Solid electrolytes for Li-ion batteries via machine learning

Abstract

In this work, machine learning (ML) techniques were employed to construct a predictive model that can be used to discover new solid state electrolytes (SE/SSE) for lithium ion batteries (LIBs). The model was built (with R2 = 0.97) based on a dataset constructed from previous works regarding ionic conductivity (IC) of solid electrolytes. After a suitable validation process, the ML-model was used to predict the IC of many compositions (∼30 K in Inorganic Crystal Structure Database (ICSD)). Interestingly, the predictions of this model, done on 145 compounds, were consistent with values of Li-phonon band center, which is used as an IC descriptor, this was then used to predict the IC vs temperature behavior of LiYS2 which is suggested as a promising SSE candidate in this work.