cond-mat0203092/phs.tex
1: %\documentstyle[preprint,aps,epsf,epsfig]{revtex}
2: \documentstyle[aps,multicol,epsf,epsfig]{revtex} 
3: %\documentclass[]{ws-p9-75x6-50}  
4: \newcommand{\xsize}{\epsfxsize=18.0cm}
5: 
6: \begin{document}
7: 
8: 
9: %\draft
10: 
11: \title{Fractal Behavior of the Shortest Path Between Two Lines in
12: Percolation Systems}
13: 
14: \author{Gerald Paul,$^\ast$ Shlomo Havlin,$^\dagger$ and H. Eugene Stanley$^\ast$} 
15: 
16: \address{$^\ast$Center for Polymer Studies and Department of Physics\\
17: Boston University, Boston, MA 02215 USA\\
18: $^\dagger$Minerva Center and Department of Physics\\
19: Bar-Ilan University, Ramat Gan, Israel}
20: 
21: \date{phs.tex ~~~ 04 March 2002 ~~~ draft}
22: 
23: \maketitle
24: 
25: \begin{abstract}
26: %\abstracts{
27: Using Monte-Carlo simulations, we determine the scaling form for the
28: probability distribution of the shortest path, $\ell$, between two lines
29: in a 3-dimensional percolation system at criticality; the two lines can
30: have arbitrary positions, orientations and lengths. We find that the
31: probability distributions can exhibit up to four distinct power law
32: regimes (separated by cross-over regimes) with exponents depending on
33: the relative orientations of the lines. We explain this rich fractal
34: behavior with scaling arguments.
35: %}
36: \end{abstract}
37: 
38: \begin{multicols}{2}  
39: \section{Introduction}
40: 
41: There has been considerable recent activity
42: \cite{Dokh,Lee60,Lee62,Grassberger,Ziff99} analyzing $P(\ell|r)$, the
43: probability distribution for the length of shortest path, $\ell$,
44: between two points separated by Euclidean distance $r$ in a percolation
45: system \cite{Neumann,Stauffer,BundeHavlin,Ben-Avraham00}. This paper extends that work
46: by determining the scaling form of the distribution of shortest paths
47: between two lines of arbitrary position, relative orientation and
48: lengths in 3-dimensional systems. These configurations are important
49: because they much more accurately model the configurations used in oil
50: recovery in which fluid is injected in one well (one of the lines in our
51: configuration) and oil is recovered at a second well (the second line in
52: our configuration); the wells may, in reality, be at arbitrary orientation
53: and of different lengths.
54: 
55: The scaling form for the 2-points configuration in which the 2 points
56: are located at $((L-r)/2,L/2,L/2),((L+r)/2,L/2,L/2)$ in a system of side
57: $L$ has been found to be \cite{Dokh}
58: %
59: \begin{equation}
60: \label{e.1}
61: P(\ell|r)\sim {1\over  r^{d_{\mbox{\scriptsize
62: min}}}}    \left( {\ell\over r^{d_{\mbox{\scriptsize
63: min}}}}\right)^{-g_{\ell}}f_1\left({\ell\over r^{d_{\mbox{\scriptsize
64: min}}}}\right)f_2\left({\ell\over L^{d_{\mbox{\scriptsize
65: min}}}}\right),
66: \end{equation}
67: %
68: where
69: %
70: %\begin{mathletters}
71: \begin{equation}
72: \label{e.2a}
73: f_1(x)=e^{-ax^\phi}
74: \end{equation}
75: %
76: and
77: %
78: \begin{equation}
79: \label{e.2b}
80: f_2(x)=e^{-bx^\psi}.
81: \end{equation}
82: %\end{mathletters}
83: %
84: The exponents $g_\ell$, $d_{\mbox{\scriptsize min}}$, $\phi$, and $\psi$
85: are universal and the constants $a$ and $b$ depend on lattice type. In
86: 3D, the values of these exponents have previously been found to be
87: \cite{Lee62} $g_\ell=2.3\pm 0.1$, $d_{\mbox{\scriptsize min}}=1.39\pm
88: 0.05$, $\phi=2.1\pm 0.5$, and $\psi=2.5\pm 0.5$. The first stretched
89: exponential function, $f_1$, reflects the fact that the shortest path
90: must always be at least equal to the distance $r$ between the two
91: points; the second stretched exponential function, $f_2$, reflects the
92: fact that the lengths of the shortest paths are bounded because of the
93: finite size, $L$, of the system.
94: 
95: We find that the scaling form for the 2-lines configuration has the same
96: form as that found for the shortest path distribution between two points
97: with the exceptions that: (i) the power law regime of the distribution
98: as represented by the term $(\ell/r^{d_{\mbox{\scriptsize
99: min}}})^{-g_{\ell}}$ in Eq.~(\ref{e.1}) is replaced with up to four
100: different power law regimes (separated by cross-over regimes) with
101: exponents depending on the relative orientations of the lines and (ii)
102: the Euclidean distance, $r$, in Eq.~(\ref{e.1}) between the two points
103: is replaced by the shortest Euclidean distance between the two lines.
104: The lengths of the lines affect the sizes of the power law regimes.
105: 
106: \section{Simulations}
107: 
108: Monte Carlo simulations were performed using the Leath method and
109: growing clusters from 2 sets of seeds---one for each line. The length of
110: the shortest path between the two lines is the sum of the chemical
111: distances from each set of seed sites to the point where a cluster
112: started at one set of seeds meets a cluster started from the other set
113: of seeds \cite{Grassberger}.  The cluster growth for a given realization
114: is terminated when the two clusters meet.  For parallel line
115: configurations, in which the probability distributions decay rapidly, we
116: use the method of Ref.\cite{Grassberger02} to obtain good statistics for
117: shortest paths that have very low probabilities.  We use the memory management
118: technique described in \cite{Paul01} to perform simulations in which the
119: growing clusters never hit a boundary of the system.
120: 
121: The clusters that are created and included in our analysis are of all
122: sizes, not just the incipient infinite cluster.
123: 
124: \section{Non-Parallel Wells}
125: \subsection{Co-Planar Lines}
126: 
127: 
128: \subsubsection{Equal Length Symmetric Lines}
129: 
130: 
131: We start by considering relatively simple configurations of the type
132: shown in Fig.~\ref{pPP1}(a) in which the lines are co-planar, of equal length and
133: are positioned symmetrically. We will study configurations in which the
134: lines are of unequal length [see Fig.~\ref{pPP1}(b)] and/or are not positioned
135: symmetrically [see Fig.~\ref{pPP1}(c)] in the following sections. In all of these
136: configurations, $r$ is the shortest Euclidean distance between the two
137: lines.
138: 
139: %\subsubsection{Power Law Regime}
140: 
141: Figure \ref{pEq} contains log-log plots of $P(\ell|r)$, the
142: shortest path distribution for $r=8$ and various values of $\theta$. We
143: have chosen $r=8$ so that the initial cutoff is present; for smaller
144: values of r, lattice effects destroy this initial cutoff. Since the
145: focus of this paper is the power law regimes, not the initial or final
146: cutoffs, in all later figures we will choose configurations with $r=1$
147: so that the extent of the power law regimes is as long as possible. The
148: exception to this will be cases in which $\theta$ is very small where
149: small $r$ introduces other lattice effects.
150: 
151: 
152: 
153: 
154: We note that after the initial peak in each distribution, there is a
155: power-law regime, the slope of which, $g_\ell(\theta)$, increases with
156: increasing $\theta$. We will call this power law regime the ``2-lines
157: regime.'' Simple scaling arguments imply that if the lengths of the
158: lines were infinite, these 2-lines regimes would continue
159: indefinitely. For finite line lengths, we would expect that, for large
160: $\ell$, the distributions would exhibit a crossover to a power law
161: regime with the same exponent as that for a configuration with two
162: points---because for large $\ell$ the long paths travel far enough away
163: from the lines that they appear to be points. In this regime, the power
164: law exponent has the value of that of two points, 2.35 \cite{Lee62}. For the plots in
165: Fig.~\ref{pEq}, in order to see the power regimes clearly, we have chosen the
166: lengths of the lines long enough that this crossover occurs after the
167: maximum value of $\ell$ in the plots.
168: 
169: In Fig.~\ref{gl}, we plot $g_\ell(\theta)$ versus $\theta$. The plot
170: suggests that $g_\ell(\theta)$ diverges as $\theta$ goes to
171: zero. We attempt to fit this function with a power of $1/sin((\theta)/2)$
172: and find the best fit with the function
173: %
174: \begin{equation}
175: \label{e1x}
176: f(\theta)={g_l(180^\circ)\over\sin(\theta/2)^{0.4}}.
177: \end{equation}
178: %
179: This form is not based on any fundamental theory; it simply has the
180: properties that $f(\theta=180^\circ)=g_{\ell}(\theta=180^\circ)$, it
181: diverges as $\theta$ goes to zero and it fits the intermediate points
182: reasonably well.
183: 
184: The crossover between the 2-line regime and the 2-point regime is
185: illustrated in Fig.~\ref{pEqCutoff} where in each panel we plot $P(\ell|r)$ for
186: fixed $\theta$, and various values of $W$. As expected, the larger the
187: length of the lines, the higher the value of $\hat\ell$, the value of
188: $\ell$ at which the crossover occurs.  Quantitative analysis of the
189: crossover behavior is given in Section III.C.  
190: 
191: 
192: 
193: \subsubsection{Point-Line Configurations}
194: 
195: 
196: 
197: 
198: We next study configurations in which one line has zero length (i.e., a
199: point) and the other is a line of finite length. This is the extreme
200: case of the configuration in which the two lines have different
201: length. We will study the case where both lines have finite length in
202: the next section. We in fact study the three configurations shown in
203: Figs.~\ref{pPtLineConfig}(a)--(c). The plots of $P(\ell|r)$ for these
204: configurations are shown in Fig.~\ref{pPtLine1}. The plots have a
205: power-law regime with exponent $-1.75$ for the configurations of
206: Fig.~\ref{pPtLineConfig}(a) and Fig.~\ref{pPtLineConfig}(c), and
207: exponent $-2.2$ for the configuration of Fig~\ref{pPtLineConfig}(b). We
208: denote this regime the ``point-line regime.'' Fig.~\ref{pPtLine2} shows
209: the crossover from point-line behavior to 2-points behavior.
210: 
211: 
212: 
213: 
214: 
215: \subsubsection{Unequal Length Symmetric Lines}
216: 
217: We can now study configurations of the type shown in Fig.~1(b) in which
218: the lines are of different lengths, $W_1$ and $W_2$. For such a
219: configuration we would expect three power-law regimes: (i) for small
220: $\ell$ such that the two lines appear to be infinite, a 2-line regime,
221: with slope dependent on $\theta$, (ii) a point-line regime, with slope
222: $-1.75$, for values of $\ell$ large enough that the shorter line appears
223: to be a point, and (iii) a 2-point regime, for even larger values of
224: $\ell$ where both lines appear to be points. Plots of $P(\ell|r)$ for such
225: configurations are shown in Fig.~\ref{pDiff} and are consistent with our
226: expectations.
227: 
228: 
229: 
230: \subsubsection{Complex Configurations (Unequal Length Non-Symmetric Lines)}
231: 
232: The last of the co-planar configurations is of the type shown in
233: Fig.~\ref{pPP1}(c). In general, based on the reasoning above, for configurations
234: of this type we would expect $P(\ell|r)$ to have four power-law
235: regimes. For the configuration shown in Fig.~\ref{pPP1}(c), in which $W_{1a}\ll
236: W_2\ll W_{1b}$, the power law regimes would be as follows: (i) a power
237: law regime corresponding to the angle $\theta$ between the segments
238: $W_{1a}$ and $W_2$, (ii) a power law regime corresponding to the angle
239: $\pi-\theta$ between segments $W_2$ and $W_{1b}$, (iii) a point-line
240: power law regime entered when $\ell\gg W_{1b}$, and (iv) the 2-points
241: regime. Figure~\ref{pOverlap} is a plot of $P(\ell|r)$ which shows this behavior.
242: 
243: \subsection{Non-Coplanar Lines}
244: 
245: 
246: 
247: For non-coplanar lines, for $l\gg r$, the fact that the lines are not
248: co-planar should be irrelevant; what is relevant is the effective angle
249: between the lines. This angle is obtained by sliding the lines toward
250: each other along the line of shortest Euclidean distance between the
251: lines (without changing their orientations) until they touch; the lines
252: are then coplanar and the angle between them determines the behavior of
253: $P(\ell|r)$. Figure~\ref{pNonCoPlanar} contains plots for two
254: configurations which illustrate this: (i) two coplanar lines with $r=1$,
255: $\theta=90^\circ$, and $W=256$, and (ii) the same configuration with the
256: second line translated out of the plane by distance 8.  We see that
257: while there is some difference in the plots for small $l$, the slope of
258: the 2-lines regime is the same for the two plots.
259: 
260: \subsection{Scaling of the Crossover Between Power Law Regimes}
261: 
262: We define the value of $\ell \cong \hat\ell$ at which $P(\ell|r)$
263: crosses over from one power-law regime to another power-law regime as
264: the value of $\ell$ where straight lines fit to the power law regimes
265: ,between which the crossover takes place, cross. In Eq.(\ref{e.1}) the
266: values of $\ell$ at which the lower and upper cutoffs occur scale
267: independently as $r^{d_{\mbox{\scriptsize min}}}$ and
268: $L^{d_{\mbox{\scriptsize min}}}$, respectively.  By extension, we would
269: expect that all characteristic values of the distribution, including
270: crossovers between different power-law regimes, would also scale as
271: $X^{d_{\mbox{\scriptsize min}}}$ where X is the length in the system which controls the
272: crossover. Thus, in analogy with the scaling of the most probable value
273: of $\ell, \ell^\ast$,
274: %
275: \begin{equation}
276: \label{e.6}
277: l^\ast=cr^{d_{\mbox{\scriptsize min}}},
278: \end{equation}
279: %
280: we would, in fact, expect that the value of $\ell$, $\hat\ell$ at which
281: $P(\ell|r)$ crosses over from 2-lines behavior to 2-points behavior
282: scales as 
283: %
284: \begin{equation}
285: \label{e.7a}
286: \hat\ell=c_1(\theta)r_{\mbox{\scriptsize max}}^{d_{\mbox{\scriptsize
287: min}}}, 
288: \end{equation}
289: %
290: where
291: %
292: \begin{equation}
293: \label{e.7x}
294: r_{\mbox{\scriptsize max}}=r+2W\sin(\theta/2)
295: \end{equation}
296: %
297: is the maximum Euclidean distance between the two lines and
298: $c_1(\theta)$ is a slowly varying function of $\theta$.  In order for a 2-lines
299: regime to exist, the 2-lines regime cutoff $\hat\ell$ must be greater
300: than $\ell^\ast$, the maximum value of the distribution. That is,
301: %
302: \begin{equation}
303: \label{e.15}
304: c_1[r+2W\sin(\theta/2)]^{d_{\mbox{\scriptsize
305: min}}}>cr^{d_{\mbox{\scriptsize min}}},
306: \end{equation}
307: %
308: which implies
309: %
310: \begin{equation}
311: \label{e.16}
312: W>{(c/ c_1)^{1/d_{\mbox{\scriptsize min}}}-r\over{2\sin(\theta/2)}}.
313: \end{equation}
314: %
315: 
316: In Figs.~\ref{pEqCutoff}(a), (b), and (c), the insets contain plots of
317: $\hat\ell$ vs.~$r_{\mbox{\scriptsize max}}$.  For $\theta=3^\circ$, the
318: scaling exponent is consistent with Eq.(\ref{e.7a}) but for $\theta=29^\circ$
319: and 180 the scaling exponent is $1.0\pm 0.1$.
320: 
321: Using the same reasoning which led to Eq. (\ref {e.7a}), we would expect
322: the crossover from point-line to 2-points behavior to scale as
323: 
324: \begin{equation}
325: \label{e.7b}
326: \hat\ell=c_2 W^{d_{\mbox{\scriptsize
327: min}}}, 
328: \end{equation}
329: %
330: because $W$ is the length which controls this crossover; as seen in
331:  Fig.~\ref{pPtLine2}, the larger the value of $W$, the larger the value
332:  of $\ell$ at which the crossover from point-line to 2-points behavior
333:  occurs.  However, as seen in the inset in Fig.~\ref{pPtLine2}, the
334:  crossover length scales with an exponent $1.0\pm 0.1$ not
335:  $d_{\mbox{\scriptsize min}}$.
336: 
337: Finally, we would expect that for different length lines, the
338: crossover from 2-lines behavior to 2-point behavior would scale as 
339: \begin{equation}
340: \label{e.7c}
341: \hat\ell=c_3(\theta)W_{2}^{d_{\mbox{\scriptsize
342: min}}}, ~~~~~~~~~~(W_2<W_1)
343: \end{equation}
344: %
345: because $W_2$ is the length which controls this crossover; as seen in
346: Fig.~\ref{pDiff}, the larger the value of $W_2$, the larger the value of
347: $\ell$ at which the crossover occurs.  Again, the insets in
348: Fig.~\ref{pDiff} indicate that the crossover scales with exponent $1.0\pm 0.1$.
349: 
350: 
351: We cannot explain why sometimes the crossover scales with
352: $d_{\mbox{\scriptsize min}}$ and sometimes it scales with an exponent
353: about 1.  It is, of course, possible that corrections-to-scaling are
354: strong and that we are not seeing the true asymptotic behavior of the
355: scaling of the crossover.  If this is the case, the question still
356: remains as to why the corrections to scaling are strong in some
357: configurations and not in others.  This area is a subject for further
358: study.
359: 
360: \section{Parallel Wells}
361: 
362: \subsection{Simple Configurations}
363: 
364: As with non-parallel wells we first consider the simple configurations
365: shown in  Fig.~\ref{pPConfig}(a) in which the parallel wells are of the same
366: length. Figure~\ref{pPar}(a) plots $P(\ell | r)$ vs $\ell$ for $r=1$ and various
367: $W$.  The initial decay of the plots increases with increasing W because
368: the longer the wells, the lower the probability for long shortest paths.
369: Eventually, all plots cross over to a power-law regime with slope
370: consistent with that for two points.  To see if this initial decay is a
371: lattice effect, Fig.~\ref{pPar32} plots of the scaled distributions
372: $r^{d_{\mbox{\scriptsize min}}}P(\ell/r^{d_{\mbox{\scriptsize min}}}|W)$
373: for various $r$ and $W$ where the aspect ratio,
374: %
375: \begin{equation}
376: \label{e.10}
377: R={W\over r},
378: \end{equation}
379: %
380: is fixed at $R=32$. Changing $r$ and $W$ but keeping $R$ fixed results
381: in scaling all lengths in the geometry by the same factor and the plots
382: collapse as expected. Again, we note the initial strong decay of the
383: distribution followed by a 2-point power-law regime.  The good collapse
384: for small $x=1/r^{d_{min}}$ indicates that the strong initial decay is
385: not a lattice effect.  
386: 
387: Because of the small values of $\ell$ at which the crossover
388: to the 2-point regime occurs it is difficult to differentiate between a
389: power-law and (stretched) exponential decay. We will proceed as if the
390: decay were either a power law with slope $\bar g(R)$ or equivalently an
391: exponential with ``effective slope'' $\bar g(R)$.
392: 
393: One might argue as follows that the initial decay for power parallel
394: lines must be exponential: Since the 2-lines regime of the probability
395: distribution for a parallel well configuration must always decay faster
396: than the 2-lines regime of a configuration with small but non-zero
397: $\theta$ and since we believe $g_{\ell}(\theta)$ goes to infinity as
398: $\theta$ goes to zero, the decay for parallel lines must be exponential
399: (i.e., faster than any power law decay). This, however, need not be the
400: case. In order for a 2-lines regime to exist, Eq.~(\ref{e.16}) must
401: hold. So as we decrease $\theta$, we must increase $W$, increasing the
402: aspect ratio $R$, to maintain a 2-lines regime. But since the effective
403: slope for parallel wells, $\bar g(R)$ increases with increasing $R$, the
404: decay can be a power law and still always have a greater slope than the
405: configuration with small but non-zero $\theta$.
406: 
407: \subsection{Complex Configurations}
408: 
409: The treatment of the more complex configurations shown in Figs.~\ref{pPConfig}(b) and
410: (c) follows that of non-parallel wells. $P(\ell)$ for configurations of
411: the type in Fig.~\ref{pPConfig}(b) would contain an initial 2-line regime with slope
412: $\bar g(R=W_2/r)$, a point-line regime, and finally a two-point
413: regime. $P(\ell)$ for configurations of the type in Fig.~\ref{pPConfig}(c), with
414: $W_{1b}\ll W_2\ll W_{1a}$ would contain an initial 2-line regime with
415: slope $\bar g(R=W_{1b}/r)$, a 2-line regime with slope $g_{\ell}(\theta=\pi)$, a point-line regime, and a 2-point regime.
416: 
417: \subsection{Quasi-Euclidean Regime}
418: 
419: When the length of the wells is very large and the distance between the
420: wells is small the behavior of the most probable shortest path between
421: the wells is the same as in a Euclidean space where p=1 and all bonds
422: are occupied. This can be seen in Fig.~\ref{pMax32} in which we plot
423: $\ell^\ast$, the most probable value of the shortest path, versus $r$
424: for various lengths $W$. For long enough wells, there is a regime of $r$
425: in which
426: %
427: \begin{equation}
428: \label{e.12x}
429: \ell^\ast=r,
430: \end{equation}
431: %
432: as one would expect in Euclidean space in which the shortest path is a
433: straight line path of occupied bonds. As also seen in Fig~\ref{pMax32}., for a
434: given well length, as $r$ increases, there is a value of $r$, $r^\ast$,
435: at which the behavior crosses over to that of 3D percolation. We can
436: develop a simple expression for $r^\ast$ as follows: The probability
437: that all bonds in a straight line path between two wells separated by
438: distance $r$ are occupied is $p_c^r$. The probability that one or more
439: bonds in the straight line path is not occupied is thus $1-p_c^r$ and
440: the probability that one or more bonds in the $W$ straight line paths
441: between the wells are not occupied is $(1-p_c^r)^W$. The probability
442: that at least one straight line path has all bonds occupied is then
443: %
444: \begin{equation}
445: \label{e.50}
446: P(r,W)=1-(1-p_c^r)^W.
447: \end{equation}
448: %
449: The shortest path will exhibit Euclidean behavior, i.e., $\ell^\ast=r$
450: when $P(r,W)$ is of the order unity. Setting $P(r^\ast,W)=a$ in
451: Eq.~(\ref{e.50}), we find
452: %
453: \begin{equation}
454: \label{e.51}
455: r^\ast={\ln(1-a^{1/W})\over\ln p_c}.
456: \end{equation}
457: %
458: In Fig.~\ref{pMax32t} we plot the observed values of $r^\ast$ and the values
459: predicted by Eq.~(\ref{e.51}) with a value of $a=0.55$ which gives the
460: best fit to the observed values.
461: 
462: \section{Relationship between Parallel Wells and ``Close to Parallel''
463: Wells}
464: 
465: For a given $r$, we expect that a configuration with small but non-zero
466: angle will have a  power-law regime slope very close to the
467: (effective) power-law regime slope of a configuration of parallel lines
468: with the same $W$. This at first leads to a seeming paradox: if we
469: increase or decrease $W$, but keep the angle of the non-parallel wells
470: fixed, the slope of the 2-lines regime of the non-parallel well
471: configuration doesn't change as discussed in Section III.A.1.  However,
472: if we consider this configuration as a parallel configuration, changing
473: W changes the aspect ratio which changes the power-law regime slope as
474: discussed in Section IV.A.  This seeming inconsistency is resolved as
475: follows: on the one hand, for a 2-lines regime to exist, W must be at
476: least as large as the value given by Eq.~(\ref{e.16}.  If $W$ is too
477: small, there will be no 2-lines regime and both the parallel and small
478: angle configurations will look like the configuration for 2 points.  On
479: the other hand, if $W$ is increased, keeping $r$ and $\theta$ fixed, the
480: greater the deviation from parallel lines and there is no reason why the
481: parallel and small-angle configurations should have the same slopes in
482: their power-law regimes.
483: 
484: Thus, only for the very small range of W for which the power-law regime
485: exists and for which the configuration with small but non-zero $\theta$
486: is ``close to parallel''(i.e. the difference between the values of $r$
487: and $r_{max}$ is small) should the slopes of the parallel configuration
488: and the configuration with small but non-zero $\theta$ be equal.  That
489: is,
490: %
491: \begin{equation}
492: \label{e.18}
493: \bar g[W/r]\approx g_{\ell}(\theta),
494: \end{equation}
495: %
496: where $\bar g(R)$ is defined in Section IV.A.
497: 
498: \section{Discussion and Summary}
499: 
500: Motivated by the need to more realistically model the geometries found
501: in oil recovery activities, we have determined the scaling form for the
502: distribution of shortest paths between two lines in 3 dimensional
503: percolation systems.  Using simple scaling arguments we explained the
504: rich fractal behavior of the shortest path in these systems.  A number
505: of open questions, however, remain:
506: 
507: \begin{itemize}
508: 
509: \item[{(i)}] From first principles, can one develop an expression for
510: $g_{\ell}(\theta)$?  An exact expression for $g_{\ell}$ for two points in
511: 2-dimensions was obtained by Ziff\cite{Ziff99} using conformal
512: invariance arguments.  Possibly this approach could be extended to find
513: $g_{\ell}$ for point-line and 2-line configurations, at least in
514: 2-dimensions.
515: 
516: \item[{(ii)}] How is the fact that the crossover from one power-law
517: regime to another does not scale with the exponent $d_{\mbox{\scriptsize
518: min}}$ explained?  Is this simply an artifact of corrections-to-scaling
519: which would disappear if we could simulate much larger systems or is the
520: scaling of the crossover actually anomalous in certain configurations?
521: 
522: \end{itemize}
523: 
524: %\subsubsection*{Acknowledgements}
525: \section*{Acknowledgements}
526: 
527: We thank Sergey Buldyrev, Nikolay Dokholyan, Youngki Lee, Eduardo Lopez,
528: Peter King, and Luciano da Silva for helpful discussions and BP-Amoco
529: for financial support.
530: 
531: \begin{references}
532: %\begin{thebibliography}{99}
533: 
534: 
535: \bibitem{Dokh}
536: N.~V. Dokholyan, Y. Lee, S. V. Buldyrev, S. Havlin, P. R. King and H. E.
537: Stanley, J. Stat. Phys. {\bf 93}, 603 (1998);
538: N.~V. Dokholyan, S. V. Buldyrev, S. Havlin, P. R. King, Y. Lee, and H.
539: E. Stanley, Physica A {\bf 266}, 53 (1999).
540: 
541: \bibitem{Lee60}
542: Y. Lee, J. S. Andrade Jr., S. V. Buldyrev, N. Dokholyan, S. Havlin,
543: P. R.  King, G. Paul, and H. E. Stanley, Phys. Rev. E {\bf 60}, 3425
544: (1999).
545: 
546: \bibitem{Lee62}
547: J. S. Andrade Jr., S. V. Buldyrev, N. Dokholyan, S. Havlin, P. R.  King,
548: Y. Lee, G. Paul, and H. E. Stanley, Phys. Rev. E {\bf 62}, 8270 (2000).
549: 
550: \bibitem{Grassberger}
551: P. Grassberger, J. Phys. A: Math. Gen. {\bf 32} 6233 (1999).
552: 
553: \bibitem{Ziff99}
554: R. M. Ziff, J. Phys. A {\bf 32}, L457 (1999).
555: 
556: \bibitem{Neumann} A. U. Neumann and S. Havlin, J. Stat. Phys. {\bf
557: 52},203 (1988).
558: 
559: \bibitem{Stauffer}
560: D. Stauffer and A. Aharony, {\em Introduction to Percolation Theory}
561: (Taylor \& Francis, Philadelphia, 1994).
562: 
563: \bibitem{BundeHavlin}
564: {\em Fractals and disordered systems}, edited by A. Bunde and S. Havlin
565: (Springer-Verlag, New York, 1996).
566: 
567: \bibitem{Ben-Avraham00} D. Ben-Avraham and S. Havlin, {\it Diffusion
568: and Reactions in Fractals and Disordered Systems\/} (Cambridge
569: University Press, Cambridge, 2000).
570: 
571: \bibitem{Grassberger02} P. Grassberger, cond-mat/0201313.
572: 
573: \bibitem{Paul01}
574: G. Paul, R. M. Ziff, and H. E. Stanley, Phys. Rev. E {\bf 64}, 026115
575: (2001). 
576: 
577: \bibitem{Leath}
578: P. L. Leath, Phys. Rev. B {\bf 14}, 5046 (1976).
579: 
580: 
581: \end{references}
582: %\end{thebibliography}
583: 
584: \end{multicols}  
585: 
586: \newpage
587: 
588: %%%%%%%%%%%%%%%%   configuration(overview)
589: 
590: \begin{figure}
591: 
592: \centerline{
593: \epsfxsize=7.0cm
594: \epsfclipon
595: \epsfbox{pPP1.eps}
596: }
597: 
598: \caption{Example configurations of two non-parallel lines which are
599: studied. (a) Simple configuration of lines of equal length. (b)
600: Configuration of lines of unequal length ($W_1>W_2$). (c) Configuration
601: in which shortest distance between lines does not terminate at the ends of
602: lines ($W_{a1}<W_2<W_{1b}$).}
603: \label{pPP1}
604: \end{figure}
605: %%%%%%%%%%%%%%%%%%
606: 
607: %%%%%%%%%%%%%%%%%   equal length - different slopes
608: \begin{figure}
609: 
610: \centerline{
611: \xsize
612: \epsfclipon
613: %\epsfbox{pEq1.eps}
614: }
615: 
616: \centerline{
617: \xsize
618: \epsfclipon
619: \epsfbox{pEq2.eps}
620: }
621: \caption{$P(\ell | r)$ vs $\ell$ for configuration of two lines of equal
622: length with $r=8$, $\theta=$ (from bottom to top) $3^\circ$ (filled
623: square), 6$^\circ$, 12$^\circ$, 20$^\circ$, 40$^\circ$, and 180$^\circ$
624: (unfilled square). The corresponding well lengths W are 4890, 2445,
625: 1224, 737, 374 and 128, respectively. The plots are normalized such that
626: the initial sections of plots are coincident.}
627: \label{pEq}
628: \end{figure}
629: %%%%%%%%%%%%%%%%%%
630: 
631: \newpage
632: %%%%%%%%%%%%%%%%%   plot of gl(theta) 
633: 
634: \begin{figure}[h]
635: 
636: \centerline{
637: \xsize
638: \epsfclipon
639: \epsfbox{pgla.eps}
640: }
641: 
642: \centerline{
643: \xsize
644: \epsfclipon
645: \epsfbox{pgl.eps}
646: }
647: 
648: \caption{$g_{\ell}(\theta)$ vs $\theta$. The solid line is a plot of
649: Eq.~(\protect\ref{e1x}).} 
650: \label{gl}
651: \end{figure}
652: %%%%%%%%%%%%%%%%%%
653: 
654: \newpage
655: %%%%%%%%%%%%%%%%%   equal length - cutoffs 
656: 
657: \begin{figure}
658: 
659: \centerline{
660: \xsize
661: \epsfclipon
662: \epsfbox{p3.eps}
663: }
664: 
665: \centerline{
666: \xsize
667: \epsfclipon
668: \epsfbox{p28.eps}
669: }
670: 
671: \centerline{
672: \xsize
673: \epsfclipon
674: \epsfbox{p180.eps}
675: }
676: \caption{$P(\ell | r)$ vs $\ell$ for configuration of two lines of equal
677: length. (a) $r=8$, $\theta=3^\circ$, $W=$ (from top to bottom) 38, 76,
678: 152, 304, 1216, and 2432 (b) $r=1$, $\theta=29^\circ$, $W=$ (from bottom
679: to top) 8, 17, 33, 66, 132, 265, and 529, (c) $r=1$, $\theta=180^\circ$,
680: $W=$ (from bottom to top) 8, 16, 32, 64, and 128. For all plots, the
681: larger the value of $W$, the larger the value of $\ell$ at which
682: behavior changes from 2-lines behavior to 2-points behavior for which
683: the slope is $-2.35$. The insets plot the crossover value, $\hat\ell$,
684: vs. $r_{\mbox{\scriptsize max}}$.}
685: \label{pEqCutoff}
686: \end{figure}
687: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
688: 
689: \newpage
690: %%%%%%%%%%%%%%%%   configuration(point line)
691: 
692: \begin{figure}
693: 
694: \centerline{
695: \epsfxsize=7.0cm
696: \epsfclipon
697: \epsfbox{pPP2.eps}
698: }
699: \caption{Example configurations in which one line is of finite length $W$
700: and one is of zero length (i.e., a point). In all cases, the shortest
701: distance between the point and the line is $r$.}
702: \label{pPtLineConfig}
703: \end{figure}
704: 
705: 
706: %%%%%%%%%%%%%%%%% point-line
707: 
708: \begin{figure}
709: \centerline{
710: \xsize
711: \epsfclipon
712: \epsfbox{pPtLne1.eps}
713: }
714: \caption{$P(\ell | r)$ vs $\ell$ for configuration of a point and a line
715: with $r=1$ and $W=128$. From top to bottom, the plots are for the
716: configurations shown in Figs.~\protect\ref{pPtLineConfig}(a), (c), and
717: (b) respectively. We see that the slopes in configurations where the
718: point is closest to the end of the line
719: [Fig.~\protect\ref{pPtLineConfig} (a) and (c)] are the same (with some
720: initial difference) and they are different from the slope in the
721: configurations in which the point is closest to the middle of the line
722: (Fig.~\protect\ref{pPtLineConfig}b).}
723: \label{pPtLine1}
724: \end{figure}
725: 
726: \begin{figure}
727: \centerline{
728: \xsize
729: \epsfclipon
730: \epsfbox{pPtLne2.eps}
731: }
732: \caption{$P(\ell | r)$ vs $\ell$ for configuration of a point and a line
733: in which the point is closest to the end of the line with $r=1$ and $W=$
734: (from bottom to top) 2, 4, 8, 16, 32, 64, and 128. For all plots, the
735: larger the value of $W$, the larger the value of $\ell\protect\cong\hat\ell$
736: at which the behavior changes from point-line behavior to 2-points
737: behavior. The inset plots $\hat\ell$ vs.~$W$.}
738: \label{pPtLine2}
739: \end{figure}
740: 
741: 
742: %%%%%%%%%%%%%%%%%%%%%
743: \newpage
744: 
745: %%%%%%%%%%%%%%%%%%%%   different lengths 
746: \begin{figure}
747: 
748: \centerline{
749: \xsize
750: \epsfclipon
751: \epsfbox{pDiff7.eps}
752: }
753: 
754: \centerline{
755: \xsize
756: \epsfclipon
757: %\epsfbox{pDiff28.eps}
758: }
759: 
760: \centerline{
761: \xsize
762: \epsfclipon
763: \epsfbox{pDiff180.eps}
764: }
765: \caption{$P(\ell | r)$ vs $\ell$ for configurations of two lines of
766: different lengths with $r=1$ and $W_1=128$. (a) $\theta=7^\circ$, $W_2=$ (from
767: top to bottom) 16, 32, and 64. Three power law regimes can be seen: the
768: 2-lines regime, the point-line regime and the 2 points regime. (b)
769: $\theta=180^\circ$, $W_2=$ (from bottom to top) 4, 8, 16, 32, 64, and
770: 128. Only the first two power law regimes can be seen: the 2-lines
771: regime and the point-line regime (the 2-points regime would require even
772: larger values of $\ell$).}
773: \label{pDiff}
774: \end{figure}
775: 
776: 
777: \begin{figure}
778: 
779: \centerline{
780: \xsize
781: \epsfclipon
782: \epsfbox{pOverlap.eps}
783: }
784: \caption{$P(\ell | r)$ vs $\ell$ for configurations of two lines of
785: different lengths which ``overlap'' [see Fig.~\protect\ref{pPP1}(c)]
786: with $\theta=7^\circ$, $r=1$, $W_{1a}=32$, $W_{1b}=128$, and
787: $W_{2}=256$. Four power law regimes are present: the 2-line
788: ($\theta=7^\circ$) regime (slope $-3.0$), the 2-line
789: ($\theta=180^\circ-7^\circ$) regime (slope $-1.2$), the point-line
790: regime (slope $-1.75$), and the 2-points regime (slope $-2.35$).}
791: \label{pOverlap}
792: \end{figure}
793: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
794: \newpage
795: %%%%%%%%%%%%%%%%   non-coPlanar
796: 
797: \begin{figure}
798: 
799: \centerline{
800: \xsize
801: \epsfclipon
802: \epsfbox{pNonCoplanar.eps}
803: }
804: \caption{$P(\ell | r)$ vs $\ell$ for configurations of two lines of
805: equal length. The co-planar configuration has $r=1$, $\theta=90^\circ$, and
806: $W=256$ and the lines are co-planar. The non-coplanar configuration is
807: obtained from the co-planar configuration by moving one of the
808: lines a distance 8 perpendicular to the plane defined by the coplanar
809: lines. One sees that for large $\ell$, the power law regimes of the two
810: plots have the same exponent.}
811: \label{pNonCoPlanar}
812: \end{figure}
813: 
814: %%%%%%%%%%%%%%%%%
815: 
816: 
817: 
818: %%%%%%%%%%%%%%%%   configuration(parallel lines)
819: 
820: \begin{figure}
821: 
822: \centerline{
823: \epsfxsize=7.0cm
824: \epsfclipon
825: \epsfbox{pPP3.eps}
826: }
827: 
828: \caption{Example configurations for parallel wells. (a) Simple
829: configuration of wells of equal length. (b) Configuration of wells of
830: unequal length ($W_1>W_2$). (c) Configuration in which shortest line
831: between end of one well does not terminate at end of other
832: well ($W_{a1}<W_2<W_{1b}$).}
833: \label{pPConfig}
834: \end{figure}
835: 
836: \newpage
837: 
838: %%%%%%%%%%%%%%%%   parallel (different aspect ratios
839: 
840: \begin{figure}
841: 
842: \centerline{
843: \xsize
844: \epsfclipon
845: \epsfbox{pParallel.eps}
846: }
847: \caption{$P(\ell | r)$ vs $\ell$ for configurations of two parallel
848: lines of equal length with $r=1$ and $W=$ (from top to bottom) 0 (two
849: points), 4, 8, 16, and 32. The slopes of the power law regimes of the
850: plots for all configurations is the same but the initial decay of the
851: plots increases with increasing $W$.}
852: \label{pPar}
853: \end{figure}
854: 
855: \newpage
856: 
857: %%%%%%%%%%%%%%%%   parallel(all with aspect ratio=32
858: 
859: \begin{figure}
860: 
861: \centerline{
862: \xsize
863: \epsfclipon
864: \epsfbox{pParallel32.eps}
865: }
866: 
867: \centerline{ 
868: \xsize 
869: \epsfclipon
870: \epsfbox{pParallel32c.eps} 
871: }
872: \caption{$P(\ell | r)$ vs $\ell$ for configurations of two parallel
873: lines of equal length with $(W,r)=$ (from top to bottom) (32,1), (64,2),
874: (96,3), (128,4), (160,5), and (196,6) (b) plots of (a) scaled with the
875: variable $x=\ell/r^{d_{\mbox{\scriptsize min}}}$. The plots in (b)
876: collapse nicely as would be expected since they all have the same
877: aspect ratio, $W/r$. The good collapse for small $x$ indicates that
878: the small $x$ behavior
879: is not a lattice effect.}  
880: \label{pPar32}
881: \end{figure}
882: 
883: 
884: \newpage
885: 
886: %%%%%%%%%%%%%%%%   parallel(plots of most probable(max) versus aspect ratio
887: 
888: 
889: \begin{figure}
890: 
891: \centerline{
892: \xsize
893: \epsfclipon
894: \epsfbox{pMax16.eps}
895: }
896: 
897: \centerline{
898: \xsize
899: \epsfclipon
900: \epsfbox{pMax32.eps}
901: }
902: 
903: \centerline{
904: \xsize
905: \epsfclipon
906: \epsfbox{pMax64.eps}
907: }
908: 
909: \centerline{
910: \xsize
911: \epsfclipon
912: \epsfbox{pMaxc.eps}
913: }
914: \caption{Most probable $\ell$ vs $r$ for configurations of two parallel lines
915: of equal length. (a) $W=16$, $r=1$, 2, 4, 8, and 14. (b) $W=32$, $r=1$,
916: 2, 3, 4, 8, and 16. (c) $W=64$, $r=2$, 4, 5, 6, 7, 8, 16, and 32. (d)
917: Combined plot of (a), (b) and (c). The upper and lower dashed lines have
918: slope $d_{\mbox{\scriptsize min}}$ (1.374) and 1.0, respectively. The
919: larger the value of $W$, the larger the value of $r$ at which scaling
920: crosses over from Euclidean behavior to fractal behavior.}
921: \label{pMax32}
922: \end{figure}
923: 
924: 
925: %%%%%%%%%%%%%%%%   parallel(theory of cross over to Euclidean behavior)
926: 
927: \begin{figure}
928: 
929: \centerline{
930: \xsize
931: \epsfclipon
932: \epsfbox{pMaxt.eps}
933: }
934: \caption{Value of $r$ at which behavior changes from Euclidean to
935: fractal, $r^\ast$, $W$. The dashed line is a prediction of
936: Eq.~(\protect\ref{e.51}).}
937: \label{pMax32t}
938: 
939: \end{figure}
940: 
941: \end{document}
942: