2D and 3D Dense-Fluid Shear Flows via Nonequilibrium Molecular Dynamics. Comparison of Time-and-Space-Averaged Tensor Temperature and Normal Stresses from Doll's, Sllod, and Boundary-Driven Shear Algorithms

TL;DR

This study compares nonequilibrium molecular dynamics algorithms for simulating shear flows, revealing differences in stress predictions and temperature ordering, with implications for modeling accuracy in fluid dynamics.

Contribution

It evaluates Doll's and Sllod algorithms against boundary-driven flows, highlighting inaccuracies in stress differences and temperature predictions in homogeneous shear simulations.

Findings

Sllod predicts opposite stress difference sign to boundary-driven flows.

Neither algorithm correctly predicts temperature ordering.

Stress differences are small but significant in 2D and 3D.

Abstract

Homogeneous shear flows (with constant strainrate du/dy) are generated with the Doll's and Sllod algorithms and compared to corresponding inhomogeneous boundary-driven flows. We use one-, two-, and three-dimensional smooth-particle weight functions for computing instantaneous spatial averages. The nonlinear stress differences are small, but significant, in both two and three space dimensions. In homogeneous systems the sign and magnitude of the shearplane stress difference, P(xx) - P(yy), depend on both the thermostat type and the chosen shearflow algorithm. The Doll's and Sllod algorithms predict opposite signs for this stress difference, with the Sllod approach definitely wrong, but somewhat closer to the (boundary-driven) truth. Neither of the homogeneous shear algorithms predicts the correct ordering of the kinetic temperatures, T(xx) > T(zz) > T(yy).