A molecular level understanding of a reaction mechanism and the computation of rates requires knowledge of the stable structures and the corresponding transition states that connect them. Temperature, pressure, and environment effects must be included to bridge the 'materials gap' so one can reasonably compare ab initio (first-principles, i.e., having no empirical parameters) predictions with experimental measurements. In this thesis, a few critical problems pertaining to ab initio modeling of catalytic systems are addressed; namely, 1) the issue of building representative models of isolated metal atoms grafted on amorphous supports, 2) modeling inorganic catalytic reactions in non-ideal solutions where the solvent participates in the reaction mechanism, and 3) bridging the materials gap using ab initio thermodynamics to predict the stability of supported nanoparticles under experimental reaction conditions.

In Chapter I, a background on first-principles modeling of heterogeneous and homogenous catalysts is provided. Subsequently, to address the problem of modeling catalysis by isolated metal atoms on amorphous supports, we present in Chapter II a sequential-quadratic programming algorithm that systematically predicts the structure and reactivity of isolated active sites on insulating amorphous supports. Modeling solution phase reactions is also a considerable challenge for first-principles modeling, yet when done correctly it can yield critical kinetic and mechanistic insight that can guide experimental investigations. In Chapter III, we examine the formation of peroxorhenium complexes by activation of H2O2, which is key in selective oxidation reactions catalyzed by CH3ReO3 (methyltrioxorhenium, MTO). New experiments and density functional theory (DFT) calculations were conducted to better understand the activation of H2O2 by MTO and to provide a strong experimental foundation for benchmarking computational studies involving MTO and its derivatives. It was found that water present in aqueous solvents plays a critical rate-enhancing role in accelerating the formation of the active sites. In Chapter IV we study olefin epoxidation catalyzed by MTO, and the strong accelerating effects in the presence of H2O. DFT calculations and experiments both support that the primary origin of the acceleration of catalytic epoxidation arises from the water-dependence of the rates of generation of the peroxorhenium complexes.

Modeling catalyst surfaces and active sites under realistic conditions is also a long-standing computational challenge due to the dynamic nature of catalytic processes. For example, catalyst surfaces can restructure under reaction conditions, nanoparticles can undergo Ostwald ripening or even disintegrate into adatom-reactant complexes, and active-site poisoning from reaction intermediates or side-products could occur. In Chapter V we focus on understanding the stability of supported nanoparticles under reaction conditions against disintegration into adatom-reactant complexes, as reactant-induced disintegration can lead to catalyst deactivation or be exploited to redisperse sintered catalysts. More specifically, to better understand the stability of TiO2(110) supported three-way catalysts rhodium, palladium, and platinum nanoparticles during NOx and CO reduction, we conducted an ab initio thermodynamics study of the feasibility for these noble metal nanoparticles to disintegrate across a large parameter space of operation temperatures, pressures of CO or NO gas, and nanoparticle sizes. Lastly, in Chapter VI, the status of DFT modeling of catalysts is briefly summarized and some remaining challenges are discussed. (Abstract shortened by UMI.).