Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-17T08:31:00.568Z Has data issue: false hasContentIssue false

Localized layers of turbulence in stratified horizontally sheared Poiseuille flow

Published online by Cambridge University Press:  19 May 2023

J. Labarbe*
Affiliation:
Aix Marseille Univ., CNRS, Centrale Marseille, IRPHE, 13384 Marseille, France
P. Le Gal
Affiliation:
Aix Marseille Univ., CNRS, Centrale Marseille, IRPHE, 13384 Marseille, France
U. Harlander
Affiliation:
Department of Aerodynamics and Fluid Mechanics, Brandenburg University of Technology, Cottbus-Senftenberg, D-03046 Cottbus, Germany
S. Le Dizès
Affiliation:
Aix Marseille Univ., CNRS, Centrale Marseille, IRPHE, 13384 Marseille, France
B. Favier
Affiliation:
Aix Marseille Univ., CNRS, Centrale Marseille, IRPHE, 13384 Marseille, France
*
†Email address for correspondence: jlabarbe@unice.fr

Abstract

This paper presents a numerical analysis of the instability developing in horizontally sheared Poiseuille flow, when stratification extends along the vertical direction. Our study builds on the previous work that originally detected the linear instability of such a configuration, by means of experiments, theoretical analysis and numerical simulations (Le Gal et al., J. Fluid Mech., vol. 907, 2021, R1). We extend this investigation beyond linear theory, investigating nonlinear regimes with direct numerical simulations. We find that the flow loses its vertical homogeneity through a secondary bifurcation, due to harmonic resonances, and describe this symmetry-breaking mechanism in the vicinity of the instability threshold. When departing from this limit, we observe a series of bursting events that break down the flow into disordered motions driven by localized shear instabilities. This intermittent dynamics leads to the coexistence of horizontal localized layers of stratified turbulence surrounded by quiescent regions of meandering waves.

Information

Type
JFM Papers
Copyright
© The Author(s), 2023. Published by Cambridge University Press
Figure 0

Figure 1.Growth rate contours in the $(k_x,k_z)$ space from the linear stability analysis of (3.2) at ${{Fr}}=2$, ${{Sc}}=1$ and over different Reynolds numbers: (a) ${{Re}}=550$, (b) ${{Re}}=580$, (c) ${{Re}}=700$. Red crosses correspond to the discretized modes present in our DNS domain (cf. § 4). White regions correspond to negative growth rates, i.e. stability.

Figure 1

Figure 2.Vertical kinetic energy curves (in logarithmic scale) over time (left-hand plots), with snapshots of the streamwise velocity fluctuation $u-U_0$ (middle) and the buoyancy perturbation $b$ (right), computed at the latest time of each simulation. Reynolds numbers are (a) ${{Re}}=550$, (b) ${{Re}}=580$, and (c) ${{Re}}=700$. Red dashed lines in the left-hand plots represent twice the growth rates of the dominant unstable mode, as predicted from linear theory. Close-up views are displayed as insets for each configuration, highlighting the nonlinear behaviour.

Figure 2

Figure 3.Hovmöller diagrams of the horizontally averaged streamwise velocity fluctuation $\langle u - U_0 \rangle _H$, computed in the $(t,z)$-plane. Positive (negative) values are displayed in red (blue). The area of constant colour on the left of each plot represents the exponential growth of disturbances. Reynolds numbers are (a) ${{Re}}=550$, (b) ${{Re}}=580$, and (c) ${{Re}}=700$.

Figure 3

Figure 4.Each curve is computed at the latest time of simulation in figure 3, once the solution has converged. (a) Horizontally averaged streamwise velocity fluctuations as a function of the vertical coordinate. (b) Spectral energy in Fourier space of the averaged velocity profiles from (a). The principal amplitude peaks are displayed with vertical lines and labelled with their associated representation defined in (4.5).

Figure 4

Figure 5.DNS at ${{Re}}=1500$. (a) Hovmöller map of streamwise mean velocity fluctuations, with the same colour code as in figure 3. (b) Vertical kinetic energy in the spirit of figure 2. The two red markers indicate the extrema of a bursting episode, to be computed in detail in figure 6. (c) Vertical minimum of the gradient Richardson number over time, computed in logarithmic scale (left $y$-axis, yellow) and viscous dissipation $\mathcal {E}$ (right $y$-axis, dark green). The yellow curve is discontinuous whenever ${{Ri}}_z$ is negative (when the flow becomes locally unstable to convection). The thin dash-dotted line delimits the Kelvin–Helmholtz instability threshold of ${{Ri}}_z=1/4$ (Howard 1961; Miles 1961).

Figure 5

Figure 6.Local measures of Froude and Richardson numbers (in logarithmic units) over the vertical coordinate, for the case ${{Re}}=1500$. Data and snapshots are computed at (a) $t=2837$, and (b) $t=3151$ (see figure 5). The middle and right-hand plots represent the streamwise velocity fluctuation $u-U_0$ and the perturbation of buoyancy $b$, respectively, both displayed at an arbitrary position along $x$. The thin dotted lines passing through each set of plots delimit the interfaces where the secondary instabilities trigger and the flow collapses into stratified turbulence.

Figure 6

Figure 7.DNS at ${{Re}}=5000$. Same plots as in figure 5. The red markers in the representation of vertical kinetic energy indicate the time coordinates used in the snapshots of figure 8.

Figure 7

Figure 8.Snapshots of the numerical computation at ${{Re}}=5000$, computed at (a) $t=2940$, and (b) $t=3600$. The left-hand plots display the streamwise velocity fluctuation $u-U_0$, whereas the right-hand plots represent the associated iso-contours of wall-normal velocity. Positive (negative) iso-contours are represented in light purple (light blue) regardless of their magnitude.