UC Berkeley

The use of theoretical chemistry techniques in the investigation of catalytic reactions has been able to provide strong insights into the inner workings of various chemical mechanisms. In tandem with experimental results, such studies often provide information by computing reaction rates with transition state theory, predicting and/or confirming various spectroscopic experiments, and elucidating the identities and structures of key stable intermediates and short-lived transition states. The present work is concerned with the application of these techniques to the study of the oxidative carbonization of toluene to toluic acid over Rh(III) and Pd(II), as well as the development of theoretical techniques to efficiently find ab initio transition states for use in such studies.

Previous work has shown that the oxidative carbonylation of toluene to form toluic acid is possible with Rh(III) and Pd(II) with acetic acid. These reactions are believed to operate via a rate-limiting electrophilic mechanism in which toluene binds to the metal complex and has a C-H bond activated. Previous works have suggested that the active catalyst for the Rh(III) system is Rh(CF3COO)3(CO)2, and Pd(CF3COO)2 for the Pd(II) system, though these were not rigorously confirmed. In this work, we properly identify the Rh(III) species as the active catalyst through a series of ab initio spectroscopic calculations with comparison to experiments. Additionally, an unprecedented interaction between an acetate and carbonyl ligand on the Rh(CF3COO)2(CO)2 catalyst is investigated and found to be the result of an unusual charge balance within the structure. Prior work has shown that using trifluoroacetic acid instead of acetic acid significantly increases the rate of reaction, without investigating further. This work demonstrates that the reaction rate passes through a maximum for intermediate strength acids, which is due to competition between the two sub-steps of the rate-limiting step. Weakly basic anionic ligands increase the positive charge on the metal center and increase the rate of toluene binding while decreasing the ability of the same ligands to accept the activated proton. A similar trend and explanation were found with a model catalyst for the Pd(II) system as the ligands were varied.

The second part of this work concerns the development of efficient transition state searching algorithms. The calculation of theoretical rate constants often employs transition state theory, but requires the user to possess the transition state structure. The local search for such a structure requires an extremely good guess, and is most practically obtained with the help of an automated guess generator. The most commonly used algorithms operate by optimizing a chain of molecular images connecting known reactant and product structures into the reaction pathway. One such routine, the Growing String Method (GSM), grows a chain of states inward from the known endpoints while optimizing these points. The original GSM algorithm relies upon cartesian coordinates with cubic splines for adding new structures to the chain, however this often leads to unrealistic images which require many steps to relax into the reaction pathway. By replacing the cartesian coordinate interpolation with Linear Synchronous Transit interpolation, the computational cost of optimizing complex reaction pathways may be cut approximately in half. Additionally, by simplifying the algorithm to focus computational effort on the location of just the transition state rather than the entire reaction pathway, the overall cost may be reduced even further. In this new method, the Freezing String Method, nodes are iteratively added to a growing chain, optimized for several steps, and then frozen in place for the remainder of the execution.