Design, Synthesis, and Screening of COFs for CO₂ Adsorption by Gaussian Process

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Effective carbon dioxide (CO2) capture is crucial for climate change mitigation.
• Covalent Organic Frameworks (COFs) show promise for CO2 adsorption, but their design is complex.
• Existing computational methods often overlook vital chemical and synthetic factors influencing real-world COF performance.

Purpose of the Study:

• To develop an integrated computational-experimental framework for predicting CO2 adsorption in COFs.
• To combine machine learning with experimental validation for enhanced COF design.
• To identify key structural and synthesis parameters influencing COF CO2 adsorption capacity.

Main Methods:

• Analysis of 240 unique COFs (617 samples) with experimentally measured CO2 adsorption data.
• Development of Gaussian Process (GP) and CatBoost machine learning models to predict CO2 adsorption.
• Incorporation of chemical structures, synthesis parameters, and measurement protocols into predictive models.
• Utilized SHAP analysis to interpret model predictions and identify influential features.

Main Results:

• The GP model exhibited superior generalization and uncertainty quantification compared to the CatBoost model.
• SHAP analysis highlighted the importance of COF type and synthesis conditions in adsorption prediction.
• Generated 5557 potential COF structures from 181 building blocks with recommended synthesis conditions.
• Experimental validation confirmed the framework's predictive accuracy for newly synthesized COFs.

Conclusions:

• The integrated computational-experimental framework provides a practical approach for optimizing COF design for CO2 capture.
• Machine learning models, particularly GP, can effectively predict CO2 adsorption by considering diverse chemical and synthesis factors.
• This framework accelerates the discovery and development of high-performance COFs for carbon capture applications.