Researchers from the University of Toronto, led by Alex Hernandez and Alan Aspuru-Guzik, have developed a new machine learning tool called Catalyst GFlowNet to aid in the design of efficient and affordable catalysts for energy storage applications. The team’s work, published in the journal Nature Communications, focuses on improving hydrogen energy storage, a crucial technology for integrating renewable energy sources into the grid.
The researchers explain that electrocatalysts are vital for converting electrical energy into chemical energy, enabling the storage of hydrogen as a fuel. However, finding cost-effective and high-performance catalysts has been a persistent challenge. To address this, the team created Catalyst GFlowNet, a generative model that uses machine learning to predict the formation and adsorption energy of crystal surfaces, ultimately designing more efficient catalysts.
In a proof-of-concept study, the researchers applied Catalyst GFlowNet to the hydrogen evolution reaction, a key process in hydrogen energy storage. The model successfully identified platinum as the most efficient known catalyst for this reaction. While platinum is already known to be an excellent catalyst, this validation demonstrates the potential of the model to accelerate the discovery of new and improved catalysts.
Looking ahead, the researchers plan to extend their approach to the oxygen evolution reaction, where current optimal catalysts are expensive metal oxides. By broadening the search space, they hope to uncover new materials that could make hydrogen energy storage more affordable and accessible. This innovative framework offers a promising avenue for the energy industry to develop novel and efficient catalysts, ultimately supporting the transition to a more sustainable energy future.
Source: Hernandez, A., et al. (2023). Catalyst GFlowNet for electrocatalyst design: A hydrogen evolution reaction case study. Nature Communications.
This article is based on research available at arXiv.