Graphical abstract

Machine Learning Interatomic Potentials for Modeling Solid-State Batteries

M. K. Phuthi, G. Wei, B. Li, S. Majumdar, V. S. C. Kolluru, N. Kumar, S. Blau, P. Canepa, M. K. Y. Chan, R. Gómez-Bombarelli, K. Persson, G. Ceder, and S. P. Ong; Chem. Mater. 38, 9623–9640 (2026).

BibTeX 
@article{Phuthi2026machine,
  author  = {Phuthi, M. K. and Wei, G. and Li, B. and Majumdar, S. and Kolluru, V. S. C. and Kumar, N. and Blau, S. and Canepa, P. and Chan, M. K. Y. and Gómez-Bombarelli, R. and Persson, K. and Ceder, G. and Ong, S. P.},
  title   = {{Machine Learning Interatomic Potentials for Modeling Solid-State Batteries}},
  journal = {Chemistry of Materials},
  volume  = {38},
  pages   = {9623--9640},
  year    = {2026},
  doi     = {10.1021/acs.chemmater.6c01051}
}
PDF
phuthi-chem-mater-2026.pdf
DOI
10.1021/acs.chemmater.6c01051
Topics
Batteries Solid-state batteries Electrochemistry Method development

Brief summary

Machine-learning interatomic potentials simulate materials at close to quantum-mechanical accuracy at a tiny fraction of the cost. This perspective examines what they bring to battery research, especially solid-state batteries: their trade-offs, the properties they can predict, their use for interfaces, reactions and materials discovery, and what the next generation needs.

Full abstract

Atomistic simulations can provide critical insights into the fundamental behavior of battery materials. A necessary input for such simulations is a description of the potential energy surface (PES). Over the past decade, machine learning interatomic potentials (MLIPs) have emerged as a powerful approach to model the PES at near-first-principles accuracy with orders of magnitude lower computational cost. In this perspective, we examine the role of atomistic simulation in battery research, especially in the context of the capabilities afforded by MLIPs. We outline their advantages and trade-offs relative to ab initio methods and discuss how they enable the prediction of key intrinsic properties relevant to battery materials, especially for solid state batteries. We further explore their potential for modeling interfaces and chemical reactivity, as well as their integration into high-throughput screening and materials discovery workflows. Finally, we highlight promising applications and identify requirements for the next generation of MLIPs to advance the study of battery materials.