NCHU team uses AI and high-entropy materials to accelerate green hydrogen catalyst development, featured on ACS Catalysis cover

A cross-disciplinary team led by  Dean Ming-Der Yang (third from left) at National Chung Hsing University integrated artificial intelligence with high-entropy materials to accelerate catalyst development, with their work featured on the cover of ACS Catalysis.

A cross-disciplinary research team from National Chung Hsing University (NCHU) has combined artificial intelligence with high-entropy materials to significantly accelerate the development of catalytic materials for green hydrogen production, with their findings featured on the cover of the journal ACS Catalysis.

The study, led by College of Engineering Dean Ming-Der Yang, Associate Dean Chih-Ming Chen and Assistant Professor Hung-Chung Li, demonstrated that an AI-driven optimization model could reduce the time required for catalyst development by approximately 99.3%, compared with conventional experimental screening methods.

As countries push toward net-zero emissions and energy transition goals, hydrogen energy has emerged as a key solution. However, identifying efficient catalyst materials for hydrogen production remains a major bottleneck.

High-entropy materials — considered a breakthrough in materials science — involve complex combinations of multiple metallic elements. While they hold promise as alternatives to precious metals in applications such as water electrolysis, the vast number of possible compositions makes it difficult to identify optimal formulations through traditional trial-and-error approaches.

To address this challenge, the NCHU team developed a machine learning-based optimization model integrating catalyst synthesis techniques. The research was led by postdoctoral researcher Chandrasekaran Pitchai and graduate student Chao-Fang Huang.

Using an XGBoost algorithm, the model analyzed optimal ratios of five metal elements — iron, cobalt, chromium, manganese and copper — and rapidly evaluated the performance of 10,626 possible high-entropy material combinations. The model achieved a prediction error rate of around 3%.

The researchers applied the model to layered double hydroxides (LDHs), a class of materials with unique structures, and demonstrated significantly improved efficiency in electrochemical water splitting for hydrogen production.

Unlike many AI studies that rely on large datasets from multiple sources, the team trained their model using 70 data points generated entirely within their own laboratory. This approach reduced uncertainties arising from inconsistencies across literature data and improved the reliability of the predictions.

“This study highlights the potential of integrating advanced materials with artificial intelligence to accelerate the design, development and application of high-entropy materials,” Chen said, noting its implications for industrial scalability.

Lee added that even small datasets can reveal critical features in complex material systems, underscoring the practical potential of AI-driven approaches.

Yang said the project was supported by Taiwan’s National Science and Technology Council under its AI research initiative. The team plans to further integrate AI with advanced materials and smart agriculture, exploring the use of renewable energy — such as solar panels installed on farms or greenhouses — to produce hydrogen via water electrolysis.

Such systems could provide a zero-carbon, sustainable energy solution, particularly for remote areas and resilient agricultural applications, he added.

The research paper is available at: https://doi.org/10.1021/acscatal.5c07303