Lithium superionic conductors are of major importance as solid electrolytes for next-generation all-solid-state lithium-ion batteries. While ab initio molecular dynamics simulations have been extensively applied to study these materials, the high computational expense have limited their application to high temperatures and short time scales. In this talk, I will discuss how machine learning interatomic potentials (ML-IAPs) offer a way to trascend these limitations while retaining near-ab initio accuracy. Trained using data generated with appropriate DFT functionals, we show that ML-IAPs can accurately reproduce the potential energy surface and key properties of complex lithium superionic conductors such as Li0.33La0.56TiO3, Li3YCl6 and Li7P3S11 while enabling linear scaling at orders of magnitude lower computational cost. Molecular dynamics using these ML-IAPs can shed new insights in transitions in diffusion mechanisms at relatively low temperatures as well as reactions occuring at the interface between electrodes and lithium superionic conductors.

Bridging length and time scales in lithium superionic conductors with machine learning

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.