UC Irvine

Optimizing matter via computational chemistry promises speed and chemical insight. However, many contemporary attempts oversimplify reaction mechanisms, use crude descriptions of materials, or utilize computational methods that lack rigor to exhaustively screen many compositions. Often these explorations of materials do not relate the properties of materials to one another or chemical concepts, and do not indicate whether an optimum composition has been discovered. To explore the possibility of quantitatively relating catalytic free energy surfaces (FES) to composition, we have computed (using DFT improved by select DLPNO-CCSD(T) calculations) FESs for two candidate reactions, and material descriptors for modular catalyst components. The electrocatalytic reduction of CO2 to formate was studied using cyclopentadienyl complexes of first-row transition metals. A surprising variety of rate-limiting steps was observed in this family, and predicted H/D KIE values can be used as a rubric to interpret new experiments. pH- and potential-dependent microkinetic models were used to generate turnover frequencies for optimization, and degree-of-rate-control analysis underscores the necessity of including multiple intermediates and transition states in the FES. Principal component analysis was utilized to reveal three degrees of freedom for catalyst optimization and the maximum turnover frequency within the composition space. Methane oxidation was studied using IrIII complexes and ligands that varied in their σ, π, steric properties and charge. Various regression models related these chemical properties to DFT-derived FESs and predicted the FESs of catalysts not yet simulated. Compositions with beneficial traits were added to the training set, refining the model iteratively. The limitations of this approach were better understood and the combination of ligand properties that minimize the C-H cleavage barrier while preventing unwanted oxidation was identified.