cond-mat0009241/spt.tex
1: %****** Start of file template.aps ****** %
2: %
3: %   This file is part of the APS files in the REVTeX 3.1 distribution.
4: %   Version 3.1 of REVTeX, July 1, 1996.
5: %
6: %   Copyright (c) 1996 The American Physical Society.
7: %
8: %   See the REVTeX 3.1 README file for restrictions and more information.
9: %
10: %
11: % This is a template for producing files for use with REVTEX 3.1.
12: % Copy this file to another name and then work on that file.
13: % That way, you always have this original template file to use.
14: %
15: %\documentstyle[preprint,prl,aps,epsfig]{revtex}
16: \documentstyle[prl,aps,multicol,epsfig]{revtex}
17: 
18: \begin{document}
19: 
20: \draft
21: %
22: \title{\lq\lq Barber pole turbulence\rq\rq in large aspect ratio
23: Taylor-Couette flow} 
24: \author{A. Prigent and O. Dauchot}
25: %
26: \address{Service de Physique de l'Etat Condens\'e, CEA Saclay, 91191 Gif
27: sur Yvette, France}
28: \date{\today}
29: \maketitle
30: 
31: 
32: \begin{abstract}
33: Investigations  of counter-rotating Taylor-Couette  flow (TCF)  in the
34: narrow  gap  limit  are  conducted   in  a  very  large  aspect  ratio
35: apparatus.  The  phase  diagram  is  presented and  compared  to  that
36: obtained  by   Andereck  {\it  et   al.}~\cite{Andereck}.  The  spiral
37: turbulence  regime is studied  by varying  both internal  and external
38: Reynolds numbers. Spiral turbulence is  shown to emerge from the fully
39: turbulent regime via a continuous transition appearing first as a {\it
40: modulated} turbulent  state, which  eventually relaxes locally  to the
41: laminar  flow. The  connection with  the intermittent  regimes  of the
42: plane Couette flow (pCf) is discussed.
43: \end{abstract}
44: % insert suggested PACS numbers in braces on next line
45: 
46: \pacs{ {47.20.Ft} %{Instability of shear flows}; 
47:        {47.20.Ky} %{Nonlinearity (including bifurcation theory)};
48:        {47.54.+r} %{Pattern selection; pattern formation}
49:        \and {47.27.Cn} %{Transition to turbulence}
50: }
51: 
52: \begin{multicols}{2} 
53: \narrowtext
54: % body of paper here
55: The  \lq\lq  barber  pole  structure  of  turbulence\rq\rq~\cite{Feyn}
56: between two counter-rotating cylinders, also called spiral turbulence,
57: is commonly  described as alternating  helical stripes of  laminar and
58: turbulent flow.   There are few quantitative studies  of this puzzling
59: regime, where  long range order coexists with  small scale turbulence.
60: In  early  studies  Coles and  Van  Atta~\cite{Coles,VanAtta,Coles-VA}
61: measured  the  spiral shape  in  the  mid-plane  perpendicular to  the
62: cylinder axis  and the pattern  rotation rate. Later Andereck  {\it et
63: al.}     reported     qualitative     observations    when     varying
64: $R_i$~\cite{Andereck}.    Hegseth    {\it   et   al.}~\cite{Heg1,Heg2}
65: described   spiral   turbulence   within   the  framework   of   phase
66: dynamics.  All these studies  were limited  by their  relatively small
67: size.  Only one  helical turbulent stripe, winding no  more than twice
68: along the cylinder axis, could  be observed. Altogether, the origin of
69: this flow pattern remains unknown.
70: 
71: Performing measurements in large  aspect ratio Taylor-Couette flow, we
72: show that the spiral turbulence bifurcates continuously 
73: from  the turbulent  flow, appearing  as  a modulated  turbulent
74: state.  After a  rapid  description  of the  experimental  set up,  we
75: present  the phase  diagram  and compare  it  to the  one obtained  by
76: Andereck  {\it  et   al.}\cite{Andereck}  with  a  different
77: cylinder radius ratio.  Then  we describe the successive steps leading
78: to the  fully turbulent flow before discussing  the origin  of spiral
79: turbulence. Finally,  we examine its breakdown  into a spatio-temporal
80: disordered  regime   similar  to  the   laminar-turbulent  coexistence
81: dynamics observed in plane Couette flow~\cite{DauDav,BottChat}. 
82: 
83: Let  us  recall  here  the  characteristics of  the  experimental  set
84: up~\cite{TCvisu}. Our  TC apparatus,  two coaxial cylinders  of radius
85: $r_i=49.09 \pm  0.005$ mm and $r_o=49.95  \pm 0.005$ mm  with a useful
86: length  $L=380 \pm  0.1$ mm,  rotating independently,  has a  gap size
87: $d=0.863  \pm  0.01$ mm,  large  aspect  ratio $\Gamma_z=L/d=442$  and
88: $\Gamma_{\theta}=\pi(r_i+r_o)/d=362$, and  a radius ratio $\eta=0.983$
89: very close  to 1.  In this paper,  length and  time units are  $d$ and
90: $d^2/\nu$.  $\eta$ being fixed, the  flow is governed by the inner and
91: outer   Reynolds   numbers  $R_{i,o}=r_{i,o}\Omega_{i,o}d/\nu$,   with
92: ${\Omega_{i,o}}$  the angular  velocities, and  ${\nu}$  the kinematic
93: viscosity.  The accuracy of $R_{i,o}$ is better than 3$\%$. 
94: 
95: We  visualize  the  flow  by  a  \lq\lq  fluorescent  lighting  \rq\rq
96: technique~\cite{TCvisu} developed  for this study.  The  water flow is
97: seeded with  Kalliroscope AQ 1000 ($6 \times
98: 30 \times 0.07  \mu m$ platelets). The inner cylinder  is covered by a
99: fluorescent  film  and  the   entire  apparatus  is  UV-lighted.   The
100: fluorescent  film  re-emits a  uniform  visible lighting,  transmitted
101: through the fluid layer: the  more turbulent the flow, the brighter it
102: appears. As  the gap is  very thin, the Kalliroscope  concentration is
103: increased up to $25\%$ by volume to enhance the contrast. A 
104: rheological study has shown that  the fluid remains Newtonian, so that
105: the  only impact  is  a viscosity  increase  up to  $\nu=1.13~10^{-6}$
106: m$^2$/s at $20^\circ$C.  The  flow is thermalized by water circulation
107: inside the  inner cylinder. At thermal equilibrium 
108: the temperature  is uniform in space  up to $0.1^\circ$C  and does not
109: vary more than $0.1^\circ$C/hour. Images and spatio-temporal diagrams
110: (temporal recording of one line along the cylinder
111: axis) are recorded by a CCD  camera. Two plane mirrors reflect the two
112: thirds of the flow hidden to  the camera so that the whole cylindrical
113: flow can be reconstructed.
114: 
115: Coles~\cite{Coles}  has  stressed   the  variety  of  flow  structures
116: observed  according to  the  path followed  in  parameter space.   The
117: experimental  procedure followed here  is the  most frequent  one. The
118: outer cylinder  is slowly accelerated to the  desired angular velocity
119: while the  inner one is  at rest. Then,  the inner cylinder  is slowly
120: accelerated  ($(1/\Omega_i).(\partial{\Omega_i}/{\partial  t})=0.03\%$
121: per  second) in  order  to ensure  quasi-static  transitions. (It  was
122: checked  that the  Taylor  vortices threshold  is  determined with  an
123: accuracy better  than $1\%$ as soon  as the accelaration  rate is less
124: than $0.05\%$).  Recording the various flow regimes in parameter space
125: $(R_i    ,R_o    )$,   gives    the    experimental   phase    diagram
126: (fig.~\ref{fig:phasediag}(a)) to be compared to that given by Andereck
127: {\it  et al.}~\cite{Andereck} for  a different  geometry ($\eta=0.878$
128: and much  smaller aspect  ratio).  The same  regimes are  observed but
129: their locations in the parameter space change.  The linear thresholds
130: increase (at $R_o=0$, the Taylor vortices appear for $R^*_i=320$ instead of
131: $R^*_i=120$). This was expected in regards of the
132: plane Couette flow limit and agrees very well with Snyder's law
133: $R^*_i~=~27[(1-(\eta))\eta]^(-5/3)$~\cite{Snyder}. 
134: 
135: \begin{figure}
136: \vspace{-1.cm}
137: \begin{center}
138: 	\hbox{
139: 	\hspace{-0.35 cm}
140: 	\psfig{figure=phasediag2.eps, width=9cm, angle=0}
141: 	}
142:         \vspace{-0.2cm}
143: 	(a)
144: 	\hbox{
145: 	\hspace{-0.30 cm}
146: 	\psfig{figure=phasediagzoom2.eps, width=9cm, angle=0}
147: 	}
148:         \vspace{-0.15cm}
149: 	(b)
150: \end{center}
151: \vspace{-0.2cm}
152: \caption{Experimental  phase diagram  ($\eta=0.983$).  (b) displays  a
153: zoom  of   the  dotted  rectangle   of  (a).  The  labels   stand  for
154: AZI~:~azimuthal flow; TVF:~Taylor  vortex flow; WVF:~wavy vortex flow;
155: SPI\&IPS:~spiral  and   interpenetrating  spiral  vortices;  WIS:~wavy
156: interpenetrating  spiral   vortices;  INT:~intermittency;  SPT:~spiral
157: turbulence; TUR:~turbulence. The arrow indicates the $R_o$ value below
158: which the AZI  flow jumps directly to INT (see  text for details). The
159: solid straight lines  in (b), show the paths  along which measurements
160: are    conducted    (A:~$R_o=-922$;   B:~$R_o=-713$;    C:~$R_o=-641$;
161: D:~$R_i=704$).}
162: \label{fig:phasediag}
163: \end{figure}
164: 
165: \noindent
166: Accordingly, the other vortex flows  (WVF, SPI, IPS...) also appear at
167: higher  $R_i$  values.    Second,  the  laminar-turbulent  coexistence
168: regions (INT  and SPT),  extend to higher  $R_o$ values.  Finally, for
169: $R_o<-725$,  the azimuthal  flow jumps  directly to  INT  whereas, for
170: $R_o>-725$, it bifurcates first to IPS or SPI.
171: 
172: Figure~\ref{fig:deroule},  shows snapshots  of the  whole  flow, along
173: paths  A  and C.  For  $R_o=-922$,  the  boundaries trigger  turbulent
174: stripes  are  at $R_i$  lower  than  the  expected linear  instability
175: threshold  (fig.~\ref{fig:deroule}(a)).  In  fig.~\ref{fig:deroule}(b)
176: periodically   organized  turbulent   stripes   have  formed:   spiral
177: turbulence is  established as a regular rotating  pattern with uniform
178: pitch angle along the cylinder axis.
179: 
180: \begin{figure}
181: \vspace{-0.5cm}
182:   \begin{center}
183:   \hbox{
184:     (d)
185:     \psfig{figure=reg900TCF6.eps, height=5.2cm, width=3.47cm, angle=0} 
186:     (h)
187:     \psfig{figure=reg650TCF4.eps, height=5.2cm, width=3.47cm, angle=0}
188:    }    
189:   \hbox{
190:     (c) 
191:     \psfig{figure=reg900TCF5bis.eps, height=5.2cm, width=3.5cm, angle=0} 
192:     (g)
193:     \psfig{figure=reg650TCF3.eps, height=5.2cm, width=3.5cm, angle=0}
194:    }
195:   \hbox{
196:     (b) 
197:     \psfig{figure=reg900TCF3.eps, height=5.2cm, width=3.5cm, angle=0} 
198:     (f)
199:     \psfig{figure=reg650TCF2bis.eps, height=5.2cm, width=3.5cm, angle=0}
200:    }
201:   \hbox{
202:     (a) 
203:     \psfig{figure=reg900TCF0.eps, height=5.2cm, width=3.5cm, angle=0} 
204:     (e)
205:     \psfig{figure=reg650TCF2.eps, height=5.2cm, width=3.5cm, angle=0}
206:    }
207: \end{center}
208: \vspace{-0.5cm}
209:     \caption{Turbulent spots and stripes along paths A and C. 
210: Left column is path A, $R_o=-922$:~$R_i=$~(a)~580, (b)~600, (c)~697, (d)~784.
211: Right column is path C, $R_o=-641$:~$R_i=$~(e)~649, (f)~668, (g)~695, (h)~717.
212: Each picture displays a 360 $^\circ $ view of the whole flow ($38$ cm
213: high and $31.4$ cm wide.} 
214:     \label{fig:deroule}
215: \end{figure}
216: 
217: \noindent
218: Further increasing $R_i$, the pattern maintains  itself and the
219: number of stripes  increases (fig.~\ref{fig:deroule}(c)). Finally, the
220: pattern disappears  smoothly as homogeneous  turbulence takes over
221: (fig.~\ref{fig:deroule}(d)). In this regime, one notices several domains of  opposite pitch  spirals.
222: For  $R_o=-641$, the azimuthal flow first bifurcates to a vortex flow,
223: within which, further increasing $R_i$,  turbulent spots follow an
224: intermittent bursting dynamics (fig.~\ref{fig:deroule}(e)).   In
225: fig.~\ref{fig:deroule}(f), the spots grow, merge and
226: build  up  the turbulent  spiral  pattern.   For 
227: illustration,  we  show,  in fig.~\ref{fig:deroule}(g),  two  opposite
228: pitch  spirals competing and  leading to a so-called ``V''-shaped
229: pattern.   In figure~\ref{fig:deroule}(h), spiral  turbulence turns
230: into  uniform   turbulence  as  for   $R_o=-922$.   Fig.~\ref{fig:dst}
231: displays the spatio-temporal dynamics  of a front between two opposite
232: spirals.  This front, moving along the cylinder axis, is eventually
233: absorbed at one of the extremities. 
234: 
235: \begin{figure}
236: \vspace{-0.8cm}
237:   \begin{center}
238:     \psfig{figure=dstsptu3b.eps, height=5cm, width=8cm, angle=0}
239:   \end{center}
240: \vspace{-0.1cm}
241:     \caption{Spatio-temporal diagram of spiral turbulence
242: displaying a front between two domains of opposite pitch
243: angle. $R_o=-838$, $R_i=707$. The z-axis display the whole flow height
244: ($38$ cm) and the total recording length is $20.48$ s.}
245:     \label{fig:dst}
246: \vspace{-0.2cm}
247: \end{figure}
248: 
249: Let  us now  consider  the previous  scenario  in reverse.  Decreasing
250: $R_i$,  spiral  turbulence appears  from  homogeneous turbulence.   At
251: $R_i=R_{ic}$,  a low amplitude  modulation, with  wave vector  $k$ and
252: frequency $\omega$, uniformly emerges  from the turbulent flow with no
253: prefered direction.   A slow front dynamics occurs  between domains of
254: opposite  pitch  until the  $z  \rightarrow  -z$  symmetry breaks  and
255: selects  a  single  uniform  spiral.  Further  decreasing  $R_i$,  the
256: modulation   amplitude  increases  until   a  more   complex  dynamics
257: occurs.  The  light  intensity,   which  in  the  first  approximation
258: corresponds to  the turbulence  level, is recorded  in spatio-temporal
259: diagrams such  as fig.~\ref{fig:dst}. In the following,  we obtain the
260: amplitude $A$,  the axial component $k_z$  of the wave  vector and the
261: frequency   $\omega$   of    the   light   intensity   modulation   by
262: demodulation~\cite{Croq}.   Owing to  the  azimuthal periodicity,  the
263: azimuthal         component         of         the         wavevector,
264: $k_\theta=2.n_\theta/(r_i+r_o)$, is  quantized, $n_{\theta}$ being the
265: number of pattern periods along one perimeter.
266: 
267: Four  paths  in the  $(R_o,  R_i)$  plan  are investigated.  We  first
268: consider the three  vertical paths A, B and  C ($R_o=-922$, $R_o=-713$
269: and $R_o=-641$).
270: 
271: \begin{figure}
272: \vspace{-0.3cm}
273:   \begin{center}
274:   \hbox{
275:     \psfig{figure=ampli.eps, height=4cm, width=4.25cm, angle=0}
276:     \psfig{figure=kz1.eps, height=4cm, width=4.25cm, angle=0}
277:        }    
278: \vspace{-0.1cm}
279:   \hbox{
280:     \hspace{1.77cm} (a) \hspace{3.77cm} (b)	
281:        }
282: \end{center}
283: \vspace{-0.5cm}
284:     \caption{The spiral turbulence amplitude (a) and vertical wave
285: number (b) vs. $\epsilon$. ($R_o=-641$ ($\ast$), $R_o=-713$
286: ($\diamond$), $R_o=-900$ ($\bullet$)).} 
287:     \label{fig:curves}
288: \end{figure}
289: 
290: \noindent
291: Fig.~\ref{fig:curves}(a) displays  the amplitude of the wave against
292: the threshold distance $\epsilon= \left|R_i-R_{ic}\right|/R_{ic}$.
293: $R^c_i$ is determined by extrapolation, assuming a
294: linear dependence of $A^2$  on  $R_i$,  as  suggested  by  the
295: general  framework  of amplitude equations. The resulting values,
296: $R_{ic}=820$ for $R_o=-922$, $R_{ic}=785$ for  $R_o=-713$ and
297: $R_{ic}=740$ for $R_o=-641$, agree with   the  observed   thresholds.
298: The axial component $k_z$ of the wavevector 
299: decreases linearly with  $\epsilon$ from  a critical  value  $k^c_z$
300: close  to 0.30  (see fig.~\ref{fig:curves}(b)). As long as  the spiral
301: is well defined, the azimuthal wave  number $n_\theta$ equals 6  along
302: paths A and  B and 5 along path C. $n_\theta$ being  constant  and
303: $k_z$  decreasing with $\epsilon$,  the spiral  pitch, $\phi=\arctan
304: (k_\theta  / k_z)$, increases with  $\epsilon$ from a critical  value
305: $\phi^c=18^\circ$ up to $40^\circ$.
306: 
307: \begin{figure}
308: \vspace{-0.5cm}
309:   \begin{center}
310:     \psfig{figure=freq1.eps, width=8cm, angle=0}
311:   \end{center}
312: \vspace{-0.5cm}
313:     \caption{The angular velocity $\omega_s$ vs.
314: $\Omega_m$, ($R_o=-641$ ($\ast$), $R_o=-713$ ($\diamond$),
315: $R_o=-900$ ($\bullet$) and $R_i=704$ ($+$)).} 
316:     \label{fig:omega}
317: \vspace{-0.0cm}
318: \end{figure}
319: 
320: We now consider the horizontal  path D with $R_i=704$. For $R_o>-630$,
321: the  flow  is fully  turbulent.  Decreasing  $R_o$, spiral  turbulence
322: emerges from  the fully turbulent  regime. The initial  azimuthal wave
323: number   $n_\theta=5$  increases   to  $n_\theta=6$   by  Eckhaus-like
324: re-arrangements.  The axial wave  vector component remains constant on
325: the  explored values  $-950<R_o<-630$.   Finally, fig.~\ref{fig:omega}
326: displays  $\omega_s=\omega/n_\theta$,  the  angular  velocity  of  the
327: spiral               turbulence              pattern,              vs.
328: $\Omega_m=\frac{1}{2}(\Omega_i+\Omega_o)$.  As seen  by Coles  and Van
329: Atta~\cite{Coles,VanAtta,Coles-VA}, the pattern  rotates with the mean
330: angular velocity of the two  cylinders ($\omega_s = \Omega_m$) and not
331: with   the  outside  cylinder   as  reported   by  Andereck   {\it  et
332: al.}~\cite{Andereck}     as    well     as    by     Goharadzeh    and
333: Mutabazi~\cite{GohMut}. In  the latter  study, the range  of variation
334: for  $\Omega_i/\Omega_o$  is  actually  too  small  to  distinguish  a
335: dependence on $\Omega_o$ from one on $\Omega_m$.
336: 
337: Altogether,  the spiral turbulence  emerges as  a wavy turbulent state
338: pattern, which bifurcates continuously  from the homogeneous turbulent flow. 
339: This wavy behavior agrees with the idea of a phase
340: dynamics analysis  suggested by Hegseth  {\it et al.}~\cite{Heg1,Heg2}
341: to describe the pitch variation, they observed. 
342: However the present study does not report any pitch variation. 
343: In their analysis, Hegseth {\it et al.}, on the basis of
344: the  mean  flow measurements  by  Coles and  Van~Atta~\cite{Coles-VA},
345: suggest  that the  spiral turbulence  induces a  backflow,  whose mean
346: axial  component is  differently  modified at  both extremities.   The
347: resulting different pitches locally imposed in the boundary conditions
348: drive the pitch  variation along the axis. In  the present study, both
349: the large  aspect ratio and  the gap thinness considerably  reduce the
350: backflow and the influence of the end boundaries.
351: 
352: For low  values of $R_i$,  we observed disconnected  turbulent domains
353: surrounded by  a laminar flow, which evokes  the transitional dynamics
354: observed   in  the   plane  Couette   flow~\cite{BotDavManDau},  where
355: turbulent domains  move, grow, decay,  split and merge leading  to the
356: so-called  spatio-temporal  intermittency (STI),  a  process in  which
357: active/turbulent regions  may invade absorbing/laminar  domains out of
358: which             disorder             cannot            spontaneously
359: emerged~\cite{ChatMan}. Considering the  vicinity of the plane Couette
360: flow     limit,     $\eta=1$,     and    following     Colovas     and
361: Andereck~\cite{ColAnd}, one would  like to know whether Taylor-Couette
362: flow actually exhibits STI and  how the flow evolves from a continuous
363: turbulent   state   modulated  in   space   and   time  to   two-state
364: coexistence.  In  a  first  attempt  to  answer  these  questions,  we
365: binarized the  flow into alternating laminar and  turbulent domains as
366: is common  in an STI  context.  This first-order  approximation should
367: not influence the  results presented below, at least  at a qualitative
368: level. 
369: 
370: \begin{figure}
371: \vspace{-0.5cm}
372:   \begin{center}
373:     \psfig{figure=lpg.eps, width=8cm, angle=0}
374:   \end{center}
375: \vspace{-0.5cm}
376:     \caption{$\lambda$ ($R_o=-713$ ($\diamond$) and $R_o=-922$
377: ($\bullet$)), the pattern wavelength and $\zeta$ ($R_o=-713$
378: ($\triangle$) and $R_o=-922$ ($\ast$)), the
379: turbulent stripes width, against $\epsilon$.} 
380:     \label{fig:lp}
381: \vspace{-0.3cm}
382: \end{figure}
383: 
384: 
385: \noindent
386: Fig.~\ref{fig:lp}   displays  the  turbulent   stripes  width,
387: $\zeta$, and the wavelength, $\lambda$ of the spiral pattern, parallel
388: to the wave vector direction. $\lambda$ increases faster than $\zeta$:
389: the minima  of the modulation becomes  more and more  distant from the
390: most active  turbulent regions. Simultaneously the  modulation amplitude
391: increases and these minima are less and less turbulent.  As a result they
392: become areas favorable to local relaxation to the laminar flow and
393: the  modulated  turbulent  state  turns  into a  periodic  pattern  of
394: alternating turbulent  and laminar  stripes separated by  fronts. This
395: totally  different  spatio-temporal dynamics  rapidly  turns into  the
396: disordered STI regime.
397: 
398: Our very large aspect ratio  TC apparatus has allowed, for the first
399: time, to visualize and quantitatively analyse the \lq\lq barber pole
400: turbulence\rq\rq at large scales.  First, this study brings quantitative
401: evidence of the spatio-temporal periodicity of the pattern and 
402: its dependance on control parameters.  Second, it
403: shows that the spiral  turbulence, previously described as alternating
404: turbulent and laminar bands,  actually appears through a modulation of
405: the homogeneous  turbulent state.  To our knowledge,  it is  the first
406: time  that  the  transition  from  homogeneous  turbulence  to  spiral
407: turbulence  is  investigated. Finally  measurements  of the  turbulent
408: stripes width relatively to the  pattern wavelength have allowed us to
409: describe  the  relaxation  mechanism  followed between  the  modulated
410: turbulent state and the laminar-turbulent coexistence.
411: 
412: We thank H. Chat\'e, P. Manneville and I. Mutabazi for
413: fruitful discussions, and C. Gasquet and V. Padilla for technical
414: assistance.
415: 
416: \vspace{-0.5cm}
417: \begin{references}
418: \vspace{-1.7cm}
419: 
420: \bibitem{Andereck}
421: C. D. Andereck, S. S. Liu and H. L. Swinney, J. Fluid Mech. {\bf 164},
422: 155 (1986).
423: 
424: \bibitem{Feyn}
425: R. P. Feynman, {\it Lecture Notes in Physics} (Addison-Wesley Reading,
426: MA, 1964), vol. 2.
427: 
428: \bibitem{Coles}
429: D. Coles, J. Fluid Mech., {\bf 21}, 385 (1965).
430: 
431: \bibitem{VanAtta}
432: C. Van Atta, J. Fluid Mech., {\bf 25}, 495 (1966).
433: 
434: \bibitem{Coles-VA}
435: D. Coles and C. W. Van Atta, Phys. of Fluids Supp., S120 (1967).
436: 
437: \bibitem{Heg1}
438: J. Hegseth, C. D. Andereck, F. Hayot and Y Pomeau, Phys. Rev. Letters,
439: {\bf 62}, 257 (1989).
440: 
441: \bibitem{Heg2}
442: J. Hegseth, C. D. Andereck, F. Hayot and Y. Pomeau, Eur. J. Mech. B, {\bf 10}, 221 (1991).
443: 
444: \bibitem{DauDav}
445: O. Dauchot and F. Daviaud, Phys. of Fluids, {\bf 7}, 335 (1995).
446: 
447: \bibitem{BottChat}
448: S. Bottin and H. Chat\'e, Eur. Phys. J. B, {\bf 6}, 143 (1198). 
449: 
450: \bibitem{TCvisu}
451: A. Prigent and O. Dauchot, submitted to Physics of Fluids (1999).
452: 
453: \bibitem{Snyder}
454: H. A. Snyder, The Physics of Fluids, {\bf 11}, 1599 (1968).
455: 
456: \bibitem{Croq}
457: V. Croquette and H. Williams, Physica D, {\bf 37}, 300 (1989).
458: 
459: \bibitem{GohMut}
460: A. Goharadzeh and I. Mutabazi, in {\it $2^{ième}$ Colloque sur le
461: Chaos temporel et le Chaos spatio-temporel}, Rouen (1998).
462: 
463: \bibitem{BotDavManDau}
464: S. Bottin, F. Daviaud, P. Manneville and O. Dauchot, Eur. Phys. Lett.,
465: {\bf 43}, 171 (1998).
466: 
467: \bibitem{ChatMan}
468: H. Chat\'e and P. Manneville, in {\it Turbulence, A tentative dictionary},
469: P. Tabeling and O. Cardoso Eds., New York, p 111-116, Plenum Press (1994).
470: 
471: \bibitem{ColAnd}
472: P. W. Colovas and C. D. Andereck, Phys. Rev. E, {\bf 55}, 2736 (1997).
473: \end{references}
474: 
475: \end{multicols}
476: 
477: \end{document}
478: %
479: % ****** End of file template.aps ******
480: 
481: 
482: 
483: 
484: 
485: 
486: 
487: 
488: 
489: 
490: 
491: 
492: 
493: 
494: