(675a) Data-Driven Massive Reaction Networks Reveal New Pathways Underlying Catalytic CO2 Hydrogenation

Although carbon dioxide (CO₂) hydrogenation using copper-based catalysts is one of the most promising ways to close the carbon cycle and, thus, reduce global warming, the exact mechanisms of the formation of products such as methanol and carbon monoxide are scarcely understood. This is attributed to the intractable computational expense in accurately modeling the large number of transition-state structures involved in the formation of hydrocarbons and oxygenates on catalyst surfaces via quantum-mechanical density functional theory (DFT) calculations. In this work, we evolve a novel strategy of combining large-scale activation barrier calculations via transition-state computations using DFT, state-of-the-art machine learning (ML) models for accurately predicting activation energies based on reaction/reactant fingerprints, algorithms for reaction network exploration, and automated microkinetic modeling (MKM) to simulate a comprehensive reaction network consisting of 9389 elementary reactions. This contrasts with typical studies available in the literature that hardly account for dozens of elementary steps. Our work also demonstrates that manually curated reaction networks are prone to human biases in considering only certain specific types of reactions in the quantum-mechanical model. In addition, we carry out carefully designed experiments to determine the precise conversion rate of CO₂ during thermochemical hydrogenation on the (111) facet of copper. Considering hundred-plus C₁-C₄ compounds, our automated approach unravels up to 40-fold higher CO₂ conversion rates, better following experimental trends, contrary to inferences from a model based on a smaller subset of 152 manually curated reactions. We also establish intermolecular hydrogen-hopping reactions to be the key to accurately simulating the CO₂ hydrogenation pathway and reveal the exact pathway for methanol and carbon monoxide. Overall, our study will significantly accelerate the study of complex chemical mechanisms, enabling better predictions and deep chemical insights, particularly into thermochemical CO₂ reduction, which is a crucial chemical conversion process for a cleaner future.