Researchers at the Indian Institute of Science developed a framework mapping 9,389 chemical reactions to turn carbon dioxide into fuel. This study, published in Nature Communications, helps scientists design better catalysts. "This surfaced only because the network was large enough to allow for it," said co-author Shivam Chaturvedi about findings.
Turning carbon dioxide from a greenhouse gas into useful fuels and chemicals could become an important part of future efforts to reduce industrial emissions. But there is a problem: the chemistry involved is far more complicated than it appears.
Researchers at the Indian Institute of Science (IISc) have now developed a computational framework that maps 9,389 elementary chemical reactions involved when CO2 reacts with hydrogen on a copper catalyst.
The study, published in Nature Communications, could help scientists better understand how these reactions unfold and design more effective catalysts for converting CO2 into useful products.
The process, known as CO2 hydrogenation, uses hydrogen and a catalyst to transform carbon dioxide into products such as methanol and carbon monoxide. Thousands of tiny chemical steps can take place on the surface of the catalyst.
Traditionally, researchers model only a relatively small number of reactions that they believe are most important. Calculating every possible reaction using quantum mechanics is extremely expensive. But this approach can leave critical reactions out of the model.
The IISc team set out to address that problem.
Researchers first created a carefully curated database of 152 reactions using quantum-mechanical calculations. They then trained machine-learning models to rapidly estimate the energy barriers of additional reactions.
At the same time, automated computational tools were used to identify possible reactions involving 105 different surface species and determine which could occur as individual reaction steps. This allowed the original network to be expanded to nearly 9,400 reactions.
The difference was significant.
When the researchers modelled CO2 conversion using only the original 152 reactions, the model predicted formic acid rather than methanol as the major product and underestimated the amount of CO2 being converted.
After thousands of previously overlooked reactions were added, the model predicted an approximately 40-fold increase in CO2 conversion and correctly identified methanol and carbon monoxide as major products, matching experimental observations.
The expanded network also revealed an unexpected pathway for hydrogen.
The researchers found that, in several important steps, hydrogen could be transferred to reaction intermediates as intact H molecules, rather than first splitting into individual hydrogen atoms. Quantum-mechanical calculations confirmed that this pathway can be particularly favourable when hydrogen reacts with oxygen-containing intermediates.
"This surfaced only because the network was large enough to allow for it," said co-author Shivam Chaturvedi, highlighting how important it can be to consider a much larger reaction landscape.
The team says the framework combines quantum mechanics, machine learning, automated reaction discovery and kinetic modelling and could potentially be adapted to other catalytic processes, including CO2 reduction using different catalysts, nitrogen reduction and water splitting.
The researchers say the approach could ultimately help scientists move closer to designing catalysts that can efficiently turn waste CO2 into valuable chemicals and fuels.