The conventional development of catalysts relies heavily on trial-and-error approaches, which are time-consuming, costly, and prone to inconsistent data reproducibility. Transitioning to a data-driven, automated paradigm is imperative for advancing precision catalyst design.
In a Perspective article published in Matter, Prof. DENG Dehui’s group from the Dalian Institute of Chemical Physics (DICP) of the Chinese Academy of Sciences (CAS), in collaboration with Dr. LI Haobo’s group from Nanyang Technological University, systematically reviewed the transformative role of artificial intelligence (AI) in heterogeneous catalyst design and synthesis, outlining future directions for AI-driven innovations in this field.

Schematic illustration of the AI-driven catalyst design and synthesis paradigm, and the development of automated chemical synthesis techniques alongside the advancements in machine learning methods. (Image by ZHANG Longhai and BING Qiming)
The Perspective highlights machine learning (ML) as a powerful tool for predicting catalyst structure-property relationships, optimizing synthesis conditions, and enabling high-throughput automated calculations and experiments. By identifying key performance descriptors, ML significantly reduces dependence on resource-intensive theoretical calculations like density functional theory, accelerating catalyst discovery. Advanced techniques such as active learning and generative models further improve the efficiency of catalyst design by prioritizing critical experiments and proposing novel catalyst candidates.
A pivotal focus lies in the AI-powered closed-loop systems that integrate automated synthesis, characterization, and optimization. These systems enhance data accuracy, minimize human error, and ensure reproducibility across the entire catalyst development cycle. The challenges persist in generalizing AI models across diverse catalytic systems, integrating multidisciplinary datasets, and improving anomaly detection in automated workflows were pointed out. The authors propose technological roadmaps to address these limitations, emphasizing cross-institutional data sharing and adaptive AI frameworks.
“This Perspective provides a blueprint for transitioning catalysis research toward fully automated and intelligent paradigms, unlocking unprecedented efficiency in catalyst development,” said Prof. DENG.
Link:
https://www.dicp.ac.cn/xwdt/kyjz/202505/t20250527_7792008.html
https://doi.org/10.1016/j.matt.2025.102138