【Frontiers in HaiChuan Chemical Technology】Progress achieved at Dalian Institute of Chemical Physics in using artificial intelligence for catalyst synthesis

Thread Content

I was invited to write a Perspective article on the use of artificial intelligence in catalyst synthesis. Publication date: 2025-05-27. Recently, the research team led by Researcher Deng Dehui from the Center for Energy and Environmental Small Molecule Catalysis (Group 509) at our institute, in collaboration with Dr. Li Haobo’s team from Nanyang Technological University in Singapore, was invited to write a Perspective article on artificial intelligence in catalyst design and synthesis. This article provides a comprehensive overview of the research progress in using artificial intelligence for the design and synthesis of heterogeneous catalysts, while also offering insights into the future directions of this field. As an indispensable key material in fields such as chemical engineering, energy, and environmental protection, the traditional development of catalysts relies on experience-driven trial-and-error methods, which suffer from issues such as long cycles, high costs, and poor data reproducibility. There is an urgent need to explore new research paradigms to achieve efficient and precise design. In recent years, the development of artificial intelligence, particularly machine learning*, and its successful application in various fields have brought significant opportunities to research in the field of catalyst design and synthesis, driving this field toward a data-driven, automated, and intelligent direction. This review systematically summarizes the latest advancements in artificial intelligence technologies within the field of catalyst design and synthesis. It elaborates on the significant applications of machine learning methods in predicting catalyst structures and properties, optimizing synthesis conditions, as well as conducting high-throughput automated experiments and characterization ； A new strategy is proposed to effectively identify the key descriptors affecting catalyst performance through machine learning*, thereby reducing the costs of traditional theoretical calculations (such as density functional theory calculations) and improving research efficiency ； It focuses on how cutting-edge technologies such as active learning* and generative models enable more efficient and precise catalyst design and synthesis processes ； Special emphasis was placed on the core value of AI-driven closed-loop automated synthesis systems in improving the data accuracy and reproducibility throughout the entire process of catalyst design, synthesis, characterization, and optimization, as well as in reducing human errors ； Finally, it explores in depth the opportunities and challenges in this field regarding the generalization capability of catalytic systems, cross-domain data integration and sharing, as well as the ability to automatically identify and analyze anomalies in experiments. It offers forward-looking perspectives on corresponding solutions and paths for technological development, providing a reference for advancing the catalysis field toward a higher level of automated and intelligent research paradigms. The Dundee team has been dedicated for a long time to the regulation of the surface properties of two-dimensional catalytic materials, as well as to the fundamental scientific research and industrial applications related to their use in the catalytic conversion of small molecules in energy and environmental fields. In recent years, the team has made a series of advances in machine learning*, artificial intelligence catalysis laboratories, and database development. A related article titled “Artificial intelligence for catalyst design and synthesis” was recently published in Matter upon invitation. The above work was supported by projects such as the **Key Research and Development Program**, the National Natural Science Foundation, and Nanyang Technological University’s Research Start-up Fund.