(244f) Designing Bimetallic Nanoparticles That Catalyze Methyl Cyclohexane Dehydrogenation Using Machine Learning

As a promising liquid organic hydrogen carrier, the catalyst design of the dehydrogenation reaction of methyl cyclohexane (MCH) is important for hydrogen storage and transport. Although Pt-based bimetallic nanoclusters (diameter < 2 nm) are used as industrial catalysts of MCH dehydrogenation [1,2], the structure-property relationship for MCH dehydrogenation catalysts needs a systematic understanding to design selective catalysts. To model reaction kinetics in a huge catalyst space, machine learning (ML) assisted first-principle computation has emerged as a powerful tool, demonstrated for heterogeneous catalytic reactions involving small molecules [3]. However, it is challenging to establish a ML model for large cyclic hydrocarbons adsorbed on small nanoclusters, due to the complex multidentate adsorption and the cost of data labelling for constructing training sets. Here, utilising physics-based features, we introduce a cost-effective and accurate ML approach using the Gaussian Process Regression (GPR) algorithm, where active learning reduces the cost of data collection. Trained on a dataset of <100 points, our GPR model achieves a 0.15 eV mean absolute error (Figure a), with excellent transferability to new elements, non-Pt bimetallic sites, and toluene-like unsaturated C₇ hydrocarbons. Combining the GPR model with a lumped microkinetic model, we simulated the turnover frequency and selectivity over diverse bimetallic active sites (Figure b). Pt-based nanoclusters with metals possessing a filled d-band like Cu emerge as effective catalysts for MCH dehydrogenation. This machine learning-accelerated screening method, extending to hydrogenation and dehydrogenation of other liquid organic hydrogen carriers, offering a versatile tool for catalyst screening over numerous bimetallic nanoparticles.

[1] Okada Y., Extended abstracts of the 9th Tokyo Conference on Advanced Catalytic Science and Technology, Fukuoka, KL14, (2022).

[2] Meng, J., Zhou, F., Ma, H. et al., Top. Catal., 2021, 64, 509

[3] Mou, T., Pillai, H.S., Wang, S. et al., Nat. Catal.,2023, 6, 122