Constructing a coarse-grained water model based on non-Markovian dissipative particle dynamics

Abstract

We constructed a coarse-grained (CG) water model based on non-Markovian dissipative particle dynamics (NMDPD) taking into account memory effects. The NMDPD equation of motion was derived from a generalized Langevin equation formulated via the Mori–Zwanzig (MZ) projection operator. We extracted a CG pair potential and memory kernels between clusters comprising 10 water molecules by means of molecular dynamics (MD) simulations. We found that the MZ-guided CG potential followed by an iterative Boltzmann inversion correction resulted in an accurate representation of both a radial distribution function and pressure. Furthermore, in contrast to Markovian DPD, the NMDPD model exploiting MZ-guided memory kernels could reproduce short-time dynamics originating from molecular collisions, which was characterized by decaying nature of a velocity autocorrelation function (VACF). The NMDPD model was also able to reasonably represent the viscosity of the MD system compared to the conventional DPD, where interaction parameters were phenomenologically tuned such that a few macroscopic properties were reproduced, leading to a significant underestimation of a viscosity or Schmidt number. Finally, the differences of the viscosity and long-time behavior of the VACF between MD and NMDPD systems implied the necessity of a more appropriate description for a one-to-one correspondence between a CG particle and a water cluster.

Keywords

• mori–zwanzig projection operator
• genralized langevin equation
• memory effect
• iterative boltzmann inversion
• dissipative particle dynamics
• molecular dynamics