CoChat
Sign Up
Cited by 968Machine Learning in Materials ScienceAdvanced Chemical Physics StudiesFusion materials and technologies

The DOI matches its Crossref record; 12 of the 12 references it cites were found in a registry or an index.

Performance and Cost Assessment of Machine Learning Interatomic Potentials

Open access
Zuo, Yunxing, Chen, Chi, Li, Xiangguo, Deng, Zhi, Chen, Yiming, Behler, Jörg, Csányi, Gábor, Shapeev, Alexander V., Thompson, Aidan P., Wood, Mitchell A., Ong, Shyue Ping·2020·The Journal of Physical Chemistry A
DOI 10.1021/acs.jpca.9b08723
Open this paper in CoChat — free accountOpen PDF

Abstract

Machine learning of the quantitative relationship between local environment descriptors and the potential energy surface of a system of atoms has emerged as a new frontier in the development of interatomic potentials (IAPs). Here, we present a comprehensive evaluation of machine learning IAPs (ML-IAPs) based on four local environment descriptors-atom-centered symmetry functions (ACSF), smooth overlap of atomic positions (SOAP), the spectral neighbor analysis potential (SNAP) bispectrum components, and moment tensors-using a diverse data set generated using high-throughput density functional theory (DFT) calculations. The data set comprising bcc (Li, Mo) and fcc (Cu, Ni) metals and diamond group IV semiconductors (Si, Ge) is chosen to span a range of crystal structures and bonding. All descriptors studied show excellent performance in predicting energies and forces far surpassing that of classical IAPs, as well as predicting properties such as elastic constants and phonon dispersion curves. We observe a general trade-off between accuracy and the degrees of freedom of each model and, consequently, computational cost. We will discuss these trade-offs in the context of model selection for molecular dynamics and other applications.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1021/acs.jpca.9b08723

Cites

12 of 12 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The scale of the GNoME dataset allows the pretrained potential to outperform existing general-purpose potentials and makes it competitive with models trained explicitly on hundreds of samples from the target data distributions.

    “In particular, we find that the scale of the GNoME dataset allows it to outperform existing general-purpose potentials (see Fig. 3d) and makes the pretrained potential competitive with models trained explicitly on hundreds of samples from the target data distributions (see Supplementary Note 6.4).”

    From Scaling deep learning for materials discovery (2023) · cites this paper as [56]

    The Conclusions discuss GAP, MTP and SNAP models trained on approximately 100–200 structures, but do not compare them with a GNoME-pretrained potential. Searches for GNoME, pretrained potentials and the claimed comparison found no passage addressing those points in our copy. (LLM, conf=0.76)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Flagged first, then by citations

Research works printed but not linked: 87

  1. 10.1103/PhysRevLett.51.1888 · DOI
  2. 10.1073/pnas.1115123108 · DOI
  3. 10.1073/pnas.0901093106 · DOI
  4. 10.1021/ct400050d · DOI
  5. 10.1021/ja00051a040 · DOI
  6. 10.1021/j100389a010 · DOI
  7. 10.1021/jp004368u · DOI
  8. 10.1063/1.4966192 · DOI
  9. 10.1103/PhysRevLett.98.146401 · DOI · Paper page
  10. 10.1103/PhysRevB.90.104108 · DOI
  11. 10.1103/PhysRevMaterials.2.013808 · DOI
  12. 10.1016/j.jcp.2014.12.018 · DOI
  13. 10.1103/PhysRevMaterials.1.043603 · DOI
  14. 10.1103/PhysRevB.98.094104 · DOI
  15. 10.1038/s41524-019-0212-1 · DOI
  16. 10.1137/15M1054183 · DOI
  17. 10.1016/j.commatsci.2017.08.031 · DOI
  18. 10.1016/j.commatsci.2018.09.031 · DOI
  19. 10.1002/qua.24836 · DOI
  20. 10.1103/PhysRevB.92.094306 · DOI
  21. 10.1021/acs.jpcc.6b10908 · DOI
  22. 10.1103/PhysRevLett.114.096405 · DOI
  23. 10.1021/acs.jctc.8b00908 · DOI
  24. 10.1038/s41467-017-00839-3 · DOI
  25. 10.1002/qua.24954 · DOI
  26. 10.1103/PhysRevB.99.064103 · DOI
  27. 10.1103/PhysRevLett.120.143001 · DOI
  28. 10.1063/1.5009502 · DOI
  29. 10.1039/C7SC04934J · DOI
  30. 10.1063/1.5024577 · DOI · Paper page
  31. Himanen, L.; Jäger, M. O. J.; Morooka, E. V.; Canova, F. F.; Ranawat, Y. S.; Gao, D. Z.; Rinke, P.; Foster, A. S. DScribe: Library of Descriptors for Machine Learning in Materials Science. arXiv:1904.08875 2019. · DOI
  32. 10.1038/s41524-019-0189-9 · DOI
  33. 10.1103/PhysRevB.46.2250 · DOI
  34. 10.1016/0168-583X(86)90243-0 · DOI
  35. 10.1088/0953-8984/15/41/004 · DOI
  36. 10.1002/qua.24890 · DOI · Paper page
  37. 10.1002/anie.201703114 · DOI
  38. 10.1021/acs.jpcb.8b06476 · DOI
  39. 10.1103/PhysRevB.99.184305 · DOI
  40. 10.1063/1.3553717 · DOI
  41. 10.1016/j.commatsci.2015.11.047 · DOI
  42. 10.1073/pnas.1602375113 · DOI
  43. 10.1021/acs.jpclett.7b00358 · DOI
  44. 10.1021/acs.jctc.8b01288 · DOI
  45. 10.1103/PhysRevB.83.153101 · DOI
  46. 10.1063/1.4997242 · DOI
  47. 10.1103/PhysRevLett.104.136403 · DOI · Paper page
  48. 10.1103/PhysRevLett.120.156001 · DOI
  49. 10.1103/PhysRevB.95.094203 · DOI
  50. 10.1103/PhysRevB.97.054303 · DOI
  51. 10.1103/PhysRevX.8.041048 · DOI
  52. 10.1021/acs.jpclett.8b00902 · DOI
  53. 10.1063/1.5017641 · DOI
  54. 10.1016/j.commatsci.2019.03.049 · DOI
  55. 10.1103/PhysRevB.99.064114 · DOI
  56. 10.1039/C8CP06037A · DOI
  57. 10.1038/sdata.2015.9 · DOI
  58. 10.1038/sdata.2016.80 · DOI
  59. 10.1016/j.commatsci.2012.10.028 · DOI · Paper page
  60. 10.1002/cpe.3505 · DOI
  61. 10.1103/PhysRevB.68.144112 · DOI
  62. 10.1103/PhysRevB.33.7983 · DOI
  63. 10.1103/PhysRevB.59.14220 · DOI
  64. 10.1103/PhysRevB.41.3316 · DOI
  65. 10.1103/PhysRevB.40.6085 · DOI
  66. 10.1021/acs.jctc.8b00770 · DOI
  67. 10.1021/acs.jpclett.9b02037 · DOI
  68. 10.1109/CVPR.2017.351 · DOI · Paper page
  69. Zhang L., 2018, Proceedings of the 32nd International Conference on Neural Information Processing Systems, 4441
  70. 10.1073/pnas.1815117116 · DOI
  71. 10.1103/PhysRevB.69.144113 · DOI
  72. 10.1103/PhysRevB.93.184101 · DOI
  73. 10.1016/j.actamat.2014.12.010 · DOI
  74. 10.1103/PhysRevB.85.214121 · DOI
  75. 10.1088/0965-0393/8/6/305 · DOI
  76. 10.1016/j.commatsci.2006.07.013 · DOI
  77. 10.1016/j.jmgm.2016.11.009 · DOI
  78. 10.1063/1.1323224 · DOI
  79. 10.1088/0370-1298/65/5/307 · DOI
  80. 10.1080/0141861021000034568 · DOI
  81. 10.1080/01418619608240729 · DOI
  82. 10.1088/0965-0393/8/2/302 · DOI
  83. 10.1063/1.4870462 · DOI
  84. 10.1080/10408436.2013.772503 · DOI
  85. 10.1103/PhysRevB.79.155107 · DOI
  86. 10.1103/PhysRevB.69.075102 · DOI
  87. 10.1126/sciadv.1603015 · DOI

Cited by

Sorted by citations

Metadata

Authors
Zuo, Y., Chen, C., Li, X., Deng, Z., Chen, Y., Behler, J., Csányi, G., Shapeev, A. V., Thompson, A. P., Wood, M. A., & Ong, S. P.
Affiliations
  • University of California San Diego (US)
  • University of Göttingen (DE)
  • Department of Engineering, University of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, U.K.
  • Skolkovo Institute of Science and Technology (RU)
  • Sandia National Laboratories (US)
Corresponding author
Ong, Shyue Ping
ORCID
Venue
The Journal of Physical Chemistry A
Publisher
American Chemical Society (ACS)
Published
2020-01-09
Volume
124
Issue
4
Pages
731-745
ISSN
  • 1089-5639 (ISSN-L)
  • 1520-5215
DOI
10.1021/acs.jpca.9b08723
OpenAlex
W3000291311
Cited by
968
References
12
Citations by year
2026: 138 · 2025: 178 · 2024: 177 · 2023: 196 · 2022: 111 · 2021: 94 · 2020: 70
FWCI
43.98
Type
journal article
Language
en
Open access
Licence
Locations
  • The Journal of Physical Chemistry A (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Apollo (University of Cambridge) (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
  • Cambridge University Engineering Department Publications Database (submittedVersion)
  • OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) (submittedVersion)
  • GoeScholar The Publication Server of the Georg-August-Universität Göttingen (Georg-August-Universität Göttingen) (submittedVersion, cc-by)
  • arXiv (Cornell University)
  • Apollo
Funders
  • Russian Science Foundation (18-13-00479)
  • Deutsche Forschungsgemeinschaft (Be3264/11-2)
  • Office of Naval Research (N00014-16-1-2621)
Topics
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Advanced Chemical Physics Studies
  • Physical Sciences › Materials Science › Materials Chemistry › Fusion materials and technologies
Keywords
machine learning interatomic potentials, atom-centered symmetry functions, soap, moment tensor, potential energy surface, elastic constants, phonon dispersion, computational cost, model selection, molecular dynamics, body-centered cubic metals, face-centered cubic metals

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review