UC Berkeley

Zeolites are crystalline microporous solids composed of corner-sharing, tetrahedrally-coordinated silicate (SiO4) units. The isomorphic substitution of a framework Si atom by an Al atom can be charge compensated by a proton, introducing Brønsted-acidic bridging-hydroxy groups, or by a metal cation, such as Na+ or Ni2+. Proton-exchanged zeolites are solid acid catalysts that are used in a large number of processes, such as hydrocarbon cracking, isomerization, and alkylation, important in the conversion of petroleum to transportation fuel. Metal-exchange zeolites are commonly used in gas adsorption, catalysis, or as ion-exchange materials. It is of great interest to predict the impact of zeolite structure and composition on the activity and product selectivity of a given reaction to screen new possible catalysts. Quantum chemical calculations can provide molecular-scale information on zeolite-adsorbate interactions, as well as model the energetic changes and dynamics of important reactions that occur within the channels and pores of zeolite catalysts. However, the application of quantum chemical calculations for the study of chemical reactions occurring in zeolites is particularly difficult because of the large number of zeolite framework atoms needed to accurately capture the long-range Coulombic and medium-range van der Waals interactions. This Thesis has focused primarily on the benchmarking and application of an electrostatically embedded combined quantum mechanics/molecular mechanics (QM/MM) methodology to the study of zeolite-catalyzed reactions.

The adsorption enthalpy of light hydrocarbon molecules in both acidic and neutral zeolite MFI has been investigated with a range of computational methods. The role of cluster model size and density functional theory methodology is examined by comparison with high quality ab initio wave function theory Møller–Plesset 2nd order perturbation theory (MP2) results and experimentally determined heats of adsorption. A hybrid QM/MM method is required to converge calculated thermochemical properties with respect to cluster model size in a manner that is computationally efficient. The accuracy of both QM and QM/MM methods is highly sensitive to choice of level of theory and cluster size. Large basis sets, large cluster models, and a density functional capable of capturing intermolecular interactions are required to achieve a chemical accuracy of 2 kcal/mol with respect to experimentally determined adsorption enthalpies and activation barriers. The computational effort for performing QM/MM simulations is considerably lower than that of similar quality QM results, and allows for the chemically accurate simulation of the energetics of chemical reactions occurring in zeolites in a manner that is computationally cost efficient.

The QM/MM method has also been utilized to study the mechanism by which Ni-exchanged Na-X zeolite is activated by propene and the mechanism for the catalytic oligomerization of propene by the activated Ni-complex in Ni-Na-X. The Ni2+ cations in as-prepared Ni-Na-X are shown to locate preferentially in inaccessible positions within the framework of Na-X, and the migration to a catalytically active position is facilitated by propene. The key activation barrier in the proposed mechanism of propene-facilitated Ni activation corresponds to the formation of a Ni-allyl complex, and is driven by enthalpic stabilization of the Ni2+ cation, which can coordinate better with propene than with the rigid Na-X lattice. The resting state of the Ni-Na-X catalyst is a propene ligated Ni-propyl organometallic complex located in the supercage of Na-X. The rate-limiting step for propene oligomerization in Ni-Na-X is carbon-carbon bond formation between coordinated propene and the Ni-propyl complex that proceeds via migratory insertion in a Cossee-Arlman type mechanism. The apparent activation energy and the predicted carbon-branching selectivity calculated from apparent free energy barriers at 453 K agree well with experimental measurements. This study is the first of its kind to examine the activation process of Ni-exchange Na-X catalysts and provides a plausible explanation for the high selectivity to dimers (>90%) during the oligomerization of propene.

The QM/MM method has been combined with the quasiclassical trajectory (QCT) method to study alkene methylation by methanol catalyzed by the zeolite H-MFI. The rate-limiting step for this reaction is the methylation of the alkene. The apparent activation energy for this step calculated ab initio density functional theory agrees well with the value obtained from experiments and previous full QM calculations. Following the ethene methylation transition state toward the products along the intrinsic reaction coordinate reveals the existence of a protonated cyclopropane (PCP+) carbocation intermediate. A similar protonated methyl-cyclopropane (mPCP+) carbocation intermediate is found for propene methylation. The intermediates produced during the alkene methylation reaction are metastable with a lifetime of roughly 1 ps obtained from QCTs initiated at the transition state for the rate-determining step. Due to the short lifetime of these intermediates, the energy in the carbocation does not achieve thermal equilibrium with the zeolite lattice before subsequent reaction occurs. The qualitative difference between product distributions obtained by static and dynamic reaction pathways suggests the pathways of zeolite-catalyzed reactions proceed through high-temperature pathways that differ from the 0 K potential energy surface. The transformation of the m-PCP+ intermediate to the longer-lived secondary 2-butyl carbocation observed during QCTs suggests that more-stable carbocations can properly thermalize and exist as reaction intermediates for longer than 1 ps.

Free energies of activation for zeolite-catalyzed reactions can be done in the blue-moon ensemble by constrained ab initio molecular dynamics simulations using the QM/MM method. We have used thermodynamic integration (TI) to study the water-assisted proton hopping in zeolite H-MFI. The free energy of activation calculated by TI is compared with that calculated with the quasi rigid-rotor harmonic oscillator (q-RRHO) recently proposed by Grimme. The estimate for the correlation time of proton hopping is improved by two orders of magnitude with the thermodynamic integration technique relative that determined using the q-RRHO approximation. In order to achieve close agreement between the predicted activation barrier for proton hopping and that determined from experimental measurements of the correlation time, it is necessary to use a level of theory that is higher than is practical for molecular dynamics. Therefore, in order to obtain a good estimate of the free energy of activation, it is necessary to use the results of thermodynamic integration to correct the estimate of the entropic contribution to the free energy obtained using the q-RHHO approximation obtained from QM/MM calculations done using a high level of theory for the QM portion of the calculations.