Image courtesy of Hartwig Lab
As the U.S. Department of Energy launches phase one of its Genesis Mission: Transforming Science and Energy with AI, a national initiative harnessing artificial intelligence and advanced computing to accelerate scientific discovery, research teams from the UC Berkeley’s College of Chemistry have been awarded funding to lead the federal initiative to double scientific productivity through artificial intelligence.
Out of thousands of proposals nationwide, the College was selected as one of just 278 teams awarded funding under the U.S. Department of Energy’s historic Genesis Mission.
Lead researcher John F. Hartwig, Dow Professor of Sustainable Chemistry at UC Berkeley’s College of Chemistry, plans, in this program, to use artificial intelligence to accelerate the discovery and understanding of catalysts for a more energy efficient chemical industry. The project tackles a major bottleneck in green chemistry: how to design better catalysts – the molecular engines behind nearly all plastics, fuels, and medicines – using AI, even when chemical data is extremely scarce.
Every day, energy-intensive chemical manufacturing powers our modern economy, producing the essential plastics, fuels, and industrial materials we rely on. Professor Hartwig’s team has pioneered some of the most widely used catalytic processes, discovering cleaner chemical reactions that drastically reduce the extreme heat, pressure, and energy required to make and break down complex molecules.
Now, through the DOE’s Genesis Mission, the Hartwig team is preparing to tackle automating the science behind catalysts. Conducted in partnership with Lawrence Berkeley National Laboratory and Cornell University, the project leverages multi-institutional expertise to translate AI catalyst predictions into real-world energy savings.
While standard AI usually fails when data is scarce, the team hopes to solve chemistry’s “small data” problem. By combining two pretrained models: a large language model and a 3D quantum-physics model (trained on 500M+ calculations), the framework will be able to predict and design custom catalysts for industrial reactions using fewer lab measurements. Combining text knowledge with 3D structures will allow the AI framework to learn from small experimental datasets and explain why a catalyst works, allowing it to accurately predict cleaner chemical reactions. This breakthrough helps transform slow, trial-and-error chemistry into a fast, precise science.
“Addressing the world’s most urgent energy challenges requires rethinking how we synthesize and recycle materials at the molecular level,” said Interim Dean Anne Baranger. “By combining Professor Hartwig’s work with advanced AI, we are turning traditional trial-and-error chemistry into predictive science that accelerates discoveries for efficient, industrial use of chemical feedstocks.”