CoChat
Sign Up
Cited by 446Machine Learning in Materials ScienceAdvanced Chemical Physics StudiesComputational Drug Discovery Methods

Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements

Open access
Takamoto, So, Shinagawa, Chikashi, Motoki, Daisuke, Nakago, Kosuke, Li, Wenwen, Kurata, Iori, Watanabe, Taku, Yayama, Yoshihiro +14 more·2022·Nature Communications
DOI 10.1038/s41467-022-30687-9PMC PMC9151783
Open this paper in CoChat — free accountOpen PDF

Abstract

Computational material discovery is under intense study owing to its ability to explore the vast space of chemical systems. Neural network potentials (NNPs) have been shown to be particularly effective in conducting atomistic simulations for such purposes. However, existing NNPs are generally designed for narrow target materials, making them unsuitable for broader applications in material discovery. Here we report a development of universal NNP called PreFerred Potential (PFP), which is able to handle any combination of 45 elements. Particular emphasis is placed on the datasets, which include a diverse set of virtual structures used to attain the universality. We demonstrated the applicability of PFP in selected domains: lithium diffusion in LiFeSO 4 F, molecular adsorption in metal-organic frameworks, an order–disorder transition of Cu-Au alloys, and material discovery for a Fischer–Tropsch catalyst. They showcase the power of PFP, and this technology provides a highly useful tool for material discovery.

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 · 10.1038/s41467-022-30687-9

Cites

8 of 8 references verified0 citation errors · 0 not checkable

Statistics

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?

  • Universal MLPs are commonly trained on datasets comprising only close-to-equilibrium structures, so caution is needed when applying them to activated events.

    “Also, commonly, the universal MLPs are trained on datasets comprising only close-to-equilibrium structures25,27–29, hence caution is needed when applying them to activated events.”

    From A reactive neural network framework for water-loaded acidic zeolites (2024) · cites this paper as [25]

    The Methods section (‘Dataset systems and structures’; page not supplied) says early datasets mainly contain optimized structures and are insufficient for reactions and diffusion, which involve structures far from equilibrium. But it also describes datasets sampling surrounding regions, while the Introduction says this paper’s own universal potential was trained on unstable structures. The source therefore supports caution about limited training coverage, not the claim that universal MLPs are commonly trained on only close-to-equilibrium structures. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

  • Universal MLPs cover large parts of the chemical space with dozens of elements but have limited coverage of the configuration space, e.g., not considering all relevant chemical reactions with associated transition states.

    “Two paradigms dominate the MLP field currently: (i) training of universal MLP that cover large parts of the chemical space with dozens of elements, but with limited coverage of the configuration space, e.g., not considering all relevant chemical reactions with the associated transition states, e.g., OC22, CHGNet, and others25–29”

    From A reactive neural network framework for water-loaded acidic zeolites (2024) · cites this paper as [25]

    The Abstract and Methods describe a potential spanning 45 elements and a training dataset with no transition-state structures, supporting the claim about broad chemical coverage and a gap in training coverage. But the Results report that the model predicts transition-state energies and evaluates chemical-reaction pathways, so the paper does not support an unqualified claim that it does not consider reactions or their transition states. The indexed copy provides section labels but no page numbers. (LLM, conf=0.85)

    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.

Cites

Flagged first, then by citations

Research works printed but not linked: 64

  1. [2]Jones JE On the determination of molecular fields.-i. from the variation of the viscosity of a gas with temperature Proc. R. Soc. Lond. Ser. A 1924 106 441 462 10.1098/rspa.1924.0081 · DOI
  2. [3]Daw MS Baskes MI Embedded-atom method: derivation and application to impurities, surfaces and other defects in metals Phys. Rev. B 1984 29 6443 6453 10.1103/PhysRevB.29.6443 · DOI
  3. [4]Finnis MW Sinclair JE A simple empirical n-body potential for transition metals Philos. Mag. A 1984 50 45 55 10.1080/01418618408244210 · DOI
  4. [5]Tersoff J Modeling solid-state chemistry: Interatomic potentials for multicomponent systems Phys. Rev. B 1989 39 5566 5568 10.1103/PhysRevB.39.5566 9948964 · DOI
  5. [6]van Duin ACT Dasgupta S Lorant F Goddard WA Reaxff: A reactive force field for hydrocarbons J. Phys. Chem. A 2001 105 9396 9409 10.1021/jp004368u · DOI
  6. [7]Senftle TP The reaxff reactive force-field: development, applications and future directions npj Comput. Mater. 2016 2 1 14 10.1038/npjcompumats.2015.11 · DOI · Paper page
  7. [9]Bartók AP Payne MC Kondor R Csányi G Gaussian approximation potentials: the accuracy of quantum mechanics, without the electrons Phys. Rev. Lett. 2010 104 136403 10.1103/PhysRevLett.104.136403 20481899 · DOI
  8. [11]Vitek, A., Stachon, M., Krömer, P. & Snášel, V. Towards the modeling of atomic and molecular clusters energy by support vector regression. In Proc. 2013 5th International Conference on Intelligent Networking and Collaborative Systems, INCOS ’13, USA, 2013 (eds Xhafa, F., Barolli, L. & Chen, X.) (IEEE Computer Society).
  9. [12]Ruddigkeit L van Deursen R Blum LC Reymond J-L Enumeration of 166 billion organic small molecules in the chemical universe database gdb-17 J. Chem. Inf. Model. 2012 52 2864 2875 10.1021/ci300415d 23088335 · DOI · Paper page
  10. [14]Smith JS Isayev O Roitberg AE Ani-1: an extensible neural network potential with dft accuracy at force field computational cost Chem. Sci. 2017 8 3192 3203 10.1039/C6SC05720A 28507695 PMC5414547 · DOI · Paper page
  11. [15]Devereux C Extending the applicability of the ANI deep learning molecular potential to sulfur and halogens J. Chem. Theory Comput. 2020 16 4192 4202 10.1021/acs.jctc.0c00121 32543858 · DOI
  12. [16]Xie T Grossman JC Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties Phys. Rev. Lett. 2018 120 145301 10.1103/PhysRevLett.120.145301 29694125 · DOI · Paper page
  13. [17]Chen C Ye W Zuo Y Zheng C Ong SP Graph networks as a universal machine learning framework for molecules and crystals Chem. Mater. 2019 31 3564 3572 10.1021/acs.chemmater.9b01294 · DOI · Paper page
  14. [19]Zitnick, C. L. et al. An introduction to electrocatalyst design using machine learning for renewable energy storage. Preprint at 10.48550/arXiv.2010.09435 (2020). · DOI
  15. [20]Chanussot L Open catalyst 2020 (oc20) dataset and community challenges ACS Catal. 2021 11 6059 6072 10.1021/acscatal.0c04525 · DOI
  16. [22]Liu, H., Dai, Z., So, D. R. & Le, Q. V. Pay attention to mlps. Preprint at 10.48550/arXiv.2105.08050 (2021). · DOI
  17. [23]Van der Ven A Aydinol MK Ceder G Kresse G Hafner J First-principles investigation of phase stability in Lixcoo2 Phys. Rev. B 1998 58 2975 10.1103/PhysRevB.58.2975 · DOI
  18. [24]Yamada Y Hydrate-melt electrolytes for high-energy-density aqueous batteries Nat. Energy 2016 1 1 9 10.1038/nenergy.2016.129 · DOI
  19. [25]Morgan D Van der Ven A Ceder G Li conductivity in li x mpo 4 (m= mn, fe, co, ni) olivine materials Electrochem. Solid State Lett. 2003 7 A30 10.1149/1.1633511 · DOI
  20. [26]He X Zhu Y Mo Y Origin of fast ion diffusion in super-ionic conductors Nat. Commun. 2017 8 1 7 10.1038/s41467-016-0009-6 28635958 PMC5482052 · DOI · Paper page
  21. [27]Jónsson, H., Mills, G. & Jacobsen, K. W. Nudged Elastic Band Method for Finding Minimum Energy Paths of Transitions 385–404 (World Scientific Pub Co Pte Ltd, 1998).
  22. [28]Henkelman G Uberuaga BP Jónsson H A climbing image nudged elastic band method for finding saddle points and minimum energy paths J. Chem. Phys. 2000 113 9901 9904 10.1063/1.1329672 · DOI
  23. [29]Recham N A 3.6 v lithium-based fluorosulphate insertion positive electrode for lithium-ion batteries Nat. Mater. 2010 9 68 74 10.1038/nmat2590 19946280 · DOI
  24. [30]Mueller T Hautier G Jain A Ceder G Evaluation of tavorite-structured cathode materials for lithium-ion batteries using high-throughput computing Chem. Mater. 2011 23 3854 3862 10.1021/cm200753g · DOI
  25. [31]Momma K Izumi F Vesta 3 for three-dimensional visualization of crystal, volumetric and morphology data J. Appl. Crystallogr. 2011 44 1272 1276 10.1107/S0021889811038970 · DOI
  26. [32]Dudarev SL Botton GA Savrasov SY Humphreys CJ Sutton AP Electron-energy-loss spectra and the structural stability of nickel oxide: an lsda+ u study Phys. Rev. B 1998 57 1505 10.1103/PhysRevB.57.1505 · DOI
  27. [33]Groom CR Bruno IJ Lightfoot MP Ward SC The Cambridge structural database Acta Crystallogr. Sect. B 2016 72 171 179 10.1107/S2052520616003954 PMC4822653 27048719 · DOI · Paper page
  28. [34]Fletcher, R. Practical Methods of Optimization, 2nd edn (Wiley, 2000).
  29. [35]Paszke, A. et al. Pytorch: an imperative style, high-performance deep learning library. In Advances in Neural Information Processing Systems, Vol. 32 (eds, Wallach, H., Larochelle, H., Beygelzimer, A., D’Alché-Buc, F., Fox, E. & Garnett, R.) (Curran Associates, Inc., 2019).
  30. [37]Dietzel PDC Johnsen RE Blom R Fjellvåg H Structural changes and coordinatively unsaturated metal atoms on dehydration of honeycomb analogous microporous metal-organic frameworks Chem. - Eur. J. 2008 14 2389 2397 10.1002/chem.200701370 18203217 · DOI
  31. [38]Furukawa, H., Cordova, K. E., O’Keeffe, M. & Yaghi, O. M. The chemistry and applications of metal-organic frameworks. Science341, 1230444 (2013). 23990564
  32. [39]Chui SS-Y Lo SM-F Charmant JPH Orpen AG Williams ID A chemically functionalizable nanoporous material [Cu3(TMA)2(H2o)3]n Science 1999 283 1148 1150 10.1126/science.283.5405.1148 10024237 · DOI
  33. [40]Liu X Wang A Zhang T Su D-S Mou C-Y Au–Cu alloy nanoparticles supported on silica gel as catalyst for co oxidation: effects of Au/Cu ratios Catal. Today 2011 160 103 108 10.1016/j.cattod.2010.05.019 · DOI
  34. [41]Li W Wang A Liu X Zhang T Silica-supported Au–Cu alloy nanoparticles as an efficient catalyst for selective oxidation of alcohols Appl. Catal. A: Gen. 2012 433–434 146 151 10.1016/j.apcata.2012.05.014 · DOI
  35. [42]Najafishirtari S Nanoscale transformations of alumina-supported AuCu ordered phase nanocrystals and their activity in co oxidation ACS Catal. 2015 5 2154 2163 10.1021/cs501923x · DOI
  36. [43]Guisbiers G Gold-copper nano-alloy, “tumbaga”, in the era of nano: phase diagram and segregation Nano Lett. 2014 14 6718 6726 10.1021/nl503584q 25338111 PMC4246003 · DOI
  37. [44]Mendoza-Cruz R Order–disorder phase transitions in Au–Cu nanocubes: from nano-thermodynamics to synthesis Nanoscale 2017 9 9267 9274 10.1039/C7NR00028F 28561093 · DOI
  38. [46]Zijlstra B The vital role of step-edge sites for both co activation and chain growth on cobalt fischer-tropsch catalysts revealed through first-principles-based microkinetic modeling including lateral interactions ACS Catal. 2020 10 9376 9400 10.1021/acscatal.0c02420 · DOI
  39. [47]Zijlstra B Broos RJP Chen W Filot IAW Hensen EJM First-principles based microkinetic modeling of transient kinetics of co hydrogenation on cobalt catalysts Catal. Today 2020 342 131 141 10.1016/j.cattod.2019.03.002 · DOI
  40. [48]Zijlstra B Coverage effects in co dissociation on metallic cobalt nanoparticles ACS Catal. 2019 9 7365 7372 10.1021/acscatal.9b01967 · DOI
  41. [49]Wang T Effect of vanadium promotion on activated carbon-supported cobalt catalysts in fischer–tropsch synthesis Catal. Lett. 2006 107 47 52 10.1007/s10562-005-9730-1 · DOI
  42. [50]Shimura K Miyazawa T Hanaoka T Hirata S Fischer–Tropsch synthesis over alumina supported cobalt catalyst: Effect of promoter addition Appl. Catal. A: Gen. 2015 494 1 11 10.1016/j.apcata.2015.01.017 · DOI
  43. [51]Yao K Herr JE Toth DW Mckintyre R Parkhill J The tensormol-0.1 model chemistry: a neural network augmented with long-range physics Chem. Sci. 2018 9 2261 2269 10.1039/C7SC04934J 29719699 PMC5897848 · DOI
  44. [53]Kirklin S The open quantum materials database (oqmd): assessing the accuracy of dft formation energies npj Comput. Mater. 2015 1 15010 10.1038/npjcompumats.2015.10 · DOI
  45. [55]Choudhary K The joint automated repository for various integrated simulations (jarvis) for data-driven materials design npj Comput. Mater. 2020 6 173 10.1038/s41524-020-00440-1 · DOI
  46. [56]Curtarolo S Aflow: an automatic framework for high-throughput materials discovery Comput. Mater. Sci. 2012 58 218 226 10.1016/j.commatsci.2012.02.005 · DOI
  47. [57]Nakata M Shimazaki T Pubchemqc project: a large-scale first-principles electronic structure database for data-driven chemistry J. Chem. Inf. Model. 2017 57 1300 1308 10.1021/acs.jcim.7b00083 28481528 · DOI
  48. [58]Chmiela, S. et al. Machine learning of accurate energy-conserving molecular force fields. Sci. Adv.3, e1603015 (2017). PMC5419702 28508076
  49. [59]Unke OT Meuwly M Physnet: a neural network for predicting energies, forces, dipole moments and partial charges J. Chem. Theory Comput. 2019 15 3678 3693 10.1021/acs.jctc.9b00181 31042390 · DOI
  50. [60]Takamoto S Izumi S Li J Teanet: universal neural network interatomic potential inspired by iterative electronic relaxations Comput. Mater. Sci. 2022 207 111280 10.1016/j.commatsci.2022.111280 · DOI
  51. [61]Chai J-D Head-Gordon M Long-range corrected hybrid density functionals with damped atom–atom dispersion corrections Phys. Chem. Chem. Phys. 2008 10 6615 6620 10.1039/b810189b 18989472 · DOI · Paper page
  52. [62]Ditchfield R Hehre WJ Pople JA Self-consistent molecular-orbital methods. ix. an extended gaussian-type basis for molecular-orbital studies of organic molecules J. Chem. Phys. 1971 54 724 728 10.1063/1.1674902 · DOI · Paper page
  53. [64]Perdew JP Burke K Ernzerhof M Generalized gradient approximation made simple Phys. Rev. Lett. 1996 77 3865 10.1103/PhysRevLett.77.3865 10062328 · DOI · Paper page
  54. [65]Kresse G Hafner J Ab initio molecular dynamics for liquid metals Phys. Rev. B 1993 47 558 10.1103/PhysRevB.47.558 10004490 · DOI
  55. [66]Kresse G Hafner J Ab initio molecular-dynamics simulation of the liquid-metal–amorphous–semiconductor transition in germanium Phys. Rev. B 1994 49 14251 10.1103/PhysRevB.49.14251 10010505 · DOI
  56. [67]Kresse G Furthmüller J Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set Comput. Mater. Sci. 1996 6 15 50 10.1016/0927-0256(96)00008-0 9984901 · DOI
  57. [68]Kresse G Furthmüller J Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set Phys. Rev. B 1996 54 11169 10.1103/PhysRevB.54.11169 9984901 · DOI · Paper page
  58. [69]Hacene M Accelerating vasp electronic structure calculations using graphic processing units J. Comput. Chem. 2012 33 2581 2589 10.1002/jcc.23096 22903247 · DOI
  59. [70]Hutchinson M Widom M Vasp on a gpu: application to exact-exchange calculations of the stability of elemental boron Comput. Phys. Commun. 2012 183 1422 1426 10.1016/j.cpc.2012.02.017 · DOI
  60. [73]Tang W Sanville E Henkelman G A grid-based bader analysis algorithm without lattice bias J. Phys.: Condens. Matter 2009 21 084204 21817356
  61. [74]Sanville E Kenny SD Smith R Henkelman G Improved grid-based algorithm for bader charge allocation J. Comput. Chem. 2007 28 899 908 10.1002/jcc.20575 17238168 · DOI
  62. [75]Henkelman G Arnaldsson A Jónsson H A fast and robust algorithm for bader decomposition of charge density Comput. Mater. Sci. 2006 36 354 360 10.1016/j.commatsci.2005.04.010 · DOI · Paper page
  63. [76]Yu M Trinkle DR Accurate and efficient algorithm for bader charge integration J. Chem. Phys. 2011 134 064111 10.1063/1.3553716 21322665 · DOI
  64. [77]Wang L Maxisch T Ceder G Oxidation energies of transition metal oxides within the gga+ u framework Phys. Rev. B 2006 73 195107 10.1103/PhysRevB.73.195107 · DOI

Cited by

Sorted by citations

Authors
Takamoto, S., Shinagawa, C., Motoki, D., Nakago, K., Li, W., Kurata, I., Watanabe, T., Yayama, Y., Iriguchi, H., Asano, Y., Onodera, T., Ishii, T., Kudo, T., Ono, H., Sawada, R., Ishitani, R., Ong, M., Yamaguchi, T., Kataoka, T., Hayashi, A., Charoenphakdee, N., & Ibuka, T.
Affiliations
  • Preferred Networks (Japan) (JP)
  • Eneos (Japan) (JP)
Corresponding author
  • Takamoto, So
  • Ibuka, Takeshi
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2022-05-30
Volume
13
Issue
1
Pages
2991-2991
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-022-30687-9
PMC
PMC9151783
OpenAlex
W4282041110
Cited by
446
References
8
Citations by year
2026: 122 · 2025: 143 · 2024: 113 · 2023: 52 · 2022: 16
FWCI
22.77
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University)
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 › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
Keywords
neural network potentials, universal interatomic potentials, computational materials discovery, atomistic simulation, virtual structure, lithium ion diffusion, LiFeSO4F, molecular adsorption, metal-organic frameworks, order-disorder transition, Cu-Au alloys, Fischer-Tropsch catalysts, machine learning interatomic potentials
MeSH terms
Metal-Organic Frameworks, Adsorption, Catalysis, Neural Networks, Computer

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