AI for Discovering Clean Energy Materials

Technological advances in clean energy depend on the improvement of materials, namely types of fuel cells, which convert fuel into electricity. These cells depend on the oxygen reduction reaction, a slow reaction that requires a catalyst to be scalable and practical for industry. The ideal electrocatalysts have been difficult to find, because fuel cell technologies employ either strong acid or strong alkaline electrolytes, which causes catalyst corrosion. This poses challenges in the discovery of new catalysts, since it requires that both the catalysts’ activity and stability be simultaneously addressed. Furthermore, the time and difficulty of finding or creating catalysts as well as predicting and testing their properties is greatly slowing the potential for progress.

Carla P. Gomes, Computer Science, and John M. Gregoire (California Institute of Technology) are working toward leveraging recent advances in high-throughput materials science and computer science to create a catalyst discovery platform with broad implications for the development of functional materials. This research is part of a large, multi-university, collaborative effort funded by the Toyota Research Institute to find better fuel cell catalysts.

Pioneering the intersection of computer and materials sciences, Gomes and Gregoire are using high-throughput experimentation to generate large-scale datasets of electrocatalysts that include more extensive and more complex catalysts. They are also developing artificial intelligence (AI) and machine learning models as well as algorithms that incorporate constraint reasoning and optimization and that are able to infer and predict composition-structure-property relationships and the viability of the catalysts. The rich database will dramatically advance the state-of-art in materials design and discovery. Their data aggregation, interpretation, inference, and prediction models will address challenges throughout materials science, computer science, and beyond.

Cornell Researchers

Funding Received

$1 Million spanning 3.5 years