Accelerated Autonomous Workflow For Antiperovskite-based Solid State Electrolytes

Abstract We developed and implemented an autonomous multi-target multi-fidelity workflow to explore the chemical space of antiperovskite materials with general formula X3AB (X = Li, Na, Mg), searching for stable high performance solid state electrolytes (SSEs) for all-solid state batteries. The workflow is based on the calculation of thermodynamic and kinetic properties, which include phase and electrochemical stability, semiconducting behavior, and ionic diffusivity. To accelerate the calculation of the kinetic properties, we implement a surrogate model able to predict the transition state structures during ionic diffusion, thus reducing the expensive calculation cost by more than one order of magnitude, keeping the error within 73 meV compared to the more accurate methods. 14 antiperovskites have been identified as possible SSE candidates. Moreover, this approach is general and chemistry neutral, so can be applied to other battery chemistries and crystal prototypes.