1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: \documentclass[12pt]{article}

3: \usepackage{psfig,epsfig}

4: \textheight=21.5cm

5: \textwidth=15.5cm

6: \topskip -3cm

7: \leftskip -2.5cm

8: \def\baselinestretch{1.6}

9: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

10: \def\bs{$\backslash$} \def\bslp{\baselineskip}    \def\d{{\rm d}}

11: \def\spg{\setcounter{page}} \def\seq{\setcounter{equation}}

12: \def\bd{\begin{document}} \def\ed{\end{document}} \def\r#1{$^{[#1]}$}

13: \def\bmp{\begin{minipage}} \def\emp{\end{minipage}}

14: \def\bcc{\begin{center}} \def\ecc{\end{center}}     \def\npg{\newpage}

15: \def\beq{\begin{equation}} \def\eeq{\end{equation}} \def\hph{\hphantom}

16: \def\beqr{\begin{\eqnarray}} \def\eeqr{\end{\eqnarray}}

17: \def\n{\noindent} \def\ni{\noindent} \def\pa{\parindent}

18: \def\hs{\hskip} \def\vs{\vskip} \def\hf{\hfill} \def\ej{\vfill\eject}

19: \def\cl{\centerline} \def\ob{\obeylines}  \def\ls{\leftskip}

20: \def\underbar#1{$\setbox0=\hbox{#1} \dp0=1.5pt \mathsurround=0pt

21:    \underline{\box0}$}   \def\ub{\underbar}    \def\ul{\underline}

22: \def\f{\left} \def\g{\right} \def\e{{\rm e}} \def\o{\over}

23: \def\vf{\varphi} \def\pl{\partial} \def\cov{{\rm cov}} \def\ch{{\rm ch}}

24: \def\la{\langle} \def\ra{\rangle} \def\EE{e$^+$e$^-$}

25: \def\bitz{\begin{itemize}} \def\eitz{\end{itemize}}

26: \def\btbl{\begin{tabular}} \def\etbl{\end{tabular}}

27: \def\btbb{\begin{tabbing}} \def\etbb{\end{tabbing}}

28: \def\beqar{\begin{eqnarray}} \def\eeqar{\end{eqnarray}}

29: \def\\{\hfill\break} \def\dit{\item{-}} \def\i{\item}

30: \def\bbb{\begin{thebibliography}{9} \itemsep=-1mm}

31: \def\ebb{\end{thebibliography}} \def\bb{\bibitem}

32: \def\bpic{\begin{picture}(260,240)} \def\epic{\end{picture}}

33: \def\akgt{\noindent{\bf Acknowledgements}}

34: \def\fgn{\noindent{\bf\large\bf Figure captions}}

35: \def\eps{\varepsilon}

36: 

37: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

38: \begin{document}

39: \begin{center}

40: 

41: {\Large

42: Two Specific Correlation Patterns as Indicators

43: 

44: for Various Random Multiplicative Cascade Processes }

45: 

46: \vskip 1cm

47: 

48: Wu Yuanfang\ \ \ Wang Yingdan\ \ \  Bai Yuting\ \ \ Liao Hongbo\ \ \ Liu Lianshou

49: 

50: \vskip 0.5cm

51: 

52: {\small \  Institute of Particle Physics, Huazhong Normal University, Wuhan 430079 China}

53: 

54: {\small FAX: 0086 27 87662646

55: \qquad email: wuyf@iopp.ccnu.edu.cn}

56: \date{ }

57: \end{center}

58: 

59: \begin{center}

60: \begin{minipage}{125mm}

61: 

62: \vskip 2cm

63: \begin{center}{\Large Abstract}\end{center}

64: 

65: 

66: \ \ \ \ It is suggested and demonstrated that

67: two specific 2-dimensional correlation patterns, fixed-to-arbitrary bin and

68: neighboring bin correlation patterns,

69: are efficient for identifying various random multiplicative cascade

70: processes. A possible application of these two correlation patterns to single event analysis in

71: Relativistic Heavy Ion Collider experiments is discussed.

72: 

73: \end{minipage}

74: 

75: \vs 1.5cm

76: 

77: {\large PACS number: 47.27.Eq, 05.40.+j, 47.54.+r, 25.75.Gz}

78: 

79: \end{center}

80: \npg

81: 

82: It is well-known that nonlinear phenomena are very popular in 

83: nature. Once a set of anomalous scaling law of probability- or

84: ``mass''-moments with respect to spatial size~\cite{multifractal}

85: is observed in experiment, we recognize that there is self-similar

86: multifractal structure. However, it is difficult to judge what kind

87: of mechanism most likely causes the results since the 

88: conventional measures have not provided sufficient information on 

89: underlying dynamics. This hinders us from further understanding and 

90: studying nonlinear physics. A useful measure for this purpose should 

91: not contain too much detail, but enough to capture the essence of

92: production mechanism. In this letter, as a first try along this

93: direction, we will suggest two specific correlation patterns which

94: manifest distinguishable characters of various random multiplicative 

95: cascade processes. They will be helpful in exploring the origin of 

96: nonlinear and correlation related phenomena.

97: 

98: A random multiplicative cascade process is the simplest example of

99: having a well-defined multifractal structure and has been used

100: extensively in various fields. For example, energy dissipation in

101: fully developed turbulence~\cite{multifractal,fluid} and successive

102: branchings in QCD parton-shower~\cite{QCD} are both this kind of

103: process and accordingly have multifractal structure. The main idea

104: of such a process is simply a series of self-similar random

105: cascades in spatial partitions. It is generally described as

106: follows: at the first step of the cascade, a given initial

107: interval with length $\Delta$ and unit probability $p_0=1$ is

108: split into $\tau$ subdivisions (``bins''). Usually, $\tau$ is an

109: integer with $\tau \geq 2$. The probability for a bin to be split

110: is determined by a weight $w$ which is distributed according to a

111: splitting function $p(w_1, w_2, \dots, w_\tau)$. Then each bin

112: obtained from the first step of the cascade is again independently

113: split into $\tau$ bins and so on. After $\nu=1, 2, ..., J$ steps,

114: the final number of bins is $\tau^{\nu}$, and the probability in a

115: specific bin $j_\nu$ is $\nu$ products of $w$ over all previous

116: generations: $p_{j_\nu}^\nu=w_{1j_1}w_{2j_2}\dots w_{\nu j_\nu}$,

117: where $j_\mu=1, 2, ..., \tau^\mu\ (\mu=1, 2, ..., \nu)$. The

118: distribution of $p_{j_\nu}^\nu$ obtained in this way fluctuates

119: violently from bin to bin, {\it cf.} Fig. 6

120: in~\cite{multifractal}.

121: 

122: By varying the splitting number and/or changing the splitting

123: function, all kinds of random multiplicative cascade processes can

124: be constructed. Especially, an additional physical constraint on

125: the splitting function, like probability conservation at each step

126: of splitting, will cause stronger correlations among the $\tau$

127: split bins than those without it. The smaller the splitting number

128: $\tau$ is, the stronger are the correlations among the ${\tau}$

129: bins. It is therefore necessary to investigate $n$-point

130: correlations in order to probe the differences among the various

131: random cascade processes. Let us start from the simplest 2-point

132: ones.

133: 

134: Why are conventional measurements for 2-point-correlation

135: moments~\cite{scaling} insufficient? This is because they are

136: usually defined at fixed correlation length with both horizontal

137: average over all bins in an event and longitudinal average over

138: all events in a sample. The fixed correlation length makes a

139: comparison between correlations with different scales impossible,

140: and the horizontal average smooths out the differences between

141: different pairs of bins. For these measurements, various random

142: multiplicative processes have the same kind of power law behavior

143: with diminishing correlation length, similar to the scale

144: invariance of ``mass'' moments. They have little help in the

145: identification of production mechanism. Hence, 2-point

146: correlation moments without horizontal averaging, but with various

147: correlation lengths, are necessary to be presented simultaneously.

148: It should be a {\it pattern} constructed by all kinds of 2-point

149: correlation moments.

150: 

151: A straightforward pattern of such a kind is the three dimensional

152: pattern of $2$-point correlation cumulants~\cite{Martin}, \beq

153: C_{K_1,K_2}=\langle \ln p_{K_1}\ln p_{K_2} \rangle_c = \langle \ln

154: p_{K_1}\ln p_{K_2} \rangle - \langle \ln p_{K_1} \rangle \langle

155: \ln p_{K_2} \rangle, \eeq \noindent where $K_1$ and $K_2$ are the

156: coordinates of two points. However, in this pattern the

157: differences among various random multiplicative cascade processes

158: are hidden in complicated background and hard to be observed, {\it

159: cf.} Fig.1, where the patterns for three different models, the

160: $\alpha$ model without probability conservation~\cite{scaling}, the

161: $p$ model with probability conservation~\cite{p} and the $c$ model

162: with both probability conservation and random probability

163: partition at each step of the splitting~\cite{c}, are presented.

164: The pattern manifests the common characteristics of random

165: multiplicative cascade processes, such as symmetry at each step of

166: splitting, the largest correlations of any bin with itself and so

167: on, but is insensitive to the basic model assumptions. Thus, it is

168: better to present the correlations of some specific two bins in a

169: two dimensional pattern so that the differences in various

170: processes are manifested.

171: 

172: A simple way to do so is to fix one bin $K_1$ and vary the left

173: one $K_2$. It is a fixed-to-arbitrary bin correlation pattern. It

174: contains correlation information at various lengths from the fixed

175: bin to all other bins, in which $\tau-1$ bins are separated from

176: the fixed one at the last cascade step and $\tau^J-\tau$ bins are

177: separated successively from the fixed one before the last step of

178: cascading. Another simple way is to fix the distance between the two

179: bins and change the position of them. Since the correlations among

180: $\tau$ bins at each step of splitting are important information,

181: the distance between two bins are  better to be as close to each

182: other as possible, {\it i.e.} $K_1=K$ and $K_2=K+1$, where $K=1,

183: 2, ..., \tau^J-1$. This is a neighboring bin correlation pattern.

184: These two correlation patterns provide useful information on

185: cascade mechanism and filter out the common characteristics of

186: random multiplicative cascade processes.

187: 

188: In the following, as a support of above arguments, the

189: correlation cumulants for three different random cascade

190: processes, the $\alpha$, the $p$ and the $c$ models, are first

191: derived in order to show how the different appearances of these

192: cumulants are caused by the model assumptions. Then it is argued

193: and demonstrated why the measurements for these two patterns

194: provide information on the model assumptions. Finally, the

195: extension and application of the patterns are discussed.

196: 

197: These three models are examples of the random binary cascade

198: processes with symmetric splitting functions

199: $p(w_1,w_2)=p(w_2,w_1)$. In this case, Eq.(1) can be written

200: simply as~\cite{Martin} \beq

201: C_{K_1,K_2}=A(J-d_2)+B(1-\delta_{d_2,0}), \eeq where

202: $d_2=J-\sum_{j=1}^J\delta_{k_1^1,k_1^2}\cdots\delta_{k_j^1,k_j^2}$

203: is the ultrametric distance, which is a measure of how many

204: generations one has to move up before a common ancestor is found;

205: \beq

206: A={\partial^2Q[\lambda_1,\lambda_2]}/{\partial\lambda_1^2}|_{\lambda=0},

207: \qquad

208: B={\partial^2Q[\lambda_1,\lambda_2]}/{\partial\lambda_1\partial\lambda_2}|_{\lambda=0}

209: \eeq are respectively the so called ``same-lineage'' cumulant,

210: which is the correlation of the common ancestor of $K_1$ and

211: $K_2$, and the ``splitting'' cumulant, which measures the

212: correlation between the two parts split first from their common

213: ancestor; \beq Q[\lambda_1,\lambda_2]=\ln \left[ \int

214: dw_{1}dw_{2}p(w_1,w_2)\exp(\lambda_{1}

215: \ln{w_{1}}+\lambda_{2}\ln{w_{2}})\right] \eeq is the branching

216: generating function (BGF).

217: 

218: For the $\alpha$ model, the $w_1$ or $w_2$ at each step of

219: splitting is two possible numbers, $1+\alpha$ and $1-\alpha$, with

220: equal probability for each of them. Hence, its splitting

221: function~\cite{Martin} can be written as: \beq p(w_1,w_2)=\frac

222: {1}{4}\biggl[\delta\biggl(w_1-(1+\alpha)\biggr)+\delta\biggl(w_1-(1-\alpha)\biggr)\biggr]

223: \biggl[\delta\biggl(w_2-(1+\alpha)\biggr)+\delta\biggl(w_2-(1-\alpha)\biggr)\biggr],

224: \eeq \noindent where $\alpha$ is a fixed parameter in the region

225: [-1, 1]. Inserting this function into Eq.(4) and then Eq.(4)

226: into Eq.(3),  the ``same-lineage'' and the ``splitting'' cumulants

227: can be easily derived: \beq A_\alpha=\frac{1}{4}\biggl[\ln

228: \frac{1+\alpha}{1-\alpha}\biggr]^2, \ \ \  B_\alpha=0. \eeq

229: \noindent $B_\alpha=0$ means that there is no correlation between

230: the two parts for every splitting. This is due to the independence

231: of $w_1$ and $w_2$ assumed in the model.

232: 

233: For the $p$ model, the summation of $w_1$ and $w_2$ keeps to be

234: $2$, with $w_1=1\pm\beta$ and $w_2= 1\mp\beta$ as the consequence

235: of probability conservation, here $\beta$ is a fixed parameter in

236: the region [-1, 1]. Its splitting function~\cite{Martin} is \beq

237: p(w_1,w_2)=\frac

238: {1}{2}\biggl[\delta\biggl(w_1-(1+\beta)\biggr)+\delta\biggl(w_1-(1-\beta)\biggr)\biggr]

239: \delta(w_1+w_2-2), \eeq \noindent and its ``same-lineage'' and

240: ``splitting'' cumulants are \beq A_p=\frac{1}{4}\biggl[\ln

241: \frac{1+\beta}{1-\beta}\biggr]^2, \ \ \ B_p=-\frac{1}{4}\biggl[\ln

242: \frac{1+\beta}{1-\beta}\biggr]^2. \eeq \noindent Due to

243: probability conservation, the ``splitting'' cumulant $B$ has sign

244: opposite to the ``same-lineage'' cumulant $A$. Moreover, since

245: $\beta$ is a unique fixed number, the derivatives of BGF with

246: respect to one of the two $\lambda$'s twice are equivalent to

247: those with respect to both of them. This is why the absolute value

248: of the ``same-lineage'' cumulant $A$ is equal to that of the

249: ``splitting'' cumulant $B$.

250: 

251: For the third $c$ model,  $w_1+w_2=1$ similar to the $p$ model.

252: Distinct from the $p$ and the $\alpha$ models, the weight $w$ is

253: allowed to be a random number {\it e.g.} $ w_1={(1+\gamma r)}/{2}$

254: and $w_2={(1-\gamma r)}/{2}$, here $r$ is a random number

255: uniformly distributed in the interval $[-1,1]$ and $\gamma$ is a

256: fixed model parameter in the region $[0,1]$.  Obviously, this

257: model is more flexible and closer to the real system with

258: probability conservation. The splitting function is: \beq

259: p(w_1,w_2)=\frac{1}{\gamma}\biggl[\theta\biggl(w_1-\frac{1-\gamma}{2}\biggr)-

260: \theta\biggl(w_1-\frac{1+\gamma}{2}\biggr)\biggr]\delta

261: (w_1+w_2-1). \eeq and its ``same-lineage'' and ``splitting''

262: cumulants are: \beq A_c=\frac{1}{\gamma}

263: \int_{\frac{1-\gamma}{2}}^{\frac{1+\gamma}{2}} (\ln

264: w_1)^2dw_1-\frac{1}{\gamma^2}

265: \biggl[\int_{\frac{1-\gamma}{2}}^{\frac{1+\gamma}{2}}\ln

266: w_1dw_1\biggr]^2, \nonumber \eeq \beq B_c=\frac{1}{\gamma}

267: \int_{\frac{1-\gamma}{2}}^{\frac{1+\gamma}{2}} \ln w_1 \ln

268: (1-w_1)dw_1-\frac{1}{\gamma^2}\biggl[\int_{\frac{1-\gamma}{2}}^{\frac{1+\gamma}{2}}\ln

269: w_1dw_1\biggr]^2. \eeq The ``splitting'' cumulant is nonzero due to

270: probability conservation. The absolute values of the

271: ``same-lineage'' and the ``splitting'' cumulants are unequal since

272: $w_1$ and $w_2$ are no longer fixed to a particular number, unlike

273: the $p$ model.

274: 

275: From the above analytic derivation, we see that the three random

276: cascade models are very different in the absolute values of their

277: ``splitting'' cumulants and in the relative values of these

278: cumulants to the ``same lineage'' cumulants. These differences 

279: are due to the basic model assumption and 

280: independent of the model parameters.

281: Therefore, a useful 2-point correlation pattern should contain

282: information on the ``splitting'' cumulant and its relation to the

283: ``same lineage'' cumulant. This requirement turns out to be two

284: possible combinations of the coefficients $A$ and $B$ in Eq.(2):

285: (1) $C_{K_1,K_2}=B$, {\it i.e.} $J-d_2=0$ and

286: $1-\delta_{d_2,0}=1$; (2) $C_{K_1,K_2}=A+B$, {\it i.e.} $J-d_2=1$

287: and $1-\delta_{d_2,0}=1$; for the first combination, $d_2=J$ and

288: $d_2\not=0$ means that common ancestor has to be at the initial

289: interval if $J > 0$. The fixed-to-arbitrary bin correlation

290: pattern happens to have $2^{J-1} \cdot (2-1)$ pairs of this kind

291: of bins. The neighboring bin correlation pattern also has a pair

292: of this kind of bins at center $K=2^{J-1}$. For the second

293: combination, $d_2=J-1$ and $d_2\not=0$ implies that common

294: ancestor locates at the first cascade step. In the

295: fixed-to-arbitrary bin correlation pattern, $2^{J-2} \cdot (2-1)$

296: pairs of bins have common ancestor at the first cascade step. The

297: neighboring bin correlation pattern again has two pairs of such a

298: kind of bins at $K=2^{J-2}$ and  $K=2^{J-2}\cdot 3$ respectively.

299: Hence the suggested two correlation patterns are useful measures

300: in identifing the above mentioned three random cascade processes.

301: 

302: We first plot the fixed-to-arbitrary bin correlation pattern for

303: the $\alpha$, the $p$ and the $c$ models in Figs. 2(a), (b) and

304: (c), respectively, with the model parameters $\alpha=0.4$,

305: $\beta=0.4$, $\gamma=0.8$, $J=6$. The ordinate is the correlation

306: strength $C_{1,K_2}$; the abscissa is the bin $K_2$ which varies

307: as $K_2=2, 3, ..., 64$, and the fixed bin is located at $K_1=1$.

308: Since the larger the distance between the two bins is, the earlier

309: the two bins are separated in cascade process, the correlation

310: strength decreases as the arbitrary bin $K_2$ goes further away

311: from the fixed one $K_1$. The number of points with the same

312: height or same correlation strength increases as  $2^{\nu-1} \cdot

313: (2-1)$ with $\nu=1, 2, ..., J$. The number $2$ comes from the

314: binary cascade. Generally, it is $\tau^{\nu-1}\cdot (\tau-1)$.

315: So by counting the number of points in each

316: correlation strength, the splitting number of the cascade process

317: can be estimated.

318: 

319: The possible correlation strengths in these three models are

320: different: (1) the weakest one in the $\alpha$ model is zero; (2)

321: there are both zero and negative ones in the $p$ model; (3)

322: negative ones, but no zero ones, are allowed in the $c$ model.

323: These differences come from the different model assumptions. The

324: absence of negative correlations implies independent splitting in

325: the model,  $B=0$, such as is the case for the $\alpha$ model. If

326: there are both negative and zero correlations in the pattern, the

327: positive one from the ``same-lineage'' cumulant $A$ must be equal

328: to the negative one from the ``splitting'' cumulant $B$. So that

329: $B < 0$ for the weakest one and $A+B=0$ for the following one, as

330: for the $p$ model.  Having negative, but no zero, correlation,

331: means that the ``same-lineage'' cumulant is different from the

332: ``splitting'' cumulant, or $A+B\not= 0$ and $B < 0$, as for the

333: $c$ model.

334: 

335: The neighboring bin correlation pattern $C_{K,K+1}$ with $K=1, 2,

336: ..., 2^J-1$ can show us the same thing. In Fig.3, the patterns for

337: the three models are presented. Here, the strongest correlations

338: are of those two neighboring bins split at the last step of

339: cascade, {\it i.e.}, $K=1, 3,..., 2^J-2^0$; the next ones happen

340: to be of those split before the last step of cascade, {\it i.e.},

341: $K=2, 6, 10, ..., 2^J-2^1$; and so on. The number of points at a

342: certain correlation strength is determined by $2^{\nu-1}$, or

343: $\tau^{\nu-1}$ in general, with $\nu=J, J-1, ..., 1$. There are

344: only positive and zero correlation spectra in the $\alpha$ model.

345: In the $p$ model, the correlation spectrum ranges over positive,

346: zero and negative values, while for the $c$ model, only positive

347: and negative correlations are allowed with no zero ones.

348: 

349: Therefore, from the measurements of these two correlation

350: patterns, we can find out by what kind of random multiplicative

351: cascade process the observed system are produced. This is a

352: first step toward the revelation of underlying mechanism of

353: multifractal phenomena. The extension of these two patterns to the

354: mixture of different random cascade processes and other mechanisms,

355: which can lead to multifractal structure, is obviously the next

356: step. Their corresponding patterns will not be so regular as those

357: of the above mentioned three processes. For the further

358: identification of the production mechanism in general, other

359: measures may be needed in addition to these two. Nevertheless,

360: having established these two patterns for some known mechanisms,

361: it is interesting to compare the patterns measured in practice

362: with them and judge if there are similar generating mechanism.

363: 

364: A possible application of these two correlation patterns is in the

365: single event analysis of Relativistic Heavy Ion Collider (RHIC)

366: experiments~\cite{rhic}. In these experiments, the probability of

367: final state particles falling into a particular bin in rapidity or

368: pseudo-rapidity can be well estimated by~\cite{wsfs}, $ p_{K_i}

369: \approx {n_{K_i}}/{N}$, where $n_{K_i}$ is the number of final

370: state particles falling into the $K_i$-th bin and $N$ is the total

371: number of particles in the event, which is typically several

372: thousand~\cite{rhicn}. At the current experimental resolution in

373: rapidity, there is almost no empty bin. In this case, it is much

374: better to investigate directly the pattern of a single event

375: without averaging over all events in the sample. The 2-point

376: correlation cumulant for a single event can be defined as: \beq

377: C_{K_1,K_2}^{(\rm e)}=\ln p_{K_1}\ln p_{K_2}-\langle \ln

378: p_{K_1}\rangle \langle \ln p_{K_2}\rangle, \eeq its average over

379: all events in a sample equals to the 2-point correlation cumulant

380: $C_{K_1,K_2}$ of Eq.(1). The whole event sample can then be

381: classified by the measured correlation patterns of single event.

382: Some events which have undergone a phase transition to Quark Gluon

383: Plasma (QGP) will most likely have very special patterns, since

384: during this transition, all kinds of correlations, the long as

385: well as the short ones, will change dramatically.

386: 

387: In this letter, we investigate two specific 2-dimensional

388: correlation patterns: the fixed-to-arbitrary bin and the

389: neighboring bin correlation patterns. The former shows

390: correlations at various lengths, and the latter has short-range

391: correlations at different positions. They are sensitive to the

392: underlying mechanisms and work equally well in

393: identifying various random multiplicative cascade processes. As an

394: example, their application to the analysis of single event of

395: current Relativistic Heavy Ion Collider(RHIC) experiments has been

396: discussed. For more-than-2-point correlations, we can also find

397: some useful correlation patterns in a similar way. The

398: measurements for all of them will hopefully open a new world in

399: the research of nonlinear and correlation related physics.

400: 

401: \vs 3mm

402: 

403: We are grateful for the helps of Dr. Jimmy MacNaughton in the

404: presentation of the paper and the good suggestions of Dr. Nu Xu.

405: This work is supported in part by the SFC under project 90103019;

406: Cross-century Talent Foundation of National Education Committee of

407: China under Grant No.1 [1998].

408: 

409: \newpage

410: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

411: \def\Journal#1#2#3#4{{#1} {\bf #2} #3 (#4)}

412: \def\CSB{\em Chinese Science Bulletin}

413: \def\NCA{\em Nuovo Cimento} \def\NIM{\em Nucl. Instrum. Methods}

414: \def\NIMA{{\em Nucl. Instrum. Methods} A} \def\NPB{{\em Nucl. Phys.} B}

415: \def\PLB{{\em Phys. Lett.}  B} \def\PRL{\em Phys. Rev. Lett.}

416: \def\PRD{{\em Phys. Rev.} D} \def\ZPC{{\em Z. Phys.} C}

417: \def\PRE{{\em Phys. Rev.} E} \def\PRC{{\em Phys. Rev.} C}

418: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

419: \vs 1cm

420: \begin{thebibliography}{99}

421: 

422: 

423: \bibitem{multifractal} G. Paladin and A. Vulpiani, {\em Phys. Reports} {\bf 4} (1987).

424: 

425: \bibitem{fluid} B. Mandelbrot, {\em J. Fluid. Mech.} {\bf 62}, 331(1974); 

426: U. Frisch, P. Sulem and M. Nelkin,  {\em J. Fluid. Mech.} {\bf 87}, 719(1978);

427: S. Lovejoy and D. Schertzer in Proc. Conf. Turbulent Shear Flows 4, eds. L. J. S. Bradbury {\it et al.}

428: (Springer, Berlin, 1984) and references therein.

429: 

430: \bibitem{QCD} R. P. Feynman, Proc. 3rd Workshop on Current Problems in High Energy Particle Theory,

431: Florence, 1979, eds. R. Casalbnoni et al., Johns Hopkins University Press, Baltimore, 1979;

432: G.Veneziano, ibid. p.15; K. Konishi, A. Ukawa and G. Veneziano, {\em Phys. Lett.} B {\bf 78},

433: 243(1978); {\em Nucl. Phys.} B {\bf 157}, 45(1979).

434: 

435: \bibitem{scaling} A. Bia\l as and R. Peschanski, {\em Nucl. Phys.} B {\bf 308}, 857(1988);

436: P. Bo\.zek, M. P\l oszajczak, R. Botet, {\em Phys. Reports} {\bf 252}, 101(1995).

437: 

438: \bibitem{Martin} Martin Greiner, Hans C. Eggers and Peter Lipa, {\em Phys. Rev. Lett.} {\bf 80},

439: 5333(1998); M. Greiner, J. Schmiegel, F. Eickemeyer, P. Lipa and H. C. Eggers, {\em Phys. Rev.} E

440: {\bf 58}, 554(1998).

441: 

442: \bibitem{p} C. Meneveau and K. R. Sreenivasan, {\em Phys. Rev. Lett.} {\bf 59}, 1424(1987).

443: 

444: \bibitem{c} Wu Yuanfang and Liu Lianshou, {\em Phys. Rev. Lett }{\bf 70}, 3197(1993).

445: 

446: \bibitem{rhic} Harris J. W., {\em Nucl. Phys.} A{\bf 566}, 277(1994).

447: 

448: \bibitem{wsfs} Wu Yuanfang and Liu Lianshou, {\em Phys. Rev.} D {\bf 43}, 3077(1991); Fu Jinghua,

449: Wu Yuanfang and Liu Lianshou, {\em Phys. Lett.} B {\bf 472}, 161(2000).

450: 

451: \bibitem{rhicn} B. B. Back {\it et al.,} {\em Phys. Rev. Lett.} {\bf 85}, 3100(2000).

452: 

453: 

454: \end{thebibliography}

455: 

456: 

457: \newpage

458: \ni{\Large\bf Figure Captions}

459: 

460: 

461: \vs 0.5cm {\pa=0pt{\ls=15mm\rightskip0mm \hs-12mm {\bf Fig.1} \

462: The correlation cumulants $C_{K_1,K_2}$ of

463:  ($a$) the $\alpha$ model,

464: ($b$) the $p$ model and the ($c$) $c$ model.

465: \par}}

466: 

467: \vs 0.5cm {\pa=0pt{\ls=15mm\rightskip0mm \hs-12mm {\bf Fig.2} \

468: The fixed-to-arbitrary bin correlation patterns $C_{1,K_2}$ of

469: ($a$) the $\alpha$ model, ($b$) the $p$ model and ($c$) the $c$

470: model, where the ordinate is correlation strength $C_{1,K_2}$ and

471: the abscissa is $K_2=2, 3, \cdots, 64$. The solid lines in the

472: figures indicate zero correlation strength.

473: \par}}

474: 

475: \vs 0.5cm {\pa=0pt{\ls=15mm\rightskip0mm \hs-12mm {\bf Fig.3} \

476: The neighboring bin correlation patterns $C_{K, K+1}$ of ($a$) the

477: $\alpha$ model, ($b$) the $p$ model and ($c$) the $c$ model, where

478: the ordinate is correlation strength  $C_{K, K+1}$ and the

479: abscissa is $K=1,2,\cdots, 63$. The solid lines in the figures

480: indicate zero correlation strength.

481: \par}}

482: 

483: 

484: \newpage

485: \begin{center}

486: \begin{picture}(300,300)

487: \put(-140,-125)

488: {

489: {\epsfig{file=fig1.ps,width=500pt,height=600pt}}

490: }

491: 

492: 

493: \put(-140,-580)

494: {

495: {\epsfig{file=fig2.ps,width=550pt,height=700pt}}

496: }

497: \end{picture}

498: \end{center}

499: 

500: \vs -4.0cm

501: \n{\hskip 9.5cm {\large Fig.1}}

502: \vs 10.cm

503: \n{\hskip 9.5cm {\large Fig.2}}

504: 

505: 

506: \newpage

507: \begin{center}

508: \begin{picture}(300,300)

509: \put(-100,-250)

510: {

511: {\epsfig{file=fig3.ps,width=500pt,height=600pt}}

512: }

513: \end{picture}

514: \end{center}

515: 

516: \vs 5.0cm

517: \n{\hskip 10cm {\large Fig.3}

518: 

519: \end{document}

520: