AI-Powered Experimental Discovery of Metal-Organic Frameworks for n/i-Butane Separation

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Separating n-butane and i-butane typically relies on energy-intensive distillation.
• Metal-organic frameworks (MOFs) show promise as efficient adsorbents but discovering optimal MOFs is challenging.
• High-throughput screening is needed to navigate the vast chemical space of MOFs.

Purpose of the Study:

• To accelerate the discovery of high-performance MOFs for n/i-butane separation using artificial intelligence (AI).
• To develop and validate an integrated descriptor system for MOF performance prediction.
• To demonstrate an efficient AI-driven material discovery paradigm.

Main Methods:

• Employed high-throughput screening integrated with artificial intelligence (AI) for MOF identification.
• Developed and validated an integrated descriptor system accessible via experiments and simulations.
• Implemented a similarity-based dataset optimization strategy for efficient AI model training.
• Utilized a neural network model to identify MOFs with superior n/i-butane separation capabilities.

Main Results:

• An integrated descriptor system outperformed existing descriptors in predicting MOF performance.
• AI model training was optimized, requiring only 10% of the database, significantly reducing costs.
• Exceptional MOFs for n/i-butane separation were identified using the AI approach.
• Synthesized SIFSIX-3-Zn demonstrated outstanding separation performance with minimal i-butane uptake.

Conclusions:

• AI-driven high-throughput screening offers an efficient paradigm for discovering advanced materials like MOFs.
• The integrated descriptor system and optimized training strategy enhance the speed and cost-effectiveness of material discovery.
• This approach presents a broadly applicable process for identifying MOFs with specific separation functionalities.