Efficient catalysts are imperative to accelerate the slow oxygen reaction kinetics for the development of emerging electrochemical energy systems ranging from room-temperature alkaline water electrolysis to high-temperature ceramic fuel cells. In this work, we reveal the role of cationic inductive interactions in predetermining the oxygen vacancy concentrations of 235 cobalt-based and 200 iron-based perovskite catalysts at different temperatures, and this trend can be well predicted from machine learning techniques based on the cationic lattice environment, requiring no heavy computational and experimental inputs. Our results further show that the catalytic activity of the perovskites is strongly correlated with their oxygen vacancy concentration and operating temperatures. We then provide a machine learning-guided route for developing oxygen electrocatalysts suitable for operation at different temperatures with time efficiency and good prediction accuracy.
Read full abstract