Pith. sign in

REVIEW 1 major objections 5 minor 1 cited by

OpenCSP claims that targeted, pressure-aware data collection—not raw model scale—is what makes machine-learned interatomic potentials reliable for crystal structure prediction under tens to hundreds of gigapascals.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

OpenCSP is an open pressure-diverse dataset and model suite that matches or beats larger universal atomistic models on high-pressure crystal structure prediction with far fewer training data.

T0 review reviewed 2026-08-04 challenge →

load-bearing objection Useful open dataset and models for high-pressure CSP, but the benchmark advantages over large models rest on a code-inconsistent evaluation that needs fixing before the comparative claims are credible. the 1 major comments →

arxiv 2509.10293 v1 pith:COYAZ7FK submitted 2025-09-12 cond-mat.mtrl-sci

OpenCSP: A Deep Learning Framework for Crystal Structure Prediction from Ambient to High Pressure

Yinan Wang , Xiaoyang Wang , Zhenyu Wang , Jing Wu , Jian Lv , Han Wang This is my paper
classification cond-mat.mtrl-sci
keywords crystal structure predictionhigh-pressure materialsmachine learning interatomic potentialsuncertainty-guided active learningenthalpy rankingvirial accuracydata-efficient trainingconcurrent learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

OpenCSP argues that the reason large atomistic models struggle at high pressure is not model size but training data: their corpora cluster near ambient equilibrium and lack explicit pressure labels. To fix this, the authors built a 1.5-million-structure dataset by random high-pressure structure search with uncertainty-guided concurrent learning, then trained three deep graph-network potentials of modest scale. They show these models match or beat much larger universal potentials in energy, force, and especially virial accuracy, and in crystal structure prediction at 50–200 GPa, despite using one to two orders of magnitude less training data. If right, this means pressure-aware data selection can substitute for brute-force scaling in extreme-condition materials discovery.

Core claim

The paper claims that a curated, pressure-resolved dataset of about 1.5 million DFT-labeled configurations—built by proposing random compressed structures, relaxing them under randomly sampled target pressures, and relabeling only the most uncertain with DFT—is sufficient to train machine-learned potentials that outperform far larger universal models in high-pressure crystal structure prediction. The strongest evidence is in the virial (pressure–volume) term: on cross-dataset tests, OpenCSP models have several times smaller virial error and a much lighter error tail, which directly benefits enthalpy ranking. In pressure-controlled relaxation, the OpenCSP models reproduce target pressures of

What carries the argument

The carrying mechanism is the dataset-construction loop, not a new architecture. Random structure proposals are relaxed at random target pressures drawn from 0–100 GPa; an ensemble of graph-network potentials estimates force or enthalpy uncertainty along the relaxation; only the most uncertain configurations go to DFT labeling; and the labeled set is folded back into training. The models are deep graph neural networks with 6–24 message-passing layers, jointly trained on energy, force, and virial, with virial treated as a first-class target because it controls the PV term in enthalpy.

Load-bearing premise

The comparisons assume that literature-reported high-pressure structures recovered within 2,000 generated candidates are the right measure of prediction quality, and that DFT labels from different electronic-structure codes are comparable enough for direct error comparison.

What would settle it

Evaluate OpenCSP and the baselines on a held-out set of high-pressure phases produced by an independent search algorithm, using identical DFT settings for all models; if OpenCSP's crystal-structure-prediction success-rate and virial advantages shrink to noise, the claim that targeted pressure sampling is the cause is not supported.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A 1.5-million-configuration, pressure-diverse dataset yields virial error several times smaller than 12M–113M-scale baselines on a held-out trajectory test, with a lighter tail of large PV errors.
  • Models trained this way can relax unseen ternary structures to target pressures of 0, 50, 100, 150, and 200 GPa with mean deviations within about 3 GPa, whereas baselines drift to errors above 80% at 50 GPa.
  • In CSP recovery tests at 50–150 GPa, OpenCSP models reach success rates roughly 20–30 percentage points higher than the baselines; at 0 GPa all models are comparable at about 60%.
  • Increasing network depth from 6 to 24 layers steadily improves energy and force accuracy and cross-dataset generalization, with diminishing returns for force and virial beyond 12 layers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This suggests that for extreme-condition modeling, targeted uncertainty-guided acquisition may beat brute-force scaling; a direct test would compare cost-per-accurate-enthalpy against large generic datasets on a fixed budget.
  • The same strategy could transfer to other structure-search generators, such as evolutionary or diffusion-based methods, or to other target properties like temperature and defect equilibria; the paper's benchmark protocol ties the result to one search pipeline, but the data-selection logic is general.
  • The virial advantage over baselines may partly reflect differences in DFT codes and pseudopotentials rather than physical accuracy alone; re-labeling all structures with one common DFT setting would test this.
  • Pressure-resolved open datasets like this one could make 'pressure MAE' a standard reporting metric for atomistic models, shifting evaluation from aggregate energy accuracy toward the enthalpy-relevant quantities that matter for high-pressure prediction.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 5 minor

Summary. The paper introduces OpenCSP, an open-source dataset of ~1.5 million DFT-labeled configurations generated by random CALYPSO structure searches under a pressure-aware DP-GEN concurrent-learning workflow, together with three DPA3-based models (OpenCSP-L6/L12/L24) trained for joint energy, force, and virial prediction. The central claim is that, despite using one to two orders of magnitude fewer training data than large universal models, OpenCSP achieves comparable or superior performance in high-pressure crystal structure prediction, particularly in virial accuracy, pressure-controlled relaxation, and CSP success rates at elevated pressures. Benchmarks are presented against MACE-MPA-0, MatterSim v1 5M, and GRACE-2L-OAM on in-distribution accuracy, MPTrj cross-dataset transfer, GNoME formation energies, pressure-constrained relaxations, and pressure-resolved CSP tasks.

Significance. If the central claim is established, OpenCSP would be a valuable community resource: it is an open, pressure-resolved dataset with explicit stress labels, and it demonstrates that targeted pressure-aware data generation can be more efficient than indiscriminate large-scale data collection. The paper's strengths include the public release of dataset and models, a realistic active-learning pipeline, held-out compositional splits for the in-distribution test, and honesty about the in-distribution versus zero-shot status of the baselines on MPTrj. However, the headline quantitative comparisons that support the high-pressure superiority claim are built on an evaluation asymmetry in DFT references, and the high-pressure CSP benchmark does not control for compositional overlap with the training pipeline. These issues must be addressed before the data-efficiency and superiority claims can be considered secure.

major comments (1)
  1. [General] The abstract and conclusion claim 'comparable or superior performance in high-pressure enthalpy ranking,' but no direct enthalpy-ranking benchmark is presented. Table IV measures pressure reproduction and §III.E measures structural matching to literature structures; neither evaluates whether the models rank competing candidate structures correctly by enthalpy at a given pressure. A direct benchmark — e.g., generating multiple candidate structures per composition, ranking them by model enthalpy, and comparing the ranking/energy ordering to DFT enthalpies computed with a single code — is needed to substantiate the enthalpy-ranking claim. Without it, the superiority claim rests on an indirect proxy.
minor comments (5)
  1. [§II] The abbreviation 'sAlex' is used without definition; please introduce it (presumably a subset of the Alexandria dataset) at first use.
  2. [§V.A] Typo: 'labled' should be 'labeled' in the description of the initial dataset.
  3. [§III.B] The claim that MatterSim results are 'true zero-shot' would benefit from a citation or verification that MPTrj structures are not included in MatterSim's 17M training set, since the paper otherwise emphasizes this distinction.
  4. [§III.D / Table IV] The phrase 'consistent, model-independent assessment' is misleading when different DFT codes are used for different models; please rephrase or, preferably, use a single common evaluator.
  5. [§X] The data availability statement lists only generic AIS Square URLs. Please provide direct and permanent identifiers (e.g., DOI) for the specific dataset and model versions used in the paper.

Circularity Check

0 steps flagged

No significant circularity: OpenCSP's performance claims rest on held-out and external benchmarks; label-source asymmetries are correctness risks, not reductions to inputs.

full rationale

OpenCSP's derivation chain is empirical rather than deductive. The models are trained on a separately generated DFT dataset (Section V.A) and then evaluated on held-out OpenCSP splits, zero-shot MPTrj structures (Section III.B), GNoME candidate structures (Section III.C), and literature high-pressure reference structures (Section III.E). The headline performance numbers are measured on data that are not part of the training labels, and no fitted parameter is renamed as a prediction. The main caveats are benchmark asymmetries rather than circularity: in Section III.B, OpenCSP errors are computed against ABACUS relabels while baseline errors use the original MPTrj labels; in Section III.C, OpenCSP uses ABACUS labels for GNoME formation energies while baselines use the originally reported GNoME energies; and in Section III.D, final relaxed pressures are evaluated with ABACUS for OpenCSP and with VASP for the baselines. These code inconsistencies could affect the magnitude of the reported advantages if ABACUS and VASP virials differ at high pressure, but they are comparability/validity concerns, not circularity: the benchmark target is not an input to the model or the fitting procedure, and the paper discloses the asymmetry. The paper also explicitly notes that MPTrj configurations are in the training sets of MACE-MPA-0 and GRACE-2L-OAM, making those baselines partially in-distribution. Self-citations to CALYPSO, DP-GEN, and DPA3 are method attributions from the same community; they do not serve as the evidence for the central performance claims, which rest on independent, held-out test sets. No equation or fitted parameter reduces to its own input, so no circular step is exhibited.

Axiom & Free-Parameter Ledger

3 free parameters · 5 axioms · 0 invented entities

The central claim rests on dataset construction choices (safe radii, selection thresholds, DFT settings) and on comparability of labels across codes. These are stated in the paper, but several are not independently grounded, so they are counted here.

free parameters (3)
  • Element-specific safe radii (rs) = Table VI, from ref [64]
    Controls minimum interatomic distance in CALYPSO initial structure generation via Eq. (1); chosen from prior literature, not fitted to OpenCSP target, but affects dataset composition.
  • Loss weight schedule for energy/force/virial = Stage 1: 0.2/100/0.02 to 20/20/1.0; Stage 2: 15/1.0/2.5
    Hand-set training hyperparameters that shape model accuracy; not fitted against the benchmark and the central claim depends on them implicitly.
  • DP-GEN selection thresholds = Force uncertainty > 1.0 eV/A discard; top 10% then random downsample to 5k or 20k; enthalpy uncertainty in iterations 83
    Sampling strategy parameters that determine which structures receive DFT labels; chosen by hand, not optimized against the downstream benchmark.
axioms (5)
  • domain assumption PBE DFT without spin polarization, noncollinear magnetism, or Hubbard U gives sufficiently accurate energy, force, and virial labels for training and benchmarking.
    Section V.A explicitly excludes these effects for consistency, limiting accuracy for strongly magnetic materials but claiming a unified electronic-structure framework.
  • domain assumption VASP and ABACUS stress tensors and energies are comparable across the benchmarked models.
    Sections III.B and III.D use different DFT codes for OpenCSP versus baselines without reporting a cross-code validation.
  • domain assumption Literature on-hull structures at 50-200 GPa are correct reference structures and are not included in OpenCSP training labels.
    Section III.E relies on refs [55-61] for reference structures; no compositional exclusion statement is given for these benchmark tasks.
  • domain assumption Randomized CALYPSO sampling explores the enthalpy landscape broadly enough that recovering the known structure within 2,000 candidates is a meaningful model-quality test.
    Section III.E uses this protocol for all models; it is a fair test across models but assumes the search protocol itself is adequate.
  • domain assumption Models trained on the 0-100 GPa stress range can extrapolate reliably to 150-200 GPa.
    Section III.D tests up to 200 GPa while Section V.A describes training data with pressures randomly sampled from 0-100 GPa; the paper argues the success of pressure reproduction supports this extrapolation.

reviewed 2026-08-04 · how reviews work

0 comments
Cite this review

Pith. "Pith review of OpenCSP: A Deep Learning Framework for Crystal Structure Prediction from Ambient to High Pressure." pith.science (2026). https://pith.science/paper/COYAZ7FK

@misc{pith2026250910293,
  author       = {Pith},
  title        = {Pith review of: OpenCSP: A Deep Learning Framework for Crystal Structure Prediction from Ambient to High Pressure},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/COYAZ7FK}},
  note         = {Machine review of arXiv:2509.10293}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

High-pressure crystal structure prediction (CSP) underpins advances in condensed matter physics, planetary science, and materials discovery. Yet, most large atomistic models are trained on near-ambient, equilibrium data, leading to degraded stress accuracy at tens to hundreds of gigapascals and sparse coverage of pressure-stabilized stoichiometries and dense coordination motifs. Here, we introduce OpenCSP, a machine learning framework for CSP tasks spanning ambient to high-pressure conditions. This framework comprises an open-source pressure-resolved dataset alongside a suite of publicly available atomistic models that are jointly optimized for accuracy in energy, force, and stress predictions. The dataset is constructed via randomized high-pressure sampling and iteratively refined through an uncertainty-guided concurrent learning strategy, which enriches underrepresented compression regimes while suppressing redundant DFT labeling. Despite employing a training corpus one to two orders of magnitude smaller than those of leading large models, OpenCSP achieves comparable or superior performance in high-pressure enthalpy ranking and stability prediction. Across benchmark CSP tasks spanning a wide pressure window, our models match or surpass MACE-MPA-0, MatterSim v1 5M, and GRACE-2L-OAM, with the largest gains observed at elevated pressures. These results demonstrate that targeted, pressure-aware data acquisition coupled with scalable architectures enables data-efficient, high-fidelity CSP, paving the way for autonomous materials discovery under ambient and extreme conditions.

Figures

Figures reproduced from arXiv: 2509.10293 by Han Wang, Jian Lv, Jing Wu, Xiaoyang Wang, Yinan Wang, Zhenyu Wang.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) Distributions of number of atoms per structure, formation energy per atom, atomic force norm, and stress (pressure) [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Relaxation success rate for each model at target pres [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: reports the CSP success rate, χ2000. At 0 GPa, all models achieved comparable performances of approximately 60%, indicating similar baseline ca￾pability in recovering previously unseen on-hull struc￾tures under ambient conditions. At high pressures (50- 150 GPa), OpenCSP models show substantial improve￾ment, outperforming baseline models by ≈ 20-30 per￾centage points. Specifically, at 50 GPa, the OpenCSP m… view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Schematic illustration of the OpenCSP dataset construction workflow based on the concurrent learning scheme. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Assessing foundational atomistic models for iron alloys under Earth's core conditions

    physics.geo-ph 2026-05 unverdicted novelty 4.0

    Foundational atomistic models reproduce some structural and dynamical properties of iron alloys under core conditions but none consistently match first-principles benchmarks due to missing explicit treatment of therma...

Reference graph

Works this paper leans on

72 extracted references · 6 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Zhang, Y

    L. Zhang, Y. Wang, J. Lv, and Y. Ma, Materials dis- covery at high pressures, Nature Reviews Materials 2, 1 (2017)

    2017

  2. [2]

    H.-K. Mao, B. Chen, J. Chen, K. Li, J.-F. Lin, W. Yang, and H. Zheng, Recent advances in high-pressure science and technology, Matter and Radiation at Extremes 1, 59 (2016)

    2016

  3. [3]

    M. Xu, Y. Li, and Y. Ma, Materials by design at high pressures, Chemical Science 13, 329 (2022)

    2022

  4. [4]

    Z. Wang, X. Luo, Q. Wang, H. Ge, P. Gao, W. Zhang, J. Lv, and Y. Wang, Advances in high-pressure materi- als discovery enabled by machine learning, Matter and Radiation at Extremes 10 (2025)

    2025

  5. [5]

    Huang, D

    Q. Huang, D. Yu, B. Xu, W. Hu, Y. Ma, Y. Wang, Z. Zhao, B. Wen, J. He, Z. Liu, et al., Nanotwinned dia- mond with unprecedented hardness and stability, Nature 510, 250 (2014)

    2014

  6. [6]

    H. Liu, I. I. Naumov, R. Hoffmann, N. Ashcroft, and R. J. Hemley, Potential high-Tc superconducting lanthanum and yttrium hydrides at high pressure, Proceedings of the National Academy of Sciences 114, 6990 (2017)

    2017

  7. [7]

    F. Peng, Y. Sun, C. J. Pickard, R. J. Needs, Q. Wu, and Y. Ma, Hydrogen clathrate structures in rare earth hydrides at high pressures: possible route to room- temperature superconductivity, Physical Review Letters 119, 107001 (2017)

    2017

  8. [8]

    Drozdov, P

    A. Drozdov, P. Kong, V. Minkov, S. Besedin, M. Ku- zovnikov, S. Mozaffari, L. Balicas, F. F. Balakirev, D. Graf, V. Prakapenka, et al., Superconductivity at 250 K in lanthanum hydride under high pressures, Nature 569, 528 (2019)

    2019

  9. [9]

    Somayazulu, M

    M. Somayazulu, M. Ahart, A. K. Mishra, Z. M. Geballe, M. Baldini, Y. Meng, V. V. Struzhkin, and R. J. Hem- ley, Evidence for superconductivity above 260 K in lan- thanum superhydride at megabar pressures, Physical Re- view Letters 122, 027001 (2019)

    2019

  10. [10]

    Jayaraman, Diamond anvil cell and high-pressure physical investigations, Reviews of Modern Physics 55, 65 (1983)

    A. Jayaraman, Diamond anvil cell and high-pressure physical investigations, Reviews of Modern Physics 55, 65 (1983)

    1983

  11. [11]

    Dubrovinskaia, L

    N. Dubrovinskaia, L. Dubrovinsky, N. A. Solopova, A. Abakumov, S. Turner, M. Hanfland, E. Bykova, M. Bykov, C. Prescher, V. B. Prakapenka, et al. , Ter- apascal static pressure generation with ultrahigh yield strength nanodiamond, Science Advances 2, e1600341 (2016)

    2016

  12. [12]

    Jeanloz, P

    R. Jeanloz, P. M. Celliers, G. W. Collins, J. H. Eg- gert, K. K. Lee, R. S. McWilliams, S. Brygoo, and P. Loubeyre, Achieving high-density states through shock-wave loading of precompressed samples, Proceed- ings of the National Academy of Sciences 104, 9172 (2007)

    2007

  13. [13]

    Wang and Y

    Y. Wang and Y. Ma, Perspective: Crystal structure prediction at high pressures, The Journal of Chemical Physics 140 (2014)

    2014

  14. [14]

    A. R. Oganov, C. J. Pickard, Q. Zhu, and R. J. Needs, Structure prediction drives materials discovery, Nature Reviews Materials 4, 331 (2019)

    2019

  15. [15]

    Y. Wang, J. Lv, P. Gao, and Y. Ma, Crystal structure prediction via efficient sampling of the potential energy surface, Accounts of Chemical Research 55, 2068 (2022)

    2068

  16. [16]

    C. J. Pickard and R. Needs, Ab initio random struc- ture searching, Journal of Physics: Condensed Matter 23, 053201 (2011)

    2011

  17. [17]

    Y. Wang, J. Lv, L. Zhu, and Y. Ma, Crystal structure prediction via particle-swarm optimization, Physical Re- view B 82, 094116 (2010)

    2010

  18. [18]

    Y. Wang, J. Lv, L. Zhu, and Y. Ma, Calypso: A method for crystal structure prediction, Computer Physics Com- munications 183, 2063 (2012)

    2063

  19. [19]

    C. W. Glass, A. R. Oganov, and N. Hansen, USPEX- evolutionary crystal structure prediction, Computer Physics Communications 175, 713 (2006)

    2006

  20. [20]

    D. C. Lonie and E. Zurek, XtalOpt: An open-source evolutionary algorithm for crystal structure prediction, Computer Physics Communications 182, 372 (2011)

    2011

  21. [21]

    J. Wang, H. Gao, Y. Han, C. Ding, S. Pan, Y. Wang, Q. Jia, H.-T. Wang, D. Xing, and J. Sun, MAGUS: ma- chine learning and graph theory assisted universal struc- ture searcher, National Science Review 10, 128 (2023)

    2023

  22. [22]

    Barducci, M

    A. Barducci, M. Bonomi, and M. Parrinello, Metady- namics, Wiley Interdisciplinary Reviews: Computational Molecular Science 1, 826 (2011)

    2011

  23. [23]

    S. Goedecker, Minima hopping: An efficient search method for the global minimum of the potential energy surface of complex molecular systems, The Journal of Chemical Physics 120, 9911 (2004)

    2004

  24. [24]

    J. Lv, Y. Wang, L. Zhu, and Y. Ma, Predicted novel high- pressure phases of lithium, Physical Review Letters 106, 015503 (2011)

    2011

  25. [25]

    Mailhiot, L

    C. Mailhiot, L. Yang, and A. McMahan, Polymeric ni- trogen, Physical Review B 46, 14419 (1992). 14

    1992

  26. [26]

    M. I. Eremets, A. G. Gavriliuk, I. A. Trojan, D. A. Dzivenko, and R. Boehler, Single-bonded cubic form of nitrogen, Nature Materials 3, 558 (2004)

    2004

  27. [27]

    Z. Wang, X. Wang, X. Luo, P. Gao, Y. Sun, J. Lv, H. Wang, Y. Wang, and Y. Ma, Concurrent learning scheme for crystal structure prediction, Physical Review B 109, 094117 (2024)

    2024

  28. [28]

    Zhang, H

    D. Zhang, H. Bi, F. Dai, W. Jiang, X. Liu, L. Zhang, and H. Wang, Pretraining of attention-based deep learn- ing potential model for molecular simulation, npj Com- putational Materials 10, 94 (2024)

    2024

  29. [29]

    Zhang, X

    D. Zhang, X. Liu, X. Zhang, C. Zhang, C. Cai, H. Bi, Y. Du, X. Qin, A. Peng, J. Huang, et al., DPA-2: a large atomic model as a multi-task learner, npj Computational Materials 10, 293 (2024)

    2024

  30. [30]

    Zhang, A

    D. Zhang, A. Peng, C. Cai, W. Li, Y. Zhou, J. Zeng, M. Guo, C. Zhang, B. Li, H. Jiang, T. Zhu, W. Jia, L. Zhang, and H. Wang, A graph neural network for the era of large atomistic models, arXiv preprint arXiv:2506.01686 (2025)

    arXiv 2025

  31. [31]

    Chen and S

    C. Chen and S. P. Ong, A universal graph deep learn- ing interatomic potential for the periodic table, Nature Computational Science 2, 718 (2022)

    2022

  32. [32]

    B. Deng, P. Zhong, K. Jun, J. Riebesell, K. Han, C. J. Bartel, and G. Ceder, CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling, Nature Machine Intelligence 5, 1031 (2023)

    2023

  33. [33]

    Batatia, P

    I. Batatia, P. Benner, Y. Chiang, A. M. Elena, D. P. Kov´ acs, J. Riebesell, X. R. Advincula, M. Asta, M. Avaylon, W. J. Baldwin, et al. , A foundation model for atomistic materials chemistry, arXiv preprint arXiv:2401.00096 (2023)

    Pith/arXiv arXiv 2023

  34. [34]

    Merchant, S

    A. Merchant, S. Batzner, S. S. Schoenholz, M. Aykol, G. Cheon, and E. D. Cubuk, Scaling deep learning for materials discovery, Nature 624, 80 (2023)

    2023

  35. [35]

    Barroso-Luque, M

    L. Barroso-Luque, M. Shuaibi, X. Fu, B. M. Wood, M. Dzamba, M. Gao, A. Rizvi, C. L. Zitnick, and Z. W. Ulissi, Open materials 2024 (omat24) inorganic materi- als dataset and models, arXiv preprint arXiv:2410.12771 (2024)

    Pith/arXiv arXiv 2024

  36. [37]

    J. Kim, J. Kim, J. Kim, J. Lee, Y. Park, Y. Kang, and S. Han, Data-efficient multifidelity training for high- fidelity machine learning interatomic potentials, Journal of the American Chemical Society 147, 1042 (2024)

    2024

  37. [38]

    H. Yang, C. Hu, Y. Zhou, X. Liu, Y. Shi, J. Li, G. Li, Z. Chen, S. Chen, C. Zeni, et al. , Mattersim: A deep learning atomistic model across elements, temperatures and pressures, arXiv preprint arXiv:2405.04967 (2024)

    Pith/arXiv arXiv 2024

  38. [39]

    Bochkarev, Y

    A. Bochkarev, Y. Lysogorskiy, and R. Drautz, Graph atomic cluster expansion for semilocal interactions be- yond equivariant message passing, Physical Review X14, 021036 (2024)

    2024

  39. [40]

    C. Zeni, R. Pinsler, D. Z¨ ugner, A. Fowler, M. Horton, X. Fu, Z. Wang, A. Shysheya, J. Crabb´ e, S. Ueda,et al., A generative model for inorganic materials design, Na- ture , 1 (2025)

    2025

  40. [41]

    Schmidt, L

    J. Schmidt, L. Pettersson, C. Verdozzi, S. Botti, and M. A. Marques, Crystal graph attention networks for the prediction of stable materials, Science Advances 7, eabi7948 (2021)

    2021

  41. [42]

    Schmidt, N

    J. Schmidt, N. Hoffmann, H.-C. Wang, P. Borlido, P. J. Carri¸ co, T. F. Cerqueira, S. Botti, and M. A. Marques, Machine-learning-assisted determination of the global zero-temperature phase diagram of materials, Advanced Materials 35, 2210788 (2023)

    2023

  42. [43]

    H. Wang, J. Schmidt, M. A. Marques, L. Wirtz, and A. H. Romero, Symmetry-based computational search for novel binary and ternary 2D materials, 2D Materials 10, 035007 (2023)

    2023

  43. [44]

    X. Luo, Z. Wang, P. Gao, J. Lv, Y. Wang, C. Chen, and Y. Ma, Deep learning generative model for crystal structure prediction, npj Computational Materials 10, 254 (2024)

    2024

  44. [45]

    Zhang, H

    Y. Zhang, H. Wang, W. Chen, J. Zeng, L. Zhang, H. Wang, and W. E, DP-GEN: A concurrent learn- ing platform for the generation of reliable deep learning based potential energy models, Comput. Phys. Commun. , 107206 (2020)

    2020

  45. [46]

    Rhodes, S

    B. Rhodes, S. Vandenhaute, V. ˇSimkus, J. Gin, J. God- win, T. Duignan, and M. Neumann, Orb-v3: atom- istic simulation at scale, arXiv preprint arXiv:2504.06231 (2025)

    Pith/arXiv arXiv 2025

  46. [47]

    Y. Park, J. Kim, S. Hwang, and S. Han, Scalable par- allel algorithm for graph neural network interatomic po- tentials in molecular dynamics simulations, Journal of Chemical Theory and Computation 20, 4857 (2024)

    2024

  47. [48]

    Riebesell, R

    J. Riebesell, R. E. A. Goodall, P. Benner, Y. Chiang, B. Deng, G. Ceder, M. Asta, A. A. Lee, A. Jain, and K. A. Persson, Matbench discovery – a framework to evaluate machine learning crystal stability predictions, arXiv preprint arXiv:2308.14920 (2024)

    Pith/arXiv arXiv 2024

  48. [49]

    B. M. Wood, M. Dzamba, X. Fu, M. Gao, M. Shuaibi, L. BarrosoLuque, K. Abdelmaqsoud, V. Gharakhanyan, J. R. Kitchin, D. S. Levine, K. Michel, A. Sriram, T. Co- hen, A. Das, A. Rizvi, S. J. Sahoo, Z. W. Ulissi, and C. L. Zitnick, UMA: A family of universal models for atoms, arXiv preprint arXiv:2506.23971 (2025)

    arXiv 2025

  49. [50]

    M. Chen, G. Guo, and L. He, Systematically improv- able optimized atomic basis sets for ab initio calcula- tions, Journal of Physics: Condensed Matter 22, 445501 (2010)

    2010

  50. [51]

    P. Li, X. Liu, M. Chen, P. Lin, X. Ren, L. Lin, C. Yang, and L. He, Large-scale ab initio simulations based on systematically improvable atomic basis, Computational Materials Science 112, 503 (2016)

    2016

  51. [52]

    A. H. Larsen, J. J. Mortensen, J. Blomqvist, I. E. Castelli, R. Christensen, M. Du lak, J. Friis, M. N. Groves, B. Hammer, C. Hargus, et al. , The atomic sim- ulation environment - a python library for working with atoms, Journal of Physics: Condensed Matter 29, 273002 (2017)

    2017

  52. [53]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set, Phys. Rev. B 54, 11169 (1996)

    1996

  53. [54]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set, Computational Materials Science 6, 15 (1996)

    1996

  54. [55]

    G. Yang, Y. Wang, F. Peng, A. Bergara, and Y. Ma, Gold as a 6p-element in dense lithium aurides, Journal of the American Chemical Society 138, 4046 (2016)

    2016

  55. [56]

    S. Shao, J. Lv, X. Li, L. Li, P. Liu, Z. Liu, C. Chen, Y. Wang, and Y. Ma, New MgSiO 4H2 phases as poten- tial primary water carriers into the deep earth, Physical 15 Review Letters 133, 214101 (2024)

    2024

  56. [57]

    M. Miao, X. Wang, J. Brgoch, F. Spera, M. G. Jackson, G. Kresse, and H. Lin, Anionic chemistry of noble gases: formation of Mg-NG (NG= Xe, Kr, Ar) compounds un- der pressure, Journal of the American Chemical Society 137, 14122 (2015)

    2015

  57. [58]

    Miao, Caesium in high oxidation states and as ap- block element, Nature Chemistry 5, 846 (2013)

    M. Miao, Caesium in high oxidation states and as ap- block element, Nature Chemistry 5, 846 (2013)

    2013

  58. [59]

    Botana and M

    J. Botana and M. Miao, Pressure-stabilized lithium cae- sides with caesium anions beyond the- 1 state, Nature Communications 5, 4861 (2014)

    2014

  59. [60]

    S. Shao, W. Zhu, J. Lv, Y. Wang, Y. Chen, and Y. Ma, The exotically stoichiometric compounds in Al-S system under high pressure, npj Computational Materials 6, 11 (2020)

    2020

  60. [61]

    F. Peng, Y. Wang, H. Wang, Y. Zhang, and Y. Ma, Stable xenon nitride at high pressures, Physical Review B 92, 094104 (2015)

    2015

  61. [62]

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

    2013

  62. [63]

    A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, et al., Commentary: The Materials Project: A materials genome approach to accelerating materials innovation, APL materials 1 (2013)

    2013

  63. [64]

    A. E. Mattsson, A lithium projector augmented wave po- tential suitable for use in V ASP at high compression and temperature, Tech. Rep. (Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States), 2012)

    2012

  64. [65]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized Gradient Approximation Made Simple, Physical Review Letters 77, 3865 (1996)

    1996

  65. [66]

    Schlipf and F

    M. Schlipf and F. Gygi, Optimization algorithm for the generation of ONCV pseudopotentials, Computer Physics Communications 196, 36 (2015)

    2015

  66. [67]

    M. J. Van Setten, M. Giantomassi, E. Bousquet, M. J. Verstraete, D. R. Hamann, X. Gonze, and G. Rignanese, The PseudoDojo: Training and grading a 85 element optimized norm-conserving pseudopotential table, Com- puter Physics Communications 226, 39 (2018)

    2018

  67. [68]

    C. Wang, R. Prange, and V. Korenman, Magnetism in iron and nickel, Physical Review B 25, 5766 (1982)

    1982

  68. [69]

    Kubler, K.-H

    J. Kubler, K.-H. Hock, J. Sticht, and A. Williams, Den- sity functional theory of non-collinear magnetism, Jour- nal of Physics F: Metal Physics 18, 469 (1988)

    1988

  69. [70]

    H. J. Kulik, M. Cococcioni, D. A. Scherlis, and N. Marzari, Density functional theory in transition-metal chemistry: A self-consistent hubbard U approach, Phys- ical Review Letters 97, 103001 (2006)

    2006

  70. [71]

    Lopanitsyna, G

    N. Lopanitsyna, G. Fraux, M. A. Springer, S. De, and M. Ceriotti, Modeling high-entropy transition metal al- loys with alchemical compression, Physical Review Ma- terials 7, 045802 (2023)

    2023

  71. [72]

    Mazitov, M

    A. Mazitov, M. A. Springer, N. Lopanitsyna, G. Fraux, S. De, and M. Ceriotti, Surface segregation in high- entropy alloys from alchemical machine learning, Journal of Physics: Materials 7, 025007 (2024)

    2024

  72. [73]

    Mazitov, F

    A. Mazitov, F. Bigi, M. Kellner, P. Pegolo, D. Tisi, G. Fraux, S. Pozdnyakov, P. Loche, and M. Ceriotti, PET-MAD, a lightweight universal interatomic poten- tial for advanced materials modeling, arXiv preprint arXiv:2503.14118 (2025)

    Pith/arXiv arXiv 2025

This paper was first reviewed by deepseek-v4-flash on August 4, 2026.