Discovering virtual Na-based argyrodites as solid-state electrolytes using DFT, AIMD, and machine learning techniques

Abstract

In surveying an extensive library of 4375 hypothetical Na-based argyrodites, we underscore the value of computational screening, noting that no Na-based argyrodite solid-state electrolyte has been successfully synthesized. We introduce a robust approach using density functional theory (DFT) calculations to identify thermodynamically and electrochemically stable candidates. By evaluating energy above the hull (E_h), formation energy (E_f), band gap (E_g), and electrochemical stability window (V_w), we narrow the set to 15 compounds via a 4-dimensional Pareto sorting. Competing materials for E_h and V_w calculations are sourced from the Materials Project, ICSD, and Google DeepMind. Connectivity-optimized graph networks validate the reliability of our calculations. Ab initio molecular dynamics (AIMD) calculations assess the room-temperature sodium ion conductivity (σ_RT) of the 15 selected entries, ultimately identifying the top 5 with promising σ_RT. This discovery of multi-compositional virtual argyrodites advances the challenge of synthesizing Na-based argyrodites.