Improving the thermodynamics and transferability of coarse-grained models

Coarse-grained (CG) simulations are becoming increasingly important for the modelling of soft matter systems, due to their ability to study large system sizes over long time scales. Common methods for parametrising CG models... [ view full abstract ]

Coarse-grained (CG) simulations are becoming increasingly important for the modelling of soft matter systems, due to their ability to study large system sizes over long time scales. Common methods for parametrising CG models include. iterative Boltzmann inversion (IBI) and force matching. While these usually give good structural accuracy, they often yield interaction potentials with poor transferability to different state points. This is mainly due to the fact that the pair potentials only include multi-body effects in an averaged out way, which is only correct at the state point at which the potential was parametrised. The thermodynamic accuracy of these CG models is often poor, particularly in reproducing free energies of mixing, which we have shown for a range of octane/benzene mixtures.

We show that parametrising CG models from multiple state points, using multi-reference IBI, can help with the problem of poor transferability in some circumstances. However, there are still issues with this approach. The resulting models do not necessarily improve upon the accuracy of calculated thermodynamic quantities relative to models derived from a single reference point. 

It has been proposed that including a correction to the pair potential, which depends on the local environment of the CG bead, could lead to improved accuracy and transferability. Here, we propose a method to parametrise such a potential, which aims to keep the structural accuracy of a CG model parametrised at a single state point, but provides more accurate thermodynamic quantities and is transferable over a range of concentrations and temperatures.