A research team led by Professor Baek Se-in at Korea University and Professor Ku Geun-ho at the Korea Institute of Energy Technology, has developed artificial intelligence (AI) for exploring promising materials and proposed a strategy for uncovering high-performance electrochemical water-splitting catalysts.
As the generation of renewable energy has recently expanded, the demand for electrolysis for the production of environmentally friendly hydrogen is increasing. In particular, water splitting through the oxygen evolution reaction (OER) is gaining attention; however, the precious metal catalysts commonly used, such as iridium and ruthenium, have drawbacks due to their high cost and scarcity.
To address this issue, the research team focused on perovskite-based catalysts capable of various elemental combinations. After selecting about 6,500 structural data points, they developed an AI model to predict stability based on the constituent elements of the materials and a neural network model to predict activity based on interatomic bonding information. They also performed simulations to secure training data.
Using the trained model, the research team predicted catalyst stability in high-voltage and highly acidic environments. Furthermore, they adjusted the neural network model with the generated data and analyzed the activity on approximately 14,000 catalyst surfaces with high accuracy. As a result, they proposed 15 candidate catalysts that induce oxygen generation better than existing oxygen evolution reaction catalysts.
Professor Baek noted, "Efficient and rapid high-performance material discovery using AI is being applied not only in the field of electrochemical catalysts but also in various environmentally friendly material sectors. This will be key to the transition to a renewable energy and green hydrogen economy."
This research was published online in the international journal of chemical engineering, the Chemical Engineering Journal, on June 23, 2025.
References
Chemical Engineering Journal(2025), DOI: https://doi.org/10.1016/j.cej.2025.165258