cond-mat0610483/text.tex
1: \documentclass[prl,preprint,showpacs,amssymb,floatfix,superscriptaddress]{revtex4}
2: \usepackage{graphicx} \usepackage{subfigure}
3: 
4: 
5: \begin{document}
6: 
7: \title{Universal Anisotropy in Force Networks under Shear}
8: 
9: \author{Srdjan Ostojic}  \altaffiliation[Present address: ]{Laboratoire de Physique Statistique, Ecole Normale Sup\'erieure, 24, rue Lhomond, 75231 Paris Cedex 05, France}\affiliation{Institute for
10: Theoretical Physics, Universiteit van Amsterdam, Valckenierstraat 65,
11: 1018 XE Amsterdam, The Netherlands}
12: 
13: \author{Thijs J. H. Vlugt} \affiliation{Condensed Matter and
14:   Interfaces (CMI), Utrecht University, P.O.Box 80.000, 3508 TA Utrecht, The Netherlands}
15: 
16: \author{Bernard Nienhuis} \affiliation{Institute for
17: Theoretical Physics, Universiteit van Amsterdam, Valckenierstraat 65,
18: 1018 XE Amsterdam, The Netherlands}
19: 
20: \date{\today}
21: 
22: \begin{abstract} 
23: Scaling  properties of  patterns formed  by large  contact  forces are
24: studied as a function of  the applied shear stress, in two-dimensional
25: static  packings  generated  from  the  force  network  ensemble.   An
26: anisotropic finite-size-scaling analysis  shows that the applied shear
27: does not  affect the universal  scaling properties of  these patterns,
28: but simply induces different length scales in the principal directions
29: of the  macroscopic stress tensor.   The ratio of these  length scales
30: quantifies the anisotropy  of the force networks, and  is found not to
31: depend on the  details of the underlying contact  network, in contrast
32: with other properties such as the yield stress.
33: \end{abstract}
34: 
35: \pacs{45.70.-n, 45.70.Cc, 46.65.+g}
36: \maketitle
37: 
38: 
39: Aggregates of macroscopic particles  such as granular materials, foams
40: and emulsions are often found  in a disordered, solid-like state whose
41: mechanical properties  have attracted  much attention in  recent years
42: \cite{jaeger:rev,degennes:rev,bouch:rev,liu98,jamming:book,radjai98,geng03,majmudar05,atman05-2,snoeijer04-1,snoeijer05,tighe05,ostojic06}.
43: Under  external  stresses,  these  systems present  a  non-zero  yield
44: threshold  which is  solely due  to  the intricate  network formed  by
45: contact forces between particles. One of the most remarked features of
46: these highly disordered {\em force networks} is the tendency for large
47: forces to align and form  branching {\em force chains}. When these appear  in  response to an external  stress, such as  a global shear, their spatial inhomogeneity is most striking \cite{radjai98,geng03,majmudar05}.
48: As the applied shear stress is increased, the large forces orient in a
49: preferred  direction, and  the  force network  becomes  more and  more
50: anisotropic (c.f.~Fig.~\ref{fig1}), up to  the point where the applied
51: stress can no longer be  sustained, and the packing yields. While this
52: qualitative  picture has  been  long  established, due  to  a lack  of
53: appropriate  analytic  and numerical  tools,  quantitative studies  of
54: packings     under    static    shear     have    been    few
55: \cite{radjai98,geng03,majmudar05,atman05-2,snoeijer05}.
56: 
57: Recently,  a  novel  characterization  has  been  introduced  for  the
58: geometrical   patterns  formed  by   large  forces   in  isotropically
59: compressed force networks \cite {ostojic06}).  The patterns, displayed by molecular
60: dynamics (MD)  simulations of  granular packings, turned  out to have
61: scaling     properties     which     are    independent     of     the
62: pressure, polydispersity  and  friction.  An intriguing  question  is
63: whether an  analysis of scaling  properties can provide  insight into
64: the organization of force networks in packings under shear. This is a
65: much more difficult problem.
66: 
67:   Creating static  packings under shear  with MD is seriously
68: hindered by local rearrangements that  seem to prevent the system from
69: reaching  a clear  mechanical  equilibrium.  To  allow for  systematic
70: examinations  of   the  effects  of   shear,  it  has   been  proposed
71: \cite{snoeijer04-1,snoeijer05}  to  ignore the  microscopic  rearrangements of  the
72: grains, and study  the ensembles of force networks  allowed on a given
73: contact  network, as  function of  macroscopic stresses.   This purely
74: statistical approach was found to account well for the properties of
75:     packings under shear such as the existence of a yield threshold 
76: \cite{snoeijer05}, as  well as for the scaling  properties of clusters
77: of large forces in packings without shear \cite{ostojic06}. 
78: It also describes remarkably well the distribution of force magnitudes
79: \cite{snoeijer04-1,tighe05,snoeijer05} and the response to an external
80: overload \cite{ostojic06-2}.
81: 
82: \begin{figure}
83: \begin{center} 
84: \begin{minipage}{0.32\linewidth}
85: \includegraphics [width=0.95\linewidth]{fig1a}
86: \end{minipage}
87: \begin{minipage}{0.32\linewidth}
88: \includegraphics [width=0.95\linewidth]{fig1b}
89: \end{minipage}
90: \begin{minipage}{0.32\linewidth}
91: \includegraphics [width=0.95\linewidth]{fig1c}
92: \end{minipage}
93: 
94: \begin{minipage}{0.32\linewidth}
95: \includegraphics [width=0.95\linewidth]{fig1d}
96: \end{minipage}
97: \begin{minipage}{0.32\linewidth}
98: \includegraphics [width=0.98\linewidth]{fig1e}
99: \end{minipage}
100: \begin{minipage}{0.32\linewidth}
101: \includegraphics [width=0.98\linewidth]{fig1f}
102: \end{minipage}
103: \caption{Effect  of  shear on  force  networks:  (a-c) Force  networks
104: obtained from the force ensemble, for three different values of shear:
105: (a)  $\tau=0$,  (b)  $\tau=0.2$,   (c)  $\tau=0.4$.   The  forces  are
106: represented  as bonds  connecting centers  of grains  of  contact, the
107: width  of  each  bond  being  proportional to  the  magnitude  of  the
108: force. The  underlying contact network is  the same for  all values of
109: $\tau$.  (d-f)  Clusters of forces  larger than a threshold  $f$ (bold
110: bonds), obtained for $f$ close to the critical value.  \label{fig1}}
111: \end{center}
112: \end{figure}
113: 
114: In this Letter,  we study the geometry of patterns  of large forces in
115: two-dimensional, static packings under  shear generated from the force
116: network ensemble.  We find that applying shear stress leaves unchanged
117: the universal  scaling properties of the patterns,  but introduces two
118: different  length scales  in the  two directions  of  principal stress
119: axes.  We characterize  the anisotropy  of the  force networks  by the
120: ratio $r$ of these  length scales.  An anisotropic finite-size scaling
121: analysis allows us  to determine $r$ as function  of the applied shear
122: stress $\tau$ and the coordination number of the packing.  In contrast with
123: other properties such  as the yield stress, $r(\tau)$  turns out to be
124: independent of the underlying contact geometry, and thus provides a universal characterization of shear-induced anisotropy.
125: 
126: 
127: 
128: 
129: {\em Force networks---} To study systematically the influence of the
130: shear  stress  on the  geometry  of  large  forces we  examined  force
131: networks  generated  from the  force  network  ensemble pertaining  to
132: packings  of  frictionless,  non-attracting particles.  This  approach
133: relies on  the fact that in  a typical packing, the  number of unknown
134: contact forces is  larger than the number of  equations for mechanical
135: balance  \cite{snoeijer04-1}.  For a  given, fixed,  contact geometry,
136: the force network ensemble consists of all force networks with repulsive
137: contact forces in mechanical balance on each grain, and
138: consistent with the tensor $\sigma$ of applied macroscopic stress. The
139: considered  contact networks  are isotropic,  and the  coordinates are
140: such     that     the      pressure     is   isotropic,
141: i.e.~$\sigma_{xx}=\sigma_{yy}$,  while the dimensionless  shear stress
142: $\tau=\sigma_{xy}/\sigma_{xx}$ is varied.
143: 
144:  The ensemble is sampled numerically using two different methods : (i)
145: the wheel-move algorithm \cite{tighe05} for ordered, hexagonal packings
146: ($z=6$); (ii) the  procedure outlined in \cite{snoeijer05} for
147: disordered packings of various  coordination numbers $z$. Both methods
148:   efficiently generate very large packings ($10^5$ and $5.10^4$
149: particles respectively), from which smaller sub-systems were extracted
150: and analyzed.   Note that  the ensemble is  non-empty only  for $\tau$
151: smaller than a maximal value $\tau_{max}(z)$, which has been identified as
152: the yield stress \cite{snoeijer05}.
153: 
154: {\em  Force chains---}  To study  the geometry  of patterns  formed by
155: large  forces, we  follow the  method introduced  in \cite{ostojic06},
156: which  we  here briefly  recall  and  then  extend to  packings  under
157: shear. The force network in a packing of frictionless particles can be
158: represented as a set of bonds connecting grains in contact, where each
159: bond  carries  the  magnitude   of  the  corresponding  contact  force
160: (c.f.~Fig.~\ref{fig1}  (a-c)).  A natural  way  to  isolate the  force
161: chains in  such a network is  to choose a threshold  $f$, and consider
162: only  the subgraph  formed by  bonds carrying  forces larger  than $f$
163: (c.f.~Fig.~\ref{fig1} (d-f)).  Varying this threshold  allows to study
164: patterns of large  forces at different scales: for  small values, most
165: of the grains remain connected, but as the threshold is increased, the
166: extracted subgraph  breaks into  disconnected clusters. The  extent of
167: each chain of forces larger than the threshold can be characterized by
168: the size of the corresponding cluster, quantified by the number $s$ of
169: mutually connected bonds.
170: 
171: In  an  ensemble  of   packings,  a  statistical  description  of  the
172: fluctuating patterns of large forces is given by the statistics of the
173: clusters obtained  at different  thresholds. Following the  methods of
174: percolation theory  \cite{stauffer:book}, the geometry  of an ensemble
175: of packings can be characterized by $P(s,f)$, the number (per bond) of
176: clusters  of  size  $s$  at  the threshold  $f$.  For  packings  under
177: isotropic pressure, it is moreover natural to describe the clusters by
178: a  single  characteristic   length,  the  cluster  correlation  length
179: $\xi(f)$, which corresponds to the  typical linear size of the 
180: clusters  at a  given threshold  (not considering  clusters percolating
181: through the whole system).
182: 
183: {\em Scaling analysis---}Analogously  to percolation and other lattice
184: models of critical phenomena, the  system is critical around the value
185: $f_c$ of  the threshold above which  no infinite cluster  is found. At
186: this  value,  $\xi(f)$  diverges  as  $|f-f_c|^{-\nu}$,  and  $P(s,f)$
187: becomes a  power-law, its moments  $\langle s^n \rangle$  diverging as
188: powers  of  $\xi(f)$:  $\langle  s^n \rangle\sim  \xi^{\phi_n}$.   For
189: systems of finite size, the correlation length is naturally bounded by
190: the  linear size  of  system, $L$, and  for  $\xi$ comparable  to $L$,  the
191: behavior of  $\langle s^n \rangle$ is  given by a  scaling function of
192: $L/\xi$ \cite{privman:book,binder89}:
193: \begin{equation}
194: \langle s^n(f,L) \rangle \sim L^{\phi_n}\Sigma_n[(f-f_c)L^{1/\nu}], \label{iso-fss}
195: \end{equation}
196: where  $\phi_n$  and  $\nu$  are  universal  critical  exponents,  and
197: $\Sigma_n$  universal  scaling functions.   For  force networks  under
198: isotropic  pressure,  $\phi_2=1.89$,  $\nu=1.65$,  and  $\Sigma_2$  is
199: independent  of  pressure,  polydispersity,  friction  and  force  law
200: \cite{ostojic06}.  From now on, we  will only consider the case $n=2$,
201: and omit the indices $n$ in the right-hand-side. Note that the scaling
202: function  $\Sigma_2$ varies  smoothly  and displays  a single  maximum
203: close to zero.
204: 
205: Equation  (\ref{iso-fss})  assumes that  the  size  of the examined
206: domains is isotropic. In the following, we will need to consider
207: %  More generally, for
208:   rectangular domains of size
209: $L_1\times   L_2$, in which case  the   scaling  function   $\Sigma$  also  depends   on the  aspect ratio  $L_2/ L_1$.   Considering  only the
210: maximum of $\langle s^2(f,L_1,L_2) \rangle$ with respect to $f$, which
211: is  equivalent  to looking  only  at  the  behavior  at  the effective
212: critical point, Eq.~(\ref{iso-fss}) reduces to
213: \begin{equation}
214: \langle s^2(L_1,L_2)\rangle _{max}=L_2^{\phi}{\bar \Sigma}\left(\frac{L_1}{L_2}\right).\label{eq:ch4-fss2}
215: \end{equation}
216: The scaling function ${\bar \Sigma}(x)$ simply expresses the fact
217:  that if $L_1\ll L_2$, the effective correlation length is set by
218:  $L_1$, so that $\langle s^2\rangle _{max}\sim L_1^{\phi}$,
219:  and correspondingly if $L_2\ll L_1$, $\langle s^2\rangle _{max}\sim
220:  L_2^{\phi}$. We therefore have (with $y=L_1/L_2$)
221: \begin{equation}
222:    {\bar \Sigma}(y) \sim\left\{
223:     \begin{array}{ll}
224:      y^{\phi} & \textrm{for $y \ll 1$}\\
225:      1 & \textrm{for  $y \gg 1$}
226:     \end{array}\right.
227:   \end{equation}
228: For $L_1\sim L_2$, there is a crossover between the two asymptotic trends, and the precise
229: behavior is determined by the scaling function.
230: 
231: {\em  Shear-induced  anisotropy---}  The  analysis  presented  so  far
232: pertains to the  case of isotropic pressure. In  packings under shear,
233: the force networks become increasingly anisotropic as the shear stress
234: is  increased,  the  large   forces  aligning  preferentially  in  the
235: direction  of  the   maximal  stress  axis  (cf.~Fig.~\ref{fig1}).  In
236: consequence,  close to the percolation threshold the clusters 
237: can  no longer be  described by  a single  length scale.   Instead two
238: correlation  lengths  $\xi_M$   and  $\xi_m$  must  be  distinguished,
239:  in the direction of  maximal and minimal stress axes respectively.  In
240: principle, at  criticality these two length scales  could diverge with
241: two different  exponents $\nu_M$  and $\nu_m$.  A  finite-size scaling
242: analysis as in  \cite{ostojic06}, but now varying independently
243: the  system sizes  $L_M$  and  $L_m$ in  the  directions of  principal
244: stresses, shows that $\nu_M=\nu_m=\nu$, the exponents $\phi$ and $\nu$
245: being equal to those of isotropically compressed packings.
246: 
247: Although  they diverge  with  the  same exponent,  the  values of  the
248: correlation lengths $\xi_M$ and  $\xi_m$ clearly differ.  This implies
249: that, to leading order,
250: \begin{equation}
251: \xi_{M}=\xi_{M}^{(0)}|f-f_c|^{-\nu} \,\,\,\mathrm{ and}\,\,\,
252: \xi_{m}=\xi_{m}^{(0)}|f-f_c|^{-\nu}. \label{eq:ch4-xi0}
253: \end{equation}
254: where the length scales  $\xi_{M}^{(0)}$ and $\xi_{m}^{(0)}$ depend on
255: the shear stress $\tau$ but not  on the threshold $f$. In the language
256: of critical phenomena, this  situation is called {\em weakly anisotropic
257: scaling},  as opposed  to the  case where  the  two correlation-length
258: exponents differ \cite{binder89}.
259: 
260:   To quantify  the anisotropy of  force networks as function  of shear
261: stress,  from Eq.~(\ref{eq:ch4-xi0}),  it appears  natural to  use the
262: anisotropy ratio $r(\tau)$ defined as
263: \begin{equation}
264: r(\tau)=\frac{\xi_{m}^{(0)}}{\xi_{M}^{(0)}}.
265: \end{equation}
266: Clearly, for $\tau=0$, $r=1$, and as $\tau$ is increased, $r$
267: decreases below one.
268: 
269: The anisotropy ratio $r$ as function of $\tau$ can be determined using
270: an anisotropic  scaling analysis. The central observation  is that, in
271: the context  of conventional critical phenomena,  a weakly anisotropic
272: system can  be made isotropic simply  by rescaling the  lengths in the
273: two  directions  of principal  axes.   In  the  present setting,  this
274: property  suggests that  if the  actual  lengths $L_M$  and $L_m$  are
275: replaced  by  properly rescaled  effective  lengths $\tilde{L}_M$  and
276: $\tilde{L}_m$, the  scaling properties  of the sheared  force networks
277: should   be   identical   to   those   of   isotropically   compressed
278: networks.  This  property  allows  us to  determine  the ratio of length scales
279: $\xi_{M}^{(0)}$ and $\xi_{m}^{(0)}$.
280:  
281: More specifically, the scaling of  the maximum of $\langle s^2\rangle$ is
282: described  by Eq.~(\ref{eq:ch4-fss2})  with an additional dependence on
283: the shear  stress $\tau$  in the right  hand side.  Our  hypothesis is
284: that the  scaling function ${\bar \Sigma}$ does  not depend explicitly
285: on  $\tau$,  and that  the  anisotropy  can  be eliminated  simply  by
286: replacing  $L_M$  and  $L_m$  by rescaled  lengths  $\tilde{L}_M$  and
287: $\tilde{L}_m$ given by
288: \begin{equation}
289: \tilde{L}_M =b_M(\tau) L_M \nonumber\,\,\, \mathrm{ and} \,\,\,
290: \tilde{L}_m=b_m(\tau) L_m  \label{eq:ch4-bs}
291: \end{equation}
292: with $b_M=1/\xi_{M}^{(0)}$  and  $b_m=
293: 1/\xi_{m}^{(0)}$,                 and                therefore
294: $r(\tau)=b_M/b_m$.  Substituting  into Eq.~(\ref{eq:ch4-fss2}),  the
295: scaling relation becomes
296: \begin{equation}
297: \langle s^2(L_M,L_m,\tau)\rangle _{max}=(b_m L_m)^{\phi}{\bar
298:   \Sigma}\left(\frac{b_M L_M}{b_m L_m}\right).\label{eq:ch4-fss3}
299: \end{equation}
300: where the dependence on $\tau$ occurs only through the scale factors
301: $b_m$ and $b_M$.
302: 
303: 
304: The  validity  of  the  scaling relation  (\ref{eq:ch4-fss3})  can  be
305: checked          directly          from         the          numerical
306: data.  Fig.~\ref{fig:ch4-b4_collapse}   illustrates  the  behavior  of
307: $\langle s^2 \rangle_{max}$ as  function of $L_m$, for three different
308: values  of  $\tau$,  and  in  each  case  three  different  values  of
309: $L_M$.  For fixed $L_M$  and $\tau$,  as $L_m$  increases, at  first a
310: clear  power-law  can be  observed.  For  larger  values of  $L_m$,  a
311: crossover occurs and $\langle s^2 \rangle_{max}$ reaches a plateau. As
312: $L_M$ is increased,  the behavior in the scaling  regime is unchanged,
313: but the crossover occurs at larger $L_m$, and the value of the plateau
314: increases.   According to  Eq.~(\ref{eq:ch4-fss2}), the  value  of the
315: crossover  should scale  as $L_M$,  and the  value of  the  plateau as
316: $L_M^{\phi}$.   Rescaling  both   axes  appropriately,  the  data  for
317: different  $L_M$  indeed collapse  on  the  same  curves as  shown  in
318: Fig.~\ref{fig:ch4-b4_collapse}~(b).
319: 
320: For different values of the shear stress, the exponents in the scaling
321: regimes appear  to be identical,  however the prefactors of  the power
322: laws  clearly depend  on $\tau$.  From  Eq.~(\ref{eq:ch4-fss3}), these
323: prefactors correspond to $b_M$ and $b_m$, which can thus be determined
324: respectively from the  value reached at the plateau  and the intercept
325: of the  power law. Replacing  $L_M$ and $L_m$ by  rescaled lengths
326: $\tilde{L}_M$ and  $\tilde{L}_m$ defined in  Eq.~(\ref{eq:ch4-bs}), if
327: the scaling function ${\bar \Sigma}$ is independent of $\tau$, all the
328: data  must  collapse  on  a  single  curve. Note  that  as  $\tau$  is
329: increased, the cross  over occurs at smaller $L_m$,  and the domain of
330: power-law  scaling  shrinks,  so  that  $b_m$ can  be  extracted  only
331: for $\tau$ sufficiently smaller than the yield stress $\tau_{max}$ (beyond which the force network ensemble is empty).
332: 
333: 
334: \begin{figure}
335: \begin{center} 
336: \includegraphics [width=0.8\linewidth]{fig2}
337: \caption{(color online) Scaling of the maximum of the second moment of cluster sizes:
338:   (a)  $\langle s^2\rangle_{max}(L_M, L_m)$   as function  of $L_m$.   Three different  values of  $L_M$ are
339:   represented  with  circles,   squares  and  triangles,  while  three
340:   different values  of the shear  stress are shown in  three different
341:   colors.   The displayed  data  corresponds to  hexagonal packing  of
342:   disks,  i.e.~ $z=6$.  (b) Collapses  of data  corresponding  to same
343:   $\tau$,      obtained     by      rescaling     the      axes     by
344:   $L_M$. The curves are colored in the order of the legend. \label{fig:ch4-b4_collapse}} 
345: \end{center}
346: \end{figure}
347: 
348: 
349: 
350: 
351: 
352: \begin{figure*}
353: \begin{center} 
354: \includegraphics [width=0.45\linewidth]{fig3}
355: \end{center}
356: \caption{(color online) The    anisotropic   scaling    function    ${\bar   \Sigma}$
357:   (cf.~Eq.~(\ref{eq:ch4-fss2})):   collapse  of   data   obtained  for
358:   different  values  of  shear  stress $\tau$  and  different  packing
359:   geometries. Each symbol   corresponds to data for a different
360:   coordination number  $z$ ($z=6$: regular,  hexagonal packing, already displayed in Fig.~
361:   \ref{fig:ch4-b4_collapse}; all others: disordered packings).  Each value of $\tau$ is represented
362:   by  a different  color.  The inset  shows  the anisotropy  parameter
363:   $r=\frac{b_M}{b_m}$ as  function of  $\tau$, for different $z$. The
364:   black solid line is a quadratic fit.\label{fig:ch4-tau_collapse}}
365: \end{figure*}
366: 
367: 
368: 
369: 
370: 
371: Fig.~\ref{fig:ch4-tau_collapse} displays the  results obtained for all
372: considered  values   of  shear  stress  $\tau$,   on  several  packing
373: geometries of different coordination  number $z$ ($z=6$ corresponds to
374: the   ordered   hexagonal  case,   while   the   other  packings   are
375: disordered). All  the data  clearly collapses on  the same  curve, in
376: agreement with the hypothesis 
377: that the scaling function  does not depend on $\tau$. As
378: postulated,  a  simple rescaling  of  the  length  scales in  the  two
379: directions  of  anisotropy is  thus  sufficient  to recover  isotropic
380: scaling in force networks under shear.The inset  of Fig.~\ref{fig:ch4-tau_collapse} displays  the anisotropy
381: parameter  $r=\frac{b_M}{b_m}$ as  function  of $\tau$.  Unexpectedly,
382: $r(\tau)$  appears to be  independent of  the coordination  number and
383: geometry of the
384: underlying  contact  network.
385: 
386: 
387: 
388: 
389: 
390: {\em Discussion---}  Our results show that applying  an external shear
391: stress  on a  force  network  does not  affect  the universal  scaling
392: properties of the force chains, but only induces two different length scales
393: in  the directions  of the  two  principal stress  axes.  While  these
394: typical lengths  and their ratio $r$  are not a priori  expected to be
395: universal,  we  find  that  they  are identical  for  various  contact
396: networks we  considered, which include the  regular hexagonal packing,
397: and disordered networks of different coordination numbers, and in this
398: sense universal.
399: 
400: %In  experimental studies  \cite{geng03,majmudar05}, the  anisotropy of
401: %force  networks  under  shear  was  described  using  the  force-force
402: %correlation function. While in  absence of shear the correlations are
403: %isotropic and  decay fast, in presence  of shear it  was observed that
404: %the  correlations in  the direction  of maximal  stress are  of much
405: %longer range  than in the  perpendicular direction. We  have evaluated
406: %the same  quantities in  networks generated from the  force ensemble.  Although the
407: %anisotropy  can  be  clearly  seen,  we  nevertheless  find  that  the
408: %correlations in the direction of  maximal stress vanish after a couple
409: %of grain diameters, even for  very large shear. As far as correlations
410: %are concerned,  the isotropy of  the underlying contact  network might
411: %thus play a role. 
412: 
413: An important remaining question is  the behavior close to the yielding
414: point. If yielding is analogous to a phase transition, as suggested by
415: the  jamming picture  \cite{liu98,jamming:book},  it could  be
416: expected that close to it a cross-over occurs, and the scaling
417: properties of clusters of  large forces change significantly. The
418: value of the yield stress $\tau_{max}$ was found to be strongly dependent on the
419: coordination number $z$ of the contact network \cite{snoeijer05}. On the other hand, we
420: find that $r(\tau)$ is completely independent of $z$ up to $\tau_{max}(z)$.
421: This observation suggests  the absence  of any diverging  or vanishing
422: length   scale  which   would  accompany   the  cross-over   close  to
423: $\tau_{max}$. Moreover,  we have not  observed any dramatic  change in
424: the scaling  properties close to $\tau_{max}$, but  additional work is
425: necessary to clarify these issues.
426: 
427: TJHV acknowledges financial support from the Netherlands Organization
428: for Scientific Research (NWO-CW) through a VIDI grant. SO is financially supported by the Dutch research organization FOM.
429: 
430: \begin{thebibliography}{19}
431: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
432: \expandafter\ifx\csname bibnamefont\endcsname\relax
433:   \def\bibnamefont#1{#1}\fi
434: \expandafter\ifx\csname bibfnamefont\endcsname\relax
435:   \def\bibfnamefont#1{#1}\fi
436: \expandafter\ifx\csname citenamefont\endcsname\relax
437:   \def\citenamefont#1{#1}\fi
438: \expandafter\ifx\csname url\endcsname\relax
439:   \def\url#1{\texttt{#1}}\fi
440: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
441: \providecommand{\bibinfo}[2]{#2}
442: \providecommand{\eprint}[2][]{\url{#2}}
443: 
444: \bibitem[{\citenamefont{Jaeger et~al.}(1996)\citenamefont{Jaeger, Nagel, and
445:   Behringer}}]{jaeger:rev}
446: \bibinfo{author}{\bibfnamefont{H.~M.} \bibnamefont{Jaeger}},
447:   \bibinfo{author}{\bibfnamefont{S.~R.} \bibnamefont{Nagel}}, \bibnamefont{and}
448:   \bibinfo{author}{\bibfnamefont{R.~P.} \bibnamefont{Behringer}},
449:   \bibinfo{journal}{Rev. Mod. Phys.} \textbf{\bibinfo{volume}{68}},
450:   \bibinfo{pages}{1259} (\bibinfo{year}{1996}).
451: 
452: \bibitem[{\citenamefont{de~Gennes}(1999)}]{degennes:rev}
453: \bibinfo{author}{\bibfnamefont{P.-G.} \bibnamefont{de~Gennes}},
454:   \bibinfo{journal}{Rev. Mod. Phys.} \textbf{\bibinfo{volume}{71}},
455:   \bibinfo{pages}{374} (\bibinfo{year}{1999}).
456: 
457: \bibitem[{\citenamefont{Bouchaud}(2003)}]{bouch:rev}
458: \bibinfo{author}{\bibfnamefont{J.-P.} \bibnamefont{Bouchaud}}, in
459:   \emph{\bibinfo{booktitle}{Les Houches, Session LXXVII}}, edited by
460:   \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{Barrat}}
461:   (\bibinfo{publisher}{EDP Sciences}, \bibinfo{year}{2003}).
462: 
463: \bibitem[{\citenamefont{Liu and Nagel}(1998)}]{liu98}
464: \bibinfo{author}{\bibfnamefont{A.~J.} \bibnamefont{Liu}} \bibnamefont{and}
465:   \bibinfo{author}{\bibfnamefont{S.~R.} \bibnamefont{Nagel}},
466:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{396}}, \bibinfo{pages}{21}
467:   (\bibinfo{year}{1998}).
468: 
469: \bibitem[{\citenamefont{Liu and Nagel}(2001)}]{jamming:book}
470: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Liu}} \bibnamefont{and}
471:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Nagel}},
472:   \emph{\bibinfo{title}{Jamming and rheology}} (\bibinfo{publisher}{Taylor \&
473:   Francis}, \bibinfo{year}{2001}).
474: 
475: \bibitem[{\citenamefont{Radjai et~al.}(1998)\citenamefont{Radjai, Wolf, Jean,
476:   and Moreau}}]{radjai98}
477: \bibinfo{author}{\bibfnamefont{F.}~\bibnamefont{Radjai}},
478:   \bibinfo{author}{\bibfnamefont{D.~E.} \bibnamefont{Wolf}},
479:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Jean}}, \bibnamefont{and}
480:   \bibinfo{author}{\bibfnamefont{J.-J.} \bibnamefont{Moreau}},
481:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{80}},
482:   \bibinfo{pages}{61} (\bibinfo{year}{1998}).
483: 
484: \bibitem[{\citenamefont{Geng et~al.}(2003)\citenamefont{Geng et~al.}}]{geng03}
485: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Geng}},
486:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Reydellet}},
487:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Clement}}, \bibnamefont{and}
488:   \bibinfo{author}{\bibfnamefont{R.~P.} \bibnamefont{Behringer}},
489:   \bibinfo{journal}{Physica D} \textbf{\bibinfo{volume}{182}},
490:   \bibinfo{pages}{274} (\bibinfo{year}{2003}).
491: 
492: \bibitem[{\citenamefont{Majmudar and Behringer}(2005)}]{majmudar05}
493: \bibinfo{author}{\bibfnamefont{T.~S.} \bibnamefont{Majmudar}} \bibnamefont{and}
494:   \bibinfo{author}{\bibfnamefont{R.~P.} \bibnamefont{Behringer}},
495:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{435}},
496:   \bibinfo{pages}{1079} (\bibinfo{year}{2005}).
497: 
498: 
499: \bibitem[{\citenamefont{Atman and {\it et al.}}(2005)}]{atman05-2}
500: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Atman}} \bibnamefont{and}
501:   \bibinfo{author}{\bibnamefont{{\it et al.}}}, \bibinfo{journal}{J. Phys.
502:   Cond. Mat.} \textbf{\bibinfo{volume}{17}}, \bibinfo{pages}{S2391}
503:   (\bibinfo{year}{2005}).
504: 
505: \bibitem[{\citenamefont{Snoeijer et~al.}(2004)\citenamefont{Snoeijer, Vlugt,
506:   van Hecke, and van Saarloos}}]{snoeijer04-1}
507: \bibinfo{author}{\bibfnamefont{J.~H.} \bibnamefont{Snoeijer}},
508:   \bibinfo{author}{\bibfnamefont{T.~J.~H.} \bibnamefont{Vlugt}},
509:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{van Hecke}},
510:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{van
511:   Saarloos}}, \bibinfo{journal}{Phys. Rev. Lett.}
512:   \textbf{\bibinfo{volume}{92}}, \bibinfo{pages}{054302}
513:   (\bibinfo{year}{2004}).
514: 
515: 
516: \bibitem[{\citenamefont{Snoeijer et~al.}(2006)\citenamefont{Snoeijer,
517:   Ellenbroek, Vlugt, and van Hecke}}]{snoeijer05}
518: \bibinfo{author}{\bibfnamefont{J.~H.} \bibnamefont{Snoeijer}},
519:   \bibinfo{author}{\bibfnamefont{W.~G.} \bibnamefont{Ellenbroek}},
520:   \bibinfo{author}{\bibfnamefont{T.~J.~H.} \bibnamefont{Vlugt}},
521:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{van
522:   Hecke}}, \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{96}},
523:   \bibinfo{pages}{098001} (\bibinfo{year}{2006}).
524: 
525: 
526: \bibitem[{\citenamefont{Tighe and {\it et al.}}(2005)}]{tighe05}
527: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Tighe}} \bibnamefont{and}
528:   \bibinfo{author}{\bibnamefont{{\it et al.}}}, \bibinfo{journal}{Phys. Rev. E}
529:   \textbf{\bibinfo{volume}{72}}, \bibinfo{pages}{031306}
530:   (\bibinfo{year}{2005}).
531: 
532: 
533: \bibitem[{\citenamefont{Ostojic et~al.}(2006)\citenamefont{Ostojic, Somfai, and
534:   Nienhuis}}]{ostojic06}
535: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Ostojic}},
536:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Somfai}}, \bibnamefont{and}
537:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Nienhuis}},
538:   \bibinfo{journal}{Nature} \textbf{\bibinfo{volume}{439}},
539:   \bibinfo{pages}{828} (\bibinfo{year}{2006}).
540: 
541: \bibitem[{\citenamefont{Ostojic and Panja}()}]{ostojic06-2}
542: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Ostojic}} \bibnamefont{and}
543:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Panja}}, \bibinfo{journal}{Phys. Rev. Lett. in press}
544:   , \eprint{cond-mat/0606349}.
545: 
546: \bibitem[{\citenamefont{Stauffer and Aharony}(1991)}]{stauffer:book}
547: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Stauffer}} \bibnamefont{and}
548:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Aharony}},
549:   \emph{\bibinfo{title}{Introduction to Percolation Theory}}
550:   (\bibinfo{publisher}{Taylor \& Francis}, \bibinfo{year}{1991}).
551: 
552: \bibitem[{\citenamefont{Privman}(1990)}]{privman:book}
553: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Privman}},
554:   \emph{\bibinfo{title}{Finite Size Scaling and Numerical Simulations of
555:   Statistical Physics}} (\bibinfo{publisher}{World Scientific},
556:   \bibinfo{year}{1990}).
557: 
558: \bibitem[{\citenamefont{Binder and Wang}(1989)}]{binder89}
559: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Binder}} \bibnamefont{and}
560:   \bibinfo{author}{\bibfnamefont{J.-S.} \bibnamefont{Wang}},
561:   \bibinfo{journal}{J. Stat. Phys.} \textbf{\bibinfo{volume}{55}},
562:   \bibinfo{pages}{87} (\bibinfo{year}{1989}).
563: 
564: \end{thebibliography}
565: 
566: \end{document}
567: