1: 

2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

3: 

4: 

5: \usepackage{graphicx}% Include figure files

6: \usepackage{dcolumn}% Align table columns on decimal point

7: \usepackage{bm}% bold math

8: 

9: 

10: 

11: \begin{document}

12: 

13: \title{Power Law Distribution of Wealth in a Money-Based

14: Model}

15: \author{Yan-Bo Xie, Bo Hu, Tao Zhou}

16: \author{Bing-Hong Wang}

17:  \email{bhwang@ustc.edu.cn}

18: \affiliation{%

19: Department of Modern Physics and The Nonlinear Science Center,\\

20: University of Science and Technology of China,\\

21: Heifei Anhui, 230026, PR China

22: }%

23: 

24: \date{\today}

25: 

26: \begin{abstract}

27: A money-based model for the power law distribution (PLD) of wealth

28: in an economically interacting population is introduced. The basic

29: feature of our model is concentrating on the capital movements and

30: avoiding the complexity of micro behaviors of individuals. It is

31: proposed as an extension of the Equ\'{\i}luz and Zimmermann's (EZ)

32: model for crowding and information transmission in financial

33: markets. Still, we must emphasize that in EZ model the PLD without

34: exponential correction is obtained only for a particular

35: parameter, while our pattern will give it within a wide range. The

36: Zipf exponent depends on the parameters in a nontrivial way and is

37: exactly calculated in this paper.

38: \end{abstract}

39: 

40: \pacs{89.90.+n, 02.50.Le, 64.60.Cn, 87.10.+e}

41: 

42: \maketitle

43: 

44: \section{\label{sec:level1}Introduction}

45: 

46: 

47: Many real life distributions, including wealth allocation in

48: individuals, sizes of human settlements, website popularity, words

49: ranked by frequency in a random corpus of text, observe the Zipf

50: law. Empirical evidence of the Zipf distribution of wealth [1-9]

51: has recently attracted a lot of interest of economists and

52: physicists. To understand the micro mechanism of this challenging

53: problem, various models have been proposed. One type of them is

54: based on the so-called multiplicative random process[10-21]. In

55: this approach, individual wealth is multiplicatively updated by a

56: random and independent factor. A very nice power law is given,

57: however, this approach essentially does not contain interactions

58: among individuals, which are responsible for the economic

59: structure and aggregate behavior. Another pattern takes into

60: account the interaction between two individuals that results in a

61: redistribution of their assets[22-25]. Unfortunately, some

62: attempts only give Boltzmann-Gibbs distribution of assets[24,25],

63: while some others[23], though exhibiting Zipf distributions, fail

64: to provide a stationary state.

65: 

66: In this paper, we shall introduce a new perspective to understand

67: this problem. Our model is based on the following observations:

68: (i) In order to minimize costs and maximize profits, two

69: corporations/economic entities may combine into one. This

70: phenomenon occurs frequently in real economic world. Simply fixing

71: attention on capital movements, we can equally say that two

72: capitals combine into one.(ii) The disassociation of an economic

73: entity into many small sections or individuals is also

74: commonplace. The bankruptcy of a corporation, for instance, can be

75: effectively classified into this category. Allocating a fraction

76: of assets for the employee's salary, a company also serves as a

77: good example for the fragmentation of capitals. Under some

78: appropriate assumptions, we shall establish a money-based model

79: which is essentially an extension of the Egu\'{\i}luz and

80: Zimmermann's (EZ) model for crowding and information transmission

81: in financial markets\cite{ez1,ez2}. The size of a cluster there is

82: now identified as the wealth of an agent here. However, analytical

83: results will show that our model is quite different from EZ's

84: \cite{ez2}, which gives PLD with an exponential cut-off that

85: vanishes only for a particular parameter. Here, a Zipf

86: distribution of wealth is obtained within a wide range of

87: parameters, and surprisingly, without exponential correction. The

88: Zipf exponent can be analytically calculated and is found to have

89: a nontrivial dependence on our model parameters.

90: 

91: This paper is organized as follows:  In section 2, the model is

92: described and the corresponding master equation is provided

93: directly. In section 3, we shall present our analytical

94: calculation of the Zipf exponent. Next, we give numerical studies

95: for the master equation, which are in excellent agreement with

96: analytic results. In section 5, the relevance of our model to the

97: real world are discussed.

98: 

99: 

100: \section{the model}

101: 

102: The money-based model contains $N$ units of money, where $N$ is

103: fixed. Though in real economic environment the total wealth is

104: quite possible to fluctuate, our assumption is not oversimplified

105: but reasonable, given that the production and consumption

106: processes are simultaneous and the resource is finite. The $N$

107: units of money are then allocated to $M$ agents (or say, economic

108: entities), where $M$ is changeable with the passage of time. For

109: simplicity, we may choose the initial state containing just $N$

110: agents, each with one unit of capital. The state of system is

111: mainly described by $n_s$, which denotes the number of agents with

112: $s$ units of money. The evolution of the system is under following

113: rules: At each time step, a unit of money, instead of an agent, is

114: selected at random. Notice that our model is much more

115: concentrating on the capital movement among agents rather than the

116: agents themselves. With probability $a\gamma/s$, the agent who

117: owns this unit of money is disassociated, here $s$ is the amount

118: of capitals owned by this agent and $\gamma$ is a constant which

119: implies the relative magnitude of dissociative possibility at a

120: macro level. After disassociation, this $s$ units of money are

121: redistributed to $s$ new agents, each with just one unit. It must

122: be illuminated that an real economic entity in most cases does not

123: separate in such an equally minimal way. However, with a point of

124: statistical view and considering analytical facility, this

125: simplified hypothesis is acceptable for original study. Now,

126: continue with our evolution rules. With probability

127: $a(1-\gamma/s)$, nothing is done. And with probability $1-a$,

128: another unit of money is selected randomly from the wealth pool.

129: If these two units are occupied by different agents, then the two

130: agents with all their money combine into one; otherwise, nothing

131: occurs. Thus, $1-a$ in our model is a factor reflecting the

132: possibility for incorporation at a macro level.

133: 

134: One may find that as $a$ is close to 1 and $\gamma$ is not too

135: small, a financial oligarch is almost forbidden to emerge in the

136: evolution of the system; but, if the initial state contains any

137: figure such as Henry Ford or Bill Gates, he is preferentially

138: protected. Note that the bankruptcy probability of moneybags is

139: inverse proportional to their wealth ranks, and the possibility of

140: being chosen is proportional to $sn_s/N$; thus, the Doomsday of a

141: tycoon comes with possibility $an_s\gamma/N$, which is extremely

142: small for large $s$. Meanwhile, the vast majority, if initially

143: poor, is perpetually in poverty, with no chance to raise the

144: economic status any way. In addition, if middle class exists at

145: first, it will not disappear or expand in the foreseeable future.

146: Again, it may be interesting to argue that when $a$ is slightly

147: above zero, the merger process is prevailing and overwhelming, and

148: all the capitals are inclined to converge. In this case, though

149: the rich are preferentially protected, the trend in the long run

150: is to annihilate them until the last. Of course, one-agent game is

151: trivial. Likewise, it is not appealing to observe the system when

152: $\gamma$ goes to 0 and $a$ to 1, since both merger and

153: disintegration are nearly impossible--in other words, all the

154: capitals are locked, thus the wealth pool is dead at any time.

155: 

156: Following Refs.\cite{ez2,x1,x2} in the case of $N\gg 1$, we give

157: the master equation for $n_s$

158: \begin{equation}

159: {\partial n_s\over\partial t}={1-a\over N}\sum_{r=1}^{s-1} rn_r(s-r)n_{s-r}

160: -2(1-a)sn_s-asn_s{\gamma\over s}

161: \end{equation}

162: for $s>1$ and

163: \begin{eqnarray}

164: {\partial n_1\over\partial t}

165: &&=-2(1-a)n_1+a\sum_{s=2}^{\infty}s^2n_s{\gamma

166: \over s}\nonumber\\

167: &&=-2(1-a)n_1+a\gamma(N-n_1)

168: \end{eqnarray}

169: where the identity

170: \begin{equation}

171: \sum_{s=1}^{\infty} sn_s=N

172: \end{equation}

173: has been used. We must point out that Eq.(1) is almost the same as

174: the master equation derived in Ref.\cite{ez2} for the EZ model

175: except for an additional factor $\gamma/s$ in the third term on

176: the right hand side of Eq.(1). Notice that this term is

177: significant because otherwise the frequency of the disintegration

178: for large $s$ agents would be too high.

179: 

180: Now we introduce $h_s=sn_s/N$, which indicates the ratio of wealth

181: occupied by agents in rank $s$ to the total wealth, and

182: $\alpha=a\gamma/2(1-a)$, that represents the maximum ratio of the

183: disintegration possibility to the merger probability in the whole

184: economic environment. Then, one can give the equations for the

185: stationary state in a terse form:

186: \begin{equation}

187: h_s={s\over 2(s+\alpha)}\sum_{r=1}^{s-1}h_rh_{s-r}

188: \end{equation}

189: and

190: \begin{equation}

191: h_1={\alpha\over 1+\alpha}

192: \end{equation}

193: According to the definition of $h_s$, it should satisfy the

194: normalization condition Eq.(3)

195: \begin{equation}

196: \sum_{s=1}^{\infty}h_s=1

197: \end{equation}

198: When $\alpha$ is less than a critical value $\alpha_c=4$ which

199: will be determined numerically in section 4, one can show that

200: Eqs.(4-5) does not satisfy the normalization condition Eq.(3).

201: This inconsistency implies that when $\alpha<\alpha_c$ the state

202: with one agent who has all the $N$ units of money becomes

203: important\cite{x1,x2}. In this case, the finite-size effect and

204: the fluctuation effect become nontrivial and the master equations

205: (1-3) is no longer applicable to describe the system\cite{x1,x2}.

206: In this paper, we shall restrict our discussion to the case

207: $\alpha>\alpha_c$.

208: 

209: \section{Analytic results}

210: 

211: When $\alpha>\alpha_c$, one can show that $h_s\to A/s^{\eta}$ for

212: sufficiently large $s$ with

213: \begin{equation}

214: \eta={\alpha\over\sum_{r=1}^{\infty} rh_r}

215: \end{equation}

216: Notice that this equation is only consistent when $\eta>2$ because

217: otherwise the sum $\sum_{r=1}^{\infty}rh_r$ would be divergent,

218: and thus $h_s\to A/s^{\eta}$ becomes an inconsistent formula.

219: 

220: The derivation of Eq.(7) is described as follows: When $s$ is sufficiently

221: large

222: \begin{eqnarray}

223: h_s&&={s\over 2(s+\alpha)}\sum_{r=1}^{s-1} h_rh_{s-r}\nonumber\\

224: &&\approx {s\over s+\alpha}(\sum_{r=1}^{s^\delta} h_{s-r}h_r

225: +h_s O({1\over s^{2\delta\eta-1-\eta}}))\nonumber\\

226: &&\approx {s\over s+\alpha}\sum_{r=1}^{s^\delta}(h_s-{dh_s\over ds}r)h_r

227: \nonumber\\

228: &&\approx (1-{\alpha\over s})[h_s\sum_{r=1}^\infty h_r-{dh_s\over ds}\sum_{r=1}

229: ^{\infty} rh_r]+h_sO({1\over s^{\delta(\eta-1)}})\nonumber\\

230: &&\approx (1-{\alpha\over s})[h_s-{dh_s\over ds}\sum_{r=1}^{\infty} rh_r]

231: \end{eqnarray}

232: where $\delta<1$ but is close to 1, $\delta(\eta-1)>1$ and

233: $2\delta\eta-1-\eta>1$.  Therefore

234: $${dh_s\over ds}=-{h_s\over s}{\alpha\over\sum_{r=1}^{\infty} rh_r}$$

235: which gives that as $s\to\infty$

236: \begin{equation}

237: h_s={A\over s^{\eta}}

238: \end{equation}

239: 

240: The value of $\sum_{r=1}^{\infty}rh_r$ can be further evaluated:

241: 

242: Introducing the generating function

243: \begin{equation}

244: G(x)=\sum_{r=1}^{\infty} x^r h_r

245: \end{equation}

246: one can rewrite Eq.(4) as

247: $$x(G'-h_1)+\alpha (G-h_1x)=xG'+\alpha(G-x)=xG'G$$

248: or

249: \begin{equation}

250: G'x(G-1)=\alpha (G-x)

251: \end{equation}

252: with the initial condition

253: \begin{equation}

254: G(0)=0

255: \end{equation}

256: Since $h_s\to A/s^\eta$ as $s\to\infty$, $G$ is only defined in

257: the interval $|x|\leq 1$.  From Eq.(6), we also have $G(1)=1$.

258: What we need to calculate is just

259: $$G'(1)=\sum_{r=1}^{\infty}rh_r$$

260: Since the left and the right hand sides of Eq.(11) are both zero at $x=1$, we

261: differentiate both sides by $x$ and obtain

262: $$G''x(1-G)+G'(1-G)-xG'^2=\alpha (1-G')$$

263: Let $x\to 1$ and one finds that $G''(1-G)$ vanishes in this limit

264: provided $\eta>2$, thus

265: \begin{equation}

266: G'^2(1)-\alpha G'(1)+\alpha=0

267: \end{equation}

268: One immediately obtains that

269: \begin{equation}

270: \sum_{r=1}^{\infty} rh_r={\alpha-\sqrt{\alpha^2-4\alpha}\over 2}

271: \end{equation}

272: and the exponent

273: \begin{equation}

274: \eta={2\over 1-\sqrt{1-4/\alpha}}

275: \end{equation}

276: which is a positive real number for $\alpha\geq 4$. Notice that

277: when $\alpha=4$, the exponent $\eta=2$. This implies that our

278: calculation is self-consistent, provided Eq.(6). In sum, we find

279: from the master equation that $h_s$ obeys PLD when $s$ is

280: sufficiently large and $\alpha>4$. It may be important to point

281: out that when $s$ is small, $h_s$ also approximately obeys the

282: PLD, and the restriction $\alpha>4$, introduced for the sake of

283: discussing master equation, can be actually relaxed. This argument

284: has been tested by the simulator investigation, which supplies the

285: gap of analytical tools and verifies the analytical outcome.

286: 

287: \section{Numerical results}

288: 

289: \begin{table}

290: \caption{\label{tab:table1} The results of $H$ for various value

291: of $\alpha$. }

292: \begin{tabular}

293: {|c|c|}\hline $\alpha$ & $H$  \\ \hline 3.0 & 0.9940886  \\

294: \hline 3.5 & 0.9997818  \\ \hline 3.6 & 0.9999214  \\ \hline 3.7 &

295: 0.9999743

296: \\ \hline 3.8 & 0.9999922  \\ \hline 3.9 & 0.9999977  \\ \hline

297: 4.0 & 0.9999995  \\ \hline 4.1 & 0.9999999  \\ \hline 4.2 &

298: 1.0000000  \\ \hline 4.3 & 1.0000000  \\ \hline 4.4 & 1.0000000

299: \\ \hline 4.5 & 1.0000000  \\ \hline 5.0 & 1.0000000  \\ \hline

300: 6.0 & 1.0000000  \\ \hline

301: \end{tabular}

302: \end{table}

303: 

304: We have numerically calculated the number

305: $$H=\sum_{r=1}^{\infty} h_r$$

306: based on the recursion formula Eq.(4) with the initial condition

307: Eq.(5).  Table.1 lists the results of $H$ for various value of

308: $\alpha$.  From Table.1, one immediately find that the

309: normalization condition is satisfied for $\alpha>\alpha_c=4$,

310: which, again, indicates consistency of related equations.

311: 

312: 

313: 

314: Fig.1-2 show $h_s$ as a function of $s$ in a log-log scale for

315: $\alpha=10$, $\alpha=4.5$, respectively.  From Fig.1, one can see

316: that $h_s$ conforms to PLD for $s\gg1$ with the exponent $\eta$

317: given by Eq.(15). Fig.2 indicates that $h_s$ observes the Zipf law

318: for nearly all $s$ with $\eta=3.0$.

319: 

320: The fitted exponents for various values of $\alpha$ are plotted in

321: Fig.3. They are given by

322: $${\ln (h_{900}/h_{1000})\over \ln (1000/900)}$$  Fig.3 also exhibits the analytic

323: results from Eq.(15). The analytic outcome fits the exponents

324: calculated from recursion quite well for $\alpha>4.2$. However,

325: when $\alpha\to 4.0$, discrepancy is obvious, since the

326: convergence of $h_s$ to the correct power law is then very slow.

327: 

328: We have also performed computer simulation, which gives excellent

329: agreement with theoretical results derived from Eqs.(4-5) for

330: $\alpha=8$ and $s\leq 10$, see Fig.4. For more about our simulator

331: investigation and further analysis for $\alpha<4$, see Ref.

332: \cite{hu}.

333: 

334: \begin{figure}

335: \scalebox{0.8}[0.75]{\includegraphics{A10}}

336: \caption{\label{fig:epsart} The dependence of $h_s$ on $s$ in a

337: log-log scale for $\alpha=10$.}

338: \end{figure}

339: 

340: \begin{figure}

341: \scalebox{0.8}[0.75]{\includegraphics{A45}}

342: \caption{\label{fig:epsart} The dependence of $h_s$ on $s$ in a

343: log-log scale for $\alpha=4.5$.}

344: \end{figure}

345: 

346: \begin{figure}

347: \scalebox{0.8}[0.75]{\includegraphics{eta}}

348: \caption{\label{fig:epsart} The calculated exponent $\eta$ for

349: different values of $\alpha$. Black squares represent the

350: numerical results of $\eta$ obtained from $h_s$ using the

351: extropolation method, see text.  The solid line represents the

352: analytic result Eq.(15).}

353: \end{figure}

354: 

355: \begin{figure}

356: \scalebox{0.8}[0.75]{\includegraphics{A8}}

357: \caption{\label{fig:epsart} $h_s$ for $\alpha=8$ from both

358: numerical calculation and computer simulation. Black stars

359: represent outcome of computer simulation for $N=2.5\times 10^{5}$,

360: $\gamma=2$ and $a=0.88889$. Total $2\times 10^6$ time steps were

361: run and the final $5\times 10^5$ time steps were used to count

362: $n_s$ statistically. The circles represent the theoretical results

363: derived from Eqs.(4-5).}

364: \end{figure}

365: 

366: 

367: \section{Discussions}

368: 

369: In this paper, we have introduced a so-called money-based model to

370: mimic and study the wealth allocation process. We find for a wide

371: range of parameters, the wealth distribution $n_s\sim

372: A/s^{\eta+1}$ with $\eta$ given by Eq.(15) for sufficiently large

373: $s$. The crucial difference between our model and the EZ model is

374: that the dissociative probability $\Gamma_d$ of an economic

375: entity, after he/she is picked up, is proportional to $1/s$ in our

376: model. However, the corresponding probability in the EZ model is

377: simply proportional to 1. This difference gives rise to divergent

378: behaviors of $n_s$. In the EZ model, $n_s\sim A/s^{2.5}

379: \exp(-\alpha s)$ for large $s$ \cite{ez2}. When $n_s$ is

380: interpreted as the number of individuals who own $s$ units of

381: assets, the choice of $\Gamma_d\sim O(1/s)$ is reasonable.

382: Actually, since at the first step, we randomly picked up a unit of

383: money, the individual who owns $s$ units of assets is picked up

384: with a probability proportional to $s$. According to the

385: observation in real economic life, large companies or rich men are

386: often much more robust than small or poor ones when confronting

387: economic impact and fierce competition. If $\Gamma_d\sim O(1)$,

388: the overall dissociation frequency would be proportional to $s$

389: which is totally unreasonable.

390: 

391: In real economic environment, capitals and agents behave similarly

392: at some point. For instance, they both ceaselessly display

393: integration and disintegration, driven by the motivation to

394: maximize profits and efficiency. This mechanism updates the system

395: every time, and gives rise to clusters and herd behaviors.

396: Furthermore, in an agent-based model, it is usually indispensable

397: to consider the individual diversity that is all too often hard to

398: deal with. When it comes to the money-based model, this micro

399: complexity may be considerably simplified. Finally, the conceptual

400: movement and interaction among capitals is not as restricted by

401: space and time as between agents. Therefore, when econophysics is

402: much more interested in the behaviors of capitals than that of

403: agents, it is recommendable to adopt such a money-based model.

404: 

405: The methodology to fix our attention on the capital movements,

406: instead of interactions among individuals, will bring a lot of

407: facility for analysis; moreover, using such random variables as

408: $\gamma$ and $a$ to represent the macro level of the micro

409: mechanism also help us find a possible bridge between the

410: evolution of the system and the protean behaviors of individuals.

411: Whether the bridge is steady or not can only be tested by further

412: investigation.

413: 

414: \begin{acknowledgments}

415: This work has been partially supported by the State Key

416: Development Programme of Basic Research (973 Project) of China,

417: the National Natural Science Foundation of China under Grant

418: No.70271070 and the Specialized Research Fund for the Doctoral

419: Program of Higher Education (SRFDP No.20020358009)

420: \end{acknowledgments}

421: 

422: \begin{thebibliography}{zz1}

423: 

424: \bibitem{zz1} G.K.Zipf, {\it Human Behavior and the Principle of Least Effort}

425: (Addison-Wesley Press, Cambridge, MA, 1949).

426: 

427: \bibitem{z2} V. Pareto, {\it Cours d'Economique Politique} (Macmillan, Paris, 1897),

428: Vol 2.

429: \bibitem{z3} B. Mandelbrot, Economietrica {\bf 29},517(1961).

430: \bibitem{z4} B.B. Mandelbrot, Comptes Rendus {\bf 232}, 1638(1951).

431: \bibitem{z5} B.B. Mandelbrot, J. Business {\bf 36}, 394(1963).

432: \bibitem{z6} A.B. Atkinson and A.J. Harrison, {\it Distribution of Total Wealth in Britain}

433: (Cambridge University Press, Cambridge, 1978).

434: \bibitem{z7} H. Takayasu, A.-H. Sato and M. Takayasu, Phys.Rev.Lett. {\bf 79},966(1997).

435: \bibitem{z8} P.W. Anderson, in {\it The Economy as an Evolving Complex System II},

436: edited by W.B. Arthur, S.N. Durlauf and D.A. Lane (Addison-Wesley,

437: Reading, MA,1997).

438: \bibitem{z9} {\it The Theory of Income and Wealth Distribution}, edited by Y.S. Brenner

439: {\it et al.},(New York, St. Martin's Press, 1988).

440: \bibitem{m1} H.A. Simon and C.P. Bonini, Am. Econ. Rev. {\bf 48},607(1958).

441: \bibitem{m2} U.G. Yule, Philos. Tran. R. Soc. London, Ser. Bc{\bf 213},21(1924).

442: \bibitem{m3} D.G. Champernowne, Econometrica {\bf 63}, 318(1953).

443: \bibitem{m4} H. Kesten, Acta Math. {\bf 131},207(1973).

444: \bibitem{m5} S. Solomon, in {\it Annual Reviews of Computational Physics II,}

445: edited by D. Stauffer (World Scientific, Singapore, 1995), p.243.

446: \bibitem{m6} M. Levy and S. Solomon, Int. J. Mod. Phys. C {\bf 7}, 595(1996).

447: \bibitem{m7} S. Solomon and M. Levy, Int. J. Mod. Phys. C {\bf 7}, 745(1996).

448: \bibitem{m8} O. Malcai, O. Biham and S. Solomon, Phys. Rev. E {\bf 60}, 1299(1999).

449: \bibitem{m9} B.B. Mandelbrot, Int. Ecomomic Rev. {\bf 1}, 79(1960).

450: \bibitem{m10} E.W. Montroll and M.F. Shlesinger, {\it Nonequilibrium Phenomena II. From

451: Stochastics to Hydrodynamics}, edited by J.L. Lebowitz, E.W. Montroll(North-Holland,

452: Amsterdam, 1984).

453: \bibitem{m11} D. Sornette and R. Cont, J. Phys. I France {\bf 7}, 431(1997).

454: \bibitem{m12} H. Takayasu and K. Okuyama, Fractals {\bf 6}(1998).

455: \bibitem{e1} Z.A. Melzak, {\it Mathematical Ideas, Modeling and Applications, Volume II of

456: Companion to Concrete Mathematics} (Wiley, New York, 1976), p279.

457: \bibitem{e2} S. Ispolatov, P.L. Krapivsky and S. Redner, Euro. Phys. J. B {\bf 2}, 267(1998).

458: \bibitem{e3} A. Dragulescu and V.M. Yakovenko, Euro. Phys. J. B {\bf 17}, 723(2000).

459: \bibitem{e4} C.B. Yang, to be published in Chinese Phys. Lett.

460: \bibitem{ez1} V.M. Egu\'{\i}luz and M.G. Zimmermann, Phys. Rev. Lett. {\bf 85},5659(2000).

461: \bibitem{ez2} R. D'Hulst and G.J. Rodgers, Euro. Phys. J. B {\bf 20},619(2001).

462: \bibitem{x1} Y.B. Xie, B.H. Wang, H.J. Quan, W.S. Yang and P.M. Hui, Phys. Rev. E {\bf 65},

463: 046130(2002).

464: \bibitem{x2} Y.B. Xie, B.H. Wang, H.J. Quan, W.S. Yang and W.N. Wang, Acta Physica Sinica

465: ({\it Chinese}), {\bf 52},2399(2003).

466: \bibitem{hu} B. Hu, Y.B. Xie, T. Zhou and B.H. Wang (unpublished).

467: 

468: 

469: \end{thebibliography}

470: 

471: 

472: 

473: \end{document}

474: