AI for Complex Catalytic Systems: High-Entropy Alloys in Electrocatalytic Acetylene Semihydrogenation

AI-driven big data set analysis offers opportunities for theoretical research on systems that combine complex catalyst materials with intricate catalytic reactions. In this study, we explore high-entropy alloys (HEAs) as potential electrocatalysts for the electrochemical semihydrogenation of acetylene. HEAs provide a variety of surface active sites due to the diverse combination of constituent elements, while the presence of bidentate dicarbon species further complicates surface interactions. By integrating density functional theory computations, geometric optimizer development, and graph neural network-based machine learning predictions, we efficiently compile a comprehensive database of 52,900 adsorption properties for AgAuCuNiPd HEA surfaces. Lasso regression and t-SNE projection reveal the distinct influences of the five metal components on adsorption and reaction properties. Using Cu as a reference, logistic regression assesses the potential for other components to surpass Cu in terms of catalytic activity and selectivity toward ethylene. Our findings suggest that, while HEAs can enhance the reaction, the ternary AgAuCu alloy achieves optimal results, indicating that high entropy is not essential. This research methodology can be extended to other complex catalytic systems, providing valuable insights into catalytic mechanisms and facilitating experimental endeavors.