Machine Learning Predictions of Methane Storage in MOFs: Diverse Materials, Multiple Operating Conditions, and Reverse Models | JoVE Visualize

Area of Science:

Materials Science
Computational Chemistry
Machine Learning

Background:

Metal-organic frameworks (MOFs) are promising materials for gas storage applications.
Predicting methane (CH4) storage capacity in MOFs is crucial for optimizing their use.
Current predictive models often lack broad applicability or require extensive input data.

Purpose of the Study:

To develop a broadly applicable machine learning (ML) model for predicting usable methane (CH4) capacities in diverse metal-organic frameworks (MOFs).
To identify novel MOF structures with superior methane storage capabilities.
To establish key structural features governing methane adsorption in MOFs.

Main Methods:

Development of a machine learning model utilizing five measurable structural features of MOFs.
Application of the model to a large database of hypothetical MOFs to screen for high-capacity candidates.
Experimental synthesis and characterization of a top-performing predicted MOF (UMCM-153).
Feature importance analysis to determine critical MOF structural descriptors for methane capacity.
Development of a reverse machine learning model for inverse design of MOFs.

Main Results:

The developed ML model demonstrates high accuracy in predicting CH4 capacities across a wide range of MOFs, outperforming less-general models.
Screening of over a million hypothetical MOFs identified several hundred candidates exceeding the benchmark MOF (UMCM-152) in CH4 capacity.
Synthesized MOF UMCM-153 exhibited superior volumetric CH4 capacity, validating the model's predictions.
Pore volume and gravimetric surface area were identified as the most influential features for CH4 capacity prediction.
A reverse ML model was successfully demonstrated for designing MOFs with targeted CH4 storage capacities.

Conclusions:

A versatile and accurate ML model has been established for predicting methane storage in MOFs, requiring minimal input data.
The study successfully identified and validated novel MOF materials with enhanced methane storage potential.
This work provides valuable insights into the structure-property relationships governing methane adsorption in MOFs and offers a powerful tool for materials discovery.