Deciphering How Anion Clusters Govern Lithium Conduction in Glassy Thiophosphate Electrolytes through Machine Learning

Glasses such as lithium thiophosphates (Li₂S-P₂S₅) show promise as solid electrolytes for batteries, but a poor understanding of how the disordered structure affects lithium transport properties limits the development of glassy electrolytes. To address this, we here simulate glassy Li₂S-P₂S₅ electrolytes with varying fractions of polyatomic anion clusters, i.e., P₂S₆^4-, P₂S₇^4-, and PS₄^3-, using classical molecular dynamics. Based on the determined variation in ionic conductivity, we use a classification-based machine-learning metric termed “softness”─a structural fingerprint that is correlated to the atomic rearrangement probability─to unveil the structural origin of lithium-ion mobility. The softness distribution of lithium ions is highly spatially correlated: that is, the “soft” (high mobility) lithium ions are predominantly found around PS₄^3- units, while the “hard” (low mobility) ions are found around P₂S₆^4- units. We also show that soft lithium-ion migration requires a smaller energy barrier to be overcome relative to that observed for hard lithium-ion migration.