Machine Learning-Guided Ab Initio Study of Na Super Ionic Conductor-Type Na4Hf2(SiO4)3 Solid-State Electrolyte for Sodium-Ion Batteries

The development of solid-state electrolytes (SSEs), which are essential for enabling safe and effective sodium-ion batteries, is being accelerated by machine learning (ML). This study evaluated the potential of a NASICON-type Na4Hf2(SiO4)3compound as a promising SSE by using ML techniques in conjunction with density functional theory (DFT). A predictive machine learning model was developed based on ionic conductivity data from the literature reported experimentally. The model was trained on log-transformed conductivity and temperature, and features were selected using statistical correlation techniques. SHAP analysis was used to identify key characteristics influencing ionic transport. Because materials databases lack specific filters for NASICON-type compounds, hypothetical structures were developed for model validation. Na4Hf2(SiO4)3was chosen due to its high expected conductivity and structural stability. DFT validation included assessments of electronic band gap and dynamical, thermal, and mechanical stability in addition to ionic conductivity, which was computed using molecular dynamics simulations. This study demonstrates the usefulness of a hybrid ML-DFT strategy in accelerating the search for solid-state electrolytes.