Scientists have uncovered thousands of previously overlooked chemical reactions that could significantly improve the conversion of carbon dioxide into valuable green fuels and essential industrial chemicals. By seamlessly combining machine learning algorithms with advanced quantum mechanics computer simulations, a dedicated research team at the Indian Institute of Science identified 9,389 individual chemical reactions occurring simultaneously on a copper catalyst. This groundbreaking computational effort has revealed unexpected chemical pathways that traditional, simplified models had completely missed for decades.
The findings, published in the prestigious journal Nature Communications, hold profound implications for the global scientific community. They could ultimately help engineers and chemists develop vastly more effective and efficient catalysts for transforming greenhouse-gas-level CO2 into commercially viable and environmentally sustainable products, most notably methanol. Furthermore, the research uncovered an entirely surprising and unconventional way that hydrogen can participate in these surface reactions, fundamentally challenging long-established assumptions about how catalytic hydrogenation processes actually operate at a molecular level.
For years, one of the most promising and heavily researched approaches to recycling carbon dioxide and mitigating climate change has been catalytic hydrogenation. This chemical process utilizes a solid catalyst to help captured CO2 react directly with hydrogen gas to form useful hydrocarbons and alcohols such as methanol. However, the underlying physical reality of this process involves an astronomically enormous number of individual microscopic reactions taking place concurrently across the complex, irregular surface of the catalyst.
Simulating every single possible molecular step using rigorous quantum mechanics principles requires immense, often prohibitive computing power. Because of these computational limitations, researchers have historically been forced to rely on heavy simplifications. They typically focus only on a narrow, highly curated selection of reactions that they intuitively believe are the most important or kinetically favorable. Unfortunately, this necessary simplification frequently leaves hundreds, or even thousands, of potentially critical reaction pathways completely out of the picture, leading to theoretical models that often fail to match real-world laboratory experiments.
"We began with a worry familiar to anyone who does mechanistic modeling: How do you know that your reaction network has not omitted the one step that matters?" says first author Anand Mohan Verma, who conducted the pivotal research as a CV Raman Postdoctoral Fellow at IISc and is now an Assistant Professor at the Motilal Nehru National Institute of Technology Allahabad.
AI Reveals Thousands of Overlooked Chemical Reactions
To successfully overcome this longstanding computational limitation and bridge the gap between theory and experiment, the research team adopted an innovative, multi-step workflow. They began by utilizing high-precision quantum mechanics simulations to construct a detailed, foundational database containing 152 baseline chemical reactions. They then trained sophisticated machine learning models to rapidly and accurately estimate the energy barriers that additional, unmapped reactions would need to overcome to proceed on the catalyst surface.
Equipped with these automated computational tools, the scientists systematically examined the vast landscape of possible reactions involving 105 distinct chemical species present on the copper catalyst surface. They meticulously determined which molecular transformations could realistically take place in a single reaction step. This ambitious, data-driven approach dramatically expanded their original, limited network from just 152 reactions to an astonishing 9,389 pathways.
The operational difference between the small model and the expanded network was immediate and profound. When the researchers relied solely on the smaller, conventional network, their computational model incorrectly identified formic acid as the primary reaction product rather than methanol. Moreover, it predicted substantially less total CO2 conversion than what physical experiments had consistently demonstrated in laboratory settings.
"When we modeled the process using the 152 reactions considered initially, the network wrongly predicted formic acid, not methanol, as the major product, and underestimated how much CO2 gets converted. Only when we expanded the network to include thousands of additional, previously overlooked reactions did the predictions fall in line with what we and others see experimentally," explains corresponding author Ananth Govind Rajan, Associate Professor in the Department of Chemical Engineering at IISc.
Expanded Model Predicts 40 Times More CO2 Conversion
When the scientists incorporated the massive, newly mapped reaction network into an advanced kinetic model tracking how quickly the chemical processes occur, the predictive power of their simulation shifted dramatically. The predicted amount of CO2 conversion increased by approximately 40 times compared with the original, restricted model, bringing theoretical mathematics much closer to physical reality.
Crucially, the expanded model also correctly identified methanol and carbon monoxide as the major products of the reaction, bringing its computational predictions into striking agreement with independent experimental findings. This validation step was not done in isolation; experimental validation was rigorously carried out by G Valavarasu and Santhosh Kotni at Hindustan Petroleum Corporation Limited’s Green Research and Development Centre, alongside Amol Amrute and his research colleagues at the Agency for Science, Technology, and Research in Singapore. Additionally, Ambedkar Dukkipati, a professor in the Department of Computer Science and Automation at IISc, contributed heavily to the development and refinement of the machine learning models that made the breakthrough possible.
A Surprising Hydrogen Pathway Challenges Conventional Chemistry
Beyond dramatically improving the quantitative predictions of CO2 conversion rates, the sheer scale of the much larger reaction network exposed a fascinating and entirely unexpected chemical mechanism involving hydrogen.
In several of the most crucial reaction steps identified by the algorithm, hydrogen was shown to transfer to intermediate chemical compounds directly as an intact molecular $H_2$ entity, rather than exclusively through the conventional route of splitting into separate, individual hydrogen atoms prior to transfer. Additional, rigorous quantum mechanics calculations were subsequently performed and confirmed that this pathway can be thermodynamically and kinetically favorable, particularly when hydrogen is transferred directly to intermediate species containing oxygen atoms.
This empirical finding directly challenges a long-standing chemical dogma regarding how hydrogen participates in heterogeneous catalytic reactions. It could provide chemical engineers with immensely valuable clues for designing better, more selective catalysts in the future.
"The idea that hydrogen can transfer as an intact molecule, without first splitting into atoms, runs against what most of us were taught," says co-author Shivam Chaturvedi, a PhD student in the Department of Chemical Engineering at IISc. "This surfaced only because the network was large enough to allow for it, and the observation held up when we went back and computed those steps explicitly. This also suggests that catalysts that interact more strongly with $H_2$ could potentially enhance pathways leading to methanol."
The researchers firmly believe that this unexpected mechanism could serve as a vital design principle, guiding the future development of novel catalysts that interact more effectively with molecular hydrogen. In turn, this could optimize the specific chemical pathways responsible for cleanly and efficiently producing methanol from captured carbon dioxide.
Furthermore, the broader computational framework developed during this study may prove to be broadly useful well beyond the specific confines of carbon dioxide hydrogenation. By successfully integrating quantum mechanics simulations, advanced machine learning, automated identification of chemical pathways, and robust models of reaction kinetics, the innovative approach could help researchers investigate a wide array of other critical industrial and environmental processes.
Potential future applications of this methodology include advanced CO2 reduction using alternative catalyst materials, sustainable nitrogen reduction for greener fertilizer production, and efficient water splitting for hydrogen fuel generation. By revealing subtle chemical steps and transient pathways that simpler, traditional models inevitably miss, the new technique offers scientists a vastly more complete, panoramic picture of complex catalytic reactions, opening up fresh opportunities to improve the chemical manufacturing processes that power modern industry.
