Machine learning accelerates catalytic applications of 2D materials

Two-dimensional (2D) materials have shown extraordinary potential in electrocatalytic reactions due to their unique structural and electronic properties. In a new review published in AI Mater., first author Chenchen Qi and colleagues from China University of Petroleum (East China), together with international collaborators, highlight how machine learning (ML) is revolutionizing the discovery, design, and performance prediction of 2D catalytic materials, providing a faster, data-driven pathway toward sustainable energy technologies.

Two-dimensional materials, such as transition metal dichalcogenides, MXenes, and graphdiyne, possess large specific surface areas, tunable electronic structures, and abundant active sites, making them promising candidates for electrocatalytic reactions including hydrogen evolution, oxygen evolution, nitrogen reduction, and carbon dioxide reduction. However, identifying optimal catalysts among vast compositional possibilities has long been a major challenge for materials scientists.

To address this issue, Chenchen Qi, a graduate researcher and the first author of the study, conducted an in-depth synthesis of recent advances in ML-assisted catalysis. Working under the supervision of Prof. Juhong Yu and Prof. Shiyu Du from the China University of Petroleum (East China), Qi collaborated with Prof. Yong Liu of the University of Colorado Denver, and researchers from Tongji University and Milky-Way Sustainable Energy Ltd.

The review, titled “Catalytic applications of 2D materials aided by machine learning,” systematically summarizes how ML methods—from supervised and unsupervised learning to deep neural networks and interpretable ML—have been applied to accelerate the screening and optimization of 2D electrocatalysts. “Machine learning enables us to uncover hidden correlations between structure and catalytic performance that were previously inaccessible through conventional approaches,” says Prof. Shiyu Du.

By integrating large materials databases, feature engineering, and algorithmic modeling, ML can predict key catalytic descriptors such as adsorption energy, d-band center, and charge transfer. These insights allow researchers to pinpoint the most efficient catalyst compositions and active sites for specific reactions, dramatically reducing experimental cost and time.

According to the authors, the convergence of ML and materials science marks a paradigm shift—from traditional trial-and-error experimentation to intelligent, data-driven catalyst design. The study emphasizes the importance of high-quality data, descriptor selection, and interpretable modeling to enhance the reliability and generalization of ML predictions.

“Data-driven models are reshaping how we design functional materials,” explains Prof. Juhong Yu. “Through ML, we can accelerate the discovery of sustainable catalysts for clean energy conversion and environmental protection.”

The review concludes that combining ML algorithms with first-principles simulations and experimental validation will further advance rational catalyst design, ultimately paving the way for efficient, low-cost, and environmentally friendly energy technologies.

Paper Information
Qi C, Wang W, Lv K, Li D, Liu Y, et al.Catalytic applications of 2D materials aided by machine learning. AI Mater. 2025(2):0014. https://doi.org/10.55092/aimat20250014