Machine learning assisted coarse-grained molecular dynamics modeling of meso-scale interfacial fluids

Area of Science:

• Computational chemistry
• Materials science
• Fluid dynamics

Background:

• Interfacial energy in meso-scale fluids is scale-dependent, posing challenges for coarse-grained (CG) modeling.
• CG models require accurate encoding of many-body atomistic interactions and heterogeneous density distributions.

Purpose of the Study:

• To construct reliable CG models for single- and two-component polymeric fluid systems.
• To accurately reproduce interfacial properties and collective behaviors using a deep CG potential scheme.

Main Methods:

• Developed CG models for polymeric fluids using a deep coarse-grained potential scheme.
• Trained models using instantaneous force data from thermal equilibrium simulations.
• Modeled each polymer molecule as a CG particle.

Main Results:

• CG models accurately reproduced void formation probability density functions in bulk.
• Models successfully predicted the capillary wave spectrum across fluid interfaces.
• CG models accurately predicted the volume-to-area scaling transition for apolar solvation energy.

Conclusions:

• The developed CG models effectively bridge molecular and continuum scales for polymeric fluids.
• The approach demonstrates fidelity in encoding molecular-level details into meso-scale collective behaviors.
• This method provides a powerful tool for studying interfacial phenomena in complex fluids.