Machine Learning-Accelerated Molecular Dynamics of Lithium-Ion Transport in Cubic LLZO

Abstract Solid-state lithium batteries (SSLBs) are gaining prominence due to their potential to offer enhanced safety and higher energy densities compared to conventional liquid electrolyte-based systems. This study focuses on lithium-stuffed garnet-type oxide, Li₇La₃Zr₂O₁₂ (LLZO), a promising solid electrolyte candidate known for its high ionic conductivity and electrochemical stability. Employing machine-learning-accelerated molecular dynamics (MLMD) simulations using the orb-models framework integrated with LAMMPS, we investigate lithium-ion transport properties, structural stability, and mechanical behavior of cubic-phase LLZO. Key material descriptors such as diffusion coefficients, radial distribution functions, non-Gaussianity parameters, elastic moduli, and lithium site occupancies were calculated. The results confirm LLZO’s potential as a fast-ion conductor. Furthermore, the mechanical analysis suggests metastability of the cubic phase, aligning with experimental observations. This computational approach underscores the utility of machine learning potentials in the predictive design of next-generation solid-state battery materials.