Machine learning potential for modelling H₂ adsorption/diffusion in MOFs with open metal sites.

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Metal-organic frameworks (MOFs) with open metal sites (OMS) are promising for gas adsorption (CO₂ capture, H₂ storage).
• Accurate computational modeling of H₂ interactions with OMS in MOFs is crucial but challenging due to limitations of classical force fields.
• Existing methods hinder high-throughput screening for novel MOF materials for adsorption applications.

Purpose of the Study:

• To develop a novel machine learning potential (MLP) for accurate simulation of H₂ adsorption in MOFs containing OMS.
• To enable efficient and accurate computational-assisted identification of MOFs for H₂ storage and low-pressure gas capture.
• To overcome the limitations of generic classical force fields in describing OMS-guest molecule interactions.

Main Methods:

• Derived a machine learning potential (MLP) for Al-soc-MOF-1d using ab initio molecular dynamics (AIMD) simulations.
• Employed MLP in molecular dynamics (MD) simulations to study H₂ binding and temperature-dependent distribution.
• Utilized MLP-Grand Canonical Monte Carlo (GCMC) simulations for H₂ sorption isotherms and MLP-MD for H₂ kinetics.

Main Results:

• Developed the first MLP capable of accurately describing H₂ interactions with OMS in MOFs.
• MLP-GCMC simulations accurately predicted H₂ sorption isotherms for Al-soc-MOF-1d, validated by experimental data.
• MLP-based MD simulations provided insights into H₂ adsorption kinetics within the MOF.

Conclusions:

• The developed MLP strategy enables accurate and efficient in silico assessment of MOFs containing OMS for H₂ adsorption.
• This approach paves the way for systematic discovery of MOFs for H₂ storage and low-pressure capture of other molecules.
• Overcomes a critical bottleneck in computational materials discovery for adsorption-related applications.