(569dv) Coverage and Facet Dependent Multiscale Modeling of O* and H* Adsorption on Pt and Ptw Catalytic Nanoparticles

Deviations in catalytic insights obtained via computational and experimental approaches for heterogeneous systems are caused by various factors. Three of the factors include: (1) low-coverage studies that permeate computational studies do not capture adsorbate-adsorbate interactions that exist at higher coverage levels where experiments happen, (2) single-facet studies yield limited results that do not cover the workings of real-life multifaceted catalysts, (3) computational studies do not consider adsorbate-induced changes in the catalyst while evaluating catalyst performance. We seek to understand and ultimately bridge the gap that results from these disparities.

Using uncomplicated adsorbates (oxygen and hydrogen) important to industrial reactions (oxidations, reductions, hydrodeoxygenation, etc.), we perform multiscale modeling—density functional theory (DFT), ab initio phase diagrams, and mean-field microkinetic models—in conjunction with molecular dynamics with machine-learned interatomic potentials to achieve our goals. Pt is renowned for hydrogenation reactions and PtW shows high activity and selectivity towards hydrodeoxygenation. We rigorously investigated the effects of O* and H* coverage on multiple facets of Pt and PtW nanoparticles. Mean-field models spanning wide coverage ranges showed that the adsorbate-adsorbate interactions are repulsive and significantly greater for O*/Pt than for H*/Pt (Figure 1). These repulsive interactions for O* on Pt(111) are both 2- and 3-body, whereas on Pt(100) and Pt(110), they are primarily 3-body. Furthermore, experimentally observed coverages and desorption temperatures were reliably estimated by computational models when the adsorbate-adsorbate interactions are included. Omitting adsorbate-adsorbate interactions over multi-faceted Pt nanoparticles overestimates the equilibrium coverages of the adsorbates across a wide range of temperatures and pressures relevant to heterogeneous catalysis. Finally, we explored promoter concentrations and placements with machine learning to obtain optimal combinations of these factors and how adsorbate effects result in catalyst reconstructions. Altogether, our work demonstrates that including coverage and facet effects improves the accuracy of computational models for heterogeneous catalysts.