Development of coarse-grained models for the simulation of soft matter systems

Abstract

This thesis aims to examine the parametrisation of coarse-grained models for the simulation of soft matter systems. The strengths and weaknesses of a range of methods are examined, and suggestions for improvements are made.
Initially, two bottom-up methods, iterative Boltzmann inversion (IBI) and hybrid force matching (HFM) are applied to a liquid octane/benzene mixture and compared to a top-down model based on a version of statistical associating fluid theory, the SAFT-γ Mie equation of state. These models are tested for their ability to represent the structure and thermodynamics of the underlying atomistic system, as well as their transferability between temperatures and concentrations. Attempts are then made to address the poor transferability of the bottom-up models using a variant of IBI, multi-state IBI (MS-IBI). MS-IBI allows concentration transferable potentials to be generated but is not successful in improving temperature transferability. The state-point dependence of pair potentials is identified as the cause of poor temperature transferability, and initial attempts to address this are discussed.
A range of coarse-grained models of the non-ionic liquid crystal TP6EO2M is examined. HFM is able to give a structurally accurate coarse-grained model; however, the difficulty of sampling all relevant configurations within an atomistic reference system appear to cause problems with calculating accurate association free energies. The new MARTINI 3 top-down force field is shown to improve upon the structural and thermodynamic properties of MARTINI 2, allowing larger system sizes to be studied. The nematic and hexagonal columnar chromonic phases are observed, and the concentration dependence seen in the experimental phase diagram is reproduced. This represents the first simulations of chromonic liquid crystal phases using systematic coarse graining.