"Explainable Machine Learning Models for Predicting Oxygen Activation Energies in Perovskites and Pyrochlores"
Solid oxide fuel cells (SOFCs) are an emerging technology used to convert fuel into electricity with up to 80% theoretical efficiency. To improve the kinetics of both oxygen transport through the solid electrolyte and the chemical reactions, SOFCs operate at high temperatures (700-1000C) at the cost of reduced durability. To reduce the necessary temperatures, these electrolytes must be replaced with new oxides with faster oxygen diffusion. Explainable machine learning methods can be used to provide physical insights, enabling rapid screening of new materials. In this work, we fit six separate models for the oxygen diffusion of two classes of oxides: perovskites and pyrochlores and compare their feature importances with each other to obtain physical insight into how these crystal structures and properties interact.
