Machine Learning-Guided Modulation of Li+ Solvation Structures towards Optimal Electrolyte Systems for High-Performance Li−O2 Battery

Abstract

As the “blood” of Li−O₂ batteries (LOBs), electrolytes with various solvation structures of Li⁺ can greatly influence the composition of solid electrolyte interphase (SEI) on anode and the growth kinetics of Li₂O₂ on cathode, and further the battery performance. However, achieving delicate modulation of the multi-composition electrolytes remains significantly challenging to simultaneously give consideration to both the anode and cathode reactions. In this work, we employed Bayesian optimization to develop advanced electrolytes for LOBs, enabling the formation of a stable inorganic-rich SEI, and modulation of Li₂O₂ morphologies. Thus obtained LOBs using the optimized dual-solvent electrolyte could deliver a discharge capacity of 14,063 mAh g⁻¹ at a current density of 500 mA g⁻¹, which is far higher than those using the single-solvent electrolytes. This study not only highlights the critical role of the solvation structure for improving the battery performance, but also provides new insights and important theoretical guidance for delicate modulation of electrolyte compositions.