cond-mat0107433/pre.tex
1: \documentclass[aps,pre,showpacs,twoside,amsfonts,twocolumn]{revtex4}
2:   
3: \usepackage{graphicx}
4: \usepackage{bm}
5: 
6: \newcommand{\e}{\mathrm{e}}
7: \renewcommand{\d}{\mathrm{d}}
8: 
9: 
10: \hyphenation{cross-links cross-linked mo-no-mer mo-no-mers}
11: 
12: 
13: 
14: 
15: \begin{document}
16: 
17: 
18: \title{Normal stresses at the gelation transition}
19: 
20: \author{Kurt Broderix}
21: \thanks{Deceased} 
22: 
23: \author{Peter M\"uller}
24: %\email{mueller@theorie.physik.uni-goettingen.de} 
25: 
26: \author{Annette Zippelius}
27: %\email{annette@theorie.physik.uni-goettingen.de}
28: 
29: \affiliation{Institut f\"ur Theoretische Physik,
30:   Georg-August-Universit\"at, D--37073 G\"ottingen, Germany}
31: 
32: \date{\today}
33: 
34: \begin{abstract}
35:   A simple Rouse-type model, generalised to incorporate the effects of
36:   chemical crosslinks, is used to obtain a theoretical prediction for
37:   the critical behaviour of the normal-stress coefficients $\Psi_{1}$
38:   and $\Psi_{2}$ at the gelation transition. While the exact
39:   calculation shows $\Psi_{2}\equiv 0$, a typical result for these
40:   types of models, an additional scaling ansatz is used to demonstrate
41:   that $\Psi_{1}$ diverges with a critical exponent $\ell = k+z$.
42:   Here, $k$ denotes the critical exponent of the shear viscosity and
43:   $z$ the exponent governing the divergence of the time scale in the
44:   Kohlrausch decay of the shear-stress relaxation function.  For
45:   crosslinks distributed according to mean-field percolation, this
46:   scaling relation yields $\ell =3$, in a accordance with an exact
47:   expression for the first normal-stress coefficient based on a
48:   replica calculation.  Alternatively, using three-dimensional
49:   percolation for the crosslink ensemble we find the value $\ell
50:   \approx 4.9$. Results on time-dependent normal-stress response are
51:   also presented.
52: \end{abstract}
53: 
54: 
55: \pacs{64.60.Ht, % Dynamic critical phenomena
56:   61.25.Hq,     %Macromolecular and polymer solutions; polymer melts; swelling
57:   61.20.Lc      %Time-dependent properties; relaxation 
58:   }
59: 
60: 
61: \maketitle
62: 
63: \section{Introduction}
64: 
65: Chemical gelation is the process of randomly introducing crosslinks
66: between the constitutents of a (macro-) molecular fluid. One way to
67: investigate the effects of the crosslinks on the fluid dynamics
68: consists in measuring the stresses the crosslinked fluid builds up
69: when subjected to a simple shear flow.  For an incompressible,
70: isotropic fluid one can experimentally access \cite{BiAr87} three
71: independent components of the stress tensor $\boldsymbol{\sigma}$: the shear
72: stress $\sigma_{xy}$ and the first and second normal stress
73: differences $\sigma_{xx}-\sigma_{yy}$ and $\sigma_{yy}-\sigma_{zz}$.
74: For static shear flows these give rise to three independent material
75: functions: the shear viscosity $\eta$ and the first and second
76: normal-stress coefficients $\Psi_{1}$ and $\Psi_{2}$. Generally
77: speaking, both Newtonian and non-Newtonian fluids possess a
78: non-vanishing shear viscosity. But, whereas for a Newtonian fluid both
79: $\Psi_{1}$ and $\Psi_{2}$ are always zero, it is precisely the
80: non-vanishing of $\Psi_{1}$ that explains a
81: number of characteristic effects known for e.g.\ polymeric liquids
82: \cite{Wei47}, see also \S{}2.3 in \cite{BiAr87}. On
83: the other hand, even for non-Newtonian fluids $\Psi_{2}$ is typically
84: found to be very small as compared to $\Psi_{1}$, and the
85: ``Weissenberg hypothesis'', $\Psi_{2}=0$, is a good approximation in
86: these cases \cite{Wei47}. It also seems that $\Psi_{2}$ is not as well
87: investigated experimentally as $\Psi_{1}$.
88: 
89: In the context of gelation one is particularly interested in the
90: dependence of these stresses on the crosslink concentration $c$.
91: Universal critical behaviour is expected to occur at the gelation
92: transition, that is, at the critical concentration
93: $c_{\mathrm{crit}}$, where the fluid (sol) undergoes a sudden phase change
94: into an amorphous solid state (gel). As far as shear stress is concerned,
95: there exist numerous experimental investigations 
96: on the static shear viscosity and on the time-dependent
97: shear-stress relaxation function. 
98: The experimentally measured values for the critical exponent $k$,
99: which governs the algebraic divergence of the shear viscosity when
100: approaching $c_{\mathrm{crit}}$ from the sol side, scatter
101: considerably and are found in the range $k\approx 0.6
102: \ldots 1.7$, see
103: e.g.\ \cite{AdDe81,MaWi88,MaAd88,DeBo93,CoGi93,ToFa01}. The origin of
104: this wide spreading is controversially  
105: discussed and eventually unclear. From a theoretical point of view
106: there exists a bunch of competing and partially contradicting scaling
107: relations which express $k$ in terms of percolation exponents. Each of
108: them relies on heuristic 
109: arguments whose validity is mostly unclear. We refer the reader to
110: \cite{BrLo99,BrLo01} for a summary and references. Here we only
111: mention the scaling relation $k=2\nu -\beta$ which was first proposed
112: by de~Gennes \cite{Gen78} and rederived by many others. Erroneously,
113: it is generally referred to as the ``Rouse expression'' for 
114: the viscosity exponent. Here, $\nu$ is the exponent governing the
115: divergence of the correlation length and $\beta$ is associated to the
116: gel fraction. For three-dimensional bond percolation one would get the value 
117: $\left.\vphantom{\big(}(2\nu-\beta)\right|_{d=3}\approx 1.35$.
118: Recently, the viscosity was \emph{exactly} determined within the Rouse
119: model for gelation in \cite{BrLo99,BrLo01}. The analysis disproves the
120: above result and shows that 
121: \begin{equation}
122:   \label{phibeta}
123:   k= \phi -\beta
124: \end{equation}
125: is the true scaling relation valid for Rouse dynamics. Here, $\phi$
126: denotes the first crossover exponent of a
127: corresponding random resistor network \cite{HaLu87,StJaOe99}. 
128: When inserting \cite{BrLo99,BrLo01} high-precision data for $\phi$ and
129: $\beta$ obtained from three-dimensional bond
130: percolation, the true Rouse value of the viscosity exponent turns out to be
131: $\left.\vphantom{\big(}(\phi-\beta)\right|_{d=3}\approx 0.71$ 
132: and agrees with simulations \cite{VePl01} on a similar model.
133: The discrepancy to de Gennes' result above can be attributed to the
134: neglect of the \emph{multi}fractal nature of percolation clusters in
135: \cite{Gen78}. Amazingly, the true Rouse value 
136: $\left.\vphantom{\big(}k\right|_{d=3}\approx 0.71$
137: differs only little from that of another proposal, $k=s$, by de Gennes
138: \cite{Gen79}, where he alluded to an analogy to the conductivity exponent    
139: $\left.\vphantom{\big(}s\right|_{d=3}\approx 0.73$ of an electrical
140: network consisting of a random mixture of superconductors and normal
141: conductors. This close agreement, however, seems to be coincidental. 
142: 
143: In contrast, the authors are not aware of any experimental or
144: theoretical studies concerning the dependence on the crosslink
145: concentration $c$ of normal stresses near the gelation transition.
146: This seems all the more surprising since there exist many
147: experiments \cite{LoMe73,VeWi90,OsIn94,VeKa94} on both the shear-rate
148: dependence of 
149: normal stresses in entangled or (temporarily) crosslinked polymeric
150: liquids in order to explain shear-thinning or shear-thickening
151: phenomena and on the time dependence of the normal-stress response to
152: particular shapes of shear strain. Theoretical explanations of these
153: experimental findings mainly rely on the analysis of transient network
154: models, see e.g.\ \cite{TaEd92,AhOs94,AhOs95,BrHo95}.
155: 
156: Even though Rouse-type models incorporate no other physical
157: interactions between monomers apart from connectivity, they serve as 
158: a standard theoretical reference in terms of which
159: experimental data are frequently interpreted. Therefore it is
160: important to test their predictions as accurately as possible.  It is
161: the purpose of this Paper to use the same generalised Rouse-type model
162: as in \cite{BrLo99,BrLo01,BrAs01} to predict the critical
163: behaviour of the normal-stress coefficients $\Psi_{1}$ and $\Psi_{2}$
164: at the gelation transition. Within this model it will turn out that
165: $\Psi_{2}$ vanishes for all $c$ and that $\Psi_{1}$ diverges with a
166: critical exponent 
167: \begin{equation}
168:   \label{ellka}
169:   \ell = k+z
170: \end{equation}
171: when approaching $c_{\mathrm{crit}}$
172: from the sol side. Here, $z$ denotes the exponent governing the divergence
173: of the time scale in the Kohlrausch decay of the shear-stress
174: relaxation function.  For crosslinks distributed according to
175: mean-field percolation (also called ``classical theory''), this
176: scaling relation yields $\ell =3$, in a 
177: accordance with an exact expression for $\Psi_{1}$ based on a replica
178: calculation. Alternatively (and more realistically), using
179: three-dimensional percolation for the crosslink ensemble we find the
180: value $\ell \approx 4.9$.  Thus, the model predicts a much more
181: pronounced divergence of $\Psi_{1}$ as compared to $\eta$ so that
182: $\Psi_{1}$ may serve as a sensitive indicator for the gelation
183: transition. 
184: We also derive results on the time-dependent
185: normal-stress response.  In particular, the Lodge-Meissner
186: rule, see e.g.\ \S{}3.4.e in \cite{BiAr87}, is shown to hold for
187: normal-stress relaxation after a sudden shearing displacement.
188: 
189: We hope that these theoretical investigations motivate corresponding
190: experimental work in order to develop more insight on normal stresses
191: in (near) critical gels.
192: 
193: 
194: 
195: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
196: %
197: \section{Model}
198: %
199: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
200: 
201: We follow a semi-microscopic approach to gelation based on a
202: Rouse-type model for $N$ monomers. The monomers are treated as point
203: particles with positions ${\mathbf{R}}_i(t)$, $i=1,\ldots,N$, in
204: three-dimensional space. The motion of the monomers is constrained by
205: $M$ randomly chosen, harmonic crosslinks which connect the pairs
206: $(i_e,i'_e)$, $e=1,\ldots,M$, of monomers and give rise to the
207: potential energy
208: \begin{equation} \label{poten}
209:   U := \frac{3}{2a^2}\:\sum_{e=1}^M \lambda_e
210:   \bigl( {\mathbf{R}}_{i_e}-{\mathbf{R}}_{i'_e} \bigr)^2
211:   =: \frac{3}{2a^2}\:\sum_{i,j}^N \Gamma_{ij}\,
212:   {\mathbf{R}}_{i}\cdot{\mathbf{R}}_{j} \,.
213: \end{equation}
214: Here, the fixed length $a>0$ models the overall inverse coupling
215: strength, whereas the individual coupling constants $\lambda_{e}$ are
216: chosen at random. Quite often, only the special case $\lambda_{e}=1$
217: has been considered previously. The second equality in (\ref{poten})
218: introduces the random $N\times N$-connectivity matrix, which encodes
219: all properties of a given crosslink realisation.
220: 
221: \begin{figure}[t]
222:   \includegraphics[width=10.5pc]{flow.eps}
223:   \caption{Homogeneous linear shear flow (\protect{\ref{flowfield}})}
224:   \label{shearflow}
225: \end{figure}
226: 
227: Following \cite{DoEd85,BiCu87,SoVi95} we employ a simple
228: relaxational dynamics
229: \begin{equation} \label{eqmotion}
230:  \zeta \left[ \partial_t R^{\alpha}_i(t)
231: - v^{\alpha}_{\mathrm{ext}}\bigl({\mathbf{R}}_i(t),t\bigr)\right] = 
232:   - \frac{\partial U}{\partial R^{\alpha}_i}(t) 
233:   + \xi^{\alpha}_i(t)
234: \end{equation}
235: without inertial term to describe the motion of the monomers in the
236: externally applied velocity field
237: \begin{equation}
238:   \label{flowfield}
239:   v^{\alpha}_{\mathrm{ext}}({\mathbf{r}},t) := \delta_{\alpha x}\,\kappa(t)\, y
240: \end{equation}
241: with a time-dependent shear rate $\kappa(t)$, see also
242: Fig.~\ref{shearflow}.  Here, Greek indices label Cartesian coordinates
243: $x$,$y$ or $z$.  A friction force with friction constant $\zeta$
244: applies if the velocity of a monomer deviates from the externally
245: applied flow field. The crosslinks exert a force $-\partial
246: U/\partial{\mathbf{R}}_i$ on the monomers, in addition to a random,
247: fluctuating thermal-noise force obeying Gaussian statistics with zero
248: mean and covariance 
249: $\langle\xi_{i}^{\alpha}(t)\xi_{j}^{\beta}(t')\rangle = 2\zeta
250: \delta_{\alpha\beta}\delta_{ij}\delta(t-t')$. Note that we have chosen
251: units in which the inverse temperature is equal to one.  Given the
252: shear flow (\ref{flowfield}), the equation of motion (\ref{eqmotion})
253: is linear and can be solved exactly for each realisation of the
254: thermal noise \cite{BrLo01}.
255: 
256: To complete the description of the dynamic model, we have to specify
257: the statistical ensemble which determines the realisations of the
258: crosslinks. We will distinguish two cases:
259: \begin{enumerate}
260: \item[(i)] Mean-field percolation (also called ``classical theory''):
261:   each pair of 
262:   monomers is chosen independently with equal probability $M/N^{2}$,
263:   irrespectively of the monomer positions in space. As a function of
264:   the crosslink concentration $c:=M/N$ , the system undergoes a
265:   percolation transition at a critical concentration
266:   $c_{\mathrm{crit}}= \frac12$.  For $c<c_{\mathrm{crit}}$ there is no
267:   macroscopic cluster, and almost all clusters are trees \cite{ErRe60,Bol98}.
268: \item[(ii)] Three-dimensional bond percolation \cite{StAh94,BuHa96}.
269: \end{enumerate}
270: For either case we assume the random coupling constants $\lambda_{e}$
271: to be distributed independently of the crosslink configuration, as
272: well as independently of each other with the same (smooth) probability
273: distribution $p(\lambda)$. Moreover, sufficiently high inverse moments
274: \begin{equation}
275:   \label{moments}
276:   P_{n}:= \int_{0}^{\infty}\!\d\lambda\, \lambda^{-n} p(\lambda)
277: \end{equation}
278: of $\lambda_{e}$ shall exist.
279: 
280: %\pagebreak[4]
281: 
282: The combined average over crosslink configurations and random coupling
283: constants will be denoted by an overbar
284: $\overline{\phantom{l}{\scriptstyle\bullet}\phantom{l}}$. Using this
285: notation, we implicitly assume that the macroscopic limit
286: $N\to\infty$, $M\to\infty$, $M/N\to c$ is carried out, too.
287: 
288: Before turning to the analysis of the model we would like to comment
289: on the fact that it describes the random crosslinking of single
290: monomers rather than of pre-formed polymers. However, this does not
291: mean that the applicability of the model is limited to the
292: description of random networks built up by polycondensation from small
293: structural units. Indeed, we expect from universality that details at
294: small length scales are irrelevant for the true critical behaviour at the
295: gelation transition so that these results will also hold for random
296: networks built from arbitrary macromolecules, as is the case in
297: vulcanization, for example. This general universality argument was
298: confirmed \cite{BrLo01} by explicit computations of the critical
299: behaviour of the shear viscosity with the mean-field distribution of
300: crosslinks.
301: 
302: 
303: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
304: %
305: \section{Stress tensor and normal-stress coefficients}
306: %
307: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
308: 
309: Due to the externally applied shear flow the crosslinks exert shear
310: stress on the polymer system, whose tensor components are given in
311: terms of a force-position correlation \cite{DoEd85,BiCu87}
312: \begin{equation}
313:   \label{stressdef}
314:   \sigma_{\alpha\beta}(t) = \lim_{t_{0}\to -\infty}
315:   \frac{\rho_{0}}{N}\,\sum_{i=1}^{N} \left\langle\frac{\partial U}{\partial
316:     R_{i}^{\alpha}}(t) \; R_{i}^{\beta}(t)\right\rangle\,.
317: \end{equation}
318: Here, $\rho_{0}$ denotes the density of monomers. In (\ref{stressdef})
319: one has to insert the explicitly known \cite{BrLo01} solutions
320: ${\mathbf{R}}_{i}(t)$ of 
321: the Rouse equation (\ref{eqmotion}) at time $t$ with initial data
322: ${\mathbf{R}}_{i}(t_{0})$ at time $t_{0}$. In order to ensure that the
323: thermal-noise average allows for the description of a possible 
324: stationary state of the system at finite times $t$, the time evolution
325: is chosen to start in the infinite past, $t_{0}\to -\infty$, thereby
326: losing all transient effects which stem from the initial data.
327: %the initial values
328: %of the solutions ${\mathbf{R}}_{i}(t)$ of (\ref{eqmotion}) are to be
329: %taken at more and more distant times $t_{0}$ 
330: %in the past in order to ensure that after averaging over the thermal
331: %noise, the system has reached a steady state at time $t$. 
332: This yields \cite{BrLo01} for the stress tensor
333: \begin{widetext}
334: \begin{equation}
335:   \label{stresstensor}
336:   \boldsymbol{\sigma} (t)  = \chi(0){\mathbf{1}} 
337:   + \int_{-\infty}^{t}\!\!\d t' \, \chi(t-t')\,
338:   \kappa(t') \left(\begin{array}{cc@{\qquad}c}
339:       2\int_{t'}^{t}\!\d s\,\kappa(s) & 1 & 0 \\
340:       1 & 0 & 0\\ 0 & 0 & 0\end{array}\right), 
341: \end{equation}
342: \end{widetext}
343: % \begin{eqnarray}
344: %   \label{stresstensor}
345: %   \boldsymbol{\sigma} (t) \! &=& \!\chi(0){\mathbf{1}} \nonumber \\
346: %   && + \int_{-\infty}^{t}\!\!\d t' \, \chi(t-t')\,
347: %   \kappa(t') \left(\begin{array}{cc@{\qquad}c}
348: %       2\int_{t'}^{t}\!\d s\,\kappa(s) & 1 & 0 \\
349: %       1 & 0 & 0\\ 0 & 0 & 0\end{array}\right), \nonumber\\
350: % \end{eqnarray}
351: where $\mathbf{1}$ denotes the $3\times 3$-unit matrix and the stress
352: relaxation function is given by 
353: \begin{equation}
354: \label{stressrelax}
355: \chi(t):=
356: \frac{\rho_{0}}{N}\; {\mathrm{Tr}} \left[ (1-E_0) \,
357: \exp\left(- \frac{6t}{\zeta a^2} \Gamma\right)\right]\,.
358: \end{equation}
359: The symbol ${\mathrm{Tr}}$ in (\ref{stressrelax}) stands for the trace
360: over $N\times N$-matrices, and $E_0$ denotes the projector on the
361: space of zero eigenvalues of $\Gamma$, which correspond to translations
362: of whole clusters. The associated eigenvectors are constant within each
363: cluster and zero outside \cite{BrGo97,BrLo01}.  Within the simple
364: Rouse model the zero eigenvalues do not contribute to shear relaxation
365: because there is no force acting between different clusters. The only
366: contribution to stress relaxation is due to deformations of the
367: clusters.
368: 
369: For a time-independent shear rate $\kappa(t) \equiv \kappa$ it is
370: customary to define a first and second normal-stress coefficient by
371: \begin{equation}
372:   \Psi_{1} := \frac{\sigma_{xx} -
373:     \sigma_{yy}}{\rho_{0}\,\kappa^{2}}\,,\qquad\quad 
374:   \Psi_{2} := \frac{\sigma_{yy} -
375:     \sigma_{zz}}{\rho_{0}\,\kappa^{2}}\,.
376: \end{equation}
377: One deduces immediately from (\ref{stresstensor}) that
378: \begin{equation}
379:  \Psi_{2}=0\,,
380: \end{equation}
381: a characteristic result for Rouse-type models. In contrast, the first
382: normal-stress coefficient $\Psi_{1}$ is non-zero
383: \begin{equation}
384:   \label{psi1}
385:   \Psi_{1} = \frac{1}{2}\,\Bigl(\frac{\zeta a^{2}}{3}\Bigr)^{2}
386:   \frac{1}{N} \; \mathrm{Tr}\,\Bigl(\frac{1-E_{0}}{\Gamma^{2}}\Bigr)
387: \end{equation}
388: and independent of the shear rate $\kappa$.
389: 
390: For a macroscopic system $\Psi_{1}$ is expected to be a self-averaging
391: quantity. Therefore we will calculate the disorder average
392: $\overline{\phantom{l}{\scriptstyle\bullet}\phantom{l}}$ of
393: (\ref{psi1}) over all crosslink realisations and all crosslink
394: strengths. To do so it is convenient to introduce the averaged density
395: \begin{equation}
396:   \label{dos}
397:   D(\gamma): =\overline{\frac{1}{N}\,\mathrm{Tr}\,
398:     \bigl[(1-E_0)\,\delta(\gamma -\Gamma)\bigr]} 
399: \end{equation}
400: of non-zero eigenvalues of $\Gamma$. Physically, $D$ describes the
401: distribution of relaxation rates in the network (in units of $\zeta
402: a^{2}/6$), as is evident from the representation
403: \begin{equation}
404:  \label{laplace}
405:   \overline{\chi}(t) = \rho_{0} \int_{0}^{\infty}\!\d\gamma\,
406:   \exp\Bigl\{-\frac{6\,t\gamma}{\zeta a^{2}}\Bigr\} D(\gamma)
407: \end{equation}
408: of the disorder average of the stress relaxation function
409: (\ref{stressrelax}).  Various properties of the eigenvalue density $D$
410: are discussed in detail in \cite{BrAs01}. The average
411: $\overline{\Psi}_{1}$ now appears as the second inverse moment of $D$,
412: \begin{equation}
413:   \label{psi1av}
414:   \overline{\Psi}_{1} = \frac{1}{2}\,\Bigl(\frac{\zeta
415:     a^{2}}{3}\Bigr)^{2} \int_{0}^{\infty}\!\d\gamma\,
416:   \frac{D(\gamma)}{\gamma^{2}}\,,
417: \end{equation}
418: while the disorder-averaged
419: static shear viscosity $\overline{\eta} :=
420: \overline{\sigma}_{xy}/(\rho_{0}\,\kappa)$ is determined
421: \cite{BrLo99,BrLo01} by the first inverse moment
422: \begin{equation}
423:   \label{viscosity}
424:   \overline{\eta} =
425:   \frac{1}{\rho_{0}} \int_{0}^{\infty}\!\d t \,\overline{\chi}(t) =
426:   \frac{\zeta a^{2}}{6}
427:   \int_{0}^{\infty}\!\d\gamma\; \frac{D(\gamma)}{\gamma}\,.
428: \end{equation}
429: 
430: At this point one can already see that $\overline{\Psi}_{1}$ serves as
431: a sensitive indicator for the gelation transition. Indeed, the Jensen
432: inequality \cite{Bau96} implies
433: \begin{equation}
434:   \overline{\Psi}_{1} \ge \frac{2\;
435:     \overline{\eta}^{\;2}}{\int_{0}^{\infty}\!\d\gamma\, 
436:     D(\gamma)}
437:   \ge 2 \;\overline{\eta}^{\;2}\,,
438: \end{equation}
439: and hence
440: \begin{equation}
441:   \label{jensen}
442:   \ell \ge 2k
443: \end{equation}
444: with $\ell$, respectively $k$, denoting the critical exponent of
445: $\overline{\Psi}_{1} \sim (c_{\mathrm{crit}} -c)^{-\ell}$,
446: respectively $\overline{\eta}\sim (c_{\mathrm{crit}} -c)^{-k}$. 
447: 
448: In the two following sections we will determine the precise Rouse value of
449: $\ell$ for the two different types of crosslink ensembles described
450: above.
451: 
452: 
453: 
454: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
455: %
456: \section{Mean-field percolation}
457: %
458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459: 
460: \begin{figure}[t]
461:   \hspace*{1.4cm}\includegraphics[width=17pc]{normal-stress.eps}
462:   \caption{First normal-stress coefficient (\protect{\ref{psi1erg}})
463:     in units of $(\zeta a^{2}/3)^{2}$ as a function of $c$ for
464:     $P_{1}=P_{2}=1$.}
465:   \label{normalstressplot}
466: \end{figure}
467: 
468: For mean-field random graphs (i) the second inverse moment of the
469: eigenvalue density $D$ was calculated in Eq.\ (38) of 
470: \cite{BrAs01} with the help of a replica approach. This gives
471: rise to the exact result
472: \begin{widetext}
473:   \begin{eqnarray}
474:     \label{psi1erg}
475:     \overline{\Psi}_{1} = \frac{1}{2}\,\Bigl(\frac{\zeta
476:       a^{2}}{3}\Bigr)^{2} c \left[  - \frac{8c^{3} - 6c^{2} - 5c 
477:         + 1}{30c(1-2c)^{3}}\,P_{1}^{2}    
478:       - \frac{4c^{2} - 3c
479:         -1}{24c(1-2c)^{2}} \, P_{2} 
480:       + \frac{5P_{2}-4P_{1}^{2}}{240c^{2}}\,\ln(1-2c)\right]\,,
481:   \end{eqnarray}
482: \end{widetext}
483: % \begin{eqnarray}
484: %   \label{psi1erg}
485: %   \overline{\Psi}_{1} &=& \frac{1}{2}\,\Bigl(\frac{\zeta
486: %     a^{2}}{3}\Bigr)^{2} c \left[  - \frac{8c^{3} - 6c^{2} - 5c 
487: %       + 1}{30c(1-2c)^{3}}\,P_{1}^{2}     \right.\nonumber\\ 
488: %     & & \left. - \frac{4c^{2} - 3c
489: %       -1}{24c(1-2c)^{2}} \, P_{2} 
490: %     + \frac{5P_{2}-4P_{1}^{2}}{240c^{2}}\,\ln(1-2c)\right]\,,\nonumber\\
491: % \end{eqnarray}
492: which is valid for all $0<c<c_{\mathrm{crit}}=\frac{1}{2}$. The inverse
493: moments $P_{n}$ were defined in (\ref{moments}). From (\ref{psi1erg})
494: we read off the critical divergence
495: \begin{equation}
496:   \overline{\Psi}_{1} \sim \Bigl(\frac{\zeta
497:     a^{2}}{3}\Bigr)^{2}\, \frac{P_{1}^{2}}{240}\; \varepsilon^{-3}\,,
498:   \qquad  \varepsilon := c_{\mathrm{crit}} - c \downarrow 0
499: \end{equation}
500: at the gelation transition, and hence the critical exponent
501: \begin{equation}
502:   \label{mfexp}
503:   \ell =3\,.
504: \end{equation}
505: For $c\to 0$ one expands
506: \begin{equation}
507:   \overline{\Psi}_{1} = \Bigl(\frac{\zeta
508:     a^{2}}{3}\Bigr)^{2}\, \frac{P_{2}}{8} \,c + {\mathcal{O}}(c^{2})\,.
509: \end{equation}
510: Fig.~\ref{normalstressplot} displays $\overline{\Psi}_{1}$ in units of
511: $(\zeta a^{2}/3)^{2}$ as a function of $c$ for the special case
512: $P_{1}=P_{2}=1$.
513: 
514: 
515: % \begin{figure}[t]
516: %   \hspace*{1.4cm}\includegraphics[width=17pc]{normal-stress.eps}
517: %   \caption{First normal-stress coefficient (\protect{\ref{psi1erg}})
518: %     in units of $(\zeta a^{2}/3)^{2}$ as a function of $c$ for
519: %     $P_{1}=P_{2}=1$.}
520: %   \label{normalstressplot}
521: % \end{figure}
522: 
523: It is the merit of the mean-field percolation ensemble that it allows
524: for a variety of exact analytical calculations. However, since the
525: probability for a crosslink to occur does not depend on the monomers'
526: positions in space, this ensemble is believed to provide a fairly
527: unrealistic description for three-dimensional gels. For this reason we
528: consider an alternative crosslink ensemble in the next section, which
529: has been successfully used \cite{StCo82} to explain static properties
530: of polymer systems.
531: 
532: 
533: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
534: %
535: \section{Three-dimensional bond percolation}
536: %
537: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
538: 
539: For this ensemble of crosslinks the second inverse moment
540: (\ref{psi1av}) of the 
541: eigenvalue density $D_{\varepsilon}$ -- note
542: that in this section we emphasize the dependence on $\varepsilon:=
543: c_{\mathrm{crit}}-c$ in the notation of various quantities -- is not
544: known analytically. In order 
545: to proceed we assume that $D_{\varepsilon}$ follows a scaling law
546: %For finite $\gamma$ the numerical data
547: %of \cite{BrAs01} suggest the form 
548: \begin{equation}
549:   \label{dscale}
550:   D_{\varepsilon}(\gamma) \sim \gamma^{\Delta -1}\,
551:   f\bigl(\gamma^{*}(\varepsilon)  /\gamma\bigr)
552: \end{equation}
553: close to the critical point and for small enough $\gamma$. It is
554: determined by a typical relaxation 
555: rate $\gamma^{*}(\varepsilon)\sim \varepsilon^{z}$, which vanishes
556: when approaching the critical point, and a scaling function $f(x)$
557: that tends to a non-zero constant for $x\to 0$ and decays faster than
558: any inverse polynomial for $x\to\infty$.  
559: In particular, this gives the power-law behaviour
560: $D_{\varepsilon=0}(\gamma) \sim \gamma^{\Delta-1}$ asymptotically for 
561: $\gamma\to 0$ at criticality, in agreement with experiments
562: \cite{WiCh86,ChWi87}. The measured exponent values, however, scatter
563: considerably, $\Delta\approx 0.4\ldots 0.8$, and seem to depend on the mass of
564: the crosslinked molecules \cite{WiMo97}. Note that on general grounds
565: the exponent $\Delta$ has to be positive, because otherwise
566: $D_{\varepsilon=0}(\gamma)$ would not be integrable at $\gamma=0$, in
567: contradiction to the definition (\ref{dos}). 
568: The scaling
569: law (\ref{dscale}) yields, via the Laplace transform (\ref{laplace}),
570: the scaling law 
571: \begin{equation}
572:   \overline{\chi}_{\varepsilon}(t) \sim \varepsilon^{z\Delta}\,
573:   g\bigl(t/\tau^{*}(\varepsilon)\bigr) 
574: \end{equation}
575: for the long-time behaviour of the stress relaxation function.
576: % For times $t$ away from zero we infer from Eq.\ (1) in
577: % \cite{BrAs01} 
578: % \begin{equation}
579: %   \overline{\chi}_{\varepsilon}(t) \sim \varepsilon^{z\Delta}\,
580: %   g\bigl(t/\tau^{*}(\varepsilon)\bigr) 
581: % \end{equation}
582: Here, the scaling function obeys $g(x) \sim x^{-\Delta}$ for $x\to 0$
583: and the typical relaxation time $\tau^{*}(\varepsilon) := \zeta a^{2}/[6
584: \gamma^{*}(\varepsilon)]\sim \varepsilon^{-z}$ diverges when
585: approaching the critical point. Precisely at the critical point one
586: finds an algebraic long-time decay $\overline{\chi}_{\varepsilon =0}(t) \sim
587: t^{-\Delta}$. Dynamical scaling relates $\Delta$ to $z$
588: and to the exponent $k$ of the shear viscosity  
589: \begin{equation}
590:   \label{expeq}
591:   \Delta = (z-k)/z\,,
592: \end{equation}
593: see e.g.\ \cite{WiMo97,BrAs01}. For $x\to\infty$ the scaling function $g(x)$
594: has to decay like a stretched exponential in order to accommodate 
595: the experimentally found \cite{MaAd89} Kohlrausch decay 
596: \begin{equation}
597:   \label{kohlrausch}
598:   \overline{\chi}_{\varepsilon >0}(t) \sim
599: \exp\{-[t/\tau^{*}(\varepsilon)]^{\alpha}\}
600: \end{equation}
601: of the stress-relaxation function in the sol phase away from
602: criticality, where $\alpha$ is a non-critical and possibly
603: non-universal exponent. We will return to (\ref{kohlrausch}) in the
604: next section.
605: 
606: From (\ref{psi1av}), (\ref{dscale}) and (\ref{expeq}) we deduce
607: $\overline{\Psi}_{1} \sim\varepsilon^{-\ell}$ for
608: $\varepsilon\downarrow 0$ with an exponent given by the scaling
609: relation
610: \begin{equation}
611:   \label{scalerel}
612:   \ell = k + z = k\, \frac{2-\Delta}{1-\Delta}\,. 
613: \end{equation}
614: Since $\Delta>0$, we have $z>k$ and the scaling relation
615: (\ref{scalerel}) is compatible with the inequality (\ref{jensen}).
616: Eq.\ (\ref{scalerel}) was obtained previously in \cite{WiIz94} from a
617: model density of relaxation times with a sharp upper cut-off. 
618: 
619: According to (\ref{phibeta}) the viscosity exponent for the Rouse-type
620: model under consideration is given by $k\approx 0.71$, when using
621: three-dimensional bond percolation to generate the crosslink ensemble.
622: Concerning $\Delta$, we are only aware of \cite{BrAs01},
623: where this exponent is determined for the Rouse model at hand without
624: any further assumptions. It was done by
625: numerical computations of the eigenvalue density (\ref{dos}) and yields
626: $\Delta\approx 0.83$. But,  
627: as compared to $k$, we suspect the numerical accuracy of the result to
628: be rather poor. Yet, using these values,
629: Eq.\ (\ref{scalerel}) predicts
630: \begin{equation}
631:   \ell \approx 4.9
632: \end{equation}
633: for the exponent of the first normal-stress coefficient
634: $\overline{\Psi}_{1}$. If, instead, one ignored the
635: \emph{multi}fractal structure of percolation clusters in employing the
636: wrong scaling relations $k=2\nu -\beta$ and $t=d\nu$, where $t$ is the
637: critical exponent of the elastic modulus in the gel phase, one would
638: arrive at the value $\Delta\approx 0.66$. This would yield the
639: considerably lower result $\ell\approx 2.8$. Thus, it is of importance
640: to improve the accuracy of the exact numerical computation of $\Delta$
641: within the Rouse model.
642: 
643: Finally, we would like to point out that for mean-field percolation
644: the scaling relation (\ref{scalerel}) is consistent with the exact
645: result presented in the preceding section. For, in this case the model
646: yields $k=0$ \cite{BrLo99,BrLo01}, $z=3$ and $\Delta=1$ \cite{BrAs01},
647: and thus (\ref{scalerel}) gives $\ell =3$ in accordance with
648: (\ref{mfexp}).
649: 
650: 
651: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
652: %
653: \section{Time-dependent normal-stress response}
654: %
655: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
656: 
657: 
658: First, let us focus on the normal-stress response to the inception of
659: a steady shear flow $\kappa(t) = \kappa_{0}\,\Theta(t)$. Here $\Theta$
660: denotes the Heaviside-unit-step function. For each realisation of the
661: crosslinks Eq.\ (\ref{stresstensor}) leads to
662: \begin{equation}
663:   \label{timenormal}
664:   N_{1}(t):= \sigma_{xx}(t) - \sigma_{yy}(t) = 2\,\kappa_{0}^{2}
665:   \int_{0}^{t}\!\d t' \, t' \chi(t')
666: \end{equation}
667: in accordance with the principle of frame invariance \cite{Lar88}.
668: Eqs.\ (\ref{stressrelax}) and (\ref{psi1}) then imply that for all
669: crosslink concentrations below $c_{\mathrm{crit}}$ the first
670: normal-stress difference increases towards its steady-state value like
671: a stretched exponential
672: \begin{equation}
673:   \overline{N}_{1}(t) = \rho_{0}\,\kappa_{0}^{2}\, \overline{\Psi}_{1} -
674:  2\,\kappa_{0}^{2} \int_{t}^{\infty}\!\d t' \, t'\, \overline{\chi}(t')
675: \end{equation}
676: with the same exponent $\alpha$ as the shear-relaxation function
677: (\ref{kohlrausch}). In contrast, for $c=c_{\mathrm{crit}}$ we deduce
678: from (\ref{timenormal}) the algebraic growth $\overline{N}_{1}(t) \sim
679: t^{2-\Delta}$ for long times, a result already known on a
680: more phenomenological basis \cite{WiMo97}.
681: 
682: Second, we consider a sudden shearing displacement $\kappa(t) =
683: E\,\delta(t)$, where $\delta$ denotes the Dirac-delta function. From
684: (\ref{stresstensor}) we infer
685: \begin{equation}
686:   \label{suddennormal}
687:   N_{1}(t) = \int_{0}^{\infty}\!\d t'\,
688:   \chi(t')\; \frac{\d}{\d t'}\,\biggl(\int_{t-t'}^{t}\!\d s\,
689:   \kappa(s)\biggr)^{2} 
690:   = E^{2}\chi(t),
691: \end{equation}
692: 
693: which, after averaging over disorder, amounts to the Kohlrausch decay
694: (\ref{kohlrausch}) in the long-time limit for systems below the
695: critical point, respectively to the algebraic decay $t^{-\Delta}$ for
696: $c=c_{\mathrm{crit}}$. Upon comparing (\ref{suddennormal}) to the
697: corresponding result $\sigma_{xy}(t)=E\,\chi(t)$ for shear stress, the
698: Lodge-Meissner rule $N_{1}(t)/\sigma_{xy}(t) = E$, see
699: e.g.\ \S{}3.4.e in \cite{BiAr87}, holds for each crosslink realisation
700: in this Rouse-type model. 
701: 
702: Third, we consider the double-step strain flow $\kappa(t)=E\,\delta(t)-
703: E\,\delta(t-t_{1})$ with $t>t_{1}>0$. In this case one can verify in an
704: analogous manner the corresponding relation
705: $N_{1}(t)/\sigma_{xy}(t) = -E$, which is known \cite{VeKa94} to be valid 
706: for class~I simple fluids.
707: 
708: 
709: 
710: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
711: %
712: \section{Outlook}
713: %
714: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
715: 
716: 
717: We hope to stimulate detailed experimental investigations on the
718: crosslink dependence of normal stresses close to the gelation
719: transition. If such experimental results were at hand, one could judge
720: the effects of the simplifications which underly the above Rouse-type
721: model, such as the neglect of the excluded-volume interaction and of
722: the hydrodynamic interaction.
723: 
724: 
725: 
726: \begin{thebibliography}{111}
727: \frenchspacing 
728: \bibitem{BiAr87} 
729:   R. B. Bird, R. C. Armstrong, and O. Hassager,  
730:     \emph{Dynamics of polymeric liquids}, vol.~1, 2nd ed.
731:     (Wiley, New York, 1987).
732: \bibitem{Wei47} 
733:   K. Weissenberg, {Nature} \textbf{159}, 310 (1947).  
734: \bibitem{AdDe81} 
735:   M. Adam, M. Delsanti, D. Durand, G. Hild, and J. P. Munch, 
736:   {Pure Appl. Chem.} \textbf{53}, 1489 (1981).
737: \bibitem{MaWi88}
738:   J. E. Martin and  J. P. Wilcoxon, {Phys. Rev. Lett.}
739:   \textbf{61}, 373 (1988).
740: \bibitem{MaAd88}
741:   J. E. Martin, D. Adolf, and J. P. Wilcoxon,  {Phys. Rev. Lett.}
742:   \textbf{61}, 2620 (1988).
743: \bibitem{DeBo93}
744:   F. Devreux, J. P. Boilot, F. Chaput, L. Malier, and M. A. V. Axelos,
745:   {Phys. Rev. E} \textbf{47}, 2689 (1993).
746: \bibitem{CoGi93}
747:   R. H. Colby, J. R. Gillmor, and M. Rubinstein, {Phys. Rev. E}
748:   \textbf{48}, 3712 (1993).
749: \bibitem{ToFa01}
750:   P. Tordjeman, C. Fargette, and P. H. Mutin, {J. Rheol.}
751:   \textbf{45}, 995 (2001). 
752: \bibitem{BrLo99}
753:   K. Broderix, H. L\"owe, P. M\"uller, and A. Zippelius, 
754:   {Europhys. Lett.} \textbf{48}, 421 (1999).
755: \bibitem{BrLo01} 
756:   K. Broderix, H. L\"owe, P. M\"uller, and A. Zippelius, 
757:   {Phys. Rev. E} \textbf{63}, 011510 (2001).
758: \bibitem{Gen78} 
759:   P.-G. de Gennes, {Comptes Rendus Acad. Sci. (Paris)}
760:   \textbf{286B}, 131 (1978). 
761: \bibitem{HaLu87} 
762:   A. B. Harris and  T. C. Lubensky, {Phys. Rev. B} \textbf{35}, 6964 (1987).  
763: \bibitem{StJaOe99} 
764:   O. Stenull, H. K. Janssen, and K. Oerding, {Phys. Rev. E} 
765:   \textbf{59}, 4919 (1999).  
766: \bibitem{VePl01} 
767:   D. C. Vernon, M. Plischke, and B. Jo\'os, \eprint{cond-mat/0102265} (2001).  
768: \bibitem{Gen79} 
769:   P.-G. de Gennes, {J. Physique (Paris) Lett.} 
770:   \textbf{40}, L-197 (1979).
771: \bibitem{LoMe73} 
772:   A. S. Lodge and J. Meissner, {Rheol. Acta} \textbf{12}, 41 (1973);
773: \bibitem{VeWi90}
774:   S. K. Venkataraman and  H. H. Winter,  {Rheol. Acta}
775:   \textbf{29}, 423 (1990).
776: \bibitem{OsIn94}
777:   K. Osaki, T. Inoue, and K. H. Ahn,  {J. Non-Newtonian Fluid Mech.}  
778:   \textbf{54}, 109 (1994).
779: \bibitem{VeKa94} 
780:   D. C. Venerus and H. Kahvand, {J. Rheol.} \textbf{38}, 1297 (1994).  
781: \bibitem{TaEd92} 
782:   F. Tanaka and S. F. Edwards,  {J. Non-Newtonian Fluid Mech.}
783:   \textbf{43}, 247, 273, 289 (1992).
784: \bibitem{AhOs94}
785:   K. H. Ahn and K. Osaki,  {J. Non-Newtonian Fluid Mech.}  
786:   \textbf{55}, 215 (1994).
787: \bibitem{AhOs95}
788:   K. H. Ahn and K. Osaki,  {J. Non-Newtonian Fluid Mech.}
789:   \textbf{56}, 267 (1995).
790: \bibitem{BrHo95}
791:   B. H. A. A. van den Brule and P. J. Hoogerbrugge,  
792:   {J. Non-Newtonian Fluid Mech.} \textbf{60}, 303 (1995).  
793: \bibitem{BrAs01} 
794:   K. Broderix, T. Aspelmeier, A. K. Hartmann, and A. Zippelius, 
795:   {Phys. Rev. E} \textbf{64}, 021404 (2001).
796: \bibitem{DoEd85} 
797:   M. Doi and  S. F. Edwards, \emph{The theory of polymer dynamics} 
798:   (Clarendon Press, Oxford, 1985).  
799: \bibitem{BiCu87} 
800:   R. B. Bird, C. F. Curtiss, R. C. Armstrong, and O. Hassager,  
801:   \emph{Dynamics of polymeric liquids}, vol.~2, 2nd ed. 
802:   (Wiley, New York, 1987).  
803: \bibitem{SoVi95} 
804:   M. P. Solf and T. A. Vilgis, {J. Phys. A} \textbf{28}, 6655 (1995).
805: \bibitem{ErRe60} 
806:   P. Erd{\H o}s and A. R\'enyi, {Magyar Tud. Akad.
807:     Mat. Kut. Int. K{\H o}zl.} \textbf{5}, 17 (1960); 
808:   reprinted in \emph{The art of counting}, edited by J. Spencer 
809:     (MIT Press, Cambridge, MA, 1973).  
810: \bibitem{Bol98} 
811:   B. Bollob\'as, \emph{Modern graph theory} (Springer, New York, 1998).  
812: \bibitem{StAh94}
813:   D. Stauffer and A. Aharony, \emph{Introduction to percolation theory},
814:   revised 2nd ed. (Taylor and Francis, London, 1994).  
815: \bibitem{BuHa96}
816:   A. Bunde and S. Havlin, in \emph{Fractals and disordered systems},
817:   edited by A. Bunde and S. Havlin (Springer, Berlin, 1996),
818:   p. 59, p. 115.
819: \bibitem{BrGo97} 
820:   K. Broderix, P. M. Goldbart, and A. Zippelius, {Phys. Rev. Lett.}
821:   \textbf{79}, 3688 (1997).
822: \bibitem{Bau96} 
823:   H. Bauer, \emph{Probability theory} (Walter de Gruyter, Berlin, 1996).  
824: \bibitem{StCo82} 
825:   D. Stauffer, A. Coniglio, and M. Adam, {Adv. Polym. Sci.} 
826:   \textbf{44}, 103 (1982).
827: \bibitem{WiCh86} 
828:   H. H. Winter and F. Chambon, {J. Rheol.} \textbf{30}, 367 (1986).
829: \bibitem{ChWi87} 
830:   F. Chambon and H. H. Winter, {J. Rheol.} \textbf{31}, 683 (1987). 
831: \bibitem{WiMo97} 
832:   H. H. Winter and M. Mours, {Adv. Polym. Sci.} 
833:   \textbf{134}, 165 (1997).  
834: \bibitem{MaAd89} 
835:   J. E. Martin, D. Adolf, and J. P. Wilcoxon, {Phys. Rev. A} 
836:   \textbf{39}, 1325 (1989).
837: \bibitem{WiIz94} 
838:   H. H. Winter, A. Izuka, and M. E. De Rosa, {Polymer Gels and Networks} 
839:   \textbf{2}, 239 (1994).
840: \bibitem{MaAd91} 
841:   J. E. Martin and D. Adolf, {Annu. Rev. Phys. Chem.}  
842:   \textbf{42}, 311 (1991). 
843: \bibitem{Lar88} 
844:   R. G. Larson, \emph{Constitutive equations for polymer melts and
845:     solutions} (Butterworths, Boston, 1988).
846: \end{thebibliography}
847: 
848: \end{document}
849: 
850: