1: %--------------------------------------------------------------------
2: % Manuscript
3: % '`Jamming under tension in polymer crazes''
4: % by J. Rottler and M.O. Robbins
5: %--------------------------------------------------------------------
6: %\documentstyle[preprint,amssymb,epsfig,prl,aps]{revtex}
7: \documentstyle[amssymb,multicol,epsfig,prl,aps]{revtex}
8:
9: \begin{document}
10: \draft
11:
12: \title{Jamming under tension in polymer crazes}
13:
14: \author{J\"org Rottler and Mark~O.~Robbins}
15:
16: \address{Department of Physics and Astronomy, The Johns Hopkins
17: University, 3400 N.~Charles Street, Baltimore, Maryland 21218}
18:
19: \date{\today}
20:
21: \maketitle
22:
23: \begin{abstract}
24: Molecular dynamics simulations are used to study a unique expanded
25: jammed state. Tension transforms many glassy polymers from a dense
26: glass to a network of fibrils and voids called a craze. Entanglements
27: between polymers and interchain friction jam the system after a fixed
28: increase in volume. As in dense jammed systems, the distribution of
29: forces is exponential, but they are tensile rather than compressive.
30: The broad distribution of forces has important implications for fibril
31: breakdown and the ultimate strength of crazes.
32: \end{abstract}
33:
34: \pacs{PACS numbers: 61.43.Fs, 62.20.Fe, 83.10.Rs}
35:
36: \begin{multicols}{2}
37: \narrowtext
38:
39: The common features of jammed systems ranging from molecular glasses
40: to granular media have sparked great
41: interest \cite{Liu2001,Ohern2001}. These systems jam as the available
42: volume becomes too small to allow relative motion of particles.
43: Unifying features of the jammed state are the presence of an
44: exponential force tail at large {\em compressive} forces and the
45: appearance of a network of forces on scales much larger than the size
46: of the constituent molecules or grains \cite{Liu1995,Coppersmith1996}.
47: The origins of these features are still debated.
48:
49: In this Letter, we consider a qualitatively different jammed state
50: that forms under {\em tension}. Many amorphous polymers expand from a
51: typical dense glass to a ``craze'' consisting of a network of fibrils
52: and voids \cite{Kramer1990}. As shown in Fig. \ref{crazepic-fig}, the
53: dense and expanded jammed states coexist at a fixed tensile stress
54: $S$, much like liquid and gas phases coexist at a fixed pressure.
55: There are no covalent bonds that keep the craze from unravelling, but
56: experiments suggest that the topological constraints called
57: entanglements that limit dynamics in polymer melts behave like
58: crosslinking bonds \cite{Kramer1990}. These experiments cannot address
59: how deformation occurs, the distribution of forces within the system,
60: or the configurations of individual chains within the intricate craze
61: structure. We describe extensive molecular dynamics simulations that
62: address all of these issues. We find that the distribution of tensile
63: forces in the craze has an exponential tail analogous to that observed
64: for compressive forces in dense jammed systems. The polymer undergoes
65: an approximately affine displacement as it deforms into the craze.
66: Expansion stops when segments that are only 1/3 of the distance
67: between entanglements are pulled taut. This factor can be understood
68: from simple geometric arguments and the assumption that entanglements
69: act like chemical crosslinks between chains.
70:
71: We study a standard coarse-grained model \cite{Puetz2000}, where each
72: linear polymer is modeled by $N$ beads of mass $m$. Van der Waals
73: interactions are modeled with a standard 6-12 Lennard-Jones (LJ)
74: potential with energy scale $\epsilon$ and length scale $\sigma$. A
75: simple analytic potential \cite{bondpot-comm,Sides2001} is used for
76: covalent bonds, with equilibrium length $l_0 = 0.96\sigma$ between
77: adjacent beads along the chain. In this fully flexible (fl) model, the
78: number of beads between entanglements (entanglement length) is
79: $N_e^{\rm fl} \approx 70$ \cite{Puetz2000}. In order to analyze
80: entanglement effects, we also consider semiflexible chains (sfl) where
81: an additional bond-bending potential stiffens the chain locally and
82: reduces the entanglement length to $N_e^{\rm sfl} \approx 30$
83: beads \cite{Sides2001,Faller2000,anglepot-comm}. We consider two
84: temperatures $T=0.1\epsilon/k_B$ and $T=0.3\epsilon/k_B$, where the
85: latter is close to the glass transition temperature. The strength of
86: adhesive interactions between beads is varied by truncating the LJ
87: force at either $r_c=1.5\sigma$ or $r_c=2.2\sigma$ \cite{Baljon2001}.
88: %****************************************************************
89: \begin{figure}
90: \epsfig{file=fig1sm.ps,width=8cm}
91: \vspace{0.1cm}
92: \caption{Close-ups of interface between dense polymer and craze for
93: (a) $T=0.1 \epsilon/k_B$, $r_c = 1.5\sigma$ (fl) and (b) $T=0.3
94: \epsilon/k_B$, $r_c = 2.2\sigma$ (sfl). The lateral dimension of both
95: panels is $64\sigma$. The diameter and spacing of the fibrils increase
96: as $T$ and/or $r_c$ increase. }
97: \label{crazepic-fig}
98: \end{figure}
99: %****************************************************************
100: An amorphous glassy state is created in a cubic simulation cell of
101: side $L$ using standard techniques \cite{Puetz2000}. The period in the
102: $z$-direction is then increased at a small constant velocity
103: \cite{Baljon2001}. After an initial nucleation period, the system
104: separates into two phases (Fig.~\ref{crazepic-fig}). The craze
105: network grows at constant tensile stress $S$ through plastic flow in a
106: narrow interfacial region called the ``active zone.'' The volume
107: occupied by the polymer increases by an ``extension ratio'' $\lambda$
108: that is remarkably insensitive to all parameters other than $N_e$.
109: For example, increasing the temperature and interaction range leads to
110: much larger fibrils in Fig.~\ref{crazepic-fig}(b) than (a), yet there
111: is no measurable change in $\lambda$. From the ratios of the initial
112: and final densities we obtain $\lambda_{\rm fl}=6.0 \pm 0.6$ and
113: $\lambda_{\rm sfl}=3.5 \pm 0.3$, independent of $N$, $T$, and $r_c$.
114: Experimental values of $\lambda$ range from about 2 to 7
115: \cite{Kramer1990}.
116:
117: Although strain rate is strongly localized at the instantaneous position of
118: the active zone, the net effect is a nearly uniform or affine
119: expansion of each region of the polymer. As shown in Figure
120: \ref{affinedisp-fig}(a), the final height $z_f$ of beads with an
121: initial height of $z_i$ is very close to $\lambda z_i$. The rms
122: variation in the final height (half the errorbar length), has a
123: relatively small constant value. For reasons discussed below, it is
124: close to $N_e/3$ for both flexible and semiflexible chains at all $T$
125: and $r_c$ studied. There are also much smaller rms displacements in
126: the $x-y$ plane. These allow chains to lower their free energy by
127: gathering into fibrils whose local density is close to that of the
128: dense glass. The magnitude of lateral displacements varies with the
129: size and spacing of fibrils, which depends on both $T$ and $r_c$. For
130: example, the lateral displacement increases from $\sim 2.5\sigma$ in
131: Fig.~\ref{crazepic-fig}(a) to $\sim 5.6\sigma$ in
132: Fig.~\ref{crazepic-fig}(b).
133:
134: The affine nature of the deformation along $z$ can also be inferred from the
135: change in the conformation of individual polymers. In the initial
136: state, chains exhibit an ideal random walk (RW) structure inherited
137: from the melt. The rms distance $\Delta r(\Delta N)$ between beads
138: that are $\Delta N$ neighbors apart scales as $\Delta r^2 =
139: l_pl_0\Delta N$, which defines the persistence length $l_p$. In the
140: dense glass, the mean-squared projection along each of the three axes
141: is equal. An affine deformation by $\lambda$ along z will only change
142: the displacements along the $z$-axis, yielding $\langle\Delta
143: z^2\rangle=\lambda^2l_pl_0\Delta N/3$.
144: Fig.~\ref{affinedisp-fig}(b) shows that the final conformation of
145: chains follows this expression at large scales. The in-plane
146: components of the displacement are not changed during crazing.
147: %****************************************************************
148: \begin{figure}[bt]
149: \epsfig{file=fig2sm.ps,width=8cm}
150: \vspace{0.1cm}
151: \caption{(a) Final vs. initial height for flexible (large slope) and
152: semiflexible (small slope) chains ($T=0.1\,\epsilon/k_B$,
153: $r_c=1.5\sigma$). Averages were performed over z-intervals of width
154: $\sigma$. Straight lines have slope $\lambda_{\rm fl}=5.9$ and
155: $\lambda_{\rm sfl}=3.5$, respectively. Error bars show the standard
156: deviation from the averages at each height and are $17\pm
157: 1\sigma$ (fl) and $9\pm 1\sigma$ (sfl). (b) Height change
158: $\Delta z$ as a function of distance $\Delta N$ from the chain
159: center. Straight lines have slope $\lambda^2l_pl_0/3$ with
160: $\lambda$ from (a). Systems at different $T,r_c$ and $N$ show
161: the same results.}
162: \label{affinedisp-fig}
163: \end{figure}
164: %****************************************************************
165: The expansion along z pulls short segments of the chains taut, so that
166: $\Delta z^2$ rises quadratically at small $\Delta N$ in
167: Fig.~\ref{affinedisp-fig}(b). The typical number of beads in a taut
168: segment, $\tilde{N}_{\rm st}$, can be obtained from the intersection
169: between the two asymptotic scaling forms: $(\tilde{N}_{\rm
170: st}l_0)^2=\lambda^2l_pl_0\tilde{N}_{\rm st}/3$, yielding
171: $\tilde{N}_{\rm st}=\lambda^2l_p/3l_0$. Inserting the observed values
172: of $\lambda$ and $l_p$ ($l_p^{\rm fl}=1.65\sigma$ and $l_p^{\rm
173: sfl}=2.7\sigma$), we arrive at $\tilde{N}_{\rm st}^{\rm fl}\simeq
174: 21\pm 4$ and $\tilde{N}_{\rm st} ^{\rm sfl}\simeq 12\pm 2$,
175: respectively. The length of taut sections can also be determined by
176: direct analysis of the chain geometry.
177: Fig.~\ref{straightsegments-fig}(a) shows the distribution $P(N_{\rm
178: st})$ of the number $N_{\rm st}$ of successive beads whose
179: displacement continues upwards or downwards. Here a bond is
180: considered up (down) if it is within 45$^\circ$ of the +z (-z) axis.
181: For both flexible and semiflexible chains, $P(N_{\rm st})$ has an
182: exponential tail with a characteristic decay length that, like
183: $\lambda$, is independent of $N$, $T$, and $r_c$. The decay lengths,
184: $\tilde{N}_{\rm st}^{\rm fl}\sim 24\pm 3$ and $\tilde{N}_{\rm st}^{\rm
185: sfl}\sim 13\pm 2$, are in good agreement with the prediction from RW
186: statistics. Essentially the same length scales appear in the decay of
187: the correlation function for the $z$-component of successive bonds,
188: Fig.~\ref{straightsegments-fig}(b), or the bond-angle correlation
189: function (not shown).
190:
191: It is not surprising that all the above lengths are comparable to the
192: deviations from a purely affine deformation in
193: Fig.~\ref{affinedisp-fig}(a), since segments of length $\tilde{N}_{\rm
194: st}$ are stretched taut from their initial RW configuration. However,
195: a very successful expression for $\lambda$ is usually derived by
196: assuming that the taut segments contain $N_e$ beads rather than $N_e/3$
197: \cite{Kramer1990}.
198: The maximum extension ratio, $\lambda_{\rm max}$, is defined
199: as the ratio between the fully stretched length $N_e l_0$
200: and the initial end-to-end distance of a RW of $N_e$ steps
201: \begin{equation}
202: \lambda_{\rm max} \equiv N_e l_0/ (l_0 l_p N_e)^{1/2} = \sqrt{l_0N_e/l_p} .
203: \label{eq:lambda}
204: \end{equation}
205: Calculated values of $\lambda_{\rm max}$ are very close to values
206: of $\lambda$ measured in experiments and in our simulations where
207: $\lambda_{\rm max}=6.5\pm 0.5$ and $3.5\pm0.3$ for flexible and
208: semiflexible chains, respectively.
209: However, it was noted in early work that fully stretched chains
210: would actually yield an even larger extension ratio
211: because the initial end-to-end
212: vector of segments of length $N_e$ is randomly oriented \cite{Kramer1983}.
213: Since the
214: mean-squared projection along any direction is only 1/3 of the total,
215: $\lambda$ would be $\sqrt{3}\lambda_{\rm max}$ for fully stretched chains.
216: The observation that
217: $\lambda \approx \lambda_{\rm max}$ implies that
218: the average length of stretched segments is only $N_e/3$, as we have shown.
219: However those segments that are initially along the $z$-axis
220: are fully stretched when $\lambda=\lambda_{\rm max}$,
221: and it appears that these few segments are sufficient to prevent
222: further elongation. Thus the factor of 1/3 results from the random
223: nature of the entangled network. Note that a recent study of chains
224: with random crosslinks \cite{Barsky2000} is consistent with
225: Eq.~(\ref{eq:lambda}), while all segments between crosslinks were
226: fully stretched in a similar study of ordered networks
227: \cite{Stevens2001}.
228:
229: %****************************************************************
230: \begin{figure}[bt]
231: \epsfig{file=fig3sm.ps,width=8cm}
232: \vspace{0.1cm}
233: \caption{(a) Probability of straight segments of length $N_{\rm st}$
234: for flexible and semiflexible chains. Thick lines:
235: $T=0.1\,\epsilon/k_B$, $r_c=1.5\sigma, N=512$ and 1048576
236: beads. Dotted lines: $T=0.3\epsilon/k_B$, $r_c=2.2\sigma, N=512$ and
237: 262144 beads. Long dashed lines: $T=0.1\epsilon/k_B$, $r_c=1.5\sigma ,
238: N=256$ and 262144 beads. (b) correlation function for the z-components
239: of the bonds for the same systems. Thin solid lines in both panels
240: show fits to an exponential decay with indicated decay lengths.}
241:
242: \label{straightsegments-fig}
243: \end{figure}
244: %****************************************************************
245: A crosslinked system will jam when there is a network of fully
246: stretched covalent bonds, but it is less clear how entanglements can
247: lead to a jammed network. The entanglement length is typically
248: determined from the response of a polymer melt to a sudden strain
249: \cite{Puetz2000,Doi1986}. The resulting time-dependent stress shows a
250: broad plateau where the polymer has relaxed on short length scales,
251: but is unable to relax at larger length scales because
252: interpenetrating loops of polymer cannot pass through each other. The
253: entanglement length corresponds to the typical length along the
254: backbone between these constraints. The stress ultimately decays by
255: the slow diffusion of polymers along their length until their ends
256: pass through the loops and release all remaining constraints.
257: Expansion of the polymer into a craze is also limited by the inability
258: of polymer loops to pass through each other and interchain friction
259: prevents the chains from unraveling by the snake-like motion that
260: allows diffusion in the melt. The tension pulling two interpenetrating
261: loops in opposite directions creates large compressive forces
262: at the entanglements. As in conventional jammed systems this increases
263: the barrier for sliding.
264: Long chains have many interpenetrating loops, and the tensions pulling in
265: each direction nearly cancel. Chains with $N<2N_e$ disentangle rather
266: than forming a craze \cite{Baljon2001}.
267:
268: The distance between entanglements is determined by the chain
269: statistics \cite{Fetters1994} which are inherited from the melt.
270: They are not sensitive
271: to temperature because there is no evolution of large scale structure
272: in the glassy phase, and they are relatively insensitive to the
273: strength of adhesive interactions and density. Thus $\lambda$ is
274: nearly constant, while the fibril structure at scales less than $N_e$
275: may vary dramatically (Fig. \ref{crazepic-fig}). We have verified
276: that $\lambda$ only depends on chain statistics by artificially
277: varying the persistence length $l_p$ in the glassy state.
278: As expected, the value of $\lambda$ depends only on $l_p$ and not
279: on the flexibility of the chains, $r_c$, or other details of the
280: interaction potential.
281:
282: As in other jammed systems, the distribution of forces in the craze
283: follows a universal curve with an exponential tail
284: (Fig.~\ref{globaltension-fig}). The tensions $f$ in the covalent bonds
285: along the chains carry most of the stress. Only 10-20\% of covalent
286: bonds are under compression ($f<0$), and this part of the distribution
287: does not change much during crazing. The tensile ($f>0$) part of the
288: distribution is exponential, and results for all systems collapse
289: after normalizing by the average $\langle f \rangle$ over the tensile
290: region. Intuitively, one might expect the crazes with the larger
291: extension ratio (and $N_e$) to exhibit a larger average tension than
292: the less stretched crazes. However, we find that $\langle f \rangle$
293: is a function of temperature $T$ and cutoff range $r_c$ only (see
294: caption) and does not depend on $N_e$. Expanding the craze elastically
295: beyond $\lambda$ increases $\langle f \rangle$, but the normalized
296: probability is unchanged until bonds begin to break.
297: %****************************************************************
298: \begin{figure}[hbt]
299: \epsfig{file=fig4sm.ps,width=8cm}
300: \vspace*{0.1cm}
301: \caption{Distribution of tension $f$ in the craze for the systems of
302: Fig.~\ref{straightsegments-fig}. The straight line is
303: $\exp{(-f/\langle f \rangle})$. Only bonds under positive tension
304: were included in calculating $\langle f \rangle$.
305: $\langle f \rangle\approx 8.2
306: \epsilon/\sigma$ for $T=0.3\epsilon/k_B$, $r_c=2.2\sigma$, $\langle
307: f\rangle\approx 7.2\epsilon/\sigma$ in all other cases.}
308: \label{globaltension-fig}
309: \end{figure}
310: %****************************************************************
311:
312: In equilibrium, the energy of any region follows the exponential
313: Boltzmann distribution because this maximizes the number of
314: microstates available to the entire system at fixed total energy. In
315: jammed systems the total force is conserved, and one may imagine that
316: the exponential distribution of local forces also arises because it
317: maximizes the number of microstates. Two more detailed models for the
318: exponential distribution have been suggested for dense jammed systems.
319:
320: The ``q-model'' \cite{Coppersmith1996} and its generalizations
321: \cite{Claudin1998} assume that stress propagates unevenly through the
322: system. The fraction of stress passed to each neighbor, $q$, is
323: chosen at random from a distribution $R(q)$. For a wide range of $R$
324: one obtains an exponential force distribution. In addition, the
325: stress is concentrated in force chains like those seen in experiments
326: on granular systems. The fibril network in a craze naturally provides
327: a branching path for transmission of stress. Force is conserved along
328: each fibril and then redistributed at the nodes where fibrils merge or
329: split. Study of the stress along chains shows that tension is
330: correlated along straight segments and then changes when the chain
331: changes direction at a node. A more detailed comparison to the q-model
332: is in progress.
333:
334: Recently, O'Hern {\it et al.} \cite{Ohern2001} have proposed an
335: alternative explanation for the exponential force tail in jammed
336: materials. In equilibrium, the separation of beads at small distances
337: is weighted by a Boltzmann-factor $\exp[-V(r)/k_BT]$, where $V(r)$ is
338: the interaction potential. The corresponding distribution of forces
339: is also nearly exponential if the force varies rapidly with $r$. The
340: repulsive part of the LJ potential satisfies this criterion, but our
341: covalent bond potential does not. We have calculated the distribution
342: of bond energies in the craze and find that it is not exponential.
343:
344: The presence of an exponential stress distribution has consequences
345: for the ultimate fracture of craze fibrils. In a minimal model often
346: employed in the literature \cite{Jones1999}, the maximum breaking
347: force $F_{\rm max}$ of a fibril composed of $n$ strands is estimated
348: by assuming that each strand carries an equal share of the load and
349: all covalent bonds break at a critical force $f_c$. This implies
350: $F_{\rm max}/f_c=n$, and all strands break at the same time. If the
351: distribution of forces among strands is exponential, the most stressed
352: strand will break at a lower $F_{\rm max}$. The force needed to
353: initiate failure can be estimated using a simple scaling argument.
354: The first strand will break when
355: $nP(f>f_c)=n\int_{f_c}^{\infty}1/\langle f\rangle \exp{(-f/\langle f
356: \rangle)df} =n\exp(-nf_c/F_{\rm max})=1$, where we have used $\langle
357: f\rangle=F_{\rm max}/n$. This implies a breaking force of $F_{\rm
358: max}/f_c=n/\ln (n)$ instead of $n$. The redistribution of load after a
359: strand breaks will cause the remaining strands to break sequentially.
360: Note that no chain scission occurs in the simulations presented here,
361: but it does occur upon further straining of the craze. A detailed
362: analysis of fibril breakdown will be presented elsewhere.
363:
364: In summary, we have shown that crazing transforms a conventional dense
365: jammed state into a unique expanded jammed state. Expansion is limited
366: by the same interlocking of polymer loops that leads to entanglements
367: in polymer melts. However, the average length of straight segments is
368: only $N_e/3$ due to the random nature of the network. Interchain
369: friction prevents disentanglement of the loops. As in dense jammed
370: systems, the distribution of forces is exponential. However, the
371: forces are tensile rather than compressive. We hope that contrasting
372: this new jammed state with conventional ones will help unravel the
373: microscopic origin of their common features.
374:
375: This work was supported by the Semiconductor Research Corporation
376: (SRC) and by the NSF grant No. DMR0083286. We thank
377: E.~J.~Kramer,H.~R.~Brown, and C.~Denniston for useful discussions.
378:
379: \begin{references}
380: \bibitem{Liu2001} A.~J.~Liu and S.~R.~Nagel (Eds.), {\it
381: Jamming and Rheology}, (Taylor \& Francis, London, 2001)
382: \bibitem{Ohern2001} C.~S.~O'Hern, S.~A.~Langer, A.~J.~Liu,
383: and S.~R.~Nagel, Phys.~Rev.~Lett. {\bf 86}, 111 (2001)
384: \bibitem{Liu1995} C.-h.~Liu, S.~R.~Nagel,
385: D.~A.~Schecter, S.~N.~Coppersmith, S.~Majumdar,
386: O.~Narayan, T.~A.~Witten, Science {\bf 269}, 513 (1995)
387: \bibitem{Coppersmith1996} S.~N.~Coppersmith, C.-h. Liu, S.~Majumdar,
388: O.~Narayan, and T.~A.~Witten, Phys.~Rev.~E {\bf 53}, 4673 (1996)
389: \bibitem{Kramer1990} E.~J.~Kramer, L.~L.~Berger, Adv. Polymer
390: Science {\bf 91/92},1 (1990).
391: \bibitem{Puetz2000} M.~P\"utz, K.~Kremer, G.~S.~Grest,
392: Europhys.~Lett. {\bf 49}, 735 (2000). The quoted $N_e$ is from the
393: plateau modulus.
394: \bibitem{bondpot-comm} The potential is $V_{\rm
395: br}(r)=-k_1(r-r_c)^3(r-R_1)$, and the constants $k_1$ and $R_1$ are
396: adjusted to fit the equilibrium bond length and to allow for bond
397: breaking when the tension exceeds 100 times the breaking force $f_{\rm
398: LJ}$ of the van der Waals bonds.
399: \bibitem{Sides2001} S.~W.~Sides, G.~S.~Grest, M.~J.~Stevens,
400: Phys.~Rev. E {\bf 64}, 050802 (2001), Macromolecules {\bf 35}, 566
401: (2002).
402: \bibitem{Faller2000} R.~Faller, F.~M\"uller-Plathe, and A.~Heuer,
403: Macromolecules {\bf 33}, 6602 (2000) and R.~Faller and
404: F.~M\"uller-Plathe, ChemPhysChem. {\bf 2 }, 180 (2001).
405: \bibitem{anglepot-comm} The bond-bending forces are modeled with
406: $V_{B}=\sum_{i=2}^{N-1}b\left(1-\frac{(\vec{r}_{i-1}-\vec{r}_{i})\cdot
407: (\vec{r}_{i}-\vec{r}_{i+1})}{|(\vec{r}_{i-1}-\vec{r}_{i})||(\vec{r}_{i}-
408: \vec{r}_{i+1})|}\right)$, where $\vec{r}_{i}$ denotes the position of
409: the $i$th bead along the chain, and $b$ characterizes the stiffness.
410: \bibitem{Baljon2001} A.~R.~C.~Baljon, M.~O.~Robbins, Macromolecules
411: {\bf 34}, 4200 (2001).
412: \bibitem{Kramer1983} E. J. Kramer, Adv. Polymer Sci. {\bf 52/53} 1 (1983).
413: \bibitem{Barsky2000} S.~Barsky and M.~O.~Robbins, unpublished
414: \bibitem{Stevens2001} M.~J.~Stevens, Macromolecules {\bf 34}, 1411
415: (2001)
416: \bibitem{Doi1986} M.~Doi and S.~F.~Edwards, {\it The Theory of Polymer
417: Dynamics}, (Oxford University Press, Oxford, 1986)
418: \bibitem{Fetters1994} L.~J.~Fetters, D.~J.~Lohse, D.~Richter,
419: T.~A.~Witten, and A.~Zirkel, Macromolecules {\bf 27}, 4639 (1994)
420: \bibitem{Claudin1998} P.~Claudin, J.-P.~Bouchaud, M.~E.~Cates, and
421: J.~P.~Wittmer, Phys.~Rev.~E {\bf 57}, 4441 (1998)
422: \bibitem{Jones1999} R.~A.~L.~Jones, R.~W.~Richards, {\it Polymers at
423: Surfaces and Interfaces}, (Cambridge University Press, Cambridge,
424: 1999).
425:
426: \end{references}
427: \end{multicols}
428: \end{document}
429: