Date of Award

Fall 2021

Document Type

Open Access Dissertation


Chemical Engineering

First Advisor

Andreas Heyden


Propane dehydrogenation is a critical process of producing propylene, an important feedstock for the chemical industrial. There are multiple key processes to produce propylene in this manner, but one of the largest processes involves non-oxidative dehydrogenation on a platinum-tin alloy based catalyst. In general, these reactions and catalytic particles are complex, with many dehydrogenation, cracking, and reforming reactions taking place during these processes on multiple surfaces. With this in mind, ab initio computational catalysis models are used to generate further insight into these catalytic processes. In this dissertation, there are three aims to be solved: identifying the most likely active surface on a catalytic platinum particle, modeling the catalyst particle itself, and understanding how alloying with platinum-tin impacts the selectivities of all surfaces studied.

For the first aim, three surfaces, Pt(100), Pt(111), and Pt(211) were used to model potential catalysts particle sites. Uncertainty quantification and Bayesian inference was applied to the developed models to understand the reported experimental quantities of interest like turnover frequencies, apparent activation energies, selectivities to propylene, and reaction orders. From this, it was found that the most likely active site was the Pt(211) model for certain simulations, and that the first dehydrogenation step of propane was rate limiting. To answer the next aim of this work, four platinum-tin surface skin models were developed to understand how tin doping affects the catalytic reaction. Four models were chosen, Pt3Sn/Pt(100), PtSn/Pt(100), Pt3Sn/Pt(111), and Pt2Sn/Pt(211). Using uncertainty analysis and Bayesian inference, it was found that the most supported model using the evidence of the calibration problem was Pt2Sn/Pt(211), which has strong evidence for this to be the model when compared to the next highest evidence model, Pt3Sn/Pt(111), which is in-line with the pure platinum model. In addition, the 1st dehydrogenation step is modeled to be the rate controlling step, however, on Pt2Sn/Pt(211), often the 2nd dehydrogenation step is rate limiting as well

Available for download on Friday, May 31, 2024