Nusselt number scaling in horizontal convection

TL;DR

This study numerically investigates horizontal convection at Prandtl number 1, revealing different Nusselt number scaling regimes with Rayleigh number in 2D and 3D, and highlighting the influence of boundary conditions and boundary layer structures.

Contribution

It provides the first detailed analysis of Nusselt number scaling in 2D and 3D horizontal convection across a wide range of Rayleigh numbers, including boundary layer characterization.

Findings

In 2D, Nu scales as Ra^{1/5} for Ra up to 10^{10} and as Ra^{1/4} for Ra > 10^{11}.

In 3D, Nu scales as Ra^{1/5} up to Ra = 10^{11.5}.

Boundary conditions significantly affect heat transport, with free-slip yielding higher Nu than no-slip.

Abstract

We report a numerical study of horizontal convection (HC) at Prandtl number $Pr = 1$, with both no-slip and free-slip boundary conditions. We obtain 2D and 3D solutions and determine the relation between the Rayleigh number $Ra$ and the Nusselt number $Nu$. In 2D we vary $Ra$ between $0$ and $10^{14}$. In the range $10^6 \le Ra \le 10^{10}$ the $Nu$-$Ra$ relation is $Nu \sim Ra^{1/5}$. With $Ra$ greater than about $10^{11}$ we find a 2D regime with $Nu \sim Ra^{1/4}$ over three decades, up to the highest 2D $Ra$. In 3D, with maximum $Ra = 10^{11.5}$, we find only $Nu \sim Ra^{1/5}$. These results apply to both free slip and no slip boundary conditions. The $Nu \sim Ra^{1/4}$ regime has a double boundary layer (BL): there is a thin BL with thickness $\sim Ra^{-1/4}$ nested inside a thicker BL with thickness $\sim Ra^{-1/5}$. The $Ra^{-1/4}$ BL thickness, which determines $Nu$, coincides with the Kolmogorov and Batchelor scales of HC. Numerical and theoretical results indicate that 3D HC is qualitatively and quantitatively similar to 2D HC. At the same $Ra$, the 3D $Nu$ exceeds the 2D $Nu$ by less than $20$%, i.e., there is very little 3D enhancement of heat transport. Boundary conditions are more important than dimensionality: the 2D free-slip solutions have larger $Nu$ than 3D no-slip solutions. Using the mechanical energy power integral of HC we show that the mean square vorticity of 3D HC is nearly equal to that of 2D HC at the same $Ra$. Thus vorticity amplification by strain-mediated vortex stretching does not operate in 3D HC.