Deciphering Coulombic Efficiency of Lithium Metal Anodes by Screening Electrolyte Properties.

Coulombic efficiency (CE) is a quantifiable indicator for the reversibility of lithium metal anodes in high-energy-density batteries. However, the quantitative relationship between CE and electrolyte properties has yet to be established, impeding rational electrolyte design. Herein, an interpretable model for estimating CE based on data-driven insights of electrolyte properties is proposed. Hydrogenbond acceptor basicity (β) and the energy level gap between the lowest unoccupied and the highest occupied molecular orbital (HOMO-LUMO gap) of solvents are identified as the top two parameters impacting CE by machine learning. β and HOMO-LUMO gap of solvents govern anode interphase chemistry. A regression model is further proposed to estimate the CE based on β and HOMOLUMO gap. Using the new solvent screened by above regression model, the Li metal anode in the pouch cell with an energy density of 418 Wh kg-1 achieves the highest CE of 99.2%, which is much larger than previous CE ranging from 70-98.5%. This work provides a reliable interpretable quantitative model for rational electrolyte design.