From surface area to functionality: data-driven insights into MIL-100(Fe) synthesis for enhanced dye removal...

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Traditional metal-organic framework (MOF) design often prioritizes generic metrics like BET surface area and crystallinity, assuming universal performance prediction.
• This approach may not yield optimal results for specific applications, necessitating a more targeted synthesis strategy.

Purpose of the Study:

• To develop and apply a machine learning-guided optimization framework for MIL-100(Fe) synthesis.
• To link synthesis parameters to key material properties and application-specific performance (methylene blue removal).

Main Methods:

• Experimental synthesis of MIL-100(Fe) using an acid-free, water-based hydrothermal route.
• Training small-data machine learning models (Gaussian process regression) to correlate synthesis parameters (temperature, time, metal-to-ligand ratio, ion concentration) with material properties (surface area, pore volume, crystallite size, crystallinity, yield) and methylene blue (MB) removal efficiency.
• Utilizing SHAP analysis to identify key synthesis parameters influencing specific properties.
• Coupling Gaussian process regression with a genetic algorithm (GA) for property-specific synthesis optimization.

Main Results:

• SHAP analysis revealed that synthesis time and metal-to-ligand ratio were dominant factors for dye removal, while temperature and time influenced surface area.
• The optimized MIL-100(Fe) for MB removal achieved 98.3% removal efficiency, a significant improvement over the baseline.
• The MB removal-optimized sample demonstrated higher adsorption capacity despite a lower surface area and crystallinity compared to the BET-optimized sample, highlighting that surface area alone does not govern performance.
• Optimized samples for different targets occupied distinct regions in the descriptor space, indicating the need for application-specific tailoring.

Conclusions:

• Machine learning-guided optimization provides a powerful framework for tailoring MOF synthesis to specific applications.
• Optimizing MOF synthesis for a target application, such as dye removal, is more effective than maximizing generic metrics like surface area.
• The study demonstrates that MIL-100(Fe) synthesis can be precisely controlled to achieve desired performance characteristics, moving beyond traditional optimization approaches.