Toward the Design of Catalysts

A chemical kinetic model, although an essential ingredient of the understanding of catalytic behavior, does not contain sufficient knowledge to predict improved catalysts.  The inherent chemistry of families of catalysts governs relationships between the various rate and equilibrium constants and thus provides the key to the optimal catalyst formulation.  We call our strategy at Purdue for identifying and utilizing this information Discovery Informatics.  This approach to catalyst design envisions building a quantitative forward model that links descriptors of the catalyst chemistry, through a microkinetic model, to catalyst performance.  While such a model has intrinsic value, that value will be powerfully leveraged if the model can be used to predict performance of new materials chosen in a guided stochastic search of the descriptor space.  In principle, comparison of the measured performance to that predicted for the “fittest” materials chosen in that inverse search then provides direction for model improvement. Iteration of the prediction, measurement, correction cycle would produce evolution toward an optimal catalyst.  In this process, the forward model becomes the repository for catalytic knowledge extracted from the experiments.

Progress in developing this approach will be illustrated with two examples: homogeneous single site olefin polymerization catalysis and heterogeneous water gas shift (WGS) catalysis.  In the first case, aryloxide and Cp(Cp^*) ligated Ti and Zr catalysts present well defined catalytic sites and a team of researchers has combined DFT computation with kinetic experiments to produce a detailed model that correlates the 1-hexene propagation rate with ion pair separation energies and the cone angles available for monomer docking.  Extensions to other systems have included a population balance kinetic model that can use molecular weight distribution data to fit additional kinetic parameters.  Heterogeneous catalyst systems add to the complexity of the design problem because of the many potential catalytic sites often available.  Recent literature has shown remarkable success using scaling laws and Brønsted, Evans, Polyani relations to identify optimal catalysts for reactions that are relatively insensitive to catalyst structure. Our work with supported Pt, Pd and Au water gas shift catalysts shows effects of support, alloying and particle size that on the one hand show the critical importance of good kinetic data in narrowing the focus for theoretical analysis and illustrate the need for “scalable” theories to account for structural details.