1: \documentclass[aps,twocolumn,floats,floatfix,showpacs,superscriptaddress]{revtex4}
2: \usepackage{epsfig}
3: \begin{document}
4:
5: \title{Dynamic Fracture Model for Acoustic Emission}
6: \author{Manuela Minozzi}
7: \affiliation{INFM UdR Roma 1 - Dipartimento di Fisica,
8: Universit\`a "La Sapienza", P.le A. Moro 2,
9: 00185 - Roma, Italy}
10: \affiliation{INFM UdR Roma 3, Dipartimento di Fisica,
11: Universit\`a di Roma 3, via della vasca navale 84, 00146
12: Roma, Italy.}
13: \affiliation{CNR Istituto di Acustica ``O. M. Corbino'', via
14: del fosso del Cavaliere 100, 00133 Roma, Italy}
15: \author{Guido Caldarelli}
16: \affiliation{INFM UdR Roma 1 - Dipartimento di Fisica,
17: Universit\`a "La Sapienza", P.le A. Moro 2,
18: 00185 - Roma, Italy}
19: \author{Luciano Pietronero}
20: \affiliation{INFM UdR Roma 1 - Dipartimento di Fisica,
21: Universit\`a "La Sapienza", P.le A. Moro 2,
22: 00185 - Roma, Italy}
23: \affiliation{CNR Istituto di Acustica ``O. M. Corbino'', via
24: del fosso del Cavaliere 100, 00133 Roma, Italy}
25: \author{Stefano Zapperi}
26: \affiliation{INFM UdR Roma 1 - Dipartimento di Fisica,
27: Universit\`a "La Sapienza", P.le A. Moro 2,
28: 00185 - Roma, Italy}
29:
30:
31: \date{\today}
32:
33: \begin{abstract}
34: We study the acoustic emission produced by micro-cracks using a two-dimensional
35: disordered lattice model of dynamic fracture, which allows to relate the acoustic response to the internal damage of
36: the sample. We find that the distributions of acoustic
37: energy bursts decays as a power law in agreement with
38: experimental observations. The scaling exponents measured in
39: the present dynamic model can related to those obtained in the
40: quasi-static random fuse model.
41: \end{abstract}
42: \pacs{62.65.+k, 46.50.+a, 63.70.+h}
43: \maketitle
44:
45: Crackling noise \cite{SET-01} is widely observed in systems as different as
46: superconductors \cite{flux}, magnets \cite{DUR-01}
47: or plastically deforming crystals \cite{MIG-01}.
48: A typical example is the acoustic emission (AE) recorded
49: in a stressed material before failure. The noise is a consequence
50: of micro-cracks forming and propagating in the material and should
51: thus provide an indirect measure of the damage accumulated in
52: the system. For this reason, AE is often used as a non-destructive
53: tool in material testing and evaluation.
54: Beside these practical applications, understanding the statistical
55: properties of crackling noise has become a challenging theoretical
56: problem. The distribution of crackle amplitudes follows a power law,
57: suggesting an interpretation in terms of critical phenomena and
58: scaling theories. This behavior has been observed in several
59: materials such as wood \cite{ciliberto}, cellular glass
60: \cite{strauven}, concrete \cite{ae} and paper \cite{paper}
61: to name just a few.
62:
63: The statistical properties of fracture in disordered media
64: are captured qualitatively by lattice models, describing the medium as
65: a discrete set of elastic bonds with randomly
66: distributed failure thresholds \cite{hr,fuse,duxbury}.
67: After each crack the stress is redistributed in the lattice
68: in a quasi-static approximation:
69: i.e. the crack velocity is much slower than stress relaxation.
70: Thus acoustic waves are not taken into account and the activity is
71: monitored by the damage evolution or by the dissipated elastic energy.
72: Numerical simulations indicate that micro-cracks propagate in avalanches
73: giving rise to an heterogeneous response. The avalanche distribution
74: is typically described by power law distributions and
75: the results are usually interpreted in the framework of phase
76: transitions \cite{hansen,zrvs,zvs,gcalda,alava}.
77: Despite the fact that critical phenomena are normally
78: associated with a certain degree of universality (i.e. the scaling
79: exponents should not depend on micro-structural details), there
80: has been so far no quantitative agreement between models and experiments.
81: A reason that could account for this discrepancy is the absence
82: of acoustic waves in most models. It is then not obvious how to relate
83: AE activity to internal avalanches.
84:
85: Dynamic lattice models have been widely used in the past to
86: analyze fracture processes \cite{dyn,marder,RAU,mar}, but although
87: acoustic waves are explicitly included, the AE signal is usually
88: not analyzed. Here we use a lattice model for dynamic fracture
89: in a disordered medium, to obtain a direct correspondence between
90: the recorded AE activity and the internal damage evolution.
91: We find that the cumulative AE amplitudes
92: are directly related --- by a power law --- to the cumulative damage.
93: Next, we measure the distribution of the AE burst energies and
94: find a power law with an exponent $\beta \simeq 1.7$ independent on
95: the loading rate. This exponent can be related to the exponent
96: describing failure avalanches in quasi-static models
97: \cite{hansen,zrvs,zvs,gcalda,alava}.
98:
99: We consider a scalar model of dynamic fractures where a
100: two-dimensional lattice is loaded in mode III \cite{marder}:
101: the lattice lies in the $(x,y)$ plane and deformation occurs
102: along the $z$ axis, so that the equations of elasticity
103: become scalar. The equation of the motion for the anti-planar
104: displacement $u$ of a site with coordinate $i,j$ is
105: \begin{equation}
106: \rho\ddot{u}_{i,j}= -K\sum_{(l,m)} (u_{i,j}-u_{l,m})-\Gamma\dot{u}_{i,j},
107: \label{eq:1}
108: \end{equation}
109: where the sum runs over the nearest neighbors $(l,m)$ of site $(i,j)$,
110: $K$ is the elastic constant, $\rho$ is the density
111: and dissipation is simulated by a viscous damping with a constant $\Gamma$.
112: In order to suppress some lattice effects, we use a 45 degree tilted
113: square lattice. A constant strain rate is imposed to the model,
114: by moving the boundary sites on two opposite boundaries at constant
115: velocity $V$ and $-V$, respectively. Periodic boundary conditions
116: are imposed in the other direction. Disorder is simulated assigning
117: randomly distributed failure threshold: a bond is removed (i.e. $K$
118: is set to zero) when $\Delta u > f_c$, and $f_c$ is uniformly distributed
119: in $[0,1]$. Notice that in the quasi-static limit
120: ($V\to 0$, $\rho\to 0$, $\Gamma\to 0$) the model reduces
121: to the random fuse model (RFM), where a lattice of fuses with
122: random threshold are subject to an increasing voltage \cite{fuse,duxbury}.
123: Due to the scalar nature of our model there is a direct mapping
124: between elastic and electric parameters \cite{hr}.
125:
126: The equation of motion (Eq.~\ref{eq:1}) is integrated numerically
127: using a fourth order Runge-Kutta method. We work with a lattice
128: of linear size $L=80$ and chose the units of space and time
129: so that $\rho=K=1$. Each time a bond is stretched
130: beyond its threshold the lattice constant is set to zero and
131: an elastic wave is emitted. Due to the anti-plane constraint
132: for the displacements, we only have transverse wave propagation
133: with sound speed $c=\sqrt{K/\rho}=1$ in our units. The damping
134: constant is chosen to be $\Gamma=0.1$ so that typical length traveled
135: by a wave is a little smaller than the lattice size. For smaller
136: values of $\Gamma$ ringing effects and reflected waves do not
137: allow to separate the single pulses and the lattice breaks at
138: once. On the other hand, excessive damping leads to very
139: small acoustic activity and the sample breaks suddenly at the edges.
140: Even if the damping constant is small reflected waves
141: can induce boundary failure, due to the rigidity of the
142: loaded edge. Thus we do not allow that bonds fail in two boundary layers
143: of length $l=5$ close to the loaded edges. This corresponds to
144: apply a load through a soft contact. The model is simulated
145: for a variety of loading velocities all much lower than the
146: sound speed $V \ll c$.
147:
148: Measuring the displacements of every lattice site and calculating
149: the forces for every time steps, we have obtained
150: the stress-strain curve for four different value of
151: the applied strain rate. In Fig.~\ref{fig:1} we show
152: that the stress is a linear function of the strain up
153: to the yield point, which precedes the total failure of the sample.
154: The applied strain rate has little effect on the linear part of the curve,
155: while it influences the curve after the yield point.
156:
157: \begin{figure}[t]
158: \centerline{\psfig{file=fig1.eps,width=8.5cm,clip=!}}
159: \caption{The stress-strain curve for different applied strain rates. }
160: \label{fig:1}
161: \end{figure}
162:
163: Monitoring the activity of some
164: particular lattice sites we have direct access to the AE
165: signal. These sites mimic the effect of transducers
166: coupled to the material in a typical AE experiment.
167: In a typical run, we record the displacements, velocities and accelerations
168: of four sites in the boundary layer and two sites in the interior.
169: Typically, AE distributions are recorded from a single site
170: and averaged over ten realizations of the disorder.
171: We have tested that the statistical properties of the signal
172: do not vary for different boundary sites, while there is
173: a clear difference between boundary and inner sites.
174: In the following, we concentrate on sites in the boundary layer,
175: in order to avoid excessive fluctuations due to failures
176: occurring on neighboring bonds in the inner region.
177:
178: An example of the typical signals recorded with our model are
179: reported in Fig.~\ref{fig:2}. A large acoustic activity is visible in the upper panel where we show the
180: local acceleration $a$ of a boundary site as a function of time.
181: We can also monitor the velocity signal which is
182: simply related to the acceleration and display the same features. In the present model,
183: it will be convenient to use the acceleration as a AE monitoring
184: tool, since the velocity has a bias induced by the external loading:
185: even in the absence of cracking the lattice has a non-vanishing velocity.
186: We define the associated cumulative energy as
187: \begin{equation}
188: E(t)\equiv \int_0^t dt' a^2(t').
189: \end{equation}
190: The behavior of the cumulative acoustic energy $E(t)$ is typically monitored
191: in AE experiments. In some cases, $E(t)$ is found to increase as a power law
192: \cite{ciliberto},
193: or exponentially in other cases \cite{paper}. In general one expects a
194: marked peak close to failure, as we also observe in Fig~\ref{fig:2},
195: obtained for $V=10^{-3}$.
196: The curve is well fitted by cubic law, $E \sim t^3$.
197:
198: \begin{figure}[h]
199: \centerline{\psfig{file=fig2.eps,width=8.5cm,clip=!}}
200: \caption{A typical signal measured in the model. The upper panel
201: shows the velocity acceleration of a site close to the boundary.
202: In the middle panel we show a magnification of a portion of the signal.
203: In the lower panel is reported the cumulative energy,
204: $E(t)\equiv \int_0^t dt' a^2(t')$, as a function of time for a single
205: realization of disorder. The dashed line follows $t^3$. }
206: \label{fig:2}
207: \end{figure}
208: \begin{figure}[t]
209: \centerline{\psfig{file=fig3.eps,width=8.5cm,clip=!}}
210: \caption{The damage $D$ evolution is displayed as a function
211: of time for different loading rates. The curves collapse onto
212: each other when plotted with respect to strain $\gamma\equiv Vt/L$.}
213: \label{fig:3}
214: \end{figure}
215:
216: A central problem in AE measurements is to correlate the recorded
217: acoustic activity with the internal damage state. In this way, AE
218: can be used as a tool for damage evaluation. In our model, we have
219: a direct access to the internal damage $D$ that can be defined
220: as the total number of failed bonds. We find that $D$ increases
221: linearly with time (see Fig.~\ref{fig:3}) apart from a rapid
222: increase very close to failure. Rescaling the curves with the loading
223: rate one sees that $D$ is in fact a linear function
224: of the applied strain $\gamma\equiv (Vt)/L$
225: (see the inset of Fig.~\ref{fig:4}).
226:
227:
228: \begin{figure}[t]
229: \centerline{\psfig{file=fig4.eps,width=8.5cm,clip=!}}
230: \caption{The average cumulative energy plotted as a function
231: of damage. The line represents a $D^3$ law which fits well the
232: curve.}
233: \label{fig:4}
234: \end{figure}
235:
236:
237: These observations thus lead to a direct scaling relation between
238: internal damage and released acoustic energy: Fig.~\ref{fig:4}
239: shows that $E$ scales as expected as $D^3$.
240: A direct consequence of this result is that
241: the measured acoustic energy is proportional to the
242: released elastic energy $E_{el}\sim KD\gamma^2 \sim D^3 \sim E$.
243:
244: A large amount of theoretical activity has been devoted in the past
245: to understand the origin of power law distributions of AE amplitudes
246: widely observed in material fracture. Most of the analysis was devoted
247: to quasi-static models, such as the the RFM, where
248: fracture was shown to occur in damage burst, distributed as
249: $P(D) \sim D^{-\tau}$ with $\tau \simeq 2.5$ \cite{hansen,zrvs}.
250: This value is in perfect agreement with the result $\tau=5/2$ obtained
251: exactly \cite{hansen1} for the exponent of the avalanche distribution of the
252: fiber bundle model (FBM) \cite{dfbm},
253: where $N$ fibers with random failure threshold
254: are loaded in parallel. It was thus conjectured that the long-range
255: stress transfer present in the RFM was equivalent to the infinite
256: range load redistribution of the FBM, placing the two models into
257: the same universality class \cite{hansen,zrvs}.
258: A similar exponent was found in a vectorial fracture model,
259: so that this class could be even broader \cite{zrvs}.
260: Comparing this result with AE experiments is problematic
261: since quasi-static models do not account for wave propagation.
262:
263: Here we can directly measure the distribution of pulse sizes
264: due to the acoustic activity. In Fig.~\ref{fig:5} we report the
265: distribution of energies for the acceleration
266: signal, defining $\epsilon\equiv a^2$.
267: In both cases the distribution
268: decays as power law with an exponent $\beta =1.7 \pm 0.1$,
269: independent on the loading rate, which only affects the
270: low part of the distribution.
271: The same law is found in the case of the velocity signal.
272: Experimental results report an exponent value in the same range,
273: even if it differs a little from one material to another:
274: for wood the exponent is $\beta =1.51 \pm 0.05$, for fiberglass $\beta =2.0 \pm 0.01$ \cite{ciliberto},
275: $\beta =1.30 \pm 0.1$ for paper \cite{paper},
276: $\beta =1.5 \pm 0.1$ for experiments on cellular glass \cite{strauven}.
277:
278: \begin{figure}[t]
279: \centerline{\psfig{file=fig5.eps,width=8.5cm,clip=!}}
280: \caption{The distribution of acoustic burst energies for different
281: driving rates. The line is a power law with exponent $\beta=1.7$.}
282: \label{fig:5}
283: \end{figure}
284:
285: Using the scaling relation between released acoustic energy
286: and damage discussed above, we can relate the exponent $\beta$ to $\tau$.
287: From $E \sim D^3$ and $D \sim t$, we expect $\epsilon \sim D^2$.
288: Substituting this expression in the equation for the probabilities
289: $P(\epsilon)d\epsilon=P(D)dD$, we obtain $\tau=1+2(\beta-1)=2.4$, which
290: is very close to $\tau=5/2$ measured in the RFM.
291: Thus we conjecture that the acoustic energy exponent measured
292: in our dynamic model is directly
293: related with the damage exponent measured in the corresponding
294: quasi-static model \cite{nota}.
295:
296:
297:
298: In conclusions, we have introduced a lattice model
299: of dynamic fracture which can be used to model AE experiments.
300: The model allows to clarify important issues in the interpretation
301: of the experiments, namely the relation between internal damage
302: and released acoustic energy. In particular, we derive direct relations
303: between the scaling behavior of failure avalanches and acoustic bursts.
304: It would be interesting to generalize this analysis to more realistic
305: situations, exploring the role of dimensionality,
306: load conditions and lattice anisotropy.
307: However, in comparing the simulated signal with experiments, we should be
308: careful about the definition of the events in the time series, since
309: the amplifier and the AE sensors could bias the recorded waveform, introducing a
310: systematic error in the data.
311:
312:
313: This work has been supported by the European Network contract FMRXCT980183
314: and the INFM center SMC.
315:
316:
317: \begin{thebibliography}{99}
318:
319: \bibitem{SET-01}
320: J.~Sethna, K.~A. Dahmen, C.~R. Myers, Nature {\bf 410} 242 (2001).
321: \bibitem{flux}
322: S. Field, J. Witt, F. Nori and X. Ling,
323: Phys. Rev. Lett. {\bf 74}, 1206 (1995).
324: \bibitem{DUR-01}
325: G. Durin and S. Zapperi,
326: Phys. Rev. Lett. {\bf 84}, 4705 (2000).
327: \bibitem{MIG-01}
328: M. C. Miguel, A. Vespignani, S. Zapperi, J. Weiss and J. R. Grasso,
329: Nature {\bf 410}, 667 (2001).
330: \bibitem{ciliberto}
331: A. Garcimart\'{\i}n, A. Guarino,
332: L. Bellon and S. Ciliberto,
333: Phys. Rev. Lett. {\bf 79}, 3202 (1997);
334: A. Guarino, A. Garcimart\'{\i}n and S. Ciliberto,
335: Eur. Phys. J. B {\bf 6}, 13 (1998).
336: \bibitem{strauven}
337: C. Maes, A. Van Moffaert, H. Frederix and H. Strauven,
338: Phys. Rev. B {\bf 57}, 4987 (1998).
339: \bibitem{ae}
340: A. Petri, G. Paparo, A. Vespignani,
341: A. Alippi and M. Costantini,
342: Phys. Rev. Lett. {\bf 73}, 3423 (1994).
343: \bibitem{paper}
344: L.I. Salminen, A.I. Tolvanen, and M. J. Alava,
345: {\em Acoustic Emission from Paper Fracture}, preprint.
346: \bibitem{hr}
347: H.J. Herrmann and S. Roux (eds.), {\em Statistical Models for
348: the Fracture of Disordered Media} (North Holland, Amsterdam, 1990).
349: \bibitem{fuse}
350: L. de Arcangelis, S. Redner and H. J. Herrmann,
351: J. Phys. Lett. (Paris) {\bf 46}, L585 (1985).
352: \bibitem{duxbury}
353: P. M. Duxbury, P. D. Beale and P. L. Leath,
354: Phys. Rev. Lett. {\bf 57}, 1052 (1986).
355: \bibitem{hansen}
356: A. Hansen and P. C. Hemmer,
357: Phys. Lett. A {\bf 184}, 394 (1994).
358: \bibitem{zrvs}
359: S.\ Zapperi, P.\ Ray, H.\ E.\ Stanley, and A.\ Vespignani,
360: Phys. Rev. Lett. {\bf 78}, 1408 (1997);
361: Phys.\ Rev.\ E {\bf 59}, 5049 (1999).
362: \bibitem{zvs}
363: S.\ Zapperi, A.\ Vespignani, and H.\ E.\ Stanley,
364: Nature (London) {\bf 388}, 658 (1997).
365: \bibitem{gcalda}
366: G. Caldarelli, F. D. Di Tolla and A. Petri
367: Phys. Rev. Lett. {\bf 77}, 2503 (1996).
368: \bibitem{alava}
369: V. I. R\"ais\"anen, M. J. Alava, R. M. Nieminen,
370: Phys. Rev. B, {\bf 58}, 14288 (1998).
371: \bibitem{dyn}
372: See J. Fineberg and M. Marder, Phys. Rep.313, 1 (1999),
373: for a review on dynamic fracture.
374: \bibitem{marder}
375: M. Marder and X. Liu,
376: Phys. Rev. Lett. {\bf 71}, 2417 (1993).
377: \bibitem{RAU}
378: T. T. Rautiainen, M. J. Alava, and K. Kaski,
379: Phys. Rev. E {\bf 51} R2727 (1994);
380: ibid {\bf 56}, 6443 (1997).
381: \bibitem{mar}
382: T. Mart\'in, P. Espa\~nol, M. A. Rubio, and I. Z\'u\~niga
383: Phys. Rev. E 61, 6120 (2000)
384: \bibitem{hansen1}
385: P. C. Hemmer and A. Hansen, J. Appl. Mech. {\bf 59},
386: 909 (1992); M. Kloster, A. Hansen and P. C. Hemmer
387: Phys. Rev E {\bf 56}, 2615 (1997).
388: \bibitem{dfbm}
389: H. A. Daniels, Proc. Roy. Soc. London {\bf A183},
390: 405 (1945);
391: S. L. Phoenix and H. M. Taylor,
392: Adv. Appl. Prob. {\bf 5}, 200 (1973).
393: \bibitem{nota} The distribution of the ``elastic'' energy
394: released in each avalanche was recently measured in
395: the quasi-static FBM and found to decay as power law with
396: exponent $\beta\simeq 1.8$ \protect\cite{paper},
397: in agreement with our results.
398: \end{thebibliography}
399:
400:
401: \end{document}
402: