45843-AC5
Catalyst Modification to Reduce Product Inhibition During High Temperature Water-Gas Shift
The simple two-step microkinetic model (CO + O* → * + CO2; H2O + * → O* + H2) was fit to a published high temperature water-gas shift data set including 189 data. The microkinetic model was formulated with five adjustable parameters: the entropies of activation for the two steps, the enthalpies of activation for the two steps, and the enthalpy of localization of an oxygen adatom on a catalytic site. (The entropy of localization on the site was estimated on the basis of changes in translational entropy and was not used as an adjustable parameter.) While this two-step microkinetic model oversimplifies the chemistry of high temperature shift, it is capable of providing a meaningful value for the enthalpy of localization of oxygen adatoms on active sites. In fact, the model provided an excellent fit to the data (correlation coefficient of 0.91). The corresponding value of the enthalpy of localization of an oxygen adatom on an active site for high temperature shift was 611 kJ/mol. The high temperature shift catalyst contains magnetite as its active component. Magnetite is an inverse spinel with some unusual properties. It is ferrimagnetic, and at operating conditions the Fe(II) and Fe(III) cations located in octahedral sites share electrons via a rapid electron hopping process. In light of these behaviors, it was not clear whether it would be possible to use small clusters as models for the active sites of magnetite for the purposes of computational chemistry. In this regard, the simple microkinetic model just described is very attractive because it only involves two surface species, the empty site and a site with an oxygen atom adsorbed, and it provides a value for the enthalpy difference between the two. The crystallographic structure of magnetite was therefore used to create three different cluster that exposed a [100], [110] and [111] surface, respectively. The cluster representing the [100] surface consisted of five Fe3O4 formula units, the [110] cluster consisted of four formula units and the [111] cluster contained six formula units plus an additional iron atom and an additional oxygen atom. Density functional theory (TZV**/B3LYP) was used to compute the geometry and energies of the vacant and oxygen-occupied sites for these three clusters. The resulting energies of localization of the oxygen adatom for the [100], [110] and [111] surfaces were 112, 378 and 668 kJ/mol, respectively. These values vary in direct proportion to the number of coordinating oxygen atoms that are missing from the active octahedral cation site. The experimental enthalpy of oxygen adatom localization resulting from the microkinetic modeling fall within the range of computed values. If one recognizes that the real catalyst is likely to expose a variety of different surface geometries, then one might interpret the experimental value to be an average over these surfaces. Using this interpretation, the computational results indicate that on average, the active octahedral cations sites are coordinated to only 3.2 oxygen anions (not counting the adsorbed adatom). Indeed, natural magnetite crystals tend to expose predominantly [111] surfaces (where surface cations would be coordinated to three anions), with a smaller fraction exposing [110] surfaces (where surface cations are coordinated to four anions). The value of 3.2 is quite reasonable being between the values for the [111] and [110] surfaces, but closer to the former. The kinetic modeling also generated experimental values for the entropies and enthalpies of activation of the two steps. One might be tempted to use computational chemistry to calculate these quantities and thereby create additional points of comparison between the cluster models and the experimental results. It is quite likely that the two steps used in the model are not elementary, and consequently the enthalpies of activation derived from the experiments could not be computed using a DFT transition state search. (It is, however valid to compare the experimental value for the heat of localization to the computational value because even in this simple model, the heat of localization is a thermodynamic quantity associated with that specific species.) On the basis of the reasonably good agreement between the experimental and computed values, more rigorous mechanistic kinetic models are now under consideration. Specifically, mechanistic kinetic models involving a redox pathway and models involving a formate pathway are being fit to experimental data. Cluster models of the surface intermediates in these models are being used to find their geometries and energetics. In addition, the effect of copper promoter upon both the experimental results and the computational results is being investigated. These second-year studies will provide a more definitive assessment of the utility of DFT cluster models to help discriminate among mechanistic pathways for high temperature shift.
