Nearly 10,000 mapped reactions reveal overlooked steps in CO₂-to-fuel conversion

Scientists at India's Indian Institute of Science (IISc) have mapped nearly 10,000 chemical reactions involved in converting CO₂ into fuels and chemicals using a copper catalyst. Traditionally, researchers only model a small number of likely reactions because simulating every possible reaction using quantum mechanics is computationally prohibitive. However, this approach can miss hundreds of important reactions.

The IISc team developed a data-driven computational framework that expands the reaction network to include nearly 10,000 elementary reactions. They began with a curated database of 152 reactions generated through quantum-mechanical simulations and used machine learning to predict activation energy barriers for additional reactions. Automated tools were also employed to predict all possible reactions involving 105 surface species and to identify single-step reactions.

When the researchers initially modeled the process using the 152 reactions, the network incorrectly predicted formic acid as the major product and underestimated CO₂ conversion. However, after expanding the network to include thousands of additional reactions, the predictions aligned with experimental observations. The expanded network correctly identified methanol and carbon monoxide as major products, and quantum calculations confirmed that hydrogen can transfer to reaction intermediates as molecular H₂, rather than only through individual hydrogen atoms.

This unexpected finding suggests that catalysts interacting more strongly with H₂ could potentially enhance pathways leading to methanol production. The framework combining quantum-mechanical simulations, machine learning, automated reaction enumeration, and kinetic modeling could also be applied to other industrial processes, such as CO₂ reduction on other catalysts, nitrogen reduction, and water splitting.