Integrating Molecular Simulations with Machine Learning to Discover Selective MOFs for CH₄/H₂ Separation.

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• The growing number of metal-organic frameworks (MOFs) presents challenges in identifying optimal materials for gas separation.
• Efficient methane/hydrogen (CH4/H2) separation is crucial for natural gas purification and hydrogen energy applications.

Purpose of the Study:

• To develop and apply an integrated molecular simulation and machine learning approach for evaluating CH4/H2 separation performance across a vast library of MOFs.
• To rapidly screen a large dataset of synthesized and hypothetical MOFs to identify high-performance adsorbents.

Main Methods:

• Grand canonical Monte Carlo (GCMC) simulations were employed to generate CH4 and H2 adsorption data for MOFs.
• Machine learning models were trained using structural, chemical, and energetic features of MOFs derived from simulation data.
• ML models were extended to hypothetical MOFs for high-throughput virtual screening.

Main Results:

• The study evaluated 126,605 distinct MOFs for their CH4/H2 separation capabilities.
• High selectivities were observed in synthesized MOFs with narrow pores and specific linkers (pyridine, histidine, imidazole).
• Hypothetical MOFs with narrow pores and carboxylate, benzoate, or cubane-based linkers showed even higher selectivities.

Conclusions:

• The integrated ML and simulation approach enables efficient identification of promising MOF adsorbents for CH4/H2 separation.
• MOFs with tailored pore sizes and linker functionalities offer superior performance compared to traditional adsorbents.
• This work paves the way for accelerated discovery of advanced materials for critical gas separation processes.