1:
2:
3:
4: %\includeonly{}
5: \input{psfig.sty}
6: \documentstyle[12pt]{article}
7: \newfont{\feff}{cmti10}
8: \topmargin=-0.4in
9: \oddsidemargin=-0.0in
10: \textheight=8.75in
11: \textwidth=6.5 in %for preview, use 6.5in to print on laserprinter
12: \def\baselinestretch{1.5}
13: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
14:
15: \begin{document}
16:
17:
18:
19: \title{Self-sustained oscillations in homogeneous shear flow. }
20:
21:
22:
23: \author{ Victor Yakhot\\
24: Department of Aerospace and Mechanical Engineering\\
25: Boston University,
26: Boston, MA 02215 }
27:
28:
29:
30: \maketitle
31:
32:
33:
34: ${\bf Abstract}.$
35: \noindent Generation of the
36: large-scale
37: coherent vortical structurs in homogeneous shear flow couples dynamical
38: processes of energy and enstrophy production. In the large rate
39: of strain limit,
40: the simple estimates of the contributions to
41: the energy and enstrophy equations result in a dynamical system, describing
42: experimentally and numerically observed
43: self-sustained non-linear
44: oscillations of energy and enstrophy. It is shown that
45: the period of these oscilaltions is independent upon the box size and
46: the energy
47: and enstrophy fluctuations are strongly correlated.
48:
49:
50: \newpage
51: Due to its seeming simplicity, the problem of
52: homogeneous shear flow has widely been used as a benchmark for
53: nu merical and experimental
54: tests of various closures for turbulence modelling. All early closures were
55: based on the Kolmogorov ideas developed for statistically steady
56: isotropic and homogeneous small-scale
57: turbulence interacting with the non-universal large-scale flow-field.
58: It became clear that to validate this physically appealing concept,
59: one had to verify and understand the symmetries and other
60: statistical properties of the small-scale velocity fluctuations in the
61: real-life
62: flows. This was the main focus of the experimental studies of
63: homogenious shear flow [1]-[6]. The interest in this model flow is also
64: related to the recent
65: numerical investigations
66: which revealed coherent structures resembling those, responsible for turbulence
67: production, in the wall- sheared flows
68: [7]-[13]. This system has also often been used for calibration of
69: various constants in semi-empirical turbulence models [15],[16].
70:
71: The problem is formulated as follows:
72: consider a flow in a cube of a side $a$, so
73: that $-a<x_{i}<a$. The velocity
74: field
75:
76: \begin{equation}
77: {\bf v}({\bf x},t)=U(y){\bf e_{1}}+{\bf u}({\bf x},t)
78: \end{equation}
79:
80: \noindent with the imposed mean velocity $<\overline{{\bf v}}>=
81: U(y){\bf e}_{1}=Sy{\bf e}_{1}$ (the definition of the averaging operation will
82: be introduced below). The vorticity is defined then:
83:
84: \begin{equation}
85: \Omega=-S{\bf e}_{3}+\omega
86: \end{equation}
87:
88: The equations of motion for the fluctuating components of velocity and
89: vorticity
90: are (density $\rho=1$):
91:
92: \begin{equation}
93: \partial_{t}{\bf u}~ +~{\bf u}\cdot\nabla {\bf u}=-Sv{\bf e}_{1}-\nabla
94: p-U(y)\partial_{x}{\bf u}+\nu\nabla^{2}{\bf u},
95: \end{equation}
96:
97: \begin{equation}
98: \partial_{t}\omega+{\bf u}\cdot\nabla\omega=\omega\cdot\nabla {\bf
99: u}-S(-\partial_{z}w{\bf e}_{3}+\partial_{x}w {\bf e}_{1}+\partial_{z}v{\bf e}_{2})-Sy\partial_{x}\omega
100: \end{equation}
101:
102: \noindent and
103: \begin{equation}
104: \nabla\cdot{\bf u}=0
105: \end{equation}
106:
107:
108: \noindent
109: The $x$, $y$ and $z$-components
110: of velocity field are denoted hereafter as $u$, $v$ and $w$, respectively.
111: Let us define the averaging
112: operations:
113:
114: $$F=<\overline{F({\bf x},t)}>=\frac{1}{T}\frac{1}{V}\int_{0}^{T}\int_{V}d{\bf
115: x}dt
116: F({\bf x},t)$$
117:
118: \noindent in the limit $V=a^{3}\rightarrow \infty; T\rightarrow \infty$. The
119: statistically steady state is assumed here. The spacial averaging is defined
120: as
121:
122: $$F(t)=\overline{F({\bf x},t)}=\frac{1}{V}\int_{V}d{\bf
123: x}
124: F({\bf x},t)$$
125:
126:
127:
128: \noindent The kinetic energy
129: equation is ($\nu\rightarrow 0$):
130:
131: \begin{equation}
132: \partial_{t}{\cal K}+\frac{1}{2}\overline{ u_{i}\nabla_{i}
133: u_{j}^{2}}=-\tau_{uv}S-\nabla_{i}\overline
134: {pu_{i}}-{\cal E}
135: \end{equation}
136:
137: \noindent with $\tau_{vu}=\overline{uv}$. The contribution
138: $\overline{U(y)\frac{\partial
139: u^{2}}{\partial x}}=0$ due to the symmetry of the problem.
140:
141:
142:
143:
144:
145:
146:
147:
148:
149: Since in a homogeneous flow all spacial derivatives of the mean
150: properties are equal to zero, the modelling is reduced to investigation of the
151: time evolution of
152: turbulent kinetic energy ${\cal K}=\overline{u_{i}^{2}/2}$ and dissipation rate
153: ${\cal
154: E}=\nu\overline{(\partial_{i}u_{j})^{2}}$.
155:
156: For small perturbations from isotropic and homogeneous state ($S\rightarrow 0$
157: and $\overline {pu}=\overline {pv}=0$),
158: the typical turbulence models, based on
159: the equilibrium ideas are [15], [16]:
160:
161: \begin{equation}
162: \partial_{t}{\cal K}=-\tau_{ij}S_{ij}-{\cal E}
163: \end{equation}
164:
165: \noindent and
166:
167: \begin{equation}
168: \partial_{t}{\cal E}=-C_{\epsilon 1}\tau_{ij}S_{ij}\frac{{\cal E}}{{\cal K}}-
169: C_{\epsilon 2}\frac{{\cal
170: E}^{2}}{{\cal K}}
171: \end{equation}
172:
173: \noindent where the Reynolds stress
174: $\tau_{ij}=\overline{v_{i}v_{j}}$. The coefficients
175: $C_{\epsilon i}=O(1)$. A simple expression, valid at the long times $t>{\cal
176: K}/{\cal E}$,
177:
178: \begin{equation}
179: \tau_{ij}\approx -\nu_{T}S_{ij}
180: \end{equation}
181:
182: \noindent
183: with turbulent viscosity $\nu_{T}\propto {\cal K}^{2}/{\cal E}$, closes the
184: set of equations (7)-(9) and defines the so called ${\cal K}-{\cal E}$ model,
185: widely used in engineering for modelling the not-too strongly sheared flows.
186: The unknown
187: magnitudes of the proportionality coefficients are typically determined in a following
188: way. Consider a flow with $S=0$. The unknown coefficient $C_{\epsilon 2}$ can
189: be found from comparing the analytic solution of the simple equations (7)-(9)
190: with experimental and numerical data. The same flow can also be used to
191: test the results of analytic theories [15]-[16]. If $S\neq 0$, solution of
192: (7)-(9) is not easy and the coefficient $C_{\epsilon 1}$ can be found from
193: comparison with the data. The solution of equation (7)-(9) with the fixed values
194: of the coefficients
195: showed a close to
196: exponential long-time growth of turbulent kinetic energy in a good agreement
197: with the outcome of direct numerical simulations [15]-[16].
198: If the shear is imposed on a decaying isotropic
199: turbulence at $t=0$,
200: the observed [16]
201: initial, short- time decay of kinetic energy is readily explained by the fact
202: that turbulent viscosity
203:
204: $$\nu_{T}\approx \int_{0}^{t}\overline{v(0)v(\tau)}d\tau$$
205:
206: \noindent which is small at short times.
207:
208: Recent numerical experiments revealed a much more complicated picture.
209: Driven
210: by a very strong shear (the criterion is derived below), in the
211: long-time limit,
212: the
213: system developed a limit cycle -like
214: strong fluctuations of the total kinetic energy about the
215: mean value $<\overline{{\cal K}}>$ [7]-[13].
216: The amplitude of these fluctuations was up to two-three times
217: that of $<\overline{{\cal K}}>$. Similar effect was observed by Borue et. al. [14]
218: in a
219: three-dimensional Kolmogorov flow driven by a steady forcing
220: ${\bf f}=(0,0, cos(x))$. Elucidation of the origin of these oscillations is
221: the goal of this paper.
222:
223: The physical process observed in both homogeneous shear and Kolmogorov flows
224: can be described in two steps: first, the shear generates both kinetic energy
225: and vortical structures leading to the access of the energy
226: production. Then, the structures become unstable and rapidly disappear with
227: the energy dissipation taking over. The process repeats itself. The
228: evolution of kinetic energy and enstrophy fluctuations in
229: 3D Kolmogorov flow, conducted by Borue et al [14], revealed extremely strong
230: correlation: the sharp spikes in the enstrophy and energy time-signals
231: were
232: almost simultanious with a slight time-lag, thus suporting the importance of coherent
233: vortical structures in the process.
234:
235: At the long times the numerical homogeneous shear flow problem (1)-(3)
236: has two very important
237: features. We can see from the equation of motion that the flow, defined on a
238: cube,
239: cannot be
240: periodic in space. Second, the integral scale ${\cal L}$ in this situation is not
241: a dynamic
242: variable which is a function of ${\cal K}$ and ${\cal E}$, but prescribed by
243: the box size, so that ${\cal L}\approx a$. This puts
244: strong constrains on
245: the modelling of various contributions to the equations (1)-(3).
246:
247: Now, we would like to establish the main characteristic length-scales. The non-universal
248: velocity fluctuations belong to the range of scales $a\approx {\cal
249: L}<r<r_{c}$
250: with the cross-over scale $r_{c}\approx
251: \sqrt{{\cal E}}/S^{\frac{3}{2}}$ are dominated by powerful
252: anisortopic
253: coherent structures (vortices). The univesal range, populated by the more or
254: less isotropic excitations, spreads over the interval $r_{c}<r<r_{d}\approx
255: (\nu^{3}/{\cal E})^{\frac{1}{4}}$. The Kolmogorov spectrum can be expected in
256: the range with the
257: total energy of quasy-isotropic fluctuations
258:
259: \begin{equation}
260: q\approx \int^{r_{c}}_{r_{d}}E(k)dk\propto
261: r_{c}^{\frac{2}{3}}- r_{d}^{\frac{2}{3}}>0
262: \end{equation}
263:
264: We can see that the inertial range shrinks to zero when the strain rate
265: becomes large. This fact, noticed in Ref. [13], defines the strong shear
266: regime.
267: In strongly anisortopic flow, the
268: simple expression (9), is invalid.
269:
270: First , let us consider the equation for $\overline{\omega^{2}}$:
271:
272: \begin{equation}
273: \frac{1}{2}\partial_{t}\overline{\omega^{2}}+\frac{1}{2}
274: \partial_{i}\overline{u_{i}\omega^{2}}=
275: \overline{\omega\cdot\omega\cdot\nabla{\bf
276: u}}-
277: S\overline{(-\omega_{z}\partial_{z}w+\omega_{1}\partial_{x}w +\omega_{2}\partial_{z}v )} -
278: \nu\overline{(\partial_{i}\omega_{j})^{2}}
279: \end{equation}
280:
281: \noindent Due to powerful, shear-generated coherent vortical structures,
282: $\overline{\omega_{y}\omega_{x}}=O(\overline{\omega^{2}})$ and the
283: contribution involving longitudinal derivative $\partial_{z} w$ can be neglected.
284: The simple dimensional considerations lead
285: to:
286:
287: $$ \overline{\omega\cdot\omega\cdot\nabla{\bf u}}=O(u_{rms}\omega^{2}/{\cal L});~~
288: S\overline{(\omega_{x}\partial_{x}w+\omega_{y}\partial_{z}v)}=O(S\overline{\omega^{2}})$$
289:
290: \noindent and denoting $A(t)=\overline{\omega^{2}}/S^{2}$ we have:
291:
292: \begin{equation}
293: \partial_{t}A(t)=\gamma(\frac{a}{{\cal L}S}\sqrt{{\cal K}}-\alpha)SA(t)
294: \end{equation}
295:
296: \noindent with all coefficients $\gamma$, $a$ and $\alpha=O(1)$.
297:
298:
299:
300:
301:
302:
303:
304:
305:
306:
307: To
308: estimate $\tau_{uv}$, we observe that in the limit of interest
309: (see below) the only
310: relevant time scale is $\tau_{c}\approx 1/S$. From the equation (3)
311: we have an estimate for the stress:
312:
313: \begin{equation}
314: \tau_{uv}=\overline{u(t)v(t)}\approx \overline{u(t)\int_{t-\tau_{c}}^{t} d\lambda
315: (-{\bf u}(\lambda)\cdot \nabla v(\lambda)-\nabla_{y}p(\lambda))}
316: \end{equation}
317:
318: \noindent where the ``initial condition $\tau_{uv}(t-\tau_{c})$ was
319: neglected for simplicity (see below). The dimensional estimate gives:
320:
321: \begin{equation}
322: \tau_{uv}=B(t)\frac{{\cal K}^{\frac{3}{2}}}{S{\cal L}}
323: \end{equation}
324:
325: \noindent where $B(t)$ is an ``anisotropy factor'' or ``order parameter''
326: characterizing the varying
327: in time strength
328: of the coherent vortical structures. The appearence of this factor is natural
329: (see below)
330: since
331: in an isotropic, non-sheared flow lacking coherent
332: vortices $B(t)=const=0$, while in the strongly sheared
333: flow these structures contribute to the
334: energy production.
335: The dissipation rate in shear flows is estimated as
336: ${\cal E}=b{\cal K}^{\frac{3}{2}}/{\cal L}$ with the coefficient $b\approx 1$
337: leading to a model equation:
338:
339: \begin{equation}
340: {\cal K}_{t}(t)\approx (B(t)-b)\frac{{\cal
341: K}^{\frac{3}{2}}}{{\cal L}}
342: \end{equation}
343:
344: \noindent or, introducing $y=\sqrt{{\cal K}}$:
345:
346:
347: \begin{equation}
348: 2\partial_{t}y(t)=
349: \frac{B(t)-b}{{\cal L}}y^{2}
350: \end{equation}
351:
352: By the virtue of (2),
353: the mean value of vorticity in homogeneous shear flow is
354: $<\overline{\Omega}>=S$. Thus, the natural measure of the
355: strength of the strongly anisotropic fluctuating coherent vortical structures is
356: the ratio $\overline{\omega}^{2}/S^{2}$. Based on these considerations ,
357: we set $A(t)=B(t)$. This result can be derived in the third-order of the
358: iteration procedure of the expression (13). Indeed, inserting an unknown
359: initial condition into (13), one can use $\tau_{uv}(t-\tau_{c})=\tau_{0}$
360: as a zero order solution. Then, after simple resummation (neglecting the first
361: -order contributions) we obtain:
362:
363: $$\tau_{uv}(t)=\tau_{0}+\int_{t-\tau_{c}}^{t}\overline{u(t){\bf u(\lambda)\cdot \nabla}
364: v(\lambda)}+
365: \int\int\int \overline{ u(t){\bf u(\lambda)\cdot\nabla u(\lambda')\cdot\nabla
366: u(\lambda '')\cdot\nabla }v(\lambda'')}d\lambda d\lambda'
367: d\lambda'' + \cdot\cdot\cdot $$
368:
369: This expression immediately gives (14) with $B(\omega)\approx
370: \frac{\omega^{2}}{S^{2}}$,
371: provided the derivatives $\partial_{i}u_{j}\approx \omega$.
372: This approximate derivation is given here to demonstrate the mechanism of
373: vorticity
374: appearence in the expression for the reynolds stress $\tau_{uv}$. It will
375: become clear below that the power of vorticity in the expression
376: $B(\omega)\propto \omega^{n}$ is unimportant.
377:
378:
379:
380: The fact that the anisotropic ordered structures can influence the magnitude and even
381: sign of the energy fluxes is known for a long time.
382: The rigorous linear
383: stability analysis developed
384: in [17]-[18], showed that even in three-dimensional flows the
385: strongly anisotropic structures (basic flows),
386: are capable of reversing the sign of the
387: energy flux, due to the ``negative
388: viscosity'' effects and lead to substantial growth
389: of a small large-scale perturbation. In the opposite limit of the
390: isotropic basic flows, the theory showed
391: generation of positive effective
392: viscosity and acceleration of the energy dissipation.
393: These features are
394: incorporated in the model (12),(15), (16):
395: indeed, we see that if $A(t)>b$, the energy
396: grows, while, when $A<b$, it decays. The model equation (12) includes the
397: well-known process of the vortex break -down: when $v^{2}>>\omega^{2}{\cal L}^{2}$,
398: the instability leads to the vortex disappearence.
399:
400: \noindent Defining dimensionless variables $Z=\frac{y}{S{\cal L}}$,
401: $Z_{0}=\frac{\alpha}{a}$ and $T=St$ gives:
402:
403:
404:
405:
406: \begin{equation}
407: 2\partial_{T}Z(T)=a(A(T)-b)Z^{2}(T)
408: \end{equation}
409:
410: \noindent and
411:
412: \begin{equation}
413: \partial_{T} A(T)=-\gamma(Z(T)-Z_{0})A(T)
414: \end{equation}
415:
416: \noindent where
417: $Z_{0}>0=const$, related to the mean amplitude of $Z(t)$.
418:
419: The equations (17),(18) have a steady-state solution
420: $Z(T)=Z_{0}$ and $A(T)=b$. A simple linear stability analysis
421: shows periodic solution when the amplitude of the perturbation is very
422: small.
423: The numerical solutions of quations (17),(18),
424: presented on Figs. 1-6, revealed
425: strong non-linear oscillations. All calculations were perforemed with
426: $Mathematica^{TM}$.
427: In a wide range of parameter variation, the
428: system generates non-linear oscillations with the shape
429: depending upon initial values
430: $Z(0)$ and $A(0)$. For a given set of parameters the frequency of
431: oscillations is proportional to the strain rate $S$.
432:
433: For the initial values of $Z(0)$ and $A(0)\approx 1$,
434: the solution shows reasonably smooth oscillations with $Z$ and $A$ being
435: somewhat out of phase (see Figs. 1,2).
436: The result supports a general physical picture of
437: the anisotropy $A(t)$ (order parameter) and energy growing (decaying)
438: together with
439: some time-lag. The energy fluctuations are by a factor 2-3 larger than
440: $S^{2}{\cal L}^{2}$. When the initial energy was doubled to $Z(0)=3.$, the
441: oscilaltions became much less symmetric with the steeper energy grows
442: (Figs. 3,4). The crucial role of the ``order parameter'' $A(0)$ is
443: demonstrated on Figs. 5,6 corresponding to $Z(0)=2$ and $A(0)=0.1$. We can see
444: the
445: the formation steep shock-like structures, somewhat resembling
446: turbulence-production bursts.
447:
448: In the range of large
449: $Z(0)$ and very small $A(0)<<1$, the solution blows up,
450: indicating the unphysicality of these initial
451: conditions corresponding to the large energy fluctuations ($u_{rms}(0)>>S{\cal
452: L}$) and small anisortopy (order) parameter ($\overline{\omega^{2}}<<S^{2}$).
453:
454: To discuss the above results, let us
455: look at this work from a somewhat different angle.
456: The Kolmogorov relation $S_{3,0}(r)=\overline{(u(x+r)-u(x))^{3}}\propto r$,
457: is a statement about constantcy of the energy
458: flux for inertial range wave numbers $k\approx 1/r>>1/{\cal L}$ of
459: isotropic and homogeneous turbulence. As $r\rightarrow {\cal
460: L}$, the structure function $S_{3,0}(r)\rightarrow 0$.
461: In strongly anisotropic flows with the
462: integral scalle ${\cal L}\approx a$, this is not so: depending on the
463: spacial distribution of velocity
464: (vorticity), the moment
465: $S_{3}({\cal L})\neq 0$. If vorticity (enstrophy)
466: is an ``order'' parameter, characterizing deviations from isotropy, then
467: $S_{3}({\cal L})\approx B(\omega){\cal K}^{\frac{3}{2}}/{\cal L}$ where
468: $B(\omega)\rightarrow 0$ when the strength of the structures
469: diminishes. This qualitative statement
470: is supported by the well-known fact that the
471: velocity field, generated by the vortex $v(r)\propto {\Gamma}\phi(r)$, where
472: the circulation
473: $\Gamma=O(\omega {\cal L}^{2})$.
474: Combined with the equation for the enstrophy, the two relations (17),(18)
475: form a dynamical system leading to strong fluctuations of both energy
476: and enstrophy. The shape of the function $B(\omega)$ does not seem to
477: influence the qualitative aspects of the process: the model
478: (17),(18) is invariant under transformation $A(t)\rightarrow A^{n}$ with a
479: simple rescaling of time.
480:
481: All this is valid when ${\cal L}\approx a$. If this is not so, the magnitude
482: of the fluctuations must substantially decrease. Indeed, if $a>>{\cal L}$,
483: then we are dealing with $N=(\frac{a}{{\cal L}})^{d}$ independent
484: systems. Here $d$ is the force dimensionalty.
485: Since the phases are crucially important , we expect the amplitude of
486: the fluctuations to decrease as $1/\sqrt{N}$. This can easily be tested on an
487: example of 3D Kolmogorov flow in a box with the side $a$
488: driven by the force ${\bf f}=(0,0,cos
489: (\frac{x}{{\cal L}}))$ by varying the forcing scale.
490:
491: To conclude: based on the equations of motion and some physical
492: considerations, we
493: propose a dynamic model, coupling vorticity (enstrophy) and energy
494: fluctuations in a
495: homogeneous shear flow. This model generates strongly correlated
496: self-sustained oscillations of both enstrophy and energy similar to those
497: observed in eperiments and direct numerical simulations.
498: The calculated time
499: -lag is similar to that observed in a numerical study of 3D Kolomogorov flow
500: by Borue et al [14].
501:
502: It is not yet clear if, properly parametrized, this simple model can mimic
503: turbulent bursts which are at the core of the energy production
504: in turbulent wall
505: flows. In case of a positive answer, the model of this kind can serve as a
506: boundary condition (``wall function'')
507: for turbulence simulations, neglecting the detailed
508: consideration of dynamics of the viscous sublayer. The achieved computational
509: economy makes this aspect of the work worth persuing.
510:
511:
512:
513:
514:
515:
516:
517:
518:
519:
520:
521:
522:
523:
524:
525:
526:
527:
528:
529:
530:
531:
532: \noindent {\bf references}\\
533: \noindent 1. S. Tavoularis and S. Corrsin, J. Fluid Mech., {\bf 104}, 331,
534: (1981)\\
535: \noindent 2. S. Tavoularis and S. Corrsin, J. Fluid Mech., {\bf 104}, 349,
536: (1981)\\
537: \noindent 3. F.H. Champain, V.G. Harris, and S. Corrsin J. Fluid Mech., {\bf 41}, 81,(1970)\\
538: \noindent 4. Rose, J. Fluid Mech., {\bf 25}, 97,(1966)\\
539: \noindent 5. S. Garg and Z. Wargaft, Phys. Fluids, {\bf 10}, 662, (1998)\\
540: \noindent 6. X. Shen and Z. Wargaft, Phys. Fluids, {\bf 12}, 2976, (2000)\\
541: \noindent 7. W. Rogers and P. Moin, J. Fluid. Mech., {\bf 176}, 33, (1987)\\
542: \noindent 8. C. Lee, J. Kim and P. Moin, J. Fluid. Mech., {\bf 216}, (1990)\\
543: \noindent 9. S. Kida and M. Tanaka, J. Fluid. Mech., {\bf 274}, (1994)\\
544: \noindent 10. K. Nomura, in IUTAM Symposium on Geometry and Statistics of
545: turbulence, Kluwer Acad. Publishers, Boston, (2002)\\
546: \noindent 11. A. Pumir, Phys. Fluids, {\bf 8}, 3112, (1996)\\
547: \noindent 12. A. Pumir and B. Shraiman (1995), Phys.Rev.Lett., {\bf 75}, 3114,(1995)\\
548: \noindent 13. P. Gualtieri, C.M.Casciola, R. Benzi, G. Amati and R.Piva,
549: arXiv:nlin.CD/0011040v2, 27Nov(2000)\\
550: \noindent 14. V. Borue and S. Orszag, J. Fluid. Mech.,.... (1996)\\
551: \noindent 15. B. Launder and D.B. Spalding, Com. Mech. in
552: Appl. Mech. Eng,. {\bf 3}, 269 (1974)\\
553: \noindent 16. V. Yakhot, S. Orszag, S. Thangam, T. Gatsky and C. Speciale,
554: Phys. Fluids A, {\bf 4}, 1510 (1992)\\
555: \noindent 17. V.Yakhot and G. Sivashinsky, Phys. Rev. A35, 815 (1987)\\
556: \noindent 18. V. Yakhot and R. Pelz, Phys. Fluids. {\bf 30}, 1272 (1987)\\
557:
558:
559:
560: ..................
561: \\
562:
563:
564: \newpage
565:
566:
567: \begin{figure}[h]
568: \centerline{\psfig{file=shearA.eps,width=5.5in,height=5.0in}}
569: \caption{Time -evolution of $Z^{2}(T)\propto {\cal K}(T)$ (higher amplitude
570: curve) and $A(T)$ vs $T$.
571: $Z(0)=1.41$,~~$ A(0)=1.3$}
572: \end{figure}
573:
574:
575: \begin{figure}[h]
576: \centerline{\psfig{file=shearA1.eps,width=5.5in,height=5.0in}}
577: \caption{Parametric plot $Z^{2}(T)$ (horizontal) vs. $A(T)$}
578: \end{figure}
579:
580:
581:
582:
583: \begin{figure}[h]
584: \centerline{\psfig{file=shearB.eps,width=5.5in,height=5.0in}}
585: \caption{Time -evolution of $Z^{2}(T)\propto {\cal K}(T)$ (higher amplitude
586: curve) and $A(T)$ vs $T$.
587: $Z_{0}=3$; $A(0)=1.3$}
588: \end{figure}
589:
590:
591:
592:
593:
594:
595: \begin{figure}[2^{b}]
596: \centerline{\psfig{file=shearB1.eps,width=5.5in,height=5.0in}}
597: \caption{Parametric plot $Z^{2}(T)$ (horizontal) vs $A(T)$}
598: \end{figure}
599:
600:
601: \begin{figure}[h]
602: \centerline{\psfig{file=shearC.eps,width=5.5in,height=5.0in}}
603: \caption{Time -evolution of $Z^{2}(T)\propto {\cal K}(T)$ (higher amplitude
604: curve) and $A(T)$ vs $T$.
605: $Z_{0}=2$; $A(0)=0.1$}
606: \end{figure}
607:
608:
609: \begin{figure}[h]
610: \centerline{\psfig{file=shearC1.eps,width=5.5in,height=5.0in}}
611: \caption{Parametric plot $Z^{2}(T)$ (horizontal) vs $A(T)$}
612: \end{figure}
613:
614:
615:
616: \end{document}
617:
618:
619:
620: