Unraveling Lithium Dynamics in Solid Electrolyte Interphase: From Graph Contrastive Learning to Transport Pathways

Abstract Fast lithium transport across the solid-state electrolyte (SSE)/lithium metal anode interface is critical for high-performance all-solid-state batteries. Uncovering the complex lithium dynamics governed by diverse local environments in the solid electrolyte interphase (SEI) is fundamental for performance optimization. A general framework for characterizing the distinct local environments remains lacking, which largely prevents the unlocking of mechanisms governing complex lithium dynamics in the SEI. Here, we develop Graph-based Estimation of ionic Transport in SEI (GET-SEI), a general framework that discovers local atomic environments without predefined labels through graph contrastive learning (GCL). Based on this scheme, lithium transition kinetics can be modeled and quantified via extended dynamic mode decomposition and transition path theory. Applied to various well-known SSE/Li systems, including sulfides (Li6PS5Cl/Li, Li10GeP2S12/Li) and oxides (Li7La3Zr2O12/Li), GET-SEI identifies dominant transport pathways and kinetic bottlenecks in each system, providing quantitative metrics for evaluating lithium transport efficiency. As novel high-performance SSEs continue to emerge, GET-SEI offers a widely applicable, interpretable tool for targeted SEI engineering.