Accelerating Discovery of Water Stable Metal-Organic Frameworks by Machine Learning

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Metal-organic frameworks (MOFs) offer versatile nanoporous materials for energy and environmental applications.
• Water stability remains a significant challenge limiting the practical use of many MOFs.

Purpose of the Study:

• To develop and validate a machine learning (ML) approach for predicting the water stability of MOFs.
• To accelerate the discovery of water-stable MOFs for practical applications.

Main Methods:

• Construction of the largest available database of MOF water stability (1133 MOFs).
• Development of ML classifiers using structural and chemical descriptors to predict water stability.
• Experimental synthesis and testing of two MOFs to validate ML predictions.

Main Results:

• ML classifiers demonstrated high prediction accuracy and transferability.
• Experimental validation confirmed the ML model's predictions.
• Application to the ARC-MOF database predicted ~130,000 water-stable MOFs out of ~280,000 candidates.

Conclusions:

• The ML approach effectively predicts MOF water stability, overcoming a key limitation.
• Identified 461 MOFs with combined water, thermal, and activation stability.
• The ML tool can streamline the exploration of stable MOFs for real-world applications.