Unraveling Anisotropic Li-Ion Transport in Li 3/8 Sr 7/16 Ta 3/4 Hf 1/4 O 3 via Machine Learning Molecular Dynamics and First-Principles Modeling

Abstract

Solid-state Li-ion electrolytes (SSLIEs) are essential components for the development of nextgeneration rechargeable lithium batteries. Among them, the perovskite-type material Li 3/8 Sr 7/16 Ta 3/4 Hf 1/4 O 3 (LSTH) has attracted significant attention owing to its excellent chemical stability, thermal robustness, and high ionic conductivity. However, the atomistic mechanisms underlying Li-ion transport in this material remain inadequately understood. In this work, we present a systematic theoretical investigation of Li-ion diffusion in LSTH using machine learning force field-based molecular dynamics (MLMD) simulations in combination with first-principles density functional theory (DFT) calculations. An accurate on-the-fly machine learning force field (MLFF) was developed and rigorously validated against DFT results. Long-timescale MLMD simulations (2 ns) were conducted over a temperature range of 300-700 K to illustrate Li-ion transport properties. The simulations reveal that anisotropic Liion migration occurs predominantly along the crystallographic z-axis, with an activation energy of 0.28 eV and a room-temperature ionic conductivity of 0.492 mS cm -1 . Complementary firstprinciples nudged elastic band calculations predict a minimum migration barrier of 0.30 eV along the z-direction, in excellent agreement with the MLMD results. Diffusion coefficients obtained from both approaches exhibit strong consistency, demonstrating the accuracy and predictive capability of the MLFF method. These findings provide detailed mechanistic insights into Li-ion transport in LSTH and highlight the effectiveness of MLFF-based simulations as a powerful approach for investigating ion-conducting materials.