Integrated Machine Learning with Molecular Simulation for Screening MOF Adsorbents toward High-Performance Separation...

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Existing machine learning (ML) models show limited accuracy for predicting CO2/CH4 separation performance in metal-organic frameworks (MOFs).
• Efficient CO2 capture and methane purification are critical for environmental and energy applications.

Purpose of the Study:

• To develop high-performance ML models for CO2/CH4 separation using MOFs.
• To identify key MOF structural features governing CO2/CH4 adsorption selectivity.

Main Methods:

• Integrated high-throughput screening with molecular simulations and a novel four-layer ensemble ML model.
• Employed Shapley Additive Explanations and Partial Dependence Plots to elucidate structure-performance relationships.

Main Results:

• Successfully screened MOFs for high-performance CO2/CH4 adsorption separation.
• Identified pore-limiting diameter (optimal 3.7-4.2 Å), metal type, and accessible surface area as priority factors.
• Found helical channel topology and nitrogen-containing MOFs with >30% oxygen to be advantageous for CO2 adsorption.

Conclusions:

• The developed ML approach significantly enhances predictive performance for MOF-based CO2/CH4 separation.
• Structure-based design principles were elucidated, guiding the selection of optimal MOFs.
• This work provides a pathway for designing advanced MOFs for efficient gas separation applications.