MOFs: Chemistry Nobel 2025
Kitagawa, Robson and Yaghi receive the 2025 Chemistry Nobel Prize for MOFs: crystals with huge internal cavities for gas, water and CO2.
Open in the interactive tree →MOFs (metal-organic frameworks) are building kits of metal nodes and organic linkers that form porous crystals with internal surfaces of thousands of square meters per gram. Robson laid out the concept in 1989, Kitagawa showed gas uptake (1997), and Yaghi coined the name and made stable structures (1995, MOF-5 in 1999). Tens of thousands of different MOFs are now known.
As of now
Nobel Prize announced 8 October 2025. Applications are being built: Svante sells filters with the MOF CALF-20 for CO2 capture, BASF manufactures MOFs industrially (Basolite C300), and prototypes harvest drinking water from desert air (about 1.8 liters per kg per day with solar heat for CAU-10-H; up to 17 liters per kg per day for a polymer-MOF prototype in humid conditions). What is needed: cheaper, more stable MOFs at tonne scale.
Open steps
- Low-cost synthesis at scale Medium AI leverageExtend water-based synthesis from CALF-20 to more MOF types at thousands of tonnes per year with cheap linkers.
- Shaping and stability in real gas Medium AI leverageForm MOFs into pellets, films or structured filters that keep capacity over thousands of humid, dusty, SO2- and NOx-laden cycles.
- From AI-designed MOF to real sample High AI leverageRaise the share of generated MOF candidates that can be made and keep their predicted pores; a 2025 preprint reported five AI-designed MOFs made in the lab.
- Matching MOFs to capture and water tasks High AI leveragePick MOFs for direct air capture and desert water harvesting by combining millions of adsorption calculations with humidity and regeneration-energy data.
Where AI could help
Medium AI leverage. AI can search millions of MOFs for stability, cost and performance, but tonne-scale synthesis and long-run cycling are engineering tests.
- Generate and rank MOFs for CO2 capture and water harvesting by stability, cost and uptake
- Machine-learned potentials that predict adsorption and humid-air stability cheaply
- Predict synthesis conditions and scale-up cost from precursors and past recipes
Shown so far
- In April 2025 a preprint described MOFGen, an agentic system combining a language model, a diffusion model and quantum-chemistry agents that generated hundreds of thousands of candidate MOFs, of which five were synthesized. source
- In August 2025 a preprint released ODAC25, nearly 60 million DFT calculations of CO2, H2O, N2 and O2 adsorption in 15,000 MOFs, with machine-learning models for direct-air-capture sorbent screening. source