Metal-Organic Framework Stability in Water and Harsh Environments from Data-Driven Models Trained on the Diverse WS24...

Area of Science:

• Materials Science
• Chemistry
• Computational Science

Background:

• Metal-organic frameworks (MOFs) are promising porous materials for gas separations and catalysis.
• Limited water stability hinders the practical application of many MOFs.
• Current design strategies for water-stable MOFs are often restrictive.

Purpose of the Study:

• To enhance the prediction of MOF water stability using machine learning.
• To expand the dataset for training machine learning models for MOF stability.
• To facilitate the discovery of novel, water-stable MOFs for various applications.

Main Methods:

• Curated a new dataset (WS24) by adding 911 MOFs with water stability labels via semiautomated manuscript analysis.
• Developed and improved machine learning models for predicting water and acidic stability.
• Integrated stability models with genetic algorithms for high-throughput screening of MOFs.

Main Results:

• Expanded the MOF water stability training data by over 400%.
• Achieved improved machine learning model performance (test ROC-AUC > 0.8) for predicting water and acidic stability.
• Identified metal- and geometry-specific design principles for robust MOFs.

Conclusions:

• The expanded dataset and improved ML models accelerate the discovery of water-stable MOFs.
• The developed tools enable rapid screening of large MOF libraries for multivariate stability.
• This work contributes to advancing MOF applications in areas like direct air capture and water treatment.