Predicting Crystal Structures and Ionic Conductivity in Li$_{3}$YCl$_{6-x}$Br$_{x}$ Halide Solid Electrolytes Using a Fine-Tuned Machine Learning Interatomic Potential

Abstract: This work demonstrates the effectiveness of fine-tuning the CHGNet universal machine learning interatomic potential (uMLIP) to investigate ionic transport mechanisms in ternary halide solid electrolytes of the Li${3}$YCl${6-x}$Br${x}$ family (x = 0 to 6), which are promising candidates for solid-state battery applications. We present a strategy for generating ordered structural models from experimentally derived disordered Li${3}$YCl${6}$ (LYC) and Li${3}$YBr${6}$ (LYB) structures. These serve as initial configurations for an iterative fine-tuning workflow that couples molecular dynamics (MD) simulations with static density functional theory (DFT) calculations. The fine-tuning process and the resulting improvements in predictive accuracy are benchmarked across energy predictions, structure optimizations, and diffusion coefficient calculations. Finally, we analyze the influence of composition (varied x) on the predicted ionic conductivity in Li${3}$YCl${6-x}$Br${x}$, demonstrating the robustness of our approach for modeling transport properties in complex solid electrolytes.