Fig. 4: Schematic of the data-driven construction of a closed-form model for the interaction of single metal atoms with the CeO2(100) support, obtained by the Bayesian Machine Scientist (BMS).
From: Data-driven models for ground and excited states for Single Atoms on Ceria
a Representative features selected via sequential feature selection (SFS) from predictive ML models. b Physical Born–Haber model previously used to describe the metal-support interaction of a Pt1/CeO2(100) SAC at fixed coordination and two charge states (energy terms: Red = Reduction; Cov = Covalent; Dist = Distortion; Coul = Coulomb; Ads = Adsorption). c Energy contributions given by each individual term of \({{{{\rm{E}}}}}_{{{{\rm{BMS}}}}}^{{{{\rm{M}}}}}\). The range of energy contribution due to the support, \({{{{\rm{E}}}}}_{{{{\rm{BMS}}}}}^{{{{\rm{Ox}}}}}\), is indicated as gray background. Relevant mOS values for term (iii) are annotated.
