1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

3: 

4: % Some other (several out of many) possibilities

5: %\documentclass[preprint,aps]{revtex4}

6: %\documentclass[preprint,aps,draft]{revtex4}

7: %\documentclass[prb]{revtex4}% Physical Review B

8: 

9: \usepackage{graphicx}% Include figure files

10: \usepackage{dcolumn}% Align table columns on decimal point

11: \usepackage{bm}% bold math

12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

13: \begin{document}

14: %------------------------------------------------------------------------------

15: \title{Roughness of Interfacial Crack Front: Correlated Percolation in

16: the Damage Zone}

17: %

18: \author{Jean Schmittbuhl\footnote{Permanent Address:

19: Departement de G{\'e}ologie, UMR CNRS 8538, Ecole Normale

20: Sup{\'e}rieure, 24, rue Lhomond, F--75231 Paris C{\'e}dex 05,

21: France. Email: schmittb@geologie.ens.fr.}

22: %

23: and Alex Hansen\footnote{Permanent Address: Department of Physics,

24: NTNU, N--7491 Trondheim, Norway. Email: Alex.Hansen@phys.ntnu.no.}}

25: \affiliation{International Center for Condensed Matter Physics,

26: Universidade de Bras{\'\i}lia, 70919--970 Bras{\'\i}lia, Distrito

27: Federal, Brazil}

28: %

29: \author{G.\ George Batrouni\footnote{Email:

30: george.batrouni@inln.cnrs.fr.}}  \affiliation{Institut Non-Lin\'eaire

31: de Nice, UMR CNRS 6618, Universit{\'e} de Nice-Sophia Antipolis, 1361

32: Route des Lucioles, F--06560 Valbonne, France} \date{\today}

33: %--------------------------------------------------------------------

34: \begin{abstract} 

35: We show that the roughness exponent $\zeta$ of an in-plane crack front

36: slowly propagating along a heterogeneous interface embeded in a

37: elastic body, is in full agreement with a correlated percolation

38: problem in a linear gradient.  We obtain $\zeta=\nu/(1+\nu)$ where

39: $\nu$ is the correlation length critical exponent. We develop an

40: elastic brittle model based on both the 3D Green function in an

41: elastic half-space and a discrete interface of brittle fibers and find

42: numerically that $\nu=1.5$. We conjecture it to be 3/2. This yields 

43: $\zeta=3/5$. We

44: also obtain by direct numerical simulations $\zeta=0.6$ in excellent

45: agreement with our prediction. This modelling is for the first time in

46: close agreement with experimental observations.

47: \end{abstract} 

48: \pacs{83.80.Ab, 62.20.Mk, 81.40.Np}

49: \maketitle

50: %--------------------------------------------------------------------

51: An important motivation for studying interfacial crack pinning

52: \cite{srvm95,sm97} is to simplify the study of the origin of the

53: scaling properties of brittle crack surfaces \cite{mpp84,bs85}.  These

54: scaling properties are seen, for example, in the height-height

55: correlation of the fracture roughness ({\it i.e.}  out-plane

56: roughness), which shows self-affinity.  That is, the conditional

57: probability density $p(x,h)$, {\it i.e.}\ the probability that the

58: crack surface passes within $dh$ of the height $h$ at position $x$

59: when it had height zero at $x=0$, shows the invariance

60: \begin{equation}

61: \label{hhscaling}

62: \lambda^\zeta p(\lambda x,\lambda^\zeta h)=p(x,h)\;,

63: \end{equation}

64: where $\zeta$ is the roughness exponent.  It is now generally believed

65: that the roughness exponent shows a universal value of about 0.80 at

66: larger scales \cite{blp90-mhhr92-sgr93-cw93-sss95}, while a lower

67: value of about 0.5 might be seen at smaller scales

68: \cite{dhbc96-dnbc97}.

69: 

70: Direct observations of the interfacial crack-pinning have been

71: performed recently. The problem consists of following the roughness of

72: a crack front moving along the flat interface between two elastically

73: connected blocks.  The experimental study of constrained crack

74: propagation between two sintered plexiglass plates presented in

75: \cite{sm97} resulted in the estimate of the in-plane roughness

76: exponent: $\zeta=0.55\pm 0.05$.  This work was followed by a longer

77: study leading to the estimate $\zeta=0.63\pm 0.03$ \cite{dsm99}.  A

78: recent study of the motion of a helium-4 meniscus along a disordered

79: substrate --- a problem closely related to the motion of a crack line

80: --- gave $\zeta=0.56\pm0.03$ \cite{prg02}.

81: 

82: Numerous models for interfacial crack propogation in heterogeneous

83: material have been proposed. The numerical simulation presented in

84: \cite{srvm95} is based on a perturbative Green function approach

85: following the quasistatic evolution of the interfacial crack front

86: position $a(x,t)$ \cite{gr89}, which is treated as a single-valued

87: function of position $x$ along the orthogonal direction to the crack

88: advancement direction, and time $t$. The linearized Green function

89: used binds together points only along the fracture front. The stress

90: intensity factor at a point $x$ along the fracture front is then found

91: to be

92: \begin{equation}

93: \label{gaorice}

94: \frac{K(x,t)}{K_0}=1+\frac{1}{2\pi}\ {\rm p.v.} \int_{-\infty}^{+\infty}

95: \frac{a(x',t)-a(x,t)}{(x-x')^2}\ dx'\;,

96: \end{equation}

97: where $K_0$ is the stress intensity factor that would result if the

98: crack were straight \cite{gr89}.  The fracture is advanced by

99: identifying the most stressed point along the fracture line and

100: advancing this by a small step.  The roughness exponent of the crack

101: front was estimated numerically to be $\zeta=0.35$, while a direct

102: dynamical renormalization group calculation gave $\zeta=1/3$ to lowest

103: order \cite{ek94}.  Higher-order corrections to this quasistatic

104: analysis increases the value of $\zeta$ to 0.48 \cite{cdw01}, while a

105: different quasistatic analytical technique suggests $0.390\pm0.002$

106: \cite{rk02}. Dynamic effects have beeen largely studied numerically

107: and analytically \cite{pr94-rf97-rf98-mr98-mr00}, in particular in the

108: form of crack front waves. They lead to an increased roughness

109: exponent compared to the initial 1/3-value up to $\zeta=0.5$ but still

110: smaller than experimental observations.

111: 

112: Hence, the situation today is that there is a large and significant

113: discrepancy between theoretical and experimental estimates: theoretical

114: estimates being considerably smaller than the experimental ones.

115: 

116: %--------------------------------------------------------------------

117: \begin{figure}

118: \includegraphics[width=2.7cm,angle=270]{rip-fig1.eps}

119: \caption{The crack front for a $128\times 128$ system. The fracture is

120: propagating from bottom to top.  The broken springs are black dots. The 

121: crack front is drawn as a white line.}

122: \label{fig1}

123: \end{figure}

124: %--------------------------------------------------------------------

125: 

126: In this letter we present a numerical calculation of the roughness

127: exponent of a crack front that propagates quasi-statically along a

128: heterogeneous interface, based on a Green function technique which

129: differs from that previously used. Indeed it does not reduce the crack

130: tip to a single tortuous line but describes the tip as a region of

131: interactions between microcracks (see Fig.~\ref{fig1}). Our technique

132: is based on the static solution of the elastic equations for the

133: deformation of the surface of an elastic half space \cite{ll58}.  The

134: local deformation, $u$, at position $(x,y)$ along the plane is related

135: to the normal stress field $\sigma$ by the expression

136: \begin{equation}

137: \label{gruf}

138: u(x,y)=\int\int G(x-x',y-y') \sigma(x',y')\ dx'\ dy'\;,

139: \end{equation}

140: where the Green function is \cite{j85}:

141: \begin{equation}

142: G(x,y)=\frac{1-s^2}{\pi e }\

143: \frac{1}{|(x,y)|}\;,

144: \label{thanksjohn}

145: \end{equation}

146: with $s$ is the Poisson ratio and $e$ the elastic constant.  For large 

147: $r=|(x-x',y-y')|$, $G\sim1/r$.

148: %The experimental system studied in \cite{sm97} consisted of inducing

149: %a mode I crack propagating along the interface between two sintered,

150: %identical elastic plates.  Hence, there is a symmetry plane along the

151: %unperturbed interface between the two plates.

152: For the sake of simplicity, as generally done for the study of contact

153: between two elastic bodies \cite{j85}, we substitute one of the

154: elastic plates by an infinitely rigid one.  The other plate is modeled

155: as an elastic block for which Eq.\ (\ref{thanksjohn}) is valid.  We

156: discretize the model, so that the forces and deformations are

157: described by the discrete version of Eq.\ (\ref{gruf}),

158: \begin{equation}

159: \label{M1}

160: u_i=\sum_j \overline{G}_{i,j} f_j\;.

161: \end{equation}

162: $\overline{G}_{i,j}$ is the Green function  Eq.\ (\ref{thanksjohn})

163: averaged over an area $b^2$,

164: \begin{equation}

165:  \overline{G}_{i,j}=\frac{1}{b^2}\ \int_{-b/2}^{+b/2} \int_{-b/2}^{+b/2}\ dx'\

166: dy' G(i-x',j-y')\;,

167: \label{avethanksjohn}

168: \end{equation}

169: where $b$ is the lattice spacing,  $u_i$ is the deformation of the

170: elastic body at site $i$, and $f_i$ the force acting at that point.

171: The indices $i$ and $j$ in Eq.\ \ref{M1} run over all $L^2$ sites.

172: 

173: The elastic block is connected to the infinitely stiff plate by a

174: discrete interface made of an array of elastic harmonic springs.  Each

175: spring is brittle and has a breaking threshold randomly drawn from a

176: uniform distribution between zero and one.  The spacing between the

177: springs is $b$ in both the $x$ and $y$ directions.  The force $f_i$

178: that an unbroken spring $i$ is carrying, is transferred over an area

179: of size $b^2$ to the soft surface and given by Hooke's law: 

180: \begin{equation}

181: \label{hooke}

182: f_i = -k(u_i-D)\;,

183: \end{equation}

184: where $k$ is the spring constant ($k=1$ for all springs). $D$ is the

185: displacement of the infinitely stiff medium, and is a function of $y$,

186: {\it i.e.} solid rotation of the stiff medium.  The quantity $(u_i-D)$

187: is, therefore, the length that spring $i$ is stretched.  We assume

188: periodic boundary conditions both in the $x$ and $y$ directions.  In

189: order to model the mode~I fracturing of the interface between the two

190: media (Double Cantilever Beam (DCB) test) in a way compatible with the

191: biperiodic boundary conditions, we let $D(y)$ be wedgeshaped with a

192: maximum at $y=0$ and $L$, and zero for $y=L/2$ ({\it i.e.} mirroring

193: conditions).  The load the system, $D(y)$ is increased and the springs

194: break one by one.  The numerical technique to solve the equations that

195: ensue is presented in detail in \cite{bhs02}.

196: 

197: %--------------------------------------------------------------------

198: \begin{figure}

199: \includegraphics[width=6.5cm,angle=270]{rip-fig2.eps}

200: \caption{Family-Vicsek scaling of the crack front roughness:

201: $w/L^{0.6}$ vs.\ $\langle a\rangle /L^{0.9}$ for different system

202: sizes: $L=16$, $32$, $64$ and $128$.}

203: \label{fig2}

204: \end{figure}

205: %--------------------------------------------------------------------

206: 

207: We show in Fig.~\ref{fig1} a typical damage front in a $128\times 128$

208: system.  We define the fracture front $a(x,t)$ in this model as the

209: set of nodes that form a continuous path separating the infinite

210: cluster of broken springs from the infinite cluster of unbroken

211: springs. This definition is similar to that of Ref.~\cite{sm97}. From

212: Fig.~\ref{fig1}, we clearly see that even if a front can be defined, an

213: extended damage zone exists. Accordingly the front does not capture

214: all the active tip of the fracture. This observation supports the

215: relevance of a non line description of the crack front.

216: 

217: In Fig.~\ref{fig2}, we show the width of the fracture front $w =

218: \sqrt{\langle a^2\rangle -\langle a\rangle^2}$ as a function of its

219: average position $\langle a\rangle$ --- which acts as a time parameter

220: in this quasi-static model, for various system sizes.  By collapsing

221: the width evolution for the different system sizes, we see that the

222: crack front follows a Family-Vicsek scaling \cite{fv85}. Two important

223: exponents are, thus, obtained: the roughness exponent

224: $\zeta=0.6$ and dynamical exponent $z=0.9$. The roughness exponent is

225: in close agreement with the experimentally obtained value

226: \cite{sm97,dsm99,ms01,prg02}, while the dynamical exponent $z$ was

227: found in Ref.~\cite{ms01} to be $1.2$.

228: 

229: Our model distinguishes itself from earlier numerical models in three

230: major ways: (1) Most previous models are based on a small perturbation

231: approach and include linearizations \cite{gr89} that are not used

232: here.  Furthermore, it was assumed in the earlier studies that (2) the

233: fracture line was a single-valued function, hence ruling out

234: overhangs.  This assumption prevents islands of unbroken bonds from

235: forming in the wake of the advancing fracture line.  (3) The

236: assumption of a single advancing fracture line also prevents the

237: formation of a damage zone in front of the fracture line.  None of

238: these three assumptions are necessary in the present model.  In order

239: to test whether assumptions (2) or (3) are responsible for the

240: difference in roughness exponent, we imposed both on the present

241: model.  No change in $\zeta$ was observed.  Hence, we conclude that it

242: is the linearization assumption that is responsable for the

243: difference.

244: 

245: %--------------------------------------------------------------------

246: \begin{figure}

247: \includegraphics[width=6.5cm,angle=270]{rip-fig3.eps}

248: \caption{Fluctuations of the density of broken bonds $W_c=(\langle

249: p_c^2\rangle -\langle p_c\rangle^2)^{1/2}$ plotted against $L$ for a

250: homogeneous parallel loading without any gradient, {\it i.e.} constant

251: D over the whole system.  The slope of the straight line is

252: $-1/\nu=-0.65$.}

253: \label{fig3}

254: \end{figure}

255: %--------------------------------------------------------------------

256: 

257: In the following, we demonstrate that the roughness exponent $\zeta$

258: is a result of a correlated percolation process in a gradient imposed

259: by the mode I loading of the system.  We start by charactarizing the

260: correlated percolation. For this goal, we consider a similar problem

261: but where the loading is obtained without any gradient, {\it i.e.}

262: horizontal and parallel displacement of the rigid medium

263: \cite{bhs02}. In this case, when the homogeneous displacement $D$ of

264: the rigid medium reaches a maximum value, $D_c$ the system goes

265: unstable, and unless $D$ is decreased again, catastrophic failure sets

266: in.  In Ref.~\cite{bhs02}, the size distribution of clusters of broken

267: springs was studied, and a power law was found with an exponent

268: $\tau=1.6$.  This value is different from ordinary percolation where

269: $\tau \approx 2.05$ \cite{sa92} and shows that correlated percolation

270: takes place.  In Fig.~\ref{fig3}, we show the fluctuations of density

271: of broken springs at $D=D_c$.  If there is a diverging correlation

272: length in the problem, these fluctuations scale as $L^{-1/\nu}$.  We

273: find that $1/\nu=0.65$, leading to $\nu=1.54$.  We conjecture that the

274: exact value is $\nu=3/2$.  Hence, the fracture process in this system

275: is in a universality class which is {\it different\/} from standard

276: percolation where $\nu =4/3$.

277: 

278: When $D$ is no longer constant, but is given by the wedge shape (DCB

279: load), the form of interactions between the springs does not change.

280: Hence, the critical properties of the model when run in the

281: ``horizontal" mode are still present under DCB loading.  The wedge

282: shape of $D$ leads to a gradient in the loading of the springs, going

283: from very high loading where the damage is large to very low loading

284: well into the still intact part of the interface.  In Fig.~\ref{fig4},

285: we show the damage profile $p(y)$, {\it i.e.} the density of broken

286: springs averaged in the $x$ direction, for systems of different sizes.

287: The profile is clearly linear.  Somewhere along this damage profile,

288: there is a line in the $x$-direction at $y=y_c$ where the damage

289: density is critical, $p(y_c)=p_c$.  In the vicinity of this line,

290: there is a critical region which is characterized by being on the edge

291: between stability and instability and corresponds to the crack

292: front. Following the arguments of Sapoval {\it et al.\/} for

293: percolation in a gradient \cite{srg85}, if $p(y)$ follows a linear

294: law,

295: \begin{equation}

296: \label{linearp}

297: p(y)=1-\frac{y-y_0}{l_y}\;,

298: \end{equation}

299: on the interval $y_0 \le y \le y_0+l_y$, where $l_y$ is the length scale

300: characterizing the width of the damage zone, then $y_c=y_0+l_y(1-p_c)$

301: is the position of the fracture front (when ignoring small corrections for

302: finite-size systems).  The critical region has a width $\xi=|y_w-y_c|$

303: which is related to the damage $p(y)$ as $\xi \sim |p(y_w)-p_c|^{-\nu}$.

304: Eliminating $|y_c-y_w|$ between these two expressions for $\xi$ gives

305: \begin{equation}

306: \xi\sim l_y^{\frac{\nu}{1+\nu}}\;.

307: \end{equation}

308: 

309: %--------------------------------------------------------------------

310: \begin{figure}

311: \includegraphics[width=6.5cm,angle=270]{rip-fig4.eps}

312: \caption{Damage profile $p(y/L)$ in reduced variable $y/L$ for three

313: different system sizes $L$.}

314: \label{fig4}

315: \end{figure}

316: %--------------------------------------------------------------------

317: 

318: Since the gradient in $D$ is inversely proportional to $L$ and the

319: average strength of the springs does not change with $L$, we have that

320: \begin{equation}

321: \label{lyL}

322: l_y \propto L\;.

323: \end{equation}

324: In Fig.\ \ref{fig4}, the reduced variable $y/L$ was used resulting in

325: data collapse for different system sizes, thus validating Eq.\

326: (\ref{lyL}).  Furthermore, the width of the fracture front, $w$ is

327: proportional to the width of the critical region $\xi\propto w$.  Hence, we

328: find

329: \begin{equation}

330: \label{finally}

331: w\sim L^{\frac{\nu}{1+\nu}} = L^{3/5}\;.

332: \end{equation}

333: where we have used our estimate $\nu=3/2$.  This result is in

334: excellent agreement with our numerical simulations and with the

335: experimental results. 

336: 

337: A similar idea has been proposed for the outplane roughness of

338: fracture surfaces where the gradient and the $\nu$ exponent are

339: different and leads to $\zeta=4/5$ in excellent agreement with the

340: experimental value 0.80 \cite{hs02}.  We finally note that Zapperi

341: {\it et al.\/} \cite{zhr00} have pointed out the connection between

342: gradient percolation and the interface fracture problem in the limit

343: of infinitely stiff plates.  We have shown in this letter that this

344: mechanism stays intact also when the plates do respond elastically.

345: However, the universality class of the corresponding correlated

346: percolation problem is different from ordinary percolation, leading to

347: the observed roughness exponent $\zeta=3/5$ rather than 4/7 for

348: ordinary percolation in a gradient.

349: 

350: This work was partially funded by the CNRS PICS contract $\#753$ 

351: and the Norwegian Research Council, NFR. J.S.\ and A.H.\  thank 

352: Fernando A.\ Oliveira

353: and the ICCMP for financial support during our stay in Bras{\'\i}lia.

354: Discussions with K.\ J.\ M{\aa}l{\o}y are gratefully acknowledged.

355: % --------------------------------------------------------------------

356: \begin{thebibliography}{10}

357: 

358: \bibitem{srvm95} J.\ Schmittbuhl, S.\ Roux, J.\ P.\ Vilotte and K.\ J.\ 

359: M{\aa}l{\o}y, Phys.\ Rev.\ Lett.\ {\bf 74}, 1787 (1995).

360: 

361: \bibitem{sm97} J.\ Schmittbuhl and K.\ J.\ M{\aa}l{o}y, Phys.\ Rev.\

362: Lett.\ {\bf 78}, 3888 (1997).

363: 

364: \bibitem{mpp84} B.\ B.\ Mandelbrot, D.\ E.\ Passoja, and A.\ J.\ Paullay,

365: Nature {\bf 308}, 721 (1984).

366: 

367: \bibitem{bs85} S.\ R.\ Brown and C.\ H.\ Scholz, J.\ Geophys.\ Res.\

368: {\bf 90}, 12575 (1985).

369: 

370: \bibitem{blp90-mhhr92-sgr93-cw93-sss95} E.\ Bouchaud, G.\ Lapasset,

371: and J.\ Plan{\'e}s, Europhys.\ Lett.\ {\bf 13}, 73 (1990); K.\ J.\

372: M{\aa}l{\o}y, A.\ Hansen, E.\ L.\ Hinrichsen, and S.\ Roux, Phys.\

373: Rev.\ Lett.\ {\bf 68}, 213 (1992); J.\ Schmittbuhl, S.\ Gentier, and

374: S.\ Roux, Geophys.\ Res.\ Lett.\ {\bf 20}, 639 (1993); B.\ L.\ Cox and

375: J.\ S.\ Y.\ Wang, Fractals, {\bf 1}, 87 (1993); J.\ Schmittbuhl, F.\

376: Schmitt, and C.\ H.\ Scholz, J.\ Geophys.\ Res.\ {\bf 100}, 5953

377: (1995).

378: 

379: \bibitem{dhbc96-dnbc97} P.\ Daugier, S.\ Henaux, E.\ Bouchaud, F.\

380: Creuzet, Phys.\ Rev.\ E {\bf 53}, 5637 (1996); P.\ Daugier, B.\

381: Nghiem, E.\ Bouchaud and F.\ Creuzet, Phys.\ Rev.\ Lett.\ {\bf 78},

382: 1062 (1997).

383: 

384: \bibitem{dsm99} A.\ Delaplace, J.\ Schmittbuhl and K.\ J.\ M{\aa}l{\o}y,

385: Phys.\ Rev.\ E {\bf 60}, 1337 (1999).

386: 

387: \bibitem{ms01} K.\ J.\ M{\aa}l{\o}y and J.\ Schmittbuhl, Phys.\

388: Rev.\ Lett.\ {\bf 87}, 105502 (2001).

389: 

390: \bibitem{prg02} A.\ Prevost, E.\ Rolley and C.\ Guthmann, Phys.\ Rev.\ B

391: {\bf 65}, 064517 (2002).

392: 

393: \bibitem{gr89} H.\ Gao and J.\ R.\ Rice, ASME J.\ Appl.\ Mech.\ {\bf56},

394: 828 (1989).

395: 

396: \bibitem{ek94} D.\ Ertas and M.\ Kardar, Phys.\ Rev.\ E, {\bf 49},

397: 2532 (1994).

398: 

399: \bibitem{cdw01} P. Chauve, P. Le Doussal and K.J. Wiese

400: Phys. Rev. Lett. {\bf 86}, 1785 (2001).

401: 

402: \bibitem{rk02} A.\ Rosso and W.\ Krauth, Phys.\ Rev.\ E, {\bf 65}, 025101

403: (2002).

404: 

405: \bibitem{pr94-rf97-rf98-mr98-mr00} G.\ Perrin and J.\ R.\ Rice, 

406: J. Mech.\ Phys.\ Solids, {\bf 42}, 1047 (1994); S.\ Ramanathan and D.\ S.\ 

407: Fisher, Phys.\ Rev.\ Lett.\ {\bf 79}, 877 (1997); S.\ Ramanathan and D.\ S.\ 

408: Fisher, Phys.\ Rev.\ B, {\bf 58}, 6026 (1998); J.\ W.\ Morrissey and J.\ R.\ 

409: Rice, J. Mech.\ Phys.\ Solids, {\bf 46}, 467 (1998); J.\ W.\ Morrissey and 

410: J.\ R.\ Rice, J. Mech.\ Phys.\ Solids, {\bf 48}, 1229 (2000).

411: 

412: \bibitem{ll58} L.\ Landau and E.\ M.\ Lifshitz, {\sl Theory of Elasticity\/}

413: (Clarendon Press, Oxford, 1958).

414: 

415: \bibitem{j85} K.\ L.\ Johnson, {\sl Contact Mechanics\/} (Cambridge Univ.\

416: Press, Cambridge, 1985).

417: 

418: \bibitem{bhs02} G.\ G.\ Batrouni, A.\ Hansen and J.\ Schmittbuhl, Phys.\

419: Rev.\ E, {\bf 65}, 036126 (2002).

420: 

421: \bibitem{fv85} F.\ Family and T.\ Vicsek, J.\ Phys.\ A, {\bf 18}, L75 (1985).

422: 

423: \bibitem{hs02} A.\ Hansen and J.\ Schmittbuhl, Cond-mat/0207360 (2002).

424: 

425: \bibitem{sa92} D.\ Stauffer and A.\ Aharony, {\it Introduction to

426: percolation theory\/} (Francis and Taylor, London, 1992).

427: 

428: \bibitem{srg85} B.\ Sapoval, M.\ Rosso and J.F.\ Gouyet, J.\ Phys.\ Lett.\

429: (France) {\bf 46}, L149 (1985).

430: 

431: \bibitem{zhr00} S.\ Zapperi, H.\ J.\ Herrmann and S.\ Roux, 

432: Eur.\ J.\ Phys.\ B {\bf 17}, 131 (2000).

433: 

434: 

435: \end{thebibliography}

436: % --------------------------------------------------------------------

437: \end{document}

438: