Deep learning and big data mining for Metal-Organic frameworks with high performance for simultaneous desulfurization...

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Flue gas treatment requires efficient methods for carbon capture and desulfurization to achieve carbon neutrality.
• Metal-organic frameworks (MOFs) show promise for adsorptive separation of CO2 and SO2.
• Developing high-performance MOFs necessitates advanced screening and predictive tools.

Purpose of the Study:

• To develop a "one-step" adsorption technology for simultaneous SO2 and CO2 capture using MOFs.
• To employ machine learning for predicting MOF performance in flue gas treatment.
• To identify key MOF properties influencing adsorption and to evaluate dynamic separation performance.

Main Methods:

• Utilized four machine learning algorithms (including Deep Factorization Machines) to predict MOF performance indicators.
• Performed sensitivity analysis to determine critical MOF descriptors for SO2 and CO2 capture.
• Developed a kinetic model to simulate breakthrough curves for dynamic adsorption separation of SO2/CO2/N2 mixtures.

Main Results:

• Deep Factorization Machines achieved a high prediction accuracy (R² = 0.95) for MOF performance.
• Adsorption heat, porosity, and open alkali metal sites were identified as key factors for SO2 and CO2 capture.
• Screening of 1000 MOFs identified 20 top candidates with significant overlap between static and dynamic performance predictions.

Conclusions:

• The integrated approach of computational screening, machine learning, and dynamic analysis accelerates the development of efficient MOF adsorbents.
• The identified MOFs demonstrate potential for simultaneous carbon capture and desulfurization in flue gas treatment.
• This work advances sustainable development goals through innovative materials for environmental remediation.