Lithium Ion Conduction in Cathode Coating Materials from On-the-Fly Machine Learning

Cathode coatings have the potential to significantly improve the performance of solid-state lithium-ion batteries, but computationally identifying materials with suitable ionic conductivity can be challenging. We demonstrate how this problem can be resolved using on-the-fly machine learning, and we validate our approach against experimental data. Based on a screen for high electrochemical stability, low interfacial reactivity and viable lithium ion conduction, we suggest two promising coating materials Li₃Sc₂(PO₄)₃ and Li₃B₇O₁₂. Li₃B₇O₁₂ to our knowledge has not been studied as an ionic conductor and is predicted to have high interfacial stability for a variety of leading electrolyte/cathode material combinations. The workflow we present accelerates the discovery of fast-to-intermediate ionic conductors and can readily be generalized to other types of materials.