Polarisable soft solvent models with applications in dissipative particle dynamics

TL;DR

This work develops explicitly polarisable soft solvent models for coarse-grained DPD simulations to capture dielectric contrasts in aqueous structured liquids. By assigning partial charges to small solvent molecules (dimers or dressed solvents), the authors let dielectric response emerge from explicit solvent structure and quantify it via the fluctuating box dipole method, Kirkwood factors, and correlation functions. They show substantial dipole-dipole correlations (g_K ≈ 0.7–0.8), causing Onsager theory to overestimate permittivity by 20–30%, while a first-order Wertheim perturbation theory with plasma corrections provides accurate estimates for mean-square dipole moments and good initial epsilon predictions. The study verifies the models across test problems, including field application, test-charge interactions, and oil/water interfacial ion desorption, and presents practical parameterisations (notably WinO-DS) for water-in-oil systems, offering a framework to design mesoscale dielectric solvents with potential extensions to true dipole representations and interfacial electrostatics.

Abstract

We critically examine a broad class of explicitly polarisable soft solvent models aimed at applications in dissipative particle dynamics. We obtain the dielectric permittivity using the fluctuating box dipole method in linear response theory, and verify the models in relation to several test cases including demonstrating ion desorption from an oil-water interface due to image charge effects. We additionally compute the Kirkwood factor and find it uniformly lies in the range gK approx 0.7-0.8, indicating that dipole-dipole correlations are not negligible in these models. This is supported by measurements of dipole-dipole correlation functions. As a consequence, Onsager theory over-predicts the dielectric permittivity by approximately 20-30 percent. On the other hand, the mean square molecular dipole moment can be accurately estimated with a first-order Wertheim perturbation theory.

Figures (10)

• Figure 1: Field lines and induced surface charges for a point charge near an air-water or oil-water interface ($\epsilon_{1} \gg\epsilon_{2}$).
• Figure 2: Polarisable soft solvent models: (a) molecular dimer comprising a bound pair of solvent beads carrying equal and opposite partial charges, (b) trimer comprising a single solvent bead dressed with a pair of tethered partial charges.
• Figure 3: Geometry for molecular dipole distribution functions: for each pair of solvent dipoles (here illustrated for the trimer model) we define the dipole moments ${\bf p}_1$ and ${\bf p}_2$, and the distance between the dipole centers ${\bf r}$.
• Figure 4: Dielectric properties of dressed solvent models in Table \ref{['tab:dsprop']}, as a function of $q^2$: (a) mean square dipole length $\langle{\Delta{\bf r}^2}\rangle$, and (b) relative permittivity $\epsilon=\epsilon_{r}/\epsilon_{b}$; the arrowed point is the WinO-DS model (Table \ref{['tab:dimtri']}), with $q=0.36$ and $\epsilon=42\pm2$. Markers with error bars are Monte-Carlo simulations (error bars not shown in (a) as the error is smaller than markers). Lines are liquid state theory predictions with (dashed lines) and without (dotted lines) the plasma reference state correction.
• Figure 5: Correlation functions for the WinO-DS model (Table \ref{['tab:dimtri']}): (a) dipole length distribution, (b) central bead pair distribution function $g_{00}(r)$ and dipole-dipole centre distribution function $g_{000}(r)$, (c) and (d) higher order dipole-dipole correlation functions $g_{110}(r)$ and $g_{112}(r)$. The functions in (c) and (d) are normalised by dividing by $\langle{{\mathbf p}^2}\rangle$. These functions are computed using matched 'Kirkwoood' boundary conditions, with $\epsilon'=42$ in the reaction field, as described in Sec. \ref{['subsec:corr']} (see also Table \ref{['tab:dsprop']}).