Pith. sign in

REVIEW 4 major objections 4 minor 67 references

Structure-Informed Learning of Flat Band 2D Materials

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Flat bands in 2D materials can be predicted from atomic structure alone, using a learnable flatness score that combines band dispersion and density-of-states information, removing the need for precomputed electronic structure.

desk verdict A genuinely new structure-only flat-band screening pipeline with real DFT-validated hits, but the discovery claim rests on positive-only examples and the training-data dependence is undersold. read the letter →

arxiv 2506.07518 v1 pith:DLDMBQJR submitted 2025-06-09 cond-mat.mtrl-sci cond-mat.mes-hall

Xiangwen Wang , Yihao Wei , Anupam Bhattacharya , Qian Yang , Artem Mishchenko This is my paper
classification cond-mat.mtrl-scicond-mat.mes-hall
keywords flatbands2Dmaterialsdiscoverymulti-modaldeeplearningstructure-basedpredictionkagomelatticetopologicalcorrelatedquantumphases
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that flat electronic bands, which drive correlated and topological quantum phases, can be predicted directly from a material's atomic geometry without first computing its band structure. It introduces a continuous, physics-motivated flatness score built from two spectral signatures—narrow bandwidth and sharp density-of-states peaks—and trains a multi-modal deep network that maps crystal structure and a short textual description to that score. The authors argue that this structure-to-flatness mapping is learnable, transferable, and scalable, enabling large-scale screening of unexplored 2D materials. They demonstrate this by identifying several new kagome-based compounds whose flat bands are confirmed by density functional theory to be topologically nontrivial.

What carries the argument

The central object is the flatness score $S_{\mathrm{total}}$ (Eqs. 1–4), a weighted combination of a bandwidth score $S_{\mathrm{bandwidth}}$ and a density-of-states score $S_{\mathrm{DOS}}$, set to zero whenever the representative band's energy span exceeds a threshold $\omega_{\mathrm{max}}$. The representative band is chosen as the narrowest reconnected dispersion curve near the Fermi level after handling band crossings. The multi-modal network fuses a line-graph geometric encoder (capturing bond lengths and angles) with a text encoder of material descriptors through bilinear attention, predicting $S_{\mathrm{total}}$ from structure alone. Downstream, a kagome-sublattice motif filter and density functional theory supplemented by elementary-band-representation analysis identify topologically nontrivial flat-band candidates.

What would settle it

Compute full DFT band structures for the top 100 unlabelled candidates that were not already validated and count how many actually have a near-Fermi band with width below the 0.3 eV threshold and a matching density-of-states peak; if the precision falls well below the roughly 95% top-150 rate reported on the training database, the flatness score is not transferring to new materials.

Watch

Extended reading notes

Core claim

The paper's central claim is that flat-band character, encoded in a continuous score combining band dispersion and density-of-states contrast, is a learnable function of the crystal structure. The authors compute this score for thousands of 2D materials using their existing band structures, train a multi-modal network that takes only atomic geometry and a textual description as input, and apply it to a previously unseen database. They report that the model ranks true flat-band materials at about 95% precision in its top 150 predictions, and that high-scoring materials form coherent clusters in the learned embedding space. DFT calculations on kagome-sublattice candidates confirm nearly dispersionless bands near the Fermi level, and band-representation analysis shows the flat bands are fragile-topological rather than trivial. The conclusion is that the flatness score provides a physically meaningful signal that lets atomic structure alone prioritise flat-band systems, removing the DFT bottleneck from early-stage discovery.

Load-bearing premise

The whole pipeline rests on the assumption that the automated flatness score, which picks a single representative band near the Fermi level and rewards both narrow dispersion and sharp density-of-states peaks, correctly identifies the bands that actually drive correlated and topological behaviour.

Editorial extensions

If this is right

  • Because no precomputed band structure is needed, an entire database of unlabelled 2D materials can be ranked in a single forward pass per structure, making flat-band screening far cheaper than DFT-based pipelines.
  • The six validated compounds (Nb3TeI7, Ta3SeI7, Ta3SBr7, Ta3TeI7, Cu3AsO4, Cu3SbO4) expand the short list of known topological flat-band materials, several of them structurally related to the experimentally studied Nb3Cl8.
  • The flatness score is continuous rather than a binary classifier, so it can prioritise 'nearly flat' systems with tunable dispersion, not only perfectly flat bands.
  • Because the scoring function is decoupled from the model, redefining the score—for example, to target exchange splittings or symmetry indicators—would let the same structure-only pipeline search for magnets, spin-liquid platforms, or other quantum phases.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's validation is concentrated on kagome-like sublattices; a natural test of transferability is to benchmark the same model against non-kagome flat-band geometries such as Lieb lattices or breathing pyrochlore layers, where the structural motif is different but flat bands still arise.
  • The score selects a single representative band per material, which may undercount systems with multiple flat bands; a multi-band score counting the number of flat bands weighted by spectral weight could be a more faithful label and likely improve precision for correlated phases.
  • Because the model consumes only structure and text, it could be coupled with a generative model to propose entirely new flat-band crystals, turning the predictor into an inverse-design objective.
  • The out-of-distribution validation uses only a small number of DFT checks; a statistically sized random sample of the top-scoring predictions from the unseen database would quantify the false-discovery rate of the screening pipeline.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The manuscript proposes a structure-only machine learning pipeline for screening two-dimensional flat-band materials. A composite flatness score S_total (Eqs. 1-4) combines bandwidth and DOS metrics computed from 2DMatPedia band structures, and a multimodal ALIGNN+RoBERTa model is trained to predict this score from crystal-structure graphs and textual descriptors. The model is applied to the unlabeled C2DB database, high scorers are filtered for kagome-like sublattices, and DFT calculations with band-representation analysis validate six candidates with nontrivial or fragile topological flat bands. The central claim is that the learned score transfers out of distribution and provides a physically meaningful screening signal without requiring precomputed electronic structures.

Significance. If the screening-precision claim were established, this would be a useful and scalable complement to existing DFT-based flat-band discovery pipelines. The paper's strengths include a clearly specified scoring protocol, a reproducible end-to-end workflow, and code/data availability. The DFT validation of Nb3TeI7, Ta3SeI7, Ta3SBr7, Ta3TeI7, Cu3AsO4, and Cu3SbO4 is a concrete positive result. However, the evidence currently establishes algorithm operation and positive examples rather than screening precision or out-of-distribution transfer; the significance of the claimed discovery pipeline depends on closing the validation gaps described below.

major comments (4)
  1. [2.4.2] The C2DB validation is positive-only. The paper states that 19 candidates with kagome-like substructures were selected and their electronic structures calculated, but it never reports how many of these 19 actually passed the flatness and orbital-projection criteria, nor how many failed. Only Cu3AsO4 and Cu3SbO4 are shown as passing examples. Without reporting the false-positive rate among the 19 candidates, and without testing high-scoring non-kagome candidates or low-scoring kagome controls, the claim that the flatness score provides a physically meaningful screening signal cannot be separated from the effect of the known-motif filter described in Section 5.3. Please provide a full pass/fail table for all 19 candidates and add appropriate control sets.
  2. [2.4.1] The claim that application to C2DB is a 'fully out-of-distribution test' is not established. The manuscript gives no statement about deduplication between the 2DMatPedia training set and the C2DB inference set. C2DB and 2DMatPedia are overlapping computational 2D-materials databases, so materials used for training may also appear in the inference pool. The authors should state explicitly whether any structural or compositional overlap was removed, and how the validation changes if overlapping entries are excluded.
  3. [5.1.2] The flatness score itself contains tunable parameters (omega_max, lambda, beta in Eqs. 1-4), and the manuscript reports that these are tuned by Bayesian optimization on the same 2DMatPedia data used to train and cross-validate the regression model. The choice of omega_max = 0.3 eV is also motivated by data-coverage statistics from the same dataset (Fig. 2d). Consequently, the reported cross-validated Pearson correlation of 0.92 is not sufficient to establish that the learned structure-score mapping generalizes; the score definition is partly fit to the target dataset. Please report how the score parameters were selected without using the test folds, or provide an independent evaluation on a held-out dataset with fixed score parameters.
  4. [2.1 / 5.1.1] The flatness score defined in Eqs. (1)-(4) scores any representative band near the Fermi level, including trivial flat bands originating from localized orbitals. The model is therefore trained to predict 'some flat band near the Fermi level,' not specifically a topologically nontrivial flat band. The paper's abstract and Discussion state that the framework identifies 'topologically nontrivial flat bands from unlabeled data,' but topological character is only assessed after the kagome-sublattice and orbital-projection filters are applied in Sections 2.3 and 2.4.2. This mismatch between the training target and the discovery claim should be addressed explicitly, and the manuscript should avoid implying that the score itself distinguishes topological from trivial flat bands.
minor comments (4)
  1. [5.3] The first sentence contains a typo: 'preformed' should be 'performed'.
  2. [2.2.1] The text contains an unresolved cross-reference 'Fig. ref-fig:dla' in the description of the model architecture; this should be corrected to a valid figure number.
  3. [2.4.1] The phrase 'using only the crystal geometry as input' is inconsistent with Section 5.2.1 and Table 1, which show that the RoBERTa text encoder receives the chemical formula, space group, lattice parameters, and other structure-derived descriptors. Please rephrase to avoid overstating the input modality.
  4. [5.1.1] The number of near-Fermi bands n used in the band-selection procedure is not stated anywhere in the paper. Since n is a tunable parameter that directly affects which band becomes the representative band, its value and sensitivity should be reported.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity; central flat-band claims rest on independent DFT validation, though BO tuning and in-database screening introduce minor self-fit.

full rationale

The derivation chain is: define a flatness score from band dispersion and DOS (Eqs. 1-4), train a multimodal model to predict that score from crystal structure alone, screen 2DMatPedia and C2DB, then validate selected candidates with fresh DFT calculations and band-representation analysis. The score is a supervised target, not a quantity equivalent to the model's structure-only prediction. The C2DB candidates Cu3AsO4 and Cu3SbO4 have no precomputed flatness labels in the model input; their DFT band structures and EBR irreps are independent, externally grounded validations. The 2DMatPedia screening in Sec. 2.3 re-examines materials whose labels were used in training, so it is closer to a retrieval or self-consistency check than a discovery, but the paper's out-of-distribution claim is anchored in Sec. 2.4 on C2DB. Hyperparameters lambda and beta in Eq. 4 are tuned via Bayesian optimization on the same labeled dataset, which is a mild self-fit, but it does not force the independent DFT results. No equation reduces to its own input, no load-bearing self-citation is used, and no uniqueness theorem or ansatz is imported from the authors' prior work. The underreporting of false positives and the lack of stated deduplication between 2DMatPedia and C2DB are precision and generalization concerns, not circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central claim depends on the flatness score, whose parameters (omega_max, lambda, beta) are fitted to data, and on the domain assumptions above. No new physical entities are introduced.

free parameters (4)
  • omega_max = 0.3 eV (chosen after testing 0.1, 0.3, 0.5 eV)
    Bandwidth threshold defining the maximum dispersion considered flat; selected to balance sample coverage and selectivity.
  • lambda (sigmoid weight) = Not reported, tuned via Bayesian Optimization
    Weight in the sigmoid combination of Sbandwidth and SDOS in Eq. 4.
  • beta (sigmoid weight) = Not reported, tuned via Bayesian Optimization
    Weight on the product term in Eq. 4.
  • n_bands (number of near-Fermi bands) = Not reported
    Number of bands selected near the Fermi level; algorithm chooses n with smallest energy difference (all if fewer than six).
assumptions (4)
  • domain assumption The band-crossing detection and segment-reconnection algorithm correctly reconstructs the representative band for each material.
    Section 5.1.1: the flattest reconnected segment is taken as representative; this may miss the physically relevant flat band if crossings are misidentified or if the band is flat only away from high-symmetry paths.
  • domain assumption The DOS reference window [-5, 5] eV is a suitable baseline for peak contrast.
    Section 5.1.2, Eq. 2-3: taken from Regnault et al., but its validity for all 2D materials is assumed.
  • domain assumption PBE-level DFT is accurate enough to identify flat bands and their topology in these candidate materials.
    Section 5.3: all validation uses PBE with and without SOC; for strongly correlated systems, PBE may misplace or distort flat bands.
  • standard math Elementary band representation analysis correctly classifies fragile topology.
    Section 2.3: the EBR criterion is a standard mathematical tool; the paper assumes its application to the specific layer group is error-free.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Structure-Informed Learning of Flat Band 2D Materials." pith.science (2026). https://pith.science/paper/DLDMBQJR

@misc{pith2026250607518,
  author       = {Pith},
  title        = {Pith review of: Structure-Informed Learning of Flat Band 2D Materials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DLDMBQJR}},
  note         = {Machine review of arXiv:2506.07518}
}
read the original abstract

Flat electronic bands enhance electron-electron interactions and give rise to correlated states such as unconventional superconductivity or fractional topological phases. However, most current efforts towards flat-band materials discovery rely on density functional theory (DFT) calculations and manual band structures inspection, restraining their applicability to vast unexplored material spaces. While data-driven methods offer a scalable alternative, most existing models either depend on band structure inputs or focus on scalar properties like bandgap, which fail to capture flat-band characteristics. Here, we report a structure-informed framework for the discovery of previously unrecognized flat-band two-dimensional (2D) materials, which combines a data-driven flatness score capturing both band dispersion and density-of-states characteristics with multi-modal learning from atomic structure inputs. The framework successfully identified multiple flat-band candidates, with DFT validation of kagome-based systems confirming both band flatness and topological character. Our results show that the flatness score provides a physically meaningful signal for identifying flat bands from atomic geometry. The framework uncovers multiple new candidates with topologically nontrivial flat bands from unlabeled data, with consistent model performance across structurally diverse materials. By eliminating the need for precomputed electronic structures, our method enables large-scale screening of flat-band materials and expands the search space for discovering strongly correlated quantum materials.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

67 extracted references · 60 canonical work pages

  1. [1]

    Nature 556(7699), 43–50 (2018)

    Cao, Y., Fatemi, V., Fang, S., Watanabe, K., Taniguchi, T., Kaxiras, E., Jarillo-Herrero, P.: Unconventional superconductivity in magic-angle graphene superlattices. Nature 556(7699), 43–50 (2018)

    work page 2018

  2. [2]

    Nature Physics16(7), 725–733 (2020)

    Balents, L., Dean, C.R., Efetov, D.K., Young, A.F.: Superconductivity and strong correlations in moir´ e flat bands. Nature Physics16(7), 725–733 (2020)

    work page 2020

  3. [3]

    Physical review letters 126(2), 027002 (2021)

    Peri, V., Song, Z.-D., Bernevig, B.A., Huber, S.D.: Fragile topology and flat-band superconductivity in the strong-coupling regime. Physical review letters 126(2), 027002 (2021)

    work page 2021

  4. [4]

    Physical review letters 121(9), 096401 (2018)

    Lin, Z., Choi, J.-H., Zhang, Q., Qin, W., Yi, S., Wang, P., Li, L., Wang, Y., Zhang, H., Sun, Z., et al.: Flatbands and emergent ferromagnetic ordering in fe 3 sn 2 kagome lattices. Physical review letters 121(9), 096401 (2018)

    work page 2018

  5. [5]

    Physical review letters 106(23), 236804 (2011)

    Neupert, T., Santos, L., Chamon, C., Mudry, C.: Fractional quantum hall states at zero magnetic field. Physical review letters 106(23), 236804 (2011)

    work page 2011

  6. [6]

    Nature communications 2(1), 389 (2011)

    Sheng, D., Gu, Z.-C., Sun, K., Sheng, L.: Fractional quantum hall effect in the absence of landau levels. Nature communications 2(1), 389 (2011)

    work page 2011

  7. [7]

    Comptes Rendus Physique 14(9-10), 816–839 (2013)

    Parameswaran, S.A., Roy, R., Sondhi, S.L.: Fractional quantum hall physics in topological flat bands. Comptes Rendus Physique 14(9-10), 816–839 (2013)

    work page 2013

  8. [8]

    Chinese Physics B 23(7), 077308 (2014) 20

    Liu, Z., Liu, F., Wu, Y.-S.: Exotic electronic states in the world of flat bands: From theory to material. Chinese Physics B 23(7), 077308 (2014) 20

    work page 2014

Show all 67 references
  1. [9]

    Nature communications 6(1), 8944 (2015)

    Peotta, S., T¨ orm¨ a, P.: Superfluidity in topologically nontrivial flat bands. Nature communications 6(1), 8944 (2015)

    2015

  2. [10]

    Advanced Functional Materials, 2416508 (2025)

    Zhang, Z., Dai, J., Wang, C., Zhu, H., Pang, F., Cheng, Z., Ji, W.: 2d kagome materials: Theoretical insights, experimental realizations, and electronic structures. Advanced Functional Materials, 2416508 (2025)

    2025

  3. [11]

    Physical review letters 114(24), 245504 (2015)

    Mukherjee, S., Spracklen, A., Choudhury, D., Goldman, N., ¨Ohberg, P., Anders- son, E., Thomson, R.R.: Observation of a localized flat-band state in a photonic lieb lattice. Physical review letters 114(24), 245504 (2015)

    2015

  4. [12]

    Physical Review B 100(23), 235145 (2019)

    Bhattacharya, A., Pal, B.: Flat bands and nontrivial topological properties in an extended lieb lattice. Physical Review B 100(23), 235145 (2019)

    2019

  5. [13]

    Proceedings of the National Academy of Sciences 108(30), 12233–12237 (2011)

    Bistritzer, R., MacDonald, A.H.: Moir´ e bands in twisted double-layer graphene. Proceedings of the National Academy of Sciences 108(30), 12233–12237 (2011)

    2011

  6. [14]

    Science 363(6431), 1059–1064 (2019)

    Yankowitz, M., Chen, S., Polshyn, H., Zhang, Y., Watanabe, K., Taniguchi, T., Graf, D., Young, A.F., Dean, C.R.: Tuning superconductivity in twisted bilayer graphene. Science 363(6431), 1059–1064 (2019)

    2019

  7. [15]

    Physical review letters 122(10), 106405 (2019)

    Tarnopolsky, G., Kruchkov, A.J., Vishwanath, A.: Origin of magic angles in twisted bilayer graphene. Physical review letters 122(10), 106405 (2019)

    2019

  8. [16]

    ACS nano 15(4), 5944–5958 (2021)

    He, F., Zhou, Y., Ye, Z., Cho, S.-H., Jeong, J., Meng, X., Wang, Y.: Moir´ e patterns in 2d materials: a review. ACS nano 15(4), 5944–5958 (2021)

    2021

  9. [17]

    Nature579(7797), 56–61 (2020)

    Chen, G., Sharpe, A.L., Fox, E.J., Zhang, Y.-H., Wang, S., Jiang, L., Lyu, B., Li, H., Watanabe, K., Taniguchi, T., et al.: Tunable correlated chern insulator and ferromagnetism in a moir´ e superlattice. Nature579(7797), 56–61 (2020)

    2020

  10. [18]

    Nature Reviews Materials 5(10), 748–763 (2020)

    Carr, S., Fang, S., Kaxiras, E.: Electronic-structure methods for twisted moir´ e layers. Nature Reviews Materials 5(10), 748–763 (2020)

    2020

  11. [19]

    Nature 584(7820), 210–214 (2020)

    Shi, Y., Xu, S., Yang, Y., Slizovskiy, S., Morozov, S.V., Son, S.-K., Ozdemir, S., Mullan, C., Barrier, J., Yin, J., et al.: Electronic phase separation in multilayer rhombohedral graphite. Nature 584(7820), 210–214 (2020)

    2020

  12. [20]

    Physical Review B—Condensed Matter and Materials Physics 82(3), 035409 (2010)

    Zhang, F., Sahu, B., Min, H., MacDonald, A.H.: Band structure of abc-stacked graphene trilayers. Physical Review B—Condensed Matter and Materials Physics 82(3), 035409 (2010)

    2010

  13. [21]

    Physical Review B 108(15), 155406 (2023)

    Park, Y., Kim, Y., Chittari, B.L., Jung, J.: Topological flat bands in rhom- bohedral tetralayer and multilayer graphene on hexagonal boron nitride moir´ e superlattices. Physical Review B 108(15), 155406 (2023)

    2023

  14. [22]

    Physical Review Materials 5(8), 084203 (2021) 21

    Liu, H., Meng, S., Liu, F.: Screening two-dimensional materials with topological flat bands. Physical Review Materials 5(8), 084203 (2021) 21

    2021

  15. [23]

    Scientific data 6(1), 86 (2019)

    Zhou, J., Shen, L., Costa, M.D., Persson, K.A., Ong, S.P., Huck, P., Lu, Y., Ma, X., Chen, Y., Tang, H., et al.: 2dmatpedia, an open computational database of two-dimensional materials from top-down and bottom-up approaches. Scientific data 6(1), 86 (2019)

    2019

  16. [24]

    Nature 603(7903), 824–828 (2022)

    Regnault, N., Xu, Y., Li, M.-R., Ma, D.-S., Jovanovic, M., Yazdani, A., Parkin, S.S., Felser, C., Schoop, L.M., Ong, N.P., et al.: Catalogue of flat-band stoichiometric materials. Nature 603(7903), 824–828 (2022)

    2022

  17. [25]

    Advanced Functional Materials 34(19), 2313067 (2024)

    Duan, J., Ma, D.-S., Zhang, R.-W., Jiang, W., Zhang, Z., Cui, C., Yu, Z.-M., Yao, Y.: Cataloging high-quality two-dimensional van der waals materials with flat bands. Advanced Functional Materials 34(19), 2313067 (2024)

    2024

  18. [26]

    Nature 566(7745), 480–485 (2019)

    Vergniory, M., Elcoro, L., Felser, C., Regnault, N., Bernevig, B.A., Wang, Z.: A complete catalogue of high-quality topological materials. Nature 566(7745), 480–485 (2019)

    2019

  19. [27]

    Science 376(6595), 9094 (2022)

    Vergniory, M.G., Wieder, B.J., Elcoro, L., Parkin, S.S., Felser, C., Bernevig, B.A., Regnault, N.: All topological bands of all nonmagnetic stoichiometric materials. Science 376(6595), 9094 (2022)

    2022

  20. [28]

    npj Computational Materials 9(1), 101 (2023)

    Bhattacharya, A., Timokhin, I., Chatterjee, R., Yang, Q., Mishchenko, A.: Deep learning approach to genome of two-dimensional materials with flat electronic bands. npj Computational Materials 9(1), 101 (2023)

    2023

  21. [29]

    npj Computational Materials 10(1), 39 (2024)

    Neves, P.M., Wakefield, J.P., Fang, S., Nguyen, H., Ye, L., Checkelsky, J.G.: Crystal net catalog of model flat band materials. npj Computational Materials 10(1), 39 (2024)

    2024

  22. [30]

    PLoS One 16(8), 0255637 (2021)

    Zhang, Y., Xu, W., Liu, G., Zhang, Z., Zhu, J., Li, M.: Bandgap prediction of two-dimensional materials using machine learning. PLoS One 16(8), 0255637 (2021)

    2021

  23. [31]

    Chemistry of Materials 30(12), 4031–4038 (2018)

    Rajan, A.C., Mishra, A., Satsangi, S., Vaish, R., Mizuseki, H., Lee, K.-R., Singh, A.K.: Machine-learning-assisted accurate band gap predictions of functionalized mxene. Chemistry of Materials 30(12), 4031–4038 (2018)

    2018

  24. [32]

    Nature Communications 13(1), 468 (2022)

    Knøsgaard, N.R., Thygesen, K.S.: Representing individual electronic states for machine learning gw band structures of 2d materials. Nature Communications 13(1), 468 (2022)

    2022

  25. [33]

    npj Computational Materials 5(1), 26 (2019)

    Dong, Y., Wu, C., Zhang, C., Liu, Y., Cheng, J., Lin, J.: Bandgap prediction by deep learning in configurationally hybridized graphene and boron nitride. npj Computational Materials 5(1), 26 (2019)

    2019

  26. [34]

    arXiv 22 preprint arXiv:2411.16483 (2024)

    Gong, W., Sun, T., Bai, H., Tsai, J.-Y., Ling, H., Yan, Q.: Graph transformer networks for accurate band structure prediction: An end-to-end approach. arXiv 22 preprint arXiv:2411.16483 (2024)

    2024

  27. [35]

    arXiv preprint arXiv:2501.03456 (2025)

    Yeh, Y.-T., Ock, J., Farimani, A.B.: Text to band gap: Pre-trained language models as encoders for semiconductor band gap prediction. arXiv preprint arXiv:2501.03456 (2025)

    2025

  28. [36]

    arXiv preprint arXiv:2502.06836 (2025)

    Lee, J., Park, C., Yang, H., Lim, S., Han, S.: Cast: Cross attention based mul- timodal fusion of structure and text for materials property prediction. arXiv preprint arXiv:2502.06836 (2025)

    2025 arXiv

  29. [37]

    JETP letters 94, 233–239 (2011)

    Heikkil¨ a, T.T., Kopnin, N.B., Volovik, G.E.: Flat bands in topological media. JETP letters 94, 233–239 (2011)

    2011

  30. [38]

    Science advances 4(11), 4511 (2018)

    Li, Z., Zhuang, J., Wang, L., Feng, H., Gao, Q., Xu, X., Hao, W., Wang, X., Zhang, C., Wu, K., et al.: Realization of flat band with possible nontrivial topology in electronic kagome lattice. Science advances 4(11), 4511 (2018)

    2018

  31. [39]

    npj Computational Materials 7(1), 185 (2021)

    Choudhary, K., DeCost, B.: Atomistic line graph neural network for improved materials property predictions. npj Computational Materials 7(1), 185 (2021)

    2021

  32. [40]

    arXiv preprint arXiv:1907.11692 (2019)

    Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., Stoyanov, V.: Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692 (2019)

    2019 arXiv

  33. [41]

    https://arxiv.org/abs/1802.03426

    McInnes, L., Healy, J., Melville, J.: UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction (2020). https://arxiv.org/abs/1802.03426

    2020 arXiv

  34. [42]

    arXiv preprint arXiv:2411.08950 (2024)

    Petralanda, U., Jiang, Y., Bernevig, B.A., Regnault, N., Elcoro, L.: Two- dimensional topological quantum chemistry and catalog of topological materials. arXiv preprint arXiv:2411.08950 (2024)

    2024 arXiv

  35. [43]

    Nature547(7663), 298–305 (2017)

    Bradlyn, B., Elcoro, L., Cano, J., Vergniory, M.G., Wang, Z., Felser, C., Aroyo, M.I., Bernevig, B.A.: Topological quantum chemistry. Nature547(7663), 298–305 (2017)

    2017

  36. [44]

    Physical review letters 123(3), 036401 (2019)

    Song, Z., Wang, Z., Shi, W., Li, G., Fang, C., Bernevig, B.A.: All magic angles in twisted bilayer graphene are topological. Physical review letters 123(3), 036401 (2019)

    2019

  37. [45]

    Nano Letters 22(11), 4596–4602 (2022)

    Sun, Z., Zhou, H., Wang, C., Kumar, S., Geng, D., Yue, S., Han, X., Haraguchi, Y., Shimada, K., Cheng, P., et al.: Observation of topological flat bands in the kagome semiconductor nb3cl8. Nano Letters 22(11), 4596–4602 (2022)

    2022

  38. [46]

    2D Materials 8(4), 044002 (2021) 23

    Gjerding, M.N., Taghizadeh, A., Rasmussen, A., Ali, S., Bertoldo, F., Deilmann, T., Knøsgaard, N.R., Kruse, M., Larsen, A.H., Manti, S., et al.: Recent progress of the computational 2d materials database (c2db). 2D Materials 8(4), 044002 (2021) 23

    2021

  39. [47]

    https://github.com/bayesian-optimization/ BayesianOptimization

    Nogueira, F.: Bayesian Optimization: Open source constrained global opti- mization tool for Python (2014–). https://github.com/bayesian-optimization/ BayesianOptimization

    2014

  40. [48]

    McInnes, L., Healy, J., Astels, S., et al.: hdbscan: Hierarchical density based clustering. J. Open Source Softw. 2(11), 205 (2017)

    2017

  41. [49]

    arXiv preprint arXiv:1912.01703 (2019)

    Paszke, A.: Pytorch: An imperative style, high-performance deep learning library. arXiv preprint arXiv:1912.01703 (2019)

    2019 arXiv

  42. [50]

    In: ICLR Workshop on Representation Learning on Graphs and Manifolds (2019)

    Wang, M.Y.: Deep graph library: Towards efficient and scalable deep learning on graphs. In: ICLR Workshop on Representation Learning on Graphs and Manifolds (2019)

    2019

  43. [51]

    the Journal of machine Learning research 12, 2825–2830 (2011)

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V.,et al.: Scikit-learn: Machine learning in python. the Journal of machine Learning research 12, 2825–2830 (2011)

    2011

  44. [52]

    Computational Materials Science 68, 314–319 (2013)

    Ong, S.P., Richards, W.D., Jain, A., Hautier, G., Kocher, M., Cholia, S., Gunter, D., Chevrier, V.L., Persson, K.A., Ceder, G.: Python materials genomics (pymat- gen): A robust, open-source python library for materials analysis. Computational Materials Science 68, 314–319 (2013)

    2013

  45. [53]

    Nature methods 17(3), 261–272 (2020)

    Virtanen, P., Gommers, R., Oliphant, T.E., Haberland, M., Reddy, T., Cour- napeau, D., Burovski, E., Peterson, P., Weckesser, W., Bright, J., et al.: Scipy 1.0: fundamental algorithms for scientific computing in python. Nature methods 17(3), 261–272 (2020)

    2020

  46. [54]

    In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pp

    Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al.: Transformers: State-of-the-art natu- ral language processing. In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing:...

    2020

  47. [55]

    Perdew, J.P., Burke, K., Ernzerhof, M.: Generalized gradient approximation made simple. Phys. Rev. Lett. 77(18), 3865–3868 (1996)

    1996

  48. [56]

    Computational Materials Science 50(8), 2295–2310 (2011)

    Jain, A., Hautier, G., Moore, C.J., Ong, S.P., Fischer, C.C., Mueller, T., Persson, K.A., Ceder, G.: A high-throughput infrastructure for density functional theory calculations. Computational Materials Science 50(8), 2295–2310 (2011)

    2011

  49. [57]

    Computational Materials Science 68, 314–319 (2013) 24

    Ong, S.P., Richards, W.D., Jain, A., Hautier, G., Kocher, M., Cholia, S., Gunter, D., Chevrier, V.L., Persson, K.A., Ceder, G.: Python materials genomics (pymat- gen): A robust, open-source python library for materials analysis. Computational Materials Science 68, 314–319 (2013) 24

    2013

  50. [58]

    Computer Physics Communications 261, 107760 (2021)

    Gao, J., Wu, Q., Persson, C., Wang, Z.: Irvsp: To obtain irreducible representa- tions of electronic states in the vasp. Computer Physics Communications 261, 107760 (2021)

    2021

  51. [59]

    Togo, A., Chaput, L., Tadano, T., Tanaka, I.: Implementation strategies in phonopy and phono3py. J. Phys. Condens. Matter 35(35), 353001 (2023) https: //doi.org/10.1088/1361-648X/acd831

    2023 doi

  52. [60]

    Togo, A.: First-principles phonon calculations with phonopy and phono3py. J. Phys. Soc. Jpn. 92(1), 012001 (2023) https://doi.org/10.7566/JPSJ.92.012001

    2023 doi

  53. [61]

    Science bulletin 67(6), 598–608 (2022)

    Gao, J., Qian, Y., Jia, H., Guo, Z., Fang, Z., Liu, M., Weng, H., Wang, Z.: Unconventional materials: the mismatch between electronic charge centers and atomic positions. Science bulletin 67(6), 598–608 (2022)

    2022

  54. [62]

    Physical Review B 103(20), 205133 (2021)

    Nie, S., Qian, Y., Gao, J., Fang, Z., Weng, H., Wang, Z.: Application of topological quantum chemistry in electrides. Physical Review B 103(20), 205133 (2021)

    2021

  55. [63]

    Aroyo, M.I., Perez-Mato, J.M., Orobengoa, D., Tasci, E., Flor, G., Kirov, A.: Crystallography online: Bilbao crystallographic server. Bulg. Chem. Commun 43(2), 183–197 (2011)

    2011

  56. [64]

    databases and crystallographic computing programs

    Aroyo, M.I., Perez-Mato, J.M., Capillas, C., Kroumova, E., Ivantchev, S., Madariaga, G., Kirov, A., Wondratschek, H.: Bilbao crystallographic server: I. databases and crystallographic computing programs. Zeitschrift f¨ ur Kristallographie-Crystalline Materials 221(1), 15–27 (2006)

    2006

  57. [65]

    Aroyo, M.I., Kirov, A., Capillas, C., Perez-Mato, J., Wondratschek, H.: Bilbao crystallographic server. ii. representations of crystallographic point groups and space groups. Foundations of Crystallography 62(2), 115–128 (2006)

    2006

  58. [66]

    Computer Physics Communications 251, 107080 (2020)

    Herath, U., Tavadze, P., He, X., Bousquet, E., Singh, S., Mu˜ noz, F., Romero, A.H.: Pyprocar: A python library for electronic structure pre/post-processing. Computer Physics Communications 251, 107080 (2020)

    2020

  59. [67]

    Computer Physics Communications 297, 109063 (2024) 25

    Lang, L., Tavadze, P., Tellez, A., Bousquet, E., Xu, H., Mu˜ noz, F., Vasquez, N., Herath, U., Romero, A.H.: Expanding pyprocar for new features, maintainability, and reliability. Computer Physics Communications 297, 109063 (2024) 25

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.