(113f) Predictive, Spectroscopy-Guided Modeling of Selectivity in Ethylene Epoxidation over Ag/?-Al2O3 Catalysts

The selectivity of a particular catalyst is dictated by surface composition and structure (i.e., reactive intermediates) which varies as a function of reaction conditions (i.e. temperature and reactant partial pressures). Understanding the complex relationship between selectivity and catalyst surface composition and structure (i.e., structure-selectivity relationships) would pave the way towards predictive theory-guided catalyst design over the conventional trial-and-error approach. We illustrate the development of structure-selectivity relationships by using ethylene epoxidation over industrial-mimic Ag/α-Al₂O₃ catalysts as a probe reaction. Ethylene oxide (EO) is a valuable chemical intermediate with downstream applications in polymers, detergents, and pharmaceuticals¹. Selectivity, the main figure of merit, is typically hindered by total oxidation of ethylene to carbon dioxide (CO₂). Despite the industrial maturity of this reaction, much confusion and uncertainty remain about the selective intermediates involved in this reaction¹.

In this work, we utilized in operando Raman spectroscopy to explore the relationship between surface composition/structure of industrial-mimic Ag/α-Al₂O₃ catalysts and EO selectivity. Our results show that EO selectivity is controlled by a complex mixture of surface oxygen species as well as carbonate species that are formed in situ and that the majority of oxygen species negatively impact EO selectivity. We also performed feature importance calculations to assign relative weights for the contribution of each measured species with selectivity to derive a model that predicts EO selectivity as a function of surface intermediate concentrations. Overall, our work provides a simplified model that can be used to predict EO selectivity as a function of operating temperature, pressure, and promoter concentrations – this work is broadly applicable to any catalyzed reaction of interest.