1: %\documentstyle[preprint,aps,epsf,epsfig]{revtex}
2: \documentstyle[aps,multicol,epsf,epsfig]{revtex}
3:
4: \newcommand{\xsize}{\epsfxsize=14.0cm}
5:
6: \begin{document}
7:
8: \draft
9:
10: \title{Beyond Blobs in Percolation Cluster Structure: The Distribution\\
11: of 3-Blocks at the Percolation Threshold}
12:
13: \author{Gerald Paul\thanks{Electronic address: gerryp@bu.edu} and H. Eugene Stanley}
14:
15: \address{Center for Polymer Studies and Department of Physics\\
16: Boston University, Boston, MA 02215 USA}
17:
18: \date{ps.tex ~~ 7 February 2002}
19:
20: \maketitle
21:
22: \begin{abstract}
23:
24: The incipient infinite cluster appearing at the bond percolation
25: threshold can be decomposed into singly-connected ``links'' and
26: multiply-connected ``blobs.'' Here we decompose blobs into objects known
27: in graph theory as 3-blocks. A 3-block is a graph that cannot be
28: separated into disconnected subgraphs by cutting the graph at 2 or fewer
29: vertices. Clusters, blobs, and 3-blocks are special cases of $k$-blocks
30: with $k=1$, 2, and 3, respectively. We study bond percolation clusters
31: at the percolation threshold on 2-dimensional square lattices and
32: 3-dimensional cubic lattices and, using Monte-Carlo simulations,
33: determine the distribution of the sizes of the 3-blocks into which the
34: blobs are decomposed. We find that the 3-blocks have fractal dimension
35: $d_3=1.2\pm 0.1$ in 2D and $1.15\pm 0.1$ in 3D. These fractal dimensions are
36: significantly smaller than the fractal dimensions of the blobs, making
37: possible more efficient calculation of percolation
38: properties. Additionally, the closeness of the estimated values for
39: $d_3$ in 2D and 3D is consistent with the possibility that $d_3$ is
40: dimension independent. Generalizing the concept of the backbone, we
41: introduce the concept of a ``$k$-bone'', which is the set of all points
42: in a percolation system connected to $k$ disjoint terminal points (or
43: sets of disjoint terminal points) by $k$ disjoint paths. We argue that
44: the fractal dimension of a $k$-bone is equal to the fractal dimension of
45: $k$-blocks, allowing us to discuss the relation between the fractal
46: dimension of $k$-blocks and recent work on path crossing probabilities.
47:
48: \end{abstract}
49:
50: %\pacs{}
51:
52: \begin{multicols}{2}
53:
54: \section{Introduction}
55:
56: Percolation is the classic model for disordered systems
57: \cite{Ben-Avraham00,Stauffer94,Bunde96}. For concreteness we will study
58: bond percolation systems in which bonds on a lattice are randomly
59: occupied with probability $p$. Clusters are defined as groups
60: of sites and bonds which are connected by occupied bonds. Clusters can
61: be decomposed into objects known as blobs. Blobs are sets of sites and
62: bonds which cannot be decomposed into disconnected sets by cutting only
63: one bond. Equivalently blobs are sometimes described as being
64: multiply-connected---there are at least two disjoint paths between each
65: point in a blob and every other point in the blob. The decomposition of
66: the entire percolation cluster into blobs has been extensively studied
67: \cite{Gyure95}, as has the distribution of sizes of blobs in the
68: backbone \cite{Herrmann84}. For both cluster and backbone blobs, the
69: fractal dimension of the blobs is the fractal dimension of the backbone.
70:
71: Here we address the questions of (i) whether there are more fundamental
72: objects into which blobs can be decomposed, and (ii) whether these
73: objects then be further decomposed. To answer these questions, we
74: employ the language of graph theory, in which sites are the vertices and
75: bonds are the edges of a graph \cite{Tutte84}.
76:
77: One can define $k$-connected graphs (or $k$-blocks) as graphs which
78: cannot be separated into disconnected subgraphs by cutting the graph at
79: fewer than $k$ vertices \cite{Tutte84,Text1}. Thus, clusters are
80: 1-blocks and blobs are 2-blocks. The natural next level of
81: decomposition of percolation systems is to decompose blobs(2-blocks)
82: into 3-blocks. By the definition above, 3-blocks are graphs which cannot
83: be decomposed by cutting the graphs at fewer than 3 vertices. From a
84: physicist's point of view, one can understand what 3-blocks are by
85: considering a blob as a resistor network with each bond being a
86: resistor. Assume one is trying to determine the resistance between 2
87: vertices of the network. One can simplify the network by using
88: Kirchoff's Laws to replace groups of sequential bonds and groups of
89: parallel bonds by single virtual bonds having resistance equivalent to
90: the bonds replaced. After this has been done as completely as possible,
91: what are left are 3-blocks. We define the mass of a 3-block as the
92: number of virtual bonds plus the number of non-replaced original bonds
93: remaining in the 3-block. Figures \ref{tpoDecomp} and \ref{preal}
94: provide examples of the decomposition of a blob into 3-blocks. It
95: has been shown \cite{Tutte84} that the decomposition of 2-blocks into
96: 3-blocks is unique.
97:
98: Determining the scaling of the distribution of the 3-blocks into which
99: the 2-blocks can be decomposed is the subject of this paper. In graph
100: theory, the sites are typically not constrained to a lattice structure,
101: and one is only concerned with the topology of the graphs; we will,
102: however, work on square and cubic lattices.
103:
104: \section{Notation}
105:
106: Because we deal with a number of different types of fractal objects, we
107: employ the following notation:
108:
109: \begin{itemize}
110:
111: \item[{(i)}] The fractal dimension of an object of type $X$ will be
112: denoted as $d_X$.
113:
114: \item[{(ii)}] The number distribution of objects of type $X$ in space
115: $Y$ of size $L$ will be denoted as $n(N_X,L_Y)$.
116:
117: \item[{(iii)}] The exponent of the power-law regime of a distribution of
118: objects of type $X$ in space of type $Y$ will be denoted as
119: $\tau_{X,Y}$.
120:
121: \item[{(iv)}] The amplitude of a distribution of objects of type $X$ in
122: space of type $Y$ will be denoted as $A_{X,Y}$.
123:
124: \item[{(v)}] We define $d_{nY}$ through the relation
125:
126: \begin{equation}
127: \label{ex10}
128: \langle n(L)\rangle\sim L^{d_{nY}},
129: \end{equation}
130: %
131: where $\langle n(L)\rangle$ is the average number of disjoint objects of
132: a given type in space $Y$.
133:
134: \item[{(vi)}] We use $0,1,2,3\ldots$ to denote $k$-blocks with
135: $k=0,1,2,3\ldots$ corresponding to Euclidean space, clusters, blobs, and
136: 3-blocks respectively. We use $B$ to denote the percolation cluster
137: backbone.
138:
139: \item[{(vii)}] Additionally, because, as noted above, objects such as
140: 3-blocks can be nested, we denote quantities that relate to all levels
141: of nesting with an asterisk. Specifically, $\tau_{X,Y}^\ast$ and
142: $A_{X,Y}^\ast$ denote the exponent of the power-law regime and the
143: amplitude of a distribution of nested objects of type $X$ at all levels
144: of nesting in space of type $Y$. Similarly, $d_{nY}^\ast$ is defined
145: through the relation
146: %
147: \begin{equation}
148: \label{ex10a}
149: \langle n^\ast(L)\rangle\sim L^{d_{nY}^\ast},
150: \end{equation}
151: %
152: where $\langle n^\ast(L)\rangle$ is the average number of nested objects
153: at all levels of nesting of a given type in space $Y$. Quantities not
154: qualified with an asterisk will denote quantities at a single level or
155: quantities which cannot be nested.
156:
157: \end{itemize}
158:
159: Using this notation, previous results are \cite{Herrmann84}
160: %
161: \begin{equation}
162: \label{ex20}
163: n(N_2,L_B)=A_{2,B}L^{d_{nB}}N_2^{-\tau_{2,B}}f_L\left({N_2\over
164: L^{d_2}}\right)
165: \end{equation}
166: %
167: for the number distribution of blobs of mass $N_2$ in the percolation
168: cluster backbone and \cite{Gyure95}
169: %
170: \begin{equation}
171: \label{ex21}
172: n(N_2,L_1)=A_{2,1}L^{d_{n1}}N_2^{-\tau_{2,1}}f_L\left({N_2\over
173: L^{d_2}}\right)
174: \end{equation}
175: %
176: for the number distribution of blobs of mass $N_2$ in the whole
177: percolation cluster. The finite-size scaling function $f_L(x)$ in
178: Eqs.~(\ref{ex20}) and (\ref{ex21}) approaches 0 when $x>1$ and is 1
179: otherwise.
180:
181: In analogy with Eqs.~(\ref{ex20}) and (\ref{ex21}) we expect the number
182: distribution of 3-blocks at all levels of nesting in a blob to be
183: %
184: \begin{equation}
185: n^\ast(N_3,L_2)=A_{3,2}^\ast L^{d_{n2}^\ast}N_3^{-\tau_{3,2}^\ast}f_c
186: \left({N_3\over c}\right)f_L\left({N_3\over L^{d_2}}\right),
187: \label{e1}
188: \end{equation}
189: %
190: where $c$ is the mass of the smallest 3-block and the finite-size
191: scaling function $f_c(x)$ approaches 0 when $x<1$ and is 1 otherwise,
192: reflecting the fact that there cannot be any 3-blocks smaller than the
193: smallest size $c$. In all dimensions and for all lattices, $c=5$. For
194: simplicity we will approximate $n^\ast(N_3,L_B)$ as
195: %
196: \begin{equation}
197: n^\ast(N_3,L_2)=\cases{
198: A_{3,2}^\ast L^{d_{n2}^\ast}N_3^{-\tau_{3,2}^\ast} & $c\leq N_3\leq aL^{d_2}$\cr
199: 0 & otherwise.}
200: \label{e2}
201: \end{equation}
202: %
203:
204: \section{Simulations}
205:
206: We perform simulations with $p=0.5$, the exact percolation threshold
207: for 2D \cite{Stauffer94,Bunde96} and $p=0.2488126$, the most precise
208: current estimate for the percolation threshold for 3D
209: \cite{LorenzXX}. We created percolation clusters which included the
210: sites $(0,L/2)$ and $(L,L/2)$ for the 2D simulations and the sites
211: $(0,L/2,L/2)$ and $(L,L/2,L/2)$ for the 3D simulations, decomposed the
212: backbones determined by these sites into blobs and then decomposed the
213: blobs into 3-blocks. We study both distributions of 3-blocks in blobs of
214: given mass, $N_2$, and distributions of 3-blocks in backbones in systems
215: of a given size, $L$. For purposes of analysis, we group together blobs
216: with mass $2^{m-1}<N_2<2^m$.
217:
218: We perform the decomposition into 3-blocks along the lines of the
219: procedure sketched in Ref.~\cite{Tutte84}. Basically, this procedure is
220: as follows: We first designate the blob that we are decomposing as the
221: 2-block graph $G$. The natural next level of decomposition is to
222: identify connected subgraphs with two or more edges that are connected
223: to $G$ at only two vertices. We denote these subgraphs
224: $G_1,G_2,G_3,\ldots$ of $G$ as 2-terminal objects. These 2-terminal
225: objects can then be replaced in $G$ by ``virtual edges,''
226: $e_1,e_2,e_3,\ldots$ Note that this process can be continued
227: recursively. That is, the subgraph $G_i$ may itself contain sub-graphs,
228: $G_{i1},G_{i2},G_{i3},\ldots$ that are connected to $G_i$ at only two
229: vertices; we then replace the subgraphs $G_{ij}$ in $G_i$ by virtual
230: edges $e_{Gij}$. The process continues until the only remaining
231: subgraphs are those that cannot be decomposed further by making cuts at
232: two vertices; these, by definition, are 3-blocks. An example of this
233: decomposition is shown in Fig.~\ref{tpoDecomp}. Other methods of
234: decompostion into 3-blocks are described in
235: Refs.~\cite{Hohberg92,Hopcroft73}.
236:
237: We perform at least 3700 realizations for each system size; for the
238: smaller system sizes for which the simulations run more quickly we
239: performed as many as $10^8$ realizations. Because, the larger the
240: systems the larger the number of 3-blocks contained in the system, the
241: statistics for the larger systems was acceptable despite the lower
242: number of realizations. We bin the results for all system sizes in
243: order to smooth the plots.
244:
245: \section{Two Spatial Dimensions}
246:
247: In this section we discuss our results for 3-blocks in 2D
248: percolation. Results in 3D are analogous and are discussed in the next
249: section.
250:
251: \subsection{3-blocks in Blobs}
252:
253: Figure~\ref{pAll3}(a) plots the distributions $P^\ast(N_3|N_2)$, the
254: probability that a 3-block contained in a blob of size $N_2$ contains
255: $N_3$ bonds, for various values of $N_2$. $P(N_3|N_2)$ is the number
256: distribution $n^\ast(N_3,N_2)$ normalized to unity. Consistent with
257: Eqs.~(\ref{e1}) and (\ref{e2}), the plots exhibit power-law regimes
258: followed by cut-offs due to the finite size of the blobs. The ``bumps''
259: in the distributions right before the cutoffs represent 3-blocks which
260: would have been larger but are truncated due to the finite size of the
261: blobs in which they are embedded. We estimate the slope of the power law
262: regimes, $\tau_{3,2}^\ast$, to be $2.35\pm 0.05$. Since
263: %
264: \begin{equation}
265: N_3\sim L^{d_3}
266: \label{e101}
267: \end{equation}
268: %
269: and
270: %
271: \begin{equation}
272: N_2\sim L^{d_2}
273: \label{e102}
274: \end{equation}
275: %
276: we expect
277: %
278: \begin{equation}
279: N_3\sim {N_2}^{d_3/d_2}.
280: \label{e103}
281: \end{equation}
282: %
283:
284: In Fig.~\ref{pAll3}(b), we show the collapsed plots in which we scale
285: the distributions by ${N_2}^{d_3/d_2}$ using the most precise published
286: estimate for $d_2$, $1.6432\pm 0.0008$ \cite{Grassberger99}. (A
287: consistent more recent estimate, $d_2=1.6431\pm 0.0006$, is given in
288: Ref.~\cite{JacobsenXX}.) Visually, we find the best collapse is
289: obtained for $d_3=1.20\pm 0.1$.
290:
291: We can also estimate $d_3$ using Eq.~(\ref{e220}) from the appendix
292: %
293: \begin{equation}
294: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
295: 3,2}}^\ast-1)=d_{n2}^\ast=d_2.
296: \label{e104}
297: \end{equation}
298: %
299:
300: Using $\tau_{3,2}^\ast=2.35\pm 0.05$ and $d_2=1.6432\pm 0.0008$, results in
301: an estimate of $d_3=1.22\pm 0.05$.
302:
303:
304:
305: \subsection{3-blocks in Backbone}
306:
307:
308: Figure~\ref{pc}(a) plots the distributions $P^\ast(N_3|L_B)$, the
309: probability that a 3-block contained in the backbone of a system of size
310: $L$ contains $N_3$ bonds, for various values of $L$. $P^\ast(N_3|L_B)$
311: is the number distribution $n^\ast(N_3,L_B)$ normalized to unity. Consistent
312: with Eqs.~(\ref{e1}) and (\ref{e2}), the plots exhibit power-law regimes
313: followed by cut-offs due to the finite size of the systems. We estimate
314: the slope of the power law regimes, $\tau_{3,B}^\ast$, to be $2.25\pm
315: 0.05$. In Fig.~\ref{pc}(b), we show the collapsed plots in which we scale
316: the distributions by $L^{d_3}$. Visually, we find the best collapse is
317: obtained for $d_3=1.15\pm 0.1$.
318:
319: Next we consider the distribution of ``top-level'' 3-blocks in the
320: backbone. Top-level 3-blocks are those not contained within another
321: 3-block. In Fig.~\ref{pc1}(a), we plot the distributions $P(N_3|L_B)$, the
322: probability that a top-level 3-block contained in the backbone of a
323: system of size $L$ contains $N_3$ bonds, for various values of $L$. The
324: plots exhibit power-law regimes followed by cut-offs due to the finite
325: size of the systems. The exponent of the power-law regimes $\tau_{3,B}$
326: is estimated to be $1.6\pm 0.05$. In Fig.~\ref{pc1}(b), we show the
327: collapsed plots, in which we scale the distributions by $L^{d_3}$. The
328: best collapse is obtained for $d_3=1.15\pm 0.1$, the same value as for
329: the distributions of 3-blocks of all levels. Thus the fractal dimensions
330: of the top level 3-blocks is the same as the fractal dimension of
331: 3-blocks of all levels but the slopes of the power law regimes are
332: different; this is seen also in Fig.~\ref{pcComb1}.
333:
334: We can also use Eq.~(\ref{e210})
335: %
336: \begin{equation}
337: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
338: 3,B}}-1)=d_{nB}={1\over\nu}.
339: \label{e106}
340: \end{equation}
341: %
342: to obtain an estimate of $d_3$. Since $d_{nB}$ is known exactly in 2
343: dimensions and has been well studied in higher dimensions and because
344: one can usually determine the slope $\tau_{3,B}$ more accurately than
345: $d_3$ can be determined by finding the best scaling collapse, we
346: determine $d_3$ more accurately by solving Eq.~(\ref{e106}) for
347: $d_3$. Using our estimate for $\tau_{3,B}$ above we find $d_3=1.25\pm
348: 0.1$. Combining this result with our earlier estimates, we make the final
349: estimate
350: %
351: \begin{equation}
352: d_3=1.20\pm 0.1.
353: \label{e900}
354: \end{equation}
355:
356:
357:
358:
359: \subsection{Why the Fractal Dimension of 3-blocks is Smaller than the
360: Fractal Dimension of the Backbone and 2-blocks}
361:
362: The fractal dimension of the 3-blocks is considerably smaller than the
363: fractal dimension, $d_B=1.6432\pm 0.0008$ \cite{Grassberger99}, of
364: 2-blocks (blobs). This is because virtual bonds are counted as one bond
365: even though they replace many bonds. This can be seen if we plot the
366: distributions $P^\ast(M_3|L_B)$, the probability that a 3-block
367: contained in the backbone of a system of size $L$ contains $M_3$ bonds
368: where we can count not the virtual bonds, but all bonds contained in a
369: 3-block. In Fig.~\ref{pe}(a) we plot $P^\ast(M_3|L_B)$ for various
370: $L$. The best collapse for these plots (Fig.~\ref{pe}(b)) corresponds to
371: a fractal dimension of $1.6\pm 0.1$ consistent with the fractal
372: dimension of 2-blocks in 2D. This can be understood as a reflection of
373: the fact that in a system of size $L$, the mass of the largest
374: 3-block(counting all bonds) can be the same as the backbone mass. This
375: is similar to the situation with blobs and backbones; the largest blob
376: in a backbone can be as large as the whole backbone, which explains why
377: the fractal dimension of blobs is the same as the fractal dimension of
378: the backbone.
379:
380: Replacing a group of bonds by a virtual bond is analogous to
381: removing dangling ends on a cluster when determining the backbone.
382:
383:
384: \section{Three Spatial Dimensions}
385:
386: Our analysis of the results of the 3D simulations proceeds in a similar
387: manner to the analysis for 2D.
388:
389: \subsection{3-blocks in Blobs}
390:
391: Figure~\ref{pAll33d}(a) plots the distributions $P^\ast(N_3|N_2)$, the
392: probability that a 3-block contained in a blob of size $N_2$ contains
393: $N_3$ bonds, for various values of $N_2$. We estimate the slope of the
394: power law regimes, $\tau_{3,2}^\ast$, to be $2.63\pm 0.05$. In
395: Fig.~\ref{pAll33d}(b), we show the collapsed plots in which we scale the
396: distributions by ${N_2}^{d_3/d_2}$ with $d_2=1.87\pm 0.03$
397: \cite{Porto97}. Visually, we find the best collapse is
398: obtained for $d_3=1.15\pm 0.1$.
399:
400: Estimating $d_3$ using Eq.~(\ref{e220}) from the appendix
401: %
402: \begin{equation}
403: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
404: 3,2}}^\ast-1)=d_{n2}^\ast=d_2.
405: \label{e304}
406: \end{equation}
407: %
408: with $\tau_{\mbox{\scriptsize 3,2}}^\ast=2.63\pm 0.05$ and $d_2=1.87\pm
409: 0.03$, results in an estimate of $d_3=1.15\pm 0.05$.
410:
411:
412: \subsection{3-blocks in Backbone}
413:
414:
415: Figure~\ref{pc3d}(a) plots the distributions $P^\ast(N_3|L_B)$, the
416: probability that a 3-block contained in the backbone of a system of size
417: $L$ contains $N_3$ bonds, for various values of $L$. We estimate
418: the slope of the power law regimes, $\tau_{3,B}^\ast$, to be $2.55\pm
419: 0.05$. In Fig.~\ref{pc3d}(b), we show the collapsed plots in which we scale
420: the distributions by $L^{d_3}$. Visually, we find the best collapse is
421: obtained for $d_3=1.15\pm 0.1$.
422:
423: Next we consider the distribution of ``top-level'' 3-blocks in the
424: backbone. In Fig.~\ref{pc13d}(a), we plot the distributions $P(N_3|L_B)$,
425: the probability that a top-level 3-block contained in the backbone of a
426: system of size $L$ contains $N_3$ bonds, for various values of $L$. The
427: exponent of the power-law regimes $\tau_{3,B}$ is estimated to be
428: $2.0\pm 0.05$. In Fig.~\ref{pc13d}(b), we show the collapsed plots, in
429: which we scale the distributions by $L^{d_3}$. The best collapse is
430: obtained for $d_3=1.15\pm 0.1$, the same value as for the distributions
431: of 3-blocks of all levels. As in 2D, the fractal dimensions of the top level
432: 3-blocks is the same as the fractal dimension of 3-blocks of all levels
433: but the slopes of the power law regimes are different; this is seen also
434: in Fig.~\ref{pcComb13d}.
435:
436: Using Eq.~(\ref{e210})
437: %
438: \begin{equation}
439: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
440: 3,B}}-1)=d_{nB}={1\over\nu}.
441: \label{e306}
442: \end{equation}
443: %
444: to obtain an estimate of $d_3$ with our estimate for $\tau_{3,B}$ above
445: we find $d_3=1.14\pm 0.1$. Combining this result with our earlier
446: estimates, we make the final estimate
447: %
448: \begin{equation}
449: d_3=1.15\pm 0.1.
450: \label{e9}
451: \end{equation}
452:
453: The simulation results notwithstanding, it would be surprising if $d_3$
454: were smaller in 3D than in 2D because, below the critical dimension
455: $d_c=6$, both the fractal dimensions of clusters and blobs increase with
456: the Euclidean dimension. This suggests that while the actual values of
457: $d_3$ may be within the bounds we have estimated, the actual values will
458: be consistent with $d_3~(\mbox{2D})\leq d_3~(\mbox{3D})$.
459:
460: As in 2D, if we do not replace two-terminal objects in a 3-block by a
461: single virtual bond, the fractal dimension of the 3-block is that of a
462: blob(see Fig. \ref{pe3d}).
463:
464: Estimates for all of the 2D and 3D exponents are summarized in Table I.
465:
466: \section{Decomposition of the Whole Percolation Cluster}
467:
468: While we have only decomposed 2-blocks that comprise the cluster
469: backbone, we could proceed similarly for all 2-blocks into which a
470: cluster is decomposed. The fractal dimension of the 3-blocks into which
471: a cluster is ultimately decomposed should be the same as the fractal
472: dimension of the 3-blocks into which the backbone is ultimately
473: decomposed. The only difference we would expect in our results would
474: be that the slope of the power-law regime of the distribution of
475: top-level 3-blocks would be given by
476: %
477: \begin{equation}
478: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize{
479: 3,1}}}-1)=d_{n1}=d.
480: \label{e209}
481: \end{equation}
482: %
483: the analogy of Eq.~(\ref{e210})
484:
485: \section{$k$-Bones and Path Crossing Probabilities}
486:
487: Just as blobs and backbones have the same fractal dimension, we can
488: identify objects analagous to backbones which have the same fractal
489: dimensions as $k$-blocks. We define a $k$-bone as the set of all points
490: in a percolation system connected to $k$ disjoint terminal points(or
491: sets of disjoint terminal points) by $k$ disjoint paths. Thus the
492: backbone is a $k$-bone with $k=2$. Just as the largest $k$-blocks into
493: which a backbone can be decomposed are 2-blocks, the largest $k$-blocks
494: into which a $k$-bone can be decomposed are $k$-blocks. The fractal
495: dimension of $k$-bones is the fractal dimension of the $k$-blocks. One
496: can see this easily by noting that if the $k$ terminal points which
497: define a $k$-bone are connected to each other, the resulting structure
498: is $k$-block.
499:
500: Recent work \cite{AizenmanXX} has identified a relationship between path
501: crossing probabilities and the fractal dimensions of percolation
502: structures. Specifically, consider the probability, $\hat P_k^P$ that in
503: an annular region the small inner circle of radius $r$ is connected to
504: the larger outer circle of radius $R$, $R\gg r$, by $k$ disjoint
505: paths. Then
506: %
507: \begin{equation}
508: \hat P_k^P\sim\left({r\over R}\right)^{\hat x_k}.
509: \label{e14a}
510: \end{equation}
511: %
512: It has been observed \cite{AizenmanXX} that $\hat x_1$ is the codimension
513: of the percolation cluster and $\hat x_2$ is the codimesion of the
514: backbone. We extend these observations to the case of general $k$
515: %
516: \begin{equation}
517: d-\hat x_k=d_k,
518: \label{e14b}
519: \end{equation}
520: %
521: where $d$ is the spatial dimension of the system. This should hold in
522: all dimensions where the annulus is now defined by two hyperspheres. It
523: has been argued \cite{JacobsenXX} that
524: %
525: \begin{equation}
526: x_k<\hat x_k<x_{2k},
527: \label{e14c}
528: \end{equation}
529: %
530: where $x_k$ is the polychromatic path crossing exponent
531: \cite{AizenmanXX} and which has been found rigorously in 2D to be
532: \cite{AizenmanXX}
533: %
534: \begin{equation}
535: x_k={1\over 12}(k^2-1).
536: \label{e14d}
537: \end{equation}
538: %
539: Using Eqs.~(\ref{e14b}), (\ref{e14c}), and (\ref{e14d}), we find in 2D
540: %
541: \begin{equation}
542: -{11\over 12}<d_3<{4\over 3},
543: \label{e14e}
544: \end{equation}
545: %
546: consistent with our estimate for $d_3$ in 2D.
547:
548: The relationship between the path crossing problem for $k=2$ and the
549: backbone dimension has been recently exploited to determine $d_B$ very
550: accurately using a transfer matrix technique \cite{JacobsenXX}. Possibly
551: similar methods can be employed to find the fractal dimension of
552: $k$-bones (and therefore $k$-blocks) with $k\geq 3$ to high precision.
553:
554: \section{Relationship to Renormalization Group}
555:
556: The process of replacing a 2-terminal object, $t$, by a single virtual
557: bond and then replacing 2-terminal objects within $t$ by single virtual
558: bonds and so on is reminiscent of the decimation process in
559: renormalization group (RG) approaches to percolation
560: \cite{Stauffer94,Bunde96,Hovi,Reynolds}. It is here, however, that the
561: similarity ends. The decimation process performed in the decomposition
562: into 3-blocks is an exact decimation performed on objects in individual
563: realizations while the RG decimation is performed on the lattice and is
564: an approximation, except for hierarchical lattices. Also, the purpose of
565: the decomposition into 3-blocks is to improve computational performance
566: and analyze the properties of substructures of the cluster while the
567: purpose of RG calculations is to find properties of percolation
568: analytically. Finally, whereas RG approaches on hierarchical lattices
569: result in objects that are finitely ramified, the decomposition into
570: 3-blocks we perform maintains the infinite ramification of the Euclidean
571: lattice.
572:
573: \section{Computational Implications}
574:
575: The fact that the fractal dimension of 3-blocks is significantly smaller
576: than the fractal dimension of 2-blocks has important computational
577: implications. We can efficiently calculate properties(e.g. resistance,
578: velocity distributions, self avoiding walk statistics) of a percolation
579: cluster or backbone as follows:
580:
581: \begin{itemize}
582:
583: \item[{(i)}] decompose the cluster or backbone into 2-blocks
584: \item[{(ii)}] decompose the 2-blocks into 3-blocks
585: \item[{(iii)}] calculate the desired properties of the 3-blocks
586: \item[{(iv)}] algebraically determine the properties of the 2-blocks from the
587: properties of the 3-blocks
588: \item[{(v)}] algebraically determine the properties of the cluster or backbone
589: from the properties of the 2-blocks.
590:
591: \end{itemize}
592:
593: In many cases the computation will require less CPU (computer
594: processing) resource when the complexity of the computation is a power
595: law or exponential of the mass of the object for which the property is
596: being calculated. By decomposition we make the mass of these objects
597: smaller. Reduced CPU resource usage is also obtained if only a
598: decomposition into 2-blocks is made although the saving are
599: less. Systems of larger size than could be treated before can now be
600: treated when we decompose into 3-blocks because the fractal dimension of
601: the 3-blocks are lower than that of the object in which they are
602: embedded; this is not true if we only decompose into 2-blocks.
603:
604: As an example of the dramatically smaller size of the largest 3-block
605: versus the size of the largest blob consider a 3D system of size
606: $L=1000$. At criticality, the largest mass blob in the backbone will be
607: of the order $L^{1.62}\approx 63,000$ while the mass of the largest
608: 3-block in the backbone will be of $L^{1.2}\approx 4000$. In
609: Fig. \ref{preal} we show an actual simulation realization in which a
610: blob of 950 bonds is decomposed into a 3-block with only 216 virtual
611: bonds, greatly reducing the computational complexity.
612:
613: \section{Discussion}
614:
615: Traditionally the decomposition of percolation systems has been to
616: decompose the system into clusters (1-blocks) and to decompose the
617: clusters into blobs (2-blocks). We extend this decomposition by
618: decomposing 2-blocks into 3-blocks. 3-blocks are especially interesting
619: because in contrast to 1- and 2-blocks, the 3-blocks have the property
620: that they can be nested. That is, 2-terminal objects, which are replaced
621: by single virtual bonds in a 3-block, can themselves contain other
622: 3-blocks. Because of this replacement of a 2-terminal object by a
623: virtual bond, the fractal dimension of 3-blocks is significantly smaller
624: than the fractal dimension of 2-blocks. As discussed in the previous
625: section, this smaller fractal dimension has important computational
626: implications for the size of percolation systems which can be analyzed
627: and the speed at which the analysis can be performed.
628:
629: In addition, within the error bars of our calculations, the values for
630: the 3-block fractal dimension appear to be identical for 2D and 3D
631: systems. Simulations of larger systems and higher dimension systems
632: could help answer whether in fact $d_3$ is independent of dimension
633: (``super-universal''). It will also be of interest to determine the
634: properties of $k$-blocks with $k>3$.
635:
636:
637: \subsubsection*{Acknowledgements}
638:
639: We thank D. Baker, L. Braunstein, S. Havlin, A. Moreira, and
640: V. Schulte-Frohlinde for stimulating discussions.
641:
642: \appendix
643: \section{Relationships among Exponents}
644:
645:
646: Here we ask if any of the fractal dimensions and power-law regime
647: exponents we have identified are related. To answer this question we
648: must first briefly review some existing results for relations between
649: other exponents.
650:
651: \subsection{Previous Results}
652:
653: It has been shown generally \cite{Huber95a,Huber95b} that, for disjoint
654: objects of type $X$ embedded in a space $Y$,
655: %
656: \begin{equation}
657: d_X(\tau_{X,Y}-1)=d_{nY}.
658: \label{e3}
659: \end{equation}
660: %
661: Equation (\ref{e3}) holds if $\tau_X<2$ or if $d_{nY}$ is equal to the
662: fractal dimension of space $Y$, $d_Y$.
663:
664: Special cases of Eq.~(\ref{e3}) have been identified previously for
665: $Y=0$, 1, and 2 corresponding to Euclidean space, percolation cluster
666: space and percolation backbone space, respectively.
667:
668: \begin{itemize}
669: \item[{(i)}] The first is the familiar scaling relation for the Fisher exponent
670: $\tau$ \cite{Stauffer94,Bunde96}
671: %
672: \begin{equation}
673: d_f(\tau-1)=d_{n0}=d \qquad\qquad\mbox{clusters},
674: \label{e4a}
675: \end{equation}
676: %
677: where $d$ is the Euclidean dimension, $d_f$ the fractal dimension of the
678: cluster, and $\tau$ the exponent of the power-law regime in the
679: distribution of cluster sizes.
680:
681: \item[{(ii)}] In Ref.~\cite{Gyure95} it was shown that
682: %
683: \begin{equation}
684: d_{\mbox{\scriptsize blob-cl}}(\tau_{\mbox{\scriptsize
685: blob-cl}}-1)=d_{n1}=d_f, \qquad \mbox{(cluster blobs)}
686: \label{e5}
687: \end{equation}
688: %
689: where $d_{\mbox{\scriptsize blob-cl}}$ and $\tau_{\mbox{\scriptsize
690: blob-cl}}$ are the fractal dimension and the exponent of the power-law
691: regime, respectively, for all blobs in the cluster.
692:
693: \item[{(iii)}] In Ref.~\cite{Herrmann84} it was argued that
694: %
695: \begin{equation}
696: d_{\mbox{\scriptsize blob-bb}}(\tau_{\mbox{\scriptsize
697: blob-bb}}-1)=d_{nB}=d_{\mbox{\scriptsize red}}, \quad
698: \mbox{(backbone blobs)}
699: \label{e6}
700: \end{equation}
701: %
702: where $d_{\mbox{\scriptsize blob-bb}}$ and $\tau_{\mbox{\scriptsize
703: blob-bb}}$ are the fractal dimension and the exponent of the power-law
704: regime, respectively, for those blobs in the backbone and
705: $d_{\mbox{\scriptsize red}}$ is the fractal dimension of
706: singly-connected red bonds in the backbone.
707:
708: \end{itemize}
709:
710: Both $d_{\mbox{\scriptsize blob-cl}}$ and $d_{\mbox{\scriptsize
711: blob-bb}}$ are equal to $d_B$, the backbone fractal dimension.
712: In (i) and (ii), Eq.~(\ref{e3}) applies because $d_{nY}=d_Y$; in (iii),
713: Eq.~(\ref{e3}) applies because $\tau_X<2$.
714:
715: \subsection{3-blocks in Blobs}
716:
717: In analogy with Eqs.~(\ref{e4a}) and~(\ref{e5}), we would expect
718: %
719: \begin{equation}
720: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
721: 3,2}}^\ast-1)=d_{n2}^\ast=d_2.
722: \label{e220}
723: \end{equation}
724: %
725: We first confirm that the total number of 3-blocks in blobs scales with
726: the exponent $d_2$. If
727: %
728: \begin{equation}
729: <n(L)>\sim L^{d_{n2}}
730: \label{e206}
731: \end{equation}
732: %
733: and
734: %
735: \begin{equation}
736: N_2\sim L^{d_2}
737: \label{e207}
738: \end{equation}
739: %
740: then we would expect
741: %
742: \begin{equation}
743: <n(N_2)>\sim L^{d_{n2}/d_2}.
744: \label{e208}
745: \end{equation}
746: %
747: Figures \ref{pNumbAll3}(a) and \ref{pNumbAll3}(b) are log-log plots of
748: $\langle n(N_2)\rangle$, the average number of all 3-blocks in a
749: blob, versus blob size $N_2$ for 2D and 3D, respectively. The straight
750: line fits with slope $1.0\pm 0.05$ are consistent with $d_{n2}=d_2$.
751: Our simulation results in 2D from Section IV, $d_3=1.20$ and
752: $\tau_{3,2}=2.35$ result in $d_3(\tau_{3,2}^\ast-1)=1.62$ close to the value
753: $d_2=1.6432$. In 3D, our simulation results from Section V, $d_3=1.15$ and
754: $\tau_{3,2}=2.63$ result in $d_3(\tau_{3,2}^\ast-1)=1.87$ identical to the value
755: $d_2=1.87$\cite{Porto97}.
756:
757: \subsection{3-blocks in Backbone}
758:
759: Because the number of top-level 3-blocks in the backbone is proportional
760: to the number of 2-blocks in the backbone, the number of top level 3-
761: blocks in the backbone should scale the same way the number of 2-blocks
762: in the backbone. For all dimensions and lattices, $d_{nB}$ has been
763: shown to be \cite{Coniglio82a,Coniglio82b}
764: %
765: \begin{equation}
766: d_{nB}=d_{red}={1\over\nu},
767: \label{e7}
768: \end{equation}
769: %
770: where $\nu$ is the exponent associated with the divergence of the
771: correlation length as $p$ approaches $p_c$ [1,2]. In 2D $1/\nu$ is
772: exactly 3/4 \cite{denNijs82,Nienhuis82}; in 3D, $1/\nu$ has been
773: estimated to be $1.143\pm 0.01$ \cite{ZiffStell,Strenski}. We would
774: expect
775: %
776: \begin{equation}
777: d_{\mbox{\scriptsize 3}}(\tau_{\mbox{\scriptsize
778: 3,B}}-1)=d_{nB}={1\over\nu}.
779: \label{e210}
780: \end{equation}
781: %
782:
783: Figures \ref{pNumbC}(a) and \ref{pNumbC}(b) are log-log plots of $\langle
784: n(L)\rangle$, the average number of top-level 3-blocks in the backbone,
785: versus system size $L$ for 2D and 3D, respectively. The straight line
786: fits with slope $0.75\pm 0.05$ and $1.14\pm 0.05$ are consistent with
787: the exact and previously estimated values for $1/\nu$ of 3/4 and 1.143
788: in 2D and 3D, respectively. Our 2D simulation results from Section IV,
789: $d_3=1.15$ and $\tau_{3,2}=1.60$ result in $d_3(\tau_{3,B}-1)=0.69$
790: close to the value $1/\nu=3/4$. For 3D, our simulation results from
791: Section V, $d_3=1.15$ and $\tau_{3,2}=2.0$ result in
792: $d_3(\tau_{3,B}-1)=1.15$ close to the value $1/\nu=1.143$.
793:
794:
795: \begin{references}
796:
797: \bibitem{Ben-Avraham00} D. Ben-Avraham and S. Havlin, {\it Diffusion
798: and Reactions in Fractals and Disordered Systems\/} (Cambridge
799: University Press, Cambridge, 2000).
800:
801: \bibitem{Stauffer94} D. Stauffer and A. Aharony, {\it Introduction to
802: Percolation Theory\/} (Taylor \& Francis, Philadelphia, 1994).
803:
804: \bibitem{Bunde96} A. Bunde and S. Havlin, in {\it Fractal and Disordered
805: Systems}, edited by A. Bunde and S. Havlin (Springer-Verlag, New York,
806: 1996).
807:
808: \bibitem{Gyure95} M. F. Gyure, M. V. Ferer, B. F. Edwards, and G. Huber,
809: Phys. Rev. E {\bf 51}, 2632 (1995).
810:
811: \bibitem{Herrmann84} H. J. Herrmann and H. E. Stanley,
812: Phys. Rev. Lett. {\bf 53}, 1121 (1984).
813:
814: \bibitem{Tutte84} W. T. Tutte, {\it Graph Theory\/} (Cambridge
815: University Press, Cambridge, 1984).
816:
817: \bibitem{Text1} The precise definition of a $k$-block is(previous
818: reference): A graph $G$ is a $k$-block if it has no $n$-separation for
819: any $n<k$. An $n$-separation of $G$ is a pair of subgraphs of $G$, $H$,
820: and $K$, where (i) $H\cup K =G$, (ii) $H$ and $K$ have exactly $k$
821: common vertices, and (iii) $H$ and $K$ each have at least $k$ edges.
822:
823: \bibitem{LorenzXX} C. D. Lorenz and R. M. Ziff, Phys. Rev. E {\bf 57},
824: 230 (1998).
825:
826: \bibitem{Hohberg92} W. Hohberg, Discrete Mathematics {\bf 109}, 133
827: (1992).
828:
829: \bibitem{Hopcroft73} J. E. Hopcroft and R. E. Tarjan, Siam
830: J. Comput. {\bf 2}, 135 (1973).
831:
832: \bibitem{Grassberger99} P. Grassberger, J. Phys. A {\bf 262}, 252
833: (1999).
834:
835: \bibitem{JacobsenXX} J. Jacobsen and P. Zinn-Justin, cond-mat/011374.
836: These authors denote $\hat x_k$ by $\tilde x_k$.
837:
838: \bibitem{Porto97} M. Porto, A. Bunde, S. Havlin, and H. E. Roman,
839: Phys. Rev. E {\bf 56}, 1667 (1997).
840:
841: \bibitem{AizenmanXX} M. Aizenman, B. Duplantier and A. Aharony,
842: Phys. Rev. Lett. {\bf 83}, 1359 (1999). These authors denote $\hat x_k$
843: by $\hat x_{\ell}^P$.
844:
845: \bibitem{Hovi} J.-P. Hovi and A. Aharony, Phys. Rev. E {\bf 56}, 172
846: (1997).
847:
848: \bibitem{Reynolds} P. J. Reynolds, H. E. Stanley, and W. Klein,
849: Phys. Rev. B {\bf 21}, 1223 (1980).
850:
851:
852: \bibitem{Huber95a} G. Huber, M. H. Jensen, and K. Sneppen, Fractals {\bf
853: 3}, 525 (1995).
854:
855: \bibitem{Huber95b} G. Huber, M. H. Jensen, and K. Sneppen, Phys. Rev. E
856: {\bf 52}, R2133 (1995).
857:
858: \bibitem{Coniglio82a} H. E. Stanley, J. Phys. A {\bf 10}, L211 (1977);
859: A. Coniglio, J. Phys. A {\bf 15}, 3829 (1982).
860:
861: \bibitem{Coniglio82b} A. Coniglio, Phys. Rev. Lett. {\bf 46}, 250
862: (1982).
863:
864: \bibitem{denNijs82} M. P. M. den Nujs, J. Phys A {\bf 15}, 199 (1982).
865:
866: \bibitem{Nienhuis82} B. Nienhuis, J. Phys. A {\bf 15}, 199 (1982).
867:
868: \bibitem{ZiffStell} R. M. Ziff and G. Stell, unpublished.
869:
870: \bibitem{Strenski} P. N. Strenski, R. M. Bradley, J. M. Debierre,
871: Phys. Rev. Lett. {\bf 66}, 133 (1991).
872:
873: \end{references}
874:
875: \end{multicols}
876:
877: \newpage
878:
879: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
880:
881: \begin{table}[htb]
882: \caption{Measured fractal dimension, measured power-law regime exponent,
883: and calculated fractal dimension for 3-blocks in 2D and 3D. The
884: calculated value of $d_3$ is determined by Eq.~(\protect\ref{e104}) for 3-blocks in a blob
885: and Eq.~(\protect\ref{e106}) for 3-blocks in the backbone.}
886: \begin{tabular}{cccc}
887: {\bf 2D}\\
888: & $d_3$ & $\tau$ & $d_3$ \\
889: & {\scriptsize MEASURED} & {\scriptsize MEASURED}
890: & {\scriptsize CALCULATED} \\ \hline
891: {\scriptsize All 3 blocks in blob} & $1.20\pm 0.1$ & $2.35\pm 0.05$ &
892: $1.22\pm 0.05$ \\
893: {\scriptsize All 3 blocks in backbone} & $1.15\pm 0.1$ & $2.25\pm 0.05$ &
894: --- \\
895: {\scriptsize Top level 3 blocks in backbone} & $1.15\pm 0.1$ & $1.60\pm 0.05$ &
896: $1.25\pm 0.1$ \\ \hline
897: {\bf 3D}\\
898: & $d_3$ & $\tau$ & $d_3$ \\
899: & {\scriptsize MEASURED} & {\scriptsize MEASURED}
900: & {\scriptsize CALCULATED} \\ \hline
901: {\scriptsize All 3 blocks in blob} & $1.15\pm 0.1$ & $2.63\pm 0.05$ &
902: $1.15\pm 0.05$ \\
903: {\scriptsize All 3 blocks in backbone} & $1.15\pm 0.1$ & $2.55\pm 0.05$
904: &
905: --- \\
906: {\scriptsize Top level 3 blocks in backbone} & $1.15\pm 0.1$ & $2.0\pm
907: 0.05$ &
908: $1.14\pm 0.1$ \\
909: \end{tabular}
910: \end{table}
911:
912: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
913:
914: \begin{figure}
915: \centerline{
916: \epsfxsize=19.0cm
917: \epsfclipon
918: \epsfbox{tpoDecomp.eps}
919: }
920: \caption{(a) Decomposition of 2-block $G$ into subgraphs $G_1$, $G_2$,
921: and $G_3$. The rightmost graph represents $G$ with the sub-graphs
922: replaced by equivalent ``virtual edges.'' (b) Subgraph $G_2$ of $G$ is
923: decomposed by identifying subgraph $G_{21}$. The rightmost graph
924: represents $G_2$ with the subgraph $G_{21}$ replaced by it equivalent
925: edge. (c) Subgraph $G_3$ of $G$ is decomposed by identifying subgraph
926: $G_{31}$. The rightmost graph represents $G_3$ with the subgraph
927: $G_{31}$ replaced by it equivalent edge. In (a), (b), and (c) virtual
928: edges are denoted by dashed lines. Note that while not shown in this
929: figure, subgraph $G_{31}$ could be further decomposed. The 3-blocks
930: contained in the graph $G$ are $G_{21}$, having 5 edges, and $G_3$
931: (with the subgraph $G_{31}$ replaced by its equivalent edge) having 8
932: edges.}
933: \label{tpoDecomp}
934: \end{figure}
935:
936:
937: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
938:
939: \newpage
940:
941: \begin{figure}
942: \centerline{
943: \xsize
944: \epsfclipon
945: \epsfbox{preal.eps}
946: }
947:
948: \caption{Example of decomposition of backbone into 3-blocks. The thin
949: lines represent the bonds in the backbone between points \{0,15\} and
950: \{31,15\} on a lattice with L=32. The backbone is composed of a few
951: single bond blobs connected to the terminal points and a single large
952: blob containing 950 bonds. The thick lines represent the virtual bonds
953: of a single top-level 3-block into which the blob has been decomposed.
954: This 3-block contains 216 virtual bonds. Some of the groups of bonds
955: replaced by virtual bonds can themselves be decomposed into lower level
956: 3-blocks and so on.}
957: \label{preal}
958: \end{figure}
959:
960:
961: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
962: % real lattice/tpos/3block
963: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
964: \newpage
965:
966: \begin{figure}
967: \centerline{
968: \xsize
969: \epsfclipon
970: \epsfbox{pAll3.eps}
971: }
972: \centerline{
973: \xsize
974: \epsfclipon
975: \epsfbox{pAll3s.eps}
976: }
977: \caption{2D (a) Distributions $P^\ast(N_3|N_2)$ of the number of 3-blocks of
978: mass $N_3$ in a blob of size $N_2$ versus $N_3$ for (from bottom to top)
979: $N_2=2^{10}, 2^{12}, 2^{14}$, and $2^{16}$. The distributions exhibit a power-law
980: regime with slope $-2.35\pm 0.05$ (b) Distributions for $N_2=2^{12}, 2^{13},
981: 2^{14}, 2^{15}$, and $2^{16}$ scaled with the value 1.20 for the fractal
982: dimension $d_3$ which gives the best collapse of the plots in (a).}
983: \label{pAll3}
984: \end{figure}
985:
986:
987: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
988:
989:
990: \newpage
991:
992: \begin{figure}
993: \centerline{
994: \xsize
995: \epsfclipon
996: \epsfbox{pc.eps}
997: }
998: \centerline{
999: \xsize
1000: \epsfclipon
1001: \epsfbox{pcs.eps}
1002: }
1003: \caption{2D (a) Distributions $P^\ast(N_3|L)$ of the number of 3-blocks of mass
1004: $N_3$ in a backbone of size $L$ versus $N_3$ for (from bottom to top)
1005: $L=16$, 32, 64, 128, 256 and 512. The distributions exhibit a power-law
1006: regime with slope $-2.25\pm 0.05$ (b) Distributions scaled with the
1007: value 1.15 for the fractal dimension $d_3$ which gives the best collapse
1008: of the plots in (a).}
1009: \label{pc}
1010: \end{figure}
1011:
1012:
1013:
1014: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1015:
1016: \newpage
1017:
1018: \begin{figure}
1019: \centerline{
1020: \xsize
1021: \epsfclipon
1022: \epsfbox{pc1.eps}
1023: }
1024: \centerline{
1025: \xsize
1026: \epsfclipon
1027: \epsfbox{pc1s.eps}
1028: }
1029: \caption{2D (a) Distributions $P(N_3|L)$ of the number of top level
1030: 3-blocks of mass $N_3$ in a backbone of size $L$ versus $N_3$ for (from
1031: top to bottom) $L=8$, 16, 32, 64, and 128. The distributions exhibit a
1032: power-law regime with slope $-1.6\pm 0.1$. (b) Distributions scaled with
1033: the value 1.15 for the fractal dimension $d_3$ which gives the best
1034: collapse of the plots in (a).}
1035: \label{pc1}
1036: \end{figure}
1037:
1038: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1039:
1040: \newpage
1041:
1042: \begin{figure}
1043: \centerline{
1044: \xsize
1045: \epsfclipon
1046: \epsfbox{pcComb.eps}
1047: }
1048: \caption{2D Distributions $P(N_3|L)$ of top level 3-blocks (filled
1049: symbols) and $P^\ast(N_3|L)$ of all-level 3-blocks (unfilled symbols). While
1050: the slopes of the power law regimes of the two types of distributions
1051: are different, the finite-size-system cutoffs are essentially
1052: superimposed, consistent with the fractal dimension of the two types of
1053: distributions being equal.}
1054: \label{pcComb1}
1055: \end{figure}
1056: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1057:
1058: \newpage
1059:
1060: \begin{figure}
1061: \centerline{
1062: \xsize
1063: \epsfclipon
1064: \epsfbox{pe.eps}
1065: }
1066: \centerline{
1067: \xsize
1068: \epsfclipon
1069: \epsfbox{pes.eps}
1070: }
1071: \caption{2D (a) Distributions $P^\ast(M_3|L)$ of the number of 3-blocks of
1072: mass $M_3$ in a backbone of size $L$ versus $M_3$ for from top
1073: to bottom) $L=16$, 32, 64, 128, 256, and 512. In $M_3$ we count not
1074: virtual bonds but all bonds in the 3-block. The distributions exhibit a
1075: power-law regime with slope $-1.8\pm 0.1$ (b) Distributions scaled with
1076: the value 1.6 for the fractal dimension $d_3$ which gives the best
1077: collapse of the plots in (a).}
1078: \label{pe}
1079: \end{figure}
1080:
1081:
1082:
1083: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1084: % 3D
1085: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1086:
1087: \newpage
1088:
1089: \begin{figure}
1090: \centerline{
1091: \xsize
1092: \epsfclipon
1093: \epsfbox{pAll33d.eps}
1094: }
1095: \centerline{
1096: \xsize
1097: \epsfclipon
1098: \epsfbox{pAll3s3d.eps}
1099: }
1100: \caption{3D (a) Distributions $P^\ast(N_3|N_2)$ of the number of 3-blocks of
1101: mass $N_3$ in a blob of size $N_2$ versus $N_3$ for (from bottom to top)
1102: $N_2=2^{11}, 2^{12}$, and $2^{13}$. The distributions exhibit a
1103: power-law regime with slope $-2.63\pm 0.05$ (b) Distributions for $N_2=
1104: 2^{11}, 2^{12}$, and $2^{13}$ scaled with the value 1.15 for the fractal
1105: dimension $d_3$ which gives the best collapse of the plots in (a).}
1106: \label{pAll33d}
1107: \end{figure}
1108:
1109: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1110: \newpage
1111:
1112: \begin{figure}
1113: \centerline{
1114: \xsize
1115: \epsfclipon
1116: \epsfbox{pc3d.eps}
1117: }
1118:
1119: \centerline{
1120: \xsize
1121: \epsfclipon
1122: \epsfbox{pcs3d.eps}
1123: }
1124:
1125: \caption{3D (a) Distributions $P^\ast(N_3|L)$ of the number of 3-blocks of
1126: mass $N_3$ in a backbone of size $L$ versus $N_3$ for (from top to bottom)
1127: $L= 8$, 16, 32, 64, and 128. The distributions exhibit a power-law
1128: regime with slope $-2.55\pm 0.1$. (b) Distributions scaled with the
1129: value 1.15 for the fractal dimension $d_3$ which gives the best collapse
1130: of the plots in (a).}
1131: \label{pc3d}
1132: \end{figure}
1133:
1134: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1135:
1136:
1137: \newpage
1138:
1139: \begin{figure}
1140: \centerline{
1141: \xsize
1142: \epsfclipon
1143: \epsfbox{pc13d.eps}
1144: }
1145: \centerline{
1146: \xsize
1147: \epsfclipon
1148: \epsfbox{pc1s3d.eps}
1149: }
1150: \caption{3D (a) Distributions $P(N_3|L)$ of the number of top level
1151: 3-blocks of mass $N_3$ in a backbone of size $L$ versus $N_3$ for (from
1152: top to bottom) $L=32$, 64, 128, 256, and 512. The distributions exhibit a
1153: power-law regime with slope $-2.0\pm 0.1$. (b) Distributions scaled with
1154: the value 1.15 for the fractal dimension $d_3$ which gives the best
1155: collapse of the plots in (a).}
1156: \label{pc13d}
1157: \end{figure}
1158:
1159: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1160:
1161: \newpage
1162:
1163: \begin{figure}
1164:
1165:
1166: \centerline{
1167: \xsize
1168: \epsfclipon
1169: \epsfbox{pcComb3d.eps}
1170: }
1171: \caption{3D Distributions $P(N_3|L)$ of top level 3-blocks (filled
1172: symbols) and $P^\ast(N_3|L)$ of all-level 3-blocks (unfilled symbols). While
1173: the slopes of the power law regimes of the two types of distributions
1174: are different, the finite-size-system cutoffs are essentially
1175: superimposed consistent, with the fractal dimension of the two types of
1176: distributions being equal.}
1177: \label{pcComb13d}
1178: \end{figure}
1179:
1180: \newpage
1181:
1182: \begin{figure}
1183: \centerline{
1184: \xsize
1185: \epsfclipon
1186: \epsfbox{pe3d.eps}
1187: }
1188: \centerline{
1189: \xsize
1190: \epsfclipon
1191: \epsfbox{pes3d.eps}
1192: }
1193: \caption{3D (a) Distributions $P^\ast(M_3|L)$ of the number of 3-blocks of
1194: mass $M_3$ in a backbone of size $L$ versus $M_3$ for (from top to bottom)
1195: $L=8$, 16, 32, 64, and 128. In $M_3$ we count not virtual bonds but all
1196: bonds in the 3-block. The distributions exhibit a power-law regime with
1197: slope $-1.87\pm 0.1$. (b) Distributions scaled with the value 1.85 for
1198: the fractal dimension $d_3$ which gives the best collapse of the plots
1199: in (a).}
1200: \label{pe3d}
1201: \end{figure}
1202:
1203:
1204: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1205: % Number dimension plots
1206: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1207:
1208: \newpage
1209:
1210: \begin{figure}
1211: \centerline{
1212: \xsize
1213: \epsfclipon
1214: \epsfbox{pNumbAll3.eps}
1215: }
1216:
1217: \centerline{
1218: \xsize
1219: \epsfclipon
1220: \epsfbox{pNumbAll33d.eps}
1221: }
1222: \caption{$\langle n^\ast(N_2)\rangle$, the average number of 3-blocks in
1223: a blob of size $N_2$ versus $N_2$ for (a) 2D and (b) 3D.}
1224:
1225: \label{pNumbAll3}
1226: \end{figure}
1227:
1228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1229:
1230: \newpage
1231:
1232: \begin{figure}
1233: \centerline{
1234: \xsize
1235: \epsfclipon
1236: \epsfbox{pNumbC1.eps}
1237: }
1238:
1239: \centerline{
1240: \xsize
1241: \epsfclipon
1242: \epsfbox{pNumbC13d.eps}
1243: }
1244:
1245:
1246: \caption{$\langle n(L)\rangle$, the average number of top level
1247: 3-blocks in a backbone of size $L$ versus $L$. (a) 2D The solid line has
1248: slope 0.75. (b) 3D The solid line has slope 1.14.}
1249: \label{pNumbC}
1250: \end{figure}
1251:
1252: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1253:
1254:
1255:
1256: \end{document}
1257:
1258: