1: \documentclass[aps,prl,twocolumn,showpacs,floatfix]{revtex4}
2: \usepackage{epsfig}
3:
4: \begin{document}
5: \title{Extended Self-Similarity and Hierarchical Structure in Turbulence}
6: \author{Emily S. C. Ching$^1$, Zhen-Su She$^{2,3}$,
7: Weidong Su$^{1,2}$ and Zhengping Zou$^2$}
8: \affiliation{$^1$Department of Physics, The Chinese University of Hong Kong,
9: Shatin, Hong Kong \\
10: $^2$ State Key Laboratory for Turbulence Research. Dept. of Mech.
11: and Engin. Sci., Peking University,
12: Beijing 100871, People's Republic of China \\
13: $^3$ Department of Mathematics, UCLA, Los Angeles, California 90095}
14: \date{\today}
15: \begin{abstract}
16: It is shown that the two remarkable properties of turbulence,
17: the Extended Self-Similarity (ESS) [R. Benzi {\it et al.}, Phy. Rev.
18: E {\bf 48}, R29, (1993)] and
19: the She-Leveque Hierarchical Structure (SLHS)
20: [Z.S. She and E. Leveque, Phy. Rev. Lett.
21: {\bf 72}, 336, (1994)] are
22: related to each other. In particular, we have shown that
23: a generalized hierarchical structure together with the most intense
24: structures being shock-like give rise to ESS. Our analysis thus suggests
25: that the ESS measured in turbulent flows is an indication of the shock-like
26: intense structures. Results of analysis of velocity measurements in a pipe-flow
27: turbulence support our conjecture.
28: \end{abstract}
29: \pacs{47.27.-i}
30: \maketitle
31:
32: Fully developed turbulence is characterized by
33: power-law dependence of the moments of velocity fluctuations.
34: It was suggested by Kolmogorov in 1941 (K41)\cite{K41}
35: that there is a constant rate of energy transfer
36: from large to small scales and that the statistical properties
37: of the velocity difference
38: across a separation $r$, $\delta v_r \!\equiv\! v(x+r)\!-\!v(x)$, depend
39: only on the mean energy transfer or equivalently the mean energy
40: dissipation rate $\epsilon$ and the scale $r$ when $r$ is within an
41: inertial range. Dimensional considerations then lead to the prediction
42: that the velocity structure functions, which are moments of the magnitude
43: of the velocity difference, have simple power-law dependence on $r$ within
44: the inertial range:
45: \begin{equation}
46: S_p(r) \equiv \langle |\delta v_r|^p \rangle \sim \epsilon^{p/3} r^{p/3}
47: \label{K41}
48: \end{equation}
49:
50: Experiments\cite{Frisch} have indicated that there is indeed power law
51: scaling in the inertial range but the scaling exponents are different from
52: $p/3$:
53: \begin{equation}
54: S_p(r) \sim r^{\zeta_p}
55: \label{inter}
56: \end{equation}
57: where $\zeta_p$ has a nonlinear dependence on $p$.
58: Such a deviation implies that the functional form of the
59: probability density function (pdf) of $\delta v_r$ depends on $r$,
60: that is, the velocity fluctuations have scale-dependent statistics.
61: Understanding this deviation from K41 is essential to our
62: fundamental understanding of the small scale statistical
63: properties of turbulence.
64:
65: Recently, Benzi {\it et al.}\cite{ESS} have discovered a
66: remarkable new scaling property: $S_p(r)$ has a power-law
67: dependence on $S_3(r)$ over a range substantially longer than
68: the scaling range obtained by plotting $S_p(r)$ as a function of $r$.
69: This behavior was named {\it Extended Self-Similarity} (ESS); its
70: discovery has enabled more accurate determination of the
71: scaling exponents $\zeta_p$, particularly at moderately high Reynolds
72: numbers assessible experimentally and numerically,
73: It was later reported that ESS is invalid for
74: anisotropic turbulent flows such as atmospheric boundary layer and
75: channel flow\cite{SS93,BST96,ABS97}. This inspires the study of
76: a generalized ESS (GESS), a scaling behavior of
77: the normalized structure functions when plotted against each
78: other\cite{BBCST95,BBCST96}, which is still valid in these anisotropic
79: flows. The validity of ESS suggests that the different order
80: structure functions have the same dependence on $r$ when
81: $r$ is near the dissipative range\cite{BCBC95,SLP96,FG01}.
82: Very recently, Yakhot argued that some mean-field approximation of
83: the pressure contributions in the Navier-Stokes equation would
84: lead to ESS\cite{Yakhot}.
85:
86: A number of phenomenological models have been proposed to
87: explain the anomalous scaling exponents $\zeta_p$.
88: Among them, a recent successful one
89: is from She and Leveque\cite{SL94} with a hypothesis of
90: a hierarchical structure (HS). When stated for the velocity structure
91: functions, the HS hypothesis reads
92: \begin{equation}
93: {S_{p+2}(r) \over S_{p+1}(r)} = A_{p+1} \left[{S_{p+1}(r) \over
94: S_{p}(r)}\right]^{\beta} [S^{(\infty)}(r)]^{1-\beta}
95: \label{SLvelocity}
96: \end{equation}
97: Here $S^{(\infty)}(r) \!\equiv\! \lim_{p \!\to\! \infty} S_{p+1}(r)/ S_{p}(r)$
98: and $0 \!<\! \beta \!<\! 1$ is a constant.
99: This hypothesis (and the similar version for the local energy dissipation)
100: was supported by experimental velocity measurements
101: taken in turbulent jets and wakes\cite{CBC95,CBBC95,CB97}.
102: Note that the boundedness of the velocity will ensure the existence of
103: $S^{(\infty)}(r)$. In fact, one can show that
104: $S^{(\infty)}(r) \!\equiv\! \lim_{p \to \infty} S_p^{1/p}$, and is equal to
105: $|\delta v_r|^{\max}$,
106: the maximum magnitude of $\delta v_r$\cite{ChingSu}.
107:
108: It was later reported that
109: the passive scalar structure functions\cite{CBC96} and
110: the local passive scalar dissipation\cite{LRCBC99},
111: the temperature structure functions\cite{Ching00}
112: and the local temperature dissipation\cite{ChingKwok00}
113: in turbulent convection, and the velocity structure
114: functions in a class of shell models\cite{FDB95,BBT96,LS97}
115: were all found to possess similar hierarchical structures.
116:
117: In this Letter, we shall show that the She-Leveque hierarchical
118: structure (SLHS) leads to GESS. In other words, the SLHS is
119: a special form of GESS. We then give a generalized form of SLHS,
120: which is equivalent to GESS, with the constant $\beta$ replaced
121: by a function of $p$. Furthermore, it is shown that if
122: $|\delta v_r|^{\max}$ is independent of $r$ for $r$ within the GESS
123: range (which will be the case when the most intense structures
124: are shock-like), the generalized SLHS (and hence GESS) will give
125: rise to ESS. We then conjecture that the observed ESS in turbulent
126: flows is an indication of the most intense structures
127: being shock-like. Consequently, we predict that in the anisotropic flows where
128: GESS but not ESS holds, $|\delta v_r|^{\max}$ has a dependence on $r$.
129: Finally, we present a systematic procedure of analysis
130: of experimental turbulent signals. Application of
131: this analysis to turbulent velocity fluctuations in a pipe flow
132: demonstrates that $|\delta v_r|^{\max}$ indeed depends on $r$
133: in the near-wall strong-shear regions where only GESS but not ESS is valid
134: but is consistent with being $r$-independent in the centerline of the pipe
135: where ESS is valid.
136:
137: First, rewrite Eq.~(\ref{SLvelocity}) as
138: \begin{equation}
139: {S_{k+1}(r) \over S_{k}(r) S^{(\infty)}(r)} = A_{k}
140: \left[{S_{k}(r) \over S_{k-1}(r) S^{(\infty)}(r)}\right]^{\beta}
141: \label{rewriteSL}
142: \end{equation}
143: for integer $k$, which implies
144: \begin{equation}
145: {S_{k}(r) \over S_{k-1}(r) S^{(\infty)}(r)} =
146: \Pi_{j=0}^{k-1} A_j^{\beta^{k-1-j}}
147: \left[{S_1 (r) \over S^{(\infty)}(r)} \right]^{\beta^{k-1}}
148: \label{step1}
149: \end{equation}
150: ultiply Eq.~(\ref{step1}) for $k = 1, 2, \ldots, n$ gives
151: \begin{equation}
152: S_n(r) = B_n
153: \left[ {S_1(r) \over S^{(\infty)}(r)} \right]^{1-\beta^n \over 1-\beta}
154: [S^{(\infty)}(r)]^n
155: \label{step2}
156: \end{equation}
157: where $B_n \!\equiv\! \Pi_{k=1}^n \Pi_{j=0}^{k-1} A_j^{\beta^{k-1-j}}$,
158: which further gives
159: \begin{equation}
160: {S_n(r) \over S_3(r)^{n/3}} = {B_n \over B_3^{n/3}}
161: \left[ {S_1(r) \over S^{(\infty)}(r)} \right]^
162: {3(1-\beta^n)-n(1-\beta^3) \over 3(1-\beta)}
163: \label{step3}
164: \end{equation}
165: Equation~(\ref{step3}) then implies the GESS property:
166: \begin{equation}
167: T_n(r) \sim T_m(r)^{\rho(n,m)}
168: \label{GESSt}
169: \end{equation}
170: which is a scaling behavior for the normalized structure functions,
171: $T_n(r) \!\equiv\! S_n(r)/S_3(r)^{n/3}$, with the normalized
172: exponents $\rho(n,m)$ even when $S_n(r)$ does
173: not have a scaling behavior in $r$. For SLHS,
174: \begin{equation}
175: \rho(n,m)= {3(1-\beta^n)-n(1-\beta^3) \over 3(1-\beta^m)-m(1-\beta^3) }
176: \label{rhoSL}
177: \end{equation}
178: We have thus proved that SLHS implies GESS.
179:
180: Earlier, Benzi et al.\cite{BBCST96} showed that
181: GESS holds if the structure functions are of the form
182: \begin{equation}
183: S_p(r) \sim g_1(r)^p g_2(r)^{H(p)}
184: \label{g1g2}
185: \end{equation}
186: for any functions $g_1(r)$, $g_2(r)$, and $H(p)$. On the other hand,
187: if Eq.~(\ref{GESSt}) with any $\rho(n,m)$ is valid,
188: we can always write the structure functions in the form of
189: Eq.~(\ref{g1g2}), say, with $g_1(r) = S_3(r)^{1/3}$,
190: $g_2(r)=S_{q^*}(r)/S_3(r)^{{q^*}/3}$, and $H(p) = \rho(p,q^*)$
191: for any chosen value of $q^*$. Thus, Eq.~(\ref{g1g2}) is
192: equivalent to GESS.
193:
194: Our demonstration here gives the functions $g_1(r)$, $g_2(r)$
195: a meaning. Although the choice of $g_1(r)$ and $g_2(r)$ is
196: not unique, the following form can be obtained from Eq.~(\ref{step2}):
197: $g_1(r)=S^{(\infty)}(r)$ and $g_2(r)=S_3(r)/[S^{(\infty)}(r)]^3$.
198: Here, $g_1(r)$ describes the $r$-dependence of very strong fluctuations
199: $S^{(\infty)}(r)$ and $g_2(r)$ describes the normalized $r$-dependence of
200: (typical) weak fluctuations [e.g. $S_3(r)$] by $[S^{(\infty)}(r)]^3$.
201: Express $g_1$ and $g_2$ this way, we have
202: \begin{equation}
203: S_p(r) \sim \left[S^{(\infty)}(r)\right]^p
204: \left\{{S_3(r) \over [S^{(\infty)}(r)]^3}\right\}^{f(p)}
205: \label{modg1g2}
206: \end{equation}
207: where $f(p)$ is a function to be discussed below. Note that
208: when $g_2(r)$ is constant, or the weak fluctuations have the
209: same $r$-dependence as the strong fluctuations, we have the K41
210: scaling. The function $f(p)$ has to satisfy various conditions.
211: By definition, $f(0)=0$, $f(3)=1$, and
212: $\lim_{p\to \infty} f(p+1)-f(p)=\lim_{p\to \infty}f(p)/p = 0$.
213: Furthermore, the boundedness
214: of the velocity restricts that $df(p)/dp \ge 0$.
215:
216: Note that Eq.~(\ref{modg1g2}) is a general expression of GESS.
217: Rewriting Eq.~(\ref{modg1g2}) in the form of Eq.~(\ref{rewriteSL}),
218: we have
219: \begin{equation}
220: {S_{p+1}(r) \over S_{p}(r) S^{(\infty)}(r)} = C_{p}
221: \left[{S_{p}(r) \over S_{p-1}(r) S^{(\infty)}(r)}\right]^{g(p)}
222: \label{generalHS}
223: \end{equation}
224: where $g(p) \equiv [f(p+1)-f(p)]/[f(p)-f(p-1)]$.
225: It is clear that SLHS corresponds to the
226: particular case of $g(p)= \beta$ or equivalently
227: \begin{equation}
228: f(p)={1-{\beta}^p \over 1-\beta^3}
229: \label{fpSL}
230: \end{equation}
231: We can list several additional possible classes of $f(p)$, which
232: include $f(p)\!=\! [(p\sigma+1)^{\alpha} \!-\! 1]/
233: [(3\sigma+1)^{\alpha} \!-\! 1]$ and
234: $f(p) \!=\! \ln(p \sigma+ 1)/\ln(3 \sigma + 1)$
235: with $\sigma \!>\! 0$ and $0\!<\! \alpha\!<\! 1$.
236: The latter is the limit of the former when $\alpha \to 0$.
237: These two cases correspond to the two results derived by
238: Novikov, Eqs. (18) and (19) in Ref.\cite{N94}, using the
239: theory of infinitely divisible distributions.
240:
241: To study when GESS would further give rise to ESS,
242: we rewrite Eq.~(\ref{GESSt}) in the following form:
243: \begin{equation}
244: {S_p(r)^{1/p} \over S_3(r)^{1/3}} \sim
245: \left[{S_q(r)^{1/q} \over S_3(r)^{1/3}}\right]^{{q \over p} \rho(p,q)}
246: \label{rewriteGESS}
247: \end{equation}
248: with
249: \begin{equation}
250: \rho(p,q) = {f(p) - p/3 \over f(q)- q/3}
251: \label{rho}
252: \end{equation}
253: Thus we see that if there exists an $p^* \ne 0$ such that
254: $S_{p^*}(r)^{1/p^*}$
255: is independent of $r$, then we would have a scaling behavior of $S_p$ vs
256: $S_3$ (and thus $S_q$ for $q \ne p$).
257: If $p^*$ is finite, $\zeta_{p^*} = 0$ which
258: implies $\zeta_m = 0$ for $0 \le m \le p^*$ because $\zeta_n$ has
259: to be a non-decreasing function of $n$ in order for the velocity field
260: to be bounded\cite{Frisch}. It further gives $\zeta_p = 0$ for all
261: values of $p$ if $\zeta_p$ is an analytic function of $p$.
262: This could be avoided only if $p^*$ approaches $\infty$,
263: which means that $S^{(\infty)}(r)$ is independent of $r$.
264: Hence, GESS together with the condition that
265: $S^{(\infty)}(r)$ or equivalently $|\delta v_r|^{\max}$
266: being independent of $r$ would give rise to ESS:
267: \begin{equation}
268: S_p(r) \sim S_3(r)^{\eta(p,3)}
269: \label{ESSt}
270: \end{equation}
271: with $\eta(p,3)\!=\! f(p)$
272: even when $S_p(r)$ does not have a power-law dependence on $r$.
273: We note that the presence of shocks with maximum velocity, $v_{\max}$,
274: in opposite directions across the shock discontinuity would give
275: $|\delta v_r|^{\max} = 2 v_{\max}$, which is thus independent of $r$.
276:
277: Using Eq.~(\ref{modg1g2}), an independence of $S^{(\infty)}$ on $r$ implies
278: \begin{equation}
279: \zeta_p = \zeta_3 f(p)
280: \label{zetap}
281: \end{equation}
282: For SLHS, $\lim_{p \to \infty} f(p) \!=\! 1/(1-\beta^3)$ hence
283: Eq.~(\ref{zetap})
284: further implies a saturation of $\zeta_p$ as $p \to \infty$.
285: A connection of the saturation of the exponents with the existence of
286: shocks was suggested earlier by Chen and Cao\cite{CC95}.
287:
288: Both ESS and SLHS have been observed in a variety of fluctuating
289: flow fields. The above analysis points to a possibility that
290: the two combined may be indicative of a property of the flow field:
291: $S^{(\infty)}(r)$ is independent of $r$. One way to
292: check the plausibility of these ideas is to examine experimental
293: velocity measurements that demonstrate GESS, and to investigate
294: whether $|\delta v_r|^{\max}$ is indeed independent of $r$ when
295: ESS is valid and otherwise dependent on $r$ when ESS is not valid.
296: Note, however, that the detectable $|\delta v_r|^{\max}$
297: in any finite measurement would almost surely underestimate
298: the true value.
299:
300: If the GESS is of the special form of SLHS, a systematic procedure
301: can be developed to perform an indirect estimate of the
302: $r$-dependence of $S^{(\infty)}(r)$ as follows.
303: First, one verifies if the SLHS is valid by performing
304: a $\beta$-test. It consists of computing the normalized structure
305: functions $T_p(r)$ and obtaining the normalized exponents
306: $\rho(p,q)$ by measuring the slopes of $\ln(T_p)$ vs
307: $\ln(T_q)$. Let $\Delta\rho(p,q)=\rho(p+1,q)\!-\!\rho(p,q)$.
308: It is easy to derive the following equation when SLHS is valid:
309: \begin{equation}
310: \label{beta-test}
311: \Delta\rho(p+1,q) =\beta \Delta\rho(p,q)
312: \!+\!{(1-\beta)(1-\beta^3)\over q(1-\beta^3)\!-\!3(1-\beta^q)}
313: \end{equation}
314: If one finds parallel straight lines when
315: ploting $\Delta\rho(p+1,q)$ vs $\Delta\rho(p,q)$
316: for a set of values of $q$,
317: we say that the data passes the $\beta-$test and the turbulent
318: flow field possesses the SLHS property. The slope and intercept
319: provides a double estimate of the constant $\beta$.
320: With the estimated $\beta$,
321: one can then construct $f(p)$ using Eq.~(\ref{fpSL}).
322:
323: For an indirect estimate of $S^{(\infty)}(r)$, we introduce
324: \begin{equation}
325: \label{Fp}
326: F_p(r,r_0)\!\equiv\! {\ln[S_p(r)/S_p(r_0)] \!-\! f(p)
327: \ln[S_3(r)/S_3(r_0)] \over p-3f(p)}
328: \end{equation}
329: From Eq.~(\ref{modg1g2}),
330: $F_p(r,r_0)$ should be independent of $p$ and
331: equal to $\log [S^{(\infty)}(r)/S^{(\infty)}(r_0)]$ for
332: $r$ within the GESS range.
333: Thus, one computes $F_p(r,r_0)$ for a fixed value of $r_0$,
334: within the GESS range, and plots it as a function of
335: $r$ for a set of values of $p$ to estimate indirectly the
336: $r$-dependence of $\ln [S^{(\infty)}(r)/S^{(\infty)}(r_0)]$.
337: In particular, if $S^{(\infty)}$ is independent of $r$, then
338: $F_p(r,r_0)=0$ for $r$ within the GESS range.
339:
340: We have applied the above procedure to analyze hot-wire measurements of
341: longitudinal velocity fluctuations in a pipe flow\cite{note}.
342: The pipe is 22.5~m long with an inner diameter of 10.5~cm, and
343: the Reynolds number is $1.35 \times 10^5$.
344: The velocity measurements studied were taken at
345: a cross section at 18.2~m away from the entrance both at the
346: centerline of the pipe and at a distance of 0.1~mm from the pipe-wall.
347: The number of measurements taken at each point is $5.76 \times 10^7$
348: and the maximum order of moments computed is $p=10$.
349: We have checked that GESS is valid for both sets of measurements but
350: ESS is valid only at the centerline.
351: The range of GESS (and also ESS when valid) is $r \approx 10 - 500$,
352: in units of the sampling time $1/48$~ms. All $r$'s quoted below will be
353: in the same units.
354: We have obtained $\rho(p,q)$ for the two locations and $\eta(p,3)$ for
355: the centerline location. In Fig.~\ref{rhoplot}, we plot
356: $\Delta \rho(p+1,q)$ vs $\Delta \rho(p,q)$
357: for some values of $q$ for both locations.
358: The data can be fitted by parallel straight lines
359: showing that the hierarchical structure is indeed of the SL
360: form in both locations.
361: We estimate the value of $\beta$ simultaneously from the slope
362: and the intercept and get $0.93 \pm 0.01$
363: and $0.85 \pm 0.01$ respectively for the centerline and the near-wall location.
364: In the inset, we see good agreement of $\eta(p,3)$ with $f(p)$
365: for the centerline measurements. This implies the $r$-independence
366: of $S^{(\infty)}$, in accord with the validity of ESS in these
367: measurements.
368:
369: \begin{figure}
370: \centering
371: \includegraphics[width=.40\textwidth]{Fig1.eps}
372: \caption{
373: $\Delta \rho(p+1,q)$ vs $\Delta \rho(p,q)$
374: for $q\!=\!1$ (squares), $q\!=\!2$ (circles), $q\!=\!4$~(triangles),
375: $q\!=\!5$~(plusses), and $q\!=\!8$ (diamonds).
376: Above the solid line (slope $\!=\! 0.93$) are the
377: centerline data and below the dashed
378: line (slope $\!=\! 0.85$) are the near-wall data (shifted for clarity).
379: In the inset, we compare $\eta(p,3)$~(circles) with
380: $f(p)$~(solid line) for the centerline data.}
381: \label{rhoplot}
382: \end{figure}
383:
384: We have next computed $F_p(r,r_0)$ with $r_0\!=\!69$ for both locations
385: and found that the data indeed collapse when $p$ is large.
386: In Fig.~\ref{Fplots}, we plot $F_p(r,r_0)$ as
387: a function of $r$.
388: We see that $F_p(r,r_0)$ for the centerline measurements is almost
389: zero for $r \!\ge\! 30$ while that for the near-wall measurements
390: shows a clear $r$-dependence throughout. Since ESS is valid for the
391: centerline but not the near-wall measurements (see the inset), these
392: results support our conjecture that the validity of ESS in turbulent
393: flows is an indication of $S^{(\infty)}$ being independent of $r$.
394:
395: \begin{figure}
396: \centering
397: \includegraphics[width=.40\textwidth]{Fig2.eps}
398: \caption{$F_p(r,r_0)$ for $p \ge 4$ for both the centerline and near-wall
399: measurements. In the inset, we show the local slope of
400: $\ln S_8(r)$ vs $\ln S_3(r)$ as a function of $\ln S_3(r)$, illustrating
401: that ESS is valid for the centerline (solid line)
402: but not the near-wall (dashed line) measurements.}
403: \label{Fplots}
404: \end{figure}
405:
406: In summary, our analysis proposes a link between a measurable
407: statistical property of turbulence, the ESS, and a property of
408: the most intense structure, the $r$-independence of
409: $|\delta v_r|^{\max}$ or $S^{(\infty)}(r)$. We have developed a
410: systematic procedure to test this conjecture
411: when the SL hierarchical structure holds.
412: The analyis of experimental pipe flow data support our conjecture:
413: the near-wall strong-shear turbulence contains more complex structures while
414: the centerline fully developed turbulence has shock-like
415: structure for $S^{(\infty)}(r)$.
416: A consequence of the $r$-independence of $S^{(\infty)}(r)$ is
417: the saturation of the scaling exponents of the velocity
418: structure functions at very high orders.
419: Further experimental and numerical tests are highly desirable.
420:
421: \begin{acknowledgments}
422: Z.Z. thanks Prof. M.D. Zhou and Y. Zhu for their help in
423: conducting the pipe-flow velocity measurements. The work at the
424: Chinese University of Hong Kong was supported by the
425: Research Grants Council of the Hong Kong SAR, China
426: (CUHK 4119/98P and CUHK 4286/00P)
427: while that at Peking University was partially supported by a grant
428: from National Natural Science Foundation (NNSF) of China and from the
429: Ministry of Education of China.
430: \end{acknowledgments}
431:
432: \begin{thebibliography}{99}
433: \bibitem{K41} A. N. Kolmogorov, C.R. Acad. Sci. URSS {\bf 30}, 301
434: (1941).
435: \bibitem{Frisch} U. Frisch, {\it Turbulence: The Legacy of
436: A. N. Kolmogorov} (Cambridge University Press, Cambridge 1995).
437: \bibitem{ESS} R. Benzi, S. Ciliberto, R. Tripiccione, C. Baudet, F.
438: assaioli, and S. Succi, Phys. Rev. E {\bf 48}, R29 (1993);
439: R. Benzi, S. Ciliberto, C. Baudet, G. Ruiz Chavarria, and R. Tripiccione,
440: Europhys. Lett. {\bf 24}, 275 (1993).
441: \bibitem{SS93} G. Stolovitzky and K. R. Sreenivasan, Phys. Rev. E
442: {\bf 48}, R33 (1993).
443: \bibitem{BST96} R. Benzi, M.V. Struglia, and R. Tripiccione, Phys. Rev. E
444: {\bf 53}, R5565 (1996).
445: \bibitem{ABS97} G. Amati, R. Benzi, and S. Succi, Phys. Rev. E {\bf 55},
446: 6985 (1997).
447: \bibitem{BBCST95} R. Benzi, L. Biferale, S. Ciliberto, M. V. Struglia,
448: and R. Tripiccione, Europhys. Lett. {\bf 32}, 709 (1995).
449: \bibitem{BBCST96} R. Benzi, L. Biferale, S. Ciliberto, M. V. Struglia,
450: and R. Tripiccione, Phys. Rev. E {\bf 53}, R3025 (1996).
451: \bibitem{BCBC95} R. Benzi, S. Ciliberto, C. Baudet, and G. Ruiz
452: Chavarria, Physica D {\bf 80}, 385 (1995).
453: \bibitem{SLP96} D. Segel, V. L'vov, and I. Procaccia, Phys. Rev. Lett.
454: {\bf 76}, 1828 (1996).
455: \bibitem{FG01} H. Fujisaka and S. Grossmann, Phys. Rev. E {\bf 63},
456: 026305 (2001).
457: \bibitem{Yakhot} V. Yakhot, Phys. Rev. Lett. {\bf 86}, 234501 (2001).
458: \bibitem{SL94} Z.-S. She and E. Leveque, Phys. Rev. Lett. {\bf 72},
459: 336 (1994); Z.-S. She, Lecture Notes in Physics, {\bf v.491}, eds. Eden
460: {\it et al.}, 28 (1997); Z.-S. She, Prog. Theo. Phys. Supp. (Japan),
461: {\bf 130}, 87-102 (1998).
462: \bibitem{CBC95} G. Ruiz Chavarria, C. Baudet, and S. Ciliberto, Phys.
463: Rev. Lett. {\bf 74}, 1986 (1995).
464: \bibitem{CBBC95} G. Ruiz Chavarria, C. Baudet, R. Benzi, and S.
465: Ciliberto, J. Phys. II {\bf 5}, 485 (1995).
466: \bibitem{CB97} R. Camussi and R. Benzi, Phys. Fluids {\bf 9}, 257 (1997).
467: \bibitem{ChingSu} E.S.C. Ching and W. Su (unpublished).
468: \bibitem{CBC96} G. Ruiz Chavarria, C. Baudet, and S. Ciliberto, Physica D
469: {\bf 99}, 369 (1996).
470: \bibitem{LRCBC99} E. Leveque, G. Ruiz-Chavarria, C. Baudet, and
471: S. Ciliberto, Phys. Fluids {\bf 11}, 1869 (1999).
472: \bibitem{Ching00} E.S.C. Ching, Phys. Rev. E {\bf 61}, R33 (2000).
473: \bibitem{ChingKwok00} E.S.C. Ching and C. Y. Kwok,
474: Phys. Rev. E {\bf 62}, R7587 (2000).
475: \bibitem{FDB95} P. Frick, B. Dubrulle, and A. Babiano, Phys. Rev. E
476: {\bf 51}, 5582 (1995).
477: \bibitem{BBT96} R. Benzi, L. Biferale, and E. Trovatore, Phys. Rev.
478: Lett. {\bf 77}, 3114 (1996).
479: \bibitem{LS97} E. Leveque and Z.-S. She, Phys. Rev. E {\bf 55}, 2789
480: (1997).
481: \bibitem{Feller} W. Feller, {\it An Introduction to Probability Theory
482: and its Applications}, 3rd Ed. Vol. 2 (Wiley, New York 1968).
483: \bibitem{N94} E. A. Novikov, Phys. Rev. E {\bf 50}, R3304 (1994).
484: \bibitem{CC95} S. Chen and N. Cao, Phys. Rev. E {\bf 52} R5757 (1995).
485: \bibitem{note} The experiment was conducted at the State Key Laboratory for
486: Turbulence Research at Peking University.
487: \end{thebibliography}
488:
489: \end{document}
490: