Hosted by: Ping Liu

Atomic simulation is able to solve more chemical problems than ever before due to advances in high performance computing and data science tools. Two case studies are presented to illustrate this point. One case study is the water-gas shift reaction catalyzed by platinum clusters deposited on metal oxide (Pt/MO2). The water-gas shift reaction is the most widely applied reaction in industry for the production of hydrogen which is an environmentally-friendly fuel. Furthermore, water-gas shift may be conducted on domestically produced shale gas, alleviating foreign dependence on oil. Three hypotheses regarding the active site for the water–gas shift reaction are the interface edge of Pt/MO2 catalysts, the interface corner, and the platinum terrace (pure platinum). The hypotheses are tested against experimental kinetic data. Uncertainties associated with density functional theory (DFT) calculations and model errors of microkinetic models of the active sites are informed and verified using Bayesian inference and predictive validation on the experiments. Our results suggest the metal oxide support is essential for the activity of the catalyst for water-gas shift, which aids the rational design of catalysts for the water-gas shift reaction. In the second case study, large data sets of organic reactions are investigated using machine learning tools. Reactions which are not labeled with a solvent are automatically labeled, which impacts automated chemical process design and may be utilized as an every-day tool by chemists. The machine learning solvent selections are tested against an expert chemist. In a broad sense, the methods developed within this work improve the reliability of atomic simulations and the incorporation of machine learning to catalysis research.