Kinetic Consequences\nof Quasi-Harmonic Entropies Calculated\nwith Machine Learning Interatomic Potentials for Microkinetic Modeling

Microporous\ncatalysts are ubiquitous in chemical processes including\nsustainable transformations of biobased feedstocks into fuels and\nfine chemicals. The mechanistic insights needed to design next-generation\nmicroporous catalysts can be obtained with <i>ab initio</i> simulations coupled with microkinetic modeling, yet active site\nconfinement complicates an accurate determination of adsorbate entropies,\nwhich, in turn, affect predictions of rate and equilibrium constants.\nIn this study, we developed a machine learning force field (MLFF)\nstrategy to rapidly predict temperature-dependent quasi-harmonic adsorbate\nentropies in zeolite Beta, reducing the number of compute-intensive <i>ab initio</i> molecular dynamics calculations needed to construct\na microkinetic model. These entropies directly impacted the kinetics\nof a model parallel reaction mechanism. We chose lactic acid dehydration\nto acrylic acid on aluminosilicate zeolite Beta to explore the pathway\ndependence of unselective product formation and initial deactivation\nmechanisms using microkinetic modeling with our MLFF entropy strategy.\nThe resulting quasi-harmonic entropy approximations led to shifts\nin steady-state coverages that impacted reaction orders and product\nselectivity. At low lactic acid partial pressures, concerted monomolecular\ndecarbonylation is favored over Brønsted acid sites, which then\nshifts at high lactic acid partial pressures to concerted bimolecular\ncondensations into lactic acid oligomers. Sequential pathways mediated\nby adsorbed alkoxide or carbonyl intermediates have no kinetic relevance\nat these conditions. These findings provide a strategy to integrate\nquasi-harmonic entropies into microkinetic modeling that is scalable\nwith reaction temperature and applicable to a wide range of catalysts\nand catalytic cycles.