Systematic softening in universal machine learning interatomic potentials

Deng, Bowen; Choi, Yunyeong; Zhong, Peichen; Riebesell, Janosh; Anand, Shashwat; Li, Zhuohan; Jun, KyuJung; Persson, Kristin A.; Ceder, Gerbrand

doi:10.1038/s41524-024-01500-6

Systematic softening in universal machine learning interatomic potentials

Journal Article · Fri Jan 10 00:00:00 EST 2025 · npj Computational Materials

DOI:https://doi.org/10.1038/s41524-024-01500-6· OSTI ID:2528028

^[1]; ^[1]; ^[1]; ^[2]; ^[3]; ^[3]; ^[1]; ^[1]; ^[1]

Univ. of California, Berkeley, CA (United States); Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)
Univ. of Cambridge (United Kingdom). Cavendish Lab.
Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States)

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.

View Accepted Manuscript (DOE)

Research Organization:: Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA (United States). National Energy Research Scientific Computing Center (NERSC); National Renewable Energy Laboratory (NREL), Golden, CO (United States)

Sponsoring Organization:: National Science Foundation (NSF); USDOE Office of Science (SC), Basic Energy Sciences (BES)

Grant/Contract Number:: AC02-05CH11231

OSTI ID:: 2528028

Journal Information:: npj Computational Materials, Journal Name: npj Computational Materials Journal Issue: 1 Vol. 11; ISSN 2057-3960

Publisher:: Nature Publishing GroupCopyright Statement

Country of Publication:: United States

Language:: English

References (53)

A derivation of the Ising model for the computation of phase diagrams Ceder, G. Computational Materials Science, Vol. 1, Issue 2 https://doi.org/10.1016/0927-0256(93)90005-8	journal	March 1993
Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set Kresse, G.; Furthmüller, J. Computational Materials Science, Vol. 6, Issue 1, p. 15-50 https://doi.org/10.1016/0927-0256(96)00008-0	journal	July 1996
Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit van de Walle, Axel Calphad, Vol. 33, Issue 2 https://doi.org/10.1016/j.calphad.2008.12.005	journal	June 2009
Python Materials Genomics (pymatgen): A robust, open-source python library for materials analysis Ong, Shyue Ping; Richards, William Davidson; Jain, Anubhav Computational Materials Science, Vol. 68 https://doi.org/10.1016/j.commatsci.2012.10.028	journal	February 2013
Density functional theory (DFT) as a powerful tool for designing new organic corrosion inhibitors. Part 1: An overview Obot, I. B.; Macdonald, D. D.; Gasem, Z. M. Corrosion Science, Vol. 99 https://doi.org/10.1016/j.corsci.2015.01.037	journal	October 2015
DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics Wang, Han; Zhang, Linfeng; Han, Jiequn Computer Physics Communications, Vol. 228 https://doi.org/10.1016/j.cpc.2018.03.016	journal	July 2018
Critical thermodynamic evaluation and optimization of the CaO–MgO–SiO2 system Jung, In-Ho; Decterov, Sergei A.; Pelton, Arthur D. Journal of the European Ceramic Society, Vol. 25, Issue 4 https://doi.org/10.1016/j.jeurceramsoc.2004.02.012	journal	April 2005
Defect engineering of ZnO: Review on oxygen and zinc vacancies Gurylev, Vitaly; Perng, Tsong Pyng Journal of the European Ceramic Society, Vol. 41, Issue 10 https://doi.org/10.1016/j.jeurceramsoc.2021.03.031	journal	August 2021
Predicting energy and stability of known and hypothetical crystals using graph neural network Pandey, Shubham; Qu, Jiaxing; Stevanović, Vladan Patterns, Vol. 2, Issue 11 https://doi.org/10.1016/j.patter.2021.100361	journal	November 2021
Data-centric approach to improve machine learning models for inorganic materials Bartel, Christopher J. Patterns, Vol. 2, Issue 11 https://doi.org/10.1016/j.patter.2021.100382	journal	November 2021
First principles phonon calculations in materials science Togo, Atsushi; Tanaka, Isao Scripta Materialia, Vol. 108 https://doi.org/10.1016/j.scriptamat.2015.07.021	journal	November 2015
Efficient creation and convergence of surface slabs Sun, Wenhao; Ceder, Gerbrand Surface Science, Vol. 617 https://doi.org/10.1016/j.susc.2013.05.016	journal	November 2013
Surface energies, adhesion energies, and exfoliation energies relevant to copper-graphene and copper-graphite systems Han, Yong; Lai, King C.; Lii-Rosales, Ann Surface Science, Vol. 685 https://doi.org/10.1016/j.susc.2019.01.009	journal	July 2019
Effect of various defects on mechanical and electronic properties of zinc-oxide graphene-like structure: A DFT study Ahangari, M. Ghorbanzadeh; Mashhadzadeh, A. Hamed; Fathalian, M. Vacuum, Vol. 165 https://doi.org/10.1016/j.vacuum.2019.04.003	journal	July 2019
Rational Composition Optimization of the Lithium-Rich Li ₃ OCl _{1– x} Br _x Anti-Perovskite Superionic Conductors Deng, Zhi; Radhakrishnan, Balachandran; Ong, Shyue Ping Chemistry of Materials, Vol. 27, Issue 10 https://doi.org/10.1021/acs.chemmater.5b00988	journal	May 2015
Accurate Fourth-Generation Machine Learning Potentials by Electrostatic Embedding Ko, Tsz Wai; Finkler, Jonas A.; Goedecker, Stefan Journal of Chemical Theory and Computation, Vol. 19, Issue 12 https://doi.org/10.1021/acs.jctc.2c01146	journal	June 2023
Expanding the Material Search Space for Multivalent Cathodes Rutt, Ann; Shen, Jimmy-Xuan; Horton, Matthew ACS Applied Materials & Interfaces, Vol. 14, Issue 39 https://doi.org/10.1021/acsami.2c11733	journal	September 2022
Performance Assessment of Universal Machine Learning Interatomic Potentials: Challenges and Directions for Materials’ Surfaces Focassio, Bruno; M. Freitas, Luis Paulo; Schleder, Gabriel R. ACS Applied Materials & Interfaces, Vol. 17, Issue 9 https://doi.org/10.1021/acsami.4c03815	journal	July 2024
Open Catalyst 2020 (OC20) Dataset and Community Challenges Chanussot, Lowik; Das, Abhishek; Goyal, Siddharth ACS Catalysis, Vol. 11, Issue 10 https://doi.org/10.1021/acscatal.0c04525	journal	May 2021
Non-equilibrium crystallization pathways of manganese oxides in aqueous solution Sun, Wenhao; Kitchaev, Daniil A.; Kramer, Denis Nature Communications, Vol. 10, Issue 1 https://doi.org/10.1038/s41467-019-08494-6	journal	February 2019
E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials Batzner, Simon; Musaelian, Albert; Sun, Lixin Nature Communications, Vol. 13, Issue 1 https://doi.org/10.1038/s41467-022-29939-5	journal	May 2022
Towards universal neural network potential for material discovery applicable to arbitrary combination of 45 elements Takamoto, So; Shinagawa, Chikashi; Motoki, Daisuke Nature Communications, Vol. 13, Issue 1 https://doi.org/10.1038/s41467-022-30687-9	journal	May 2022
A charge-density-based general cation insertion algorithm for generating new Li-ion cathode materials Shen, Jimmy-Xuan; Horton, Matthew; Persson, Kristin A. npj Computational Materials, Vol. 6, Issue 1 https://doi.org/10.1038/s41524-020-00422-3	journal	October 2020
Predicting stable crystalline compounds using chemical similarity Wang, Hai-Chen; Botti, Silvana; Marques, Miguel A. L. npj Computational Materials, Vol. 7, Issue 1 https://doi.org/10.1038/s41524-020-00481-6	journal	January 2021
High-throughput calculations of charged point defect properties with semi-local density functional theory—performance benchmarks for materials screening applications Broberg, Danny; Bystrom, Kyle; Srivastava, Shivani npj Computational Materials, Vol. 9, Issue 1 https://doi.org/10.1038/s41524-023-01015-6	journal	May 2023
Robust training of machine learning interatomic potentials with dimensionality reduction and stratified sampling Qi, Ji; Ko, Tsz Wai; Wood, Brandon C. npj Computational Materials, Vol. 10, Issue 1 https://doi.org/10.1038/s41524-024-01227-4	journal	February 2024
Pretraining of attention-based deep learning potential model for molecular simulation Zhang, Duo; Bi, Hangrui; Dai, Fu-Zhi npj Computational Materials, Vol. 10, Issue 1 https://doi.org/10.1038/s41524-024-01278-7	journal	May 2024
Cartesian atomic cluster expansion for machine learning interatomic potentials Cheng, Bingqing npj Computational Materials, Vol. 10, Issue 1 https://doi.org/10.1038/s41524-024-01332-4	journal	July 2024
Cation-disordered rocksalt-type high-entropy cathodes for Li-ion batteries Lun, Zhengyan; Ouyang, Bin; Kwon, Deok-Hwang Nature Materials, Vol. 20, Issue 2 https://doi.org/10.1038/s41563-020-00816-0	journal	October 2020
Lithium superionic conductors with corner-sharing frameworks Jun, KyuJung; Sun, Yingzhi; Xiao, Yihan Nature Materials, Vol. 21, Issue 8 https://doi.org/10.1038/s41563-022-01222-4	journal	March 2022
Scientific discovery in the age of artificial intelligence Wang, Hanchen; Fu, Tianfan; Du, Yuanqi Nature, Vol. 620, Issue 7972 https://doi.org/10.1038/s41586-023-06221-2	journal	August 2023
Cooperative origin of proton pair diffusivity in yttrium substituted barium zirconate Du, Peng; Chen, Qianli; Fan, Zhijun Communications Physics, Vol. 3, Issue 1 https://doi.org/10.1038/s42005-020-00464-5	journal	November 2020
CHGNet as a pretrained universal neural network potential for charge-informed atomistic modelling Deng, Bowen; Zhong, Peichen; Jun, KyuJung Nature Machine Intelligence, Vol. 5, Issue 9 https://doi.org/10.1038/s42256-023-00716-3	journal	September 2023
A universal graph deep learning interatomic potential for the periodic table Chen, Chi; Ong, Shyue Ping Nature Computational Science, Vol. 2, Issue 11 https://doi.org/10.1038/s43588-022-00349-3	journal	November 2022
Recent advances and outstanding challenges for machine learning interatomic potentials Ko, Tsz Wai; Ong, Shyue Ping Nature Computational Science, Vol. 3, Issue 12 https://doi.org/10.1038/s43588-023-00561-9	journal	December 2023
Unified graph neural network force-field for the periodic table: solid state applications Choudhary, Kamal; DeCost, Brian; Major, Lily Digital Discovery, Vol. 2, Issue 2 https://doi.org/10.1039/D2DD00096B	journal	January 2023
A climbing image nudged elastic band method for finding saddle points and minimum energy paths Henkelman, Graeme; Uberuaga, Blas P.; Jónsson, Hannes The Journal of Chemical Physics, Vol. 113, Issue 22, p. 9901-9904 https://doi.org/10.1063/1.1329672	journal	December 2000
Commentary: The Materials Project: A materials genome approach to accelerating materials innovation Jain, Anubhav; Ong, Shyue Ping; Hautier, Geoffroy APL Materials, Vol. 1, Issue 1 https://doi.org/10.1063/1.4812323	journal	July 2013
An efficient algorithm for finding the minimum energy path for cation migration in ionic materials Rong, Ziqin; Kitchaev, Daniil; Canepa, Pieremanuele The Journal of Chemical Physics, Vol. 145, Issue 7 https://doi.org/10.1063/1.4960790	journal	August 2016
Modeling Intercalation Chemistry with Multiredox Reactions by Sparse Lattice Models in Disordered Rocksalt Cathodes Zhong, Peichen; Xie, Fengyu; Barroso-Luque, Luis PRX Energy, Vol. 2, Issue 4 https://doi.org/10.1103/PRXEnergy.2.043005	journal	October 2023
Phonon softening near topological phase transitions Yue, Shengying; Deng, Bowen; Liu, Yanming Physical Review B, Vol. 102, Issue 23 https://doi.org/10.1103/PhysRevB.102.235428	journal	December 2020
Cluster expansions of multicomponent ionic materials: Formalism and methodology Barroso-Luque, Luis; Zhong, Peichen; Yang, Julia H. Physical Review B, Vol. 106, Issue 14 https://doi.org/10.1103/PhysRevB.106.144202	journal	October 2022
From ultrasoft pseudopotentials to the projector augmented-wave method Kresse, G.; Joubert, D. Physical Review B, Vol. 59, Issue 3, p. 1758-1775 https://doi.org/10.1103/PhysRevB.59.1758	journal	January 1999
First-principles theory of ionic diffusion with nondilute carriers Van der Ven, A.; Ceder, G.; Asta, M. Physical Review B, Vol. 64, Issue 18 https://doi.org/10.1103/PhysRevB.64.184307	journal	October 2001
Factors that affect Li mobility in layered lithium transition metal oxides Kang, Kisuk; Ceder, Gerbrand Physical Review B, Vol. 74, Issue 9 https://doi.org/10.1103/PhysRevB.74.094105	journal	September 2006
Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties Xie, Tian; Grossman, Jeffrey C. Physical Review Letters, Vol. 120, Issue 14 https://doi.org/10.1103/PhysRevLett.120.145301	journal	April 2018
First-Principles Determination of the Soft Mode in Cubic ZrO 2 Parlinski, K.; Li, Z. Q.; Kawazoe, Y. Physical Review Letters, Vol. 78, Issue 21 https://doi.org/10.1103/PhysRevLett.78.4063	journal	May 1997
Island Nucleation in Thin-Film Epitaxy: A First-Principles Investigation Fichthorn, Kristen A.; Scheffler, Matthias Physical Review Letters, Vol. 84, Issue 23 https://doi.org/10.1103/PhysRevLett.84.5371	journal	June 2000
The effect of lattice vibrations on substitutional alloy thermodynamics van de Walle, A.; Ceder, G. Reviews of Modern Physics, Vol. 74, Issue 1 https://doi.org/10.1103/RevModPhys.74.11	journal	January 2002
New developments in the Inorganic Crystal Structure Database (ICSD): accessibility in support of materials research and design Belsky, Alec; Hellenbrandt, Mariette; Karen, Vicky Lynn Acta Crystallographica Section B Structural Science, Vol. 58, Issue 3 https://doi.org/10.1107/S0108768102006948	journal	May 2002
Phase Equilibria in the System CaO-MgO Doman, R. C.; Barr, J. B.; McNALLY, R. N. Journal of the American Ceramic Society, Vol. 46, Issue 7 https://doi.org/10.1111/j.1151-2916.1963.tb11737.x	journal	July 1963
Extending the limit of molecular dynamics with ab initio accuracy to 10 billion atoms Guo, Zhuoqiang; Lu, Denghui; Yan, Yujin Proceedings of the 27th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming https://doi.org/10.1145/3503221.3508425	conference	March 2022
First-principles Phonon Calculations with Phonopy and Phono3py Togo, Atsushi Journal of the Physical Society of Japan, Vol. 92, Issue 1 https://doi.org/10.7566/JPSJ.92.012001	journal	January 2023

Similar Records

Cross-functional transferability in foundation machine learning interatomic potentials

Journal Article · Tue Dec 31 19:00:00 EST 2024 · npj Computational Materials · OSTI ID:3014169

Robust training of machine learning interatomic potentials with dimensionality reduction and stratified sampling

Journal Article · Sun Feb 25 19:00:00 EST 2024 · npj Computational Materials · OSTI ID:2311933

Modeling the Behavior of Concentrated Aqueous HNO3 Using Machine Learning Interatomic Potentials

Journal Article · Tue Jan 06 19:00:00 EST 2026 · Journal of Chemical Physics · OSTI ID:3013819

Related Subjects

36 MATERIALS SCIENCE

Systematic softening in universal machine learning interatomic potentials

Citation Formats

References (53)

Similar Records

Related Subjects