1: \documentclass[showpacs,preprintnumbers,showkeys,preprint,prb,floatfix]{revtex4}
2: %\documentclass[showpacs,preprintnumbers,showkeys,preprint,prb,floatfix,twocolumn]{revtex4}
3: %\documentclass[showpacs,preprintnumbers,showkeys,preprint,prb,twocolumn]{revtex4}
4: %\documentstyle[graphicx,epsf,twocolumn,prl,aps]{revtex}
5:
6: \usepackage{epsfig}
7: \usepackage{amsmath}
8: \usepackage{amssymb}
9: \usepackage{bm}
10:
11: \begin{document}
12:
13: \title{Crack Growth Laws from Symmetry}
14:
15: \author{James P. Sethna}
16: \affiliation{Laboratory of Atomic and Solid State Physics, Clark Hall, Cornell
17: University, Ithaca, NY 14853-2501}
18: \email[Corresponding author, Electronic address:]{sethna@lassp.cornell.edu}
19: \homepage{http://www.lassp.cornell.edu/sethna/sethna.html}
20: \date{\today}
21:
22: \begin{abstract}
23: Reviewing work done in collaboration with Jennifer Hodgdon, we derive
24: the most general crack growth law allowed by symmetry for mixed-mode
25: three-dimensional fracture. We do so using the system developed in
26: condensed matter physics to derive new laws that emerge on long length
27: and time scales. In our derivation, we provide a symmetry interpretation
28: for the three modes of fracture, we rederive the law giving the crack
29: growth direction in two dimensional fracture, and we derive a growth law
30: appropriate for three dimensional simulations. We briefly discuss
31: related work subsequent to ours, incorporating disorder, dynamics, and
32: other internal variables.
33: \end{abstract}
34:
35: \pacs{46.35.+z, 62.20.Fe, 83.60.La}
36: \keywords{fracture, crack, growth, evolution, mixed-mode, linear stability
37: analysis}
38: % fracture, crack, nucleation, growth, thermal, evolution, factory roof
39:
40: \maketitle
41:
42: \def\ddx#1#2{ {\frac{\partial #2}{\partial #1}} }
43: \def\kI{K_{\rm I}}
44: \def\kII{K_{\rm II}}
45: \def\kIII{K_{\rm III}}
46: \def\nhat{\hat n}
47: \def\bhat{\hat b}
48: \def\that{\hat t}
49: \def\Xvec{\vec X}
50: \def\Rb{R_b}
51: \def\Rt{R_t}
52: \def\dadotb#1#2{ \frac{\partial #1}{ \partial s} \cdot #2 }
53:
54: In this paper, I review work with Jennifer Hodgdon\cite{Hodgdon} using
55: symmetry principles to derive the growth law for mixed-mode, curved cracks
56: for isotropic materials in three dimensions. In the process, I'll attempt
57: to illustrate the general approach that condensed-matter physicists have
58: developed to derive new phenomenological laws for macroscale physical phenomena.
59:
60: The physical laws governing the behavior of many physical systems can
61: be derived using symmetry. Basically, there are five steps, which are
62: not necessarily applied sequentially:
63:
64: \noindent{\it I. Pick an order parameter field.} The order parameter
65: field is a state variable, which at a point $\bf x$ summarizes the
66: current state of the material in the local neighborhood of $\bf x$.
67:
68: \noindent{\it II. Use the symmetries of the problem.}
69:
70: \noindent{\it III. Imagine writing the most general possible law.}
71:
72: \noindent{\it IV. Simplify the theory using small parameters, power
73: counting, etc.} Usually one will need to expand in gradients of the
74: order parameter: the theory then describes only behavior at long
75: wavelengths. In dynamical evolution laws, one will naturally also expand
76: in time derivatives, specializing to low frequencies. One often expands
77: in powers of the order parameter.
78:
79: \noindent{\it V. Solve the theory.} In many cases, this is straightforward,
80: analytically or computationally. In other cases temperature, dirt, or noise
81: introduce fluctuations that remain important on all length scales (such
82: as near continuous phase transitions), and renormalization-group methods
83: may be needed.
84:
85: Landau introduced this system to derive the forms of free energies and
86: near-equilibrium dynamics of materials with broken symmetries. His
87: methods have been applied extensively to exotic liquid crystals,
88: superconductors and superfluids, magnetic systems, ferroelectrics,
89: incommensurate systems, martensites, and in spatially extended dynamical
90: systems. In many of these systems, the phenomenological Landau theory
91: was effectively used long before a microscopic mechanism or model was
92: developed. More recently, essentially the same procedure has been used to
93: derive evolution equations in systems which are far from equilibrium.
94:
95: How shall we proceed to apply these ideas to cracks?
96:
97: \bigskip\bigskip
98: \noindent {\bf I. Pick an order parameter field.}
99: \medskip
100:
101: \noindent
102: The region of interest in fracture is the immediate vicinity of the crack front:
103: in general, a curved line in space (figure \ref{CrackTemplate}). Let us
104: parameterize this curve by arc
105: length $s$, so the curve is $\Xvec(s)$. The geometry of the crack edge
106: demands not only the coordinates of this line, but also the orientation of
107: the crack surface as it approaches the line. We can define a unit tangent
108: vector $\that = d\Xvec/ds$ to the crack edge, the direction of crack
109: growth $\nhat$ (perpendicular to $\that$), and the normal to the crack plane
110: $\bhat=\that \times \nhat$. The functions $\Xvec(s)$ and $\nhat(s)$ are enough
111: to determine the geometry of the crack; we use them as the order parameter
112: for the problem.
113:
114: \begin{figure}[thb]
115: \epsfig{file=Fig/CrackTemplate.ps, width=3truein}
116: \caption{\label{CrackTemplate}
117: {\bf Order parameter for crack growth.}
118: The crack is parameterized by arclength $s$; the crack edge is a curve
119: $\Xvec(s)$ growing in direction $\nhat(s)$. The tangent to the crack
120: edge $\that=d\Xvec/ds$ and the normal to the crack plane
121: $\bhat=\that\times\nhat$ are derivable from $\Xvec$ and $\nhat$.}
122: \end{figure}
123:
124: Of course, the crack growth rates will depend strongly on the applied
125: stresses at the crack front. These depend on the loading far away, as well
126: as the shape of the crack surface behind the growing crack front. In special
127: circumstances (nearly straight crack fronts in
128: mode~I\cite{Cotterell,RamanathanFisher1}) we can write closed-form
129: expressions for these stresses
130: in terms of integrals over $\Xvec(s)$ and $\nhat(s)$; more generally these
131: stresses can be calculated from elastic theory, {\it e.g.} using finite
132: element analysis \cite{IngraffeaWeb,Wawrzynek}. We thus assume that
133: the stresses as a function of $s$ are given.
134:
135: \bigskip\bigskip
136: \noindent {\bf II. Use the symmetries of the problem.}
137: \medskip
138:
139: \noindent
140: We assume that our material is isotropic, and will consider only
141: quasi-static fracture. On length scales short compared to the curvatures
142: of the crack and compared to the gradients of the
143: stresses at the crack front, we have two independent discrete symmetries
144: for the local crack geometry: reflection $\Rb$ in the plane of the crack
145: (taking $\bhat$ to $-\bhat$) and reflection $\Rt$ in the plane
146: perpendicular to the crack front $\that$. There is a third symmetry,
147: a 180$^\circ$ rotation about the axis $\nhat$, which is the product of
148: the two other symmetries.
149:
150: We can use these symmetries first to simplify the characterization of
151: the quasistatic stresses at the crack tip. As for all linear
152: problems with symmetries, we may decompose a general solution of the
153: linear elastic problem into solutions whose displacement (and strain)
154: fields are odd or even under the two symmetries $R_b$ and $R_t$.
155: For an uncracked material whose strain is constant along $\that$,
156: a general elastic solution can be decomposed into multipoles: for
157: each power $n$ of the distance $r$ to the $\that$ axis, there are six
158: elastic solutions whose strains vary as $r^n$. (For example, there
159: are six solutions with constant strain $n=0$, corresponding to the six
160: independent coefficients of the elastic strain tensor.)
161: There are three solutions which are even in $R_b$ and $R_t$, and one each
162: of the other three possibilities. For the medium with a crack, three of these
163: elastic solutions are not allowed by the condition that there be no
164: traction at the crack surface. Instead, there are three new classes of
165: solutions with strain fields depending on half-integer powers of $r$,
166: with non-zero displacements across the crack
167: surface\cite{Ruina,HodgdonThesis}.
168:
169: What are all these elastic solutions? The solutions which involve
170: strains varying as $r^n$ for $n>0$ are large at the boundaries of the
171: sample but vanish quickly at the crack front. They are important for
172: solving for elastic deformations in complex geometries, but are
173: irrelevant for crack growth except when the sample size is comparable to
174: the size of the nonlinear zone at the crack front. The solutions which
175: have $n\le -1$ have diverging total energy at the crack tip. These solutions
176: are important for matching the boundaries of the nonlinear zone to the
177: elastic theory, but will die away at distances comparable to the size
178: of the nonlinear zone. Their magnitudes thus are determined by the
179: local geometry and stress at the crack front, and hence can also be ignored.
180: We are left with three ordinary elastic strains $n=0$ and three
181: strain fields with crack opening displacements $n=-1/2$. These latter fields are
182: the familiar three modes of fracture (table \ref{tab:FractureModes}).
183:
184: \begin{table}
185: %\begin{ruledtabular}
186: \begin{tabular}{|c|c|c|c|}
187: \hline
188: &$\Rb$ & $\Rt$ & \\
189: \hline
190: \vbox to 1.2truein{\vfil\hbox{Mode I, $K_{I}$}\vfil}
191: &\vbox to 1.2truein{\vfil\hbox{Even}\vfil}
192: &\vbox to 1.2truein{\vfil\hbox{Even}\vfil}
193: & \epsfig{file=Fig/CrackModeI.ps, width=1.5truein}\\
194: \hline
195: \vbox to 0.6truein{\vfil\hbox{Mode II, $K_{II}$}\vfil}
196: &\vbox to 0.6truein{\vfil\hbox{Odd}\vfil}
197: &\vbox to 0.6truein{\vfil\hbox{Even}\vfil}
198: & \epsfig{file=Fig/CrackModeII.ps, width=1.5truein}\\
199: \hline
200: \vbox to 0.7truein{\vfil\hbox{Mode III, $K_{III}$}\vfil}
201: &\vbox to 0.7truein{\vfil\hbox{Odd}\vfil}
202: &\vbox to 0.7truein{\vfil\hbox{Odd}\vfil}
203: & \epsfig{file=Fig/CrackModeIII.ps, width=1.5truein}\\
204: \hline
205: &Even & Odd & No crack opening\\
206: \hline
207: \end{tabular}
208: \caption{\label{tab:FractureModes}Symmetries and Modes of Fracture.}
209: %\end{ruledtabular}
210: \end{table}
211:
212: In our earlier work\cite{Hodgdon}, we ignored the $n=0$ elastic
213: stresses near the crack tip. It's likely that their effects are relatively
214: small, since they aren't intensified near the crack tip, and we will ignore
215: them in this review as well. It would be straightforward to incorporate
216: the overall strains into the equations of motion, for a more complete theory.
217:
218: There is one more symmetry in our problem: an artificial symmetry which
219: is introduced by our theoretical description. It is not always
220: convenient to use arc-length $s$ to parameterize the crack. In
221: particular, it makes the growth laws nonlocal: even for a stationary
222: portion of the crack, if a far-away region of the crack with smaller $s$
223: changes in length the functions $\Xvec(s)$ and $\nhat(s)$ will shift
224: sideways. Of course, we can use any parameterization that's convenient.
225: Changing parameterizations in a formal sense is analogous to changing
226: the choice of gauge in electromagnetism \cite{SteveLanger}. For roughly
227: straight cracks parallel to an axis $z$, we would likely use $z$-gauge
228: with $\Xvec(z)$ and $\nhat(z)$; for loops we would likely use
229: $\theta$-gauge. Changing the equation of motion from one gauge to
230: another is called a {\it gauge transformation}. There is thus a {\it
231: gauge symmetry}: the physical evolution of the crack must be unchanged,
232: or {\it gauge invariant}, when we change from one parameterization to another.
233:
234: \begin{figure}[thb]
235: \epsfig{file=Fig/CrackGauge.ps, width=3truein}
236: \caption{\label{CrackGauge}
237: One must choose a parameterization of the crack edge, $\lambda$ (choosing
238: a gauge). The form of the equations of motion will change for different
239: choices for the parameterization, but physical quantities like the location
240: of the crack front must be unchanged (gauge invariant).}
241: \end{figure}
242:
243: \bigskip\bigskip
244: \noindent {\bf III. Write the most general possible law.}
245: \medskip
246:
247: \noindent
248: In two-dimensional fracture, we can immediately write down the most
249: general possible form for the crack velocity and turning rate. The
250: position $\Xvec$ of the crack moves forward along $\nhat$ with velocity
251: $v$ which depends on the stress intensity factors. The velocity does
252: not change sign under $\Rb$ or $\Rt$, so $v$ cannot have terms odd in
253: $K_{II}$ or $K_{III}$, giving us the first law
254: \begin{equation}
255: \partial {\bf X}/\partial t = v(K_I,K_{II}^2,K_{III}^2) {\bf \hat n}.
256: \label{eq:2Dv}
257: \end{equation}
258: Similarly, the rate of rotation of the crack surface changes sign under
259: reflection in the plane of the crack $\Rb$, and doesn't change sign under
260: $\Rt$ so it must have one factor of $K_{II}$ times an arbitrary function
261: of $K_{I}$, $K_{II}^2$, and $K_{III}^2$:
262: \begin{equation}
263: \partial {\bf \hat n}/\partial t = -f(K_I,K_{II}^2,K_{III}^2)
264: K_{II} {\bf \hat b}.
265: \label{eq:2Df}
266: \end{equation}
267:
268: In three dimensions, we must incorporate possible dependences of the
269: equations of motion on the local curvatures of the crack front and the
270: gradients of the stress intensity factors. To linear order in gradients,
271: Hodgdon found
272: \begin{eqnarray}
273: \partial x / \partial t =& v \nhat + w \that \cr
274: \partial \nhat / \partial t =& - \left[ \ddx{s}{v} +
275: w \ddx{s}{\that} \cdot \nhat \right] \that +
276: \left[ - f \kII + g_{\rm I} \kIII \ddx{s}{\kI} +
277: g_{\rm II}\kII \kIII \ddx{s}{\kII} + g_{\rm III}\ddx{s}{\kIII} + \right. \cr
278: &\qquad \left. h_{tb} \dadotb{\that}{\bhat} +
279: h_{nt} \kII \dadotb{\nhat}{\that} +
280: (h_{nb}\kII\kIII + w) \dadotb{\nhat}{\bhat} \right] \bhat ,\cr
281: \label{eq:3D}
282: \end{eqnarray}
283: where $f$, $g_{\alpha}$, and $h_{ij}$ are functions of $\kI$, $\kII^2$, and
284: $\kIII^2$, and the velocity $v$ can be a function of these and a number of
285: gradient terms\cite{Hodgdon}.
286:
287: \bigskip\bigskip
288: \noindent {\bf IV. Simplify the theory using small parameters,
289: power counting, etc.}
290: \medskip
291:
292: \noindent
293: By focusing on quasi-static fracture, and by assuming that the
294: equations of motion involve only first derivatives in time, we're already making
295: use of the small ratio of the velocity of the crack growth to the
296: natural material velocities (velocity of sound, surface relaxation, etc.)
297: By confining our attention to the external strains that go as $r^{-1/2}$
298: and $r^0$ we made the assumption
299: that the nonlinear zone of the crack is small compared to the system size
300: (so-called K-dominant fracture, since the three stress intensity factors
301: dominate). In three dimensions, by keeping terms to linear order in gradients,
302: we have again assumed the nonlinear zone is small compared to curvatures
303: and changes in stresses.
304:
305: We can use the small size of the nonlinear zone a third time to establish the
306: relative sizes of different terms in our equations of motion. In our
307: two dimensional equations of motion (\ref{eq:2Dv}) and (\ref{eq:2Df}),
308: the functions $v$ and $f$ have different units: $v/K f$ has units of
309: length, where $K$ is any one of the stress intensities. It is natural to
310: assume that this length will be set by the size of the nonlinear zone,
311: and in any case it is part of our approach to assume that this length is small.
312:
313: \bigskip\bigskip
314: \noindent {\bf V. Solve the theory.}
315: \medskip
316:
317: \noindent
318: It is well known that cracks loaded with a mixture of mode~I and more~II
319: will turn abruptly. In our formulation, the function $f$ governs the
320: rate at which the crack turns. Hodgdon used results of Cotterell and
321: Rice \cite{Cotterell} to show that if the angle of the crack differs
322: from the angle that makes $K_{II}=0$ by a small amount $\Delta \theta$,
323: then $K_{II} = K_I \Delta \theta/2$. Combined with the two growth laws
324: above, we find that $\Delta \theta \sim \exp(-f K_I x/2 v)$, the crack
325: turns to pure mode~I exponentially with a material-dependent decay
326: length of $2 v / f K_I$: precisely twice the characteristic length scale
327: we assumed was small! Thus we derive the {\it principle
328: of local symmetry}\cite{GS}, which says that a mode~II fracture will
329: turn abruptly (that is, on a length comparable to the nonlinear zone
330: size) until it is pure mode~I.
331:
332: There were two other rules in the literature for picking the crack growth
333: direction: one maximizing the energy release\cite{Wu}, and one moved in
334: the direction of minimum strain energy density\cite{Sih}. Cracks grown
335: by these different rules (move forward by a small step size $\Delta x$,
336: recalculate the stress intensity factors, repeat) all gave rather similar
337: predictions for the shapes. Our analysis did not make any assumptions about
338: microscopic mechanisms, so we can apply it to a hypothetical material which
339: behaved according to these other crack growth rules.
340: We conclude that these other rules will end up forcing the crack to turn
341: until $K_{II}=0$, yielding in all cases the principle of local symmetry!
342: Instead of a nonlinear zone size, the turning radius is given by the
343: step size of the algorithm $\Delta x$. Our analysis predicts that {\it all
344: three growth rules are equivalent} in the limit of small stepsize/turning
345: radius.
346:
347: \bigskip
348: \noindent {\bf Where to go from here?}
349: \medskip
350:
351: \noindent
352: Hodgdon's analysis is now a decade old. What came next?
353:
354: First, Hodgdon used the three-dimensional growth law (equation \ref{eq:3D})
355: to investigate the stability of crack growth under mixtures of mode~III
356: and mode~I. All of the new 3D materials functions $g_X$, $h_{yz}$, and $w$ in
357: equation (\ref{eq:3D}) multiply gradients, so they aren't unusually large
358: like $f$.
359: Hodgdon found that mixed-mode fracture can be stable or unstable depending
360: on details: in particular, for $g_I>0$ steady-state mode~III cracks are
361: unstable to small perturbations. Mode~III fracture is unstable experimentally
362: to the formation of a ``factory roof'' morphology, with ramps followed by
363: cliffs. With the computers at that time, we couldn't get into the nonlinear
364: regime needed to test whether our theory leads to the same jagged solutions
365: as seen experimentally: Hodgdon's linear stability analysis remains
366: unpublished\cite{HodgdonThesis}.
367:
368: Second, real materials are dirty: not only impurities and inclusions,
369: but also the random fracture strengths introduced by polycrystallinity
370: suggest the incorporation of randomness in the evolution equations.
371: Introducing randomness is also motivated by the experimental observation
372: of Bouchaud\cite{Bouchaud}, who finds roughness on all scales, with
373: power-law height-height correlation functions. Ramanathan and
374: Fisher\cite{RamanathanFisher1} (see also \cite{Nattermann}) studied the
375: effects of incorporating disorder into a dynamics similar to that
376: described here. They discovered that it does indeed predict a fracture
377: surface which is rough on all length scales, but with much weaker
378: randomness than that observed experimentally (logarithmic rather than
379: power law). It seems likely that the experimental roughness, at least in
380: ductile and intergranular fracture, reflects void growth and the
381: presence of grains and inclusions\cite{Anderson}.
382:
383: Thirdly, many effects of dynamics on crack growth have been ignored in
384: our discussion. The theory as described here is appropriate, perhaps, for
385: fatigue crack growth where mass and inertia are not important. New
386: phenomena arise as the crack speeds up (such as the mirror-mist-hackle
387: morphology transitions\cite{Fineberg}). Incorporating the inertial
388: dynamics of the growing front does lead to interesting traveling wave
389: solutions\cite{FrontWaves}.
390:
391: Finally, there is the likely possiblity that there are other important
392: degrees of freedom that are important, and need to be incorporated into the
393: order parameter describing the current state of the crack tip.
394: There are recent indications that the curvature of the crack tip\cite{Argonne}
395: and the effects of surface tension and diffusion\cite{LobkovskyKarma,Brener}
396: may be important.
397:
398:
399: \begin{acknowledgments}
400:
401: This work was supported by NSF KDI-9873214 and NSF ACI-0085969\/.
402: We thank Nick Bailey for helpful discussions.
403:
404: \end{acknowledgments}
405:
406: \begin{thebibliography}{99}
407:
408: \bibitem{Hodgdon} Jennifer A. Hodgdon and James P. Sethna, {\sl Phys. Rev. B}
409: {\bf 47}, 4831 (1993), James P. Sethna, invited talk at the 10th International
410: Conference on Fracture, December 2001.
411:
412: \bibitem{Cotterell} B. Cotterell and J. R. Rice, {\sl Int. J. Frac.}
413: {\bf 16}, 155 (1980).
414:
415: \bibitem{RamanathanFisher1} S. Ramanathan, D. Ertas, and D. S. Fisher,
416: {\sl Phys. Rev. Lett.} {\bf 79}, 873 (1997).
417:
418: \bibitem{IngraffeaWeb} Cornell Fracture Group, http://www.cfg.cornell.edu/.
419:
420: \bibitem{Wawrzynek} P. A. Wawrzynek and A. R. Ingraffea, {\it Discrete
421: Modeling of Crack Propagation: Theoretical Aspects and Implementation Issues
422: in Two and Three Dimensions}, Ph.D. Thesis, Cornell University, 1991.
423:
424: \bibitem{Ruina} C. Y. Hui and Andy Ruina, {\it Int. J. Frac.} {\bf 72}
425: 97 (1995).
426:
427: \bibitem{HodgdonThesis} J. A. Hodgdon, {\it Three-Dimensional Fracture:
428: Symmetry and Stability}, Ph. D. Thesis, Cornell University, 1993.
429:
430: \bibitem{SteveLanger} S. A. Langer, R. E. Goldstein, and D. P. Jackson,
431: {\sl Phys. Rev. A} {\bf 46}, 4894 (1992).
432:
433: \bibitem{GS} R. V. Gol'dstein and R. L. Salganik {\sl Int.
434: J. Frac.} {\bf 10}, 507 (1974).
435:
436: \bibitem{Wu} C. H. Wu, {\sl J. Appl. Mech.} {\bf 45}, 553 (1978).
437:
438: \bibitem{Sih} G. C. Sih, p. xv in {\it Mechanics of Fracture 2: Three
439: Dimensional Crack Problems} (eds. M. K. Kassir and G. C. Sih), Noordhoof
440: International, Netherlands, 1975.
441:
442: \bibitem{Nattermann} P. F. Arndt and T. Nattermann, ``A New Criterion for
443: Crack Formation in Disordered Materials'', cond-mat/0012113, accepted
444: for publication in {\sl Phys. Rev. B.}
445:
446: \bibitem{Bouchaud} E. Bouchaud, {\sl J. Phys. Condens. Matter} {\bf 9},
447: 4319 (1993) and references therein.
448:
449: \bibitem{Anderson} T. L. Anderson, {\it Fracture Mechanics, Fundamentals
450: and Applications}, CRC Press, Boca Raton, 1991, chapter 5.
451:
452: \bibitem{Fineberg} J. Fineberg, S. P. Gross, M. Marder, and H. L. Swinney,
453: {\sl Phys. Rev. Lett.} {\bf 67}, 457 (1991).
454:
455: \bibitem{FrontWaves} S. Ramanathan and D. S. Fisher, {\sl Phys. Rev. Lett.}
456: {\bf 79}, 877 (1997); J. R. Rice, Opening Lecture at the 20th International
457: Congress of Theoretical and Applied Mechanics, Chicago, Aug. 27, 2000.
458:
459: \bibitem{Argonne} I. S. Aranson, V. A. Kalatsky, V. M. Vinokur, {\sl Phys.
460: Rev. Lett.} {\bf 85}, 118 (2000).
461:
462: \bibitem{LobkovskyKarma} A. Lobkovsky and A. Karma, ``Tip splitting
463: instability in a phase field model of mode III dynamic fracture'', APS
464: meeting presentation Q27.002, March 2002; A. Lobkovsky, condensed-matter
465: seminar, Cornell University, spring 2002.
466:
467: \bibitem{Brener} Efim A. Brener, Robert Spatschek, ``Fast crack propagation
468: by surface diffusion'', cond-mat/0204056.
469:
470: \end{thebibliography}
471:
472: \end{document}
473:
474: