Data-driven models for ground and excited states for Single Atoms on Ceria

Abstract Ceria-based single-atom catalysts present complex electronic structures due to the dynamic electron transfer between metal atoms and semiconductor oxide support. Understanding these materials implies retrieving all states in ensembles, which can be limiting if done via density functional theory. Here, we propose a data-driven approach obtain parsimonious model identifying appearance of charge for single (SAs). We first constructed database (701) configurations group 9–11 metals on CeO 2 (100). Feature Selection based predictive Elastic Net Random Forest models highlights eight fundamental variables: atomic number, ionization potential, size, coordination, metal–oxygen bond strengths, surface strain, Coulomb interactions. With variables Bayesian algorithm yields an expression adsorption energies SAs ground low-lying excited states. Our work paves way towards understanding structure complexity metal/oxide interfaces.