(272e) A Systematic Coarse-Graining of Molecular Dynamics Simulations: Thermodynamic and Transport Properties

Coarse-graining techniques have recently attracted great interest for providing descriptions at a mesoscopic level of resolution that preserve fluid thermodynamic and transport behaviors with a reduced number of degrees of freedom (DOFs) and hence less computational effort. One fundamental question arises:  how well and to what extent can a “bottom-up” developed mesoscale model recover the physical properties of a molecular scale system? To answer this question, we explore both the thermodynamic and transport properties of a systematically-developed coarse-grained (CG) model that represents an intermediate mesoscale fluid between the atomistic and continuum scales [1,2]. The model is developed using the iterative Boltzmann inversion (IBI) technique to determine a CG potential for a (1-ϕ)N mesoscale particle system, where ϕ is the fraction of removed particles from an atomic system.  The uniqueness theorem guarantees a one to one relationship between the radial distribution function (RDF) and such effective pairwise potentials, but we find that RDFs are insensitive to the long-range part of the IBI-determined potentials, which provides significant flexibility in further matching other properties.

We subsequently propose a reformulation of IBI that enables simultaneous matching of the RDF and the fluid pressure. This new method mainly changes the attractive tail region of the CG potentials, and it improves the isothermal compressibility relative to pure IBI. We also find that there are optimal interaction cutoff lengths for the CG system, as a function of ϕ, that are required to attain an adequate potential while maintaining computational speedup. Dynamical properties such as the self diffusion coefficient and viscosity cannot be matched directly during coarse-graining by modifying the pair interaction. Instead, one can introduce a dissipative and random forces characterized by a friction coefficient γ, which becomes an additional parameter in the CG model that can be tuned.  Using the Galilean-invariant dissipative particle dynamics thermostat, we show that a value of γ for each degree of coarse-graining ϕ can be found for which both viscosity and diffusion match the reference LJ liquid. Importantly, we show that Stokes-Einstein behavior persists for these coarse models and offers a useful perspective for interpreting the dynamics of mesoscale CG models.

References:

[1] Fu, C.; Kulkarni, P.; Shell, M. S.; Leal, L. G. J. Chem. Phys. 137, 164106 (2012).

[2] Fu, C.; Kulkarni, P.; Shell, M. S.; Leal, L. G. J. Chem. Phys. 2013, Submitted.