Smarter search for fuel cell catalysts using machine learning

Researchers at Science Tokyo have reported that a computational method that combines production AI and atomic simulation can identify a promising platinum alloy catalyst structure for hydrogen fuel cells. Their approach addresses long-standing challenges in catalyst design and consistently generates high-performance candidates from several material combinations.

Polymer electrolyte fuel cells (PEMFCs) are a promising clean energy technology that can generate electricity by combining hydrogen and oxygen with water. However, their performance relies heavily on a chemical step known as the oxygen reduction reaction (ORR), which requires efficient catalysts to react at practical rates. Platinum (Pt) remains the standard ORR catalyst in PEMFCs due to its excellent electrochemical properties, but its high cost and scarcity are barriers to large-scale adoption. As a result, researchers turned to platinum-based alloys as a cheaper alternative while maintaining strong catalytic performance.

However, designing these alloy-based catalysts is far from straightforward. Due to the large number of possible atomic arrangements in alloy materials, it is impractical to test all candidates through experiments or computational techniques such as density functional theory. At the same time, catalysts must meet multiple requirements. They must be highly responsive to ORR, but also stable under real operating conditions. Most machine learning-based approaches address these properties separately and therefore lack the ability to propose atomic structures that simultaneously satisfy both criteria. How can I search for the right alloy design more efficiently?

In a recent study, Associate Professor Atsushi Ishikawa of the Department of Environmental Sociology at Tokyo University of Science, along with graduate student Yasushiro Wakamiya, developed a new strategy to address this challenge. Their works published in magazines npj calculation materials introduced a method on April 14, 2026 that combines atomistic simulation and generative artificial intelligence to design alloy catalysts for ORR.

The proposed approach relies on two important tools. The first is the Neural Network Potential (NNP) model. It is a machine learning model trained on quantum mechanical calculations that can quickly estimate key material properties. The second is a generative model known as a conditional variational autoencoder (CVAE), which can suggest new atomic structures based on desired properties. In this case, the model was trained to target both low overpotentials (a measure of catalytic activity) and low alloying energies (a measure of stability).

The workflow operates as an iterative loop, where the NNP model evaluates the performance of the proposed alloy and CVAE refines them and feeds them back to the NNP stage. By repeating this process multiple times, the alloy gradually migrates to a higher performing configuration. Application of this method to a Pt-nickel alloy produced structures that simultaneously satisfied the overpotential and formation energy criteria. Remarkably, this model rediscovered known design principles by itself, such as how a platinum-rich surface layer can enhance ORR activity.

Additionally, the team demonstrated the versatility of the workflow by extending it to multiple alloy systems such as Pt-titanium and Pt-yttrium. “This study demonstrated that the combination of atomic calculations and CVAE provides a general computational screening method that can generate new alloy surface structures that meet both activity and stability criteria from limited initial data,” explains Ishikawa.

The researchers believe their framework could have a wide range of applications beyond fuel cell catalysts. “The newly developed workflow has the potential to be applied to a wide range of materials challenges, including water electrolysis for hydrogen production, battery electrode materials, and catalysts for chemical processes,” concludes Ishikawa.

This research could help accelerate the development of sustainable energy technologies by enabling faster and more targeted exploration of complex material spaces.

***

About Tokyo University of Science (Science Tokyo)

Tokyo University of Science (Science Tokyo) was established on October 1, 2024 through the merger of Tokyo Medical and Dental University (TMDU) and Tokyo Institute of Technology (Tokyo Institute of Technology), with the mission of “promoting science and human well-being and creating value for society and society.”