Silver Oxidation and Oxygen Speciation in Ethylene Epoxidation

Supported Ag nanoparticles (NPs) are commercial catalysts used for partial oxidation of ethylene to ethylene oxide (EO). Despite over 60 years of research and industrial optimization, the scientific community is still debating the oxygen species that are responsible for selective and unselective pathways as well as the state of the Ag catalyst. The aim of this work was to apply a unified computational and experimental approach to critical aspects of EO reactions: the structure of the oxidized Ag, the oxygen speciation and its effects on EO formation. We initially report that Ag surface is oxidized and contains different types of surface and subsurface atomic O, as well as adsorbed dioxygen species. Our calculations on silver oxide slab, a model for the oxidized surface, show that dioxygen species, specifically peroxo species, is a major contributor to EO formation. However, the modeled surface was an approximation of a working catalyst. Next, we search for an improved model of the surface that would capture additional catalyst characteristics. We combined ab initio methods and active learning protocols to generate a machine learning force field capable of simulating oxidized Ag and different types of oxygen. We examine the content of subsurface O atoms at which we observe adsorbed dioxygen species, which will be used to estimate O coverages during EO reaction conditions.