Identifying electronic descriptors to predict work functions of perovskite electrodes

The ability to predict work functions of semiconductor electrodes is critical to the development of photocatalytic and electrocatalytic systems. Understanding the compositional and structural dependence of the interfacial electronic structure of semiconductors could enable us to screen and select appropriate functional materials for electronic applications. We address this problem with a focus on perovskite oxides (ABO₃) in the metastable cubic phase by enumerating the possible compositions for A- and B-site cations. We analyze the computed work functions by means of statistical learning using elemental and bulk chemical descriptors. The resulting descriptor-based model not only delivers accurate predictions of computed work functions using a limited number of features, but also allows us to interpret the complex correlations between electronic levels and surface structure across a vast chemical space.