Atomistic modeling of bulk and grain boundary diffusion in solid electrolyte Li 6 PS 5 Cl using machine-learning interatomic potentials

Harnessing the potential of solid-state electrolytes is crucial for advancing all-solid-state lithium-ion batteries, yet the role of grain boundaries (GBs) in ionic conductivity remains poorly understood. The authors present an active learning approach to develop machine-learning interatomic potentials with high accuracy, enabling large-scale, long-term simulations of complex GBs in polycrystalline Li${}₆$PS${}₅$Cl. Intercage diffusion of Li ions near GBs, triggered by the ``cage-opening effect'' of GBs, is notably observed. These findings offer key insights into optimizing ionic transport in solid electrolytes, paving the way for improved performance in next-generation battery technologies.