Relative entropy indicates an ideal concentration for structure-based coarse graining of binary mixtures

Area of Science:

• Computational chemistry
• Materials science
• Statistical mechanics

Background:

• Coarse-graining (CG) methods aim to simplify complex molecular systems.
• Representability and transferability are key metrics for CG model accuracy.
• Current methods often focus on refining algorithms rather than parametrization strategies.

Purpose of the Study:

• To investigate if systematic selection of parametrization state points can enhance CG model representability and transferability.
• To identify an optimal concentration for deriving effective interactions in binary mixtures.
• To assess the transferability of CG models across different mixture compositions.

Main Methods:

• Application of the inverse Monte Carlo (IMC) approach, a structure-based CG method.
• Derivation of effective interactions for binary mixtures of size-mismatched Lennard-Jones (LJ) particles.
• Utilizing relative entropy to quantify information loss and identify optimal parametrization concentrations.

Main Results:

• A specific concentration was identified where IMC-derived potentials exhibit optimal structural representability and transferability for LJ mixtures.
• Relative entropy successfully pinpointed this optimal concentration by measuring information loss.
• Similar trends in transferability were observed for n-hexane/n-perfluorohexane mixtures, validating the findings.

Conclusions:

• Systematic selection of the parametrization state point is a viable strategy to improve CG model performance.
• The optimal concentration enhances both representability and transferability of derived potentials.
• The findings offer a new perspective on optimizing CG model development beyond methodological refinements.