1: %\documentclass[aps,prl,twocolumn]{revtex4}
2: %\documentstyle[aps,prl,multicol,epsf]{revtex}
3: %\documentstyle[aps,prl,multicol,epsf,graphics]{revtex}
4: %\documentstyle[aps,prl,twocolumn,epsf,graphics]{revtex}
5: %\documentstyle[pra,aps,eqsecnum,epsf]{revtex}
6: \documentstyle[aps,pra,multicol,psfig]{revtex}
7: %\usepackage{graphicx}
8: \begin{document}
9:
10: %\title{Disorder dominated critical behavior of the random Potts model}
11: %\title{Exact critical exponents of the 2d random Potts model in the large-$q$ limit}
12: \title{Phase transition in the 2d random Potts model in the large-$q$ limit}
13:
14: \author{J-Ch. Angl\`es d'Auriac$^{1}$ and
15: F. Igl\'oi$^{2,3}$}
16:
17: %\affiliation{
18: \address{
19: $^1$ CNRS-CRTBT B. P. 166, F-38042 Grenoble, France\\
20: $^2$ Research Institute for Solid State Physics and Optics,
21: H-1525 Budapest, P.O.Box 49, Hungary\\
22: $^3$ Institute of Theoretical Physics,
23: Szeged University, H-6720 Szeged, Hungary
24: }
25:
26: \date{\today}
27:
28: \maketitle
29:
30: \begin{abstract}
31: Phase transition in the two-dimensional $q$-state Potts model
32: with random ferromagnetic couplings in the large-$q$ limit is
33: conjectured to be described by the isotropic version of the infinite
34: randomness fixed point of the random transverse-field Ising spin
35: chain. This is supported by extensive numerical studies with a
36: combinatorial optimization algorithm giving estimates for the
37: critical exponents in accordance with the conjectured values:
38: $\beta=(3-\sqrt{5})/4$, $\beta_s=1/2$ and $\nu=1$. The specific
39: heat has a logarithmic singularity, but at the transition point there are
40: very strong sample-to-sample fluctuations. Discretized randomness results
41: in discontinuities in the internal energy.
42: \end{abstract}
43:
44: %\maketitle
45:
46: \newcommand{\bc}{\begin{center}}
47: \newcommand{\ec}{\end{center}}
48: \newcommand{\be}{\begin{equation}}
49: \newcommand{\ee}{\end{equation}}
50: \newcommand{\beqn}{\begin{eqnarray}}
51: \newcommand{\eeqn}{\end{eqnarray}}
52:
53: \begin{multicols}{2}
54: \narrowtext
55:
56: Quenched disorder could often change the properties of phase transitions
57: but exact information is scarce about the singularities in random fixed points,
58: in particular about isotropic classical systems with short range interactions.
59: For most of random classical systems - such as in spin glasses and random ferromagnets -
60: the critical behavior is governed by conventional random fixed
61: points (CRFP-s), in which the strength of disorder remains finite under renormalization.
62:
63: There is, however, a class of strongly anisotropic two-dimensional (2d)
64: random models, in which the disorder is strictly
65: correlated in 1d and there are exact results about
66: their critical behavior. These models, such as
67: the McCoy-Wu model\cite{mccoywu}, are closely related to random quantum spin chains,
68: and their critical behavior is
69: usually governed by an infinite randomness fixed point\cite{2drg} (IRFP), in which - under
70: renormalization - the strength of disorder growths without limits.
71: In an IRFP the singularities are primary determined by disorder effects and for almost
72: every studied, strongly disordered, 1d quantum (or 2d classical) systems the critical exponents
73: are the same as in the universality class of the random
74: transverse-field Ising model\cite{fisher} (RTIM).
75:
76: Having in mind the known relation between the critical behavior of 2d classical and
77: corresponding 1d quantum systems\cite{kogut} with non-random couplings one might ask the
78: question if a similar correspondence exists for random systems and in particular what is the
79: 2d isotropic classical counterpart of the RTIM. Some hints about a possible 2d isomorphism
80: of the RTIM is presented in Ref.\onlinecite{profile} in which the operator profiles of the RTIM
81: are found to be conformally invariant, the property of which should be shared with
82: some 2d isotropic random system, which - by definition - could obey conformal
83: symmetries\cite{cardy}.
84:
85: In this Letter we conjecture that the possible candidate for this r\^ole is
86: the $q$-state Potts model\cite{Wu} with random ferromagnetic bonds (random bond Potts model - RBPM)
87: in the large-$q$ limit and present the following arguments. First, we note that in the above
88: limit the high-temperature expansion of the
89: RBPM is dominated by a single diagram\cite{JRI01}, $\cal F$, so that the critical behavior is primary
90: determined by disorder effects as in the IRFP of the RTIM. Second, we have performed extensive
91: numerical calculations, in which $\cal F$ is exactly determined by a very efficient combinatorial
92: optimization algorithm\cite{aips02} and the obtained bulk and surface magnetization scaling dimensions,
93: $x$ and $x_s$, respectively, are found in good agreement with the corresponding exact values for
94: the RTIM\cite{fisher}:
95: %
96: \be
97: x=\frac{3-\sqrt{5}}{4},\quad x_s=\frac{1}{2}\;.
98: \label{exponents}
99: \ee
100: %
101: Third, we point out on apparent topological similarities between the ground state of
102: the RTIM and the fractal structure of the optimal graph, $\cal F$, which could then explain
103: the isomorphism between the two problems.
104:
105: In the following we introduce the RBPM by the Hamiltonian:
106: %
107: \be
108: -\frac{H}{kT}=\sum_{\langle ij \rangle} K_{ij} \delta(\sigma_i,\sigma_j)\;,
109: \label{hamilton}
110: \ee
111: %
112: where the spin variable at site $i$ is $\sigma_i=0,1,\dots,q-1$, $K_{ij}>0$
113: are random ferromagnetic couplings and the summation runs over nearest neighbor
114: pairs. In the random cluster representation the partition function of the system
115: is expressed in terms of $v_{ij}=\exp K_{ij}-1$ as\cite{JRI01}
116: %
117: \be
118: Z=\sum_F q^{C(F)} \prod_{ij \in F} v_{ij}\;,
119: \label{Z}
120: \ee
121: %
122: where the summation runs over all subset of bonds, $F$, and $C(F)$ is the number of
123: connected components of $F$, counting also the isolated sites. Having the
124: parameterization, $v_{ij}=q^{\alpha_{ij}}$ the partition function is expressed as
125: %
126: \be
127: Z=\sum_F q^{f(F)},\quad f(F)=C(F)+ \sum_{ij \in F} \alpha_{ij}\;,
128: \label{Z-q}
129: \ee
130: %
131: which in the large-$q$ limit is indeed dominated by the largest contribution,
132: $f^*={\rm max}_F f(F)$, and the partition function is
133: asymptotically given by $Z=N q^{f^*}$ where the number of optimal sets (OS-s), $N$, is
134: likely to be one.
135:
136: One of the few rigorous results about random systems is due to Aizenman and Wehr\cite{aizenmanwehr}
137: who states that in the 2d RBPM the internal energy is continuous at the phase transition point for
138: any $q$, if the probability distribution of the couplings is absolutely continuous.
139: In numerical calculations\cite{pottsmc,pottstm} one usually studies the properties of the
140: system at the phase transition point, which is known by duality\cite{dom_kinz}, and
141: universal, i.e. disorder independent, critical behavior
142: has been observed even for atomistic (c.f. bimodal) distributions.
143: The scaling dimension, $x$, is a monotonously increasing function of $q$, but its saturation value
144: in the large-$q$ limit is difficult to be estimated due to strong
145: logarithmic corrections\cite{jacobsenpicco}.
146: The correlation length exponent is close to $\nu=1$, for any
147: value of $q$. Note, however, that by transfer matrix calculation Jacobsen and Cardy\cite{pottstm}
148: have obtained $\nu<1$ for the bimodal distribution of disorder, which violates the rigorous
149: upper bound, $\nu \ge 2/d$\cite{ccfs}.
150:
151: In the large-$q$ limit the critical properties of the RBPM are related to the structure
152: of the clusters in the OS in close analogy with percolation\cite{staufferaharony}.
153: In the paramagnetic phase there are only finite clusters in the OS
154: and the linear extent of the largest clusters is used to define the correlation length, $\xi$.
155: In the ferromagnetic phase there is an infinite cluster and the ratio of lattice points
156: belonging to it is related to the (finite) magnetization of the RBPM.
157: Finally, at the phase transition point the largest (infinite) cluster is a fractal, and its
158: fractal dimension, $d_f$, is related to the magnetization scaling dimension as $d=d_f+x$.
159: Similarly, the sum of the surface
160: fractal dimension of the percolating cluster, $d_f^s$, and the anomalous dimension of the
161: surface magnetization, $x_s$, gives the Euclidean dimension of the surface:
162: $d^s=1=d_f^s+x_s$.
163:
164: In Ref.\onlinecite{JRI01} the OS was approximatively calculated by the simulated annealing
165: method. Results, obtained on relatively small lattices (up to $24
166: \times 24$), such as $x=0.17-0.19$ were consistent with the conjectured value in
167: Eq.(\ref{exponents}), however, due to
168: the relatively large error and also due to the lack of precise estimates about another
169: exponents we could not make a definite statement about the universality class
170: of the transition.
171:
172: In the present Letter we apply a recently developed combinatorial optimization
173: algorithm\cite{aips02} with which we can determine the {\it exact} OS in strongly
174: polynomial time, i.e. the time of computation does not depend on the form of the
175: disorder. With this algorithm we could treat far larger systems as before,
176: averages and distributions were calculated for systems with $L=32,~64,~128$ and $256$
177: over at least thousand disorder realizations.
178: (We have also analyzed a few samples of $512 \times 512$).
179: As a consequence our estimates of the critical exponents become much more accurate than before
180: and we studied, at the first time, also the surface properties of the RBPM.
181:
182: In most of the calculation the couplings and thus the parameters, $\alpha_{ij}$,
183: were taken from a bimodal distribution:
184: $P(\alpha)=[\delta(w+\Delta w - \alpha) + \delta(w-\Delta w - \alpha)]/2$
185: with $w > \Delta w > 0$, so that the distance of the critical temperature, $t$,
186: is measured by $t \approx 1 - 2w$. Having the parameter $\Delta w=1/3$
187: the microscopic length-scale in the problem\cite{JRI01},
188: $l_c \approx (2 \Delta w)^{-2}$, was not too large. We also used the continuous
189: (uniform) distribution: $P_u(\alpha)=1/(2 u)$, for $0 \le \alpha < u$ and zero otherwise.
190: Here the distance from the critical point
191: is given by $t \approx 1-u$. To calculate bulk (surface) quantities
192: we applied periodic (open) boundary conditions (b.c.-s).
193:
194: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG. 1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
195: \begin{figure}[b]
196: \centerline{\psfig{file=fig1.ps,height=12.75cm}}
197: \vskip 0.2truecm
198: \caption{top: OS for a typical disorder realization at the critical point on
199: a $256 \times 256$ lattice with periodic boundary conditions. Percolating and finite clusters
200: are marked with gray (green) and dark (red) lines, respectively. middle: Enlargement of a square
201: proportion of size $48 \times 48$ in the upper-middle part of the OS to illustrate
202: self-similarity. bottom: The connectivity structure of the OS along a line, which
203: consists of six connected units (``spins'') and five open units (``bonds'').}
204: \label{FIG01}
205: \end{figure}
206: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
207:
208: First, we considered the behavior of the system at the transition point, when the cluster
209: structure of a typical OS is shown in Fig. \ref{FIG01}. As illustrated in the middle
210: of Fig. \ref{FIG01} the OS is self-similar and its topology, being isotropic,
211: can be conveniently represented by the connectivity structure (CS) of the OS
212: at a given line, as shown in the bottom of Fig. \ref{FIG01}. We are going to analyze
213: the CS after presenting the numerical results.
214:
215: We have checked that asymptotically half of the sites of the OS are
216: isolated: the probability
217: $P(L)$ that a given site belongs to a cluster having two or more sites, behaves like
218: $P(L) \simeq 1/2 \left( 1 + 1/(2\ln L) \right)$.
219: We have also checked that the largest
220: connected cluster is indeed a fractal (and not a multi-fractal). Its fractal dimension, $d_f$, is calculated
221: by direct dimensional analysis and by finite-size scaling of the average mass of the largest
222: cluster. Here we present an analysis of the probability distribution function,
223: $R(m,L)$, which measures the fraction of clusters having a size at least $m$ and which,
224: according to scaling theory\cite{staufferaharony}, should asymptotically behave as:
225: %
226: \be
227: R(m,L)=m^{-\tau} \tilde{R}(m/L^{d_f})\;,
228: \label{distr_m}
229: \ee
230: %
231: with $\tau=(2-d_f)/d_f$. With the conjectured value of $d_f=2-x=(5+\sqrt{5})/4=1.809$
232: we obtained a very good scaling collapse, which is shown in Fig.\ref{FIG02}.
233:
234: To characterize the accuracy of the collapse we have
235: measured the surface of the overlap of the scaled curves with varying value of $d_f$.
236: As shown in the inset a) to Fig. \ref{FIG02} the best collapse is indeed around the
237: conjectured value, so that our estimate $x=0.190(5)$ has a relatively small error.
238:
239: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG. 2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
240: \begin{figure}[b]
241: \centerline{\psfig{file=fig2.ps,height=5.75cm,angle=-90}}
242: \vskip 0.2truecm
243: \caption{Scaling collapse of the reduced probability distribution function, $R(m,L) m^{\tau}$,
244: in Eq.(\ref{distr_m}) with $d_f=(5+\sqrt{5})/4=1.809$. Inset a): Surface of the
245: collapse region for different values of $d_f$. The arrow shows the conjectured value.
246: b) Average mass of surface sites of the largest cluster vs. linear size of the system
247: in a log-log plot. The slope of the straight line corresponds to $d_f^s=0.495(10)$.}
248: \label{FIG02}
249: \end{figure}
250: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
251:
252: Next, we considered the surface magnetization properties of the model by
253: using open b.c.-s and measuring the average mass of surface sites of the largest cluster, $m_s$.
254: As shown in the inset b) to Fig. \ref{FIG02} it has an asymptotic size dependence
255: $m_s \sim L^{d_f^s}$, with $d_f^s=0.495(10)$, which is in very good agreement with the
256: conjectured value of $d_f^s=1-x_s=1/2$, see Eq.(\ref{exponents}).
257:
258: In the following we analyze in more details the fractal structure of the OS and point out
259: the topological similarities with the ground state wave function of the RTIM, which could
260: explain the appearance of the same critical exponents in the two problems. As already noted
261: in the bottom of Fig.\ref{FIG01} the CS consists of connected units (CU-s)
262: (corresponding to spins in the RTIM) of variable length, $l_s$, and moments, $\mu$, and of
263: open units (OU-s) (corresponding to bonds in the RTIM) of variable
264: length, $l_b$. If two neighboring CU-s, with parameters $l_s^1$, $\mu^1$ and $l_s^2$, $\mu^2$,
265: separated with an OU of $l_b$, belong to the same cluster, it is merged to an effective
266: CU (represented by a connecting line in Fig. \ref{FIG01}), with length
267: $l_s^{12}=l_s^1+l_s^2+l_b$ and moment $\mu^{12}=\mu^1+\mu^2$. This is precisely
268: equivalent of a strong bond decimation in the RTIM, which is one ingredient of the
269: strong disorder renormalization group (SDRG) method\cite{mdh,fisher}. Similarly, if a CU
270: with $l_s$ and $\mu$ and with neighboring bonds of lengths: $l_b^1$ and $l_b^1$, is isolated,
271: it does not contribute to any larger cluster, therefore - at larger length-scales - can be
272: eliminated (represented by an overgoing line in Fig. \ref{FIG01}) and the new effective
273: bond has a length of $l_b^{12}=l_b^1+l_b^2+l_s$. This process then equivalent of a strong
274: field decimation in the SDRG for the RTIM. Thus we can conclude that for any CS of the OS
275: in the RBPM one can construct an equivalent ground state of the RTIM,
276: and one can give a set
277: of couplings, $J_i$, and transverse fields,$h_i$, for which the given ground state is realized.
278: We recall that in the RTIM the bulk (surface) magnetization is related to the
279: average moment of a bulk (surface) effective spin\cite{fisher} in close analogy with the
280: computation of the same quantities in the OS of the RBPM.
281: If we now assume that in the two problems the statistics of the appearance of
282: states with equivalent topology is also similar, we arrive to the conjectured
283: relation about the critical exponents in Eq.(\ref{exponents}).
284:
285: In the following we turn to study the energy-density of the model outside the transition point.
286: To do this one should determine the OS at different temperatures,
287: however, in a finite system of size, $L$, there are only a finite
288: number of different OS-s, their typical
289: number being $L$, independently of the type of disorder. This result
290: can be interpreted that two neighboring OS-s differ in average by one
291: line ($\sim L$) of edges. For a given sample
292: the free energy is a piece-wise linear function of $t$ and the
293: internal energy is a step-like function having typically
294: $L$ steps. Averaging over disorder the average internal energy
295: becomes continuous for continuous distributions (see. Fig.\ref{FIG03}), whereas for discrete
296: distributions some discontinuities, located at special isolated points,
297: remain. (The discontinuous behavior of the internal energy in this
298: case is somewhat analogous to that of the magnetization of the random-field Ising model at $T=0$
299: having the same bimodal distribution\cite{RFIM}.)
300: There is a discontinuity at the phase-transition (self-dual) point, as we
301: illustrate with the sequence of finite-size latent heats:
302: $[\Delta E(32)]_{\rm av}=0.0456(90)$, $[\Delta E(64)]_{\rm av}=0.0477(52)$,
303: $[\Delta E(128)]_{\rm av}=0.0484(28)$ and $[\Delta E(256)]_{\rm av}=0.0474(14)$
304: for the bimodal distribution with $\Delta w=1/3$, which approach a
305: finite value in the thermodynamic limit. The discontinuities in the
306: internal energy are due to degeneracies, which are connected also to finite
307: clusters. For example at the self-dual point any lattice point which
308: has two couplings - one weak and one strong - to an existing cluster, could
309: either be connected or disconnected. Thus the jumps in the
310: internal energy are related not exclusively to the largest cluster.
311: The contribution of the largest cluster only, which has a diverging size $\xi$,
312: is expected to cause singularity in the specific heat
313: at the two sides of the transition point.
314:
315: This true singularity of the
316: specific heat, which is expected to be independent of the type of
317: disorder, will be investigated numerically in the following.
318: Here, as usual, the thermodynamic limit should be taken
319: first and then approach the transition point.
320: In our numerical study we observed that close to the transition point,
321: in the order of $|t| \sim 1/L$, there are large sample to sample
322: fluctuations. As an illustration in the inset a) to Fig.\ref{FIG03}
323: we have shown the distribution of the finite-size specific heat,
324: $C^1_v$, defined for a given sample as the ratio of the distance
325: between the first two energy steps and the corresponding temperature
326: difference, just at the right of the transition point. The
327: distribution, as shown in the inset has a broad, power-law tail
328: and its second moment does not exist. This fact represents the
329: strong randomness character of the transition. Using continuous
330: distribution of disorder the strong fluctuation regime close to the
331: transition point remains qualitatively the same. Therefore, to determine
332: the thermodynamic singularity of the specific heat we analyzed its
333: behavior for $|t|>1/L$. As presented in the inset b) to Fig.\ref{FIG03}
334: the specific heat has a logarithmic singularity of the form of
335: $C_v(t) \sim (ln|t|)^{\epsilon}$, with $\epsilon \le 1$.
336: Thus the specific heat exponent is $\alpha=0$ and
337: the correlation length exponent is $\nu=1$. This latter result is the half
338: of the similar exponent of the RTIM in the strongly anisotropic IRFP\cite{fisher}.
339:
340: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% FIG. 3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
341: \begin{figure}[b]
342: \centerline{\psfig{file=fig3.ps,height=5.75cm}}
343: \vskip 0.2truecm
344: \caption{Average internal energy for the continuous
345: distribution. Inset a): distribution of the finite-size latent heat
346: at the right side of the transition point for the bimodal
347: distribution for $L=32$ (left) and $L=128$ (right). Inset b): the
348: average specific heat for the continuous
349: distribution as a function of $\ln |t|$ in the region $Lt \gg 1$.}
350: \label{FIG03}
351: \end{figure}
352: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
353:
354: To conclude our investigations have shown that the critical behavior
355: of the RBPM in the large-$q$ limit is dominated by strong disorder
356: effects and possibly related to the IRFP of the RTIM. The
357: conjectured exact values of the critical exponents are checked by
358: extensive numerical calculations. If an asymptotically exact
359: RG treatment - in the same spirit as the SDRG for the RTIM - can be
360: constructed, should be clarified by future research.
361:
362: We are indebted to R. Juh\'asz,
363: M. Preissmann, H. Rieger, A. Seb\H o and L. Turban for stimulating discussions.
364: This work has been supported by the Hungarian National
365: Research Fund under grant No OTKA TO34183, TO37323,
366: MO28418 and M36803, by the Ministry of Education under grant No FKFP 87/2001,
367: by the EC Centre of Excellence (No. ICA1-CT-2000-70029).
368:
369: \begin{references}
370:
371: \bibitem{mccoywu}
372: B.M. McCoy and T.T. Wu,
373: Phys. Rev. {\bf 176}, 631 (1968); {\bf 188}, 982(1969);
374: B.M. McCoy, Phys. Rev. {\bf 188}, 1014 (1969).
375:
376: \bibitem{2drg}
377: O. Motrunich, S.-C. Mau, D.A. Huse and D.S. Fisher, Phys. Rev. B{\bf 61}, 1160 (2000).
378:
379: \bibitem{fisher}
380: D.S. Fisher, Phys. Rev. Lett. {\bf 69}, 534 (1992);
381: Phys. Rev. B {\bf 51}, 6411 (1995).
382:
383: \bibitem{kogut}
384: J. Kogut, Rev. Mod. Phys. {\bf 51}, 659 (1979).
385:
386: \bibitem{profile}
387: F. Igl\'oi and H. Rieger, Phys. Rev. Lett. {\bf 78}, 2473 (1997).
388:
389: \bibitem{cardy}
390: For a review see: J.L. Cardy, in {\it Phase Transitions and Critical Phenomena},
391: edited by C. Domb and J.L. Lebowitz (Academic Press, London, 1987), Vol. 11.
392:
393: \bibitem{Wu}
394: F.Y. Wu, Rev. Mod. Phys. {\bf 54}, 235 (1982).
395:
396: \bibitem{JRI01}
397: R. Juh\'asz, H. Rieger, and F. Igl\'oi, Phys. Rev. E{\bf 64}, 056122 (2001).
398:
399: \bibitem{aips02}
400: J.-Ch. Angl\`es d'Auriac, F. Igl\'oi, M. Preissmann, and A. Seb\H{o},
401: J. Phys. A{\bf 35}, 6973 (2002).
402:
403: \bibitem{aizenmanwehr}
404: M. Aizenman and J. Wehr, Phys. Rev. Lett. {\bf 62}, 2503
405: (1989); errata {\bf 64}, 1311 (1990).
406:
407: \bibitem{pottsmc}
408: M. Picco, Phys. Rev. Lett. {\bf 79}, 2998 (1997); C. Chatelain and B. Berche,
409: Phys. Rev. Lett. {\bf 80}, 1670 (1998); Phys. Rev. E{\bf 58} R6899 (1998);
410: {\bf 60}, 3853 (1999); T. Olson and A.P. Young, Phys. Rev. B{\bf 60}, 3428 (1999).
411:
412: \bibitem{pottstm}
413: J.L. Cardy and J.L. Jacobsen, Phys. Rev. Lett. {\bf 79}, 4063 (1997),
414: J.L. Jacobsen and J.L. Cardy, Nucl. Phys. B{\bf 515}, 701 (1998).
415:
416: \bibitem{dom_kinz}
417: W. Kinzel and E. Domany, Phys. Rev. B{\bf 23}, 3421 (1981).
418:
419:
420: \bibitem{jacobsenpicco}
421: J.L. Jacobsen and M. Picco, Phys. Rev. E{\bf 61}, R13 (2000);
422: M. Picco (unpublished).
423:
424: \bibitem{ccfs}
425: J. T. Chayes, L. Chayes, D. S. Fisher and T. Spencer,
426: Phys. Rev. Lett. {\bf 57}, 299 (1986).
427:
428: \bibitem{staufferaharony}
429: For a review see, Stauffer and A. Aharony, {\it Introduction to Percolation Theory},
430: (Taylor and Francis, London) (1992).
431:
432: \bibitem{mdh}
433: S.K. Ma, C. Dasgupta and C.-K. Hu, Phys. Rev. Lett. {\bf 43}, 1434 (1979);
434: C. Dasgupta and S.K. Ma, Phys. Rev. B{\bf 22}, 1305 (1980).
435:
436: \bibitem{RFIM}
437: J. C. Angl\`es d'Auriac and Nicolas Sourlas.\\
438: Europhys. Lett. {\bf 39}, 473 (1997).
439:
440:
441: \end{references}
442: \end{multicols}
443: \end{document}
444:
445: