Understanding Missing Entropy in Coarse-Grained Systems: Addressing Issues of Representability and Transferability.

Area of Science:

• Computational Chemistry
• Materials Science
• Statistical Mechanics

Background:

• Coarse-grained (CG) models simplify complex systems by reducing degrees of freedom, enabling efficient simulations.
• Current CG models often struggle with transferability across different thermodynamic conditions due to approximations in representing the potential of mean force (PMF).
• The PMF is a complex, multidimensional quantity typically approximated by pairwise additive potentials, lacking fundamental principles for broad applicability.

Purpose of the Study:

• To develop transferable CG models by explicitly decomposing the CG PMF into energy and entropy contributions.
• To establish a framework that formally connects the entropy lost in CG representations to the CG configurational entropy.
• To enable CG models to accurately predict system behavior across varying temperatures and compositions.

Main Methods:

• Investigated the explicit energy-entropy decomposition of the CG PMF for liquid systems.
• Represented the entropic component of the CG PMF using additive pairwise contributions, linked to CG configurational entropy.
• Developed combining rules to approximate cross-interactions in mixtures from bulk CG PMFs, ensuring transferability across compositions.

Main Results:

• Demonstrated that the entropic component of the CG PMF can be represented by additive pairwise contributions, closely related to CG configurational entropy.
• Achieved transferability of CG interactions across different temperatures by incorporating an explicit additive entropic term.
• Showcased transferability across composition states, including bulk liquids and their mixtures, validated by structural correlations in predicted CG models.

Conclusions:

• The energy-entropy decomposition provides a physically grounded method for constructing transferable CG models.
• This approach offers a compact representation of CG entropy and a solution to the transferability problem in CG modeling.
• The findings pave the way for predictive multiscale modeling by enhancing the reliability and applicability of CG simulations.