Density functional theory-free descriptor for the practical discovery of perovskite catalysts

Abstract

Descriptors play vital roles in aiding the understanding of complicated catalytic processes and guiding the design of novel catalysts. Conventional descriptors, such as adsorption energy and d-band center, are based on domain knowledge, i.e., the Sabatier principle, with typical volcano scaling. Quantitative determination of those descriptors requires either experimental measurements or density functional theory (DFT) calculations, inhibiting the practical and efficient design of new catalysts. Recently, domain-knowledge-based descriptors and prevalent volcano scaling have been challenged via a symbolic regression (SR) approach, deriving a DFT-free descriptor composed of easily accessible material parameters. This review briefly summarizes the development of descriptors in the field of catalysis and introduces our proposal to construct a DFT-free descriptor to accelerate catalyst design. This descriptor will then guide experiments to successfully synthesize new oxide perovskites, outperforming the current state-of-the-art. The DFT-free descriptor and SR approach bridge the gap between computational materials scientists and experimentalists, exemplifying how computational materials science can accelerate materials discovery in experiments.