Ab Initio Prediction of Proton Exchange Barriers for Alkanes at Brønsted Sites of Zeolite H-MFI

Abstract

A hybrid of high level and low level quantum mechanics (QM) methods has been employed to predict intrinsic and apparent energy barriers for the direct proton exchange mechanism of methane, ethane, propane, n-butane, and i-butane on Brønsted sites of H-MFI. The specific hybrid MP2:PBE+D2 + ΔCC implementation used is known to yield the so-called “chemical accuracy” (±4 kJ/mol). Whereas the apparent enthalpy barriers decrease with increasing C number from 104 to 63 kJ/mol, in line with the decreasing heat of adsorption, the intrinsic enthalpy barriers are constant within 124–127 kJ/mol at 500 K. For methane, ethane, propane, and n-butane, we find the expected agreement of apparent barriers with activation energies from batch recirculation reactor experiments. The activation energies derived from NMR experiments (103–113 kJ/mol) are similarly constant as the predicted intrinsic barriers but systematically lower. For i-butane the predicted intrinsic and apparent barriers for the direct proton exchange step are the same as for n-butane with deviations of 2–5 kJ/mol, while the experiments yield values that are 50–60 kJ/mol lower, far outside the estimated range of combined experimental and computational uncertainty (±14 kJ/mol). A change to the indirect proton exchange mechanism, in which a hydride ion is transferred between the alkane and a tert-butyl carbenium ion can be excluded, because we confirm previous findings that the barrier for dehydrogenation that would create a tert-butyl cation from i-butane is much too high, 188 and 132 kJ/mol for the intrinsic and apparent enthalpy barriers, respectively, at 500 K. The possible role of extraframework- and framework-bound alumina species is discussed.

On the Accuracy of Density Functional Theory in Zeolite Catalysis

“…In this work we strictly evaluate the energy error for a given T46‐cluster derived from a PBE‐D3 optimized periodic structure, but stress that reoptimization of the structures at a different level of theory will affect these errors. Approaches that optimize periodic structures using more accurate forces from cluster models show that deviations for proton‐exchange reactions of alkanes are on the order of 1 kJ/mol between optimized and single point calculations …”

“…Approaches that optimize periodic structures using more accurate forces from cluster models show that deviations for proton-exchange reactions of alkanes are on the order of 1 kJ/mol between optimized and single point calculations. [31] The acid-catalyzed reactions that we will use in the first part employing H-SSZ-13 are shown in Scheme 1 and Scheme 2. These reactions have been largely investigated previously.…”

“…In order to perform an extensive benchmark analysis, several different models of Cu‐SSZ‐13 have been tested and compared with different levels of theory, similar as described for H‐SSZ‐13. While CCSD(T) is an established benchmark method for the acid catalyzed reactions studied above, this is less clear for Cu‐SSZ‐13. D1 and T1 diagnostics were checked (see the SI Table S8), and do not indicate multireference character.…”

“…Similarly, the accuracy of DFT calculations for these materials is also highly interesting . While these sizes limit the computational method to DFT, Sauer and co‐workers have introduced a hierarchical cluster approach, that represents an elegant method allowing to employ higher level calculations as well . In this approach, nonperiodic cluster models of the active site are cut out of the zeolite structure and used to supplement DFT with higher level methods such as MP2 or CCSD(T) calculations as shown exemplary in Figure .…”

“…[10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26] While these sizes limit the computational method to DFT, Sauer and co-workers have introduced a hierarchical cluster approach, that represents an elegant method allowing to employ higher level calculations as well. [27][28][29][30][31] In this approach, nonperiodic cluster models of the active site are cut out of the zeolite structure and used to supplement DFT with higher level methods such as MP2 or CCSD(T) calculations as shown exemplary in Figure 1. [27,29,30,[32][33][34] This methodology can be used to obtain accurate reaction energies for all problems associated with zeolite chemistry, hence giving a guideline of the accuracy of DFT calculations in zeolite catalysis.…”