|
High-entropy alloy electrocatalysts screened using machine learning informed by quantum-inspired similarity analysis: computational prediction and experimental synthesis
Y. Chang, I. Benlolo, Y. Bai, C. Reimer, D. Zhou, H. Zhang, H. Matsumura, H. Choubisa, X.-Y. Li, W. Chen, P. Ou, I. Tamblyn, and E.H. Sargent Matter, 7, 11, 3728-3755 (2024) Discovering new catalyst materials for clean energy applications typically requires an enormous number of expensive quantum mechanical calculations, creating a bottleneck in materials discovery. In this work, we combine machine learning with a novel quantum-inspired similarity analysis to dramatically accelerate the screening of high-entropy alloy electrocatalysts. By incorporating information about the similarity of different atomic adsorption sites into a graph neural network, we cut the required training data in half while maintaining prediction accuracy. The computational predictions were validated experimentally, successfully identifying a new five-metal alloy catalyst for the oxygen reduction reaction that outperforms the industry-standard platinum-on-carbon catalyst, demonstrating how physics-informed machine learning can speed up the discovery of new catalytic materials for fuel cells and other electrochemical technologies. |
