As authors, we aim to address one of the most pressing challenges in sustainable energy: how to accelerate the discovery of efficient electrocatalysts for green hydrogen production.

Key Highlights

• Paradigm Shift in Catalyst Design: Traditional electrocatalyst discovery relies on costly experiments and DFT simulations. This review highlights graph neural networks (GNNs) as a transformative approach, enabling direct learning from atomic graphs without manual feature engineering.
• Comprehensive Methodological Framework: The article systematically categorizes GNN architectures—invariant, equivariant, and Transformer-based models—and compares them with traditional machine learning approaches, emphasizing their superior ability to capture complex geometric and topological information.
• Applications in HER and OER: GNNs have been successfully applied to design electrocatalysts for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER), identifying high-performance non-precious metal catalysts, metastable IrO₂ phases, and efficient single-atom catalysts.
• Integration with AI and Databases: Open Catalyst datasets (OC20/OC22) and physics-informed GNN models are accelerating high-throughput screening, reducing computational costs by over tenfold compared to DFT.
• Future Outlook: The review calls for physics-informed GNNs, multi-objective optimization frameworks, and closed-loop R&D systems integrating computational prediction, automated synthesis, and real-time feedback.

Significance

• Accelerating Green Hydrogen Technology: By reducing reliance on expensive experiments and simulations, GNN-driven design can significantly shorten the development cycle for electrocatalysts, supporting global decarbonization goals.
• Bridging AI and Physical Sciences: This work exemplifies how AI for Science can integrate deep learning with fundamental physics, paving the way for rational catalyst design and autonomous discovery.

Authors

Wenhao Dong, Qi Wang, Liping Ren, Jinjia Wei, Shaohua Shen*, Jie Chen*

Affiliations: Xi’an Jiaotong University; State Key Laboratory of Multiphase Flow in Power Engineering; Xi’an Jiaotong University Suzhou Academy