Machine Learning-Assisted Discovery of Propane-Selective Metal-Organic Frameworks

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Metal-organic frameworks (MOFs) are increasingly utilized in materials discovery and prediction.
• Efficient separation of propane/propylene (C3H8/C3H6) is crucial for industrial applications.
• Computational screening of existing MOFs can accelerate the identification of suitable materials.

Purpose of the Study:

• To develop a machine learning-assisted strategy for screening C3H8-selective MOFs.
• To identify high-performance MOFs for propane/propylene separation from the CoRE MOF database.
• To experimentally validate the performance of promising candidate MOFs.

Main Methods:

• Applied machine learning algorithms, including random forest (RF), to screen the CoRE MOF database.
• Evaluated MOFs based on predicted C3H8 selectivity.
• Experimentally verified the separation performance and hydrolytic stability of the top-performing MOF.

Main Results:

• The random forest algorithm achieved the highest accuracy in predicting MOF performance.
• The JNU-90 MOF was identified as a top-performing material for C3H8/C3H6 separation.
• JNU-90 exhibited benchmark selectivity and excellent hydrolytic stability, enabling direct C3H6 production.

Conclusions:

• Machine learning provides an effective strategy for accelerating the discovery of MOFs for challenging separations.
• JNU-90 shows significant potential for energy-efficient propane/propylene separation.
• This approach can expedite the development of advanced MOFs for industrial applications.