Researchers at the Indian Institute of Science combined machine learning with quantum chemistry simulations to map nearly 10,000 reactions in CO2 hydrogenation on a copper catalyst. The expanded network corrected earlier models that pointed to formic acid instead of methanol, and it raised predicted CO2 conversion by roughly 40 times to match lab results. The team also found evidence that molecular hydrogen can transfer directly to oxygen-containing intermediates, a pathway that standard models had assumed away. The work is early-stage and does not mean CO2-to-fuel plants are imminent. But it gives catalyst researchers a more complete reaction map to work from, which could lower the cost of making methanol and other chemicals from captured CO2. That matters for carbon utilization projects that depend on cheaper catalytic conversion.
