Learning techniques for designing solid-state lithium-ion batteries with high thermomechanical stability

This work is an implication of machine learning (ML) techniques to predict thermomechanically compatible cathode and electrolyte materials for lithium-ion batteries (LIBs). The ML models were trained on a dataset of 5,578 materials, using the coefficient of thermal expansion (CTE) as the target property. The optimized extra trees model was first evaluated using 10-fold-cross-validation and then the model was further validated using an experimental dataset. Additionally, density function theory (DFT) calculations were performed on MgBe₁₃ and MgPd₂ in which the electron localization function (ELF) plots show an agreement between the predicted CTE and the overall type of bonding supported by the melting temperature being the most important feature in the model. After validation, the model was used to predict the CTE for 25 K materials. Based on the predictions, cathode and electrolyte materials were screened and analyzed for thermomechanical compatibility using CTE.