(323f) Bayesian Analysis for Identifying the Active Sites for Platinum Catalyzed Propane Dehydrogenation: Bridging Experiments, Density Functional Theory, and Reactor Modeling.

Microkinetic studies based on parameters obtained from density functional theory can elucidate reaction mechanisms and help identify potential active sites for different catalysts. Catalysts are complex, with multiple surfaces present on nanoparticles. Through uncertainty quantification and Bayesian analysis, catalyst models can be discriminated against each other and directly compared to experimental data, elucidating the active site on catalyst particles.

In this study, the non-oxidative propane dehydrogenation (PDH) to propylene was investigated over a series of three surface models, Pt(100), Pt (111), and Pt(211), using density functional theory calculations and microkinetic modeling. We performed calculations using an ensemble of four functionals to generate mean and variances for each state's energy present in the microkinetic model. In addition, we developed a second model using the ensembles generated by the Bayesian Error Estimation Functional with van der Waals interactions (BEEF-vdW). When propagating the uncertainty in the energy computed by density functional theory, it was found that there was a wide distribution of kinetic parameters. Using uncertainty analysis, we calibrated our model outputs, such as turnover frequency, reaction orders in propane and hydrogen, apparent activation energies, and selectivity to propylene, on three different experimental studies. We checked for model fit using a goodness-of-fit test after calibrating and validating each surface model to published experimental data. After rejecting all models that failed the “goodness-of-fit” test, we then evaluated the evidence for each surface against each other using Bayes Factor and Jeffreys’ Scale. Dependent on the experimental simulation and functional system used, there was evidence that Pt(111) and Pt(211) could be the active site. In addition, a particle model was developed using all three surfaces, and it was found that propylene was further consumed by some of the surfaces, leading to deep dehydrogenation and propylene cracking.