Molecular Modeling-Based Machine Learning for Accurate Prediction of Gas Diffusivity and Permeability in...

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Gas diffusion in metal-organic frameworks (MOFs) is critical for applications like membrane separations but difficult to measure experimentally.
• Predicting gas diffusion accurately is essential for designing efficient MOF-based materials.

Purpose of the Study:

• To create an efficient computational framework for predicting gas diffusivities in a large number of MOFs.
• To evaluate the performance of MOF membranes for industrially relevant gas separations.
• To identify key material properties for designing next-generation MOFs.

Main Methods:

• Integrated high-fidelity molecular dynamics (MD) simulations with machine learning (ML) models.
• Trained ML models on MD data to predict diffusivities of CO2, N2, O2, CH4, and H2 in over 18,000 MOFs.
• Developed an interactive web interface for MOF diffusivity prediction and analyzed membrane separation performance.

Main Results:

• ML models accurately predicted gas diffusivities within minutes using accessible MOF structural and guest properties.
• Evaluated membrane-based separation performance for seven industrially important gas pairs.
• Identified top-performing MOF membranes with high selectivity and permeability.

Conclusions:

• The computational framework enables rapid prediction of gas diffusion in MOFs, facilitating material selection.
• Molecular fingerprinting revealed critical chemical properties for designing advanced MOF membranes.
• This approach accelerates the discovery of MOFs for efficient gas separation technologies.