Mapping 10,000 reactions uncovers overlooked CO₂-to-fuel conversion steps

Researchers at the Indian Institute of Science have developed a novel computational framework that maps nearly ten thousand elementary chemical reactions involved in converting carbon dioxide into fuels and chemicals on copper catalysts. Published recently in Nature Communications, the study addresses a longstanding bottleneck in mechanistic modeling, where computational limits traditionally force scientists to simulate only a fraction of possible reaction pathways, often omitting critical steps. The team, led by associate professor Ananth Govind Rajan and first author Anand Mohan Verma, began with a curated database of 152 reactions derived from quantum mechanical simulations. To overcome scaling limitations, the researchers integrated machine learning models trained to rapidly predict activation energy barriers, alongside automated tools that enumerated all feasible single-step reactions among 105 surface species. This data-driven approach expanded the reaction network to 9,389 steps, fundamentally altering predictive accuracy. When tested against traditional models, the expanded network corrected a significant discrepancy. Earlier simulations incorrectly identified formic acid as the primary product and severely underestimated carbon dioxide conversion rates. Incorporating the comprehensive reaction map increased predicted CO2 conversion by approximately 40-fold and accurately forecast methanol and carbon monoxide as the dominant outputs, aligning precisely with experimental data validated by collaborators at Hindustan Petroleum Corporation Limited and the Agency for Science, Technology and Research in Singapore. Beyond improved accuracy, the expanded network uncovered a previously unrecognized mechanistic pathway. The analysis revealed that hydrogen can transfer directly to reaction intermediates as intact molecular hydrogen, bypassing the conventional requirement of atomic dissociation. This molecular hydrogen transfer proves particularly favorable for oxygen-containing intermediates. Co-author Shivam Chaturvedi noted that this discovery challenges traditional catalytic theory and suggests that engineered catalysts with stronger hydrogen-molecule interactions could further optimize methanol synthesis routes. The framework synthesizes quantum mechanical calculations, machine learning prediction, automated reaction enumeration, and kinetic modeling into a unified pipeline. Researchers emphasize that this methodology is not limited to CO2 hydrogenation. The same computational architecture can be adapted to model other industrially vital catalytic processes, including nitrogen reduction, water splitting, and carbon dioxide conversion over alternative catalyst materials. By systematically capturing overlooked reaction steps, the study provides a scalable blueprint for accelerating the design of next-generation industrial catalysts and advancing carbon capture and utilization technologies.