Harnessing Physics-inspired Machine Learning to Design Nanocluster Catalysts for Dehydrogenating Liquid Organic Hydrogen Carriers

Using liquid organic hydrogen carriers for the trans-oceanic shipment of hydrogen requires selective and low-cost dehydrogenation catalysts. Machine learning methods can accelerate the discovery of these catalysts. The state-of-the-art machine learning methods are however limited by challenges associated with building predictive models for large cyclic intermediates that adsorb and react on low-symmetry active sites. Focusing on methyl cyclohexane dehydrogenation to toluene, an industrially relevant hydrogen carrier, we introduce a machine learning approach to accelerate the design of selective and cost-effective catalysts. Using inputs to a gaussian process regression model that are inspired by physical theories of chemisorption, we predict the adsorption energies of large hydrocarbon intermediates that are encountered during methyl cyclohexane dehydrogenation. Across bimetallic active sites of nanoclusters having varied shapes and compositions, our model yields mean absolute errors of 0.11 – 0.25 eV on test sets and utilizes under 100 datapoints per reaction intermediate. This model is integrated with a microkinetic model to identify promising catalysts. Modifying Pt nanoclusters with IB, IIB, and post-transition elements like Cu and Sn increases dehydrogenation rates, reduces unselective reactions, and lowers Pt utilization, consistent with prior experiments. This work presents a scalable, and efficient framework for designing bimetallic catalysts for dehydrogenating hydrogen carriers.