Insights into Ethylene Deep Oxidation on Pd–Au Bimetallic Catalysts: From Surface Dynamics to Pathway Selectivity

Deep oxidation of ethylene represented the primary side reaction during vinyl acetate synthesis over the Pd–Au catalysts. Because of the complexity of the deep oxidation process and the uncertain surface structures of bimetallic catalysts, the process was not well understood. We developed a robust Pd–Au–C–H-O machine-learning force field (MLFF) by employing a pretraining and fine-tuning strategy. Deep-potential molecular dynamics (DPMD) simulations were then used to elucidate the dynamic surface reconstruction of Pd–Au nanoparticles (NPs) under operando conditions and the heterogeneous reaction pathways on Pd–Au bimetallic surfaces. By integration of density functional theory (DFT) calculations with our MLFF framework, the accessible time and length scales of conventional atomistic simulations were well extended. Our results demonstrated that dynamic reconstruction of Pd–Au NPs enriched the Pd content on the surface, forming contiguous Pd ensembles that served as active sites for ethylene deep oxidation. Surface Pd concentration dictated the selectivity between dehydrogenation and oxidation pathways: high Pd coverage favored dehydrogenation to low H/C intermediates and carbon deposition, whereas Au moderated the Pd activity by balancing dehydrogenation and oxidation. This complex problem involving the coupling of adsorbents, active surfaces, and catalytic reactions can potentially be solved through these multiscale simulation strategies. This work provides mechanistic insights into the deep oxidation of ethylene and offers guidelines for the rational design of bimetallic catalysts. Moreover, it highlights the broad potential of MLFF-assisted molecular dynamics to advance studies in heterogeneous catalysis and reveal dynamic adsorbate-catalyst coupling mechanisms.