Stable All-Solid-State Lithium Metal Batteries Enabled by Machine Learning Simulation Designed Halide Electrolytes

Solid electrolytes (SEs) with superionic conductivity and interfacial stability are highly desirable for stable all-solid-state Li-metal batteries (ASSLMBs). Here, we employ neural network potential to simulate materials composed of Li, Zr/Hf, and Cl using stochastic surface walking method and identify two potential unique layered halide SEs, named Li2ZrCl6 and Li2HfCl6, for stable ASSLMBs. The predicted halide SEs possess high Li+ conductivity and outstanding compatibility with Li metal anodes. We synthesize these SEs and demonstrate their superior stability against Li metal anodes with a record performance of 4000 h of steady lithium plating/stripping. We further fabricate the prototype stable ASSLMBs using these halide SEs without any interfacial modifications, showing small internal cathode/SE resistance (19.48 Ω cm2), high average Coulombic efficiency (∼99.48%), good rate capability (63 mAh g-1 at 1.5 C), and unprecedented cycling stability (87% capacity retention for 70 cycles at 0.5 C).