Proceedings of MATSUS Spring 2025 Conference (MATSUSSpring25)
DOI: https://doi.org/10.29363/nanoge.matsusspring.2025.556
Publication date: 16th December 2024
Finding affordable, durable, and efficient catalysts is crucial for advancing green hydrogen production and carbon dioxide conversion, both vital for mitigating climate change. Currently, the discovery of new catalysts is hindered by the gap between computational predictions and experimental outcomes. Traditional catalyst discovery is a slow, trial-and-error process relying on decades of expertise. Analyzing a single material takes time, so different parts of the material design space are often studied separately by various research groups. This results in reproducibility challenges, making it difficult to build on previous findings and draw comprehensive conclusions. Progress requires large, diverse experimental datasets that are reproducible and tested under industrially relevant conditions to bridge the gap in discovery. In my talk, I will highlight the development of the Open Catalyst Experiments OCx24 dataset, featuring more than 500 samples synthesized and characterized by XRF and XRD and tested on gas diffusion electrodes in zero-gap electrolysis for CO2 reduction and hydrogen evolution. To find correlations with experimental outcomes and to perform computational screens, we calculated DFT-verified adsorption energies for six adsorbates on approximately 20,000 inorganic materials, requiring 685 million AI-accelerated relaxations.