Hostname: page-component-7b7567c8c-z4zkx Total loading time: 0 Render date: 2026-11-16T16:27:43.049Z Has data issue: false hasContentIssue false

Large-scale characteristics of a stably stratified turbulent shear layer

Published online by Cambridge University Press:  28 September 2021

Tomoaki Watanabe*
Affiliation:
Department of Aerospace Engineering, Nagoya University, Nagoya 464-8603, Japan
Koji Nagata
Affiliation:
Department of Aerospace Engineering, Nagoya University, Nagoya 464-8603, Japan
*
†Email address for correspondence: watanabe.tomoaki@c.nagoya-u.jp

Abstract

Implicit large eddy simulation is performed to investigate large-scale characteristics of a temporally evolving, stably stratified turbulent shear layer arising from the Kelvin–Helmholtz instability. The shear layer at late time has two energy-containing length scales: the scale of the shear layer thickness, which characterizes large-scale motions (LSM) of the shear layer; and the larger streamwise scale of elongated large-scale structures (ELSS), which increases with time. The ELSS forms in the middle of the shear layer when the Richardson number is sufficiently large. The contribution of the ELSS to velocity and density variances becomes relatively important with time although the LSM dominate the momentum and density transport. The ELSS have a highly anisotropic Reynolds stress, to a degree similar to the near-wall region of turbulent boundary layers, while the Reynolds stress of the LSM is as anisotropic as in the outer region. Peaks in the spectral energy density associated with the ELSS emerge because of the slow decay of turbulence at very large scales. A forward interscale energy transfer from large to small scales occurs even at a small buoyancy Reynolds number. However, an inverse transfer also occurs for the energy of spanwise velocity. Negative production of streamwise velocity and density spectra, i.e. counter-gradient transport of momentum and density, is found at small scales. These behaviours are consistent with channel flows, indicating similar flow dynamics in the stratified shear layer and wall-bounded shear flows. The structure function exhibits a logarithmic law at large scales, implying a $k^{-1}$ scaling of energy spectra.

Information

Type
JFM Papers
Copyright
© The Author(s), 2021. Published by Cambridge University Press
Figure 0

Figure 1.Schematic diagram of numerical simulations of a temporally evolving, stably stratified shear layer.

Figure 1

Figure 2.Premultiplied energy spectrum $k_x E_u$ of initial streamwise velocity fluctuations in three simulations of run 2. The spectrum $E_u$ is calculated with Fourier transform in the $x$ direction, and is plotted against the wavelength $\lambda _x=2{\rm \pi} /k_x$. Here, $k_x E_u$ is normalized with the r.m.s. of velocity fluctuations $u_{rms}$.

Figure 2

Table 1.Parameters of DNS and LES. Grid sizes $\varDelta _x$ and $\varDelta _y$ are constant while $\varDelta _{z}$ varies as a function of $z$. Table presents $\varDelta _{z}$ at $z=0$ and $\delta _u/2$, where $\delta _u$ is the shear layer thickness at $t=320$. The number of simulations repeated with different initial velocity fluctuations is denoted by $N_S$.

Figure 3

Figure 3.Density $\bar {\rho }$ on an $x$–$z$ plane at (a) $t=60$ and (b) $t=200$ in run 2. Only a small part of the computational domain is shown here.

Figure 4

Figure 4.Streamwise velocity fluctuations $\bar {u}'$ on the horizontal plane at the centre of the shear layer in run 2: (a) $t =2$; (b) $t =60$; (c) $t =120$; (d) $t =180$; (e) $t =240$; (f) $t =320$. Figure (f) depicts the size of $\delta _u$ and $10\delta _u$, where $\delta _u$ is the shear layer thickness defined with the mean velocity profile.

Figure 5

Figure 5.Comparisons between DNS (run 1) and LES (run 2). (a) Vertical profiles of mean velocity and density at $t=100$ and 240. (b) Temporal variations of root-mean-squared velocity fluctuations at $z=0$ and shear layer thickness $\delta _u$. Grey broken lines represent $u_{rms}$ in five simulations of LES.

Figure 6

Figure 6.(a) Energy spectrum of streamwise velocity $E_u$ at the centre of the shear layer at $t=60$, 140 and 320 in LES (run 2). Direct numerical simulation results at $t=140$ and 320 are also shown for comparison. Grey lines are $E_u$ in each simulation of LES (nine simulations). (b) Vertical profiles of turbulent kinetic energy dissipation rate at $t=140$ in DNS ($\varepsilon ^{(k)}$ in run 1) and LES ($\varepsilon _{G}^{(k)}+\varepsilon _{S}^{(k)}$ in run 2). The dissipation terms of grid scales ($\varepsilon _{G}^{(k)}$) and subgrid scales ($\varepsilon _{S}^{(k)}$) are also shown in the figure.

Figure 7

Figure 7.Temporal variations of (a) shear layer thickness $\delta _u$ and (b) r.m.s. streamwise velocity fluctuations at $z=0$ for ${Re}=40\,000$ (runs 3–6).

Figure 8

Table 2.Statistical properties of the stably stratified shear layer in LES. Values of $L_{O}$, $\eta$, ${Re}_b$ and $Ri_g$ are taken at the centre of the shear layers.

Figure 9

Figure 8.Premultiplied energy spectra of velocity and density, $k_{x}E_{\alpha }$ ($\alpha =u, v, w$ or $\rho$) in run 2 (${Re}=2000$, $Ri=0.06$), plotted against the streamwise wavelength $\lambda _x=2{\rm \pi} /k_x$ and the vertical location $z$: (a,d,g,j) $t=140$; (b,e,h,k) $t=240$; (c,f,i,l) $t=320$. (a–c) Streamwise velocity $E_u$; (d–f) spanwise velocity $E_v$; (g–i) vertical velocity $E_w$; (j–l) density $E_\rho$. The colour contours are normalized by the peak values at specific times.

Figure 10

Figure 9.Energy spectra of streamwise velocity at $z=0$ at $t=140, 240$ and 320 in runs 3–6 $({Re}=40\,000)$: (a) $Ri=0$ (run 3); (b) $Ri=0.02$ (run 4); (c) $Ri=0.06$ (run 5); (d) $Ri=0.1$ (run 6). For comparison of the spectra at different $t$, $k_x E_{u}$ is normalized by its maximum value $[k_x E_{u}]_{max}$ at each time instance.

Figure 11

Figure 10.(a) Energy spectra of turbulent kinetic energy ($E_k$) and density ($E_\rho$) at $t=260$ with the definition of short and long peak-wavelengths. The spectra multiplied by $k_x$ are normalized by turbulent kinetic energy $k_T= (\langle {\bar {u}^\prime}{^2}\rangle +\langle {\bar {v}^\prime}{^2}\rangle +\langle {\bar {w}^\prime}{^2}\rangle )/2$ or density variance $\langle {\bar {\rho }^\prime}{^2}\rangle$. (b) Temporal evolution of short and long peak-wavelengths in $E_k$ and $E_\rho$. These results are taken at the centre of the shear layer in run 2.

Figure 12

Figure 11.(a) Variation of $\xi _b$ and $\eta _b$ between $z=0$ and $0.5\delta _u$ at $t=140$ and 320 in run 2. An enlarged figure is also shown for clarity. (b) Plots of $\xi _b$ and $\eta _b$ associated with large and small peak-wavelengths of the turbulent kinetic energy spectrum at $z=0$ and $t=320$ in run 2. Three thin broken lines correspond to an axisymmetric state with one large eigenvalue, $\eta _b=\xi _b$, or one small eigenvalue, $\eta _b=-\xi _b$, and a two-dimensional state $\eta _b=(1/27+2\xi _b^3)^{1/2}$. Figures also include $\xi _b$ and $\eta _b$ obtained in the turbulent boundary layer at the momentum thickness Reynolds number of ${Re}_{\theta }=13\,000$ in Watanabe et al. (2019b) and at the centre of the non-stratified, fully developed turbulent mixing layer (Takamure et al.2018).

Figure 13

Figure 12.Fractional contributions of fluctuations associated with long and short peak-wavelengths ($F_L$ and $F_S$) of $E_k$ to (a) velocity and density variances, (b) vertical turbulent fluxes of streamwise velocity and density. (c) Reynolds stress associated with long and short peak-wavelengths, $R_{ij}^{(L)}$ and $R_{ij}^{(S)}$, for $(i,j)=(1,1)$ and $(1,3)$. Figure (d) shows $F_L$ and $F_S$ for the dissipation rates of turbulent kinetic energy and density variance. These results are taken at $z=0$ in run 2.

Figure 14

Figure 13.Spectral energy budgets of (a,b) streamwise velocity, (c,d) spanwise velocity, (e,f) vertical velocity and (g,h) density at $t=140$ in run 2. Results are shown for (a,c,e,g) $z=0$ and (b,d,f,h) $z=0.4\delta _u$. Each term in the energy budget equations is plotted against the streamwise wavelength $\lambda _x=2{\rm \pi} /k_x$.

Figure 15

Figure 14.Same as figure 13, but at $t=240$.

Figure 16

Figure 15.The rate of the change in energy spectra defined as $T_{\alpha }^{-1}=(\partial E_{\alpha }/\partial t)/E_{\alpha }$ for $\alpha =u, v, w$ and $\rho$ at $z=0$ in run 2: (a) $t=140$; (b) $t=240$.

Figure 17

Figure 16.Temporal variations of (a) production term $\hat {P}^{(u)}$ of $E_u$ and (b) pressure term $\hat {P}_{r}^{(w)}=\hat {\varPi }^{(w)}+\hat {T}_{P}^{(w)}$ of $E_w$ on the centreline ($z=0$) in run 2. The terms $k_{x}\hat {P}^{(u)}$ and $k_{x}\hat {P}_{r}^{(w)}$ are normalized by their maximum values, $[k_{x}\hat {P}^{(u)}]_{max}$ and $[k_{x}\hat {P}_{r}^{(w)}]_{max}$.

Figure 18

Figure 17.Structure functions of (a) streamwise velocity $\langle (\Delta u)^2\rangle$, (b) spanwise velocity $\langle (\Delta v)^2\rangle$ and (c) vertical velocity $\langle (\Delta w)^2\rangle$ at $t=320$. The broken lines represent $A \ln (r/\delta _u)+B$ with $A$ and $B$ obtained with the least square method applied for $z=0$ and $0.23\delta _u$. (d) Premultiplied energy spectrum of streamwise velocity, $k_xE_u$, at $t=320$. These results are taken at $z/\delta _u=0$, $0.23$ and $0.46$ in run 2.

Figure 19

Table 3.Computational parameters of DNS used for the assessment of the domain size effects.

Figure 20

Figure 18.Density profiles on (a) an $x$–$z$ plane and (b) the horizontal plane at $z=0$ at $t=60$ in run 7.

Figure 21

Figure 19.The same as figure 18 but for run 1: (a,b) an $x$–$z$ plane; (c,d) the horizontal plane at $z=0$. Profiles (a,c) are taken at $t=40$ while (b,d) are at $t=60$.

Figure 22

Figure 20.Pressure fluctuations $p'=p-\langle p\rangle$ normalized by the r.m.s. fluctuations $p_{rms}$ on the horizontal plane at $z=0$ in run 1: (a) $t=40$; (b) $t=60$. The same region is visualized as in figure 19.

Figure 23

Figure 21.(a) Shear layer thickness $\delta _u$ and momentum thickness $I_u$. Vertical profiles of (b) mean streamwise velocity $\langle u\rangle$ and density $\langle \rho \rangle$, and (c) mean velocity and density gradients, $\partial \langle u\rangle /\partial z$ and $\partial \langle \rho \rangle /\partial z$, at $t=240$.

Figure 24

Figure 22.Temporal evolutions of statistics: variances of (a) streamwise velocity, (b) spanwise velocity, (c) vertical velocity, (d) density, (e) turbulent kinetic energy dissipation rate, (f) vertically integrated production and buoyancy flux of the turbulent kinetic energy budget. The results in (a–e) are taken at $z=0$. In (a–d) the present DNS results are compared with DNS in Watanabe et al. (2017) with $(L_x,L_y,L_z)=(48,28,80)$.