You are on page 1of 55

Rubber Elasticity of Polymer Networks: Theories

G . H e i n r i c h 1, E. S t r a u b e 2 a n d G . H e l m i s 1
T e c h n i s c h e H o c h s c h u l e ,,Carl S c h o r l e m m e r " L e u n a - M e r s e b u r g
S e k t i o n P h y s i k 1, S e k t i o n W e r k s t o f f - u n d V e r a r b e i t u n g s t e c h n i k 2
DDR-4200 Merseburg, Otto-Nuschke-Str.

The present state of development of the statistical mechanics of rubber elasticity is reviewed and
analysed, starting from some problems and controversial results drawn from recent experimental
progress in this area. Attention is focused on the tube model as a mean field approach to the statistical
mechanics of polymer systems with topology conservation. In particular, a new model for simulating
the topological constraints in polymer networks and melts is presented which allows the order and
the deformation dependence of the tube dimensions to be calculated. Conclusions resulting from the
description of large-strain and small-strain behaviour of dry, completely crosslinked networks are
discussed and compared with experimental data where all modes of deformation usually employed
can be described with similar accuracy. Further, the concept of relaxed microscopic deformation
much smaller than the macroscopic deformation of the sample is introduced, which allows explanation
of mechanical and thermodynamic properties as well as the scattering results for networks at higher
swelling degrees. Similarly, constraint release effects are expected to be responsible for the different
experimental results collected for end-linked networks and for networks prepared by cross-linking
of long primary chains. The different degrees of completeness of crosslinking have to be considered
as the main reason for these differences.

List of S y m b o l s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2 The Topology of Networks . . . . . . . . . . . . . . . . . . . . . . 37


2.1 T h e T o p o l o g y o f C r o s s l i n k s . . . . . . . . . . . . . . . . . . . . 38
2.2 T h e T o p o l o g y o f E n t a n g l e m e n t s . . . . . . . . . . . . . . . . . . 39

3 Theory of Phantom Networks . . . . . . . . . . . . . . . . . . . . . 46

4 Theory of Real Networks . . . . . . . . . . . . . . . . . . . . . . . 49


4.1 Non-topological "Interaction" Concepts . . . . . . . i . . . . . . 49
4.2 The Theory of Restricted Junction Fluctuations . . . . . . . . . . . 49
4.3 The Langtey-Graessley Concept . . . . . . . . . . . . . . . . . . 51
4,4 The "Single-Chain" Approach . . . . . . . . . . . . . . . . . . 53
4.5 Network Theories Based on Topological Invariants . . . . . . . . . . 54
4.6 The Tube Model for Randomly Crosslinked Networks . . . . . . . . 57

5 Results and Comparison with Experiments . . . . . . . . . . . . . . . . 63


5.1 S t r e s s - S t r a i n B e h a v i o u r . . . . . . . . . . . . . . . . . . . . . 64
5.2 S m a l l - S t r a i n B e h a v i o u r a n d t h e F r o n t F a c t o r P r o b l e m . . . . . . . . 67
34 G. Heinrich, E. Straube und G. Helmis

5.3 Swelling Properties of Polymer Networks . . . . . . . . . . . . . . 73


5.4 SANS and Microscopic Deformation . . . . . . . . . . . . . . . . 79
6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . 82
7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

List of Symbols

A microstructure factor
at,c activity of the solvent over a swollen cross-linked polymer
al,u activity of the solvent over an uncrosslinked polymer system
CI, C 2 Mooney-Rivlin parameters
c 1~ c 2 reduced Mooney-Rivlin parameters
do undeformed tube radius
d~ deformed tube radius (g = x, y, z)
F elastic free energy
FT elastic free energy of a network with a special topology
f functionality of the crosslinks
G shear modulus
Gc shear modulus contribution arising from chemical crosslinks
G~ maximum possible contribution of entangled chains to the modulus
GN shear modulus contribution arising from topological constraints
Go plateau modulus of an uncrosslinked polymer system
g front factor
g~ reduced shear modulus
H Hamiltonian
I~ topological invariant
11, 12, 13 invariants of the deformation tensor
t 1, generalized invariants of a generalized deformation tensor
k scattering vector
k Boltzmann constant
L contour length of a macromolecule
L contour length of a network strand
1 statistical segment length
M number of chemical crosslinks
M number-average molecular weight of a network chain
Mn number-average molecular weight of a primary chain
Mw weight-average molecular weight of a primary chain
Ms molecular weight of a statistical segment
m Gauss linking number
N Number of primary chains
NA Avogadro number
Nk number of network chains
Ns number of statistical segments per macromolecule
N, number of statistical segments per network chain
NF Flory number [ = number of network chains in the volume (1Lc)3/2]
Rubber Elasticity of Polymer Networks: Theories 35

Nsl number of slip-links


n number of spatial neighbours of a network junction
np number density of polymer chains
ns number density of statistical segments
n1 number of moles of solvent in the swollen network
R(S), r(s) configurations of a macromolecule
k(s), ~(s) mean configurations of a macromolecule
Re end-to-end distance of a network chain
R .gl radius of gyration of a chain in the reference (undeformed) state
Rgll radius of gyration of a network chain parallel to the stretching
Rg± radius of gyration of a network chain perpendicular to the stretching
s(k) scattering function
S chain arc length
T temperature, topology
T trapping factor
V,V 0 volume
71 molar volume of the diluent
V2 polymer volume fraction in a swollen gel
v, polymer volume fraction corresponding to the equilibrium swelling degree
v~'* cross-over volume fraction from Gaussian to excluded-volume behaviour
of the network chains
o
volume fraction of polymer in the solution prior to crosslinking
Wel elastic potential
Ws sol fraction
WT probability of a special topology
Z canonical partition integral
(or [3) parameter of the constraint release effect in polymer networks
F dimensionless parameter of the strength of topological constraints
parameter of the Flory-Kgstner theory characterizing the departures from
affine transformation of the shapes of domains
n "memory" factor
~4 parameter of the Flory-K~stner theory characterising the strength of
restrictions on junction fluctuations
k deformation ratio of the sample, deformation tensor
linear isotropic extension ratio of swollen networks
~mi e
microscopic deformation ratio
tac chemical crosslink density
elastic contribution of the chemical potential of the diluent in swollen gels
Ve network chain density
cycle rank (= number of independent circuits in the network)
Qp polymer density (g/mol)
G nominal stress (equilibrium value of the elastic force measured in uniaxial
deformation divided by the undeformed cross-sectional area of the sample)
(TYM Mooney stress [= o/(L -- ~-2)]
36 G. Heinrich, E. Straube and G. Helmis

1 Introduction
Rubber elasticity is a unique phenomenon which has attracted the attention of many
researchers and which has been reviewed in a considerable number of papers 1-7)
Although its entropic nature was discovered long ago, a deep understanding is today,
still, a task of great scientific and technological importance. Rubber-like polymer
networks belong to the best-understood bulk polymer systems, and progress in this
field continues to influence the development of knowledge of other polymer systems
to a remarkable degree. Examples are polymer melts and partially crystalline polymers.
The physics of rubber elasticity is characterized by a great variety of approaches,
models and concepts. It is the aim of this review to summarise the situation and to
formulate a comprehensive picture of this topic from the point of view of the authors.
In the introduction, this situation will be illustrated by a few examples of current
problems and developments.
The simplest model of rubber-like behaviour is the phantom network model.
The term "phantom" is used to emphasize that the configurations available to each
strand are assumed to depend on the positions of the junctions only. Consequently,
the configurations of one chain are independent of the configurations of neighbouring
strands. For many purposes, the strands can be treated as Gaussian random coils.
Even in this simplest case, an exact solution is not a trivial task as will be outlined in
Sect. 3.
The phantom network can account qualitatively for many properties ofcrosslinked
elastomers, but the quantitative explanation of basic properties is wrong, For example,
stress-strain properties, especially in simple extension, show departures from the
phantom network results even at extension ratios covered by the Gaussian chain
model. The explanation of these departures, phenomenologically described by the
famous Mooney-Rivlin Eq. (1)

cy0v) = 2C, Q ~ - ;k-2) + 2Cz(1 - - ) - - 3 ) (1)

was the starting point for a great number of studies of the statistical mechanics of
networks. Section 4 of this review deals with this problem extensively. In Eq. (1),
cy denotes the tensile force per unit unstretched area and ~, is the tensile stretch ratio.
Related problems which have received more interest recently are the swelling
dependence of the C2 term in the Mooney-Rivlin equation, and the swelling depend-
ence of the product of the dilation ratio and of the elastic contribution to the chemical
potential of the solute in a swollen network. Classical network theories predicted
either constant or monotonically increasing values, whereas the experiments give
a sharp and pronounced maximum 8-12)
In recent experimental studies by small angle neutron scattering (SANS)13-15)
and deuterium magnetic resonance (2H NMR) i6-19), the mechanisms of deformation
and orientation of network chains has received increasing attention. Experiments
by SANS on end-linked poly(dimethylsiloxane) (PDMS) networks have shown that
the changes of network chain dimensions are smaller than the values given by all
phantom network theories. These results may be expected to be of importance beyond
the physics of networks, because they point to the existence of constraint release
mechanisms which are not due to single chain reptation processes. The suggestions
Rubber Elasticity of Polymer Networks: Theories 37

put forward by several authors about these phenomena are discussed together with
the abovementioned swelling behaviour in Sect. 5.
A review on rubber elasticity has to take account of the current controversy con-
cerning the experimental results and their treatment and the conclusions drawn from
investigations made on PDMS end-linked "model" networks. There are two main
points in dispute. Several authors 20-24) assume that the end-linked networks are really
perfect networks with constant and known functionalities of the junctions and a mono-
disperse chain length distribution, whereas other authors 25-32) point out that the
number of dangling ends depends very much on the degree of conversion of the
functional groups. It is beyond the scope of this review to decide the controversial
question concerning the completeness of reaction in the investigated networks, but
it should be noted that networks made by crosslinking of long primary chains may' be
expected to be almost completely crosslinked. The term "complete crosslinking"
means not only the absence of any finite cluster, but also that the number of paths
connecting two network junctions increases exponentially with increasing distance
between the junctions. In accordance with the different opinions concerning the net-
work structure, different conclusions on the proper approach to the statistical
mechanics of networks have been drawn. The experimental results of Flory, Mark
and coworkers z0-24,33-35) obtained for networks produced by end-linking seem to
be in agreement with the theory due to K~istner 36-40) and Flory and Erman 41-44)
of restricted junction fluctuation, On the other hand, Graessley, Macosko and
others 25-3z~ came to the conclusion that the topological constraints caused by the
uncrossability of the network chains affect all chain segments equally, as in polymer
melts, and may be taken into account to a good approximation with the help of tube-
type models.
From this rough outline of some examples of current problems in the physics of
rubber elasticity, it is clear that it is important to have a well-founded statistical-
mechanical theory of equilibrium properties of rubber-elastic networks. Consequently,
first junction and entanglement topology are described and discussed. Then a section
briefly reviews the theory of the phantom network. In the following two sections,
theories of equilibrium properties of networks and a comparison of theoretical
results with experimental data are presented.

2 The Topology of Networks


Rubber-like networks are in general disordered systems. As in many macromolecular
systems, several types of disorder exist, each of which demands a different treatment.
Small-scale disorder on the level of the co-monomer sequence and/or the configura-
tions or conformations of the monomeric units may be absorbed into the random
flight statistics of the most often used Gaussian chain model. The basic characteristics
of rubber-like networks are the large-scale topology of crosslinks and the topology
of entanglements. Both types of large scale topology are fixed by the crosslinks.
Thus, to understand rubber elasticity we need a statistical mechanics taking into
account topology conservation at all changes of state (especially under deformation)
of the system.
If we confine ourselves to equilibrium properties and isothermal conditions, the
38 G. Heinrich, E. Straube and G. Hetmis

free energy F is the key quantity of the theory. For disordered systems, F is given by
a structure average over the contributions of all possible realisations (topologies)
of the system 4s, ~6):

V0d = ~ WvFT(?~) (2)


T

The sum in Eq. (2) extends over the topologies of the network, w:r is the deformation-
independent a priori probability of the occurrence of the topology T and Fv(?~) is
the free energy of a network with the topology T; FT0~) depends on the deformation
tensor )~ and on further external parameters as, for example, the swelling degree.
An expression for FT(k ) is given by the canonical partition integral

ZT(~.) = f dF exp (-~3H), [~ = (kT) - 1 (3)


T

and

ErO Q = - - k T In [ZT(k)] (4)

Equation (3) means that the integrations have to be performed over those parts of
the configuration space that are compatible with the given topology T; H is the
Hamiltonian of the system. From Eqs. (2)-(4) it follows that the equilibrium theory
of network properties involves at least three steps:
- - Classification and description of the topology plus determination of W-r
- - Calculation of FT(~ )
- - Averaging over the topologies
The expressions for w T and F~(£) resulting from the first two steps are in most cases
of interest so complicated that a direct evaluation of the sum in Eq. (2) is hardly
possible. This problem can be overcome in two ways:
-- The sum is approximated by F~(£), the contribution of a representative (most
probable, giving the largest contribution) realisation of the ensemble.
-- Application of the replica trick of Edwards 4v.,8), which is based on the identity

d d d
In Z = lim Z n = lim exp (n In Z) = lim dnn (t + n In Z + O(n2))
n40 ~ .~0 ~ °~0
(5)

and uses the fact that it is much easier to average the power Z n of the partition integral
of n copies of the system than to average In Z itself.
It should be noted that in the case of the phantom network (see Sect. 3) the free
energy of any "typical" representative with a given number of chemical crosslinks
shows the same behaviour.

2.1 The T o p o l o g y o f Crosslinks

Numbering the crosslinks by i = 1. . . . . M and the chains between the crosslinks by


k : l . . . . . Nk, the crosslink topology of a network of Gaussian chains composed
Rubber Elasticityof PolymerNetworks:Theories 39

of statistical segments with segment length 1 may be characterised by the off-diagonal


elements of the symmetric M x M reduced Kirchhoff matrix K

{ ~ } f connected )
Kij = 3/(2 Nij ) for crosslinks i and j direetly { ,~
tunconnectea)
by a network chain with N~j segments. (6)

The N k n o n - z e r o matrix elements Kij (i < j) contain all the information necessary
for the theory of the equilibrium properties of the phantom network. The determina-
tion of the distribution of the N~j has been the aim of numerous studies (see e.g.
Refs. 49-51~) dealing with the problems of network formation. These questions are
of particular interest for networks formed by reactions of prepolymer chains or of
multifunctional monomers.
For networks formed by vulcanisation of long primary chains, the assumption of
random crosslinking with equal probability of any realisation with Z N~j -~ constant
is the least arbitrary and most widely used.
Model networks formed by perfect end-to-end linking of monodisperse chains
are characterised by Nij = N~ = constant.
Using the definition of K~j, the partition integral (3) acquires the form

ZT(~) = f exp (-V/kT) I-I 6(R~ - ).Rb,~) dR1 ... dRM (71
i

with

V/kT = ~ Kkl(Rl - Rk) 2 (8)


k<l

The product of the 5-functions takes into account the crosslinks fixed on the boundary
of the sample. The macroscopic deformation of the*sample is introduced by the
displacement of these boundary points Rb. ~~ LRb, i.
The concept of the phantom network is completely defined by Eqs. (6)-(8). Its
treatment is outlined in Sect. 3. It is worth noting that non-reduced connectivity
matrices, describing both the connection of the segments in the linear parts of the
network and the connection of the chains by the crosslinks, are a suitable basis for
a theory of the dynamical properties of the phantom network. The basic theory" and
applications can be found in Refs. 52,53~

2.2 The Topology of Entanglements

The treatments of crosslink topology and entanglement topology are very different.
In contrast to the phantom case, it is impossible to express the restrictions caused
by the conservation of the entanglement topology by any weight function depending
on the coordinates of the configuration space of the network chains only. The con-
figuration space splits into accessible and inaccessible regions, divided by hyper-
surfaces that form the boundaries of areas in the configuration space with
T = constant.
40 G. Heinrich, E. Straube and G. Helmis

One way to classify and describe the topology of chain systems is to introduce
topological invariants. Until now (see Sect. 4.5), the topological invariants related to
the properties of loop integrals in electromagnetism have been used exclusively in
network physics. For two closed chain contours C and C a the functional

I,a= ~
'# d s d s ' #,(s)-?a(s') x V l r , _ r (9)

has integer values, depending on the number of entanglements of the two contours.
Figure 1 gives examples for the entanglement situation and the corresponding values
of the invariants. The relationship between the formulation of the invariants and
electromagnetism suggested a gauge-field approach to Brereton and Shah 54), Elder-

Co~ C# C,~ Cfl


[~=0 I~L~=I

l~ =2 I~=0
Fig. 1. Examples of the lowest invariants for two entangled chains

field 551 and Tanaka 56) Entanglement topologies of chains have been classified and
discussed mainly in the context of biopolymer problems by Frank-Kamenetskij
and Vologodskij 57)
Topological invariants offer in principle a rigorous approach to the statistical
mechanics of entangled polymer systems, but this approach suffers from a number
of disadvantages. First of all, the invariants are so complex that the use of topological
invariants for more than two chains is hardly possible 58-63). This restriction would
correspond in physics of simple liquids to a restriction of cluster expansion theories
to the contributions of the binary cluster integral. The limitation to invariants involv-
ing a small number of chains only confines the theory in the strict sense to systems
near the critical concentration for the onset of the entangled behaviour. This restric-
tion is important because typical rubber-like networks are distinguished by a high
degree of chain overlap. The Flory number NF, the number of chains within the coil
volume (R2) 3/2 of one chain, is given by

N F = (nt 3) N~/2 (10)

where n s = QpNA/M s is the number density of statistical segments, Qp the density of


the network, N g the Avogadro number and M s the molecular mass of a statistical
Rubber Elasticity of Polymer Networks: Theories 41

Table 1. Reduced segment density and the ratio (molecu-


lar mass of a network chain)/(number of statistical seg-
ments) for some common rubbers

Polymer n,l3 a M (g/mol)

Poly(dimethylsiloxane) 2.12 301


cis,trans-Polybutadiene 5.09 105
cis-Polybutadiene 3.77 70
1,2-Polybutadiene 4.09 236
cis-Polyisoprene 3.37 122

" Values of the statistical segment length are taken from


Ref. 64~

segment. Table 1 shows values for the reduced segment density n~l3 and for M~
= M J N c, the ratio between the molecular mass o f a network chain and its number
of statistical segments, for some c o m m o n rubbers. The data show that N v ~> 1 holds
in bulk systems with the exception o f crosslink densities beyond the usual values for
rubbers.
It has to be added that even considerations of two-chain invariants or one chain-one
obstacle invariants 65, 66) are too complex to be treated rigorously. Consequently,
perturbation expansions 67) and/or variational methods 60,68,69) have to be used,
which in most cases introduces serious and partially unknown approximations into
the theory. Finally, it must be noted that the invariants according to Eq. (9) cannot
differentiate between the situation without entanglements and a much more com-
plicated one also giving I ~ = 0. Furthermore, the "simple" formulation by loop
integrals is not very suitable for the entanglement situation in networks where the
contour o f one network chain can be closed in general by many paths passing through
the network.
With regard to the sketched problems and the obtained results (see Sects. 4 and 5),
it has to be concluded that a direct approach to rubber elasticity starting with Eqs.
(2)-(4) and with the topological invariants has not as yet been able to provide a satis-
factory theoretical basis for the understanding of rubber elasticity and for the treat-
ment o f experimental results. Consequently, a number of approaches which may be
summarised as mean field approximations have been developed.
The physical basis of the mean field approach to the entanglement contribution
to rubber elasticity is the large degree of coil interpenetration. For N v ~> 1, the
restrictions on the configurations of one chain under consideration are strong and
are caused by contributions of a number o f chains of the order of N v. Consequently,
it may be expected that the fluctuations o f the topological constraints are small and
that effects o f large-scale correlations are screened out so that the topological con-
straints are well simulated by a constraining potential. This picture was suggested by
Edwards 7o, 71) in one o f his first papers on bulk polymers. A chain is confined to the
neighbourhood o f the initial or mean configuration by a restoring (usually harmonic)
potential. Following the rules of statistical mechanics, this confining potential has
to be determined self-consistently. This problem is in general not much simpler than
the investigation o f the entanglement contributions to the free energy discussed above.
However, in the case o f N v ~> 1 it can be expected that the contributions to the
42 G. Heinrich, E. Straube and G. Helmis

confining potential are not controlled by the global topology of all chains in the coil
volume but by the local topology of the chains in the neighbourhood of the chain
under consideration.
The mean field approach can be applied in different stages of elaboration. In the
first stage, models are introduced that contain additional free parameters which are
not determined by microscopic theory. Using these models, the influence of the con-
straints on network properties has been calculated and discussed. Box models 72-78)
slip-link models 79, 80), constraining springs 81), constrained junction fluctuation 36 - 44~
and different tube models 70,71,82-87) are predominantly used. The main charac-
teristics of these models and the resulting network properties are outlined in more
detail in Sect. 4. Here some general features will be discussed which give insight
into the mechanisms of the influence of topology conservation on chain configurations
and network properties. An almost trivial but important result of these investigations
is the finding that only the deformation dependence of the effective constraints on
the chain configurations influences the stress-strain properties of networks.
Tube models, and their simpler versions, the box models, are distinguished by the
advantage that the constraints act equally on each segment of the chains. This assump-
tion corresponds much more directly to the physically well settled idea of the mean
field approach than an introduction of discrete constraints as in the models of con-
strained junction fluctuation 36-44) and of constraining springs sl) in slip-link
models 79, 80, 88) or in the Langley concept 89) of trapped entanglements. Additionally,
strength and deformation dependence of the constraints may be expressed in the case
of tube models in a very instructive way by the square root of the mean square deviation
of the position of a segment from its average position:

d = ( [ g (s) - - k.(s)]2) 1/2 (11)

In deformed systems, it is preferable to work in the principal axes system of the


deformation tensor and to consider the component d of the mean square deviations
in the directions of the principal axes. In phenomenologicat as well as in microscopic
approaches almost exclusively the power law

d = do~.~ (12)

is used. In Eq. (12), d o is a measure of the strength of the constraints whereas the
exponent v controls their deformation dependence.
Among the cited models, the tube model with harmonic constraints seems to be
the most natural. Furthermore, it allows an approximate self-consistent calculation
of the strength and the deformation dependence of the constraining potential. The
following model is used in Refs. 90-92).
- - The configurations of the chains are described by space curves R(s), where s is
the arc length of the chain.
- - The distribution of the mean configurations R(s) obeys the random walk behaviour

P(R(s)) - exp { - 3/21 ~L, ds[d/~(s)/ds] 2} (1 3)

where L = N¢I is the contour length of a network chain.


Rubber Elasticity of Polymer Networks: Theories 43

- - The statistical weight of a configuration R(s) of a chain with a mean configuration


(tube axis) R(s) is given by

I L¢
P(R(s) I/~(s)) ,-~ exp - ~ ds {(3/21) [dR(s)/ds] 2

(14)
+,=x...z -

The strength of the constraints is determined in this model by the prefactors of


[ R ( s ) - I~ (s)]2. It is assumed that for the undeformed isotropic system the con-
straining potential is independent of the direction of the constrained chain. Con-
sequently, the constraining potential has to be diagonal in the main axis system of
the deformation tensor in the case of external deformation.
The constraint model defined above has the advantage that it gives a network
model which can be investigated within the framework of functional integration,
which yields closed analytical expressions for the free energy.
The idea of the calculation of the constraining potential follows directly from
the mean field view of the problem : the constraining potential is given by the change
of free energy of the surrounding chains as a function of the displacement of a segment
of the chain under consideration from the corresponding tube axis. Figure 2 illustrates
the model and the physical meaning of the quantities which occur in Eqs. (13) and
(14). The results will be applied in Sect. 4 to the calculation of the free energy of
a network. Here, some general features of the dependence of the potential parameters

yl

Y(Lc/2}>_Y o
f" .,~
/ \

_rls)/
\
_toE / ELc~,

Fig. 2. Entanglementmodel. Configuration of the constrained chain: mean, -- ....... special.


Allowed configuration of a constraining chain: . . . .
44 G. Heinrich, E. Straube and G. Hetmis

on the network parameters will be discussed, which lead to a classification of networks


with respect to the relative importance of crosstink and entanglement contributions.
The simplest case is that of high crosslink density or small coil interpenetration
(N v >~ 1). In this case, the restrictions on the configurations of network chains caused
by the crosslinks dominate, and the constraints acting on the constraining chains
may be omitted in the course of the calculation of the constraining potential. With
the assumption of an affine displacement of the crosslink positions with the deforma-
tion of the sample, Eq. (12) was obtained 91) with

do/(tL ) 1/z = N ; 1/4 and v = -- 1/4 (I 5)

As expected from scaling considerations, the constraints are controlled by the chain
or crosslink density only. An estimate of the contributions by the constraints acting
on the constraining chains gives N v - 20 as the upper limit for applicability of Eqs.
(12) and (15). The fluctuations of the crosslinks increase the numerical value of
d o by the factor p = (1 + ( A R 2c ) / ( R c2) ) l/2 9a). The ratio of the mean square crosslink
fluctuations and the mean square network chain end-to-end distance ( A R ~ ) / ( R 2)
= 1/2 for the tetrafunctional phantom network gives an estimate of this effect. Longer
network chains give higher degrees of coil interpenetrations and consequently entang-
lement and constraint effects result. Therefore, the constraints acting on the con-
straining chains have to be taken into account in a completely self-consistent manner.
An analytic expression for d u has been derived in the case that effects of the con-
straints predominate over the crosslink contribution, i.e. in the melt case 92). The
assumption of affine displacement of the tube axis of the constraining chains again
gives Eq. (12) with

do/1 -~ (n13) -1/3 and v = 1/3. (16)

This result is not in agreement with the concentration dependence of the tube diameter
or the entanglement spacing deduced from plateau modulus data of polymer melts or
concentrated solutions. The reason is that the model used, together with the affine
displacement of the tube axis, does not take into account the occurrence of the first
relaxation process according to the Doi/Edwards 94) terminology. As a result of the
first relaxation process, parts of the chain of the order of the tube dimension are in
equilibrium, i.e. their free energy does not change relative to the undeformed state.
This may be taken into consideration in the foregoing calculations by simply intro-
ducing an effective number of degrees of freedom fo~ the constraining chains. The
number of degrees of freedom is reduced by the factor 1/do, which is the number of
statistical segments between "entanglements". The corrected theory gives 92)

d0/1 = 8.5(ns13)-1/2 and v = 1/2 (17)

The ratio do/1 is controlled by the reduced segment density only. Using these values
in the equation for the plateau modulus

G°N = 4/5kTnplL/do2 (18)


Rubber Elasticity of PolymerNetworks: Theories 45

(where np is the number density of potymer chains) yields

G ° = 0.0t lkT(n 13)21-3 (19)

This result agrees with the scaling predictions of Edwards and Graessley 95). Even
the numerical prefactor following from the numerical prefactors in Eqs. (17), (18)
is almost identical to the value obtained by a least squares fit of experimental data 95).
The case of a moderately crosslinked network, where strong topological constraints
and the influence of the crosslinks have to be considered, could not be solved rigor-
ously 96, 97) For the concentration and the deformation dependences of d the same
relations as for the melt case were obtained. The numerical prefactor has to be fitted
to stress-strain data (see Sect. 4).
Uniaxial extension experiments yield the relation

do/1 "~ A(nsl3) -1/2 (20)

where the prefactor A ranges from 2 to 4. The wide range of values of the prefactor
results from the possible influence of constraint release processes (see end of this
section). It has to be added that a number of phenomenological considerations yield
relation (12) with different values of v. An affine deformation of the tubes (v = 1)
was suggested long ago by Edwards 7o). It gives the famous Mooney-Rivlin equation
[Eq. (1)] as recently rediscovered by Khokhlov and Ternovskii 98) Marrucci 74)
employed the model of constant tube volume. Box models with rectangular boxes
and a broad range of values of the parameter v (from v = --1 to v = + 1) have been
deduced from simple entanglement models by Thomas et al. 72, 7s). Similar models
have been successfully investigated by Gayiord 75-77)
Sketching a comprehensive picture of the topological constraints in the molecular
field approximation, we observe two characteristic features:
- The tube dimensions are of the order and mostly somewhat greater than the
-

statistical segment length.


-- The tube dimensions are deformation dependent.
This deformation dependence is the main source of the deviations of network properties
from phantom network results. Anticipating the results of the comparison of theory
with experiment in Sect. 5 it can be stated that the data for the tube dimensions
according to Eqs. (12), (15), (20) lead to a very satisfactory description of the uniaxial
stress-strain behaviour of dry rubbers. The scattering data, collected for end-linked
networks and problems with the explanation of data on biaxial extension as well as
the strong dependence of the C 2 term of the Mooney-Rivlin equation on swelling
suggest that the assumption of equality of macroscopic deformation and of the
deformation of the chains does not work in this case. This observation may be under-
stood qualitatively by considering that the time scale of the stress-strain experiment
is beyond the long time relaxation peak which is typical for moderately crosslinked
networks (compare with Ferry 99) and Havr/mek loo)). This means that constraint
release processes may take place and the assumption of an affine microscopic deforma-
tion )~mic ~- ~" which works very well in the plateau region, is not applicable.
46 G. Heinrich, E. Straube and G. Helmis

A rough quantitative explanation of all these phenomena can be achieved by the


ansatz

Lmlc = )~ (21)

The exponent [3 < 1 depends for example on the swelling degree and/or on the
number of network defects. Equation (21) resolves the contradictions observed for
swollen or imperfect networks (see Sect. 5.3). The mechanisms leading to Eq. (21)
are not yet completely clear. Some hints come from a paper of Gordon and Scho-
walter t01), who obtained relation (21) by treating elements sliding relative to deformed
surroundings.

3 Theory of Phantom Networks


This section gives a short survey of the statistical mechanics of phantom networks.
For a more comprehensive study of this field, the reader is referred to more detailed
reviews x- 5)
As outlined in Sect. 1, a phantom network is defined as a network with the fictitious
property that chains and junctions can move freely through one another without
destroying the connectivity of the network. Usually, models of rubber elastic networks
are built up from Gaussian chains and the topology of 6onnectivity is completely
described by the reduced Kirchhoff matrix of Eq. (6). However, Staverman 5) pointed
out that for a network with a given Kirchhoff matrix, the model has to be completed
by additional assumptions.
Three types of phantom networks can be distinguished: free phantom networks,
fixed phantom networks, and localised phantom networks 5). The first type is without
any constraints and will consequently collapse. The second type is a phantom network
with some junctions fixed in space. As a result, it is subjected to contraints that do
not really exist. The most natural phantom network model is the last variant, in which
the equilibrium positions of all segments are determined by suitable boundary con-
ditions without any need for segments or junctions being fixed.
The last two models have been investigated extensively by Flory 102) He considered
networks forming part of a larger network with fixed junctions outside the network
under consideration. The crosslinking points of the inner network were assumed
to fluctuate freely, or to have fixed positions in space. In both cases, the inner network
can be considered as localized by boundary conditions. In particular, Flory lo2)
considered a network formed in two hypothetical steps. First, a giant acyclic molecule
is formed by joining all chains via the available multifunctional junctions; such a
tree-like molecule can be characterized by v + 1 ~ v junctions plus chain ends
(i.e. number of labelled points). Figure 3 shows an example of the network after
the first step. In the second step ~ additional connections are formed by the reaction
of 2~ unreacted functionalities, which reduces the number of labelled points to
approximately N K - - {. The number ~ is called the cycle rank, which can be defined
as the number of connections that have to be cut in order to reduce the network with
cycles to a tree. Flory distinguishes two limiting cases: a) the free-fluctuation limit
and b) the limit of full suppression of junction fluctuations. These two limiting cases
Rubber Elasticityof PolymerNetworks:Theories 47

Fig. 3. Exampleof a network with a cyclerank { = 7

correspond to the two "classical" approaches to the phantom network problem, the
James-Guth (JG)Jo3} a) and the Hermans-Flory-Wall (HFW)lo4-1ov} b) theories.
In case a), the mean values of the chain end-to-end vectors are displaced affinely
with the principal extension ratios ~ (la = x, y, z) specifying the macroscopic strain.
The fluctuations about these mean values are independent of the sample deformation.
Consequently, in the free-fluctuation limit, the transformation of the actual chain
vectors is not affine in the L's. The elastic free energy change for deformation results
in the expression

1
AF = -~ ~kT ( ~ ) ~ - 3) (22)

Equation (22) holds for phantom networks of any functionality, irrespective of their
structural imperfections. In case b), fluctuations of junctions are assumed to be
suppressed fully. The junctions themselves are considered to be firmly embedded in
the medium and their position is transformed affinely with the macroscopic strain.
This leads to the free energy expression for an f-functional network possibly contain-
ing free chain ends

~kT In (~---~) (23)


(2 f- 1)

In the perfect network, the cycle rank ~ is connected to the number of junctions M
and the number of chains by the relation

~= NK--M

where the number of chains is given by

1
N K = ~ Mf

The second term ,-~ In ,~x~,y~z in Eq. (23) represents the contribution of the entropy
of mixing of ~(f/2 -- 1) -1 junctions over the volume V relative to the reference
48 G. Heinrich, E. Straube and G. Hehnis

volume Vo. The validity of this term has been the subject of controversy in the lite-
rature and seems to be an artefact of the phantom concept. Some insight may be
obtained in the following way. If one uses the analogy between the path-integral for-
mulation of polymer configuration and Brownian particle motion 46), a statistical
thermodynamics of rubber phantom networks may be developed which is equivalent
to the Gibbs formalism for conventional statistical mechanics 48). In that frame, it
may be noted that the existence of the logarithmic term may be connected with assump-
tions about the polymer density profile in rubbers 6o) If one describes the phantom
network as a long phantom chain in a box (NL1 > V2/3), a choice between cyclic
boundary conditions and zero density conditions outside the walls may be made
when solving the equation for the statistical weight of the chain configurations.
Deam and Edwards 6o) used the replica technique to analyse both cases a) and b)
with regard to the existence of the logarithmic term. The original H F W treat-
ment lo4-1o7~ [case b)] assumes that the crosslinks are fixed in space and are randomly
distributed over the volume of the rubber. This is an even-density assumption corre-
sponding to cyclic boundary conditions for the Green's function of the network which
leads to expression (23). This means that each crosslink eliminates one degree of
freedom and the volume VoU k is available for all chains, Deam and Edwards 60)
pointed out that this is only correct if the system has not gelled and, consequently,
the volumeV0 l-I k, is available to each chain. In a strong gel, however, the chain

segments are localised so that the available space is determined by the crosslinking
density and segment number density.
In contrast, the original James-Guth treatment t03) [case a)] assumes that there
are two types of crosslinks, one type is fixed at the boundary of the rubber and the
other is free to move inside the volume, In the path integral approach 6o) of this
model, a density distribution with the polymer piled up at the centre of the box results
as a consequence of the zero-density boundary conditions outside the walls. Then
the free energy expression no longer contains the logarithmic term and leads to
Eq. (22) with ~ = M for f = 4. The two approaches may be interpreted as Fourier
terms of the polymer density where the H F W theory includes a k = 0 mode whereas
that of JG does not.
Of course, phantom chains are not realistic, and intra- and interchain forces favour
a uniform density distribution. Therefore, one must use a model which incorporates
the uniform density assumption but allows the junction to fluctuate about its affinely
deformed mean position, as is realised in the phantom case a).
The localisation of crosslinks can also be accomplished by introducing a harmonic
constraining potential 47'6°'83"84), where its strength is a variational parameter
which is found to be proportional to crosslinking density and independent of the
strain condition. Apart from a wasted-loops correction factor, the elastic free energy
results in an expression like formula (22).
It should be noted that the discrepancy between expressions (22) and (23) is revealed
only in deformation accompanied by volume changes (e.g. swelling). Otherwise,
under conditions of nearly constant density the incompressibility condition N ~', = t
holds and the contribution from the logarithmic term vanishes.
Rubber Elasticity of Polymer Networks: Theories 49

4 Theories of Real Networks


4.1 Non-topological 'qnteraction" Concepts
The shortcomings of the phantom network concepts have stimulated a number of
attempts to find theoretically a more satisfactory elastic free energy function to
describe the properties of elastomers at different states of deformation. The efforts to
explain the real network behaviour by a special non-topological mechanism can be
divided into four types. The first group considered intra- and intermotecular ef-
fects 108-111). With these assumptions it is hard to explain values of the C z Mooney-
Rivlin parameter which are of the order of the corresponding C 1 parameter.
The second group thought that the packing conditions or the short-range order
and structure effects in amorphous polymers could provide an explanation 112-119)
The ordering effects in polymer networks can be a useful means of discussing particular
structure-property relationships in some special polymers. However, the ordering
effects cannot be generally operative in the majority of polymer networks.
Another group tried to explain the Mooney-Rivlin-like stress-strain behaviour
by different non-affine network chain deformation mechanisms 118-120)
The models in which the departures from classical network behaviour are assumed
to be caused by the compressibility of the dry polymer 1zl) or by the dependence of
the mean square end-to-end distance of the "free" chain on the intemal pressure 122)
fall into another category. These theories fail to explain the observed order of magni-
tude of the C z term, or the proposed mechanism seems to be a very special one. A
detailed discussion has been published by the present authors in an earlier review 123~
More recently, Kilian et al. 124-126) have treated the stress-strain behaviour of
rubber with the aid of the van der Waals equation of state. Without assuming any
special microscopic model, this approach proposes the picture of a conformational
gas with weak interaction. The van der Waals parameters are related to finite chain
extensibility or global interactions between the chains. Both parameters can be satis-
factorily fitted to the experimental data. In the first attempt to interpret the van der
Waals parameters on the basis of an appropriate microscopic model the chain ex-
tensibility parameter was identified with parameters of the primitive path model
proposed by Edwards 82)

4.2 The Theory of Restricted Jtmction Fluctuations


The concept of restricted junction fluctuations was proposed in papers by F10ry
and Erman 41-44), K~istner 36-39), and Ronca and Allegra 40~. The main assumption
is that the fluctuations of the network junctions about their mean positions are re-
stricted by topological interactions, but the chains themselves can be handled as
phantom chains in the configuration space of the network, i.e. they do not feel any
topological constraints. Flory uses the term "entanglement" to denote the inter-
spersion of chains and junctions in ways that render them inseparable. This term
includes both the effect of entanglements and steric constraints. This view is quite
different from the models in which the action of entanglements is completely equiva-
lent to a chemical crosslinkage in its effect on the equilibrium stress. Flory and
Erman 4t -44) assume that all constraints directly affecting the backbone of network
50 G, Heinfich, E. Straube and G. Hehnis

chains may be subsumed in the restrictions imposed on junctions. To clarify the


foundations of the model and to facilitate a comparison between this model and the
tube model described below, the model of restricted junction fluctuations will be
discussed from a more general point of view.
To focus on the mean field approach (see Sect. 2), the statistical weight of any con-
figuration of the network may be expressed as a product of the contributions of the
(restricted) network chains. Starting with (in our opinion) more natural models for
restricted chains, a model with restricted junction fluctuations can be derived by inte-
grating over the configuration space of the network chains. The constraints acting
on the network chains are transformed in this way into a mean force potential acting
on the junctions. This mean force potential is approximated within the model of
restricted junction fluctuation by the domain accessible to each junction.
The so-called domain of constraints is assumed to be cubical am or spherical 41).
The degree of constraint on fluctuations is affected by the degree of deformation. All
centres of domains are distributed with respect to the mean positions of the junctions
in the phantom network and the mean positions of the domains are assumed to trans-
form affinely with the macroscopic strain. Further, the shapes of domains are assumed
to transform non-affinely due to some relaxation of the constraints. The theory
results in an elastic free energy change

1
AF = ~ CkT[(I1 - - 3) + I* + I*'1 (24)

where 11 = ~2 X~ (first strain invariant) comes from the phantom result (see Eq. (22)).

The terms I* and I** take the forms

M
I* (1 + g~) B~ (25)
¢

I** = - M Z In [(1 + B,) (1 + g,B,)]

where

B = ( X -- 1)(1 + L --~2)/(1 + g.)2

Here, × and ~ are material parameters; × is proposed to be proportional to the degree


of interpenetration of chains and junctions, and it defines the strength of restrictions
on junction fluctuations. The value × = 0 corresponds to the free-fluctuation limit
and for × -~ o0 the affine network is obtained. The parameter ~ ( 4 1) characterizes
departures of the shapes of domains from the affine transformation assumption
and reflects the effect of structural inhomogeneities on the network structure 43)
Although its presence is not as critical as that of ×, comparison of experiment with
the theory of restricted junction fluctuations shows that it is necessary 2~
Rubber Elasticity of Polymer Networks: Theories 51

From Eq. (24), the reduced stress

~(~.) %kT
13"M ~. (~%-)% - 2 ) - V0 (I + ATOM) (26)

results, where AcyM comes from the strain dependence of I* and I** (its analytical
expression is not given here). At large extensions and at high degrees of swelling zk~M
vanishes. The relation G = lira aM(L) yields the small-strain modulus, which in the
case of perfect networks can be written as

Nk M (27)
G = ( v - - hPc) k T , vc = V - ' lac = V "

The parameter h, which was introduced by Graessley 27), correlates with the para-
meter ×; h = 0 corresponds to complete suppression of junction fluctuation and
h = 1 corresponds to the free fluctuation limit. Intermediate values account for
intermediate junction fluctuation.
Equation (27) shows the main limitation of the theory of restricted junction fluc-
tuation. The topological contribution to the modulus is limited by the value of the
fixed phantom network. As an example, the relation C z > C 1 is impossible in the
case f = 4. On the other hand, experimental studies on networks produced by end-
linking of chains have been interpreted successfully within the theory of restricted
junction fluctuation 2o-z3}. As already discussed in Sect. 2, this may be explained as
the effect of the relaxation of the topological constraints (see also Sect. 5.2).

4.3 The Langley-Graessley Concept

In the small-strain limit Graessley and co-workers 25-28) extended the model of
restricted junction fluctuation by adding to Eq. (27) a contribution tbr so-called
trapped entanglements to explain the large topological contributions to the shear
modulus of networks produced by sulphur, peroxide or radiation crosslinking 89)
of chains:
G = (v -- hla¢) kT + GeT e (28)

The second term is taken from the Langley theory 89). It describes the topological
chain-chain interaction and is said to arise from permanently trapped entanglements
in the network. The Langley concept assumes that during the crosslinking process
a portion of two chain entanglements is trapped and acts as a physical junction.
Here T denotes the trapping factor of Langley ag) which can be viewed as the probabil-
ity that a pairwise entanglement of network strands is elastically active. It has limiting
values of zero at the gel point and unity for a perfect network without dangling ends.
The maximal modulus G e = eRT is assumed to have a value close to G °, the plateau
modulus of the corresponding linear polymer system before crosslinking. This con-
clusion was drawn from experimental data 27-32,99,127). It is supported by the result
of Sect. 5 that the topological contribution to the total modulus has the same universal
52 G. Heim'ich,E. Straube and G. Hehnis

dependence on the chain density as the plateau modulus G ° 92,96). Usually e or G e


is treated as a fit parameter, while T e is the central quantity of this approach.
Along with T , further structural parameters of networks, e.g. the sol fraction, w ,
and the concentration of elastically active network chains in the gel, v g = v J(1 -- w),
can be calculated using the theory of branching processes based on cascade sub-
stitution 2"118'I28-132t The theory of branching processes uses the formalism of
probability generating functions to describe the distribution of reacting units in
different reacting states, and it seems to be the most fruitful method today 133~. Its
usefulness has been exemplified in a number of problems (e.g. Refs. 128-i31) referring
to polyurethane model networks and epoxy resins), including the often neglected
effects ofcyctisation (e.g. Refs. ~34-136)). A detailed description of the theory is beyond
the scope of this review.
It should be mentioned that recently Dossin and Graessley 27) used equations
developed by Pearson and Graessley 13v) to derive asymptotic expressions for v , g~
and T in randomly crosslinked networks where the sol fraction is negligible. In this
case, nearly all crosslinks will have two or more paths leading to the network. Under
these conditions the expressions for structural parameters become relatively simple 2v).
Values for v , gc and T of end-linked PDMS networks have often been calculated 6,25,
26,29, 31,32) from the sol fraction data by using the recursive method of Miller, Macosko
and Vall~s 13s-140)
A few further points concerning the physical basis of the Langley-Graessley concept
should be noted.
a) The Langley-Graessley concept is a phenomenological one and does not give any
information on the large-strain dependence of the topological contributions to the
free energy. Consequently, a separation of crosslink and topological contributions
using the different /brms of their stress-strain behaviour is impossible within this
concept. Any estimate of the topological contribution depends on the assumptions
made about the relation of the network structure and the contribution to the modulus
due to the chemical crosslinks. The Flory term, - - h~tkT in Eq. (28), is an example
of this phenomenological choice.
b) The Langley-Graessley concept seems most suitable in the case of networks where
the stress-strain behaviour is nearly phantom-like but the network modulus is in-
fluenced by entanglement effects. This is especially the case for incompletel3; cross-
linked networks (cf. Sect. 2). There, the trapping factor T takes into account the
dependence of the entanglement contribution on the topology of crosslinking, includ-
ing network imperfections.
c) Comparing the Langley-Graessley concept with the tube model outlined in Sect. 2,
the picture of trapped entanglements may be viewed as a special case of the mean
field concept of topological constraints 90-92) The tube model is mainly based on
the assumption of dominating local constraints. It fails increasingly if constraint
release effects allow local constraints to relax. Thus, it may be expected that the
remaining global entanglements can be described successfully with the help of the
trapping factor. This suggestion is supported by a conclusion of Queslel and Mark 6~
who treated experimental data of various authors on end-linked PDMS networks
by means of a modification of the usual Langley-Graessley plot. They concluded
that the mean distance between trapped entanglements is large (8t00 g/mol) and so
it can be only one or two entanglements per network chain.
Rubber Elasticity of Polymer Networks: Theories 53

Some further relations'between the Langley-Graessley concept and the tube model
will be discussed in Sect. 5.

4.4 The "Single-Chain" Approach

A number of theoretical models use a single-chain approach to simulate topological


constraints in real polymer networks. The basic idea is that one starts from the sta-
tistical mechanics of a single network chain which is subjected to a spatial domain
of constraints. The constraining potential is introduced in a heuristic manner and
cannot be calculated within the frame of the chosen model self-consistently. Hence,
the strength of the topological interaction must be characterized by best-fit para-
meters of the model.
We distinguish three ways to simulate the configurational confinement of network
strands. In the slip-link model 79,80), each strand threads its way through a number
of small rings. Such an approach has also been used successfully in calculating the
viscoelastic properties of polymer melts 94). The topological contributions are caused
by the orientation of the subchains between the slip-links, i.e. the slip-link model
may be called an alignment model.
The second approach employs simple tube or channel models in which each strand
is confined within a configurational channel. Different assumptions have been made
concerning the deformation dependence of the tube dimensions 72- 77,8l ) For example,
Thomas et al. 72,73) analyzed the configurational properties of a three-chain network
model with each chain confined within a rectangular cylinder. The deformation de-
pendence of the lateral dimensions of the cylinder was discussed separately in the two
cases of the action of entanglements and of packing effects. As in most of the three-
chain models, in both cases one-third of the cylinders were assumed to be orientated
along one of the principal directions of strain, and the end-to-end chain separation
changes affinely. The channel cross-section is affected by the deformation dependence
of the local constraints. A simple picture of the action of pure entanglements leads to
an expression like Eq. (12) with v = --1. Packing effects result in v = 1. Gaylord vv)
used a similar model with the value v = --1/2.
The third version starts from the notion that a basic topology is fixed in the rubber
and primitive paths for all network strands are established s2) when crosslinking takes
place. The primitive path can be regarded as the shortest path between chain ends
that does not violate topological restrictions. On deformation, this path is lengthened,
and this process increases the free energy relative to models in which the topology
has been ignored.
Gottlieb and Gaylord 78) compared to experimental data the predictions of several
constraining models (including the theory of restricted junction fluctuations) of rubber
elasticity for the uniaxial stress-strain response (extension and compression) of cross-
linked networks. They found that the slip-link model so) does not describe the elastic
behaviour satisfactorily. Further, it could be demonstrated that uniaxial strain ex-
periments are not sensitive enough to distinguish between the models. An example
of this fact is the results of the modification of the slip-link model offered by Graess-
ley ~9) who introduced the concept of surplus segments into this model. There, by a
proper choice of the extent of surplus segments a good agreement with the data of
54 G. Heinrich,E. Straube and G. Hetmis

uniaxial stress-strain experiments was achieved. In Sect. 5, the problem of ambiguity


will be discussed in more detail when comparing different uniaxial stress-strain rela-
tions of real network theories with experimental results by Kawabata et al. 141) It
follows from this analysis that other modes of strain (e.g. biaxial 73)) and solvent
effects need to be studied in order to test the network models more profoundly.

4.5 Network Theories Based on Topological Invariants

In all network theories based on topological invariants (see Sect. 2), it has been assum-
ed that the retractive force exhibited by polymeric networks comes not only from the
entropic forces generated in deformed elastically active chains, but also from topolo-
gical interactions among entangled molecules.
A pioneering approach was taken by Deam and Edwards 60). Using Eq. (7) and
introducing entanglements by adding an extra constraint such that the topological
invariant Eq. (.9) be conserved, the configurational integral for a given set of M cross-
linking points contains the Dirac delta functions 5(I(R m~) -- I(R)):

N M L L
Z = j" e -H/'zF l~ l l Y J" 8(R,(s,)- R~(s',)) ds, as I
(M ~<~ i = 1 0 0

N
x l~ [I,,(R(°)) - I,a(R)] l l DR~(s)

H = - --
21 ~=1
0

where R ~°~denotes the undeformed configurations and DR(s) indicates the measure
of functional integration. The ensemble averaging has been performed by using the
replica method (cf. Eq. (5)). This technique will be described in Sect. 4.6 in more
detail. Further, the expression 6(AI) has been rewritten as

(sin (AI/2)]
exp [In 8(AI)] _ In \- ~ -j

The averaging of all the accessible configurations was performed by introducing


collective variables and expanding the sine, where it is argued that many entangle-
ments are present and each contributes a small effect, Further, the effect of the topo-
logical invariants is absorbed in a variational parameter which determines the strength
of a trial harmonic molecular field potential. The same trial potential is used to simu-
late the contributions of the chemical crosslinks. Further, the system is assumed to
be in a state of complete thermal equilibrium with all states accessible and the topology
of chain connectivity by chemical crosslinks is frozen at some initial time. The cal-
Rubber Elasticity of Polymer Networks: Theories 55

culation of the elastic free energy as the average over all possible structures produces
the simple phantom term plus contributions from entanglements:

F(~.)=kT{M~-~ - N ~ C(nfl3)
Z(~-~2). + 1

+(6C)3,2N~(n~13)l/2I~(~_~
• )]3/2} + 1 + F'(~.) (29)

Here C is a numerical constant --~1 and F' denotes an additional contribution which
is different in the two cases of highly and slightly crosslinked rubbers. In Eq. (29)
N s is the total number of segments.
Highly crosslinked rubber:

C2kTI
F'(;~) - - - Ns '(nsl3)(~ ~ ; 2 ) [ I - ( ~ k~2) 2] (30)
31c

Lightly crosslinked rubber:

x[~(~.22 + 1)](1-)~-2)} (31)

(1 is the cutoff length ~ 1).


In the case of highly crosslinked rubbers, the topological contributions to the elastic
free energy according to Eqs. (29) and (30) depends on the segment density and is
independent of the crosslink density. This is a surprising result because this independ-
ence of topological contributions ofcrosslink density would be more expected in the
opposite case of slightly crosslinked systems made by crosslinking primary chains
of high molecular weight. This expectation follows from the observation that the
topological contribution G N to the total modulus (see Sect. 5 and Figs. 10 and 11)
of such networks hardly depends on the contribution of crosslinks.
Iwata 67,142~ extended the Deam-Edwards model by starting from a complete
basic subset {I'~} e {I~} as a set of independent topological invariants of the system;
{ I j denotes the set of all I~ defined for all the possible pairs of loops in the network.
Considering a network as having a simple cubic lattice regular connection pattern,
an affine relative displacement of centres of mass of the cells under macroscopic
deformation is assumed. The calculations have been performed numerically. The
elastic free energy obtained is mainly determined by a topological term due to the
topological interaction between the strands. The classical term plays a minor role
only. This result agrees qualitatively with the approach presented by Graessley and
Pearson 143), who calculated the stress-strain relations of an ensemble of polymers
with one subset of simply entangled pairs of loops and a second subset of non-ent-
angled pairs of ring molecules,
56 G. Heinrich, E. Straube and G. Helmis

The Iwata model probably overestimates the entanglement contribution to the


rubber elasticity. This conclusion results from the observation that the topological
contribution to the network modulus and the contribution arising from chemical
crosslinks are approximately of the same order of magnitude 26-30). A statistical
mechanical treatment of a long closed strand entangled with an elastic background
network has been performed by Brereton and Filbrandt 144) using the topological
invariant Eq. (9) averaged over all background configurations filling a macroscopic
box with constant segment density. The contribution of the entanglements to the
modulus is found to be governed by the ratio of the winding number, m, and an equi-
librium winding number, (m2) m, giving the relation C 2 ~ (1 ...... mZ/(m2)). A positive
C 2 term of the Mooney-Rivlin equation has been interpreted as meaning that these
systems are under-entangled with respect to ( m 2 ) 1/2. This approach may be fruitful
as the starting point of a theory for networks consisting of cycles trapped by linear
chains which passed through them prior to being end-linked 14~- ~46~or of networks
having no crosslinks (e.g. "Olympic" or "chain mail" networks). But at present the
approach of Brereton and Filbrandt 144) is troubled by a number of approximations,
especially concerning the assumption of the affine deformation of the network strands
with the deformation of the sample.
A very interesting way to explain i]eviations from the phantom network theory is
the approach proposed by Ball et al. ssl, who modelled the topological constraints
by sliding links which make contacts between the network chains. Formally, this
approach is based on the Deam-Edwards concept. Assuming that the chemical
junctions are free to fluctuate about their mean positions the calculation of the elastic
free energy change leads to the expression

- 311 + In (1 + q)]} (32)

where N I denotes the number of sliplinks andq is a relative measure of the freedom
of a link to slide. Forrl = 0, the expression (32) predicts the usual free energy change
of a phantom network containing M chemical junctions plus Ns~ entanglements.
Ball et al. assumed that each sliplink can slide over distances comparable .with the
spacing of the centres of its topologically neighbouring links. Minimizing AF results
in q = 0.2343.
Edwards and Vitgis extended the model tb the case of the high deformation limit 147)
Thirion and Weil 148) considered rt as an adjustable material parameter which is
obtained by comparing the theoretical expression for various types of strain with
experimental data. The investigation of the connection between Nst and actual
structural characteristics of networks (number of effective strands, chemical junctions,
fraction T e of trapped entanglements calculated according to the Miller-Macosko
theory 13s)) ted to the empirical relation Ns~/V = G°Te/kT. Such a result agrees
qualitatively with the authors' expression for the topological contribution G N of
the shear modulus, as pointed out in (Sect. 5) of this review. There, an effective
Rubber Elasticity of Polymer Networks: Theories 57

number of physical junctions - - representing the topological constraints - - can


be defined according to N,op/V = GN/kT, where G N contains the parameter 13(0 < 13
_<_ 1) which serves as a measure describing the differences between microscopic and
macroscopic deformation ratios (see Sect. 5 and Eq. (21)).
Queslel 149) combined the Ball model with the Flory model of restricted junction
fluctuations to explain the origin of junction and trapped entanglement contributions
to the total modulus exhibited by networks. There, the total contribution to the
reduced force is considered to be the sum of the full Ftory term (see Sect. 4.2) and the
entanglement contribution of the Ball term. Such an expression has the practical
advantage that it is valid over the entire range of deformation, but it seems to be an
artificial construction.
Summarizing the approaches reviewed in this section, it has to be concluded that
to date the free energy cannot be deduced by a rigorous statistical mechanical theory
starting from basic assumptions about topological invariants.
A new, valuable type of experiment are computer simulations of network proper-
ties. Computer simulations are welt established both in equilibrium and non-equi-
librium physics of systems of linear chains and have been used for the study of net-
works and melts. The effect of the topological constraints on the stress-strain beha-
viour of tetrafunctional networks with a regular structure has been investigated
by Elyashevich and Remeev is0,151) using Monte Carlo methods to generate Marko-
vian chains on a regular lattice. The simulation procedure assumes that transitions
between configurational states that cannot occur without chain intersection are
forbidden. The results show for uniaxial extension and compression the usual "ex-
perimental" behaviour described adequately by the Mooney-Rivlin equation with
finite C 2 values in the case of extension and very small C 2 values for compression,
Computer simulations of network properties which have to be extended to other
types of deformations and to be confirmed by different methods of simulation may
be of particular importance for network theories because models close to the models
investigated in theoretical work are used in these simulations. They offer a basis for
the study of both network formation processes and the properties of polymer net-
works.

4.6 The Tube Model for Randomly Crosslinked Networks

Using the harmonic molecular field picture (Sect. 2.2; Eqs. (13) and (14)) to simulate
the local topological constraints in polymer networks, the elastic free energy of a
randomly crosslinked network may be written as

.~=NL

F()~)=-kTffP(l~(s)l°glfexp(-H/kT)[D~i'~'~'(-Ro.(s)]}o
M 1
X [D~i, ~[}(/~(s))] I-I ~ dsl ds~ (33)
i=l
58 G. Heinrich, E. Straube and G. Helmis

with the Hamiltonian


£,o .~

H/kT - 2 .31
o
f /~2(s) ds + ~
f'
0
~ JR,(s) - )~,l~(s)] 2 ds (34)

and the measures

D~si,s~/(R(s)) - DR(s) l-[ 6(R(s0 - R(s'i)), (35)


i

Di~i,~/~(/~(s)) - D/~(s) [ I 8(R(sl) -- R(s;)) (36)


i

(8: Dirac delta function).


In Eq. (33), the network may be envisaged as a giant macromolecule of contour
length £a = NL. A uniform polymer density will be produced by an appropriate
distribution of the R(s). Further calculations of the elastic free energy are performed
by using the replica technique sketched in Sect, 2. In the first step, a set of n + 1
coordinates R (~) (~ = 0, 1. . . . . n) is introduced where the R ~°) are identified with the
mean constraining tube axis configurations in the undeformed sample. The set
{R(~)}, ~ = 1. . . . . n, corresponds to the n deformed sample copies. It is beyond the
scope of this review to explain the calculations in detail. For the basic steps of the
replica technique, the reader is referred to Ref. 60) where the concept is examined in
the frame of phantom network theories. Similarly, in our work a unitary transforma-
tion has to be introduced to single out the "centre of mass" coordinate from the relative
variables of the (n + 1)th system. The transformation does not alter the expression
for the configurational integral and makes it possible to simulate the action of chemical
crosslinks with the same harmonic potential term as has been done for the action of
local topological constraints. The strength of such an additional potential is calculated
using variational principles ls2) and minimizing the free elastic energy expression.
Finally, the functional integrals are directly integrable by using the corresponding
Green's functions of the undeformed and deformed replica systems. The first Green's
function (the solution of the free diffusion equation inside a box of volume
V [ I (1 + n)~) 1/2 for cyclic boundary conditions) is approximately a constant term

representing the mean density of chains. It is mainly responsible for the classical
(phantom) part in the final elastic free energy expression. The second Green's function
corresponds to the Green's function of the Schr6dinger equation of the harmonic
oscillator. Finally, one obtains the expression

F(k) ~ kT ~ )~ - ~, In (*if2) + X w. (37)

with

Wl~2 = W~2 + 6/d~ (38 a)

6M { [ (6Md~-2qt/2~1/2
(38 bl
W~N-i~ 1+ I+\-Ni-L-/ J J
Rubber Elasticity of PolymerNetworks: Theories 59

w is the contribution due to the crosslinks. It has been calculated by using variational
principles 15z)and minimizing the free elastic energy. In the case of slightly and mode-
rately crosslinked networks (strong topological constraints, 1Lc/d~ >> 1), one gets the
expression

w~ ~ •/M 1
Iq-i-L d.
(38 c)

In the contrasting case of highly crosslinked networks (M/N ,> 1) the expression
(38 d) results:

w = w = 6M/(N1L) (38 d)

d is given by Eqs. (11) and (15) or (20).


Although the replica technique has been successfully used in the physics of disorder-
ed "quenched" systems, it is mathematically ill-founded and examples are known
where the replica trick does not give physically sound results. For example, it has
been found 153) that an Ising spin glass with infinite-range interaction when treated
using the replica trick acquires a negative entropy at low temperatures. This behaviour
might arise from the interchange of the thermodynamic limit and the limit n ~ 0.
In a recent paper 15,~, it could be clarified that problems with the replica trick may
occur whenever the theory for a general integer value of n does not have the same
symmetries as the theory for n = 0.
Considering the deficiencies of the replica method mentioned above, independent
approaches are highly valuable if they are able to treat models which are physically
identical or equivalent to the tube model Eq. (33). One method is the method of
stochastic fields O(r(s)). The main idea consists in simulating the topological con-
straints which have been modelled in Eq. (33) by the restoring potential ,-~r(s)2
[with r(s) = R ( s ) - ~A~(s)] by a more general stochastic potential qb(r(s))84). The
configurational integral or any other average of interest is then obtained by averaging
over the stochastic potential:

(..-)0 = I DO(r) p(O(r)) ... (39)

Performing this procedure with the Gaussian distribution

P(~(r))~expI--fffd3rd3r'@(r')q~(r,r')-l~(r)l
1
(40)

leads to a partition function which contains in the Hamiltonian (Eq. (34)) the restoring
potential

~¢, .La
f ds f ds' qo(r(s)-- r(s')) (41)
0 0

where q~(r -- r') is the correlator of the stochastic field. The functional integral
Eq. (33) can be evaluated in closed form by means of the Green's function of the
60 G. Heinrich, E. Straube and G. Helmis

memory oscillator 84) if two different but physically equivalent assumptions are
introduced :
a) The potential Op(r) is expanded in a Taylor series up to second order:

* ~ ~ [A.r~(s) + B~r~(s)2]

(i.e. ~ is replaced by the potential of a harmonic oscillator under the action of a con-
stant force). Further, the A are assumed to have a Gaussian distribution whereas
fixed values B~ = 1/d~ (cf. Eq. (34)) are chosen for the oscillator strength in the
deformed system.
b) The correlator ~p(r(s) - - r(s')) is expanded in a Taylor series up to second order.
The influence of the crosslinks has been considered, as in the replica case above,
by an additional term added to the oscillator strength, which was determined by a
variational method.
In both cases an expression for the free energy results which is identical with
Eq. (37) derived by the replica method 84) From the congruent results obtained by
different theoretical methods for physically equivalent models, it can be concluded
that Eq. (37) is representative for network models with harmonic constraining poten-
tials. Equation (37) will be used extensively in all further discussions concerning the
properties of the tube model of polymer networks,
If one reduces the free elastic energy expression of Eq, (37) to the special case of
strong topological constraints (1LJd 2 ~> 1) and to an external mechanical deformation
of an incompressible sample, the elastic free energy change reads

F(,)-- 3) + 5 - - 3 (42)

1 IL~ 1 IL~
with F - and =
dg d2
V6 V6

Here F serves as a dimensionless measure of the strength of the topological constraints.


The free-fluctuation phantom result follows from Eq. (42) in two ways. Firstly, in
the limit F = 0 (do ~ vc), that is, for negligible topological constraints, this limit
has to be viewed as the phantom result in the true sense. Secondly, in the case of
deformation-independent tube dimension d r = d o the free energy differences also
simulate phantom chain behaviour. Using the relation

d =do~ (43)

a wide variety of relations between tile crosslink and the topological contributions
to the network moduli and of the stress-strain behaviour can be obtained. As already
discussed in Sect. 2, the parameter v describes different constraint mechanisms
while the parameter 0 < [3 < 1 takes into account different extents of constraint
release processes.
Rubber Elasticity of Polymer Networks: Theories 61

It is worth noting that the elastic free energy according to Eqs. (42) and (43) can
be deduced from a phenornenological point of view using the two-network picture.
The elastic potential w~ = AF/V is assumed to consist of two additive contributions

Wel---..,----Gc]~1(2 ) +-29"rll(n)
n
(44)

The phenomenological parameter n is then related by Eqs. (42) and (43) to the expo-
nents v and [3 by n = --2vl3. ~l(n) characterizes the first invariant of the generalized
deformation tensor

b , = -(B."/2
v - ~,,)/n (45)

where B v denotes the Finger's deformation tensor

~x~(t) 0xv(t)
(46)
B,,,, = ~ 0x,(O) ~x,(0)

In Eq. (46), the {x(t)} are the spatial (deformed) coordinates and the partial differ-
entiation is performed with respect to the material (undeformed) coordinates. Ex-
pression (45) was introduced by Blatz et al. 1557and Ogden 156), independently, who
adopted the idea of a generalized strain measure to predict stress-strain relations of
crosslinked samples of elastomers under various types of deformation. Representa-
tion in the principal axes system yields for the components of the Lagrangian stress
tensor 157)

b = ()~ -- 1)/n (47)

where the ;~ (g = x, y, z) are the principal stretch ratios. The strain invariant then
takes the form

tl(n) = ~ b. (48)

As an illustration of what has been accomplished with the tube model, some
examples of stress-strain relations resulting from Eqs. (42) and (43) will be presented.
It has already been mentioned in Sect. 2 that the simplest assumption, affine de-
formation of the tubes d = d0X., yields the Mooney-Rivlin equation (1). The value
v = 1/2 was obtainetl by a microscopic model 92,96) which is briefly discussed in
Sect. 2. It is suitable for the description of moderately but almost completely cross-
linked networks (e.g. sulphur-, peroxide-, or radiation-crosslinked NR, PB and
PDMS chains of very high degree of polymerization).
The stress-strain relation for uniaxial deformations of an incompressible sample
(~"x = ~'' ~y = )t*Z ~-- ~--1/2) has the form

~(~)
°M ---)~_ ~-2 -- Gc + GNf(k) (49)
62 G. Heinrich, E. Straube and G. Helmis

with

2(~ ~/2 -- L - ~)
f(~) - 13(~2_ ~_~) , f(1) = 1 (50)

Here

~(;~) =

denotes the nominal stress and c~Mis called the reduced or Mooney stress.

M
Gc = -V kT (51)

is the shear modulus contribution caused by chemical network junctions.

1 1 1a
GN = ~_ J3~ns kT (52)
4 V6 do2

(where n S is the segment number density and t is the length of the statistical segment)
can be considered as the contribution to the total shear modulus G resulting from
topological constraints, where the total modulus is given by

G - lim erM = G c + GN (53)


~,~1

In Sect. 5.2, the connection between G N and the plateau modulus G ° of the non-
crosslinked sample will be pointed out (G~ ~ G°).
The case v = --1/4, [3 = 1, obtained under assumptions valid for high crosslink
densities, results in the uniaxiat stress-strain behaviour

% = G c + G;~i'(~0 (54)

with
1 1 12
G~= 8 1/~66 n~ d~ kT (55)

and
()~l,,z _ )~- i/4)
f'QQ = 4 k2 ~_ 1 (56)

It should be noted that Tschoegl us) introduced the elastic potential Wel leading to
Eq. (54) phenomenologically to obtain reasonable agreement with experimental
uniaxial stress-strain data on some natural and poly(butadiene) rubbers.
Rubber Elasticity of Polymer Networks: Theories 63

The value v = --1 was obtained by a simple entanglement model by Thomas


et al. 72,73~. The limiting case v = --1, 13 = 1 describes a network with the modulus
G = G c + G N (see Eq. (42)), but with a phantom-like stress-strain relation; 13 = 0
obviously yields a phantom network with G n = 0. In the intermediate range 0 < [3 < 1,
deviations from the phantom behaviour with G N ~ 132 are observed.
The last examples confirm that in the case o f nearly phantom-like stress-strain
relations additional information or assumptions are necessary to obtain reliable
results on the topological contributions from stress-strain measurements, because at
present no microscopic theory o f the constraint release parameter is available.
To elucidate the different possible ways o f interpreting constraint mechanisms,
it seems worthwhile to consider the connection between the parameter 13 and the
trapping factor T e o f Langley. I f one compares Eqs. (19) and (52), the following
relation is obtained:

G N - 0.1313272G° (57)

It ~ rtmelt/rlnetw°tk
~a 0 /u 0 denotes the ratio of the lateral tube dimensions in the melt
and in the corresponding network. Equations (17) and (20) yield 7 ~_ 2 to 4 which
leads to the relations G N - (0.5 - - 2) [32 . G ° and T e ~ 132. These considerations
differ somewhat from the conclusions o f Gaylord 77) and Marrucci 74) who assumed
the relation 7 z = T and, hence, 3' < 1. Without discussing constraint release effects
they expected that a dilution effect o f network defects would lead to a cross-sectional
area o f the tube in the network larger than the cross-sectional area of the tube surround-
ing the same chain in the non-crosslinked system. In contrast, our results show an
additional effect o f the crosslinks to the strength of the topological constraints and,
consequently, 7 > 1. This contradiction is solved by setting ~tC,yJord/~,..... i = yl3.
The relation T ~ 132 has already been assumed to reflect the influence of network
imperfections on mechanical properties ~59) Experimentally, this assumption can
be confirmed by comparing the parameters C 2 of end-linked and randomly crosslinked,
but otherwise identical, networks. To the authors' knowledge, such data are not
available. Otherwise, a qualitative confirmation may be obtained from the reported
comparison 16o) o f uniaxial compression moduli and extractable fractions o f poly-
(butadiene) networks prepared by three different types of crosslinking reaction (the
crosslinking reaction is either an end-linking mechanism if the polymer precursor
is a telechelic material, or a usual addition mechanism on the double bonds along the
polymer backbone, or both mechanisms simultaneously). In this way, networks
with different contents o f dangling ends are obtained. It is found t60) that the last
type with few dangling ends has an "aromatic"-type structure and the highest modulus,

5 Results and Comparison with Experiments

In the foregoing section a number of theories and their characteristic results were
presented. Corresponding to their importance, the mechanical properties and
especially the stress-strain behaviour at large extensions were emphasised in the
discussion. This section concentrates on comparing the results o f the tube model
64 G. Heinrich, E. Straube and G. Helmis

with experimental results. Besides the mechanical properties at small and large defor-
mations, swelling and small angle neutron scattering (SANS) will be considered.

5.1 Stress-Strain Behaviour


Stress-strain measurements at uniaxial extension are the most frequently performed
experiments on stress-strain behaviour, and the typical deviations from the phantom
network behaviour, which can be observed in many experiments, provided the most
important motivation for the development o f theories o f real networks. However, it
has turned out that the stress-strain relations in uniaxial deformation are unable to
distinguish between different models. This can be demonstrated by comparing Eqs.
(49) and (54) with precise experimental data o f Kawabata et al. 141) on uniaxially
stretched natural rubber crosslinked with sulphur. The corresponding stress-strain
curves and the experimental points are shown in Fig. 4. The predictions o f both

1.0 .....

,
M Pa
j

j°/°
0.5 / °
Fig. 4. Stress-strain behaviour in uniaxial

o
/ extension: O experimental values by Kawa-
bata et al. 141~;_ theoretical results
for moderately crosslinked networks (Eq.
(49), 0.1 < I~ < 1) and for highly cross-
linked networks (Eq. (54)). The network
moduti are determined by a least-squares
fit of the Mooney stress. The resulting
° ,
curves are practically indistinguishable
1 2 3
=-

equations work pretty welt and are, in fact, practically indistinguishable. Consequently,
other types o f strain, like biaxial deformation, shear, etc., and swelling properties
need to be studied in order to test special models.
Figure 5 demonstrates the different behaviour resulting from Eqs. (49) and (54)
in the case o f uniaxial compression. We also tested the elastic potential (Eq. (44))
in the two cases v = 1/2 and v . . . . . 1/4 by comparing the corresponding stress-
strain relations with biaxial extension experiments 14i~ which cover relatively small
as well as large deformation regions for an isoprene rubber vulcanizate. In the rectan-
Rubber Elasticity of Polymer Networks: Theories 65

2.4
Pa

2.2

I Z0
x
j 1.a
/ -
1.6 i°
,7 Fig. 5. Mooney stress at uniaxial exten-

/
o
sion/compression : © experimental values
1.4 by Rivlin and Saunders 16~) for natural
rubber; - - Eq. (49), 13 = 1 ; . . . . .
Eq. (49), [~ = 0.4; . . . . . . . Eq. (49),
, i
13= 0.1; . . . . Eq. (54)
t20 1 2 3 4
~.-~ ,~

gular c o o r d i n a t e system t a k e n a l o n g the p r i n c i p a l axes o f strain, the principal stresses


for a n incompressible b o d y are expressed as functions o f the principal stretch ratios as

0
- - We,

2GN 1
[(~)~ - ~?~], (58)

0
- - Wel

-1 2 2 2 --1 2GN 1
= -(~1~.2) ]+---- [(~., ~.2)~ - ~4 p]
13 x2
for slightly or m o d e r a t e l y crosslinked r u b b e r s (n = --13), a n d

ch(~'l, ~'2) = -1 2 2 2 -1

+ 4 G N, X1-1[)J1/2 - - 0tlk2) -1/2]


(59)

for a highly crosslinked n e t w o r k (n = 1/2).


I n Fig. 6, the stress-strain curves oa(X 1, X2) are c o m p a r e d with the experimental
d a t a o f K a w a b a t a et at. 141) Both e q u a t i o n s reflect the d e p e n d e n c e o f o 1 on X~ a n d
66 G. Heinrich, E. Straube and G. Hetmis

M Pa

1.0

1
0.5
Fig. 6. Stress-strain behaviour %(Xp k2)
in biaxial extension: 6, o-, ?, -o, o-
experimental values by Kawabata et
al. 14~ for natural rubber. Theoretical
results (X1 = const.): - - moderately
crosslinked network (Eq. (58), ~ ~ 0.25--
0.33); . . . . highly crosslinked network
[Eq. (59)]. Guide lines: limiting case of
uniaxial extension 2 2 = 211/2, and sym-
0 metric biaxiat extension 21 = )-z
0 1 2 3

L2, but the quantitative agreement is worse than in the uniaxial case. Particularly
for large deformations, Eq. (58) greatly overestimates the dependence of cr1 on Xz,
if the case Xmi~ = X (e.i. 13 = 1) is considered. Reasonable agreement is obtained for
much smaller values 13 ~ 0,25-0.33. Then the fit of the uniaxial data is almost un-
changed whereas the fit of the biaxial data is considerably improved. The measured
stresses seem to be interpolation between the two versions of the treatment of the
constraints investigated here. The stress-strain relations in compression in Fig. 5
show an analogous picture. This behaviour can be expected if one considers the
assumptions which are made in Sect. 2 when developing the theory of the tube para-
meters. The expressions for the tube dimensions of Eqs. (58) and (59) correspond
to the limiting cases of highly and slightly crosslinked networks, respectively. Con-
sequently, the experimental results in Ref. 1417 which are obtained for a moderately
crosslinked network with a chain molecular mass of the order of 104 g/mol should
agree with the predictions for the limiting cases. As already mentioned in Sect. 2,
an explanation o f the swelling behaviour o f networks (see Sect. 5.3) requires the
assumption of the constraint release mechanisms. The discussion of the biaxial
extension experiments indicates that constraint release processes may also take place
in dry moderately crosslinked networks. However, at the present stage of knowledge
it is impossible to differentiate between effects resulting from a more complicated
deformation dependence of the constraints in moderately crosslinked networks and
from constraint release effects. Stress-strain investigations (including biaxial extension)
combined with relaxation spectroscopy on the same samples may clarify this situation.
Rubber Elasticity of Polymer Networks: Theories 67

5.2 Small-Strain Behaviour and the Front "Factor Problem

An old point of controversy in rubber elasticity theory deals with the value of the so-
called front factor g = Aq which was introduced first in the phantom chain models
to connect the number of elastically effective network chains per unit volume and the
shear modulus by G = AqkTv c. We use the notation of Rehage 1621 who clearly
distinguishes between A and rl. The factor A is often called the microstructure factor.
One obtains A = 1 in the case ofaffine networks and A = 1 - - 2/f(f = functionality)
in the opposite case of free-fluctuation networks. The quantityrl is called the memory
factor and is equal to the ratio of the mean square end-to-end distance of chains in
the undeformed network to the same quantity for the system with junction points
removed. The concept of the memory factor permits proper allowance for changes of
the modulus caused by changes of experimental conditions (e.g. temperature, solvent)
and the reduction of the modulus to a reference state 1631 However, in a number of
cases a clear distinction between the two contributions to the front factor is not
unambiguous. Contradictory results were obtained even in the classical studies.
According to Hermans 1041, Flory 1051 and Wall 1061 there should be no change of
chain dimensions, i.e. q = 1. James and Guth 103,1641 assumed that a contraction
takes place during the crosslinking process and estimatedq ~ 1/2. This interpretation
has to be considered erroneous. The assumptions of the model correspond to the
free-fluctuating network with q = 1, and the James-Guth result is related to the
microstructure factor A. The problem of the memory factor is still under discussion
in the literature.
Rehage and Oppermann 163, 1651assumed that the value ofq depends on the mobility
of the network chains. They argued that, for example, end-linked PDMS chains relax
fast enough during crosslinking to yield a vatueq ~ t/2, whereas in networks made
from long primary chains the network strands do not attain equilibrium configurations
and rl = 1 can be expected. Staverman 41 assumed q = 1, as he considered that in
the most probable state of phantom networks the chains are in the 0-state. Further,
he considered that in highly swollen networks the geometry of the equilibrium positions
of the junctions may be considered to approach the most probable state of the cor-
responding phantom network. This assumption is based on the observation of Rempp
et al. 166) that for swollen networks the front factor depends on the swelling solvent
and not on the conditions of crosslinking. Allen 1671 offered a direct approach to
the memory factor problem. He reported neutron scattering measurements of chain
dimensions of poly(ethylene oxide) in the melt doped with perdeutero chains and
crosslinked using a trifunctional isocyanate. This experiment was carried out for
networks made from samples with two different molecular weights of chains and
2 2
yielded Rgc/R o = 0.60 + 0.05, where Rgc is the root mean square of the radius of
gyration of a chain crosslinked into a network and Rgo is the corresponding value of
the free chain in the melt. For the free chain the relation R~o = (1/6) R0a (R0 is the
root-mean-square end-to-end distance of the free chain) holds but Rggcis not necessarily
equal to (1/6)R 2 and depends on the chosen model 1671. Thus, one has to be careful
to give an unambiguous value for the memory factor from direct neutron scattering
experiments.
Summarizing this rough review of discussions of the memory factor problem it
can be concluded thatrl = 1 seems to be a reasonable assumption in the case of net-
68 G. Heinrich, E. Straube and G. Hehnis

works crosslinked in bulk. Consequently, the assumption of random flight behaviour


of the chains made in most theoretical works is justified. The discussion of the front
factor problem reduces then to discussion of the microstructure factor and of the
relation between junction density and its contribution to the network modulus.
An additional discussion of the dilution effects of loose ends, ineffective nodes and
cycles wilt be omitted here. A detailed discussion may be found in Refs. 6,163)
An estimation of the junction contribution to the network modulus can be performed
most unambiguously if topological and junction contributions give different stress-
strain behaviour. Then, the parameters Gc and GN or the Mooney-Rivlin parameters
C 1 and C z can be determined directly from the stress-strain data. Most of the experi-
mental material is still represented and discussed in terms of the Mooney-Rivtin
equation. Therefore, the front factor problem of real networks will be discussed
within the framework of this equation.
We therefore propose to define an effective front factor cl by the equation

~M _ cr0~)
2Gc v kT(~. -- L-2) = ci + cz~-~ (60)

An expression for the left hand side of Eq. (60) for moderately crosslinked networks
follows from Eq. (49)

cYM - 0.5 + 0.5 GN


2Gc G~ fiX) (61)

1.5 GN/Gc=J"
7 2.0 -
/
1.5

T
,'L.'.O
1.0
1.0

"x.
05

0.1
0.5

0.1 0.5 1.0


,q-1 --

Fig. 7. Reduced Mooney stress crM/2Gc vs strain showing the dependence on the strength of the
constraints GN/G c

Figure 7 shows the dependence ofcrM/(2Gc) on ;L- i for different ratios GN/Gc. A least-
squares fit of data in the easily measurable deformation range 0 ~-1 = 0.5, 0.6, 0.7,
0.8, 0.9) yields Cl and c2 as a function of the strength of topological constraints GN/Gc.
Rubber Elasticity of Polymer Networks: Theories 69

1,0

0.5
0
L
t~
Fig, 8. Reduced Mooney- Rivlin parameters
c1 and c2 as functions of the strength of the
I I constraints GN/Gc (see Eq. (62))
0 I 2
GN/ Gc

Figure 8 shows the dependence o f the parameters c 1 and c 2 on the topological con-
straints. A least-squares fit yields the relations

cl ~- 0.5 + 0.02GN/Gc = 0.5 + 0.04C 2


(62)
c2 ~ 0.5GN/Gc
In the limit GN/Gc = 0, we obtain the value cl = 0.5, i.e. the free-fluctuation result.
Clearly, the extrapolated front factor c 1 - - in principle -- offers no upper limit
and would be greater than 1 for (of course unrealistic) very large GN/Gc. In the case
of usual values GN/Gc ~ 1, the difference between c 1 and the free-fluctuation
phantom limit result cl = 0.5, if the functionality f is 4, plays only a minor role.
On the other hand, the theory of restricted junction fluctuations limits cl to values
smaller (or equal to) 1, and the upper bound corresponds to the affine-deformation
p h a n t o m limit where complete suppression o f junction fluctuation has been assum-
ed 4~ -44). Furthermore, c 2 passes through a maximum for values o f c~ in the range
between 0.5 and 1, and the sum of the reduced parameters c I and c2, the reduced
shear modulus gr = G/vckT ~ c~ + %, is limited to the range 0.5 < gr < 1 36)
The different predictions of the tube model with deformation-dependent constraints
and o f the model o f restricted junction fluctuations concerning the cl - - cz.relation
are shown in Fig. 9.
Experimental data on poly(dim~hylsitoxane) networks crosslinked by the end-
linking reaction 168,169) yielded for all samples with Ms > 104 g/mol values of the
reduced shear modulus in the range gr > 1 36, 97) From this result, it may be concluded
that the c 2 values o f the theory o f restricted junction fluctuation are much too small,
i.e. the predictions o f this theory remain close to the results of the phantom network
with fixed junctions as is expected from the structure of the theory itself. The tube
model with deformation-dependent constraints yields unlimited cz values and allows
a description o f the small-strain behaviour o f polymer networks in agreement with
the experimental material. The results show particularly that the network behaviour
m a y be described by the following parameters only: crosstinking density, chain density
70 G. Heinrich, E. Straube and G. Hehnis

/// 0.8

1.2

0.6

1,0 3///
J
l
0.8

0.2

0.6

0.5
0.5 1.0 1.5
CI+C 2

Fig, 9, Reduced Mooney-Rivlin parameters cl and ca as functions of the reduced shear modulus
gr ~ Cl -1- C2: . . . . . . CI; ..... C2. 1: moderately crosslinked network (Eq. (49), [3 = 1); 2: highly
crosslinked network (Eq. (54)) ; 3 : Flory-K~tstner theory. Experimental values ( 0 cl, C) %) by Mark
and Sullivan i68) for end-linked tetrafunctional PDMS networks

and functionality. This may be shown most convincingly by scaling relations between
ideal and constraint contributions to the network moduli.
In Sect. 2, we pointed out that for the plateau modulus of polymer melts an excellent
quantitative agreement was achieved with the scaling representation of experimental
results deduced by Graessley and Edwards 95) (see Eq. (19)). Similary, from Eqs.
(20, 51, 52) in the case of moderately crosslinked networks (NF >> 1) the relation

GN 1
GcG ° G c ~, z= 1 (63)

follows where the expression 92, 95)

G ° = 0.01 lkT1-3(nfl3) 2

has been used. Equation (63) can be tested experimentally by double logarithmic
plotting of its left hand side, against Gc using the data of different stochastically cross-
linked PB, N R and P D M S rubbery net works. Some values of Gc and GN were deter-
mined from stress-strain experiments using Eq. (49) directly for the force-strain
dependence 93,97) Other values of Gc and GN could be determined from literature
values 27,170~ of the Mooney-Rivlin parameters C1 and C2 according to the following
fitting procedure 97):
~2
dX[c%R(X ) -- c(X)] 2 --* min

with the Mooney-Rivlin stress 0"MR(~,) according to Eq. (1) and a(2) given by Eq. (49).
Rubber Elasticity of Polymer Networks: Theories 71

x
x PDMS
x-~x +ONR
× ~'x,, "PB
10
IMPa-1
J~

0.1 I , ~ J , ,,~,1 , ~ , ~ ,,,,I ~ , ,


0.01 0.1 M Pa 1
Gc ,-

Fig. 10. Scaling relation for the network moduli (cf. E q . ( 6 3 ) ) . The experimental data from Refs.
93 ( O , + ) , 97 ( O , + ) , 27 ( 0 ) , 170 ( X ) and the fitting procedure are described in the text

The parameters 21 and 22 represent the smallest and the largest deformation ratios
of the corresponding uniaxial stretching experiment. Figure 10 shows that a least-
squares fit gives a straight line characterized by z = 0.93 (with a correlation coefficient
of 0.98). Further, plotting all the shear moduli contributions GN directly against the
corresponding plateau values of non-crosstinked polymers, the relation GN ~ G °
becomes clear (Fig. 11) where in our case the constant of proportionality is 0.55.

2
MPa

t 1
,_4

0.5 1.0 MPa 1.5

F i g . 11. Relationship between the plateau m o d u l u s o f the uncrosslinked system and the constraint
contribution to the network modulus. Experimental data f r o m F i g . 10
72 G, Heinrich, E. Straube and G. Helmis

The very satisfactory fit of the scaling predictions obtained by the tube model
with deformation-dependent constraints to experimental results for various rubbers
at small strains strongly supports the basic assumptions of this approach. Further,
it seems worth noting that the small-strain behaviour of rubber-elastic networks
considered here is well described by the theoretical results for the case of low crosslink
density. This case shows a strong resemblance to that of the melt; as there, the influence
of topological constraints on the mechanical properties of the networks is important
and is well reflected by the molecular field approach of the tube model. As mentioned
already in Sect. 4.2, these conclusions are almost identical with the view of Graessley
and co-workers 27,2s) and Macosko and co-workers 2~,26), who also extensively
used the relationship between melt and network properties in order to interpret their
experimental results.
The foregoing discussions of the front factor problem and of the relations between
topological and junction contributions are based on the separation of these contribu-
tions using their different stress-strain behaviour. In the case of networks with large
topological contributions and phantom-like stress-strain behaviour, approaches
based on the trapping factor concept were successfully applied. Ilavsk~ and Du~ek 12s)
represented small-strain and large-strain data of polyurethane model networks of
poly(oxypropylene)triols and diisocyanate by extremely small values of the parameters
C2 but with GN/Gc > 1. The results were successfully interpreted in the fi'ame of
of the Langley-Graesstey concept 12s) with the microstructure factor A = (f - - 2)/f
-- 1/3 as the most realistic assumption. This is in agreement with the results of Ref. 6~
and contradicts some papers (e.g. Refs. 171,172)) which assumed no topological
contributions to the modulus. The interpretation of the data of Ilavsk~, and Dugek lzs)
in the context of the present authors' tube model is possible with v = - - 1 and fl ~ 1.
Finally, the relation G c = AkTv c has to be discussed. It is important for the deter-
mination of the molecular mass of elastically active chains from elastometric data.
The tube model always gives Eq. (51). The limiting case of the fixed phantom network
is not obtained even for do --, 0. To deduce the limiting case of the affine phantom
network we need to abandon the basic assumptic ns of our model that the constraints
acting on any segment are equal. This can be realized by introducing a tube radius
•do(s) which depends on the arc length parameter s of the chain, and allows one to
distinguish between the restrictions of segment fluctuations and junction fluctuations,
respectively. Then, in the case of unrestricted chains and vanishing junction fluctuation
the simple replica partition function (see Sect. 2)

Z(n) =
(0; ds f ds' I-I 6(R~)(s) - R~)(s'))
0 ~t=O ?
has to be evaluated. The averaging must be performed with the phantom-chain
Green's function for the uniform density assumption (compare to the discussion
in Sect. 3). This leads to the affine phantom limit 6o~
In the general case of restricted chains and partial junction fluctuation, the averaging
of the junction position fluctuation ~ (R (~*°~ - - ~.R(°)) has to be performed with
the Green's function of the harmonic oscillator where the integrals are limited to
finite length scales of the order ofd o. This gives additional contributions to the partition
flmction which have to be taken into account when using the variational principle
Rubber Elasticity of Polymer Networks: Theories 73

(cf. Sect. 4). Without any proof, it is expected that, using this idea, the microstructure
factor scales with d / ( I L ) 1/2; d < d o characterises the range of the junction fluctua-
tion. This is consistent with some heuristic arguments made by K/istner 3s) who
discussed the joint introduction of restricted junction and segment fluctuations. He
came to the following two important conclusions:
a) All influences on crosslinks and network strands can be assumed to be decoupled
and to be additive.
b) The front factor depends only on the restriction of junction fluctuations 3s).

A= A =1--( 1 ~ erfK /] (64)

K = (ILj/2

In accordance with the tube model, Eq. (64) assumes a length scale d c to be available
for junctions to fluctuate. For example, in the case of natural rubber the value of A
according to Eq. (64) changes from A g 0.55 to A ,~ 0.85 when the molecular weight
of network strands increases from M = 4 × 103 g/mot to M = 40 × 103 g/tool.

5.3 Swelling Properties of Polymer Networks


In the study of swollen networks, two problems are of major importance: The depend-
ence of the stress-strain properties on the solvent or polymer fraction and the mixing
contributions to the free energy of the network or the elastic contribution to the
chemical potential. Latest research seems to provide an improved insight into some
special effects which are typical for swollen and completely crosslinked networks,
and for unswollen (and swollen) incompletely crosslinked networks. The relaxation
on the deformation dependence of topological constraints, which leads to a constraint
release, is one of them.
The present experimental situation concerning swelling properties of polymer
networks is characterised by the following features:
1) The phenomenological Mooney-Rivlin constant C 1 is essentially independent of
swelling, whereas C 2 decreases very significantly with decreasing volume fraction,
v v of the polymer in the swollen network tvo). In particular, the empirical relation
C2(v2) : C2(1)v2k with 1/3 -< k -< 4/3 may be quoted 17o) In addition, it is known
that C 2 in general becomes zero at v2 -~ 0.2 to 0.25.
2) Vapour-sorption experiments on different polymer plus solvent systems s-12)
have shown that the elastic component of the solvent chemical potential exhibits
a maximum, contrary to the phantom network theories or the Mooney-Rivlin equa-
tion. Furthermore, evidence has been found that the localisation and height of the
maximum is dependent upon the nature of the diluent.
3) Measurements on swollen PS networks by SANS have shown that the radius of
gyration of the elastic chains is the same as in bulk, although the network was swollen
by a factor of 3 173,174) Additionally, a decrease of the radius of gyration of PS network
chains in cyclohexane has been observed if the swelling degree is increased by raising
74 G. Heinrich, E. Straube and G. Hehnis

the temperature. These effects and the SANS observation ~75~ that the molecular
deformation of bulk networks induced by strain is also much less pronounced than
that calculated for phantom models have led to the proposition that large macroscopic
deformations could be accompanied by topological rearrangement of junctions
(microscopic molecular deformation) at a scale larger than the mesh size 176)
To explain the effects listed above, we start with the molecular field approach of
Sect. 4. The experimental analysis of stress-strain isotherms of swollen polymer
networks is most easily carried out in terms of the reduced stress

--
o00 v~;3
--2
(3"M ~, - - k- 2 = C 1 + C2(v2) }v.- 1 (65)

where ~(X) is the force per cross-sectional area of the unswollen, undistorted sample,
and X is the deformation ratio of the sample relative to the swollen state; v2 = V0/V
denotes the volume fraction of the polymer, V o is the polymer volume in the dry
reference state, V = V0 + n~V~ is the swollen network volume with n 1 the number
of moles of solvent and V1 the partial molar volume of the diluent.
The introduction of the swelling dependence of the lateral tube dimensions according
to Eq. (20)

do(v2) = dov21/2 (66)

into Eq. (62) for the reduced Mooney-Rivlin parameters c t and c2 yields the relations 97,
177)

q ( v 2) = 0.5 + v21%(1)- 0.5]

= 0.5 + v2 × 0.02 GN (67)


Gc

C2(V2) = V2C2(1 ) = V2 × 0,5 GN (68)


Gc

Hence, from Eq. (67) it follows that c, should be almost independent of the swelling
degree. The ratio c2(v2)/c2(1) = C2(v1)/C2(1) decreases linearly with swelling (k = 1).
This prediction is in agreement with the experimental behaviour (see Fig. 12) only in
the case of small swelling degrees v2 >~ 0.5. The nonlinear decrease and the above-
mentioned results of SANS investigations leads to the conclusion that processes of
constraint release that are dependent on swelling degree take place. As already
mentioned, these processes can be described by a microscopic network chain defor-
mation much smaller than the macroscopic sample deformation.
A first quantitative approach 177)to the explanation of swelling behaviour based on
constraint release has been provided by the relation

k2llliC = [t - or(v2)]k 2 + ot(v2) 0 -<- cz -<- 1 (69)


Rubber Elasticity of Polymer Networks: Theories 75

instead o f the ansatz (21). The parameter ~zhas the same physical meaning as the para-
meter 13 introduced in Eq. (21). Scaling considerations lead in the case of completely
crosstinked networks to the relation

~ do/IN~/Z (70)

The parameter c¢ is determined by the ratio o f the allowed fluctuations to the end-to-
end distance o f a network chain. The swelling dependence of~ results from the swelling
dependence of d o according to Eq. (66)

0ffv2) = %v~ 1/z (71)

In this approach, it has been assumed that, as well as any kind of network defects,
the softening of the network by swelling (see Eq. (71)) facilitates the constraint release
and increases the value of the parameter ~.
The parameter Xmi* replaces the macroscopic deformation ratio only in the contribu-
tions to the free energy by the topological constraints. This can be shown rigorously
using the replica method. It turns out that the "ideal" contribution to the free energy
of the network ~ ~ k 2 is caused solely by the deformation o f the boundary of

the network and is not influenced by any assumptions concerning the deformation of
the network chains or processes of junction rearrangements. Introduction o f Eqs.
(69) and (71) into Eq. (62) results in a nonlinear decrease of the C z value with increasing
swelling degree. Figure 12 shows that by" an appropriate choice o f % a very satisfactory
agreement with the experimental results can be obtained. Summarising the results
obtained so far for swollen networks, we conclude that in order to understand both
the mechanical properties and the SANS results x3-15) the constraint release effects
have to be considered.
N o w let us consider the vapour-sorption investigations on swollen networks based
on the determination o f a L ¢, the activity o f the solvent over a swollen crosslinked

1.0 , , , - ~

t 0.5 $7~"

, /o /, , ,

0 0,2 O.Z, 0.6 0.8 1.0


V 2 ~

Fig. 12. Swelling dependence of the Mooney-Rivlin parameter C2 : O experimental values by Gumbrelt
et at. i~8) Theoretical results: - - % = 0; - % = 0.5; . . . . . % = 0.6 (compare with
Eqs. (68)-(71))
76 G. Heinrich, E. Straube and G. Helmis

polymer, and al.., the corresponding activity over an non-crosslinked but otherwise
identical polymer at the same concentration of the solvent. As already mentioned,
Gee et al. 8~and later Eichinger et al. 9-12) have shown that the results for the chemical
potential of the diluent is at variance with conventional theories of rubber elasticity.
The product of the elastic contribution g~,~ of the diluent and the linear isotropic
extension ratio ki = v21/3 = [Vo/(Vo + n1~1)]-1/3

Lilal,el - 1 1 (8 AF) = LiR_TTIn ( a1"¢~} (72)


~71 3Vo ~'i ~ ~ i fiT 91 \ a l , u/

passes through a maximum. Phantom network theories and also the Mooney-Rivlin
equation predict either constancy or a monotonic increase with swelling degree.
These theoretical results were obtained by making the following assumptions:
1) The elastic and the mixing contributions to the free energy" are additive.
2) The mixing contributions are equal for crosslinked and uncrosslinked systems.
Using the same assumptions and expression (37) for the elastic contribution to
the free energy and taking into account the results of the SANS measurements on
swollen networks 174) by the choice (kmi~)i = 1, we obtain

F 3 3 NIL
= ~ M)~2 + _r7 + 3M log do(v2) + const. (73)
k--~ V6 do2(V2)

and

kiRT ( a l , c) MkT
--In = -- g~(k0 = Gcq00~i), (74)
Vl \al,u/ Vo
with

3 3 R2f
q~()~i) = 1 + ~_~.~2 )q-5 (75)
Z 2~ do2 2

and R 2e the mean square end-to-end distance of a network chain in the dry state and f
the functionality. Expression (75) exhibits a maximum at the isotropic extension
ratio

(5 R2 f ) l / 3
0~i)max = -6Y/2 d 2 (76)

2 2 1/2] were obtained by Gaylord 76.77)


Qualitatively similar results [(~'i)ma, -~ (R~/do)
and Marucci 7,) using a three-chain tube model with each tube aligned along one of
the principal axes of strain (see Sect. 4.3). There, it was assumed that the tube length,
and therefore the end-to-end distance of the network chains, changes affinely, while
the tube cross section deforms in accordance with the assumption of conservation
of tube volume. These particular assumptions are obviously not crucial, the same
result is obtained in our model when there is no relaxation of the topological con-
Rubber Elasticityof Polymer Networks: Theories 77

straints, i.e. when (Lmi~)i = Li is assumed. Similar results were also obtained by
Flory 178~within the theory of restricted junction fluctuation. Brotzman and Eichin-
ger 11,12) compared Flory's predictions with measurements on a series of end-linked
and well-characterised PDMS networks swollen in cyclohexane and benzene. The
result is typical for theories resting on the assumptions formulated above. Whilst
they succeed more or less in explaining the value of the swelling degree at the maximum
all these theories mentioned so far fail to explain the maximum value of qo(Li). This
conclusion has been supported by Gottlieb and Gaylord 180), who used experimental
swelling data s -12) on PDMS and N R to compare some theoretical models 41-44, 74,
75, vg, so, 8z, 88) of rubber elasticity. Quantitative agreement between experiments and
tube-like constraint-models 74, 75, v9) requires physically unrealistic parameter values
(e.g. negative values for the number of steps in a chain).
Using for example the data available for the sample EL 2 in the paper of Brotzman
and Eichinger 12), the present authors calculated the swelling function according to
Eq. (74). The height of the maximum turns out to be only slightly higher than the
corresponding result of Flory's theory, i.e. the particular form of the elastic free
energy does not significantly influence the height of the maximum, and the reasons
for the differences have to be found in other contributions. We conclude that the
second premise introduced above, the assumption that the translational part of the
free enthalpy of mixing is the same in both crosslinked and uncrosslinked material
is wrong. Upon more careful consideration, we have to take into account that for
given values of the volume V and moles of solvent n~, the free enthalpy of mixing
differs in both states. This difference is caused by the presence of the network junctions,
which leads to a decrease of the number of degrees of freedom of the system. Con-
sequently, different swelling-dependent contributions to the thermodynamic potentials
restflt due to the swelling dependence of the volume (or the number of sites in the
Flory-Huggins theory) available to the network chains.
These suggestions are supported by the observation that the equilibrium swelling
degree corresponds to the overlap volume fraction v* of uncrosslinked chains with
the molecular mass M r lsl). From this correspondence it can be concluded that
upon swelling a disinterspersion process takes place. This conclusion is supported
by recent scattering results. The occurrence of the maximum can be related to a two-
step disinterspersion process where the maximum corresponds to a cross-over point
in the disinterspersion process from Gaussian highly interpenetrated chains in the
dry network vz = 1 to Gaussian chains partially separated from one another at
v~* and completely swollen and separated chains at v*.
In a preliminary approach 97) the swelling dependence of the mixing contribution
to the free enthalpy was simulated by a simple lattice gas model that roughly took
into account the different volumes available to the network chains at different stages
of swelling and disinterspersion. This approach yields an additional contribution
,-~Gcqb(v2) to the elastic chemical potential where

(1)(V2) = ~--(V2/V~ $) V22/3 In [1 -(v**/v2)] for 1=> v2~> v**


(77)
[-(v**/vz) v2z/3 In [1 - ( v * * / v z ) ] for v~* >~ v2 > v*

Details will be published in Ref. 18z)


78 G. Heinrich, E. Straube and G. Hetmis

1.0
Nmm-2 ?

r o

Fig. 13. Swelling dependence of the ratio


~ 0.5 of the activitiesof the solventin crosslinked
and uncrosslinked polymers: O experi-
mental results by Brotzman and Eichin-
,*4
ger 12) (sample EL2); • swelling equili-
brim. Theories with equal mixing contri-
// butions in crosslinked and uncrosslinked
systems: . . . . . . . Flory-K/istner theory;
. . . . affine swellingof the chains (% = 0);
unchanged end-to-end distance
on swelling (% = 1); . . . . . . theory which
takes into account disinterspersion effects
t _ _ (Eqs. (75)-(77))

v22/3 = ~ 2 =

Figure 13 shows that this approach yields the location and the height of the observed
maximum for the sample EL 2 of Ref. 12) surprisingly well. The approach presented
above still uses the additivity of mixing and elastic contributions to the thermo-
dynamic potentials. Therefore, observed shifts of the location and of the height of
the maximum that depend on the kind of solvent 11,lz) have to be explained by non-
additive contributions. Such contributions have been investigated by Ball and
Edwards 183) using the replica approach (see Sect. 2). The investigations support the
existence of non-additive contributions but suffer from the fact that the influence
of topological constraints has not been included. A comparison of the results with
experimental data seems hardly possible or useful because it has to be concluded
from the results presented so far that the swelling dependence of the topological
constraints are mainly responsible for deviations from the "ideal" behaviour.
On the basis of uniaxial compression experiments on swollen gels, Nagy 184)
discussed the structure of the mixing contributions with the help of thermodynamic
considerations. His assumption of a vanishing swelling pressure in the dry state of
networks crosslinked in bulk yields a maximum of the chemical potential of the solvent
that depends on the polymer volume fraction. Mechanisms that may lead to different
mixing contributions in crosslinked and uncrosslinked systems are not considered
in Ref. 18~)
To conclude the discussion of the swelling properties of polymer networks, it
should be noted that the swelling is only one (but the most important) case of an
external influence on a network. Pressure, and especially temperature, variations are
Rubber Elasticity of Polymer Networks: Theories 79

of interest and can be investigated on the given theoretical basis. These problems are
in principle somewhat simpler due to the absence of mixing contributions. Pressure
and temperature variations enter the theory via the dependence of both the model
parameters (e.g. the segment length) and the sample dimensions on these quantities.
Consequently, detailed knowledge of the dependence of these parameters on pressure
or temperature is needed and the results are controlled by the particular behaviour
of the polymer under consideration. A discussion of these problems goes beyond the
scope of this review.
Networks, crosslinked in the swollen state and investigated at different swelling
degrees than that at crosslinking (e.g. in the dry state) are another example of problems
related to the swelling problem discussed above. In this case the theories given above
can be used if at all states of dilution at crosslinking and investigation of network
properties the network chains may be treated as random flight chains (this assumption
is made in most of the network theories). Then the change of chain dimensions may
be absorbed into the memory factor as discussed in Sect. 5.2. This condition confines
the polymer concentration at crosslinking to values above the crossover to random
flight behaviour c**. Whereas in swelling experiments the simultaneous occurrence
of chain swelling and chain disinterspersion has to be supposed, the de-swelling of
networks crosslinked in solution has to be assumed to be accompanied by chain
compression leading to a supercoiling of chains and by chain interpenetration. It is
expected that comparison of the behaviour of bulk and solution crosslinked networks
will advance the understanding of the swelling properties of polymer networks,
which still suffers from a number of shortcomings.

5.4 SANS and Microscopic Deformation

As already mentioned, SANS on partially deuterated networks allows a direct


measurement of polymer chain deformation (stretching or swelling). This makes
SANS a unique tool for assessing the validity of theoretical models of polymer
networks.
For uniaxially stretched networks, the molecular deformations are characterised
by the radii of gyration Rgfl an R l respectively parallel and perpendicular to the
stretching direction. From the small-angle scattering function, the molecular defor-
mation of a stretched elastic phantom chain has been calculated for three cases:
1) Free-fluctuation phantom network ls5).

Rg,, ( f + 2 + (f - 2) ~.2)~/2
R,__ =
(78)
__R'i= (f+2+(f-2) Z.~l)'/2
Rgi 2f

where Rgi is the radius of gyration for unstretched isotropic network chains, £ is
the extension ratio, and the network is assumed to be incompressible.
80 G. Heinrich, E. Straube and G. Hetmis

2) Affine junction displacement 173) ("end-to-end pulling" deformation):

Rgi
(79)
R g ± _ (~- + l~ 1/2

3) Affine deformation: This model assumes that the deformation of each configura-
tion of the chains is affine in the macroscopic deformation. It is not compatible with
known classical theories of rubber elasticity.

Rgl[/Rg i = 2v (8o)
Rg±/R i -- 2~,-3/2

The experimental situation can be summarised as follows: Crosslinked and swollen


polystyrene gels exhibit a chain deformation which is less than that given by all
phantom network models 1,.4). Early experiments by Clough et al. 186) on radiation
crosslinked polystyrene networks had been interpreted to be consistent with the free-
fluctuation phantom scattering law. Most of the scattering experiments which were
performed on uniaxially stretched end-linked siloxane networks showed much smaller
molecular deformations than would be predicted even from Eq. (78)la-15L To
clarify this situation, Ullman 187)used the Flory model of suppressed junction fluctua-
tions to extend Eq. (78) to the case of a partially free- fluctuation phantom network.
He reached the conclusion that the reduction in junction fluctuation cannot account
for chain deformations less than that expected from a phantom network. From this
result, the following conclusions could be drawn 124.,188).
1) The concept of network unfolding and junction rearrangement 176)seems to provide
a qualitative explanation for small chain extensions. Obviously, the amount of
network unfolding depends on network preparation, degree of sample deformation
and quantity of solvent in the network.
2) The scattering from randomly crosslinked networks built up from long primary
chains is very different from that of end-linked networks. The former shows much
greater sensitivity to network deformation and is much less sensitive to network
functionality. The latter seems to be very sensitive to incomplete crosslinking which
changes the elastic properties of a network substantially. We believe that the concept
of microscopic deformation introduced in Sect. 2 provides a useful quantitative
basis for the construction of the scattering law of topologically constrained polymeric
networks.
Using the Warner-Edwards replica method 189), the scattering function S(k, 2,0
= (ST(k, 2~))T(k is the scattering vector and ~ the sample deformation) was calculated
in Ref. 190~by averaging the scattering law ST of a single network topology over the
set of all possible topologies (according to Sect. 2, the topology T includes the
topology of crosslinking and of the tube-like configurational constraints). In agree-
ment with the considerations in Sect. 2.2 and 5.3, the tube axis configurations are
assumed to change according to the microscopic deformation law R(s) ~ )~mik(s).
Rubber Elasticity of Polymer Networks: Theories 81

In the SANS case (q =- kZ12NJ6 < 1), the limit d ~ 0 leads to the scattering law
in the stretching direction

S(k**, ~v) ~ 1 - - kl)Rg**


22 q ( ~ )
= 1 -- -~ (8t)

which is equal to the scattering law of tetrafunctionat phantom networks in the free-
fluctuation limit iss). Equation (81) then gives Eq. (78) for the special case of f = 4.
The more interesting case of topologically constrained moderately crosslinked
networks (M/N ~ 1, 1L/doz >> M/N) gives 97,19o)

S(kll, )Q -~ 1 -- qA 2 (82)
3

with

M/N ( t M/N (83)


A2 L2mi¢- 3 F( ,mic ) 1+ r(•mle)/ ( '2mic --

and

1 1L
FO~m,~) - - t / ~ d z , d,, = do(~,mi~),~/2

Here ~lf = X is the stretch ratio parallel to the principal strain axis. The result

RglI/R i = A (84)

leads to the final conclusion that the radius of gyration of a network chain changes
more or less according to the microscopic deformation ratio, i.e. much tess than the
applied macroscopic strain. The microscopic deformation depends on the parameter [3
(or ~ as in Sect. 5.3) which describes the global rearrangement of crosslinks 176~and
the release of topological constraints by relaxation of the deformation dependence
of the tube dimensions. For completely crosslinked networks, [3 (or c0 is related
to typical network lengths like d o and ]//iL (Sect. 2), and hence it is correlated with
the Flory number N v defined in Sect. 2. The interpretation proposed by Bastide et
al. 176~differs only slightly. There, the parameter 13 correlates with the ratio of the
total number of spatial neighbours to the number of topological neighbours, n/f,
[[3 ~ (n/0-1]. The larger the value of n/f, the more likely is deformation by rearrange-
ment of junctions and disinterpenetration of network chains. Furthermore, as already
mentioned in the theory of rubber elasticity, the parameter 13also depends on network
preparation and serves as an indicator of the completeness of the crosslinking reaction.
In the case of moderately crosslinked and topologically restricted networks, we
obtain results which may be used to explain the low (non-affine) values of molecular
deformation observed experimentally (Fig. 14). Values of eLclose to 1 indicate that
the elastic properties of networks prepared by the end-linking reaction are very
82 G. Heinrich, E. Straube and G. Hehnis

f ...... I'
?
/<-
1.31 / /
/ /

1 i.2 /
/
/
/
./.
/./
I

7 111
J
/ /( / ./1/
"V V /' Z" /."~" /"
./ +/ , ./ 1. f ~ . i

10 "'~e"~'~"- t
1,0 1.5 2.0

Fig. 14. SANS properties of networks: Experimental results on end-linked PDMS 13): + Mc =
= 6100 g/mol, (2) Mc = 10500 g/mol, It M = 23000 g/tool and crosslinked in solution (v° = 0.7).
Theoretical results: . . . . . . affine phantom network (Eq. (80)); . . . . . . affine junction displace-
ment (Eq. (79)); . . . . . . . freely fluctuating phantom network (Eq. (78)); . . . . Eq. (84) with % = 0.8,
M~ = 23000 g/mol; - - - - Eq. (84) with % = 0.9, M0 = 23000 g/tool

sensitive to the completeness o f reaction. Such types o f networks exhibit phantom-


like behaviour.

6 Summary and Conclusions

In this article we have reviewed and analysed the present stage of development of
the statistical mechanics o f rubber elasticity, taking into account the controversial
results and recent progress in this area. We have concentrated on the tube model as
a mean field a p p r o a c h to the statistical mechanics o f polymer systems with topology
conservation. There are two main reasons for the preference of the tube model.
Firstly, rubber-like polymer networks are characterised by such a high degree o f coil
interpenetration that a more rigorous treatment than the mean field approach, i.e.
the application o f methods relying on topological invariants, needs a great deal of
elaboration for the complicated system. Secondly, the molecular field approach
allows a self-consistent treatment and provides (to a considerable extent) a correct
or at least very satisfactory description o f the properties o f polymer networks.
In order to remove the main differences and confusions mentioned in Sect. 1,
the relations between the tube model and other approaches which have been used
successfully, such as the model o f restricted junction fluctuations and the concept
o f trapped entanglements, have been discussed in detail and a number o f new insights
have been achieved.
Summarizing the results, the following conclusions can be drawn:
1) The thermodynamic potentials and the resulting stress-strain relations can be
expressed by" two contributions. One contribution corresponds to the p h a n t o m net-
Rubber Elasticity of Polymer Networks: Theories 83

work result and an additional part comes from the topological constraints caused
by the uncrossability of the network chains.
2) Closed theoretical results have been derived in the case of dry networks made
from long primary chains. They satisfactorily fit the experimental data both for small
strains and for large strains in the range covered by the Gaussian chain model. All
modes of deformation usually employed can be described with similar accuracy.
For this system, scaling relations of the dependence of the topological contribution
on the network parameters, such as crosslink and segment number density, have been
deduced which agree with available data for PB, NR and PDMS networks. Scaling
relations for the limiting case of a polymer melt coincide completely with results
derived phenomenologically.
3) The deformation dependence and the magnitude of the constraint contributions
to the thermodynamic potentials and to the stress-strain relations are controlled by
the values and the deformation dependence of the tube dimensions.
4) Networks which contain a large number of network defects (which facilitate
constraint release processes) can be described successfully within the concept of trapped
entanglements acting analogously to chemical crosslinks. It has been shown that
the typical results of the trapped entanglement concept can be obtained within the
framework of the tube model if the results of a simple entanglement model that
describes only the unrelaxable entanglements are introduced.
5) The swelling dependence of the constraint contribution shows two different
regions. For small swelling degrees (v2 > 0.5), the behaviour may be described by
the swelling dependence of the tube dimensions caused by the decrease of the chain
or segment density on swelling. The explanation of the mechanical and thermodyna-
mical properties as well as those of the scattering results on networks at higher swelling
degrees requires constraint release mechanisms similar to constraint release processes
in defect-containing networks. The concept of relaxed microscopic deformations much
smaller than the macroscopic deformations of the sample was introduced. It yields
a very satisfactory description of the results considered.
6) Constraint release effects are expected to be responsible for the different experi-
mental results collected for end-linked networks and for networks made by the cross-
linking of primary chains much longer than the resulting network chains. The different
degree of completeness of the networks has to be considered as the main reason for
these differences.
7) The constraint release concept introduced is supported by scaling considerations
resting on the assumption that the ratio between the tube dimensions and the chain
end-to-end distance controls these properties. In swollen networks, the cross-over
phenomena known from the physics of concentrated and intermediately concentrated
polymer solutiol~s influence the topological constraints significantly. A detailed
microscopic understanding is not yet possible.
8) The tube model presented here yields values of the front factor of the crosslink
contribution close to the front factor of the free-fluctuating phantom network.
It is felt that the stronger constraints acting on the crosslinks have to be simulated
by tube dimensions that depend on the distance from the crosslinks. In this way,
the crossover from the free-fluctuating to the fixed phantom network value of the
front factor, characteristic for the model of restricted junction fluctuation, can also
be reproduced by tube models.
84 G. Heinrich, E. Straube and G. Helmis

C o m p a r i n g the a d v a n t a g e s and s h o r t c o m i n g s o f the various concepts, the tube


m o d e l m a y be c o n s i d e r e d as the m o s t flexible and the one m o s t suited to successive
i m p r o v e m e n t . T h e m a i n task r e m a i n i n g is the d e v e l o p m e n t o f a q u a n t i t a t i v e theory
o f the t o p o l o g i c a l constraints for the w h o l e range o f n e t w o r k parameters o f interest.
In particular, this t h e o r y should c o n t a i n a t h e o r y or at least a m o r e refined m o d e l
o f the c o n s t a i n t release processes p r o p o s e d in p o l y m e r networks.
Acknowledgement: T h e a u t h o r s are grateful to Prof. K. Du~ek for his c o m m e n t s on
the p r e l i m i n a r y version o f the m a n u s c r i p t and for very helpful discussions.

7 References
1. Treloar, L. R. G. : 'The Physics of Rubber Elasticity"~, 3rd Ed., Clarendon Press, Oxford 1975
2. Du~ek, K., Prins, W. :Adv. Pol. Sci. 6, 1 (1969)
3. Flory, P. J. : "'Principles of Polymer Chemistry", Cornell University, Ithaca, NY 1953
4. Staverman, A. J. : -'Thermodynamics of Polymers" in Handbuch der Physik, Vol. XIII, ed. by
S. Fltigge, Springer-Verlag, Berlin, G6ttingen, Heidelberg I962
5. Staverman, A. J. : Adv. PoL Sci. 44, 73 (1982)
6. Queslel, J. P., Mark, J. E. : Adv. Pot. Sci. 65, 135 (1984)
7. Burchard, W. : Ber. Bunsenges. Phys. Chem. 89, 1154 (1985)
8. Gee, G., Herbert, J. B. M., Roberts, R. C.: Polymer 6, 541 (1965)
9. Yen, L. Y., Eichinger, B. E. : J. Pol. Sci., Pol. Phys. Ed. 16, 121 (1978)
10. Brotzman, R. W., Eichinger, B. E. : Macromotecules t4, 1445 (1981)
11. Brotzman, R. W., Eichinger, B. E. : Macromolecules 15, 531 (t982)
12. Brotzman, R. W., Eichinger, B. E. : Macromolecules 16, 1131 (1983)
13. Beltzung, M., Picot, C., Rempp, P., Herz, J. : Macromolecules 15, 1594 (1982)
14. Beltzung, M., Herz, J., Picot, C. : Macromolecules 16, 580 (1983)
15. Beltzung, M., Picot, C., Herz, J. : Macromolecules 17, 663 (1984)
16. Gronski, W., Stadler, R., Jacobi, M. M. : Macromotecules 17, 74t (1984)
17. Deloche, B., Samulski, E. T. : Macromolecules 14, 575 (1981)
18. Deloche, B., Beltzung, M., Herz, J.: J. Phys. Lett. 43, 763 (1982)
19. Dubault, A., Deloche, B., Herz, J. : Polymer 25, 1405 (1984)
20. Mark, J. E. : Adv. Pol. Sci. 44, t (1982)
21. Erman, B., Flory, P. J. : Macromolecules 15, 806 (1982)
22. Erman, B., Wagner, W., Flory, P. J. : Macromolecules 13, 1554 (1980)
23. Flory, P. J., Erman, B.: J. Pol. Sci., Pol. Phys. Ed. 22, 49 (1984)
24. Queslel, J. P., Mark, J. E. : J. Pol. Sci.. Pol. Phys. Ed. 22, 1201 (1984)
25. Gottlieb, M., Macosko, C. W., Lepsch, T. C.: J. Pol. Sci., Pol. Phys. Ed. 19, 1603 (1981)
26. Gottlieb, M., Macosko, C. W., Benjamin, G. S., Meyers, K. O., Merrill, E. W. : Macromolecules
14, 1039 (1981)
27. Dossin, L. M., Graessley, W. W.: Macromolecules 12, 123 (1979)
28. Pearson, D. S., Graessley, W. W. : Maeromolecules 13, 1001 (1980)
29. Meyers, K. O., Bye, M. L., Merrill, E. W.: Macromolecules I3, 1045 (1980)
30. Kirk, K. A., Bidstrup, S. A., Merrill, E. W., Meyers, K. O.: Macromolecules 15, 1123 (t982)
31. Vall6s, E. M., Rost, E. J., Macosko, C. W.: Rubber Chem. Technol. 57, 55 (t984)
32. Macosko, C. W., Benjamin, G. S. : Pure Appl. Chem. 53, 1505 (1981)
33. Pak, H., Flory, P. J. : J. Pol. Sci., Pot. Phys. Ed. 17, 1845 (1979)
34. Erman, B.: J. Pot. Sci., Pol. Phys. Ed. 19, 829 (1981)
35. Erman, B. : J. Pol. Sci., Pol. Phys. Ed. 21, 893 (1983)
36. Kgstner, S. : Faserforschung u. Textiltechnik/Z. Polymerforschung 27, 1 (1976)
37. Kgtstner, S. : Acta Polymerica 31,444 (1980)
38. Kgstner, S. : Colloid Polym. Sci. 259, 499, 508 (1981)
39. K~istner, S. :Polymer 20, 1327~(t979)
40. Ronca, G., Altegra, G. : J. Chem. Phys. 63, 4990 (1975)
Rubber Elasticity of Polymer Networks: Theories 85

41. Flory, P. J. : J. Chem. Phys. 66, 5720 (1977)


42. Erman, B., Flory, P. J. : J. Chem. Phys. 68, 5363 (1978)
43. Ftory, P. J., Erman, B.: Macromolecutes 15, 800 (1982)
44. Flow, P. J. : Polymer 20, 1317 (1979)
45. Ziman, J, M. : "Models of Disorder '', Cambridge University Press, Cambridge 1979
46. Freed, K. F. : Adv. Chem. Phys. 22, t (1972)
47. Edwards, S. F. : Statistical mechanics of rubber, in: Polymer Networks: Structural and Mechani-
cal Properties (A. J. Chompff, S. Newman, eds.), Plenum Press, New York 1971
48. Edwards, S. F.: Proc. R. Soc. London, Ser. A 351, 397 (1976)
49. Gels, Gelling Processes (Cone proc.): Faraday Discuss. 57, 1 (1974)
50. Gordon, M. : Proc. R. Soc. London, Ser. A 268, 240 (1962)
51. Mikeg, J., Du~ek, K. : Macromoteeules 15, 93 (1982)
52. Martin, J. E., Eichinger, B. E. : J. Chem. Phys. 69, 4588, 4595 (1978)
53. Martin, J. E., Eichinger, B. E. : Macromoleeules 13, 626 (t980)
54. Brereton, M. G., Shah, S. : J. Phys. A 13, 2751 (1980)
55. Etderfietd, D. J.: J. Phys. A 15, 1369 (1982)
56. Tanaka, F. : PROB. Theor. Phys. 68, 164 (1982)
57. Frank-Kamenetskij, M. D., Vologodskij, A. V.: Usp. Fiz. Nauk 134, 641 (1981)
58. Edwards, S. F.: Proc. Phys. Soc. 91, 513(1967)
59. Edwards, S. F.: J. Phys. A l, 15 (1968)
60. Deam, R. T., Edwards, S. F.: Phitos. Trans. R. Soc. London, Ser. A 280, 3t7 (1976)
61. Des Cloizeaux, J., Mehta, M. L. : J. Phys. (Paris) 40, 665 (1979)
62. Des Cloizeaux, J., Ball, R.: Commun. Math. Phys. 80, 543 (1981)
63. Des Cloizeaux, J. : J. Phys. (Paris) 42, L 433 (1981)
64. Aharoni, S. M. : Macromoleeules 16, 1722 (1983)
65. Saito, N., Chen, Yi-der: J. Chem. Phys. 59, 3701 (1973)
66. Prager, S., Frisch, H. L. : J. Chem. Phys. 46, 1475 (1967)
67. Iwata, K. : J. Chem. Phys. 76, 6363, 6375 (1982)
68. Tanaka, F. : Rep. PROB. Pol. Phys. Jpn. 27, 299 (1984)
69. Tanaka, F. : J. Phys. Soc. Jpn. 53, 2205 (1984)
70. Edwards, S. F. : Proc. Phys. Soc. 92, 9 (1967)
71. Edwards, S. F. : J. Phys. C 2, 1 (1969)
72. Thomas, F., Straube, E., Helmis, G. : Plaste Kautsch. 22, 4l 1 (t975)
73. Thomas, F., Straube, E., Helmis, G. : Plaste Kautsch. 25, 577 (1978)
74. Marrucci, G. : Macromolecules 14, 434 (198t)
75. Gaylord, R. J. : Polym. Eng. Sci. 19, 263 (1979)
76. Gaylord, R. J. : Polym. Bull. 8, 325 (1982)
77. Gaylord, R. J. : Polym. Bull. 9, t86 (1983)
78. Gottlieb, M,, Gaylord, R. J. : Polymer 24, 1644 (1983)
79. Graessley, W. W. : Adv. Polym. Sci. 47, 67 (t982)
80. Marrucci, G. : Rheol. Acta 18, 193 (1979)
81. Priss, L. S.: J. Polym. Sci., Polym. Symp. 53, 195 (1975)
82. Edwards, S. F.: Br. Polym. J. 9, 140 (1977)
83. Heinrich, G., Straube, E., Helmis, G. : Z. Phys. Chem. (Leipzig) 258, 361 (1977)
84. Heinrich, G., Straube, E., Helmis, G. : Z. Phys. Chem. (Leipzig) 260, 737 (1979)
85. Heinrich, G., Straube, E., Helmis, G. : PIaste Kautsch. 26, 361 (1979)
86. Heinrich, G. : Z. Phys. Chem. (Leipzig) 261, 188 (1980)
87. Heinrich, G. : Plaste Kautsch. 29, 437 (1982)
88. Bail, R. C., Doi, M., Edwards, S. F., Warner, M.: Polymer 22, 1010 (1981)
89. Langley, N. R. : Macromolecules 1, 348 (1968)
90. Heinrich, G2, Straube, E., Helmis, G. : Polym. Bull. 3, 399 (1980)
91. Heinrich, G., Straube, E. : Acta Polymerica 34, 589 (1983)
92. Heinrich, G., Straube, E. : Acta Polymerica 35, 115 (1984)
93. Heinrich, G., Hube, H. : Plaste Kautsch. 31,329 (1984)
94. Doi, M., Edwards, S. F. : J. Chem. Soc., Faraday Trans. 2, 74, 1789, 1802, 1818 (1978); ibid. 75,
38 (1979)
95. Graessley, W. W., Edwards, S. F. : Polymer 22, 1329 (1981)
86 G. Heinrich, E. Straube and G. Helmis

96. Heinrich, G., Straube, E. : Polym. Bull. 17, 247 (1987)


97. Heinrich, G. : Doctoral Dissertation B, Tech. Hochsch. Leuna-Merseburg 1985
98. Khokhlov, A. R., Ternovskii, F. F.: Soy. Phys.-JETP 90, 1249 (1986)
99. Ferry, J. D. : "Viscoelastic Properties of Polymers", 3rd Ed., Wiley, New York 1980
100. Havr/mek, A. : Potym. Bull. 8, 133 (1982)
101. Gordon, R. J., Schowalter, W. R. :Trans. Soc. Rheol. 16, 79 (1972)
102. Ftory, P. J.: Proc. R. Soc. London, Ser. A 351, 35t (I976)
103. James, H. M., Guth, E.: J. Chem. Phys. 11,455, 472 (1943); ibid. 15, 669 (1947)
104. Hermans, J. J. : Trans. Faraday Soc. 43, 591 (1947)
t05. Flory, P. J.: J. Chem. Phys. t8, 108(1950)
106. Wall, F. T., Flory, P. J.: J. Chem. Phys. 19, 1435 (1951)
107. Wall, F. T.: J. Chem. Phys. ]0, 132, 487(I942); ibid. 11,527(1943)
108. Krigbaum, W. R., Kaneko, M. : J. Chem. Phys. 45, 1505 (1966)
109. Flory, P. J., Hoeve, C. A. J., Ciferri, A.: J. Pot, Sci. 34, 337 (1959)
1t0. James, H. M.: J. Chem. Phys. 15, 651 (1947)
111. Wang, M. C., Guth, E, : J. Chem. Phys. 20, 1144 (1952)
112. Volkenstein, M. V., Gotlib, J. J., Pti~,yn, O. B. : Vysokomol. Soedin. l, 1056, 1069 (1959)
113. Blokland, R.: Elasticity and Structure of Polyurethane Networks. Thesis, Tech. Hochsch.
Delft, University Press Rotterdam (1968)
114. Schwarz, J. : Colloid Polym. Sci. 25t, 2t5 (1973)
115. Schwarz, J.: Polym. Bull. 5, 151 (1981)
t 16. Di Marzio, E. A. : J. Chem. Phys. 36, 1563 (1962)
117. Jackson, J. C., Shen, M. C., McQuarne, D. A. : J. Chem. Phys. 44, 2388 (1966)
118. Dobson, G. R., Gordon, M. : J. Chem. Phys. 43, 705 (1965)
119. Tatara, Y. : J. Chem. Phys. 6t, 1147 (1974)
120. Thirion, P., Chasset, R. : C. R. Acad. Sci., Set. C 264, 958 (1967)
121. Eichinger, B. E. : Macromolecules 5, 496, 647 (t972)
122. Staverman, A, J. :J. Pol. Sci., Pol. Symp. 51, 45 (1975)
123. Heinrich, G., Straube, E., Helmis, G. : Acta Polymerica 31, 275 (1980)
124. Kilian, H.-G. : Polymer 22, 209 (1981); Vilgis, Th., Kilian, H.-G. : Polymer 25, 71 (1984)
125. Kilian, H.-G. : Colloid Polym. Sci. 259, 1089 (1981)
126. Eisele, U., Heise, B., Kilian, H.-G., Pietralla, M. : Angew. MakromoI. Chem. 100, 67 (1981)
127. Tschoegt, N. W., Gurer, C. : Macromolecules 18, 680 (1985), Gurer, C., Tschoegl, N. W.: Ma-
cromolecules 18, 687 (1985),
128. Ilavsk3), M., Du~ek, K. : Polymer 24, 981 (1983)
129. Dugek, K,, Ilavsk), M. in: Elastomers and Rubber Elasticity, Am. Chem. Soc. Symp. Ser.
130. Du~ek, K., tlavsk2~, M. : J. Pol. Sci., Pot. Phys. Ed. 21, 1323 (1983)
131. Ilavsk3~, M., Dugek, K. :J. Pol. Sci., Pol. Phys. Ed. 22, 265 (1984)
132. Burchard, W. : Adv. Pol. Sci. 48, 1 (1983)
133. Dugek, K~: Br. Polym. J. 17, 185 (1985)
134. Dugek, K., Vojta, V.: Br. Polym. J. 9, 164(1977)
135. Lloyd, A. C., Stepto, R. F. T. : Br. Polym. J. t7, 190 (1985)
136. Dugek, K,, Gordon, M., Ross-Murphy, S. B. : Macromolecules ll, 236 (1978)
t37. Pearson, D. S., Graessley, W. W. : MacromoIecules tl, 528 (1978)
138. Miller, D. R., Macosko, C. W. : Macromolecules 9, 206 (1976)
139. Miller, D. R., Valle~, E. M., Macosko,'C. W.: Polym. Eng. Sci. 19, 272 (1979)
140. Valleg, E. M., Macosko, C. W. : Macromolecules 12, 673 (1979)
141. Kawabata, S., Matsuda, M., Tei, K., Kawai, H.: Macromolecules 14, 154, 1688 (1981); ibid.
I5, 160 (1982)
142. lwata, K. : J. Chem. Phys. 83, 1969 (1985)
143. Graesstey, W. W., Pearson, D. S.: J. Chem. Phys. 66, 3363 (1977)
144. Brereton, M. G., Fitbrandt, M.: Polymer 26, 1134 (1985)
145. Garrido, L., Mark, J. E., Clarson, S. J., Semtyen, J. A. : Polym. Commun. 26, 53 (1985)
146. Garrido, L., Mark, J. E., Clarson, S. J., Semlyen, J. A. : Polym. Commun. 26, 55 (1985)
147. Edwards, S. F., Vilgis, Th. : Polymer 27, 483 (1986)
t48. Thirion, P., Weil, T. : Polymer 25, 609 (1984)
149. Queslel, J. P. : Rubber Chem. Technol. 57, 145 (1984)
Rubber Elasticity of Polymer Networks: Theories 87

150. Elyashevich, A. M., Remeev, J. S. : Preprints of the conf. Rubber 84, Moscow 1984, Preprints
p. 3
151. Remeev, I. S., Elyashevich, A. M. : Vysokomol. Soedin., Ser. A 27, 629 (1985)
152. Feynman, R., Hibbs, A. R. : ~Quantum Mechanics and Path Integrals", McGraw-Hill Book
Company, New York t965
153. Sherington, D., Kirkpatrick, S. : Phys. Rev. Lett. 35, 1792 (1975)
154. Verbaarschot, J. J. M., Zirnbauer, M. R.: J. Phys. A 17, 1093 (1985)
155. Blatz, P. J., Sharda, S. C., Tschoegl, N. W.: Trans. Soc. Rheol. 18, 145 (1974)
156. Ogden, R. W.: Rubber Chem. Technol. 46, 398 (1973)
157. Chang, W. V., Bloch, R., Tschoegl, N, W. : A New Measure of Strain to Describe the Mechanical
Response of Elastomer Networks, in: Chemistry and Properties of Crosslinked Polymers (S. S.
Labana ed.), Academic Press, New York 1977
158. Tschoegl, N. W. : Phenomenological Aspects of the Deformation of Elastomeric Networks,
Abstracts of 9th Europhysics Conference on Macromolecular Physics, Jabtonna, Poland 1979
159. Heinrich, G., Hube, H. : Plaste Kautsch. 33, 88 (1986)
160. Friedmann, G., Brossas, J. : Eur. Polym. J. 20, 1151 (1984); J. Appl. Polym. Sci. 30, 755 (1985)
161. Rivtin, R. S., Saunders, D. W. : Philos. Trans. R. Soc. London, Ser. A 243, 251 (1951)
Rehage, G. : Ber. Bunsenges. Phys. Chem. 81, 969 (1977)
162. Rehage, G. : Ber. Bunsenges. Phys. Chem. 81,969 (1977)
163. Oppermann, W. : Ph.D. Thesis, Technical University Clausthal (1979)
164. Guth, E. : J. Polym. Sci., Pt. C 12, 89 (1966)
165. Oppermann, W., Rehage, G. :Colloid Polym. Sci. 259, 1177 (1981)
166. Rempp, P., Herz, E., Borchard, W. : Adv. Pol. Sci. 26, 105 (1978)
167. Allen, G.: Proc. R. Soc. London, Ser. A 351, 361 (1976)
168. Mark, J. E., Sullivan, J. L.: J. Chem. Phys. 66, 1006(1977)
169. Mark, J. E., Rahatkar, R. R., Sullivan, J. L. : J. Chem. Phys. 70, 1794 (1974)
170. Mark, J. E. : Rubber Chem. Technol. 48, 495 (1975)
171. Gordon, M., Roberts, K. R.: Polymer 20, 681 (1979)
172. Allen, G., Holmes, P., Walsh, D. J. : Faraday Discuss. Chem. Soc. 57, 19 (1974)
173. Duplessix, R.: Thesis, Universit~ Louis Pasteur, Strasbourg (1975)
174. Benoit, H., Decker, R., Dupplessix, R., Picot, C., Rempp, P., Cotton, J. P., Farnoux, B.,
Jannik, G.: J. PoL Sci., Pol. Phys. Ed. 14, 2119 (1976)
175. Richards, R. W., Maconnachie, A.: Annual Report of the Institute Laue-Langevin, p. 380
(1980)
176. Bastide, J., Picot, C., Candau, S.: J. Macromot. Sci., Phys. 19, 13 (t981)
177. Helmis, G., Heinrich, G., Straube, E. : Wiss. Z. Tech. Hochsch. Leuna-Merseburg 26, 461 (1984)
178. Gumbrell, S. M., Mullins, L., Rivlin, R. S. : Trans. Faraday Soc. 48, 1495 (1953)
179. Flory, P. J. : Macromolecutes 12, 119 (1979)
180. Gottlieb, M , Gaylord, R. J. : Macromolecules 17, 2024 (1984)
181. De Gennes, P. G.: Scaling Concepts in Polymer Physics, Cornell University Press, Ithaca,
NY 1979
182. Heinrich, G., Straube, E., Helmis, G. : Wiss. ZI Tech. Hochsch. Leuna-Merseburg 29, 259 (1987),
special issue for the IUPAC meeting Macro 87, Merseburg 1987
183. Ball, R. C., Edwards, S. F. : Macromolecutes 13, 748 (1980)
184. Nagy, M. : Colloid Polym. Sci. 263, 245 (1985)
185. Pearson, D. S.: Macromolecules 10, 696 (1977)
186. Clough, S. B., Maconnachie, A., Allen, G. : Macromolecules 13, 774 (1980)
187. Ullman, R. : Macromolecules 15, 582 (1982)
188. Ullman, R.: Macromolecules 15, 1395 (1982)
189. Warner, M., Edwards, S. F.: J. Phys. A 11, t649 (1978)
190. Heinrich, G , Straube, E.: Polym. Bull, 17, 255 (1987)

Editor: K. Dugek
Received October 28, 1986

You might also like