Predicting hydrogen storage in MOFs via machine learning.

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Hydrogen (H2) storage is critical for clean energy technologies.
• Metal-organic frameworks (MOFs) offer tunable structures for gas adsorption.
• Predicting MOF performance for H2 storage requires extensive screening.

Purpose of the Study:

• To predict the H2 capacities of a large dataset of MOFs using machine learning.
• To identify novel MOFs with superior H2 storage potential.
• To understand the key structural features governing H2 uptake in MOFs.

Main Methods:

• Machine learning (ML) models were trained using 7 structural features.
• A diverse dataset of 918,734 MOFs from 19 databases was analyzed.
• Feature importance analysis was performed to identify key predictors of H2 capacity.

Main Results:

• 8,282 MOFs were identified as having potential to exceed state-of-the-art H2 capacities.
• These promising MOFs exhibit low densities, high surface areas, void fractions, and pore volumes.
• Pore volume and void fraction were found to be the most critical features for gravimetric and volumetric H2 uptake, respectively.

Conclusions:

• ML effectively predicts H2 capacities in MOFs, enabling rapid discovery of high-performance materials.
• Hypothetical MOFs with specific structural characteristics show significant promise for H2 storage.
• Accessible ML models facilitate rapid screening of MOFs using minimal structural data.