cond-mat0308291/lim.tex
1: %\documentclass[11pt]{article}
2: \documentclass[prl,aps,preprint]{revtex4}
3: \topmargin -0.5 cm
4: \textwidth 16.0 cm
5: \oddsidemargin -0.0 cm
6: \linespread{1.6}
7: %\renewcommand{\baselinestretch}{1.2}
8: \begin{document}
9: \setcounter{page}{1}
10: \date{\today}
11: \title{Anisotropy and percolation threshold in a multifractal support}
12: 
13: \author{   L. S. Lucena, J. E. Freitas, G. Corso, and R. F. Soares. } 
14: 
15: \affiliation{ International Center for Complex Systems and    
16:  Departamento de F{\'\i}sica Te\'orica e Experimental,      
17:  Universidade Federal do Rio Grande do Norte, Campus Universit\'ario   
18:  59078 970, Natal, RN, Brazil.}
19: 
20: 
21: \begin{abstract}
22:  
23: Recently a multifractal object, $Q_{mf}$, was proposed 
24: to study percolation properties in a multifractal support. 
25: The area  and the number of 
26: neighbors of the  blocks of $Q_{mf}$ show a non-trivial behavior.
27: The value of the probability of occupation at the percolation 
28: threshold, $p_{c}$, is a function of $\rho$, a parameter 
29: of $Q_{mf}$ which is related to its anisotropy. 
30: We investigate the relation between $p_{c}$ 
31: and the average number of neighbors of the blocks as well as the  
32: anisotropy of $Q_{mf}$. 
33: 
34: %
35: \end{abstract}
36: 
37: \maketitle
38: 
39: \section{1- Introduction}
40: 
41: Due to the work of many physicists, and particularly to the 
42: contributions of Dietrich Stauffer, percolation theory has 
43: became a powerful tool in Science describing phenomena in 
44: many areas as geology, biology, magnetism, or social phenomena 
45:  \cite{stauffer,perc3}. Despite the enormous success of 
46: percolation it has been a theory studied in a   
47: support (lattice) that has a single dimension. 
48: The only references relating percolation and multifractality 
49: concern to the multifractal properties of some quantities of the 
50: spanning cluster at the percolation threshold \cite{multperc1,multperc2}.
51: 
52: Recently, a model to study percolation in a multifractal was 
53: proposed \cite{last} in the literature. In fact, the authors have 
54: created an original multifractal object, $Q_{mf}$, and an efficient 
55: way to estimate its percolation properties. In this work we study in 
56: detail the method to estimate  $p_c$ for this multifractal and 
57: discuss the relation between  $p_c$, some topologic characteristics, 
58: and the anisotropy of $Q_{mf}$. 
59: 
60: The multifractal object we develop, $Q_{mf}$, is an intuitive generalization of 
61: the square lattice \cite{last}. Suppose that in the construction of the square 
62: lattice we use the following algorithm: take a square  of size $1$ and 
63: cut it symmetrically with  vertical and  horizontal lines. Repeat this process 
64: $n$-times; at the $n^{th}$ step we have a regular square
65: lattice with $2^n \times 2^n$ 
66: cells. The setup algorithm of $Q_{mf}$ is quite similar, the main difference 
67: is that we do not cut 
68: the square in a symmetric way. In section $2$ we 
69: explain in detail this algorithm. 
70: 
71: The development of $Q_{mf}$ has a twofold motivation. 
72: Firstly, there are systems like oil reservoirs  
73: that show multifractal properties \cite{Hermann}
74:  and are good candidates to be modeled by such 
75: object. Secondly, there is indeed a much more general scope: 
76: we want to study percolation phenomena 
77: in lattices that are not regular, but that 
78: are multifractal in the geometrical 
79: sense. It is important to know how site percolation transition  happens in 
80:  lattices in which the cells vary in size and also in the number of neighbors.
81: 
82: 
83: 
84: In this work we analyse some geometric and topologic properties  of  
85: the percolation cluster generated  on $Q_{mf}$ at the percolation threshold. 
86:  The paper is organized as follows: in 
87:  \S2 we present the process of construction 
88: of $Q_{mf}$ and the algorithm for the estimation of $p_c$,
89: in \S3 we show the numerical simulations concerning  the 
90: percolation threshold $p_c$ and the topologic properties of $Q_{mf}$; and  
91:  finally in \S4 we present our final remarks and comments.
92: 
93: 
94: 
95: \section{2- The Model}
96: 
97: In this section we show the process of building of the multifractal $Q_{mf}$ 
98: and the key concepts to estimate $p_c$. 
99:  We start with a square of linear size $1$ and a partition parameter
100: $0<\rho<1$. For reasons that will be clear
101:  later, $\rho = \frac{s}{r}$, where $s$ and 
102: $r$ are integers.  We call the multifractal built from this 
103: parameter $\rho$ as the "$\rho-Q_{mf}$", or "$(r,s)-Q_{mf}$". 
104: 
105:   
106: 
107: The first step, $n=1$, consists of two sections of 
108: the square: a vertical and an horizontal. 
109: Initially the square is cut in two pieces of 
110: area $\frac{r}{s+r}= \frac{1}{1+\rho}$ and 
111: $\frac{s}{s+r}= \frac{\rho}{1+\rho}$ by a vertical line.
112: This process is shown in figure \ref{fig1}(a), where we use as an example 
113: $\rho = \frac{s}{r} = \frac{2}{3}$.
114:  The horizontal cut  in which we use the same partition $\rho$ 
115:  is shown in figure \ref{fig1}(b). 
116: The first partition of the square generates
117: four rectangular blocks: the largest one of area $(\frac{\rho}{1+\rho})^2$,
118: two  of area $\frac{\rho}{(1+\rho)^2}$ and the
119: smallest one of area $(\frac{1}{1+\rho})^2$. 
120: The difference between the largest area and the smallest one increases 
121: as $\rho \rightarrow 0$, moreover, the blocks get 
122: more and more stretched in this limit. 
123: Therefore $\rho$ measures the anisotropy of 
124:  $Q_{mf}$. 
125: 
126: 
127: In the second step, $n=2$, we 
128: repeat the same process of vertical and horizontal 
129: sections as in step $1$. Generically we get $2^{2 n}$ blocks after 
130: the $n^{th}$-step. 
131: The partition process produces a set of blocks with a variety of areas. 
132:  We call a set of all elements with the same area as 
133: a $k$-set. 
134:  At the $n^{th}$-step of the algorithm the 
135: partition of the square in  blocks
136:  follows the binomial rule:
137: \begin{equation}
138:          A= \sum_{k=0}^n \> C_k^n \> \left(\frac{\rho}{1+\rho}\right)^k 
139: \left(\frac{1}{1+\rho}\right)^{n-k} = 
140: \left(\frac{\rho + 1}{\rho + 1}\right)^n = 1.
141: \label{bino2}
142: \end{equation}
143: The number of elements of a $k$-set is $C_k^n$.  
144: In reference \cite{last} we see that as $n \rightarrow \infty$ all the $k$-sets 
145: determine a monofractal whose dimension is $d_k = lim_{n \rightarrow \infty} \frac{log \> C_n^k \>
146: \> s^k \>r^{(n-k)} }{log \> (s+r)^\frac{n}{2} }$. In this limit, the ensemble of all
147: $k$-sets engenders the multifractal object $Q_{mf}$ itself. 
148: 
149: Figure \ref{fig2} shows a picture of $Q_{mf}$ 
150: for $\rho = \frac{1}{3}$ and $n=4$. We use 
151: the following code color: blocks of equal area have the same tonality. In 
152: other words, all the blocks of a same $k$-set 
153: share a common gray-tone. 
154: The general view of the object shows an anisotropic, heterogeneous lattice with 
155: a non-trivial topology. The anisotropy of $Q_{mf}$ will change the percolation 
156: threshold as we investigate in the next section. Before 
157: that we study in more detail the problem of percolation. 
158: 
159: 
160: 
161: The main subject of this work is the study of percolation properties of  
162: $Q_{mf}$. To perform such a task we develop a percolation algorithm.  
163: The percolation algorithm for  $Q_{mf}$
164: start mapping this object into the
165: square lattice. The square lattice should be large enough that
166: each line segment of $ Q_{mf}$
167: coincides with a line of the lattice, this condition 
168: imposes that $\rho$ is a rational number 
169: (it means, $r$ and $s$ are integers). 
170: Therefore we consider that the square lattice is more
171: finely divided than $Q_{mf}$. In this way all blocks
172:  of the multifractal are
173: composed by a finite number of cells of the square lattice. 
174: To explain the percolation algorithm we suppose that  $Q_{mf}$
175: construction is at step $n$.
176: We proceed the percolation algorithm by choosing at
177: random one among the $2^{2 n}$ blocks of
178: $ Q_{mf}$ independent of its size or number of neighbors. 
179: Once a block is chosen all the  cells
180: in the square lattice corresponding to this block are considered as
181: occupied. Each time a block of $ Q_{mf}$ is chosen
182: the algorithm check if the occupied cells at the underlying square 
183: lattice are connected in such a way to form a spanning  percolation
184: cluster. The algorithm to check percolation is similar to the one used in
185:  \cite{ziff,freitas0,freitas2,freitas3}. 
186: 
187: \section{3- Numerical Results}
188:  
189: In this section we show numerical results concerning the percolation 
190: threshold. Before the numerics we introduce some definitions.  
191: We call lattice the square lattice underlying  $Q_{mf}$. 
192: Following the literature \cite{stauffer} we call $p$ the probability 
193: of occupation of 
194: a lattice site. $R_L$  is the probability that for a site occupation $p$ there exists a 
195: contiguous cluster of occupied sites which crosses completely  the square lattice 
196: of size $L$.  
197: $p_c$ is the probability of occupation at the percolation threshold. 
198: There are several ways \cite{ziff} to define $R_L$. We use two of them: 
199: $R_L^e$  is the probability that there exits a cluster crossing  
200: either the horizontal or the vertical direction, and $R_L^b$ is the 
201: probability that there exits a cluster crossing  
202: both directions. At the limit 
203: of infinite lattice size $R_L^e$ and $R_L^b$ converge to a common value 
204: for the square lattice case. Besides we call $p_c^e$ the value of 
205: $p_c$ estimated from $R_L^e$, it means, the average $p_c$ over lattices that 
206: percolate in one direction, the horizontal or the vertical. And, similarly, 
207: $p_c^b$ the value of $p_c$ estimated from $R_L^b$. 
208: 
209: 
210:   
211: 
212: Figure \ref{fig3} illustrates the behavior of $p_c^e$ and $p_c^b$ for $7$ different 
213: lattice sizes. Two typical values of $\rho$ are used: $\rho = \frac{2}{3}$ 
214: (solid line) and  $\rho = \frac{1}{4}$ (dashed line). In \ref{fig3} 
215: (a) we show both $p_c$ versus 
216: $\frac{1}{L}$, the upper values corresponds to $p_c^b$ 
217:  and the bottom values to $p_c^e$. The lattice size, $L=(r+s)^n$, 
218: corresponds to  $4 \leq n \leq 10$, in both cases $r+s=5$.  The data of the 
219: upper branch of the figure  collapse into 
220: a single curve, diversly from the lower branch. In other words, 
221: the curves of $(2,3)-Q_{mf}$ and $(4,1)-Q_{mf}$ share roughly the same 
222: $p_c^e$ but diverse $p_c^b$. The reason for this feature is the strong 
223: anisotropy of  $(4,1)-Q_{mf}$ compared to the case 
224: $(2,3)-Q_{mf}$. The anisotropy of the percolation 
225: cluster does not affect $p_c$ in both  
226:  directions, but it affects $p_c$ in one direction.  
227: In fact, the percolation in both 
228: directions comes from an average over both directions, in such situation, 
229: any eventual anisotropy of the percolation cluster vanishes  
230: because of the average. 
231: 
232: An useful way to define  $p_c$ is to take the average 
233: value $p_{c_{ave}} = \frac{p_c^e + p_c^b}{2}$.  
234: In figure \ref{fig3}(b) we plot $p_{c_{ave}}$ versus $\frac{1}{L}$ 
235: corresponding to the same data figure \ref{fig3}(a). We observe in this figure that 
236: both curves converge to a saturation value that is not the same. 
237: The difference between the two cases is  related to the curve 
238: of $p_c^b$. The anisotropy (stretching of the blocks) due to  
239: $\rho$ implies an anisotropy in the percolation cluster. Such anisotropy 
240: determines that the percolation cluster does not have a 
241: correlation length independent of the direction. Because the percolation 
242: cluster is anisotropic the symmetry between $p_c^e$ and $p_c^b$ should 
243: fail. Therefore, we expect that for smaller $\rho$ (more anisotropy in 
244: the multifractal) a greater difference in $p_{c_{ave}}$ will appear. 
245: Figure \ref{fig4} confirm this tendency for diverse values of $\rho$. 
246: 
247: Figure \ref{fig4} shows  $\bar p_c$ versus $\rho$. To obtain  $\bar p_c$ we  
248: make an average of $p_{c_{ave}}$ after the saturation process, it means, 
249: for $n \geq 8$. The  parameter  
250:  $\rho$ is indicated in the figure. 
251: These values are also shown in Table $I$.  The straight line 
252: in the figure is the linear fitting of the data.
253:  We focus our attention on two 
254: features of the figure: the general tendency of decreasing 
255: $p_c$ with $\rho$, 
256: and the anomalous case $(3,1)-Q_{mf}$. The main tendency of 
257: decreasing $p_{c}$ with $\rho$ we have discussed in connection 
258: with anisotropy.  We comment the anomalous situation  
259: of $(3,1)-Q_{mf}$ in relation to 
260:  topologic properties of $Q_{mf}$ which  are analyzed in 
261: what follows. 
262: 
263: 
264: The object $Q_{mf}$ is build of a set of blocks $i$ with different areas,
265:  $A_i$, and number of neighbors, $\zeta_i$. The number  $\zeta_i$ is a central 
266: quantity in the study of topology.  
267: A topologic quantity, that is important in the investigation of the percolation 
268: threshold, is $\zeta_{ave}$ which is an average of $\zeta_i$  over 
269:  $Q_{mf}$. It means $\zeta_{ave} = \frac{\sum_{i} \zeta_i }{N} $ 
270: where $N=2^{2\>n}$ is the total 
271: number of blocks and the sum in the numerator is performed over the full 
272: multifractal.  Typically $\zeta_{ave}$ is a number that does not 
273: increase with  $n$, but attains a constant value after  
274: $n \approx 6$. We plot in Table $I$ this topologic quantity for $n=8$ 
275: and several values of $\rho$. 
276: 
277: \centerline{\it Table I}
278: \begin{tabbing}
279:   $(s,r)$      \hspace{0.6cm}        \= (1,1) \hspace{0.7cm} \= (4,3) \hspace{0.7cm}\=      (3,2) \hspace{0.7cm} \= (2,1) \hspace{0.7cm} \= (5,2)     \hspace{0.7cm} \=(3,1)      \hspace{0.7cm} \= (4,1)   \hspace{0.7cm} \= (5,1)\\
280: \hspace{0.2cm} $p_{c}$  \hspace{0.6cm} \= 0.5929 \hspace{0.4cm} \= 0.5262  \hspace{0.4cm} \= 0.5262  \hspace{0.4cm} \= 0.5256   \hspace{0.4cm} \=  0.5252   \hspace{0.4cm} \= 0.5253  \hspace{0.4cm} \= 0.5243   \hspace{0.5cm} \= 0.5241 \\
281: \hspace{0.2cm} $ \zeta_{ave} $   \hspace{0.6cm} \= 999 \hspace{0.6cm} \= 5.436 \hspace{0.7cm} \= 5.436  \hspace{0.6cm}  \= 5.434 \hspace{0.7cm} \= 5.436    \hspace{0.6cm} \= 5.426  \hspace{0.7cm} \= 5.436  \hspace{0.6cm} \= 5.436 \\
282: 
283: \end{tabbing}
284: 
285: We observe in the row of $ \zeta_{ave} $ that all these values are roughly the same, 
286: the exception corresponds to the case of the $(3,1)-Q_{mf}$. 
287: The conclusion we take is that the variation of the average 
288: number of neighbors causes the fluctuation in percolation threshold observed in 
289: figure \ref{fig4}. The $p_c$ 
290: of the $(3,1)-Q_{mf}$ is a little bit greater than the other cases 
291: because it has a smaller average number of neighbors. 
292: In other words, because each 
293: block of the $(3,1)-Q_{mf}$, in the average, has less neighbors, 
294: it percolates with more difficulty than 
295: the tendency among its group. 
296:  
297: 
298: 
299: \section{4- Final Remarks}
300: 
301: To summarize we analyze in this work the role of the anisotropy and the 
302: average number of neighbors of $Q_{mf}$ on its 
303: percolation threshold, $p_c$. The multifractal object is 
304: composed by a set of blocks with different areas and number of neighbors. 
305: As  the parameter defining $Q_{mf}$,  $\rho$, goes to 
306: zero, the multifractal becomes more and more anisotropic. This 
307: anisotropy reflects in the percolation cluster creating an asymmetry 
308: between $p_c^e$ and $p_c^b$. 
309: 
310: The anisotropy of  $Q_{mf}$ is evident when we compare 
311: $p_c^e$ and $p_c^b$.  The observed 
312: curves of $p_c^e$ show a similar behavior, in contrast to 
313: the curves of  $p_c^b$. Actually, the measure of $p_c^b$ makes 
314: an average over both directions which erases any anisotropic 
315: effect of the percolation cluster. This erasing effect does not 
316: exist when we measure $p_c^e$. The anisotropy of  $Q_{mf}$ 
317: decreases with $\rho$ (with $\rho \rightarrow 0$ the blocks 
318: became more stretched). 
319:  
320: 
321: A special case in our analysis is the  $(3,1)-Q_{mf}$. This 
322: case is singular compared to others analyzed cases. 
323: For a same number of blocks, 
324: the $(3,1)-Q_{mf}$ case  has less neighbors than the multifractal blocks  
325: corresponding to other values of $\rho$. 
326: This phenomenon is intrinsic to the topology 
327: of  $Q_{mf}$.  It implies that the $(3,1)-Q_{mf}$  
328: is less connected, and, as a consequence, it shows more difficulty to  
329: percolate. Therefore the case of the $(3,1)-Q_{mf}$ 
330: has a percolation threshold 
331: slighter greater then its neighbors in the $\rho$ sequence. 
332: In a future work we intend to 
333: study in more detail the effect of other topologic properties on 
334: the percolation properties. 
335: 
336: 
337: 
338: 
339: \vspace{2cm}
340: 
341: The authors gratefully acknowledge the financial support of Conselho Nacional
342: de Desenvolvimento Cient{\'\i}fico e Tecnol{\'o}gico (CNPq)-Brazil, 
343: FINEP and CTPETRO.
344: 
345: 
346: \vspace{1cm}
347: \begin{thebibliography}{99}
348: %
349: %
350: %\bibitem{perc1} A. Bunde and H. Havlin (eds.) {\it Fractals and Disordered Systems}
351: %(Springer, Berlin, 1991).
352: 
353: \bibitem{stauffer} D. Stauffer and A. Aharony,
354: {\it Introduction to Percolation Theory}, (Taylor $\&$ Francis, London, 1994)
355: %
356: 
357: \bibitem{perc3} M. Sahimi, {\it Applications of Percolation Theory},
358:  (Taylor \& Francis, Bristol, 1994).
359: 
360: \bibitem{multperc1} K.P.N. Murthy, L.K. Gallos, P. Argyrakis, K. N. Kehr, 
361: Phys. Rev E, {\bf 54} 6922 (1996).  
362: 
363: \bibitem{multperc2} O. Stenull and H. K. Jenssen, Phys. Rev. E, {\bf 65} 045104 
364: (2002). 
365: 
366: \bibitem{last} G. Corso, J. E. Freitas, L. S. Lucena, R. F. Soares, 
367: cond-mat/0212530.
368: 
369: \bibitem{Hermann} F. Herrmann, {\it A Scaling Medium Representation,
370: a Discussion on Well-logs, Fractals and Waves.} PhD Thesis,
371: Delft University of Technology, (1997).
372: 
373: 
374: \bibitem{ziff} M. E. J. Newman and R. M. Ziff, Phys. Rev. Lett. {\bf 85}
375: 4104 (2000). 
376: %
377: %\bibitem{bp2} L. S. Lucena, L. R. da Silva and S. Roux, Physica A {\bf 266}
378: %86 (1999).
379: 
380: \bibitem{freitas0} J. E. de Freitas, L. S. Lucena and S. Roux, Physica A
381: {\bf 266} 81 (1999).
382: 
383: 
384: \bibitem{freitas2} J. E. de Freitas and L. S. Lucena,
385: IJMPC, {\bf 11} 1581 (2000).
386: 
387: \bibitem{freitas3} J. E. de Freitas, L. S. Lucena, and S. Roux, Phys. Rev. E, 
388: {\bf 64}, 051405 (2001). 
389: 
390: %\bibitem{Hewet} T. A. Hewett, {\it Fractal Distribution of Reservoir Heterogeneity and
391: %Their Influence on Fluid Transport,} 61${th}$ Annual Technical Conference and Exhibition on the
392: %Society of Petroleum Engineers, (1998).
393: %
394: 
395: 
396: %
397: 
398: \end{thebibliography}
399: %
400: 
401: \vspace{1cm}
402: 
403: \centerline{FIGURE LEGENDS}
404: 
405: \begin{figure}[ht]
406: \begin{center}
407: \caption{ The initial step, $n=1$, in the formation of $Q_{mf}$. 
408: In (a) a vertical line cut the
409: square in two pieces according to  $\rho$. 
410: Two horizontal lines sectioning the rectangles by
411:  the same ratio are depicted in (b). The underlying square lattice 
412: is depicted  with thick lines. }
413: \label{fig1}
414: \end{center}
415: \end{figure}
416: %
417: \begin{figure}[ht]
418: \begin{center}
419: \caption{The object $Q_{mf}$ for $\rho = \frac{1}{3}$ and $n=4$, we 
420: introduce a square inside the figure to help the visualization. }
421: \label{fig2}
422: \end{center}
423: \end{figure}
424: %
425: \begin{figure}[ht]
426: \begin{center}
427: \caption{In (a) we show $p_c$ versus $\frac{1}{L}$ for  
428: $\rho = \frac{2}{3}$ (solid line) and 
429:  $\rho = \frac{1}{4}$ (dashed line). The upper values correspond to $p_c^b$ 
430:  and the lower values to $p_c^e$; it is used $4 \leq n \leq 10$. 
431: In (b) we plot $p_{ave}$ 
432: versus $\frac{1}{L}$ for the same data. }
433: \label{fig3}
434: \end{center}
435: \end{figure}
436: %
437: %
438: \begin{figure}[ht]
439: \begin{center}
440: \caption{The  $\bar p_c$ versus $\rho$. 
441: The chosen values 
442: of $\rho$ are indicated in the figure. The straight line 
443: corresponds to the linear fitting.}
444: \label{fig4}
445: \end{center}
446: \end{figure}
447: %
448: 
449: %
450: %
451: 
452: %
453: 
454: %
455: 
456: 
457: \end{document}
458: 
459: 
460: 
461: 
462: