Presentation Information
[2P05]Composition optimization of high-entropy alloy catalysts using machine learning
*Koki Otsuka1, Koji Shimizu1,2, Anh Khoa Augustin Lu1,3, Satoshi Watanabe1 (1. Graduate School of Engineering, The University of Tokyo, 2. Materials DX Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Japan, 3. Research Center for Materials Nanoarchitectonics, National Institute for Materials Science (NIMS), Japan)
High-Entropy Alloys (HEAs) are promising as catalysts, but designing optimal compositions is challenging due to the vast number of possibilities. The present research addresses this challenge by introducing a machine learning (ML) framework for efficient catalyst design. Our framework utilizes an improved graph neural network to predict key catalyst properties. This model guides the search for optimal compositions, which is performed efficiently using Monte Carlo sampling and directed by Bayesian optimization. For the hydrogen evolution reaction, our framework successfully identified compositions containing Palladium (Pd) and Cobalt (Co) as top candidates, which aligns with experimental results. Furthermore, multi-objective optimization balanced catalyst activity and stability to identify promising candidates.