Accelerated Screening of Zeolites for Methanol-to-Propylene Conversion Using Machine Learning with Interpretable...

Area of Science:

• Catalysis
• Materials Science
• Chemical Engineering

Background:

• Methanol-to-propylene (MTP) is a key industrial process for propylene production.
• Zeolite catalysts are crucial for MTP, but controlling propylene selectivity remains a challenge.
• Understanding MTP selectivity mechanisms is vital for catalyst development.

Purpose of the Study:

• To develop an interpretable function for assessing propylene/ethylene (P/E) selectivity in MTP.
• To implement a machine learning (ML) method for high-throughput screening of zeolite frameworks for MTP.
• To experimentally validate ML predictions and identify superior zeolite catalysts.

Main Methods:

• Density functional theory (DFT) and grand canonical Monte Carlo (GCMC) simulations were used to build the ML model.
• An interpretable function was developed to assess P/E selectivity.
• Machine learning facilitated high-throughput screening of zeolite frameworks.
• Experimental validation was performed using Ga-MFI and Ga-DDR catalysts.

Main Results:

• The ML model accurately predicted P/E selectivity, validated by experimental data.
• Ga-MFI catalyst achieved a P/E ratio of 12.3 and an operational lifespan of 121.3 hours.
• Ga-MFI significantly outperformed Ga-DDR in both P/E ratio and lifespan.
• The study confirmed the effectiveness of the zeolite screening strategy.

Conclusions:

• The developed interpretable function and ML method enable efficient screening of zeolite catalysts for MTP.
• Ga-MFI demonstrates exceptional performance for MTP, offering high propylene selectivity and stability.
• This work provides a pathway for designing advanced zeolite catalysts for improved MTP processes.