Compatible observable decompositions for coarse-grained representations of real molecular systems

Area of Science:

• Computational chemistry
• Multiscale modeling
• Statistical mechanics

Background:

• Coarse-grained (CG) observable expressions often differ from their fine-grained (FG) counterparts, posing a representability challenge.
• Existing methods lack clear procedures for deriving CG observables from numerical FG simulations.
• The mathematical representability problem in CG modeling requires practical solutions for observable determination.

Purpose of the Study:

• To propose novel methods for determining CG observable expressions from FG systems.
• To develop systematically improvable and compatible CG observables.
• To address the bottom-up coarse-graining of real FG systems efficiently.

Main Methods:

• Introducing minimization targets for CG observables.
• Developing multiscale compatible observable decomposition (MS-CODE) and relative entropy compatible observable decomposition (RE-CODE).
• Utilizing local, data-efficient decomposition of observable contributions with new basis sets and one-body terms.

Main Results:

• Demonstrated the application of MS-CODE and RE-CODE to CG models of methanol and acetonitrile.
• Successfully calculated pressure for methanol models and potential energy for methanol and acetonitrile models.
• Showcased the compatibility and systematic improvable nature of the derived CG observables.

Conclusions:

• MS-CODE and RE-CODE provide effective bottom-up approaches for deriving CG observable expressions.
• The proposed methods offer a practical solution to the representability problem in CG modeling.
• These techniques facilitate accurate and efficient multiscale simulations.