Propane Dehydrogenation on Platinum Catalysts: Identifying the Active Sites through Bayesian Analysis

Uncertainty quantification, Bayesian statistics, the reported experimental literature, and density functional theory are synthesized to identify the active sites for the non-oxidative propane dehydrogenation on platinum catalysts. This study tests three different platinum surface models as active sites, Pt(100), Pt(111), and Pt(211), and two different methodologies for generating uncertainty, using data from four density functional theory functionals and data from the BEEF–vdW ensembles. By comparing these three surface facets using two uncertainty sources, a total of six different computational models were evaluated. Three experimental data sets, with varying numbers of reported observables, such as turnover frequencies, selectivity to propylene, apparent activation energy, and reaction orders, are calibrated and validated for these six models. This study finds no evidence for Pt(100) as the dominant active facet and finds that Pt(211) has some evidence for being the most relevant active site on the catalyst. In addition, all four functional models were excluded from final data analysis due to poor “goodness-of-fit”. In contrast, the BEEF–vdW model with ensembles (BMwEs) was found to pass “goodness-of-fit” for most of the models tested. Finally, for both Pt(111) and Pt(211), this study finds that the majority of simulations found the kinetically rate-controlling step the first dehydrogenation step from propane to C3H7*.