| Abstract Scope |
Understanding Li-ion transport through the solid electrolyte interphase (SEI) is essential for designing stable, high-performance lithium-metal batteries. In this work, density functional theory (DFT) nudged elastic band calculations were employed to systematically investigate Li-ion diffusion in eight major inorganic SEI components, including LiF, LiCl, LiBr, LiI, Li₂O, Li₂S, Li₃N, and Li₂CO₃. Both bulk and grain-boundary transport pathways were evaluated across homogeneous and heterogeneous interfaces to establish a unified atomistic diffusion dataset. To accelerate exploration of complex SEI structures, a graph-learning framework combining a path-aware graph variational autoencoder and graph neural network was developed to predict migration barriers from local structural environments. The framework achieved test-set Rē values exceeding 0.93 for both grain and grain-boundary diffusion. Analysis of the learned representations reveals how composition, interfacial structure, and grain-boundary energetics jointly govern Li-ion transport, enabling computational discovery and design of SEI materials with improved ionic conductivity and interfacial stability. |