Multiple-resolution scheme in finite-volume code for active or passive scalar turbulence

TL;DR

This paper introduces a multiple-resolution finite-volume scheme for scalar turbulence, allowing separate, optimized grids for velocity and scalar fields, leading to computational speed-up without sacrificing accuracy.

Contribution

The authors develop a finite-volume multiple-resolution algorithm that efficiently simulates scalar turbulence by using different grids for velocity and scalar fields, especially beneficial for high Schmidt numbers.

Findings

Significant speed-up in scalar turbulence simulations.

Maintains accuracy of scalar and velocity fields.

Applicable to both active and passive scalar turbulence.

Abstract

In scalar turbulence it is sometimes the case that the scalar diffusivity is smaller than the viscous diffusivity. The thermally-driven turbulent convection in water is a typical example. In such a case the smallest scale in the problem is the Batchelor scale $l_b$, rather than the Kolmogorov scale $l_k$, as $l_b = l_k/Sc^{1/2}$, where Sc is the Schmidt number (or Prandtl number in the case of temperature). In the numerical studies of such scalar turbulence, the conventional approach is to use a single grid for both the velocity and scalar fields. Such single-resolution scheme often over-resolves the velocity field because the resolution requirement for scalar is higher than that for the velocity field, since $l_b<l_k$ for $Sc>1$. In this paper we put forward an algorithm that implements the so-called multiple-resolution method with a finite-volume code. In this scheme, the velocity and pressure fields are solved in a coarse grid, while the scalar field is solved in a dense grid. The central idea is to implement the interpolation scheme on the framework of finite-volume to reconstruct the divergence-free velocity from the coarse to the dense grid. We demonstrate our method using a canonical model system of fluid turbulence, the Rayleigh-B\'enard convection. We show that, with the tailored mesh design, considerable speed-up for simulating scalar turbulence can be achieved, especially for large Schmidt (Prandtl) numbers. In the same time, sufficient accuracy of the scalar and velocity fields can be achived by this multiple-resolution scheme. Although our algorithm is demonstrated with a case of an active scalar, it can be readily applied to passive scalar turbulent flows.