Domestic researchers have developed a generative artificial intelligence (AI) model that produces desired porous materials. This is expected to significantly enhance the possibility of integrating AI and material science by solving the challenges in the field of porous materials.
A research team led by Professor Kim Ji-han from the Korea Advanced Institute of Science and Technology (KAIST) announced on the 23rd that it has developed an AI model called "MOFFUSION" that generates metal-organic frameworks (MOFs) with desired physical properties. The results of this research were published in the international journal "Nature Communications" on the 2nd.
Generative AI is an AI model that directly creates new data, including large-scale language models like ChatGPT and image generation software DALL·E. However, it has not yet been sufficiently utilized in the field of material development. In particular, the complex structure of porous materials has made it challenging to apply generative AI.
To express the structure of metal-organic frameworks more efficiently in the AI model, the researchers adopted a method that uses three-dimensional modeling techniques to represent their pore structures. This approach significantly increased the structural generation efficiency to 81.7%, surpassing the generation efficiencies reported by other existing models.
Users can input the desired material characteristics into the AI model in various forms such as numbers, categories, and text. For example, if a user inputs a textual description of the desired material's property, such as "a structure with a hydrogen adsorption amount of 30g/L," the model selectively generates the corresponding material. The research team noted that this greatly improves the usability and convenience of AI models in material development.
Professor Kim Ji-han stated, "Developing materials with desired properties is the greatest goal and a long-standing research topic in the field of materials," adding that "the technology developed this time will promote the adoption of generative AI in the field of porous material development."
Reference material
Nature Communications (2025), DOI: https://doi.org/10.1038/s41467-024-55390-9