Mechanisms of Ethylene Epoxidation over Silver from Machine Learning-Accelerated First-Principles Modeling and Microkinetic Simulations

We employed machine learning-augmented density functionaltheory(DFT) thermodynamic calculations to assess the stability of differentAgO( x ) structures under catalytic ethyleneepoxidation reaction conditions. We found that there are multipleAgO( x ) surface motifs that could co-existunder the relevant conditions. These included Ag surface oxides (e.g.,AgO_p(4 x 4) and Ag1.83O) and atomicoxygen-covered Ag(111) surfaces. Furthermore, we employed DFT calculationsto evaluate the energetics of different reaction mechanisms by whichethylene and oxygen can react on these surfaces. These studies revealedseveral energetically viable reaction pathways for ethylene epoxidation.Microkinetic modeling analyses, based on the DFT-calculated reactionpathways, showed that ethylene epoxidation can proceed on all surfacesand that multiple pathways, including those involving Langmuir-Hinshelwoodand Eley-Rideal mechanisms, could be involved in selectiveand unselective reactions. The diversity of mechanisms that we discoveredin the context of the relatively simple ethylene epoxidation reactionon Ag suggests that the richness and complexity of surface chemistryare most likely a rule rather than an exception in heterogeneous catalyticchemical transformations on metal surfaces and that the concept ofa single or even a dominant mechanism and reaction intermediates mightneed to be revisited for many reactions.