Active Site of Catalytic Ethene Epoxidation: Machine-Learning Global Pathway Sampling Rules Out the Metal Sites

TLDR: In this paper, a machine-learning-based reaction exploration was used to reveal the true active site of an Ag-based catalyst in ethene oxidation on both Ag(100) and Ag(111) metal surfaces.

Abstract: Ethene epoxidation on Ag-based catalysts, an important heterogeneous catalytic reaction with large-scale global wide production, has generated continuous debate on the active site of epoxidation over the past 60 years. The controversy is not only on the roles of the phase transition from Ag metal to oxidation but also on the necessity of the minority crystal facets of the Ag metal, i.e., Ag(100). Herein, we identify, via a machine-learning reaction exploration, that ethene oxidation on the Ag metal surfaces has three low-energy pathways and the most important one, the dehydrogenation of oxometallacycle intermediate (OMC-DH), is entirely overlooked previously. By computing the free energy profile and performing microkinetics simulation, we show that irrespective of the reaction conditions the dehydrogenation path is always dominant for ethene oxidation on both Ag(100) and Ag(111) metal surfaces (>90%), which rationalizes the low selectivity to combustion products (CO2 and H2O) in low oxygen pressure experiments and rules out the chance of Ag metal phases being the active site of ethene epoxidation under industrial conditions (high O2 pressures). The universal presence of the OMC-DH pathway and the general low selectivity on metal sites are then confirmed by evaluating this mechanism on different catalysts. Our results highlight the power of machine-learning-based reaction exploration for resolving the complex reaction network and also point the direction to reveal the true active site of Ag-based catalyst in ethene oxidation.