(145f) SAFT-γ Coarse Grained Models for the Molecular Simulation of Complex Fluids with a Top-Down Methodology

An application of the “top-down” concept for the development of accurate coarse-grained intermolecular potentials of complex fluids is presented. With the common “bottom-up” procedure, coarse-grained models are constructed from a suitable simplification of a full detailed atomistic representation, and minor refinements to the intermolecular parameters are made by comparison with limited experimental where necessary. By contrast in the top-down approach, a molecular-based equation of state is used to obtain an effective coarse-grained intermolecular potential that reproduces the macroscopic experimental thermophysical properties over a wide range of conditions. In order to demonstrate our procedure, coarse-grained models for carbon dioxide (CO₂), light green-house gases, n-alkanes, alkylbenzene water and alkylpolyoxyethylene non-ionic surfactants are obtained from a recent implementation of the statistical associating fluid theory of variable range (SAFT-VR) employing a Mie (generalised Lennard- Jones) potential. For example, the parameters of a single-site Mie model of CO₂ and a coarse-grained Mie-site model of the linear alkanes are estimated by optimising the equation of state’s description of the experimental vapour-pressure and saturated liquid density data. Our SAFT-γ coarse-grained models are used in Monte Carlo molecular simulation to assess the adequacy of the description of the fluid phase behaviour and properties which were not used to develop the potential model such as the enthalpy of vaporisation, interfacial tension, density profiles, supercritical densities and second-derivative thermodynamic properties. The accuracy of the CG SAFT-γ models are found to be of similar quality to that of more sophisticated all-atom or united atom intermolecular potentials. The models for the aqueous solutions of alkylpolyoxyethylene surfactants provide a good description of these complex micellar solutions at a fraction of the computational cost of the more detailed models.