IISc researchers map thousands of CO2-to-methanol reactions using machine learning and copper catalysts

Researchers at the Indian Institute of Science have developed a computational framework that maps nearly 10,000 chemical reactions involved in converting carbon dioxide into methanol-like fuels. The team started with 152 reactions from quantum-mechanical simulations, then trained machine learning models to predict activation energy barriers and automatically expand the network to cover 105 surface species. The larger network corrected earlier model errors and predicted a 40-fold increase in CO2 conversion, with methanol and carbon monoxide as major products. The work was validated with Hindustan Petroleum Corporation Limited's Green Research and Development Centre and a Singapore research agency. The approach could help scientists design better catalysts for turning captured CO2 into useful chemicals and fuels, which matters for any serious effort to make low-carbon fuels at scale. Still, this is a modeling result, not a working reactor, so the practical path to commercial production remains open.