Researchers at Stony Brook University have leveraged machine learning to identify six promising solvents for carbon dioxide electroreduction, a process that uses clean electricity to convert CO2 into useful fuels and products. The work, led by Ph.D. researcher Kuldeepsinh Raj and principal investigator Professor Nav Nidhi Rajput from the Department of Materials Science and Chemical Engineering, aims to address the climate challenge posed by rising CO2 levels.
Carbon dioxide is a major driver of climate change, and finding efficient ways to transform it into valuable resources is a critical frontier in sustainability. Electroreduction relies on dissolving CO2 in a solvent before it can be converted, but traditional solvents often suffer from low CO2 solubility or poor energy efficiency. By screening thousands of candidates computationally, the team narrowed the field to half a dozen solvents with optimal properties.
The machine learning model evaluated factors such as CO2 solubility, electrochemical stability, and cost. The six shortlisted solvents are expected to enhance reaction rates and reduce energy demands compared to conventional alternatives. Specific numerical performance metrics were not disclosed in the source article.
If validated experimentally, these solvents could make industrial-scale CO2 conversion more economically viable, potentially allowing power plants or factories to turn emissions into synthetic fuels, plastics, or chemicals. The approach also demonstrates how data-driven methods can accelerate materials discovery, shrinking years of lab work into months.
"We were able to screen a vast chemical space that would have been impossible to explore manually," Rajput said. The team now plans to test their top candidates in lab-scale electrochemical cells.