1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %
3: % Thu Dec 9 17:11:13 CET 1999
4: %
5: % TITLE : Shear Effects in Non-Homogeneous Turbulence
6: % AUTHOR : F. Toschi, E. L\'ev\^eque and G. Ruiz-Chavarria
7: %
8: % 1) First version (date: Thu Dec 9 17:11:13 CET 1999)
9: % 2) Second revised version (date: Tue Jun 20 14:01:38 CEST 2000)
10: %
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12:
13: \documentstyle[epsfig,cite,aps,prl]{revtex}
14:
15: \twocolumn
16:
17: \newcommand{\be}{\begin{equation}}
18: \newcommand{\ee}{\end{equation}}
19: \newcommand{\bea}{\begin{eqnarray}}
20: \newcommand{\eea}{\end{eqnarray}}
21: \newcommand{\lp}{\left(}
22: \newcommand{\rp}{\right)}
23: \newcommand{\la}{\left\langle}
24: \newcommand{\ra}{\right\rangle}
25: \newcommand{\bm}[1]{\mbox{\boldmath $#1$}}
26: \newcommand{\bv}{{\mbox{\bm v}}}
27: \newcommand{\br}{{\mbox{\bm r}}}
28: \newcommand{\brhat}{\mbox{\bm {\hat r}}}
29: \newcommand{\bx}{\mbox{\bm x}}
30: \newcommand{\bnabla}{\mbox{$\bm \nabla$}}
31: \newcommand{\bdelta}{\mbox{$\bm \delta$}}
32: \renewcommand{\S}{{\cal S}}
33:
34: \begin{document}
35:
36: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname
37: @twocolumnfalse\endcsname
38:
39: \title{Shear Effects in Non-Homogeneous Turbulence}
40:
41: \author{F. Toschi$\;^{(1)}$, E. L\'ev\^eque$\;^{(2)}$
42: and G. Ruiz-Chavarria$\;^{(3)}$}
43: \address{
44: ${(1)\;}$University of Twente, Department of Applied Physics,\\
45: P.O. Box 217, 7500 AE Enschede, The Netherlands\\
46: INFM, Unit\'a di Tor Vergata, Roma, Italy\\
47: ${(2)\;}$Laboratoire de Physique CNRS, ENS de Lyon,
48: 69364 Lyon cedex 07, France \\
49: ${(3)\;}$Departamento de F\'{\i}sica,
50: Facultad de Ciencias, UNAM, 04510 Mexico D.F., Mexico
51: }
52:
53: \date{\today}
54: \maketitle
55: \vspace{0.2cm}
56: %\baselineskip 8ex
57: \begin{abstract}
58: %\baselineskip 8ex
59: Motivated by recent experimental and numerical results,
60: a simple unifying picture of intermittency in turbulent shear flows
61: is suggested.
62: Integral Structure Functions (ISF), taking
63: into account explicitly the shear intensity,
64: are introduced on phenomenological grounds.
65: ISF can exhibit a universal scaling behavior, independent of
66: the shear intensity.
67: This picture is in satisfactory agreement with both
68: experimental and numerical data.
69: Possible extension to convective turbulence and implication on
70: closure conditions for Large-Eddy Simulation
71: of non-homogeneous flows are briefly discussed.
72: \vskip0.2cm
73: \end{abstract}
74:
75: PACS: 47.27-i, 47.27.Nz, 47.27.Ak
76: \vskip 0.2cm
77: ]
78:
79: %%%%%%%%%%%
80:
81: Statistical properties of turbulent flows are
82: usually characterized in terms
83: of the scaling behavior of velocity Structure Functions (SF).
84: These quantities are
85: defined as the statistical moments of longitudinal velocity increments
86: across a separation $\br$ at the location $\bx$:
87: ${D}_p(\bx,\br)=\la \delta v(\bx,\br) ^p \ra$.
88: In homogeneous and isotropic turbulence,
89: ${D}_p(\bx,\br)$ only depends on the distance (or scale) $r$.
90: Experimental and numerical observations support the idea
91: that ${D}_p(r)$ display universal power-law dependence on $r$
92: in the so-called inertial range, i.e. ${D}_p(r) \sim r^{\zeta_p}$.
93: Universality refers here to the scaling exponents $\zeta_p$ being
94: independent of the stirring process of turbulence.
95: The $\zeta_p$ values are found to be in disagreement
96: with Kolmogorov's linear prediction (K41) $\zeta_p=p/3$ \cite{k41}.
97: The understanding of this correction to K41, usually
98: referred to as intermittency, has stimulated many phenomenological
99: and theoretical works during the last 40 years
100: (see \cite{Frisch} for a recent review).
101:
102: %%%%%%%%%%%%
103:
104: Only recently, interests in understanding intermittency
105: in non-homogeneous turbulent flows have started to emerge
106: (see
107: \cite{danaila,Toschi_prl,Toschi_pof,iuso,experimentos,Chilla_etc,Wesfreid,Benzi_shear}).
108: The major point is to understand how the phenomenology of
109: intermittency is modified, or can be extended,
110: in case of non-homogeneous flows.
111: A common characteristic of such flows, e.g. wall-bounded flows,
112: is the presence of a non-zero mean velocity gradient (usually called shear).
113: Note that shear does not necessarily imply inhomogeneity, e.g. in homogeneous shear,
114: straining and rotational flows.
115:
116: %%%%%%%%%%%%
117: Our investigation starts from the following key observations:
118: i) In presence of a strong shear, intermittency, defined as
119: the deviation of scaling exponents $\zeta_{p}$
120: from the linear law, is larger than in homogeneous and
121: isotropic turbulence.
122: ii) Relative scaling exponents, measured in very different flows
123: but in positions where the shear is strong enough,
124: seems to be very similar (universal).
125:
126: Data in Fig.~\ref{expo} fully confirm our two key observations.
127: They come from very different situations: near the wall
128: in a channel flow numerical simulation \cite{Toschi_prl,Toschi_pof}
129: and experiment \cite{iuso},
130: in the logarithmic sublayer
131: of a boundary layer flow \cite{experimentos},
132: near a strong vortex \cite{Chilla_etc}, in the wake of a cylinder \cite{Wesfreid}
133: and in a Kolmogorov flow \cite{Benzi_shear}.
134:
135: %%%%%%%%%%%%
136: \begin{figure}[!t]
137: \hskip -.7cm
138: \epsfig{file=fig1.eps,width=\hsize}
139: \caption{Intermittency corrections to scaling exponents,
140: $\zeta_p/\zeta_3-p/3$,
141: in homogeneous and isotropic turbulence ($\Box$),
142: and several turbulent shear flows ($\triangle$).
143: Intermittency corrections are significantly larger in presence of
144: shear and display universality.
145: }
146: \label{expo}
147: \end{figure}
148: %%%%%%%%%%%
149:
150: In order to provide a theoretical unifying framework
151: for non-homogeneous turbulent shear flows, we start
152: from the Navier-Stokes (N-S) equations.
153: The velocity field can be decomposed into
154: a mean value (average is meant on time)
155: plus a fluctuating part:
156: $\bv(\bx;t)={\overline {\bv}}(\bx)+\bv'(\bx;t)$.
157: It yields the usual Reynolds decomposition
158: \be
159: D_t v'_i+\S_{ij}(\bx) v'_j +
160: v_j' \partial_j v'_i - {\overline {v_j' \partial_j v'_i}}
161: = - \partial_i p' + \nu \Delta v'_i
162: \label{reynolds}
163: \ee
164: with $D_t=\lp \partial_t + {\overline v}_j \partial_j \rp$.
165: The shear is defined as
166: $\S_{ij}\lp\bx\rp= \partial_j {\overline v_i\lp\bx\rp}$
167: and will depend on the mean flow geometry.
168: In regions where $\S_{ij}=0$, e.g.
169: very far from the boundaries, turbulence can be considered as homogeneous.
170: Otherwise, the shear term $\S_{ij}(\bx) v'_j$ must be taken into
171: account. The results
172: displayed in Fig.~\ref{expo} show that the presence of this term
173: modifies significantly the statistical properties of turbulence.
174: In order to better hilight the physical implication of the shear,
175: we consider together the second and third terms
176: of the l.h.s. of (\ref{reynolds}),
177: defining the following
178: Integral Structure Functions (ISF):
179: \be
180: {\tilde D}_p(z,r)= \la \lp \delta v(z,r)^3
181: + \alpha r\cdot \S(z) \cdot \delta v(z,r)^2 \rp^{p/3}\ra,
182: \label{ISF}
183: \ee
184: where $\alpha$ is an empirical prefactor of order one.
185: These ISF are expected to take into account shear effects,
186: particularly at large scales (see next paragraph)
187: and to display a universal behavior.
188: We consider here {\em generic} situations in which
189: the shear reduces to $\S(z)=\partial_{z} {\overline v_{x}}$.
190: Such situation occurs near a rigid wall, where the principal
191: mean-velocity component aligns in the $x$-direction,
192: parallel to the boundary
193: \cite{Tennekes}.
194: The shear $\S(z)$ characterizes the variation of $v_{x}$
195: along the $z$-direction, i.e. as one moves off the wall.
196: Finally, we consider increments in the direction of the mean flow, i.e.
197: orthogonal to the shear direction.
198:
199: ISF reduce on two different SF when either
200: the first or the second term dominates.
201: These two terms will exactly balance at scale $L_{\S}$
202: such that
203: $ {{\delta v(L_{\S})} / {L_{\S}}} = \alpha~\S.
204: \label{Ls} $
205: At scales $r \ll L_{\S}$, shear effects become negligible and
206: homogeneous and isotropic scalings are expected.
207: Kolmogorov's scaling then yields
208: $\delta v(r) \sim \epsilon^{1/3} r^{1/3}$, where
209: $\epsilon$ denotes the mean energy dissipation rate.
210: By extending this similarity relation to the scale $r=L_{\S}$, one
211: obtains the usual dimensional estimate for the
212: shear length scale $L_{\S} \sim \lp \epsilon/\S^3 \rp^{1/2}$ \cite{Hinze}.
213: In the logarithmic layer of a plane near-wall flow
214: it yields $L_\S(z) \sim z$ \cite{experimentos}.
215: Roughly speaking, $L_{\S}$ can be viewed as the size of small-scale
216: eddies, whose turn-over time equals the shear time scale $1/\S$,
217: imposed by the flow geometry and the stirring process
218: of turbulence at large scales.
219: Note that this estimate of the shear length-scale stems out
220: from dimensional analysis.
221: In practice, there may be a prefactor in the expression of $L_s$: this
222: is taken into account by the coefficint $\alpha$.
223: From previous reasoning it follows that
224: ${\tilde D}_p(r) \sim {D}_p(r)$ for $r\ll L_s$ and
225: ${\tilde D}_p(r) \sim \lp r \S\rp^{p/3} {D}_{2p/3}(r)$ for $r\gg L_s$.
226:
227: Another way to see that the central objects, in presence of
228: shear, are the ISF, comes from the generalization of Yaglom's
229: equation to homogeneous-shear flows, i.e. with $\S(z)=\S$
230: \cite{Hinze,Struglia_tesi}:
231: \bea
232: \nonumber -{4 \over 5}\epsilon r &=& {D}_3(r) - 6\nu {{d~{D}_2(r)} \over
233: {dr}} +\\
234: &-&{2\S\over {r^4}}\int^r_0{dx \; x^4 \;\overline {v_x'(x_0)\cdot v_z'(x_0+x)}}.
235: \eea
236: Supposing that this relation can be
237: generalized, along the same line of idea of the
238: Kolmogorov's Refined Similarity Hypothesis \cite{k62}, one obtains
239: \be
240: \delta v^3(r) + \alpha r \cdot \S \cdot \overline {v_x'(x) \cdot v_z'(x+r)}
241: \sim \varepsilon(r) \cdot r,
242: \ee
243: where $\varepsilon(r)$ denotes the coarse-graining of the energy dissipation field
244: $\epsilon=\frac{\nu}{2} \sum_{i,j} \lp\partial_i v_j\rp^2$, at
245: scale $r$.
246: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
247: \begin{figure}[!b]
248: \hskip -.5cm
249: \epsfig{file=fig2.eps,width=\hsize}
250: \caption{In the turbulent boundary layer at the distance $z^+=102$.
251: ${D}_3$ and ${D}_6$ ($\triangle$) are compared with ${\tilde D}_3$
252: and ${\tilde D}_6$ ($\star$).
253: The scale $r$ has been renormalized by the characteristic shear length-scale $L_s$.
254: The solid lines passing through ${\tilde D}_3$ and ${\tilde D}_6$
255: indicate the expected {\it homogeneous and isotropic} power-laws,
256: respectively $\zeta_3=1$ and $\zeta_6=1.78$.
257: For comparison, the dashed line has slope $2$.
258: }
259: \label{comp1}
260: \end{figure}
261: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
262: Simplifying, on pure dimensional grounds, the velocity-cross correlation
263: $\overline {v_x' v_z'}$, with $\delta v(r)^2$
264: one ends up again with ${\tilde D}_p(r)$.
265: Furthermore, we propose the following Refined
266: Similarity Hypothesis
267: $$
268: %\label{grksh}
269: {\tilde D}_p(r)= \la \lp \delta v(r)^3 + \alpha r \cdot \S \cdot
270: \delta v(r)^2 \rp^{p/3}
271: \ra \sim \la \varepsilon(r)^{p/3}\ra r^{p/3}.
272: $$
273: \noindent
274: This formulation is consistent, in the limiting cases of strong and
275: negligible shear, with some recent findings (see
276: \cite{Toschi_prl,Toschi_pof}). In addition to that, the ISF should
277: be able to abridge smoothly between these two limiting regimes, i.e.
278: \bea
279: \label{rksh1}r\ll L_{\S}\;&:&\;\;\;\;{D}_p(r) \sim \la
280: \varepsilon(r)^{p/3}\ra \cdot r^{p/3}\\
281: \label{rksh2}{\rm and} \;\;\;\; r\gg L_{\S}\;&:&\;\;\;\;{D}_p(r) \sim \la
282: \varepsilon(r)^{p/2}\ra.
283: \eea
284: Eqn. (\ref{rksh1}) is in agreement with the restoration of
285: homogeneity and isotropy at small scales.
286: For $r \gg L_{\S}$, eqn.
287: (\ref{rksh2})
288: gives ${D}_2(r)\sim const$ (modulo possible logarithmic corrections), yielding
289: for the energy spectrum $E(k)\sim k^{-1}$ (a relation
290: suggested long time ago in \cite{Hinze}).
291:
292: In the previous picture, it is assumed that the shear length scale
293: $L_\S$ remains larger than the dissipation length scale $\eta$.
294: The dissipation field
295: $\varepsilon(r)$ is then expected to display the same scaling properties as
296: in homogeneous and isotropic turbulence.
297: However, in regions where $L_\S \lesssim \eta$ (very
298: close to the wall)
299: it is expected that shear effects act down to the dissipative scale
300: and therefore can modify the scaling behavior of $\varepsilon(r)$.
301: The scalings of ${D}_p(r)$
302: should then also change according to (\ref{rksh2}).
303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
304: \begin{figure}[!t]
305: \hskip -.7cm\epsfig{file=fig3.eps,width=\hsize}
306: \caption{From the boundary layer experiment,
307: third-order structure function at various
308: distances from the wall: $z^+=37$ ($\circ$), $z^+=124$ ($\triangle$)
309: and $z^+=233$ ($\Box$).
310: The scaling properties of ${D}_3$ (SF) do depend on the distance $z^+$.
311: On the contrary, ${\tilde D}_3$ (ISF) displays
312: the same scaling behavior for all $z^+$.
313: %The shear value, entering in the definition of ${\tilde S}_3$,
314: %has been estimated from the mean velocity profile.
315: %The coefficient $\alpha=1/5$ is fixed.
316: The dashed line has slope $1$. The curves have been shifted
317: vertically for convenience.
318: }
319: \label{alls3}
320: \end{figure}
321: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
322: The relevance of Integral Structure Functions for describing the
323: scaling properties of non-homogeneous shear flows is now tested
324: on both experimental and numerical data.
325: The experiment, performed in the recirculating wind tunnel of ENS-Lyon,
326: consists in a turbulent boundary-layer flow over a smooth horizontal plate (see \cite{experimentos}
327: for details about the experimental apparatus).
328: Velocity measurements are carried out at various elevations from the
329: plate in the logarithmic turbulent sublayer \cite{Tennekes}.
330: Numerical results are obtained from a direct simulation
331: of the Navier-Stokes equations
332: in a rectangular channel flow (see \cite{Toschi_prl} for
333: details).
334:
335:
336: In Fig.~\ref{comp1}, ${D}_3$ and ${D}_6$,
337: measured in the logarithmic boundary sublayer, are compared with the
338: corresponding ${\tilde D}_3$ and ${\tilde D}_6$.
339: The shear $\S(z)=\partial_{z} {\overline v_{x}}$
340: has been estimated from the mean velocity profile.
341: In standard non-dimensional variables \cite{Schl}, namely $z^+=v_*z/\nu$ and
342: $ v^+=\bar v_x/v_*$ where $v_*$ is the characteristic
343: velocity of the viscous sublayer, our data are well fitted
344: by the logarithmic profile
345: $ v^+(z^+)={(1/{\kappa})}\log(z^+)+B $.
346: We obtain $\kappa\simeq 0.4$ and $B\simeq 5.26$
347: in agreement with previously
348: reported results \cite{experimentos}.
349: For what concerns the coefficient $\alpha$, all our results have been obtained with the fixed
350: value $\alpha \approx 0.2$.
351: Note that the constant $\alpha$ is not (a priori) intended to be universal
352: but may strongly depend on the geometry and stirring process of the flow.
353: However, it is expected to remain of order unity.
354: In practice, the value of $\alpha$ has been extracted from data
355: by requiring that one third-order ISF scale as $r$ (see next paragraph).
356: Corresponding estimates of $L_s$ are indicated in all figures.
357: Finally, one must point out that velocity increments have been estimated
358: in the direction of the mean flow by use
359: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
360: \begin{figure}[!t]
361: \hskip -.7cm
362: \epsfig{file=fig4.eps,width=\hsize}
363: \caption{In the experimental boundary layer at $z^+=102$.
364: SF and ISF, compensated by homogeneous and isotropic scalings,
365: are displayed for p=$1\dots 6$.
366: }
367: \label{compensated}
368: \end{figure}
369: \noindent
370: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
371: of the Taylor hypothesis.
372: Both ${D}_3$ and ${D}_6$ exhibit
373: a power-law dependence on $r$ but
374: the scaling exponents are clearly different from those observed
375: in homogeneous and isotropic (h-i)
376: turbulence, respectively, $\zeta_3 \simeq 1$ and
377: $\zeta_6 \simeq 1.78$.
378: On the other hand, the corresponding ISF exhibit power-laws in good
379: agreement with h-i scalings (up to very large scales).
380:
381: In Fig.~\ref{alls3}, third-order SF and ISF are displayed
382: for various distances from the wall.
383: We notice that scaling behavior of ${D}_3$ changes with $z^+$.
384: On the contrary, all the corresponding ${\tilde D}_3$ display
385: the same power-law scaling with exponent $1$.
386: We recall that
387: the coefficient $\alpha$ is kept constant and the shear is
388: estimated from the mean velocity profile; there is
389: no adjustable parameter.
390:
391: We now report a sharper test:
392: SF and ISF, compensated by h-i power-law scalings,
393: are displayed in Fig.~\ref{compensated}
394: for $p=1,\dots,6$ at distance $z^+=102$.
395: A departure from h-i scalings is clearly observed for SF.
396:
397: On the contrary, ${\tilde D}_p(r)/r^{\zeta_p}$ exhibits a plateau up
398: to very large scales, indicating that ISF roughly behave as $r^{\zeta_p}$.
399: In other terms, ISF compensate shear effects
400: and restore, via the extra term
401: $\alpha r\cdot \S \cdot \delta v(r)^2$, the h-i scalings.
402: Finally, the same test, made on numerical data,
403: is reported in Fig.~\ref{test_nu} for the sake of comparison.
404: Data are obtained at distance $z^+=25$ from the wall, i.e.
405: where the shear is strong. Results are in reasonable
406: agreement with those of Fig.~\ref{compensated},
407: despite the lower resolution of the numerical simulation.
408: We would like to underline the major points of our study.
409: Our description relates the scaling properties of velocity fluctuations
410: in non-homogeneous shear flows, to those
411: of the coarse-grained dissipation rate $\varepsilon(r)$.
412: Provided that the shear length-scale remains much larger than the
413: dissipative length-scale, we are inclined to believe
414: that the scaling properties of $\varepsilon(r)$ remain flow-independent;
415: the dissipation process, operating on very
416: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
417: \begin{figure}[b]
418: \hskip -.5cm\epsfig{file=fig5.eps,width=\hsize,height=6cm}
419: \caption{As in Fig.~\ref{compensated} but from numerical data at $z^+=25$.
420: }
421: \label{test_nu}
422: \end{figure}
423: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
424: \noindent
425: small-scales,
426: is mainly insensitive to the presence of the shear.
427: We then claim that the scalings of velocity fluctuations
428: are different in sheared regions only
429: because the similarity between $\delta v(r)$
430: and $\varepsilon(r)$ changes form (see (\ref{rksh1}) and (\ref{rksh2})).
431: When $L_S\lesssim\eta$, shear effects acting down to
432: dissipative scales are expected to modify the scaling behavior
433: of $\varepsilon(r)$.
434: Nonetheless, we observe in Fig.~\ref{expo} that relative scalings
435: exponents of velocity structure functions remain universal.
436:
437: The introduction of ISF relies on quite simple dimensional arguments.
438: However, they have proved to be valuable (preliminary) tools in order
439: to capture (at leading order)
440: shear effects on the scaling behavior of velocity
441: structure functions.
442: We believe that some refinements in the definition
443: of ISF may be possible,
444: e.g. considering the cross correlation
445: $\overline {v_x' v_z'}$ instead of $\delta v(r)^2$: however
446: ISF are of practical interest as they are easily accessible
447: experimentally.
448:
449: A possible extension of ISF applies to thermal turbulent
450: convection, the buoyancy term $a g \delta T(r)$
451: playing a similar role of $\S \cdot \delta v(r)$.
452: In that case, the ISF would read
453: $${\tilde D}^{\mbox{\scriptsize RB}}_p(r)= \la \lp\delta v(r)^3 +
454: \alpha r \cdot a g \delta T(r) \cdot \delta v(r)\rp^{p/3}\ra$$
455: and are expected to take into account buoyancy effects at
456: scales larger than the Bolgiano length-scale.
457:
458:
459: An important application of our findings concerns Large Eddy Simulations (LES).
460: Usual eddy viscosity models are known to fail close to the boundary.
461: As one moves near the boundary,
462: the shear length-scale $L_\S$ becomes smaller and smaller.
463: While $L_\S$ remains larger than the cut-off scale $\Delta$ of the LES,
464: usual closure conditions, based on homogeneous
465: and isotropic turbulent dynamics, remain acceptable.
466: However, when $L_\S$ becomes comparable or smaller than $\Delta$,
467: shear effects must be taken into account and the closure condition
468: should be modified.
469: An alternative consists in decreasing the mesh-size near the
470: wall so that $L_\S$ always remains larger than $\Delta$.
471: More simply, our study suggests to consider
472: $\lp \delta v(r)^3+ \alpha r \S \delta v(r)^2 \rp^{1/3}$
473: instead of $\delta v(r)$ in closure relations.
474: Along this line of idea,
475: the Smagorinski's closure condition \cite{smago}
476: can be generalized in order
477: to uniformly take into account shear effects.
478: The eddy viscosity $\nu_{{\mbox{\footnotesize eddy}}}$ then reads
479: $$\nu_{{\mbox{\footnotesize eddy}}}= C_s~
480: \Delta^2\cdot \lp{\overline S}+ \alpha \S \rp,$$
481: where the extra term $\alpha \S$ takes care of shear effects.
482: $C_s$ is the empirical constant
483: of the Smagorinski's closure and $\overline S$ denotes
484: the rate of strain on scale $\Delta$.
485:
486: \noindent
487: {\bf Acknowledgments:}
488: F. T. would like to thank S. Ciliberto for his kind hospitality at
489: ENS-Lyon.
490: G. R.-C. and E. L. acknowledge the ECOS committee and CONACYT
491: for their financial support under the project No M96-E03.
492: G. R.-C. also acknowledge DGAPA-UNAM for partial support under the project
493: IN-107197.
494:
495: \begin{thebibliography}{99}
496: \bibitem{k41} A. N. Kolmogorov, {\em CR. Acad. Sci. USSR} {\bf 30}, 301 (1941);
497: {\em CR. Acad. Sci. USSR} {\bf 31}, 538 (1941);
498: {\em CR. Acad. Sci. USSR} {\bf 32}, 16 (1941).
499: \bibitem{Frisch}
500: U. Frisch, {\em Turbulence: the legacy of A. N. Kolmogorov},
501: Cambridge University Press, Cambridge (England) (1995).
502: \bibitem{danaila}
503: L. Danaila, F. Anselmet, T. Zhou and R. A. Antonia,
504: {\em A Generalization of Yaglom's Equation
505: which accounts for the Large-Scale Forcing in Heated Grid Turbulence}
506: submitted to J. Fluid Mech. (1999).
507: \bibitem{Toschi_prl}
508: F. Toschi, G. Amati, S. Succi, R. Benzi and R. Piva,
509: Phys. Rev. Lett., {\bf 82}, 5044 (1999).
510: \bibitem{Toschi_pof}
511: R. Benzi, G. Amati, C. M. Casciola, F. Toschi
512: and R. Piva, Phys. of Fluids {\bf 11}, 1284
513: (1999).
514: \bibitem{iuso}
515: G. Iuso, R. Camussi, M. Onorato, to appear in Phys. Rev. E (1999).
516: \bibitem{experimentos}
517: G. Ruiz-Chavarria, S. Ciliberto, C. Baudet and E. L\'ev\^eque,
518: {\em Scaling Properties of the Streamwise
519: Component of Velocity in a Turbulent Boundary Layer},
520: to appear in Physica D (2000).
521: \bibitem{Chilla_etc}
522: F. Chill\'a and J.-F. Pinton, {\em Advances in Turbulence VII},
523: U. Frisch Ed. (Kluwer
524: Academic Publishers), 211-214 (1998).
525: \bibitem{Wesfreid}
526: E. Gaudin, B. Protas, S. Goujon-Durand,
527: J. Wojciechowski, and J. E. Wesfreid,
528: Phys. Rev. E {\bf 57}, R9 (1998).
529: \bibitem{Benzi_shear}
530: R. Benzi, M. V. Struglia and R. Tripiccione, Phys. Rev. E {\bf 53}, 6 (1996).
531: \bibitem{Tennekes}
532: H. Tennekes and J. L. Lumley, {\it A first course in turbulence}, MIT Press
533: Cambridge (Massachussets) (1972).
534: \bibitem{Hinze}
535: O. J. Hinze, {\em Turbulence}, Mc Graw-Hill, New York (1959).
536: \bibitem{Struglia_tesi}
537: M.V. Struglia, ``Ph.D. thesis'', Universit\'a di Roma, Tor Vergata, Roma, Italy (1996).
538: \bibitem{k62} A. N. Kolmogorov, {\em J. Fluid Mech.} {\bf 62}, 82 (1962).
539: \bibitem{Schl}
540: H. Schlichting, {\it Boundary Layer Theory}, McGraw-Hill New-York (1968).
541: \bibitem{smago}
542: J. Smagorinski, Mon. Weather Rev. {\bf 91}, 99-164 (1963).
543: \end{thebibliography}
544:
545: \clearpage
546: \end{document}
547:
548: \begin{center}
549: \begin{figure}[!t]
550: \hskip -.7cm
551: \epsfig{file=fig1.eps,width=\hsize}
552: \caption{Intermittency corrections to scaling exponents,
553: $\zeta_p/\zeta_3-p/3$,
554: in homogeneous and isotropic turbulence ($\Box$),
555: and several turbulent shear flows ($\triangle$).
556: Intermittency corrections are significantly larger in presence of
557: shear and display universality.
558: }
559: \label{expo}
560: \end{figure}
561: \end{center}
562:
563: \begin{center}
564: \begin{figure}[!b]
565: \hskip -.5cm
566: \epsfig{file=fig2.eps,width=\hsize}
567: \caption{
568: In the turbulent boundary layer at the distance $z^+=102$.
569: ${D}_3$ and ${D}_6$ ($\triangle$) are compared
570: with ${\tilde D}_3$
571: and ${\tilde D}_6$ ($\star$).
572: The scale $r$ has been renormalized by
573: the characteristic shear length-scale $L_s$.
574: The solid lines passing through ${\tilde D}_3$ and ${\tilde D}_6$
575: indicate the expected {\it homogeneous and isotropic} power-laws,
576: respectively $\zeta_3=1$ and $\zeta_6=1.78$.
577: For comparison, the dashed line has slope $2$.
578: }
579: \label{comp1}
580: \end{figure}
581: \end{center}
582:
583:
584: \begin{center}
585: \begin{figure}[!t]
586: \hskip -.7cm\epsfig{file=fig3.eps,width=\hsize}
587: \caption{From the boundary layer experiment,
588: third-order structure function at various
589: distances from the wall: $z^+=37$ ($\circ$), $z^+=124$ ($\triangle$)
590: and $z^+=233$ ($\Box$).
591: The scaling properties of ${D}_3$ (SF) do depend on the distance $z^+$.
592: On the contrary, ${\tilde D}_3$ (ISF) displays
593: the same scaling behavior for all $z^+$.
594: The dashed line has slope $1$. The curves have been shifted
595: vertically for convenience.
596: }
597: \label{alls3}
598: \end{figure}
599: \end{center}
600:
601: \begin{figure}[!t]
602: \hskip -.7cm
603: \epsfig{file=fig4.eps,width=\hsize}
604: \caption{In the experimental boundary layer at $z^+=102$.
605: SF and ISF, compensated by homogeneous and isotropic scalings,
606: are displayed for p=$1\dots 6$.
607: }
608: \label{compensated}
609: \end{figure}
610: \noindent
611:
612: \begin{figure}[b]
613: \hskip -.7cm\epsfig{file=fig5.eps,width=\hsize}
614: \caption{Same plot as in Fig.~\ref{compensated} from
615: the numerical channel flow at $z^+=25$.
616: }
617: \label{test_nu}
618: \end{figure}
619:
620: \end{document}
621: