Massive Dataset Accelerates Metal-Organic Framework Screening for Direct Air Capture

Scientists have unveiled Open DAC 2025 (ODAC25), a monumental open dataset containing nearly 60 million density functional theory (DFT) single-point calculations, designed to revolutionize the search for efficient sorbent materials for direct air capture (DAC). This comprehensive resource addresses a critical bottleneck in the field: the scarcity of large, carefully curated training datasets required for sophisticated machine learning models. The ODAC25 dataset encompasses adsorption behaviors of carbon dioxide ($ ext{CO}_2$), water ($ ext{H}_2 ext{O}$), nitrogen ($ ext{N}_2$), and oxygen ($ ext{O}_2$) across approximately 15,000 different metal–organic frameworks (MOFs). By providing such an extensive and high-resolution computational foundation, researchers can now screen MOFs for DAC applications at a level of detail previously unattainable.

To maximize the versatility and utility of the ODAC25 dataset, the research team incorporated several advanced features into their computational modeling. The dataset includes the study of functionalized MOFs, the simulation of high-energy adsorbate placements, and the generation of synthetically derived framework structures. Furthermore, the calculations were not limited to static models; they applied engineering intuition regarding the complex phenomenon of adsorbate-induced MOF flexibility. The ODAC25 dataset significantly improves upon the accuracy of DFT calculations established in previous studies, providing a robust foundation for predictive modeling in materials science.

Leveraging this massive computational library, the researchers have also released state-of-the-art machine-learned interatomic potentials that were specifically trained on the ODAC25 data. The integration of high-fidelity DFT data with advanced machine learning techniques enables a powerful predictive cycle: researchers can now rapidly screen vast chemical spaces of MOFs, identifying promising candidates for DAC with unprecedented efficiency.

The ODAC25 dataset fundamentally changes this landscape by providing the necessary data volume and quality to train predictive algorithms effectively.

Subject of Research: The release of the ODAC25 dataset, a massive collection of computational data on gas adsorption in metal–organic frameworks, to advance direct air capture technology.

Article Title: An open dataset of gas adsorption in deformable metal–organic frameworks for direct air capture

Article References: Sriram, A., Brabson, L. M., Yu, X., Choi, S., Abdelmaqsoud, K., Moubarak, E., de Haan, P., Löwe, S., Brehmer, J., Shuaibi, M., Kitchin, J. R., Welling, M., Zitnick, C. L., Ulissi, Z., Medford, A. J., & Sholl, D. S. (2026). An open dataset of gas adsorption in deformable metal–organic frameworks for direct air capture. Nature Chemical Engineering. https://doi.org/10.1038/s44286-026-00445-5

Image Credits: AI Generated

DOI: 10.1038/s44286-026-00445-5

Keywords: Metal-Organic Framework, Direct Air Capture, Machine Learning, DFT, Gas Adsorption, Environmental Science, Chemical Engineering, Sustainable Technology, open, dataset, adsorption, deformable