trade cycle is a proxy for measuring working capital. The net trade cycle is measured by the following formula. NTC i,t = (Account Receiveables i,t / Sales i,t )* 365 days + (Inventories i,t / Sales i,t ) *365 days -(Account payables i,t / Sales i,t )*365 day Furthermore, in this study the group of financially constrained firms or financially unconstrained firms based on dividends, cash flow, cost of external financing, and interest coverage. The score is 1 for a financially constrained firm and 0 financially unconstrained firm. The estimates of the models are as follows: Simplify the equation: Based equation (1), the optimal degree of working capital can be measured by determine the turning point obtained from: -E1/2E2 If there is maximum turning point form equation (1), then equation (2) can be estimated. The optimal degree of working capital can be measured by determine the turning point obtained from equation (2):
RESULT AND DISCUSSION
The sample in this study was a non-financial companies listed on the Indonesia Stock Exchange during the period 2010-2019. Data were obtained from OSIRIS database. During sample selection, firms that have negative equity and delisted were removed. Therefore, the final sample consisted of 435 firms. This work used a panel regression model to test the hy-pothesis. Using panel data allows us to minimize unobservable heterogeneity and eliminate the risk of obtaining biased (Greene, 2000). This work used a Hausman test to choose using a fixed-effect or random-effect model. Table 2 provides summary variables of this study.
Hypothesis 1 Testing
The results show a non-linear effect on company performance when using ROA and ROE as proxies for performance variables. That means that there is a non-linear effect of working capital on company performance. The NTC coefficient is significantly positive, and the coefficient of NTC2 has a significant negative value which means that the form of the effect is an u-shaped inverted curve that shows the maximum point. In line with hypothesis 1 shows that the firm has an optimal degree of working capital. The optimal degree of working capital can be calculated using the maximum point formula (-E1/2E2).
Hypothesis 2 Testing
The results show that the NTC coefficient is significantly positive and the coefficient of NTC2 has a significant negative value which unconstrained firms. That obtained from -E1/ 2E2 dan -(E1G1)/2(E2+G2). The results supported H2 that financially constrained firms will have a lower degree of optimal working capital than financially unconstrained firms.
CONCLUSION
The results show that working capital has a nonlinear effect (U-shaped inverted) on the performance when using ROA and ROE as performance measurement, which means there is a maximum point. The optimal degree of working capital for the ROA and ROE as performance is 271.23 days and 267.20 days, respectively. This study answers the contradictions of previous studies, which showed a positive effect of working capital on firm perfor-
Then further research can confirm the findings of whether a firm that has a low degree of working capital will increase its performance if working capital is extended and vice versa by the following research suggestions: 1. For further research, it can be used by separating the sample first based on the firm's working capital. 2. For further research, the optimal degree of working capital can be related to other factors such as the business life cycle or bargaining power. Besides, the results also show that there is an interaction impact between the financial constraints dummy and working capital when using financially constrained firms grouped based on cash flow, size, and dividend pay-out ratio using ROA and ROE as performance. The next step is calculating the optimal degree of working capital for financially constrained and unconstrained firms. The results show that financially constrained firms have lower optimal working capital than financially unconstrained firms. This shows that the benefits of working capital are more used by financially unconstrained firms.
This study has implications for management to consider the optimal degree of working capital. First, working capital can have a positive impact on firms performance, but when a firm has working capital that exceeds the optimal degree, it will harm firm performance. So | 2022-04-16T15:06:49.598Z | 2021-11-01T00:00:00.000 | {
"year": 2021,
"sha1": "d313f7c645baa722b57ff65a54325e9978a54e85",
"oa_license": "CCBYSA",
"oa_url": "https://journal2.unusa.ac.id/index.php/BFJ/article/download/1970/1513",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "b343f34a14e9a3f96ccf6609ddcca2b16583824b",
"s2fieldsofstudy": [
"Economics",
"Business"
],
"extfieldsofstudy": []
} |
252647987 | pes2o/s2orc | v3-fos-license | Takotsubo cardiomyopathy in a female presenting with status asthmaticus: a case report and review of literature
Background Takotsubo cardiomyopathy (TCM) is a non-ischemic syndrome characterized by transient acute left ventricular dysfunction as evident on transthoracic echocardiography. It can often mimic myocardial ischemia and is characterized by the absence of angiographic evidence of obstructive coronary artery disease. Reports of Takotsubo syndrome in elderly with asthma exacerbations have been noted. Case presentation We describe a case of TCM in a 68-year-old female who presented with acute shortness of breath secondary to status asthmaticus. Her electrocardiogram showed ST segment elevations in multiple coronary artery distributions and mildly elevated troponin levels. Coronary angiography showed no significant stenosis of the coronary arteries with left ventriculography that showed systolic apical ballooning with a 10% ejection fraction, consistent with TCM. Conclusions Takotsubo syndrome should be considered in the differential diagnosis of patients presenting with status asthmaticus and elevated troponin levels on admission. Patients should be asked about the use of beta agonist prior to admission. A thorough literature review including a summary of 11 previously published case reports of TCM with acute asthma exacerbations has been presented. Supplementary Information The online version contains supplementary material available at 10.1186/s43044-022-00310-9.
Background
Takotsubo cardiomyopathy (TCM) is an acute cardiac dysfunction that typically represents hypokinesis of the apical segment of the left ventricle (LV) beyond a single coronary artery territory [1]. It presents with typical features of myocardial ischemia including central tightening chest pain and shortness of breath, electrocardiographic changes which mimic coronary artery disease, and minimal release of myocardial enzymes in the absence of angiographically significant coronary artery stenosis. The condition is relatively rare and is found in about 1-2% of all patients with suspected acute coronary syndromes (ACS) [2]. It is commonly seen among postmenopausal women [3]. The underlying pathophysiology of this acute cardiac entity remains largely unclear, but is often associated with physical or emotional stress, hence the term "stress cardiomyopathy" [4]. Catecholamine surge-induced cardiomyocyte injury in response to emotional or physical stress has been postulated as its primary pathophysiology [5]. The list of potential triggers associated with TCM continues to expand as the disease gains increasing recognition among internists and cardiologists. The condition is usually benign, and reversible, with nearly full recovery expected in around 6-8 weeks [6]. Rarely, it may be complicated by life-threatening sequelae including acute cardiogenic shock, lethal ventricular arrhythmias, or ventricular wall rupture [7]. In this report, we present a case of TCM secondary to asthma exacerbation along with a review of the published case reports (Table 1) [8][9][10][11][12][13][14][15][16][17][18]. This report emphasizes the importance of the awareness of the potential association between status asthmaticus and TCM.
Case presentation
A 68-year-old female presented to the Emergency Department with chief complaint of shortness of breath, wheezing, and cough for past 3 days. Prior to the hospital visit, she was seen by her primary care provider who prescribed her clarithromycin and Medrol dose pack but she had no significant benefit. She denied any episodes of chest pain or fever.
Her medical history was significant for asthma since childhood, hypertension, hypothyroidism, and left lower extremity deep venous thrombosis 15 years back. The patient reported an eight-pack-year smoking history. There was no personal history of coronary artery disease, prior myocardial infarction, or hyperlipidemia. She had no history of unusual emotional stress prior to admission.
On examination, patient was afebrile with temperature of 36.3 °C, heart rate of 116 beats/minute, blood pressure of 102/75 mm Hg, and respiratory rate of 22 breaths/minute. Bilateral wheezes were heard on chest auscultation. In the emergency room, patient was placed on bilevel positive airway pressure (BiPAP) to maintain oxygen saturation. Appropriate management for asthma exacerbation was started and patient received methylprednisolone 125 mg IV one time along with repeated albuterol nebulizers, Ipratropium-albuterol inhalers, and magnesium sulfate. Initial Chest radiograph revealed mild flattening of diaphragm without any acute process, cardiomegaly or effusions. Electrocardiogram (EKG) on admission showed mild ST elevation in the inferior and anterolateral leads as well as a prolonged QT interval (0.486 s) (Fig. 1A). Initial troponins at the time of admission were normal (0.023 ng/ml). Her pro-BNP was 103 pg/ml. Her erythrocyte sedimentation rate was 7 mm/h and her C-reactive protein was 1.34 mg/l. Patient was transferred to critical care unit for further management, on BiPAP and was treated for acute asthma exacerbation. A repeat EKG obtained 1.5 hours later showed Left axis deviation but no significant changes from the initial EKG (Fig. 1B). Troponins were trended overnight and continued to rise but were moderately elevated (3.97 ng/ml and 4.65 ng/ml). An EKG the following morning 18 hours later showed a new onset Left bundle branch block (LBBB) (Fig. 1C). She came off the BiPAP and was transitioned to Nasal cannula, was feeling improved the morning of the Day 2. That evening, she started to deteriorate again requiring BiPAP.
An echocardiogram (ECHO) was obtained at this time, which revealed an Ejection fraction (EF) of 24%, with severely depressed LV function but normal wall thickness and shape (Additional file 1: Fig. S1). A prior ECHO done eight months back showed an EF of 65%, without any regional wall motion abnormalities. Due to her worsening respiratory status, new EKG changes from that morning and increasing troponin, she was taken to the Cardiac Catherization lab urgently. Cardiac catheterization revealed normal coronary angiogram, an EF of 10%, and Left Ventricular End Diastolic Pressure of 35 mm Hg. A pro-BNP resulted that time was 20,242 pg/ml. Left ventriculogram showed akinetic apex, anterior and inferior walls with mild dilation of the left ventricle supporting the diagnosis of TCM (Fig. 2). An intra-aortic balloon pump (IABP) was placed for circulatory support, however vasopressors and inotropes were not required.
The patient was started on carvedilol and enalapril and she was monitored in the critical care unit until the IABP was removed. The rest of the hospital course was unremarkable on the medicine floors and she was discharged after being weaned off supplemental oxygen. She opted out of anticoagulation. Follow-up ECHO 9 weeks later revealed normal left ventricular function with EF of 59%.
Discussion
We presented a case of TCM in a 68-year-old female presenting with asthma exacerbation. Within 24 hours of presentation, she developed a new onset LBBB on EKG and had mild troponin elevation. Her respiratory status improved the morning of the day 2, but then deteriorated again the night of day 2. An ECHO showing low EF, increasing troponin, continuing ST changes, and worsening respiratory status prompted an emergent cardiac catheterization for high-risk NSTEMI. The cardiac catheterization revealed findings that were highly suspicious of Takotsubo syndrome.
Currently, the diagnostic criteria adopted by the Heart Failure Association and European Society of Cardiology are the most widely used in the diagnosis of TCM. In addition to the above definition, criteria includes an absence of atherosclerotic disease that would explain the presentation, new and reversible EKG changes (ST elevation/depression, LBBB, and/or QTc prolongation), elevated Brain natriuretic peptide (BNP) or natriureticpro-BNP, slightly elevated troponins and recovery of the ventricular function on imaging at 3 to 6 months. The International Takotsubo Diagnostic Criteria has been recently proposed and is similar to the criteria put forth by the Heart Failure Association and European Society of Cardiology with the exception that infectious myocarditis must be ruled out [19]. Myocarditis requires a biopsy for a definitive diagnosis; however a biopsy for this purpose is not commonly done in our facility. Clinically, myocarditis presents with chest pain, which our patient did not report. Other findings suggestive of myocarditis include global dilation on ECHO, elevated inflammatory markers (ESR, C-reactive protein, etc.), and non-specific changes on EKG, which were not seen in our patient. A LBBB is associated with a worse prognosis in patients diagnosed with myocarditis. Given its similarity in presentation to ACS, it is imperative to differentiate TCM from ACS. There are certain findings on EKG and imaging that can help differentiate TCM from ACS, although are not definitive. One finding on EKG that is 100% specific for TCM is ST elevation in AVR combined with ST elevation in more than 2-3 anteroseptal leads (V1-V6). This finding has been reported to have a positive predictive value of 100%, a negative predictive value of 52%, and a sensitivity of 12% [20]. Another finding that may suggest TCM as opposed to ACS is the differences in elevation of cardiac troponins. In TCM, troponins are typically elevated less than 10 ng/ml. BNP and natriuretic-BNP are typically much higher in TCM as compared to ACS. An ECHO may be able to demonstrate the classic apical ballooning form, however, it cannot be differentiated from ischemia involving the apex. Cardiac magnetic resonance imaging has also been used to differentiate TCM from ACS [21]. Delayed gadolinium enhancement, which helps to demonstrate myocardial edema, tends to match the area of akinesis in TCM during the acute phase of illness while it is more localized to the involved coronary artery in ACS.
The pathophysiology of asthma-induced TCM is not well understood. It is hypothesized that the asthma attack itself leads to excessive production of catecholamines and Neuropeptide Y, and it may trigger apical cardio-depression precipitating TCM [22]. However, it is not clear if elevated levels of these substrates are a reflection of the pathophysiology or if this is the cause. However, direct exposure of myocardium to catecholamine is deemed to be highly toxic and may cause cellular damage, altered cellular metabolism, and contraction band necrosis [23,24]. Further, there is evidence to suggest that selective beta-2-agonists may induce apoptosis of cardiac myocytes due to production of reactive oxygen species [25,26]. Other possible mechanisms proposed include the possibility of increased beta receptors in certain areas of the myocardium (9), changes in G-protein signaling from Gs (stimulatory) to Gi (inhibitory) leading to negative inotropic response (11), and in the event of increasing beta agonist use, the loss of selectivity of these drugs to lung tissue leading to vascular spasm in the myocardium (13).
The main complication of TCM is cardiogenic shock resulting in circulatory failure. The management is supportive, with severe cases requiring Mechanical Circulatory Support with Bridge to Recovery. It is important to determine if a left ventricular outflow tract obstruction (LVOTO) is present (as determined by ECHO). In the case of LVOTO, the obstruction must be relieved prior to treating the heart failure. Given risk of arrhythmias, continuous EKG monitoring is recommended for at least 48 h.
We performed a review of literature and found 11 other cases that describe asthma exacerbation as a trigger for TCM (Table 1). Six out of the 11 cases we reviewed were post-menopausal females. None of the cases, including ours, exhibited the specific EKG finding for this syndrome. However, there is only one other case to our knowledge that presented with a new LBBB [13]. All cases were treated with a loading dose of steroids and inhaled bronchodilators for management of asthma exacerbation. In our study, additionally, repeated doses of magnesium sulfate were used for the management of status asthmaticus. Other notable treatments used in prior reports included combinations of epinephrine/ketamine and aminophylline. Our case presented with similar troponin elevations, ECHO findings, and cardiac catheterization findings as seen in previous reports.
We suspect that our case of TCM was induced by the physiological stress of an asthma exacerbation or possibly by the administration of beta agonistic drugs. However, the former seems more likely as our patient had EKG changes (ST Elevations in the inferior and anterior-lateral leads) on admission that were not present on a prior EKG and she did not report any chronic and excessive use of inhaled beta agonists. We came to the conclusion that her asthma exacerbation lead to the development of TCM immediately on or prior to admission. The new LBBB on day 2 prompted the cardiac catheterization. The ST elevations and LBBB resolved completely after the catheterization.
Conclusions
This case highlights the importance of maintaining a high suspicion for TCM in post-menopausal women with asthma exacerbations, who have persistent symptoms despite treatment with bronchodilators and steroids, even in the absence of an emotional trigger. | 2022-10-02T14:21:19.737Z | 2022-10-01T00:00:00.000 | {
"year": 2022,
"sha1": "66b58e33d4a61dd52ab1a7f11829abe61258a064",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Springer",
"pdf_hash": "66b58e33d4a61dd52ab1a7f11829abe61258a064",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
253137461 | pes2o/s2orc | v3-fos-license | An Experimental Analysis of the High-Cycle Fatigue Fracture of H13 Hot Forging Tool Steels
In this study, the axial fatigue behaviour of hot forging tool steels at room temperature was investigated. Fatigue tests were performed on two steels within the same H13 specification. The fatigue tests were carried out in the high-cycle fatigue domain under normal conditions. These tests were also performed on specimens in contact with a corrosive medium, applying stress values that led to the high-cycle fatigue domain under normal conditions for the sake of comparison. Both materials showed similar fatigue strengths when they were tested under normal conditions. In contrast, corrosion fatigue lives were much lower than in normal tests and differed significantly between the two steels. Crack initiation was triggered by microstructural and surface defects in the normal tests, whereas the formation of corrosion pits caused crack initiation in the corrosion fatigue tests. Moreover, a fracture surface analysis revealed dissimilar crack propagation areas between both steels, which suggested that both steels had different fracture toughness. These results were in line with the differences observed between the carbide and grain sizes of both of the material microstructures.
Introduction
Hot forging tool steels present excellent mechanical properties, such as extraordinary yield and tensile strengths even at high temperatures [1,2] and a fairly good balance between hardness and fracture toughness when they are subjected to an appropriate thermal treatment [3]. Hence, these steels are extensively used for manufacturing forging dies, pressure die casting tools, and extrusion tools [4]. However, in spite of their outstanding mechanical properties, these steel tools suffer from severe damage when they are subjected to thermal [5], mechanical [6], and tribological [7] fatigue loads, which cause the initiation and propagation of cracks and eventually the fracture of these tools [8]. The fracture of hot forging tools is a crucial issue for many industrial sectors since it requires large annual investment in maintenance and tool replacement [9]. In order to reduce these costs and maximise the replacement times of forging tools, a fundamental understanding of the fatigue strength and the fatigue fracture mechanisms of hot forging tool steels is required.
The fatigue behaviour of hot forging tool steels depends strongly on their mechanical properties, especially on the hardness and the fracture toughness. The surface hardness of a material improves its fatigue strength, as a higher hardness results in a higher resistance to the local plastic deformation required to open a crack [10]. Since fatigue cracks usually start at the surface of a component, a high surface hardness delays crack initiation and thus extends the fatigue life of the component [11]. On the other hand, the fracture toughness provides the resistance to support loads in the presence of cracks or defects, so a higher fracture toughness improves the fatigue crack propagation behaviour of the material.
The mechanical properties of hot forging tool steels can be varied substantially by modifying the variables of the heat treatment they are subjected to. The heat treatment of these steels often consists of a quenching stage followed by two or even three tempering stages. The effects of the quenching stage are controlled by the selection of an austenitising temperature. An increase in the austenitising temperature promotes grain growth, which causes a decrease in the yield strength (YS) and ultimate tensile strength (UTS) of the steel according to the Hall-Petch equation [12]. Moreover, higher austenitising temperatures also favour the dissolution of primary carbides that precipitate during the tempering stages as smaller secondary carbides, thus enhancing the hardness of the steel [13]. As for the tempering stages, it has been observed that higher tempering temperatures result in lower values of the UTS and the hardness, whereas the fracture toughness is improved [14]. During tempering, a softening process takes place where the martensite microstructure decomposes and carbides precipitate [4]. The presence of smaller carbides and their homogeneous distribution is beneficial for the fracture toughness of the steels [3].
Most previous investigations into the fatigue behaviour of hot forging tool steels have aimed to characterise the performance of these alloys in the low-cycle fatigue (LCF) domain [15,16]. However, the results of the high-cycle fatigue (HCF) characterisation of hot forging tool steels remain scarce in the literature. Shinde et al. [17] performed rotating bending HCF tests on conventionally heat-treated H13 steel. As a result of these tests, a fatigue strength of 566 MPa was obtained at 10 7 cycles. Likewise, Korade et al. [18] evaluated the fatigue behaviour of H21 steel using rotating bending HCF tests up to 10 7 cycles. The resultant fatigue strength for conventionally heat-treated specimens was found to be 560 MPa.
Work environments can worsen the fatigue behaviour of steels by causing corrosion damage. For instance, Papageorgiou et al. [19] analysed the failure mechanisms of an H13 steel hot forging die that failed earlier than expected during its operation. This failure analysis revealed that the cooling agent utilised in the working process of this die presented a high salt concentration and caused a severe corrosion attack, which increased the surface roughness of the die and thus accelerated the damage process. The corrosion fatigue phenomenon can be triggered by many factors, such as contact with water, condensation of moist air, or adsorption of gases [20]. Therefore, understanding the interaction between corrosive environments and fatigue performance is also important to guarantee a high durability of hot forging tool steel components.
Numerous investigations have dealt with the corrosion fatigue analysis of various types of steels. Yongmei et al. [21] performed axial fatigue tests on maraging steel specimens within a chamber filled with a NaCl solution. The maraging steel showed a dramatic decrease in the number of cycles to failure in the tests carried out in saline bath tests compared to conventional fatigue tests. Amongst these results, it was also observed that a lower loading frequency and a lower stress ratio decreased the fatigue lives of the specimens. Ebara [22] analysed the effect of the corrosive medium concentration on the corrosion fatigue results of a stainless steel, observing that a higher concentration of NaCl in the corrosive medium affected fatigue strength and crack growth rates negatively. May et al. [23] evaluated the failure mechanisms of a martensitic stainless steel submerged in a saline solution. The origin of cracks was attributed to the rupture of local passive films in the surface of the tested specimens, leading to an increased corrosion attack that favoured crack initiation. Nevertheless, no results regarding the corrosion fatigue of hot forging tool steels have yet been presented in the literature.
The purpose of this work was to analyse the fatigue behaviour and fracture mechanisms of AISI H13 steel, which is one of the most common hot forging tool steels. This analysis included a study of the dissimilar effects of the microstructure on the results Materials 2022, 15, 7411 3 of 20 of the corrosion fatigue (CF) tests when compared with the results of the conventional fatigue tests. In order to achieve this goal, a microstructural analysis was performed on H13 samples from two different manufacturers that underwent analogous heat treatments. Hardness and tensile tests were carried out on both steels, since mechanical properties have an important influence in the fatigue behaviour. Next, axial fatigue tests were performed at room temperature in the HCF domain under normal conditions. Using a saline solution as a corrosive medium and applying the same stress values that led to HCF failure under normal conditions, CF tests were carried out. Finally, fracture surfaces were analysed to identify the mechanisms that led to the failure of the specimens.
Materials
The materials analysed here were two hot forging tool steels from different manufacturers within the same AISI H13 specification. H13 steels were selected to carry out this research as this specification is one of the most utilised for hot forging tool manufacturing. The H13 designation is also referred to in the literature as 1.2344 or SKD61 [24]. The chemical compositions in weight percentages of both the H13 steels, namely A and B, are displayed in Table 1. It can be observed that both steels have considerably similar chemical compositions and that H13 steels have high contents of chromium, molybdenum, and vanadium. The content of chromium increases the resistance to oxidation and high temperatures, whereas molybdenum improves the hardenability of the steel and vanadium enhances its strength and toughness [25]. Both H13 steels were manufactured using an electro-slag remelting (ESR) process. The ESR process has been demonstrated to be an effective process for obtaining highstrength steels with a high degree of cleanliness, smaller inclusions, and enhanced fatigue strengths [26]. According to the assessment of the statistical influence of the ESR process on the fatigue behaviour of H13 steels, it was proved that fatigue lives were significantly enhanced for the refined steel specimens [27].
The H13 steels analysed in this study were obtained from actual crankshaft forging dies. These dies were austenitised, quenched in water, and then tempered twice to attain a hardness of 46 HRC. The A-steel forging die was austenitised to 1025 • C for 2.5 h and tempered twice at temperatures of 570 • C and 605 • C for 4 h and 5 h, respectively. Similarly, the B-steel forging die was austenitised to 1020 • C for 2.5 h and tempered twice for 4 h and 5 h at a temperature of 600 • C. The heating procedure during the stages of the heat treatments was adapted in order to ensure a homogeneous distribution of temperature in the whole of the forging dies. Such recommended heat treatment procedures for H13 steel can be found in NADCA no. 229.
Specimen Dimensions
Steel blocks were obtained from a crankshaft forging die, as represented in Figure 1a. These forging dies were already subjected to the heat treatment specified previously. Next, the specimens were machined from the steel blocks as per the geometry shown in Figure 1b, where the dimensions are in millimetres. The main axis of the specimens was aligned with the main axis of the crankshaft. These specimens were used for tensile and axial fatigue testing. Figure 1b, where the dimensions are in millimetres. The main axis of the specimens was aligned with the main axis of the crankshaft. These specimens were used for tensile and axial fatigue testing. It should be pointed out that the specimen size can also affect the fatigue test results, as it raises the likelihood of presenting more and larger microstructural defects that may lead to premature crack initiation [28]. Multiple studies have been carried out where larger specimens showed significantly lower fatigue lives than smaller ones under the same loading conditions [29].
It is well known that surface roughness weakens fatigue strength as rough profiles present sharp stress concentrators that lead to premature crack initiation [30]. Therefore, the fatigue specimens were polished in order to remove any surface defects that may have originated throughout the manufacturing stage and to obtain a surface roughness below Ra = 0.2 µm.
Characterisation Methods
A microstructural analysis of the two studied steels was carried out on a JEOL JSM-6010 LA scanning electron microscope (SEM). This analysis was complemented with Xray diffraction (XRD) to compare the diffraction patterns of both of the steels and to find any potential difference between their microstructures. The XRD analysis was performed using a PANalytical X'Pert Pro X-ray diffractometer by means of monochromatic Cu-Kα radiation (wavelength 1.54 Å) over the 2θ range of 25-130° with a step size of 0.02°. In order to enable a clear observation of the microstructure, samples of both of the steels were saw-cut using a refrigerant consisting of a mixture of oil and water so that heating and microstructural changes were prevented. Next, the steel samples were ground with silicon carbide sandpaper using water as a lubricant and then mirror-polished with 6 μm and 1 μm diamond suspensions. Finally, the samples were etched at room temperature with a Nital 5% solution for 22 s. Surface microhardness measurements were taken from the same samples that were used for the microstructural analysis using an HMV-G21 Series micro Vickers hardness tester. The applied load was 4.903 N (500 g) for 10 s. The Vickers hardness test was repeated 10 times for each steel and the hardness was obtained as the average of all the measurements.
Tensile tests were performed using a universal testing machine with a load capacity of 250 kN in order to obtain the UTS and YS values for each steel. These tests were displacement-controlled at a rate of 5 mm/min until fracture of the specimens. The UTS was calculated as the maximum load reached during the test over the initial cross-section area of the specimen, whereas the YS was obtained as the normal stress that corresponded to a plastic normal strain of 0.2%. It should be pointed out that the specimen size can also affect the fatigue test results, as it raises the likelihood of presenting more and larger microstructural defects that may lead to premature crack initiation [28]. Multiple studies have been carried out where larger specimens showed significantly lower fatigue lives than smaller ones under the same loading conditions [29].
It is well known that surface roughness weakens fatigue strength as rough profiles present sharp stress concentrators that lead to premature crack initiation [30]. Therefore, the fatigue specimens were polished in order to remove any surface defects that may have originated throughout the manufacturing stage and to obtain a surface roughness below Ra = 0.2 µm.
Characterisation Methods
A microstructural analysis of the two studied steels was carried out on a JEOL JSM-6010 LA scanning electron microscope (SEM). This analysis was complemented with X-ray diffraction (XRD) to compare the diffraction patterns of both of the steels and to find any potential difference between their microstructures. The XRD analysis was performed using a PANalytical X'Pert Pro X-ray diffractometer by means of monochromatic Cu-Kα radiation (wavelength 1.54 Å) over the 2θ range of 25-130 • with a step size of 0.02 • . In order to enable a clear observation of the microstructure, samples of both of the steels were saw-cut using a refrigerant consisting of a mixture of oil and water so that heating and microstructural changes were prevented. Next, the steel samples were ground with silicon carbide sandpaper using water as a lubricant and then mirror-polished with 6 µm and 1 µm diamond suspensions. Finally, the samples were etched at room temperature with a Nital 5% solution for 22 s.
Surface microhardness measurements were taken from the same samples that were used for the microstructural analysis using an HMV-G21 Series micro Vickers hardness tester. The applied load was 4.903 N (500 g) for 10 s. The Vickers hardness test was repeated 10 times for each steel and the hardness was obtained as the average of all the measurements.
Tensile tests were performed using a universal testing machine with a load capacity of 250 kN in order to obtain the UTS and YS values for each steel. These tests were displacement-controlled at a rate of 5 mm/min until fracture of the specimens. The UTS was calculated as the maximum load reached during the test over the initial cross-section area of the specimen, whereas the YS was obtained as the normal stress that corresponded to a plastic normal strain of 0.2%.
A Walter-Bai LFV-25 servo hydraulic dynamic test machine was used to perform the fatigue tests. This machine has a dynamic load capacity of ±25 kN, a maximum stroke of ±50 mm, a frame stiffness of 200 kN/mm, and a servo actuator accuracy of ISO 7500 class 0.5. This machine allows one to hold specimens with steel clamps hardened up to 60 HRC, which makes it suitable for testing hot forging tool steels. Load-controlled axial fatigue tests were carried out at room temperature according to the ISO 1099 standard [31].
In all the tests, the load profile was sinusoidal with a loading frequency of 10 Hz and a stress ratio equal to zero. In order to study the CF behaviour of these steels, the mid-section of the specimens was placed in contact with a sponge soaked with a corrosive medium throughout additional fatigue tests. The setup of these corrosion fatigue tests is shown schematically in Figure 2. The corrosive medium was a sodium chloride (NaCl) solution with a concentration of 0.1 M. All the specimens were tested at different load ranges up to either specimen failure or 10 7 loading cycles (run-out). In the case of specimen failure, the output of the test was the number of cycles to failure. The load values applied to the specimens were selected to obtain fatigue lives within the HCF domain (10 4 -10 7 cycles) under normal conditions. The same load values were applied to the specimens subjected to CF tests for the sake of comparison. The least squares method was applied for the data fitting of the fatigue results to obtain the Wöhler diagrams. As a result of all the combinations of the type of material (A-type and B-type steels) and test conditions (normal HCF and CF), four Wöhler diagrams were considered.
A Walter-Bai LFV-25 servo hydraulic dynamic test machine was used to perform the fatigue tests. This machine has a dynamic load capacity of ±25 kN, a maximum stroke of ±50 mm, a frame stiffness of 200 kN/mm, and a servo actuator accuracy of ISO 7500 class 0.5. This machine allows one to hold specimens with steel clamps hardened up to 60 HRC, which makes it suitable for testing hot forging tool steels. Load-controlled axial fatigue tests were carried out at room temperature according to the ISO 1099 standard [31]. In all the tests, the load profile was sinusoidal with a loading frequency of 10 Hz and a stress ratio equal to zero. In order to study the CF behaviour of these steels, the mid-section of the specimens was placed in contact with a sponge soaked with a corrosive medium throughout additional fatigue tests. The setup of these corrosion fatigue tests is shown schematically in Figure 2. The corrosive medium was a sodium chloride (NaCl) solution with a concentration of 0.1 M. All the specimens were tested at different load ranges up to either specimen failure or 10 7 loading cycles (run-out). In the case of specimen failure, the output of the test was the number of cycles to failure. The load values applied to the specimens were selected to obtain fatigue lives within the HCF domain (10 4 -10 7 cycles) under normal conditions. The same load values were applied to the specimens subjected to CF tests for the sake of comparison. The least squares method was applied for the data fitting of the fatigue results to obtain the Wöhler diagrams. As a result of all the combinations of the type of material (A-type and B-type steels) and test conditions (normal HCF and CF), four Wöhler diagrams were considered. Fracture surface images were obtained using a Nikon SMZ1000 microscope and a JEOL JSM-6010 LA SEM.
Microstructure
The samples of both of the steels were taken using SEM analysis. Microstructure images of these samples are displayed in Figure 3 at different magnifications. These samples were mirror-polished and etched with an acid solution (Nital 5%) in order to obtain a clear observation of the microstructures. Fracture surface images were obtained using a Nikon SMZ1000 microscope and a JEOL JSM-6010 LA SEM.
Microstructure
The samples of both of the steels were taken using SEM analysis. Microstructure images of these samples are displayed in Figure 3 at different magnifications. These samples were mirror-polished and etched with an acid solution (Nital 5%) in order to obtain a clear observation of the microstructures. As can be seen in Figure 3a,b, the grain boundaries of the steels were revealed due to the acid attack. Furthermore, these images suggest that the grain size in the B-type steel was slightly smaller than that in the A-type steel. In Figure 3c,d, the typical martensitic laths with intermetallic carbides can be observed for both materials. Some grain boundaries are highlighted with dashed lines for clarity. The microstructural features of both materials resembled each other to some extent. However, the A-type steel showed a higher volume of carbides than the B-type steel. These carbides were significantly more abundant As can be seen in Figure 3a,b, the grain boundaries of the steels were revealed due to the acid attack. Furthermore, these images suggest that the grain size in the B-type steel was slightly smaller than that in the A-type steel. In Figure 3c,d, the typical martensitic laths with intermetallic carbides can be observed for both materials. Some grain boundaries are highlighted with dashed lines for clarity. The microstructural features of both materials resembled each other to some extent. However, the A-type steel showed a higher volume of carbides than the B-type steel. These carbides were significantly more abundant and coarser, and they adopted grain boundaries as their preferential position. The images shown in Figure 3e,f allowed us to confirm that the A-type steel was richer in primary carbides, whereas the B-type steel presented a lower number of primary carbides but a higher content of secondary carbides. The primary carbides had round and polygonal shapes, with sizes between 100 and 500 nm, whilst the secondary carbides were significantly smaller. The usual carbides observed in hot forging tool steels are chromium (M 7 C 3 and M 23 C 6 ), molybdenum (M 6 C), and vanadium (M 8 C 7 and MC) [32].
As a result of the microstructural analysis, differences between the two studied steels were observed. A possible reason for the bigger grain size of the A-type steel could be due to the slightly higher austenitising temperature that this steel was subjected to. However, since the discrepancy between the austenitising temperatures of both steels was minimal, it seems more likely that different cooling rates were applied to the steels during quenching. Increasing the quenching cooling rate would cause a decrease in the average grain size, as well as a lower fraction and size of carbides. The reason for this phenomenon could be due to the insufficient time given to the carbides to coarsen at high cooling rates [33]. Therefore, it is likely that the B-type steel was subjected to a higher cooling rate than the A-type steel during the quenching stage of the heat treatment.
XRD
XRD analysis was carried out to identify the phases in both of the materials and to detect any potential differences between their microstructures. The results of the XRD analysis are displayed in Figure 4, where the crystallographic plane indexes are indicated for each diffraction peak. It was observed that both steels presented a characteristic diffractogram of a martensitic microstructure, with no signs of the austenitic phase [34,35]. Furthermore, the positions of the diffraction peaks matched perfectly, which unequivocally proved that both steels were constituted by identical phases.
Materials 2022, 15, x FOR PEER REVIEW and coarser, and they adopted grain boundaries as their preferential position. The i shown in Figure 3e,f allowed us to confirm that the A-type steel was richer in pr carbides, whereas the B-type steel presented a lower number of primary carbides higher content of secondary carbides. The primary carbides had round and poly shapes, with sizes between 100 and 500 nm, whilst the secondary carbides were s cantly smaller. The usual carbides observed in hot forging tool steels are chromium and M23C6), molybdenum (M6C), and vanadium (M8C7 and MC) [32].
As a result of the microstructural analysis, differences between the two studied were observed. A possible reason for the bigger grain size of the A-type steel could to the slightly higher austenitising temperature that this steel was subjected to. How since the discrepancy between the austenitising temperatures of both steels was mi it seems more likely that different cooling rates were applied to the steels during qu ing. Increasing the quenching cooling rate would cause a decrease in the average size, as well as a lower fraction and size of carbides. The reason for this phenomenon be due to the insufficient time given to the carbides to coarsen at high cooling rate Therefore, it is likely that the B-type steel was subjected to a higher cooling rate th A-type steel during the quenching stage of the heat treatment.
XRD
XRD analysis was carried out to identify the phases in both of the materials detect any potential differences between their microstructures. The results of the analysis are displayed in Figure 4, where the crystallographic plane indexes are ind for each diffraction peak. It was observed that both steels presented a characteris fractogram of a martensitic microstructure, with no signs of the austenitic phase [ Furthermore, the positions of the diffraction peaks matched perfectly, which une cally proved that both steels were constituted by identical phases.
Hardness
Hardness measurements were obtained through Vickers tests for both of the The hardness of the A-type steel was 455 HV, and for the B-type steel it was 458 HV combined standard uncertainties of 6 HV and 11 HV, respectively. Both materials sh
Hardness
Hardness measurements were obtained through Vickers tests for both of the steels. The hardness of the A-type steel was 455 HV, and for the B-type steel it was 458 HV, with combined standard uncertainties of 6 HV and 11 HV, respectively. Both materials showed mean values considerably close to 46 HRC (≈460 HV), as specified. No significant distinctions were observed between the measurements performed near the surface and in the interior of the samples.
Both of the steels presented an equal hardness despite the different grain sizes and carbide distributions observed in the microstructural analysis. As a result of their equal hardness, both of the steels should have a similar resistance to crack initiation and should thus show comparably similar fatigue lives in HCF tests under normal conditions.
Tensile Tests
The specimens of both types of steel were subjected to tensile tests. The load applied to each specimen was recorded as a function of the axial displacement, which was increased at a rate of 5 mm/min during the test. The resultant UTS of each steel was calculated dividing the maximum load reached during the test over the minimum cross-section area of the specimen. The UTS of the A-type steel was 1470 MPa, whereas the UTS of the B-type steel was 1580 MPa, with combined standard uncertainties of 40 MPa and 50 MPa, respectively. Therefore, the mean UTS of the B-type steel was 7.5% higher than that of the A-type steel. The YS of each steel was obtained as the stress value that caused a plastic strain of 0.2%. The YS of the A-type steel was 1430 MPa, whereas the YS of the B-type steel was 1480 MPa, with combined standard uncertainties of 40 MPa and 50 MPa, respectively. In this case, the mean YS of the B-type steel was only 3.5% higher than the YS of the A-type steel.
The modest differences in the UTS and YS values of the steels were consistent with the dissimilar grain sizes observed in the microstructural analysis as the B-type steel grains were found to be marginally smaller than those of the A-type steel. Despite these minor differences between the two types of steel, the results of these tensile tests demonstrated that the H13 steels were able to provide remarkable mechanical strengths.
Fatigue Tests
Axial fatigue tests were carried out at room temperature. All the tests were loadcontrolled using a sinusoidal waveform with a frequency of 10 Hz and a stress ratio equal to zero. The output of all the fatigue tests was either the number of cycles to failure or a run-out at 10 7 cycles. Each specimen was subjected to a unique maximum stress selected so that the corresponding point in the Wöhler diagram fell within the HCF domain of the material. Likewise, the CF tests were developed under analogous conditions, although in these tests the surface of the specimen test section was placed in contact with a 0.1 M NaCl aqueous solution. Nine specimens of each steel were subjected to HCF tests, whereas six specimens of each steel underwent the CF tests. As a result of all the combinations of the type of material and test conditions, four datasets were considered.
The least squares method was applied to each of the four datasets in order to describe the HCF performance according to the Basquin exponential law for all the combinations of the type of steel and test conditions [36]. Each dataset was fitted to Equation (1): where σ MAX is the maximum stress applied on the fatigue test, N is the number of cycles to failure, a is a coefficient of fatigue strength, and b is the exponent of fatigue strength.
As a result of this regression fit, the coefficients presented in Table 2 were obtained. The coefficients of determination R 2 were added to this table to evaluate the goodness of fit in each case. The results of all the fatigue tests are displayed in Figure 5 as Wöhler diagrams. In these diagrams, the maximum axial stress in MPa is plotted versus the number of cycles to failure of the specimen using a semi-logarithmic scale. The filled dots shown in the diagrams represent the specimens tested to HCF and the blank dots represent specimens that were tested to CF. Moreover, the green square-shaped dots denote the fatigue test results of the A-type steel specimens, whereas the blue diamond-shaped dots indicate the fatigue test results obtained from the B-type steel. Run-out tests are indicated with an arrow pointing rightwards at 10 7 cycles. As for the tendency curves calculated in the regression analysis, the HCF curves are shown as continuous lines, whilst the CF curves are shown as dashed lines. The results of all the fatigue tests are displayed in Figure 5 as Wöhler diagra these diagrams, the maximum axial stress in MPa is plotted versus the number of to failure of the specimen using a semi-logarithmic scale. The filled dots shown diagrams represent the specimens tested to HCF and the blank dots represent spec that were tested to CF. Moreover, the green square-shaped dots denote the fatigu results of the A-type steel specimens, whereas the blue diamond-shaped dots indica fatigue test results obtained from the B-type steel. Run-out tests are indicated with row pointing rightwards at 10 7 cycles. As for the tendency curves calculated in the r sion analysis, the HCF curves are shown as continuous lines, whilst the CF curv shown as dashed lines. As can be observed in Figure 5, the fatigue behaviour under normal condition similar for both H13 steels. Both steels were able to resist 10 7 loading cycles when were subjected to a maximum stress of 980 MPa (68% and 66% of the YS of the A-typ B-type steel, respectively). Nevertheless, the fact that few A-type steel specimens below that stress, while B-type steel specimens did not, suggests that the actual f limit of the A-type steel could be marginally lower than that of the B-type steel. Th tentially better fatigue limit of the B-type steel would be in good agreement with th crostructural analysis carried out in this work, as the A-type steel showed a higher dance of coarser carbides than the B-type steel. Finer carbides should hinder crack agation, whereas coarser carbides and inclusions help promote crack initiation an detrimental to fatigue behaviour [17].
The scatter of the HCF results obtained here could be attributed to the bimod tribution of fatigue life in the transition between HCF and very-high-cycle fatigue (V domains. This transition between domains extends within a certain range of stress material. Within this range of stress, the probability of fracture in the HCF doma creases as the applied load decreases [37]. Moreover, microstructural heterogeneitie a wide distribution of defects, such as coarse carbides and inclusions, may raise the s of fatigue results within the transition between HCF and VHCF domains [38]. As can be observed in Figure 5, the fatigue behaviour under normal conditions was similar for both H13 steels. Both steels were able to resist 10 7 loading cycles when they were subjected to a maximum stress of 980 MPa (68% and 66% of the YS of the A-type and B-type steel, respectively). Nevertheless, the fact that few A-type steel specimens failed below that stress, while B-type steel specimens did not, suggests that the actual fatigue limit of the A-type steel could be marginally lower than that of the B-type steel. The potentially better fatigue limit of the B-type steel would be in good agreement with the microstructural analysis carried out in this work, as the A-type steel showed a higher abundance of coarser carbides than the B-type steel. Finer carbides should hinder crack propagation, whereas coarser carbides and inclusions help promote crack initiation and are detrimental to fatigue behaviour [17].
The scatter of the HCF results obtained here could be attributed to the bimodal distribution of fatigue life in the transition between HCF and very-high-cycle fatigue (VHCF) domains. This transition between domains extends within a certain range of stress of the material. Within this range of stress, the probability of fracture in the HCF domain decreases as the applied load decreases [37]. Moreover, microstructural heterogeneities and a wide distribution of defects, such as coarse carbides and inclusions, may raise the scatter of fatigue results within the transition between HCF and VHCF domains [38].
The methodology utilised in this study to develop the CF tests, despite being rather simple, proved to be effective and economical. An abrupt decrease in fatigue lives in the CF tests as compared with the HCF tests was perceptible. Moreover, this difference between the fatigue lives in the test conditions was more noticeable at lower stress values than at higher ones. As a result of these differences, the exponents of the fatigue strength calculated for the CF tests were much lower (more negative) than those of the datasets obtained from the HCF tests. Lower exponents of fatigue strength resulted in steeper Wöhler diagrams. Therefore, no signs of fatigue limit were observed for the CF tests within the stress levels that they were subjected to in this study. All of the mentioned observations of the CF curves were in accordance with the contributions of other researchers [39,40].
The CF lives of the B-type steel were approximately twice as long as those of the A-type steel. Unlike the HCF tests, the Wöhler diagrams of both steels obtained from the CF tests remained virtually parallel and did not overlap. It should also be pointed out that the data were much less scattered for the CF tests than for the HCF tests. The coefficients of determination, displayed in Table 2, were higher than 0.95 for the CF tests, whereas they did not exceed the value of 0.85 for the HCF tests. It was not surprising that the data scatter was reduced in the CF tests since the mechanism that triggered fatigue failure in the CF tests was the formation of corrosion pits rather than microstructural features. The corrosion pits acted as sharp stress concentrators that eventually opened a crack in the surface of the specimen that was subjected to fatigue loads. In contrast, crack initiation in HCF tests was attributed to the presence of surface defects and inclusions, so crack initiation required more time to be developed. Therefore, the presence of microstructural defects and inclusions seemed to play a minor role in the CF tests as compared to the HCF tests as the specimen failure was advanced substantially in the CF tests. In order to further understand these results, an analysis of the fracture surfaces of the specimens that failed in the fatigue tests was carried out.
Fracture Analysis
Fracture surface analysis was carried out on several specimens that failed after the fatigue tests. First, images of the whole fracture surfaces were taken using a stereoscopic microscope. Next, a deeper fracture surface analysis was performed using SEM. SEM is a powerful tool with a great depth of focus which permitted us to obtain good-quality images at a high magnification.
Representative fracture surface images of the specimens tested under normal conditions are presented in Figures 6-8. The images show the characteristic zones of a typical axial fatigue fracture with no stress concentration [41], namely, the fatigue crack propagation zone, the final fracture zone, and the shear lip. The remarkable symmetry and crack directionality of these fracture surfaces confirmed that crack initiation took place at a single point of the surface of the specimens. After the crack was opened, it started growing at a rate dependent on the applied stress intensity range. Once the crack reached a size such that the remaining cross-section was no longer able to support the applied load, a sudden fracture of the specimen occurred. Therefore, it was not surprising that, for lower stress amplitude values, the fatigue crack propagation area increased in size whilst the final fracture area decreased. As a result of the overall fracture development process, the fatigue crack propagation zones were comparatively smoother than the final fracture zones. The shear lips indicated the ultimate failure location of the specimens, and they extended throughout the perimeter of the final fracture areas. Materials 2022, 15, x FOR PEER REVIEW 11 of 20 Some differences between the fracture surfaces of both steels were observed. It was noted that for the same applied stress range, the size of the crack propagation areas and the shear lip widths were greater in the B-type steel than in the A-type steel. This difference in the crack propagation areas was in line with the results obtained in the tensile tests, as the B-type steel presented higher YS and UTS values than the A-type steel. Under the same fatigue test conditions, the material with higher UTS is expected to present a Some differences between the fracture surfaces of both steels were observed. It was noted that for the same applied stress range, the size of the crack propagation areas and the shear lip widths were greater in the B-type steel than in the A-type steel. This difference in the crack propagation areas was in line with the results obtained in the tensile tests, as the B-type steel presented higher YS and UTS values than the A-type steel. Under the same fatigue test conditions, the material with higher UTS is expected to present a Some differences between the fracture surfaces of both steels were observed. It was noted that for the same applied stress range, the size of the crack propagation areas and the shear lip widths were greater in the B-type steel than in the A-type steel. This difference in the crack propagation areas was in line with the results obtained in the tensile tests, as the B-type steel presented higher YS and UTS values than the A-type steel. Under the same fatigue test conditions, the material with higher UTS is expected to present a Some differences between the fracture surfaces of both steels were observed. It was noted that for the same applied stress range, the size of the crack propagation areas and the shear lip widths were greater in the B-type steel than in the A-type steel. This difference in the crack propagation areas was in line with the results obtained in the tensile tests, as the B-type steel presented higher YS and UTS values than the A-type steel. Under the same fatigue test conditions, the material with higher UTS is expected to present a lower final fracture area. Moreover, the B-type steel fracture surfaces looked smoother than those of the A-type steel. The distinctions observed in the fracture surface roughness of both steels were attributed to their dissimilar microstructures, since the B-type steel proved to have a lower grain size and finer carbides than the A-type steel.
All the corrosion fatigue specimens failed due to surface crack initiation. The fracture surfaces of two specimens from the corrosion fatigue tests are shown in Figure 9. These specimens were tested to a maximum stress of 750 MPa (52% and 51% of the YS of A-type and B-type steel, respectively), which means that they would have been run-outs if they were tested under normal conditions. Essentially, these images revealed almost the same features as those of normal fatigue tests. However, it was noted that failure took place due to the development of a corrosion pit on the surface of the specimens. The corrosion pits are indicated with arrows in Figure 9. The formation of these corrosion pits accelerated the crack initiation process and thus decreased the fatigue lives of these specimens dramatically. In this case, the size of the crack propagation area was significantly greater in the B-type steel than in the A-type steel. lower final fracture area. Moreover, the B-type steel fracture surfaces looked smoother than those of the A-type steel. The distinctions observed in the fracture surface roughness of both steels were attributed to their dissimilar microstructures, since the B-type steel proved to have a lower grain size and finer carbides than the A-type steel. All the corrosion fatigue specimens failed due to surface crack initiation. The fracture surfaces of two specimens from the corrosion fatigue tests are shown in Figure 9. These specimens were tested to a maximum stress of 750 MPa (52% and 51% of the YS of A-type and B-type steel, respectively), which means that they would have been run-outs if they were tested under normal conditions. Essentially, these images revealed almost the same features as those of normal fatigue tests. However, it was noted that failure took place due to the development of a corrosion pit on the surface of the specimens. The corrosion pits are indicated with arrows in Figure 9. The formation of these corrosion pits accelerated the crack initiation process and thus decreased the fatigue lives of these specimens dramatically. In this case, the size of the crack propagation area was significantly greater in the B-type steel than in the A-type steel. Fracture surface images with a high magnification were obtained using SEM. Under normal test conditions, most of the specimens revealed that crack initiation took place at a single point on the surface of the specimens. Examples of specimens that presented surface crack initiation are shown in Figures 10 and 11. The predominant mechanism of the crack initiation region was a transgranular fracture. Crack initiation was likely to occur in a surface defect of these specimens, such as a small scratch or an indentation that acted as a notch and raised the stress locally. Intermetallic carbides could have also intervened as stress concentrators that triggered crack initiation in these steels as they are heterogeneities present in the material [42]. Crack opening was driven by shear stress at this point. After the crack reached a size of a few micrometres, it started to propagate perpendicularly to the tension load applied to the specimen [43]. Underneath the surface, even though the fractures were predominantly transgranular, signs of intergranular fracture could be observed at higher magnifications. Fracture surface images with a high magnification were obtained using SEM. Under normal test conditions, most of the specimens revealed that crack initiation took place at a single point on the surface of the specimens. Examples of specimens that presented surface crack initiation are shown in Figures 10 and 11. The predominant mechanism of the crack initiation region was a transgranular fracture. Crack initiation was likely to occur in a surface defect of these specimens, such as a small scratch or an indentation that acted as a notch and raised the stress locally. Intermetallic carbides could have also intervened as stress concentrators that triggered crack initiation in these steels as they are heterogeneities present in the material [42]. Crack opening was driven by shear stress at this point. After the crack reached a size of a few micrometres, it started to propagate perpendicularly to the tension load applied to the specimen [43]. Underneath the surface, even though the fractures were predominantly transgranular, signs of intergranular fracture could be observed at higher magnifications. Unlike most of the tested specimens, two A-type steel specimens failed due to internal crack initiation. The fracture surfaces of these specimens are shown in Figures 12 and 13. In these specimens, crack initiation occurred at an inner point of the material and extended radially. Furthermore, a circular dark area can be distinguished in Figures 12a and 13a, which is commonly known as the fish-eye region. At the centre of these circular areas, the presence of an inclusion or a cluster of inclusions was revealed. The size of these inclusions was greater than 10 µm, and they might have led to premature crack initiation on the sample. These inclusions were mainly oxides of calcium, magnesium, and aluminium from the electrode, which survived the ESR manufacturing process due to their high melting point [44]. No internal crack initiation or inclusions were observed in the B-type steel specimens. Unlike most of the tested specimens, two A-type steel specimens failed due to internal crack initiation. The fracture surfaces of these specimens are shown in Figures 12 and 13. In these specimens, crack initiation occurred at an inner point of the material and extended radially. Furthermore, a circular dark area can be distinguished in Figures 12a and 13a, which is commonly known as the fish-eye region. At the centre of these circular areas, the presence of an inclusion or a cluster of inclusions was revealed. The size of these inclusions was greater than 10 µm, and they might have led to premature crack initiation on the sample. These inclusions were mainly oxides of calcium, magnesium, and aluminium from the electrode, which survived the ESR manufacturing process due to their high melting point [44]. No internal crack initiation or inclusions were observed in the B-type steel specimens. Unlike most of the tested specimens, two A-type steel specimens failed due to internal crack initiation. The fracture surfaces of these specimens are shown in Figures 12 and 13. In these specimens, crack initiation occurred at an inner point of the material and extended radially. Furthermore, a circular dark area can be distinguished in Figures 12a and 13a, which is commonly known as the fish-eye region. At the centre of these circular areas, the presence of an inclusion or a cluster of inclusions was revealed. The size of these inclusions was greater than 10 µm, and they might have led to premature crack initiation on the sample. These inclusions were mainly oxides of calcium, magnesium, and aluminium from the electrode, which survived the ESR manufacturing process due to their high melting point [44]. No internal crack initiation or inclusions were observed in the B-type steel specimens. On the fracture surfaces analysed in this study, the so-called fish-eye morphology was present, but no fine granular area (FGA) was observed [45]. The darkness of the fisheye was ascribed to the absence of air contact with the fracture surface until the crack reached the specimen surface [46]. Moreover, the size of the fish-eye area was strongly correlated with the depth of the inclusion that initiated the main crack [47], which is in accordance with the observations shown in Figures 12 and 13.
The likelihood of internal crack initiation depends on the applied stress range and the average size of inclusions. When these inclusions reach a size of several micrometres, they can be treated as effective cracks [48]. At lower stress values, the stress concentration due to the surface defects may not be as high as at inclusions, so the likelihood of internal crack initiation is enhanced. In fact, mathematical models have been proposed that estimate the fatigue strength of steels in the VHCF regime using the inclusion area as one of the main parameters of the model [49]. Even though internal crack initiation is a usual fracture mechanism of the VHCF regime, internal crack initiation has also been observed at cycles to failure as low as 10 5 cycles [50].
Plastic deformation in the stable crack growth area of an A-type steel and a B-type steel is shown in Figures 14 and 15, respectively. Striations of less than 1 µm in width can On the fracture surfaces analysed in this study, the so-called fish-eye morphology was present, but no fine granular area (FGA) was observed [45]. The darkness of the fisheye was ascribed to the absence of air contact with the fracture surface until the crack reached the specimen surface [46]. Moreover, the size of the fish-eye area was strongly correlated with the depth of the inclusion that initiated the main crack [47], which is in accordance with the observations shown in Figures 12 and 13.
The likelihood of internal crack initiation depends on the applied stress range and the average size of inclusions. When these inclusions reach a size of several micrometres, they can be treated as effective cracks [48]. At lower stress values, the stress concentration due to the surface defects may not be as high as at inclusions, so the likelihood of internal crack initiation is enhanced. In fact, mathematical models have been proposed that estimate the fatigue strength of steels in the VHCF regime using the inclusion area as one of the main parameters of the model [49]. Even though internal crack initiation is a usual fracture mechanism of the VHCF regime, internal crack initiation has also been observed at cycles to failure as low as 10 5 cycles [50].
Plastic deformation in the stable crack growth area of an A-type steel and a B-type steel is shown in Figures 14 and 15, respectively. Striations of less than 1 µm in width can On the fracture surfaces analysed in this study, the so-called fish-eye morphology was present, but no fine granular area (FGA) was observed [45]. The darkness of the fish-eye was ascribed to the absence of air contact with the fracture surface until the crack reached the specimen surface [46]. Moreover, the size of the fish-eye area was strongly correlated with the depth of the inclusion that initiated the main crack [47], which is in accordance with the observations shown in Figures 12 and 13.
The likelihood of internal crack initiation depends on the applied stress range and the average size of inclusions. When these inclusions reach a size of several micrometres, they can be treated as effective cracks [48]. At lower stress values, the stress concentration due to the surface defects may not be as high as at inclusions, so the likelihood of internal crack initiation is enhanced. In fact, mathematical models have been proposed that estimate the fatigue strength of steels in the VHCF regime using the inclusion area as one of the main parameters of the model [49]. Even though internal crack initiation is a usual fracture mechanism of the VHCF regime, internal crack initiation has also been observed at cycles to failure as low as 10 5 cycles [50].
Plastic deformation in the stable crack growth area of an A-type steel and a B-type steel is shown in Figures 14 and 15, respectively. Striations of less than 1 µm in width can be distinguished in these specimens, which revealed the slip plane during crack propagation. These striations were often located in individual grains within a fracture surface caused by transgranular plastic deformation. The width of the striations can be associated with the crack growth rate during the fatigue test. Secondary cracks can also be seen in the fracture surfaces, usually perpendicular to the direction of the main crack propagation. These secondary cracks are also evidence of a high crack propagation rate. Furthermore, it was observed that the main crack propagated through non-coplanar grains. be distinguished in these specimens, which revealed the slip plane during crack propagation. These striations were often located in individual grains within a fracture surface caused by transgranular plastic deformation. The width of the striations can be associated with the crack growth rate during the fatigue test. Secondary cracks can also be seen in the fracture surfaces, usually perpendicular to the direction of the main crack propagation. These secondary cracks are also evidence of a high crack propagation rate. Furthermore, it was observed that the main crack propagated through non-coplanar grains. The SEM images also revealed a more gradual transition of the crack propagation through non-coplanar grains in the B-type steel (Figure 15a) as compared to the A-type steel (Figure 14a). Intermetallic carbides could have acted as obstacles for crack propagation [17], deviating the main crack to different non-coplanar grains during crack growth. Therefore, the presence of coarser carbides in the A-type steel specimens would explain why the fracture surfaces of these specimens were rougher than those of the B-type steel.
In summary, fracture surface analysis was carried out to further understand and explain the results obtained from the fatigue tests of hot forging tool steels. Most specimens presented a unique crack initiation site located on the surface. Moreover, it was demon- be distinguished in these specimens, which revealed the slip plane during crack propagation. These striations were often located in individual grains within a fracture surface caused by transgranular plastic deformation. The width of the striations can be associated with the crack growth rate during the fatigue test. Secondary cracks can also be seen in the fracture surfaces, usually perpendicular to the direction of the main crack propagation. These secondary cracks are also evidence of a high crack propagation rate. Furthermore, it was observed that the main crack propagated through non-coplanar grains. The SEM images also revealed a more gradual transition of the crack propagation through non-coplanar grains in the B-type steel (Figure 15a) as compared to the A-type steel (Figure 14a). Intermetallic carbides could have acted as obstacles for crack propagation [17], deviating the main crack to different non-coplanar grains during crack growth. Therefore, the presence of coarser carbides in the A-type steel specimens would explain why the fracture surfaces of these specimens were rougher than those of the B-type steel.
In summary, fracture surface analysis was carried out to further understand and explain the results obtained from the fatigue tests of hot forging tool steels. Most specimens presented a unique crack initiation site located on the surface. Moreover, it was demon- The SEM images also revealed a more gradual transition of the crack propagation through non-coplanar grains in the B-type steel (Figure 15a) as compared to the A-type steel (Figure 14a). Intermetallic carbides could have acted as obstacles for crack propagation [17], deviating the main crack to different non-coplanar grains during crack growth. Therefore, the presence of coarser carbides in the A-type steel specimens would explain why the fracture surfaces of these specimens were rougher than those of the B-type steel.
In summary, fracture surface analysis was carried out to further understand and explain the results obtained from the fatigue tests of hot forging tool steels. Most specimens presented a unique crack initiation site located on the surface. Moreover, it was demonstrated that a few A-type steel specimens failed due to internal cracks originating at inclusion locations. The sizes of the crack propagation areas of the B-type steel were significantly lower and smoother than those of the A-type steel. These differences were attributed to the coarser carbides observed in the A-type steel through the microstructural analysis. Overall, these observations suggest that the more homogeneous microstructure of the B-type steel provided a higher resistance to fracture than the A-type steel in the presence of cracks.
Discussion
In the present work, the axial fatigue behaviour within the HCF domain of two H13 steels has been studied. As a novel contribution, the CF results of these H13 steels have been analysed for the first time. On the one hand, a microstructural analysis was carried out prior to the fatigue tests. The microstructural analysis revealed some dissimilarities between the steels studied here regarding their grain sizes and carbide distributions. These dissimilarities anticipated potential differences in the fatigue behaviour of both materials, despite belonging to the same steel specification. On the other hand, the phase compositions, hardness, and the specimens utilised in the fatigue tests were identical for both steels. Hence, this work has allowed us to analyse the effect of some microstructural features on the fatigue behaviour of hot forging tool steels under different work environments.
Both steels presented reasonably similar results in the HCF tests under normal conditions. The HCF results were presented as the maximum stresses supported in axial tests when the stress ratio was zero. If these results were converted to stress amplitudes, the resultant values would be approximately 75-85% of those reported by Shinde et al. [17] for H13 or by Korade et al. [18] for H21 in rotating bending tests. This proportion aligns with the experimental differences observed between the axial and bending fatigue tests. The different results of both of the test methodologies are due to the fact that a higher proportion of the specimen's volume is subjected to the maximum stress in the axial fatigue tests as compared to the rotating bending fatigue tests [51].
The corrosion fatigue lives of the H13 steels were dramatically lower than the fatigue lives obtained under normal conditions. Moreover, this reduction in the fatigue life results was more evident as the applied stress range was decreased. At higher stress levels, the mechanical damage due to the applied stress was much higher than the corrosion damage; thus, the fatigue lives were found to be similar regardless of the test environment. At lower stress levels, corrosion damage was reported to overcome mechanical damage as the formation of corrosion pits takes place [21]. The corrosion fatigue behaviour of the H13 steels observed in the present study was in good agreement with the findings of other studies regarding the corrosion fatigue of martensitic steels [22,23].
Interestingly, the coefficients of determination obtained for each material and test condition were higher for the CF tests than for the HCF tests. These results meant that the data scatter was greater in the conventional HCF tests. In the conventional HCF tests, microstructural defects are known to be an important source of scatter [38]. Crack initiation is appreciably sensitive to the size distribution of defects and the microstructural heterogeneities of the specimen. Under such conditions, the time invested in crack initiation represents almost the totality of the fatigue lives in the HCF domain.
In contrast, crack initiation occurred due to the formation of corrosion pits in the CF tests. The formation of corrosion pits advanced the failure of specimens substantially. The loading frequency plays a key role in the development of corrosion pits. Lower loading frequencies proved to result in lower corrosion fatigue lives [21]. If two specimens were subjected to CF tests at different loading frequencies, the one with the lower frequency would undergo a lower number of loading cycles than the other at the same time instants. The fatigue lives of the specimens in the CF tests should depend strongly on the time required to develop a corrosion pit, as corrosion pits are the mechanisms that trigger crack initiation in these tests. The kinetics of corrosion pit formation depend mostly on the loading frequency [52], the salt concentration [53], the temperature, and the pH of the corrosive medium [54]. Since all these variables remained identical for all the specimens, the crack initiation times of the CF tests are believed to have been virtually alike. As a result, similar crack initiation times led to lower data scatter.
Another feature that should be emphasised is the difference observed between the crack propagation areas of both the steels when they were tested at the same stress level. The A-type steel presented smaller crack growth regions than the B-type steel. Furthermore, this difference between crack growth regions was more remarkable at lower stresses. This observation suggested that, for a certain crack size, the B-type steel was able to support a higher load than the A-type steel. The fact that the B-type steel showed marginally higher strength than the A-type steel accords with the tensile test results. In the tensile test results, the B-type steel had a UTS that was 7.5% higher and a YS 3.5% that was higher than those of the A-type steel. Nevertheless, this discrepancy between the YS and UTS values did not seem significant enough to explain the large difference between the crack growth region sizes of both steels at low stress ranges (see Figure 9). These different crack propagation sizes suggested that the B-type steel had higher fracture toughness than the A-type steel, as fracture toughness characterises the material's resistance to crack propagation.
The potentially higher fracture toughness of the B-type steel could also explain its better performance as compared to the A-type steel in the CF tests. It has been stated that all the CF tests should reveal similar crack initiation times as this time depends on the kinetics of the development of corrosion pits. The fact that the B-type steel presented fatigue lives that were approximately twice as high as those of the A-type steel in all the CF tests should be attributed to longer crack propagation times. The longer crack propagation times of the B-type steel specimens would also align with the larger extension of their crack propagation areas. Moreover, the presence of coarser carbides and inclusions in the A-type steel is believed to have accelerated the fractures [17].
The mechanical properties of hot forging tool steels are strongly controlled by the parameters of the heat treatments that these steels are subjected to. Such mechanical properties include the UTS, hardness, and fracture toughness [3,4,[12][13][14]. These properties are a consequence of the grain size and carbide distribution obtained at the end of the heat treatment. Slight discrepancies in the austenitising or tempering temperatures, times, or heating/cooling rates could have been factors that led to the dissimilar microstructures of the two types of steel. In this study, it is likely that the B-type steel was subjected to a higher cooling rate than the A-type steel during the quenching stage of the heat treatment, as this fact would explain the lower grain size and finer carbides of the B-type steel.
In summary, hot forging tool steels present outstanding static and dynamic mechanical properties. Nevertheless, an aggressive environment may result in the premature failure of these steels under cyclic loads. We encourage the study of the crack initiation and propagation phenomena of hot forging tool steels under different working environments in future research. Understanding the fatigue fracture mechanisms of these steels is crucial to guaranteeing the correct design of hot forging dies, thereby also saving resources invested in the maintenance of such tools.
Conclusions
For the first time, this study presents the comparative results of H13 steels subjected to axial high-cycle fatigue tests under normal conditions and corrosion fatigue tests. The understanding of these damage processes in hot forging tool steels is fundamental to ensuring the extensive lives of forging tools, especially under aggressive environments.
The two H13 steels studied here, namely, the A-type and the B-type, were found to have identical microstructural phases and hardness values, with only slight differences in their YS and UTS values. Nevertheless, the B-type steel had a lower grain size than that of the A-type steel. Moreover, carbides in the A-type steel were coarser and more abundant than those in the B-type steel. These discrepancies were attributed to marginal variations in the heat treatments of both materials.
The results of all the axial fatigue tests were fitted to the Basquin equation to mathematically represent the relationship between the maximum stress of the specimens, σ MAX , and the number of cycles to failure, N, for each material and test condition. As a result, the fatigue behaviours of the A-type and B-type steel under normal conditions were modelled by σ MAX = 2240.0 N −0.057 R 2 = 0.850 and σ MAX = 1711.7 N −0.038 R 2 = 0.673 , respectively, whereas their corrosion fatigue behaviours were modelled by σ MAX = 16, 662.6 N −0.269 R 2 = 0.964 and σ MAX = 11, 808.1 N −0.227 R 2 = 0.971 , respectively.
Both of the H13 steels showed outstanding fatigue strengths in conventional highcycle fatigue tests, reaching values of maximum stress as high as 980 MPa at 10 7 cycles. However, the corrosion fatigue lives of the two types of steel were much lower than the conventional high-cycle fatigue lives, showing no signs of fatigue limit for the same range of applied stress. The significant decrease in the exponents of fatigue strength in the corrosion fatigue tests demonstrated that the effect of the corrosive agent on the fatigue lives was more noticeable at lower stress values. Furthermore, the coefficients of determination for the Basquin equations were much higher in the CF tests than in the HCF tests, which indicated a lower tendency of data scatter in the CF tests. The crack initiation stage in the conventional HCF tests involved greater randomness as it depended strongly on the size and distribution of the microstructural defects. In contrast, it is believed that the crack initiation times were similar in all the CF tests, as they mainly relied on the time required to create a corrosion pit due to exposure of the steels to the corrosive medium.
Interestingly, both of the steels presented similar fatigue behaviour under the conventional HCF tests, whereas their CF lives were significantly different. The corrosion fatigue lives in the B-type steel were twice as high as those in the A-type steel. This difference between the corrosion fatigue lives was in line with the greater extension of the crack propagation regions in the B-type steel as compared to the A-type steel. Overall, the higher corrosion fatigue lives of the B-type steel can be attributed to its higher resistance to crack propagation.
The results obtained in the present study have emphasised the important influence of the steel microstructure on the fatigue behaviour and fracture toughness. The specimens with coarser carbides and higher grain sizes were found to have lower fracture strengths during crack propagation. The effect of fracture toughness on the steel durability was more noticeable in the CF tests than in the conventional HCF tests. Therefore, we propose analyses of the crack growth rate of hot forging tool steels under different aggressive environments as worthwhile research to be carried out in the future. | 2022-10-27T15:35:34.034Z | 2022-10-22T00:00:00.000 | {
"year": 2022,
"sha1": "5049be4712db272855655988013d6ab169524387",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1996-1944/15/21/7411/pdf?version=1666425679",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "67a154b8160ab26313779d8498f55e989d83c421",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": []
} |
232105200 | pes2o/s2orc | v3-fos-license | Continuity properties of Lyapunov exponents for surface diffeomorphisms
We study the entropy and Lyapunov exponents of invariant measures $\mu$ for smooth surface diffeomorphisms $f$, as functions of $(f,\mu)$. The main result is an inequality relating the discontinuities of these functions. One consequence is that for a $C^\infty$ surface diffeomorphisms, on any set of ergodic measures with entropy bounded away from zero, continuity of the entropy implies continuity of the exponents. Another consequence is the upper semi-continuity of the Hausdorff dimension on the set of ergodic invariant measures with entropy bounded away from zero. We also obtain a new criterion for the existence of SRB measures with positive entropy.
Introduction
Entropy and Lyapunov exponents play a major role in the study of differentiable dynamical systems, and their dependence on the measure and the map is of great interest. This dependence is sometimes continuous, but not always (for entropy, see [33,35,12,9,17,10], and for Lyapunov exponents, see [44,20,6,7,1,47]). While there are many works relating the values of the entropy to the values of the Lyapunov exponents [43,39,31,27], the relation between the (dis)continuity of these objects as functions of the measure and the diffeomorphism has not yet been studied. The purpose of this work is to fill this gap, in the smooth two-dimensional case. For instance, we show: For other consequences, including the upper semi-continuity of the unstable dimension and of the Hausdorff dimension of ergodic measures with positive entropy, see Section 1.5.
These results follow from inequalities between the multiplicative size of the defects in continuity of the entropy and the top Lyapunov exponent. These inequalities, which are the main results of this work, are described in detail in the next section.
Main results
Throughout this paper, M is a two-dimensional compact C ∞ Riemannian manifold without boundary. Let Diff r (M ) denotes the class of C r diffeomorphisms on M (see §3.1).
In the C ∞ case, these functions are semi-continuous. Specifically, suppose f k , f ∈ Diff ∞ (M ), ν k are ergodic f k -invariant measures, f k → f in C ∞ and ν k → µ weak- * . Then We call these quantities the discontinuity ratios, and think of them as measures for the difference between the two sides of the inequalities in (1). We will provide inequalities relating the discontinuity ratio of the entropy to the discontinuity ratio of λ + . (It is enough to consider λ + , because λ − (f, µ) = −λ + (f −1 , µ).) 1.1. The ergodic C ∞ case. Our results are simplest and strongest when the maps are C ∞ and the limiting measure is ergodic: Theorem A. For every k ≥ 1, let f k ∈ Diff ∞ (M ) and let ν k be an f k -ergodic invariant measure. Suppose -lim k λ + (f k , ν k ) and lim k h(f k , ν k ) exist and are positive, -f k converge in the C ∞ topology to a diffeomorphism f ∈ Diff ∞ (M ), -ν k converge weak- * to a probability measure µ (necessarily f -invariant).
If µ is f -ergodic, then lim The following result is an immediate consequence of this and (1), and was the original aim of our work: Corollary 1.1. For every k ≥ 1, let f k ∈ Diff ∞ (M ) and let ν k be an f k -ergodic invariant measure. Suppose f k → f in the C ∞ topology, and ν k → µ weak- * where µ is an f -ergodic invariant measure with positive entropy. If h(f k , ν k ) → h(f, µ), then: The result mentioned in the introduction is the special case f 1 = f 2 = · · · = f . Df Since M is compact, λ(f ) is independent of the choice of a Riemannian metric · x .
Theorem B. Fix r > 2. For every k ≥ 1, let f k ∈ Diff r (M ) and let ν k be an f k -ergodic invariant measure. Suppose -lim k→∞ λ + (f k , ν k ) and lim k→∞ h(f k , ν k ) exist and are positive, -f k → f in the C r -topology, -ν k converge weak- * to a probability measure µ. If µ is f -ergodic and has positive entropy, then λ + (f, µ) .
As we will explain in Section 1.4, the smoothness index r above does not need to be an integer.
1.3. The non-ergodic case. The assumption that the limiting measure is ergodic is often difficult to check, and we now explain what can be said in its absence (a more general but also more technical result, Theorem D, will be given in section 7).
Theorem C. Fix r > 2. For every k ≥ 1, let f k ∈ Diff r (M ) and let ν k be an f k -ergodic invariant measure. Suppose -lim k→∞ λ + (f k , ν k ) and lim k→∞ h(f k , ν k ) exist and are positive, -f k → f in the C r -topology, -ν k → µ weak- * for some f -invariant probability measure µ (perhaps non-ergodic).
Note that for C ∞ diffeomorphisms, the term (λ(f ) + λ(f −1 ))/(r − 1) can be replaced by zero, because the theorem can be applied with r arbitrarily large. The decomposition µ = (1 − β)µ 0 + βµ 1 depends on the sequences (f k ) k≥1 , (ν k ) k≥1 , and not just on their limits. We give a heuristic description of this decomposition in Section 2.2.
1.4. Additional comments. We now supplement Theorems A, B and C by some examples, comments, strengthenings, and generalizations. The proofs can be found in Section 8.
Variant inequality.
Theorem A does not use the symmetry between a diffeomorphism and its inverse. When 0 < −λ − (f, µ) < λ + (f, µ), this symmetry yields a sharper bound: 1.4.3. Sequences of non-ergodic measures. Our results can be extended to the case when the invariant measures ν k are not ergodic, but this requires stronger assumptions on the Lyapunov exponents of ν k , which in the non-ergodic case are functions and not constants. See Corollary 8.5.
1.4.4.
Lifted version. f induces a dynamical system f on the projective tangent bundle M , see Section 3.2. It turns out that µ = (1 − β)µ 0 + βµ 1 is a projection of a decomposition of a limit point µ := lim ν ki , where ν k are lifts of ν k to f -invariant measures on M . The decomposition of µ contains more information than the decomposition of µ, and leads to a stronger statement, Theorem D, in Section 7. This strengthening is essential to the proof of Corollary 8.5 on the case when ν k are not ergodic. 1.4.5. Convergence of C r -diffeomorphisms. In finite differentiability, Theorem D allows a weaker convergence assumption (denoted by f k r−bd −→ f , see Section 3.1), and r does not have to be an integer. 1.4.6. Entropy upper semi-continuity in the C r case. It is well-known that for C r -diffeomorphisms the entropy may fail to be upper semi-continuous, but the defect in upper semicontinuity can be bounded (see the discussion in Section 1.6 below). This bound manifests itself in Theorems B and C in the expression λ(f )+λ(f −1 ) r−1 . But our proof gives a slightly stronger bound λ( f ) r−1 , in terms of the dynamics of f on the projective tangent bundle, see §3.2, §3.5, and Theorem D.
In some special cases, even this stronger bound can be improved. For instance, if λ + (f k , ν k ) → λ + (f, µ), then which is stronger than the assertion of Theorem B. When f 1 = f 2 = · · · = f , (5) is a refinement of a classical inequality of Yomdin and Newhouse; it follows from bounds on the tail entropy, which were first written explicitly in [10], and which are consequences of the Downarowicz variational principle [16]. When the sequence (f k ) is non-constant, it follows from a bound on robust tail entropy, which can be shown using techniques in [11]. We thank David Burguet for explaining this to us. Corollary 1.6. For every k ≥ 1, let f k ∈ Diff ∞ (M ) and let ν k be an f k -ergodic invariant measure. Suppose f k → f in the C ∞ topology, ν k → µ weak- * , and lim h(f k , ν k ) exists and is positive. Then there are ergodic components µ , µ of µ satisfying λ + (µ ), λ + (µ ) > 0 and Proof. Consider the decomposition µ = (1−β)µ 0 +βµ 1 and take suitable ergodic components of µ 1 .
Notice that it is essential in this proof to be able to deal with non-ergodic limits, since we have no control of lim ν k .
1.6. Related works. In this paper we relate the continuity properties of the entropy to those of the Lyapunov exponents. The continuity of these objects has been studied separately before in several works, which we now recall.
Entropy. In general, the entropy map (f, µ) → h(f, µ) is not lower semi-continuous, even in the uniformly hyperbolic case. For example, it is easy to construct sequences of atomic measures (with zero entropy) on a basic set, which converge to limits with positive entropy.
However, for C ∞ diffeomorphisms on compact manifolds, the entropy map is upper semicontinuous: This is due to Newhouse [35]. For C r diffeomorphisms with finite r, even upper semi-continuity may fail (for examples in dimension four see [33], and for examples in dimension two see [14]). However, the (additive) defect in semi-continuity: can be bounded from above by min(λ(f ), λ(f −1 ))/r, using Yomdin theory [35,12,10,11]. A subject of more recent interest is the loss of semi-continuity due to non-compactness. This has been studied for countable Markov shifts [22,23], geodesic flows on non-compact homogeneous spaces [18,25], and geodesic flows on non-compact manifolds with negative sectional curvatures [21,42].
Lyapunov exponents. The top Lyapunov exponent map (f, µ) → λ + (f, µ) varies continuously for uniformly hyperbolic systems on surfaces. It even depends analytically on the diffeomorphism f [44]. Moreover if µ max is the unique measure of maximal entropy of a mixing Anosov surface diffeomorphism, then [24] (see also [41,45] In the non-uniformly hyperbolic case, the situation is different. For example, [6] proves that among conservative systems the Lyapunov exponents of the volume measure are discontinuous when the diffeomorphism varies in the C 1 -topology, unless they vanish.
We are not aware of other general results on the continuity of the Lyapunov exponents for general non-uniformly hyperbolic surface diffeomorphisms.
By contrast, much is known on the continuity of Lyapunov exponents of random products of independent identically distributed SL(2, R) matrices, as functions of the underlying Bernoulli process, see [20,7]. More general Hölder continuous matrix cocycles with holonomies are considered in [2], and a higher-dimensional extension has been announced in [47].
Dimension. L.-S. Young gave the famous formula (6) for the dimension of hyperbolic invariant measures in [49] in terms of the entropy and the Lyapunov exponents of the measure. For further dimension theoretic properties of hyperbolic invariant measures, see [4] and [3]. The continuity of the dimension of invariant sets and measures for hyperbolic systems have been considered in numerous works, for instance [37] proves that basic sets on surface have a Hausdorff dimension which varies continuously with the diffeomorphism, [5] proves that the supremum of the Hausdorff dimensions of ergodic measures on such a basic set is attained by a measure of maximal dimension and [3] discusses some non-uniformly hyperbolic cases.
A heuristic overview of the proof
All our results follow from Theorem C, and the remainder of the paper is dedicated to the proof of this theorem. Here we give a heuristic overview of the proof, in the special case when f 1 = f 2 = · · · = f is a C ∞ diffeomorphism.
2.1. The origin of the discontinuities in λ + . As Furstenberg discovered, the Lyapunov exponents are easier to study in terms of the projective dynamics f (x, E) = (f (x), Df x (E)) on the projective tangent bundle Indeed by Ledrappier's work, a Lyapunov exponent of an f -ergodic measure µ is simply the integral of the continuous function with respect to the lift of µ to the bundle of the associated Oseledets spaces.
Suppose ν k are ergodic measures with positive entropy such that ν k → µ weak * , and suppose for the moment that µ is ergodic and with positive entropy. Since dim(M ) = 2, ν k have two simple Lyapunov exponents, and there are exactly two ergodic lifts ν + k and ν − k , one carried by the bundle E u of unstable Oseledets spaces, and the other carried by the bundle E s of the stable Oseledets spaces. (The third bundle E 0 associated to the zero exponent has measure zero for all lifts of ν k .) Hence, Suppose ν + k converge weak-star on M to an f -invariant probability measure µ (this is true for a subsequence). Since ϕ : M → R is continuous, The limiting measure µ is a lift of µ, but this does not have to be the lift of µ to E u , µ + . If It is certainly possible that µ = µ + : The Oseledets bundle E u carrying the lifts ν + k is not necessarily bounded away from E s , and some mass 0 ≤ ρ ≤ 1 on E u can escape to E s .
Escape of mass to E s is reflected in long stretches of time when ν k -typical orbits do not experience the exponential growth of E u -directions predicted by λ + . Instead, they see, temporarily, exponential decay at rate λ − , cancelling some of the previous growth. If µ + , µ − denote the two ergodic lifts. we must have µ = (1 − ρ) µ + + ρ µ − and thus, In the language of Theorem C (and since µ = µ 1 by ergodicity), the discontinuity ratio β is: (A different description of β will be given below.) So if µ is ergodic, then β is a function of ρ, whence of the amount of mass which escapes to E s . In the case where µ is not ergodic, the different ergodic components of µ have to be considered, and some of them may have zero Lyapunov exponents. The way in which ν ktypical orbits approximate those ergodic components determine the possible cancellations. So if µ is not ergodic, then β may depend on the entire sequence (ν k ), not just on its limit µ.
2.2.
Neutral blocks, the decomposition of µ, and the parameter β. Recall the measurable f -invariant decomposition X = E s ∪ E u ∪ E 0 defined by the Oseledets theorem according to the sign of the limit 1 n log Df n x | E . It has full measure with respect to any f -invariant measure. To get quantitative estimates, we select compact subsets K * ⊂ E * , for * ∈ {s, u, 0}, from which the contraction, expansion, or "central" behavior of the sequence Df n x | E are uniformly controlled, and such that the µ-measure of K := K s ∪ K u ∪ K 0 is close to 1. Since each E * is invariant, we can choose these compact sets to be nearly invariant: Points in a very small neighborhood stay close for a long time.
Hence, if x 0 is a ν + k -typical point for some very large k, its orbit under f spends nearly all its time close to K and every visit in a small neighborhood of K s ∪ K 0 is the beginning of a long period of uniform contraction (or weak expansion/contraction). One expects no entropy creation not only during this period, but also during the "recovery period" which follows, i.e., until the expansion predicted by the Lyapunov exponent of ν k cancels this period of contraction (or weak expansion/contraction).
We select such long time intervals along the orbit of x 0 in the following greedy way. Fixing α > 0 small and L large, an (α, L)-neutral block is a maximal interval of integers (n 0 , . . . , n 0 + ) such that ≥ L and Df n f n 0 (x0) | E u ≤ exp(α(n − n 0 )) for all 0 < n ≤ .
We will check that indeed, there is very little if any entropy creation during neutral blocks. Our estimates will be in terms of the distribution of these long neutral blocks. Let x k = (x k , E u (x k )) be ν + k -generic points and let N α,L ( x k ) denote the union of all (α, L)neutral blocks of the orbit of x k . In Section 6 we show that it is possible to choose a subsequence k i → ∞ so that following limits make sense weak- * on M for ( ν + k1 × ν + k2 × · · · )a.e. ( x k1 , x k2 , . . .): Notice that the sum of the two limits in the brackets is a.s. ν + ki , because this is the limit of the empirical measure of x ki , and x ki are all a.s. ν + ki -generic. So The measures m 0 , m 1 are f -invariant. We will see that ϕd m 0 = 0, and that m 1 is carried by E u . The decomposition µ = βµ 1 + (1 − β)µ 0 in Theorem C is defined by where π : M → M is the natural projection. Note that β is indeed the discontinuity ratio lim k λ + (f, ν k )/λ + (f, µ 1 ) and the quantity 1 − β coincides with the fraction of the time spent in maximal neutral blocks. The measures µ 0 , µ 1 and β depend not just on µ, but also on the way the measures ν + k accumulate on µ.
2.3.
Upper bound on the entropy. To complete the proof of the theorem it remains to show that lim k→∞ h(f, ν k ) ≤ βh(f, µ 1 ). This is the heart of the proof, and where most of the difficulties lie. We use Ledrappier-Young Theory and Yomdin Theory.
• Ledrappier-Young theory bounds h(f, ν k ) by the exponential rate of growth of the minimal number of (n, )-balls needed to cover a definite fraction of a local unstable manifold W u loc (x k ), where x k is a fixed ν k -typical point and the scale ε tends to zero. The "fraction" is measured using the conditional measure ν u x k of ν k on W u loc (x k ). In particular, it suffices to follow points • Yomdin theory provides tools for controlling the number of (n, ε)-balls needed for such covers, for C r maps. Instead of working with (n, ε)-balls, one works with parametrized pieces of unstable manifolds which lie inside (n, ε)-balls and which have uniformly bounded C r size, and Yomdin Theory allows to bound the number of such pieces. Here the regularity assumptions on f come into play. The expression (3) is due to Yomdin theory (see section 2.4).
Let us sketch our argument for the upper bound on the entropy using the neutral blocks. Since the unstable lift of ν u x k -almost every point is ν + k -typical, neutral blocks represent roughly a fraction 1 − β of their time. During a neutral block, typical points on a small piece of f n (W u loc (x k )) do not separate much, therefore this piece remains small (or can be kept small by a subdivision into a small exponential number of pieces). For the rest of the time, these subcurves follow the ergodic components of µ 1 , hence they experience entropic separation at an exponential rate given by h(f, µ 1 ). Since the time outside neutral blocks is a proportion β of the total time, this leads to the bound This argument explains the link between the entropy bound and the semicontinuity defect of the Lyapunov exponents.
This sketch glosses over several difficulties. We will only comment on the main issue: How to use non-expansion of the linearization Df at (x k , E u (x k )) during a neutral block, to infer non-expansion of the map f itself on a small piece of W u loc (x k ) during this neutral block. The difficulty is in controlling Df on (x k , E u (x k )) for x k close to x k .
2.4.
Control of the expansion during neutral blocks. This is one of the most delicate points in the proof. To deduce the non-expansion of the small piece of the unstable manifold containing this point, we need to know that not only the diameter of this curve is small but that its tangent is almost constant too. This forces us to work with pieces of W u loc (x k ) whose lifts to M are also small: The size in the fiber of M measures the variability of the tangent directions.
How small is small enough? To use information on Df to control what happens on W u loc (x k ), we need the fluctuations of the tangent direction along any piece to be smaller than some ε > 0, determined (mostly) by the modulus of continuity of Df . Using the uniform continuity of the measurable unstable bundle on a set of large measure, we find an ε > 0 such that if the diameter of the projection to M is less than ε, then the fluctuation of the tangent is smaller than ε.
The price we pay for this solution is that we need to work with different scales in M and along the fibers of the bundle M → M . This leads us to introduce fibered (n, ε, ε)-balls, and to work with Yomdin theory for f : M → M . When dealing with diffeomorphisms of finite regularity, there is an additional price to pay: If f is C r , then f is only C r−1 , and this accounts for the extra term from Yomdin theory
2.5.
Organisation of the paper. The different ingredients of the proof appear as follows in the text. Section 3: background on tangent dynamics and Lyapunov exponents. Section 4: results from Ledrappier-Young and Yomdin theories on the entropy in differentiable dynamics. Section 5: reparametrization lemmas estimating the entropy from neutral blocks and other time intervals. Section 6: neutral decomposition of typical orbits. Section 7: proof of the technical version of our main theorem. Section 8: proof of the remaining statements.
A remark on style. Our constructions, estimating entropy for a sequence of measures converging to a nonergodic one, require many parameters. We have chosen to make the dependences as explicit as possible to help the reader check that there is no circular argument.
2.6. Standing notations for the duration of the paper. We collect here some notations that we will use frequently below.
• |X| or Card(X): the cardinality of a set X.
• M is a compact Riemannian C ∞ manifold without boundary, with tangent bundle T M , tangent spaces T x M , and Riemannian norm · x . Derivatives of maps f : M → M are denoted by Df :
Tangent dynamics and the semi-continuity of Lyapunov exponents
Let M be a smooth compact Riemannian surface without boundary.
3.1.
Review of the C r size of maps. Let U be an open subset of R n .
Given k ∈ N, we say that a map F : U → R d is C k if for all ω ∈ (N ∪ {0}) n such that |ω| := ω 1 + · · · + ω n = k, the partial derivative ∂ ω F := ∂ ω1+···+ωn F ∂ ω1 x 1 · · · ∂ ωn x n exists and is continuous on U . For any compact subset K ⊂ U , we then define the C k size Given α ∈ (0, 1), we say that a map F is C α if the following quantity is finite for any compact set K ⊂ U , Given r > 1 which is not an integer, we decompose it as r = k + α, with k = r and α ∈ (0, 1). We say that F is C r if it is C k and each partial derivative ∂ ω F , |ω| = k is C α . For any compact set K ⊂ U , we define the C r size Let Ω be a compact subset of R n which is equal to the closure of its interior (we mostly need [0, 1] n ). A map F : Ω → R d is C r if F has a C r extension to an open neighborhood of Ω. In this case, the C r size of F on Ω is This (finite) quantity is independent of the extension of F to the neighborhood of Ω. Notice that the C r size of a constant function is zero.
A C r structure on a smooth manifold N is defined by a maximal atlas A with C r changes of coordinates. A smooth manifold equipped with a C r structure A is called a C r manifold. A finite subset of A which covers N is called a C r atlas of N .
Let N 1 , N 2 be two compact C r manifolds (later this will be M , M or the circle S 1 ), and let A i be finite C r atlases of N i . Let Ω be a compact subset of N 1 equal to the closure of its interior. We say that f : Again, the constant map has size zero.
The quantity f C r depends on the choice of atlases A i , but if N i are compact, then finite atlases induce equivalent C r sizes. In case N 1 = S 1 , we will always use the Euclidean atlas.
Suppose f k , f ∈ Diff r (M ) and 1 ≤ r < ∞. We will say that f k converges to f uniformly in a C r -bounded way, if f k → f uniformly, and sup k≥1 f k C r < ∞. We write in this case The Arzela-Ascoli theorem implies the following.
Lemma 3.1. Let N 1 , N 2 be compact C r manifolds, and f, f 1 , f 2 , . . . : N 1 → N 2 be a collection of C r maps such that (f k ) converges to f uniformly, and sup k f C r < ∞. Then (f k ) converges to f in the C -topology for any < r, ∈ N.
Thus, if for some real r > 1 s.t. r ∈ N, f, f 1 , f 2 , · · · ∈ Diff r (M ) where M is a compact manifold, then f k r−bd −→ f implies that f k → f in the C r -topology.
The projective tangent bundle. Let
It can also be viewed as the image of {v ∈ T x M : v x = 1} by the two-to-one map v → Span{v}. These identifications allow us to endow P x M with a topology and with a smooth structure, and to identify the tangent spaces T E (P x M ) with {w ∈ T x M : w ⊥ E}. We can also pull back the induced Riemannian inner product on {w ∈ T x M : w ⊥ E} to an inner product on T E (P x M ). This endows P x M with a Riemannian structure. The resulting Riemannian distance on P x M is simply dist(E 1 , E 2 ) = | (E 1 , E 2 )|. With this structure, P x M is isometric to the circle with perimeter π.
The projective tangent bundle (or just "projective bundle") of M is the bundle ( M , π, M ) where π : M → M is the natural projection π(x, E) = x, and M is a smooth compact three-dimensional manifold. We endow it with the Riemannian metric √ ds 2 + dθ 2 , where ds is the length element on M and dθ is the length element on P x M . Points in M will be denoted by x = (x, E).
If f is of class C r , then f is of class C r−1 . Notice that π • f = f • π, and ).
Every f -invariant probability measure ν on M projects to an f -invariant probability measure ν on M given by We call ν the projection of ν, and ν a lift of ν. In what follows, when we "lift", we always mean an f -invariant lift. The following lemma is a well-known consequence of the compactness of M . (1) Every f -invariant probability measure ν has at least one lift ν.
(2) If ν is f -ergodic and ν lifts ν, then a.e. ergodic component of ν is a lift of ν.
Hence every ergodic f -invariant probability measure has at least one ergodic lift.
By the subadditive ergodic theorem, the largest Lyapunov exponent also satisfies Throughout this paper, an ergodic invariant probability measure µ is called hyperbolic if one of its Lyapunov exponents is positive, and the other is negative (sometimes this is called hyperbolic of saddle-type). If µ is hyperbolic, we sometimes write λ u = λ + , λ s = λ − , E u = E + and E s = E − .
3.4.
Semi-continuity of Lyapunov exponents. We will use the dynamics of the projective tangent bundle to study the semi-continuity properties of (f, µ) → λ + (f, µ) (and by symmetry of (f, µ) → λ − (f, µ)). The principal tool is the function Notice that if f is a C 1 diffeomorphism, then ϕ is bounded and uniformly continuous. We will make frequent use of the following identity: This is because E is a one-dimensional subspace of T x M , and therefore by the chain rule (9) presents the subadditive cocycle log Df n | E for f as an additive cocycle for f . See [28,Prop. 5.1 on p. 328] for a proof of a more general fact (and [19,Lemma 8.7] for the first use of a related idea). (2) If µ has two different Lyapunov exponents, then it has exactly two ergodic f -invariant lifts: Moreover ϕd µ ± = λ ± (f, µ).
For any f -invariant probability measure, it will be convenient to denote When µ is hyperbolic, the lifts µ + , µ − are called the unstable and stable lifts of µ.
Proof. If µ has equal Lyapunov exponents, (1) follows from eq. (9) and the ergodic theorem. Otherwise, by Oseledets theorem, there are two a.e. defined sections . Every f -invariant probability measure carried by the graph of an invariant section x → E x is ergodic, and coincides with M δ (x,Ex) dµ(x). So (1) and (2) follow from Lemma 3.2 and (9) (see [28]).
Proof. By a general Borel construction, the graphs of E + , E − are measurable. There are unique lifts µ + , µ − of µ to graph(E + ), graph(E − ), and it is easy to check using the identity that µ ± are f -invariant. Now let µ be an arbitrary lift of µ and consider its ergodic decomposition µ = µ ξ dm. For almost every ξ, the ergodic measure µ ξ = µ ξ • π −1 has two different exponents, hence, by Lemma 3.3, its lift µ ξ is some combination a In particular, any lift µ is carried by the union of the graphs of E + and E − .
is continuous as a function on Diff 1 (M ) × {probability measure on M }. Eq. (8) now give us the following "folklore" fact: The next result computes the defect in continuity λ + (f, µ) − lim sup λ + (f k , ν k ) in terms of the dynamics on the projective bundle. It relates the defect in continuity to the escape of some of the mass of the lifts to graph(E + ) to the vicinity of graph(E − ).
We split the set of ergodic components µ ξ (ξ ∈ Ω) by considering whether they are carried by the invariant line bundles E + or E − or by a subset of M where these bundles are not defined: Theorem 3.6 (Defect in continuity). Let M be a compact smooth boundaryless surface.
Notice that the defect in continuity originates at Ω − , the set of ergodic components of lim ν + k which are carried by graph(E − ). This confirms the heuristic that discontinuity in Lyapunov exponents is due to the asymptotic escape of mass from graph(E + ), which carries ν + k , to graph(E − ), which carries µ ξ for ξ ∈ Ω − .
3.5.
A bound for the asymptotic dilation of f .
Proof. Working locally in charts, we identify the iterates f n locally with diffeomorphisms F i : U i → R 2 defined on open subsets of R 2 . We choose the charts so that the change of coordinates distorts the metric by a factor of less than 2. The lift f n ∈ Diff 1 ( M ) is identified with: In what follows, we omit the first factor. The differential of F i can be computed in a straightforward way (writing a * b for the scalar product of two vectors in R 2 ): Thus, . We apply this to some iterate of f on M , remembering the distortion in the metric: Let 0 < ε < 1/4. Fix n an integer so large that 32 ≤ e εn/4 and By dilating the metric on M , we can ensure that D 2 f n 1/n sup ≤ (ε/4)e λ(f ) without changing the asymptotic dilations. Thus λ( f ) ≤ λ(f ) + λ(f −1 ) + ε As ε is arbitrarily small, the claim follows.
These computations allow the following control of the C r−1 size of the lift of a C r diffeomorphism.
Proof. We first consider the kth derivative of f in charts for the maximal integer k ≤ r. A straightforward induction on the integer k ≥ 2 based on eq. (12) shows that the (k − 1)th differential of F at some point (x, v) ∈ M can be written as a linear combination of terms: where p, α 1 , β 1 . . . , α j , β j , γ are integers, and α 1 , β 1 , . . . , γ ≤ k. The coefficients of this linear combination depend only on k.
If k = r, the claim is immediate. If α := r − k > 0, recall that the C r size is the sum of the C k size and α-Hölder size of the k-th derivative. A further computation using the above expression gives the required bound for the Hölder constant of order α of D k g.
Entropy formulas and reparametrizations
We saw in last section that the defect in continuity of (f, µ) → λ + (f, µ) can be described in terms of the canonical lift f : M → M . In this section we develop tools for studying the Specifically, we will show that the entropy of hyperbolic measures on M can be studied in terms of the exponential rate of growth in C r -complexity of f n • σ, where σ : [0, 1] → M is the curve σ(t) = (σ(t), R.σ (t)) and σ : [0, 1] → M is a smooth parameterization of a local unstable manifold.
4.1.
Review of entropy and the ergodic decomposition. This section collects several classical facts on the entropy theory of non-ergodic measures. For proofs and details, see [15, chap. 13].
Consider a compact metric space X together with a continuous map T preserving an invariant Borel probability measure µ and the σ-algebra X of Borel subsets of X. The ergodic decomposition of µ with respect to T is: where µ x := lim n→∞ 1 n 0≤k<n δ T k x (the weak- * limit exists almost everywhere by the ergodic theorem).
The map x → µ x is µ-measurable with respect to the σ-algebra I of invariant measurable subsets; for µ-a.e. x ∈ X, µ x is a T -invariant and ergodic Borel probability measure and for every Borel µ-absolutely integrable u, x → µ x (u) belongs to L 1 (µ) and µ(u) = X µ x (u) dµ(x). The metric entropy of µ and µ x are related by e., is measurable with respect to the µ-completion of I. Let ξ n (x) denote the atom of ξ n which contains x. The Shannon-McMillan-Breiman theorem, in its version for non-ergodic measures [15, (13.4)], states that, In addition, we have the following identity: [15, (13 We call h(T, µ) the essential supremum entropy of µ:
Bowen and Katok entropy formulas.
Let T : X → X be a continuous map on a compact metrix space X. An (n, )-Bowen ball is a set of the form The (n, ε)-covering number of a subset Z ⊂ X, is Bowen [8] defined the topological entropy of a (possibly non-invariant) set Z ⊂ X for T to be and showed that the topological entropy of Katok gave a similar formula for the metric entropy of an invariant measure. Let µ be an invariant probability measure. For every γ ∈ (0, 1), let He showed that if µ is ergodic, then h(T, µ) = lim λ→1 lim ε→0 lim sup n→∞ 1 n log r T (n, ε, µ, γ). Katok's proof in [26] also works in the non-ergodic case, if we replace the usual Shannon-McMillan-Breiman Theorem by (13). The result is that for a general (possily non-ergodic) invariant probability measure µ, Here h(T, µ) is the essential supremum entropy from (14).
4.3.
The lift to the projective tangent bundle preserves entropy. This is a standard consequence of the following theorem [30].
Thus by the variational principle, in the setup of the theorem, h top (T ) = h top (S). Proof. We check condition (17) for T = f , S = f , and apply the previous theorem. M is a topological bundle over M , and its fibers P x M are homeomorphic to circles. The map f : For every ε > 0, one can find partitions ξ x of P x M into a bounded number of arcs with diameter at most ε. It is easy to see that Recall the natural projection π : M → M , π(x, E) = x. The fibered ball with center x ∈ M and scales ε, ε > 0 is the set B( x, ε, ε) := { y ∈ M : d( x, y) < ε and d( π( x), π( y)) < ε}.
Suppose f ∈ Diff 1 (M ) and f is the canonical extension of f to M . The fibered (n, ε, ε)-Bowen ball with center x ∈ M , size (ε, ε) and length n is the set and d(f k ( π( x)), f k ( π( y))) < ε}.
The (n, ε, ε)-covering number of a subset Z ⊂ M is the minimal number of fibered (n, ε, ε)-Bowen balls whose union contains Z. It is denoted by Similarly, given an ergodic measure µ of f and a number 0 < γ < 1, the (n, ε, ε, γ)covering number of µ is the minimal number of fibered (n, ε, ε)-Bowen balls whose union has µ-measure at least γ. It is denoted by r f (n, ε, ε, µ, γ).
Proof of the Claim. Let ε, α > 0. Note first that Recall that a set-valued function F from a topological space X to the set of subsets of a topological set Y is called upper semi-continuous, if for every E ⊂ Y closed, {x ∈ X : F (x) ∩ E = ∅} is closed. The continuity of π and the compactness of M implies that It follows that if π −1 (x) is contained in some open set U (say the union of a minimal cover by fibered Bowen balls), then π −1 (y) is contained in U for all y sufficiently close to x. Hence there is an r x > 0 such that Using a compactness argument, we see that n * := sup{n x : x ∈ M } is finite and that one can arrange for ε * := inf{r x : x ∈ M } to be positive.
4.5.
Curves and C r reparametrizations. The entropy of a diffeomorphism can be related to the exponential rate of growth in C r complexity of the iterates of a local unstable manifold.
To do this we need to control the curvature, and for this purpose it is useful to lift the curve to projective bundle M and study its iterations there. Here we develop the tools needed for doing this.
In this case, it has a canonical lift σ : Here and throughout, R.
The curve has diameter less than (ε, ε) if To say that a curve has finite C r size implies that it is regular and C r .
Remark 4.7. If a curve σ is parametrized by length (i.e., σ (t) = 1 for all t), then it has size at most ( σ C r , C σ C r−1 ) for some constant C > 0 which depends on the choice of C r−1 atlas of M used to define C r−1 size (see section 3.1).
Cutting sufficiently finely, we can obtain covering by pieces with affine reparametrizations with C r size as small as we wish.
Yomdin measured the C r complexity of a curve (more generally a set) by counting how many reparametrized pieces with C r size less than 1 are needed to cover it. We adapt this to the projective dynamics: Let f ∈ Diff r (M ) and let σ be a regular C r curve. We will be interested in families of reparametrizations which remain bounded in C r size after application of f n for certain n. Specifically, fix numbers ε, ε > 0, an integer N ≥ 1, and T ⊂ [0, 1]. Definition 4.9. A reparametrization ψ of σ is (C r , f, N, ε, ε)-admissible up to time n, if there exists an increasing sequence (n 0 , n 1 , . . . , n ) such that We call the integers n j the admissible times.
which is (C r , f, N, ε, ε)-admissible up to time n .
We claim that 0 = n 0 < n 1 < · · · < n +m = n + n are admissible times for ψ • φ. That the gaps are no larger than N is clear. If j = + 1, . . . , + m, then f nj • ψ • φ is a regular curve with size < (ε, ε), because φ ∈ R ψ and R ψ is admissible. If j ≤ , then using the fact that φ = c with c a constant s.t. |c| ≤ 1, we find that Consider g ∈ U * , ε, ε ∈ (0, ε * ), a regular C r curve σ, and a reparametrization ϕ of σ which is (C r , g, N, ε, ε)-admissible up to time n. Then for any (x, E) ∈ σ[0, 1], Proof. M is compact, so one can find ε * > 0 and a small C 2 -neighborhood U * of f in Diff r (M ) such that for all g ∈ U * and x, y ∈ M satisfying d( x, y) < ε * , Let us consider g ∈ U * , a curve σ and a reparametrization ϕ as in the statement, with admissibility times n 0 , . . . , n . One gets for any 0 ≤ i < , which immediately implies the conclusion of the lemma.
(the second inequality follows from the first). Notice that y ∈ C ρ (R).
All this shows that σ(T ) ⊂ y∈Cρ(R) 4.6. Yomdin estimates. In this section we discuss a converse to Lemma 4.12: Covers by Bowen balls generate admissible reparametrizations with cardinality of the same order of magnitude. This result is much more delicate than Lemma 4.12, and requires Yomdin's Theorem [48]. Here is the tool we need from Yomdin's work, in a form adapted to our setup.
Theorem 4.13 (Yomdin). Given real numbers 2 ≤ r < ∞ and Q > 0, there are Υ = Υ(r) > 0 and ε Y = ε Y (r, Q) > 0 with the following properties. For every 1], and T := {t ∈ [0, 1] : g( σ(t)) ∈ B( g( x), ε, ε)}. there exists a family R of reparametrizations of σ over T such that (1) for every ψ ∈ R, g • σ • ψ is a regular C r curve with C r size at most (ε, ε), Proof. The reader who will compare this theorem to Yomdin's original statement in [48] will find that the two results are nearly the same, except for the following differences: (1) Yomdin considered the more general case of σ : [0, 1] → M whereas we restrict to = 1; (2) Yomdin did not specify that all reparametrizations are affine as we do; (3) Our result allows r to be real, not just an integer; and (4) We use (n, ε, ε) balls in M , whereas Yomdin used (n, ε) balls in M .
Each family of reparametrizations R i generates a cover of T by the intervals ψ([0, 1]), ψ ∈ R i . Without loss of generality, the interiors of these intervals are pairwise disjoint (otherwise discard some of them and shrink the rest by composing the reparametrizations by affine contractions). We define is an affine diffeomorphism. The image of this new family of reparametrizations contains the intersection of the images of R 1 and R 2 , hence R is a family of reparametrizations of σ over T .
As the reparametrizations are affine and contracting, we see that of the theorem holds. Next, using the order structure on the interval, it is not difficult to show that |R| ≤ Corollary 4.14 (Existence of admissible reparametrizations). For 2 ≤ r < ∞, Q > 0, let Υ(r), ε Y (r, Q) be the constants from Theorem 4.13. Suppose • g ∈ Diff r (M ) and Q r,N (g) < Q, Then there exists a family R of reparametrizations of σ over T , which is (C r , g, N, ε, ε)admissible up to the time n, and with cardinality |R| ≤ C(r, g)r g (n, ε, ε, g • σ(T )), where C(r, g) = Υ(r) Proof. Fix n ≥ 1 and divide with remainder n = qN + p, q ≥ 0, p = 0, . . . , N − 1.
Step 0. If p = 0 move to step 1. Otherwise proceed as follows.
Fix 1 ≤ i ≤ . Yomdin's theorem for g p , σ and x i gives a family of reparametrizations R 0 of σ over T 0 i := {t ∈ [0, 1] : g p ( σ(t)) ∈ B( g p ( x i ), ε, ε)}, which is (C r , g,N, ε, ε)-admissible up to time p (the admissible times are 0, p), and such that we have a reparametrization up to time n, and we stop.
Step 1. Fix ψ ∈ R 0 and apply Yomdin's theorem to g N , g p • σ • ψ, and g p ( x i ). The result is a family R ψ of reparametrizations of g p • σ • ψ over is an admissible family of reparametrizations of σ over T i up to time N + p, with admissible times 0, p, N + p and cardinality sup .
If q = 1, we have a reparametrization up to time n, and we stop.
Otherwise we continue as before to a "step 2" which applies Yomdin's theorem to g N , g N +p • σ • ψ and g N +p ( x i ).
Eventually, at step q, we arrive to a family of reparametrizations R q over T i which is admissible up to time qN + p = n, and which has cardinality Taking the union over i = 1, . . . , , we obtain the family of reparametrizations over T i ⊃ T , as required. 4.7. Entropy and growth of C r complexity of unstable manifolds. In this section f is a C r diffeomorphism, r > 1, of a surface M and ν is an ergodic hyperbolic probability measure of saddle type, ie λ s := λ − (f, ν) is strictly negative, and λ u := λ + (f, ν) is strictly positive. Pesin's Unstable Manifold Theorem says that ν-a.e. x belongs to an unstable manifold W u (x), which is an injectively immersed C r curve and is characterized as: A measurable partition ξ is subordinated to the unstable lamination W u of ν if for νalmost every x ∈ M , the atom ξ(x) is a neighborhood of x inside the curve W u (x) and ξ is increasing: Every atom of f (ξ) is a union of atoms of ξ. By [29], such measurable partitions exist.
Since ξ is measurable, Rokhlin's disintegration theorem applies, and for ν-a.e. x there exists a probability measure ν u x on ξ(x) so that The family {ν u x } is not unique, but given ξ, any two families like that are equal outside a set of x of measure zero. Therefore it is not a serious abuse of terminology to call ν u x the conditional measure on ξ(x).
In this section we use the entropy theory of Ledrappier and Young [27] and especially the following corollary established by Zang (see [50,Remark 1.8]) to show that the entropy of ν can be bounded by the exponential rate of growth of the C r complexity of the curve f n (W u loc (x)), as quantified in the previous section using admissible C r reparametrizations up to time n.
The difference between this result and Ledrappier-Young theory is that Zang assumes C r smoothness for some r > 1 and hyperbolicity, whereas Ledrappier and Young assume C 2 smoothness, but no hyperbolicity.
is independent of n.
Main reparametrization lemmas
This section collects our main technical results on the existence of admissible families of reparametrizations of pieces of unstable manifolds.
The point is to produce families with cardinality as small as possible. The first result provides admissible families of reparametrizations of local unstable manifolds, with cardinality controlled in terms of the entropy. The second result, which is much more subtle, produces much smaller families of reparametrizations for the subset of the local unstable stable where there is little expansion up to some iterate, see Definition 6.1. (7), and µ is a (possibly non-ergodic!) f -invariant probability measure, which projects to an f -invariant measure µ. Q r,N (f ) is given by (20) in the previous section, h is the essential entropy (14), and λ( f ) is the asymptotic dilation of f , see (2) and §3.5.
The statements of the following two propositions should be formally understood as stating the existence of functions N 1 , n 1 , γ 0 , and N 0 with values in (0, ∞) such that the following stated properties hold.
-R is (C r , g, N, ε, ε)-admissible up to time n, -|R| ≤ exp n h(f, µ) + λ( f ) r−1 + η , The following and key estimate applies to the part of the local unstable manifold which does not (initially) see much expansion. More precisely, suppose g is a diffeomorphism with canonical lift g, and let α > 0. An orbit segment with length n is a string ϑ := ( x, g( x), . . . , g n−1 ( x)).
The proofs of these two propositions may be skipped at the first reading. -Fix an integer N 1 = N 1 (r, f, η) ≥ 1 such that for all N ≥ N 1 , This is possible since λ( f ) = lim N →+∞ -Set ε Y := ε Y (r, Q) as in Yomdin's Theorem 4.13.
-Let U 1 = U 1 (f, η, N, n, ε, ε) ⊂ Diff r (M ) be a small enough C 2 neighborhood of f in Diff r (M ) such that the lift g of any g ∈ U 1 satisfies: and so that every fibered (n, ε 2 , ε 2 )-Bowen ball for f is contained in a fibered (n, ε, ε)-Bowen ball for g.
-Let V = V(f, η, N ) be a C 2 neighborhood of f in Diff r (M ) such that for g ∈ V, D g k sup ≤ e η/10 D f k sup (k = 1, . . . , N ).
Step 3 (Control of N iterates starting near K # ).
Claim 5.4. Suppose n > n * /γ. Any orbit segment ( x, g( x) . . . , g n−1 ( x)) which spends a proportion of time larger than 1 − γ in U 0 can be decomposed into: (a) orbit segments of length n * with initial point in W 0 , (b) orbit segments of length N with initial point in W # , (c) orbit segments of length 1, of total number less than 2γn.
( W 0 and W # are not necessarily disjoint, so the decomposition may not be unique.) Proof. A decomposition as in the statement is completely characterized by the increasing sequence of times (n 0 , n 1 , . . . , n ), where g ni ( x) is the initial point of the i-th segment. (So n 0 = 0, n = n.) We set n 0 = 0 and define the sequence inductively. Assuming that n i < n has already been defined, we set (a) n i+1 := n i + n * if n i + n * ≤ n and g ni ( x) ∈ W 0 \ W # , (b) n i+1 := n i + N if n i + N ≤ n and g ni ( x) ∈ W # , (c) n i+1 := n i + 1 otherwise. We stop when n i+1 = n.
Since n * ≥ N , the times n i such that g ni ( x) ∈ U 0 but which are not associated to case (a) or (b) must satisfy n i > n − n * . Since n > n * /γ and since ( x, . . . , g n−1 ( x)) spends a proportion of time larger than 1 − γ in U 0 , the set of times n i corresponding to case (c) has size smaller than 2γn.
The sequence of times θ := (n 0 , n 1 , . . . , n ) obtained in the previous claim is called type of a decomposition. Recall that H(t) = t ln 1 t + (1 − t) ln 1 1−t . Claim 5.5. There exists n H := n H (γ) such that for all n > n H , the number of possible types θ is less than exp(H(4γ)n).
Proof. By our choices of N, N 0 and n * , we have n * , N > 1/γ. Hence there can be at most γn times n i such that n i+1 − n i ∈ {n * , N }. Since there are also at most 2γn times n i such that n i+1 − n i = 1, we must have ≤ 3γn .
Step 6 (Definition of U 0 , n 0 ). We fix the last parameters of our construction. Recall the C 2 neighborhoods of f in Diff r (M ) introduced in Lemma 4.11 and eqs. (29), (30), (32).
Step 7 (An inductive scheme). Now we fix some g ∈ U 0 with Q r,N (g) < Q, a regular C r curve σ with C r size at most (ε, ε) and n > n 0 . We need to bound the minimal cardinality of a family R n of reparametrizations of σ which are (C r , g, N, ε, ε)-admissible up to time n, over the set For each type θ = (n 0 , . . . , n ), we introduce the corresponding subset , g( x), . . . , g n−1 ( x)) has type θ .
Then T is the union of T θ over all possible type θ. Fixed some type θ = (n 0 , . . . , n ). We will build by induction a family R θ ni of reparametrizations ψ of σ over T θ satisfying the following properties: (i) admissibility: R θ ni is (C r , g, N, ε, ε)-admissible up to time n i ; (ii) small cardinality: if i ≥ 1, (iii) small length: for each ψ ∈ R θ ni and any (x, E) ∈ σ • ψ([0, 1]), At the end of the construction, one obtains a family R θ n := R θ n over T θ which is admissible up to time n. Then one can take the union over all θ and finish the construction.
We begin the construction by defining R θ 0 := {Id}: This meets our requirements because n 0 = 0 and σ has C r size at most (ε, ε). Now we assume by induction that R θ ni has been constructed, and we build R θ ni+1 . The construction uses the concatenation procedure described in Lemma 4.10. For each ψ ∈ R θ ni we have to build a family R ψ of reparametrizations of the curve g ni • σ • ψ over ψ −1 (T θ ) with the following properties: (i') R ψ is (C r , g, N, ε, ε)-admissible up to time n i+1 − n i , The family R θ ni+1 := {ψ • ϕ, ψ ∈ R θ ni ϕ ∈ R ψ } then satisfies (i-iii) above.
-Given a type θ = (n 0 , . . . , n ), an integer i ∈ {0, . . . , − 1} and a reparametrization ψ ∈ R θ ni , the construction of the families R ψ depends on which of the following cases from Claim 5.4 holds for n i : The three cases are discussed in steps 8-10 below.
Step 9 (Case (b)): In this case n i+1 −n i = N and σ (T ) ⊂ W # . We combine Lemma 4.11 with (33) and get that, for each (x, E) ∈ g −ni • σ ([0, 1]), One can thus subdivide [0, 1] into intervals I 1 , . . . , I m with m ≤ e η 5 N ≤ e 3η 10 N , such that: We can focus on the intervals I j such that I j ∩ T = ∅. Fixing such an I j , there exists On the other hand since σ (T ) ⊂ W # , eq. (31) implies: We have shown that the image g N • σ (T ∩ I j ) is contained in a (ε, ε)-ball. We can apply Yomdin's Theorem 4.13 and obtain a family R j of reparametrizations ϕ of σ over T ∩ I j with cardinality at most Υ D g N 1/(r−1) sup such that each curve g N • σ • ϕ has C r size at most (ε, ε). Consequently, R j is (C r , g, N, ε, ε)-admissible up to time N .
The union R ψ := j R j is thus a family of reparametrizations ϕ of σ over T which is (C r , g, N, ε, ε)-admissible up to time N . By (36) they satisfy the bound (iii') of the induction scheme (Step 7) on the length of σ • ϕ. Combining the bounds on m and |R j |, one bounds the cardinality of R ψ by |R ψ | < Υe which by (27), (29) and N > 1 γ , is bounded by exp( λ( f ) r−1 N + 6η 10 N ) as required.
Step 11 (Completion of the proof ). Steps 7-10 provide the construction of the family R θ n for each type θ. The inductive bounds (ii) for |R θ ni |/|R θ ni−1 | imply where A c (θ) is the number of times n i belongs to case (c) for the type θ. Let R n denote the union of R θ n over all possible types θ. This is a family of reparametrizations of σ over T = θ T θ , which is (C r , g, N, ε, ε)-admissible up to time n.
Since A c (θ) ≤ 2γn for all θ (by Claim 5.4), since the number of types θ is bounded by exp(H(4γ)n) (by Claim 5.5), and since H(4γ) < η 10 (by our choice of γ, see (27)), this gives This concludes the proof of Proposition 5.3.
The neutral decomposition
Let f be a homeomorphism on a compact metric space X. We denote the point mass measure at x ∈ X by δ x . Given N ⊂ N, let The weak- * limit points of (µ N x,n ) n≥1 are called the N-empirical measures of x. Definition 6.1. Suppose ϕ : X → R is continuous, α > 0 and L ≥ 1. An (α, L)-neutral block of (x, f, ϕ) is an interval of integers (n 0 , n 0 + 1, . . . , n 1 − 1) s.t.
Proof of the claim. Fix some countable dense set E ⊂ (0, 1]. By compactness and a diagonal argument, there is an increasing sequence k i → ∞ such that the following limits exist in the weak- * topology: ∀(α, L) ∈ E × N lim p→∞ χ ki α,L ν ki = m α,L . Let us check that this can be extended to all (α, L) ∈ (0, 1] × N, maybe after passing to a subsequence. Indeed, select a countable family (u j ) of nonnegative continuous functions which generate a countable dense algebra over Q in C 0 (M ) (the space of continuous real-valued functions on M with the supremum norm). Fix L and u j . The function α ∈ E → m α,L (u j ) is non-decreasing on E, and therefore extends uniquely to a leftcontinuous function α ∈ [0, 1] → m α,L (u j ). The discontinuity points form a countable set D L,j . Again by monotonicity with respect to α, at every α ∈ [0, 1) \ D L,j . By a further extraction of a subsequence, we ensure that χ α,L ν ki converge for all (α, L) in the countable set L∈N,j≥1 D L,j .
We return to the proof of Proposition 5.2. To simplify notation, from now on (ν k ) will denote the subsequence (ν ki ).
Neutral blocks have length at least L, therefore for every continuous function u, we have This proves items (i) and (ii).
Item (iii) is a simple consequence of the construction.
We turn to (iv). For any function ψ : X → R, we define For every x, we decompose N k α,L (x) ∩ [0, ∞) into maximal disjoint intervals: Since ν k is ergodic, for ν k -a.e. x, Each interval [a i , a i + b i ) is a maximal (α, L)-neutral block except possibly the initial one, if it contains 0. The first block contributes C 0 (x)/n → 0 to the limit. The other blocks are all maximal neutral blocks, and satisfy the bounds ) ≤ αb i . The first inequality comes from the maximality of the block, the second is the definition of neutrality. Summing over i = 1, . . . , j, we obtain the bounds Since each such complete neutral block has length at least L, there are j ≤ n/L maximal (α, L)-blocks in [0, a j + b j ). Dividing by a j + b j ≥ i≤j b i and discarding some nonnegative terms from the lower bound, we obtain in the limit j → ∞, − sup x,k ϕ k (x)/L < (χ α,L ν k )(ϕ k ) ≤ α.
for k large enough and ν k -a.e. x. Finally, let V denotes the set of j such that f j (x) ∈ V k 0 (K 0 ) \ N k α1,L1 . Then by eq. (40), for k large enough and ν k -a.e. x, We will show that (41)-(43) lead to a contradiction. By definition of V k 0 (K 0 ), each j ∈ V is the last element of an (α 0 , L 0 )-neutral block I(j) with length ≤ K 0 (we do not claim that this block is maximal). Let We claim that the upper asymptotic density d(I ) := lim sup 1 n |I ∩ [0, n)| is less than γ/100. To see this note that if j ∈ V and I(j) ∩ N 1 = ∅, then j ∈ N 1 (by definition of V) and since L 0 < L 1 , I(j) contains the last element of a maximal sub-interval of N 1 . The interval with length 2K 0 centered at this last element must contain I(j). Since the number of maximal sub-intervals of N 1 ∩ [0, n] is bounded by n/L 1 , the upper asymptotic density of I is no more than 2K 0 /L 1 < γ/100.
Proof of the main theorem
We recall the notation λ( f , µ) := M log Df x | E d µ(x, E). In this section, we prove the following stronger version of Theorem C.
Theorem D. Fix a real number r > 2. For every k ≥ 1, let f k ∈ Diff r (M ) and let ν k be an f k -ergodic measure. Let ν k be an f k -ergodic lift satisfying λ( f k , ν k ) = λ + (f k , ν k ) such that: -the limits lim k λ + (f k , ν k ) and lim k h(f k , ν k ) exist and lim k λ + (f k , ν k ) ≥ 0, -f k r−bd −→ f for some f ∈ Diff r (M ) (i.e. f k → f uniformly and sup k f k C r < ∞), → µ for some f -invariant probability measure µ on M , perhaps non-ergodic.
Note that when ν k is hyperbolic, the measure ν k above is simply the unstable lift ν + k . 7.1. Reductions. We assume the setting of Theorem D. There is no loss of generality in assuming that r is finite, since the C ∞ case follows from the C r case by letting r → ∞. By Lemma 3.1 and since f k f k → f in the C 2 topology and f k → f in the C 1 topology.
In particular, each measure ν k is hyperbolic, i.e. has one positive and one negative Lyapunov exponent. Note that it is enough to prove the theorem for any convenient further subsequence.
7.2. The decomposition of the limiting measure. Theorem D is stated in terms of the properties of a special decomposition µ = (1 − β)µ 0 + βµ 1 of the µ = lim ν k . In this section we construct β, µ 0 and µ 1 .
The idea is to apply Proposition 6.2 to a suitable sequence of measures. By Ruelle's inequality and the reduction to the case h(f k , ν k ) > 0, ν k must be f k -hyperbolic of saddle type. Let ν + k denote the unstable lift of ν k to M , and let f k , f be the lifts of f k , f to M . Define ϕ k , ϕ : M → R by We apply Proposition 6.2 to M , f k , ϕ k , ν + k . (The proposition is applicable, because by eq. (47), f k → f in Diff 1 ( M ) and ϕ k → ϕ uniformly on M , and because by Lemma 3.3 and Ruelle's inequality, ϕ k d ν k = λ + (f k , ν k ) ≥ h(f k , ν k ) > 0.) Proposition 6.2 gives us a subsequence {k i } and two finite positive measures m 0 , m 1 with the following properties.
The limit µ is f -invariant and lifts µ.
It follows that µ 1 -a.e. x has at least one positive Lyapunov exponent. Assume by contradiction that the claim is false, then there is an f -invariant set Ω of positive µ 1 -measure such that every x ∈ Ω has two (possibly equal) positive Lyapunov exponents. Recall the following well-known fact: Proof. Pesin's local stable manifold theorem [38,Thm 2.2.1] implies that µ 1 -almost every point x admits a neighborhood U x and constants C > 0 and κ ∈ (0, 1) such that for any y ∈ U x and n ≥ 0, d(f −n (x), f −n (y)) ≤ Cκ n . Since the orbit of x is recurrent, this implies that the forward orbit of x converges towards a periodic orbit O and in fact must coincide with that periodic orbit, again by recurrence. As a consequence, x is a hyperbolic sink so Df −N Thus lim k→∞ λ + (f k , ν k ) = βλ + (f, µ 1 ), as required.
7.4. Proof of Theorem D part (2). We now come to the heart of the proof of Theorem D.
Step 1 (The decomposition µ 1 = a c µ 1,c ). We decompose µ 1 into invariant measures µ 1,c all of whose ergodic components have nearly the same entropy.
(b) For all f k with k > k 0 , for any regular curve σ with C r size at most (ε, ε), there exists a family of reparametrizations R of σ over σ −1 ( U 1,c ) such that (b1) R is (C r , f k , N, ε, ε)-admissible up to time n 1,c := n 1 + c, For all f k with k > k 0 , for any different 1 ≤ c, c ≤ , and for any 0 ≤ j ≤ n 1,c , Construction. For each c, we apply Proposition 5.1 to f , µ 1,c and to the parameters η, γ, ε, ε, N and n = n 1,c . This gives an open set U 1,c s.t.
By assumption, the measures µ 1,c (for 1 ≤ c ≤ ) are mutually singular and there exist pairwise disjoint f -invariant measurable sets X c such that µ 1,c (X c ) equals one when c = c , and zero otherwise. Using (a'), one constructs compact sets This inequality remains true if one replaces f by f k with k large enough and the compact sets K 1,c by small enough neighborhoods U 1,c . We may choose those neighborhoods so that µ(∂ U 1,c ) = 0. Replacing each U 1,c by its intersection with U 1,c , we obtain sets satisfying both the conclusion (*) of Proposition 5.1 and: for k large enough. We replace the sets U 1,c by these new U 1,c . Moreover (a') and (b') are preserved.
Step 6 (Decomposition of orbits into orbit segments). Recall that the orbit segment of f k with length t and initial point x is the string ( x, f k ( x), . . . , f t−1 k ( x)). It is associated with the measure An orbit segment will be called neutral if (0, 1, . . . , t − 1) is an (α, L)-neutral block of ( x, f k , ϕ k ) as defined in Section 6, i.e. if t ≥ L and if x = (x, E) satisfies: Using the open sets U 0 , U 1,c and the integers n 1,c defined at steps 2 and 3, we introduce + 2 classes of orbit segments ( x, f k ( x), . . . , f t−1 k ( x)): (a) Segments with color 1 ≤ c ≤ : orbit segments such that x ∈ U 1,c and t = n 1,c . (b) Blank segments: neutral orbit segments such that µ t f k , x ( U 0 ) ≥ 1 − γ. (c) Fillers: orbit segments with length t = 1. The class of an orbit segment as above can be recognized from its length t: If t = 1, it is a filler, if t ∈ [n 1 + 1, n 1 + ], it is colored with color t − n 1 , and if t is larger than L, then it is blank, see (53). So these + 2 classes are disjoint.
Lemma 7.5. For all k > k * and for ν + k -a.e. x, there exists n k ( x) ∈ N such that all the orbit segments ( x, f k ( x), . . . , f n−1 k ( x)) with n ≥ n k ( x) can be decomposed into: (a) colored segments of total length at most βa c n + γn, for each color c, (b) blank segments of total length at least (1 − β)n − 4γn, (c) fillers of total length at most 6γn.
Proof. By the reduction in section 7.1, the ergodic measures ν k have positive entropy, and therefore the ν k measure of f k -periodic points is zero. Thus it is sufficient to consider nonperiodic x only. Orbit segments of non-periodic points can be identified with the non-ordered sets of points they contain without any loss of information, because there is only one way to order them to get an orbit segment. We will therefore feel free to abuse terminology and treat orbit segments as sets, subject to the usual set-theoretic operations.
We call the sequence (t 0 , . . . , t m ) the type of the decomposition since it determines not only how the orbit segment is divided but to which class each segment belong.
By analogy with Section 6, a neutral sub-segment of ϑ is called maximal, if it does not lie in a strictly longer neutral sub-segment of ϑ. Let S neut (ϑ) denote the collection of all maximal neutral sub-segments of ϑ. It is not difficult to see that every neutral sub-segment Let C c denote the union of all colored segments with color c; let B denote the union of all blank segments; and let F denote the union of all fillers.
(b) Blank segments. By definition, every blank segment is neutral, so B ⊂ Neut, and |B| = |Neut| − |Neut \ B|. By (52) and the definition of m 0 , The set Neut \ B is the union of the maximal neutral orbit segments which visit U 0 with frequency less than 1 − γ. Thus γ · |Neut \ B| < m 0 ( M \ U 0 )n. By (52) and the bound (c) Fillers. By construction, a filler is a segment of length one ( y) such that one of the following holds: (i) y does not belong to a colored segment or to a segment in S neut (ϑ); (ii) y belongs to a segment in S neut (ϑ), but this segment is not a blank segment; (iii) y belongs to a segment of length n 1,c which begins at U 1,c , but it fails to be a colored segment because it extends beyond the right endpoint of ϑ; (iv) y belongs to a segment of length n 1,c which begins at U 1,c , but it fails to be a colored segment because it intersects an element of S neut (ϑ).
The fillers of type (i) belong to Neut c \ c U 1,c , so their cardinality is bounded by eq. (56): The fillers of type (ii) belong to Neut\B. As we saw above this means that their cardinality is less than γ(3 − β)n < 3γn.
The number of fillers of type (iii) is clearly bounded by the maximum length of a colored segment max c n 1,c = n 1 + ≤ n 1 + . This can be assumed to be less than γn when n is large enough.
It remains to control the fillers of type (iv). Fix ϑ 0 ∈ S neut (ϑ), and suppose y belongs to a colored segment which intersects ϑ 0 . All colored segments have lengths at most n 1 + , therefore y must belong to one of two segments of length n 1 + adjacent to the endpoints of ϑ 0 . This gives the following bound for the number of fillers of type (iv): 2(n 1 + )·|S neut (ϑ)|. Recalling that S neut (ϑ) consists of disjoint sub-segments of ϑ, each with length at least L, we find that |S neut (ϑ)| ≤ n L .
Thus by (53), the number of fillers of type (iv) is at most 2(n1+ ) L n < γn.
It follows that the total length of the fillers is |F| < 6γn.
Step 7 (A bound on the number of decomposition types). In the previous step we decomposed orbit segments of typical points with length n large enough into colored segments, blank segments and fillers. Let θ = (t 0 , t 1 , . . . , t m ) be the type of the decomposition, see (55) and the discussion which follows it. Here we bound the number of possible types. As always, let H(t) := t log 1 t + (1 − t) log 1 1−t for 0 < t < 1. Claim 7.6. There exists n H := n H (γ) such that the number of types of decompositions of all f k -orbit segments as in Lemma 7.5 with length n > n H and k arbitrary is at most exp[nH(10γ)].
Proof. A decomposition of an orbit segment with length n has -at most γn blank segments (because these have lengths ≥ L > 1/γ), -at most γn colored segments (because these have lengths ≥ n 1 + 1 > 1/γ), -and at most 6γn fillers (by Lemma 7.5). This gives a total of at most 8γn segments.
Step 8 (Conditional measures and choice of N k , F k ). The measures ν k are assumed to be f k -ergodic, and by the reductions in section 7.1 they have positive entropy. So by Ruelle's inequality, each ν k is a hyperbolic measure. As explained in Section 4.7, one can introduce a measurable partition subordinated to the unstable lamination of ν k and associate to it a system of conditionals probability measures ν u k,x . We fix N k ≥ 1 and a Borel set F k ⊂ M with ν k (F k ) > 1 2 such that for every point x ∈ F k and for the diffeomorphism f k : -x has a well-defined unstable manifold, an immersed C r curve W u (x) ⊂ M ; -ν u k,x is well-defined and x belongs to the support of the restriction of ν u k,x to F k ; -x := (x, E u (x)) satisfies Lemma 7.5 with n k ( x) ≤ N k . In particular for each n ≥ N k , the orbit segment ( x, f k ( x), . . . , f n−1 k ( x)) has a decomposition as in Lemma 7.5. Let θ = θ(x, n) be the type of decomposition.
Step 9 (Construction of reparametrizations). Choose a point x ∈ F k which satisfies Corollary 4.16.
Let σ : [0, 1] → W u (x) be a regular C r -curve which parametrizes a neighborhood of x in W u (x) in the intrinsic topology, and which has C r size at most (ε, ε). By the choice of F k , T := σ −1 (F k ) has positive measure for ν u k,x . Fix n ≥ N k , and let ε, ε and N be as in step 2. Our aim is to construct a particular family of reparametrizations R n of σ over T , which is (C r , f k , N, ε, ε)-admissible up to time n. In later steps, we will estimate the cardinality of R n and use Corollary 4.16 to obtain the upper bound for h(f k , ν k ) which completes the proof of the theorem.
We begin by fixing a type θ := (t 0 , t 1 , . . . , t m ) with t m = n, and constructing a family of reparametrizations R θ n of σ admissible up to time n over the set T θ := σ −1 {y ∈ F k with type θ}.
8.1. Discontinuities: construction of the Example 1.2. Let us recall that two transitive hyperbolic sets K 1 , K 2 are homoclinically related if a stable manifold of K 1 has a transverse intersection point with an unstable manifold of K 2 and a stable manifold of K 2 has a transverse intersection point with an unstable manifold of K 1 . In this case there exists a transitive hyperbolic set that contains K 1 and K 2 .
If O is a periodic orbit, we will denote by µ O the invariant probability measure supported on O. We say that a sequence of periodic orbits (O k ) converges weak- * to a measure µ, if the sequence of measures (µ O k ) converges weak- * towards µ.
We say that a C ∞ diffeomorphism f 0 belongs to the Newhouse domain if there exist an attracting region U where | det Df 0 | < 1, a transitive hyperbolic, locally maximal set K ⊂ U (not reduced to a periodic orbit) and a C ∞ neighborhood U of f 0 such that for any diffeomorphism f ∈ U the hyperbolic continuation of K (still denoted by K) admits a stable manifold and an unstable manifold with a non-transverse intersection. The Newhouse domain is open by definition, and non-empty by [34].
We prove the following more precise version of Example 1.2: Proposition 8.1. The Newhouse domain in Diff ∞ (M ) contains a dense G δ subset of diffeomorphisms f with the following property. For any pair of numbers 0 < α ≤ β ≤ 1, there is a sequence of ergodic measures (ν k ) converging weak- * to a measure µ with h(f, µ) > 0 and such that: lim h(f, ν k ) = αh(f, µ) and lim λ + (f, ν k ) = βλ + (f, µ).
Remark 8.2. One can choose for µ any invariant probability measure with positive entropy, ergodic or not, and carried by the hyperbolic set K associated with the Newhouse domain of f .
There is a dense G δ subset of the Newhouse domain in Diff ∞ (M ), made of diffeomorphisms f with the following property. For any periodic orbit P contained in the hyperbolic set K associated to f , there exists a sequence of hyperbolic periodic orbits O k homoclinically related to P which converge weak- * towards P and satisfy λ + (O k ) → 0.
Proof. Let f 0 ∈ U. By an application of Baire's argument, it is enough to find f C ∞ close to f 0 with a periodic orbit O homoclinically related to P which is weak- * close to P and has a top Lyapunov exponent close to 0. We sketch the proof which uses classical arguments on the behavior near homoclinic tangencies, and we refer to [36] for further details. In order to simplify the presentation, we assume that P is fixed and the eigenvalues 0 < λ < 1 < µ of Df (P ) are positive. By dissipation, λ · µ < 1.
Since the stable (resp. unstable) manifold of P is dense in the stable (resp. unstable) lamination of K, and since f 0 belongs to the Newhouse domain, one can perturb f 0 in such a way that P exhibits a quadratic homoclinic tangency z ∈ W s loc (P ). One can also assume that the eigenvalues λ, µ are non-resonant, so that by Sternberg's theorem, there exists a smooth chart on a neighborhood U [−1, 1] 2 of P , where the dynamics is linear: On [−1, 1] × [−µ −1 , µ −1 ], f coincides with the map L : (x, y) → (λ · x, µ · y). The local manifolds W s loc (P ) and W u loc (P ) coincide with {y = 0} and {x = 0}. Moreover z has a preimage z ∈ W u loc (P ) by an iterate f N and one denotes by T the map induced by f N from a neighborhood of z to z. The unstable manifold at z is locally a graph {(x, ϕ(x))} and by a suitable rescaling of the axis of U , one can require that D 2 ϕ 1 near z.
Let us fix δ > 0 small. When n is large one considers a rectangle where a is chosen such that z = (0, a · µ n ). Note that C −1 ≤ a · µ n ≤ C where C depends on the Sternberg linearization domain U , but not on n.
The rectangle R is mapped by f n+N to a thin curved rectangle T • L n (R) whose width is of the order of δλ n , hence much smaller than the width of R. One perturbs f near z in such a way that the transition map T is composed with a vertical translation. The tip of the image can thus be adjusted to be at distance L · a from the rectangle R where L is a large constant independent from n. Therefore f n+N (R) crosses {y = 0} and also R with a slope s close to L · a (since D 2 ϕ 1). See Figure 1. Moreover R ∩ f n+N (R) contains a periodic point q whose unstable direction is dilated at the period by a factor of the order of s exp(−λ + n) L · a · µ n , which is close to a large constant (comparable to L). As the period n + N of q can be chosen arbitrarily large, the unstable Lyapunov exponent of q is close to 0.
Note that the unstable manifold of q crosses f n (R) along its largest dimension (see Figure 1), hence crosses W s loc (P ). The local stable manifold of q is a graph which crosses R horizontally. The image f n (W s loc (P )) if close to f n (R), crosses R, and then the local stable manifold of Q. Hence P and the orbit of q are homoclinically related. As the n first iterates of q belong to the linearization domain U , the orbit of q spends an arbitrarily large proportion of time in any neighborhood of P , as the period n + N goes to infinity, proving that the invariant probability measure supported on the orbit of q gets arbitrarily close to P in the weak- * topology.
We will also need the following fact: Sketch of proof. This is routine, even if we could not locate an exact reference. Observe that it is enough to show this for a transitive subshift of finite type Σ. Given an invariant probability measure on Σ, approximate it by a Markov measure with finite memory N . Taking N sufficiently large, we can make this approximation arbitrarily close, both weak- * and in entropy. By a small modification of the transition probabilities we can make the measure fully supported on Σ, and therefore ergodic.
Proof of Proposition 8.1. For convenience, we fix some distance d on the space of Borel probability measures of M , compatible with the weak- * topology. Let f be a diffeomorphism with a locally maximal transitive hyperbolic set K as given by Lemma 8.3. Since K is not reduced to a single periodic orbit, it carries invariant probability measures with positive entropy. We choose any one of them. Lemma 8.3 yields a sequence (O k ) k≥1 of hyperbolic periodic orbits homoclinically related to K such that d(µ O k , µ) < 1/k and |λ + (O k )| < 1/k.
The sequence (ν k ) k≥1 is as claimed.
Note that if f k ∈ Diff ∞ (M ) and f k → f in C ∞ , then the corollary applies for all r > 2, and Property (c) becomes lim k→∞ h(f k , ν k ) ≤ βh(f, µ 1 ).
Note also that the conclusions (a)-(e) are the same as in Theorem D, except for the extra term λ(f )/r on the right hand side. See the following remark on this term. Remark 8.6. Our proof relies on discretizing the ergodic decompositions of the measures ν k and applying Theorem D to the atoms thus defined and taking a limit. This limiting process is responsible for the term λ(f )/r in the entropy estimate in Property (c).
Using the decomposition in the projective bundle is the key to avoid any such loss in the Lyapunov estimate eq. (b) and is therefore essential for our proof of this generalization.
Proof. Let P(M ) denote the set of Borel probability measures on M , and let d be the L 1 -Wasserstein distance over P(M ). This distance is compatible with the weak- * topology and satisfies d( We fix some ε > 0 and discretize the ergodic decompositions ν k = X ν k,ξ dP k (ξ).
Note that λ + (f, x) ≥ 0 for µ-a.e. x since otherwise the ergodic decomposition of µ would contain a source as an atom and therefore ν k would contain the same atom with uniform weight for large k, in contradiction to our assumption lim inf k min(λ + (f k , x), 0)dν k (x) = 0. Let Z := { x ∈ M : λ + (f, π( x)) = 0}.
since the probability measure µ1( ·∩Z) µ1(Z) has both zero entropy and zero top Lyapunov exponent. This concludes the proof of the Corollary 8.5. | 2021-03-04T09:19:25.096Z | 2021-03-03T00:00:00.000 | {
"year": 2021,
"sha1": "920cac5c4592d8ebdbf4cb5c5252ce12a7a7d31f",
"oa_license": null,
"oa_url": "http://arxiv.org/pdf/2103.02400",
"oa_status": "GREEN",
"pdf_src": "Arxiv",
"pdf_hash": "920cac5c4592d8ebdbf4cb5c5252ce12a7a7d31f",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Mathematics"
]
} |
74248146 | pes2o/s2orc | v3-fos-license | Effectiveness of single and double phototherapy on indirect hyperbilirubinemia in neonates
Background Hyperbilirubinemia is a common problem in full term newborns and phototherapy is the most widespread treatment for lowering bilirubin concentration in neonates. Double phototherapy could increase the effectiveness of treatment. Objective To compare the effectiveness of single and double phototherapy and increasing spectral irradiance for decreasing serum bilirubin levels in neonates for indirect hyperbilirubinemia. Methods An open, randomized, controlled trial was conducted at H. Adam Malik and Pirngadi Hospitals, Medan, from May to December 2009. Subjects were divided into two groups, those who received single phototherapy (n=30) and those who received double phototherapy (n=30) treatments. We included term newborns with neonatal jaundice in the first week of life. Serum bilirubin and average spectral irradiation levels were measured at baseline and after 12 hours and 24 hours of phototherapy treatment. Results The mean total bilirubin levels of the single and double phototherapy groups at the beginning of therapy were 17.6 mg/ dL (SD1.41) and 17.5 mg/dL (SD 1.32), respectively, with no significant difference between values. During the study period the sum of average spectral irradiance by double phototherapy was significantly higher than that of single phototherapy (P < 0.05). A significantly greater decrease in bilirubin levels was observed in the double phototherapy group at 12 hours and 24 hours of phototherapy compared to the single phototherapy group (P = 0.001). Conclusion Double phototherapy is more effective than single phototherapy in reducing bilirubin levels in jaundiced newborns. [Paediatr Indones. 2011;51:316-21].
the back of the baby increase the dose by delivering the same amount of irradiance per square centimeter of skin to a larger skin surface area. 7ur study aimed to compare neonatal serum bilirubin levels and spectral irradiance for single and double phototherapy at 12 and 24 hours of treatment.
Methods
We conducted a randomized, controlled, open trial in Haji Adam Malik and Pirngadi Hospitals, Medan, Indonesia from May to December 2009.All babies admitted to the special care nursery for uncomplicated neonatal jaundice and requiring phototherapy were eligible for the study.The need for phototherapy was determined using the American Academy of Pediatrics guidelines for management of jaundice in healthy term newborns.We excluded infants with serum bilirubin levels close to the exchange transfusion limit, increased direct bilirubin, hemolytic diseases and congenital anomalies.This study was approved by the medical ethics Committee, university of Sumatera utara medical School.
Sample size needed was estimated to be 30 neonates per group.In total, 60 babies were included.Demographic and laboratory data were obtained, including age, gender, hemoglobin and albumin levels.The babies included in the study were randomized to receive either single phototherapy (control group) or double phototherapy.Subjects remained in their assigned group until after 24 hours of phototherapy.Light intensity was measured as spectral irradiance (µW/cm 2 / nm) using a Dale 40 light intensity meter (USA).
Phototherapy units used were manufactured by Tessna, USA.The phototherapy units utilized five compact blue fluorescent lamps (Toshiba 20WT52), with irradiation of 6.6 µW/cm2 by radiometer.Distance between the phototherapy unit and baby was standardized at 45 cm above and 10 cm below the baby.We used blue light photolights (Toshiba 20WT52) with irradiation of 6.6 µW/cm2 by radiometer (Dale).
Primary outcome measures were the mean differences in serum bilirubin levels at baseline, 12 hours and 24 hours.Secondary outcome measures were spectral irradiance of single phototherapy (control group) and double phototherapy.At baseline and at 12 and 24 hours of phototherapy, serum bilirubin levels and spectral irradiance were measured.
The safety of both methods was assessed and compared by monitoring body temperature, hydration status (monitored clinically and by weight measurement), skin problems (such as rashes) and gastrointestinal problems (such as loose stools or feeding intolerance).
Associations between phototherapy types and serum bilirubin levels were analyzed by Student's t-test.We analyzed data with SPSS version 15.0.Statistical significance was accepted as P < 0.05 with a 95% confidence interval.
Results
out of 66 neonates with hyperbilirubinemia, 5 with direct hyperbilirubinemia were excluded, for a total of 61 subjects enrolled in our study (Figure 1).However, 1 subject dropped out due to blood sample damage.
Subjects were divided into two groups of 30 each.one group received single phototherapy and the other group received double phototherapy.Infants' characteristics, including gender, age, initial bilirubin levels before phototherapy, albumin and hemoglobin levels are shown in Table 1.There were significant decreases from the initial bilirubin levels to those measured after 12 and 24 hours in the double phototherapy group.For the single phototherapy group, a significant decrease was observed from the initial to the 24 hour bilirubin level only, but not in the 12 hour measurement (Table 2).
A significantly greater decrease in bilirubin levels was observed in the double phototherapy group at 12 hours and 24 hours of phototherapy compared to the single phototherapy group (P = 0.001) (Table 3).
Figure 2 shows significant differences in irradiance between single and double phototherapy (P < 0.05) at initial, 12 hours and 24 hours of phototherapy.The total irradiation 45 cm above and 10 cm below the neonate's body in the double phototherapy group was approximately 29.2 µW/cm2/nm.
We observed the adverse effect of hyperthermy (T > 37.5ºC) in 8 (13.3% of total) subjects from both the single and double phototherapy groups, 3 and 5 subjects, respectively.Other potential side effects, such as diarrhea and dehydration, were not found during monitoring.
Discussion
The average age of subjects in our study was 4 to 5 days.It is known that elevated bilirubin levels often peak in the first week of life. 3Hyperbilirubinemia is a serious, and potentially life-threatening condition.This condition is the main reason for hospital revisits by full term infants in their first week of life. 8 significant decrease of bilirubin level between baseline and 12 hours and baseline and 24 hours was seen in the double phototherapy group, with an average decrement of 6.5 mg/dl (P=0.001) and 10.1 mg/dL (P=0.001),respectively.For the single phototherapy group, a significant decrease in bilirubin level was observed only between baseline and 24 hours, with an average decrement of 3.8 mg/dL (P= 0.001).
Double phototherapy using blue light with wavelength 430-490 nm and spectral irradiance of ≥ 30 µW/cm2/nm (measured with radiometer or estimated by placing the baby directly under the light or widening the exposed surface) is effective for reducing bilirubin levels. 9,10Double phototherapy was more effective than single phototherapy in our study.Newborns receiving double phototherapy had a larger surface area exposed to constant irradiance thereby increasing their total dose of phototherapy compared to those undergoing single phototherapy.This increased dose causes the production of more lumirubin. 11However, merely increasing the surface area exposed, while maintaining the same total energy output to the skin, does not improve the efficacy of phototherapy. 12,13A randomized, clinical trial in Thailand showed that double phototherapy was safer and more effective in reducing bilirubin levels compared to single phototherapy. 14Double phototherapy is an alternative mode of intensive phototherapy that is effective, economical and easy to use.
A study in Brazil compared the effectiveness of double phototherapy with total irradiance 75.6 µW/ cm2/nm and pharmacotherapy, and found that double phototherapy was better and safer in reducing bilirubin levels with minimal side effects. 15][18] light intensity is the factor that determines the effectiveness of phototherapy.Higher light intensity reduced bilirubin levels faster.A study in england showed that phototherapy with maximum irradiance and wide exposure shortened the duration of phototherapy. 22n our study, the nearest light source was 10 cm from the neonates' bodies.][21] Blue light of wavelength 425-475 nm is the best light type to reduce indirect bilirubin levels. 23,24lue light is very effective because it has a shorter wavelength compared to other visible light, with the exception of purple.Wavelength is inversely proportional to energy level; the shorter the wavelength, the greater energy produced. 24n our study we used blue light.light intensity was measured at baseline, and at 12 and 24 hours of phototherapy.There was a significant difference in spectral irradiance between the two groups at all time points (P = 0.001).To increase photolight intensity for both groups, we placed the neonates supine and varied positions every 3 hours during phototherapy.However, a clinical trial in Israel comparing 14 neonates in alternating positions to 16 others in a supine position while using single phototherapy showed a significantly greater decrease in bilirubin levels in the supine group after 24 hours of phototherapy. 25eonates treated for high bilirubin levels can also suffer from dehydration and may require additional fluid intake. 26Neonatal maturity, adequate caloric intake, photolight unit temperature, distance between the neonate and photolight, and incubator rate of heat loss are all potential factors in increasing neonatal body temperature, environmental temperature, insensible water loss, as well as respiratory rate and blood flow to the skin.Increased peripheral blood flow can increase fluid loss, requiring adjustment by administration of intravenous fluids. 26,27Changes in skin, such as rash, darker skin colour and burning can be seen if infants are overexposed to fluorescent light.A study in the Netherlands found that during intensive phototherapy, a 20% increment of total fluid requirement may prevent increased body temperature. 28ody temperature and fluid administration were strictly monitored.Fluid intake was given every 2 hours, and was increased by 10 -20% of the total fluid requirement.In breastfed neonates, phototherapy was withheld during breastfeeding. 29,30In our study, we found hyperthermia (T > 37.5 o C) in 3 (5% of total) neonates in the single phototherapy group and 5 (8.3% of total) in the double phototherapy group.
A limitation of our study was not including maternal characteristic data associated with hyperbilirubinemia in neonates.
In conclusion, we found a significantly greater decrease in bilirubin levels in the double phototherapy group at 12 hours and 24 hours of phototherapy compared to the single phototherapy group.Double phototherapy is more effective than single phototherapy in reducing bilirubin levels in jaundiced newborns.
Figure 2 .
Figure 2. Comparison of spectral irradiance in single and double phototherapy at initial, 12 hours and 24 hours of phototherapy
Table 2 .
Mean decreases in serum total bilirubin levels between initial and 12 hours, and between initial and 24 hours in the double and single phototherapy groups
Table 3 .
Bilirubin levels at baseline, and after 12 and 24 hours of phototherapy | 2019-03-12T13:05:36.269Z | 2011-12-31T00:00:00.000 | {
"year": 2011,
"sha1": "ed040e8b93d9a62316cef3dcbaa5a620b2ff3a42",
"oa_license": "CCBYNCSA",
"oa_url": "https://paediatricaindonesiana.org/index.php/paediatrica-indonesiana/article/download/857/704",
"oa_status": "GOLD",
"pdf_src": "Anansi",
"pdf_hash": "ed040e8b93d9a62316cef3dcbaa5a620b2ff3a42",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
224769117 | pes2o/s2orc | v3-fos-license | Evidence of an Effect of Gaming Experience on Visuospatial Attention in Deaf but Not in Hearing Individuals
Auditory cortex in congenitally deaf early sign language users reorganizes to support cognitive processing in the visual domain. However, evidence suggests that the potential benefits of this reorganization are largely unrealized. At the same time, there is growing evidence that experience of playing computer and console games improves visual cognition, in particular visuospatial attentional processes. In the present study, we investigated in a group of deaf early signers whether those who reported recently playing computer or console games (deaf gamers) had better visuospatial attentional control than those who reported not playing such games (deaf non-gamers), and whether any such effect was related to cognitive processing in the visual domain. Using a classic test of attentional control, the Eriksen Flanker task, we found that deaf gamers performed on a par with hearing controls, while the performance of deaf non-gamers was poorer. Among hearing controls there was no effect of gaming. This suggests that deaf gamers may have better visuospatial attentional control than deaf non-gamers, probably because they are less susceptible to parafoveal distractions. Future work should examine the robustness of this potential gaming benefit and whether it is associated with neural plasticity in early deaf signers, as well as whether gaming intervention can improve visuospatial cognition in deaf people.
INTRODUCTION
Without technical intervention, congenitally profoundly deaf individuals have little opportunity to process sound. As a result, auditory cortex reorganizes to process other types of information, including visual cognition (Cardin et al., , 2018Ding et al., 2015;Twomey et al., 2017;Holmer et al., 2019; for reviews, see Alencar et al., 2019;Cardin et al., 2020), possibly offering deaf individuals the potential to outperform their hearing peers in this domain (Cardin et al., 2018). However, deaf children sometimes have difficulty achieving expected performance in academic skills, such as reading and math (Qi and Mitchell, 2012), and may not realize their potential as adults . Performance on some visuospatial tasks, in particular those tapping into visuospatial perception and attentional processes, have been shown to be altered in deaf individuals (for reviews, see Bavelier et al., 2006;Rudner et al., 2009). In hearing individuals, visuospatial perception and attention have been reported to shift as a function of gaming experience (for recent meta-analyses, see Wang et al., 2016;Bediou et al., 2018; also see, Kristjánsson, 2013;Powers et al., 2013, for critical reviews). One study has also reported improved inhibition control in deaf individuals after playing a first-person shooter game one hour per day for 16 weeks (Nagendra et al., 2017). The aim of the present, cross-sectional, study was to investigate the combined effect of deafness and naturally occurring gaming experience on visuospatial attention.
Changes driven by congenitally deafness seem to be limited specifically to attentionally demanding aspects of visuospatial processing (Bavelier et al., 2006). Visual processing is supported by dorsal and visual neural streams. The dorsal visual stream supports processing of "where" a stimulus is and how it moves while the ventral stream supports identification of "what" the stimulus is. Both "what" and "where" processing becomes attentionally demanding in the presence of task-irrelevant information. Armstrong et al. (2002) reported evidence of an influence of deafness on the function of the dorsal visual stream. Effects of deafness are manifested in altered processing of motion in the visual periphery (Bavelier et al., 2001;Armstrong et al., 2002;Bosworth and Dobkins, 2002;Fine et al., 2005) as well as some aspects of peripheral attention in deaf individuals (Bavelier et al., 2001;Proksch and Bavelier, 2002;Colmenero et al., 2004;Dye et al., 2007;Hauser et al., 2007). In addition, detection of changes outside foveal vision seems to be faster in deaf than in hearing individuals (Loke and Song, 1991;Chen et al., 2006;Dye et al., 2009), suggesting that stimuli outside the fovea are more likely to challenge attentional control in deaf populations, at least when those stimuli are of relevance to solving the task (Bavelier et al., 2006;Belanger and Rayner, 2015). For the ventral stream, Armstrong et al. (2002) showed no effect of deafness, whereas others showed altered effects of deafness in both ventral and dorsal streams (Weisberg et al., 2012;Samar and Berger, 2017). Because the dorsal stream is susceptible to effects of deafness, with increased attentional resources used for processing of stimuli in the periphery, deaf individuals might perform worse than hearing individuals on visual tasks where stimuli outside the fovea need to be suppressed.
Working memory, the active storage of representations for ongoing processing, and attentional control, the selection of stimulus to focus on in processing, limits performance on cognitive tasks (Oberauer, 2019). For the processing of stimulusrich displays and subsequentially presented stimuli, working memory is recruited and demands on attentional control are high. Although verbal working memory is similar for deaf and hearing individuals (Boutla et al., 2004;Andin et al., 2013), deaf individuals have better visuospatial working memory than hearing peers when assessed on a dynamic sequence tapping task, such as the Corsi Block-Tapping Test (Wilson et al., 1997;Geraci et al., 2008;Lauro et al., 2014). Similar results have been shown with a card-pair matching task . This behavioral advantage may well reflect enhanced dorsal stream processing. On a static visual working memory task, however, the performance of deaf individuals has been reported to be worse than for hearing individuals (Lauro et al., 2014). It is likely that this reflects compromised ventral stream processing (cf. Samar and Berger, 2017).
In the Flanker task (Eriksen and Eriksen, 1974), the participant needs to suppress static distractors presented outside the fovea while making a decision on a target stimulus presented in the center of the visual field. Thus, it is a task requiring visuospatial attentional control for selective monitoring of what is visually present Unsworth et al., 2015). This means that the Flanker task probably taps both dorsal and ventral visual stream functions and this notion is supported by empirical data (Lange-Malecki and Treue, 2012;Perry and Fallah, 2014;McDermott et al., 2017). A slowing of performance on the task is typically observed as the incongruence between response selection for a target stimulus and flanking distractors increases (Eriksen and Eriksen, 1974;Rueda et al., 2004;Sladen et al., 2005;Dye et al., 2007), indicating a conflict in determining what the target is. The standard task typically has two response keys, corresponding to two different targets, and incongruence is achieved by presenting flanking stimuli that correspond to the non-target response key. In other trials, flanking stimuli are congruent with the target stimulus, which leads to faster responses. The difference in response times between incongruent and congruent trials is an indicator of visuospatial attentional control (Rueda et al., 2004), and with an increase in attentional allocation to stimuli outside the fovea in deaf individuals (Bavelier et al., 2006) as well as altered ventral stream processing (Weisberg et al., 2012;Samar and Berger, 2017), an incongruence effect is likely to be stronger for deaf compared to hearing individuals. Thus, despite superior performance on some tasks related to the dorsal stream, deaf individuals are more distracted by flanking stimuli in a Flanker task than hearing participants Dye and Hauser, 2014), irrespective of sign language skill (Proksch and Bavelier, 2002;Dye et al., 2007; also see Bosworth and Dobkins, 2002;Dye et al., 2009). This does, however, align with the notion of changed ventral stream processing in deaf compared to hearing individuals shown in some studies, since the Flanker task poses a challenge in maintaining control of what (i.e., ventral) is presented on the screen, rather than where (i.e., dorsal) stimuli are located.
Visuospatial attentional control is a domain that has been reported to be improved by gaming experience (Wang et al., 2016;Bediou et al., 2018). In fact, a recent meta-analysis (Bediou et al., 2018), indicated robust effects of gaming experience on top-down attentional control tasks, including Flanker tasks. Greenwood and Parasuraman (2016) argue that in the initial stages of cognitive training the dorsal stream is recruited through a bottom-up process of distraction suppression, but as the need for distraction suppression is reduced with increasing skill, functional disconnection of the dorsal stream occurs. Thus, reduced load on dorsal stream function as a result of cognitive training may make attentional resources available for transfer to other tasks. Nagendra et al. (2017) reported improved performance of deaf individuals on a Stroop color-word task, as indexed by shorter response latency, after a video gaming intervention. In a Stroop color-word task, participants have to shield themselves from interference effects when the color and the word do not match (Scarpina and Tagini, 2017), in a manner analogous to the Flanker task. However, although the Stroop task is visual, interference effects are semantico-lexical rather than visuospatial.
Previous studies on hearing populations suggest that effects of gaming experience on visuospatial attention might be restricted to specific type of games. In particular, action video games (AVGs) have been suggested to be faciliative (Wang et al., 2016;Bediou et al., 2018). AVGs are described as fast paced, to rely on flexible use of visuospatial attention, and involve dealing with a multitude of objects on screen simultaneously. However, different criteria for labeling games are used in the literature, and what qualifies as an AVG and what does not, is not easily determined (see Bediou et al., 2018). Importantly, types of games other than AVGs have also been reported to improve cognition, and it has been suggested that specific changes in cognition are to be expected for specific type of games (i.e., near-transfer effects, Oei and Patterson, 2013). This notion is similar to the idea that differences in visuospatial attention between deaf and hearing individuals are specific and experience-based (Bavelier et al., 2006;Samar and Berger, 2017). Here, we wanted to investigate this association by comparing performance on a Flanker task of deaf individuals who report they play video or computer games, to those who report that they do not play such games.
In the present study, we predict the negative effect on response times of distracting stimuli in a Flanker task to be greater for deaf than hearing individuals (see e.g., Dye et al., 2007). However, as gaming experience has been shown to improve visuospatial attentional control (Bediou et al., 2018), and gamers are expected to show less interference from incongruent flankers than nongamers, we predict that gamers will outperform non-gamers on the Flanker task.
Participants
We included 16 early deaf (9 female) and 24 hearing (12 female) participants. All had normal or corrected-to-normal visual acuity and normal contrast sensitivity, as measured by Snellen chart (McGraw et al., 1995) and Pelli-Robson contrast sensitivity chart (Pelli and Robson, 1988), respectively. Due to recruitment constraints, deaf participants (M = 35.1, SD = 7.6, range 22-48) were on average almost 9 years older than the hearing participants (M = 26.5, SD = 7.5, range 19-40) and this difference was statistically significant, t(22.2) = 3.44, p = 0.002, ε = 0.64. However, there was no statistically significant difference between groups in non-verbal cognitive ability, t(12.3) = 0.91, p = 0.38, ε = 0.25, as measured on the Visual puzzles subset from WAIS-IV (Wechsler, 2008). All participants had completed at least high school (minimum of 12 years); six deaf and seven hearing participants had a university degree.
Deaf participants used Swedish Sign Language (Svenskt teckenspråk; STS) as their primary language. Nine were deaf from birth and the remaining seven were between 6 months and 3 years old when their deafness was confirmed. Five had deaf parents who signed with them from birth, and the rest started to learn sign language as soon as their deafness was discovered, and their parents started to use STS. For nine participants this was before the age of 3, and for one participant, this was in preschool years. One participant did not specify when they started using sign language.
Gaming Experience
To classify participants as a gamer or a non-gamer, participants answered a questionnaire (see Supplementary Appendix A; for similar procedures, see e.g. Rudner et al., 2015;Unsworth et al., 2015) on their gaming habits. Since the literature on gaming effects on visuospatial attention is limited to hearing populations, and we know little of whether reported effects generalize to deaf populations, assignment by self-report was applied instead of more extensive, and costly, longitudinal designs. Participants were asked how often (0 = Not at all, 1 = Less than once per week, 2 = One to three days per week, 3 = Four to six days per week, 4 = Every day, or 5 = Several times, every day) they had been playing computer and/or console games (including games on handheld consoles) during the last 6 months. We did not assess whether gaming intensity varied during this period, or if this period was a representative example of the individual's general gaming pattern. Based on self-reported gaming experience, participants were then categorized as a gamer or a non-gamer. All participants who reported having played any type of game on a computer or console or both during the last six months were defined as gamers (i.e., response categories 1-5). All participants who reported not playing computer or console games at all during the last 6 months were defined as nongamers (i.e., response category 0). Among hearing participants, 12 (2 female) were categorized as gamers and 10 (8 female) as a non-gamers (two female participants did not report gaming experience), and among deaf participants, there were 8 gamers (3 female) and 8 non-gamers (6 female). Of the deaf gamers, 4 reported playing only console games and 1 played only computer games, the rest played both, and of the hearing gamers, 5 played console games only, 3 only computer games, and the rest played both computer and console games. We did not make sub-groups based on the type of games participants played (see Supplementary Appendix B for a list of the games participants reported playing). This was partly due to the small sample size, but also because the previous literature on gaming effects almost exclusively include hearing populations.
The Flanker Task
In the Flanker task (Eriksen and Eriksen, 1974), participants had to decide whether a target stimulus, which was an arrow presented at the center of a computer screen (e.g., Dye et al., 2007;Unsworth et al., 2015), pointed left or right, and respond by pressing the corresponding button on the keyboard. Specifically, if the target stimulus was an arrow pointing left, the participant was instructed to press the left Shift key (marked with an arrow pointing to the left drawn on a piece of self-adhesive paper) and if the target stimulus was an arrow pointing right, the participant was instructed to press the right Shift key (marked with an arrow pointing to the right drawn on a piece of self-adhesive paper). In each trial, the target stimulus was flanked by two arrows on each side. Congruent trials had flankers pointing in the same direction as the target (e.g.,←←←←←) and incongruent trials, in the opposite direction (e.g.,←←→←←). The participant was instructed to ignore the flanker arrows and respond to the direction of the target arrow. A trial began with a fixation point presented in the middle of the screen for 550 ms, which was immediately followed by a horizontal array, 8 cm wide, of five equally sized and equally spaced black arrows. The array remained on the screen for 2100 ms, after which the screen went blank for 800 ms before the start of the next trial. For an overview of the structure of the task, see Figure 1. The task was administered on a 12 laptop computer using presentation software DMDX version 5.1.4.2 (Forster and Forster, 2003) and the distance between the participant's face and the screen was approximately 60 cm. Participants responded to 48 trials in total, with an equal number of congruent and incongruent trials. In half of the trials within each condition, the target pointed to the left, and in the other half to the right. The order of presentation was randomized for each participant. The dependent variable was average response time in ms on trials to which a correct response was given (both for congruent and incongruent trials).
Swedish Sign Language Sentence Repetition Test
To rule out inadequate sign language skills as an explanation for the results in the present study, deaf participants' STS skill was assessed on the Swedish Sign Language Sentence Repetition Test (STS-SRT, Schönström, 2014a,b). The STS-SRT is an adaptation of an American Sign Language sentence repetition test (ASL-SRT, Hauser et al., 2008) used to measure global sign language fluency of deaf adults. The STS-SRT is a reliable and valid test of STS skills in adults who have used STS since childhood (Schönström, 2014b). The test consisted of 31 trials with filmed STS sentences produced by a deaf native signing man. The sentences varied in length and in difficulty. The participant was instructed to watch the sentences and to reproduce them exactly as signed in the video clips, including the vocabulary and grammatical markers used. Before testing started, participants practiced on three sentences to make sure that they had understood the procedure. On each trial in the actual test, the participants saw a video clip presented on a laptop (12 screen), and were given approximately 8 seconds to repeat the sentence before the next trial started. The front camera on the laptop was used to film responses. Responses were scored based on a guideline with instructions for each trial on a later occasion (Schönström, 2014b). For a response to be scored as correct, the participants had to reproduce the sentence exactly as it was performed. The dependent variable was number of correctly reproduced sentences (maximum = 31). Testing time was approximately 10 minutes.
Procedure
Participants were tested individually in a quiet room. Participants provided written informed consent before behavioral testing commenced. This study is part of a larger project and testing started with screening of visual acuity and visual contrast, before a cognitive test battery, including tests of episodic long-term memory, lip-reading ability, and phonological skill, in addition to the test of non-verbal cognitive ability (Visual puzzles, Wechsler, 2008), STS skill (STS-SRT, Schönström, 2014a) and the Flanker task reported here, was administered. Before the test battery was administered, participants performed one motor speed task and a physical matching task (Holmer et al., 2016) to become familiar with the set-up of the computerized testing. Testing took approximately 60 minutes in total. For deaf participants, an accredited STS interpreter was present during testing and provided verbatim translation of instructions. In a second part of the larger project, participants performed an fMRI experiment not reported here.
Statistical Analysis
First, descriptive statistics and frequencies for control and background variables were calculated, and the distribution of response times from the Flanker task were visually inspected. Due to the small sample size with associated potential threats of nonnormality and low power, robust statistical methods were applied (Erceg-Hurn and Mirosevich, 2008;Wilcox, 2017). Statistical analysis was performed in RStudio version 1.2.5042 (RStudio Team, 2020), running R version 4.0.0 (R Core Team, 2020). Group comparisons on control and background variables: age, non-verbal cognitive ability, STS skill (only deaf participants), and gaming habits for gamers, were performed using yuen t-tests with the yuen function from package WRS2 (Mair and Wilcox, 2020). As an estimate of effect size the explanatory measure of effect size ε is reported, with values of 0.1, 0.3, and 0.5 corresponding to small, medium, and large effects (Mair and Wilcox, 2020). After that, Wilcox (2017) bbwtrim function was used to perform a robust mixed ANOVA with one within-group factor: Congruency (congruent, incongruent), and two betweengroup factors: Group (deaf, hearing) and Gaming (gamer, nongamer), on response time (in ms) from the Flanker task. Effect size estimates ε for main effects of the ANOVA were calculated with the yuen function for between group effects and the yuend function for the within group effect, both from package WRS2 (Mair and Wilcox, 2020). Main effects were followed up by comparing means between levels of the factor, and simple main effects were followed up by comparing percentile bootstrapped confidence intervals, estimated using the onesampb function from WRS2 (Mair and Wilcox, 2020). To investigate associations between age and non-verbal cognitive ability and performance on the Flanker task, robust correlations were calculated with the pbcor function from WRS2 (Mair and Wilcox, 2020). The default value of a trim proportion of 0.2 was applied in all robust analyses. Due to a technical issue, the result was missing for one deaf non-gamer on the Flanker task. One hearing gamer and one hearing non-gamer performed on chance level on the Flanker task, indicating that they did not follow instructions. The mean performance of the sub-group that the participant belonged to was used for these three participants in analyses to maximize statistical power.
Characteristics of Deaf and Hearing Gamers and Non-gamers
Descriptive statistics on background variables for deaf and hearing gamers and non-gamers are reported in Table 1. Deaf participants demonstrated proficiency in STS skills, as assessed on the STS-SRT (mean performance was on par with mean performance from a previously tested group, M = 17.7 och SD = 4.9, Schönström, 2014a,b). No statistically significant differences on any background variables were seen between deaf gamers and non-gamers: age, t(6.6) = 0.00, p = 1.00, ε = 0.00, nonverbal cognitive ability, t(8.7) = 2.08, p = 0.07, ε = 0.61, and STS skill, t(10) = 0.67, p = 0.52, ε = 0.31. Similarly, hearing gamers and non-gamers did not differ on background variables: age, t(8.7) = 0.00, p = 1.00, ε = 0.07, and visual puzzles, t(11.6) = 0.26, p = 0.80, ε = 0.13. Thus, there were no underlying differences on background variables between gamers and non-gamers in either of the two groups.
Flanker Task
As expected, deaf gamers (M = 98%, SD = 5.8) and non-gamers (M = 98%, SD = 4.2), as well as hearing gamers (M = 99%, SD = 1.4, after exclusion of the participant who performed at chance level) and non-gamers (M = 99%, SD = 2.3, after exclusion of the participant who performed at chance level) performed close to ceiling on accuracy on the Flanker task. Thus, response times for almost all trials were included in the analysis (see Table 2 for descriptive statistics). The mixed robust ANOVA for response times in Flanker showed a main effect of congruency, Q = 74.1, p < 0.001, ε = 0.32, gaming, Q = 5.40, p = 0.02, ε = 0.41, and of Group, Q = 5.09, p = 0.02, ε = 0.41. Response time was faster for congruent (M = 539 ms, SD = 110) than incongruent (M = 597, SD = 114) trials, and gamers (M = 541 ms, SD = 112) responded faster than non-gamers (M = 598 ms, SD = 102), and hearing (M = 557 ms, SD = 108) responded faster than deaf (M = 594 ms, SD = 108). There was a statistically significant interaction between group and gaming, Q = 8.89, p = 0.003 (see Figure 2). Investigation of the confidence intervals for the group by gamer interaction, indicated that deaf gamers, 95% CI [475 ms, 562 ms], responded faster than deaf non-gamers, 95% CI [626 ms, 739 ms], and on par with hearing gamers, 95% , responded faster than deaf non-gamers, but no difference was observed in comparison to hearing gamers. Thus, the main effect of gaming experience was explained by a group-specific effect for deaf participants that eliminated any difference in processing efficiency across groups. Besides the interaction between Group and Gaming, interactions were not statistically significant (all ps > 0.05). Thus, our predictions that deaf individuals are more distracted and that gamers are less distracted by incongruent flanking stimuli were not supported. Non-verbal cognitive ability, r pb = −0.21, p = 0.19, and age, r pb = 0.23, p = 0.16, were not associated with response time on the Flanker task, and it is thus unlikely that these variables strongly influenced the pattern of results.
DISCUSSION
In the present study, we investigated the effect of naturally occurring gaming experience on visuospatial attentional control in early deaf signers. We predicted longer response times on the Flanker task for deaf compared to hearing participants and that this difference would be most apparent for incongruent trials. We also predicted that gamers would show less interference from flankers than non-gamers and outperform them on the Flanker task, especially for incongruent trials.
Our predictions were partially supported by the results. While both deaf and hearing groups had longer response times on FIGURE 2 | Response time (in ms, y-axis) for gamers and non-gamers (x-axis) for deaf and hearing participants. Error bars represents 95% confidence intervals.
incongruent than congruent trials, the deaf group did not show longer response times than the hearing group specifically on incongruent trials. Instead, the deaf group responded slower on both congruent and incongruent trials. Across groups, deaf non-gamers responded slower than hearing non-gamers, while there was no significant difference in performance between deaf gamers and hearing participants. Further, an effect of gaming was only observed in the deaf group, and we did not find evidence of a specific effect of gaming on incongruent trials.
Although there was a statistically significant main effect of group on performance on the Flanker task, this effect was explained by longer latencies for deaf non-gamers compared to the other participants. Deaf gamers performed similar to hearing participants. With enhanced visuospatial perception in deaf compared to hearing participants under some circumstances (Loke and Song, 1991;Chen et al., 2006;Dye et al., 2009), worse performance on tasks demanding control of visuospatial attention might seem contradictory. However, these seemingly contradictory findings might be explained by differences in ventral versus dorsal stream processing, and their relative contribution to the behavioral task (Samar and Berger, 2017). Proksch and Bavelier (2002) proposed that congenital deafness alters visuospatial attention in such a way that more attentional resources are used for processing stimuli outside central vision (also, see Bavelier et al., 2006). In a visuospatial perception task designed to invoke dorsal stream functions, this will lead to better ability to, e.g., detect stimuli in the periphery (e.g., Dye et al., 2009), but in a task that relies more on ventral stream processing, and suppression of dorsal stream elements, performance might be impaired (e.g., Dye et al., 2007). Like Lauro et al. (2014), here we used a static task that could be argued to rely on ventral stream processing, and in line with what Lauro et al. (2014) reported, we saw worse performance in deaf compared to hearing individuals. Thus, our results lend further behavioral support to the notion of potentially altered ventral stream processing in deaf populations (Weisberg et al., 2012;Samar and Berger, 2017). On the other hand, we did not find evidence that deaf participants are more distracted by incongruent flanking stimuli than hearing participants. In line with previous data (e.g., Dye et al., 2007), we reasoned that the effect of incongruency would become stronger as a consequence of the redistribution of visuospatial attention. It is likely that the small sample in combination with the complexity of the design might have been at play here. To maximize power and minimize bias due to potential non-normality in the data, robust methods were used in analysis. Although this was likely to be the best analytic approach for the purposes of the present study, the results are still constrained by the available data. In addition to the limited amount of individuals, the Flanker task only included 24 congruent and 24 incongruent trials. This number of trials is similar to what others have used (i.e., 30 for each type in Unsworth et al., 2015), but more trials are likely to produce more stable estimates when averaging within individual, with reduced noise in the analysis as a result (Brysbaert, 2019). These factors: small sample, complex design, and small number of trials, are likely to have reduced the probability of detecting a group by congruency interaction. Thus, we cannot rule out the possibility that deaf individuals are more distracted by incongruent flanking stimuli in a Flanker task than hearing individuals.
Based on the present study we suggest that deaf individuals with recent gaming experience reveal a level of visuospatial attentional control similar to that revealed by hearing individuals in a task that presumably draws upon ventral stream processing. To our knowledge, only one previous study has investigated effects of gaming on cognition in a deaf population (on a Stroop color-word task, Nagendra et al., 2017), and that study also reported a positive effect. Our findings extend the results of Nagendra et al.'s (2017) study, by showing an effect of gaming in another executive domain. Importantly, the effect of gaming in the present study was not simply driven by sign language proficiency, since sign language skills did not differ between deaf gamers and non-gamers. Greenwood and Parasuraman (2016) argue that cognitive training leads to functional disconnection of the dorsal stream, releasing attentional resources for transfer to other tasks. Because a specific effect of gaming is found only for deaf individuals with potentially enhanced dorsal stream skills, one interpretation is that this group has more resources to transfer as a result of the cognitive training inherent in gaming. A potential group-specific effect of gaming experience in deaf individuals needs to be followed up in future work. In particular, combining behavioral and brain imaging measures will help us illuminate potential alterations in dorsal and/or ventral stream processing. Related to this, an effect should also be compared between congenitally deaf individuals and individuals with acquired deafness.
Previous studies in hearing individuals have reported effects of gaming on the kind of attentional control demanded by a Flanker task (Bediou et al., 2018). However, here we did not see any effect of gaming in the hearing group, and there was no significant interaction between gaming and congruency. Although it might be the case, as some argue, that gaming experience does not lead to any meaningful effects on cognitive functions in hearing individuals (Kristjánsson, 2013;Powers et al., 2013), the present study had some limitations that might explain why our results were not in line with our prediction. As already mentioned, statistical power was restricted due to the small sample size, another issue might be that our definition of a gamer was not as strict as definitions applied in previous studies in the literature (e.g., Bediou et al., 2018). Further, selfreported gaming habits during the last six months determined group assignment. In hearing individuals, there is evidence to suggest that gaming effects vary as a function of gaming genre (however, see a discussion on issues in defining genres in Bediou et al., 2018). In particular, action video games (AVGs) seem to have the most robust effects (Wang et al., 2016;Bediou et al., 2018). Gamers in the present study played a wide variety of games (see Supplementary Appendix B), ranging from simple puzzle games (not typically categorized as AVGs, e.g., Tetris) to firstperson shooters (commonly categorized as AVGs, e.g., Counterstrike), and there was also variability in what type of platform they preferred for playing games (i.e., some played games on stationary consoles, others on a computer, and yet others on both these types of platforms). Self-report measures are convenient, but they do not always reflect actual behavior, and this is true also in the case of gaming experience (Kahn et al., 2014). Besides the potentially low correspondence to actual behavior, the temporal resolution of the self-report measure included here was coarse. It is possible that effects of video games on visuospatial attention are transient (similar to effects of gaming on attitudes, e.g., Sestir and Barthalow, 2010), which might have then influenced our results.
As two examples, we do not know whether participants in one group had more recent gaming experience than the participants in the other group, or if participants had played for only a limited period during the time for which they reported their habits. Our approach was, however, intentional and motivated by a number of factors. Most importantly, we did not find any previous study on the effect of gaming experience on visuospatial attention in deaf individuals, but plenty of evidence to suggest that visuospatial processing differs between deaf and hearing individuals (Bavelier et al., 2006). Thus, we had little reason to assume that findings from hearing populations would be exactly the same for deaf individuals. However, since we did find an effect in deaf individuals, and saw that groups reported similar gaming habits, this could mean that effects of gaming experience on visuospatial attentional control are observed with a lower dose of exposure in this population. One explanation for this could be that the mechanisms are somewhat different across groups, and more malleable to visuospatial experience for deaf individuals. It is reasonable to assume that effects arising from gaming experience are constrained by baseline levels across tasks, and with different baselines in visuospatial attention across deaf and hearing populations, the pattern across groups is influenced by task selection. Oei and Patterson (2013) suggest that game characteristics constrain transfer, and here we propose that the characteristics of the gamer will produce similar constraints. It is thus important to further investigate the role of different types of gaming experiences in visuospatial perception, and visuospatial attention in particular, in deaf individuals. Experimental designs are a way forward, with active manipulation of gaming experience, although that might become more and more challenging with gaming turning into a mainstream leisure activity in society. As an alternative, using fine-grained correlational approach, for example, by following participants over a longer period of time and using active measures of gaming experience, such as ecological momentary assessment (Kirchner and Shiffman, 2008), might be useful in future studies. Also, the longevity of gaming effects on cognition is something that needs to be addressed in such work.
CONCLUSION
Visuospatial attention is altered by early deafness. The results of the present study show better visuospatial attentional control in deaf signers who play video games than those who do not. Gaming experience may help harness the changes in visuospatial attention displayed by deaf individuals for better attentional control. Thus, gaming might be a useful intervention for shielding deaf children from potential visuospatial distractions.
DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/Supplementary Material.
ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the Regional Ethical Review Board, Linköping, Sweden (dnr 2016/344-31). The participants provided their written informed consent to participate in this study.
AUTHOR CONTRIBUTIONS
EH, MR, and JA designed the study. JA collected data. KS scored performance on the STS-SRT. EH performed the data analysis. All authors were involved in the interpretation of results as well as preparing and finalizing the manuscript, after a draft version was prepared by EH.
FUNDING
This work was funded by a grant to MR from Vetenskapsrådet (dnr 2015-00929). | 2020-10-20T13:11:24.757Z | 2020-10-20T00:00:00.000 | {
"year": 2020,
"sha1": "fcc2276f8b6de2fa872dc04bb0fd4cabbf9836fa",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.3389/fpsyg.2020.534741",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "fcc2276f8b6de2fa872dc04bb0fd4cabbf9836fa",
"s2fieldsofstudy": [
"Psychology"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
14066890 | pes2o/s2orc | v3-fos-license | Infliximab reduces Zaprinast-induced retinal degeneration in cultures of porcine retina
Background cGMP-degrading phosphodiesterase 6 (PDE6) mutations cause around 4 to 5% of retinitis pigmentosa (RP), a rare form of retinal dystrophy. Growing evidence suggests that inflammation is involved in the progression of RP. The aims of this study were to corroborate the presence of high TNFα concentration in the eyes of RP patients and to evaluate whether the blockade of TNFα with Infliximab, a monoclonal anti-TNFα antibody, prevented retinal degeneration induced by PDE6 inhibition in cultures of porcine retina. Methods Aqueous humor from 30 patients with RP and 13 healthy controls were used to quantify the inflammatory mediators IL-6, TNFα, IL-1β, IL-10 by a multiplex enzyme-linked immunosorbent assay (ELISA) system. Retinal explants from pig were exposed to Zaprinast, a PDE6 inhibitor, for 24 hours in the absence or the presence of Infliximab. Cell death was evaluated by TUNEL assay. The number and distribution of caspase-3 positive cells, indirect poly(ADP)ribose polymerase (PARP) activation and glial fibrillary acidic protein (GFAP) content were visualized by immunolabeling. Antioxidant total capacity, nitrites and thiobarbituric acid reactive substances (TBARS) formation were determined to evaluate antioxidant-oxidant status. Results IL-6 and TNFα concentrations were higher in the aqueous humor of RP patients than in controls. Infliximab prevented retinal degeneration, as judging by the reduced presence of TUNEL-positive cells, the reduction of caspase-3 activation and also reduction of glial activation, in an ex vivo model of porcine retina. Additionally, Infliximab partially reduced oxidative stress in retinal explants exposed to Zaprinast. Conclusions Inflammatory mediators IL-6 and TNFα were elevated in the aqueous humor of RP patients corroborating previous studies suggesting sustained chronic inflammation. Our study suggests that TNFα is playing an important role in cell death in an ex vivo model of retinal degeneration by activating different cell pathways at different cell layers of the retina that should be further studied. Electronic supplementary material The online version of this article (doi:10.1186/s12974-014-0172-9) contains supplementary material, which is available to authorized users.
Background
Retinitis pigmentosa (RP) is a common form of rod-cone dystrophy, constituting the largest Mendelian genetic cause of blindness in the developed world. Patients with RP typically loose night vision in adolescence, peripheral vision in young adulthood, and central vision later in life due to progressive loss of rod and cone photoreceptor cells. Photoreceptor cell death starts with rod photoreceptor degeneration and eventually cone cell death that is the major problem affecting RP patients, because it leads to loss of central vision [1]. More than 60 genes, including phosphodiesterase 6 (PDE6) subunit genes, have been identified to date that, when mutated, cause different forms of non-syndromic RP [2][3][4][5][6][7].
Although RP is a genetic disease, increasing evidence in patients and animal models suggests that oxidative stress and inflammation, especially TNFα, contribute to its pathogenesis, independently of the genes mutated [8][9][10]. Some reports show the presence of sustained chronic inflammatory reaction including elevated TNFα levels in the eyes of RP patients [11] and rd10 mice [12]. TNFα is a pleiotropic cytokine essential for the induction and maintenance of the inflammatory immune responses [13] that is also upregulated in inflammatory ocular diseases, including Adamantiades-Behcet disease [14], retinal vascular tumors [15], neovascular age-related macular degeneration [16], uveitis [17], glaucoma [18] and ischemic retinopathy [19].
TNFα mediates a broad range of cellular activities, including proliferation, survival, differentiation, inflammation and cell death. In the retina, TNFα is likely to be secreted from activated macrophages, astrocytes, microglial cells and retinal Müller glial cells. TNFα can trigger several well-characterized death-promoting (caspase-dependent and caspase-independent cell death) and survival-promoting pathways, depending upon the predominating signaling pathway in the particular cell type [20]. TNFα binding to cell surface receptors such as TNFR1 mediates activation of initiator caspases (caspase-8, caspase-10) and finally triggers cleavage of effector caspases (extrinsic pathway of cell death) [21]. TNFα is also involved in the intrinsic pathway of cell death that is initiated by cellular and DNA damage which particularly involves mitochondria. Finally, TNFα can also activate a subset of programmed necrosis called necroptosis. The mechanism that leads cells to undergo apoptosis or necroptosis and the mechanism that mediates the execution of necroptosis still remains unclear. The poly(ADP-ribose) polymerase (PARP) pathway can also activate this mode of programmed necrosis. PARP-1 activation in response to excessive DNA damage results in the massive synthesis of poly(ADP-ribose) polymers (PAR), NAD + depletion and subsequent release of apoptosis inducing factor (AIF) from mitochondria, which translocates to the nucleus where it forms an active DNA-degrading complex (caspase-independent pathway). The PARP pathway has been considered as an integral part of TNF-induced necroptosis; however, it has been recently described that both pathways represent distinct and independent routes to programmed necrosis [22].
The mechanisms responsible for photoreceptor cell death in RP are still unclear. However, increasing evidence suggests that inflammation [11,12,23,24] and especially TNFα could contribute to the pathogenesis of RP. Therefore, inhibition of TNFα and downstream cellular signaling mechanisms, following interaction of TNFα with its receptors, could be a possible target in the treatment of retinal neurodegenerative disorders such as RP.
In the current study we found that IL-6 and TNFα were increased in the aqueous humor of RP patients. We also observed that pharmacological inhibition of TNFα with Infliximab, a specific monoclonal antibody against TNFα, prevented retinal degeneration in cultures of porcine retina exposed to Zaprinast. This model reproduces some events of the degeneration found in murine models of RP caused by non-functional PDE6 [25]. We also observed in our model a reduction of caspase-3 activation, GFAP reactivity and partially oxidative stress, caused by Infliximab treatment. These results suggest that inflammation, especially TNFα upregulation, is playing an important role in retinal degeneration and, importantly, that strategies that promote its blockade could be promising therapies.
Participants in the study
Human samples were obtained, informed consent from all subjects previously having been given. The procedure was in accordance with the tenets of the Declaration of Helsinki and was approved by the IRB of La Fe University Hospital (Valencia, Spain). Thirty adult patients with typical forms of RP characterized by an elevated final dark-adaptation threshold, retinal arteriolar narrowing, and a reduced and delayed electroretinogram were enrolled in the study. Thirteen Caucasian patients suffering from cataracts without any other ocular disease served as controls. Further details of the patients enrolled in the study are shown in Table 1.
Patients diagnosed of RP were recruited from Retina Comunidad Valenciana -Asociación Afectados por Retinosis Pigmentaria and also from the department of Ophthalmology of La Fe University Hospital (Valencia, Spain). Healthy controls were recruited from La Fe University Hospital (Valencia, Spain).
Ophthalmic examination
The best-corrected visual acuity (BCVA) and automated visual field (VF) were measured in RP patients as previously described [8]. Individual data for each patient is shown in Additional file 1: Table S1. Macular edema secondary to RP was only present in one patient.
Aqueous humor extraction
Aqueous humor samples from 30 RP patients and from 13 patients with cataracts without any other ocular disease (controls) were collected as previously described [8]. Undiluted aqueous humor samples were collected from each patient, placed in sterile tubes, and stored immediately at −80°C until use. All specimens were assayed to evaluate cytokine concentration in a double-blind arrangement with respect to their group. For each patient, aqueous humors were collected from the eye with the more severe retinopathy.
Cytokine levels in aqueous humor
The concentrations of cytokines in aqueous humor were measured using a multiplex enzyme-linked immunosorbent assay (ELISA) system. To measure the concentrations of IL-1β, IL-6, IL-10 and TNFα, the SearchLight Custom Human Cytokine-Inflammation Q-Plex Array (Aushon Biosystems, MA, USA) was used. Array was used according to the manufacturer's instructions. The signal of the cytokine array was determined by a cooled CCD camera (Fujifilm, Tokyo, Japan) using chemiluminescence. SearchLight CCD Imaging and Analysis System were used to quantify cytokine concentrations. The cytokine levels were expressed as pg/mL.
Porcine retinal explant cultures
Seventy eyes (both left and right eyes from each animal) from small miniature pigs aged 3 to 7 months were obtained from the local slaughterhouse. Neuroretinal explants were carried out as recently described [25]. Treatments were added the day of the culture and maintained for 24 hours. To inhibit PDE6 and induce retinal degeneration, we used a final concentration of 100 nmol/L Zaprinast [25,26]. Zaprinast (Sigma-Aldrich, Madrid, Spain) was diluted in dimethyl sulfoxide (DMSO) (AppliChem, Darmstadt, Germany). The equivalent amount of DMSO was added to the culture medium of controls. To evaluate the possible neuroprotective effect of TNFα blockade we used Infliximab (2 μg/mL, alone or combined with Zaprinast) as TNFα blocker (Remicade®, Schering-Plough, Madrid, Spain). Infliximab is a chimeric human immunoglobulin G1 with a mouse variable fragment having high TNFα affinity and neutralizing capacity.
TUNEL assay
To evaluate apoptosis the terminal deoxynucleotidil transferase dUTP nick and labeling (TUNEL) assay was used as previously described [25]. The apoptotic (TUNEL-positive) nuclei per field were counted in at least three fields per retinal explant using NIS-Elements imaging software (NIKON Data are expressed as mean ± SEM. For co-localization of cleaved caspase-3 (combined with Alexa Fluor 647) and PAR (combined with Alexa Fluor 488 (Invitrogen, Life Technologies, Madrid, Spain)) staining was followed by TUNEL staining.
caspase-3 activity assay caspase-3 activity was measured with a colorimetric tetrapeptide (DEVD-pNA) cleavage assay kit following the manufacturer's instructions (Bio-Vision, Mountain View, CA, USA). Total retinal protein was extracted from retinal explants and measured by the bicinchoninic acid (BCA) protein assay. caspase-3 activity was expressed as arbitrary units (au)/mg of protein.
Nitrites and nitrates (NOX) determination
Intracellular nitrites (stable end-product of nitric oxide (NO)) and nitrates (NOX) were measured in retinal explants by spectrophotometric GRIESS reaction using nitrate reductase [28]. The tissue NOX levels were expressed as nmol/mg protein.
Oxidative stress evaluation
Retinal explants were assayed for total antioxidant capacity (TAC) and thiobarbituric acid reactive substances (TBARS) formation as indicator of malonyldialdehyde (MDA) formation.
Retinal explants were homogenized in 5 mM phosphate buffer pH 7, 0.9% NaCl, 0.1% glucose, centrifuged at 10,000 × g for 15 minutes at 4°C, and then the supernatants were used to determine TAC and TBARS. Protein concentrations were measured by the BCA protein assay.
TAC was measured using a commercial kit (Cayman Chemical, Ann Arbor, MI, USA) [29]. The tissue TAC levels were expressed as nmol/mg protein.
MDA levels were detected by a colorimetric method involving thiobarbituric acid (TBA) adduct formation (Cayman Chemical, Ann Arbor, MI, USA). Tissue TBARS levels were expressed as nmol/mg protein.
Values for caspase-3 activity, NOX and oxidative markers are given as the mean ± SEM of at least eight different cultures. For each experiment samples were measured in duplicate.
Statistical analyses
All statistical analyses were done using R software (version 2.15.3) (Foundation for Statistical Computing, Vienna, Austria). Multivariate analysis of covariance (MANCOVA) and multiple linear regression models were used to analyze human data. For parametric data, ANOVA followed by Newman-Keuls post hoc test was used. For non-parametric data, Kruskal-Wallis test followed by Dunn's Multiple Comparison test was used. Significance levels were set at α =0.05.
Increased levels of TNFα and IL-6 in aqueous humor of RP patients
We performed a multiplex ELISA to determine the concentration of TNFα, IL-6, IL-1β and IL-10 in aqueous humor of RP patients. IL-1β and IL-10 were below detectable levels. Descriptive statistics of the results of the measurements of IL-6 and TNFα are shown in Table 2. We performed a MANCOVA with the results of TNFα and IL-6 as dependent variables while disease, age and gender were taken as predictive variables.
This analysis revealed that RP significantly increased inflammatory mediators IL-6 and TNFα in aqueous humor (P = 0.03) (See Additional file 2: Table S2). We found no statistical evidence for gender or age effects. Further analysis of each of the response variables indicated that IL-6 is increased in RP patients (P = 0.018). TNFα showed a tendency to increase in RP patients (P = 0.09). We assessed the possible association between inflammatory status (measured as TNFα and IL-6 levels) and stage of the disease (measured as VF and BCVA values) using a MANOVA with VF, BCVA, sex and age as predictors and TNFα and IL-6 levels as response variables. Our results showed no evidence of association between VF and BCVA and inflammatory status (P = 0.09 for VF and P = 0.94 for visual acuity). Additionally, we also analyzed separately the associations among these predictor variables and each of the two cytokine using linear models. In these analyses we found a statistically significant association between higher VF values and higher levels of TNFα (P = 0.03) ( Figure 1).
Infliximab prevents Zaprinast-induced cell death in cultured porcine retina
We previously described that PDE6 inhibition by Zaprinast triggered retinal degeneration and induced oxidative stress and inflammatory mediators such as TNFα and IL-6 in cultured porcine retina after 24 hours. In particular, TNFα and IL-6 content increased to twice control content [25]. We tested whether incubation with 2 μg/mL Infliximab, a TNFα blocker, for 24 hours prevented Zaprinast-induced retinal degeneration. Firstly, we checked the effect of Infliximab on the TNFα signaling cascade. As we could not measure TNFα, because Infliximab interferes with the ELISA assay as previously described [30], we evaluated its receptor TNF-R1 whose activation is involved in multiple apoptotic pathways. TNF-R1 relative expression increased up to 1.36 ± 0.08 arbitrary units (ANOVA Newman-Keuls post-test, P <0.0001) in Zaprinast-treated explants compared to control explants (1.00 ± 0.08 arbitrary units). However, Infliximab normalized Zaprinast-induced overexpression of TNF-R1 (0.90 ± 0.08 arbitrary units, ANOVA Newman-Keuls post-test, P <0.0001). No significant changes were found in explants treated only with Infliximab (0.91 ± 0.06 arbitrary units).
As mentioned above, TNFα can trigger programmed cell death by activating the extrinsic and intrinsic apoptotic pathways that converges on the execution pathway, which is initiated by the cleavage of caspase-3 [31]. The activity of caspase-3 in Zaprinast-treated explants was 2.3 ± 0.2 au/mg protein (ANOVA Newman-Keuls post-test, P <0.01) and 1.3 ± 0.2 au/mg protein in control explants. Infliximab almost normalized caspase-3 activity (1.7 ± 0.2 au/mg protein) compared to Zaprinast-treated explants (ANOVA Newman-Keuls post-test, P <0.05) and the percentage of cleaved caspase-3 positive cells (1.1 ± 0.3%) compared Figure 2B). We have previously observed an over activation of poly(ADP-ribose) polymerase (PARP) in our model of porcine retinal degeneration [25]. Moreover, other authors have described similar results in other animal models of retinal degeneration [32,33]. Therefore, we investigated whether TNFα mediated cell death via the PARP pathway. Accumulation of poly(ADP-ribose) polymers (PAR) was used to analyze indirectly PARP activity indirectly. Immunostaining of PAR revealed a significant accumulation of these polymers in ONL and outer segments (OS) in Zaprinast-treated explants (Kruskal-Wallis, Dunn's post-test, P <0.05) that were not prevented by Infliximab treatment (Table 3 and Figure 2C). Infliximab treatment increased PAR accumulation at all cell layers of Zaprinast-treated explants. Thus, the inhibition of TNFα by Infliximab is not causally linked to PARP activation, and therefore does not prevent the secondary PAR accumulation.
To determine whether cleaved caspase-3 or PAR accumulation co-localize with TUNEL-positive cells, we performed triple labeling (Figure 3). In Zaprinast-treated explants PAR immunostaining co-localized with TUNELpositive cells in some cells of ONL, in a few cells of the INL and in several cells of GCL. This co-localization disappeared after Infliximab treatment in ONL and GCL but remained in a subset of cells of the INL. PAR accumulation remained high, and even increased, at all cell layers, although the number of TUNEL-positive cells decreased.
However, caspase-3 positive cells did not co-localize with TUNEL-positive cells except for a subset of cells in INL in Zaprinast-treated explants. Co-localization of caspase-3 with TUNEL-positive cells disappeared after Infliximab treatment but increased co-localization of caspase-3 with PAR in INL.
Infliximab ameliorates Zaprinast-induced glial activation in cultured porcine retina
Gliosis commonly involves upregulation of the intermediate filament protein, GFAP, in Müller glial cells. We studied whether Zaprinast-induced retinal degeneration was accompanied by altered glial reactivity, and if it was the case, whether the blockade of TNFα could prevent it.
In control explants, GFAP were located in the inner half of the retinal Müller cells and their endfeet (GCL layer). However, Zaprinast-treated explants exhibited strong GFAP-positive staining of Müller cells. After PDE6 inhibition, GFAP was massively upregulated throughout the retinal explant. After Infliximab treatment the GFAP-positive labeling was significantly decreased (Figure 4).
Infliximab partially prevents Zaprinast-induced oxidative stress in cultured porcine retina cGMP accumulation induces oxidative stress in murine models of retinal degeneration [34] as it does in our model of porcine retina treated with Zaprinast [25]. To explore whether Infliximab also prevented Zaprinast-induced oxidative damage in cultured porcine retina, we measured intracellular nitrite formation (iNOX), as stable NO metabolite, TBARS content as indicator of MDA and total antioxidant capacity (TAC).
As shown in Figure 5, Infliximab normalized TAC but did not prevent oxidative stress in Zaprinast-treated explants. Total antioxidant capacity returned to control level (230 ± 15 μmol/mg protein, ANOVA Newman-Keuls post-test, P <0.05) ( Figure 5A), but TBARS formation ( Figure 5B) and intracellular NOX ( Figure 5C) remained high after the blockade of TNFα.
Discussion
Abnormal pathological pathways such as oxidative stress and inflammation, including upregulation of TNFα, have been described in retinal neurodegenerative diseases both affecting the outer retina, such as RP and age-related macular degeneration (AMD), and the inner retina, such as glaucoma and ischemic retinopathy [35][36][37][38][39]. Low-grade inflammation is present in AMD and glaucoma. For instance, in AMD, many mediators of chronic low-grade inflammation such as C-reactive protein, immunoglobulins, and acute phase molecules, the complement-related proteins, autoantibodies, macrophage infiltration and microglial activation have been found [40]. In glaucoma, microglial activation and an inflammatory response involving Toll-like receptors (TLRs), complement molecules and cytokines, such as TNFα and IL-1β, is associated with secondary phase of the disease [41]. Much less is known about the inflammatory response to retinal ischemic-reperfusion (IR) injury. However, pro-inflammatory gene upregulation, accumulation of leukocytes, and microglial activation is found following IR in rodent retinas [42].
In RP, retinal degeneration is caused by various mutations that result in rod death followed by gradual death of cones [43]. Growing evidence suggests that, regardless of the causative mutation, neuroinflammation contributes to photoreceptor degeneration [44,45]. For instance, different animal models of RP (rds mice, rd1 mice, P23 rats, RCS rats) carrying mutations in different genes (Prph2, PDE6, Rho, Mertk) show signals of an inflammatory process [23,[46][47][48][49]. In early stages of retinal degeneration the photoreceptor cells and surrounding cells, such as microglia, respond to unfavourable conditions with the production of cytokines, chemokines, growth factors, and so on, in an attempt to protect neurons and to preserve retinal function. As disease progresses, sustained inflammatory mediators and others such as oxidative stress may exacerbate photoreceptor cell death and RP progression.
Early studies suggested the presence of immune reactivity in RP patients, including the presence of retinal autoantibodies in blood and lymphocytes in vitreous humor. However, these results were variable, maybe due to the inherent genetic heterogeneity of this disease [36]. Afterwards, microglial activation, a common hallmark of both inherited and induced retinal degeneration, was described in RP patients and murine models of RP [12,45,[50][51][52][53]. It has been shown that microglial activation leads to proliferation, followed by migration to damaged sites and release of cytokines (TNFα, IL-1α, IL-1β) chemokines, neurotrophins, glutamate, NO, superoxide anions and prostaglandins to repair tissue damage. Although these events are triggered to prevent cell damage, sustained high levels of these molecules, especially cytokines, can cause progressive neurodegeneration. In models of RP, microglial activation coincides, or precedes, the peak of photoreceptor cell death and with high levels of TNFα [12,44,45,50,54,55] that seems to be toxic for photoreceptor cells in vitro [23]. Besides, microglial inhibition reduces photoreceptor cell death, TNFα content and improves visual function [46].
In our human study we confirmed (1) the presence of high levels of TNFα and IL-6 in aqueous humor in a larger population of RP patients than previously reported [11]; and we observed that (2) RP patients with higher TNFα values show better visual function (visual field). The conflictive positive correlation, between TNFα and better visual function, may be due to the different stage of the disease of the patients. It has been shown that an increase of proinflammatory markers, including TNFα, in mice models of RP occurs just before photoreceptor cell loss [12]. Therefore, it is tempting to speculate that at early onset of RP, when proinflammatory markers are elevated, visual function is better in patients, and after these stages patients lose visual function in parallel with TNFα decrease. In any case, these conflicting, and interesting, results strongly suggest that further studies are needed for clarification.
In the last few years, TNFα has been widely recognized as an attractive therapeutic target for the treatment of retinal diseases. Different types of monoclonal antibodies against TNFα, such as Infliximab, Adalimumab, Certolizumab pegol and Golimumab, or circulating receptor fusion protein, such as Etanercept, have been used to treat glaucoma [56,57], ischemic retinopathy [58] or AMD [59].
The role of TNFα in photoreceptor degeneration and the possible therapeutic use of antibodies against TNFα in the treatment of RP or other retinal degenerations remain quite unexplored. Based on previous studies we decided to evaluate the potential protective effect of the blockade of TNFα in an experimental porcine model of retinal degeneration. In a previous report we demonstrated that this porcine model recapitulated some aspects, especially those related to oxidative stress and inflammation, of the retinal degeneration observed in small animals after PDE inhibition [60,61] and RP patients [8,11]. Sustained elevation of intracellular cGMP in porcine retinal explants triggered different downstream effectors of cell death related to caspase-dependent mechanisms (caspase-3) and caspase-independent mechanisms (calpain-2 and probably PARP activity) [25].
Our current study demonstrated that retinal degeneration accompanied by upregulation of TNFα and IL-6, GFAP and oxidative damage was ameliorated by blocking TNFα with Infliximab. Under our experimental conditions, Infliximab reduced retinal degeneration in all cell layers, mainly in the ONL, by decreasing the number of TUNEL-positive cells, supporting the idea that inflammation plays an important role in the processes of cell death.
We found that Infliximab reduced caspase-3 activity and the number of cleaved caspase-3 positive cells across the different cell layers, especially at the INL. Co-localization studies of caspase-3 and PAR with TUNEL assay suggested that TNFα is promoting cell death through caspase-independent mechanisms in ONL and GCL and caspase-dependent mechanisms in INL. TNF signaling can lead to cell death to two distinct outcomes, each of which is initiated by different signaling complexes: the apoptosis mode and the necrosis mode.
The apoptosis mode includes the extrinsic pathway, mainly mediated by caspases, and the intrinsic or mitochondrial pathway, that rely on the balance between the proapoptotic and the anti-apoptotic proteins from the Bcl-2 family. Both pathways converge on the same execution pathway. The execution pathway is initiated by the cleavage of caspase-3 and results in DNA fragmentation and cell death.
We measured indirect activation of PARP through quantification of PAR accumulation. We found an upregulation of PAR due to PDE6 inhibition. However, blockade of TNFα did not prevent PAR accumulation but also increased it. PAR polymers are mainly degraded by poly (ADP-ribose) glycohydrolase (PARG) enzymes, some of them activated by caspase-3 cleavage [62]. On the other hand, PARP can be inactivated by caspase-3 cleavage [63]. Therefore, the inhibition of caspase-3 induced by Infliximab could inhibit PARG activity and prevent PARP inactivation thus exacerbating PAR accumulation at all cell layers of retinal explants. These results support that PARP pathway is independent of TNFα-associated pathways in this experimental model of retinal degeneration ( Figure 6).
These results were supported by previous reports in which reactive gliosis (GFAP overexpression) induced by exogenous TNFα was prevented by Adalimumab, other monoclonal anti-TNFα, in a similar model of organotypic culture of porcine neuroretina [64]. Activated Müller cells can release antioxidants, growth factors, and cytokines, including TNFα, contributing to retinal regeneration or to neurodegeneration. Müller cells are activated in models of RP [49,[65][66][67][68] resulting in overexpression of GFAP, translocation of Müller cell bodies to the outer retina and thickening of their processes [69].
As previously shown, retinal degeneration induced by PDE inhibition was accompanied by oxidative stress in porcine retinas [25]. This is consistent with the idea that oxidative stress is also contributing to the progression of RP in animal models [70][71][72] and RP patients [8]. In the current study, we demonstrated that Infliximab partially prevented antioxidant defense depletion but not oxidative stress markers. Infliximab normalized the total antioxidant capacity in Zaprinast-treated explants, but it failed to return TBARS and NOX to control levels. In retinas of rd10 mice antioxidant treatment reduced inflammatory mediators and photoreceptor cell loss [12]. Based on these data, it is tempting to speculate that the low Infliximab effect could be due to oxidative stress preceding upregulation of inflammatory mediators [12,[73][74][75]. The other possibility is that Infliximab affects other oxidative stress markers that we did not measure in this study. Anyway, further studies will be needed to explore this issue in more depth.
In summary, our results corroborate that RP patients have ocular inflammation and that TNFα plays an important role in the retinal degeneration induced by PDE6 inhibition in cultured porcine retinas. The mechanisms of cell death vary in the distinct cell layers. TNFα is involved in retinal degeneration through caspase-3 activation, caspase-independent mechanisms and reactive gliosis. Our data suggest that other unknown molecules must be contributing to TNFαmediated cell death in this model. On the other hand, PARP activation is independent of TNFα signaling and it is probably responsible for a future cell death in ONL. The existence of several distinct pathways that trigger programmed cell death implies that an efficient protection requires their simultaneous interruption via combined therapies.
The experimental model of organotypic culture has its own limitations because it involves transection of the optic nerve and mechanical retinal detachment causing retrograde retinal ganglion cell degeneration. To minimize this problem, we have used detached retinas as controls. Moreover, the model cannot recapitulate the whole chronic nature of the degeneration, but we believe that it could be useful for studying some aspects related to the retinal degeneration. In our case, we believe that it may provide a helpful model to design and assay some treatments, such as Infliximab, thus replacing or reducing animal experiments. The use of this model allowed us to evaluate the effect of Infliximab faster and more cheaply than using the available in vivo models of RP.
The current model of retinal degeneration allowed us to describe an interesting and, in our opinion, neuroprotective Figure 6 Diagram showing the possible mechanism of Infliximab in the porcine retinal degeneration model. PDE6 inhibition induces cGMP accumulation and triggers retinal degeneration. The degeneration is accompanied by upregulation of inflammatory mediators, PARP pathway, reactive gliosis and oxidative stress markers. According to the current study, TNFα may be involved in the retinal degeneration by increasing caspase-3 activation and reactive gliosis. Infliximab may prevent cell death by inhibiting caspase-dependent pathways that converge in caspase-3 activation in the INL. Infliximab also may prevent cell death by caspase-independent pathways that remain unclear in the ONL and GCL. Moreover, Infliximab may exacerbate PARP over activation probably through the caspase-3 inhibition. This over activation could contribute to the future cell death. cGMP: cyclic GMP; GCL, ganglion nuclear layer; GFAP: glial fibrillary acidic protein; INL, inner nuclear layer; NO: nitric oxide; ONL, outer nuclear layer; PAR: poly(ADP-ribose) polymers; PARG: poly(ADP-ribose) glycohydrolase; PARP: poly(ADP)ribose polymerase; PDE6: phosphodiesterase 6; TAC: total antioxidant capacity; TBARS: thiobarbituric acid reactive substances; TNFα: tumor necrosis factor alpha. effect of Infliximab that strongly encourages further exploration using other experimental models. Due to the importance of the inflammatory process in the pathogenesis of several retinal degenerative conditions such as RP, AMD, ischemic retinopathy, or glaucoma, targeting inflammation could be a promising therapeutic strategy. In particular, TNFα blockers could be a new therapeutic strategy for the treatment of RP and other retinal degenerative conditions.
Additional files
Additional file 1: Table S1. Individual data for each patient with retinitis pigmentosa (RP).
Additional file 2: Table S2. MANCOVA in aqueous humor from retinitis pigmentosa (RP) patients and healthy controls.
Competing interests
The authors declare that no competing financial or non-financial interests exist.
Authors' contributions CMFC carried out biochemical determinations, performed histological analysis and helped to write the draft. LOG carried out organotypic cell cultures and helped to biochemical and histological analysis. DH carried out statistical analysis and helped to revise the draft. DS obtained human samples and carried out ophthalmic examination. JMM participated in the design of the study and helped to revise the manuscript. RR conceived the study, designed and coordinated the study, performed cytokine determinations in human samples, analyzed data and wrote the draft. All authors read and approved the final manuscript. | 2017-06-27T03:03:34.612Z | 2014-10-10T00:00:00.000 | {
"year": 2014,
"sha1": "b8a86c1c9e8d6ceab4410d109792524834e8ec99",
"oa_license": "CCBY",
"oa_url": "https://jneuroinflammation.biomedcentral.com/track/pdf/10.1186/s12974-014-0172-9",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "b8a86c1c9e8d6ceab4410d109792524834e8ec99",
"s2fieldsofstudy": [
"Biology",
"Medicine"
],
"extfieldsofstudy": [
"Biology",
"Medicine"
]
} |
257483105 | pes2o/s2orc | v3-fos-license | Cutting time & height improve carbon and energy use efficiency of the forage–food dual-purpose ratoon rice cropping
Forage–food dual-purpose ratoon rice cropping (FFRR) is used to balance forage demands with ratoon rice grain yields, that is whole plant (stem and sheath, panicles) cuttings in the main season are used as forage, and rice in the regeneration season is used as food. In this study, the local ratoon rice production system as the control, we were carried out the field experiment of cultivation practices (cutting time and cutting height), and investigated the system productivity, economic benefits, carbon footprints and energy use efficiency. The energy use efficiency, energy productivity and energy profitability increased with cutting time delay, and cutting height decreased. Significant differences of these index were observed among the treatments for cutting time and cutting height (p < 0.05). Carbon efficiency and carbon sustainability index was increase with cutting time delay, and there was significant difference among the treatment of cutting time in 2018 (p < 0.05), the minimum carbon footprint of FFRR was 78.6 kgCO2 t−1 averagely at the cutting time of 30 days after the flowering stage. In 2018, the maximum net income of FFRR was 30,577 CNY hm−2 at a cutting time of 30 days after the flowering stage while the stubble height was 10 cm, and dependent on the forage yield of the main crop; in 2019, the maximum net income of FFRR was 27,326 CNY hm−2 at a cutting time of 10 days after the flowering stage while the stubble height was 10 cm, and dependent on the grain yield of the ratoon crop. Therefore, the optimal cultivation practice of the FFRR (cutting at 30 days after the flowering stage and with a stubble height of 10 cm) showed higher carbon and energy use efficiency, economic benefits of the FFRR were fluctuated with the price of forage of the main crop and rice grain of the ratoon crop.
Introduction
It needs to adjust its agricultural production system to balance cereal production and livestock farming [1]. According to statistics, the food (e.g., rice and wheat grain) consumption per person declined from 138.9 kg in 2013 to 117.9 kg in 2019. In particular, rice was food for more than 60% of the population in China, and the annual consumption of red meat (pork, beef, and mutton) per person grew from 19.8 kg in 2013 to 26.9 kg in 2019 (China Statistical Yearbook). China's red meat consumption will continue to grow into the future and stimulate the development of livestock farming. Large amounts of forage are needed to feed domestic animals; consequently, rice straw has been widely perceived as an important source of forage. The average planting area of the rice (Oryza sativa L.) production system is 6 million hm − 2 , and this system needs to meet the developing food structure needs of Chinese diets.
Ratoon rice is a second rice crop produced from the stubble left behind after the main crop has been harvested [2], and there is more than 1 million hm 2 of ratoon rice cropping in southern China [3]. Rice ratooning is associated with a significantly larger net energy and cost-benefit ratio and substantially lower global warming potential than are middle-and double-season rice cultivation techniques [4]. Reference [5] reported that developing ratoon rice as forage could increase forage yields without placing extra demand on fields in subtropical and temperate rice areas, but it does not balance cereal production with forage production for livestock farming. Based on the need for forage and improvements in ratoon rice grain yields, consequently, we propose a new ratooning rice cropping model, a forage-food dual-purpose ratoon rice cropping (FFRR) model, to balance forage demands with high ratoon rice grain yields.
Cutting time and stubble height are two important factors affecting the rice yield and quality of the ratoon crop, and affecting forage (straw) yield of the main crop. For example, Ref. [6] reported that significant differences in ratoon rice grain yield were observed with cutting time and cutting height of the main crop, and that the highest grain yield was found at a stubble height of 40 cm. Reference [7] reported that the rice quality (such as the amount of head rice and protein content) of main crops were significantly lower than those of ratoon crops. However, Previous studies was concentrated on grain yield of the ratoon crop by different cutting time and stubble height, and the report about the effect of forage yield on different cutting time and stubble height was few. Therefore, the scientific hypotheses of the study were that how balancing rice straw as forage and rice grain as food according to the cutting time and stubble height of the main crop would be beneficial for the development and improvement of ratoon rice production.
In Hunan Province, the cutting time and stubble height in the ratoon rice production system are about 30 days after the flowering stage and 30 cm, the annual grain yield is about 12-15 t hm − 2 , and the ratio of the grain yield between the main crop and ratoon crop is 7:3 or 8:2. However, poor rice quality (high temperatures during the filling stage) of the main crop and a lower grain yield of the ratoon crop have affected rice producers' income, although ratoon rice attains a higher annual grain yield, net energy yield, and net economic return than middle-season rice. In this study, based on a new ratooning rice cropping model (FFRR), compared with the traditional ratoon rice production system, with regard to total annual yields, energy, profits, and environmental footprint, the goal of this study was to investigate the effects of cultivation practices (cutting time and cutting height) on the forage yield of the main crop (the whole plant exclude stubble) and the rice grain yield of the ratoon crop to provide direction for agricultural practice.
Ratoon rice planting
FFRR was established in Yiyang City, Hunan Province, (29 • 08 ′ N, 112 • 26 ′ E) in 2018-2019. Its basic climate condition was characterized by annual average sunshine time of 1643.3 h, annual average temperature of 16.9 • C, the coldest month (January) has an average temperature of 4.3 • C and the hottest month (July) an average temperature of 29.1 • C, frost-free period of 264 days, Fig. 1. Schematic diagram of the forage-food dual-purpose ratoon rice cropping in the experiment. accumulated temperature (≥10 • C) of 5240 • C, and an average rainfall of 1240.8 mm, concentrating in the period from May to September. The rice cultivars were Xiangliangyou 900 (XLY900) that was widely planting in the Hunan Province, and the planting area was 1 hm 2 . The detail of these rice cultivar characteristics was posted on the China rice data center (www.ricedata.cn).
Sowing was done on March 29th, and the seed quantity was 22.5 kg hm − 2 . At a seedling age of 30 days, machine transplanting was performed (on April 29th). The nitrogen (N) application rate was 370 kgN hm − 2 (220 kgN hm − 2 for the main crop and 150 kgN hm − 2 for the ratoon crop), in accordance with the local high-yield and high-quality ratooning rice cultivation system. The P 2 O 5 and K 2 O application rates were 90 and 225 kg hm − 2 , respectively.
Begin with the flowering stage of the main crop, three cutting times (10, 20, and 30 days after the flowering stage) were tested. Two cutting heights (stubble of 10 and 30 cm) were used per cutting time, respectively, these variants were applied in three replications with an area of 500 m 2 per the treatment (Fig. 1). Begin with 10 days after the flowering stage in the main crops, rice plant exclude the rice stubbles, as forage, was harvested using a combine harvester. The local ratoon rice production system as the control, at the mature stage of the main crop, the rice grain was harvested using a combine harvester, and the cutting time and heights were 30 days after the flowering stage and 30 cm stubble. At the mature stage of the ratoon crop, the rice grain was also harvested using a combine harvester.
Pesticides, including those used to prevent rice planthopper, and damage from rice borers. Herbicides was used for controlling barnyard grass, and fungicide, including those used for seed disinfection and sheath blight controlling. Other cultivation measures followed the local high-yield and high-quality ratooning rice cultivation system.
Data records
Records of every input and output in the FFRR were kept as raw data for assessing the carbon budget, energy, and economics. The input and output data, including labor (adult men), fertilizers, diesel fuel, electric motors, plastics (general), pesticides, fungicides, herbicides, irrigation water, rice seed, rice straw yield, and rice grain yield were calculated and recorded. The inputs of the FFRR were the same in 2018 and 2019. The output data in 2018-2019 were used as average values. The carbon transformity [8,9] and energy transformity [10] and the prices of materials during the years were as listed in Table 1. Experimental research and field studies on cultivated plants, including the collection of plant material, must comply with relevant institutional, national, and international guidelines and legislation.
Carbon footprint and related indices
The carbon transformity (Table 1) was used to calculate the carbon footprints (CFs) of the various inputs. Rice straw, grain, and root were used to calculate the carbon quantity of the various outputs. The harvest index of the ratoon crops was 0.45, the ratio of root to aboveground portion was 0.10, and the transformation coefficient of biomass and its carbon content was 0.42 (grain) and 0.38 [11]. The summations of the inputs and outputs were represented as the total carbon input and output, respectively. The indices included Table 1 The carbon and energy transformity, the price of the materials in the forage-food dual-purpose ratoon rice cropping. crop primary productivity (CPP = the sum of all outputs), carbon footprint (CFy = CFs/system productivity), carbon efficiency (carbon output/carbon input), and carbon sustainability index (CSI = [carbon output/carbon input] − 1).
Economic benefits and related indices
Refer to the method of [8] the economic indices used in the study were gross income (GI), net income (NI), output/input ratio, cost-benefit ratio (NI/input), and labor productivity (NI/quantity of labor input).
Data analysis
Means were calculated using Microsoft Excel 2007 software. We performed a factorial analysis of variance and a least squares difference to test for statistically significant differences between days after the flowering stage and stubble height using Statistix 8.0.
Inputs of the FFRR
Three metrics, energy input, equivalent carbon emissions for agricultural inputs, and inputs based on money, were used to describe the inputs of the FFRR ( Table 2). The total energy equivalent of the FFRR was 33,993.0 MJ hm − 2 . The energy equivalent of the nonrenewable industrial subsidiary source (F) was 343,316.1 MJ hm − 2 , accounting for 88.9% of the FFRR, and that of the renewable source (R) and renewable organic subsidiary source (R1) was 3825.5 MJ hm − 2 . The energy equivalent of N was 22,714.3 MJ hm − 2 , which accounted for 66.2% of the FFRR.
The equivalent carbon emission for the agricultural inputs of the FFRR was 1424.0 kgCO 2 eq. hm − 2 , and that of the nonrenewable industrial subsidiary source (F) was 1058.9 kgCO 2 eq. hm − 2 , accounting for 71.3% of the FFRR. The carbon footprint of fertilizer (including N, P 2 O 5 , and K 2 O) was 806.2 kgCO 2 eq. hm − 2 , accounting for 60.3% of the total. Inputs based on money of the FFRR were 14,113.4 CNY hm − 2 .
Outputs of the FFRR
The straw yield of the main crop increased with cutting time delay and decreased with cutting height increase. A significant difference was observed among different cutting times (T, p < 0.01) and cutting heights (H, p < 0.01), but the interaction produced no significant difference between T and H. The maximum straw yields of the main crop were 15,438 kg hm − 2 in 2018 and 18,192 kg hm − 2 in 2019 at 30 days after the flowering stage while the stubble height was 10 cm. However, the variation in grain yield of the ratoon crop was opposite the straw yield result of the main crop. The grain yield of the ratoon crop decreased with cutting time delay and increased with cutting height increase. The maximum grain yields of the ratoon crop were 6633 kg hm − 2 in 2018 and 7199 kg hm − 2 in 2019 at 10 days after the flowering stage while the stubble height was 30 cm, and significantly higher (p < 0.01) than that of the local ratoon rice production model. Table 2 Estimates of energy inputs, equivalent carbon emissions for agricultural inputs used and input based on money in the forage-food dual-purpose ratoon rice cropping. The numerical value of the bracket was percentage The average content of crude protein in the rice straw was 9.35% (range from 8.05% to 10.66%), and the average total crude protein was 1630 kg hm − 2 (from 1348 to 1911 kg hm − 2 ), while that of the local ratoon rice production model was 1229 kg hm − 2 averagely. The variance analysis revealed that there were no significant differences in crude protein content among the treatments of T, H, and T × H (p > 0.05, Table 3); and significant differences in total crude protein among the treatments of T (p < 0.05).
Energy budget of the FFRR
The maximum total output energy of the FFRR was 274,099 MJ hm − 2 in 2018 and 283,654 MJ hm − 2 in 2019 with the treatment of 30 days after the flowering stage while the stubble height was 10 cm, and the variation in net energy was the same as the variation in the total output energy ( Table 4). The variance analysis showed that there were significant differences in total output energy and net energy among the cutting time and cutting height treatments (p < 0.05, Table 4). Compared with the local ratoon rice production model, FFRR was shown high total output energy and net energy.
The energy use efficiency, energy productivity and energy profitability increased with cutting time delay and cutting height decreased (Table 4). Significant differences of these index were observed among the treatments for cutting time and cutting height (p < 0.05). The interaction between T and H also revealed no significant differences. Compared with the local ratoon rice production model, FFRR was shown high energy use efficiency, energy productivity and energy profitability.
Carbon budget of the FFRR
The maximum CPP of the FFRR were 11,065 kgC hm − 2 in 2018 and 10,629 kgC hm − 2 in 2019 in the treatment of 30 days after the flowering stage while the stubble height was 10 cm, and there were no significant differences among the treatments of T and H (Table 5). There was a significant difference in carbon footprint among the treatments of days after flowering stage. The minimum carbon footprint of the FFRR was 74.8 kgCO 2 t − 1 in 2018 and 71.4 kgCO 2 t − 1 in 2019 in the treatment of 10 days after the flowering stage while the stubble height was 30 cm, while there was significant difference (p < 0.05) between the treatment of 30 days after the flowering stage while the stubble height was 10 cm and the local ratoon rice production model (Table 5).
Carbon efficiency and carbon sustainability index was increase with cutting time delay, and there was significant difference among the treatment of cutting time in 2018 (p < 0.05). The carbon efficiency and carbon sustainability index of the FFRR was 9.94 and 26.33,
Table 3
Variation of straw yield of main rice and grain yield of ratoon rice among different cutting time and cutting height in the forage-food dual-purpose ratoon rice cropping. Different small letter between means were assessed with the least significant difference (LSD0.05) at p ≤ 0.05. * means significant at the 0.05 level, ** means significant at the 0.01 level, and ns means the correlation is not significant. The same below.
respectively in 2018, and 9.55 and 25.25, respectively in 2019 with the treatment of 30 days after the flowering stage while the stubble height was 10 cm (Table 5).
Economic benefits of the FFRR
The net income (NI) of the FFRR was 30,577 CNY hm − 2 in 2018 and 21,367 CNY hm − 2 in 2019 in the treatment of 30 days after the flowering stage while the stubble height was 10 cm, and there were significant differences among the treatments of T and H (Table 6). A high NI depended mainly on a high rice grain yield of the ratoon crop. The grass income of the FFRR was 45,816 CNY hm − 2 in 2018 and 36,605 CNY hm − 2 in 2019 in the treatment of 30 days after the flowering stage and a stubble height of 10 cm, and there was significant difference among the treatment of cutting time in 2019 (p < 0.05). Similarly, a high cost-benefit ratio and labor productivity were realized in the treatment of 30 days after the flowering stage and a stubble height of 10 cm in 2018. Economic benefits of the local ratoon rice production model were lower (p < 0.01) than that of the treatment of 30 days after the flowering stage while the stubble height was 10 cm.
Improving system productivity of the FFRR in practice
In this study, we performed an on-farm assessment of system productivity among different cutting times and stubble heights of rice. The findings of this study show that the straw yield of the ratoon crop increased with a cutting time delay and stubble height increase. The average maximum straw yield was 18.1 t hm − 2 in the treatment of 30 days after the flowering stage and a stubble height of 10 cm (Table 4). Similarly, Ref. [5] reported that Zhunliangyou 608 could be used as ratoon rice in subtropical and temperate rice planting areas to produce good-quality forage by using a 30 cm stubble height for the first season when the grain had reached 80% maturity. However, forage yield rather than forage quality is more important for livestock farming in southern China, especially considering the supply shortage of fresh forage during winter; therefore, a lower stubble height is beneficial because of the higher forage yield. Our data in the forage quality experiment reported by Ref. [12] was indicated that whole plant rice silage could be used to partially replace whole plant corn silage in dairy cows diet, which could improve the immune capacity and reduce feeding costs of dairy cows and provide a good foundation for obtaining higher economic benefits. Compared with the local ratoon rice production model, the FFRR *means significant at the 0.05 level, ** means significant at the 0.01 level, and ns means the correlation is not significant.
was produce forage and grain simultaneously for supporting the development of livestock farming and maintaining the supply of the food. Cutting time and stubble height are two important factors affecting the rice yield and quality of ratoon crops. Reference [6] reported significant differences in ratoon crop grain yield between different cutting times and cutting heights and found a maximum ratoon crop grain yield at a stubble height of 40 cm in America. [11], working in Guangdong Province, southern China, reported that a 5 cm stubble height prolonged the ratoon crop growth duration and increased the spikelet number per panicle. In Hunan Province, central China, the cutting time and stubble height in the ratoon rice production system are about 30 days after the flowing stage and 30 cm, respectively. Our results indicate that the grain yield of the ratoon crop decreased with a cutting time delay and stubble height decrease ( Table 4). The highest grain yield of the ratoon crop occurred with a cutting time of 10 days after the flowering stage and a cutting height of 30 cm. Therefore, the most suitable cutting time and cutting height depend on local climate conditions.
In FFRRs systems, improving the agricultural practices of optimal cutting time and cutting height can help balance the rice straw yield of the main crop for forage and the rice grain yield of the ratoon crop for unprocessed grain. The results of this study help to fill knowledge gaps about the performance of FFRR systems in China. Consequently, farmers can select the optimal strategy to satisfy the demands for unprocessed grain, forage, or both.
Improved practices improve the energy and carbon budgeting of FFRR systems
For cleaner production technology, reducing the carbon footprint, improving energy consumption and gas emissions, and maintaining soil health simultaneously are the major targets for fulfilling the sustainable production goals of agriculture [9]. Ratoon rice exhibits a significantly higher net energy ratio and benefit-to-cost ratio and a substantially lower yield-scaled global warming potential than the other two cropping systems (i.e., double-season and middle-season rice [4]). In this study, based on the same energy inputs, compared with the local ratoon rice production model, equivalent carbon emissions for the agricultural inputs used, as well as the inputs based on money, the high system productivity (including straw yield and ratoon rice yield) achieved by improving cutting time and cutting height led to high net energy and EUE. However, there was no alteration of CPP, possibly because the CPP of each rice cultivar was relatively constant under the same ecological conditions and cultivation. Thus, there was no effect by cutting time or cutting height. Different small letter between means were assessed with the least significant difference (LSD0.05) at p ≤ 0.05. * means significant at the 0.05 level, ** means significant at the 0.01 level, and ns means the correlation is not significant.
Advantages of FFRR systems
In FFRR systems, the rice main crop (including the rice stem and grain) is harvested to help satisfy the forage need for large domestic animals and the rice ratoon crop is harvested to help satisfy the demand for high-quality unprocessed grain to feed people. The ability to increase the forage yield by improving the cutting time and cutting height when there is enough high-quality unprocessed grain or decrease the forage yield when unprocessed grain is insufficient is a major advantage of the system, compared with the local ratoon rice production model. Thus, FFRR is an effective way to ensure both forage security and unprocessed grain security, especially in relatively impoverished regions of Southeast Asia, Africa, and South America.
Conclusions
Compared with the local ratoon rice production model, FFRR was shown high total output energy, net energy, energy use efficiency, energy productivity and energy profitability. There was significant difference (p < 0.05) between the treatment of 30 days after the flowering stage while the stubble height was 10 cm and the local ratoon rice production model. Our data was shown that economic benefits of the local ratoon rice production model were lower (p < 0.01) than that of the treatment of 30 days after the flowering stage while the stubble height was 10 cm. Therefore, FFRR systems can be optimized effectively by selecting the ideal cutting time and cutting height, from the perspectives of carbon footprint and economic benefits, harvesting at 30 days after the flowering stage and with a stubble height of 10 cm was the optimal cultivation practice of the FFRR.
In this study, only one high-yielding cultivar were studied according to local high-yielding cultivation technology, there was significant difference of rice aboveground biomass and grain yield among the cultivars, moreover, fertilizer (for example: nitrogen rates and its application) management was affected significantly rice aboveground biomass and grain yield, further research was needs to testify the conclusions from more rice genotype or cultivars (high-yielding or high-quality) and nitrogen management.
CRediT author statement
Chen Yuanwei, Zheng Huabin, Wang Weiqin, Tang Qiyuan: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Different small letter between means were assessed with the least significant difference (LSD0.05) at p ≤ 0.05. * means significant at the 0.05 level, ** means significant at the 0.01 level, and ns means the correlation is not significant.
Funding
The study was financially supported by the Earmarked Fund for China Agriculture Research System (No. CARS-01-27). Funds from the Ministry of Agriculture & Rural affairs.
Data availability statement
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Institutional review board statement
Not applicable.
Informed consent statement
Not applicable.
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. | 2023-03-13T05:06:50.821Z | 2023-02-24T00:00:00.000 | {
"year": 2023,
"sha1": "2cd6b45ac3b6de246c5bb9c7ec5a0ab4b25ab296",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1016/j.heliyon.2023.e14042",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "2cd6b45ac3b6de246c5bb9c7ec5a0ab4b25ab296",
"s2fieldsofstudy": [
"Agricultural And Food Sciences"
],
"extfieldsofstudy": [
"Medicine"
]
} |
233392506 | pes2o/s2orc | v3-fos-license | Excel Add-In Program for Computation of Thermodynamic Properties of R134a and R718 Refrigerants
Thermodynamic properties of R134a (1,1,1,2Tetrafluoroethane) and R718 (water) have been developed as a Microsoft Excel add-in called ThermoAnalysis. The mathematical correlations for the thermodynamic properties of R134a and R718 were formulated from well-known equations of state and used to develop ThermoAnalysis based on a computer program in Microsoft Excel Visual Basic for Application language. ThermoAnalysis provides thermodynamic properties for the saturated regions for R134a and R718 refrigerants, and the subcooled and superheated regions of R718 refrigerant. The calculated values are accurate compared to the standard properties tables for refrigerants. A typical thermo-fluid problem was discussed to illustrate how ThermoAnalysis can be used for practical problem solving. Generated properties’ data can be easily used in the Excel spreadsheet for process analysis, simulation and design of systems that use R134a and R718. ThermoAnalysis is handy for both students and practicing thermo-fluid engineers.
INTRODUCTION
Thermodynamic properties such as specific volume, entropy, temperature, enthalpy, pressure, and internal energy of pure substances are important in the analysis and design of various thermodynamic systems. A pure substance is one that is homogeneous in chemical composition and invariable in chemical aggregation [1]. For a pure substance, there are three measurable properties: pressure (P), specific volume (ν) and temperature (T). The equation that is used to express the relationship between these properties in the gaseous phase is called the equation of state of the pure substance. A functional p-ν-T relationship of a gas can either be theoretical, generalized or an empirical equation fitted from experimental data. The simplest theoretical equation of state is the ideal or perfect gas equation of state representing the behaviour of the gas at low pressures (tending to zero) and high temperatures. Theoretical equations can be used to generate the property data of substances that represents their physical behaviour. However, realistic property data are usually determined from partially-empirical methods and are either provided as charts or tables. Reading properties data from tables and charts takes time and is prone to error. It becomes even more strenuous and time-consuming when trying to read a particular data point that is not explicitly stated and requires interpolation between two points. This limitation of the tables and charts has necessitated and motivated the development of computer packages for obtaining properties data of substances.
The widespread use of computer in modern-day engineering training has rendered the use of thermodynamic property tables and charts obsolete [2]. Consequently, a number of computer programs have been developed to automate the process of obtaining property data. Taftan Data [3] developed Thermo Utilities v3.5, an MS Excel add-ins software package that can be used to design, analyze or optimize power plants, air conditioning systems and other chemical processes. Lemmon,et al. [4] developed a NIST Reference Fluid, Thermodynamic and Transport Properties (REFPROP) software which provides tables and plots of the thermodynamic and transport properties of industrially important fluids and their mixtures with an emphasis on refrigerants and hydrocarbons. REFPROP can only compute results for saturated properties of refrigerants considered. Tan and Chua [5] developed a Java programming application (Java Applet) for the computation of thermodynamic properties of steam and R134a refrigerants. They presented correlation formulae and results for saturated region but none was presented for superheated and sub-cooled regions, although they stated that their software can compute results for these regions with much less accuracy compared to the saturated region. Java applets for thermodynamic properties may not be available to the majority of users because it requires internet connectivity. On the other hand, Microsoft Office is readily available and it would be advantageous to have Microsoft embedded software for thermodynamic properties instead of Java applets. In ref.
[6] a user-friendly MS Excel add-in for the thermodynamic properties of R152a called ThermoProp_R152a was developed. The study showed that the Excel add-in makes it possible to use Excel spreadsheet for direct process analysis, simulation and system design.
In this paper, we present a user-friendly MS Excel add-in package for the thermodynamic properties of R134a and R718 (called 'ThermoAnalysis'), which can aid students and practicing engineers to solve problems in the area of refrigeration and air-conditioning systems. ThermoAnalysis software covers the saturated region of both R134a and R718, and the subcooled and superheated region of R718. An important feature of the present software is that it is an MS Excel-based program. Therefore, its results can be easily applied for spreadsheet problem-solving which can aid realtime simulation of practical problems in classroom settings. thermo-properties was developed based on the mathematical correlations. In implementing the algorithm, attention was given to the accuracy of the numerical solution for the mathematical correlations and a maximum relative error bound of 0.01% was used. The algorithm was converted into a computer code in Microsoft Excel Visual Basic for Application (VBA) to develop ThermoAnalysis. An advantage of Excel VBA is that it allows for the application of familiar and user-friendly interfaces to implement an add-in software package. Also, because of its inherent connection with other MS Excel tools, Excel VBA enables direct use of Excel spreadsheet capabilities to generate properties table and for process analysis.
Mathematical correlations
To determine the thermodynamic properties of a refrigerant or any pure substance, the following minimum experimental data/correlations are required The other properties to be calculated are the internal energy ( ), enthalpy ( ) and entropy ( ). Equations (2 -5) are used to calculate for the isothermal changes in , , in the gaseous phase. The MBWR equation of state provides the most accurate fit of thermodynamic data for R134a over a wide range of temperatures and pressures. The data fit and calculation of constants for HFC-134a were performed at the National Institute of Standards and Technology (NIST) [4]. All constants were calculated in SI units which is consistent with the unit system of the present work. The MBWR equation of state is given as follows:
Saturation liquid volume ( ):
The saturation liquid volume was evaluated using the following equation: where is in , is in and .
Latent heat of vaporization ( ):
The latent heat of vaporization was evaluated using the following equation: Differentiating equation (7) with respect to temperature and substituting it into equation (2), and then substituting the resulting equation into equation (6) gives the vapour phase enthalpy. The vapour phase internal energy was then obtained using the equation: . Substituting the differential of equation (7) with respect to temperature into equation (5), we have the vapour phase entropy. Water as a refrigerant (R718) is one of the oldest refrigerants used for refrigeration applications because of its availability and excellent thermo-chemical properties. The most accurate equations that model the thermodynamic properties of water are given by the International Association for the Properties of Water and Steam (IAPWS) for different regions which cover the following temperature and pressure range [9,10]. For the purpose of this work, the following equations were used:
Saturated densities
The density of the saturated liquid was calculated from: with The density of the saturated vapour was calculated from: with Saturated specific enthalpy and specific entropy Auxiliary equations: with The specific enthalpy of the saturated liquid was calculated from: Equation (15) yields the specific enthalpy of the saturated liquid when used in conjunction with Equations (10), (11), and (13). The specific enthalpy of the saturated vapour was calculated from: Equation (16) yields the specific enthalpy of the saturated liquid when used in conjunction with Equations (10), (12), and (13). The specific entropy of the saturated liquid was calculated from: Equation (17) yields the specific enthalpy of the saturated liquid when used in conjunction with Equations (10), (11), and (14). The specific entropy of the saturated vapour was calculated from: Equation (18) yields the specific enthalpy of the saturated liquid when used in conjunction with Equations (10), (12), and (14). These equations are valid from the triple point to the critical point. This corresponds to 273.16 K ≤ T ≤ 647.096 K Sub-cool region The basic equation for this region is a fundamental equation for the specific Gibbs free energy . This equation is expressed in dimensionless form as , and can be represented as: where and with , . The coefficient and exponents and can be found in [9, 10].
Superheat region
The basic equation for this region is a fundamental equation for the specific Gibbs free energy . This equation is expressed in dimensionless form, , and it is separated into two parts, an ideal-gas part, , and a residual part, , such that it is expressed as: The ideal-gas part, , and the residual part, , of the dimensionless Gibbs free energy equation are given as: where and with , . The coefficients and and exponents , and can be found in [9,10]. All other thermodynamic properties can be derived from the basic equations by using the appropriate combination of the dimensionless Gibbs free energy ( ) and its derivatives. Relations between the relevant thermodynamic properties, and its derivatives can be found in refs. [9, 10].
Statement Algorithm for implementation of ThermoAnalysis
The thermodynamic property analysis was carried out in the following steps: Start 1. Input data (i)choose the refrigerant for analysis (ii)choose the phase of the refrigerant (saturated liquid, saturated vapour, mixed vapour-liquid, superheated vapour or sub-cooled liquid); (iii) specify the desired (unknown) property (specific enthalpy , specific entropy, specific volume, temperature, pressure , specific internal energy, or quality); (iv) choose the corresponding dependent variable(s) which could be any one variable (for saturated states) or two variables (for other states); (v) enter the value(s) of the dependent variable(s) 2. Compute the desired unknown property; 3. Output the desired data (property name and numerical value); 4. Use the output for process analysis, if desired;
Generate table of properties, if desired;
Stop.
The above algorithm was used to develop ThermoAnalysis.
Using ThermoAnalysis 2.3.1 Installation
To start with, the ThermoAnalysis software must be installed as an Excel Add-in using the standard procedure of installing an Add-ins. The following steps can be used to install ThermoAnalysis. a) Save the Thermoanalysis excel file as an excel AddIns. b) Go to File, Options, then Add-Ins c) On the displayed interface, go down to manager, then click GO… d) On the new displayed interface, check the Thermoanalysis option and click OK. Now, the software is ready for use.
Application
After installation, depending on the Microsoft Excel version used, 'Add-Ins' or 'ThermoAnalysis' would show on the menu bar. If 'Add-Ins' reflects on the menu bar as shown in Figure 2, then clicking on the Add-Ins menu displays the 'ThermoAnalysis' menu. On the other, if 'ThermoAnalysis' is displayed on the menu bar after installation, clicking on it would display the 'ThermoAnalysis' menu. After clicking on the ThermoAnalysis menu, a drop-down menu list appears requiring the user to choose the refrigerant for analysis (either R134a or R718). By clicking on the desired refrigerant, the user interface shown in Figure 3 appears. Figure 3 is the user interface for R134a and a similar interface is applicable for R718. The user interface is very simple to use and like most Microsoft interfaces, requires the user to select the relevant options by clicking and entering the input values in the spaces provided. Clicking the 'Read' button computes the selected property while the 'Transfer' button transfers the computed results to Excel worksheet.
Validation of ThermoAnalysis
The tabulated data generated from the software for the saturated properties of R134a and R718 are shown in the appendix. These data agree with standard properties tables for R134a and R718 as shown by the relative percentage difference charts in Figures 4 and 5. The reference data used to plot the RPD chart R134a were obtained from ref. [7] while that for R718 were obtained from steam tables in ref [12]. The RPD charts show that the deviations incurred in this work are within acceptance limits of engineering purposes.
Example of problem-solving using ThermoAnalysis
To demonstrate the application of ThermoAnalysis, we consider a typical thermodynamic problem on refrigeration that can be used in the classroom setting.
Problem statement
An R134a hermetic (directly-coupled motor) reciprocating compressor with 4 percent clearance is to be designed for 7.5 [TR] capacity at 4 o C evaporating and 40 o C condensing temperatures. The compression index is assumed to be 1.15 [-]. The pressure drops at suction and discharge valves were be assumed as 0.2 and 0.4 bar respectively. Determine: (a) Power consumption of the compressor (b) COP of the cycle (c) Volumetric efficiency of the compressor.
Solution
The solution process and the results obtained using ThermoAnalysis are show in Table 1.
CONCLUSION
A software package, called ThermoAnalysis, that can compute the thermodynamic property data for R134a and R718 was developed as an MS Excel Add-ins using Excel Visual Basic for Applications. A main feature of ThermoAnalysis is that it can compute data for the sub-cooled and superheat regions for R718, which is not readily available in similar programs [4,5]. The software package, also takes advantage of the computational and spreadsheet capabilities of MS Excel to generate property tables as shown in the appendix. Although the generated data showed some deviations from experimental data and previous works, the deviations are well within the acceptable range for engineering purposes. Finally, we demonstrate how ThermoAnalysis can be used to solve refrigeration and airconditioning problems using a typical example that is suitable for classroom exercise. This implies that ThermoAnalysis would be a good teaching and learning resource for courses on 0 0 thermodynamics, refrigeration and air-conditioning. Furthermore, it can be a useful and cost-effective resource in actual engineering professional practices that involves the use of properties data.
DECLARATION
The author declared that there is no potential conflict of interest with respect to the research, authorship, and/or publication of this article.
ACKNOWLEDGMENT
This work is dedicated to the memory of Late Prof. C.O.C. Oko, under whose guidance much of the work was conducted. No specific fund was received for the research and publication of this work. | 2020-12-24T09:11:48.433Z | 2020-12-17T00:00:00.000 | {
"year": 2020,
"sha1": "2c28af16c72ef3179a0323296d535e0f625bcd6b",
"oa_license": null,
"oa_url": "https://doi.org/10.5120/ijca2020920936",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "c71bc749ab8d0e5da768139a3705c293ad641a6d",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": []
} |
222135566 | pes2o/s2orc | v3-fos-license | How Does the Campus Environment Influence Everyday Physical Activity? A Photovoice Study Among Students of Two German Universities
Background: Sedentary time is high among university students. Prolonged sitting time and reduced physical activity is linked to a number of health risks, therefore interventions to increase options for physical activity on campuses are of high public health relevance. Evidence about the influence of the campus environment on movement and sedentary behavior of students is scarce. This study explores how the structural and social environment of two University campuses are related to students' everyday physical activity. Methods: We used the photovoice method to get a thorough insight into students' daily life on campus. We recruited a total of 46 University students in two German cities (University 1: n = 22, University 2: n = 24). They were asked to take ≥15 photos of places and situations on their respective campus that facilitate or hinder them to be physically active. The pictures were discussed with the participants in 10 focus groups. Focus group discussions were audio-recorded, transcribed and analyzed using content analysis. Results: Both universities do not exploit their potential of fostering daily physical activity on campus, according to the photos and discussions of the participating students. The vast green spaces offer no cues for movement: easily accessible equipment for sports (fixed or mobile) is lacking, walkways are partially hidden, and the facilities discourage from cycling to and on campus. Social norms induce participants to keep sitting during lectures and learning time. It was also pointed out that indoor hallways and foyers could be put to better use with regard to physical activity. The Photovoice project raised the participants' awareness of how the context influences their movement behavior, and helped them come up with solutions to make physical activity easier for students of their respective universities. Conclusion: The studied campuses discouraged students from being physically active by missing out on opportunities—indoors and outdoors—for fostering movement, such as designating the greens for games or walks, or providing sufficient lockers for biking gear. The results can serve as a basis to plan custom-made public health interventions.
INTRODUCTION
Sedentary behavior has been identified as a major health risk. There is a growing number of studies indicating a relationship between prolonged sitting time and health risks, such as an increased risk for all-cause mortality, in addition to several chronic conditions, e.g., diabetes, different cancers, or cardiovascular disease (1). Germany is among those countries with a fairly high sitting time (2). In adults, higher socioeconomic status is related to prolonged sitting in the workplace (3,4). Within universities, sedentary behavior has been a social norm which is ingrained in most typical procedures, structures, and behavioral patterns of these institutions: Students are sitting during lectures, in seminars and in the library, additionally during lunchtime, in the cafeteria and often during breaks between sessions. Sedentary time encompasses on average 34 h per week for University students while staying on the campus, and this does not yet include (passive) transportation, study time at home or leisure time (5). Only 50% of young adults in Germany are physically active for 150 min or more per week, as is recommended by the World Health Organization (6).
It is therefore important to increase options for movement in the daily campus life, in order for students to be able to reduce their sedentary time. Studies from other settings, e.g., schools, show that environmental contexts shape children's and adolescents' everyday life and are likely to influence their (health) behavioral patterns. Quantitative research has shown, for example, that physical activity of students is positively correlated with a larger total school campus size, playground areas and facilities for physical activity (7)(8)(9)(10). Few qualitative studies on that subject shed light on the potential ambivalence of these correlations, as some students reported that they were "too old" for playgrounds and considered "safe play spaces boring" (10).
Studies on structural barriers and facilitators for physical activity experienced by University students are scarce. Arzu et al. showed that the perceived lack of time is a major barrier for physical activity among University students (11). Apart from that, other barriers or facilitators for movement in the University environment have hardly been investigated yet. Understanding the contextual influences that shape sedentary or movement behavior of University students is critical for planning needsbased interventions. Evaluations of interventions that aimed at increasing physical activity among University dwellers show that these approaches tend to be effective, but were mainly conducted in US American colleges, which show significant differences compared to German universities (12). It becomes clear that we need to better understand how the environments within which University students spend a lot of their time might act to enhance or constrain sedentary behavior and physical activity (13).
Therefore, this study intended to explore and understand how factors of the structural and social environment in University campuses influence physical activity and sedentary behavior of students. We chose the Photovoice technique, a participatory action research first described by Wang and Burris (14,15). In Photovoice, the participants are provided with cameras (or use the photograph function of their mobile phones) to identify and document resources, barriers and facilitators with regard to a certain topic or behavior (e.g., food security, physical activity, living with HIV) in their surroundings. Participants are then brought together as a group in order to discuss a selection of the photos taken. These discussions contextualize the photographic data, and have the participants identify common themes and concepts. This process is not only useful to researchers or health promoters, but may also benefit the participants. They may develop critical awareness of their environment and the role that this environment plays in influencing behavior patterns (16). The Photovoice technique is based on Paulo Freire's theories on participatory education, and Feminist theory focusing on giving voice to the disadvantaged (14). It helps people use visual evidence to recognize and voice their problems and potential solutions to researchers, which in turn communicate these concerns and suggestions to policy makers (16). The ultimate aim of the Photovoice technique is to bring about social change in a setting. Since its inception, the Photovoice methodology has been applied to a variety of populations, places, health issues, and disciplines. Some studies have performed Photovoice with adolescents, in order to capture their perspectives on health and well-being (17), healthy eating and active lifestyle barriers (18), or sexual health information, alcohol and drugs (19).
We performed a Photovoice project intending • to actively engage University students in documenting and discussing their campus surroundings and daily lives with regard to physical activity and sedentary behavior • to explore factors in the campus context that hinder or facilitate physical activity and active living
Context of the Photovoice Study
We performed a Photovoice study with students from two German campus universities [total number of students enrolled University 1 (U1): 13,500 and University 2 (U2): 21,000] from May to July 2018. In these two universities, researchers had received funding to implement campus-based measures that promote and facilitate physical activity for students (project "Smart Moving"), and intended to found stakeholder groups with students, lecturers and University administration staff to plan these measures. The Photovoice study was meant to obtain an insight into the daily life of the students, and into the role that the campus context plays in their physical activity. Thereby, we intended to obtain ideas for adequate interventions (=needs assessment) which could inform the stakeholder groups. It was also meant to raise the participants' critical awareness of environmental influences on their movement behavior, and thereby also possibly motivate some participants to take part in further activities (i.e., recruiting them for the stakeholder group).
Research Participants
In qualitative research, a sample's statistical representativeness is not a prime requirement, especially when the research aims at understanding social processes (20). A strategy to ensure rigor in qualitative data collection is using systematic, or purposive sampling, e.g., theoretical sampling, by identifying specific groups of people who possess certain characteristics that are relevant to the social phenomenon being studied (20). In our study, we strove at including (1) a comparable number of male and females in the sample, as views on physical activity may be gender-related, (2) a broad range of different study programs that participants were enrolled in, in order to obtain a balanced composition with regard to expertise, areas usually frequented during study time, and time schedules. Thus, we intended to minimize bias in the composition of the sample. We recruited students at both universities with the aim of recruiting a minimum of 20 students per University. We used bulletins, flyers, University mailing lists, personal contacts and a snowballing technique.
Participants had to fulfill the following criteria: They needed to be (a) enrolled students of the respective universities, (b) able to take digital pictures, and (c) willing to participate in a 1-1.5 h focus group discussion.
The recruitment was terminated when theoretical saturation was reached, i.e., no more new themes came up in the focus group discussions (see below).
Sample
We recruited 22 students (14 female, eight male) in University 1, and 24 students (14 female, 10 male) in University 2. The average age was 23.6 ± 2.2 years (U1), and 23.8 ± 2.8 years (U2), respectively. Four students did not live within the town of the University, but on the outskirts, one student lived further away (1 h drive to and from campus). Most of the interviewees were of German nationality, one was from Tanzania. They were enrolled in 14 different study programs, ranging from tourism and history to medicine and engineering. Most frequent study subjects were psychology (n = 8), sport science (n = 6), and law, teaching and economics (each n = 5). (see Table 1).
Procedure
Students contacting the project team because they had learnt of the study (recruitment strategies: see above) were provided with further information. The participants received a list of specific questions ("Where and when am I physically active during my daily life on campus? What prevents me from being physically active during my daily life on campus?"), which they were asked to consider when taking their pictures. The participants were asked to take at least 15 pictures with their smartphones of places and/or situations where physical activity during their daily life was considered to be easily possible, or not/hardly possible, or made easy or difficult. They were specifically asked to include not only sports, but any physical activity (e.g., taking the stairs instead of the elevator, etc.). To ensure guidance, students were invited to contact the project team any time in case of doubts or questions. As smart phones equipped with cameras are an integral part of the University students' life, training with regard to taking pictures was deemed unnecessary.
Participants were asked to send the photos to the researchers via e-mail. The received photo files were then printed on a larger scale (13 × 18 cm) and laid out during the focus group discussion. The photos were equipped with post-its containing letters and numbers. The participants were asked to mention those letters and numbers during the focus group when talking about a specific picture in order to make the picture distinguishable for the researchers during the analysis.
We performed ten focus groups (five in each University) with 3-5 participants per group. The focus group discussions were guided by prompts (see Table 2). Based on the SHOWED proceeding recommended for the Photovoice approach, the participants were first asked to explain what can be seen on the picture, and what was happening in the photographed scene, in order to understand the specific focus of the participants (21). The participants then described how the picture connected to physical activity, and how the participants appraise this situation. For example, stairs can encourage physical activity (as opposed to elevators), but may also render physical activity difficult (when being an obstacle for using the bike). Finally, the students were asked to come up with ideas how the situation could be changed or used to make physical activity on campus easier (21,22). After conducting three focus groups in both universities, no more novel picture motives came up. As the discussion in the focus groups evolved around the images provided by the students, the same topics were covered in the discussion. Two more focus groups were conducted in each University, to ensure theoretical saturation was reached. The recruitment was stopped subsequently (23).
Informed Consent and Confidentiality, Ethics Approval
The study was approved by the Ethics Committee of the University of Regensburg (18-896-101). Written consent was obtained from all study participants in accordance with the ethics approval. The participants were informed that they could opt out at any time of the process. They received an expense allowance of 45 Euro for their time spent.
Analysis
All focus groups were audio recorded and subsequently transcribed verbatim. All transcripts were de-identified before analysis. The focus group (FG) transcripts were numbered chronologically (FG 01 -FG 10) and classified according to respective University (U1, University 1; U2, University 2) before de-identification. Two researchers coded the transcripts independently, using content analysis (24,25). The researchers extracted themes and topics with regard to environmental barriers and facilitators of physical activity, following an inductive approach. The transcripts were re-read repeatedly, and the themes and topics were compared between the different focus groups to identify cross-cutting themes that came up in several group discussions. The pictures were not explicitly analyzed, but taken as a reference for the text (26). They received a deidentified number according to the student who took the photo (S1-S22 in U1, S23-S46 in U2). To increase the scientific rigor of the analysis, the independent assessments of transcripts by the two researchers were compared; differences were discussed until consensus was reached. In addition, "negative" or "deviant" views were examined with specific thoroughness (20).
RESULTS
During the analysis, several pathways in which the University context shapes physical activity and/or sedentary behavior emerged, and different barriers and facilitators to an active lifestyle could be identified (see Table 3). The participants also came up with solutions to overcome certain barriers.
Active Transport to the Campuses Could Be Enhanced by Improving the Campus Environment for Cyclers and Their Gear
At both universities, the participants explained that the campus could easily be reached by bicycle from centers and living quarters of the respective towns. Cycling paths were provided throughout the towns, encouraging students to take the bike.
There is an old railway trail, which was tarred some time ago. It's perfect for riding the bike, because it is set aside from the road. For me, that is a reason to take the bike to campus. You are out in the nature, and you can clear your mind. (U1 FG2) Participants reported difficulties in parking their bikes. They especially pointed out the lack of sheltered bike racks preventing bikes from getting wet in rain in University 2.
I have a fairly good bike and I don't want it to stand in the rain all day. This really prevents me from going to the University, for example at the weekend, to study in the library. (U1 FG5) University 1 offers a bike station in which students can pump up tire and do little repair works; this is very much appreciated by the participants and considered a factor encouraging them to bike to campus. Bad weather (e.g., rain) itself does not so much deter students from cycling to the campus, but the lack of adequate FIGURE 1 | Participants reported that they needed to carry with them a lot of gear, which is difficult to store on campus (Photo: U1 Student S3). storage room for wet clothing, rain jackets etc. does, according to the focus group discussions (A: Figure 1).
In University 2, there are many stairs all over the outdoor area, and they are considered a major obstacle to riding the bike on campus (B: Figure 2). This might even prevent some students from taking the bike to campus in the first place. The participants suggested that bypasses for bikes or ramps should be indicated more clearly in order to help cyclers navigate the campus more easily.
There are 'hidden paths' to circumnavigate the stairs. You get to know them when you take the bike and enjoy exploring your surroundings. But you have to find out for yourself, there are no signs indicating how stairs can be circumnavigated. (U2 FG7) The Vast Outdoor Areas Motivate to Sit and Relax, Rather Than to Move and Be Active; Cues for Physical Activity Are Lacking On both campuses, there are wide lawns between the buildings. Footpaths linking buildings are valued for their location within the green spaces, and are sometimes even used for detours. However, in University 2, students complain that despite the vast grounds of the campus, attractive walking trails are lacking. They felt that the campus area could be put to better use if special pathways encouraged students to stroll around at different sites.
I would like to have a walkway where I can walk and contemplate, or can stroll around with other people. Somehow, this University campus does not seem inviting for taking a walk. There is no footpath with special spots along the way, e.g. inspiration for meditation or something similar. . . . I often feel somewhat lost on campus, even though I have been studying here for seven years. I would love to go for a walk to clear my mind sometimes, but I don't feel comfortable doing that, because I'm always afraid of getting lost, or not being back in time. (U2 FG9) University 2 also disposes of a central plastered area built in an Amphitheater style, which proves to be a barrier to physical activity: the Amphitheater stairs are used to sit rather than to walk, and people avoid moving through or within the half-oval as they feel observed by the people sitting above them (C: Figure 3).
In University 1, on the other hand, there is a great circular path in the center of the campus ("rondel") that invites students to take walks around while talking to friends, drinking coffee, or even learning (D: Figure 4). I know some people who even take their study scripts with them and stroll around the "rondel", for example when they need to memorize something. (U1 FG1) The "rondel" offers a great opportunity to walk "laps". I do that all the time and because the length is manageable and visible, you might bring yourself to walk one more than actually intended. (U1 FG 4) Other than that, the green spaces and campus squares invite students to sit down and relax, rather than move, according to the participants (E: Figure 5).
Further barriers are uncertainties about the campuses' usage policies of the green spaces, e.g., whether (ball) games are banned.
Here's the problem: it is not clearly communicated what you are allowed to do on those vast green spaces. What is forbidden and what is permitted? Can we play soccer? (U2 FG8) The participants regret that there is no sports equipment, e.g., balls, rackets etc. available for use or hiring. Bringing this equipment to campus is considered cumbersome due to the lack of storage options. There are opportunities for exercise available, like a slackline park or a bouldering tower, but they are situated close to the sports center at both universities. Especially at University 2, the sports center is rather separated from the rest of the campus, thus the offers are partially unknown or considered too far away to use during breaks. The participants felt that fixed outdoor movement options on the main campus (rather than in the area of the sports center), e.g., slacklines, table tennis, or table soccer, would encourage many students to be more physically active during breaks. They pointed out that easy access and playful character of these options would be beneficial.
Close to the philosophic faculty building, there are trees where you could easily install some slacklines. I really like using slacklines, I really think I would do it more often if it was closer by. I would say: "Guys, let's go and do some slacklining. . . " (U2 FG7) [Possibilities for being physically active] -it is necessary for those to be known by all students. Maybe slacklines, ping pong tables. It is necessary to distribute the sports opportunities over the campus, rather than centralize everything around the sports center. (U1 FG3) It would be great if you were able to claim: "I'm not really being physically active in the sense of doing sports, but doing something fun, which in turn entails being physically active." (U1 FG3)
Sports and Activity Areas Are Part of the Campuses, but Out of the Scope of Many Students
The sports centers within the universities are highly appreciated, but several aspects prevent students from making the most of the available offers, according to the participants. One reported barrier is the physical distance, especially in University 2, where the sports center is separated from the rest of the campus by a street which can be crossed via a long bridge. The (spatial) distance is also linked to a lack of transparency about options and offers for movement that can be found in and around the sport centers, as was reported about both universities.
For me, all these sportive offers are a mystery. We don't get information about that. We simply don't know what we are allowed to use and how. I heard it was possible to rent sports equipment? (U1 FG2) I study sports, and only I knew about this (sports) event (a fun competition in dodge ball). This event is free and you can try out everything. I thought it is so sad, on campus there are so many posters etc. about parties etc. but nothing about this sport event. You can see that physical activity is not regarded important. (U2 FG6)
The Design of Stairs and Staircases Influences Their Use
When discussing pictures of the interior of the University buildings, the participants acknowledged that using the stairs is a good opportunity for physical activity. It transpired that students use stairs (instead of elevators) with different frequency in different buildings, depending on the architecture. In buildings with wide, open staircases, participants report to feel encouraged to use the stairs (F: Figure 6), or use them "automatically" without noticing. In other buildings, the staircases are hidden or less accessible, which leads to a more frequent elevator use.
In the XY building [a building with rather narrow staircases behind heavy doors], there are two elevators and I'm really the only one taking the stairs there. Sometimes people are waiting upstairs until people have ridden downstairs with the elevator and the elevator is coming up again, instead of taking the stairs! (U2 FG8) Participants also took pictures of wide spaces and hallways within the universities, which completely lack encouragement for physical activity (G: Figure 7). Participants regard this as "lost" space and would appreciate inspiration for physical activity.
Sedentary Behavior During Lectures and Learning Is Taken for Granted-Narrow Spaces Dispel Ambitions for More Movement
Within libraries, sedentary behavior is considered inevitable by the participating students. The library rooms themselves offer no space for movement, according to the participants, and students avoid leaving the library (e.g., for active breaks) because In both universities, stand-up tables are available within the library, but several barriers prevent their regular use: these tables are scarce, occupied by others most of the time, and/or often hidden in corners and thus not easily found. Moreover, stand-up tables are not adjustable to a sitting position. Students working on these tables are thus obligated to remain standing the entire time, which is exhausting for them. The participants suggest a more flexible solution.
Likewise, the participants explain that they feel obliged to stay seated during lectures. Most lectures last 1.5 h, and even if lecturers make a 5-or 10-min break in the middle of this period, the long narrow rows of chairs will render it difficult for students to leave the lecture hall for a short walk, according to the participants.
Principal Findings
The photo voice study revealed that in neither of the two universities studied, the vast potential for increasing everyday physical activity on the University campus was exploited. According to the participants, the extensive green spaces between campus buildings offer no cues for movement: neither fixed nor mobile equipment for games or activities is (easily) available, and attractive walking trails are lacking or well-hidden. Instead, the lawns and meadows are used for sitting and relaxing. It became clear that cyclers struggle to find parking and storage options for bikes and biking gear as well as ramps and pathways free of steps when moving around campus, discouraging them from taking the bike to the campus in the first place. Restricting the options for physical activity (e.g., slackline parks) to the sports centers, as is the case in both universities, was considered both a physical and a psychological barrier to utilizing these offers. The participants did not feel well informed about many aspects regarding physical activity options on the campuses, e.g., they did not know the usage policy of lawns, or were not aware of options or events offered by the sports centers. The participants made many suggestions how the outdoor areas on campus could be changed for the better.
They were less optimistic when discussing the indoor situation. They reported how they felt obliged to keep sitting during lectures and learning time, and doubted that options for physical activity could be improved in the narrow spaces of lecture halls and libraries. The architectural design of staircases within buildings was regarded as mainly responsible for using-or not using-the elevators instead of walking. On the other hand, the participants pointed out that the wide hallways, foyers and courtyards, which were present all over the two campuses, could be put to better use with regard to physical activity. It became clear in the focus groups that the Photovoice project raised the participants' awareness of how the context influences their movement behavior, and helped them come up with solutions to make physical activity easier for students of their respective universities.
Strengths and Weaknesses
Generally, the study design proved adequate for the focus of the study. The pictures helped the researchers get a more thorough insight into students' lives, and also helped the participants remember all the topics with regard to physical activity during the group discussions, as they served as a reminder. On the other hand, discussing photographed scenes and situations also prompted conversations about general topics and problems that were not linked to physical activity, e.g., complaining about construction sites on campus or malfunctioning doors; this required a clear focus and rigor of the facilitator.
We succeeded in recruiting students from a range of study programs, and could thereby capture heterogeneous views on physical activity on the campus. Nevertheless, it is not representative for the total number of students enrolled in those study courses. There was a slight predominance of students from sport related subjects (U1) and psychology (U2), respectively. Moreover, there were slightly more female participants. As participants were not only asked to take photos, but also to participate in a focus group discussion, outgoing and talkative people might have been attracted more; reserved and less talkative individuals may have been discouraged from participating.
A bias of the results may arise from the fact that the Photovoice project was conducted in summertime. This may have increased the participants' awareness of and focus on outdoor areas and outdoor activities, at the cost of the indoor context. For a more thorough understanding of the campus environment and its influence on physical activity, it may be advisable to perform additional data collections in winter. Still, students spend a lot of time indoors also in summer, e.g., during lectures, while learning in the library, or eating in the canteen; therefore, there was a substantial number of pictures taken inside the buildings that served as an adequate basis for discussions in the focus groups.
The study was conducted in two campus universities. The findings pertain to specific, unique built environments and cannot simply be transferred to other (campus) universities. It is interesting, however, that there were many similarities in both two locations, e.g., as to the usage of green spaces and hallways, the role of the sports center, sedentary behavior during lectures, and barriers for cyclers. Therefore, the results may serve as a catalog of potential contextual factors which may be worth considering when planning interventions for physical activity, as green spaces, sport centers, libraries and staircases can be found in any University.
Comparison With Other Studies
There are only a few studies using a Photovoice approach to explore the setting of a University campus with regard to physical activity. Joy et al. used photo elicitation with volunteer University members (students n = 11, employees n = 14) to identify healthy eating and active lifestyle barriers and supports. Similar to our findings, participants reported that on campus, physical activity was principally possible, but dysfunctional (e.g., overgrown) walkways were regarded as barriers (18).
Deliens et al. focused on determinants of physical activity and sedentary behavior in Belgian University students using semi-structured focus groups. Unlike our findings, lack of time was named as the most important barrier to physical activity. Besides that, findings mainly corresponded with ours, e.g., lack of information on sports offers, or access to sports gear (27). This study has not used a Photovoice approach to contribute to the focus group discussion. Therefore, participants may have neglected factors connected to architectural design or layout of the University; in our study, students claimed that taking the pictures has raised the awareness of those environmental barriers/ facilitators.
In a study employing Photovoice sessions with female Hispanic adolescents, the participants identified some barriers to physical activity in the built environment of the community; these referred mainly to unaesthetic features (e.g., dirt, graffiti, vandalism), perceived lack of safety, or poor public transport to sports facilities (28). These aspects were not brought up by the participants of our study, hinting at the privileged status of the selected University campuses as compared to some (socially disadvantaged) neighborhoods.
Implications for Policy and Practice
The Photovoice study helped identify starting points for environmental interventions that could foster physical activity among campus students. Creating and/or signposting attractive walks (including information on length of footpaths and time needed to walk along these paths) may "nudge" students to stroll over campus for talking with each other and/or winding down in breaks, rather than sitting. In addition, students may be more inclined to cycle to classes when campuses offer better options for moving around by bike, and for storing bikes and gear. Indeed, many studies have shown that creating new infrastructure for walking and cycling, including bike parking opportunities, were related to increased physical activity (29), although these data mainly refer to community and city designs. Few studies have focused on University campuses. Horacek et al. (30) analyzed the infrastructure of 13 US-American college campuses and showed that walkability and bikeability of the campus (i.e., the safety, quality and comfort of paths) was related to college students' physical activity. A study in a Hongkong University demonstrated that increasing and repairing the pedestrian networks improved the students' walking behavior (31).
In both studied universities, the outdoor areas dispose of vast green spaces that were currently mainly used for sitting and relaxing; cues for movement were reported to be missing. The Photovoice participants emphasized that easily accessible activity areas, as well as a sports equipment sharing program, could encourage students to get physically active in breaks. According to a number of studies in schools, improving playground structures and designs as well as providing game equipment could increase the moderate or moderate-vigorous physical activity in children during recess, although the results were mixed (32). It is not clear if these results can be transferred to the University setting and University students, respectively; experimental studies on the effect of gaming equipment and exercise facilities in universities are lacking. It is interesting that the activity options that the participants suggested included gaming and "fun" sports (i.e., slacklines, table soccer, or bouldering sites) rather than "classical" fitness or team sports (e.g., outdoor workout equipment, soccer, basketball). This aspect would be worth exploring further before deciding on offers and changes in infrastructures, for example by performing a survey among University students on their preferred activity and exercise types. A survey among Emirati University students revealed that indeed the vast majority preferred activities "with a fun element, " although the activities in question were not restricted to the campus setting (33). Probably, nonathletic and athletic students may also differ with regard to their activity preferences.
Whereas, breaks and recreation between classes and learning phases provide the main opportunities for University-based physical activity, interventions may also target the sedentary behavior prevalent in lecture halls and libraries. The participants stated they felt uneasy leaving their textbooks, notes and laptops behind. Therefore, on-site exercise measures seem appropriate, for example sit-stand desks, or activity breaks in classrooms. In a review on standings desks in school classrooms, Minges et al. (34) show that those desks reduce sitting time, whereas the results with regard to physical activity are mixed. Reducing sedentary time may be even more significant in the University sector, though, where students sit even longer than in school. Activity breaks can reduce a prolonged sitting time and increase physical activity, as several studies among school children have shown (35)(36)(37)(38). To our knowledge there are no studies focusing on University students. Therefore, we do not know if activity breaks can be implemented in classes and lectures of universities; reluctance of lecturers to offer exercises in breaks, and/or reluctance of students to participate may be barriers.
The Photovoice study could highlight a range of environmental factors-indoors and outdoors-which influence sedentary behavior and physical activity of University students. This catalog may serve as a base for aspects to be considered when analyzing the campus setting before implementing measures. Systematic approaches to gather information pertaining to the campus environment are scarce. Horacek et al. (39) developed a tool to assess the built environment of college campuses; its components include bike racks, stairwell direction prompts, exercise spaces, available equipment, and quality and safety of paths. According to the results of our study, such an audit should be supplemented with factors relating to indoor situations in buildings and libraries, storage facilities, as well as more general aspects pertaining to the layout of campus grounds (stairs, rondels etc.) (39). Murphy et al. (40) published a study protocol indicating that there is work underway to create a comprehensive audit tool for examining the environment, provision, and support offered by Irish universities for students' participation in physical activity. This may give further hints as to environmental factors relating to physical activity of students.
CONCLUSION
The Photovoice method proved to be an adequate approach for thoroughly analyzing the environmental factors that influence physical activity and sedentary behavior, seen through the lens of those individuals spending a lot of time in this environment. It helped identify a wide range of starting points for health-related interventions. Future research needs to focus on questions around how students can be actively involved in planning and advocating for (structural) interventions that change campuses to be healthier places which support movement and active lifestyles. Analyzing the effectiveness of such interventions will also warrant further research.
DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethics Committee of the University of Regensburg (18-896-101). The patients/participants provided their written informed consent to participate in this study.
AUTHOR CONTRIBUTIONS
JvS conducted the focus groups, coded and analyzed the material, and was a major contributor in writing the manuscript. JL set up the study design, commented on the interview guide, and was a major contributor in writing this manuscript. JR set up the study design, developed the interview guide, and conducted focus groups. JC conducted focus groups in one of the universities. JH and ST helped to set up and conduct the focus groups in one of the universities. All authors contributed to the article and approved the submitted version.
FUNDING
The project was funded by the public health insurance company TK (Techniker Krankenkassen) and supported by the KErn (Competence Center for Nutrition). | 2020-10-06T13:17:48.139Z | 2020-10-05T00:00:00.000 | {
"year": 2020,
"sha1": "196f89380d86996f27c6871c1197f102005e2223",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fpubh.2020.561175/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "196f89380d86996f27c6871c1197f102005e2223",
"s2fieldsofstudy": [
"Education"
],
"extfieldsofstudy": [
"Medicine",
"Psychology"
]
} |
261098692 | pes2o/s2orc | v3-fos-license | A comprehensive evaluation in clinic and physiologically‐based pharmacokinetic modeling and simulation to confirm lack of cytochrome P450–mediated drug–drug interaction potential for pomotrelvir
Abstract Pomotrelvir is a new chemical entity and potent direct‐acting antiviral inhibitor of the main protease of coronaviruses. Here the cytochrome P450 (CYP)–mediated drug–drug interaction (DDI) potential of pomotrelvir was evaluated for major CYP isoforms, starting with in vitro assays followed by the basic static model assessment. The identified CYP3A4‐mediated potential DDIs were evaluated clinically at a supratherapeutic dose of 1050 mg twice daily (b.i.d.) of pomotrelvir, including pomotrelvir coadministration with ritonavir (strong inhibitor of CYP3A4) or midazolam (sensitive substrate of CYP3A4). Furthermore, a physiologically‐based pharmacokinetic (PBPK) model was developed within the Simcyp Population‐based Simulator using in vitro and in vivo information and validated with available human pharmacokinetic (PK) data. The PBPK model was simulated to assess the DDI potential for CYP isoforms that pomotrelvir has shown a weak to moderate DDI in vitro and for CYP3A4 at the therapeutic dose of 700 mg b.i.d. To support the use of pomotrelvir in women of childbearing potential, the impact of pomotrelvir on the exposure of the representative oral hormonal contraceptive drugs ethinyl estradiol and levonorgestrel was assessed using the PBPK model. The overall assessment suggested weak inhibition of pomotrelvir on CYP3A4 and minimal impact of a strong CYP3A4 inducer or inhibitor on pomotrelvir PK. Therefore, pomotrelvir is not anticipated to have clinically meaningful DDIs at the clinical dose. These comprehensive in vitro, in clinic, and in silico efforts indicate that the DDI potential of pomotrelvir is minimal, so excluding patients on concomitant medicines in clinical studies would not be required.
INTRODUCTION
Although the worst outcomes from the ongoing coronavirus disease 2019 (COVID-19) endemic have been mitigated by vaccines, the variable uptake of vaccines and the continued emergence of new and highly transmissible variants, for which currently available vaccines may be less effective, have led to the need for direct-acting antiviral treatments effective against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections. 1 Although new therapeutics for COVID-19 have improved patient outcomes, additional treatment options are needed, especially in patients who take medications for comorbid conditions. 2,3Concomitant use of the most commonly prescribed oral treatment within the United States, ritonavir-boosted nirmatrelvir (PAXLOVID™), is contraindicated with a number of commonly prescribed drugs that are highly dependent on cytochrome P450 (CYP) 3A for clearance or are potent CYP3A inducers. 4,5omotrelvir (PBI-0451) is an orally bioavailable, directacting, antiviral inhibitor of the main protease of coronaviruses and is being developed as a stand-alone agent for the treatment of COVID-19. 6A phase I, first-in-human (FIH) (Clini calTr ials.gov:NCT05011812) study showed that pomotrelvir is well tolerated, with good oral bioavailability and dose-linear, single-and multiple-dose pharmacokinetic (PK) exposures over a >20fold dose range when administered with food. 7n the past two decades, physiologically-based PK (PBPK) modeling has become an accepted approach to inform drug-drug interaction (DDI) risk and reduce DDI studies.The use of PBPK modeling is encouraged by the regulatory authorities and the framework to include PBPK in the new drug application submission is documented in guidance from the US Food and Drug Administration (FDA). 8Herein, a comprehensive evaluation of the CYP enzyme-mediated DDI potential of pomotrelvir in in vitro and clinical studies along with PBPK modeling is presented.
In vitro assays
The DDI potential of pomotrelvir was evaluated in standard in vitro experiments including CYP enzyme
WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?
The most prescribed oral treatment for coronavirus disease 2019 (COVID-19) in the United States, ritonavir-boosted nirmatrelvir, is contraindicated for coadministration with a number of commonly prescribed drugs because of the potential for drug-drug interactions (DDIs).Pomotrelvir is a new chemical entity being developed for treatment of COVID-19.
WHAT QUESTION DID THIS STUDY ADDRESS?
The cytochrome P450 (CYP) enzyme-mediated DDI potential was evaluated in in vitro and clinical studies.These results, along with human pharmacokinetic data from the first-in-human study, were used to develop a physiologically-based pharmacokinetic (PBPK) model to simulate the DDI potential for pomotrelvir as a perpetrator for various CYP enzyme activities and as a victim of CYP3A4.
WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?
Pomotrelvir's lack of potential for clinically meaningful CYP enzyme-mediated DDIs in humans was predicted early in its development.These results supported the decision to remove typical restrictions on the use of concomitant medications from the phase II trial's inclusion/exclusion criteria.
HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?
The favorable DDI profile suggests a potential advantage of pomotrelvir over commonly prescribed COVID-19 oral treatments.The strategy of applying PBPK modeling and simulation allowed for early DDI potential prediction and avoided the unnecessary exclusion of patients in clinical trials and potentially additional clinical DDI trials, thereby reducing unnecessary exposure to DDI study drugs in healthy volunteers.phenotyping, inhibition, and induction studies, followed by the assessment using the basic static model as described in the FDA's Guidance. 9(Detailed methods are in Appendix S1.)
Clinical DDI study with midazolam and ritonavir
Three DDI cohorts were included in the FIH study (Clini calTr ials.gov:NCT05011812) to evaluate the impact of multiple doses of pomotrelvir on the PK of midazolam (a sensitive probe substrate for CYP3A4 10 ) and the impact of coadministering ritonavir (a potent, mechanism-based inhibitor of CYP3A4 11 ) on the PK of pomotrelvir.Informed consent was obtained from all individual participants included in the study before any screening procedures.The study protocol was reviewed and approved by the institutional review board at the study site (Auckland City Hospital, Auckland, New Zealand).
Participants
Eight healthy adults (with an approximately even distribution between males and nonpregnant, nonlactating females) aged between 18 and 59 years with body mass indexes ≥19.0 and ≤30.0 kg/m 2 were enrolled in each of the DDI cohorts.Participants were in good general health as determined by the investigator at a screening evaluation performed no more than 28 days before the scheduled first dose.Medications and food contraindicated with midazolam or ritonavir were prohibited for the participants in these cohorts.
Study design for midazolam DDI cohort
This was a crossover design with 1 mg midazolam oral solution (1 mg/mL) administered alone on Day 1 and coadministered with a supratherapeutic dose of pomotrelvir at 1050 mg (three 350 mg tablets, twice daily [b.i.d.] from Days 2 through 11) on Days 6 and 11 under fed conditions.All treatments were administered with 240 mL of water following an overnight fast and immediately after fully completing a standard meal.Blood samples for PK profiles (18 timepoints, up to 24 h postdose) were collected on Days 1, 6, and 11 for midazolam and Days 6 and 11 for pomotrelvir, also at the time corresponding to the morning predose on Days 4 and 8 for pomotrelvir.Plasma concentrations of midazolam and 1-hydroxy-midazolam (1-OH-midazolam) were determined using a validated liquid chromatography with tandem mass spectrometry (LC-MS/MS) method with a quantitation dynamic range from 0.1 to 100 ng/mL.
Study design for ritonavir DDI cohorts
Two cohorts with 100 mg ritonavir once daily oral coadministration with pomotrelvir suspension at a single dose of 20 mg or multiple doses of 50 mg (once daily for 10 days).The PK of pomotrelvir from these two DDI cohorts was compared with that from the single dose of 100 mg pomotrelvir suspension alone.The treatment procedure was the same as that for the midazolam DDI cohort.Blood samples for PK profiles (18 timepoints, up to 24 h postdose) were collected on Days 1, 5, and 10 and at the time corresponding to the morning predose on Days 4 and 8, if applicable, for pomotrelvir and ritonavir.Plasma concentration of pomotrelvir was determined using a validated LC-MS/ MS method with a quantitation dynamic range from 10 to 10,000 ng/mL.
PK and statistical analysis
All PK and statistical analyses were conducted by using SAS Version 9.4 or higher (SAS Institute) or Phoenix Win-Nonlin Version 8.3 or higher (Certara).The PK analyses of plasma midazolam, 1-OH-midazolam, and pomotrelvir were performed by a noncompartmental method, and the PK parameters, including maximum observed concentration (C max ), area under the curve [AUC] from time 0 to the last quantifiable concentration (AUC last ), and AUC from time 0 to infinity (AUC inf ), were calculated.The effect of pomotrelvir on midazolam PK was assessed by comparing midazolam and 1-OH-midazolam plasma exposure (C max and AUC) on Day 1 versus 6 and Day 1 versus 11.To assess the effect of ritonavir on pomotrelvir PK, the dosenormalized plasma exposure of pomotrelvir (C max /dose and AUC/dose) was compared between the treatments of pomotrelvir alone (100 mg, single dose) and coadministered with ritonavir (20 mg pomotrelvir + 100 mg ritonavir, single dose; 50 mg pomotrelvir + 100 mg ritonavir, once daily for 10 days).Treatment difference was expressed using geometric mean ratios (GMRs) and their 90% confidence intervals (CIs).
Safety and tolerability assessment
Safety and tolerability were assessed throughout the study period by monitoring and recording adverse events, physical examination findings, clinical laboratory tests, vital signs, urine drug and alcohol assessments, calculated creatinine clearance, serum/urine pregnancy tests (females of childbearing potential only), and 12-lead electrocardiogram results.SARS-CoV-2, hepatitis B virus, hepatitis C virus, and HIV-1 testing were performed during screening.
PBPK
A PBPK model for pomotrelvir based on in vitro and in vivo information on the metabolism and PK of pomotrelvir was constructed in the Simcyp Simulator (V21 Release 1) following the workflow shown in Figure S1.The model was developed to simulate plasma concentration-time profiles of pomotrelvir following a single dose and repeat dosing in healthy subjects and then applied to evaluate the likely impact of administration of strong (itraconazole), moderate (fluconazole and erythromycin), and weak (cimetidine) CYP3A4 inhibitors and strong (rifampicin) and moderate (efavirenz) CYP3A4 inducers on the PK of pomotrelvir and the potential for CYP1A2-, CYP2B6-, CYP2C8-, CYP2C9-, CYP2C19-, and CYP3A4-mediated DDIs with pomotrelvir as a perpetrator (inhibitor or inducer for the CYP enzyme activity).
Distribution
The PBPK model for pomotrelvir was developed using a full-body PBPK model.Tissue-to-plasma partition coefficient values were predicted using the methodology described by Poulin and Theil 12 and updated by Berezhkovskiy 13 using LogP or LogD (lipid/water partitioning coefficient) and protein binding as input.
Absorption
Absorption was described by a mechanistic absorption model (Simcyp Advanced Dissolution, Absorption and Metabolism [ADAM] model).In human colon adenocarcinoma clone 2 (Caco-2) and Madin Darby canine kidney-breast cancer resistance protein (BCRP) cell lines, pomotrelvir exhibits poor apparent permeability with measurements of 0.36 × 10 −6 cm/s and 1.01 × 10 −6 cm/s, respectively, and a large efflux ratio.These in vitro measurements did not adequately predict absorption.Pomotrelvir clinical PK appeared dose proportional over the doses studied when administered with food.In vivo, ritonavir (a potent P-glycoprotein and BCRP efflux transporter inhibitor), when coadministered at 100 mg once a day with pomotrelvir, did not impact the absorption of pomotrelvir.Thus, the mechanistic effective permeability model was used to predict regional permeability from LogP and molecular weight without considering efflux transporters.
Metabolism
In vivo, pomotrelvir is subject to nonhepatic and likely presystemic hydrolysis, which was thought to contribute ~50% of the apparent clearance of pomotrelvir following oral dosing.As no in vitro data were available to describe this elimination process, presystemic hydrolysis was taken into account by assuming incomplete absorption as was predicted using the ADAM model (predicted absorption fraction at 700 mg is 0.65).
In vitro, pomotrelvir was mainly metabolized by CYP3A4 with CYP2C8, contributing to a minor extent (fm CYP3A4 = 88%, fm CYP2C8 = 11% [fm, fraction metabolism]) (Appendix S1).A clinical study was performed in combination with ritonavir.The dose-normalized effect of ritonavir on the systemic exposure (AUC) of pomotrelvir was minimal, indicating the contribution of CYP3A4 to the metabolism of pomotrelvir was less than 88%.Using a mean apparent oral clearance value of 33.19 L/h derived from clinical studies where pomotrelvir was investigated after an oral dose of 300 mg administered with food (FIH study) and a fm CYP3A4 value of 25% to match the ritonavir DDI study, the retrograde model (extrapolation from in vivo data) was used to estimate an unbound intrinsic clearance (Cl int,u ) to match the observed data.The Cl int,u value was apportioned to CYP3A4 and non-CYP3A4 to obtain estimates of 0.376 μL/min/pmol CYP3A4 and 137.91 μL/min/mg protein (additional undefined Cl int,u ).The additional undefined value represents the hydrolysis and any CYP2C8 metabolism.Renal clearance (CLr) was negligible in preclinical species so was assumed negligible (Cl r = 0) in the model.
Changes in enzyme kinetics attributed to enzyme inhibition and induction
Following incubation with human liver microsome (HLM), pomotrelvir exhibited weak to moderate inhibition of CYP1A2, CYP2C8, CYP2C9, CYP2C19, CYP2D6, and CYP3A4 (Appendix S1).The concentration to reduce enzyme activity to 50% of the uninhibited values were adjusted to an inhibition constant (K i ; concentration required to produce half maximum inhibition) value using the Cheng-Prusoff equation and incorporated into the model.Pomotrelvir is a mechanism-based inhibitor and inducer of CYP3A4, therefore autoinhibition and autoinduction were integrated in the PBPK model.It should be noted that mechanism-based inhibition (MBI) is typically overpredicted and that in vitro appearance K i (μM) values often have to be optimized to recover observed DDIs.The fumic (fraction unbound in HLM) value for CYP3A4 MBI parameters was optimized to correctly predict the observed PK after multiple doses of pomotrelvir and to account for the observed impact of the coadministration of pomotrelvir on the exposure of midazolam.
In in vitro induction experiments in human hepatocytes at concentrations from 0.3 to 30 μM, pomotrelvir increased CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19, and CYP3A4 mRNA by at least twofold (Appendix S1).Given that no clear dose response was observed in the CYP1A2, CYP2C8, CYP2C9, and CYP2C19 in vitro induction studies, it was not possible to calculate a concentration with the maximum induction effect and a concentration with 50% of the maximum induction effect for these CYPs with a high level of confidence.Thus, only induction parameters for CYP2B6 and CYP3A4 were included in the model.
All final parameters used for the simulation of pomotrelvir kinetics are listed in Table 1.
Population characteristics
The healthy volunteer (HV) population available within the Simcyp Simulator (V21 Release 1) was used in this analysis.Except for demographic data collected in the clinic (FIH study), all other parameter values for the HV population are the same as those used for the Caucasian population.Default Simcyp parameter values for creating a virtual North European Caucasian population (physiological parameters including liver volume, blood flows, and enzyme abundances) have been described previously. 14
Simulations
For the model development and verification, 10 simulated trials of virtual subjects with characteristics that matched (according to the number, age range, and proportion males and females) with those of the subjects used in each clinical study were generated.
To evaluate the likely impact of administration of CYP3A4 inhibitors and inducers, 10 virtual trials of 10 healthy subjects (50% female) aged 20-50 years receiving a single oral dose of 700 mg pomotrelvir in the absence of perpetrator and on the seventh day of 16 days of oral dosing of perpetrator were generated.
To assess the CYP1A2-, CYP2B6-, CYP2C8-, CYP2C9-, CYP2C19-, and CYP3A4-mediated DDIs, victim (impacted by inhibitors or inducers for the CYP enzymes) drugs were administered as a single dose with and without coadministration of pomotrelvir.All simulations were performed under steady-state conditions for pomotrelvir.A total of 10 virtual trials of 10 healthy subjects (50% female) aged 20-50 years receiving a single oral dose of 150 mg substrate in the absence of pomotrelvir and on the seventh day of 11 days of dosing of pomotrelvir (700 mg b.i.d.) were generated.
In a recent publication, Kilford et al. 15 recommended a sensitivity analysis using an appropriate range of K i values when no significant DDIs are predicted to assess the worst-case scenario for DDI risk.
Additional DDI simulations were therefore performed after reducing the unbound K i values by 10-fold.The objective of this sensitivity analysis was to account for the worst-case scenario of potentially inaccurate enzyme K i estimates from in vitro experiments and other uncertainties.
The impact of pomotrelvir on the exposure of the representative oral hormonal contraceptive drugs ethinyl estradiol and levonorgestrel was evaluated in two separate simulations.In each of the two simulations, 10 virtual trials of 10 healthy subjects (100% female) aged 18 to 45 years receiving a single oral dose of 0.035 mg ethinyl estradiol, or 0.27 mg levenorgestrel, in the absence of pomotrelvir and on the seventh day of 11 days of dosing of pomotrelvir (700 mg b.i.d.) were generated.
In vitro assays
In vitro data indicate that pomotrelvir is a substrate for human CYP3A4 and CYP2C8, which are responsible for 88% and 11% of the hepatic metabolism of pomotrelvir, respectively.No inhibitory effect on CYP2B6, CYP2C19, or CYP2D6 was observed for pomotrelvir in vitro, whereas a weak to moderate inhibition for CYP1A2, CYP2C9, and CYP2C8 and time-dependent inhibition (TDI) for CYP3A4 was shown.In addition, pomotrelvir has shown an induction effect for CYP2B6 and CYP3A4, but has no clear concentration-dependent induction effect on CYP1A2, CYP2C8, CYP2C9, or CYP2C19 in vitro (detailed results are in Appendix S1).With the weak to moderate potential inhibition or induction effect observed for pomotrelvir in vitro and the results from the basic static model analysis, the DDI potential mediated by CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2C19, and CYP3A4 enzymes were further evaluated in PBPK modeling and simulation and in clinic for CYP3A4 because of the potential complex DDI effect that could involve TDI and induction for pomotrelvir as a perpetrator as well as a victim.
Clinical midazolam study
The mean plasma concentration-time profile of midazolam and its metabolite 1-OH-midazolam are plotted in Figure 1.All predose concentrations of midazolam were below the lower limit of quantitation on all occasions, demonstrating that the wash-out period was adequate.After a single dose of midazolam, mean plasma concentrations of midazolam were slightly higher, and the concentration of 1-OH-midazolam was lower, following coadministration of pomotrelvir on Days 6 and 10 than administered alone on Day 1.The plasma exposures of midazolam and 1-OH-midazolam were compared between that after coadministration with a supratherapeutic dose at 1050 mg b.i.d.pomotrelvir versus administration alone.The GMR with 90% CI for AUC last , AUC inf , and C max (Table 2) showed that the systematic exposure of midazolam increased, and 1-OH-midazolam decreased, by <50% after coadministration with pomotrelvir compared with midazolam alone.
Clinical ritonavir study
The pomotrelvir FIH study included two DDI cohorts to evaluate the potential impact of ritonavir, an inhibitor of CYP3A4, on pomotrelvir PK.In one of the cohorts, pomotrelvir 20 mg was coadministered with 100 mg ritonavir as a single dose, whereas in the other cohort, pomotrelvir 50 mg was coadministered with ritonavir 100 mg once daily for 10 days.A comparison of pomotrelvir plasma exposure following coadministration of 100 mg ritonavir with that after pomotrelvir administered alone at a 100 mg single dose (the evaluated pomotrelvir dose level that was the closest to that tested in the DDI cohorts in the FIH study) was performed, and the GMR with 90% CI for AUC last /dose, AUC inf /dose, and C max /dose of pomotrelvir coadministered with ritonavir versus pomotrelvir alone was estimated (Table 3).The data demonstrated that the dose-normalized systemic exposure of pomotrelvir increased modestly (0.92-1.54-fold) in the presence of ritonavir.
Model development and verification
Simulated PK profiles, AUC inf , AUC over the dosing interval, and C max of pomotrelvir following single and repeat oral doses of pomotrelvir (100 mg to 2100 mg) to healthy subjects were in reasonable agreement with the observed data (all within twofold, the majority within 1.5-fold) from the FIH clinical study (Table S3, Figure 2).The simulated AUC inf and C max following the coadministration of a single dose of 100 mg ritonavir with 20 mg pomotrelvir, or following repeated daily coadministration of 100 mg ritonavir with 50 mg pomotrelvir, were in reasonable agreement with the observed data (all within twofold difference) given the observed clinical data variability (Table S4).The model was also able to accurately predict the effect of pomotrelvir (1050 mg b.i.d.) on the exposure of midazolam (Day 6 and Day 11) to be within 1.3-fold of the observed data (data not shown).
Pomotrelvir as perpetrator
In this PBPK analysis, no clinically significant DDIs were predicted following the administration of multiple doses of the proposed pomotrelvir daily dose of 700 mg b.i.d. with the sensitive probe substrates for CYP1A2, CYP2B6, CYP2C8, CYP2C9, and CYP2C19.The sensitivity analysis with 10-fold reduction of K i for these CYP enzymes predicted a weak inhibition effect on CYP2C8 only with the simulated geometric mean AUC inf and C max ratio (with vs without pomotrelvir) at 1.55 and 1.36, respectively.Moreover, consistent with the clinical DDI data, a weak inhibitory effect was predicted in silico with the sensitive CYP3A4 substrate midazolam (Table 4).
No clinically significant DDIs were predicted following administration of multiple doses of 700 mg b.i.d.PBI-0451 with ethinyl estradiol and levonorgestrel.Specifically, the simulated AUC inf and C max values, with and without pomotrelvir, were <1.20fold under simulated steady-state conditions for these oral contraceptives (Table 4).
Pomotrelvir as victim
This analysis indicated that the simulated changes in plasma exposure of pomotrelvir during administration with strong inhibitors of CYP3A4 (itraconazole) to healthy subjects are weak.No clinically significant DDIs were predicted with moderate and weak CYP3A4 inhibitors.A weak impact on the exposure of pomotrelvir was predicted in the presence of strong or moderate CYP3A4 inducers (Table 5).
DISCUSSION
The CYP enzyme-mediated DDI potential for pomotrelvir was evaluated in in vitro studies, where the CYP isoforms that are responsible for hepatic metabolism of pomotrelvir were identified.In addition, the inhibitory and inducing effects of pomotrelvir on major CYP enzymes were evaluated.The inhibitory effect observed in vitro was interpreted using a basic model, which was anticipated to overestimate the effect because the model was based on the maximum clinical pomotrelvir exposure (C max ).Furthermore, a dynamic mechanism-based PBPK model was successfully developed with a validation showing similar PK results between the simulated data and that observed in clinic and applied to predict the clinical DDI potential that could be caused by pomotrelvir being a perpetrator or a victim.In addition, the potential DDI of pomotrelvir through the CYP3A4 pathway was evaluated in the clinic by coadministration with midazolam and ritonavir.
Pomotrelvir DDI potential as a victim
In vitro data indicated that pomotrelvir was a substrate for human CYP3A4 (major, fm 88%) and 2C8 (minor, fm 11%), whereas the contribution of the CYP isoforms 1A2, 2A6, 2B6, 2C9, 2C19, 2D6, 2E1, and 3A5 to pomotrelvir metabolism was negligible.Despite pomotrelvir being turned over in vitro by CYP3A4, clinically upon single or multiple doses of coadministration with ritonavir, pomotrelvir had a minimal increase (<50%) in exposure, suggesting that pomotrelvir is not a sensitive substrate for CYP3A4 in humans that would be associated with a clinically significant DDI with inhibitors or inducers of this pathway.Based on the available knowledge on the clearance mechanism and the role of CYP3A4 and CYP2C8, the clinical DDI liability attributed to CYP2C8 is not expected to be significant.Overall, the results suggest that pomotrelvir is not a sensitive substrate of CYP3A4 or CYP2C8 in humans and is not expected to have substantial changes in exposure when
Pomotrelvir DDI potential as a perpetrator
Pomotrelvir showed the potential to reversibly and timedependently inhibit as well as induce CYP3A4 in vitro, whereas midazolam had a minimal (<50%) increase in exposure in humans when codosed with pomotrelvir at multiday doses of 1050 mg b.i.d., thus suggesting a potential weak inhibitory net effect of pomotrelvir for CYP3A4.
Based on in vitro and PBPK modeling and simulation assessment, no clinically significant DDIs were predicted following the administration of multiple doses of the proposed daily pomotrelvir dose of 700 mg b.i.d. with sensitive probe substrates for CYP1A2, CYP2B6, CYP2C8, CYP2C9, or CYP2C19.In addition, no clinically significant DDIs were predicted with the oral hormonal contraceptive drugs ethinyl estradiol and levonorgestrel.The evaluation of the two main metabolites, PBI-0451A and S4-Q1, did not result in reversible inhibition for the activity of the tested CYP enzymes in vitro at clinically relevant concentrations (Appendix S1).Overall, these results suggest that pomotrelvir at the proposed clinical dose of 700 mg b.i.d. is not anticipated to have clinically meaningful DDI potential with the substrates of major CYP enzymes.
Study limitation
The PBPK model was developed at the early phase of drug development, when the absorption, distribution, metabolism, and excretion of the compound in human was not yet fully understood, therefore there were a number of uncertainties with the current model.The mechanisms underlying the elimination of pomotrelvir still need to be clarified.It was believed that presystemic hydrolysis played a significant role in the clearance of pomotrelvir, but the relative contribution of this pathway was unclear.The calculation of CYP3A4 fm was influenced by the weak DDI observed with ritonavir, but hampered by the variability in the clinical data and the noncrossover study design.In addition, there was some indication of induction in human hepatocytes with CYP1A2 and the CYP2C enzymes, which could counteract any inhibition observed with these enzymes; however, quantifiable data were not available to add these mechanisms into the model.Despite the limitations during the modeling, the recent human mass balance study using radiolabeled pomotrelvir 16 suggested that nonhepatic hydrolysis of pomotrelvir is the major route of metabolism and clearance for pomotrelvir.Minimal urinary excretion of intact pomotrelvir and a low percentage of phase I oxidative metabolites observed in excreta support the settings of negligible urinary excretion and low CYP3A4 fm in the PBPK model.Based on the comprehensive in vitro, in vivo (in clinic), and in silico (PBPK) evaluations, pomotrelvir is anticipated to have no clinically meaningful CYP enzyme-mediated DDI potential at a clinical dose of 700 mg b.i.d.This favorable DDI profile supported the recommendation to not exclude any patients as a result of CYP-mediated DDI concerns in pomotrelvir phase II clinical trials.Given the role of non-CYP-mediated mechanisms in the clearance of pomotrelvir, additional clinical DDI studies may not be necessary.The DDI assessment approach of including small DDI cohorts in the FIH study coupled with PBPK modeling and simulation provides an early read on DDI potential in early drug development.This approach supported removing restrictions for use of concomitant medications in clinical trials and also reduced the need for additional DDI clinical studies, thereby reducing unnecessary exposure to DDI study drugs in HVs.
F I G U R E 1
Mean (standard deviation) plasma concentrations of (a) midazolam and (b) 1-hydroxy-midazolam versus time when administered alone (solid circle) or coadministered with pomotrelvir (Day 6 with open triangle, Day 11 with open square).
T A B L E 3
Comparisons of POMO dose-normalized plasma exposure following coadministration with 100 mg RTV versus POMO alone.
F I G U R E 2
Linear (a) and log-linear (b) simulated and observed plasma concentration-time profiles of the first oral dose (Day 1) of 150 mg pomotrelvir every day (QD) and 225, 700, and 1050 mg pomotrelvir twice daily (BID) for 10 days in healthy subjects.Depicted are simulated (lines) and observed data (circles, n = 8; first-in-human study).The gray lines represent the mean values of simulated individual trials, the dashed lines represent the 5th and 95th percentiles, and the solid black line represents the mean data for the simulated population (n = 80).
92 (0.67, 1.26) 1.54 (1.13, 2.1)1.34(0.96,1.87) a 1.37 (1.05, 1.79) b : AUC inf , area under the curve from time 0 to infinity; AUC last , area under the curve from time 0 to the last quantifiable concentration; CI, confidence interval; C max , maximum observed concentration; GMR, geometric mean ratio; PK, pharmacokinetics; POMO, pomotrelvir; RTV, ritonavir; SD, single dose.GMR was for [AUC last /Dose of (50 mg PBI + 100 mg RTV, Day5)]/[AUC inf /Dose (100 mg POMO, SD)].b GMR was for [AUC last /Dose of (50 mg PBI + 100 mg RTV, Day10)]/[AUC inf /Dose (100 mg POMO, SD)]. Abbreviationsa Summary of simulated geometric mean AUC inf and C max ratios for sensitive CYP probe substrates and representative oral contraceptive drugs in the absence and presence of repeat oral dosing of 700 mg pomotrelvir twice daily in healthy subjects (physiologically-based pharmacokinetic evaluation).Summary of simulated geometric mean AUC inf and C max ratios for pomotrelvir in the absence and presence of CYP3A4 inhibitors and inducers in healthy subjects following single oral dosing of 700 mg pomotrelvir (physiologically-based pharmacokinetic evaluation).
T A B L E 4Abbreviations: AUC inf , area under the curve from time 0 to infinity; AUC t , area under the curve over the dosing interval (time = 24 h); C max , maximum observed concentration; CYP, cytochrome P450; GMR, geometric mean ratio; K i,u , unbound enzyme competitive inhibition constant.aRatio was calculated using AUC t (AUC t , area under the curve over the dosing interval [time = 24 h]).T A B L E 5 | 2023-08-25T06:17:24.734Z | 2023-08-23T00:00:00.000 | {
"year": 2023,
"sha1": "4d817c8b7047453c4e69dd42399b26b932bb8966",
"oa_license": "CCBY",
"oa_url": "https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/psp4.13034",
"oa_status": "GOLD",
"pdf_src": "Wiley",
"pdf_hash": "40a3d568ac91eeff73a00f09f2c5fa058cf08a1c",
"s2fieldsofstudy": [
"Medicine",
"Biology",
"Chemistry"
],
"extfieldsofstudy": [
"Medicine"
]
} |
252764442 | pes2o/s2orc | v3-fos-license | Addictive behavior and incident gallstone disease: A dose–response meta-analysis and Mendelian randomization study
Background Previous studies have suggested associations between addictive behavior and gallstone disease (GSD) risk, yet conflicting results exist. It also remains unclear whether this association is causal or due to confounding or reverse associations. The present study aims to systematically analyze the epidemiological evidence for these associations, as well as estimate the potential causal relationships using Mendelian randomization (MR). Methods We analyzed four common addictive behaviors, including cigarette smoking, alcohol intake, coffee, and tea consumption (N = 126,906–4,584,729 participants) in this meta-analysis based on longitudinal studies. The two-sample MR was conducted using summary data from genome-wide associations with European ancestry (up to 1.2 million individuals). Results An observational association of GSD risk was identified for smoking [RR: 1.17 (95% CI: 1.06–1.29)], drinking alcohol [0.84 (0.78–0.91)], consuming coffee [0.86 (0.79–0.93)], and tea [1.08 (1.04–1.12)]. Also, there was a linear relationship between smoking (pack-years), alcohol drinking (days per week), coffee consumption (cups per day), and GSD risk. Our MRs supported a causality of GSD incidence with lifetime smoking [1.008 (1.003–1.013), P = 0.001], current smoking [1.007 (1.002–1.011), P = 0.004], problematic alcohol use (PAU) [1.014 (1.001–1.026), P = 0.029], decaffeinated coffee intake (1.127 [1.043–1.217], P = 0.002), as well as caffeine-metabolism [0.997 (0.995–0.999), P = 0.013], and tea consumption [0.990 (0.982–0.997), P = 0.008], respectively. Conclusion Our study suggests cigarette smoking, alcohol abuse, and decaffeinated coffee are causal risk factors for GSD, whereas tea consumption can decrease the risk of gallstones due to the effect of caffeine metabolism or polyphenol intake.
Introduction
Addictive behavior causes a major public health concern, and it has a massive, long-term impact on human suffering and societal costs (1). Gallstone disease (GSD) is one of the most common problems in the digestive tract and a major public health issue worldwide. The incidence of GSD continues to rise (around 10-20% of all adults in Europe), and its etiology remains to be understood (2). The pathogenesis of GSD involves environmental triggers, genetic predispositions, and behavioral factors; and the major pathogenetic factors, including abnormal cholesterol metabolism and slow intestinal motility are related to metabolic syndrome (3). Addictive behavior is of increasing interest as it is one of the leading contributors to the global burden of GSD and can be modified to achieve a desired preventive effect (4). Therefore, it is imperative to understand the relationship between common addictive behaviors and incident GSD, including cigarette smoking, alcohol drinking, coffee intake, and tea consumption.
Epidemiological investigations have consistently shown that current smoking, alcohol drinking, and coffee consumption play a key role in the incidence of GSD (5-7). A previous metaanalysis of 10 studies (N = 4,213,482) has provided evidence that smokers have an estimated 11% increased risk of GSD per 10 cigarettes per day compared to non-smokers (8). Another meta-analysis conducted by Wang et al. (9) that involved 14 studies (N = 316,028) has identified a significant non-linear trend of GSD risk reduction associated with the increment of drinking alcohol (up to about 30 g per day). In addition, one meta-analysis based on six studies (N = 227,749) has observed a dose-dependent association of coffee consumption with GSD [0.95 (0.91-1.00), P = 0.049] (10). These meta-analyses, despite their large sample sizes, have several limitations. First, there lack analyses for different types of alcohol (liquor, beer, and wine) and coffee (caffeinated and decaffeinated). Also, not all addictive behaviors have been comprehensively examined (e.g., pack-year smoking and drinking days per week). Second, the majority of evidence is cross-sectional, and the observational nature of conventional epidemiological studies hinders causal inference hampered by confounding or reverse causality (11).
Mendelian randomization (MR) fills the gap of making causal inferences by using single nucleotide polymorphism (SNP) as an instrumental variable (IV) since SNPs are usually established before the development of disease and therefore independent of confounders (12). Indeed, Yuan et al. have found that smoking is causally associated with GSD risk (13). However, tobacco smoking is a highly addictive behavior that contains large amounts of substances, such as nicotine, cannabis, and exposure to tobacco smoke (ETS) the causality of them with GSD has not yet been investigated. As for drinking alcohol, although common alcohol use was not significantly causally associated with the GSD risk, we additionally analyzed the causality between problematic alcohol use (PAU) and the risk of GSD. Moreover, a recent genomewide association study (GWAS) of caffeine intake has identified additional SNPs associated with coffee or tea consumption, which can be used as IVs for further MR (14). Note that the effect of consuming tea on the GSD risk lacks systematic evaluation.
The current study aims to comprehensively evaluate the relationship between these common addictive behaviors and the GSD risk. We first summarized the evidence in one updated meta-analysis only including a longitudinal study. Data from the meta-analysis was further tested for potential dose-response relationships and by trial sequential analysis (TSA) to check if the present evidence is conclusive. We then explored a putative causal association of tobacco smoking, alcohol use, caffeine intake, and tea consumption with the risk of GSD using a two-sample MR design.
Search strategy and meta-analysis
Our meta-analysis has been registered at PROSPERO (CRD42020179076) and following PRISMA checklists. We searched PubMed and Embase databases for studies published before January 2021, and references to the retrieved articles were manually searched for additional information (Supplementary Table 1). The flow chart is presented in Supplementary Figure 1. GSD was defined as gallstones diagnosed by ultrasonography or a history of cholecystectomy; participants without gallstones or cholecystectomy were considered as the Frontiers in Nutrition 02 frontiersin.org control group (15). Longitudinal studies, including nested casecontrol, cohort, and randomized controlled trials, provided sufficient data for calculating the effect sizes with 95% confidence intervals (CI) and were eligible for our analysis (see Supplementary Table 2). If the person-years of subgroup GSD cases were not reported, we calculated the proportion of new total cases for each group (dividing the exact number of GSD by RR) and multiplied the proportion by total personyears as described previously (16). Two authors (Y.B. and X.W.) extracted data back-to-back from identified articles in current research, and disagreement was solved by consensus. DerSimonian and Laird's random-effect meta-analysis was applied to summarize the association between addictive behaviors and GSD when I 1 exceeded 50%; otherwise, a fixedeffect meta-analysis was conducted (17,18). Heterogeneity sources were explored by conducting subgroup analyses. Funnel plots were drawn to demonstrate the possible publication bias if asymmetry were observed, and the bias would be further tested after combining with Egger's and Begg's test results (19). The pooled effect was adjusted by Duval and Tweedie's trim-andfill method to account for publication bias (20). Sensitivity was evaluated by omitting each estimate at one time to see to what extent a single study could influence the overall risk estimate. Pooled analyses were done using Comprehensive Meta-Analysis version 3.0 (Biostat, Englewood, NJ, USA).
Dose-response analysis
To investigate whether the dose of addictive substances intake was associated with GSD, we conducted Greenland and Longnecker's method using linear and non-linear models (21). The mean amount was used to assign the exposure levels for each risk estimate. For the open-ended lower boundary, the level was assumed to be zero, and non-taken was considered as the reference category. For the open-ended upper boundary, the highest level was assigned to 1.5 or 1.2 times the lower boundary of the category (22). In this study, we further tested a dosedependent association of GSD with smoking status (cigarettes per day and pack-year smoking), consuming alcohol (drinking grams per day, alcohol intake times per week, and drinking days per week), and intaking caffeine (coffee or tea consumptioncups per day). These statistical analyses were done with the use of STATA 16.0 (StataCorp. College Station, TX, USA).
Trial sequential analysis
TSA was applied to evaluate the sufficiency of the total sample size of a meta-analysis to investigate the associations. A cumulative Z-curve exceeds the trial sequential monitoring 1 http://www.ctu.dk/tsa limit or the required information size, suggesting conclusive evidence (23). TSA was conducted by the program version 0.9 beta. 2 All statistical significance were determined by P < 0.05.
Genetic instruments selection and outcome data sources
SNPs showing genome-wide significance (P < 5.0 × 10 −8 ) and with R 2 < 0.1 identified by LDlink 2 were used as IVs for lifetime smoking (i.e., ever and never smokers, smoking duration, heaviness, and cessation in ever smokers were taken into account) (24). The selection of IVs for smoking initiation (including ever-smoking, current-smoking, and smoking cessation) and common alcohol drinking, for (PAU, considering both alcohol use disorder and measures of problematic drinking), and for caffeine intake (the caffeine content per cup was multiplied by the number of cups of tea or coffee) were retrieved from three GWASs, respectively (14, 25, 26). All study populations were European descendants. The strength of instruments used in this study has been previously described, and an F-statistic larger than 10 was regarded as a strong instrument (27). Details are available in Supplementary Tables 3, 4.
Mendelian randomization analysis
For our MR study, the multiplicative random-effect inverse variance weighted (IVW) method was used to estimate the causal associations between addictive behaviors and GSD risk. In sensitivity analysis, the MR-Egger regression was used to identify and correct for the horizontal pleiotropy, the weighted median method provides the estimates when SNPs accounting for more than half of the weight are valid, and the maximum likelihood method maximizes the likelihood of the model based on the causal association (29)(30)(31). The p-value of the MR-Egger intercept was used to indicate potential horizontal pleiotropy, and Cochrane's Q-value was used to evaluate the heterogeneity among those SNPs for each addictive behavior (32). In this study, the large sample size allowed us to gain sufficient power (all were greater than 80%) for conclusive estimation of the associations between addictive behaviors and incident GSD. The analyses were performed via the MR-Base 3 using the R package "TwoSampleMR" (version 4.0.3, R Foundation for Statistical Computing, Vienna, Austria). Here, the causal association would be considered statistically significant when a Bonferroni corrected P-value was less than 0.013 (correcting for four exposures, including tobacco smoking, alcohol drinking, coffee, and tea consumption). A p-value < 0.05 was regarded as the marginal significance. Table 1 and Supplementary Figure 2). The associations were directionally consistent when stratified by sex, ethnicity, underwent cholecystectomy, or according to different types of addictive behaviors (Supplementary Figures 3-7). Of note, a significant increment in GSD risk was associated with smoking only in males (1.15 [1.11-1.20]), and current smokers increased about 7% risk of GSD compared to former smokers. Consuming coffee was significantly associated with a decrement of GSD risk [0.87 (0.79-0.96)] in females only. Although an association with GSD [0.84 (0.82-0.87)] was found in caffeinated coffee, it was not statistically significant in decaffeinated coffee ( Table 1). We also assessed the potential publication bias, and the adjusted funnel plot is shown in Supplementary Figure 8. Then a sensitivity analysis suggested that one of each included study did not influence the overall estimate of the meta-analysis (Supplementary Figure 9). Moreover, there were significant differences across all dose levels of cigarette smoking and alcohol intake with the risk of GSD.
Meta-analysis
Here, our dose-response meta-analysis showed a non-linear relationship between GSD risk with daily smoking per 10 cigarettes [1.10 (1.08-1.12), P nonlinearity ≤ 0.01]. We further detected a linear association that an increment of pack-years of smoking increased the risk of GSD [1.01 (1.01-1.01), P = 0.08] (Figure 1). However, there was no significant association between alcohol consumed grams per day and GSD, despite a non-linear relationship being found. A significant non-linear association of alcohol intake times per week was observed for GSD risk reduction with an RR of [0.81 (0.67-0.99), P ≤ 0.01)] per 5 units. We further found an increment of days per week of alcohol drinking decreased the GSD risk with a linear inverse association [0.96 (0.94-0.97), P = 0.89]. As for caffeine consumption, a potential linear association was detected between coffee cups per day and GSD risk [0.95 (0.94-0.96), per 1 cup]. Despite a non-linear relationship between GSD risk with consuming tea-cups per day (P = 0.01), we found no significant association. In addition, the risk of GSD increased by 4% and 8% with every 5 and 10 pack-years increments in cigarettes-smoking; while the risk was reduced by 20% and 23% per five units increment in alcohol-drinking days per week and coffee-cups per day (Supplementary Table 5).
Trial sequential analysis
In the TSA of our meta-analysis, the cumulative Z-curve crossed trial sequential monitoring and/or conventional boundary and penalized tests adjusted Z-curves also presented similar results, denoting that this evidence was robust and conclusive (Supplementary
Mendelian randomization analyses
As shown in Figure 2 and Supplementary Figure 11, our MR found that genetically predicted current smoking and PAU both were associated with an increased risk of GSD, while genetically predicted tea consumption was associated with a decreased risk of GSD. However, there was no genetic association between smoking cessation, common alcohol use, coffee consumption, and the risk of GSD.
The causal association between cigarette smoking and incident gallstone disease
In the IVW method, using lifetime-smoking associated 120 independent SNPs as IVs, we found that it had a causal effect on diagnosed cholelithiasis (OR: 1.008, 95% CI: 1.003-1.013, P = 0.001) and patients underwent cholecystectomy Overview of the design and main findings in this Mendelian-randomization study. Assumption 1 indicates that the genetic instruments are significantly genome-wide associated with these addictive substances of interest. Assumption 2 indicates that our genetic instruments should not be associated with confounders. Assumption 3 indicates that genetic instrument affect these outcomes only via the exposures. The directly protective effect on the risk of gallstone disease with tea consumption Consuming tea (an additional source of caffeine mainly in black tea) was significantly negatively associated with GSD risk in IVW ( Table 2)
Discussion
As summarized in this study, the meta-analysis based on a longitudinal study indicates that addictive behaviors are significantly associated with the incidence of GSD. Compared with never smokers, current smokers have a positive dosedependent response to GSD risk, and the evidence is further verified by the MR analysis. There are negative dose-response relationships between common alcohol use, coffee intake, and the GSD risk. However, the results of MR do not confirm the causal relationship between them. The novel finding of this study is that alcohol abuse may be causally associated with an increment in GSD risk, whereas tea consumption has a protective effect on the GSD risk in Europe.
Smoking has been shown to alter lipid metabolism, and the abnormal synthesis of bile may cause cholesterol supersaturation for the formation of gallstones (33). Consistent with a previous dose-response meta-analysis (8), the risk of GSD is found to be increased by smoking (cigarettes per day) with a non-linear relationship in our study. Moreover, a finding suggests that it is a linear dose-response association between pack years of smoking and risk of GSD and is further subjected to causality. It is also notable that the causal association of GSD is not significant in smoking cessation.
To date, evidence linking alcohol drinking with GSD is controversial (9). In this meta-analysis, we verify a negatively non-linear dose-response association between drinking alcohol grams per day and GSD risk, consistent with Cha et al. (34), but the strict study design and dose definition are used in our study. We conclude that the association of GSD risk reduction appears to reach the limit when the dose is higher than 45 g/day, and this finding (J-shaped) is similar to that of Figueiredo et al. (5), while the appropriate dose of alcohol-intake protects against GSD awaits future study. Concerning the types of alcohol, an RR of GSD is 0.82, 0.84, and 0.87 (liquor, beer, and wine) as the alcohol concentration decreased. One possible explanation is that alcohol may reduce cholesterol levels, improve HDL-C levels, and promote the secretion of bile acid, which in turn may inhibit gallstone formation (35). Meanwhile, we also propose two possible explanations for a non-causality found between common alcohol use and GSD risk in this MR. One possibility is that there may be a mediation effect for liver cirrhosis in the relationship. Some studies indicate that alcohol drinking increases the risk of liver cirrhosis, which has a close correlation with incident GSD (36,37). Second, it may have a potential non-linear relationship between them that moderate drinking decreases the risk of GSD, whereas problematic drinking increases the GSD risk. Interestingly, high coffee consumption was associated with a decrement in GSD risk. An MR also suggested a causal relationship between them in the Danish cohort (38). But, our results of MRs do not support such putative causality from a larger sample size in UKB, which agrees with a finding reported by Yuan et al. (13). In addition to population differences, one intriguing possibility is that self-reported coffee consumption includes decaffeinated coffee, coffee beverages, and others, which may weaken the effect of caffeine (39). Furthermore, this study provides the first report, to our knowledge, of a negative correlation of the GSD risk with drinking tea in Europe.
Tea consumption as another addictive behavior with caffeine intake (including black and green tea) has been associated with a GSD risk decreased in both genders within the population of Asia, and caffeine can stimulate cholecystokinin secretion and release bile acids into the intestine (40,41). Our MR verified this causal association in a European population, and tea polyphenol (mainly found in green tea) was also found to be causally associated with GSD risk, despite only one instrument being used. Certainly, more studies need to be done in the future. Current knowledge shows that polyphenols may accelerate bowel movements, and promote lipolysis and absorption, which in turn decreases morbidity in GSD (42).
Here, some plausible mechanisms are explored for the causal associations between addictive substance use and GSD risk. For the association between active smoking or ETS and the risk of GSD, the nicotine-dependence may be a key factor in this relationship. Of note, electronic cigarette has not been reported in GSD-related research. In addition, caffeine and tea polyphenols are the most commonly consumed psychostimulants, and they both causally decrease the risk of GSD in our MR. However, coffee consumption (including decaffeinated coffee or other beverage) is not associated with GSD, which may weaken the effect of caffeine.
There is high heterogeneity in our meta-analysis, and existing research regarding this topic is relatively fewer, which may have yielded publication bias. For example, the positive association between tea consumption and GSD in the American population might be due to the smaller number of studies included, which is an important limitation of our study. The second limitation is vertical pleiotropy, which could be shown to mediate the effect within a relationship between exposure and outcome. Another limitation is that we could not explore a non-linear relationship using this MR approach. As for the heterogeneity in a different population, our IVs were all identified in GWAS of a European-origin sample, although these instruments can only explain the percent of 0.24-1.72 (smoking cessation, caffeine-intake, etc.) in total estimated heritability, which limits the generalizability of our finding to diverse populations.
Conclusion
In conclusion, tobacco smoking, PAU, and decaffeinated coffee directly confer high risks of GSD; nonetheless, habitual caffeine intake and tea consumption may have a protective role against GSD due to an effect of caffeine metabolism or polyphenol intake. Accordingly, we infer that changing addictive behavior may be necessary for reducing the risk of GSD.
Data availability statement
The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.
Author contributions
BZ managed the project and study design. XW and YB read and abstracted the studies included in the meta-analysis. XJ and YB analyzed the data in the Mendelian randomization study. MZ and YB prepared the tables and figures. DG, MT, and YB did the statistical analyses. YB drafted the manuscript with HC, XS, YW, XW, and XJ. All authors reviewed and approved the article.
Funding
This work was supported by the National Natural Science Foundation of China (81874283 and 81903398). The introduction of talents of Sichuan University (YJ2021112) and Medical Youth Innovation Research Project of Sichuan Province (Q21016). The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. | 2022-10-10T13:46:41.964Z | 2022-10-10T00:00:00.000 | {
"year": 2022,
"sha1": "8f0d4c23214d487267074cd2c0e1247fa9a7f487",
"oa_license": null,
"oa_url": null,
"oa_status": null,
"pdf_src": "Frontier",
"pdf_hash": "8f0d4c23214d487267074cd2c0e1247fa9a7f487",
"s2fieldsofstudy": [
"Medicine"
],
"extfieldsofstudy": []
} |
86804495 | pes2o/s2orc | v3-fos-license | Mobile Robot Navigation in Indoor Environments: Geometric, Topological, and Semantic Navigation
The objective of the chapter is to show current trends in robot navigation systems related to indoor environments. Navigation systems depend on the level of abstraction of the environment representation. The three main techniques for representing the environment will be described: geometric, topological, and semantic. The geometric representation of the environment is closer to the sensor and actuator world and it is the best one to perform local navigation. Topological representation of the environment uses graphs to model the environment and it is used in large navigation tasks. The semantic representation is the most abstract representation model and adds concepts such as utilities or meanings of the environment elements in the map representation. In addition, regardless of the representation used for navigation, perception plays a significant role in terms of understanding and moving through the environment.
The answer to these questions refers to localization to determine where the robot is, pathplanning to know how to reach other places from where the robot is, and navigation to perform the trajectory commanded by the path-planning system. This approach has ruled robot navigation paradigm and many successful solutions have been proposed. However, the advances in technical developments allow a wider thought of mobile robot navigation giving a solution for a "bigger" problem. This solution would give answer to two additional questions, which are: • How is this place like?
• How is the structure of the environment I am in?
The answer to these new questions focuses on the importance of perception to determine the place and the elements of the place where the robot is and the importance of modeling the environment to determine the structure of the place and infer the connections between places. When robots are integrated in human environments and live together with humans, perception and modeling gain importance to enable future tasks related to service robots such as manipulation or human-robot interaction. In this chapter, we give a conceptual answer to the traditional robot navigation question and to the questions proposed above.
Robot navigation has been tackled from different perspectives leading to a classification into three main approaches: geometric navigation, topological navigation, and semantic navigation. Although the three of them differ in their definition and methods, all of them have the focus of answering the same questions.
From the beginning, authors have focused on generating metric maps and moving through the map using metric path planners. The most well-known algorithm to build metric maps is simultaneous localization and mapping (SLAM) as proposed by Bailey & Durrant-Whyte [2]. Wang et al. [3] use SLAM and Rapidly-exploring Random Tree (RTT) planning with Monte Carlo localization to drive a wheelchair in indoor environments. Pfrunder et al. [4] use SLAM and occupancy grids to navigate in heterogeneous environments. However, as larger maps are considered, it is computationally expensive to keep metric maps and other authors have focused on topological representations. Fernández-Madrigal et al. [5], design a hierarchical topological model to drive a wheelchair and perform reactive navigation and path-planned navigation. Ko et al. [6] works with topological maps where nodes are bags of visual words and a Bayesian framework is used for localization.
Regarding topological navigation, since the first developments, the global conception of the system has attracted the interest of several authors. In Kuipers & Levitt [7], the authors presented a four-level hierarchy (sensorimotor interaction, procedural behaviors, topological mapping, and metric mapping) to plan trajectories and execute them. Giralt et al. [8] defined a navigation and control system integrating modeling, planning, and motion control and stated that the key to autonomous robot was system integration and multisensory-driven navigation. Mataric [9] defined a distributed navigation model whose main purposes were collision-free path-planning, landmark detection, and environment learning. In Levitt [1], Levitt established a visual-based navigation where landmarks are memorized and paths are sequences of landmarks. These authors tackled for the first time robot navigation concepts from a topological approach and some of them noticed the importance of perception in the navigational processes.
At the last level of the semantic navigation paradigm, the ability to reason and to infer new knowledge is required. In today's world of robotics, there is a general tendency to implement behavioral mechanisms based on human psychology, taking as example the natural processing of thought. This allows a greater understanding of the environment and the objects it contains. This trend is also valued in the field of mobile robot navigation. Navigators have been increasing their level of abstraction over time. Initially, navigation was solved with geometric navigators that interpreted the environment as a set of accessible areas and areas that were not accessible. Then, the concept of node and the accessibility between nodes was introduced, which allowed increasing the level of abstraction generating graphs and calculating trajectories with algorithms of graphs. However, the level of abstraction has increased a step further, introducing high-level concepts that classify the rooms according to more complex and abstract data such as the utility and the objects they contain.
Another important aspect is the collection of the information of the environment, which has to be available for the navigation systems, among other tasks carried out by mobile robots. This information is provided by a perceptual system, therefore non-trivial problems appear related to object detection and place recognition. One of the most challenging issues in scene recognition is the appearance of a place. Sometimes, the same place may look different or different places may look similar. Also, the position of the robot and the variation of its point of view can affect the identification of the place when it is revisited. Environment models and the way to define the navigation tasks can be remodeled taking into account new technologies and trends. This will provide more autonomy to mobile robots and will help the interaction with humans in their usual environments. In this trend, vision is the main sensor for navigation, localization, and scene recognition. Place recognition is a very well-known challenging problem that not only has to do with how a robot can give the same meaning that a human do to the same image, but also with the variability in the appearance of these images in the real world. Place recognition has a strong relation with many major robotics research fields including simultaneous localization and mapping.
Geometric navigation
Geometric navigation consists of moving the robot from one point of the environment to another one, given those points by its coordinates in a map. The map of the environment is classically represented by a grid of points, and the trajectory between two points is a sequence of points the robot must reach in the given order. The controller of the robot must reach the next point of the path closing a loop with the encoder information about its position (distance and angle). One of the main objectives in the geometric navigation researches is the pathplanning task from an initial point to a final point, creating an algorithm able to find a path ensuring completeness.
Although the path-planning task is widely treated in mobile robots, computational capacity has increased exponentially and more complex algorithms can be developed. Algorithms try to find the shortest or fastest path while maintaining safety constraints. Another challenge is to try to get solutions that provide smoother trajectories, trying to imitate the human trajectories.
In the literature, many different algorithms can be found. One of the most important precedents were the works of LaValle [10], where a classification into two big groups depending on the way information is discretized is proposed: combinatorial planning and samplingbased planning. Combinatorial planning constructs structures containing all the necessary information for route planning; see De Berg et al. [11]. Sampling-based planning is based on an incremental representation of space and uses collision-free algorithms for path search. Here are some of the algorithms most used in the path-finding problem.
Deterministic algorithms
One of the first approaches tries to get all the possible paths between two points and choose the shortest one. Two example algorithms use potential fields and Voronoi diagrams.
Potential fields
The potential fields method is based on reactive planning techniques, which can be used to plan locally in unexplored environments. This method assigns to the obstacles similar characteristics that an electrostatic potential might have; in this way, the robot is considered as a particle under the influence of an artificial potential field that pulls it toward a target position, Figure 1a. The generation of trajectories is due to an attractive field toward the final position and another repulsive one with respect to the obstacles [12].
In this way, navigation through potential fields is composed of three phases: • Calculation of the potential acting on the robot using the sensor data.
• Determination of the vector of the artificial force acting on the robot.
• Generation of the movement orders of the robot. On the one hand, the attractive potential must be a function of the distance to the final destination, decreasing when the robot approaches this point. On the other hand, the repulsive potential should only be influenced when the robot is at a dangerous distance from obstacles. Therefore, the potential fields method allows to be performed in real time, as a local planner, considering obstacles only when the robot is at a minimum distance from them.
The problem that exists in this type of method is the local minimums [13], places where the potential is null but it is not the final position.
Voronoi diagram
The generation of Voronoi diagrams seeks to maximize the distance between the robot and the obstacles, looking for the safest path between two points in the space [14]. In this way, the diagram is defined as the locus of the configurations that are at the same distance from the obstacles.
The algorithm divides the space into sections formed by vertices and segments that fulfill the cost function of maximum distance between obstacles, Figure 1. Then, the trajectory is sought from an initial point to the objective. For a more realistic representation, obstacles are considered as polygons, since physically an obstacle is not a point.
One way of building the Voronoi diagram is using image-processing methods (skeletonization) based on the method of Breu [15]. These present a linear (and therefore asymptotically optimal) time algorithm in order to calculate the Euclidean distance of a binary image.
The algorithm is built with the pixels that result after performing the morphological operations on the image, with the nodes being the points where the lines that pass through the pixels of the image intersect.
A trajectory generated by Voronoi diagrams has the disadvantage that it is not optimal from the point of view of length traveled, it can also present a large number of turns [16]. Moreover, this method is not efficient for more than two dimensions.
Probabilistic algorithms
Over the years, a large number of solutions for the navigation problem have been presented using algorithms with random components, specifically in generic environments and with a large number of dimensions.
PRM algorithm
Probabilistic roadmap (PRM) is a trajectory planner that searches the connectivity of different points in the free space from a starting point to the final goal avoiding collisions with obstacles, using random sampling methods [17].
If the environment in which the robot is located is very complex, this type of algorithm allows finding a solution without a large computational requirement, so it is used in environments with a large number of dimensions.
One of the main problems of PRM is that the solutions it finds do not have to be the optimal trajectory. Also, since the way it generates the nodes is completely random, it produces a nonhomogeneous distribution of the samples in the space. These two disadvantages are seen in Figure 2a.
Rapidly exploring random tree algorithm
Rapidly exploring random tree (RRT) [18] provides a solution by creating random branches from a starting point. The collision-free branches are stored iteratively and new ones are created until the target point is reached. The algorithm is started with a tree whose source is a single node, the starting point. In each iteration, the tree expands by selecting a random state and expanding the tree to that state. The expansion is performed by extending the nearest node of the tree to the selected random state, which will depend on the size of the selected step. The algorithm creates branches until the tree takes a certain extension approaching the goal.
The size of the step is an important parameter of the algorithm. Small values result in slow expansion, but with finer paths or paths that can take fine turns [19].
Fast marching square algorithm
Fast marching square (FM 2 ) is a path-planning algorithm, that searches for the optimal path between two points. It uses the fast marching method (FMM) as a basis for calculation. It is a modeling algorithm of a physical wave propagation.
The fast marching method uses a function that behaves similar to the propagation of a wave. The form that this wave propagates, following the Eikonal equation, from an initial point to reach a goal position is the most efficient way in terms of time to reach it. The fast marching algorithm calculates the time (T) that the front of the wave, called interface, spends to reach each point of the map from a starting point. The FMM requires as previous step a The trajectories obtained using FMM present two drawbacks: great abruptness of turns and trajectories very close to the obstacles. This makes it impossible to use the algorithm as a trajectory planner in real robotics. The main change in FM 2 [20] solves these problems by generating a speed map that modifies the expansion of the wave taking into account the proximity to obstacles.
Topological navigation
Topological navigation refers to the navigational processes that take place using a topological representation of the environment. A topological representation is characterized by defining reference elements of the environment according to the different relations between them. Reference elements are denominated nodes and the relations between them are characterized as arches. The aim of topological navigation is to develop navigational behaviors that are closer to those of humans in order to enhance the human-robot understanding. Summing up the main implications of topological navigation [21]: • Topological navigation permits efficient planning. However, as they are based on relations, they do not minimize distance traveled or execution time.
• Topological navigation does not require precise localization.
• It is easy to understand by humans due to its natural conception.
• A complex recognition model is needed.
• Huge diagrams are involved in large environments and diagrams scale better than geometrical representations. Mobile Robot Navigation in Indoor Environments: Geometric, Topological, and Semantic Navigation http://dx.doi.org/10.5772/intechopen.79842 Topological representation classifications, as the one proposed by Vale & Ribeiro [22], differentiate mainly two ways of representing the environment: topological representations based on movements, for example the works developed by Kuipers & Byun [23], and topological representations based on geometrical maps, as proposed by Thrun [24]. These two conceptions differ mainly in the spatial relation between the real world and its representation. Regarding topological maps based on geometry, an exact relation between the environment and the representation is mandatory. Every topological node is metrically associated with a position or place in the environment, whereas, in topological representations based on movements, it is not necessary to have a metrical correspondence with the elements of the environment. An example of the same environment represented as a topological map based on geometry and as a topological map based on movements is shown in Figure 4.
In topological representations based on geometry, nodes normally correspond to geometrical positions (characterized as x; y; θ ð Þ) that correspond to geometrical events such as junctions, dead-ends, etc. and arches correspond to the geometrical transition between positions. In topological representations based on movements, the relation between nodes is more abstract as it does not have a geometrical meaning, being a qualitative relation instead of a quantitative one. That is why arches can be associated to different and more complex behaviors. In addition, nodes are associated to important sensorial events which can be determined for each application ranging from important geometrical events to landmarks and objects.
Although there are substantial differences between topological representations based on geometry and topological representations based on movements, both of them share the definition of the required modules so the system works efficiently. These modules will be explained in the following subsections, which are: modeling of the environment, planification and navigation, and perception interface.
Modeling of the environment as a topological graph
A topological graph, as defined by Simhon & Dudek [25], is a graph representation of an environment in which the important elements of an environment are defined along with the transitions among them. The topological graph can be given to the system by the user or can be built through exploration. Exploration strategies differ if the system works with a topological representation based on geometry or a topological representation based on movements. In the case of using representations based on geometry, exploration and map acquisition strategies such as the ones explained previously for geometrical representations can be used adding a processing stage to extract the relevant geometrical positions. In the case of representations based on movements, strategies like Next Best View, as proposed in Amigoni & Gallo [26], or Frontier Exploration, as in the work proposed by Arvanitakis et al. [27] can be used. Generally, these representations are translated into a text file that lists the information of the graph in order to maximize the graph efficiency. In Figure 5, an example of a map text file structure and its graphical interpretation is shown. In this example, a topological representation based on movements is used so the position of the nodes in the graph is not necessarily linked to the position of reference elements in the environment.
This text file contains all the information required for navigation. Nodes are ordered according to an identifier and they are associated with their corresponding event type (odometry, marker, closet, etc.). Arches are associated with the behavior or translational ability that the robot has to perform (GP, go to point; R, turn; etc.).
A model of the environment can be formed by several topological graphs containing different information or representing different levels of abstraction, in that case the environment is modeled as a hierarchical topological graph.
Topological planification and navigation
Topological navigation behaviors are determined mainly by the path-planning and navigation (meaning strict motor abilities execution) strategies as considered in Mataric [28]. The different nodes conforming the path are associated with the topological place where the robot perceives a sensorial stimulus or where it reaches a required position. In addition, while navigating, the behavior of the robot is subjugated to real-time perception and movement in the environment.
Topological planification is in charge of finding the topological route that the robot has to follow; it has to determine the path that the robot has to perform in order to move between two nodes. The path is a subgraph of the original graph of the environment. In order to plan a trajectory in a topological representation of the environment, a graph path-planning algorithm has to be used. There are many algorithms in the literature for this purpose, such as Dijkstra's algorithm, Skiena [29]. The main objective of Dijkstra's algorithm is to obtain the shortest path between nodes in a graph according to some heuristics or cost function. Given an initial node, it evaluates adjacent nodes and chooses the node that minimizes the cost function. This process is iterated until the goal node is reached or every connection between nodes has been explored.
Using traditional algorithms such as Dijkstra, many modifications can be implemented to establish heuristics that fit better to real environments. For example, the cost of a translation between nodes can be varied according to previous executions or pursing a specific global behavior, such as the personality factor proposed in Egido et al. [30].
The navigator is in charge of performing the path and reaching the goal node analyzing events and abilities or positions and transitions. An ability is the order the robot has to execute to reach coming nodes and it is intrinsically related to motor control. An event is the sign indicating the robot has reached a node, through sensorial information or odometrically in the case of representations based on geometry. A navigator can be based on a single behavior or on multiple behaviors. Topological representations based on geometry are all based on a single behavior that can be understood as "Going to point" or "Reaching next position." Topological representations based on movements can be based on single behaviors also or define a set of behaviors that would be used for optimizing the transitions between nodes. A multiplebehavior system contains several strategies and abilities in order to perform navigation behaviors and it should enable the inclusion of new strategies. Some of the common abilities implemented are Go to point, Turn, Contour following, and Go to object. Navigation systems are complemented with reactive obstacle avoidance modules to guarantee safe operation.
Managing the information of the environment: perception interface
If the topological navigation system is designed to work with multiple exteroceptive and proprioceptive events and it has to handle them simultaneously, these events have to be managed carefully. In order to manage the information that will be assigned to the nodes, a new module is needed. In this chapter, we will refer to this module as the perception interface.
The perception interface is decoupled from the system translating the specific information of each perception to a general structure which interacts with the other modules of the topological system, as shown in Figure 6. When translating the information to a general structure, the power of the system is multiplied exponentially, as adding new perception types becomes very simple. Perceptions are mainly based on vision, such as object detection and landmark detection, but some perceptions such as magnetic signals, ultrasounds, or proprioceptive perceptions such as odometry can be used. Another advantage of the perception interface is the possibility of establishing priorities and relations between perceptions easily.
Semantical navigation
The current tendency in robotics is to move from representation models that are closest to the robot´s hardware such as geometric models to those models closer to the way how humans reason, with which the robot will interact. It is intended to bring closer the models the way robots represent the environment and the way they plan to the way the humans do. The current trend is to implement behavior mechanisms based on human psychology. Robots are provided with cognitive architectures in order to model the environment, using semantics concepts that provide more autonomy, and which helps the navigation to be robust and more efficient.
In theory, any characteristic of the space can be represented on a map. But in general, it tends to identify a map with geometric information more or less complemented with additional information. When a mobile robot builds a map, the techniques used generally ignore relevant descriptive information of the environment and quite close to the process of made by humans, such as the navigation of the environment, the nature of the activity that there is, what objects it contains, etc.
The problem of the construction of semantic maps consists of maps that represent not only the occupation and geometric appearance of the environment but the same properties. Semantics is needed to give meaning to the data, mainly to make the simplest interpretation of data. The application of this idea to the construction of maps for mobile robots allows a better understanding of the data used to represent the environment and also allows an exchange of information between robots or between robots and people, if needed, easily. It also allows to build more accurate models and more useful environment models.
Semantic navigation is another step to the logical representation, it implies an additional knowledge about the elements of the world and it allows the robot to infer new information. For instance, the root can perform this type of associations: "There is a photocopier in this room-> there must be paper nearby".
Semantic navigation allows the robot to relate what it perceives to the places in which it is located. This way, an environment model is managed using the objects and on the concepts that they represent. All this information is used to classify the place where objects are located and it is used to reach a given location or a specific target. Therefore, the robot can find places or rooms Vasudevan & Siegwart [31] semantically related with the target, even if it is in a littleexplored or unknown environment. This navigation level allows to complete the information that is necessary from a partial knowledge of the environment.
Semantic navigation is related to classification methods of the environment object and places, which provides more effectiveness, as is described in Vasudevan & Siegwart [32]. An example of place identification using a Naive Bayes Classifier to infer place identification is shown in Duda et al. [33] and Kollar & Roy [34]. These works show that relations between object-object and objet-scenario can be used to predict the location of a variety of objects and scenarios.
Knowledge representation of the environment
One of the most important issues is how the robot models the environment. In semantic navigation, a large amount of information related to the environment and to the objects of the environment is used in the environment representation. For this representation, an ontology can be used to define the concepts and the links between objects in the environment. Figure 7a shows the concepts of the ontology and its relations.
An ontology design through a database
The ontology design can be implemented using different tools. In this work, a relational model using a database is proposed. Figure 7b shows a relational database diagram scheme. This scheme is used to understand the elements of the system and its relations. Tables used to model the ontology are described below.
Basic information tables
Basic information tables are used to store the environment elements. Four elements have been considered: physical rooms (real locations sensorially perceived by the robot), conceptual rooms (the type of room that the robot can recognize), physical objects (objects perceived by sensors), and conceptual objects (each semantic information of the object that the robot must recognize). With these tables, objects and places have been modeled. More tables are needed in order to complete the relations and information of the environment that are described below.
Links between primary tables
To complete the model, the following links between the primary tables are needed: Objects are always in rooms; so, between PhysicalRoom and PhysicalObject tables, there exists a link. And a physical room may have an indeterminate amount of physical objects. In table PhysicalObject, the room associated to each object is also stored.
Tables that allow searches by semantic proximity
As an example of the implementation of the tables, Figure 7b shows the design of the database. This database contains several tables that manage important information to help the robot to find objects and relations. With the information of this • Interaction: Objects interact with other objects. This table can handle any type of interaction, although tests have been performed with a limited number of them. The interactions taken into account are: BE_A, when an object is a subtype of another object; IS_USED_WITH, when an object is used with an object; and IS_INSIDE_OF, when usually an object is inside another object.
• Utility: All objects have one or more utilities. The tendency is to group objects that have the same utility (or something related) in the same places. Also, concepts are created regarding kinds of room depending on what they are used for.
• Meaning: The actions or utilities that an object has often are associated with a specific meaning. The goal of the navigation may also be oriented to emotions or places that give a feeling, such as calm or fun. For example, read is an action quiet.
• Characteristic: Objects may have features to best define the concept or even differentiate them from others. For example, on the one hand, water may be cold, warm, or hot. And that implies differences in location. Cold water may be found in the refrigerator or in a fountain, but hot water would come out from the faucet.
These tables include the semantic information of the objects in the environment that can help the robot to locate on it. For example, considering the utility, it can be said that a computer is used to work. The robot can go to an office if it needs to locate a computer. This relational database model provides a simple way to define and manage semantic knowledge using simple queries without the need to define rules as it has been described in previous works (Galindo et al. [35], Galindo et al. [36]).
Semantic information management
Once the environment and its components are modeled, the next step is to manage the information in order to accomplish navigation tasks. Robots need to localize in the environment and calculate the path to the targets defined with semantical information. In the real robot, semantic targets must be transferred to the topological and geometrical levels in order to complete the movement of the robot.
Semantic target obtention
Similar to the previous navigation system, a way to establish how to reach a target is needed.
In the proposed semantic system is defined the concept of Semantic Position as a contextual information unit related to the robot's position. Semantic Position contains attributes that correspond to a room code (the target room in this case) and to an object code (the target object in this case). It is therefore understood that a position is obtained in relation to a semantic reference point, which can be either objects or rooms. The target of the robot is defined as a Semantic Position.
Depending on the information available in the database about the requested destination and the type of destination (object or room), several options can be found in order to establish the semantic path: • If the destination is a known object in a known room, the robot has all the information to reach the Semantic Target.
• If the destination is an unknown room, a semantic exploration routine must be used to manage the semantic information to locate a room with similar semantic information of the destination.
• If the destination is an unknown object in a known room, the robot would try to explore it to get an object with similar semantic information in the room to define it as a Semantic Target.
• If the destination is an unknown object in an unknown room, all information of the object and room in the database must be used to get an object in a room which matches with the semantical information of the destination.
To get the Semantic Target, several consultations to the database must be done. One query is used to obtain the specific object instances that have already been stored, other query to know the physical room where a specific object is found, and a last query to match a room with a specific type of object.
In addition, this system allows the search of destinations by semantic proximity. For example, using the knowledge of Figure 8a if the requested objective is to do something fun, the answer would be to go to the computer-1. This is because the computer-1 is an object associated with the computer concept. The computer concept is associated with the watching movies utility and the watching movies utility is associated with the fun concept. Using the knowledge of Figure 8b, searches for objects with characteristics and actions can be done. If the destination is to drink something cold, the system recognizes cold water as a valid objective. The cold water is related to the refrigerator, the refrigerator is related to the kitchen, and the kitchen is related to Room-1. Then, the system goes to Room-1.
Semantic identification of a place
A place, in this case a room, can be identified as a vector of detected objects. The object observed from the environment must be analyzed and with a query we can get a list of the types of room where the object can be found. If an object is found, it defines the room where it is located. For instance, the bed is an object that can only be in a bedroom. If the result of the previous query returns several records, it is because the object is not discriminatory enough to be able to conclude any identification, and more additional semantic information and more queries are needed to get the destination room.
Example
In this example, the robot is asked for several targets in a home. In the experiment, the robot has the information provided by the previous exploration of the environment. This information is stored in several tables: 1a is the table with physical objects information, 1a is the table with physical objects information and 1b is the table with deducted knowledge.
The objective is to use semantic information to get the semantic target, that is attached to a topological or to a geometrical target to which the robot must move ( Table 1). First, the robot is asked for the kitchen. Checking the table PhysicalConceptualRoom with a query, the result is Room-1. Repeating the query, in the tables, there are no more kitchens in the database, so the process finishes.
In a second experiment, the robot is asked for a chair. In the initial environment, there are three real objects of the CHAIR type. In the query, the robot identifies three chairs: Chair-1, Chair-2, and Chair-3. The robot gets the information of the rooms in which the chairs are and moves to the first option. If it indicates that this is not the correct room, the robot moves to the rest of the options. This process is described in Figure 9.
In case the object has no match with the observed physical object, it is necessary to ask again for a chair, after having discarded the chairs from the previous test. The operation sequence is shown in Figure 10a. In Figure 9, another query on chair is shown, and the robot starts to explore searching in new types of rooms where chairs could be found.
(a) Content of the
Perception model for navigation
Talking about autonomous robots implies carrying out movements safely and having a complete knowledge of the environment. These elements define the capabilities of action and interaction between the environment, humans, and robots. Tasks performed by mobile robots such as navigation, localization, planning, among others, can be improved if the perceptual information is considered. So, the general idea is detecting and identifying meaningful elements (objects and scenes) of the environment. There are different ways to obtain the information about the environment. One of them consists of detecting recognizable features of the environment (natural landmarks) using several types of sensors. The detection of artificial landmarks, based on the work of Fischler & Elschlager [37], can be used to acquire a representation of the environment. However, due to the technological advances in 3D sensors (e.g., RGB-D sensors) and according to Huang [38], vision has become the main sensor used for navigation, localization, and scene recognition. Vision provides significant information about the objects present in a place; at the same time, it is capable of providing semantic information of the scene where it is. Scene recognition is a very well-known challenging issue that deals with how robots understand scenes just like a human does and the appearance variability of real environments. Therefore, regardless of the type of navigation used, whether geometrical, topological, or semantic, place recognition and object identification play a significant role in terms of representation of the environment.
Object recognition
To perform several tasks in common indoor environments, mobile robots need to quickly and accurately verify and recognize objects, obstacles, etc. One of these crucial tasks is to move safely in an unknown environment. Autonomous robots should be able to acquire and hold visual representations of their environments.
The most important stages in an object recognition model are: feature extraction and prediction. As for feature extraction, the identification of significant aspects of different objects that belong to the same class, independently of the appearance variabilities, such a scaling, rotation, translation, illumination changing, among others, is crucial to obtain a suitable representation of the objects present in a scene. Some techniques are based on local and global descriptors such as the works presented by Bay et al. [39] and Hernández et al. [40], or a combination of both of them (Hernandez-Lopez et al. [41]). Also, in other approaches such as the work presented by Csurka et al. [42], visual vocabularies (e. g., bag of words) are commonly used to create a proper representation of each object.
Regarding prediction, through the vectors created from the extracted features, it is possible to learn these characteristics in order to identify objects that correspond with each class. In the literature, different classification techniques based on machine learning such as nearest neighbor classifier, neural networks, AdaBoost, etc., have been proposed depending on the kind of extracted features (Pontil & Verri [43]). Machine learning is a field of computer science that includes algorithms that improve their performance at a given task considering the experience. In this way, support vector machine (SVM) is one of the most helpful classification algorithms. The aim of SVM is to generate a training model that is capable of predicting the target classes of the test dataset, considering only its attributes.
Generally, an object recognition system that works in real time is divided into two stages: offline and online. Offline stage includes all the processes to reduce the execution time and guarantee the efficiency of the system, which are image preprocessing, segmentation, feature extraction, and training process. Online stage refers to the processes carried out in real time with the goal to ensure the interaction between the robot and the environment. Mainly, the processes included in the online stage are image retrieval and classification.
Object detection can be very useful for a navigation system since it allows the robot to relate what it perceives to the scenes in which it is. For this reason, it is necessary to consider that the designed systems and the sensors are not perfect. Therefore, the object detection model has to incorporate uncertainty information. The uncertainty management is a relevant aspect in an object recognition system because it allows to represent the environment and its elements in a more realistic way. The uncertainty calculation can be determined considering the dependency relations between different factors. Some of the factors are: the accuracy of the model that can be determined empirically and a factor based on the outcomes of the classification algorithm. Also, other factors can include the influence of other elements of the environment, for example the distance during the detection process. Finally, considering these factors makes it possible to obtain a robust object recognition model to serve as an input to a mobile robot.
Place recognition
The growth of service robotics in recent years has created the needed for developing models that contribute to robots being able to adequately handle information from human environments. A semantic model can improve the high-level tasks of a robot, such as, semantic navigation and human-robot and robot-environment interaction. According to Wang et al. [44], semantic navigation is considered as a system that takes into account semantic information to represent the environment and then to carry out the localization and navigation of the robot.
In view of the above, place recognition deals with the process of recognizing an area of the environment in which there are elements (objects), actions are developed, and robotenvironment and human-robot interaction is possible. The scene-understanding issue can be defined as a combination between scene recognition and object detection. There are approaches that try to solve the scene recognition problem through computer vision algorithms, including the creation of complex feature descriptors (Xie et al. [45]), and a combination of feature extraction techniques (Nicosevici & Garcia [46] and Khan et al. [47]), among others. Moreover, if it is desired to get a robust model as close as possible to reality, the incorporation of environmental data and errors of the sensors is needed.
Some approaches are based on machine learning and select support vector machine as classification technique. Commonly, this type of place recognition model is composed by the following processes ( Figure 11): image preprocessing, that includes the selection of the datasets and the initial preprocessing of the images to obtain the training model; feature extraction, that can implement some techniques such as bag of words, local and global descriptors, among others; training process, where the parameters of the classifier are defined and the model of the scene is generated; prediction process that generates the final results of the possible scene where the robot is located; and ,finally, a last process called reclassification has to be considered in which it is possible to generate a relationship between the place recognition model and the elements (objects) of the environment.
In this way, the influence of the objects in the scene can improve, worsen, and correct the final results on where the robot is located. The process implies the adjustment of the probabilities of being in a place and therefore the management of uncertainties. To do that, it is necessary that the place recognition model and the object recognition system work simultaneously in real time. The uncertainty can be modeled from a set of rules based on learning to determine the probability of co-occurrence of the objects. Also, it is important to incorporate the prior information of the place and object recognition models. Finally, it is possible to apply different theorems such as Bayes to determine the final relation between objects and scenes, and the last result about where the robot is. The adjusted prediction has to be available as input for relevant tasks such as localization, navigation, scene understanding, and human-robot and robot-environment interaction. All this contributes to the main goal that is to have an autonomous and independent robot, with a wide knowledge of the environment.
Conclusion
The aim of this chapter was to describe different approaches for global navigation systems for mobile robots applied to indoor environments. Many researches and current developments are focused on solving specific needs for navigation. Our objective is to merge all these developments in order to classify them and establish a global frame for navigation.
Robot navigation has been tackled from different perspectives leading to a classification into three main approaches: geometric navigation, topological navigation, and semantic navigation. Although the three of them differ in their definition and methods, all of them have the focus on driving a robot autonomously and safely.
In this chapter, different trends and techniques have been presented, all of them inspired by biological models and pursuing human abilities and abstraction models. The geometric representation, closer to the sensor and actuator world, is the best one to perform local navigation and precise path-planning. Topological representation of the environment, which is based on graphs, enables large navigation tasks and uses similar models as humans do. The semantic representation, which is the closest to cognitive human models, adds concepts such as utilities or meanings of the environment elements and establishes complex relations between them. All of these representations are based on the available information of the environment. For this reason, perception plays a significant role in terms of understanding and moving through the environment.
Despite the differences between the environment representations, we consider that the integration of all of them and the proper management of the information is the key to achieve a global navigation system. Figure 11. Schematic of a general scene recognition model. Vision data are the initial input of the recognition model. Then, the metric data and object information can be included into the model to update the final outcome of the possible place where the robot can be. | 2019-03-28T13:14:17.233Z | 2018-11-05T00:00:00.000 | {
"year": 2019,
"sha1": "b851cd53b2be15513fb803f4961d6c2694163e1e",
"oa_license": "CCBY",
"oa_url": "https://www.intechopen.com/citation-pdf-url/63790",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "b52e86e8261d541b62a877525e63dc57911a1069",
"s2fieldsofstudy": [
"Computer Science"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
483479 | pes2o/s2orc | v3-fos-license | Memory performance on the Auditory Inference Span Test is independent of background noise type for young adults with normal hearing at high speech intelligibility
Listening in noise is often perceived to be effortful. This is partly because cognitive resources are engaged in separating the target signal from background noise, leaving fewer resources for storage and processing of the content of the message in working memory. The Auditory Inference Span Test (AIST) is designed to assess listening effort by measuring the ability to maintain and process heard information. The aim of this study was to use AIST to investigate the effect of background noise types and signal-to-noise ratio (SNR) on listening effort, as a function of working memory capacity (WMC) and updating ability (UA). The AIST was administered in three types of background noise: steady-state speech-shaped noise, amplitude modulated speech-shaped noise, and unintelligible speech. Three SNRs targeting 90% speech intelligibility or better were used in each of the three noise types, giving nine different conditions. The reading span test assessed WMC, while UA was assessed with the letter memory test. Twenty young adults with normal hearing participated in the study. Results showed that AIST performance was not influenced by noise type at the same intelligibility level, but became worse with worse SNR when background noise was speech-like. Performance on AIST also decreased with increasing memory load level. Correlations between AIST performance and the cognitive measurements suggested that WMC is of more importance for listening when SNRs are worse, while UA is of more importance for listening in easier SNRs. The results indicated that in young adults with normal hearing, the effort involved in listening in noise at high intelligibility levels is independent of the noise type. However, when noise is speech-like and intelligibility decreases, listening effort increases, probably due to extra demands on cognitive resources added by the informational masking created by the speech fragments and vocal sounds in the background noise.
INTRODUCTION
Speech understanding requires the interplay of top-down and bottom-up processes. Top-down processes include cognitive abilities that allow speech perception and comprehension (Davis and Johnsrude, 2007;Besser et al., 2013), while bottom-up processes include the perception of sound and the ability to hear. Hearing can be regarded as a mainly passive function that provides access to the auditory world via perception of sounds. Listening can then be viewed as a higher order function that requires intention and attention (Kiessling et al., 2003;Pichora-Fuller and Singh, 2006). Every day we hear many sounds, but we only listen to some of them. We hear the hum from the refrigerator but we may listen attentively to the news on the radio. Consequently, listening is required when heard information is to be processed for comprehension and to be remembered. However, the processes involved in listening, intention and attention, load on cognitive resources and therefore demand expenditure of effort (Kiessling et al., 2003;Pichora-Fuller and Singh, 2006).
In favorable listening conditions the speech signal is intact and understanding is implicit and automatic (Rönnberg, 2003;Rönnberg et al., 2008Rönnberg et al., , 2013. However, when listening takes place in adverse conditions, a mismatch between the input from the speech signal and the phonological representations that are stored in long term memory may occur. Then explicit processing is needed for speech recognition. Thus, having a good cognitive capacity facilitates speech recognition in adverse listening conditions (Edwards, 2007;Akeroyd, 2008;Avivi-Reich et al., 2014). Adverse conditions may arise due to signal degradation caused by an unfamiliar speaker, competing background sounds, signal processing in a hearing aid, or hearing impairment (Stenfelt and Rönnberg, 2009;Mattys et al., 2012). Therefore, more cognitive resources appear to be needed when listening in noise than in quiet (Larsby et al., 2005;Pichora-Fuller and Singh, 2006;Edwards, www.frontiersin.org Akeroyd, 2008;Mishra et al., 2013a;Ng et al., 2013a). Even though low levels of noise can be beneficial for speech perception of weak signals through stochastic resonance (Moss et al., 2004), for well audible and clear speech noise result in worse speech perception that load the cognitive resources. These cognitive resources may include working memory and executive functions . Working memory is the ability to temporarily store and process information (Baddeley, 2000). During speech comprehension, executive functions are required to update working memory with new information and simultaneously remove old information (Miyake et al., 2000). It has been suggested that both working memory and updating processes are involved in disambiguating degraded speech and inferring absent information when listening takes place in adverse conditions (Rudner et al., 2011b). This may compensate for speech understanding difficulties (Rönnberg et al., 2008Rudner et al., 2011a;Mishra et al., 2013a). However, it seems that the relation between speech perception in noise and working memory capacity (WMC) is stronger when speech is masked by a fluctuating masker compared to stationary noise George et al., 2007;Lunner and Sundewall-Thoren, 2007;Rudner et al., 2009Rudner et al., , 2011aRönnberg et al., 2010;Koelewijn et al., 2012;Zekveld et al., 2013). An explanation for this might be that individuals with greater cognitive capacity are better able to utilize the short periods with increased signal-to-noise ratio (SNR) to infer information that is masked when the noise is louder (Duquesnoy, 1983), but they might also be better to inhibit the distracting effect of the noise.
Cognitive resources are consumed in the act of listening, which in turn leaves fewer resources to process the auditory information at a higher level (Rudner and Lunner, 2013). The residual cognitive resources after successful listening has taken place are referred to as cognitive spare capacity (Mishra et al., 2010;Rudner et al., 2011a). It has been shown that cognitive spare capacity is sensitive to processing load relating to both memory storage requirements (Mishra et al., 2013a,b) and background noise (Mishra et al., 2013a). Rönnberg et al. (2014) showed an effect of SNR with decreased memory performance in poorer SNR for individuals with normal hearing and high WMC, using the Auditory Inference Span Test (AIST). This test is designed to measure the ability to apply different levels of cognitive processing to auditory information as an objective measure of listening effort. These levels are designed to load differently on working memory and the executive function of updating. When background noise level increased the memory performance decreased, even though speech intelligibility levels were better than 90% (Rönnberg et al., 2014). This suggests that more cognitive resources were engaged in listening when background noise increased, which reduced residual resources needed to remember the auditory information. However, this was only true for individuals with greater WMC. This indicated that the test might be too difficult for individuals with less WMC, and that the extra demands the noise put on the cognitive system did not further decrease the overall low memory performance. Other studies have showed an effect of improved memory performance for hearing impaired individuals with high WMC when a noise reduction algorithm was used (Ng et al., 2013a). This suggests that background noise affects memory performance for individuals with normal hearing as well as individuals with hearing impairment, but that this effect is dependent on task difficulty as well as the individual's WMC.
Limited WMC is gradually consumed by increasing processing demands when listening takes place in adverse conditions, leaving fewer resources to process and store information (Pichora-Fuller and Singh, 2006;Schneider, 2011), or in other words, leading to less cognitive spare capacity (Rudner et al., 2011a;Rudner and Lunner, 2014). Therefore, an individual with higher WMC is likely to cope better with adverse listening conditions than an individual with lower WMC (Lunner, 2003;Larsby et al., 2005;Pichora-Fuller and Singh, 2006;Foo et al., 2007;Pichora-Fuller, 2007;Rudner et al., 2009;Schneider, 2011). When a modulated masker is used, this difference is expected to be more pronounced (Koelewijn et al., 2012;Zekveld et al., 2013). Depending on the SNR, the modulated noise can divide the speech signal into intelligible and unintelligible parts. This is because the modulated noise contains short periods where the masker has low magnitude resulting in higher SNRs, where speech recognition is better, which in turn might lead to a release from masking of the target speech (Festen and Plomp, 1990). The cognitive processes, WMC and updating ability (UA), store and update unidentified disjointed parts of the speech signal, caused by the modulated masker, in working memory until the speech information can be resolved. Consequently, an individual with greater cognitive capacity is likely to be more capable to decode speech embedded in a modulated masker and thereby better speech recognition. As processing continues, the contents of working memory are continually updated with new information and old pieces of information are discarded (Rudner et al., 2011b). Therefore, an individual with greater cognitive capacity will perform better on a task that tests storage and processing of auditory information compared to an individual with fewer cognitive resources. More specifically, in easy listening conditions with low cognitive loads, there would neither be a significant performance difference between individuals with high or low WMC, nor between individuals with high or low UA, since task demands are low. However, in adverse listening conditions or when task demands require more cognitive processes, as updating information or processing of information in working memory, individuals with higher cognitive capacity are likely to perform better. Finally, when the masker is modulated, the difference in AIST performance between individuals with high cognitive capacity and individuals with low cognitive capacity is likely to be greater than in steady-state noise (Koelewijn et al., 2012;Zekveld et al., 2013).
The aim of the present study was for the first time to test whether type of noise influences listening effort measured using the AIST (Rönnberg et al., 2011) at high speech intelligibility levels. AIST performance was expected to be best in amplitude modulated noise (AMN) compared to steady state noise (SSN) and the international speech test signal (ISTS) when intelligibility was at equal level for all noise types. We also expected AIST performance to decrease with increasing noise level, as also shown by Rönnberg et al. (2014). Furthermore, we expected that participants with better cognitive capacity, i.e., higher WMC and better UA, would show better AIST performance than participants with worse cognitive capacity, similar to Rönnberg et al. (2014). Also, Frontiers in Psychology | Auditory Cognitive Neuroscience participants with high cognitive capacity were expected to perform better than participants with lower cognitive capacity on AIST tasks presented at poorer SNRs in modulated noise with high memory and processing demands.
PARTICIPANTS
Twenty participants with normal hearing thresholds, 11 women and 9 men, with a mean age of 35 years (SD: 4.4, range 28-42) accepted to be part of the study. They were all native Swedish speakers. Baseline audiometry was done (in a sound treated room according to ISO 8253-1:2010) to verify the inclusion criteria of hearing thresholds better than or equal to 20 dB HL for the frequencies 250-4000 Hz in both ears. These frequencies were used as inclusion criteria since there is little information in the speech material used above these frequencies. Three participants did not have normal hearing for all frequencies (125-8000 Hz). One participant had a threshold of 30 dB HL at 6000 Hz at the worst ear, one participant 35 dB HL at 6000 Hz and 40 dB HL at 8000 Hz at the worse ear, and one participant 30 dB at 125 Hz at the worse ear. The participants had self-reported normal visual acuity (after correction), and no tinnitus problems. All had participated in a previous study (Rönnberg et al., 2014). The study was approved by the Regional Ethical Review Board in Linköping.
MATERIALS
The AIST test (Rönnberg et al., 2011(Rönnberg et al., , 2014 uses five-word matrixtype sentences in Swedish, the Hagerman sentences (Hagerman, 1982;Hagerman and Kinnefors, 1995). These sentences always have the same structure: name, verb, number, adjective, item. For example "Anna has four new gloves," see Figure 1. The tokens for each category are selected from a closed set of 10 items. Thus, the Hagerman sentences have low redundancy, which makes it impossible to predict any of the words from the context provided in the sentence.
Three noise types were used in the experiment. One of these was the original speech-shaped steady state noise (SSN) by Hagerman (1982) which has the same long-term average spectrum as the speech material. The second noise type (AMN) was the same as SSN but amplitude modulated with a modulation frequency of 5 Hz and a modulation depth of 20 dB. The third noise type was the ISTS (Holube et al., 2010), which consists of six voices reading a story in six different languages. These recordings are cut into 500 ms segments, which are then randomized and concatenated. This method ensures a natural speech signal that is largely nonintelligible.
The test was administered at three different SNRs targeting a speech intelligibility of above 90% but below 100%, see Figure 2. This ensured reasonably good speech recognition, while the noise level theoretically caused a challenging listening situation. In a previous study (Rönnberg et al., 2014), the AIST was administered in SSN at three SNRs (−2, −4, and −6 dB). These SNRs corresponded to the average speech intelligibility levels of 97, 96, and 91% in SSN. Ten participants with normal hearing, none of whom took part in the present study, were recruited to determine SNRs for the same three
FIGURE 1 | Schematic of the Auditory Inference Span Test (AIST).
A sub-list of three Hagerman sentences with SQ are shown. These are then followed by three memory load level (MLL) questions, all of these belong to the same MLL. MLL 2 questions are shown. speech intelligibility levels: 97% (SNR1), 96% (SNR2), and 91% (SNR3) for the target sentences embedded in AMN and ISTS. Matching speech intelligibility levels between noise types enabled comparison in AIST performance between noise types, and also made for a very conservative test of differences in listening effort across noise types and SNRs. The SNRs were obtained by altering the noise level, while holding the speech level constant. The sound was presented bilaterally through headphones.
AUDITORY INFERENCE SPAN TEST
The AIST is a dual-task hearing-in-noise test, combining auditory and memory processing (Rönnberg et al., 2011). The participants' task is to recall and process the information from the sentences and respond in a three-alternative forced-choice procedure. In the present study, a total of nine sentences, all belonging to the same original list (Hagerman, 1982) of ten sentences, were presented consecutively in each noise type at each SNR. This was to keep speech intelligibility balanced, and to avoid duplicate answer alternatives. To verify speech recognition, one word from each sentence was probed immediately after the presentation [this will be termed sentence question (SQ)]. The accuracy and timing of the responses www.frontiersin.org FIGURE 2 | Signal-to-noise ratio threshold for speech intelligibility levels 91, 96, and 97%.
to these questions were recorded. The AIST was administered in accordance with the standard procedure (Rönnberg et al., 2011). After each sub-list of three sentences, the participant was prompted to answer three sequentially presented multiple choice questions about the information given in the sentences, see Figure 1. These questions were designed to engage one of three levels of cognitive processing, called memory load levels (MLLs). Only one MLL was probed at a time, using three different questions. The multiple choice alternatives were names, numbers, or items. The order of presentation of MLLs was balanced between participants to avoid order effects. MLL 1 tapped into memory storage by asking the participant to recall which of three given words occurred in the sentences presented, e.g., "Which of the following items was used in the sentences." This type of question could be answered simply by scanning information held in working memory. MLL 2 also tapped into memory storage but also required updating, e.g., "What item did Britta have?" This type of question could be answered by scanning the sentences to find the appropriate name, updating working memory to maintain the relevant sentence and then scanning the sentence to find the relevant item. Consequently, MLL 2 made greater demands on working memory storage and updating than MLL 1. MLL 3 was the most cognitively demanding level. It required storage and updating of information in working memory, as well as processing of the information from all three sentences presented, e.g., "Which item was there most of ?" This type of question could be answered by scanning the sentences for the relevant information and comparing between sentences to find the information that met the criterion. After that, memory could be updated to retain the appropriate sentence and identify the correct answer. Thus, MLL 3 made greater cognitive demands than MLL 2, specifically in terms of working memory storage, comparing characteristics and updating. Correct responses related equally often to the first, second, and third sentences and a balancing procedure ensured that this applied across conditions and participants. The AIST score was the number of questions that were correctly answered for each MLL in each noise type at each SNR.
COGNITIVE TESTS
The reading span test (RS; Rönnberg et al., 1989;Daneman and Merikle, 1996) is a well-established test of working memory (Unsworth and Engle, 2007). A short version in Swedish, with a maximum score of 28, was used in the present study (Rönnberg et al., 2014). Grammatically correct three-word sentences were presented, one word at the time, on the computer screen. Half of the sentences were reasonable and half were absurd. After each sentence, the participant was asked to judge whether it made sense or not. After each set of between 2 and 5 sentences, the participant's task was to recall in serial order either the first or the last words of each of the sentences in the set. The prompt "first" or "last" was provided only after set presentation was complete. The reading span score was the number of correctly recalled words.
The letter memory test (LM) evaluates the executive function of updating (Miyake et al., 2000). Lists of consonants were presented with capital letters one at a time on the computer screen, and the participant's task was to recall the last four letters in the correct order. The length of the lists was either 5, 7, 9, or 11 letters long, and the presentation order was randomized. Thus, list length could not be accurately predicted. The letter memory score was the number of the four target letters that were correctly recalled in serial order for each list.
SET UP AND TEST PROCEDURE
The AIST experiment was administered with an application developed in Matlab (R2013a; Rönnberg et al., 2014). Visual stimuli were presented on a 14 computer screen, and auditory stimuli via an M-Audio FireWire 410 audio interface through a pair of Sennheiser HDA 200 headphones with the speech level calibrated to an output level of 60 dB SPL. The testing took place in a single session in a quiet room. Even if the room was not sound attenuated, the test environment was deemed quiet enough not to affect the tests conducted. Before the test started, the participants read written instructions as a complement to instructions given orally by the test supervisor. The total testing time was at most 30 min.
STATISTICAL ANALYSES
The data collected in this study were analyzed together with AIST performance in SSN as well as cognitive measurements of the participants collected in a previous study (Rönnberg et al., 2014). Repeated measures analyses of variance were performed on accuracy scores generated by the AIST. Bonferroni adjustment for multiple comparisons was applied as appropriate. To determine effects of other measurements on AIST performance, Pearson's correlation analyses were used. These analyses started with total AIST score (pooled over noise type, SNR, and MLL), then AIST performance in each noise type Frontiers in Psychology | Auditory Cognitive Neuroscience (pooled over SNR and MLL), AIST performance in each SNR (pooled over noise type and MLL), and AIST performance in each MLL (pooled over noise type and SNR), and then AIST performance in each SNR in each noise type (pooled over MLL). All statistic calculations were performed using IBM SPSS Statistics 22.
COGNITIVE TESTS
Mean performance on the RS was 16.2 (SD = 3.7, max = 28), and mean performance on the LM was 36 (SD = 5.2, max = 48), see Table 1. There was no statistically significant correlation between RS and LM scores (r = 0.25, p = 0.29).
SPEECH INTELLIGIBILITY
Speech intelligibility data collected in the previous study (Rönnberg et al., 2014) were reanalyzed in the current study. A repeated measures ANOVA with one within group variable, SNR (SNR1, SNR2, SNR3) showed a significant effect of SNR [F(2,38) = 27.5, p < 0.001, η 2 p = 0.59]. Post hoc test showed a significant decrease in speech intelligibility levels between SNR1 and SNR2 (p = 0.035), between SNR1 and SNR3 (p < 0.001), as well as between SNR2 and SNR3 (p < 0.001). Speech intelligibility data was not collected in this study and thus speech intelligibility levels for AMN as well as for ISTS are based on the equalization data obtained from 10 subjects prior to the current study.
AIST performance and reading span score
A significant positive correlation was found between total AIST performance and reading span score (r = 0.712, p < 0.001), showing that a higher reading span score was associated with better www.frontiersin.org general AIST performance (see Table 3). As shown in Table 3, reading span score also correlated positively with AIST performance in all three noise types, in all three SNRs, as well as with all three MLLs. More specifically in SSN, reading span score correlated with AIST performance in SNR1. In the modulated noise types (AMN and ISTS), reading span score correlated with AIST performance in SNR2 as well as SNR3.
AIST performance and letter memory score
Letter memory score did not significantly correlate with total AIST performance (see Table 3). The only significant correlation between Letter memory score and AIST performance was found between Letter memory score and AIST performance in SNR1 (r = 0.495, p < 0.05). As shown in Table 3, Letter memory score correlated with AIST performance in SNR1 for the modulated noise types (AMN and ISTS).
Sentence questions
When SQ performance was pooled over SNRs the mean score was 26.8 (SD = 0.4) in SSN, in AMN the mean score was 26.8 (SD = 0.5), and in ISTS it was 25.7 (SD = 1.4), maximum score was 27, see Table 4 and Figure 4A. A repeated measures ANOVA with two within group variables, noise type (SSN, AMN, ISTS) and SNR (SNR1, SNR2, SNR3), showed a significant effect of noise type [F(2,38) = 12.79, p < 0.001, η 2 p = 0.40], but there was only a tendency toward significant effect of SNR [F(2,38) = 2.59, p = 0.088, η 2 p = 0.12]. Post hoc tests revealed a significantly better SQ performance in SSN than in ISTS (p = 0.006), as well as in AMN compared to in ISTS (p = 0.004), but there was no significant difference in SQ performance between SSN and AMN. A significant two-way interaction between noise type and SNR was found [F(4,76) = 2.96, p = 0.025, η 2 p = 0.14]. Analyses of simple main effects revealed significant differences in SQ performance between SNRs for ISTS [F(2,38) = 3.35, p = 0.046, η 2 p = 0.15], but only a tendency toward significant effect for SSN [F(2,38) = 2.84, p = 0.071, η 2 p = 0.13] and no effect for AMN. Post hoc tests showed a significant decrease in SQ performance in ISTS between SNR1 and SNR3 (p = 0.047), as well as a tendency toward significant difference between SNR1 and SNR2 (p = 0.074), but there was no significant difference between SNR2 and SNR3. Performance on SQs did not significantly correlate with WMC or with UA.
When response times, see Table 4 and Figure 4B, was assessed in a repeated measures ANOVA with two within group variables, noise type (SSN, AMN, ISTS), SNR (SNR1, SNR2, SNR3), a significant effect of noise type [F(2,38) = 5.48, p = 0.008, η 2 p = 0.23] was revealed as well as a significant effect of SNR [F(2,38) = 5.94, p = 0.006, η 2 p = 0.24]. Post hoc tests showed a significant increase in response time between SSN and ISTS (p = 0.045), but there were no significant differences between SSN and AMN, or between AMN and ISTS. Post hoc tests also showed a significant increase in response time between SNR1 and SNR3 (p = 0.010), but there were no significant differences between SNR1 and SNR2, or between FIGURE 3 | (A) Mean AIST performance in each noise type pooled over SNRs and MLLs. The maximum score was 27. Chance level was at 9. (B) Mean AIST performance for each MLL pooled over noise types and SNRs. The maximum score was 27. Chance level was at 9. (C) Mean AIST performance in each noise type and in each SNR pooled over MLLs. The maximum score was 9. Chance level was at 3.
SPEECH INTELLIGIBILITY
Speech intelligibility levels in SSN in the present study were identified in a larger study cohort (Rönnberg et al., 2014). The speech intelligibility levels in AMN and ISTS were matched to the speech intelligibility levels in SSN prior to the study to provide equal intelligibility levels between noise types. Even though performance on SQ is not a measure of speech intelligibility, it is nevertheless an indication of how well the participant has heard the sentence. The accuracy on SQs supported the estimated speech intelligibility levels used.
Noise types
It was hypothesized that the average AIST performance would differ between noise types, even though mean speech intelligibility levels were held constant. The poorest AIST performance was expected to be found in SSN, while the best AIST performance was expected to be found in AMN. However, contrary to expectations there were no statistical significant differences in memory performance between the noise types (see Figure 3C). Mishra et al. (2013a) showed an increased cognitive spare capacity, as measured by improved memory performance, in ISTS compared to SSN, using lists of numbers between 13 and 99 as targets. This was not the case in the present study. The reason for this might be that the vocal sounds and speech fragments add an additional informational masking interfering more with the speech information in the sentences compared to the numbers used by Mishra et al. (2013a). This in turn would add more demands on the cognitive system leading to less cognitive spare capacity. The AMN contains short periods with less noise which might make it possible to achieve the same speech intelligibility level as for SSN but with less cognitive demands (Duquesnoy, 1983), but there was no statistical significant improved memory performance in AMN compared to SSN or ISTS (see Figure 3C). This suggests that for young adults with normal hearing, in SNRs targeting 90% speech intelligibility or better, the type of noise is not of importance for memory performance of the information in the sentences.
Signal-to-noise ratio
Speech intelligibility levels were matched between all noise types at SNR1, as well as at SNR2 and at SNR3 (see Figure 2). Therefore, the amount of amplitude change of the noise between SNR1 and www.frontiersin.org
FIGURE 4 | (A)
Mean sentence question (SQ) performance for SNR type in each noise type. The maximum score was 9. Chance level was at 3. (B) Mean response time (in seconds) for SQ questions for each SNR in each noise type.
SNR2, as well as between SNR2 and SNR3, differed between noise types, i.e., SNR1 was different in different noise types but corresponded to the same speech intelligibility level (see Figure 2). Access to the information in the sentences is essential for accurate AIST performance. Since all SNRs gave a mean speech intelligibility level of 90% or better, access to the information was not appreciably limited at any of the SNRs (see Figure 2). Based on the previous study (Rönnberg et al., 2014), we hypothesized that a decreased SNR would force an increase in cognitive processing of auditory information, leading to less cognitive spare capacity resulting in reduced AIST performance. The tendency toward a statistically significant effect of SNR on AIST performance (see Tables 1 and 2; Figure 3C) suggested that the cognitive spare capacity, as measured by memory performance on AIST, was reduced by increasing noise level. Similar results have also been found in other studies (Mishra et al., 2013a;Ng et al., 2013a,b;Rönnberg et al., 2014). However, in the present study, increasing noise level only reduced AIST performance when ISTS was used as background noise. This suggests that increasing background noise at the high intelligibility levels used in the present study only influences listening effort when noise is speech-like (see Figure 3C).
When listening in AMN, young adults with normal hearing are likely to be able to utilize the short periods with increased SNR to infer information that is masked when the noise level is louder (Duquesnoy, 1983) which would give rise to release from masking (Festen and Plomp, 1990). As a result, the decrease in SNR for AMN might not be particularly more demanding when listening in SNRs targeting 90% speech intelligibility or better. Nevertheless, for ISTS, the noise level seemed to have an impact on the cognitive processes involved leading to less cognitive spare capacity and decreased memory performance on AIST (see Tables 1 and 2; Figure 3C). Even if the ISTS is largely non-intelligible (Holube et al., 2010), the voices and speech fragments in ISTS may promote informational masking (Francart et al., 2011) which would add to the cognitive load since ISTS will interfere with the Hagerman sentences at different linguistic levels (Tun et al., 2002;Brouwer et al., 2012). Consequently, since ISTS adds more cognitive load, AIST performance in ISTS is more sensitive to decreased SNR than in the other noise types. As a result, the decrease in AIST performance with worse SNR in ISTS cannot be explained by reduced intelligibility alone since SNR did not significantly affect AIST performance in SSN or in AMN.
Interestingly, the correlations with WMC, i.e., reading span score, indicated that WMC had an impact on performance in AIST when presentation took place in SSN with SNR1, but not with the other SNRs (see Table 3). A reason for this might be that SSN masks the signal at worse SNRs, and when the signal becomes inaudible, a greater WMC does not improve speech intelligibility. On the other hand, when SNR is better and the signal is only partly masked by the SSN, a greater WMC might facilitate speech intelligibility by storing partly heard sounds of the speech signal until these can be disambiguated. The relation between speech recognition in noise and WMC is more evident in modulated noise where individuals with high WMC have better speech recognition in noise performance compared to individuals with less WMC George et al., 2007;Zekveld et al., 2013), which might also explain the relation between WMC and AIST performance in SSN. For the modulated noise, WMC was of importance for memory performance when the SNR was more demanding (see Table 3). This suggests that when listening takes place in more troublesome listening conditions, such as increased SNR and modulated noise, WMC is more occupied with listening, and individuals with higher cognitive capacity are likely to have more cognitive spare capacity after listening and consequently perform better on the memory task than individuals with less cognitive capacity. Consequently, individuals with greater cognitive capacity will probably experience less listening effort than individuals with less cognitive capacity. On the other hand, when listening takes place in modulated noise in SNR1, the listening condition might be described as fairly simple which explains why, the extra WMC capacity did not add an additional advantage.
Another way to explain the correlations between AIST performance and WMC is based on attention. One may expect that a person with a higher WMC is better to filter out the desired signal (speech) and suppress the unwanted signal (noise) than a person with worse WMC. There are indications of such mechanisms in the literature. In an auditory brainstem response measurement it was found that the neural amplitude increased when focusing on the signal and decreased when adding a cognitive load (distractor; Sorqvist et al., 2012). This modulation of the neural response was correlated with the persons WMC. Other studies have indicated that attention and WMC correlates with spatial speech recognition performance in adults (Neher et al., 2011) and that attention supports language processing in children (Astheimer et al., 2014). However, there are other studies that have found correlation between WMC and speech perception that is unrelated to attention skills (Tamati et al., 2013). The current study did not measure attention per se, but it is very plausible that a better WMC facilitated auditory attentional filtering of the sentence and thereby improved both speech recognition and ability to store the information crucial for AIST performance.
Updating ability, i.e., Letter memory score, did not correlate with total AIST performance (see Table 3). However, having a greater UA improved AIST performance in SNR1, more specifically for SNR1 in the modulated noise types (AMN and ISTS) but not in SSN. This is consistent with the previous study where no interactions were found between AIST performance and SNRs when UA was used as a between-group variable and SSN was used as masker (Rönnberg et al., 2014). In the modulated noise types, at the best SNR, listening is fairly undemanding why having a higher UA facilitates performance on AIST. However, when the SNR gets worse, there was no effect of UA on AIST performance. Nevertheless, there was an effect of WMC on AIST performance in worse SNRs suggesting that in more troublesome listening conditions WMC is of more importance for listening than UA. WMC improves memory performance in SSN in the easiest SNR, but UA does not improve memory performance. However, in modulated noise, WMC facilitates memory performance in the worst SNR, while UA improves memory performance in the best SNR.
Memory load level
Auditory Inference Span Test accuracy was, as expected, a function of MLL (see Table 1; Figure 3B), where performance decreased with increasing level of memory load (Mishra et al., 2013a,b;Rönnberg et al., 2014). As in the previous study (Rönnberg et al., 2014), there were no significant difference in performance on MLL2 and MLL3. Even though performance at MLL2 and MLL3 is low, performance on both MLLs are clearly above chance level. The results suggested that regardless of MLL, WMC improves memory performance on AIST. A similar effect was found in a previous study (Rönnberg et al., 2014). Also, in the previous study (Rönnberg et al., 2014) an interaction between MLL and UA showed a benefit of high UA on questions demanding more updating of information, i.e., MLL 2. This relation was not found to be significant in the present study (see Table 3).
Response time
Response times on MLL questions were registered in the AIST process. These response times on MLL questions were not included in the analyses. The reason for this was that the measure of response time started when the question was presented on the computer screen and continued until an answer had been given, and the test had continued to the next question. Consequently, the time it took to read and comprehend the question was part of the measured response time. However, there is a difference in the complexity of the questions, why differences in response time might be due to differences in the amount of time it took to read and comprehend the question. Nevertheless, response times on MLL questions might be analyzed when pooled over the three MLLs. It was expected that response times then would be dependent on SNRs and noise types. However, no statistically significant effect of SNR or of noise type was not found. Pooled response times on MLL questions did not change with listening conditions. Consequently, response time on AIST was not deemed to be a useful measure.
SENTENCE QUESTIONS
Performance on SQs decreased in ISTS compared to SSN and AMN, and there was an effect of SNR in ISTS but not in SSN or AMN, see Figure 4A. Since SQ might be considered a measure of speech recognition in the sense that the question probes that the sentence was heard, even if the three-choice procedure facilitates performance by giving possible answer alternatives as well as having a chance level of 33%, the results suggested that the general speech intelligibility levels were at the expected levels above 91% (Rönnberg et al., 2014). However, the effect of SNR only found in ISTS might suggest that speech intelligibility levels were not perfectly matched between noise types. Nevertheless, the results might also imply that speech-shaped noise in these rather favorable SNRs did not load the cognitive system to such a degree as the vocal sounds and speech fragments in ISTS did, and consequently there was no effect of SNRs for SSN and AMN on SQ accuracy. Even if ISTS is largely non-intelligible (Holube et al., 2010), it may cause additional informational masking (Francart et al., 2011) and consequently add to the cognitive load since the masker interferes with the speech material at different linguistic levels (Tun et al., 2002;Brouwer et al., 2012).
The analyses of SQ response times were based on response times correct answers as well as for incorrect answers, as there was no statistically significant difference in response time between correct and incorrect answers. Response time on SQs was an effect of noise type, with longer response times in ISTS compared to www.frontiersin.org SSN and AMN. There was also an effect of SNR with increasing response times in SNR3 compared to SNR1, see Figure 4B. The results suggest that more processing was needed in the more problematic listening conditions (in ISTS compared to SSN, and in SNR3 compared to SNR1) and that this processing takes longer, with longer response times as a result. It seems likely to assume that the longer response time is a measure of listening effort. SQ response time correlated with WMC and not with UA. Contrary to expectations that having a greater WMC would imply faster access time to information stored in working memory and a shorter time to retrieve the position of the correct answer alternative, instead the results showed that greater WMC rather meant longer response times. The results suggested that individuals with greater WMC spent more time reading the answer alternatives and pondering the answer; however, they did not gain from this extra time spent when considering accuracy on SQ questions. Also, having a higher WMC implies having more information held in working memory, resulting in more information to scan which would require a longer time to find the matching answer.
THE COGNITIVE MEASUREMENTS
Both the RS and the LM are delivered in visual modality, unlike the AIST which is delivered in auditory modality with visually presented multiple choice responses. This is a strength of the study, since the measurements of WMC and of UA are independent on the individual's hearing status. Furthermore, the AIST is intended to be used in the hearing aid fitting process to assess listening effort, then it is of even greater importance that the measurement of the individual's cognitive capacity is unaffected by the hearing status.
CLINICAL IMPLICATION
Performance on AIST can be expected to be lower for individuals with hearing impairment than for individuals with normal hearing. A hearing impairment decreases the signal fidelity (Plomp, 1978;Pichora-Fuller and Singh, 2006), which in turn increases the cognitive involvement in listening and consequently leaves less cognitive capacity for memory storage (Rudner et al., 2011b;Picou et al., 2013) which would be measurable with the AIST. It is well established that successful hearing aid fitting needs to take individual differences in cognitive capacity into account . Hitherto, cognitive measures such as reading span have been used to demonstrate associations with ability to repeat and recall speech. The advantage of a test such as AIST is that it has the potential to measure the listening effort expended by the individual under different sets of listening conditions in which noise types, SNR and potentially hearing aid settings can be manipulated. This will allow better hearing aid fitting in the future and provides an important tool for the development of better hearing aids.
CONCLUSION
The results suggest that for young adults with normal hearing the cognitive spare capacity is reduced when background noise consists of voices and the SNR decreases. However, when speech intelligibility levels are kept constant, different masker types do not have different effects on cognitive spare capacity, at least not for intelligibility levels above 90%. | 2016-05-12T22:15:10.714Z | 2014-12-22T00:00:00.000 | {
"year": 2014,
"sha1": "95b6b7cb6eb56ada05599199dad6e5aef9c78430",
"oa_license": "CCBY",
"oa_url": "https://www.frontiersin.org/articles/10.3389/fpsyg.2014.01490/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "95b6b7cb6eb56ada05599199dad6e5aef9c78430",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Psychology",
"Medicine"
]
} |
14872485 | pes2o/s2orc | v3-fos-license | Supernova 1993J: a veiled pulsar in a binary system?
The recent report of a binary companion to the progenitor of supernova 1993J may provide important clues for identifying the nature of the nascent compact object. Given the estimates of the progenitor mass, the potential power source is probably a pulsar rather than an accreting black hole. If there is a pulsar, one would expect the rotational luminosity to be stored and reprocessed by the supernova remnant, but no pulsar nebula has yet been seen. The lack of detection of an X-ray synchrotron nebula should be taken as strong evidence against the presence of a bright pulsar. This is because absorption by the surrounding supernova gas should be negligible for the light of the companion star to have been detected. A model is developed here for the luminosity of the pulsar nebula in SN 1993J, which is then used to predict the spin-down power of the putative pulsar. If one exists, it can be providing no more than about 6 x 10^{39} erg/s. With an initial rotation period of 10-30 ms, as extrapolated from young galactic pulsars, the nascent neutron star can have either a weak magnetic field, B<10^{11} G, or one so strong, B>10^{15} G, that its spin was rapidly slowed down. The companion star, if bound to the neutron star, should provide ample targets for the pulsar wind to interact and produce high-energy gamma-rays. The expected non-pulsed, GeV signal is calculated; it could be detected by current and future experiments provided that the pulsar wind velocity is similar to that of the Crab Nebula.
INTRODUCTION
Supernova 1993J was discovered on 28 March 1993 in the spiral galaxy M81 (at a distance of 3.63 Mpc; Freedman et al. 1994). Although it is factor of 65 further away than SN 1987A, it is relatively close for a SN and has been the subject of intense observational campaigns in a number of wavelengths regions (Chevalier 1997, and references therein). Shortly after the explosion, its spectrum was found to contain hydrogen lines -a type II supernova -but within a few weeks strong helium lines developed. The spectrum looked then more like a type Ib supernova, as though the envelope of hydrogen had lifted to reveal a helium layer (Filippenko et al. 1993). This, together with the observed peculiarities observed in the lightcurve, led many to conclude that the dying star, identified as a non-variable red supergiant in images taken before the explosion (Aldering et al. 1994), must have lost a considerable amount of its outer envelope of hydrogen gas to a companion before it exploded (Nomoto et al. 1993;Podsiadlowski et al. 1993;Woosley et al. 1994). But no companion had been seen until recently when the brightness of SN 1993J dimmed sufficiently that its spectrum showed the features of a massive star superimposed on the supernova (Maund et al. 2004).
Given the estimates of the progenitor mass, the remnant is probably a neutron star rather than a black hole (Nomoto et al. 1993;Podsiadlowski et al. 1993;Woosley et al. 1994), and the neutron star is generally expected to manifest as a pulsar. The newly born neutron star is characterized by its initial rotation rate and magnetic dipole moment. It will spin down, generating radiation and accelerating charged particles at the expense of its rotational energy. We expect the reprocessing of rotational energy to produce a bright flat-spectrum synchrotron nebula, but this soft emission has not been observed.
Here we developed a simple shocked wind model for the luminosity of the pulsar nebulae in SN 1993J (Section 2), which is then compared with the current luminosity limits in order to place bounds on the dipole emission of the nascent neutron star and its birth properties (Section 4). We also investigated the impact of the soft photon field radiation of the binary companion on the central pulsar wind (Section 3), along with the types of observation that would help to unambiguously demonstrate whether or not the system is disrupted as a result of the supernova. Our conclusions are discussed in Section 4.
A MODEL FOR THE PULSAR NEBULA IN SN 1993J
Neutron stars formed in core collapse SN explosions are thought to emit radiation as pulsars, the Crab (Rickett & Seiradakis 1982) being the canonical example. Pulsars are generally presumed to lose rotational energy by some combination of winds of relativistic particles and magnetic dipole radiation at the frequency of the neutron star's rotation. In the generic pulsar model, the field is assumed to maintain a steady value, and the luminosity declines as the spin rate slows down (Shapiro & Teukolsky 1983). The spindown law is then given by Ω(t) = Ω0(1 + t/tΩ) −1/2 , so that LΩ(t) = LΩ,0(1 + t/tΩ) −2 , where tΩ ≃ 10 3 s I45B −2 p,15 P 2 0,−3 R −6 6 is the characteristic time scale for dipolar spindown, Bp,15 = Bp/(10 15 G) is the dipolar field strength at the poles, R6 = R/10 6 cm is the radius of the light cylinder, P0,−3 is the initial rotation period in milliseconds, and LΩ,0 ≃ 10 49 erg s −1 B 2 p,15 P −4 0,−3 R 6 6 . The power output of the pulsar is typically assumed to be LΩ, regardless of the detailed process by which the dipole emission is converted to the energy of charged particle acceleration and high-energy electromagnetic radiation. The pulses themselves, even when seen in the gamma-rays, constitute an insignificant fraction of the total energy loss. Large fraction of the power output, on the other hand, goes into a bubble of relativistic particles and magnetic field surrounding the pulsar (Chevalier & Fransson 1992). This bubble gains energy from the pulsar, and loses it in synchrotron radiation losses and in doing work on the surrounding supernova gas, sweeping it up and accelerating it.
Hydrodynamic Evolution
The density structure of the supernova gas is determined by the initial stellar structure, as modified by the explosion (Arnett 1988). The density structure has been relatively well determined for models of SN 1987A -it shows an inner flat profile outside of which is a steep power-law decrease with radius (e.g. Shigeyama & Nomoto 1990). Chevalier & Soker (1989) approximated the velocity profile by an inner section ρi = Ar −m t m−3 and an outer section ρ0 = Br −n t n−3 , where m = 1 and n = 9 for SN 1987A. The two segments of the density profile are assumed to be continuous and they intersect at a velocity vt. For an explosion with total energy E and mass Mt, one has where SN 1987A was an unusual type II supernova because of its small initial radius. This is in contrast to SN 1993J where the early lightcurve indicated that the progenitor star was more extended, about ten times larger than that of SN 1987A. The lack of a plateau phase in this case is attributed to the low mass of the hydrogen envelope, about 0.1 to 0.6 M⊙ (Nomoto et al. 1993;Podsiadlowski et al. 1993;Woosley et al. 1994), although the initial stellar mass was probably ∼ 15M⊙. Woosley et al. (1994), for example, estimated that the final presupernova star 1 had a helium and heavy element core of ∼ 3.71M⊙, a low density hydrogen envelope of 0.2M⊙, and a radius of about 4.3 × 10 13 cm.
The core collapse of the progenitor of 1993J resulted in the deposition of about 3 × 10 50 erg of kinetic energy per solar mass (Woosley et al. 1994). The expanded density profile shows an outer steep power-law region with n ∼ 9 and an inner, relatively flat region (Blinnikov et al. 1998 and Fig. 4 therein). The bend in the profile occurs at a velocity of ∼ 2000 km s −1 . The inner density profile is clearly not a single power law in r. However, it is likely that hydrodynamic instabilities, similar to those believed to occur in SN 1987A (e.g. Fryxell et al. 1991), lead to mixing and smooth out any sharp features in the density profile. Thus, a reasonable smooth, flat inner profile may be a good approximation. In the outer parts of the density structure, on the other hand, radiative transfer effects are more important for the explosion of an extended star and can lead to the formation of a dense shell in the outer layers. The shell is, however, expected out in the steep power-law region of the density profile, where the pulsar bubble is not likely to reach. In what follows, we assume m = 1 or 2 and n = 9.
The first stage of evolution involves the interaction of the pulsar bubble with the inner density section. With the assumptions that the pulsar luminosity is constant during this phase (i.e. tage ≤ tΩ) and the supernova gas is swept up into a thin shell of mass M and velocity v, the radius of the pulsar bubble can be written as (Chevalier & Fransson 1992) km s −1 for m = [1, 2], so it is likely that the shell remains within the inner part of the density profile.
The density of the uniform gas shell is The presence of the companion implies that the supernova gas should be transparent to optical radiation (i.e. the optical depth to electron scattering should be less than unity), which gives E > 10 49 M 2 t,5 t −2 age,11 for m = 1 and LΩ ≥ 10 40 E −2 51 M −9/2 t,5 t −7 age,11 erg s −1 for m = 2. For X-ray energies ≥ 10 keV, electron scattering also provides the main opacity. However, in this case the bound electrons contribute as well as the free ones, so that the optical depth is always given by electron scattering irrespectively of the degree of ionization. At lower X-ray energies ǫ ≥ 3 keV, photoionization provides additional opacity (Bahcall et al. 1970) and may delay the time at which the envelope becomes transparent by an additional factor ∼ 5(ǫ/1keV) −3/2 (this estimate is for a predominantly neutral envelope of pure hydrogen). The delay is somewhat larger if the envelope is rich in heavy elements. The matter that is swept up by the pulsar bubble is subject to Rayleigh-Taylor instability and may form filaments (Vishniac 1983). The optical depth to electron scattering could then be much less if the envelope is sufficiently irregular that some lines of sight to the pulsar traversed relatively little envelope mass. Under such favourable circumstances, the opacity of the envelope is a less serious problem (except at late times when grain formation may occur in the supernova ejecta and dust absorption could significantly increase the opacity at visual wavelengths).
Radio measurements suggest that any pulsar nebula in the center of SN 1993J is fainter than ≈ 10 38 erg s −1 (Bietenholz et al. 2003). The material immediately surrounding the putative pulsar is, however, expected to be totally opaque until Mt,5T at which time the radio flux (ν9 = ν/1 GHz) will suddenly appear. Here T4 = T /10 4 K is the temperature of the swept-up material. The lack of detection of a synchrotron nebula at radio wavelengths should not be taken as strong evidence against the presence of a bright pulsar (Bahcall et al. 1970). On the other hand, the lack of an X-ray nebula indicates that, if one exists, its radiative power must be less than LX ≈ 2 × 10 38 erg s −1 (Zimmermann & Aschenbach 2003).
Emission Model
A compelling model for the optical/X-ray properties of the Crab Nebula was developed by Rees & Gunn (1974). In this model, the central pulsar generates a highly relativistic, particle dominated wind (with Lorentz factor γw) that passes through a shock front and decelerates to match the expansion velocity set by the outer nebula. The emission from the pulsar bubble is thought to provide a larger luminosity source than the radiative shock front itself (Chevalier & Fransson 1992). The wind particles acquire a power-law energy spectrum of the form N (γ) ∝ γ −p (for γ ≥ γm, where γ is the particle Lorentz factor and γm is its minimum value) in the shock front and radiative synchrotron emission in the down stream region (Chevalier 2000). Under the basic simplification that the emitting region can be treated as a one zone (i.e. no spatial structure in the nebula), and assuming that a balance between injection from the shock front and synchrotron losses is established, the number of radiating particles at a particular γ is given by N (γ) = γ p−1 m (γwβB 2 ) −1 LΩγ −(p+1) (Chevalier 2000), where B is the magnetic field in the emitting region, and β = 1.06 × 10 −15 cm 3 s −1 .
In what follows, it is assumed that the energy density in the emitting region (which is approximately determined by the shock jump conditions) is divided between a fraction ǫe in particles and a fraction ǫB in the magnetic field. These efficiency factors are constrained by ǫe + ǫB = 1. With this assumption, the magnetic field in the emitting region is B = (6ǫB LΩ/r 2 s c) 1/2 , where rs is the shock wave radius. The electron energy is radiated at its critical frequency ν(γ) = γ 2 (qeB/2πmec), where qe and me are the electron charge and mass, respectively. If the electrons and positrons cool rapidly by synchrotron radiation, as thought to be the case for the Crab Nebula, the luminosity produced from the pulsar power is given by (Chevalier 2000), where φ(p) = 1 2 ( p−2 p−1 ) p−1 ( 6ψ 2 c ) (p−2)/4 and ψ = 2.8 × 10 6 in cgs units. If the particle spectrum is similar to that of the Crab Nebula (p = 2.2), the X-ray luminosity (ν = 10 18 Hz) becomes where cgs units are used. We note that the X-rays from a putative pulsar nebula in SN 1993J with LΩ ∼ 10 39 erg s −1 could provide the additional energy input required to reproduce the observed Hα luminosity at ∼350 days (Houck & Fransson 1996). The above relation can be used to predict the spin-down power, LΩ, of the putative pulsar in wind nebula where a pulsar has not yet been observed (Chevalier 2000). In the case of SN 1993J, the value rs can be determined by the condition that the envelope should be transparent to optical radiation (Section 2.1). The need for a particle dominated shock suggests that ǫe ≥ 0.5. We set ǫB = ǫe = 0.5, although there is little dependence to ǫB. The value of γw is difficult to constraint. Here we use γw ∼ 3 × 10 6 , the value typically assumed for the Crab Nebula (Kennel & Coroniti 1984). A better estimate of γw is given in Section 3. When these parameters are substituted into equation (5), the XMM upper limit of LX ≈ 2 × 10 38 erg s −1 (0.3-10 keV; Zimmermann & Aschenbach 2003) yields LΩ ≤ 2 × 10 39 erg s −1 . On the basis of the ASCA data, Kawai et al. (1998) found the relation log LX = (33.42 ± 0.20) + (1.27 ± 0.17) log(LΩ/10 36 ), where LX is the nebular luminosity in the 1-10 keV range, from which we estimate LΩ ≤ 5 × 10 39 erg s −1 . This estimate is in fair agreement with our predicted value.
THE ROLE OF BINARITY
At the time of the explosion, the progenitor of 1993J has a mass of 5.4 M⊙ (with a helium-exhausted core of 5.1 M⊙), the secondary has a mass of 22 M⊙. The orbital period of the system is ∼ 25 yr, and the companion has an orbital velocity of ∼ 6 km s −1 (Maund et al. 2004). There may have been a large kick imparted by the explosion mechanism, and that would be very interesting to study, but baring that, let us assume the companion star is bound in an eccentric orbit with the newly born pulsar 2 . The bulk of the pulsar energy LΩ would be primarily in the form of a magnetically driven, highly-relativistic wind consisting of e − , e + and probably heavy ions with Lw ≃ ζLΩ and ζ ≤ 1.
Under the foregoing conditions, the relativistic wind (which is likely to be undisturbed by the presence of the binary companion 3 since LΩ ≫ v 2 wṀw ) would escape the compact remnant while interacting with the soft photon field of the companion with typical energy θ * = kT * /(mec 2 ) ∼ 3 × 10 −6 (Maund et al. 2004). The scattered photons whose energy is boosted by the square of the bulk Lorentz factor of the magnetized wind (i.e. Θ * = 2γ 2 w θ * ∼ 6 × 10 6 γ 2 w,6 ) propagate in a narrow γ −1 w beam owing to relativistic aberration. The rate of energy loss of a relativistic particle moving in a radiation field with an energy density w * ≃ L * /4πa 2 c is about mec 2 dγ/dt ≈ −w * σT cγ 2 in the Thompson limit (Landau & Lifshitz 1975). Here L * ∼ 10 5 L⊙ is the luminosity of the optical star and a is the binary separation (a ∼ 25 AU ∼ 200R * ; Maund et al. 2004). The total luminosity of scattered hard photons LΓ is then equal to the total particle energy losses in the course of motion from the pulsar to infinity LΓ = Lw ∆γ denotes the efficiency in extracting energy from the relativistic outflow (Chernyakova & Illarionov 1999). The resulting radiation pressure on electrons in the ejecta will brake any outflow whose initial Lorentz factor exceeds some critical value γ lim ≤ a R * ( L Ω L * ) 1/2 ∼ 3 × 10 2 L 1/2 Ω,39 , converting the excess kinetic energy into a directed beamed of scattered photons. With γw ∼ γ lim , equation (5) yields LΩ ≤ 6×10 39 erg s −1 .
As the stellar companion emits a black body spectrum, of effective temperature θ * , the local photon energy density is given by where ̟ is the soft photon energy in units of mec 2 . The scattered photons are boosted by the square of the Lorentz factor so that the local spectrum has a black body shape enhanced by γ 2 w . As can be seen in Fig. 1, the resulting spectrum is the convolution of all the locally emitted spectra (i.e. ∞ 0 dn̟[r, ̟]) and it is not one of a blackbody. Note that Klein-Nishina effects are important for incoming photon energies such that θ * γw > ∼ 1. The maximum energy of the scattered photons in this regime is γwmec 2 . We note that the total luminosity emitted by the relativistic outflowing wind through the Compton-drag process in the direction of the observer could be highly anisotropic and may change periodically during orbital motion for an eccentric (or a highly inclined) orbit. The time dependence, in this case, will be very distinctive and such effects should certainly be looked for.
DISCUSSION
Despite presumptions that a neutron star may have been created when the progenitor star of SN 1993J exploded and its core collapsed, no pulsar has yet been seen. If one exists, its radiative power must be less than LΩ ≈ 6 × 10 39 erg s −1 or lower if the pulsar nebula has a high radiative efficiency. The well-known Crab pulsar and its nebula for comparison, put out 2 × 10 38 erg s −1 ; and originally, when it was spinning faster, the luminosity might have been seven times greater still. Fig. 2 shows the pulsar spin-down luminosity divided by 4π times the square of the distance, the total pulsar energy output at Earth. The ten gamma-ray pulsars (including candidates) are shown as large circles. The solid curves show the dipole emission of the putative pulsar in 1993J (at t ∼ 11 years) for various assumptions regarding its initial period and magnetic field strength. The pulsar spinning down by magnetic dipole radiation alone, with initial rotation periods of 10-30 ms -as extrapolated for galactic young pulsars -can have a spin-down luminosity below ∼ 6 × 10 39 erg s −1 (see shaded region in Fig. 2) only if either the magnetic field is relatively weak ≤ 10 11 G or if it is so strong (i.e. ≥ 10 15 G) that the pulsar luminosity decays rapidly. If weak magnetic moments could be ruled out in the near future by X-ray and infrared observations, then we find that, if undetected, the putative pulsar in SN 1993J could be a magnetar.
With a spectrum similar to that of the Crab (see Fig. 2), the pulsed emission is likely to be below the detection limit of current instruments and may deprive us of the opportunity to witness the emergence of a gamma-ray pulsar in the immediate future. GeV radiation with luminosities high enough to be detected with ARGO 4 , and the Veritas 5 experiment now under construction, could be produced by the interaction of the pulsar wind with the soft photon field of its companion provided that the system remains bound and γw ≥ 10 6 (i.e. similar or larger to than inferred for the Crab pulsar). Its detection will surely offer important clues for identifying the nature of the progenitor and possibly constraining whether or not kicks have played an important role.
If there is not a neutron star in SN 1993J, could there be a black hole instead? Had matter fallen back onto the nascent neutron star (with mass of 1.4 M⊙) on a timescale of 100 seconds to a few hours after the explosion, enough mass may have been accreted to push the object over the minimum thought to be necessary for the creation of a black hole. Matter would continue to accrete onto the black hole, but the resulting radiation would be trapped in the outflow so that the escaping luminosity would be small. Because the cores of massive stars increase with initial stellar mass, this picture would be more plausible if the initial mass of the SN 1993J progenitor star, ∼15 M⊙, were close to the mass limit above which stars collapse directly to black holes (e.g. Fryer 1999).
Even if the neutron star is rotating only slowly or has a weak magnetic field, one would expect surrounding material to fall back onto it, giving rise to a luminosity < ≃ L Edd ∼ 3 × 10 38 erg s −1 (see Fig. 1). The remaining possibility of a weak pulsar with little surrounding mass will be difficult to rule out. The thermal emission from a newly formed neutron star gives a luminosity of ∼ 3 × 10 34 erg s −1 . The neutron-star emission should have a characteristic soft X-ray emission, but its detection will be arduous. If future observations fail to identify a pulsar in 1993J, then the youngest pulsar that we know will still be PSR J0205+6449 (Camilo et al. 2002), associated with SN 1181. . Pulsar observability as measured by the spin-down energy seen at Earth. Small dots represent pulsars with no gamma-ray emission, while large filled symbols show all known gamma-ray pulsars (including candidates; Thompson 2003). The power-output of the pulsar wind, at the distance and age (here assumed to be 11 years) of SN 1993J, is shown for B = 10 11 , 10 12 , 10 13 , 10 14 and 10 15 G (from left to right) as a function of the assumed initial period. The putative pulsar in SN 1993J, spinning down by magnetic dipole radiation alone, can have a wind nebula X-ray luminosity below the XMM limit if L Ω ≤ 6 × 10 39 erg s −1 (shaded region). | 2014-10-01T00:00:00.000Z | 2004-07-02T00:00:00.000 | {
"year": 2004,
"sha1": "e7a79062d83f625a09e03f5e326dbc113ac9880d",
"oa_license": null,
"oa_url": "https://academic.oup.com/mnras/article-pdf/353/1/L7/18662438/353-1-L7.pdf",
"oa_status": "BRONZE",
"pdf_src": "Arxiv",
"pdf_hash": "e7a79062d83f625a09e03f5e326dbc113ac9880d",
"s2fieldsofstudy": [
"Physics"
],
"extfieldsofstudy": [
"Physics"
]
} |
253543895 | pes2o/s2orc | v3-fos-license | Comprehensive Analysis of Phaseolus vulgaris SnRK Gene Family and Their Expression during Rhizobial and Mycorrhizal Symbiosis
Sucrose non-fermentation-related protein kinase 1 (SnRK1) a Ser/Thr protein kinase, is known to play a crucial role in plants during biotic and abiotic stress responses by activating protein phosphorylation pathways. SnRK1 and some members of the plant-specific SnRK2 and SnRK3 sub-families have been studied in different plant species. However, a comprehensive study of the SnRK gene family in Phaseolus vulgaris is not available. Symbiotic associations of P. vulgaris with Rhizobium and/or mycorrhizae are crucial for the growth and productivity of the crop. In the present study, we identified PvSnRK genes and analysed their expression in response to the presence of the symbiont. A total of 42 PvSnRK genes were identified in P. vulgaris and annotated by comparing their sequence homology to Arabidopsis SnRK genes. Phylogenetic analysis classified the three sub-families into individual clades, and PvSnRK3 was subdivided into two groups. Chromosome localization analysis showed an uneven distribution of PvSnRK genes on 10 of the 11 chromosomes. Gene structural analysis revealed great variation in intron number in the PvSnRK3 sub-family, and motif composition is specific and highly conserved in each sub-family of PvSnRKs. Analysis of cis-acting elements suggested that PvSnRK genes respond to hormones, symbiosis and other abiotic stresses. Furthermore, expression data from databases and transcriptomic analyses revealed differential expression patterns for PvSnRK genes under symbiotic conditions. Finally, an in situ gene interaction network of the PvSnRK gene family with symbiosis-related genes showed direct and indirect interactions. Taken together, the present study contributes fundamental information for a better understanding of the role of the PvSnRK gene family not only in symbiosis but also in other biotic and abiotic interactions in P. vulgaris.
A better understanding of the signalling pathways that affect the productivity and sustainability of such crop systems is particularly important. A gene family such as SnRK with diverse regulatory mechanisms might play a pivotal role in biotic and abiotic interactions of legumes, and greater knowledge of these aspects may help in crop improvement. In the present study, we sought to identify and understand the diversity of the SnRK gene family in P. vulgaris and elucidate their differential expression patterns under rhizobial and mycorrhizal symbiotic conditions.
Phylogenetic Analysis of PvSnRK Family Genes
Multiple sequence alignment of PvSnRKs and AtSnRKs was performed using ClustalW. Based on the alignments, phylogenetic analysis of the aligned sequences was carried out using Molecular Evolutionary Genetics Analysis (MEGA XI) with the neighbour-joining (NJ) method and the JTT + I + G substitution model with 1000 bootstrap replicates and default parameters [52].
Conserved motifs in the PvSnRK gene family in P. vulgaris were identified using the Multiple Expectation Maximization for Motif Elicitation (MEME) online program (http://meme.sdsc.edu/meme/itro.html, accessed on 2 October 2021) with the following parameters: number of repetitions = any, maximum number of motifs = 20; and optimum motif length = 6 to 100 residues. The gene structure of the PvSnRK gene family was analysed using the Gene Structure Display Server online program (GSDS: http://gsds.cbi.pku.edu.ch, accessed on 10 October 2021) [53].
Sequences 2 kb upstream of PvSnRKs were downloaded from the Phytozome database. The plant transcriptional regulatory map (http://plantregmap.gao-lab.org/, accessed on 25 October 2021) was used to analyse the promoter sequences.
Calculation of Ka/Ks and Dating of Duplication Events
To identify putative orthologues between two different species (A and B), each sequence from species A was searched against all sequences from species B using BLASTN; each sequence from species B was also searched against all sequences from species A. The two sequences were defined as orthologues when reciprocal best hits were each within ≥300 bp of the two aligned sequences. The duplication period (Million Years ago, MYA) and divergence of each PvSnRK gene were calculated using the following formula: T = Ks/2λ (λ = 6.56 × 10 −9 ) [54]. The calculation of Ka and Ks was performed using TBtools. Graphs were developed with R studio with the R commander package [55].
Transcriptome Profiling and RT-qPCR Analysis
Data on differential expression of SnRK genes in P. vulgaris tissues under nitrogen treatments and after inoculation with Rhizobium tropici (CIAT899) were obtained from the PvGEA website (https://plantgrn.noble.org/PvGEA/, accessed on 12 January 2022). Previously, we performed global transcriptome profiling in P. vulgaris L. cv. Negro Jamapa roots colonized by Rhizophagus irregularis spores or Rhizobium tropici strain CIAT899 [57]. The present study uses the same transcriptomic data to obtain expression profiles of PvSnRK family genes under both types of symbiotic conditions. Heatmaps were constructed with fold-change values applying the R package (https://www.r-project.org/, accessed on 14 January 2022).
To validate the RNA-seq data, we surface-sterilized P. vulgaris L. cv. Negro Jamapa seeds and germinated them as described by Nanjareddy et al. [57]. Two-day-old germinated seedlings were transplanted into sterile vermiculite and inoculated with R. irregularis or R. tropici according to Nanjareddy et al. [58]. Total RNA preparation and RT-qPCR analysis were carried out according to Quezada et al., 2019 [59].
Identification of PvSnRK Protein Orthologues in P. vulgaris
A BLAST search was carried out using Arabidopsis SnRKs as a reference, and a total of 42 genes were identified based on conserved domains in each sub-family (Table S1). The sub-family PvSnRK1 was identified by the domain KA1 (PF02149), the PvSnRK2 sub-family by the OST domain [60], and the PvSnRK3 sub-family by the NAF/FISL domain (PF02149). The genes were named PvSnRK1.1, PvSnRK1.2, PvSnRK2.1-PvSnRK2.11 and PvSnRK3.1-PvSnRK3.29 based on sequence homology with the Arabidopsis SnRK family. The amino acid length of the 42 PvSnRK gene family members ranged from 310 aa (PvSnRK2.11) to 528 aa (PvSnRK1.2), corresponding to molecular weights of 60.54 to 35.67 kDa (Table 1). The theoretical isoelectric point of PvSnRKs (PI) ranged from 4.7 to 9.24, with PvSnRK1 sub-family members showing a basic PI, the PvSnRK2 sub-family being mostly acidic (4.7-6.65) and the PvSnRK3 sub-family being slightly acidic to highly basic (6.4-9.37). Subcellular localization analysis was carried out using ProtComp v.9.0, and the results showed PvSnRK1s to localize to the extracellular space; PvSnRK2s mostly localized to the nucleus and plasma membrane and PvSnRK3 sub-family members to the plasma membrane (Table 1). However, the subcellular localization analysis through different software showed some variation, as depicted in Table S3.
Phylogenetic and Structural Analyses
To determine the evolutionary relationship among Arabidopsis and Phaseolus SnRK superfamily genes, a phylogenetic tree was constructed using the protein sequences of 42 PvSnRK genes and 34 AtSnRK genes using the neighbour-joining (NJ) method with 1000 bootstrap replications (Figures 1 and S2). The accession numbers or locus IDs of the SnRK genes are listed in Table 1 and Supplementary Table S2. The resulting tree categorized the PvSnRKs and AtSnRKs into four clades, indicating that the ancestral genes of these two clades diverged before Brassicaceae and Fabaceae separated. In the P. vulgaris phylogeny, clade I contains PvSnRK2 sub-family members represented by the OST domain, clade II contains the PvSnRK1 sub-family containing the KA1 domain, and clades III and IV contain the NAF/FISL domain and 3 and 26 members of the PvSnRK3 sub-family, respectively ( Figure S1). The combined phylogenetic analysis of Phaseolus and Arabidopsis SnRK proteins also showed a similar distribution, whereby the proteins from two species appear scattered across the branches of the evolutionary tree, suggesting that they experienced duplications after the lineages diverged. The theoretical isoelectric point of PvSnRKs (PI) ranged from 4.7 to 9.24, with PvSnRK1 sub-family members showing a basic PI, the PvSnRK2 sub-family being mostly acidic (4.7-6.65) and the PvSnRK3 sub-family being slightly acidic to highly basic (6.4-9.37). Subcellular localization analysis was carried out using ProtComp v.9.0, and the results showed PvSnRK1s to localize to the extracellular space; PvSnRK2s mostly localized to the nucleus and plasma membrane and PvSnRK3 sub-family members to the plasma membrane (Table 1). However, the subcellular localization analysis through different software showed some variation, as depicted in Table S3.
Phylogenetic and Structural Analyses
To determine the evolutionary relationship among Arabidopsis and Phaseolus SnRK superfamily genes, a phylogenetic tree was constructed using the protein sequences of 42 PvSnRK genes and 34 AtSnRK genes using the neighbour-joining (NJ) method with 1000 bootstrap replications ( Figure 1, Figure S2). The accession numbers or locus IDs of the SnRK genes are listed in Table 1 and Supplementary Table S2. The resulting tree categorized the PvSnRKs and AtSnRKs into four clades, indicating that the ancestral genes of these two clades diverged before Brassicaceae and Fabaceae separated. In the P. vulgaris phylogeny, clade I contains PvSnRK2 sub-family members represented by the OST domain, clade II contains the PvSnRK1 sub-family containing the KA1 domain, and clades III and IV contain the NAF/FISL domain and 3 and 26 members of the PvSnRK3 sub-family, respectively ( Figure S1). The combined phylogenetic analysis of Phaseolus and Arabidopsis SnRK proteins also showed a similar distribution, whereby the proteins from two species appear scattered across the branches of the evolutionary tree, suggesting that they experienced duplications after the lineages diverged. To examine the evolution of Phaseolus SnRK genes, their chromosomal distribution was determined. PvSnRK superfamily genes are distributed across 10 pairs of homologous chromosomes among 11 pairs in the P. vulgaris genome. PvSnRK1 genes are located on chromosomes 4 and 8 and PvSnRK2 genes on chromosomes 2, 3, 6 and 8. PvSnRK3s are located on nine chromosomes, where chromosomes 1 and 11 are the exceptions, and chromosome 3 contains the highest number of PvSnRK genes (nine PvSnRKs), followed by chromosome 8 (eight PvSnRKs). Chromosomes 1, 4, 5 and 7 each have one gene each, as shown in Figure 2 and Table 1.
protein sequences of SnRK family genes of two species. The phylogenetic tree was constructed using MEGA XI software with the neighbour-joining tree method with 1000 bootstrap values.
To examine the evolution of Phaseolus SnRK genes, their chromosomal distribution was determined. PvSnRK superfamily genes are distributed across 10 pairs of homologous chromosomes among 11 pairs in the P. vulgaris genome. PvSnRK1 genes are located on chromosomes 4 and 8 and PvSnRK2 genes on chromosomes 2, 3, 6 and 8. PvSnRK3s are located on nine chromosomes, where chromosomes 1 and 11 are the exceptions, and chromosome 3 contains the highest number of PvSnRK genes (nine PvSnRKs), followed by chromosome 8 (eight PvSnRKs). Chromosomes 1, 4, 5 and 7 each have one gene each, as shown in Figure 2 and Table 1. Intron-exon analysis was carried out to obtain better insight into the structure of PvSnRK genes. PvSnRK gene family members exhibited a great variation, from 1 to 14 introns, as shown in Figure 3 and Table 1. In the PvSnRK1 sub-family, PvSnRK1.1 and PvSnRK1.2 have 9 and 10 introns, respectively, and all members of the PvSnRK2 sub-family have 8 introns each. A great variety in introns numbers was found in the sub-family PvSnRK3, with 16 PvSnRK3 members having no introns, PvSnRK3.19 having 1 intron, PvSnRK3.21 and 3.28 having 11 introns, and PvSnRK3.4 having 14 introns; the remaining eight PvSnRK3 sub-family members have 13 introns. This divergence in introns numbers indicates that exon gain, and loss occurred during evolution of the PvSnRK gene family. These findings are corroborated by the clades in the phylogenetic analysis, where clade I contains all PvSnRK2 members, clade II has PvSnRK1 individuals, and PvSNRK3 members are divided into clade III, with genes comprising 11 and 14 introns. Finally, clade IV includes all the remaining members of PvSNRK3 ( Figure S1). Intron-exon analysis was carried out to obtain better insight into the structure of PvSnRK genes. PvSnRK gene family members exhibited a great variation, from 1 to 14 introns, as shown in Figure 3 and Table 1. In the PvSnRK1 sub-family, PvSnRK1.1 and PvSnRK1.2 have 9 and 10 introns, respectively, and all members of the PvSnRK2 sub-family have 8 introns each. A great variety in introns numbers was found in the sub-family PvSnRK3, with 16 PvSnRK3 members having no introns, PvSnRK3.19 having 1 intron, PvSnRK3.21 and 3.28 having 11 introns, and PvSnRK3.4 having 14 introns; the remaining eight PvSnRK3 sub-family members have 13 introns. This divergence in introns numbers indicates that exon gain, and loss occurred during evolution of the PvSnRK gene family. These findings are corroborated by the clades in the phylogenetic analysis, where clade I contains all PvSnRK2 members, clade II has PvSnRK1 individuals, and PvSNRK3 members are divided into clade III, with genes comprising 11 and 14 introns. Finally, clade IV includes all the remaining members of PvSNRK3 ( Figure S1). A search for conserved motifs in all 42 PvSnRK proteins using the MEME program revealed a total of 25 conserved motifs, named from 1 to 25. The identified motifs were annotated in Pfam, and the details of the putative motifs are shown in Table S4. Motifs 1-4 are designated protein kinase domains and are found in all three sub-families. Motif 8, a kinase domain associated with kinase1 and KA1, and motifs 10, 11, and 20, which encode an NAF domain, are only present in PvSnRK3. Ubiquitin-associated domain 20 was found in PvSnRK1 and PvSnRK3. Furthermore, motifs 5 and 13, designated protein superfamily kinase domains, were found only in the PvSnRK1 and PvSnRK3 sub-families ( Figure 4). The remaining motifs were not annotated functionally. A search for conserved motifs in all 42 PvSnRK proteins using the MEME program revealed a total of 25 conserved motifs, named from 1 to 25. The identified motifs were annotated in Pfam, and the details of the putative motifs are shown in Table S4. Motifs 1-4 are designated protein kinase domains and are found in all three sub-families. Motif 8, a kinase domain associated with kinase1 and KA1, and motifs 10, 11, and 20, which encode an NAF domain, are only present in PvSnRK3. Ubiquitin-associated domain 20 was found in PvSnRK1 and PvSnRK3. Furthermore, motifs 5 and 13, designated protein superfamily kinase domains, were found only in the PvSnRK1 and PvSnRK3 sub-families ( Figure 4). The remaining motifs were not annotated functionally.
Ka/Ks and Gene Duplication
To further explore evolutionary constraints on Phaseolus PvSnRK genes, synonymous (Ks) and nonsynonymous (Ka) substitutions per site and their ratio (Ka/Ks) and divergence time of paralogous and orthologous SnRK family genes were calculated for AtSnRK orthologues of PvSnRKs (Tables 2 and S5). The Ka/Ks ratio among all SnRK sequences was lower than 1, indicating purifying selection. These Ka/Ks ratios suggest the conservation of SnRK homologues in terms of both sequence and biological function [61].
Ka/Ks and Gene Duplication
To further explore evolutionary constraints on Phaseolus PvSnRK genes, synonymous (Ks) and nonsynonymous (Ka) substitutions per site and their ratio (Ka/Ks) and divergence time of paralogous and orthologous SnRK family genes were calculated for AtSnRK orthologues of PvSnRKs (Table 2 and Table S5). The Ka/Ks ratio among all SnRK sequences was lower than 1, indicating purifying selection. These Ka/Ks ratios suggest the conservation of SnRK homologues in terms of both sequence and biological function [61].
Cis-Elements in Promoter Regions of PvSnRKs
To determine the gene expression pattern of PvSnRKs, the 2 kb region upstream of the CDS was analysed using the PlantRegMap database. Among all transcription factors recorded, ERF and MYB were found to be the most abundant. PvSnRK3.7 contained the greatest number of cis-elements (607) in the examined regulatory region, with 286 ERF binding sites and 49 MYB and 44 C2H2 sites. PvSnRK 3.20 has 338 TF sites; this was followed by PvSnRK 3.19 with 318 TFs, with 111 TFs being ERF TFs and 56 being bHLH TFs, and PvSnRK 1.2, with 312 TFs, with 100 being NAC TFs, 45 being MYB TFs and 40 being ERF TFs ( Figure 5, Table S6, Figure S3). The most abundant TFs, ERF, C2H2, bHLH, NAC and MYB identified in PvSnRKs were also found by symbiosis related studies in other species such as M. truncatula and L. japonicus (Table 3).
Cis-Elements in Promoter Regions of PvSnRKs
To determine the gene expression pattern of PvSnRKs, the 2 kb region upstream of the CDS was analysed using the PlantRegMap database. Among all transcription factors recorded, ERF and MYB were found to be the most abundant. PvSnRK3.7 contained the greatest number of cis-elements (607) in the examined regulatory region, with 286 ERF binding sites and 49 MYB and 44 C2H2 sites. PvSnRK 3.20 has 338 TF sites; this was followed by PvSnRK 3.19 with 318 TFs, with 111 TFs being ERF TFs and 56 being bHLH TFs, and PvSnRK 1.2, with 312 TFs, with 100 being NAC TFs, 45 being MYB TFs and 40 being ERF TFs ( Figure 5, Table S6, Figure S3). The most abundant TFs, ERF, C2H2, bHLH, NAC and MYB identified in PvSnRKs were also found by symbiosis related studies in other species such as M. truncatula and L. japonicus (Table 3). bHLH Controls a diverse processes from cell proliferation to cell lineage establishment Nodule organogenesis [64] bZIP Plant bZIP proteins preferentially bind to DNA sequences with an ACGT core Nodule organogenesis [65] C2H2 The majority of such proteins characterized to date are DNA-binding transcription factors, and many have been shown to play crucial roles in the development of plants, animals and fungi Symbiosome development [66] Dof Regulation of gene expression in processes such as seed storage protein synthesis in developing endosperm, light regulation of genes involved in carbohydrate metabolism, plant defense mechanisms, seed germination, gibberellin response in post-germinating aleurone , auxin response and stomata guard cell specific gene regulation No report bHLH Controls a diverse processes from cell proliferation to cell lineage establishment Nodule organogenesis [64] bZIP Plant bZIP proteins preferentially bind to DNA sequences with an ACGT core Nodule organogenesis [65] C2H2 The majority of such proteins characterized to date are DNA-binding transcription factors, and many have been shown to play crucial roles in the development of plants, animals and fungi Symbiosome development [66] Dof Regulation of gene expression in processes such as seed storage protein synthesis in developing endosperm, light regulation of genes involved in carbohydrate metabolism, plant defense mechanisms, seed germination, gibberellin response in post-germinating aleurone, auxin response and stomata guard cell specific gene regulation No report No report MYB transcription factors promote expression of genes involved in cell proliferation and differentiation. ERFs are involved in regulation of developmental processes in response to stimuli, and NAC, C2H2 and bHLH are involved in the pathogen response, cell proliferation and development.
GO Analysis
Gene Ontology analysis of all PvSnRK genes showed them to be involved in biological processes and molecular functions, but not cellular components. The biological processes involved related to PvSnRKs are cellular signalling, phosphorylation, and metabolic processes such as protein, macromolecule, and phosphorous metabolism. Among molecular functions, the majority are associated with binding and catalytic activities, followed by transferase and nucleotide-binding functions (Figure 6a). RT-qPCR data are the averages of three biological replicates (n > 9). The statistical significance of differences between mycorrhized and nodulated roots was determined using an unpaired two-tailed Student's t-test (** p < 0.01). Error bars represent means ± Standard error mean (SEM).
Expression Profiles of PvSnRK Genes in Different Tissues
Differential expression data for PvSnRK genes were obtained from PvGEA: Common Bean Gene Expression Atlas and Network Analysis (https://plantgrn.noble.org/PvGEA/, accessed on 16/11/2021). The expression patterns of all 42 PvSnRK members were analysed in 25 different tissues, including leaves, stems, shoots, pods, seeds, roots (inoculated and uninoculated with Rhizobium) and nodules, at different developmental stages and treatments (Figure 6b, Table S7). The lowest expression of all PvSnRK genes was found in young flower, seed and shoot tissues; at least one of the 42 genes was expressed in the remaining tissues. Among nodulation-related tissues, most of the PvSnRK genes showed high expression, except for the PvSnRK1 sub-family, in whole roots separated from fix+ nodules collected at 21 days after inoculation (PvRE). Pre-fixing (effective) nodules collected at 5 days after inoculation (PvN5) showed very high expression of only PvSnRK3. 5 and PvSnRK3.29 (Figure 6b).
Expression Patterns of PvSnRK Genes under Symbiotic Conditions
Additionally, differential expression patterns of PvSnRKs under rhizobial and mycorrhizal symbiotic conditions were analysed by transcriptomics at the early infection stages 5 dpi and 7 dpi, respectively. Most of the genes encoding PvSnRK1s and PvSnRK3s Candidate genes were selected and corresponding transcript accumulation under mycorrhized and nodulated conditions was quantified by RT-qPCR. RT-qPCR data are the averages of three biological replicates (n > 9). The statistical significance of differences between mycorrhized and nodulated roots was determined using an unpaired two-tailed Student's t-test (** p < 0.01). Error bars represent means ± Standard error mean (SEM).
Expression Profiles of PvSnRK Genes in Different Tissues
Differential expression data for PvSnRK genes were obtained from PvGEA: Common Bean Gene Expression Atlas and Network Analysis (https://plantgrn.noble.org/PvGEA/, accessed on 16 November 2021). The expression patterns of all 42 PvSnRK members were analysed in 25 different tissues, including leaves, stems, shoots, pods, seeds, roots (inoculated and uninoculated with Rhizobium) and nodules, at different developmental stages and treatments (Figure 6b, Table S7). The lowest expression of all PvSnRK genes was found in young flower, seed and shoot tissues; at least one of the 42 genes was expressed in the remaining tissues. Among nodulation-related tissues, most of the PvSnRK genes showed high expression, except for the PvSnRK1 sub-family, in whole roots separated from fix+ nodules collected at 21 days after inoculation (PvRE). Pre-fixing (effective) nodules collected at 5 days after inoculation (PvN5) showed very high expression of only PvSnRK3.5 and PvSnRK3.29 (Figure 6b).
PvSnRK Protein-Protein Interaction Network Prediction
A protein interaction network was constructed for PvSnRK proteins based on orthologues in Arabidopsis using the STRING database with the highest bit score (0.9). While predicting interactions among the PvSnRK proteins, we used experimentally proven, coexpressed and co-occurring protein interactions, revealing a total of 40 interacting proteins. The majority of them are phosphoprotein phosphatase (PPP) family proteins, PP2A (protein phosphatase PP1-α catalytic subunit), PP2CA/PP2CB (protein phosphatase 2A catalytic subunit α/β isoform), PPP2RA/PPP2RB (protein phosphatase 2A 65 kDa regulatory subunit A β isoform) and PPP1C (protein phosphatase PP1-α catalytic subunit)-type proteins. Phosphatases are proteins involved in substrate recognition, plant signalling pathways such as stress regulation, light, pathogen defence and hormonal signalling, the cell cycle, differentiation, metabolism, and plant growth.
The next important interactions were with cell cycle proteins and cyclin B proteins. Cyclin B proteins have been implicated in plant growth and development. All PvSnRK protein sub-families were found to interact with PPP family members and cyclin B proteins through PKGII, indicating their involvement in various cellular and developmental processes. SnRK2 sub-family members were found to specifically interact with PP2C proteins (Figure 7). were found to be upregulated under both symbiotic conditions, though PvSnRK2 was less induced during symbiosis with rhizobia. It was particularly interesting to find highly induced PvSnRK1.2 in comparison to PvSnRK1.1, and among the PvSnRK3 sub-families, PvSnRK3.10 was highly induced, followed by PvSnRK3.7, PvSnRK3.20, and PvSnRK3. 22. Under mycorrhizal inoculation conditions, PvSnRK1s and PvSnRK3s were all induced, and some PvSnRK2 genes were also induced. In the SnRK1 sub-family, PvSnRK1.2 was more induced; among PvSnRK2 genes, PvSnRK2.6, PvSnRK2.7 and PvSnRK2.11 were least induced. In the PvSnRK3 sub-family, PvSnRK3.1, PvSnRK3.2, PvSnRK3.3, PvSnRK3.4, PvSnRK3.5 and PvSnRK3.6 were less induced than other members (Figure 6c & 6d).
PvSnRK Protein-Protein Interaction Network Prediction
A protein interaction network was constructed for PvSnRK proteins based on orthologues in Arabidopsis using the STRING database with the highest bit score (0.9). While predicting interactions among the PvSnRK proteins, we used experimentally proven, co-expressed and co-occurring protein interactions, revealing a total of 40 interacting proteins. The majority of them are phosphoprotein phosphatase (PPP) family proteins, PP2A (protein phosphatase PP1-α catalytic subunit), PP2CA/PP2CB (protein phosphatase 2A catalytic subunit α/β isoform), PPP2RA/PPP2RB (protein phosphatase 2A 65 kDa regulatory subunit A β isoform) and PPP1C (protein phosphatase PP1-α catalytic subunit)-type proteins. Phosphatases are proteins involved in substrate recognition, plant signalling pathways such as stress regulation, light, pathogen defence and hormonal signalling, the cell cycle, differentiation, metabolism, and plant growth.
The next important interactions were with cell cycle proteins and cyclin B proteins. Cyclin B proteins have been implicated in plant growth and development. All PvSnRK protein sub-families were found to interact with PPP family members and cyclin B proteins through PKGII, indicating their involvement in various cellular and developmental processes. SnRK2 sub-family members were found to specifically interact with PP2C proteins (Figure 7).
Figure 7.
In silico prediction of protein-protein interactions among SnRK1s, SnRK2s and SnRK3s using the Cytoscape tool based on the Pearson correlation coefficients of the relative expression of the gene. Each node represents a protein, and each edge refers an interaction. The red-coloured box represents SnRK1s, blue represents SnRK2s, and green represents SnRK3s.
Prediction of Coregulatory and Interaction Networks of PvSnRK and Symbiotic Genes
The symbiotic interaction between legumes and rhizobia is unique and involves a set of common symbiotic genes that regulate root nodule symbiosis and mycorrhizal Figure 7. In silico prediction of protein-protein interactions among SnRK1s, SnRK2s and SnRK3s using the Cytoscape tool based on the Pearson correlation coefficients of the relative expression of the gene. Each node represents a protein, and each edge refers an interaction. The red-coloured box represents SnRK1s, blue represents SnRK2s, and green represents SnRK3s.
Prediction of Coregulatory and Interaction Networks of PvSnRK and Symbiotic Genes
The symbiotic interaction between legumes and rhizobia is unique and involves a set of common symbiotic genes that regulate root nodule symbiosis and mycorrhizal symbiosis. Another 190 genes have been implicated in regulating symbiotic interactions. A total of 200 genes known to be involved in symbiosis were chosen to predict an interaction network with each of the PvSnRK sub-families in the STRING database. Interaction between PvSnRK1 sub-family members and symbiosis-related proteins shows that PvSnRK1.1 and PvSnRK1.2 do not interact directly with any of the symbiosis-related genes. However, they do interact with TOR, PI3K and ATG-RP, which interact with other symbiotic genes represented by 54 nodes and 119 edges. The network mostly involves nucleoporins (NUPs), RAB proteins, auxin-responsive ARP proteins and many proteins involved in vesicle transport (Figure 8a). The largest PvSnRK3 sub-family, with 29 members, interacts with TOR, PI3K and ATG-RP at the first level, similar to PvSnRK1 members. The predictions showed 102 nodes and 319 edges, indicating a larger network. CCS interacts with nucleoporins (NUPs), and through GNBI, they interact with several RPSs (ribosomal proteins) and a variety of symbiosis-related proteins. Network prediction showed that PvSnRK3 sub-family members may be involved in regulating symbiosis in Phaseolus via different signalling pathways ( Figure 9). The largest PvSnRK3 sub-family, with 29 members, interacts with TOR, PI3K and ATG-RP at the first level, similar to PvSnRK1 members. The predictions showed 102 nodes and 319 edges, indicating a larger network. CCS interacts with nucleoporins (NUPs), and through GNBI, they interact with several RPSs (ribosomal proteins) and a variety of symbiosis-related proteins. Network prediction showed that PvSnRK3 sub-family members may be involved in regulating symbiosis in Phaseolus via different signalling pathways ( Figure 9).
Figure 9.
In silico prediction of protein-protein interactions among SnRK3 subfamily proteins with symbiosis-related proteins using the Cytoscape tool based on the Pearson correlation coefficients of the relative expression of the gene. Each node represents a protein, and each edge refers an interaction. The different colours represent the different gene families.
Discussion
The SnRK gene family, serine/threonine kinases, and its orthologues in animals and yeast are highly conserved. In plants, they have been identified as regulators of abiotic and biotic stress [80][81][82]. In recent years, genome-wide analysis of this gene family in an array of both model and economically important plant species has focused primarily on abiotic stress. Phaseolus vulgaris is the most important legume consumed by humans worldwide, as it is an affordable source of proteins, vitamins, minerals, antioxidants, and bioactive compounds [83]. Climate change has impacted the world's crop yield and quality, leading to socioeconomic and food insecurity [84]. The yield quantity and quality of N2-fixing crops can be improved by several agronomic practices, such as irrigation, sowing density and Rhizobium application. Since common bean does not need exogenous N fertilizer application, productivity is cost effective. Any efforts towards the betterment of rhizobial association to improve N fixation in crops such as Phaseolus should be undertaken. The present investigation encompasses the identification, classification, and analysis of the expression patterns of the SnRK gene family under symbiotic conditions. Genome-wide identification studies have been carried out, with early reports documenting the presence of various numbers of members in both monocots and dicots. A total of 39, 114, 30, 34, 26, 149, 46, 48 and 44 SnRK genes have been identified in A. thaliana [1], Brassica napus [30], C. sativus [31], E. grandis [32], F. ananassa [33], T. aestivum [34], H. vulgare [35], O. sativa [17] and B. distachyon [36], respectively. In Phaseolus, we identified 42 SnRK genes with 2 SnRK1s, 11 SnRK2s and 29 SnRK3s with the characteristic domains of each sub-family. As in any other species, the SnRK2 and SnRK3 sub-families in Phaseolus are larger than SnRK1, supporting the view that these two sub-families originated from duplication of SnRK1 [35]. The nonmotile nature of plants exposes them to biotic and abiotic factors, and plants adopt such changes by expanding their genes and gene families. Gene and genome duplications are important events that contribute to polyploidy and genome evolution. One or multiple polyploidies are prevalent in angiosperms [85,86] and explain the large number of SnRK2 and SnRK3 sub-family members in all reported plant species.
In Phaseolus, the SnRK gene family is distributed on 10 of 11 chromosomes. This distribution is unlike in other species, in which all chromosomes in the genome contain at
Discussion
The SnRK gene family, serine/threonine kinases, and its orthologues in animals and yeast are highly conserved. In plants, they have been identified as regulators of abiotic and biotic stress [80][81][82]. In recent years, genome-wide analysis of this gene family in an array of both model and economically important plant species has focused primarily on abiotic stress. Phaseolus vulgaris is the most important legume consumed by humans worldwide, as it is an affordable source of proteins, vitamins, minerals, antioxidants, and bioactive compounds [83]. Climate change has impacted the world's crop yield and quality, leading to socioeconomic and food insecurity [84]. The yield quantity and quality of N2-fixing crops can be improved by several agronomic practices, such as irrigation, sowing density and Rhizobium application. Since common bean does not need exogenous N fertilizer application, productivity is cost effective. Any efforts towards the betterment of rhizobial association to improve N fixation in crops such as Phaseolus should be undertaken. The present investigation encompasses the identification, classification, and analysis of the expression patterns of the SnRK gene family under symbiotic conditions. Genome-wide identification studies have been carried out, with early reports documenting the presence of various numbers of members in both monocots and dicots. A total of 39, 114, 30, 34, 26, 149, 46, 48 and 44 SnRK genes have been identified in A. thaliana [1], Brassica napus [30], C. sativus [31], E. grandis [32], F. ananassa [33], T. aestivum [34], H. vulgare [35], O. sativa [17] and B. distachyon [36], respectively. In Phaseolus, we identified 42 SnRK genes with 2 SnRK1s, 11 SnRK2s and 29 SnRK3s with the characteristic domains of each sub-family. As in any other species, the SnRK2 and SnRK3 sub-families in Phaseolus are larger than SnRK1, supporting the view that these two sub-families originated from duplication of SnRK1 [35]. The nonmotile nature of plants exposes them to biotic and abiotic factors, and plants adopt such changes by expanding their genes and gene families. Gene and genome duplications are important events that contribute to polyploidy and genome evolution. One or multiple polyploidies are prevalent in angiosperms [85,86] and explain the large number of SnRK2 and SnRK3 sub-family members in all reported plant species.
Each of the SnRK sub-families has a characteristic domain; however, the N-terminal kinase domain is highly conserved across species and sub-families. Exon-intron structural diversification and motif composition play an important role in the evolution and function of many gene families [87]. PvSnRK1 sub-family members have 10-11 exons, such as BdSnRK1s, EgrSnRK1s and CsSnRK1s. All PvSnRK2s have nine exons, similar to most HvSnRKs, HcSnRK2s, EgrSnRK2s, VvSnRK2s, AtSnRK2s and BdSnRK2s, indicating the conserved nature of these sub-families. The sub-family PvSnRK3 can also be subdivided into exon-rich and exon-poor types, as can all other species reported thus far. Reports suggest the origin of the SnRK3 sub-family from green algae, and the intron-poor group first appeared in seed plants [88]. These results are consistent with phylogenetic tree studies showing that SnRK exon-intron numbers are highly conserved during the evolution of each sub-family. The phylogenetic tree and exon-intron structure showed that most paralogous gene pairs contain the same exon number, though some gene pairs have different exon numbers. Motif analysis revealed a close evolutionary relationship within sub-groups due to the conserved nature of motif composition among sub-families. The motif structure of each sub-family might define the biological function in which they are involved. The gene structure and sequence conservation were similar to most of the previously studied species [1,17,[30][31][32][33][34][35][36].
Gene expression is regulated by external factors that are perceived through signalling mechanisms. Such signals activate specific transcription factors that, when combined with cis-acting elements in the promoter regions of genes, alter gene expression. In most genomewide analysis studies of SnRK gene families, detection of cis-acting elements has focused on abiotic stress conditions in which the presence of hormone-, salt-, temperature-, and drought-specific cis-elements [1,17,[30][31][32][33][34][35][36]. As the aim of our investigation was to predict the possible role of the PvSnRK gene family in symbiosis, we analysed symbiosis-related cis-elements as demonstrated in Medicago and L. japonicus. All PvSnRK gene sub-families contain symbiosis-related cis-elements. Among all cis-elements, ERF and MYB are the most abundant, followed by C2H2 and bHLH.
Ka/Ks analysis showed that PvSnRK gene family duplications either occurred slowly or are highly conserved [89]. Gene Ontology studies revealed that the PvSnRK genes function mostly in molecular and biological processes, specifically in cellular and metabolic processes followed by nucleotide binding activity. When we analysed expression of PvSnRK gene family members in different tissues, the lowest expression of any PvSnRK was found in aerial tissues such as young pods, flowers, seeds, and shoots. Most of the SnRKs were found to be expressed in root or root nodules at some stage of their development. Furthermore, transcriptomic data under early symbiotic conditions showed elevated expression levels of PvSnRK1 s and PvSnRK3s, with some members being more highly induced than others. For some genes, these expression patterns were found under both root nodule and mycorrhizal symbiotic conditions, suggesting a possible role for PvSnRKs in the establishment of symbiotic relationships.
To predict possible interaction networks among the PvSnRKs and PvSnRKs with symbiosis-related genes, we carried out in situ interaction network building using the STRING database and Cytoscape. The results were interesting, as most of the PvSnRKs interact among themselves, and interaction is mediated by master regulators of cellular processes such as TOR and PKGII. Through these proteins, PvSnRKs interact with several protein phosphatases and cell cycle-regulating cyclins. We chose symbiosis-related genes based on a previously published article and identified a total of 200 genes in the Phaseolus genome. PvSnRK1s mostly interact with other major proteins, such as TOR and PI3Ks, which are connected to the NUPs, RAB proteins, ARPs and proteins involved in vesicle transport. On the other hand, PvSnRK2s interact with PKGII, which interact with the cell cycle regulatory proteins cyclins, CDC, and APC. In contrast, PvSNRK2s shows very few symbiotic gene interactions. Of the largest PvSnRK sub-families, PvSnRK3 interacts with NUPs, RPSs and many symbiotic genes.
Taken together, our extensive analysis of the PvSnRK gene family revealed structural conservation of genes across species and possible functional conservation as well. Furthermore, the principle aim of this study was to understand the putative role of PvSnRKs in regulating symbiotic interactions between Phaseolus and Rhizobium or Phaseolus; mycorrhiza are promising, as some genes contain specific cis-elements and showed transcript upregulation in response to symbionts. Finally, the identified in situ protein-protein interactions may help in predicting candidate genes for functional characterization to obtain a clear picture of the regulatory mechanisms involved. | 2022-11-15T18:40:51.348Z | 2022-11-01T00:00:00.000 | {
"year": 2022,
"sha1": "d33daea1638131bfe37b8266a6d8c66f6addfe69",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4425/13/11/2107/pdf?version=1668345873",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "3e1f528eebff62906f6559f9738db8b0c533beb3",
"s2fieldsofstudy": [
"Environmental Science",
"Biology"
],
"extfieldsofstudy": [
"Medicine"
]
} |
216516172 | pes2o/s2orc | v3-fos-license | MILP-PSO Combined Optimization Algorithm for an Islanded Microgrid Scheduling with Detailed Battery ESS Efficiency Model and Policy Considerations
This paper presents the optimal scheduling of a diesel generator and an energy storage system (ESS) while using a detailed battery ESS energy efficiency model. Optimal scheduling has been hampered to date by the nonlinearity and complexity of the battery ESS. This is due to the battery ESS efficiency being a multiplication of inverter and battery efficiency and the dependency of an inverter and any associated battery efficiencies on load and charging and discharging. We propose a combined mixed-integer linear programming and particle swarm optimization (MILP-PSO) algorithm as a novel means of addressing these considerations. In the algorithm, MILP is used to find some initial points of PSO, so that it can find better solution. Moreover, some additional algorithms are added into PSO to modify and, hence, improve its ability of dealing with constraint conditions and the local minimum problem. The simulation results show that the proposed algorithm performs better than MILP and PSO alone for the practical microgrid. The results also indicated that simplification or neglect of ESS efficiency when applying MILP to scheduling may cause a constraint violation.
Introduction
In 2015, the Hawaii legislature adopted Act 97 featuring a renewable portfolio standard of 100 percent by 2045, making Hawaii the first U.S. state to make a legally binding commitment to produce all of its electricity from indigenous renewable sources. Act 97 also established interim targets of 30, 40, and 70 percent by 2020, 2030, and 2040, respectively [1]. The Hawaii legislature subsequently adopted Act 200 into law, concluding that microgrids can provide valuable services to the public utility electric grid, including many ancillary services, such as frequency and voltage control, and thereby can increase public safety and security [2]. Section 4 of Act 200 notes potential opportunities for microgrid demonstration projects at the Natural Energy Laboratory of Hawaii Authority (NELHA) that test advanced technologies and market concepts that can facilitate microgrid development consistent with the purpose of Act 200. Accordingly, NELHA is planning a microgrid that consists of photovoltaic (PV) generation, ESS, and an energy management system (EMS) to enhance energy resiliency, self-sufficiency and reliability through the capability of operating in an islanded mode during outages of the Hawaii Island electrical grid.
NELHA's configuration is consistent with remote and islanded renewable energy source (RES)-connected microgrids and distribution networks that use diesel generators, ESSs, and RESs, such
System Configuration
The algorithm has been developed to manage critical loads at the Hawaii Ocean, Technology and Technology (HOST) Park, which is managed by NELHA on Hawaii island. From Ho'ona Bay to Wawaloli Beach of the ocean-facing portion of the HOST Park, NELHA consists of three main electricity loads-a research campus, a farm compound, and a 55-inch pump station. While all three loads would benefit from inclusion in a microgrid, the three loads are separated by multiple meters and along a distribution feeder line that is operated by Hawaiian Electric Light Company, the Hawaii island utility. The 55-inch pump station location has been selected for development of the initial microgrid because Energies 2020, 13,1898 3 of 17 of uncertain tariff and interconnection implications of an advanced hybrid microgrid. Figure 1 is a line diagram of the microgrid that consists of the 750 kW diesel generator, the 600 kW PV generation system, the 500 kW/567 kWh ESS, and the pump station with an average load of 297 kW. The PV generation system, ESS, and microgrid are currently under development and they are expected to be operational in 2021.
Energies 2020, 13,1898 3 of 17 electricity loads-a research campus, a farm compound, and a 55-inch pump station. While all three loads would benefit from inclusion in a microgrid, the three loads are separated by multiple meters and along a distribution feeder line that is operated by Hawaiian Electric Light Company, the Hawaii island utility. The 55-inch pump station location has been selected for development of the initial microgrid because of uncertain tariff and interconnection implications of an advanced hybrid microgrid. Figure 1 is a line diagram of the microgrid that consists of the 750 kW diesel generator, the 600 kW PV generation system, the 500 kW/567 kWh ESS, and the pump station with an average load of 297 kW. The PV generation system, ESS, and microgrid are currently under development and they are expected to be operational in 2021.
Formulation for MILP
The proposed algorithm utilizes MILP for finding partial initial search points of the PSO model. Both of the models are separately formulated in this paper since the MILP is suitable for a linear model and the PSO is suitable for a non-linear model.
The purpose of the developed algorithm is to extend the island mode operation time by more efficiently fueling the diesel generator. This can be achieved by optimizing the efficiency of the energy generation and ESS system. In an islanded microgrid with a diesel generator, minimizing the fuel consumption of the diesel generator is the energy efficient way to operate. Hence, the objective function is expressed as: where the subscript t is the index for operation time intervals, PD is the diesel generator output, CD is the generation cost per hour of the diesel generator, ΩT is the set of time periods, and Δt is the duration of each time interval. The diesel generator cost at time t, CD,t, is expressed as: where UD is the diesel generator on/off status and a, b, and c are the cost coefficients and derived by using the curve fitting tool in MATLAB based on the diesel generator fuel consumption data [16].
Formulation for MILP
The proposed algorithm utilizes MILP for finding partial initial search points of the PSO model. Both of the models are separately formulated in this paper since the MILP is suitable for a linear model and the PSO is suitable for a non-linear model.
The purpose of the developed algorithm is to extend the island mode operation time by more efficiently fueling the diesel generator. This can be achieved by optimizing the efficiency of the energy generation and ESS system. In an islanded microgrid with a diesel generator, minimizing the fuel consumption of the diesel generator is the energy efficient way to operate. Hence, the objective function is expressed as: min where the subscript t is the index for operation time intervals, P D is the diesel generator output, C D is the generation cost per hour of the diesel generator, Ω T is the set of time periods, and ∆t is the duration of each time interval. The diesel generator cost at time t, C D,t , is expressed as: where U D is the diesel generator on/off status and a, b, and c are the cost coefficients and derived by using the curve fitting tool in MATLAB based on the diesel generator fuel consumption data [16]. The value of a, b, and c are 32,000, 210, and 0.097, respectively, where the unit of C D is Korean Won (KRW) per hour. The quadratic diesel generator cost curve is split into 10 linear pieces and, hence, to form piecewise linear cost function since the cost has to be transformed into the linear function. Consequently, the objective function for the MILP is expressed as: where S is the slope of every linearly split part derived from Equation (2), the subscript l is the index for the slope, and Ω L is the set of slopes. Table 1 shows the value of each slope. The constraint conditions to be satisfied ∀t ∈ Ω T and ∀l ∈ Ω L are l∈Ω L P D,l,t + P ESS,t + P PV,t = P L,t , − P ESS,max ≤ P ESSchg,t ≤ 0, where P ESSdis and P ESSchg are the discharge and charge output power of ESS, respectively, P PV is the output power of PV, P L is the power of load demand, P D,l,min and P D,l,max are the minimum and maximum power of diesel generator at lth interval, respectively, P ESS,max the maximum power of ESS, SOC min and SOC max are the minimum and maximum value of SOC, respectively, and P ESS is the output power of ESS. P D,l,min and P D,l,max ∀l ∈ Ω L are 0 and 75 kW, respectively, since the range of P D is split into 10 linear pieces. The SOC of ESS at time t is expressed as: where η ESSdis and η ESSchg are the efficiencies of ESS when discharging and charging, respectively. For the MILP formulation, they are considered to be constants.
Formulation for PSO
The formulation has to be different from that of the MILP, since the PSO is adopted to consider the detailed model of the system. For the PSO, instead of Equation (3), the objective function is expressed, as: min Energies 2020, 13, 1898
of 17
Note that U D,t is multiplied to the whole diesel fuel cost, C D,t , instead of being multiplied only to the coefficient a because of the characteristic of the PSO. P D,t could be selected as non-zero value, even if U D,t is zero, which is physically impossible, since the PSO is heuristically searching method. Accordingly, to prevent this error, U D,t is multiplied to the whole diesel fuel cost, so that the diesel generation cost becomes definitely zero if the diesel generator is turned off (U D,t = 0). The constraint conditions Equations (6)- (11), (13) are unnecessary for the PSO, since P D,t and P ESS,t are now considered as a single variable each. Hence, the following constraints are needed: where P D,min and P D,max are zero and the rated power of diesel generator, respectively. Most importantly, the SOC formulation considering the ESS efficiency has to be changed. In previous works [3][4][5][6][7][8][9][10][11][12][13][14][15], the ESS efficiency has been neglected or simplified, whereas it is considered in detail in this paper, which is the key contribution of this paper. The net ESS efficiency is the multiplication of the inverter efficiency η inv and the battery efficiency η BT and it is expressed as: where η BTchg and η BTdis are the charge and discharge efficiency of the battery, respectively. Note that, in order to consider the net ESS efficiency in detail, both inverter and battery efficiencies have to be considered as a function of output power. The generic inverter efficiency is referred to [17] and it can be divided into several linear equations, as (in per unit): where P ESS = P ESSdis when discharging and P ESS = P ESSchg when charging. The energy efficiency for rechargeable battery is expressed as [18] (in per unit): where P BTchg and P BTdis are the charge/discharge power flow of the battery from the inverter and they are expressed as: Refer to Equation (20), let η inv be expressed as a general form of linear function with coefficient C 1 and C 0 , that is: Energies 2020, 13, 1898 6 of 17 Subsequently, by substituting Equations (21)- (25) into Equations (18) and (19), the net ESS efficiency is represented as (in per unit): Note the per unit value of CAP is 1 and hence is omitted in Equations (26) and (27). By substituting Equations (14), (26), and (27) into (12), all of the constraint conditions can be represented in terms of controllable variables (P ESS , P D , and U D ) and parameters.
Generation Scheduling Algorithm
As shown in Equations (14), (26), and (27), in the detailed ESS efficiency model, the SOC constraint is represented as a highly nonlinear and complex function of P ESS . Hence, numerical optimization techniques, such as LP or quadratic programming, are not applicable to the problem. It is inevitable to use heuristic optimization techniques to take account of the detailed model of the ESS efficiency and PSO is used in this paper. However, the heuristic optimization techniques also have some disadvantages of performance dependency on initial points and imperfection of reflecting constraint conditions into the algorithm. We propose a MILP-PSO coordinated algorithm for an optimal scheduling of a diesel generator-ESS islanded microgrid to solve these problems.
PSO Algorithm
The PSO algorithm is working based on the following equations [17]: where V is the velocity vector, w is the weight factor, c 1 and c 2 are the acceleration constants, rand 1 and rand 2 are random numbers in the range between 0 and 1, X is the position vector, X lb and X gb are the local and the global best positions for X, respectively, w max and w min are the maximum and the minimum values of w, respectively, i is the particle index, k is the iteration index, and N is the total number of iterations. In this paper, the position vector includes P ESS , P D , and U D at every time slot. Refer to [19][20][21], the maximum value of the sum of c 1 and c 2 should be 4 and an effective initial value for each is 2. Based on these initial values, we have chosen c 1 , c 2 , w max and w min as 1, 1, 1, and 0.1, respectively, as grounds for trial-and-error simulation results. It is noticeable that there are some competitive offline and online parameter control methods that alleviate the user of this time-consuming phase while providing superior results [22][23][24][25]. An application of parameter control methods remains as a future work, as this paper is focused on the detailed model of battery ESS and the application of MILP for finding a part of PSO initial searching points.
Although PSO has been developed primarily as unconstrained optimization methods [26], it is considered to be a good alternative for solving constrained optimization problems. The most common approach for solving a constrained optimization problem is the use of a penalty function [27]. The problem with this approach is that no other method exists other than trial-and-error for defining pertinent penalty functions [28]. If the penalty values are high or low, the solutions could be trapped in a local minimum or could not be well detected. Furthermore, the solution of the optimal ESS scheduling usually lies on the limit of the SOC constraint, since the ESS will continuously charge and Energies 2020, 13, 1898 7 of 17 discharge power until its SOC reaches the maximum or the minimum value. The tendency of the solution to lie on the constraint limit makes finding solutions much more difficult. We add a simple algorithm that forces the variables to stay within the feasible solutions into the PSO to alleviate the penalty function problem. Figure 2 shows the algorithm flow chart.
Although PSO has been developed primarily as unconstrained optimization methods [26], it is considered to be a good alternative for solving constrained optimization problems. The most common approach for solving a constrained optimization problem is the use of a penalty function [27]. The problem with this approach is that no other method exists other than trial-and-error for defining pertinent penalty functions [28]. If the penalty values are high or low, the solutions could be trapped in a local minimum or could not be well detected. Furthermore, the solution of the optimal ESS scheduling usually lies on the limit of the SOC constraint, since the ESS will continuously charge and discharge power until its SOC reaches the maximum or the minimum value. The tendency of the solution to lie on the constraint limit makes finding solutions much more difficult. We add a simple algorithm that forces the variables to stay within the feasible solutions into the PSO to alleviate the penalty function problem. Figure 2 shows the algorithm flow chart. A 'Penalty Function Relaxation (PFR)' component is added to the general PSO algorithm, as shown in Figure 2. The purpose of the PFR algorithm is simply to relax the burden of the penalty function by forcing the variables to stay within the feasible solution range. All of the constraints, except that of the SOC, are handled by the proposed part of the algorithm. This is because the SOC is a highly nonlinear function of PESSchg and PESSdis in consideration of the detailed model of battery ESS, as shown in Equations (14), (26), and (27). Consequently, maintaining the SOC within a feasible range while updating the position vector can no longer be achieved by simple rules. A more complex algorithm, which understands the nonlinear relationship between PESS and SOC, is required to update the SOC in such a manner to keep it within a feasible range. In this case, we only used the penalty function for the SOC constraint due to its simplicity. The sequence of the PFR algorithm is as follows: A 'Penalty Function Relaxation (PFR)' component is added to the general PSO algorithm, as shown in Figure 2. The purpose of the PFR algorithm is simply to relax the burden of the penalty function by forcing the variables to stay within the feasible solution range. All of the constraints, except that of the SOC, are handled by the proposed part of the algorithm. This is because the SOC is a highly nonlinear function of P ESSchg and P ESSdis in consideration of the detailed model of battery ESS, as shown in Equations (14), (26), and (27). Consequently, maintaining the SOC within a feasible range while updating the position vector can no longer be achieved by simple rules. A more complex algorithm, which understands the nonlinear relationship between P ESS and SOC, is required to update the SOC in such a manner to keep it within a feasible range. In this case, we only used the penalty function for the SOC constraint due to its simplicity. The sequence of the PFR algorithm is as follows: (1) After updating V and X, P D is forced to be set as the net load (P L − P PV − P ESS ) in order to balance the generation and load. (2) Afterwards, all of the variables are checked as to whether if they violate the upper or lower limit. If so, they are constrained to their closest limit value. However, by adjusting the violated variables, the generation and load equality constraint could be violated. (3) Hence, this time, P ESS is forced to be set as the net load (P L − P PV − P D ) in order to balance the generation and load again. For sequence in Equation (2), other rules may be applied. For instance, instead of binding the violated variable on the limit, it can be readjusted to be within the limit. In such a case, the following equation can be applied to the violated variables: where X i,viol denotes the amount of violation by the variable X i . Comparative analysis of both methodologies yielded results showing little variation between the two, but suggested that the methodology of sequence (2) performed slightly better than that of Equation (31). Nonetheless, we were unable to render a definitive assessment of superiority between the two methodologies, because the PSO algorithm contains a random variable that might yield a slightly different result in each simulation. In many cases of the optimal scheduling problem, including this paper, some variables tend to be bound on their upper or lower limit. For instance, the SOC could operate at its minimum and/or maximum value for at least one cycle (the simulation results show that it has minimum value at a certain time slot) and/or the output of ESS could operate at its minimum and/or maximum value. Therefore, we have adopted the methodology that is presented in sequence (2) in this paper.
Note that, due to the sequence in Equation (4), the proposed algorithm can avoid being formulated as the mixed-integer PSO (MIPSO) problem, which makes the formulation more complicated, since it contains both discrete and continuous variables. In other words, U D is not contained in the position vector X and is simply acquired by the sequence in Equation (4). This is possible due to the characteristic of the diesel generator cost function in Equation (2), which is included in the objective function of the PSO. Due to the existence of the cost coefficient a, once being operated, the diesel generator tends to generate its power as much as possible in order to achieve cost efficient operation.
The PFR algorithm cannot deal with the SOC constraint condition in Equation (12), since it is very complex in terms of the controllable variables. Hence, it is unavoidable to use the penalty function to take account of the SOC constraint. The penalty function f p is presented as: where k p is the penalty function coefficient and SOC viol,t is the amount of SOC violation at t in per unit. The penalty function is formulated by trial-and-error, as referred to [28], and k p is set to 100. By applying PFR algorithm, all of the constraints, except the SOC constraint, can be excluded from the penalty function.
PSO Coordination with MILP
We propose a methodology to find proper initial points to alleviate such difficulties in finding a solution since the penalty function makes finding optimal solution difficult. To find initial points of PSO, MILP is adopted, since it is the most common optimization technique for the optimal scheduling. To adopt MILP, the problem is linearized as in Equations (3)- (14). The SOC constraint in Equation (12) is linearized by setting η ESS to a constant number, as the previous works [3][4][5][6][7][8][9] have done. However, in this paper, we have applied 11 different values to η ESS ranging from 90% to 100% with 1% interval. Hence, 11 different particles are acquired from MILP as the initial points of the PSO. The rest of the initial points are working the same as with the regular PSO. Hence, the proposed algorithm can take advantages from both MILP (by fixing some initial points near the local optimal value) and PSO (by retaining the chance of searching an entire feasible area). Figure 3 shows the flow chart of getting the initial points of PSO by using MILP. The set of solutions resulting from MILP is used as a portion of the initial points of the PSO, as illustrated in Figure 3. The rest of the initial points of the PSO are randomly generated from the uniformly distributed values within the allowable range. For instance, if the population of PSO N P is 1000, 11 are acquired from the algorithm shown in Figure 3 and the rest of the population is randomly selected. proposed algorithm can take advantages from both MILP (by fixing some initial points near the local optimal value) and PSO (by retaining the chance of searching an entire feasible area). Figure 3 shows the flow chart of getting the initial points of PSO by using MILP. The set of solutions resulting from MILP is used as a portion of the initial points of the PSO, as illustrated in Figure 3. The rest of the initial points of the PSO are randomly generated from the uniformly distributed values within the allowable range. For instance, if the population of PSO NP is 1000, 11 are acquired from the algorithm shown in Figure 3 and the rest of the population is randomly selected.
Simulation Environment
We used the real measured data of NELHA's 17.5-kW PV system and the pump station load shown in Figure 4 to test the algorithm. Since the 600-kW PV system of the target microgrid shown in Figure 1 has not been installed yet, we used the 17.5-kW PV data and scaled it up to 600 kW. Four cases were tested to present two different weather conditions-a sunny day (Cases 1 and 2) and a cloudy day (Cases 3 and 4)-and two different SOC initial values-40% (Cases 1 and 3) and 60% (Cases 2 and 4). For all cases, an island mode starts at 15 h. Figure 5 shows the PV generation and the pump station load data for Cases 1-4.
Go to PSO initialization in Fig. 2 and set X 1 , …, X 11 as part of initial X YES NO Figure 3. Flow chart of acquiring the initial points of PSO by using mixed-integer linear programming (MILP).
Simulation Environment
We used the real measured data of NELHA's 17.5-kW PV system and the pump station load shown in Figure 4 to test the algorithm. Since the 600-kW PV system of the target microgrid shown in Figure 1 has not been installed yet, we used the 17.5-kW PV data and scaled it up to 600 kW. Four cases were tested to present two different weather conditions-a sunny day (Cases 1 and 2) and a cloudy day (Cases 3 and 4)-and two different SOC initial values-40% (Cases 1 and 3) and 60% (Cases 2 and 4). For all cases, an island mode starts at 15 h. Figure 5 shows the PV generation and the pump station load data for Cases 1-4. shown in Figure 4 to test the algorithm. Since the 600-kW PV system of the target microgrid shown in Figure 1 has not been installed yet, we used the 17.5-kW PV data and scaled it up to 600 kW. Four cases were tested to present two different weather conditions-a sunny day (Cases 1 and 2) and a cloudy day (Cases 3 and 4)-and two different SOC initial values-40% (Cases 1 and 3) and 60% (Cases 2 and 4). For all cases, an island mode starts at 15 h. Figure 5 shows the PV generation and the pump station load data for Cases 1-4. Three different optimal scheduling methods are tested to prove the effectiveness of the proposed algorithm. The following a are brief explanation about the tested algorithms: MILP: MILP is adopted to solve the scheduling problem. The problem is linearized in a piecewise fashion to adopt MILP. Eleven results are acquired by changing ηESS from 90% to 100% with 1% interval. PSO: A general PSO algorithm is applied. All of the initial points are randomly selected. The rest of the algorithm is the same as explained in Section III-A, except that it has no Equations (1) and (3) sequences in PRF algorithm and, hence, has a different penalty function that takes account of the generation and load balance constraint. The penalty function is expressed, as: where kp is set to 100,000 grounds for trial-and-error. The exponential form used in Equation (32) is not applicable here, since we experienced that the solution cannot be found by applying exponential form with the generation and load balance constraint. Three different optimal scheduling methods are tested to prove the effectiveness of the proposed algorithm. The following a are brief explanation about the tested algorithms: • MILP: MILP is adopted to solve the scheduling problem. The problem is linearized in a piecewise fashion to adopt MILP. Eleven results are acquired by changing η ESS from 90% to 100% with 1% interval. • PSO: A general PSO algorithm is applied. All of the initial points are randomly selected. The rest of the algorithm is the same as explained in Section III-A, except that it has no Equations (1) and (3) sequences in PRF algorithm and, hence, has a different penalty function that takes account of the generation and load balance constraint. The penalty function is expressed, as: where k p is set to 100,000 grounds for trial-and-error. The exponential form used in Equation (32) is not applicable here, since we experienced that the solution cannot be found by applying exponential form with the generation and load balance constraint. Since MILP is applied to the linear model while the others are applied to the detailed model (a practical model that considers the detailed model of the net ESS efficiency), it is not appropriate to directly compare those results. To make a fair comparison between algorithms, the same detailed model evaluates all of the solutions acquired from the algorithms. Figure 6 shows the methodology of solution evaluation among three algorithms. Since MILP is applied to the linear model while the others are applied to the detailed model (a practical model that considers the detailed model of the net ESS efficiency), it is not appropriate to directly compare those results. To make a fair comparison between algorithms, the same detailed model evaluates all of the solutions acquired from the algorithms. Figure 6 shows the methodology of solution evaluation among three algorithms. Table 2 shows the simulation results of the operation cost and the sum of SOC violation at each time slot for all cases by applying different methods shown in Figure 6. Figure 7 shows the simulation results of the ESS power and SOC for each time slot using the proposed MILP-PSO algorithm. Careful observation reveals initial values of the SOC in Figure 7(a-d) are not exactly at 40% or 60%. This is because these are not actually initial values, but are the values after the first time slot. The SOCs at the first time slot of Cases 1 and 2 are slightly higher than 40% and 60%, respectively, since the ESS charges in the first time slot of Cases 1 and 2, as shown in Figure 7a,b. Similarly, the SOCs at the first time slot of Cases 3 and 4 are slightly lower than 40% and 60%, respectively, since the ESS discharges in the first time slot of Cases 3 and 4, as shown in Figure 7c,d. Table 2 shows the simulation results of the operation cost and the sum of SOC violation at each time slot for all cases by applying different methods shown in Figure 6. Figure 7 shows the simulation results of the ESS power and SOC for each time slot using the proposed MILP-PSO algorithm. Careful observation reveals initial values of the SOC in Figure 7(a-d) are not exactly at 40% or 60%. This is because these are not actually initial values, but are the values after the first time slot. The SOCs at the first time slot of Cases 1 and 2 are slightly higher than 40% and 60%, respectively, since the ESS charges in the first time slot of Cases 1 and 2, as shown in Figure 7a,b. Similarly, the SOCs at the first time slot of Cases 3 and 4 are slightly lower than 40% and 60%, respectively, since the ESS discharges in the first time slot of Cases 3 and 4, as shown in Figure 7c,d. Applying MILP to the ESS scheduling problem causes the SOC violation problem if the practical plant (ESS) is highly nonlinear, as shown in Table 2. When considering the ESS efficiency to be 95% or 100% as most previous works have done definitely contributes to the SOC violation. In the case of 100% ESS efficiency, Table 2 shows that the sum of SOC violations for 9 h with 15-min. time intervals ranges from 85.0% to 142.9%, depending on the case studies. In the case that applies the general PSO, the SOC violation is even worse and the cost is also high. This is because the general PSO is hard to handle with the constraint conditions and it is highly dependent on the initial condition.
Simulation Results and Discussion
Decreasing the ESS efficiency can avoid the risk of violating SOC, but it increases the operational cost. Moreover, it is impossible to choose a proper constant ESS efficiency value since it depends on the operational condition, such as the initial SOC, load, and PV generation. It is shown as the bold letters of Table 2-for Cases 1 and 2, the proper constant ESS efficiency is 93%, whereas it is 94% for Cases 3 and 4. Yet, these results do not represent the best choices for Cases 1 and 2. By using MILP-PSO, the cost can be even further decreased without the SOC violation, i.e., the cost is decreased by 0.53% and 0.60% for Case 1 and 2, respectively, when compared to the best solution of the conventional MILP.
The simulation results vary every time they are conducted since the PSO and MILP-PSO contains random parameters in their algorithms. Accordingly, the results of PSO and MILP-PSO in Table 2 are the average values of a thousand different simulation results. The variation of MILP-PSO results is very small since we have upgraded it, as shown in Figures 2 and 3. Table 3 shows the minimum, maximum, and standard deviations of the costs for Cases 1 and 2. Even the maximum costs among 1000 different MILP-PSO simulation for Cases 1 and 2 are smaller than those of the best solutions of the MILP, as shown in Table 3. Decreasing the ESS efficiency can avoid the risk of violating SOC, but it increases the operational cost. Moreover, it is impossible to choose a proper constant ESS efficiency value since it depends on the operational condition, such as the initial SOC, load, and PV generation. It is shown as the bold letters of Table 2-for Cases 1 and 2, the proper constant ESS efficiency is 93%, whereas it The population number N p and the total iteration number N of the PSO algorithm is adjustable. Additional simulations have been conducted for Case 1 to analyze the effect of changing those numbers on the results and the results are shown in Figure 8. Figure 8a shows the result costs and the computation Energies 2020, 13, 1898 14 of 17 times by changing N p to 100, 300, 500, and 1000, while N is fixed as 1000. As N p increases, the cost is reduced, but the reduction rate is very small while the computation time increases drastically. Similarly, by increasing N to 500, 1000, 1500, and 2000 while fixing N p as 1000, the computation time is drastically increased, as shown in Figure 8b. In this case, the cost is slightly increased even if N is increased from 1500 to 2000. This is due to the PSO containing random variables that can yield slightly different results each time. Consequently, the increment profile of the population number and the iteration number causes too large a computation time increment relative to the small size of the cost reduction.
Energies 2020, 13,1898 14 of 17 increases drastically. Similarly, by increasing N to 500, 1000, 1500, and 2000 while fixing Np as 1000, the computation time is drastically increased, as shown in Figure 8b. In this case, the cost is slightly increased even if N is increased from 1500 to 2000. This is due to the PSO containing random variables that can yield slightly different results each time. Consequently, the increment profile of the population number and the iteration number causes too large a computation time increment relative to the small size of the cost reduction.
(a) (b) Since this algorithm is about to be applied to the real test-site where the scheduling time interval is 15 min. and is expected to yield its result every time interval, the computation time should be significantly less than 15 min. This is most importantly seen in emergencies, such as an unexpected transition to island mode when the algorithm has to reschedule and yield its solution rapidly. Grounds for this constraint, we expect that the algorithm has to output the result within 1-2 min. Hence, we have selected Np = 1000 and N = 1000. However, this value is adjustable by operators, depending on their own time constraint.
Scope of the Study
We have considered the net ESS efficiency in detail vis-a-vis the application of the combined MILP-PSO algorithm and model, but acknowledge several assumptions that simplify a portion of the overall model. We have not conducted a power flow analysis on the 55-inch pump system or the applied power flow equations to our model. We also did not consider the startup cost of the diesel generator, since the primary purpose of this paper is to show that the SOC violation and the premature solution are solely caused by the consideration of ESS efficiency as a constant value. The authors deemed that the addition of these particular complex factors would obscure the findings on the critical impact of accurately considering the ESS efficiency as we have clearly demonstrated.
Moreover, it is worthy to try to adopt other metaheuristic algorithms, such as the differential evolution and genetic algorithm. Comparing the superiority between all of the metaheuristic algorithms is out of scope of this paper, but adopting other types of algorithms with a parameter adaptation method could yield an improved performance. These issues remain as future works.
Conclusions
In this paper, the detailed model of ESS efficiency is reflected into the ESS-diesel generator scheduling algorithm for an islanded microgrid. A combined MILP-PSO algorithm is developed to overcome the limitation of the conventional, independent MILP and PSO algorithms, which has resulted in greater efficiency of the diesel back-up generator. Actual electrical consumption data from the 55-inch pump at NELHA are used to simulate microgrid performance to prove the effectiveness of the proposed algorithm. The simulation results show that applying MILP or PSO Since this algorithm is about to be applied to the real test-site where the scheduling time interval is 15 min. and is expected to yield its result every time interval, the computation time should be significantly less than 15 min. This is most importantly seen in emergencies, such as an unexpected transition to island mode when the algorithm has to reschedule and yield its solution rapidly. Grounds for this constraint, we expect that the algorithm has to output the result within 1-2 min. Hence, we have selected N p = 1000 and N = 1000. However, this value is adjustable by operators, depending on their own time constraint.
Scope of the Study
We have considered the net ESS efficiency in detail vis-a-vis the application of the combined MILP-PSO algorithm and model, but acknowledge several assumptions that simplify a portion of the overall model. We have not conducted a power flow analysis on the 55-inch pump system or the applied power flow equations to our model. We also did not consider the startup cost of the diesel generator, since the primary purpose of this paper is to show that the SOC violation and the premature solution are solely caused by the consideration of ESS efficiency as a constant value. The authors deemed that the addition of these particular complex factors would obscure the findings on the critical impact of accurately considering the ESS efficiency as we have clearly demonstrated.
Moreover, it is worthy to try to adopt other metaheuristic algorithms, such as the differential evolution and genetic algorithm. Comparing the superiority between all of the metaheuristic algorithms is out of scope of this paper, but adopting other types of algorithms with a parameter adaptation method could yield an improved performance. These issues remain as future works.
Conclusions
In this paper, the detailed model of ESS efficiency is reflected into the ESS-diesel generator scheduling algorithm for an islanded microgrid. A combined MILP-PSO algorithm is developed to overcome the limitation of the conventional, independent MILP and PSO algorithms, which has resulted in greater efficiency of the diesel back-up generator. Actual electrical consumption data from the 55-inch pump at NELHA are used to simulate microgrid performance to prove the effectiveness Energies 2020, 13,1898 15 of 17 of the proposed algorithm. The simulation results show that applying MILP or PSO independently to the islanded microgrid causes an SOC violation in the practical microgrid and also contributes to premature solutions that hinder efficiency. However, by applying the proposed combined MILP-PSO algorithm, less fuel is used and operational costs can be decreased by 0.53-0.6% by eliminating the SOC violation.
From a resilience standpoint, this greater efficiency results in longer operation of the back-up systems in islanded mode. Therefore, optimized resource scheduling, as described herein, can help to fulfill the intent of Hawaii's Act 200 regarding microgrids adding valuable services to the grid and energy resiliency.
Conflicts of Interest:
The authors declare no conflict of interest. Local and global best positions for X, respectively. | 2020-04-16T09:05:29.623Z | 2020-04-13T00:00:00.000 | {
"year": 2020,
"sha1": "fb96bb64659be06726a91b2095db1aed00125339",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/1996-1073/13/8/1898/pdf",
"oa_status": "GOLD",
"pdf_src": "MergedPDFExtraction",
"pdf_hash": "619c4580265f5789d9872a56d9ba19c62db04b89",
"s2fieldsofstudy": [
"Engineering"
],
"extfieldsofstudy": [
"Computer Science"
]
} |
244938649 | pes2o/s2orc | v3-fos-license | Gallic Acid-Containing Gelatin-Based Nonwoven Mat with Synergistic Photodegradation and Photoindication Function for Reducing Nicotine
Cigarette smoking is a popular habit that has negative health consequences for populations. In this study, we developed a gallic acid-containing, gelatin-based nonwoven mat with photodegradation and photoindication functions. This could react with sidestream cigarette smoke and simultaneously inhibit the activity of the microbe growth in the air. The results of a fluorescence emission spectrum evidenced this photoindication function. Neither the nicotine nor gallic acid showed a redshift emission spectrum. However, the emission spectrum of the nonwoven mat exhibited the redshift and increased in intensity after absorbing the sidestream cigarette smoke. In this spectral evidence, the natural polymer played a key role in the photoindication function’s display because it could dissolve the nicotine of the sidestream cigarette smoke and cause it to react with the gelatin structure. The high performance liquid chromatography–mass spectroscopy results indicated that the gallic acid and ultraviolet (UV) light enhanced the absorption of nicotine and nicotine-like derivatives, which were dissolved by the Tween 80 of nonwoven mat. The liquid paraffin and Tween 80 could oxidize, dehydrogenate, and demethylate the nicotine that was absorbed by the gelatin nonwoven mat. In conclusion, the nonwoven mat developed in this study provided the functions to filter the nicotine of sidestream smoke and activate the photoindication property by absorbing 365-nm UV light.
Introduction
Cigarette smoking is a popular habit that causes negative health consequences worldwide, including pulmonary diseases, cardiovascular diseases, and cancer [1]. When one smokes a cigarette, nicotine in the mainstream smoke is breathed into the lungs and the sidestream smoke pollutes the nearby air [2]. Some medical research has suggested that nicotine is the major addictive component in tobacco and inhaling excessive nicotine can cause limb peripheral vasoconstriction, rapid heartbeat, elevated blood pressure, platelet aggregation, cardiovascular blockage, stroke, and other cardiovascular diseases [3]. Nicotine is the major compound emitted in cigarette smoke and is soluble in chloroform, ether, ethanol, oil, and water. The common nicotine-like derivatives include parent nicotine (163), nornicotine (149), nicotine-N-oxide (179), cotinine (177), and trans-3 -hydroxycotinine (193); these are the major components of nicotine [4].
Considering the deleterious influence of cigarette smoking on human health as well as the importance of lung health given the risk of COVID-19 infection, we developed a nonwoven mat that can simultaneously provide the functions of indicating air pollutant levels and purifying the air. Moreover, to apply this nonwoven mat in daily life, we designed a small, compact, wearable device for the user to clip onto cloth, rather than the type of bulky air purifier commonly used in household environments. The benefit of this new compact wearable device is that it can help immobilize the porous and thin nonwoven mat and provide a visible indication of specific air pollutant levels in situ. When the mat user perceives a color change in the nonwoven mat, they can either actively leave the sidestream smoke area or temporarily remain, using the device to help reduce the pollutant levels until the user can relocate to a place with cleaner air. A piece of light and thin nonwoven material is the basis for providing the photodegradation and photoindication functions; thus, this material should be biodegradable and environmentally friendly. Researchers have discovered that when some polyphenoic acids, such as ellagic acid [5,6], are directly bound to certain molecules, their molecular configuration changes and markedly enhances their ultraviolet (UV) absorption (maximum absorption wavelength 280 nm). Furthermore, ellagic acid then emits a fluorescence spectrum at approximately 400-450 nm. To strike a balance between biodegradability and ease of electrospinning protocol, we adopted a natural polymer gelatin, which is a tasteless, colorless, solid substance derived from collagen, as the main constituent of the electrospun nonwoven mat due to its array of advantages, such as biological origin, nonimmunogenicity, biodegradability, and biocompatibility [7,8]. The biodegradability of gelatin could be used to dispose of the nonwoven mats that adsorb nicotine. In some studies, researchers have reported that gelatin nanofibers exhibit outstanding properties of surface area and functionality, rendering them suitable in the development of biomimicking artificial extracellular matrices, wound dressing materials with antibacterial properties, and drug delivery matrices [9,10]. We followed Huang et al. in screening a series of polyphenol compounds to render the photoindication and photodegradation functionalities of the nonwoven mat. Among the polyphenolic compounds used, gallic acid reportedly exerts numerous therapeutic effects, including antimicrobial, antioxidant, and anticancer activities [11,12]. In some studies, gallic acid was used as an antifungal agent [13] and an antibacterial drug [14]. It is also widely noted in the plant kingdom and is largely found in free-form or as a derivative in various food sources, such as nuts, tea, grapes, and sumac [15,16]. Other sources include gallnuts, oak bark, honey, assorted berries, pomegranate, mango, and other fruits, vegetables, and beverages. Therefore, the proposed gallic acid-containing, gelatin-based nonwoven mat was designed and fabricated using the electrospinning technique. In this study, a series of photoluminescence spectra and high-performance liquid chromatography-mass spectroscopy (HPLC-MS) were used to develop and verify a nonwoven mat possessing photodegradation and photoindication functions. The porous membrane composed of nanofibers could react with sidestream smoke and inhibit the activity of microbes.
To determine the optimal recipe for fabrication of the gallic acid-containing, gelatinbased nonwoven mat, two types of nonwoven mats were prepared using different nicotinedissolving recipients. One nicotine-dissolving recipient was liquid paraffin, which was dissolved with PEG (polyethylene glycol) (Sigma) by using formic acid (Sigma), and the other was dissolved using hot deionized (DI) water (1:2 measure). First, the fish gelatin was weighed and then dissolved using formic acid (120 mg dissolved in 200 µL formic acid) and subsequently stirred together with the liquid paraffin-PEG mixture (150 µL/20 mg) or Tween 80 solution (33 µL Tween 80, stock solution mixed with 67 µL DI water). Second, the 100 µL of 0.587 M gallic acid solution, which was dissolved with ethanol (100 mg dissolved in 1 mL ethanol), was pipetted to the previous mixture and stirred well. Finally, the aforementioned mixture was fed by a syringe pump and electrospun to generate two distinct types of gallic acid-containing, gelatin-based nonwoven mats.
Adsorption Measurement of Nicotine from the Sidestream Cigarette Smoke
The nonwoven mat was weighed and divided into three equal parts that were placed in a container full of sidestream smoke, as illustrated by the schematic in Scheme S1. The space of the container was segmented into two parts by using a cardboard barrier; one of the chambers was equipped with the cigarettes and 365-nm UV flashlight and the other one only contained the cigarettes. Then, gallic acid-containing, gelatin-based nonwoven mats were fastened to the roof of each chamber with a retaining clip. After the sidestream smoke was vented through interstices of the box, the test samples, which were named according to their ingredients and test parameters, were fixed and tested. The experimental groups were abbreviated as follows: fGglp, F-fGglp, UV-F-fGglp, fGgT, F-fGgT, and UV-F-fGgT (fG: fish gelatin, g: gallic acid, lp: liquid paraffin, T: Tween 80, F: fumigated, UV: 365-nm UV light).
Experimental Analyses
The fiber structure of the nonwoven mats was observed using scanning electron microscopy (SEM; JSM-6500F, JEOL), and all relevant statistical results were calculated using Image J (National Institutes of Health, Bethesda, MD, USA). Fluorescence emission spectra of the nonwoven mats were fumigated by cigarettes in Scheme S1 and acquired using a fluorescence spectrophotometer (Jasco, FP-8500). The nonwoven mats were set to a measurement block and excited by 365 nm UV light. Fluorescence emission spectra were obtained from a solid matrix because we had hoped to simulate the actual situation in the device and avoid the quenching property of the liquid sample. To determine the amount of nicotine adsorption and the degradation by-product, the experimental samples were loaded into individual bottles, each containing 4 mL of methanol, and then the nonwoven mats were soaked in methanol for 4 h to dissolve the nicotine and degradation by-products. After 4 h, the methanol extract solution in every bottle was filtered (0.22-µm filters) and submitted for HPLC-UV. In the HPLC-UV spectrum, the mixture was separated into individual components and then calculated based on the integral area of the UV absorption peak. Finally, the integral area was transformed to the absorption dose by using the standard curve and Equation (1).
Absorption dose = C(dilution ratio)(amount of methanol) 1000 (1) Equation (1): The transformed equation of absorption dose, where C is the concentration that was calculated using the standard curve (ppm).
To identify the gallic acid, gelatin, and nicotine derivatives, the nonwoven mats were soaked in DI water, and the solutions were then filtered using 0.22-µm filters. The HPLC-MS results were provided by the National Tsing Hua University Precision Instru- To secure the thin and lightweight nonwoven mat, we designed a type of wearable equipment that included a powerful yet compact blower (30 mm × 30 mm × 3.8 mm, 9500 rpm ± 20%, 88.18 Pa, 72 mA; Invni Tech Developing Corp.), as shown in Scheme S2, a UV light-emitting diode (5 mm, 365-370 nm, 3.6-3.8 V, 20 mA; Koodyz Technology), and a piece of gallic acid-containing, gelatin-based nonwoven mat. The wearable equipment could create an indraft to filter sidestream smoke though the nonwoven mat, and then the filtrated pollute could be captured, dissolved, and used to derive the fluorescence emission for photoindication. Two clean boxes were prepared with content agar plates, as shown in Table S4. One of the boxes was prepared with the wearable equipment and nonwoven mat, which could flow the air and filter the pollute, and the other was prepared with the same type of blower to flow the air and blow the dust to the surface of agar. Then, the plates were cultured for 48 h in the incubator.
Results and Discussion
The structure of the nonwoven mat was evaluated based on the SEM results. The porous structure could increase the contact area, allowing the mat to obstruct more sidestream smoke. Although the nicotine-dissolving polymer helped provide the photoindication property, it could clog the pores and decrease the fiber morphology formation, causing suboptimal porosity. By contrast, gallic acid was a noteworthy material that improved the gelatin fiber structure and enhanced the porosity regardless of whether paraffin or Tween 80 was used (Tables S1 and S2). This suggests that the addition of gallic acid not only induces the photoindication function but also alleviates the side effects of the nicotine-dissolving recipients (i.e., paraffin or Tween 80). In the SEM results, the major pore size distributions that were expected to filter and absorb the pollution of sidestream smoke were as small as 0.05 µm 2 . Figure 1E illustrates that liquid paraffin coated the fibers and clogged up the pores, destroying the porous structure and generating pore size distributions of approximately 0.025-0.05 µm 2 ( Figure 1F). The width of fibers and the porosity was markedly superior when the nonwoven mat was prepared using gelatin, gallic acid, and liquid paraffin recipe pairs, as shown in Figure 1G,H. Before the electrospinning, the liquid paraffin was mixed with the gelatin solution, which increased the viscosity of the mixture and simplified the generation of the electrospun nonwoven mat. In contrast, when Tween 80, rather than liquid paraffin, was used as the nicotine-dissolving polymer, the resultant nonwoven mat was composed of fibers but exhibited a bead morphology that covered the pores, markedly reducing the porosity ( Figure 2E,F). Gallic acid could be used to provide photodegradation functionality and strike a balance between the viscosity of the solution and the integrity of the electrospun nonwoven mat. Therefore, the generation of beads was greatly reduced by adding gallic acid when the nonwoven mat was prepared using the stock solution (consisting of gelatin, gallic acid, and Tween 80). The results of the gallic acid-containing nonwoven mats could not only ensure the photoindication function but also enhance the fiber-forming property of the gelatin, as shown in Figure 2G,H. The fluorescence emission spectra results could be used to evidence the process of photoindication. The nicotine-dissolving polymer materials (i.e., liquid paraffin and Tween 80) were the key components in the nonwoven mat, responsible for adsorbing the nicotine for fluorescence emission. To observe the change in fluorescence property between nicotine, nicotine-dissolving polymer materials, and 365 nm UV light, the equal parts of nonwoven mats were recorded to measure the disparity in 365 nm UV. The equal parts of nonwoven mats could emit fluorescence after the fumigation step. Then, the nonwoven mats were used to detect the distribution of emission wavelength and the disparity of intensity by fluorescence spectrophotometer, as shown in Figures 3-6. In Figure 4, the quantitative emission spectra acquired from the fGglp (fish gelatin/gallic acid/liquid paraffin) nonwoven mat with different proportions of added paraffin proved the fluorescence intensity and wavelength before and after the sidestream smoke was absorbed. The cigarette smoke caused a purple emission light redshift at approximately 10-30 nm and simultaneously enhanced the intensity of the blue emission light at 468 nm. Figure 6 indicates the emission spectrum acquired from the fGgT (fish gelatin/gallic acid/Tween 80) nonwoven mat system, which was irradiated by a UV light (365 nm, UV absorption maximum wavelength of gallic acid) and simultaneously adsorbed the sidestream smoke to confirm the effect of UV irradiation on the activation of photoindication and the photodegradation property. The purple emission peak (400 nm) exhibited redshift at approximately 29 nm before and after UV exposure. Notably, the fluorescence did not disappear after the fumigation; conversely, the emission peak intensity at 429 and 467 nm was considerably enhanced ( Figure 6B). This result reveals that the nicotine-dissolving polymer (i.e., Tween 80) played a key role in dissolving the sidestream-smoke nicotine, possibly performing a similar function to that of the ellagic acid, which altered the conformation of the chemical structure after reacting with some small molecules. When the nonwoven mat was irradiated by 365-nm UV light and adsorbed the sidestream smoke, the fluorescence emission was activated and markedly greater (see Figure 6B, pink curve) than that of the nonexposed nonwoven mat (see Figure 6B, orange curve).
HPLC-UV was used to determine the efficiency of liquid paraffin and Tween 80 in the nonwoven mat for enhancing nicotine capture. The amount of nicotine and gallic acid was determined using the standard curves that were established using the integration area of the UV peak absorption versus concentration (ppm; Figure 7). As can be seen in the figure, two equations were derived from the standard curves for gallic acid and nicotine, respectively. As the components of our nonwoven mats included gallic acid, Tween 80, liquid paraffin, and the adsorbate (i.e., nicotine), we detected the releases of those components in the fGglp (fish gelatin/gallic acid/liquid paraffin) and fGgT (fish gelatin/gallic acid/Tween 80) nonwoven mats. In Figure 8, the analyte peak elutions at 0.89, 4.20, and 4.53 min in the HPLC-UV spectra were attributed to gallic acid, nicotine, and liquid paraffin (or Tween 80), respectively. For the fresh nonwoven mats (control group, Figure 8A,D), subtle peaks were observed at 4.20 min, suggesting that these peaks were attributed to the adsorbate nicotine after the nonwoven mat had been used. The peak of nicotine could be separated and observed at 4.20 min with the UV absorption detection at 259 nm, as analyzed using the HPLC-UV system. The peak areas acquired from various samples were calculated (using the standard curves and linear equation) as 152.25 ± 6.59 µg (F-fGglp), 180.10 ± 16.08 µg (UV-F-fGglp), 101.15 ± 16.15 µg (F-fGgT), and 162.91 ± 17.23 µg (UV-F-fGgT; Table S3). Table S3 and Figure 8B,C,E,F clearly show that the content of adsorbed nicotine could be increased by 18-61% when the UV light was irradiated on the nonwoven mats during the sidestream smoke fumigation period. The peaks of the liquid paraffin and Tween 80 groups were notably present at 4.53 min. When Figure 8A,D was compared with the others, the peaks at 4.53 min in Figure 8B,C,E,F were stronger than Figure 8A,D as presented in Table S3. These results could be determined by the correlation of nicotine with liquid paraffin (or Tween 80). Noteworthily, because liquid paraffin and Tween 80 were not dissolvable by mobile phase methanol, only the nicotine could be dissolved, which suggested that part of the liquid paraffin and Tween 80 could be dissolved and extracted when interacting with nicotine. To further verify this hypothesis, the extraction liquid from the nonwoven mat was analyzed using both HPLC-UV and HPLC-MS data. The results of HPLC-MS provided the necessary information to precisely understand the varieties of nicotine derivatives that were captured or photodegraded from the nicotine in the sidestream cigarette smoke. The HPLC-MS result suggested that the intensity of nicotine was greater when the nonwoven mat was irradiated by UV light (Figure 9C-E and D-F, before and after). When the nonwoven mat was prepared with a liquid paraffin system (Figure 9), the HPLC-MS result indicated that both nicotine and nicotine derivative molecules could be extensively captured upon application of both gallic acid and 365 nm UV light. The nicotine derivatives observed were as follows: parent nicotine (m/z = 163), nornicotine (m/z = 149), nicotine-N-oxide (m/z = 179), cotinine (m/z = 177), and trans-3 -hydroxycotinine (m/z = 193), which were referred by references [3,4]. Notably, the inclusion of gallic acid in the nonwoven mat produced substantially greater adsorption of nicotine derivatives ( Figure 9D,F). When the mat was irradiated with UV light, the amount of nicotine adsorbate was markedly higher and the intensity of nicotine derivatives (especially trans-3 -hydroxycotinine) was higher ( Figure 9D,F). Figure 10 also displays the results for the fGgT nonwoven mat, which replaced liquid paraffin with Tween 80 ( Figure 10D); when this nonwoven mat absorbed the smoke, we noted that the Tween 80 also dissolved the nicotine and the signal of the nicotine derivatives, ( Figure 10E). However, when the nonwoven mat was irradiated by 365 nm UV light, the nonwoven mat could adsorb more smoke, which also enhanced the signal of nicotine derivatives, possibly due to the gallic acid-induced photodegradation ( Figure 10F). In Figure 10F, the signal intensity of the norcotinine and nicotine-N-oxide derivatives are obviously increasing. The results also indicate a markedly higher nicotine intensity compared with that shown in Figure 10E. In general, these results may indicate that the gallic acid addition not only enhances the absorption property of the nonwoven mat but also enables it to exert a photodegradation function in conjunction with suitable UV light irradiation. In accordance with the fluorescence emission spectra and HPLC-MS results, we theorize that the fluorescence emission mechanism of the nonwoven mat was strongly tied to the chemical structure of the gallic acid and the nicotine molecules that were dissolved by liquid paraffin or Tween 80.
The mechanism of the photoindication function of the nonwoven mats could be explained by the fluorescence emission spectra. When the gelatin mat was irradiated by 365-nm UV light, a dominant purple emission peak was observed at approximately 400-439 nm, with the other shoulder peak observed at a blue emission wavelength of approximately 468 nm. This purple emission mixed again, concealed the 400-439-nm emission, and made the observed color present when close to navy blue light. Regarding another aspect of the photodegradation property, we discovered that, because nicotine is a greasy pollutant, the addition of nicotine-dissolving polymer materials (i.e., liquid paraffin or Tween 80) could help adsorb smoke and react with nicotine after the sidestream smoke fumigated the gelatin mat to raise the adsorption ability. Subsequently, the nicotine contained in the sidestream smoke caused the purple emission peak to exhibit redshift close to the 468-nm emission, and then changed the observed color from purple-blue to pale blue. According to the fundamental mechanism of the fluorescence property, the π-π* (orbital) of the conjugated system was a major transition type that contributed to the following fluorescence emission. After the nonwoven mat absorbed the nicotine, the conjugated system of gallic acid inside provided a transition path that produced the fluorescence. When the 365-nm UV light provided energy to the conjugated system to excite the electron, the electrons would then relax by releasing energy in the form of fluorescence and returning to the ground state with a different vibrational energy level. When the nonwoven mat was prepared using Tween 80 or liquid paraffin, more nicotine could be captured, and the fluorescence intensity was greater due to the increase in the conjugated system. Although gallic acid and UV light with an accurate absorption wavelength were the key parts of the nonwoven mat developed with photoindication and photodegradation, their addition in the electrospinning with gelatin also provide benefits for fiber and porous structure formation. The HPLC-UV results suggest that liquid paraffin and Tween 80 could be dissolved when they captured nicotine molecules. Conversely, no relevant elution peaks were observed in the chromatogram when the nonwoven mats were not fumigated with sidestream smoke. On the basis of the fluorescence spectra, we can speculate that the liquid paraffin and Tween 80 in this study captured nicotine that was adsorbing on the surface of fibers, exhibiting a noteworthy fluorescence phenomenon that was based on the excitation of the electrons of the conjugated system by 365-nm UV light. Moreover, gallic acid was excited, and part of the excited state was transformed from a singlet state to a triplet state. The triplet state of the gallic acid performed two functions: enhancement of the UV absorption and photodegradation of the pollutant [2]. The aforementioned mechanism is depicted in Scheme 1. Thereafter, when the nonwoven mats were soaked in methanol, a part of liquid paraffin and Tween 80 that combined with the nicotine was extracted and subsequently detected by HPLC-UV. The absorbed dose of nicotine was quantitively determined and is displayed in Table S3. Notably, the gallic acid could be leached out and detected. However, the amount of adsorbed nicotine was much greater than that of the added gallic acid. This result reveals that the remaining part of the nicotine was extracted by the nicotine-dissolving polymer in the fiber (i.e., paraffin or Tween 80). We also confirmed that the combination of 365-nm UV light, liquid paraffin or Tween 80, and gallic acid produced the two optimal parameter configurations. UV-F-fGglp and UV-F-fGgT nonwoven mats absorbed twice the dose of nicotine compared to the other groups. Obviously, gallic acid and nicotine-dissolving polymer work synergistically to capture more nicotine molecules. On the basis of the HPLC-MS results, we conclude that the gallic acid and UV light were associated with greater absorption of nicotine and simultaneously generated a series of nicotine derivatives when liquid paraffin and Tween 80 dissolved the nicotine and transferred it to the extract solution. We assert that the triplet state of gallic acid caused the photodegradation of nicotine to generate a series of nicotine derivatives. According to this evidence, we used a nonwoven mat that included gallic acid and a nicotine-dissolving polymer to adsorb smoke under 365-nm UV light irradiation. The HPLC-MS results showed the signals of cotinine (m/z = 177) and trans-3 -hydroxycotinine (m/z = 193), which is reportedly the major metabolite of nicotine [4]. Apart from adsorption and degradation, the gelatin-based nonwoven mats fabricated in this study provided filtration functions to capture the nicotine from sidestream smoke and activate the photoindication property through 365 nm UV light irradiation. As the nonwoven mat developed in this study was a freestanding film, a portable fan-type device was designed to actively channel the sidestream smoke to be filtrated and alert the user to the health risk of the surrounding air quality through the photoindication function and the 365 nm UV light. In Table S4, some of the pollutants adhered to the control plate with dust when the air was blown by the device, and then the control plate was cultured in an incubator for 48 hr. Almost 1064 units of bacterial colonies grew in the control plate. Conversely, there were not any bacterial colonies that appeared in the other agar plate. The result could prove the filtrate functionality of the nonwoven mats. HPLC-UV was used to determine the efficiency of liquid paraffin and Tween 80 in the nonwoven mat for enhancing nicotine capture. The amount of nicotine and gallic acid was determined using the standard curves that were established using the integration area of the UV peak absorption versus concentration (ppm; Figure 7). As can be seen in the figure, two equations were derived from the standard curves for gallic acid and nicotine, respectively. As the components of our nonwoven mats included gallic acid, Tween 80, liquid paraffin, and the adsorbate (i.e., nicotine), we detected the releases of those components in the fGglp (fish gelatin/gallic acid/liquid paraffin) and fGgT (fish gelatin/gallic acid/Tween 80) nonwoven mats. In Figure 8, the analyte peak elutions at 0.89, 4.20, and 4.53 min in the HPLC-UV spectra were attributed to gallic acid, nicotine, and liquid paraffin (or Tween 80), respectively. For the fresh nonwoven mats (control group, Figure 8A,D), subtle peaks were observed at 4.20 min, suggesting that these peaks were attributed to the adsorbate nicotine after the nonwoven mat had been used. The peak of nicotine could be separated and observed at 4.20 min with the UV absorption detection at 259 nm, as analyzed using the HPLC-UV system. The peak areas acquired from various samples were calculated (using the standard curves and linear equation) as 152.25 ± 6.59 μg (F-fGglp), 180.10 ± 16.08 μg (UV-F-fGglp), 101.15 ± 16.15 μg (F-fGgT), and 162.91 ± 17.23 μg (UV-F-fGgT; Table S3). Table S3 and Figure 8B,C,E,F clearly show that the content of adsorbed nicotine could be increased by 18-61% when the UV light was irradiated on the nonwoven mats during the sidestream smoke fumigation period. The peaks of the liquid paraffin and Tween 80 groups were notably present at 4.53 min. When Figure 8A,D was compared with the others, the peaks at 4.53 min in Figure 8B,C,E,F were stronger than Figure 8A,D as presented in Table S3. These results could be determined by the correlation of nicotine with liquid paraffin (or Tween 80). Noteworthily, because liquid paraffin and Tween 80 were not dissolvable by mobile phase methanol, only the nicotine could be dissolved, which suggested that part of the liquid paraffin and Tween 80 could be dissolved and extracted when interacting with nicotine. To further verify this hypothesis, the extraction liquid from the nonwoven mat was analyzed using both HPLC-UV and HPLC-MS data. The results of HPLC-MS provided the necessary information to precisely understand the varieties of nicotine derivatives that were captured or photodegraded from the nicotine in the sidestream cigarette smoke. The HPLC-MS result suggested that the intensity of nicotine was greater when the nonwoven mat was irradiated by UV light (Figure 9C-E and D-F, before and after). When the nonwoven mat was prepared with a liquid paraffin system (Figure 9), the HPLC-MS result indicated that both nicotine and nicotine derivative molecules could be extensively captured upon application of both gallic acid and 365 nm UV light. The nicotine derivatives observed were as follows: parent nicotine (m/z = 163), nornicotine (m/z = 149), nicotine-N-oxide (m/z = 179), cotinine (m/z = 177), and trans-3′-hydroxycotinine (m/z = 193), which were referred by references [3,4]. Notably, the inclusion of gallic acid in the nonwoven mat produced substantially greater adsorption of nicotine derivatives ( Figure 9D,F). When the mat was irradiated with UV light, the amount of nicotine adsorbate was markedly higher and the intensity of nicotine derivatives (especially trans-3′-hydroxycotinine) was higher ( Figure 9D,F). Figure 10 also displays the results for the fGgT nonwoven mat, higher nicotine intensity compared with that shown in Figure 10E. In general, these results may indicate that the gallic acid addition not only enhances the absorption property of the nonwoven mat but also enables it to exert a photodegradation function in conjunction with suitable UV light irradiation. In accordance with the fluorescence emission spectra and HPLC-MS results, we theorize that the fluorescence emission mechanism of the nonwoven mat was strongly tied to the chemical structure of the gallic acid and the nicotine molecules that were dissolved by liquid paraffin or Tween 80. The mechanism of the photoindication function of the nonwoven mats could be explained by the fluorescence emission spectra. When the gelatin mat was irradiated by 365nm UV light, a dominant purple emission peak was observed at approximately 400-439 nm, with the other shoulder peak observed at a blue emission wavelength of approximately 468 nm. This purple emission mixed again, concealed the 400-439-nm emission, and made the observed color present when close to navy blue light. Regarding another aspect of the photodegradation property, we discovered that, because nicotine is a greasy pollutant, the addition of nicotine-dissolving polymer materials (i.e., liquid paraffin or Tween 80) could help adsorb smoke and react with nicotine after the sidestream smoke fumigated the gelatin mat to raise the adsorption ability. Subsequently, the nicotine contained in the sidestream smoke caused the purple emission peak to exhibit redshift close to the 468-nm emission, and then changed the observed color from purple-blue to pale blue. According to the fundamental mechanism of the fluorescence property, the π-π* (orbital) of the conjugated system was a major transition type that contributed to the following fluorescence emission. After the nonwoven mat absorbed the nicotine, the conjugated system of gallic acid inside provided a transition path that produced the fluorescence. When the 365-nm UV light provided energy to the conjugated system to excite the
Conclusions
In this study, nonwoven mats composed of gelatin, gallic acid, and nicotine-dissolving polymers (liquid paraffin or Tween 80) were designed and developed to efficiently filtrate and adsorb nicotine, and nicotine-dissolving polymers were able to dissolve the nicotine from sidestream smoke, causing it to react with the gelatin structure and alerting users to the health risk of the air environment. Regarding the nicotine-dissolving polymers (liquid paraffin or Tween 80), they could not only capture the nicotine molecules, but also oxidize, dehydrogenate, and demethylate the nicotine to generate nicotine derivatives that included cotinine and trans-3′-hydroxycotinine. When those nonwoven mats were fumigated with sidestream smoke and simultaneously irradiated by 365-nm UV light, the adsorbed dose of nicotine increased by 18-61% compared with the mats not containing the gallic acid or nicotine-dissolving polymers.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1, Scheme S1: Experimental steps of the nonwoven mats fumigated with sidestream smoke, Scheme S2: Powerful yet compact blower attached to portable equipment, Scheme S3: Experimental steps of the pollute filtrate. (A) The combination of the wearable equipment and nonwoven mat, (B) The control plate, (C) The filter plate, Table S1: The porosity of nonwoven mat (fG: fish Gelatin, g: gallic acid, lp: liquid paraffin), Table S2: The porosity of nonwoven mat (fG: fish Gelatin, g: gallic acid, T: Tween
Conclusions
In this study, nonwoven mats composed of gelatin, gallic acid, and nicotine-dissolving polymers (liquid paraffin or Tween 80) were designed and developed to efficiently filtrate and adsorb nicotine, and nicotine-dissolving polymers were able to dissolve the nicotine from sidestream smoke, causing it to react with the gelatin structure and alerting users to the health risk of the air environment. Regarding the nicotine-dissolving polymers (liquid paraffin or Tween 80), they could not only capture the nicotine molecules, but also oxidize, dehydrogenate, and demethylate the nicotine to generate nicotine derivatives that included cotinine and trans-3 -hydroxycotinine. When those nonwoven mats were fumigated with sidestream smoke and simultaneously irradiated by 365-nm UV light, the adsorbed dose of nicotine increased by 18-61% compared with the mats not containing the gallic acid or nicotine-dissolving polymers.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/polym13234245/s1, Scheme S1: Experimental steps of the nonwoven mats fumigated with sidestream smoke, Scheme S2: Powerful yet compact blower attached to portable equipment, Scheme S3: Experimental steps of the pollute filtrate. (A) The combination of the wearable equipment and nonwoven mat, (B) The control plate, (C) The filter plate, Table S1: The porosity of nonwoven mat (fG: fish Gelatin, g: gallic acid, lp: liquid paraffin), Table S2: The porosity of nonwoven mat (fG: fish Gelatin, g: gallic acid, T: Tween 80), Table S3: The integral area and absorbed dose of experimental groups, Table S4: The intensity of nicotine derivatives in Figure 9C-F (fG: fish Gelatin, g: gallic acid, lp: liquid paraffin, F: fumigated, UV: 365 nm), Table S5: The intensity of nicotine derivatives in Figure 10C-F (fG: fish Gelatin, g: gallic acid, T: Tween 80, F: fumigated, UV: 365 nm). | 2021-12-08T16:11:54.955Z | 2021-12-01T00:00:00.000 | {
"year": 2021,
"sha1": "40507260111b386e78f1404b72b23bd3e414d6ae",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2073-4360/13/23/4245/pdf",
"oa_status": "GOLD",
"pdf_src": "PubMedCentral",
"pdf_hash": "8500ec63a995f8de24c320254effae87c0498017",
"s2fieldsofstudy": [
"Chemistry",
"Materials Science",
"Medicine"
],
"extfieldsofstudy": [
"Medicine"
]
} |
133870925 | pes2o/s2orc | v3-fos-license | Silicomanganese and Ferromanganese Slags Treated with Concentrated Solar Energy †
Solar energy when properly concentrated offers a great potential in high temperature applications as those required in metallurgical processes. Even when concentrated solar energy cannot compete with conventional metallurgical processes, it could find application in the treatment of wastes from these processes. These by-products are characterized by their high metallic contents, which make them interesting as they could be a raw material available in the own factory. Slags are one of these by-products. Slags are most of them disposed in controlled landfill with environmental impact, but also with economic impact associated to the storing costs and the metallic losses. Here we propose the treatment of ferromanganese and silicomanganese slags with concentrated solar energy with the purpose of evaluating the recovery of manganese from these slags.
Introduction
Solar energy has taken growing interest in the last decades with the development of renewable energy sources, and particularly when properly concentrated, this energy source can be used in high temperature applications, as those that can be read in Fernández-González et al., 2018 [1], and, for instance, in our case, we used solar energy in the synthesis of calcium aluminate cement [2] and in the iron metallurgy [3].
Slags are usually recycled in the own process or disposed in controlled landfill [4][5][6][7][8], but they contain significant quantities of metal and recovering it from the slag could be interesting since the economic and environmental point of view.This paper is dedicated to the utilization of concentrated solar energy in synthesizing ferromanganese, and in the treatment of ferromanganese and silicomanganese slags.
Materials and Methods
Experiments were carried out in a 1.5 kW vertical axis solar furnace located in Odeillo and belonging to the PROMES-CNRS (Procédés, Matériaux et Énergie Solaire-Centre National de la Recherche Scientifique).The solar furnace is based on that the sun radiation strikes a heliostat, which tracks the sun and reflects the radiation towards a 2 m. in diameter parabolic concentrator (Figure 1).This parabolic concentrator makes solar radiation converge at a focal point of 12-15 mm. in diameter.In this way, solar radiation (700-1100 W/m 2 ) is concentrated by four orders of magnitude.A shutter was used to control the power applied to the sample.Mixtures of Fe2O3, MnO2 and carbon (laboratory quality reagents; different carbon excesses, 10, 15, 25 and 40% wt.), were loaded in mullite crucibles of 75 mm in length, 12 mm in width and 8 mm in depth (see in [9] a full description of the same process used in the case of Fe2O3 + C mixtures), having as objective synthesizing ferromanganese.A series of thermocouples were located at the bottom of the crucible (outside) to register the temperature (max.temperature >1200 °C).Crucible displaced at a controlled speed below the focal point to treat all the material loaded in the crucible.
Mixtures of ferromanganese and silicomanganese slags (industrial slags, Table 1), some with calcium carbonate to liberate manganese from them, were prepared separately in crucibles of tabular alumina with 55 mm in height, 30 mm in upper diameter, 25 mm in lower diameter and 3 mm in thickness of the crucible walls.The crucible was located below the focal point (15 mm in diameter) and covered with a glass hood to avoid the projections in the parabolic concentrator.Samples comprised four different types of mixtures of slags: ferromanganese (Table 1), silicomanganese (Table 1), ferromanganese with limestone and silicomanganese with limestone.The additions of limestone were performed in an attempt of forming silicates of calcium instead of silicates of manganese and thus liberating manganese from the slag.Initial slags were analyzed through x-ray diffraction, and in the case of FeMn slag it reported glaucochroite ((Ca, Mn)2SiO4), gehleneite (Ca2Al2SiO7) and manganosite (MnO), while in the case of SiMn slag it showed gehleneite (Ca2Al2SiO7), kirschsteinite (CaFeSiO4), melilite (Ca8Al6MgSi5O28) and manganese oxide (IV), in order of abundance.
Results
Samples were analyzed using x-ray diffraction technique in powdered materials.
Synthesis of Ferromanganese
Partially reduced iron and manganese oxides were detected.Manganese oxide (II)(MnO) and manganese oxide (II, III) (Mn 2+ Mn 3+ 2O4) are the partially reduced manganese oxides as the initial manganese oxide was pyrolusite (MnO2).Partially reduced iron phases are not typical, but when they appear they do as magnetite (Fe3O4) and maghemite (γ-Fe2O3, which is like a magnetite with vacancies) while the initial iron oxide was in the form of hematite (Fe2O3).Typical phases, major components in the quantitative analysis discounting amorphous phases, were iwakiite and jacobsite ((Mn 2+ , Fe 2+ )(Fe 3+ , Mn 3+ )2O4), and the manganese oxides.
Treatment of Ferromanganese Slag
Calcium carbonate additions were not sufficient to produce any effect in the slag and thus phases identified in one and another type of sample were similar.In this case, as we will see in the case of the silicomanganese slags, silicates are the main phase in the slag.Gehlenite (Al2Ca2O7Si), glaucochroite (CaMn 2+ SiO4) and kirschsteinite (CaFeSiO4) were identified, the same as manganese oxides (MnO, MnO2, Mn2O3 and Mn 2+ Mn 3+ 2O4).Other phases identified in these samples were: brownmillerite (Ca4(Al, Fe)2O10), fayalite ((Fe 2+ )2SiO4), melilite (Ca8Al6MgSi5O28) and yoshiokaite (Ca(Al, Si)2O4).
Treatment of Silicomanganese Slag
The same as in the other case, calcium carbonate additions were not sufficient to produce any effect in the slag and thus phases identified in one and another type of sample were similar.Gehlenite (Al2Ca2O7Si) and kirschsteinite (CaFeSiO4) are the main phases.Manganese oxides (MnO, Mn2O3 and Mn 2+ Mn 3+ 2O4) are also representative in the x-ray diffraction analyses.Other phases identified in the samples were melilite (Ca8Al6MgSi5O28), manganese aluminate oxide (Mn2AlO4) and glaucochroite (CaMn 2+ SiO4).
Discussion
Synthesis of ferromanganese: Ferromanganese was not detected during the experiments.Partially reduced mixed manganese and iron oxides are detected, as we found jacobsite and iwakiite in the samples as main constituents (Mn 2+ , Fe 2+ )(Fe 3+ , Mn 3+ )2O4 (substitutions and vacancies as expressed), while in the initial mixture we had Mn 4+ and Fe 3+ .Regarding the Mn and Fe individual oxides, we found MnO and Mn3O4 (Mn 2+ Mn 3+ 2O4), and Fe3O4 (FeOFe2O3) and γ-Fe2O3 Working under ambient atmosphere is unfavorable for the reduction reactions (even when using carbon excesses) because carbon was burned during the process and left the sample without fully reducing the load.Thermal decomposition allows the partial reduction of the mixture, but neither this method nor the carbon (nor the combined effects) allows obtaining the ferromanganese (Fe, Mn).Reactions solidsolid are difficult, and the expected solid-gas (carbon monoxide, Boudouard mechanism) reaction is minimized because hot gas tends to leave the sample without reducing because of the venetian blind (shutter) that allows the entrance of air from outside of the building causing circulation of air.The presence of a glass hood connected to a reducing or inert atmosphere would have been positive for the obtaining of ferromanganese.Temperature is also problematic, as if the temperature is increased to maximum values we did not improve the results, but we achieve the melting of the crucible and the appearance of silicates.
Ferromanganese and silicomanganese slags: Results are not satisfactory, complex silicates are the main phases identified together with manganese silicates and oxides (these in lower quantities).Calcium carbonate additions (6 % wt.) did not play role in the treatment of the slags because it should have destroyed the silicates and liberate Mn.The treatment is also limited by the depth that is possible to reach (20 mm in depth).Increasing the additions could have improved the results, the same as the stirring during the process.If Mn would have been transformed into oxide (well-developed and with proper size), the sample could have been grinded and milled, and manganese oxides could have been recovered through gravimetric methods.These unfavorable results could make more interesting alternatively hydrometallurgical processes for manganese residues, as for instance, that used by Fernández-González et al., in the treatment of anodic lodes and scrapings from the zinc electrolytic process [10].Maybe the addition of any reductant reagent or even increasing the quantity of calcium carbonate (or lime if calcined) would liberate manganese from the silicates present in the slag.Anyway, and as opposed to the iron by-products where iron can be removed from them through magnetic methods, manganese should be separated through gravimetric methods, and this requires a proper development and growth of the manganese phase to obtain it with the proper size for the gravimetric separation.
Conclusions
Concentrated solar thermal was used to obtain of ferromanganese.The lack of reducing atmosphere impeded the presence of reducing conditions that could have reduced iron and manganese oxides to obtain ferromanganese.This way, only partially reduced phases, as iwakiite and jacobsite, were detected.
Concentrated solar thermal was also used to treat silicomanganese and ferromanganese slags although with negative results.Manganese was not liberated from the slag as oxide (although oxides were identified) or any other phase.Additions of calcium carbonate were not enough to promote the destruction of manganese silicates and thus liberating this element.Further studies should be necessaries to evaluate the possibility of recovering manganese from these slags.
Figure 1 .
Figure 1.Experimental equipment used in the experiments.Parabolic concentrator and experimental device (right) and heliostat (left). | 2019-04-27T13:12:14.813Z | 2018-11-15T00:00:00.000 | {
"year": 2018,
"sha1": "cde4b919f22a574be74d6d55acfc1839662cb2d3",
"oa_license": "CCBY",
"oa_url": "https://www.mdpi.com/2504-3900/2/23/1450/pdf?version=1542255534",
"oa_status": "HYBRID",
"pdf_src": "ScienceParseMerged",
"pdf_hash": "cde4b919f22a574be74d6d55acfc1839662cb2d3",
"s2fieldsofstudy": [
"Materials Science"
],
"extfieldsofstudy": [
"Environmental Science"
]
} |
259213499 | pes2o/s2orc | v3-fos-license | Drug-induced change in transmitter identity is a shared mechanism generating cognitive deficits
Cognitive deficits are long-lasting consequences of drug use, yet the convergent mechanism by which classes of drugs with different pharmacological properties cause similar deficits is unclear. We find that both phencyclidine and methamphetamine, despite differing in their targets in the brain, cause the same glutamatergic neurons in the medial prefrontal cortex of male mice to gain a GABAergic phenotype and decrease expression of their glutamatergic phenotype. Suppressing drug-induced gain of GABA with RNA-interference prevents appearance of memory deficits. Stimulation of dopaminergic neurons in the ventral tegmental area is necessary and sufficient to produce this gain of GABA. Drug-induced prefrontal hyperactivity drives this change in transmitter identity. Returning prefrontal activity to baseline, chemogenetically or with clozapine, reverses the change in transmitter phenotype and rescues the associated memory deficits. This work reveals a shared and reversible mechanism that regulates the appearance of cognitive deficits upon exposure to different drugs.
We then examined the expression of the GABA vesicular transporter (VGAT) and VGLUT1 in mCherry + neurons that co-expressed or gained GABA. We used fluorescent in situ hybridization (FISH) to detect transcripts for mCherry, for the GABA synthetic enzyme (GAD1), and for either VGAT or VGLUT1.
To reveal changes in expression levels, we selectively decreased the amplification for VGAT and VGLUT1 to obtain punctate staining (Fig. 1, E and F). In the PL of PCP-treated mice, the density of neurons expressing both mCherry and GAD1 (mCherry + /GAD1 + ) was 1.6-fold higher than in controls ( fig. S3F), mirroring the PCP-induced increase in number of mCherry + /GAD67 + neurons (Fig. 1C). Across treatments, mCherry + /GAD1 + neurons expressed VGAT at the level of GABAergic neurons (labeled with GAD1 and not mCherry) ( Fig. 1G and fig. S3, A, B and D). At the same time, the expression level of VGLUT1 in mCherry + /GAD1 + neurons decreased by ~55% compared to that of glutamatergic cells expressing mCherry and not GAD1 in PCP-treated mice (Fig. 1H) and by 46% in saline controls (fig. S3, C and E). Thus, both the glutamatergic neurons that gained GABA after PCP-exposure, as well as those expressing GABA in drug-naïve conditions, are characterized by high expression levels of VGAT and lowered expression levels of VGLUT1.
We next asked whether PCP-treatment affects the transmitter phenotype of PL GABAergic neurons. No difference was observed in the number of neurons expressing GABA and not mCherry between PCP-treated animals and controls (8594±340 vs 8837±271) (fig. S1, E and F). However, PCP and other NMDA receptor antagonists have been shown to reduce expression of GAD67 and parvalbumin in prefrontal cortex parvalbumin-positive (PV + ) interneurons (16,17). Because variability in the number of GABAergic neurons scored could have prevented the detection of loss of GABA from a small number of PV + neurons, we quantified the number of PV + neurons expressing GAD67 after PCP-treatment by permanently labeling them with a PV CRE ::TdTomato mouse line. PCP caused 223±45 TdTomato + neurons to stop expressing GAD67 ( fig. S4). This result suggests that, except for its effect on PV + neurons, PCP does not affect the GABAergic phenotype of PL interneurons.
To investigate whether glutamatergic neurons that have gained GABA contribute to cognitive deficits, we selectively suppressed GABA expression in PL glutamatergic neurons by injecting a Credependent adeno-associated virus (AAV) expressing shRNA for GAD1 (AAV-DIO-shGAD1-GFP or AAV-DIO-shScr-GFP as control) in the PL of VGLUT1 CRE mice before exposure to PCP (Fig. 1, I and J). shGAD1 suppressed GABA expression in transfected cells ( fig. S5, A to C), and reduced the number of PL neurons co-expressing GAD1 and VGLUT1 transcripts in both PCP-and saline-treated mice to half of that in saline-ShScr controls (Fig. 1K). Having efficiently overridden PCP-induced gain of GABA, we examined the impact of shGAD1 on PCP-induced behavior. While shGAD1 did not affect PCP-induced hyperlocomotion on the first day of treatment, it prevented appearance of locomotor sensitization after a 10-day PCP-exposure ( Fig. 1L and fig. S5, D and E), indicating that PCP-induced gain of GABA is required for sensitization to the acute locomotor effect of the drug. We next focused on deficits in recognition and working memory, since these behaviors are affected by repeated exposure to PCP (13,18) and are regulated by the PL (19,20). shGAD1 prevented PCP-induced impairments of recognition memory in the novel object recognition test (NORT) (Fig. 1M) and deficits in spatial working memory in the spontaneous alternation task (SAT) (Fig. 1N). Neither Glutamatergic neurons that co-express GABA or gain it after treatment with either drug were most prevalent in layer 2/3 and layer 5 of the PL (Fig. 1D, Fig. 2C, and fig. S7). These PL layers innervate the nucleus accumbens (NAc) (23), which modulates behaviors that are affected by repeated intake of PCP or METH (24)(25)(26). To determine whether neurons that switch transmitter identity project to the NAc, we injected fluoro-gold (FG) into the NAc of VGLUT1 CRE ::mCherry mice, treated them with PCP, METH or saline, and screened the PL for mCherry + /GABA + neurons expressing the retrograde tracer ( fig. S8, A to C). In both PCP-and METH-treated mice, ~0.9% of FG + neurons were mCherry + /GABA + . Such cells were less frequent PCP+PCP; P+M: PCP+METH). (K) Experimental protocol to learn whether serial administration of PCP and METH changes the transmitter phenotype of the same number of neurons as PCP alone, causes neurons that have gained GAD1 to lose it, or enables other neurons to gain GAD1. (L) Quantification of these neurons in mice treated as described in (K) (n=4 to 5 mice per group). Scale bar, (G) 20 µm. Statistical significance (**P<0.01, ****P<0.0001) was assessed using unpaired t-test or Mann Whitney test (B), Kruskal-Wallis followed by Dunn's test (D and E), and two-way ANOVA followed by Tukey's test (I and L). Data are mean ± SEM. Further statistical details are presented in Table S2.
in controls (~0.3% of the total number of FG + neurons) ( fig. S8D), indicating that neurons changing transmitter identity with drug-treatment project to the NAc.
Since both PCP and METH affect the transmitter phenotype of PL glutamatergic neurons that have the NAc as a shared downstream target, we asked whether both drugs change the transmitter identity of the same cells. If PCP and METH changed the transmitter identity of different cells, administering the two drugs one after the other should induce gain of GABA in neurons that have not gained it after treatment with the first drug. To determine whether this was the case, we genetically labeled neurons expressing GABAergic markers during the interval between the delivery of PCP and METH, using VGAT FLP ::CreER T ::TdTomato cON/fON mice in which neurons expressing VGAT at the time of tamoxifen administration are permanently labeled with TdTomato (Fig. 2F). We first injected tamoxifen in salinetreated controls and determined that TdTomato labels neurons co-expressing VGLUT1 and GAD1 in drugnaïve conditions with 77% efficiency and 79% specificity (Fig. 2G, and fig. S9).
We then used this labelling approach to distinguish neurons expressing GAD1 in drug-naïve mice from those gaining it upon drug-exposure, by administering mice with PCP after saline-and tamoxifentreatment (Fig. 2H). PCP administration increased the total number of PL VGLUT1 + /GAD1 + neurons 2-fold compared to controls (1188±23, saline+PCP; 582±27, saline+saline) (Fig. 2I), in line with previous findings (Fig. 1K). We detected no differences in the number of VGLUT1 + /GAD1 + /TdTomato + neurons (441±43, saline+PCP; 447±34, saline+saline) and VGLUT1 + /TdTomato + neurons (99±61, saline+PCP; 120±8, saline+saline) between saline+PCP mice and saline+saline controls (Fig. 2, G to J). Changes in these numbers would have indicated that drug-treatment caused some glutamatergic neurons co-expressing GAD1 in drug-naïve conditions to lose expression of GAD1. The results indicate that PCP induces expression of GAD1 in a population of PL neurons that were not previously expressing it, without affecting the transmitter phenotype of cells co-expressing GAD1 and VGLUT1 in drug-naïve conditions. We next used VGAT FLP ::CreER T ::TdTomato cON/fON mice to determine if PCP and METH cause the same neurons to switch transmitter phenotype. Mice were treated first with PCP followed by tamoxifen administration, and then treated with either saline, METH or PCP (Fig. 2K). Across treatment groups the total number VGLUT1 + /GAD1 + neurons was unchanged (1169±46, PCP+saline; 1273±69, PCP+PCP;
Drug-induced prelimbic hyperactivity mediates the change in transmitter phenotype and linked cognitive deficits
Demonstration that both PCP and METH have the same effect on the transmitter phenotype of the same PL glutamatergic neurons prompted investigation of the underlying mechanism of drug action.
Increased neuronal activity can cause neurons to change the transmitter they express (8,27,28). Could PCP and METH induce alterations in PL activity that mediate the switch in PL neuron transmitter phenotype? Both PCP and METH increased c-fos expression in PL glutamatergic neurons by 3.8-and 3.5fold after a single injection and by 2.6-and 3.7-fold throughout a 10-day treatment ( fig. S10, A to D, and F and G). This PL hyperactivity was still present 2 days after the end of drug-treatment (fig. S10, I to K). To determine whether this increase in activity promoted the switch in transmitter phenotype, we tested whether suppression of PL hyperactivity would prevent glutamatergic neurons from gaining GABA.
Glutamatergic cells in the PL receive perisomatic inhibition from local PV + interneurons, which do not show changes in c-fos expression after administration of PCP or METH ( fig. S10, E and H). We hypothesized that chemogenetic activation of PV + neurons would suppress drug-induced hyperactivity of glutamatergic cells (29,30). To test this idea, we expressed the chemogenetic receptor PSAML-5HT3HC in mPFC PV + neurons and administered PSEM 308 immediately before acute injection of either PCP or METH We then combined chemogenetic activation of PL PV + interneurons with either PCP-or METHadministration for the duration of drug-treatment (Fig. 3, A and B). The number of VGLUT1 + /GAD1 + neurons in the PL of PCP-and METH-treated mice that received PSEM 308 was half of that of mice that did not (586±18 and 611±23 vs 1262±66 and 1222±12) and was indistinguishable from that of saline-treated controls (Fig. 3, C and D). Thus, suppression of drug-induced PL hyperactivity is sufficient to prevent glutamatergic neurons from switching their transmitter identity, indicating that hyperactivity mediates the change in transmitter phenotype.
We now tested whether blocking the change in transmitter phenotype through chemogenetic activation of PV + neurons was sufficient to prevent drug-induced cognitive deficits. In mice treated with PSEM 308 , drug-induced hyperlocomotion was absent on both the first and last days of treatment (fig. S13, A to C, and fig. S14, A to C), consistent with suppression of acute drug-induced hyperactivity of PL glutamatergic neurons (31). Suppressing PL activity prevented both PCP-and METH-induced appearance of memory deficits in both the NORT and the SAT (Fig. 3, E to H, fig. S13, F and G and fig. S14, F and G), without influencing exploratory behaviors ( fig. S13, D and E, and fig. S14, D and E). The results indicate that chemogenetic activation of PV + neurons in the PL has no effect on the behavior of saline controls and specifically affects the performance of drug-treated mice. Table S3. GABAergic phenotype for at least 11 days of drug washout (Fig. 3D, and PCP+saline group in Fig. 2L). As PCP-and METH-induced memory deficits are also long-lasting (13,32), we asked whether the persistence of behavioral deficits is linked to retention of the GABAergic phenotype and whether both can be reversed.
Normalizing prelimbic neuron activity after drug-exposure reverses the change in transmitter identity and the associated behavioral alterations
Clozapine, a powerful antipsychotic drug, reverses PCP-induced deficits in the NORT (13), leading us to investigate whether it also reverses the change in transmitter phenotype. VGLUT1 CRE ::mCherry mice that received PCP displayed 1.94 fold more mCherry+/GABA+ PL neurons than controls, 17 days after the end of PCP-treatment, indicating that neurons had maintained the newly acquired GABAergic phenotype ( fig. S15, A to C). In mice that received clozapine treatment after PCP, the number of mCherry + /GABA + neurons was reduced compared to that of mice treated with PCP alone (559±55 vs 1124±94) and not different from that of saline-treated controls ( fig. S15, A to C). Clozapine did not affect the number of mCherry + /GABA + neurons in saline-treated mice, suggesting that this drug selectively reverses the PCPinduced change in glutamatergic neuron transmitter identity. We confirmed that clozapine rescued PCPinduced memory deficits in the NORT and SAT, without affecting the behavioral performance of controls (fig. S15, D to J).
We next investigated the mechanisms underlying clozapine-induced reversal of the switch in transmitter identity. Because clozapine suppresses the acute PCP-induced increase in PL c-fos expression (33,34), we asked whether reversal of the gain of GABA depends on neuronal activity. After drug-treatment, the number of c-fos + glutamatergic neurons was 2.1-and 2.8-fold higher in PCP-and METH-treated mice compared to controls (fig. S10, I to K). Administration of clozapine after PCP-treatment returned c-fos expression to baseline ( fig. S16), suggesting that PL hyperactivity during drug-washout is necessary to maintain the newly acquired transmitter phenotype. If this were the case, suppressing PL hyperactivity after the transmitter switch has occurred should reverse the change. To test this hypothesis, we chemogenetically activated PV + neurons for 10 days to normalize c-fos expression in the PL of PCP-and METH-treated mice after the change in transmitter phenotype had taken place (fig. S17). More than 3 weeks after the end of drug-treatment, glutamatergic neurons in the PL of both PCP and METH-treated mice still displayed the drug-induced GABAergic phenotype (Fig. 3, I to K). Normalizing PL activity decreased the number of VGLUT1 + /GAD1 + neurons in the PL of PCP and METH-treated mice to the level of controls (588±27 and 575±9 vs 1225±38 and 1071±42) (Fig. 3, I to K). Thus, PL neuronal activity maintains the transmitter switch once it has been induced. Chemogenetically activating PL PV + interneurons after the change in transmitter phenotype had occurred also rescued memory deficits in the NORT and SAT and suppressed locomotor sensitization to both PCP and METH (Fig. 3, L to O, fig. S18 and fig. S19). Overall, these data show that suppressing PL hyperactivity following drug-exposure reverses the change in transmitter phenotype and the associated behavioral alterations. Table S4.
PCP, METH, and other addictive substances increase phasic firing of dopaminergic neurons in the ventral tegmental area (VTA) (35,36) and increase the levels of extracellular dopamine (DA) in the striatum and prefrontal cortex (37,38). Could DA signaling be a common mediator for the PCP-and METH-induced transmitter switch? To address this question, we tested whether suppressing the activity of VTA prevented the increase in the number of PL VGLUT1 + /GAD1 + neurons (Fig. 4, A to D). These results show that drug-induced increase in activity of VTA dopaminergic neurons is required for PL neurons to change their transmitter phenotype upon treatment with PCP or METH.
It remained unclear, however, whether DA signaling alone is sufficient to induce PL neurons to switch transmitter phenotype, or whether non-dopaminergic effects of PCP or METH are involved. Phasic firing of VTA neurons can be mimicked by optogenetic stimulation of VTA dopaminergic neurons (39). We asked whether repeated optogenetic stimulation of VTA dopaminergic neurons is sufficient to induce PL glutamatergic neurons to switch transmitter identity. We expressed ChR2-YFP (or YFP as control) in VTA DAT CRE neurons and implanted an optic fiber above the VTA (Fig. 4, E and F Expression of c-fos in PL glutamatergic neurons was also increased by 2.7-fold, resembling the effect of a single dose of PCP or METH ( fig. S22, A to D). We then exposed mice to 1h of VTA stimulation per day for 10 days and analyzed the transmitter phenotype of PL glutamatergic neurons (Fig. 4E). Remarkably, the number of VGLUT1 + /GAD1 + was 1.7-fold higher in ChR2-expressing mice compared to controls (Fig. 4, G and H, and fig. S22, E to G), demonstrating that phasic firing of dopaminergic neurons in the VTA changes the transmitter phenotype of PL glutamatergic neurons. These findings establish DA signaling as a common mediator for PCP-and METH-induced gain of GABA in PL glutamatergic neurons and suggest that exposure to other addictive substances that activate the VTA could produce similar effects.
DISCUSSION
We show that a change in the transmitter identity of PL glutamatergic neurons is a shared mechanism underlying both PCP-and METH-induced cognitive deficits. Both drugs cause the same PL neurons to acquire a new transmitter phenotype characterized by expression of GABA, GAD67, and VGAT, combined with lower levels of VGLUT1. Other PL neurons show the same transmitter phenotype in drug naïve conditions, as suggested by earlier studies (40,41). This change in transmitter phenotype causes locomotor sensitization and memory deficits in the NORT and SAT, consistent with the involvement of the PL and the NAc, which receives input from PL neurons that change their transmitter identity (19,20,(24)(25)(26)42). Given the role of GABAergic long-range projections in modulating brain oscillation and synchronization (43), gain of GABA by PL neurons projecting to the NAc may contribute to the reduction of NAc firing rates and disruption of cortex-accumbens synchronization after PCP-treatment (44).
Neuronal activity mediates this drug-induced change in transmitter identity, as expected for activitydependent neurotransmitter switching (8,27,28), and is necessary to maintain the newly acquired transmitter phenotype after the end of drug-treatment. Midbrain cholinergic neurons that change transmitter identity in response to sustained exercise spontaneously revert to expression of their original transmitter within a week of cessation of the stimulus (8). In contrast, PL glutamatergic neurons maintain their GABAergic phenotype for more than 3 weeks after the end of drug-treatment and the linked cognitive deficits are long-lasting (13,32). c-fos expression in the PL increases after acute treatment with PCP or METH (33,45), as well as after one hour of phasic stimulation of VTA dopaminergic neurons, and remains elevated for at least two weeks ( fig. S16, fig. S17, and 46, 47). Suppressing drug-induced hyperactivity during or after drug-treatment, respectively, prevents or rescues the switch in transmitter phenotype and the coupled behavioral alterations, indicating that PL hyperactivity is necessary to produce and maintain these changes. Because exposure to either PCP or METH decreases expression of PV and GAD67 (17,48), impaired function of PFC PV + interneurons may contribute to PL glutamatergic neuron hyperactivity and maintenance of the newly acquired transmitter phenotype. Chronic treatment with clozapine also reduces c-fos expression in the PL of PCP-treated mice and reverses PCP-induced changes in transmitter phenotype and behavior. This effect may be mediated by increased inhibitory input to PL glutamatergic neurons (49).
DA signaling is necessary for PL neurons to change their transmitter identity, because suppression of dopaminergic hyperactivity in the VTA during PCP-or METH-treatment prevents it. Optogenetic stimulation of phasic firing of dopaminergic neurons in the VTA is sufficient to cause PL hyperactivity and induce the switch in transmitter phenotype. Many addictive substances promote phasic firing of these dopaminergic neurons (50). Reproducing this firing pattern optogenetically enhances DA release in the NAc (51) and replicates some of the neuroplastic and behavioral effects of cocaine, including conditioned place preference and acquisition of self-administration that can lead to compulsive-like behavior (39,51). This stimulation only models some aspects of drug intake (52) and does not account for the non-dopaminergic mechanisms through which multiple drugs of abuse differentially impact brain function and behavior.
Evidence that multiple drugs of abuse acutely promote mPFC hyperactivity (53,54), and that stimulation of VTA dopaminergic neurons is sufficient to change the transmitter identity of PL neurons, raises the possibility that drugs other than METH and PCP may induce cognitive deficits by switching the transmitter phenotype of PL glutamatergic neurons. | 2022-06-21T13:22:24.347Z | 2023-07-31T00:00:00.000 | {
"year": 2024,
"sha1": "7c1f6b8598d3f0aa68732aced6ee15f5a624781f",
"oa_license": "CCBY",
"oa_url": null,
"oa_status": null,
"pdf_src": "ScienceParsePlus",
"pdf_hash": "2bbe203c02edb9a0c270d5c5fe0a54b15510ea2b",
"s2fieldsofstudy": [
"Biology",
"Psychology"
],
"extfieldsofstudy": [
"Biology"
]
} |
252810209 | pes2o/s2orc | v3-fos-license | Development and implementation of a Direct Surface Description method for free surface flows in OpenFOAM
The solution procedure for two phases in OpenFOAM suffers from unphysical velocity oscillations at the free surface between the two phases. It is likely that this problem also exists in other two-phase Computational Fluid Dynamics (CFD) codes. We aim to solve this by imposing boundary conditions directly on the free surface. We have taken the first step towards a new two-phase solution method by first addressing the water phase alone. It is a free surface modelling method based on merging concepts from two existing methods: (1) A single-phase free surface method and (2) the solution method used in OpenFOAM. The underlying motivation is to enable more accurate estimation of wave induced load distributions from wave crest impacts on offshore structures. This first and foremost requires an accurate prediction of the kinematics near the free surface. We present a solution method with boundary conditions directly on the free surface, thereby the name: Direct Surface Description (DSD). Additionally it is the first time that the isoAdvector algorithm is combined with a single phase free surface method. The implementation is made in OpenFOAM, but may also relevant to other codes as well. First a still water level simulation is presented to illustrate the unphysical behaviour of the existing solvers and validate the behaviour of the DSD method. The second test case is a moderately steep stream function wave in intermediate water depth. The DSD method is validated and compared to the existing solution methods of OpenFOAM: interFoam and interIsoFoam . We present a detailed comparison of surface elevations and velocity profiles. This is followed by a convergence study including wave height, velocity and phase shift. Additionally the influence of the Courant–Friedrichs–Lewy (CFL) number is studied. The stream function wave case demonstrates that the DSD method accurately predicts the free surface elevation and velocity fields without free surface undulations or oscillatory velocity fields. The convergence study underlines an increased accuracy of the DSD method. Finally, a 2D and a 3D showcase with breaking waves are presented to show that the DSD method is capable of simulating more complex and realistic cases.
Introduction
The observation of extreme breaking waves from offshore structures at intermediate water depth in the Danish North sea has made the industry question the safety level and failure probability of the existing structures.Safe operations are important within the offshore industry, which deals with decommissioning and maintenance of oil and gas platforms as well as construction and maintenance of offshore wind turbines.Detailed analyses of wave loads from both breaking and nonbreaking waves are key to reduce the uncertainty of the wave load on individual beams in offshore structures.
Much research have investigated wave loads on monopiles, oriented both vertically as well as horizontally and substantial knowledge is available for the load on fully submerged cylinders.For semisubmerged structures like monopiles or jacket structures it is often 25 years (OpenFOAM-v1912, 2019.We chose to implement our work in OpenFOAM, because (1) it is popular in both academia and industry, (2) it has an active community and (3) the code is open-source and maintained with official releases every six months.The official release of OpenFOAM-v1912 provides two solvers called interFoam and interIsoFoam which are the common choices for simulation of water waves.
The numerical method in interFoam and interIsoFoam solves the governing equations with smooth density differences near the free surface.This is a well accepted method for smoothly varying densities, such as stratified ocean flows.This method was successfully used to study mixing of saline layers due to bridge piers by Jensen et al. (2018).However, the density jump at the interface between water and air is very large and the approach of solving the flow directly with density differences might not be adequate, as the density gradient is unresolved.The white cells are air cells.Wroniszewski et al. (2014) compared the performance of several codes including interFoam in a test case with a progressive solitary wave and highlighted that interFoam over-predicts the velocity at the wave crest relative to the analytical solution.The same problem was also highlighted by Roenby et al. (2017), Amini Afshar (2010), Tomaselli (2016) and Larsen et al. (2019).Amini Afshar (2010) and Larsen et al. (2019) noted a second problem in interFoam, which is generation of wiggles in the interface between air and water.A third problem is the development of spurious velocities in the air region above the free surface, which may adversely affect the velocities in the water near the crest.The spurious air velocity problem has received most attention as indicated by Larsen et al. (2019), who mentioned nine articles published in the time span from 1999 to 2008.All these studies attribute the development of spurious velocities to the surface tension model, however more recently it was shown that spurious air velocities can be generated without surface tension modelling (Wemmenhove et al., 2015;Vukčević, 2016;Vukčević et al., 2017).Larsen et al. (2019) followed a stream function wave for 100 wave periods in a cyclic domain.The study provided a detailed description of the OpenFOAM setup and aimed to establish a best practice for the interFoam solver.They showed that interFoam can propagate waves quite accurately over long distances using specific settings for the numerical schemes, which resulted in a numerical diffusive balance.However, even in the optimised simulation oscillations in the velocity field close the free surface are present.The study also highlighted that interFoam is very sensitive to the settings of the numerical schemes and the applied CFL number, which they recommend to be less than 0.05 to obtain reasonable predictions of the wave kinematics.
The solver interIsoFoam (Roenby et al., 2018) was based on the interFoam solver, where the algebraic advection algorithm Multidimensional Universal Limited Explicit Solver (MULES) (Ubbink, 1997;Rusche, 2002) was substituted with the geometrical advection algorithm isoAdvector developed by Roenby et al. (2016).isoAdvector has been further improved by Scheufler and Roenby (2018), who improved the prediction of the orientation of the iso surfaces on both structured and unstructured grids.The method was introduced in OpenFOAM-v2006 and we have implemented our work in OpenFOAM-v1912.Therefore, we have not been able to study the impact of the new interface reconstruction method in the present work.The improved interface reconstruction is however expected to improve the free surface advection and thereby also improve the overall performance of the DSD solvers.interIsoFoam has been used to propagate a stream function wave by Roenby et al. (2017Roenby et al. ( , 2018) ) and Larsen et al. (2019).The studies showed that isoAdvector improved the prediction of the surface elevation and eliminated the wiggly interface observed with interFoam.However, the velocity prediction by interIsoFoam has severe over-and underprediction, which is much larger than with interFoam.Roenby et al. (2017) used a fine resolution of 20 cells per wave height to get acceptable results, where significant errors could still be observed in both the surface elevation and velocity profile.Vukčević et al. (2017) showed that the continuous density field creates an imbalance in the momentum equation, which causes spurious velocities in absence of surface tension.The discontinuity of the density field was captured with a newly developed Ghost Fluid Method (GFM).Originally the GFM method was developed by Fedkiw et al. (1999), Fedkiw and Liu (2001) and Fedkiw (2001).The GFM method in combination with a novel level set method was shown to accurately propagate waves over long distances by Vukčević (2016), however no clear validation, e.g.velocity profiles, was given for the velocity field.
More recently, GFM was combined with isoAdvector by Vukčević et al. (2018) where they studied a progressive stream function wave train with intermediate water depth and moderate steepness.The study only presented results from a mesh with 26 cells per wave height.This is a relatively fine resolution compared to the literature referenced earlier, where resolutions from 2 to 30 cells per wave height have been seen.They compared different components of the stream function solution, and reported a first-order surface elevation error of 3.81% after eight wave periods and a first-order horizontal velocity error of 4.2% after seven wave periods.The GFM method have also been used for prediction of wave loads by Jasak et al. (2015), Gatin et al. (2018), Liu et al. (2020) and Gatin et al. (2020).More recently the GFM method have also been implemented by Peltonen et al. (2020), who suggests an alternative formulation for reconstructing the free surface position from the volume fraction field of water.Other related work on the GFM method includes Meyer et al. (2016), Wemmenhove et al. (2015), Queutey and Visonneau (2007), Huang et al. (2007), Desjardins et al. (2008) and Egan and Gibou (2020).
The original Marker and Cell method (MAC) and later Surface Markers methods (Harlow and Welch, 1965;Nichols and Hirt, 1975;Chen et al., 1991Chen et al., , 1995;;Christensen and Deigaard, 2001;Raad and Bidoae, 2005) only resolved the fluid domain, which was also the case for the original Volume of Fluid Method (VOF) by Hirt and Nichols (1981).Many of these methods were developed for structured grids.Nielsen (2003) and Christensen (2006) developed and used the CFD code NS3, which only solves the governing equations in the water region and models the behaviour at the interface.The interest in a single fluid free surface method for OpenFOAM was already mentioned by Jacobsen et al. (2012), because the air flow induced by the wave motion is irrelevant in many engineering applications.Jacobsen et al. (2012) observed that the CFL criterion is sometimes dominated by velocities in the air, which restricts the time step and thereby increases the simulation time.Furthermore, it is also mentioned that only solving the governing equations in the water region would reduce the cell count significantly and thereby substantially reduce the simulation time.
The Direct Surface Description (DSD) method has been inspired by Nielsen (2003) and Christensen (2006), however the methods differ in: (1) the discretisation of the Laplacian operator in the pressure equation, (2) how the continuity equation and momentum equation is coupled and (3) how the fields are extrapolated near the free surface.
The DSD method has been implemented in the general polyhedral FVM framework of OpenFOAM, whereas NS3 is limited structured hexahedral meshes.The general polyhedral discretisation in OpenFOAM limits the implicit discretisation to the two immediate neighbours of each cell face, which gives a compact stencil for the Laplacian operator.NS3 uses a larger, higher-order stencil for the Laplacian operator, which is valid on structured hexahedral meshes.Another difference between NS3 and the DSD method is the advection scheme, where NS3 applies the algebraic Compressive Interface Capturing Scheme for Arbitrary Meshes (CICSAM) scheme (Ubbink and Issa, 1999) and the DSD method uses isoAdvector.
The DSD method eliminates the air phase, which is treated like a void phase with constant pressure.In the present work simple cases with a well defined free surface and theoretical solution are considered for validation and comparison to the existing solvers.The ultimate goal is to use the DSD method to predict wave loads on structures from breaking wave impacts.During such simulations pockets of air may be occluded by the water phase, leading to a compression of the air phase and impulsive loading.It should be emphasised that this cannot be captured by the method in its present form, because the air phase is treated as a void phase with constant pressure.In order to capture these effects the method must be extended to simulate the flow in both the water and air phase.Furthermore, the assumption of incompressibility of the air phase needs to be relaxed meaning that the air phase should be modelled as an compressible phase.This is beyond the scope of the present paper, where the main focus is to solve the issue with the spurious velocities near the free surface.However, it is certainly an aspect which needs to be studied in the future.
The objective of the present work is to develop an efficient single phase method for free surface flows in OpenFOAM based on Christensen (2006) and Vukčević et al. (2017), which can accurately predict the free surface elevation, pressure fields and the wave kinematics without resolving the boundary layer near the free surface.Furthermore, we aim to study: (1) the influence of different temporal integration methods (2nd order Adams-Moulton, 2nd order explicit Runge-Kutta and 4th order explicit Runge-Kutta), (2) the convergence, (3) the influence of the CFL number and (4) the computational speed.
Mathematical model
This section introduces the mathematical background for the numerical methods presented.A nomenclature is provided in Appendix A.
Governing equations
The present study analyses a flow of two incompressible Newtonian fluids in a gravitational field, where the fluids are separated by a sharp interface.The density, , is eliminated from the general form of the continuity equation, when the fluids are incompressible and is piece-wise constant.The continuity equation for an incompressible fluid is where is the velocity vector.In our application for ocean waves the heavy phase is water with a density of = 1000 kg∕m 3 and the light phase is air with a density of = 1 kg∕m 3 .Conservation of momentum for each fluid is described by the incompressible Navier-Stokes equations: where is time, is the kinematic viscosity, is the gravitational acceleration vector, and is the total pressure.The gravity term from Eq. ( 2) is reformulated to 1 ∇ ( ⋅ ) assuming a constant density field.The momentum equation is rewritten to The density is isolated in a single term accounting for both pressure and gravity.The expression inside the gradient on the right hand side of Eq. ( 3) is denoted ℎ = − ⋅ and called the relative pressure.It is the total pressure minus the hydrostatic pressure.Sometimes this expression is also called the dynamic pressure, however this is only valid if is determined with respect to the still water level.The final form of Eq. (2) becomes The equation for ℎ is presented later in Section 3.1.4after the discretisation has been introduced.
Interface capturing
The interface between the two phases is captured with the VOF method originally introduced by Hirt and Nichols (1981).The volume fraction of the water phase is computed by Hence, = 1 indicates that a cell is filled with water and = 0 that a cell is filled with air.Cells with 0 < < 1 represent the interface.The VOF advection equation for the water phase is derived from the continuity equation Eq. (1) and reads According to Hirt and Nichols (1981), a VOF method needs to provide: 1.A numerical description of the location and shape of the free surface 2.An algorithm for the time evolution of the free surface 3. A scheme for imposing the desired free surface boundary conditions in the discretisation of the governing equations interFoam is based on a diffusive interface description, which is not in agreement with the original VOF method.This is handled with a ''work-around'' solution introducing the compressibility algorithm MULES.The interIsoFoam method uses an advection algorithm which ensures that the volume fraction field remains sharp and well defined.However, neither interFoam nor interIsoFoam provide a scheme for imposing free surface boundary conditions, the third condition above.
The interIsoFoam solver is missing a scheme for imposing boundary conditions at the free surface.To implement boundary conditions in the discretisation of the governing equations the location of the free surface is needed.Cells with > 0.5 are denoted wet and cells with < 0.5 are denoted dry.The free surface location along the vector spanning from the face owner cell to the face neighbour cell is defined as where is the normalised location of the free surface between and , see defined as whereas Christensen (2006) defined as a conditional expression where is the face normal vector and is the face tangential vector.The gradient of the volume fraction normal and tangential to the cell face is defined as Fig. 1(a) illustrates the projections performed in Eq. (10).
Free surface boundary conditions
At the interface between two phases the kinematic and dynamic free surface boundary conditions must be considered.The kinematic boundary condition (Batchelor, 1967) states that the velocity field is continuous across the free surface: Here the light phase is air and the heavy phase is water.The boundary layer around the free surface due to the kinematic condition is difficult to resolve in numerical simulations of ocean waves due to the needed spatial resolution.The kinematic boundary condition also makes the continuity equation valid for two incompressible phases.The dynamic boundary condition follows from momentum conservation, where the forces acting on the fluid at the interface are in equilibrium.Following Tuković and Jasak (2012) the tangential stress balance is where is the unit normal vector on the interface pointing from heavy towards light fluid, ∇ = ∇ − ⋅ ∇ is the surface gradient operator, , = ( − ) ⋅ is the tangential velocity component, is the identity tensor and is the surface tension coefficient.The normal stress balance yields a relative pressure jump across the free surface that reads where = −∇ ⋅ is twice the mean curvature of the interface.
The first term models a pressure jump due to surface tension.The second term −2( − )∇ ⋅ models a pressure jump due to the normal viscous force at the interface expressed in terms of the surface divergence of the interface velocity according to Chen et al. (2000).
The last term ( − ) ⋅ models a pressure jump caused by the density jump at the free surface.The free surface boundary conditions are simplified through assumptions (Vukčević et al., 2017).The tangential stress balance in Eq. ( 12) is neglected, as it is of minor importance for free surface flows at high Reynolds numbers (Huang et al., 2007).The normal stress balance in Eq. ( 13) simplifies to assuming a continuous viscosity field and neglecting the effect of surface tension, which is valid for high Weber number flows.The simplified free surface conditions yields an additional condition, which follows from inspection of the terms in the momentum equation in Eq. ( 4) (Vukčević, 2016).Since the temporal, convective and diffusive terms are continuous, the pressure gradient term on the right hand side must also be continuous: As the DSD method only models the flow in the water, we need to reformulate the boundary conditions at the free surface.The dynamic boundary condition from Eq. ( 14) is reformulated and total pressure in the air is approximated with a constant total pressure: = 0.The flow is driven by pressure gradients, and assuming = 0 in the air The total pressure at the free surface on the air side is set to = 0.The final relative pressure boundary condition is then which defines an inhomogeneous Dirichlet condition that must be enforced at the location of the free surface.It is not possible to derive a free surface boundary condition for the velocity at the free surface when only the water phase is considered.The movement of the free surface should correspond to the velocity at the free surface, but that does not provide a boundary condition.Instead we use linear spatial extrapolation from internal cell centres to compute velocities at face and cell centres outside the water domain needed in the discretisation.Fig. 2 illustrates the definition of a cell centre and a face centre and the extrapolation procedure is described in Section 3.2.4.The air faces of the surface cells receives an extrapolated velocity, which is used to evaluate the divergence term on right hand side of the pressure equation, which is presented later in Section 3.1.4.The air cell faces are those cell faces that have an air cell as its face neighbour.The predicted pressure is thereby directly affected by the extrapolated velocities through the continuity constraint.
Numerical method
During our review of the literature around OpenFOAM, we have noticed that the discretisation of the governing equations is often presented in a very compact form.For the sake of completeness, our description includes a broader introduction to the discretisation in OpenFOAM followed with specific details on our development.The first part describes the fundamental interpolation and discretisation of the momentum equations, the continuity equation and the derivation of the pressure equation.The numerical discretisation is presented with the implicit time integration used in OpenFOAM.Especially the details of the derivation of the pressure equation is often hidden in the compact notation.The knowledge about these details is necessary to understand the underlying numerical method and the developed new method.The second part is dedicated to aspects of the numerical method and discretisation of the free surface boundary specific to the DSD method.The first section describes the discretisation of the pressure gradient at the free surface.The second section describes the movement of the free surface.The third section defines the extrapolation practice used at the free surface boundary in the DSD method.The fourth section describes how the least squares gradient calculation has been modified in OpenFOAM for cells at the free surface.Lastly, the fifth section describes the explicit Runge-Kutta time integration of the momentum and pressure equations.Flowcharts for the solvers dsdFoam, dsdrkFoam and interIsoFoam can be found in Appendix B.
Interpolation
In OpenFOAM linear interpolation determines face values from cell values by where is a general dependent variable and the overbar indicates linear interpolation.assumed to intersect the face centre.It is possible to apply a skewness correction, 1 when ⃖⃖⃖⃖⃖⃖ ⃗ does not intersect the face centre, but it is not used in this work.
Momentum equation
The time integration in OpenFOAM is performed according to Section 6.2.4 (Ferziger and Perić, 2002).The momentum equation in Eq. ( 4) is integrated over a time step centred around : The integral is approximated with the mid-point rule, which yields The equation is divided by and integrated over a control volume : The mesh is assumed to be static in the present derivation.
Temporal term.The implicit time derivative is discretised by secondorder Adams-Moulton scheme. 2 OpenFOAM uses a finite difference formulation of the temporal discretisation, which is described in Section 13.4.2(Moukalled et al., 2016).The velocity at − and −− 0 is denoted 0 and 00 .The scheme is derived by expressing 0 and 00 with a Taylor series expansion around .The discretised form of the acceleration term for non-uniform time steps reads where the algebraic coefficients are defined as The first time step in a simulation is discretised with the first-order forward Euler scheme.
1 Specified by the keyword skewCorrected in OpenFOAM.
2 Known by the keyword backward in OpenFOAM.
Convection term.The volume integral of the convection term is transformed to a surface integral with Gauss' theorem.The surface integral is approximated with a summation over the cell faces, where the face integral is approximated with the mid-point rule: The summation over cell faces is split in two sums for the owned and neighbouring cell faces to account for the face normal vector always pointing from owner cell P towards neighbour cell N. Hence the above cell face summation assumes the following split, which is valid for all face summations.
The volumetric flux defined at the cell faces is defined as * = ⋅ * .The surface area normal vector is given by = , where is the face area.When the convection term is treated implicitly, the term is linearised by computing * explicitly from the previous time step or iteration.This is indicated with superscript * .The nonlinearity is resolved by the outer iteration loop in the PIMPLE algorithm in OpenFOAM.The PIMPLE algorithm consists of an outer and inner loop.The outer loop is a series of Picard iterations and the inner loop is the Pressure-Implicit with Splitting of Operators (PISO) algorithm (Issa, 1986;Tuković and Jasak, 2012).
Diffusion term.The discretisation of the diffusion term in Eq. ( 26) follows Jasak (1996) and Muzaferija (1994). is the surface of the control volume.
The surface normal gradient ⋅ (∇) is approximated by where the orthogonal face area vector is computed as . This is called the over-relaxed approach (Jasak, 1996).The magnitude of the surface area vector is denoted by two vertical lines as | |.The operation is defined as Pressure gradient term.The volume integral of the pressure gradient term is transformed to a surface integral with Gauss' theorem.The surface integral is approximated by a sum over the cell faces, where the integral of each face is given by the mid-point approximation: When the pressure gradient term is left undiscretised in the momentum equations the volume integral yields
Continuity equation
The continuity equation from Eq. ( 1) is integrated in time and the integral is approximated using the mid-point rule: Eq. ( 30) is integrated over via Gauss' theorem as a sum over the cell faces, and the integral over each cell face is approximated using the mid-point rule:
Pressure equation
The PISO algorithm implemented in OpenFOAM is described by Tuković and Jasak (2012), Tuković et al. (2018) and Uroić (2019).The following derivation is presented for static meshes.The pressure equation is derived from the discretised continuity equation Eq. ( 31) and a partially discretised momentum equation, where the pressure gradient is not discretised.This form of the momentum equation is commonly known as the semi-discretised form in the literature.The semidiscretised momentum equation presented in Eq. ( 32) is derived from Eq. ( 21) after substituting the discretisation presented in Eqs. ( 18), ( 22), ( 24), ( 27) and ( 29).Finally the semi-discretised momentum equation is divided by to normalise the equation.This leads to where ∑ is the off-diagonal contributions from the neighbouring cells.The diagonal matrix coefficient for cell ( ) corresponds to in Eq. ( 23) and is The off-diagonal matrix coefficient from each cell neighbour is given by and the source term is given by The cell centre velocity, , is expressed from the momentum equation in Eq. ( 32), which leads to The essence of the momentum interpolation method is to express the momentum equation at the cell faces to mimic a staggered grid.This is obtained by interpolating the collocated grid coefficients in Eq. ( 36) to the cell faces.The resulting expression for the cell face velocity using the momentum interpolation method is The above expression is needed in Eq. ( 31), the discretised continuity equation.The terms ( ) are found from linear interpolation of neighbouring cells to cell faces indicated by the overbar.The pressure gradient term The face velocity of the previous time steps 0 and 00 are described by Tuković and Jasak (2012), which explains why small time steps may lead to pressure oscillations and how the solution of the pressure equation is made independent of the time step size using a method developed by Yu et al. (2002).The cell face flux satisfies the discretised continuity equation, whereas the cell centre velocity interpolated to the face centres ( ) does not satisfy the discretised continuity equation.The idea is to correct by the difference between the face normal where is the face normal vector and is the face area.The pressure equation is obtained by substituting the face velocity from Eq. ( 37) in the discretised continuity equation Eq. ( 31): After the pressure equation is solved, the volumetric face flux is computed by and the cell centre velocity is computed by the expression in Eq. ( 36).
To the authors impression, understanding the connection between the theoretical background and the source code in relation to the pressure-velocity coupling is a challenge for a majority of the Open-FOAM users.We hope that linking the presented theory to the variable names used in OpenFOAM will be useful especially for those who are new to or not familiar with the OpenFOAM source code.To highlight the connection between the presented theory and the implementation in OpenFOAM-v1912, 0 and 00 from Eq. ( 38) are inserted in Eq. ( 40), which yields The former equation is rearranged to obtain: in the third line is a correction to the flux from the discretisation of acceleration in the second line, which ensures that the solution is independent of time step size and eliminates pressure oscillations in the solution for very small time steps (Tuković and Jasak, 2012).
The fifth line describes the flux from the pressure gradient.Since ⋅ = the third and fourth line can be further simplified and the expression is rearranged to obtain as given in Box I.The OpenFOAM code terminology is indicated with the braces, where phiHbyA is the uncorrected volumetric flux computed by linear interpolation of HbyA in Eq. ( 36) and rAUf* fvc::ddtCorr(U, phi) is the correction from the source terms of the discretised acceleration term.
Discretisation of cell centred pressure gradient
The cell-centred gradient in cell from Eq. ( 28) forms a sum over values at the cell faces, which is obtained by linear interpolation: As previously mentioned the sum needs to be separated in two summations, due to the sign convention of the surface area normal vector .The sum ∑ includes all the cell faces where is the owner of cell face .Likewise, ∑ ℎ includes all the cell faces where is the neighbour of cell face .The difference in sign comes from the positive sign convention of the surface area normal vector from owner cell towards neighbour cell .However, the Gauss discretisation of the cell gradient assumes that the cell face normal vectors point away from the discretised cell .Hence, the contribution from face to the discretised gradient in is positive, whereas the contribution from face to the discretised gradient in is negative.
The discretisation includes the pressure boundary condition at the free surface using extrapolated pressures in the neighbouring air cells.When the owner cell is water the extrapolated value ℎ, at neighbour cell is given by linear extrapolation: The contribution from face to the computed gradient in cell can be written as follows after inserting ℎ, = ℎ, in Eq. ( 44): When the neighbour cell is water the extrapolated value ℎ, at owner cell is given by: The contribution from face to the computed gradient in cell can be written as follows after inserting ℎ, = ℎ, in Eq. ( 44): The negative sign comes from the aforementioned sign convention of the surface area normal vector from towards .
Discretisation of face normal pressure gradient
The pressure gradient term from Eq. ( 39) at a fully submerged face is discretised with the same scheme used for the diffusion term in Eq. ( 27).The non-orthogonal correction term is omitted in the following derivation, since the only change is the application of the pressure gradient scheme from the previous section.The orthogonal part of the face normal pressure gradient is: At free surface faces where cell is wet and cell is dry, the Dirichlet condition for the relative pressure from Eq. ( 17) gives the following expression for the pressure gradient term: Similarly when cell is dry and cell is wet, the Dirichlet condition ℎ, is implemented by the expression: The density at the cell face is that of the wet phase, i.e = .
Free surface advection
The governing advection equation presented in Eq. ( 6) is solved using the isoAdvector algorithm (Roenby et al., 2016).It geometrically captures the change of the VOF field over a time step based on a fixed velocity translation of an iso-surface reconstructed from the volume fraction field .In brief, the isoAdvector algorithm updates the volume fraction field by Eq. ( 52).
is a list of the cell faces for a single cell. is the total volume of water flowing from the current cell during one time step through face into the neighbouring cell. is approximated from , and in Eq. ( 53), where is the volumetric cell face flux through face , is the surface area vector at face and () is the cell face area of face that is submerged in water as function of time.
Furthermore, it is assumed that the iso-surface, reconstructed from , moves with a constant normal velocity that is representative for the movement of the iso-surface during the time step from to +1 .
When isoAdvector is used in combination with an implicit time integration the solver consists of two loops: An outer and an inner loop.The outer loop iterates over the movement of the free surface and the non-linearity in the convection term in the momentum equations.The inner loop is an implementation of the PISO algorithm for a given number of correction steps.The implicit solver employs an explicit Euler extrapolation in time for the velocity field used to advect the free surface, when the solver passes the first outer corrector.If only a single outer corrector is used the advection of the free surface is expected to be first-order accurate.If the solver employs more than a single outer corrector the velocity field advecting the free surface is found by a central difference approximation between the previous velocity field and the newest estimate of the velocity field to the new time step.This should lead to second-order accuracy.For the implicit method DSD solver and the explicit DSD solvers , and used to advect the free surface are estimated from the previous time instance in the first outer iteration: In all the subsequent outer iterations in the implicit DSD solver and are estimated with a central difference approximation, and is equal to the newest prediction from the previous outer iteration: After the prediction of the new field in each outer iteration, the velocity and volumetric flux fields are reset to the value of the previous outer iteration: The DSD Runge-Kutta solvers use an explicit first-order Adams-Bashforth prediction of the input parameters and to isoAdvector.
Extrapolation
Whenever values are required in face centres or cell centres at the other side of the free surface in the air region, a linear extrapolation based on known cell centre values and gradients is performed.The extrapolation is given in Eq. ( 57).Here subscript indicates the target coordinate point of the extrapolation.The subscript indicates the points, that the extrapolation is based on.It is assumed that the gradient is known at the points , where the extrapolation is performed from.
The expression is a weighted linear extrapolation based on the field value and field gradient in the selected extrapolation cells .The linear extrapolation performed from each extrapolation point is given by the expression . The weight 1∕| − | increases the importance of the extrapolation points close to the target point and decreases the importance of the extrapolation point far away from the target point.The extrapolation is combined with different conditions for the selection of the points included in to yield different extrapolation practices.These conditions are constructed to ensure that extrapolation is only performed from cells in the water region.The selection of extrapolation points can be divided in four cases: 1. Extrapolation to new water cell centres, which have entered the water region during the advection step, where the cell shares at least one cell face with the water region at the former time instance.2. Extrapolation to new water cell centres, which have entered the water region during the advection step, where the cell only shares an edge line or a vertex point with the water region at the former time instance.3. Extrapolation to new water cell faces, which have entered the water region during the advection step.4. Extrapolation of fields at previous time instances used in the discretisation of ∕.
Fig. 3 illustrates the first three extrapolation cases (1-3) from the list above.The first case represents the case where an air cell has entered the fluid region during the advection of the free surface.The target 1 is the cell centre of the new fluid cell and the extrapolated value is based on the field value and the field in gradient the cells marked with 1.These cells shares a cell face with the new water cell 1.
The second case also represents the case where an air cell has entered the fluid region during the advection of the free surface.The target 2 is the cell centre of the new fluid cell and the extrapolated value is based on the field value and the field gradient in the cell marked with 2.This cell only shares a vertex with the new water cell 2.
The third case is an example of the extrapolation procedure, when new face values are needed in new water cells.In this case cells marked by 3 are used to estimate the value at the new cell face marked by 3. The fourth extrapolation case is used in the discretisation of the acceleration term in the momentum equation, when cells with no previous time history are encountered.The velocity field at the previous time steps are extrapolated according to the new location of the free surface.This means that extrapolated values will be assigned to cells, which have changed from air to water during the time interval from the previous time instance to the new time instance.In relation to this it is assumed that cells, which have entered the water region, are limited to a region of a single layer of cells from the old time to the new time.The single layer of cells is illustrated in Fig. 3, where the green cells mark the single layer next to the fluid region marked with blue cells.This restricts the CFL number to < 0.5, however this is not considered a problem, since decreasing the CFL number is a more effective way to get more accurate results compared to increasing the mesh resolution (Larsen et al., 2019).
Gradient calculation at the free surface
The least squares gradient calculation in OpenFOAM is programmed to include all cell neighbours in the calculation (Jasak and Weller, 2000).However, the DSD method does not compute any values in the air, hence the dry neighbouring cells next to the free surface should not contribute to the gradient calculation.Therefore, we have modified the least squares gradient calculation in order to only account for the neighbouring cells categorised as water.
The normal least squares gradient is determined by minimising the error of a plane fitted to a cloud of points, which include cell and its immediate neighbours .The velocity gradient in cell is computed by where is a symmetric 3 × 3 matrix for the three dimensional case.
The × 3 matrices and , the × 3 distance matrix and the × weighting matrix are defined as where is the number of neighbour cells, ⃖ ⃗ is the 1 × 3 velocity vector in cell centre , and ⃖⃖⃖⃖⃖⃖ ⃗ is the 1 × 3 vector spanning from cell centre to cell neighbour centre .
The idea behind this scheme is to evaluate the gradient with the standard least squares method in all cells and then correct the estimated gradients in the free surface cells.The free surface cells are the cells containing one or more free surface cell faces.This is illustrated in Fig. 1(b), where the cell faces along the red line are free surface faces.
For each free surface cell in the water the immediate neighbours inside the water region are identified.A distance weighted average of the neighbouring gradients is computed and assigned to the free surface cell using the extrapolation procedure presented in Section 3.2.4.If cell has one or more fully submerged cells, i.e. cells with no free surface faces, then these cells will be used to extrapolate from.This is extrapolation case 1 in Fig. 3 where the cells marked with 1 are used to estimate value in 1.For some surface cells there will be no immediate neighbour cells fully submerged in water.This is extrapolation case 2 in Fig. 3, where the cell marked with 2 is used to estimate value in 2. The extrapolation of the gradient fields corresponds to a weighted zero gradient condition on the velocity gradient field.In other words, the velocity field is expected to continue to change in the same way in cell P as it does in the immediate neighbouring cells in water.The scheme algorithm can be summarised in the following steps: 1. Compute standard least squares gradient in all cells.
Table 1
Butcher tableau for n-stages, RK2 and RK4. n-stages 2. Extrapolate ∇ according to extrapolation procedure from Section 3.2.4,where is simply replaced by ∇. 3. Set gradient to zero in all air cells.This scheme has proven to be very robust when applied to complex free surface flow simulations.The scheme extracts as much information as possible from the nearest cells, when limiting the search radius to one layer of cells from the target cells.
Explicit Runge-Kutta time integration
The explicit Runge-Kutta scheme can be defined by a Butcher tableau, which enables us to write an algorithm that can handle any explicit Runge-Kutta time integration.The Butcher tableau for an explicit Runge-Kutta scheme with stages, the RK2 scheme and the RK4 scheme are given in Table 1.
Each row in the table represents a single stage in the Runge-Kutta method.After evaluating all stages the final prediction is made from a linear combination of the stages.For explicit Runge-Kutta methods the diagonal elements are zero, which means that the first stage simply evaluates the function ( + 1 , 1 ) = ( , ) from the solution at the last time step.The coefficients and are defined according to the chosen Runge-Kutta scheme, while the coefficients are computed by where is the row index and is the column index in the Butcher tableau.
The implementation of the explicit Runge-Kutta time integration together with an equation for the pressure follows (Sanderse and Koren, 2012).We solve an equation of the form The acceleration term is discretised with the forward Euler scheme, and the pressure gradient term is left undiscretised to form an equation for the pressure solved at each Runge-Kutta stage.After the pressure has been found its contribution is added to the intermediate velocity and volumetric flux at the end of each Runge-Kutta stage.The function (, ) is evaluated from the solution at the previous Runge-Kutta stages or previous time step.The general form of the equation at stage is The unknown velocity is isolated, leading to The intermediate velocity, ũ , is corrected by the pressure gradient after the pressure equation has been solved.The continuity equation must be satisfied at each stage, which yields the pressure equation: The constant works like a Lagrangian multiplier, which can be neglected without affecting the result, as long as it is also removed from the pressure gradient updating the velocity (Sanderse and Koren, 2012).
After the Runge-Kutta stages have been completed, the final prediction of the velocity to the new time step is found by The pressure equation derived from Eq. ( 65) is After the pressure has been found, the velocity and flux are corrected to give the final prediction to the new time +1 .The discretisation of the pressure equation follows the same discretisation procedures presented in Sections 3.1.4and 3.2.2.
Results and discussion
The DSD method has been implemented in two solvers.The first solver is an implicit solver dsdFoam using the second-order Adams-Moulton scheme to discretise the acceleration.The second solver is an explicit solver dsdrkFoam with a Butcher tableau based explicit Runge-Kutta time integration of the pressure-velocity coupling and a first-order Adams-Bashforth prediction of the input variables to the free surface advection algorithm isoAdvector.The solver dsdrkFoam is tested with a second and fourth order Runge-Kutta scheme.The second order scheme is Heun's method and the short name used throughout the paper is DSD-RK2.The fourth order scheme is the low cost RK4 scheme and will be known by the name DSD-RK4.The short name for the implicit solver is DSD.This leads to the three different numerical methods which are compared to interFoam and interIsoFoam.
The solvers are used to simulate two 2D test cases: (1) A still water level and (2) A progressive wave train described by stream function theory.
Still water level test
It may seem trivial to simulate a still water level, but the case is in fact quite challenging since it exposes any imbalances in the numerical model.The test case is a still water level in an inclined square of 1 × 1 m with 50 × 50 cells.We simulate 1 s using a constant time step of = 0.001 s.The case is simulated with interFoam (IF), interIsoFoam (IIF), DSD, DSD-RK2 and DSD-RK4.
Before we developed the DSD method the starting point for our research was the GFM method (Vukčević et al., 2017).The solver GFM is our implementation of the GFM method in OpenFOAM combined with isoAdvector (Vukčević et al., 2018).The only modification is the formulation of the inverse distance from Eq. ( 9).The solver GFM-RK4 is our implementation of the GFM method combined with an explicit fourth-order Runge-Kutta time integration for the pressure-velocity coupling.
Fig. 4 presents the simulated velocity magnitude fields at = 1 s for the seven solution algorithms.The free surface contour is given by a red line.The DSD and GFM methods provide very low velocities caused by the discrepancies between the true position of the free surface and the reconstructed position from the volume fraction field.When the inverse distance is specified based on the known location of the free surface the resulting velocities are of the order 10 −11 m∕s for the DSD solvers and 10 −12 m∕s for the GFM solvers.The maximum velocity in the air is 4.32 m∕s for interIsoFoam and 1.81 m∕s for interFoam.
The air region from the interFoam and interIsoFoam simulations have been clipped out to highlight the effect in the water cells.The imbalance is caused by the linear interpolation of the density field in interIsoFoam and interFoam (Vukčević et al., 2017).However, after the air has been accelerated the linear interpolation of the velocity field accelerates the momentum transfer to the water column through the convection term, as the boundary layer is not resolved.This test shows a general problem with the interFoam and interIsoFoam methodologies.
Progressive surface gravity wave based on stream function wave theory
The stream function wave test case from Vukčević et al. ( 2018) is adopted to enable a qualitative comparison to the performance of their model, which combines the GFM method with isoAdvector.
The domain is 13 corresponding to 70.32 m and the height is 2 m.The water depth is ℎ = 1 m and the still water level is located at = 0 m.The regular wave train is initialised with the stream function wave theory (Fenton, 1988), generated by the waves2Foam toolbox (Jacobsen et al., 2012).The first and last 1.5 are relaxation zones, where the theoretical solution is gradually blended with the numerical solution at the inlet and vice versa at the outlet.
The wave height is = 0.3 m, the wave period = 2 s and the initial phase shift is = rad.The stream function wave is based on a zero Eulerian time mean current.The corresponding OpenFOAM settings are specifyEuler true; and eulerVelocity 0;.The gravitational constant is = 9.81 m∕s 2 .The stream function theory is based on 32 coefficients and the solution is found with 10 iterations.
The derived wave properties are the wave length = 5.409 m, the wave number = 1.162 m −1 , the angular frequency = 3.142 s −1 and the wave celerity = 2.704 m∕s.The steepness of the wave is moderate with a value of ∕ = 0.05546.The relative depth of ℎ = 1.162 tells that the wave is travelling on an intermediate water depth between the shallow and deep water limits: 0.25 < ℎ < 3.14.
The case is simulated on five different meshes with a constant time step that decreases when the mesh is refined to keep the initial CFL number constant at approximately CFL = 0.09.The solvers are capable of handling variable time steps.The time step is kept constant to enable direct comparison between simulations from different solvers.The simulated duration of time is 40 s corresponding to 20 wave periods.The properties of each mesh case is given in Table 2, which also gives the resolution of wave with the number of cells per wave height ∕ and the number of cells per wave length ∕.The cell aspect ratio is = 1.An overview of the applied boundary conditions is provided in Table 3.
The acceleration term ∕ can be discretised with different discretisation schemes.For DSD and interIsoFoam, it is discretised with the second-order Adams-Moulton scheme.The second-order Adams-Moulton scheme cannot be selected in OpenFOAM when using the The boundary condition on the top patch for velocity and pressure are not used in the solvers that only consider the water region, instead we apply the free surface boundary conditions at location of the free surface.b There is no boundary condition on the fluid volume fraction field alpha.water,which is defined in the entire domain.
interFoam.As an alternative a 50∕50% blend of the implicit Euler and Crank-Nicolson schemes is employed.The DSD-RK2 and DSD-RK4 employs a different time integration method using an explicit Runge-Kutta scheme, which is described in Section 3.2.6.The OpenFOAM case settings for interFoam and interIsoFoam can be found at the git repository: https://gitlab.gbar.dtu.dk/jrkp/Article1.git.
Surface elevation
Fig. 5 presents the surface elevation centred around the last wave crest identified within = 10−11 at ∕ = 20.The focus is the shape of the wave, whereas the wave phase error is examined in Section 4.2.4.The wave is propagating from left to right.The black line is the numerical surface elevation and the red line is the theoretical surface elevation.Selected subplots in the figure contain a 300% zoom on the wave crest in a small box.The DSD solvers are better at maintaining the wave height and form compared to interFoam and interIsoFoam.
Local undulations on the free surface are observed in the solution by interFoam.The undulations and the wave length of the undulations decrease in magnitude when the mesh is refined.The cause for the undulations seems to be the MULES advection algorithm, since the undulations cannot be observed in the surface elevation predicted by interIsoFoam.The only difference between the two solvers is the advection algorithm.
The surface elevation predicted by the DSD solvers is smooth without oscillations.It is observed that the explicit DSD solvers DSD-RK2 and DSD-RK4 are slightly more accurate than the implicit solver DSD.The two explicit Runge-Kutta solvers show almost identical results.However, later in Section 4.2.4 a convergence study of the errors is presented, which reveals that DSD-RK4 is the most accurate solver with respect to the surface elevation.
Fig. 6 presents the surface elevation in the interval 9 − 11 at ∕ = 20 for all cases and solvers.The last cycle of identified crests and troughs have been plotted with dots in colours matching the legend.The theoretical crest and trough envelope are shown as black lines and the theoretical crest and trough positions are indicated with red and blue circles.This qualitative comparison clearly illustrates the deficiency of interFoam.It consistently produces a wiggly surface elevation in the cases ∕ = [3, 6, 12, 15], and halves the wave height for ∕ = 1.5.The coarse cases illustrate that the DSD solvers are better at maintaining the wave height also on coarse meshes.In terms of phase shift interFoam and interIsoFoam are also outperformed by the solvers using the DSD method.
In case ∕ = 6 for the interIsoFoam solver the wave crest envelope is observed to oscillate.We have studied this further and found that all cases and solvers have some degree of wave reflection with a period of ∕2.Fig. 7 shows the envelope of the surface elevation with zoomed plots of the crest region at the top and the trough region at the bottom.The oscillations at the crest and trough are in phase with each other and they have a wave length of ∕2.This shows that the phenomenon is a first harmonic reflection.The magnitude of the oscillations in the wave height is around 0.33% of the specified wave height.This is well below 1%, which is the reported amount of reflection generated by waves2Foam for a relaxation zone of 1.5 (Jacobsen et al., 2012).
Velocity fields
The velocity field given by the stream function wave theory is shown in Fig. 8. Figs.9-13 presents the 2D velocity field from case ∕ = 6 at = [10.5− 11.5] and ∕ = 19.5 for the five solution methods.
The velocity field predicted by interFoam appears acceptable in the water.Still, the velocity field in the air shows that interFoam has some fundamental instability issues when simulating surface gravity waves.As illustrated with the still water level test, the air velocities can significantly effect the fluid velocities near the free surface during a simulation over a longer period.The velocity field in the water predicted by interIsoFoam is clearly faster in the upper region of the wave compared to the theoretical solution.Furthermore, comparing the velocity vectors to interFoam they are pointing more forwards at the front of the wave.Very large velocities can be observed in the air showing that accurate advection of the free surface alone cannot eliminate the problems in the air.On the contrary, the problem seems to have worsened.A possible explanation for the intensified velocities is as follows: The geometrical advection algorithm does not smear the volume fraction field as the MULES algorithm.The combination of a sharp advection algorithm with a continuous pressure and velocity field over the free surface seems to make the numerical solution more unstable and incorrect.The reason may be that the two modelling philosophies are not consistent, one part of the method in interIsoFoam keeps the surface sharp, whereas the other tends to smear the solution.
The velocities in the air is obviously not a problem for the DSD solvers, where the air phase is eliminated.For the relatively coarse mesh with 6 cells per wave height small differences between the DSD solvers are observed.
Velocity profiles under the wave crest
From the references in the introduction we have observed that OpenFOAM solvers are often validated against the surface elevation and less or even no effort is spent on validating velocity fields.The wave kinematics are in many cases very important to derived analyses, e.g.wave forces and wave run-up on structures.
Fig. 14 presents the horizontal velocity profile under the last wave crest identified in = 10−11 at ∕ = 20.The velocity data are taken from the column of cells closest to the wave crest.The wave crest has been estimated by a parabolic interpolation of the surface elevation.The offset between the location of wave crest and the position of the column of cells with velocity data is accounted for in the theoretical velocity profile in order to match the wave phase of the theoretical profile to the numerical profile.The small boxes show a 250% zoom of the velocity profile just below the free surface.The interFoam solver underpredicts the velocity over the entire water column for the coarse cases, which is also expected since the wave height is significantly reduced.Near the wave crest interFoam has a tendency of overshooting at the crest and underestimating in the region just below the crest.Furthermore, the velocity profile contains small oscillations, which may be caused by the wiggly surface elevation.interIsoFoam overestimates the crest velocity by 20 − 30%, which is maintained with a finer resolution.Just below the overestimation the velocity is underestimated.
The DSD solvers have a tendency to slightly overpredict the velocity near the wave crest.The velocity profile is also underpredicted over the entire depth for coarsest resolution, which is linked to the wave height decrease.The DSD solvers have a tendency to overshoot the velocity at the top and undershoot the velocity at the bottom.The deviations quickly reduces as the mesh is refined and there are almost no visible deviations once the mesh resolution reaches ∕ = 12.Later in Section 4.2.4 the velocity errors are studied in more detail.The small boxes show a 400% zoom of the velocity profile just below the free surface.interFoam and interIsoFoam underpredict the vertical velocity and the velocity profile suffers from oscillations, which attenuates as the mesh is refined.It can be noted from the zoomed plots of the velocity profile near the free surface, that interFoam and interIsoFoam also produce kinks in the profile for the fine mesh resolutions.The zoom on the DSD solvers emphasises how well the solution agree with the theory.To conclude the DSD solvers provide more accurate velocity profiles without unphysical oscillations compared to interFoam and interIsoFoam.Note that a resolution of 15 cells per wave height is in fact only a medium fine resolution with respect to the range used in the literature.
Simulation time and convergence
This section presents a study of the solver performance with respect to accuracy and simulation time.The mean and maximum error of the velocity magnitude profile (((||)), ((||))), the wave height error (((||))) and the wave phase error (((| ℎ ∕|))) are included in the convergence study.
Data is extracted below the wave crest in the interval = 10 − 11 during the last 5 s of each simulation.Fig. 16a presents the convergence of the maximum error of the velocity magnitude profile.Fig. 16b presents the convergence of the mean error of the velocity magnitude profile.
Fig. 17a presents the convergence of the mean error of the wave height.Fig. 17b presents the convergence of the mean error of the phase shift.The convergence rate of the interFoam solver is unsteady and in a single case the maximum error of the velocity magnitude profile increases from the second last to the last case.The expected behaviour is a constantly decreasing error.It also seems that there is a tendency for the convergence rate of interFoam to flatten out as the mesh resolution is increased, however a larger study on several different test cases is required before a conclusion can be made.
For interIsoFoam the convergence rate of the maximum velocity magnitude profile error attenuates and reaches an almost constant error, which is undesirable.However, the mean error of the velocity and wave height decreases with a fairly constant rate between −1 and −2.Since the Finite Volume method is based on linear interpolation, it is expected that the error should decrease with a slope of −2 in a double logarithmic plot.
The DSD solvers show a convergence rate closer to −1 than −2 for the maximum velocity error.However, for the mean velocity, wave height and phase shift the convergence rate is −2 and even closing in on −3 for the wave height error with DSD and DSD-RK2.This is better than what could have been expected for the explicit DSD solver, since it uses a first order extrapolation in time for the input variables to isoAdvector.The explicit and implicit DSD solvers are almost equally accurate across the presented mesh resolutions, and they are significantly more accurate than interFoam and interIsoFoam.
For the finest mesh case the wave height error is 18 times lower with DSD-RK2 compared to interFoam for the same simulation.In addition the DSD-RK2 solver quarters the CPU time for the current case in comparison to interFoam as seen from Fig. 18.The phase shift error in Fig. 17b is expressed as a fraction of the wave length.The interFoam solver gives the lowest phase error at the second coarsest mesh, but then the error increases again with further mesh refinement.
The interIsoFoam solver got the largest error among all the solvers and the error converges with a slope around −1, which indicates first order accuracy.The DSD solvers make an accurate prediction for the coarsest case.This is interpreted as a coincidence.The error increases in the second coarsest mesh and then starts to decrease with constant rate with a slope around −2.This indicates that the DSD solvers are second order accurate.Fig. 18 presents the computational time normalised by the computational times from DSD-RK2.In terms of computational speed interFoam and interIsoFoam are slightly faster on the coarse meshes.However, as the mesh is refined the simulation time relative to DSD-RK2 increases and for the finest mesh interFoam and interIsoFoam are around four times slower than DSD-RK2.The DSD solver gradually performs better as the mesh resolution increases.The DSD-RK4 is roughly 1.5 times slower than DSD-RK2 across all cases.
Effect of CFL number varying the time step
This section presents the effect of the CFL number on the accuracy.Running simulations with higher CFL numbers through a larger time step leads to lower computational cost.However, as the CFL number increases, the errors of the temporal discretisation may increase and change convergence or the numerical method may become unstable.The CFL number is varied by changing the time step size keeping everything else constant.Fig. 19 presents the variation of the maximum velocity error, mean velocity error, wave height error and phase shift error as function of the CFL number.The interIsoFoam solver crashed for the case with an initial = 0.18, because the velocities had increased enough to make the solution unstable.The interFoam is more stable.It was able run the simulations up to = 0.38.It should be stressed that the simulations only crashed because a fixed time step was used.It is also observed that the error levels only gradually increases, which indicates that the cause of the crash can be found in the air phase region of the simulation.The implicit DSD solver was able to run up to = 0.3, where the error suddenly increases very rapidly.The explicit DSD solvers are capable of producing results for all the numbers up to = 0.45, however the error increases significantly after = 0.23.The accuracy of both the implicit and explicit DSD solvers are generally observed to be several times better than interFoam and interIsoFoam up to = 0.23.
The only exception being the implicit DSD solver at = 0.23 for the maximum velocity error in Fig. 19a and the phase shift error in Fig. 19d.This suggests that running simulations with a CFL number up to 0.2 is an acceptable choice for the DSD solvers.
Decreasing the CFL number generally reduces the errors for the DSD solvers.The only exception being DSD-RK4 and DSD-RK2 for the maximum velocity error and DSD-RK4 for the phase shift error.In these cases the error is almost constant, however with a small increase as the CFL number decreases.The origin of this behaviour has not been further investigated, because the increase is very small.For interFoam the maximum velocity error decreases when lowering the CFL number as also indicated by Larsen et al. (2019), where = 0.05 is recommended for simulations performed with interFoam.It is interesting to note that our convergence study shows that the wave height error and the mean velocity error for interFoam do not decrease as the CFL number decreases.The reason for the constant or increasing mean velocity error and wave height error could be that the wave travels a shorter distance of 10 compared to 100 in Larsen et al. (2019).In their work the wave height is observed to be fairly constant in the first 10.
Experiences from our implementation of the Ghost Fluid Method
A study of the GFM method combined with isoAdvector according to Vukčević et al. (2018), that we implemented in OpenFOAM-v1912 was the initial starting point for the development of the DSD method.This included a study of the stream function wave propagation case.Fig. 20 presents the velocity field from a simulation performed with our implementation of the GFM method combined with isoAdvector, and Fig. 21 presents the velocity field simulated using our implementation of the GFM method solver combined with a fourth-order explicit Runge-Kutta time integration of the pressurevelocity coupling.The velocity field near the free surface cells at the crest is smeared and the velocity is overpredicted under the wave.This observation was confirmed by Vuko Vukčević (personal communication, October 20, 2021).It was suggested that not taking into account the density jump in the viscous term could result in this behaviour.However, the shown simulations do not include diffusion in the numerical implementation, and the velocity field is still smeared.Several other aspects could also cause the problem: (1) the temporal discretisation, (2) the convective discretisation and (3) the assumption of a continuous velocity field at the free surface.
The temporal term, (1), approximates the acceleration using two previous time steps.When the cells near the free surface change from water to air cells or vice versa, the changes in the velocity field can be very sudden.The convective term, (2), is discretised using the linear upwind scheme, which combines the implicit upwind scheme with an explicit correction based on the cell centre velocity gradients.The velocity gradients will for some cells be affected by sudden changes in velocity field direction over the free surface.Furthermore, the convective term uses the volumetric flux, which is computed by linear interpolation of the velocity field.This could also contribute to the error.The assumption of a continuous velocity field, (3), allows a single continuity equation for the entire domain.This means that one should resolve the boundary layer close to the free surface in order to get a correct solution.That would require very fine mesh and it is often not possible in practical engineering applications.
Our results with the GFM method points in the same direction as the results presented by Vukčević et al. (2018), where the first order wave crest decayed with 3.81% after 8 wave lengths for the finest case with 28 cells per wave height and 140 cells per wave length.The first order horizontal velocity component had an error of 4.2% after 7 wave lengths.The horizontal velocity field presented in Fig.
29 (Vukčević et al., 2018) shows that there is a smooth transition of the horizontal velocity at the wave crest, which is also the case for our implementation.
No direct comparisons of velocity profiles were given by Vukčević et al. (2018).Even though the two implementations, (Vukčević et al., 2018) and ours, give similar results we cannot rule out the possibility of differences in the implementation.
Complex free surface flows
The main focus of the present paper has been to validate the velocity fields for simple cases.However, the DSD solvers are indeed capable of simulating complex free surface flows in 2D as well as 3D.The DSD solvers have been used to simulate various 2D cases with breaking waves (spilling and plunging), a dam break and a drop impact in a liquid pool.In 3D the DSD method has been used to simulate regular waves passing cylindrical and square vertical piles and a breaking phase focused wave impacting a vertical cylinder.Fig. 22 shows a snapshot from a simulation of a breaking wave generated with a single piston stroke.The simulated case is based on the case from Wei et al. (2018).The simulation is performed with DSD-RK2.The presented snapshot also shows that the numerical method can handle occluded air pockets without any problems.A movie of the simulation is available from the supplementary material Video S1.The movie shows that DSD-RK2 is capable of simulating a plunging breaking wave event including the post breaking region with a chaotic and violently changing free surface shape.DSD-RK2 was also used to simulate a 3D case of a phase focused breaking wave impact on a vertical cylinder.Fig. 23 presents a single time frame from the simulation, where the focused wave have just impacted the vertical cylinder.A movie of the simulation is available from Video S2 in the supplementary material.Hence, the DSD solvers are fully capable of handling violent and chaotic flows in 2D as well 3D.These simulations are solely intended to show that the DSD method can handle complex free surface simulations.Future work will focus on validating the accuracy of the DSD solvers on cases involving complex free surface flows.
Conclusion
The present study has developed and implemented a method for free surface flows in the OpenFOAM CFD library.The focus was on improving the description of the wave kinematics in the upper part of the water column for instance for analysing wave interaction with local parts of a structure.The problems using an approach with a varying density combined with a VOF-like propagation of the free surface were illustrated with an inclined square box with calm water.The two methods using a varying density, interFoam and interIsoFoam, clearly generated spurious velocities at the free surface in an else calm water.This phenomenon might be the underlying problem in getting reliable wave kinematics in the upper part of the water column.The present study has demonstrated that the new DSD method predicts a smooth surface elevation without any wiggles as seen from interFoam results.Furthermore, the velocity field is smooth over the entire depth.The DSD solvers are more accurate than the existing solvers interFoam and interIsoFoam from OpenFOAM-v1912, with a lower computational cost.For the finest mesh tested DSD-RK2 is 4.5 times faster than interFoam and several times more accurate.The DSD-RK4 solver is 1.5 times slower than DSD-RK2.It is as accurate as DSD-RK2 in some cases and less accurate in other cases.DSD and DSD-RK2 are very equal to each other in accuracy with minor variations.
We have presented an extensive study of the accuracy of the numerical methods for both constant and varying CFL numbers, which underlines the increased accuracy for the DSD solvers.The study focused on a single test case to enable a more extensive study, which can be widened to include other wave conditions with different steepness and relative depth.
From our own implementation of the GFM method, we have found that accurately predicting the pressure is not sufficient to obtain accurate prediction of the velocity kinematics near the wave crest.In the work by Vukčević et al. (2018) the presented results using 26 cells per wave height gave errors of 3.81% for the surface elevation.This is significantly higher than using our DSD-RK2 solver with 15 cells
Fig. 1 .
Fig. 1.Illustration of projection of a scalar gradient on a cell face and discretisation of the free surface.
Fig. 3 .
Fig. 3. Illustration of extrapolation case 1-3.The blue cells indicate water cells before the advection step giving the free position at the new time instance.The green cells illustrate a single layer of air cells next to the fluid region.After the advection step the new cells entering the fluid region is assumed to be within the green layer of cells.The white cells are air cells.
Fig. 4 .
Fig. 4. Velocity magnitude after 1000 time steps.Case: Inclined box with still water level.Mesh: 50 × 50, = 0.001 s.The air region of the interFoam and interIsoFoam simulation have been clipped out to highlight the effect in the cells filled by water.
Fig. 5 .
Fig. 5. Surface elevation centred around the last wave crest identified in interval = 10 − 11 at ∕ = 20. is the position of the theoretical wave crest.
Fig. 7 .
Fig. 7. Envelope of surface elevation from the stream function wave simulated with DSD-RK4 using the finest resolution of ∕ = 15.
Fig. 8 .Fig. 9 .
Fig. 8. Velocity magnitude vector plot around wave crest given by the stream function wave theory.
Fig. 16 .
Fig. 16.Average error under the last wave crest within = 10 − 11 over the last 5 s of each simulation.(a) Maximum error of the velocity magnitude profile (b) Mean error of the velocity magnitude profile.
Fig. 17 .
Fig. 17.Average error under the last wave crest within = 10 − 11 over the last 5 s of each simulation.(a) Error of the wave height (b) Phase shift error ℎ ∕.
Fig. 19 .
Fig. 19.Error under the last wave crest within = 10 − 11 averaged over the last 5 s of the simulation.(a)Maximum error of the velocity magnitude profile (b) Mean error of the velocity magnitude profile (c) Error of the wave height (d) Wave crest phase shift error ℎ ∕.
Fig. 15
Fig.15presents the vertical velocity profile below the zero down crossing of the last wave identified in = 10 − 11 at ∕ = 20.The small boxes show a 400% zoom of the velocity profile just below
Table 2
Case properties.
Table 3
Boundary conditions. | 2022-10-11T17:11:10.288Z | 2022-10-01T00:00:00.000 | {
"year": 2022,
"sha1": "1c33662989f11e863e8f65395a9f159f140a129c",
"oa_license": "CCBY",
"oa_url": "https://doi.org/10.1016/j.coastaleng.2022.104227",
"oa_status": "HYBRID",
"pdf_src": "ScienceParsePlus",
"pdf_hash": "e877758c8c627d7ab4d8c1b21894368e66d2a290",
"s2fieldsofstudy": [
"Computer Science",
"Engineering"
],
"extfieldsofstudy": []
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.