Skip to main content
Computing, Environment and Life Sciences

Using Machine Learning Force Fields to Model Battery Structure

Argonne is developing a physics-informed machine learning approach to discover compatible cathodes, thereby overcoming barriers to commercialization of solid-state lithium-ion batteries.

Volume contraction of cathodes during delithiation is hampering the development and commercialization of solid-state lithium-ion batteries The contraction causes cracking and delamination between the cathode particles and the solid-state electrolyte. Modeling these sophisticated structures has required the development of machine learned force fields (MLFFs) capable of simulating the volume contraction with orders-of-magnitude lower cost than traditional methods.

MLFFs, combined with active machine learning strategies, can simulate larger volumes of material than could ever be done before with high accuracy. In this project, we leverage the MLFFs to find the lowest energy configurations for various dopants and then to estimate the volume change with delithiation. The discovery of an appropriate dopant and its optimal composition could enable next-generation batteries and revolutionize vehicle electrification and other applications.