(382a) A Combined Bottom-up and Top-Down Coarse-Graining Approach to Develop Polymer Models

Copolymers are widely used in many commercial products due to advantageous physical, chemical, or thermal properties. Coarse-grained molecular dynamics (CG-MD) simulations can be used to efficiently explore the effect of their chemical composition and sequence. Most CG polymer force fields have been developed using a combination of structural data from all-atom simulations, and refinement of nonbonded parameters through iterative MD or Monte Carlo (MC) simulation to reproduce experimental data. SAFT-γ Mie, a molecular group-contribution equation of state, is an efficient alternative where experimental data on vapor-liquid equilibrium (VLE) is used to directly parameterize a nonbonded Mie potential.

We have developed a hybrid bottom-up/top-down coarse-graining framework that adds bonded potentials derived from all-atom MD simulations to fused-sphere SAFT-γ Mie homopolymer chains parameterized to reproduce VLE data. The shape factor parameter in the fused-sphere model allows for bead overlap, preserving structural consistency with the all-atom model. We find that the optimal Mie parameters for a range of fixed shape factor values provide nearly equal representations of the target VLE data. Empirical correlations between shape factor and optimal Mie parameters thus represent a locus of solutions for monomer chemical fragments. Applying bond and angle potentials derived from the all-atom reference system, we determined the shape factor which best reproduces all-atom polymer density through a series of NPT MD simulations.

In addition to homopolymers, we have also extended the fused-sphere SAFT-γ Mie approach to copolymers, and evaluated the transferability of shape factors for chains of various sequence distributions ranging from diblock to random. In this work, we present models developed using our combined top-down/bottom-up coarse graining approach for both homopolymers and a complex poly[(vinyl alcohol)-co-(vinyl butyral)] (PVB) system. For the copolymer, we found that a set of shape factors determined from separate homopolymer models does not in general accurately reproduce random copolymer density when standard SAFT-γ Mie mixing rules are used. However, performing density matching for a 50:50 copolymer, with only one of the homopolymer shape factors fixed, results in a model which correctly predicted copolymer density over a wide range of chemical compositions and sequences. Our model was able to accurately reproduce experimental glass transition temperature and solubility parameter as a function of chemical composition for random PVB copolymers. Additionally, we demonstrate successful reverse-mapping from fused-sphere SAFT-γ Mie copolymer melts to the all-atom model from which bonded potentials were derived, and study the adsorption of PVB on a realistic silica surface model.