Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T04:10:23.474Z Has data issue: false hasContentIssue false

Reactive scalars in incompressible turbulence with strongly out of equilibrium chemistry

Published online by Cambridge University Press:  15 March 2022

Wenwei Wu
Affiliation:
UM-SJTU Joint Institute, Shanghai Jiao Tong University, Shanghai 200240, PR ChinaUniv. Lille, ULR 7512 Unité de Mécanique de Lille - Joseph Boussinesq (UML), F-59000 Lille, FranceUniv. Lille, Univ. Littoral Cote d'Opale, CNRS, Laboratory of Oceanology and Geosciences, UMR LOG 8187, Wimereux, France
Lipo Wang*
Affiliation:
UM-SJTU Joint Institute, Shanghai Jiao Tong University, Shanghai 200240, PR China
Enrico Calzavarini
Affiliation:
Univ. Lille, ULR 7512 Unité de Mécanique de Lille - Joseph Boussinesq (UML), F-59000 Lille, France
François G. Schmitt
Affiliation:
Univ. Lille, Univ. Littoral Cote d'Opale, CNRS, Laboratory of Oceanology and Geosciences, UMR LOG 8187, Wimereux, France
*
†Email address for correspondence: lipo.wang@sjtu.edu.cn

Abstract

We study the statistical properties of scalar fields undergoing reversible chemical reactions in a turbulent environment by means of numerical simulations. To produce strong chemical fluctuations in a wide region of the domain, an original flow configuration has been proposed, where the species are supplied from buffer boundaries with adjustable thickness, while the flow is developed homogeneous and isotropic turbulence in a periodic domain. With the presence of the mean scalar gradient in the bulk region, the strength of turbulent advection is comparable with the chemical source, which is quantified by the Damköhler number. Our analysis focuses on the global and spatial properties of the reactive scalars in their statistically steady regime. We show how for the case of a second-order reaction such features can be connected to the properties of a non-reactive scalar field advected in the same system. Analytical predictions of the scalar moments in the fast reaction regime agree satisfactorily with the direct numerical simulation results. In comparison with the existing results of the isotropic turbulence case, we conclude that the scalar correlation is jointly determined by both the chemical source and the flow configuration. Moreover, the chemical reaction also plays an important role in determining the scalar energy spectra.

Information

Type
JFM Papers
Copyright
© The Author(s), 2022. Published by Cambridge University Press
Figure 0

Figure 1.Schematic diagram of the flow configuration and computational domain. For the scalar quantities, the periodic boundary conditions are set along $x$ and $y$ directions, while a Dirichlet boundary condition is used along the $z$ direction. The shadowed layers near the boundaries are the ‘buffer layers’ generated artificially, in which the quantities of scalars are close to the preset boundary values, as defined in (2.6). The part between buffer layers is denoted as the bulk region. Such a set-up is statistically stationary and ensures the local positiveness of scalar concentrations because it evolves directly the scalar concentration field rather than its fluctuation with respect to a mean background profile.

Figure 1

Table 1.Non-dimensionalized parameters for the simulations: $Re= u'H/\nu$ is the Reynolds number based on large scale, where $u'$ is the single-component r.m.s. velocity, $H$ is the size of the simulation domain, $\nu$ is the viscosity; $Re_{\lambda } = u'\lambda /\nu$ is the Taylor scale $\lambda$-based Reynolds number; $Sc$ is the Schmidt number ($\nu /D$); $N^3$ is the number of total grid points; $|\boldsymbol {k}|_{max}\cdot \eta$ is the resolution condition, where $|\boldsymbol {k}|_{max}$ is the maximum amplitude of the wavenumber kept by the dealiasing procedure, $\eta$ if the Kolmogorov length scale; $\tau _\eta$ is the Kolmogorov time scale; $\varGamma =Da_1/Da_2$, with $Da_1$ and $Da_2$ as the Damkholer numbers for forward and backward reactions, respectively; $L_I$ is the integral length scale; $T_I$ is the integral time scale; $\Delta t$ is the numerical time step.

Figure 2

Figure 2.The reaction rates computed from the mean quantities as functions of $z$. The solid lines are for the forward reaction rates $Da_1\langle R_1R_2\rangle$ and the dashed lines are for the backward reaction rates $Da_2\langle P\rangle$. A clear difference can be observed. Vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 3

Figure 3.The three-dimensional instantaneous snapshots of (a) $R_1$, and (b) reaction rate ($Da_1R_1R_2-Da_2P$) on the isosurface of $R_1=0.5$, under the condition of $\varGamma =10$.

Figure 4

Figure 4.Evolution of the spatial averages of scalar concentrations and r.m.s. of scalar fluctuations in the bulk region with $z\in [\delta, 1-\delta ]$ for the case of $\varGamma = 10$. Time is normalized by the integral time $k/\epsilon$ with $k = 3u'^2/2$. The dashed vertical lines mark the initial time for the computation of statistical quantities.

Figure 5

Figure 5.Numerical profiles of (a) the mean, and (b) the r.m.s. of $T$ with different buffer layer thickness $\delta$. The vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 6

Figure 6.(a) Turbulent diffusivity calculated from (4.1). (b) Theoretical prediction of the mean of $T$ (dashed lines) compared with the DNS results (solid lines with the same colour). The vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 7

Figure 7.Dependence of p.d.f.s at $z=1/2$ on $\varGamma$ for (a) $R_1$ and $T$, and (b) $P$. The insert panel in (b) plots the peak of $p_P$ as a function of $\varGamma$.

Figure 8

Figure 8.Probability density functions of (a) $R_1$ under the condition of $\varGamma =10$ and (b) $T$ at different positions.

Figure 9

Figure 9.Probability density functions of the reaction rate ($Da_1R_1R_2-Da_2P$) under the conditions of different $\varGamma$, at the position of $z=1/2$.

Figure 10

Figure 10.The instantaneous two-dimensional snapshot of reaction rate at the position of $z=1/2$. The upper row (a–c) corresponds to $\varGamma =100$; the lower row (d–f) corresponds to $\varGamma =1$. The first column (a,d) shows the net reaction rate ($Da_1R_1R_2-Da_2P$); the second column (b,e) shows the forward reaction rate ($Da_1R_1R_2$); the third column (c,f) shows the backward reaction rate ($Da_2P$).

Figure 11

Figure 11.The mean (solid lines) and r.m.s. (dashed lines) of (a) $R_1$ compared with $T$; (b) $P$ under the conditions of different $\varGamma$ as functions of position ($z$). The main panel of (c) shows the mean profile of $P$ normalized by its maximum, whose function as $\varGamma$ is plotted in the inset plot. There is a perfect superposition for all $\varGamma$ values. In all the plots, the vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 12

Figure 12.Dependence of (a) the scalar means and (b) the scalar fluctuations on $\varGamma$. There is no scaling behaviour in different flow regions.

Figure 13

Figure 13.The scalar mean (a) $\langle R_1\rangle$ and (b) $\langle P\rangle$ as functions of $z$ obtained from theoretical analysis (dashed lines) based on (4.15) and DNS (solid lines with the same colours). The grey dashed lines are from the theoretical prediction at infinitely large $Da_1$ according to (4.12). We see that the prediction for $\varGamma =\infty$ is close to the curves for $\varGamma =100$ and also the predictions for large $\varGamma$ are close to the DNS results when $\varGamma =10$, 30, 100. The vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 14

Figure 14.The scalar fluctuations (a) $\langle R_1'^2\rangle ^{1/2}$ and (b) $\langle P'^2\rangle ^{1/2}$ as functions of $z$ obtained from theoretical analysis (dashed lines) and DNS (solid lines with the same colours). The grey dashed lines are from the theoretical prediction at infinitely large $Da_1$ according to (4.13). We see that the prediction for $\varGamma =\infty$ is close to the curves for $\varGamma =100$ and also the predictions for large $\varGamma$ are close to the DNS results when $\varGamma =10$, 30, $100$. The vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 15

Figure 15.Direct numerical simulations of the correlation coefficients between (a) $R_1$ and $R_2$; (b) $R_1$ and $P$ as functions of $z$, for different $\varGamma$ cases. The vertical dotted lines mark the interfaces between the buffer layers and the bulk region.

Figure 16

Figure 16.The ratio of the energy spectra of (a) $R_1$ at $z=\frac {1}{4}$; (b) $R_1$ at $z=\frac {1}{2}$; (c) $R_1$ at $z=\frac {3}{4}$; (d) $P$ at $z=\frac {1}{4}$; (e) $P$ at $z=\frac {1}{2}$; (f) $P$ at $z=\frac {3}{4}$ to the energy spectrum of $T$ at the same $z$.

Figure 17

Figure 17.Two-dimensional coherency spectra between $R_1$ and $R_2$ at the positions of (a) $z=1/4$, (b) $z=1/2$ and (c) $z=3/4$.