Systematic softening in universal machine learning interatomic potentials
Source abstract
Abstract
Abstract Machine learning interatomic potentials (MLIPs) have introduced a new paradigm for atomic simulations. Recent advancements have led to universal MLIPs (uMLIPs) that are pre-trained on diverse datasets, providing opportunities for universal force fields and foundational machine learning models. However, their performance in extrapolating to out-of-distribution complex atomic environments remains unclear. In this study, we highlight a consistent potential energy surface (PES) softening effect in three uMLIPs: M3GNet, CHGNet, and MACE-MP-0, which is characterized by energy and force underprediction in atomic-modeling benchmarks including surfaces, defects, solid-solution energetics, ion migration barriers, phonon vibration modes, and general high-energy states. The PES softening behavior originates primarily from the systematically underpredicted PES curvature, which derives from the biased sampling of near-equilibrium atomic arrangements in uMLIP pre-training datasets. Our findings suggest that a considerable fraction of uMLIP errors are highly systematic, and can therefore be efficiently corrected. We argue for the importance of a comprehensive materials dataset with improved PES sampling for next-generation foundational MLIPs.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Machine Learning in Materials Science · X-ray Diffraction in Crystallography · Nuclear Materials and Properties
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (67)
Show source-provided references
1. Wang, H. et al. Scientific discovery in the age of artificial intelligence. Nature 620, 47–60 (2023).10.1038/s41586-023-06221-2 ↗
2. Wang, H., Zhang, L., Han, J. & Weinan, E. Deepmd-kit: A deep learning package for many-body potential energy representation and molecular dynamics. Comput. Phys. Commun. 228, 178–184 (2018).10.1016/j.cpc.2018.03.016 ↗
3. Batzner, S. et al. E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nat. Commun. 13, 2453 (2022).10.1038/s41467-022-29939-5 ↗
4. Guo, Z. et al. Extending the limit of molecular dynamics with ab initio accuracy to 10 billion atoms. In Proc. of the 27th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming 205–218 (2022).10.1145/3503221.3508425 ↗
5. Musaelian, A., Johansson, A., Batzner, S. & Kozinsky, B. Scaling the leading accuracy of deep equivariant models to biomolecular simulations of realistic size. In Proc. International Conference for High Performance Computing, Networking, Storage and Analysis. 1–12 (2023).10.1145/3581784.3627041 ↗
6. Gasteiger, J., Groß, J. & Günnemann, S. Directional message passing for molecular graphs. ICLR, arXiv preprint arXiv:2003.03123 (2020).
7. Chen, C. & Ong, S. P. A universal graph deep learning interatomic potential for the periodic table. Nat. Comput. Sci. 2, 718–728 (2022).10.1038/s43588-022-00349-3 ↗
8. Deng, B. et al. Chgnet as a pretrained universal neural network potential for charge-informed atomistic modelling. Nat. Mach. Intell. 5, 1031–1041 (2023).10.1038/s42256-023-00716-3 ↗
9. Choudhary, K. et al. Unified graph neural network force-field for the periodic table: solid state applications. Digit. Discov. 2, 346–355 (2023).10.1039/D2DD00096B ↗
10. Batatia, I. et al. A foundation model for atomistic materials chemistry. arXiv preprint arXiv:2401.00096. https://arxiv.org/abs/2401.00096 (2023).
11. Takamoto, S. et al. Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements. Nat. Commun. 13, 2991 (2022).10.1038/s41467-022-30687-9 ↗
12. Merchant, A. et al. Scaling deep learning for materials discovery. Nature 1–6 (2023).
13. Zhang, D. et al. Pretraining of attention-based deep learning potential model for molecular simulation. npj Comput. Mater. 10, 94 (2024).10.1038/s41524-024-01278-7 ↗
14. Riebesell, J. et al. Matbench discovery – an evaluation framework for machine learning crystal stability prediction. arXiv preprint arXiv:2308.14920. https://arxiv.org/abs/2308.14920 (2023).
15. Focassio, B., Freitas, L. P. M. & Schleder, G. R. Performance assessment of universal machine learning interatomic potentials: Challenges and directions for materials’ surfaces. ACS Appl. Mater. Interfaces. https://doi.org/10.1021/acsami.4c03815 (2024).10.1021/acsami.4c03815 ↗
16. Yu, H., Giantomassi, M., Materzanini, G., Wang, J. & Rignanese, G.-M. Systematic assessment of various universal machine-learning interatomic potentials. Mater. Genome Eng. Adv. 2, 3 (2024).
17. Jain, A. et al. Commentary: The materials project: A materials genome approach to accelerating materials innovation. APL Mater. 1, 011002 (2013).10.1063/1.4812323 ↗
18. Xie, T. & Grossman, J. C. Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties. Phys. Rev. Lett. 120, 145301 (2018).10.1103/PhysRevLett.120.145301 ↗
19. Luo, S., Chen, T. & Krishnapriyan, A. S. Enabling efficient equivariant operations in the fourier basis via gaunt tensor products. The Twelfth International Conference on Learning Representations. https://openreview.net/forum?id=mhyQXJ6JsK (2024).
20. Cheng, B. Cartesian atomic cluster expansion for machine learning interatomic potentials. npj Comput. Mater. 10, 157 (2024).10.1038/s41524-024-01332-4 ↗
21. Ko, T. W., Finkler, J. A., Goedecker, S. & Behler, J. Accurate fourth-generation machine learning potentials by electrostatic embedding. J. Chem. Theory Comput. 19, 3567–3579 (2023).10.1021/acs.jctc.2c01146 ↗
22. Chanussot, L. et al. Open catalyst 2020 (OC20) dataset and community challenges. ACS Catal. 11, 6059–6072 (2021).10.1021/acscatal.0c04525 ↗
23. Obot, I., Macdonald, D. & Gasem, Z. Density functional theory (DFT) as a powerful tool for designing new organic corrosion inhibitors. part 1: An overview. Corros. Sci. 99, 1–30 (2015).10.1016/j.corsci.2015.01.037 ↗
24. Han, Y. et al. Surface energies, adhesion energies, and exfoliation energies relevant to copper-graphene and copper-graphite systems. Surf. Sci. 685, 48–58 (2019).10.1016/j.susc.2019.01.009 ↗
25. Sun, W., Kitchaev, D. A., Kramer, D. & Ceder, G. Non-equilibrium crystallization pathways of manganese oxides in aqueous solution. Nat. Commun. 10, 573 (2019).10.1038/s41467-019-08494-6 ↗
26. Fichthorn, K. A. & Scheffler, M. Island nucleation in thin-film epitaxy: A first-principles investigation. Phys. Rev. Lett. 84, 5371–5374 (2000).10.1103/PhysRevLett.84.5371 ↗
27. Gurylev, V. & Perng, T. P. Defect engineering of zno: Review on oxygen and zinc vacancies. J. Eur. Ceram. Soc. 41, 4977–4996 (2021).10.1016/j.jeurceramsoc.2021.03.031 ↗
28. Broberg, D. et al. High-throughput calculations of charged point defect properties with semi-local density functional theory—performance benchmarks for materials screening applications. npj Comput. Mater. 9, 72 (2023).10.1038/s41524-023-01015-6 ↗
29. Ahangari, M. G. et al. Effect of various defects on mechanical and electronic properties of zinc-oxide graphene-like structure: A DFT study. Vacuum 165, 26–34 (2019).10.1016/j.vacuum.2019.04.003 ↗
30. Kang, K. & Ceder, G. Factors that affect li mobility in layered lithium transition metal oxides. Phys. Rev. B 74, 094105 (2006).10.1103/PhysRevB.74.094105 ↗
31. van de Walle, A. Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit. Calphad 33, 266–278 (2009).10.1016/j.calphad.2008.12.005 ↗
32. Cation-disordered rocksalt-type high-entropy cathodes for Li-ion batteries. Nat. Mater. 20, 214–221 (2021).10.1038/s41563-020-00816-0 ↗
33. Barroso-Luque, L. et al. Cluster expansions of multicomponent ionic materials: Formalism and methodology. Phys. Rev. B 106, 144202 (2022).10.1103/PhysRevB.106.144202 ↗
34. Zhong, P., Xie, F., Barroso-Luque, L., Huang, L. & Ceder, G. Modeling intercalation chemistry with multiredox reactions by sparse lattice models in disordered rocksalt cathodes. PRX Energy 2, 043005 (2023).10.1103/PRXEnergy.2.043005 ↗
35. Ceder, G. A derivation of the ising model for the computation of phase diagrams. Comput. Mater. Sci. 1, 144–150 (1993).10.1016/0927-0256(93)90005-8 ↗
36. DOMAN, R. C., BARR, J. B., McNALLY, R. N. & ALPER, A. M. Phase equilibria in the system cao—mgo. J. Am. Ceram. Soc. 46, 313–316 (1963).10.1111/j.1151-2916.1963.tb11737.x ↗
37. Jung, I.-H., Decterov, S. A. & Pelton, A. D. Critical thermodynamic evaluation and optimization of the CaO–MgO–SiO2 system. J. Eur. Ceram. Soc. 25, 313–333 (2005).10.1016/j.jeurceramsoc.2004.02.012 ↗
38. Deng, Z., Radhakrishnan, B. & Ong, S. P. Rational composition optimization of the Lithium-Rich Li3OCl1–x Br x anti-perovskite superionic conductors. Chem. Mater. 27, 3749–3755 (2015).10.1021/acs.chemmater.5b00988 ↗
39. Du, P. et al. Cooperative origin of proton pair diffusivity in yttrium substituted barium zirconate. Commun. Phys. 3, 200 (2020).10.1038/s42005-020-00464-5 ↗
40. Rutt, A. et al. Expanding the material search space for multivalent cathodes. ACS Appl. Mater. Interfaces 14, 44367–44376 (2022).10.1021/acsami.2c11733 ↗
41. Shen, J.-X., Horton, M. & Persson, K. A. A charge-density-based general cation insertion algorithm for generating new li-ion cathode materials. npj Comput. Mater. 6, 161 (2020).10.1038/s41524-020-00422-3 ↗
42. Rong, Z., Kitchaev, D., Canepa, P., Huang, W. & Ceder, G. An efficient algorithm for finding the minimum energy path for cation migration in ionic materials. J. Chem. Phys. 145, 074112 (2016).10.1063/1.4960790 ↗
43. Ven, A. V. D., Ceder, G., Asta, M. & Tepesch, P. D. First-principles theory of ionic diffusion with nondilute carriers. Phys. Rev. B 64, 184307 (2001).10.1103/PhysRevB.64.184307 ↗
44. Walle, A. V. D. & Ceder, G. The effect of lattice vibrations on substitutional alloy thermodynamics. Rev. Mod. Phys. 74, 11–45 (2002).10.1103/RevModPhys.74.11 ↗
45. Togo, A. & Tanaka, I. First principles phonon calculations in materials science. Scr. Mater. 108, 1–5 (2015).10.1016/j.scriptamat.2015.07.021 ↗
46. Yue, S. et al. Phonon softening near topological phase transitions. Phys. Rev. B 102, 235428 (2020).10.1103/PhysRevB.102.235428 ↗
47. Parlinski, K., Li, Z. Q. & Kawazoe, Y. First-principles determination of the soft mode in cubic ZrO2. Phys. Rev. Lett. 78, 4063–4066 (1997).10.1103/PhysRevLett.78.4063 ↗
48. Togo, A. First-principles Phonon calculations with Phonopy and Phono3py. J. Phys. Soc. Jpn. 92, 012001 (2023).10.7566/JPSJ.92.012001 ↗
49. Togo, A., Chaput, L., Tadano, T. & Tanaka, I. Implementation strategies in phonopy and phono3py. J. Phys.: Condens. Matter 35, 353001 (2023).
50. Wang, H.-C., Botti, S. & Marques, M. A. L. Predicting stable crystalline compounds using chemical similarity. npj Comput. Mater. 7, 12 (2021).10.1038/s41524-020-00481-6 ↗
51. Pandey, S., Qu, J., Stevanović, V., John, P. S. & Gorai, P. Predicting energy and stability of known and hypothetical crystals using graph neural network. Patterns 2, 100361 (2021).10.1016/j.patter.2021.100361 ↗
52. Bartel, C. J. Data-centric approach to improve machine learning models for inorganic materials. Patterns 2, 100382 (2021).10.1016/j.patter.2021.100382 ↗
53. Belsky, A., Hellenbrandt, M., Karen, V. L. & Luksch, P. New developments in the Inorganic Crystal Structure Database (ICSD): accessibility in support of materials research and design. Acta Crystallogr. Sect. B 58, 364–369 (2002).10.1107/S0108768102006948 ↗
54. Ko, T. W. & Ong, S. P. Recent advances and outstanding challenges for machine learning interatomic potentials. Nat. Comput. Sci. 1–3 (2023).10.1038/s43588-023-00561-9 ↗
55. Frey, N. C. et al. Neural scaling of deep chemical models. Nat. Mach. Intell. 1–9 (2023).10.26434/chemrxiv-2022-3s512 ↗
56. Qi, J., Ko, T. W., Wood, B. C., Pham, T. A. & Ong, S. P. Robust training of machine learning interatomic potentials with dimensionality reduction and stratified sampling. npj Comput. Mater. 10, 43 (2024).10.1038/s41524-024-01227-4 ↗
57. Xu, K. et al. How neural networks extrapolate: From feedforward to graph neural networks. arXiv preprint arXiv:2009.11848 (2020).
58. Barroso-Luque, L. et al. Open materials 2024 (omat24) inorganic materials dataset and models. arXiv preprint arXiv:2410.12771 (2024).
59. Sun, W. & Ceder, G. Efficient creation and convergence of surface slabs. Surf. Sci. 617, 53–59 (2013).10.1016/j.susc.2013.05.016 ↗
60. Ganose, A. et al. atomate2. https://github.com/materialsproject/atomate2 (2024).
61. Henkelman, G., Uberuaga, B. P. & Jónsson, H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys. 113, 9901–9904 (2000).10.1063/1.1329672 ↗
62. Jun, K. et al. Lithium superionic conductors with corner-sharing frameworks. Nat. Mater. 21, 924–931 (2022).10.1038/s41563-022-01222-4 ↗
63. Kingma, D. P. & Ba, J. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014).
64. Kresse, G. & Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6, 15–50 (1996).10.1016/0927-0256(96)00008-0 ↗
65. Kresse, G. & Joubert, D. From ultrasoft pseudopotentials to the projector augmented-wave method. Phys. Rev. B 59, 1758–1775 (1999).10.1103/PhysRevB.59.1758 ↗
66. Ong, S. P. et al. Python materials genomics (pymatgen): A robust, open-source python library for materials analysis. Comput. Mater. Sci. 68, 314–319 (2013).10.1016/j.commatsci.2012.10.028 ↗
67. Deng, B. Wbm high energy states. figshare. Dataset (2024). https://figshare.com/articles/dataset/WBM_high_energy_states/27307776.