Insights into computational methods for surface science and catalysis
Abstract
The fundamental understanding of both the reactions at catalytic surfaces and the ways in which these surfaces change throughout a catalytic cycle and lifetime are important for both academic and industrial disciplines. To develop these understandings on complex catalytic systems, ultra-high vacuum techniques such as molecular beam studies, temperature programmed desorption, reflection-absorption infrared spectroscopy and Auger electron spectroscopy can be used to study the simplest interactions between gas molecules and surfaces. These interactions can be studied from a bottom-up approach to learn about the system in question, upon which additional complexities can be added. To parallel these experimental techniques, a number of computational methods can be used to support findings and guide new experiments. Ab-initio electronic structure calculations allow for a better understanding of adsorbate-surface interactions, while long timescale dynamic simulations provide insight into the time evolution and kinetics of catalysts and catalytic surfaces. Empirical and machine-learning guided potentials can be developed to lessen computational cost while retaining accuracies comparable to ab-initio calculations. Fitting such potentials ultimately allows for larger calculations to be performed and longer timescales to be simulated. The above methods will be applied to a number of industrially and academically relevant catalytic systems, including studying the interaction of H₂ and CO with Cobalt based Fischer-Tropsch catalysts and the interaction between hydrogen and palladium surfaces. Additionally, the development of a machine learning package to fit and use interatomic potentials will be discussed.