From prediction to materials design: machine learning in electrocatalytic water splitting.

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Machine learning (ML) shows promise for accelerating electrocatalyst discovery in water splitting.
• Current ML applications are hindered by data quality, thermodynamic bias, and limited experimental validation.

Purpose of the Study:

• To survey recent ML-guided efforts in oxygen and hydrogen evolution reactions.
• To highlight ML applications on complex material systems like layered double hydroxides and metal-organic frameworks.
• To identify limitations and propose future directions for ML in electrocatalyst design.

Main Methods:

• Review of supervised models, ML-density functional theory (DFT) workflows, and interpretable surrogates.
• Analysis of ML applications in composition screening, dopant optimization, and descriptor discovery.
• Identification of bottlenecks in datasets, theoretical models, and experimental correlation.

Main Results:

• ML is used for rapid screening and optimization of electrocatalyst compositions.
• Non-linear descriptors beyond traditional volcano plots are being discovered.
• Challenges remain in dataset size, kinetic/stability representation, and bridging theory-experiment gaps.

Conclusions:

• Integrating active learning, operando data, and explainable AI can create closed-loop, synthesis-aware workflows.
• ML can serve as a mechanistically meaningful tool for designing robust, earth-abundant water-splitting electrocatalysts.
• Addressing current limitations will enhance ML's impact on sustainable energy technologies.