Assessing the Accuracy of Bayesian-Optimized CGMD in Predicting Polymer Miscibility

Abstract Coarse-grained molecular dynamics (CGMD) presents a computationally efficient alternative to all-atom (AA) simulations for modeling polymer systems, yet the accuracy of its predictions for emergent thermodynamic properties remains a challenge. This study builds upon our previous work, which introduced a Bayesian optimization framework for refining Martini3 force field, to validate the transferability of these optimized models to multi-component systems. We demonstrate that CG force fields, optimized against single-chain properties (density and radius of gyration), can qualitatively estimate the phase behavior of both homopolymer blends and diblock copolymer systems. Bayesian-optimized topologies for polyvinylidene fluoride (PVDF), polystyrene (PS), and poly(methyl methacrylate) (PMMA) were evaluated through direct simulation of binary blend morphologies and a PMMA-PEO (poly(ethylene oxide)) diblock copolymer system. The framework successfully predicted macroscopic phase separation in the immiscible PS-PMMA system and miscibility in the PVDF-PMMA system, with resulting density profiles showing qualitative agreement with experimental observations. Flory–Huggins interaction parameters ( $$\chi $$ χ ) estimated from Hildebrand solubility parameters ( $$\delta $$ δ ) correctly rank relative miscibilities and align qualitatively with literature trends. Our results show that optimizing for fundamental properties such as density and radius of gyration (achieving <10% error) implicitly captures the necessary thermodynamic contributions to predict complex, multi-component phase behaviors. Validation of miscibility prediction across two canonical homopolymer blends, combined with efficient parameterization of a diblock copolymer architecture, provides compelling evidence that Bayesian-optimized coarse-grained molecular dynamics is a robust and predictive tool for polymer thermodynamics, effectively bridging the gap between computational efficiency and predictive fidelity in materials discovery.