In this work we present a detailed \textit{ab initio} study of the carbonylation reaction of methoxy groups in the zeolite mordenite, as it is the rate determining step in a series of elementary reactions leading to ethanol. For the first time we employ full molecular dynamics simulations to evaluate free energies of activation for the reactions in side pockets and main channels. Results show that the reaction in the side pocket is preferred and, when dispersion interactions are taken into account, this preference becomes even stronger. This conclusion is confirmed using multiple levels of density functional theory approximations with (PBE-D2, PBE-MBD, and vdW-DF2-B86R) or without (PBE, HSE06) dispersion corrections. These calculations, that in principle would require several demanding molecular dynamics simulations, were made possible at a minimal computational cost by using a newly developed approach that combines thermodynamic perturbation theory with machine learning.
Ferrierite is industrially used as a catalyst for the dehydration of isobutanol into butenes and the critical part of this transformation is presumably catalyzed by Bronsted acid (BA) sites located on its external surface. In this work, we present a DFT investigation of a complete reaction network over the T1O3 BA site at the (100) surface of this zeolite. Exploration of reaction mechanisms by means of the static approach revealed that the transformation towards all products proceed in two steps - dehydration and deprotonation. Free energy static calculations showed that the dehydration step is rate determining for all competing reaction channels. Free energetics of the dehydration reaction was therefore investigated using ab initio molecular dynamics accelerated by machine learned forcefields (MLFFs) with accuracy controlled via a correction scheme based on machine learning perturbation theory. The activation free energies for transformations towards branched and linear products computed using the PBE+D2 functional (111.3 and 125.7 kJ/mol, respectively) suggest a preference of the former reaction channel. Separating dispersion correction from MLFFs allowed us to efficiently recompute free energies using other dispersion corrections (no correction, D3, D3(BJ) and D4) and confirm that this factor does not affect prediction on the preferred reaction channel.Our results are in agreement with our previous investigation of analogous reactions catalyzed by bulk chabazite, but not with experimental selectivities observed for ferrierite, which suggests that the considered BA sites are not at the origin of the specific selectivity of ferrierite for linear alkenes.