UC Berkeley

The selective oxidation and ammoxidation of light olefins comprises a 5 million ton per year industry, and is responsible for making possible products from nitrile rubber to Plexiglas to acrylic paint. The industrial catalyst of choice for such reactions is based on bismuth molybdate, and was first patented in the 1950s. In the intervening decades, a significant body of research has been done on bismuth molybdate-based catalysts, and yet a surprising amount is still not known about how these catalysts work. This Thesis has focused primarily on developing new methods for studying bismuth molybdates and related catalysts in order to gain new insight into the means by which the physical and electronic structures of the active sites in these catalysts give rise to their catalytic activity.

The mechanism by which propene is oxidized on the (010) surface of Bi2Mo3O12 has been investigated using the RPBE+U variety of density functional theory (DFT). The location of the active site was determined, and the calculated barrier for the rate-determining step at this site found to be in good agreement with experimental results. Calculations revealed the essential roles of bismuth and molybdenum in providing the geometric and electronic structure responsible for catalytic activity at the active site, and suggested that catalytic activity could be further enhanced by substitution with a more reducible element.

In order to accurately model substitution of an additional reducible element in to Bi2Mo3O12 using DFT, more sophisticated approaches than RPBE+U were required. Two more advanced density functionals, M06-L and HSE, were examined. The HSE functional was found to be too expensive for practical use on extended systems like bismuth molybdate catalysts. The accuracy of the M06-L functional for lattice constants and geometries, reaction energies and barriers, electronic structures, and non-covalent interactions was investigated, and compared results from the RPBE+U method. The M06-L functional was found to be superior to RPBE+U for lattice constants, reaction energies, and non-covalent interactions, and as good as or better than RPBE+U for electronic structures. Use of the M06-L functional was therefore determined to be preferable to use of RPBE+U for use in the study of substituted bismuth molybdate catalysts.

Calculations employing the M06-L functional were combined with physical characterization using diffuse reflectance UV-VIS, x-ray photoelectron, and x-ray near edge absorption spectroscopies in order to understand the effect of substitution of vanadium for molybdenum on the activation energy for propene oxidation in catalysts of formula Bi1-x/3V1-xMoxO4. In these catalysts, substitution of vanadium for molybdenum has been observed to lower the apparent activation barrier for propene oxidation. It was found that the lower activation barrier for propene oxidation over mixed vanadate-molybdate catalysts is a consequence of the smaller difference between the catalyst conduction band edge energy and the energy level of the highest occupied molecular orbital in propene. The lower conduction band edge energy in mixed vanadate-molybdate catalysts is related to the energies of and degrees of mixing between the V 3d and Mo 4d orbitals comprising the conduction band. Both of these observations suggest general principles that may be of relevance to a variety of mixed metal oxide catalyst systems.

An improved procedure for synthesizing bismuth molybdate and bismuth vanadate catalysts was also developed. This procedure involved a two step templating process: a structured mesoporous carbon was templated from KIT-6 or MCM-48 mesoporous silica, and the structured mesoporous carbon in turn used as a template during synthesis of the metal oxide catalyst. Catalysts produced by the double templating process had surface areas of 14-17 m²/g, a large improvement on the < 1 m²/g surface area produced by conventional synthesis techniques.