Vous êtes sur la page 1sur 11

Statistical Physics of Polymer Networks

Sergey Panyukov
P.N. Lebedev Physical institute, Russian Academy of Sciences,
Leninskii prospect, 53, Moscow, Russia, 117924

Abstract. In this report we review the most important concepts in statistical physics
of polymer networks. We present the field theoretical approach to the theory of phantom networks and discuss main applications of this approach to calculate the scattering
spectra of weakly charged, randomly cross-linked polymer gels. We study the phenomenon of microphase separation in the model of cross-linked block copolymer. The
main models of topological entanglements in polymer networks are reviewed and a new
Slip-Tube Model for nonlinear elasticity of entangled polymer networks is proposed.

INTRODUCTION
Understanding the molecular mechanisms of rubber elasticity remains one of the
most important unsolved problems of polymer physics. In this report we review
what we believe are the most significant of physical concepts and put them together
into a coherent molecular picture of rubber elasticity.
Major difficulties in elaborating a microscopic theory of polymer networks (see
[1]) are due to the fact that a network represents a solid exhibiting some mechanical properties absent in polymer liquids. Traditional methods of molecular theory
of solids have been worked out for crystals and can not be applied to deal with
polymer networks. Atoms in crystals are localized in regular lattice, whereas mean
positions of monomers of an amorphous polymer network are of disorder character prescribed by the network topological structure. This structure forming in the
course of the synthesis of a network imposes certain constraints on the position
of its monomers. There are two kinds of such constraints arising either due to
cross-linking of linear chains or due to topological entanglements of chains caused
by their disinterpenetration. For short enough network chains the effect of such entanglements can be neglected. In "Phantom Network Model" the chains can freely
penetrate through each other. Classical theory of James and Guth is based on the
hypothesis of affine deformation of the network. Small angle neutron scattering experiments shown that this hypothesis is violated on mesoscopic scales. To describe
the network behavior on all spatial scales we developed the strict mathematical
description of the Edwards model of network with statistically cross-linked chains

CP519, Statistical Physics, edited by M. Tokuyama and H. E. Stanley


2000 American Institute of Physics l-56396-940-8/00/$17.00

111

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

[2], section I. We show that the affinity becomes broken on length scales smaller
than the affine length RaffVariation of thermodynamic conditions to which charged gels are subjected, can
result in the appearance of complex patterns on the gel surface. Recent neutron
scattering experiments support our prediction [3] that charged gels undergo microphase separation which leads to the reorganization of their density profile on
microscopic length scales. Thus the onset of microphase separation from the homogeneous state of cross-linked blends has been well understood. On the other
hand the structure of the microphase separated state has been studied in much less
detail. There is a qualitative difference between translationally invariant solutions
or melts and polymer networks with broken translational invariance due to randomness of cross-linking process. It is therefore very interesting to find out how
this broken translational invariance affects the microphase separation process in
networks. This important question is studied in section II.
While our theory does not include some of the features of real polymer gels such
as entanglement contributions to elasticity and the non-Gaussian character of real
chains, it does capture what we consider to be the most important characteristics of
polymer gels: the frozen randomness of their structure introduced by the statistical
character of their preparation, and the interplay between short range ("liquid")
osmotic and long range ("solid") elastic forces. Moreover, entanglement effects
can be included by a proper generalization of the present model which accounts
for the effective "tube" introduced by the topological constraints. Real chains can
not penetrate through each other and the realistic model of the polymer networks
should take into account these topological constraints. The models with constrained
chain fluctuations are reviewed in section III. In section IV we build upon all the
above ideas and propose our new model of rubber elasticity - the slip-tube model.

SOLUTION OF THE EDWARDS PHANTOM


NETWORK MODEL

Since no explicit algorithm of setting up of the topology of polymer network of


macroscopic size are not available, the replica trick has been introduced by Edwards
[4]. The main idea consists in reducing the problem of calculation of the free energy
F of a network with frozen-in chemical structure to find the thermodynamic fipotential of a so called "replica system" :

Here p, and In z are chemical potentials of monomers and cross-links, respectively.


The coordinates x = {x^\ . . . , x^mH of the point of the replica system are a set of
coordinates of m replicas x^ (i = 0, . . . , ra) each of which being defined in ordinary
three-dimensional space. The 0-th replica describing the system where the network
has been formed while other m identical replicas characterizing the conditions of

112

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

experiment. The key to the microscopical theory of polymer networks gives the
field theoretical representation for fi-potential of the replica system [5]

exp [-(Ufa, z)/T\ = Jb(p exp {-H [<p]}

(2)

where the Hamiltonian H is given by

(3)
6 is the monomer size and w are effective virial coefficients in the conditions of
network preparation and experiment. This effective Hamiltonian is a straightforward extension of the <>4 zero-component field theory of a polymer chain with
excluded volume to the 3(1 -f m) dimensional replica space.
We argue that the solid state of polymer network is described by the order parameter (p(x) with spontaneously broken translational symmetry of the Hamiltonian
and find such ground state solution. The fluctuations of the field <p(x) around this
ground state are connected with corresponding density variations in real physical
space:
1. Inter-replica fluctuations originate from the randomness of cross-linking process and describe static density inhomogeneities of the network. We show
that they are small for networks prepared far from the "cross-link saturation
threshold".

2. Intra-replica fluctuations, which describe "excluded volume" density fluctuations both in synthesis and experiment conditions. A renormalization group
approach for dealing with such problems in the case of polymers with fixed
structure was developed in Ref. [6]. We adapted these methods to the present
case and obtained a scaling description of the thermodynamics of gels swollen
in good solvents.
We found that the equilibrium swelling state of gels cross-linked away from the
cross-link saturation threshold is much more concentrated than the c* state [7] and
that there are many other network chains in the volume spanned by a chain which
connects two adjacent cross-links. This means that the intuitive picture suggested
by the c* theorem in which there is only one chain in the volume occupied by
the average mesh, has to be replaced by one in which there are many chains per
mesh volume and which corresponds to a semi-dilute solution of interpenetrating
network chains.
We performed thermodynamic analysis of phase transitions in charged gels in
poor solvents [8] and shown that two types of phase transitions are possible. A

113

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

gel can undergo a phase transition into a new homogeneous and isotropic phase
by expelling the solvent and changing its volume. This type of a transition has
no analogue in binary liquids. The second type of transition which can take place
in gels as well as in liquids, is phase separation into two coexisting phases of different compositions. However, the presence of elastic forces results in important
differences between the thermodynamics and the kinetics of phase separation in
gels and in binary liquids. Phase separation in liquids proceeds through nucleation
or spinodal decomposition and results in the formation of coexisting bulk phases,
each of which is isotropic and homogeneous. In gels, the nucleation of a new bulk
phase is strongly suppressed by the fact that the formation of a such nucleus must
be accompanied by the deformation of the surrounding elastic medium. At sufficiently high degree of ionization the new phase can be formed as a thin anisotropic
layer on the surface of the gel. The stability of the surface phase is studied and it
is shown that a dilute surface face is unstable against long wavelength fluctuations
with wave vectors oriented along the surface of the gel.
Applying field theoretical methods to calculate the structure factor of neutral and
weakly charged, randomly cross-linked polymer, we find that it can be written as the
sum of two terms: the correlator of static inhomogeneities, Cq, that characterizes
the statistical properties of the inhomogeneous equilibrium density profile of the
gel, and the correlator of thermal fluctuations, Gq, about this equilibrium [3]:
a q = r>
i r*
b
Cq +
Gq

(4.}
W

We found explicit expressions for both these correlators. The presence of static inhomogeneities gives rise to observed stationary speckle patterns in light scattering
from gels. When the gel is stretched, the anisotropy of the inhomogeneous equilibrium density profile leads to enhanced scattering in the direction of the stretching
and to the appearance of "butterfly" patterns in iso-intensity plots, see Fig. 1 a.

FIGURE 1. Iso-intensity lines of scattering: on fluctuations of total monomer density - a, on


deuterated chains of the network - b. Arrows indicate the direction of the uniaxial network
stretching.

114

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

We also calculated [9] the amplitude of scattering from deuterated chains (with
fraction <) of the network stretched by factors of A2-

5rq = *(C, + Gq) + ^(l-^)5S^

(5)

Here S is the structure factor of uncross-linked deuterated chains, which depends


only on q = |q|. Eq. (5) predicts that the scattering will have "lozenge" type of
the anisotropy, see Fig. 1 b. These predictions, and also non-monotonous dependence of the scattered intensity on the degree of cross-linking (for gels prepared in
good solvent and studied in poor solvent), have been recently confirmed by light
scattering experiments [10]. The scattering experiments on NIPA gels are now in
progress.

II

MICROPHASE SEPARATION IN POLYMER


NETWORK

We study the microphase separation phenomenon using the model of block


copolymer that is cross-linked in a homogeneous state forming a multicomponent
network. Upon cooling such network, the repulsion between different components
may lead to a microphase separation into regions rich in each of these components.
If random forces in network are neglected, the microphase separation results in an
ordered microstructure. For simplicity, we consider a strongly segregated lamellar
mesophase which, in ideal state, is a one-dimensional lattice along z-direction.
We had shown [11] that microphase separation in polymer networks is significantly different from that in uncross-linked polymers. The main reason for this
difference is the existence of quenched random forces [12] acting on lamellar planes
from the network. These forces destroy the long range order in the microphase separated networks. The short range order is still presented, but the lamellar planes
are rotated with respect to each other by random forces (see Fig. 2). The lengths
scales at which this rotation reaches angles of order unity define the orientational
correlation lengths. The orientational order of lamellar planes is well defined only
in the asymmetric correlation regions of the size x~y<& z with
&-&AT 3 / 4 ,

&~6JV

(6)

where N is the number of monomers of the network strand. On large length scales
the correlations between these rotations are lost leading to random bending of
lamella. The scattering from such structure is spherically symmetric with a peak
independent on the direction of the scattering wavevector.
Upon anisotropic deformation of the network, normals to the lamella planes
tend to turn into the direction of the weakest stretching. The deviations of the
normals from this preferred direction decreases with increasing anisotropy. The
scattering from these anisotropically deformed microphase separated networks has
two pronounced peaks in the weakest stretching direction. The height of these

115

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

FIGURE 2. The distorted conformation of lamellae under the influence of random forces from
the network. The correlation region is sketched by the rectangle with dimensions r and [r (the
orientations! correlation lengths).

peaks rapidly increases and their width decreases with anisotropy. This is the
signature of a long range orientational order. The lamellar planes can be rotated
by changing the direction of weakest stretching.

Ill

CONSTRAINED NETWORK MODELS

One of the standard models that attempts to account for the topological intermolecular interactions is the "Constrained-Junction Model" [13]. It represents the
topological interactions between network chains by imposing additional restrictions
on the fluctuations of junctions by an additional harmonic potential acting on the
cross-links of the phantom network. The confining potential can be represented
by confining virtual chains connecting cross-links to the elastic background, which
deforms affinely with the network, see Fig. 3. The space available for chain fluc-

FIGURE 3. Constrained Junction model. Cross-links of network chains are connected to the
elastic nonfluctuating background by virtual chains (thin lines) containing ra monomers each.

116

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

tuations changes upon network deformations, since chains move further apart or
closer together. To take this effect into account the potential acting on the crosslinks is assumed to change with deforming network. This implies that the number of
monomers of confining virtual chains (see Fig. 3) varies with elongation coefficients
A as
m

= m0A2

(7)

In the Constrained-Junction Model only junction fluctuations are constrained,


the fluctuations of other monomers of the chain are almost unchanged. This is not
a realistic representation of real topological interactions in polymer networks. The
realistic model of the topological constrains should take into account the fact that
topological interactions act along the whole chain and restrict the fluctuations of
all monomers of the chain. The first successful way of representing these topological entanglements is the "Confining Tube Model" by Edwards [14]. In this model
the amplitude of fluctuations of chains is independent of the network deformation.
Although the Edwards tube model captures the physics of the continuous distribution of constraints along the chain, this assumption contradicts both theoretical
and experimental observations. The confining virtual chains were introduced only
to restrict the fluctuations of network chains and they can not carry any real stress.
In the Edwards tube model the stress is carried both by real network chains and
by confining virtual chains.
An accurate estimate of the free energy of a chain confined to the deformation
dependent tube was made within the framework of the "Non-Affine Tube Model"
[15]. The main idea of this model is that the effect of entanglements is formed
collectively by many neighboring chains and thus, it can be described by an effective
mean field vr. There are two ways how to introduce this topological field in the
theory:
1. In the replica field theory formalism we add the term in Hamiltonian, Eq. (3):

(8)

The field theory, Eqs. (3) and (8), can be solved analogously to the Phantom
Network model [16].

2. We can attach every monomer of the chain to the elastic background by a


confining virtual chain.
This model is similar to the classical Edwards tube model. There are two major
differences between the two tube models: In the Edwards tube model the points
of attachments of confining virtual chains to the elastic background are placed
along the well-defined contour of the tube. In the non-affine tube model they
are randomly distributed in space in order to guarantee the Gaussian statistics
of the network chain at the preparation conditions. In the Edwards tube model

117

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

105

9104
810*
7104

6104

0.5 1 I/A
0.4

0.6 0.8 1

FIGURE 4. Comparison of the experimental data (solid circles) with predictions of the nonaflme
and the slip-tube models on the Mooney representation for uniaxial deformation data of PDMS
network. Dashed line - fit of the data by the nonaffine model, solid line is fit by the slip-tube
model. The redistribution parameter gz is shown in the insertr.

the amplitude of fluctuations of confining virtual chains is independent of network


deformation. In the non-affine tube model it changes proportionally to the network
deformation. This implies that the number of monomers of confining virtual chains
in the non-affine tube model m = mo A2, see Eq. (7). It was shown this is the only
possible assumption consistent with the microscopic definition of the stress tensor:
the confined virtual chains make no contribution to the network stress.

The attractive feature of the Confining Tube Model is that it provides an unified
physical picture of the deformation of both phantom and entangled networks. This
picture is based on the separation of liquid-like and solid-like degrees of freedom.
Both networks deform affinely on scales larger then the affine length J?a//. Deformation of both networks on length scales smaller than Ra// can be described by
the elongation of the individual chains. The major difference between the Phantom Network and the Non-Affine Tube models is that in phantom networks Raf/
changes affinely with the network deformation while in our tube model it changes
non-affinely Ra/f ~ A3/2 for A > 1 and Raf/ ~ A1/2 for A < 1. The tube diameter
a of the non-affine tube model also changes non-affinely a ~ A1/2 with the network
deformation.
The predictions of the non-affine tube model are in much better agreement with
experiments on uniaxial compression than the Mooney-Rivlin expression. Nevertheless, the maximum of the normalized stress on the Mooney plot predicted by
this model is not as pronounced as in some experiments (see Fig 4).

118

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

FIGURE 5. Slip-Tube model. Confining potential acting on network chains is represented by


virtual chains attached to the elastic nonfluctuating background at one end and ending with
virtual slip-links at the other. These slip-links can slide along the network chain, but can not
pass through each other.

IV

SLIP-TUBE MODEL

In the non-affine tube model the virtual chains are permanently attached to
the network chains. This eliminates the possibility of slippage of the network chain
and the redistribution of its monomers along the contour of the tube. This slippage
mechanism leads to the main modes of stress relaxation in polymer melts - reptation
and tube length fluctuations. It is natural to expect that the redistribution of chain
length along the tube contour will also be important in anisotropically deformed
highly entangled networks. We modify the non-affine tube model to include these
degrees of freedom. In the "Slip-Tube model" [17] the junction points between
network chains and virtual chains are replaced by slip-links attached to fluctuating
ends of virtual chains, see Fig. 5. These slip-links are allowed to slide along the
contour of the network chains but are not allowed to pass through each other. Note
that the same confining slip-tube but with free boundary conditions at the ends of
the chain becomes the classical reptation model with tube length fluctuations.
The effect of the redistribution of chain length along the tube contour has been
earlier studied within the framework of the slip-link models. The major difference
between slip-link and tube models is the assumptions made in these models about
the nature of topological entanglements. In tube models the confining potential
acting on a given network strand is assumed to be formed by many neighboring
strands. This justifies the use of the mean field description of topological entanglements within the framework of tube models. In slip-link models each entanglement
is postulated to be formed by a pair of chains.
We show that the redistribution parameter gz = 3NZ/N (Nz is the number of
chain monomers along ^-direction) of the stored length between different directions
of deformation decreases monotonically with reciprocal deformation I/A, see insert
in Fig. 4. In the case of strongly compressed network, A < 1, all of the chain
length is stored only in the plane perpendicular to the direction of compression. In
the case of strong stretching, A ^> 1, the chain is pulled out from perpendicular

119

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

FIGURE 6. Universal plot for the reduced stress in uniaxially deformed networks. Solid line theoretical predictions of the Slip-Tube model, Eq. 9.

directions into the stretching one.


The dependence of the stress a on the elongation coefficient A for the uniaxially deformed network is usually represented in the form of the Mooney stress
/*(1/A) = cr/(A I/A 2 ). One can see from Fig. 4 that the slip-tube model is in
much better agreement with experiments than nonaffine model. Our solution of
the slip-tube model suggests a useful comparison between the theory and experimental data in the form of the universal plot. It can be obtained by subtracting the
phantom modulus Gc from the Mooney function /* and dividing the difference by
the entangled modulus Ge. The best collapse of all the data to the universal curve
gives a pair of adjustable parameters Gc and Ge for each network. The result of this
procedure is shown in Fig. 6. Solid square and diamond ~ natural rubber, other
symbols - different PDMS networks. All six sets of experimental data collapse very
well onto a single theoretical universal plot
1.84A1 + 0.84A-3/2

(9)

Deviations from this universal curve are smaller than fluctuations within each individual set of data. The excellent agreement between the theory of the slip-tube
model and experiment demonstrates that it captures the essential physics of rubber
elasticity. Note that the most detailed check of this theory can be done using the
data of scattering experiments on entangled networks. The theoretical structure
factor for the slip-tube model can be obtained by the direct renormalization of

120

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

the structure factor of the nonaffine tube model [18]. So the matter is for future
experiments...

UNSOLVED PROBLEMS
In this report we presented main results of the statistical mechanical analysis
of neutral and weakly charged polymer networks. Interplay between disorder and
Coulomb interaction in strongly charged polymer networks can lead to the formation of glass-like structures. The description of such networks remains thus far an
open problem.
Large scale relaxation processes in phantom polymer networks had been studied
in Ref. [12]. While small scale inhomogeneities can be observed in speckle pattern
experiments, the theory of the relaxation of these short wave length modes has not
been developed yet.
The description of dynamics of the slip-tube model remains one of the most interesting unsolved problem in the theory of topological entanglements. It is important
to understand the relaxation processes in strongly deformed entangled polymers.

REFERENCES
1.
2.
3.
4.

Edwards S.F., Vilgis T.A. Rep. Prog. Phys. 51, 243 (1988).
Beam R.T. and Edwards S.F., Philos. Trans. R. Soc. London, A 280, 317 (1976).
Panyukov S., Rabin Y., Macromolecules 29, 7960 (1996).
Edwards S.F. in. Proc 4th International Conference on Amorphous Materials, Douglas R.W., Ellis B., eds., Wiley PubL, NY, 1970.
5. Panyukov S.V., JETP Lett. 55, 608 (1992); Panyukov S., Rabin Y., Physics Reports
269, I (996).
6. Panyukov S.V., Sov. Phys. JETP 67, 930 (1988).
7. de Gennes P.-G., Scaling Concepts in Polymer Physics, Cornell University Press,
Ithaca, NY, 1979.
8. Panyukov S., Rabin Y., Macromolecules 29, 8530 (1996).
9. Panyukov S.V., Sov. Phys. JETP 75, 347 (1992).
10. Ikkai F., Shibayama M., Phys. Rev. E 56, R51 (1997).
11. Panyukov S., Rubinstein M., Macromolecules 29, 8220 (1996).
12. Panyukov S.V. JETP 76, 808 (1993).
13. Erman, B., Mark J.E. Structure and Properties of Rubberlike Networks, NY, Oxford:
Oxford University Press, 1997.
14. Edwards S. F. Proc. Phys. Soc. (London), 92, 9 (1967).
15. Rubinstein M.; Panyukov S.V. Macromolecules, 30, 8036 (1997).
16. Panyukov S.V., Sov. Phys. JETP 67, 2274 (1988); 69, 342 (1989).
17. Rubinstein M.; Panyukov S.V. Macromolecules (1999).
18. Panyukov S., Potemkin L, J. Phys. I (France) 7, 1 (1997).

121

Downloaded 03 Aug 2005 to 152.2.174.22. Redistribution subject to AIP license or copyright, see http://proceedings.aip.org/proceedings/cpcr.jsp

Vous aimerez peut-être aussi