Improving electrolyte performance for target cathode loading using an interpretable data-driven approach

Higher loading of active electrode materials is desired in batteries, especially those based on conversion reactions, for enhanced energy density and cost efficiency. However, increasing cathode loading in electrodes can cause performance depreciation, which can be alleviated by a compatible electrolyte design. In this work, a data-driven approach based on a graph-based deep learning model is adopted to screen high-performing electrolytes for a multi-electron redox-mediated Li-ICl battery and extract design rules custom to different cathode loadings. The model is trained with an experimental dataset of electrolyte formulations and battery capacity, with the inclusion of additional cell-level variables like cathode loadings. The approach brings about an additional 20% increment in the specific capacity of the battery over capacities obtained from the experimental optimization. The study resulted in an electrolyte with a high specific capacity of 250 mAh/g (at 1 mA/cm2) and excellent rate capability at the targeted 45 wt % cathode loading.