Universal-neural-network-potential molecular dynamics for lithium metal and garnet-type solid electrolyte interface

Abstract

Abstract All-solid-state Li-metal batteries can conceivably improve the safety and extend the driving ranges of electric vehicles. In this regard, the garnet-type solid electrolyte Li7La3Zr2O12 (LLZ) has garnered considerable attention because of its high Li-ion conductivity and nonreactivity towards molten Li metal. Here, we perform molecular dynamics (MD) simulations using a universal neural network potential (UNNP) to analyse the Li-ion exchange at the LLZ/Li interface at the atomic scale. The UNNP-MD calculations show that Li ions traverse the LLZ/Li interface and that excess Li ions relative to the stoichiometric composition accumulate in an approximately 1 nm-thick zone near the LLZ phase interface, signifying the formation of a space-charge layer. Electronic structural analysis of the UNNP-MD-derived configuration, performed using density functional theory calculations, reveals band bending near the LLZ phase interface and the simultaneous suppression of Li metal reduction. These findings can help expedite the development of rationally designed all-solid-state Li-metal batteries.