cond-mat0307604/lb.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: \documentstyle[aps]{revtex}
3: %----------------------------------------------------------------
4: \begin{document}
5: \draft
6: \title{
7: Mean-field and fluctuation analyses 
8: of a forced turbulence 
9: simulated by the lattice Boltzmann method 
10: }
11: \author{
12: W. Sakikawa and O. Narikiyo
13: }
14: \address{
15: Department of Physics, 
16: Kyushu University, 
17: Fukuoka 810-8560, 
18: Japan
19: }
20: \date{
21: July 16, 2003
22: }
23: \maketitle
24: %----------------------------------------------------------------
25: \begin{abstract}
26: On the basis of the lattice Boltzmann method 
27: we have done a numerical experiment of a forced turbulence 
28: in real space and time. 
29: Our new findings are summarized into two points. 
30: First 
31: in the analysis of the mean-field behavior of the velocity field 
32: using the exit-time statistics  
33: we have verified Kolmogorov's scaling and Taylor's hypothesis 
34: for the first time 
35: in the simulation for the Navier-Stokes equation. 
36: Second 
37: in the analysis of the intermittent velocity fluctuations 
38: using a non-equilibrium probability distribution function 
39: and the wavelet denoising 
40: we have clarified that the coherent vortices sustain 
41: the power-law velocity correlation in the non-equilibrium state. 
42: \end{abstract}
43: %----------------------------------------------------------------
44: \vskip 18pt
45: \section{Introduction}
46: \label{intro}
47: 
48: Understanding turbulence has not been attained over millenniums. 
49: However, new approaches to turbulence\cite{Frisch} 
50: have been invented successively. 
51: 
52: Recently turbulence data in nature and laboratories 
53: are analyzed from new view points. 
54: One\cite{exit-time-stat} is based on the exit-time statistics 
55: and discusses Kolmogorov's scaling\cite{Frisch}. 
56: The other\cite{PDF-space} 
57: on the basis of the R\'enyi-Tsallis statistics\cite{RT1,RT2} 
58: discusses the probability distribution functions (PDF). 
59: 
60: Our study starts from the application of the above two schemes 
61: to our numerical simulation data 
62: and aims to obtain some new insight of the turbulence. 
63: 
64: Before reporting our study of the turbulence 
65: we briefly review here 
66: our strategy approaching intermittency problems 
67: in other non-equilibrium systems. 
68: In the numerical study 
69: of the critical spin-state of the Ising model\cite{SN1} 
70: we have observed 
71: that the degree of the non-Gaussianity of the R\'enyi-Tsallis PDF 
72: represents that of the non-equilibrium. 
73: On the other hand, 
74: in the numerical study of a supercooled liquid\cite{SN2} 
75: near the glass transition 
76: we have observed 
77: that the spatial distribution of the coherently moving regions 
78: is well analyzed by the singularity spectrum of multifractal. 
79: Since the R\'enyi-Tsallis statistics can describe 
80: the systems with multifractality, 
81: the above two observations are closely related. 
82: 
83: In these two systems 
84: the correlation length of the fluctuations becomes divergently large 
85: and we can expect the scale-invariance. 
86: A scale-invariant system has multifractal nature. 
87: On the other hand, 
88: we can also expect the scale-invariance 
89: in the turbulent systems at high Reynolds number\cite{Frisch}. 
90: Thus we adopt the same strategy 
91: to intermittency problem in turbulence. 
92: 
93: The next section describes our simulation method. 
94: In the following sections the simulation data are analyzed 
95: at the mean-field level at first 
96: and subsequently taking fluctuations into account. 
97: In the final section a brief summary is given. 
98: 
99: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
100: 
101: \section{Lattice Boltzmann simulation}
102: \label{LBsimulation}
103: 
104: We simulate a forced turbulence in real space and time 
105: and adopt the lattice Boltzmann method\cite{LB2,LB1} 
106: as one of the easiest ways for such a purpose. 
107: 
108: The distribution function $f_i({\vec r},t)$ 
109: of the particle with the velocity ${\vec c}_i$ 
110: at position ${\vec r}$ and time $t$ 
111: obeys the lattice Boltzmann equation\cite{LB2,LB1} 
112: %============================================
113: \begin{equation}
114: f_i({\vec r}+{\vec c}_i,t+1) - f_i({\vec r},t)
115: = - \omega [ f_i({\vec r},t) - f_i^0({\vec r},t) ],
116: \label{f}
117: \end{equation}
118: %============================================
119: where we have adopted the Bhatnagar-Gross-Krook (BGK) model 
120: for the collision term 
121: and the time step has been chosen as unity. 
122: The position ${\vec r}$ of the particle 
123: is restricted on the cubic lattice 
124: and the lattice spacing is chosen as unity. 
125: The local equilibrium distribution $f_i^0({\vec r},t)$ is assumed 
126: to be given as 
127: %============================================
128: \begin{equation}
129: f_i^0({\vec r},t)
130: = w_i \rho [ 1 + 3 ({\vec c}_i\cdot{\vec u}) 
131:                + {9 \over 2} ({\vec c}_i\cdot{\vec u})^2 
132:                - {3 \over 2} ({\vec u}\cdot{\vec u}) ].
133: \label{f0}
134: \end{equation}
135: %============================================
136: The density $\rho$ and the velocity ${\vec u}$ of the fluid 
137: are given by the sum of the contributions of the particles as 
138: $\rho = \sum_i f_i({\vec r},t)$ and 
139: $\rho {\vec u} = \sum_i f_i({\vec r},t) {\vec c}_i$, respectively. 
140: Using the multiscale analysis 
141: the Navier-Stokes equation, 
142: which is supposed to be able to describe turbulence, 
143: is derived from the present BGK lattice Boltzmann equation 
144: and the corresponding viscosity $\nu = (1/\omega - 1/2)/3$. 
145: 
146: In the following 
147: we use a 15-velocity model and ${\vec c}_i$ is chosen as 
148: ${\vec c}_1 = (0,0,0)$, 
149: ${\vec c}_i = (\pm1,0,0),(0,\pm1,0),(0,0,\pm1)$ 
150: for $i=2,3,\cdot\cdot\cdot,7$ 
151: and 
152: ${\vec c}_i = (\pm1,\pm1,\pm1)$ for $i=8,9,\cdot\cdot\cdot,15$. 
153: Then the weight factor $w_i$ for the local equilibrium distribution 
154: is determined as 
155: $w_1 = 2/9$, 
156: $w_i = 1/9$ for $i=2,3,\cdot\cdot\cdot,7$ and 
157: $w_i = 1/72$ for $i=8,9,\cdot\cdot\cdot,15$. 
158: The relaxation frequency $\omega$ is chosen as $\omega=1.94$. 
159: 
160: In order to simulate forced turbulence 
161: we add the forcing term ${\vec g}_i\cdot{\vec F}({\vec r},t)$ 
162: due to an applied body force ${\vec F}({\vec r},t)$ 
163: to the right hand side of Eq.\ (\ref{f}). 
164: In order to reproduce the Navier-Stokes equation 
165: we should choose as ${\vec g}_i = {\vec c}_i/10$ 
166: for the present 15-velocity model\cite{LB2}. 
167: For simplicity we apply a solenoidal force in $y$-direction\cite{LB1}, 
168: ${\vec F}=(0,F_y(z,t),0)$, and $F_y(z,t)$ is a Gaussian white noise 
169: whose variance $\sigma_F$ is a function of the lattice coordinate 
170: in $z$-direction, 
171: $\sigma_F = |0.01 \times \sin(2\pi k_f z / L)|$, 
172: where $L$ is the linear dimension 
173: of the simulation region of cubic box. 
174: We have chosen as $L = 200$ so that $1 \leq x,y,z \leq 200$. 
175: We have adopted the periodic boundary condition. 
176: Although we have chosen as $k_f = 4$, 
177: the results of the simulation is insensitive to the choice. 
178: 
179: In the initial state of the simulation 
180: the fluid density is uniform, $\rho = 1$, and 
181: the spatial distribution of the fluid velocity ${\vec u}$ 
182: is chosen to be random. 
183: 
184: In Fig.\ \ref{fig1} 
185: we show a snapshot of the vorticity field, 
186: $\vec \omega = \nabla \times \vec u$, 
187: in order to visualize a turbulent structure. 
188: We see an inhomogeneous distribution of the vortices. 
189: 
190: The Reynolds number $Re$ for our simulation\cite{LB2,LB1} is 
191: estimated as $Re = \sqrt{ \langle {\vec u}^2 \rangle } L / 2\pi\nu$ 
192: and time-dependent as shown in Fig.\ \ref{fig2}. 
193: Here $\langle \cdot\cdot\cdot \rangle$ represents the spatial average. 
194: In our simulation 
195: we can realize a turbulent state for relatively small Reynolds number, 
196: since we add random force. 
197: 
198: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
199: 
200: \section{Mean-field behavior}
201: \label{MF}
202: 
203: In this section 
204: we discuss the mean-field aspects of the velocity field. 
205: 
206: Turbulence is one of the typical phenomena 
207: with multiscale motions. 
208: Each phenomena of a scale strongly couples with 
209: all the other scales of turbulent motion. 
210: In order to analyze such a system 
211: a scale-dependent entropy, the so-called $\epsilon$-entropy, 
212: works well\cite{exit-time-stat}. 
213: For example, the time series of the velocity fluctuation in turbulence 
214: leads to a non-trivial scaling relation of the $\epsilon$-entropy, 
215: $ h(\epsilon) \propto \epsilon^{-3} $, 
216: expected from Kolmogorov's scaling\cite{exit-time-stat}. 
217: The existing experimental data are consistent 
218: with this scaling\cite{exit-time-stat}. 
219: In this section we try to show the consistency of the scaling 
220: with our numerical experiment simulating the Navier-Stokes equation. 
221: 
222: We focus our attention to the time series of 
223: the $y$-component $u_y({\vec r},t)$ of 
224: the fluid velocity ${\vec u}({\vec r},t)$ 
225: as shown in Fig.\ \ref{fig3}, 
226: since our system is anisotropic due to the forcing. 
227: This turbulent signal is characterized by the exit-time 
228: $\tau(\epsilon;{\vec r})$ at position ${\vec r}$. 
229: At the exit-time $t=\tau(\epsilon;{\vec r})$ 
230: the first exit satisfying the condition, 
231: $|u_y({\vec r},t_0+\tau) - u_y({\vec r},t_0)| > \epsilon/2$, 
232: occurs. 
233: Here the time $\tau$ is measured from $t_0$ and $t_0=1000$ in our simulation 
234: regarding the states at the first 1000 time-steps as transitional. 
235: 
236: The mean of the exit-time, 
237: $\tau(\epsilon) \equiv \langle \tau(\epsilon;{\vec r}) \rangle$, 
238: is related to the $\epsilon$-entropy, $h(\epsilon)$, 
239: as\cite{exit-time-stat} 
240: %============================================
241: \begin{equation}
242: h(\epsilon) \propto {1 \over \tau(\epsilon)}. 
243: \label{h}
244: \end{equation}
245: %============================================
246: In the region 
247: where Kolmogorov's scaling, the 4/5 law\cite{Frisch,exit-time-stat}, holds 
248: the velocity difference between two points behaves as 
249: %============================================
250: \begin{equation}
251: \langle|{\vec u}({\vec r}+{\vec R},t_0)-{\vec u}({\vec r},t_0)|\rangle 
252:      \propto |{\vec R}|^{1/3} . 
253: \label{space-scaling}
254: \end{equation}
255: %============================================
256: Using Taylor's hypothesis\cite{Frisch,exit-time-stat}, we obtain 
257: %============================================
258: \begin{equation}
259: \langle|{\vec u}({\vec r},t_0+\tau)-{\vec u}({\vec r},t_0)|\rangle 
260:      \propto \tau^{1/3},
261: \label{time-scaling}
262: \end{equation}
263: %============================================
264: as a mean-field description. 
265: In spite of the absence of the global flow in our simulation 
266: Taylor's hypothesis is valid in a local sense\cite{JPV} 
267: as shown in the following. 
268: Namely in the mean-field description 
269: the velocity difference in temporal and spatial directions 
270: have the same fractal scaling exponent.
271: With Eq.\ (\ref{time-scaling}) the definition of the exit-time leads to 
272: $\tau(\epsilon)^{1/3} \propto \epsilon$. 
273: From Eq.\ (\ref{h}) we can conclude 
274: that $h(\epsilon) \propto \epsilon^{-3}$ 
275: in the above mean-field description\cite{exit-time-stat}. 
276: 
277: In accordance with the above scaling analysis 
278: our simulation data\cite{SN3} for $1/\tau(\epsilon)$ 
279: shown in Fig.\ \ref{fig4} 
280: have the scaling region for $\epsilon > \epsilon_0$ 
281: where $\epsilon_0 \sim 0.04$. 
282: Thus by our simulation we have established 
283: that $h(\epsilon) \propto \epsilon^{-3}$. 
284: For $\epsilon \ll \epsilon_0$ 
285: the exit in the simulation occurs within one time step 
286: and the curve saturates at small $\epsilon$. 
287: These behaviors are consistent 
288: with the experimantal data\cite{exit-time-stat}. 
289: 
290: In Fig.\ \ref{fig5} 
291: the result of the similar analysis in spatial direction is shown. 
292: In place of $\tau(\epsilon)$ we measure $r(\epsilon)$ 
293: where at the exit-length $r=r(\epsilon;{\vec r})$ 
294: the first exit satisfying the condition, 
295: $|u_y(x+r,y,z,t) - u_y(x,y,z,t)| > \epsilon/2$, 
296: occurs at a time $t$. 
297: Here $r(\epsilon) = \langle r(\epsilon;{\vec r}) \rangle$. 
298: Because of the smallness of the number of the available data 
299: in space direction, 200, in comparison with that in time direction, 7000, 
300: the scaling region is narrow. 
301: Since $1/\tau(\epsilon)$and $1/r(\epsilon)$ 
302: have the same scaling exponent, 
303: Taylor's hypothesis is confirmed to be valid. 
304: 
305: In conclusion we have confirmed 
306: Kolmogorov's scaling and Taylor's hypothesis at the same time. 
307: In contrast to the evaluation of the energy spectrum, 
308: which is usually employed for testing Kolmogorov's scaling 
309: and determined by the two-point correlation function, 
310: Kolmogorov's scaling is easily observed in the $\epsilon$-entropy, 
311: since it is a mean-field description and fluctuations are averaged out. 
312: Such a mean-field scaling is a unifractal description and 
313: fluctuations can be taken into account in a multifractal analysis 
314: as shown in the following. 
315: 
316: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
317: 
318: \section{Fluctuations: qualitative observation}
319: \label{quality}
320: 
321: We have discussed the mean-field aspects of the velocity field 
322: in the preceding section. 
323: Next we discuss the fluctuations of the velocity field. 
324: In this section we make a qualitative observation 
325: using the correlation maps. 
326: 
327: In Fig.\ \ref{fig6} we show the space correlation. 
328: The correlation decays as the length scale is increased. 
329: In Fig.\ \ref{fig7} we show the time correlation. 
330: The correlation decays as the time scale is increased. 
331: At small length or time scale 
332: the strong correlation leads to a non-equilibrium state 
333: which is identified as the coherent vortex in the following section. 
334: 
335: We see a multiscale structure in the correlation map, 
336: Fig.\ \ref{fig8}, 
337: which is one of the characteristics of turbulence 
338: neither periodic nor random. 
339: 
340: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
341: 
342: \section{Fluctuations: non-equilibrium nature}
343: \label{non-equilibrium-nature}
344: 
345: We have discussed the qualitative aspects of the velocity fluctuation 
346: in the preceding section. 
347: Next we try to quantify the fluctuation 
348: by two measures. 
349: One is the non-equilibrium parameter used in this section and 
350: the other is the singularity spectrum in the next section. 
351: 
352: In order to analyze the distribution of the fluctuation 
353: we use the R\'enyi-Tsallis PDF\cite{RT1,RT2} 
354: %============================================
355: \begin{equation}
356: P(X) = P_0 \cdot [ 1 + {q-1 \over 2\sigma_X^2} X^2 ]^{1 \over 1-q}, 
357: \label{PDF-RT}
358: \end{equation}
359: %============================================
360: for a variable $X$. 
361: Here the parameter $q$ represents 
362: the degree of non-equilibrium\cite{SN1}. 
363: To be precise, 
364: the deviation of $q$ from unity 
365: is the measure for the degree of non-equilibrium, 
366: since the equilibrium Gaussian distribution 
367: is expressed by Eq.\ (\ref{PDF-RT}) 
368: in the limit of $q \rightarrow 1$. 
369: 
370: The PDF in Eq.\ (\ref{PDF-RT}) 
371: is derived from the R\'enyi or Tsallis entropy 
372: by using the maximum entropy principle\cite{AA}. 
373: The Tsallis entropy is non-extensive, 
374: while the R\'enyi entropy is extensive. 
375: The functional form of the distribution is independent of 
376: the choice of the entropy, the R\'enyi or Tsallis, 
377: so that our result is independent of the extensivity of the entropy. 
378: 
379: We show the PDF for the velocity difference 
380: between two points in Fig.\ \ref{fig9} and 
381: the vorticity difference in Fig.\ \ref{fig10}. 
382: The $q$-value of the PDF is length-scale dependent. 
383: Similarly the PDF for the velocity or vorticity difference 
384: between two times at a point is time-scale dependent. 
385: In any case 
386: the degree of non-equilibrium is larger 
387: at smaller spatio-temporal scales 
388: where the correlation survives. 
389: As shown later on, 
390: the formation of the coherent vortex structure 
391: sustains the strong correlation leading to non-Gaussian PDF. 
392: At larger scales the contribution of the structureless 
393: incoherent background among vortices dominates. 
394: This incoherency leads to Gaussian PDF. 
395: The Gaussian PDF is a parabola in the semi-logarithmic plane 
396: as Fig.\ \ref{fig9} and Fig.\ \ref{fig10}. 
397: 
398: Although our data in Fig.\ \ref{fig9} 
399: are not enough to discuss the scaling between $q$ and $r$ quantitatively, 
400: we have to derive such a relation in our future study. 
401: Already a scaling relation between $q$ and $r$ 
402: has been proposed\cite{Beck} 
403: under the assumption of a cascade picture. 
404: On the other hand, another scheme 
405: with a $r$-independent $q$-value 
406: has been proposed\cite{AA} 
407: under the assumption of another cascade picture. 
408: In contrast to these two approaches 
409: our analysis has nothing to do with cascades. 
410: 
411: In Fig.\ \ref{fig11} 
412: we show the PDF for the vorticity itself 
413: taken from a snapshot of the vorticity field. 
414: This PDF is also non-Gaussian. 
415: The Gaussian components can be removed from the snapshot 
416: by using the wavelet denoising technique\cite{FSK}. 
417: It can be seen from the coincidence 
418: between the left and right panels in Fig.\ \ref{fig12} 
419: that the non-Gaussian components corresponds to the vortices. 
420: Here the Gaussian component is filtered out 
421: by the wavelet denoising in the right panel. 
422: This visualization is not new, 
423: since it has already been done 
424: in the study of the coherent vortex simulation\cite{FSK}. 
425: Our new contribution is to clarify the origin of 
426: the scale-dependent $q$-value 
427: in terms of the coherent vortex structure. 
428: The strong correlation leading to the non-Gaussian PDF 
429: at smaller scales is sustained in the inner region of the vortices, 
430: while the PDF at larger scales than the vortex size tends to Gaussian. 
431: In other words the inner region is coherent 
432: and the outer is incoherent. 
433: 
434: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
435: 
436: \section{Fluctuations: multifractal nature}
437: \label{multifractal-nature}
438: 
439: We have discussed the local nature of the vortex 
440: as the elemntary excitation in the preceding section. 
441: In this section we discuss the spatial distribution 
442: of the vortices in a global point of view. 
443: 
444: First we make a qualitative observation. 
445: Already as seen in Fig.\ \ref{fig12} 
446: the spatial distribution of the coherent vortices is intermittent. 
447: In Fig.\ \ref{fig13} 
448: we also see a intermittent behavior 
449: for the derivative of the vorticity. 
450: Similar intermittent behavior is observed 
451: in experimental data\cite{MSKF}. 
452: 
453: Next we quantify the intermittency 
454: by introducing the measure, $\mu_{k}(l)$, defined as the probability 
455: of finding a coherent vortex in the $k$-th cubic box 
456: of the linear dimension $l$ in a snapshot of the vorticity field. 
457: The total number of the boxes $N_L = (L/l)\times(L/l)$ 
458: where $L$ is the linear dimension of the simulation region. 
459: The measure behaves as 
460: %----------------------------------------------------------------
461: \begin{equation}
462: \mu_{k}(l) \propto l^{\alpha_k},  
463: \end{equation}
464: %----------------------------------------------------------------
465: for small $l/L$. 
466: The number density $N(\alpha)$ for the exponent $\alpha_k$ is 
467: defined as 
468: %----------------------------------------------------------------
469: \begin{equation}
470: N(\alpha) \equiv \sum_k \delta(\alpha - \alpha_k) 
471:           \propto l^{-f(\alpha)}.  
472: \end{equation}
473: %----------------------------------------------------------------
474: This relation defines the singularity spectrum, 
475: $f(\alpha)$, which is the key quantity to qualify multifractal systems. 
476: Using the normalized $q$-th moment $\mu_{k}(q,l)$ 
477: of the measure $\mu_{k}(l)$, 
478: %----------------------------------------------------------------
479: \begin{equation}
480: \mu_{k}(q,l) = \{\mu_{k}(l)\}^{q} 
481:              / \sum_{k'=1}^{N_L} \{\mu_{k'}(l)\}^{q}, 
482: \end{equation}
483: %----------------------------------------------------------------
484: the sigularity spectrum $f(\alpha)$ is obtained 
485: by the following formulae,\cite{Schreiber} 
486: %----------------------------------------------------------------
487: \begin{equation}
488: \alpha(q) = \sum_{k=1}^{N_L} \mu_{k}(q,l) \ln \mu_{k}(l) / \ln (l/L),  
489: \end{equation}
490: %----------------------------------------------------------------
491: and 
492: %----------------------------------------------------------------
493: \begin{equation}
494: f(\alpha(q)) = \sum_{k=1}^{N_L} \mu_{k}(q,l) \ln \mu_{k}(q,l) / \ln (l/L).
495: \end{equation}
496: %----------------------------------------------------------------
497: 
498: In Fig.\ \ref{fig14} 
499: we show the singularity spectrum 
500: for the coherent vortex. 
501: For simplicity we have used the definition $N_k/N_L = l^{\alpha_k}$ 
502: where $N_k$ is the number of lattice points 
503: satisfying the condition $\omega_z > \langle \omega_z \rangle$. 
504: The spectrum is similar to the typical one 
505: measured for the energy-dissipation rate 
506: by experiments\cite{MSKF} or simulations\cite{HY}. 
507: 
508: In the last part of this section 
509: we give some speculations. 
510: As seen in our previous study\cite{SN2} 
511: the width of the singularity spectrum 
512: depends on the degree of intermittency 
513: so that we expect some Reynolds-number dependence 
514: of the spectrum. 
515: It has been discussed by many authors\cite{Sreenivasan} 
516: and should be claryfied in future systematic study. 
517: In our present study 
518: the density of the vortex is dilute 
519: as seen in Fig.\ \ref{fig12}. 
520: Thus the singularity spectrum in Fig.\ \ref{fig14} 
521: describes the spatial distribution of relatively free vortices. 
522: As the Reynolds number is increased 
523: the density increases\cite{KM}. 
524: In this case 
525: the interaction among vortices becomes important and  
526: the correlation length of the vorticity fluctuation becomes large. 
527: While we can observe the vortex 
528: only as an individual elementary excitation 
529: in our numerical experiment, 
530: some collective excitation is expected to dominate 
531: at higher Reynolds number. 
532: In the limit of divergently large Reynolds number 
533: the correlation length becomes divergently large 
534: so that we can expect full scale-invariance. 
535: We can find a resemblance to the case of the scaling theory\cite{deGennes} 
536: in polymers where an ideal scaling relation is realized for dense solutions 
537: where polymers are strongly entangled. 
538: In the limit of high Reynolds number 
539: each boxes counting the coherent vortex 
540: will be filled by almost equal number of vortices 
541: so that intermittency will disappear and 
542: unifractal Kolmogorov's scaling will prevail. 
543: 
544: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
545: 
546: \section{Summary}
547: \label{summary}
548: 
549: Numerical simulation data, in real space and time, 
550: for a forced turbulence 
551: on the basis of the lattice Boltzmann method 
552: have been analyzed by unifractal and multifractal schemes. 
553: 
554: Our new findings are summarized into two points. 
555: First in the unifractal analysis using the exit-time statistics 
556: we have verified Kolmogorov's scaling and Taylor's hypothesis 
557: at the same time. 
558: Second in the analysis 
559: using the R\'enyi-Tsallis PDF and the wavelet denoising 
560: we have clarified that the coherent vortices sustain 
561: the power-law velocity correlation in the non-equilibrium state. 
562: 
563: Finally in the multifractal analysis 
564: it is clarified 
565: that the intermittent distribution 
566: of the coherent vortices in space-time 
567: is described as a multifractal. 
568: 
569: This work was supported in part 
570: by a Grand-in-Aid for Scientific Research 
571: from the Ministry of Education, Culture, Sports, Science 
572: and Technology of Japan. 
573: 
574: %----------------------------------------------------------------
575: \begin{references}
576: 
577: \bibitem{Frisch}
578: U. Frisch, 
579: Turbulence, 
580: Cambridge University Press, Cambridge, 1995.
581: 
582: \bibitem{exit-time-stat}
583: M. Abel, L. Biferale, M. Cencini, M. Falcioni, D. Vergni, 
584: A. Vulpiani, 
585: Physica D 147 (2000) 12.
586: 
587: \bibitem{PDF-space}
588: F.M. Ramos, R.R. Rosa, C.R. Neto, M.J.A. Bolzan, 
589: L.D.A. S\'a, H.F.C. Velho, 
590: Physica A 295 (2001) 250.
591: 
592: \bibitem{RT1}
593: C. Beck, F. Schl\"ogl, 
594: Thermodynamics of chaotic systems, 
595: Cambridge University Press, Cambridge, 1993.
596: 
597: \bibitem{RT2}
598: C. Tsallis, 
599: Nonextensive Statistical Mechanics and Its Applications 
600: (eds. S. Abe, Y. Okamoto), 
601: Springer-Verlag, Berlin, 2001, p.3. 
602: 
603: \bibitem{SN1}
604: W. Sakikawa, O. Narikiyo,
605: J. Phys. Soc. Jpn. 71 (2002) 1200.
606: 
607: \bibitem{SN2}
608: W. Sakikawa, O. Narikiyo,
609: J. Phys. Soc. Jpn. 72 (2003) 450.
610: 
611: \bibitem{LB2}
612: D.A. Wolf-Gladrow, 
613: Lattice-Gas Cellular Automata and Lattice Boltzmann Models, 
614: Springer-Verlag, Berlin, 2000.
615: 
616: \bibitem{LB1}
617: S. Succi, 
618: The Lattice Boltzmann Equation for Fluid Dynamics and Beyond, 
619: Clarendon Press, Oxford, 2001.
620: 
621: \bibitem{SN3}
622: The first simulation data for a smaller system ($100^3$) 
623: to confirm Kolmogorov's scaling 
624: has been reported by us in an unpublished preprint, 
625: cond-mat/0208094.
626: 
627: \bibitem{JPV}
628: M.H. Jensen, G. Paladin, A. Vulpiani, 
629: Phys. Rev. A 45 (1992) 7214.
630: 
631: \bibitem{AA}
632: T. Arimitsu, N. Arimitsu, 
633: Physica A 295 (2001) 177.
634: 
635: \bibitem{Beck}
636: C. Beck, 
637: Physica A 277 (2000) 115.
638: 
639: \bibitem{FSK}
640: M. Farge, K. Schneider, N. Kevlahan, 
641: Phys. Fluids 11 (1999) 2187.
642: 
643: \bibitem{MSKF}
644: C. Meneveau, K.R. Sreenivasan, P. Kailasnath, M.S. Fan, 
645: Phys. Rev. A 41 (1990) 894.
646: 
647: \bibitem{Schreiber}
648: M. Schreiber, 
649: Computational Physics, 
650: (eds. K.H. Hoffmann and M. Schreiber), 
651: Springer-Verlag, Berlin, 1996, p.147.
652: 
653: \bibitem{HY}
654: I. Hosokawa, K. Yamamoto, 
655: J. Phys. Soc. Jpn. 59 (1990) 401.
656: 
657: \bibitem{Sreenivasan}
658: K.R. Sreenivasan, 
659: More Things in Heaven and Earth: 
660: A Celebration of Physics at the Millenium, 
661: (ed. B. Bederson), 
662: Springer-Verlag, New York, 1999, p.644.
663: 
664: \bibitem{KM}
665: S. Kida, H. Miura, 
666: J. Phys. Soc. Jpn. 67 (1998) 2166.
667: 
668: \bibitem{deGennes}
669: P-G. de Gennes, 
670: Scaling Concepts in Polymer Physics, 
671: Cornell University Press, Ithaca, 1979.
672: 
673: \end{references}
674: 
675: \vskip 15pt
676: %----------------------------------------------------------------
677: \begin{figure}
678: \caption{
679: The snapshot of the $z$-component of the vorticity, $\omega_z$, 
680: in the $xy$-plane at $z=100$ and $t=4000$. 
681: }
682: \label{fig1}
683: \end{figure}
684: 
685: \vskip -15pt
686: 
687: \begin{figure}
688: \caption{
689: The Reynolds number $Re$ as a function of time $t$. 
690: }
691: \label{fig2}
692: \end{figure}
693: 
694: \vskip -15pt
695: 
696: \begin{figure}
697: \caption{
698: The $y$-component of the fluid velocity $u_y$ 
699: as a function of time $t$ at ${\vec r}=(100,100,100)$. 
700: }
701: \label{fig3}
702: \end{figure}
703: 
704: \vskip -15pt
705: 
706: \begin{figure}
707: \caption{
708: The inverse of the mean exit-time $\tau(\epsilon)$. 
709: The exits are measured at $200^3$ lattice points for $1000 \le t \le 8000$. 
710: Here the dots and straight line represent the simulation data 
711: and $\epsilon^{-3}$ law, respectively. 
712: }
713: \label{fig4}
714: \end{figure}
715: 
716: \vskip -15pt
717: 
718: \begin{figure}
719: \caption{
720: The inverse of the mean exit-length $r(\epsilon)$. 
721: The exits are measured at $200^3$ lattice points at $t=6000$. 
722: Here the dots and straight line represent the simulation data 
723: and $\epsilon^{-3}$ law, respectively. 
724: }
725: \label{fig5}
726: \end{figure}
727: 
728: \vskip -15pt
729: 
730: \begin{figure}
731: \caption{
732: The space correlation of the velocity fluctuation at $t=4000$. 
733: The horizontal and vertical axes represent $u_y(x,y,z,t)$ and 
734: $u_y(x,y+y_{\rm corr},z,t)$, respectively. 
735: The space difference $y_{\rm corr}$ is taken as 
736: $y_{\rm corr}=1$ for the left, 
737: $y_{\rm corr}=4$ for the center and 
738: $y_{\rm corr}=30$ for the right. 
739: }
740: \label{fig6}
741: \end{figure}
742: 
743: \vskip -15pt
744: 
745: \begin{figure}
746: \caption{
747: The time correlation of the velocity fluctuation 
748: at $\vec r =(100,100,100)$. 
749: The horizontal and vertical axes represent $u_y(\vec r,t)$ and 
750: $u_y(\vec r,t+t_{\rm corr})$, respectively for $1000 \le t \le 8000$. 
751: The time difference $t_{\rm corr}$ is taken as 
752: $t_{\rm corr}=10$ for the left, 
753: $t_{\rm corr}=100$ for the center and 
754: $t_{\rm corr}=500$ for the right. 
755: }
756: \label{fig7}
757: \end{figure}
758: 
759: \vskip -15pt
760: 
761: \begin{figure}
762: \caption{
763: The correlation map for the velocity difference between two times 
764: $t_i$ and $t_j$ 
765: for $1000 \le t_0+t_i \le 8000$ and $1000 \le t_0+t_j \le 8000$ 
766: with $t_0=1000$. Namely $0 \le i \le 7000$ (horizontal axis) and 
767: $0 \le j \le 7000$ (vertical axis). 
768: The dot is marked at $(i,j)$ if $D(i,j)<\theta$ 
769: where $D(i,j)=|u_y(\vec r,t_i)-u_y(\vec r,t_j)|$ 
770: and $\theta=0.2D_{\rm max}$ 
771: with $D_{\rm max}$ being the maximum of $D(i,j)$. 
772: The position $\vec r$ is fixed at $(100,100,50)$. 
773: }
774: \label{fig8}
775: \end{figure}
776: 
777: \vskip -15pt
778: 
779: \begin{figure}
780: \caption{
781: The unnormalized PDF $P(X)$ for the velocity difference 
782: $X=|u_y(x,y+\Delta y,z,t)-u_y(x,y,z,t)|$ 
783: between two points $(x,y+\Delta y,z)$ and $(x,y,z)$ at $t=4000$. 
784: The difference in $y$-direction is chosen as $\Delta y = 1, 4, 30$. 
785: The data are fitted by the R\'enyi-Tsallis PDF with 
786: $q = 1.218, 1.164, 1.100$ for $\Delta y = 1, 4, 30$, respectively. 
787: }
788: \label{fig9}
789: \end{figure}
790: 
791: \vskip -15pt
792: 
793: \begin{figure}
794: \caption{
795: The unnormalized PDF $P(X)$ for the vorticity difference 
796: $X=|\omega_y(x,y+\Delta y,z,t)-\omega_y(x,y,z,t)|$ 
797: between two points $(x,y+\Delta y,z)$ and $(x,y,z)$ at $t=6000$. 
798: The difference in $y$-direction is chosen as $\Delta y = 1, 4, 30$. 
799: The data are fitted by the R\'enyi-Tsallis PDF with 
800: $q = 1.240, 1.209, 1.179$ for $\Delta y = 1, 4, 30$, respectively. 
801: }
802: \label{fig10}
803: \end{figure}
804: 
805: \vskip -15pt
806: 
807: \begin{figure}
808: \caption{
809: The unnormalized PDF $P(X)$ for the vorticity 
810: $X=|\omega_y(x,y,z,t)|$ at $t=6000$. 
811: The data are fitted by the R\'enyi-Tsallis PDF with $q = 1.392$. 
812: }
813: \label{fig11}
814: \end{figure}
815: 
816: \vskip -15pt
817: 
818: \begin{figure}
819: \caption{
820: The snapshot of the $z$-component of the vorticity, $\omega_z$, 
821: in the $xy$-plane at $z=100$ and $t=4000$. 
822: In contrast to Fig.\ \ref{fig1} 
823: the regions satisfying the condition, 
824: $\omega_z > \langle \omega_z \rangle + 2\sigma \equiv \Theta$, 
825: are depicted in the left panel. 
826: Here $\langle \omega_z \rangle$ is the spatial average 
827: and $\sigma$ is the standard deviation. 
828: In the right panel 
829: the high vorticity regions satisfying the same condition, 
830: $\tilde \omega_z > \Theta$, are depicted 
831: where $\tilde \omega_z$ is the non-Gaussian component 
832: after the wavelet denoising filtering out the Gaussian component. 
833: By the restriction of the algorithm of the wavelet transform 
834: we have used only 128$\times$128 lattice points. 
835: }
836: \label{fig12}
837: \end{figure}
838: 
839: \vskip -15pt
840: 
841: \begin{figure}
842: \caption{
843: The spatial distribution of the vorticity gradient, 
844: $\Delta\omega_z = \omega_z(x+1,y,z,t) - \omega_z(x,y,z,t)$, 
845: in $x$-direction ($1 \le x \le 200$) with $y=z=100$ at $t=6000$. 
846: }
847: \label{fig13}
848: \end{figure}
849: 
850: \vskip -15pt
851: 
852: \begin{figure}
853: \caption{
854: The singularity spectrum of the coherent vortex with $l/L = 1/10$. 
855: The data correspond to $-90 \le q \le 95$. 
856: }
857: \label{fig14}
858: \end{figure}
859: %----------------------------------------------------------------
860: \end{document}
861: %----------------------------------------------------------------
862: \bye
863: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
864: