Quantum Informed Machine-Learning Potentials for Molecular Dynamics Simulations of CO₂'s Chemisorption and Diffusion...

Area of Science:

• Materials Science
• Computational Chemistry
• Chemical Engineering

Background:

• Metal-organic frameworks (MOFs) show promise for CO2 capture due to high uptake and selectivity.
• Computational identification of optimal MOFs is challenging due to simulation limitations.
• First-principle simulations offer accuracy but are computationally expensive; classical methods lack accuracy.

Purpose of the Study:

• To develop quantum-informed machine-learning force fields (QMLFFs) for accurate and efficient atomistic simulations of CO2 in MOFs.
• To overcome the computational cost and accuracy trade-offs in MOF simulation for gas separation.
• To enable reliable in silico evaluation of gas molecule chemisorption and diffusion in MOFs.

Main Methods:

• Development of quantum-informed machine-learning force fields (QMLFFs).
• Atomistic simulations using QMLFFs for CO2 adsorption in MOFs.
• Molecular dynamics simulations of CO2 in Mg-MOF-74 to predict binding free energy and diffusion.

Main Results:

• QMLFFs achieve ~1000x higher computational efficiency than first-principle methods while maintaining quantum-level accuracy.
• QMLFF-based simulations accurately predict CO2 binding free energy landscapes in Mg-MOF-74.
• Simulations yield CO2 diffusion coefficients in Mg-MOF-74 close to experimental values.

Conclusions:

• QMLFFs offer a powerful approach for accurate and efficient in silico screening of MOFs for CO2 capture.
• This method bridges the gap between computational feasibility and accuracy in MOF simulations.
• The QMLFF approach facilitates the design of advanced materials for gas separation and purification.