(584ab) Modeling Catalytic CO2 Hydrogenation with Automatically Generated Mechanisms: Evaluating Products across a Wide Search of Hypothetical Surfaces

Current levels of carbon dioxide (CO₂) in the atmosphere have negative climate implications, yet the stability of CO₂ presents a barrier to its conversion. Heterogeneous catalysis offers a promising route for producing valuable products through CO₂ hydrogenation, but a characteristic challenge in this realm is breaking the Anderson-Schulz-Flory distribution to achieve higher selectivity for longer-chain products. Catalysts capable of this have been discovered, but associated reaction mechanisms are not well-defined. Developing detailed mechanisms for CO₂ hydrogenation on successful catalysts can help identify driving chain growth steps to support catalyst improvement or discovery.

This study presents a computational workflow for searching a wide range of hypothetical catalysts and automatically generating detailed microkinetic models for potential CO₂ hydrogenation on each. Unique metals can be described by the binding energies of carbon, oxygen, and hydrogen for adsorbate species, and mechanisms were proposed for 363 combinations of these. Each mechanism was generated by the software Reaction Mechanism Generator (RMG) and simulated using the software Cantera to predict product distributions along a plug flow reactor (PFR) with CO₂ and hydrogen (H₂) feed streams. Bimetallic catalysts with a shared gas phase and multiple surface site types were simulated as combinations of these mechanisms to expand the range of catalysts further.

Simulated consumption of CO₂ and H₂ was used to identify the most “active” hypothetical metals under experimentally relevant conditions. Corresponding mechanisms included reactions for experimental products, but agreement between simulated and observed conversion and selectivity varied within this “active” space. While confidence in the “best” mechanism describing CO₂ hydrogenation on a particular catalyst is dependent on the experimental data available to validate resulting predictions, these findings indicate that descriptors such as binding energies and bimetallic combinations can serve as a starting point for suggesting catalysts capable of converting CO₂ into desired products.