Modeling Catalyzed Reactions on Metal-Doped Amorphous Silicates: The Case of Niobium-Catalyzed Ethylene Epoxidation

Metal-doped amorphous silicates are promising materials for heterogeneous catalysis because they are easy to make and their properties can be tuned for specific reactions. However, the amorphous nature of the surface makes the characterization difficult, and improve-ments are based on a trial and error approach. Density functional theory simulations can in principle provide direct structure-property relations, but the sampling of the active site configuration space is not possible via brute force. In this contribution, we use the Nb-catalyzed epoxidation of ethylene as a test reaction to analyze various aspects of the modeling that need to be taken into account for simulations of effective reaction rates. We show that each site can host a variety of transition state structures that represent the same reaction event, but that can differ considerably in reaction barrier. Furthermore, many different sites need to be sampled. We then use machine learning to identify the most important descriptors of the bare active site that correlate directly with the energy barrier. Although our test set is too small for quantitative predictions of reaction rates, we discuss what the important features of a very active site are that can drive the kinetics in the real material.

Automated summary

Metal-doped amorphous silicates are promising materials for heterogeneous catalysis due to their easy fabrication and tunable properties. However, their amorphous surface makes characterization challenging, and improvements rely on a trial and error approach. In this study, we use the Nb-catalyzed epoxidation of ethylene as a test reaction to analyze various aspects of simulation modeling for effective reaction rate predictions. We find that each site can have multiple transition state structures with varying reaction barriers, and that sampling of multiple sites is necessary. Machine learning is employed to identify key descriptors of the active site that correlate with energy barriers.