Integrating Molecular Dynamics and Machine Learning for Solvation‐Guided Electrolyte Optimization in Lithium Metal Batteries

Abstract Optimizing liquid electrolytes is essential for achieving long‐term cycling stability and high safety in lithium metal batteries. However, severe side reactions and lithium dendrite formation during repeated cycling lead to low Coulombic efficiency (CE) and limited lifespan. Herein, a rapid and cost‐effective strategy that integrates high‐throughput molecular dynamics simulations with machine learning predictions is proposed, further validated through experimental evaluation. A mixed electrolyte composed of LiFSI (LiN(SO 2 F) 2 ) as the main salt, DEE (1,2‐diethoxyethane) as the solvent, and LiNFS (LiC 4 F 9 SO 3 ) as an additive achieves a significantly improved CE of 98.32%. Key molecular descriptors are identified for each performance label, and the most accurate model is selected through rigorous benchmarking. The optimal region reveals a preference for medium‐to‐high salt concentrations; low C, O, and N content; and high F content in salts, along with high C and low O content in solvents. This framework enables reusable and resource‐efficient modeling for targeted electrolyte design and accelerated optimization.