1: \documentclass[prl,aps,preprint,showpacs,showkeys]{revtex4}
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5: \usepackage{hyperref}
6: \usepackage{latexsym}
7: \usepackage{color}
8: \topmargin -0.5 cm
9: \textwidth 16.0 cm
10: \oddsidemargin -0.0 cm
11: \linespread{1.6}
12: \renewcommand{\baselinestretch}{1.2}
13:
14: \begin{document}
15: %\setcounter{page}{1}
16: \date{\today}
17: \title{A Fractal Space-filling Complex Network}
18:
19: \author{D. J. B. Soares$^{1,4}$, J. Ribeiro Filho$^{1,2}$,A. A. Moreira$^1$,
20: D. A. Moreira$^3$, and G. Corso$^{3,4}$ }
21:
22: \affiliation{ $^1$
23: Departamento de F\'{\i}sica, Universidade Federal do Cear\'a,
24: 60451-970 Fortaleza, CE, Brazil}
25: \affiliation{ $^2$
26: Curso de Matem\'atica, Universidade Estadual Vale do Acara\'u 62040-370,
27: Sobral, CE, Brazil}
28: \affiliation{ $^3$ Departamento de F{\'\i}sica Te\'orica e Experimental,
29: Universidade Federal do Rio Grande do Norte, Campus Universit\'ario,
30: 59078 970 Natal, RN, Brazil}
31: \affiliation{ $^4$ Departamento de Biof{\'\i}sica e Farmacologia,
32: Centro de Bioci\^encias,
33: Universidade Federal do Rio Grande do Norte, Campus Universit\'ario
34: 59072 970, Natal, RN, Brazil}
35:
36: \hspace{1cm}
37:
38: \begin{abstract}
39:
40: We study in this work the properties of the $Q_{mf}$ network which is
41: constructed from an anisotropic partition of the square, the multifractal tiling.
42: This tiling is build using a single parameter $\rho$, in the limit of
43: $\rho \rightarrow 1$ the tiling degenerates into the square lattice that is
44: associated with a regular network.
45: The $Q_{mf}$ network is a space-filling network with the
46: following characteristics: it shows a power-law distribution of
47: connectivity for $k>7$ and it has an high clustering coefficient when compared
48: with a random network associated. In addition the $Q_{mf}$ network
49: satisfy the relation $N \propto \ell^{d_f}$ where $\ell$ is a typical length
50: of the network (the average minimal distance) and
51: $N$ the network size. We call $d_f$ the fractal
52: dimension of the network. In the limit case of
53: $\rho \rightarrow 1$ we have $d_f \rightarrow 2$.
54:
55: \end{abstract}
56: %.......................................
57:
58:
59: \pacs{89.75.Da, 89.75.Hc, 61.43.Hv}
60:
61: \keywords{networks, fractals, space-filling, multifractal tiling, small-world}
62:
63: \maketitle
64:
65: \section{I - Introduction}
66:
67: The last years have seen an increasing interest in network studies
68: in physics \cite{barabasi,strogatz}. Despite graph
69: theory have been a research topic of mathematics and science
70: of computation, the physicists have driven their attention
71: to networks that show distribution of connectivity, $P(k)$, following a
72: power law and that present small world effect: $\ell \propto ln N$, for
73: $\ell$ a typical network distance and $N$ the size of the network.
74: A recent article \cite{nature} points the difficulty to put together these
75: two aspects in a common broad scale-free framework, it means, a fractal
76: paradigm. In fact, a real fractal should have $N \propto \ell^{d_f}$ for
77: $d_f$ the fractal dimension, and not a logarithm dependence between
78: $N$ and $\ell$. In this work we present a network that have the following
79: characteristics: it is fractal, $N \propto \ell^{d_f}$,
80: it is scale-free $P(k) \propto k^{-\gamma}$ for a large range of $k$, and
81: it is complex in the sense of having large clustering coefficient.
82: The scaling relation $N \propto \ell^d$ is trivially fulfilled for
83: regular lattices where $d$ is an integer, the topologic dimension of the space.
84: The complex fractal network we explore in this paper has a non integer $d_f$.
85:
86:
87: Recently networks embedded in metric spaces
88: have been investigated in the literature \cite{Soares} because of
89: the large applications of networks that effectively occupy a volume in
90: 3-dimensions. In addition, motived by microprocessors design
91: space-filling networks \cite{andrade} have been studied.
92: We cite these new trends in networks because the distinguished
93: network we analyze in this paper
94: is embedded in a metric space and furthermore it is space filling.
95: In fact our network has a geometrical inspiration, it comes
96: from a partition of the square. Indeed
97: this network is originated from a singular tiling
98: that has the additional property of being a multifractal partition of the square
99: \cite{corso1}.
100:
101: We follow the previous literature
102: \cite{corso1,lim,randi1,randi2,multphy} and call this object the
103: multifractal tiling, $Q_{mf}$.
104: The $Q_{mf}$ tiling was developed in the context of modeling transport
105: and percolation in heterogeneous porous media.
106: In fact, a broad set of irregular and
107: heterogeneous systems are model in the literature using a multifractal
108: approach. We cite systems in geology \cite{Riedi,Hermann},
109: atmospheric science \cite{Muller1} and chemistry \cite{stan}.
110: Oil reservoirs are complex anisotropic structures whose
111: treatment have been challenged science and technique because of its
112: non trivial geometry.
113: Inspired in the description of oil reservoirs it was developed
114: \cite{corso1} the $Q_{mf}$ tiling.
115: In reference \cite{lim} an exhaustive study of the
116: percolation threshold of $Q_{mf}$ was performed, in
117: \cite{randi1} a random version of $Q_{mf}$ was created, in
118: \cite{randi2} some of its percolation critical exponents were
119: found, and in \cite{multphy} a numerical study of its coordination
120: number is done.
121:
122: In this work we explore the network properties of
123: the $Q_{mf}$, it means its topology, the study
124: of the connections (neighborhood) among the cells of the tiling.
125: The paper is organized as follows.
126: In section $2$ we describe in some detail
127: the process of construction of the $Q_{mf}$ tiling
128: and show its more important properties.
129: In section $3$ we show the main
130: results concerning the properties of the network:
131: the distribution of connectivity, the scaling of the
132: the average minimal distance between two sites and the
133: clustering coefficient.
134: Finally in section $4$ we present our conclusions,
135: discuss the main implications of our results and compare
136: the properties of the $Q_{mf}$ network with other networks in
137: the literature.
138:
139: \section{II - The multifractal object}
140:
141: The multifractal tiling is a peculiar partition of the square.
142: It is interesting to think about the $Q_{mf}$ tiling in contrast to
143: the square lattice.
144: The square lattice can be constructed using the following algorithm:
145: take a square and cut it symmetrically with horizontal and vertical
146: lines. This procedure produces four square cells.
147: Repeat this procedure $n$ times inside each new block and you have finally
148: a square lattice with $2^n$ cells. The $Q_{mf}$ object is generated
149: in a similar way as the square lattice above described, but instead of using a
150: symmetric partition we perform horizontal and vertical sections following
151: a given ratio.
152:
153:
154: In Fig. \ref{fig1} we exemplify the five initial steps of the construction
155: of the multifractal for the parameter $\rho=1/3$, or $(s,r)=(1,3)$.
156: In \ref{fig1} (a) we show, $n=0$, the
157: initial square that by convenience we assume of size $L=1$.
158: In (b), $n=1$, a vertical cut is performed and two rectangles are formed.
159: We call this a $(s,r)=(1,3)$ object because the square is divided in
160: $4$ parts such that $1$ part stays at one side and $3$ parts at the
161: other side. In (c), $n=2$, two
162: horizontal lines are drawn using the same section rate as before.
163: At this level the initial square generates
164: four rectangular blocks. Using as the area unit a square of size
165: $\epsilon =1/(s+r)$, the largest block has area $r^2$,
166: there are two blocks of areas $r s$ and the
167: smallest block has area $s^2$.
168: In (d) and (e), $n=3$ and $n=4$, respectively,
169: the same procedure is repeated inside the initial four blocks.
170: In reference \cite{new_phys} it is explored the group of eight possibilities
171: of cutting a square lattice with a given ratio. In this work, as in other
172: papers about $Q_{mf}$ \cite{corso1,lim,randi1,randi2,multphy},
173: it is followed the recipe of Fig. \ref{fig1}.
174:
175: \begin{figure} \begin{center}
176: \includegraphics[width=3cm]{Fig1a.ps}
177: \includegraphics[width=3cm]{Fig1b.ps}
178: \includegraphics[width=3cm]{Fig1c.ps}
179: \includegraphics[width=3cm]{Fig1d.ps}
180: \includegraphics[width=3cm]{Fig1e.ps}
181: \caption{
182: \baselineskip=12pt
183: The first five steps in the construction in the multifractal tiling $Q_{mf}$
184: for $(s,r)=(1,3)$. }
185: \label{fig1}
186: \end{center}
187: \end{figure}
188:
189: \begin{figure} \begin{center}
190: \includegraphics[width=7cm]{Fig2a.ps}
191: \includegraphics[width=7cm]{Fig2b.ps}
192: \caption{
193: \baselineskip=12pt
194: A $Q_{mf}$ multifractal tiling and a $Q_{mf}$ network,
195: for $\rho=1/2$ and $n=6$.
196: In (a) the original tiling and in (b) the corresponding network. }
197: \label{fig2}
198: \end{center}
199: \end{figure}
200:
201: We remark that the number of blocks at step $n$ is $2^n$. These blocks do
202: not have all the same area, we call the
203: subsets of blocks of same area by a $k$-set.
204: It is easy to check that the block area distribution
205: follows a binomial distribution and the number of $k$-sets is
206: $k=n+1$. This fact implies that the $Q_{mf}$ has the
207: remarkable property: in the limit of $n \rightarrow \infty$
208: the area of its forming blocks follows a multifractal distribution
209: \cite{corso1}.
210: The spectrum of fractal distributions comes from a box counting
211: reasoning:
212: \begin{equation}
213: D_X= \lim\limits_{\epsilon \rightarrow 0}
214: \frac{\log \> N(X)}{\log \> (1/\epsilon)}
215: \label{NN}
216: \end{equation}
217: for $N(X)$ the number of unitary cells
218: of size length $\epsilon$ that cover the set of blocks of a
219: given area $X$. Once the initial square is
220: partitioned $n$ times, the size of the unitary cell is
221: $\epsilon = 1/(s+r)^{n}$.
222: For each $k$-set the total area of blocks (using $\epsilon$ area units) is done by:
223: \begin{equation}
224: N_k = C_n^k \> \> s^k \>r^{(n-k)},
225: \label{NNN}
226: \end{equation}
227: where $C_n^k$ is the binomial coefficient that express
228: the number of elements $k$-type, and
229: $s^k \> r^{(n-k)}$ is the area of each element of this set.
230: We put together all these elements to have the fractal dimension of
231: each $k$-set:
232: \begin{equation}
233: D_k = \lim\limits_{\epsilon \rightarrow 0} \frac{ \log N_k }{ \log (1/\epsilon)} =
234: \lim\limits_{n \rightarrow \infty} \frac{\log (\> C_n^k \>
235: \> s^k \>r^{(n-k)})}{\log \> (s+r)^n }.
236: \label{boxcou2}
237: \end{equation}
238: This distribution show a concave shape with a maximum at $k=\rho n$.
239: The case $r=s=1$ is degenerated. In this situation the
240: subsets of the lattice are composed uniquely by square cells of the same area.
241: Therefore the tiling is formed by a single subset of dimension $2$.
242:
243: In Fig. \ref{fig2} (a) it is shown an example of $Q_{mf}$ construction for
244: $(s,r) = (1,2)$ and $n=6$. In Fig. \ref{fig2} (b) we build a network
245: corresponding to this tiling. The nodes of the network are the $2^n$ blocks
246: of the $Q_{mf}$
247: and the vertices are established according to a neighborhood criterium. These
248: last figures offer a glimpse of the metric heterogeneity and
249: the topology of the multifractal.
250: In the next section we explore in detail the network properties of this
251: class of objects.
252:
253: \section{III - Results}
254:
255: We start the analysis of the properties of the $Q_{mf}$ network discussing
256: its distribution of connectivity, $P(k)$. In Fig. \ref{cum} we show the cumulative
257: sum of $P(k)$ versus $k$ for several values of $\rho$ as indicated in the figure.
258: The option for the cumulative sum instead of $P(k)$ itself is
259: due to the strong fluctuation of the data. Fig. \ref{cum} confirms the results
260: of a previous work \cite{multphy}. For low $k$, typically $k<7$, the
261: curve suggests an exponential behavior and above this threshold the network
262: depicts a scale-free behavior. The values of the exponents $\gamma$ of
263: the power-law $P(k) \propto k^{-\gamma}$ are indicated in Table \ref{tab1} for
264: several $\rho$, trivially the values of $\gamma$ are estimated from
265: the slopes of the curve of the cumulative
266: sum decreased by one. The exponent $\gamma$
267: goes to an asymptotic limit for large $N$ \cite{multphy}.
268: We observe that in the limit of $\rho \rightarrow 1$ the $Q_{mf}$ tiling
269: gets more symmetric and at $\rho=1$ the $Q_{mf}$ degenerates into the square
270: lattice. For the regular square lattice $P(k)$ is a delta of Dirac centered at $k=4$
271: and the cumulative sum a step function. Fig. \ref{cum} corroborates this
272: idea, the skewed curves in the $\rho \rightarrow 1$ limit anticipate
273: the phase transition at $\rho=1$.
274:
275:
276:
277:
278:
279: \begin{figure}
280: \begin{center}
281: \includegraphics[width=7cm]{Fig3.eps}
282: \caption{
283: \baselineskip=12pt
284: The cumulative degree distribution for several values of $\rho$
285: and $N=2^{16}$. For $k > 7$ the distributions
286: approaches a power-law form. The exponent $\gamma$
287: controlling the decay of the distributions decreases in the limit of
288: $\rho \rightarrow 0$, the full set of exponents is show in Table \ref{tab1}.}
289: \label{cum}
290: \end{center}
291: \end{figure}
292:
293: We explore the distance characterization of the $Q_{mf}$ network in
294: Fig. \ref{fig4} where we display the behavior of the
295: average shortest distance for all couple of distinct vertices
296: of the network, $\ell$, versus
297: network size, $N$. The simulation is performed for some values of $\rho$ as
298: indicated in the figure. For all $\rho$ examined the data show a power-law
299: behavior $N \propto \ell^{d_f}$. We find $2 < d_f < 4$,
300: the full set of $d_f$ is shown in Table \ref{tab1}. Two limit cases
301: are interesting. The limit $\rho \rightarrow 1$, which corresponds to
302: the square lattice, has $d_f \rightarrow 2$ as it is expected in
303: a bidimensional space. The opposit limit
304: $\rho \rightarrow 0$, which is
305: associated with very anisotropic structures, shows large $d_f$. We
306: cannot affirm that $4$ is an asymptotic threshold, further
307: numerical investigation should test this hypothesis.
308: We remark that the $Q_{mf}$ network does not
309: follow a small world relationship $\ell \propto \ln N$ that is common
310: to most of power-law and random like networks.
311:
312:
313: \begin{figure}
314: \begin{center}
315: \includegraphics[width=7cm]{Fig4.eps}
316: \caption{
317: \baselineskip=12pt
318: In (a) it is displayed the average distance
319: $\ell$ as a function of N for several values of
320: $\rho$. }
321: \label{fig4}
322: \end{center}
323: \end{figure}
324:
325: An analysis of the clustering coefficient, $C$, versus network size,
326: $N$, is shown in Fig. \ref{fig5}. The general view of this figure points to a
327: stable behavior of $C$ in the limit of large $N$.
328: The dispersion of $C$ among $\rho$ is not large, the numerics show
329: $C = 0.37 \pm 0.01$. Smaller values of $\rho$,
330: however, show a significant larger $C$.
331: The discussion about $C$ is intriguing once
332: we compare the numerical values of $C$ with the clustering coefficient of a
333: random network associated to the $Q_{mf}$ network. An associated random
334: network is defined as a network with the same $N$ and $<k>$ of the original
335: network (we do not compare our results with a random network with a same
336: $P(k)$ because such random network would alterate the space filling
337: characteristics that we are interested in). For a random network
338: $C=<k>/N$, in the case of our network: $<k>$ is a constant
339: number smaller than $6$ and $N$ a number that can grow without limit. As a
340: consequence the associated random network has $C \rightarrow 0$ in the limit
341: $N \rightarrow \infty$.
342: Therefore the $Q_{mf}$ network has a $C$ that is infinitely larger
343: than the clustering coefficient
344: of the associated random network. Because of the high $C$ and
345: the power-law behavior of distribution of connectivity we call the
346: $Q_{mf}$ network a complex network.
347:
348:
349: \begin{figure}
350: \begin{center}
351: \includegraphics[width=7cm]{Fig5.eps}
352: \caption{
353: \baselineskip=12pt
354: In the clustering coefficient $C$ as function of N for
355: the same set of values of $\rho$ of the previous figure. }
356: \label{fig5}
357: \end{center}
358: \end{figure}
359:
360:
361: In Table \ref{tab1} it is shown some parameters related to the
362: $Q_{mf}$ network for several values of $\rho$.
363: Most of these data was already commented in the text.
364: We focus now on the average connectivity $< k>$ of the network.
365: For all $\rho$ studied we have $< k> \sim 5.43$ which characterizes
366: a sparse network. This is not surprising, since
367: there is a result in topology that shows that
368: for two dimensions the average coordination number of a tiling
369: cannot exceed $6$.
370: The average connectivity confirms Fig. \ref{cum}
371: where we can see that the majority of
372: vertices are situated in the range: $4\leq k \leq 6$.
373: Otherwise we note that most of interesting results concerning the distribution
374: of connectivity, in special the power-law behavior, are satisfied only in
375: the range $k>7$. Indeed, the $Q_{mf}$ network, as most of complex networks, also
376: have hubs that determine the distinguished characteristics of the network.
377:
378:
379: \begin{table}[ht]
380: \caption{The average quantities: clustering coefficient, $C$, minimal distance, $ \ell$, and
381: connectivity, $< k> $; the slope $\gamma$ of the distribution of connectivity
382: and the fractal dimension $d_f$.
383: The data corresponding to the average parameters
384: are estimated for $N=2^{16}$.}
385: \vskip 0.5in
386: \begin{center}
387: \label{tab1}
388: \begin{tabular}{|c|c|c|c|c|c|c|}
389: \hline $\rho$ & $(3,4)$ & $(2,3)$ & $(1,2)$ & $(1,3)$ & $(1,4)$ & $(1,8)$
390: \\ \hline
391: $ C$ & $0.3603$ & $0.3610$ & $0.3703$ & $0.3735$ & $0.3769$ & $0.3777$
392: \\ \hline
393: $ \ell$ & $122.02$ & $115.26$ & $96.43$ & $75.12$ & $57.12$ & $41.70$
394: \\ \hline
395: $< k > $ & $5.4357$ & $5.4357$ & $5.4338$ & $5.4275$ & $5.4357$ & $5.4357$
396: \\ \hline
397: $\gamma$ & $16.6$ & $11.1 $ & $6.4$ & $4.2 $ & $3.4$ & $2.9$
398: \\ \hline
399: $d_f$ & $2.09$ & $2.14$ & $2.25$ & $2.52 $ & $2.94$ & $3.79$
400: \\ \hline
401: \end{tabular}
402: \end{center}
403: \vspace{0.5cm}
404: \end{table}
405:
406:
407: \section{IV - Conclusion}
408:
409: In this work we explore some properties of a space filling network
410: that come from a multifractal partition of the square lattice, the
411: $Q_{mf}$ network.
412: An analysis of the distribution of the connectivity, $P(k)$, assures that the
413: $Q_{mf}$ network shows a power-law tail that is more
414: accentuated as increases the anisotropy of the underlying $Q_{mf}$ tiling.
415: Roughly the power-law tail of $P(k)$ starts at $k \sim 7$. We remark that
416: there is no regular lattice in $2$ dimensions with $k>6$ and typical
417: Voronoi latices have an exponential small number of
418: vertices in this range. In addition the $Q_{mf}$ network has a clustering
419: coefficient that approaches a constant, $C = 0.37 \pm 0.01$,
420: that does not depend on $N$. This fact is in contrast to random
421: networks that (for a constant $<k>$) have $C \propto N^{-1}$. Because
422: the value of $C$ is much larger than the value of $C$ of the associated
423: random network we call the $Q_{mf}$ network a complex network.
424:
425: The most interesting aspect of the $Q_{mf}$ network concerns its
426: fractal behavior. For the average minimal path $\ell$ we
427: observe that $N \propto \ell^{d_f}$ for the fractal dimension, $d_f$.
428: The simulations show that $2 < d_f < 4$.
429: The lower limit correspond to the
430: case $\rho =1$ where the multifractal tiling degenerates
431: into the square lattice. In this situation the slope $\gamma$
432: (from the power law $P(k) \propto k^{-\gamma}$)
433: increases dramatically, this situation corresponds to the $P(k)$ of
434: the square lattice that is of the form of a Delta of Dirac.
435: The opposite limit, $\rho =0$,
436: corresponding to very anisotropic tilings, presents comparatively small
437: values of $\gamma$.
438:
439:
440: We point that, diversly from \cite{nature},
441: the $Q_{mf}$ network shows an actual fractal behavior
442: $N \propto \ell^{d_f}$ that is obtained without any renormalization
443: artefact. In the reference \cite{nature} a ingenious procedure
444: is used to calculate two fractal dimensions $d_B$ and $d_f$. The
445: first dimension depends on a suitable embedding in a metric space and a box counting
446: methodology. The second is based on network distance and a mass (number
447: of vertices) inside a given radios. In our case, we have calculated $d_f$ in
448: the standard way, the box-counting, however, depends on the methodology
449: we use to make the embedding of the $Q_{mf}$ object. The simplest embedding
450: is the $Q_{mf}$ lattice itself that is a 2-dimensional object and as a result
451: $d_B=2$. Note that the multifractal property of $Q_{mf}$ appears when we
452: consider the subsets of blocks of same area, if we disregard the
453: area set a bidimensional tilling assumes the trivial topologic dimension, $d_f=2$.
454:
455:
456: The $Q_{mf}$ tiling is indeed a remarkable mathematical object, from a
457: metric perspective it is a multifractal: it is formed by a
458: denumerable quantity of sets of different areas each one with
459: a given fractal dimension. In a topologic perspective the
460: connections among the vertices (the cells of the tiling)
461: form a fractal network. We remark that regular networks
462: (generated from lattices for instance) and the Bethe tree satisfy the
463: criterium $N \propto \ell^{d}$,
464: but these structures are regular.
465: For the best knowledgement of the
466: authors the $Q_{mf}$ network is the only case of a
467: true fractal, scale-free and with high clustering coefficient.
468:
469:
470: \vspace{1cm}
471:
472: \centerline{\bf Acknowledgments}
473:
474: The authors gratefully acknowledge the financial support of Conselho Nacional
475: de Desenvolvimento Cient{\'\i}fico e Tecnol{\'o}gico (CNPq)-Brazil,
476: FINEP and Programa PET-SESU/MEC. D. J. B. S. Thanks to D. R. de Paula.
477:
478:
479: \begin{thebibliography}{9}
480: %
481: %
482: \bibitem{barabasi} R. Albert and A-L Barab\'asi, Rev. Mod. Phy. {\bf 74}, 47
483: (2002).
484:
485: \bibitem{strogatz} S. H. Strogatz, Nature, {\bf 410}, 268 (2001).
486:
487: \bibitem{nature} C. Song, S. Havlin and H. A. Makse, Nature,
488: {\bf 433} 392 (2005).
489:
490: \bibitem{Soares} Danyel J. B. Soares, Constantino Tsallis, A. M. Mariz
491: and L. R. da Silva, Europhys. Lett., {\bf 70}, pp 70-76 (2005).
492:
493: \bibitem{andrade} J. S. Andrade,Jr, H. J. Hermann, R. F. S Andrade,
494: and L. R. da Silva, Phys. Rev. Lett. {\bf 94}, 018702 (2005).
495:
496: \bibitem{corso1} G. Corso, J. E. Freitas, L. S. Lucena, and R. F. Soares,
497: Phys. Rev. E. {\bf 69}, 066135 (2004).
498:
499: %\bibitem{newman05} M. E. J. Newman, Social Networks {\bf 27}, 39 (2005).
500:
501: \bibitem{lim} L. S. Lucena, J. E. Freitas, G. Corso, and R. F. Soares,
502: Brazilian Journal of Physics {\bf 33}, 637 (2003).
503:
504: \bibitem{randi1} M. G. Pereira, G. Corso, L. S. Lucena, and J. E. Freitas
505: Chaos, Solitons and Fractals, {\bf 23}, 1105 (2004).
506:
507: \bibitem{randi2} M. G. Pereira, G. Corso, L. S. Lucena, and J. E. Freitas,
508: International Journal of Modern Physics C, {\bf 16} 317 (2005).
509:
510: \bibitem{multphy} G. Corso, J. E. Freitas and L. S. Lucena, Physica A,
511: {\bf 342 }, 214 (2004).
512:
513: \bibitem{Riedi} R. H. Riedi, {\it Multifractals and wavelets: a
514: potential tool in Geophysics,} Proceedings of the $68$th SEG
515: Meeting, New Orleans, Louisiana, USA, (1998).
516:
517: \bibitem{Hermann} F. Herrmann, {\it A Scaling Medium Representation,
518: a Discussion on Well-logs, Fractals and Waves.} PhD Thesis,
519: Delft University of Technology, (1997).
520:
521: \bibitem{Muller1} J. Muller
522: Annales Geophysicae - Atmospheres Hydrospheres and Space Sciences
523: {\bf 11}, (6): 525-531 (1993).
524:
525: \bibitem{stan} H. E. Stanley and P. Meakin, Nature {\bf 355},
526: 405 (1988).
527:
528: %\bibitem{science} M. E. J. Newman, Phys. Rev E. {\bf 64}, 016131 (2001).
529:
530:
531: %\bibitem{Doro} S. N. Dorogovtsev and
532: %J. F. F. Mendes, Adv. Phys., {\bf 51 }, 1079 (2002).
533:
534: \bibitem{new_phys} G. Corso and L. S. Lucena, Physica A (to appear) (2005).
535:
536:
537: %\bibitem{barthelemy} M. Barth\'elemy, Phys. Rev. Lett., {\bf 91 },
538: %189803-1 (2003).
539:
540:
541: \end{thebibliography}
542:
543:
544:
545: \end{document}
546:
547:
548:
549:
550:
551: