Non-Arrhenius Li-ion transport and grain-size effects in argyrodite solid electrolytes

Abstract: Argyrodite solid electrolytes, such as Li$_6$PS$_5$Cl, exhibit some of the highest known superionic conductivities. Yet, the mechanistic understanding of Li$^+$ transport in realistic argyrodite microstructures -- where atomic-scale mechanisms interplay with continuum-scale dynamics at grain boundaries -- remains limited. Here, we resolve Li$^+$ transport in silico by developing accurate machine-learning potentials via closed-loop active learning and embedding the potentials in a multiscale modeling framework that integrates molecular dynamics with finite element simulations. We show that bulk diffusion barriers scale linearly with anion radius. Grain boundaries have opposite effects depending on the bulk -- enhancing Li$^+$ diffusion in low-diffusivity phases but suppressing it in fast-diffusing ones. Simulations of polycrystalline Li$_6$PS$_5$I reveal non-Arrhenius transport behaviors consistent with experiments. Grain-size-dependent predictions indicate that grain refinement improves intergranular contacts in argyrodites without compromising superionic conductivity, while nanosizing can activate ionic transport in electrolytes lacking intrinsic superionic behavior. Our findings highlight the decisive role of microstructure in developing solid electrolytes.