As scientists and engineers work to decarbonize our planet, one energy carrier has emerged as a possible substitute for fossil fuels: hydrogen. When used as a fuel, hydrogen does not directly produce any greenhouse gases—it only produces water. Hydrogen can also be used to store and deliver energy produced from other sources.
However, scientists must use electrocatalysts to promote the reactions needed to make hydrogen a viable fuel source. Currently, these catalysts are made from expensive materials like platinum and palladium.
Engineers have tried for decades to find new materials that can provide the same reactions needed for a much lower cost, but such experiments involve a lengthy trial-and-error process in the lab.
The answer could be found using machine learning, says Rui Ding, an Eric and Wendy Schmidt AI in Science Postdoctoral Fellow co-advised by Prof. Junhong Chen, a leading expert in nanomaterials and water technology, and Prof. Yuxin Chen, who specializes in AI. He should know: he spent five years conducting these trial-and-error experiments for his PhD. But during the COVID-19 pandemic, when he was locked down in his apartment in Wuhan, China, he was unable to return to the lab. So, he switched his focus to AI and machine learning, applying these techniques to materials science research.
Now at the University of Chicago Pritzker School of Molecular Engineering (PME), Ding found that there was a need for a comprehensive review of machine learning techniques that could aid scientists in discovering new electrocatalyst materials. He, Chen, and a group of collaborators reviewed more than 150 papers in this area, evaluating various reactions, potential materials, and applicable machine-learning methods.
The result, published in Chemical Society Reviews, offers a meta-level analysis of the current research landscape, serving as a valuable guide for scientists and industry professionals aiming to advance this area of study.
“If you want to find new electrocatalysts, this paper gives you everything you need to start,” Ding said.
Analysis of each electrocatalyst reaction
The paper offers overviews for each reaction needed to use hydrogen as an energy source. The first two, hydrogen evolution reaction (HER) and oxygen evolution reaction (OER), split a water molecule into hydrogen and oxygen. Electricity is then stored in hydrogen molecules.
The second two, hydrogen oxidation reaction (HOR) and oxygen reduction reaction (ORR), convert hydrogen and oxygen into water. These reactions are necessary for energy release.
For each of these reactions, Ding and his co-authors offer an overview of potential material categories that could be used in electrocatalysts, and offer suggested machine-learning techniques that might help identify the best materials.
“If you want to find new electrocatalysts, this paper gives you everything you need to start.”
Rui Ding, Eric and Wendy Schmidt AI in Science Postdoctoral Fellow
“We provide a meta-level analysis of all the publications in the field: what techniques have been used, how much data you would need to use these techniques, which material categories have been tested, and which are the best,” Ding said. “It gives people very clear statistics and a hands-on tutorial on which machine learning tool they could use to benefit their research.”
Next, Ding is working to create a model that would allow researchers to test many different material categories at once, speeding up the process even more.
“I began as an experimental scientist, but when I changed directions and learned AI techniques, I knew that was the future,” Ding said. “Problems like this require an interdisciplinary approach—chemistry, materials science, and AI. UChicago has a very vibrant ecosystem where ideas across disciplines can collide. That’s how we are able to have creative ideas to try different approaches.”
Other authors on the paper include Yuxin Chen, Jianguo Liu, Yoshio Bando, and Xuebin Wang.
Citation: “Unlocking the Potential: Machine Learning Applications in Electrocatalyst Design for Electrochemical Hydrogen Energy Transformation,”Ding et al, Chemical Society Reviews. October 9, 2024. DOI: https://doi.org/10.1039/D4CS00844H
Funding: Eric and Wendy Schmidt AI in Science Postdoctoral Fellowship, National Natural Science Foundation of China, Jiangsu Provincial Natural Science Foundation, Jiangsu Provincial Key Research and Development Program