Machine learning assisted high-throughput screening of transition metal single atom based superb hydrogen evolution electrocatalysts
Published in Journal of Materials Chemistry A, 2022
Recommended citation: Umer, M., Umer, S., Zafari, M., Ha, M., Anand, R., Hajibabaei, A., Abbas, A., Lee, G., & Kim, K. S. (2022). Machine learning assisted high-throughput screening of transition metal single atom based superb hydrogen evolution electrocatalysts. Journal of Materials Chemistry A, 10(12), 6679-6689. https://doi.org/10.1039/D1TA09878K
Producing hydrogen by splitting water needs a catalyst, and the best ones rely on platinum and other metals that are scarce and costly. An alternative anchors individual metal atoms onto a thin supporting sheet, using a fraction of the metal — but the sheet can be made of several different materials, the atom can be any of dozens of metals, and the support can be flawed or chemically seasoned in many ways. Around 364 such combinations were designed and computed here from quantum mechanics, across supports including graphene, carbon nitride and boron nitride, yielding twenty candidates that bind hydrogen in the ideal range and seven that are very nearly perfect. Machine learning models trained on those calculations then showed that a candidate’s performance can be anticipated from simple electronic and geometric properties, without repeating the full calculation each time.