Estimating hydrodynamic quantities in the presence of microscopic fluctuations (Q855020)

The paper discusses the estimation of hydrodynamic quantities in the presence of microscopic fluctuations. Such spontaneous fluctuations appear in molecular simulations of fluid flows including a variety of particle-based algorithms like molecular dynamics, particle-in-cell, direct simulation Monte Carlo, dissipative particle dynamics. In such simulations, the quantities of interest are not the precise trajectories of the particles, but rather the hydrodynamic variables such as density, fluid velocity, temperature, and pressure. In distinction to macroscopic systems, the number of particles in simulations typically is of order 10 to 100. For this reason, the spontaneous fluctuations in a volume element are significant and statistical samples are taken. Mechanical variables in a cell \({\mathcal M}=\rho, {\mathbf J}, K\) are considered, where \(\rho\), \({\mathbf J}\), and \(K\) is the instantaneous density of mass, momentum, and kinetic energy, respectively. The sample mean of a mechanical variable is \[ \langle {\mathcal M} \rangle_S = \frac{1}{S} \sum\limits_{j=1}^S {\mathcal M}_j \;, \] where the subscript \(j\) indicates individual samples. The mean value is given by \[ \bar {\mathcal M} = \lim_{S \to \infty} \langle {\mathcal M} \rangle_S \;. \] The principal point is that hydrodynamic variables must be defined in terms of mechanical variables. Because these relations are nonlinear and because fluctuations of mechanical variables are correlated, care must be taken to avoid introducing a bias when evaluating means, variances, and correlations of hydrodynamic variables. It is important to understand that, for a hydrodynamic variable \({\mathcal H}\), the mean is defined in terms of the means of mechanical variables \[ \bar {\mathcal H} = {\mathcal H} \left( \bar \rho, \bar {\mathbf J}, \bar K \right) \neq \lim_{S \to \infty} \langle {\mathcal H}(\rho, {\mathbf J}, K) \rangle_S \;. \] The purpose of the paper is to establish the correct construction for measuring hydrodynamic variables and to point out some common errors that lead to biased results. The unbiased estimates are formulated and some incorrect approaches are presented as cautionary warnings. The expressions are verified by numerical simulations both in thermodynamic equilibrium and in nonequilibrium steady state.