Data-Driven Universal Model for Predicting Thermodynamics and Kinetics for Na-Ion Battery

The emergence and recent development of artificial intelligence has made it possible to develop foundational models for predicting the global chemical space of materials for Na-ion batteries. In this work, we aim to demonstrate the creation of a foundational machine learning model capable of predicting the phase stability of several classic Na-ion battery systems, including NASICON, Na metal oxides, and Prussian blue analogs. This model will be based on a comprehensive dataset covering typical species and stoichiometries across electrodes and electrolytes in Na-ion batteries. In addition to the thermodynamic predictor, we will also showcase the integration of kinetic Monte Carlo simulations to capture the kinetic growth of these materials in aqueous solution.