EPFL Pinpoints Key Step in Oxygen Evolution Reaction

Using advanced simulations and machine learning, researchers at EPFL have identified the details of the first key step in the oxygen evolution reaction. This reaction is a major bottleneck in the process of producing clean hydrogen.

Hydrogen is considered a clean fuel because burning it produces only water. It can store and deliver energy without carbon emissions.

One method for generating hydrogen is water splitting, which uses sunlight to separate water into hydrogen and oxygen. While this technique could make hydrogen more accessible, it remains inefficient despite years of research.

The Bottleneck in Hydrogen Production

The first step, called proton-coupled electron transfer (PCET), has long been a bottleneck in hydrogen production. In this step, a proton and an electron move together to help break apart water molecules. This makes PCET a key part of producing both hydrogen and oxygen.

The oxygen evolution reaction (OER) takes place at the surface where water meets a light-absorbing material, such as bismuth vanadate (BiVO₄). This reaction is slow. BiVO₄ helps remove electrons and protons from water molecules, leading to the formation of oxygen gas. It is a commonly used material for supporting this slow but important stage.

The Key Step: Proton-Coupled Electron Transfer

PCET, where a proton and an electron move together, has long been a key bottleneck in splitting water into hydrogen and oxygen. This step is essential because it helps break apart water molecules.

While the thermodynamics of PCET are better understood, its exact mechanism remains unclear. One challenge is the complex and disordered movement of water molecules on the surface of BiVO₄. Many previous studies have either overlooked this factor or used methods that lacked the time resolution or precision to capture it. As a result, researchers still have an incomplete understanding of how BiVO₄ works and how it might be improved.

Advanced Simulations Reveal the Dance

Yong-Bin Zhuang and Alfredo Pasquarello, two researchers at EPFL, have developed a new approach to better understand the problem. By combining machine learning algorithms, trained to simulate complex quantum calculations, with long-timescale molecular dynamics simulations, they were able to capture the full interaction between atoms and electrons at the BiVO₄–water interface.

Focusing on the first step of the OER, the initial PCET event, they found that the proton moves first, followed by the electron. This sequence determines the rate of the reaction.

The researchers built a detailed atomic model of the BiVO₄–water interface and used machine learning to predict atomic interactions accurately. This approach allowed them to run simulations for much longer periods (up to 30 nanoseconds) than standard quantum methods. It also enabled them to analyze hundreds of thousands of atomic configurations while maintaining stable and consistent results.

By closely tracking key features such as the proton’s location and the shifting position of the “hole” (the missing electron), they were able to observe the full PCET process. To ensure reliability, they used multiple independently trained machine learning models to cross-check their results.

What Makes Hydrogen Production Slow?

The calculations showed that the slowest, rate-limiting step is the direct transfer of a proton from a water molecule on the BiVO₄ surface to a nearby oxygen atom. The electron (or “hole”) moves only after the proton has shifted.

The study also found that this direct transfer dominates the process. It does not rely on an alternative pathway involving extra water molecules. This finding supports earlier research showing that proton transfer is the main bottleneck in the reaction.

With this confirmed, future efforts can focus on speeding up the proton transfer step. This might involve modifying the surface of BiVO₄ or using additives to stabilize key structures. The study also sets a new standard for simulating complex reactions at material interfaces. It shows that machine learning can balance computational cost with high accuracy.

The research was supported by the Swiss National Supercomputing Centre (CSCS).

Journal Reference:

Zhuang, Y.-B., et al. (2025). Mechanism of First Proton-Coupled Electron Transfer of Water Oxidation at the mathematical equation–Water Interface. Angewandte Chemie – International Edition. doi.org/10.1002/anie.202507071