Dynamic Monte Carlo methods have become effective tools for studying processes that fall between the time and length scales investigated by molecular dynamics and continuum mechanics. The molecular-level resolution of the technique affords a detailed analysis of complex interactions such as adatom diffusion and catalytic reactions. An efficient Monte Carlo method capable of modeling arbitrarily complex reaction mechanisms has been developed. The current implementation maintains detailed information about the current state of the catalytic surface and only updates the information on a local scale, greatly increasing the efficiency of the algorithm. This allows for facile investigation of large, complex mechanisms with minimal effort and competitive solution times. The implementation is first discussed, and then its application to several model systems is shown, including comparisons to explicit integration solutions.

The combined insight gained from laboratory and computational experiments has historically provided detailed information about the complex interactions present in many heterogeneous catalytic systems. Although this information has been effectively used to interpret and explain experimental observations, it unfortunately has not, in general, been incorporated into kinetic models of the reacting systems. In particular, catalytic nonuniformity is one such phenomenon which is often ignored when kinetic models are constructed despite there being much experimental and computational evidence indicating its presence. Catalytic nonuniformity has been therefore incorporated into both microkinetic models and dynamic Monte Carlo simulations of catalytic reacting systems, and the performance and applicability of each solution method has been investigated for several model systems exhibiting different causes of nonuniform behavior.

In recent years, the use of mechanistic modeling to identify the underlying kinetics of complex systems has increased greatly. One of the challenges to kinetic modeling is the construction of a model which can capture the essential chemistry of a system while a manageable size is retained. The rate-based generation of mechanistic models is an attractive approach because kinetically significant species can be determined and selectively included in the final mechanism. An algorithm for the rate-based generation of reaction mechanisms developed previously(1) was improved and used to construct a compact mechanistic model for low-pressure tetradecane pyrolysis. Though thousands of species and reactions were generated, only a small portion of these (2% of species and 20% of reactions) was deemed necessary and incorporated into the final model. Experimental data were used to determine frequency factors for a subset of the reaction families, while all other kinetic parameters were set on the basis of literature values. With no adjustment to the optimized frequency factors, the mechanistic model was able to accurately predict reactant conversions and product yields for varying reaction conditions and initial reactant loadings. It was also observed that increasing the quantity of species initially seeded resulted in a smaller mechanism that had comparable fitting and predicting abilities as those of the models seeding only the reactant. Subsequent regeneration of the reaction mechanisms using the optimized values for the frequency factors resulted in smaller models with comparable capabilities.

Methylidyne is a key intermediate in hydrocarbon growth reactions over transition metal surfaces. However, experimental data characterizing its interaction with various surfaces are scarce. Therefore, to deepen our understanding of the chemisorption of methylidyne and to quantify its role in the catalytic formation of hydrocarbons, we have calculated the binding energy of methylidyne on Ni(111) and Co(0001) surfaces using density-functional theory within the generalized gradient approximation and the full-potential linear augmented planewave (FP-LAPW) method. The dependence of the binding energy on both the adsorption site and the surface coverage was significant on both surfaces. The binding energy of CH adsorbed in the fcc threefold hollow site of Ni(111) decreased by 1.1eV as the surface coverage increased from 0.25 to 1.0ML, whereas the increase of 0.74eV observed over Co(0001) for the hcp threefold hollow site was slightly less pronounced. The density-of-state plots revealed that rehybridization of the carbon atom interacting with the surface occurred, and the electronic changes induced upon chemisorption were similar for both surfaces.

Recently, the synthesis of ammonia over ruthenium-based catalysts has become an industrially viable process. Unfortunately, investigations of ammonia synthesis over ruthenium are scarce, particularly in comparison to the number of studies carried out over iron. To begin to fill this void, we have performed a series of electronic density-functional theory (DFT) calculations to investigate the effect of particle size and surface structure on ammonia synthesis over ruthenium. Our study has focused on the dissociative adsorption of dinitrogen, which is thought to be the rate-determining step in the synthesis, on both single-crystal surfaces and spherical clusters of ruthenium. The equilibrium adsorbate geometries were remarkably similar on both the single-crystal surfaces and the spherical clusters studied. The binding energy of dinitrogen in the end-on state exhibited a strong dependence on ruthenium surface atom coordination, being much stronger on atoms with low coordination. The main difference between the two single-crystal surfaces studied was the ability of the open Ru(112̄0) face to stabilize a low-energy side-on dinitrogen state, while the close-packed Ru(0001) face could not. It is likely that this stable side-on state provides a low-energy dissociation pathway.

Novel modifications were made to the core components of the algorithms for rate-based generation of reaction mechanisms', including introducing thermodynamic constraints into the estimation of the controlling rate parameters and an alternative approach for determining the species included in the final mechanism. Once implemented, the adapted rate-based building criterion was successfully employed to construct a compact mechanistic model for low-pressure tetradecane pyrolysis. Though thousands of species and reactions were generated, only a small portion of these were deemed necessary and incorporated into the final model. Experimental data were used to determine frequency factors for a subset of the reaction families, while all other kinetic parameters were set based on the literature. The final optimized values for the frequency factors were consistent with literature, and the model was able to accurately fit experimental data from different reaction conditions. With no adjustment to the optimized frequency factors, the mechanistic model for tetradecane pyrolysis was able to accurately predict reactant conversions and product yields for varying reaction conditions. Both relative trends and the actual values were predicted correctly over a wide range of reactant conversions and initial reactant loadings.

Nonuniformity, the variation in heats of adsorption and kinetic parameters with changes in coverage or adsorption site, is present in many heterogeneous catalytic reaction systems. Unfortunately, there have been few investigations addressing the incorporation of nonuniformity into kinetic models of catalytic systems. The main reason for this disparity is the difficulty encountered in establishing a quantitative assessment of the effect of nonuniformity on reaction kinetics. Indeed, many systems in which nonuniformity is known to be present are well described by uniform kinetic models. To begin to get a better understanding of the effect nonuniformity has on reaction kinetics, both model and real systems in which nonuniformity is known to be present have been studied using microkinetic modeling. The model system was a simple condensation mechanism involving only one surface reaction. Two types of nonuniformity, adsorbate-adsorbate interactions and biographic heterogeneity, were incorporated into the kinetic parameters of the model, and two sets of nonuniform data were generated. Uniform models were fit to transient nonuniform data, and their accuracy was assessed. The uniform models were then used to predict steady-state reaction rates. While the uniform models performed fairly well in fitting the transient nonuniform data, they overpredicted steady-state reaction rates by roughly a factor of 20 regardless of the type of nonuniformity used to generate the data. The insights gained from the model system were used to model the reaction of carbon monoxide and hydrogen to methane and water (methanation). The nonuniform methanation model included adsorbate-adsorbate repulsive interactions and sites created by the adsorption of carbon monoxide upon which only hydrogen could adsorb. The nonuniform model did an excellent job of fitting and predicting transient methanation data. While it could not predict batch reactor methanation rates within an order of magnitude, it did capture the trends in rate with changing temperature and pressure which had eluded other microkinetic models of methanation reported in the literature.

Ruthenium has long been known to be an effective catalyst for ammonia synthesis. However, compared to the traditional iron-based catalysts, studies on ruthenium-based catalysts are limited. The rate determining step of ammonia synthesis, the dissociative adsorption of dinitrogen, has been shown to be extremely structure sensitive on both iron and ruthenium catalysts. To study this structure sensitivity on ruthenium, density functional theory calculations were performed on Ru(001) and Ru(110) clusters. End-on, side-on, and dissociated adsorption stales were investigated on both surfaces. While the Ru(110) cluster could stabilize all three adsorption modes, a minimum energy structure for the side-on adsorption on Ru(001) could not be found. It is likely that this side-on mode can provide a low energy pathway to the dissociated state, thereby resulting in faster dissociative adsorption on Ru(110).