Hierarchical high-throughput screening of alkaline-stable lithium-ion conductors combining machine learning and first-principles calculations

The advancement of solid-state batteries depends on the development of lithium-ion conductors that exhibit both high ionic conductivity and stability across a wide range of electrochemical and chemical conditions. In this paper, we investigate the chemical factors that control the stability of Li-NASICONs and garnets in highly alkaline aqueous environment. While this is of general importance, it is particularly important for the operation of Li-air cells with humidified air. Humid air promotes the formation of LiOH as the discharge product, creating a highly alkaline environment on the surface of cathode and solid-state electrolyte. In this work, we combine machine learning and first-principles calculations to conduct a high-throughput computational screening of alkaline-stable oxide-based Li-ion conductors in order to better characterize the tradeoff between the various relevant properties. We evaluate the material stability in terms of pH, voltage, and species present in the environment (LiOH and H2O) across a vast range of chemical compositions with NASICON and garnet crystal structures. We utilize the CHGNet universal machine learning interatomic potential for pre-screening, followed by DFT calculations. Such a hierarchical screening procedure enables the evaluation of over 320,000 chemical compositions, encompassing nearly the entire periodic table. From this set 209 alkaline-stable NASICON and garnet compounds are selected as final candidates. We identify the specific cation substitutions that improve alkaline stability in NASICON and garnet compounds, and reveal the underlying mechanism. We also discover the trade-offs for designing alkaline-stable Li-ion conductors, highlighting the need to carefully optimize compositions so that it can simultaneously enhance all the material properties required for practical battery applications.