The Entanglement Concept in Polymer Rheolgy

The Entanglement Concept in
Polymer Rheology
WILLIAM W . GRAESSLEY
Chemical Engineering a n d Materials Science Departments, Northwestern University,
Evanston, Illinois 60201, U.S.A.
Table of Contents
1
Introduction
2.
2.1.
2.2.
2,2.1.
2.2.2.
2.2.3.
2.2.4.
2.2.5.
2.3.
2.4.
2.5.
General Considerations . . . . . . . . . . . . . . . . . . . . . . . .
Large Scale Configuration on Concentrated Systems
. . . . . . . . . . .
Theoretical Aspects . . . . . . . . . . . . . . . . . . . . . . . . . .
Evidence o n C h a i n Dimensions from Physical Measurements . . . . . . . .
Collapsed Coil Theories . . . . . . . . . . . . . . . . . . . . . . . .
Evidence o n Chain Dimensions from Chemical Properties . . . . . . . . .
Extended Configuration Theories
. . . . . . . . . . . . . . . . . . .
Short-Range Organization in Concentrated Systems . . . . . . . . . . . .
Structural Aspects of Flow a n d Flow History . . . . . . . . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
7
7
9
12
13
15
16
17
18
3.
Material Properties of Viscoelastic Liquids
. . . . . . . . . . . . . . . .
19
4.
4.1.
4.2.
4.3.
4.3.1.
4.3.2.
4.4.
4.5.
Molecular Models in Polymer Rheology . . . . . . . . . . . . . . . . .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Necklace Models . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spring-Bead Models
. . . . . . . . . . . . . . . . . . . . . . . . .
Configurational Relaxation
. . . . . . . . . . . . . . . . . . . . . .
Response to Flow
. . . . . . . . . . . . . . . . . . . . . . . . . .
Properties of Alternative Models . . . . . . . . . . . . . . . . . . . .
Discussion
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
26
26
27
30
31
35
37
5.
5.1.
5.2.
5.3.
5.3.1.
5.3.2.
5.3.3.
5.3.4.
5.3.5.
5.4.
5.4.1.
5.4.2.
5.4.3.
5.4.4.
5.4.5.
Experimental Results on Linear Viscoelastic Behavior . . . . . . . . . . .

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Very Dilute Solutions . . . . . . . . . . . . . . . . . . . . . . . . .
Concentration Effects a n d Intermolecular Interaction . . . . . . . . . . .
Parameters M e a s u r i n g Intermolecular Interactions . . . . . . . . . . . .
Viscosity Correlations
. . . . . . . . . . . . . . . . . . . . . . . .
Relationships between Low a n d High Concentration Regimes . . . . . . . .
A n o m a l o u s Flow Properties at Intermediate Concentrations . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Concentrated Solutions a n d Melts
. . . . . . . . . . . . . . . . . . .
Zero-Shear Viscosity t/o . . . . . . . . . . . . . . . . . . . . . . . .
Properties of the Plateau M o d u l u s G . . . . . . . . . . . . . . . . . .
Properties of the Terminal Z o n e Spectrum . . . . . . . . . . . . . . . .
The Steady State Compliance j o . . . . . . . . . . . . . . . . . . . . .
Effects of Molecular Weight Distribution
. . . . . . . . . . . . . . . .
38
38
38
40
42
43
46
47
48
48
48
54
58
60
70
Chain Configuration in Amorphous Polymer Systems
............
3
6
W. W. Graessley:
6.6.
Molecular Entanglement Theories of Linear Viscoelastic Behavior

Formal Molecular Theories . . . . . . . . . . . . . . . .
Network Theories
. . . . . . . . . . . . . . . . . . . .
Entanglement Friction
. . . . . . . . . . . . . . . . . .
Theories of Entanglement Friction
. . . . . . . . . . . . .
Discussion of the Friction Theories . . . . . . . . . . . . .
Theories of the Terminal Relaxation Spectrum . . . . . . . .
Theories of the Characteristic Molecular Weights . . . . . . .
Relationships a m o n g Mo, Me, and Mg . . . . . . . . . . . .
Relationship between the Characteristic Molecular Weights
and Molecular Structure . . . . . . . . . . . . . . . . . .
Effects of Polydispersity on r/o and 3,0
. . . . . . . . . . . .
7.
7.1.
7.2.
7.3.
7.4.
7.4.i.
7.4.2.
7.5.
7.6.
7.6.1.
7.6.2.
7.6.3.
Entanglement in Cross-linked Systems

. . . . . . . . . . . . . . .
Rubber-like Elasticity in Networks
. . . . . . . . . . . . . . . .
Tests of Theoretical Modulus V a l u e s - - M o d e l Networks . . . . . . .
Tests of Theoretical Modulus Values--Statistical Networks . . . . .
Determination of Entanglement Density in Networks . . . . . . . .
Theories of Entanglement Trapping . . . . . . . . . . . . . . . .
Experimental Results . . . . . . . . . . . . . . . . . . . . . .
Relaxation Processes in Networks
. . . . . . . . . . . . . . . .
Theoretical Analysis of Topological Constraints . . . . . . . . . .
Effects of Topological Classification . . . . . . . . . . . . . . . .
Topological Models for Entangling Interactions
. . . . . . . . . .
Network Models with Some Non-Classical Character . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
100
101
105
107
110
110
114
117
118
118
121
123
8.
8.1.
8.2.
8.2.1.
8.2.1.1.
8.2.1.2.
8.2.2.
8.2.2.1.
8.2.2.2.
8.3.
8.4.
8.5.
8.5.1.
8.5.2.
8.5.3.
8.5.4.
8.5.5.
8.5.6.
Non-linear Viscoelastic Properties

. . . . . . . . . . .
Introduction
. . . . . . . . . . . . . . . . . . . .
Shear Rate Dependence of Viscosity
. . . . . . . . . .
Correlation of Experimental t/(;?) Behavior . . . . . . . .
F o r m of the Reduced Viscosity Function . . . . . . . . .
Onset of Shear Rate Dependence . . . . . . . . . . . .
Theories of Shear Rate Dependence of Viscosity . . . . . .
Intramolecular Contributions
. . . . . . . . . . . . .
Intermolecular Contributions
. . . . . . . . . . . . .
First Normal Stress Function
. . . . . . . . . . . . .
Second Normal Stress Function
. . . . . . . . . . . .
Non-linear Response in Time-dependent Shearing Flows
Stress Development at the Onset of Steady Shearing Flow
Stress Decay at the Termination of Steady Shearing Flow
Stress Relaxation from Sudden Strains of Large Amplitude
Shear Recovery after steady Shearing Flow
. . . . . . .
Small Oscillations Superimposed on Steady Shearing Flow
Discussion
. . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
125
125
126
128
132
134
139
139
143
148
15t
153
153
154
154
155
155
156
6.
6.1.
6.2.
6.3.
6.3.1.
6.3.2.
6.4.
6.5.
6.5.1.
6.5.2.
9.
Conclusions
10.
Acknowledgments
11.
Nomenclature
12.
References
.
.
.
.
.
.
.
.
.
. . . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
.
.
.
.
.
.
.
.
.
72
74
77
78
79
86
88
96
96
. . . . .
. . . . .
97
99
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . .
. . . . . .
. . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
157
159
160
164
1. Introduction
The concept of chain entanglement first arose more than 40 years ago from
attempts to explain the mechanical properties of amorphous polymers above the
glass transition temperature. In 1932 Busse (I) noted that if unvulcanized rubber
is subjected to a large deformation, held for a short time and released, it recovers
its original shape almost completely, while if held for a long time it flows and
recovers only partially when released. He distinguished between the weak van der
Waals forces between molecules (flexible fibrous units in his terminology) over
most of their length which offer almost no resistance to motion, and a few
widely-separated strong interactions which serve, for short times at least, to bind
the structure into a three-dimensional network. He attributed these strong interactions to a physical interlocking of the molecules which can slip to new equilibrium positions if given time, and he distinguished such temporary links from
the permanent chemical linkages provided by vulcanization.
In 1940 Treloar (2) pointed out that such physical coupling or entanglement
might indeed be expected, given the long flexible nature of individual molecules
and a random, interpenetrating arrangement in the solid. Regions where molecules
were looped through one another might offer high resistance to deformation for
a time, but the loops would eventually slip or be removed and reformed by
random thermal motion. He commented further that most of the observations on
unvulcanized rubber--large, recoverable deformations for short times, stress
relaxation and viscous flow for long times--could be satisfactorily explained if
one regarded entanglements as isolated regions of high viscosity, interconnected by
freely extensible molecular segments.
In 1944, Flory (3) noted that the moduli of cross-linked butyl rubbers
generally differ somewhat from values calculated from the crosslink density
according to the kinetic theory of rubber elasticity. In many cases, the modulus
also depends on the primary (uncross-linked) molecular weight distribution of the
polymer. He attributed both observations to three kinds of network defects: chain
ends, loops, and chain entanglements. The latter are latent in the system prior to
cross-linking and become permanent features of the network when cross-links
are added.
W.W. Graessley:
It was argued that such structures would behave as additional network

constraints, and thereby increase the equilibrium modulus of the network.
In 1946 Green and Tobolsky (4) considered the properties of a rubberlike
network in which the junctions are being continually broken and reformed. They
showed that under certain circumstances the stress could be expressed as a
function of the history of the strain, and that such typical viscoelastic properties
as viscous flow and stress relaxation were natural consequences. Buchdahl (5) in
1948 suggested that the shear rate dependence of viscosity in polymer melts could
be accounted for by a net reduction of entanglement coupling in steady flow.
Nielsen and Buchdahl (6) in 1950 estimated Me, the molecular weight between
coupling points, using creep and stress relaxation data and the modulus
equation from the theory of rubber elasticity. For polystyrene melts they obtained
Me = 10000-40000.
In 1952 Bueche (7) proposed the first model for the extra frictional drag
conferred by entangling interactions and used it to calculate the molecular weight
dependence of melt viscosity for linear chains. The result [as revised in 1956 (8)]
was qoOCM 3"5 for highly entangled chains, in good agreement with the empirical
relation tloOCM 3"4 (9, 10), found at high molecular weights.
In 1955 Ferry, Landel, and Williams (11) applied the Rouse-Bueche molecular
model to concentrated systems, and suggested that the largest relaxation times
(those associated with the motion of portions of the chain longer than the
entanglement spacing) were simply shifted to longer times by entanglement drag,
the short relaxations being essentially unaffected. According to this view the
plateau region in stress relaxation is created by a simple splitting of the
relaxation time distribution. The plateau modulus for high molecular weight
polyisobutylene gave Me = 6700. It was noted that this value was smaller but
nevertheless of the same order of magnitude as the molecular weight at the
break in the qo vs M curve (M~ = 17000 for polyisobutylene).
In t956, Lodge (12) and Yamamoto (13) showed that a wide range of optical
and mechanical properties of deforming polymers could be described quantitatively through the concept of a transient network of flexible random coil
elements, the junction points being specifically identified as chain entanglements. A mechanical constitutive equation for such networks was developed, and a theoretical basis was laid for flow birefringence as a means of
evaluating mechanical properties in concentrated polymer liquids (12, 13a). In
1958, Bagley and West (13b) pointed out the close connection between prominent
shear rate effects in the viscosity and the presence of entanglements.
During this same period, the equilibrium stress-strain properties of well
characterized cross-linked networks were being studied intensively. More complex
responses than the neo-Hookean behavior predicted by kinetic theory were
observed. Among other possibilities it was speculated that, in some unspecified
way, chain entanglements might be a contributing factor.
Two major developments since the mid-1950's have greatly influenced the
entire field of rheology. First, continuum theories have shown that, subject to
very broad and physically plausible assumptions about mechanical response,
many seemingly independent viscoelastic properties of a material are in fact
related. These relationships have been confirmed experimentally and should hold
The Entanglement Concept in Polymer Rheology

regardless of the specific molecular mechanisms for relaxation and flow. Second,
a considerable body of viscoelastic data has been accumulated on well
characterized polymers in concentrated solutions and in bulk. The studies show
that properties which depend principally on the longest relaxation times of the
system vary systematically with concentration, molecular weight and molecular
weight distribution. Broadly speaking, the manner of these variations is independent of the chemical structure of the repeating unit.
Molecular theories, based in large part on ideas about chain entanglements,
have been constructed to explain certain of these observations. The theories
must still be regarded as tentative and incomplete. They are based, first of all, on
reasonable but still incompletely accepted ideas about chain organization in
concentrated solutions and melts. Secondly, they deal with the response of individual chains or pairs of chains in a smoothed medium, rather than with an entire
interacting ensemble. Finally, they circumvent the deep mathematical difficulties
of the central problem, interaction between mutually "uncrossable" sequences of
chain elements, by approximations which are not easy to evaluate. The purpose
of this review is to summarize the present status of entanglement theories and
the data upon which they are based.
Chain entanglement is probably best regarded as a special type of intermolecular interaction, affecting mainly the large-scale motions of the chains, and
through them, the long time end of the viscoelastic relaxation time spectrum.
Evidence for the existence of entangling interactions comes entirely from
mechanical properties, specifically, the flow properties of solutions and melts and
the equilibrium mechanical properties of networks. The effects appear most
prominently at high polymer concentrations and high molecular weights in polymer liquids and large primary chain lengths and low cross-linking densities in the
case of networks. Molecular weight distribution and long chain branching
influence all flow properties, profoundly in some cases, and their effects are
especially important in the entanglement regime. Data are accumulating on these
influences, and a critical review of them would undoubtedly be useful. However,
to keep the current review to a manageable length, and to focus attention on the
entanglement phenomenon itself, we will deal here mainly with the behavior of
linear polymers with narrow molecular weight distribution, making only
occasional reference to polydispersity and branching effects. For similar reasons
we wilt omit discussion of elongational flows and behavior observed in more
complex flows, such as die swell.
Quantitative evidence regarding chain entanglements comes from three principal sources, each solidly based in continuum mechanics: linear viscoelastic
properties, the non-linear properties associated with steady shearing flows, and
the equilibrium moduli of crosslinked networks. Data on the effects of molecular
structure are most extensive in the case of linear viscoelasticity. The phenomena
attributed to chain entanglement are very prominent here, and the linear
viscoelastic properties lend themselves most readily to molecular modeling since
the configuration of the system is displaced for equilibrium only slightly by the
measurement.
In steady shear flows, only the shear-rate dependence of viscosity is well
documented in terms of molecular structure effects. Molecular theories are more
W.W. Graessley:
difficult because the configuration of the system is appreciably displaced from

equilibrium. Indeed the simplest and most attractive molecular models of linear
visoeleastic behavior predict no shear rate dependence whatsoever for the viscosity.
Entanglement concepts enter here in a more indirect and qualitative way: to
explain the existence of a prominent dependence of viscosity on shear rate at high
concentrations and molecular weights, and to account for and predict the forms of
the other non-linear properties. One must, of course, be careful not to attribute
more than necessary specifically to chain entanglement. The viscosity of polymer
systems depends on shear rate even when intermolecular interactions are absent,
and very simple molecular models predict elasticity in polymer liquids without
additional assumptions about entanglement.
Modulus data on crosslinked systems would seem to offer the most direct
method for studying entanglement effects. Certainly, from the standpoint of
molecular modeling, the advantages of equilibrium properties are clear. However,
the structural characterization of networks has proven to be very difficult, and
without such characterization it is almost impossible to separate entanglement
contributions from those of the chemical crosslinks alone. Recent work suggests,
however, that these problems are not insurmountable, and some quantitative
results are beginning to appear.
The sections on experimental results from linear viscoelasticity (Section 5),
entanglement theories in linear viscoelasticity (Section 6), entanglements in networks (Section 7) and non-linear viscoelasticity (Section 8) comprise the major
portion of the review. Section 2 concerns the current evidence on configuration
and arrangement of chains in concentrated polymer liquids. Section 3 is a brief
summary of phenomenological equations in linear viscoelasticity and steady
shearing flow. Section 4 is a summary of molecular models, principally the springbead models, which provide the framework for discussing entanglement effects.
However, in all cases involving experimental data we have tried to carry the
correlations as far as possible before introducing specific molecular interpretations.
This article builds upon an earlier review on the same subject by Porter and
Johnson in 1966 (14), and on the recent treatise on viscoelasticity in polymers by
Ferry (15). We have generally tried to maintain the same nomenclature as the latter.
Recent reviews on the relation between the zero-shear viscosity and molecular
structure (16), crosslinked networks (17), and flow birefringence (18) in this same
journal cover portions of the subject. We have tried to minimise redundancy with
these works while at the same time making the review reasonably self-contained.
2. Chain Configuration in Amorphous Polymer Systems

2.1. General Considerations
Random crosslinking reactions lead quite efficiently to network formation,
requiring that the individual chains have relatively open configurations which
provide plentiful intermolecular contacting. On the other hand, crosslinked networks can undergo large and reversible deformations, a property which demands
a rest state of relatively unextended configurations for the network strands. The
random coil configuration, such as that observed in dilute solutions, fulfills both
requirements. The volume fraction of segments from any individual chain is small
throughout its own pervaded volume. Since the space in concentrated systems
is densely filled with polymer segments, random coils must overlap and interpenetrate extensively. Intermolecular contacts are therefore plentiful for all
repeating units in each chain. At the same time, the end-to-end distance is small
compared to contour length in random coils, so considerable extension is possible
without the breaking of chemical bonds.
Despite its plausibility and the apparent absence of driving forces to produce
other large scale arrangements, the random coil configuration has been questioned
on various grounds. It seems worthwhile to review the evidence at this point,
since most molecular theories for amorphous polymers are based on the random
coil picture.
2.2. Large Scale Configuration in Concentrated Systems

2.2.1. Theoretical Aspects
The random coil model is firmly established for flexible, linear molecules
of high molecular weight in dilute solution (19, 20). Characteristic dimensions
such as the mean end-to-end distance and radius of gyration are controlled by both
local chain structure (bond angle and steric restrictions) and long-range interactions. The latter arise from the mutually excluded volume of segments widely
separated along the chain contour. The magnitude of the perturbation depends
also on the thermodynamic interaction between polymer segments and solvent.
The net perturbation can be reduced to zero (the theta condition) by an
appropriate choice of solvent and temperature.
In unperturbed random coils, the distribution of end-to-end distances for tong
chains is Gaussian:
F(xt,x2,xa)dxidx2dx3=
3
2n (r2)
)a/2
3(x~ + x 2 + x 2)
exp-
2(r2 )
dxldx2dx 3
(2.1)
in which F(xl, x2, xa)dxl dx 2 dxa is the fraction of chains with end-to-end coordinate differences in the range
xl, xl+dXl; x2, x2+dx2;
x3+dxa,
and (r 2) is the mean square end-to-end distance:

( r 2 > = I ~ I(x~ + x~ + x~)F(x~,x2, xa)dx, dx2dx3"
W.W. Graessley:
For random coils, (r 2) is directly proportional to the contour length. If n is the

number of main chain atoms in the chain, (r z) = an. The parameter a is relatively
insensitive to ~nvironment (21), and has been calculated for a number of polymers
from strictly intramolecular considerations using the rotational isomeric model
(22). The root-mean-square distance of segments from the center of gravity of
the coil is called the radius of gyration S. The quantity S 3 is an approximate
measure of the pervaded volume of the coil. For Gaussian coils,
S2 = (r2)/6.
(2.2)
For dilute solutions in good solvents the net excluded volume is positive, and
coil dimensions are expanded beyond their unperturbed values. The expansion
ratio a,
~2 = ( r Z ) / ( r Z ) o '
(2.3)
increases with chain length (19). A repulsive potential exists between chains in
good solvents because of volume exclusion, so the distribution of molecular
centers of gravity in dilute solutions is probably not quite random.
It is generally supposed that coil dimensions in good solvents decrease with
increasing concentration, the intramolecular effects of excluded volume on dimensions gradually being offset by a forced increase in intermolecular contacts.
How rapid the contraction with increasing concentration is not known with
certainty, since unambiguous methods for determining coil dimensions in the
moderate concentration range are not available. The calculations of Fixman and
Peterson (23) and Yamakawa (24) suggest that dimensions in good solvents contract rather rapidly through the dilute range until the coils begin to overlap
(c[q] ~ 1). Above this range the local density of polymer segments becomes
uniform, and the molecular centers again become randomly distributed in the
solution, According to Fixman and Peterson (23), the coils continue a gradual
concentration at higher concentrations, the expansion factor e 2 _ 1 eventually
becoming inversely proportional to polymer concentration. The actual magnitude
of ez _ 1 in the concentrated region of in undiluted polymers is not given by the
theory. Computer simulations of volume-excluding polymers (25, 26) support the
idea of coil concentration with increasing concentration in the dilute range.
In Flory's theory of the excluded volume (27), the chains in undiluted
polymer systems assume their unperturbed dimensions. The expansion factor in
solutions is governed by the parameter (-Z)/v, v being the molar volume of
solvent and ~ the segment-solvent interaction (regular solution) parameter. In undiluted polymers, the solvent for any molecule is simply other polymer molecules.
If it is assumed that the excluded volume term in the thermodynamic theory
of concentrated systems can be applied directly to the determination of coil
dimensions, then :~ is automatically zero but v is very large, reducing the expansion to zero.
Taken literally, however, this viewpoint leads to rather strange variations of

dimensions at intermediate concentrations. Thus (27),
~5
K{1
(2.4)
in which K is a concentration independent parameter for the given polymersolvent system. From the standpoint of any polymer molecule in the system the
surroundings consist of a mixture of solvent and other polymer molecules. The
mean interaction coefficient with this two-component environment might
reasonably be taken to be Zo(1 -q~), where Xo is the interaction coefficient with
pure solvent as the surroundings and q~ is the volume fraction of polymer. The
molar volume of the environment is
v = 1/(q~/% + (1 - ~o)/vs) ,~, vs/(l - ~o),
v~and Vpbeing the molar volumes of pure solvent and polymer, and Vp~>vs. Thus:
K 1
ct5 - ct3 = -~-[-~- - Xo(1- ~o)](1 - q~).
(2.5)
Accordingly, coil dimensions in typical good solvents (~0 = 0.4) should expand at
intermediate concentrations, pass through a maximum near ~0= 0.35 and finally
subside slowly toward the theta dimensions at very high concentrations. Even
without the modification of Z, the molar volume effect makes ~5 _ ~3 proportional to 1 -q~, which suggests that coil dimensions are still only partially contracted even out to rather high concentrations.
2.2.2. Evidence on Chain Dimensions from Physical Measurements

Krigbaum and Godwin (28) have estimated the end-to-end distance of individual chains in undiluted polystyrene, using low angle X-ray scattering. Polystyrene
molecules (.h~rn = 87000) which had been tagged at both ends by silver atoms
were dispersed at a concentration of 5% in untagged polystyrene. Analysis of the
excess scattering yielded (r 2) 1/2 = 269/~. The value in 0-solvents for this molecular
weight is 235 A.
Very recently, results from low angle neutron scattering with undiluted
mixtures ofdeuterated and ordinary polymers have begun to appear. In these cases
all units along the chain contribute to the excess scattering, and the radius of
gyration is obtained. Cotton et al. (29) have measured the dimensions of a rather
10
W.W. Graessley:
low molecular weight (M = 8000) deuterated polystyrene in ordinary polystyrene,

obtaining S = 28-30 ]~ (S o = 26 ~). Ballard et al. (30) dispersed ordinary polystyrene (Mw=97100; Mw/Mn= 1.06) to a concentration of 5% in a matrix of
93% deuterated polystyrene (M, = 99200; Mw/Mn = 1.1), obtaining S = 90 A
(So = 84 A).
Kirste et al. (31) dispersed ordinary polymethyl methycrylate (M,v= 250000;
Mw/Mn = 1.36) to a concentration of 1.2% in completely deuterated polymethyt
methacrylate (M, = 250000; M w / M , = 2.1). They report S = 126 A, compared to
So = 110 A, and S = 170 A in dilute solutions of an athermal solvent. The same
workers have recently studied the concentration dependence of the excess
scattering (constructing in effect a Zimm plot) and obtain S = 121 A at infinite
dilution in the deuterated matrix, with A2 ~ 0 and a molecular weight (by neutron
scattering measurements alone) Mw = 220000 (32). Moreover, they have determined dimensions for a series of molecular weights (Mw ranging from 80000 to
1000000) and find S ~ M '49, although within their limits of error the 0-solvent
exponent of 0.5 fits also.
The results of the scattering experiments are shown in Table 2.1. They leave
little room for doubt as to the essential correctness of Flory's deduction about
chain dimensions in undiluted polymers.
Table2.1. Chain dimensions in undiluted amorphous polymers obtained by scattering

experiments
Polymer
)14w
Polystyrene
87000~
Polystrene
Polystrene
Polymethyl
methacrylate
Mw/Mn
S ]~
unreported 269b
but probably
narrow
8000 unreported 28-30
97100 1.06
250000
1.36
So ~
Method
235 b
X-ray scattering from

chains with Ag-tagged
ends (8)
Neutron scattering from
deuterated chains (10)
Neutron scattering using
deuterated substrate (10a)
Neutron scattering using
deuterated substrate (t 1)
26
90
84
130
110
" Only/~rn = 87000 was reported.

b (r2)llz rather than S was measured.
Bueche et al. (33) determined chain dimensions indirectly, through measurements of the diffusion coefficient of C13-tagged polymers in concentrated
solutions and melts.The self-diffusion coefficient is related to the molar frictional
coefficient N~n(o through the Einstein equation:
D-
RT
Nan~o
(2.6)
11
in which Na is Avogadro's number, n is the number of main chain atoms per

molecule, and (0 is the frictional coefficient per main chain atom. The viscosity
of concentrated solutions or melts in the free-draining Bueche-Rouse model (7) is:
,7 -
Nan(oc(r 2>
36M
(2.7)
Elimination of the frictional coefficient between Eqs.(2.6) and (2.7) yields
Dt 1
cRT
(r2>
36M "
(2.8)
For polystyrene fractions in diethyl phthalate solution ( 3 0 0 0 0 < M < 200000;

0 . 3 < c < 0 . 6 gm/ml) the group on the left had an average value of 1.6 x 10 -18
( _ 50%). In dilute solution (r2>o/36M is 1.27 10 -18 for polystyrene (21). No
systematic variations with concentration, molecular weight or temperature were
apparent, the scatter of the data being mainly attributable to the experimental
difficulties of the diffusion measurements. The value of Dtl/cR T for an undiluted
tagged fraction of poly(n-butyl acrylate) in pure polymer was found to be
2.8 x 10-18. The value of <r2>o/36M is 1.6 x 10-18, based on dilute solution data
for other acrylate polymers (34). Thus, transport behavior, like the scattering
experiments, supports random coil configuration in concentrated systems, with
perhaps some small expansion beyond 0-dimensions.
Semi-quantitative arguments in favor of the random coil have been made on
the basis of mechanical behavior. For example, Bueche and co-workers (35)
estimated molecular dimensions from an analysis of fracture envelopes in rubbery
polymers. They suggested that the elongation at break corresponds to a maximum
extension of strands between cross-links or entanglement junctions. The end-toend distance for strands in the undeformed state (r2> 1/2 was obtained from the
molecular weight between junction points, estimated from other experiments, and
the macroscopic deformation at break. Four different polymers were examined;
the values obtained for (r2> in each case fell within 30% of corresponding
0-solvent values. Later work by Smith and Frederick (36) on a wide variety of
vutcanizates showed that Bueche's equations indeed gave correct orders of
magnitude for (r2>, but the residual variations among polymers did not
correlate very well with independently measured 0-dimensions.
Hoffman (37) has offered a variety of circumstantial evidence supporting the
random coil model. In A-B block copolymers of styrene and butadiene, for
instance, the characteristic dimension of the dispersed phase particles depends on
the molecular weight of blocks in the dispersed phase according to:
D = K M 'Ss .
(2.9)
12
W.W. Graessley:
The exponent would be 0.33 for collapsed coils, 0.5 for random coils, and 1.0 for
fully extended chains. The observed values of D lie within a factor of three of the
dilute solution values for (r2) 1/2 of molecules with the same molecular weight
as the blocks.
2.2.3. Collapsed Coil Theories

Other chain conformations have been suggested for concentrated systems.
Maron and co-workers (38) have proposed that the molecules collapse into
increasingly compact coils as concentration increases, with relatively little interpenetration and segment mixing even at very high concentrations. They base this
conclusion on solution viscosities (0-20% polymer) in several polymer systems,
analyzed with an equation which had previously been found useful in correlating
viscosity in suspensions of spherical particles. In the work on suspensions, the
viscosity had been found to be principally a function of the volume fraction of
solvent, expressed as 1 - e~. In the case of suspensions q~ is the volume fraction
of particles and e is a factor interpreted in terms of the maximum packing
density of spheres. In suspensions, e is a constant, equal to 1.35. In polymer
solutions the empirically deduced value of e is a decreasing function of concentration with the form:
e0
(2.10)
l+(~o-~)q~
where ~o is the volume fraction of polymer in the system, eo is [q] Q/2, e0~ is
approximately 4.0, Q is the density of undiluted polymer, and [q] is its intrinsic
viscosity in the solvent in question. Particle-like flow properties are indeed
observed in the dilute to moderately concentrated range for polymer solutions
(see Section 5). Moreover the idea of coil contraction with increasing concentration, due to cancellation of excluded volume in good solvents, is quite a
reasonable idea, as noted earlier. However, the authors go on to infer from this
result that coil contraction continues unabated into the regime of concentrated
solutions and melts, with relatively little intermolecular contact except at the coil
periphery at all concentration levels.
Maron has incorporated this collapsing coil picture into a thermodynamic
theory of polymer solutions (39).The free energy of mixing for n 1 moles of solvent
and nz moles of polymer at constant volume turns out to be the Flory-Huggins
expression with two modifications, an additional term containing the contraction
ratio e/Co and a somewhat different handling of the interaction coefficient )~, the
latter having nothing directly to do with coil contraction:
AGM=RT[n~tn(1-q~)+n21nq)+n21n@o
+ zn~q~].
(2.11)
The EntanglementConcept in Polymer Rheology
13
This equation was then applied to a variety of experimental data on solvent

activity as related to vapor pressure (40), osmotic pressure in dilute and concentrated solutions (41), and light scattering (42).
Both modifications affect the analysis of dilute solution behavior, and it is
difficult to judge how much the e/eo term is actually needed. In any case, as the
authors themselves point out (41), the e/e0 term makes an entirely negligible
contribution to solvent activity in concentrated solutions. For example, simple
calculations yield a contribution of approximately 1% in a 10% solution of natural
rubber in benzene at 30 C (M = 500000, [t/]Q = 250, ~(= 0.4). It is therefore clear
that thermodynamic measurements can furnish no evidence for or against
continued collapse in concentrated solutions.
Aharoni has stated that the observed rates of crystallization in polymers are
inconsistent with the times required for random-coil molecules to separate themselves from the melt, and claims this as support for the collapsed coil model (43).
No numerical comparisons are given, and it is difficult therefore to judge the basis
for his assertion.
2.2.4. Evidence on Chain Dimensions from Chemical Properties

Vollmert and Stutz (44) suggest that they have found support for a partially
collapsed coil in concentrated solutions from an extensive study of the reactions
between functional groups located on different polymer molecules. They prepared
copolymers of n-butyl acrylate, one with 5% alcohol functions, the other with 5 %
acid chloride functions. Equal amounts of these copolymers were dissolved in
benzene and allowed to mix for several hours at room temperature. Reaction to
form ester cross-linkages was then brought about by heating to 70 C. The extent
of reaction was much lower than when the copolymer containing alcohol
functions was cross-linked with equivalent amounts of a monomeric dichloride.
To separate out segregation effects due to thermodynamic incompatibility,
gelation experiments (on the same acrylic substrate but with an alcohol-isocyanate
curing system this time) were conducted over a range of copolymer composition
and extrapolated to 0% reactive functions. Segregation effects are rather clearly
evident even with relatively low copolymer contents. The extrapolation is
therefore a difficult and sensitive one. The values obtained by the authors are
given in Table 2.2.
Table 2.2. Extents of reaction in acetone between alcohol and isocyanate functions on
differentpoly(n-butylacrylate)molecules,extrapolated to 0% reactive functions
Polymerconcentration(%) Extentof reaction (%)
10
15
20
25
30
2
5
9
t3.5
17
14
W.W. Graessley:
The authors propose that these extents of reaction reflect the degree of coil
overlap at each polymer concentration. They conclude that each molecule in
concentrated solution has a segregated core, and that the intermolecular contact
occurs only near the periphery. As the concentration increases the overall
molecular size must shrink correspondingly. [We note in passing that fluorescence
and fluorescence quenching by groups attached to different polymer molecules
appears to be a promising alternative method for analyzing segregation in concentrated systems (45), although no results applicable to questions of coil
dimensions have appeared as yet.]
It seems impossible to reconcile such collapsed coil models with a wealth of
other observations on chemical reactivity in amorphous polymers. For example,
any significant collapse of molecular size in concentrated systems should cause
deviations from the predictions of random cross-linking theory. The theory
assumes that all repeating units in a molecule have equal probabilities of undergoing cross-link formation. Only intermolecular cross-links can be effective in
developing a gel structure. Collapsed coils would result in appreciable amounts
of intramolecular cross-linking, and the intermolecular cross-linking density
would be proportional to the molecular surface a r e a , M 2/3, rather than total
molecular weight M as in the case of intermingling coils. Gel point dosages Rg,
corrected for chain scission and small amounts of end-linking, are shown in
Table 2.3 for polystyrene films which were cross-linked by gamma radiation (46).
Table 2.3. Gel points in polystyrenecross-linked by gamma radiation

~t
Rg(megarads) RgMw
Rg~12w/a
0.0562 10 6
0.082
O.186
0.329
0.362
0.915
1.87
381.0
256.0
105.0
58.4
63.6
22.8
12.0
2.1 107
2.1
1.9
1.9
2.3
2.1
2.2
5.6 105
4.9
3.4
2.8
3.2
2.2
1.8
2.3
1.8
2.1
1.4
1.90
3.37
11.9
6.33
The total number of cross-links per gram is directly proportional to radiation

dose. Either Rg/~rwo r Rg(/~w)2/3 should be independent of initial molecular weight,
depending on whether all repeating units or only those near the surface of the
coils have equal probabilities for intermolecular cross-link formation. Complete
coil collapse is clearly ruled out by the results. Moreover, detailed calculations
(46) show that a trend in the product RgMwdue to systematic differences in intramolecular cross-linking, should be observed even for only partially collapsed
random coils. If the coil radii in bulk polystyrene were reduced by even as much
15
as a factor of two from their 0-solvent values, a different of 15% between the
extremes of molecular weight would be expected. No such trend is observed.
The frequency of intramolecular cross-linking can be estimated from the
material balance on Hz in linear polyethylene irradiation. The total number of
cross-links (obtained from gel point determination) plus the number of unsaturated
links formed (obtained from infrared measurements) minus the number of chain
scission reactions (obtained from the limiting solubility at high radiation doses)
should be equal to the number of H2 molecules evolved, assuming all cross-links
are intermolecular. There does indeed appear to be a discrepancy in the material
balance for irradiation in the melt state, implying some intramolecular linking,
although there is some dispute about its actual magnitude (47, 48). In any case,
the discrepancy appears to be independent of molecular weight, an unlikely
circumstance for collapsed coils since intramolecular cross-linking would surely
vary with molecular weight for that configuration.
More quantitative chemical evidence for random coil configuration comes
from cyclization equilibria in chain molecules (49). According to the random coil
model there must be a very definite relationship among the concentrations of
x-mer rings in an equilibrated system, since the cyclization equilibrium constant
K~ should depend on configurational entropy and therefore on equilibrium chain
and ring dimensions. Values of (r2)/M deduced from experimental values on Kx
for polydimethylsiloxane, both in bulk and in concentrated solution, agree very
well with unperturbed dimensions deduced from dilute solution measurements(49).
2.2.5. Extended Configuration Theories

Kargin and co-workers (50) have published extensively on supermolecular
structures in crystalline polymers. Based principally on the surface morphologies
observed by electron microscopy, Kargin has proposed that even in amorphous
polymers the chains are aligned in loose bundle-like structures which persist over
distances which are large compared to molecular dimensions. Direct supporting
evidence for such expanded chain configurations has not been forthcoming
however.
Tager and co-workers (51) have invoked bundle structures to explain
correlations between the vfscosities of concentrated polymer Solutions and the
thermodynamic interactions between polymer and solvent. They note, for example,
that solutions of polystyrene in decalin (a poor solvent) have higher viscosities
than in ethyl benzene (a good solvent) at the same concentration, and quote
a number of other examples. Such results are attributed to the ability of good
solvents to break up the bundle structure; the bundles presumably persist in poor
solvents and give rise to a higher viscosity. It seems possible that such behavior
could also be explained, at least in part, by the effects of solvent on free volume
(see Section 5). Berry and Fox have found, for example, that concentrated
solution data on polyvinyl acetate in solvents of quite different thermodynamic
interaction could be reduced satisfactorily by free volume considerations alone
(16). Differences due to solvent which remain after correction for free volume
16
w . w . Graessley:
effects might reasonably be attributed to association, although not necessarily

to bundle structures.
Tager and Dreval (52) have also noted that the shear rate dependence of
viscosity in concentrated solutions varies with the thermodynamic character of
the solvent. In appropriately reduced form, the onset of shear rate dependence
appears to be independent of solvent character (53), but at high shear rates the
viscosity levels off in poor solvents while continuing to decrease in good
solvents (52). The bundle picture is again invoked, the argument apparently being
that some bundles are so strong in poor solvents that they prevent the viscosity
from falling beyond on a certain point. Viscosity anomalies have been reported
by others for solutions of moderate concentration near the theta temperature
(54, 55). However, in these cases the anomalies were found in company with
critical point concentration fluctuations (54) or with the actual separation of a
second phase (55).
2.3. Short-Range Organization

Local str ucturat features have been postulated for amorphous polymer systems,
based on the asymmetry of chain-like molecules. Flory (56) has shown that
molecular asymmetry in itself is no barrier to a dense random packing of the
chains are sufficiently flexible. Robertson (57) suggests, however, that some
degree of local alignment is required simply to accomodate linearly connected
sequences in the rather limited space available. Unfortunately, Calculations of
local cooperative effects are extremely difficult and sensitive to specific assumptions
about available packing arrangements.
Small nodular regions (D ~ 50-150 A) have been observed by electron microscopy in thin films of amorphous but crystallizable polymers such as polyethylene
terephthate (58) and polycarbonate (59). Scattering regions of approximately the
same size have also been detected by low-angle X-ray diffraction (60). Yeh has
observed similar structures in atactic polystyrene (61). The nodule diameter in the
latter polymer is approximately 30 A, corresponding to a grouping of about
300 repeating units. Yeh also finds evidence for strongly diffracting regions by
dark field examination of the amorphous halo (62). He has proposed that there
are small regions even in non-crystallizable polymers where the chain segments
are highly ordered, and suggests that the properties of amorphous polymers may
in fact be strongly influenced by such structures (63). The packing in these regions
cannot be much different from the remainder of the system, however, since even
in polycarbonates the density difference can be no more than about 2% (39).
Flory has recently summarized the experimental evidence pertaining to local
correlation and their effects on chain dimensions (49). There is experimental
support for local alignment from optical properties such as stress-optical
coefficients in networks (both unswelled and swelled in solvents of varying
asymmetry), and from the depolarization of scattered light in the undiluted state
and at infinite dilution. The results for polymers however, turn out to be not
greatly different from those for asymmetric small molecule liquids. The effect of
17
these correlations on thermodynamic and mechanical properties appears to be

very small in non-polar polymers. Molecular asymmetry of the swelling solvent
has no discernable effect on the modulus of polyethylene (64) and polyisoprene
(65) networks, after the usual corrections for variations with swelling ratio are
made. The temperature coefficient for chain dimensions d ln(r2)/dT obtained
from modulus measurements on networks agrees very well with values from dilute
solution measurements (49). The entropy of melting and the activity of solvent in
concentrated solutions appear to be understandable without postulating special
polymeric association or alignment effects (66).
Thus, although some degree of local organization may indeed occur in
amorphous systems, and may even have some effect on the mechanical properties
of polymers in the glassy state, the influence on the mechanical properties of
melts, concentrated solutions and networks appears to be negligible.
2.4. Structural Aspects of Flow and Flow History

Molecular theories of flow behavior are applied on the assumption that the
macroscopic velocity field can be considered to apply without modification right
down to the molecular scale. In continuum theories the components of relative
velocity in an arbitrarily small neighborhood of any material point are taken
to be linear functions of the spatial coordinates measured from that point, i.e.,
the flow is assumed to be locally homogeneous. The local velocity field is calculated from the macroscopic velocity field. This property of local homogeneity
of flow is an obvious prerequisite for any meaningful macroscopic analysis, and
perhaps the fact that analyses are at all successful and that flow properties can be
determined which are independent of apparatus geometry constitutes a fair test
of the assumption.
Certain systems may violate the assumption. Locally inhomogeneous deformation has been observed in thin films of atactic polystyrene below the glass
temperature (67). Melts formed from polyvinyl chloride particles retain a
particulate appearance in some cases even after extrusion (68). Collins and
Krier (68a) have shown that molecular weight changes during flow have
little influence on apparent viscosity in polyvinyl chloride below 190-200 C.
At higher temperatures the viscosity responds immediately when molecular
weight changes. The clear implication is a change in the local flow pattern near
200 C from particulate flow to molecular flow.
Schreiber and co-workers have noted very persistent history effects in linear
polyethylenes (69). Fractions which have been crystallized from dilute solution
required times of the order of hours in the melt state at 190 C in order to attain
a constant die swell behavior upon subsequent extrusion. The viscosity on the
other hand reached its ultimate value almost immediately. The authors concluded
from this result that different types of molecular interactions were responsible for
elastic and viscous response. However, other less specific explanations might also
suffice, since apparent viscosity might be relatively intensitive to the presence of
incompletely healed domain surfaces, while die swell, requiring a coordinated
motion of the entire extrudate, might be affected by planes of weakness. It would
18
W.W. Graessley:
be interesting to repeat such experiments in a plate-cone instrument where elastic

response can be measured more directly.
Ballman and co-workers have used carbon particles to determine flow patterns
for polystyrene melts in plate-cone and capillary viscometers (70). Complex
patterns, rather than the simple flow expected, were observed for high molecular
weight samples. These may have been caused, however, by differences in viscosity
between adjacent layers of pure melt and melt with suspended particles.
Mooney earlier had suggested a kind of ball-bearing flow for unvulcanized
rubber in plate-cone viscometers, based on an observed rapid rate of diffusion of
dye in the direction of the velocity gradient (71, 72). Flow instabilities of various
kinds, melt fracture in capillary flow and spontaneous extrusion from the gap in
plate-cone flows, are known to be associated with high shear rates in viscoelastic
materials. Judged from the combination of high shear rates and viscosities in the
Mooney experiments, it seems likely in retrospect that the rapid mixing behavior
was obtained under conditions of unstable flow. Most recent studies of theological
properties seek to avoid conditions which lead to such flows. Locally homogeneous
deformations apparently can be achieved in systems which are initially uniform
if the deformations or deformation rates are sufficiently small. The onset of flow
instabilities appears to depend on flow geometry, and occur at much higher shear
rates for capillary flows than plate-cone flows.
A related question concerns the effect of prior flow history on the physical
structure of the system. Flow would be expected to produce a partial alignment of
the chains. This orientation can be frozen in by rapid cooling but should be lost
through configurational relaxation when the sample is reheated to the melt state.
Wissbrun has noted a reduction in both viscosity and die swell in branched potymethylene oxide after an initial extrusion (73). Prolonged annealing restored only
part of the original behavior, and a more complete recovery of properties was
only achieved by dissolving and re-precipitating the polymer.
Schreiber and co-workers have found some indication of molecular fractionation during capillary flow, with smaller molecules migrating preferentially
towards the wall and larger molecules toward the center line (74). This conclusion
has been disputed recently by Porter and co-workers however (75).
Numerous experiments have shown that recovery from previous flow history
is smooth and relatively rapid in polymer melts and solutions. Recovery is
especially rapid for systems of narrow molecular weight distribution. Agreement
between different investigators, and agreement in rheological information obtained
in different flow geometries also suggest that effects of prior history, although
crucial in solid state properties, is not a significant problem in most rheological
studies.
2.5. Summary
Both direct measurements and a wealth of compelling indirect evidence
support the random coil, perhaps slightly expanded beyond 0-dimensions, as the
correct description of large-scale chain configuration in amorphous systems of
flexible, relatively non-polar polymers. It is certain that neither greatly contracted
19
nor highly extended configurations can be reconciled with these data. Evidence
does exist for short-range intermolecular correlations which are probably related
to molecular asymmetry. The degree of orientation correlation is similar in extent
to that in systems of asymmetric small molecules. Its effect on the mechanical
and thermodynamic properties of polymer systems appears to be negligible in the
cases so far investigated. Judged by agreement among investigators using different
samples and flow geometries, the effects of preparation history on flow properties
is negligible in narrow distribution linear polymers, and the flow pattern itself
is locally homogeneous in the absence of various types of viscoelastic instability.
3. Material Properties of Viscoelastic Liquids

This section summarizes results of the phenomenological theory of viscoelasticity as they apply to homogeneous polymer liquids. The theory of incompressible simple fluids (76, 77) is based on a very general set of ideas about
the nature of mechanical response. According to this theory the flow-induced
stress at any point in a substance at time t depends only on the deformations experienced by material in an arbitrarily small neighborhood of that point in all
times prior to t. The relationship between stress at the current time and deformation history is the constitutive equation for the substance.
The stress tensor describes the forces transmitted to an element of material
through its contacts with adjacent elements (78). Traction is the force per unit
area acting outwardly on the material adjacent to a material plane, and transmitted through its contact with material across the plane. If the components of
traction are known for any set of three planes passing through a point, the
traction across any plane through the point can be calculated. The stress at a
material point is determined by an assembly of nine components of traction, three
for each plane. If the orientations of the three planes are chosen to be normal
to the coordinate directions of a rectangular Cartesian coordinate system, the
Cartesian components of the stress are obtained:
FPli
12 P131
(3.1)
LP31 P32 P33J

The stress tensor is symmetric, so Pij = Pji. Furthermore, addition of an isotropic pressure has no effect on the physical properties of incompressible substances, so only differences among the normal components of stress are governed
by the constitutive equation. Thus, there are at most only five independent
terms of rheological significance in p, three shear components (pl 2, Pl 3, and P23, for
example) and two differences in the normal components (Pl i -P22 and P22 -P33,
20
W.W. Graessley:
for example). In incompressible simple fluids these attributes of the stress are
governed only by the history of the relative deformation gradient. The property
of fading memory requires that recent strains relative to the current configuration
be weighted more heavily in determining the current stress than strains in the
more distant past.
The response of simple fluids to certain classes of deformation history can be
analyzed. That is, a limited number of material functions can be identified which
contain all the information necessary to describe the behavior of a substance in
any member of that class of deformations. Examples are the viscometric or steady
shear flows which require, at most, three independent functions of the shear rate
(79), and linear viscoelastic behavior (80, 81) which requires only a single function,
in this case a relaxation function. The functions themselves must be determined
experimentally for each substance.
The utility of the simple fluid theory lies in the plausibility and generality of
its assumptions about how materials behave and in the exactness with which its
conclusions are worked out. In particular, one is inclined to believe, as a
working hypothesis and in the absence of contradictory evidence, that the theory
is general enough to encompass the behavior of homogeneous polymeric liquids.
On this basis the role of molecular theories is a complementary one, to provide
forms for the material functions and to account for their systematic change with
molecular structure and temperature.
Most rheological data on polymer liquids of known structure has been obtained in simple shearing deformations. The velocity field for homogeneous simple
shear in rectangular Cartesian coordinates may be expressed:
/91 = ~(t)X2
(3.2)
V2 = 0
/)3 = 0
/
3
in which the coordinates xi(i = 1, 2, 3) are measured from some suitably chosen
fixed origin, and in which 1 denotes the direction flow, 2 is the direction normal
to the shearing planes, and 3 is the third or neutral direction. The shear rate,
~}(t), characterizes the flow field; the current stress depends on the history of the
21
shear deformation 7(t):

t
7(0 = ?'(0) + S f~(s)ds
(3.3)
i.e., the values of the shear deformation at all past times.

Symmetry arguments suffice to show that P13 and P23 are identically zero for
this deformation field (79):
(3.4)
P22
0
P33
Thus, the effects of deformation on the stress tensor reduce to the effect of the
values of y ( t - s ) from s = 0 to s = oo on the values of Px2 (=P21), Pll -P22, and
P22 - P3a at time t. The shear stress will be denoted by a in all further discussions.
For steady shear flow, the shear rate ~ is constant for all past time. Since deformation history now depends only on the parameter 3, the stress components
become functions of ~ alone:
G=~(;);
(3.5)
Pl 1 -- P22 = Nl = ~PI(~)~2
(3.6)
p ~ - p~3 = N2 = ~ 2 ( ; ) ; ~.
(3.7)
I~-~--~.pz~= ~r
=
P.
D-1
3
P3~
IP~2
The symbols N1 and N 2 denote the normal stress functions in steady state shear
flow. Symmetry arguments show that the viscosity function q(~) and the first and
second normal stress coefficients ~z(~) and ~z(~) are even functions of ~;. In the
22
W.W. Graessley:
limit of zero shear rate q(~) and ~1(??) become constants, termed the zero-shear
viscosity r/o and the zero-shear normal stress coefficient 7/1(0) respectively.
Of major interest in this review are q(~) and 7~1(0) for which a large quantity
of data has now been accumulated on well-characterized polymers. Some limited
information is also available on the shear rate dependence of 7~1. The second
normal stress function has proved to be rather difficult to measure; N) appears to
be negative and somewhat smaller in magnitude than NI (82).
Viscoelastic behavior is classified as linear or non-linear according to the
manner by which the stress depends upon the imposed deformation history (80).
Insteady shear flows, for example, the shear rate dependence of viscosity and the
normal stress functions are non-linear properties. Linear viscoelastic behavior
is obtained for simple fluids if the deformation is sufficiently small for all past
times (infinitesimal deformations) or if it is imposed sufficiently slowly (infinitesimal rate of deformation) (80, 83). In shear flow under these circumstances, the
normal stress differences are small compared to the shear stress, and the expression for the shear stress reduces to a statement of the Boltzmann superposition
principle (15, 81):
~r(t) = S G(s) ~(t - s) ds.
(3.8)
The shear stress relaxation modulus of the fluid, G(t), is a monotonically decreasing function of time, with G(oe)= 0. If the fluid initially at rest is given a
small shear deformation 7o at t = 0, the shear stress at later times becomes simply:
g(t) = ?o G(t).
(3.9)
The value G(0) is called the instantaneous modulus, and will be denoted by G.
It turns out that stress relaxation following a simple shear deformation is
seldom employed experimentally. A more common technique is to measure the
steady state response to small sinusoidal deformations as a function of angular
frequency o9. The dynamic storage modulus G'(o)) and loss modulus G"(~o) in small
sinusoidal deformations are related to G(t):
G'(co) = 09 ~ G(t) sinogt dr,
(3.10)
G"(~o) = ~o ~ G(t)cos lot dt.
(3.11)
The instantaneous modulus is given by:
Go = 2
~ G"(co)dlnc9.
(3.12)
The Entanglement Concept in Polymer Rheoiogy
23
In creep and creep recovery experiments, the stress is imposed and the
deformation is observed. Equation (3.8) can be inverted to describe those cases:
7(t) = ~ J(s) gr(t -s) ds.
(3.13)
The shear creep compliance, J(t), is related to the relaxation modulus through:
t
(3. t4)
t = ~ G(t - s) J(s) ds.

0
If a sufficiently small constant shearing stress ao is imposed on the material at

t = 0, the subsequent deformation is given by
(3.15)
7(0 = J(t) a o .
At long times, the deformation rate becomes constant, and the creep function
approaches a straight line:
(3.16)
J(t) = jo + t/~lo
in which ~/o is the zero shear viscosity and jo is the steady state recoverable
shear compliance, or more simply the steady state compliance. If the stress is
removed at a time to, the liquid recoils and a portion of the total deformation
is eventually recovered. If to lies in the region of steady state deformation, the total
recoverable shear or elastic recoil yr is
7r = jo tr0.
(3.17)
Both jo and r/o can be evaluated from the response in other linear viscoelastic
experiments also.
rlo = S
G ( t ) d t = lim G"((0_____)),
t~O
(3.18)
(0
c~3
j o = ~-o ! tG(t) dt = ~
lim G'((0)
~/0 oJ--,0
(0 2
(3.19)
24
W.W. Graessley:
Behavior in steady state shearing flows likewise provide values of these parameters
(83):
r/o = ~im r/(~),
(3.20)
1
jo = 2~/o
2 Jim ~1('~).
(3.21)
The factor of two in Eq. (3.21) was a matter of some dispute until recently, but it
now seems well established (84).
The linear viscoelastic properties are often expressed in terms of an auxiliary
function, the relaxation time distribution, H(z); H(z)din z is the portion of the
initial modulus contributed by processes with relaxation times in the range
lnz, l n z + d l n z :
G(t) = S H(z) e -'/'dlnz.

--
(3.22)
ct3
This alternative representation is entirely arbitrary of course, but its use is well
established in the literature. It offers some advantage in discussing the connections
between the macroscopic and microscopic descriptions of viscoelasticity. The
parameters, ~/o, jo, and G O can be expressed in terms of the moments of H(z):
GO= ~ H(z) din z,
(3.23)
--00
rlo = ~ z H(z) d In z,
(3.24)
z 2 H(z) dlnz
s =
- ~
~.
(3.25)
[ ~ H(z)dlnz]
Highly entangled systems, especially those of narrow molecular weight

distribution, are characterized by a set of relaxations at long times (terminal
relaxations) which are more or less isolated from the more rapid processes. The
modulus associated with the terminal processes is called the plateau modulus
G. Because qo and j o depend on weighted averages over H(z), their values are
controlled almost completely by the terminal processes. These experimental
25
parameters can therefore be combined to define average relaxation times in the

terminal region:
S zH(z)dlnz
-
z,, -
oo
- rlo/G,
(3.26)
Zw~ - ~
= r/oJ ,
~ "cH(z)dlnz
(3.27)
~ n ( z ) d In z
S z2H(z)dlnz
--O7)
and
l
27w
= j~ G}.
(3.28)
"gn
The mean times z, and Z~vwill be called the number-average and weight-average
relaxation times of the terminal region, and zw/z, can be regarded as a measure
of the breadth of the terminal relaxation time distribution, It should be
emphasized that these relationships are merely consequences of linear viscoelastic behavior and depend in no way on assumptions about molecular behavior.
The observed relationships between properties such as r/o, jo, and G and
molecular parameters provides the primary evidence for judging molecular
theories of the long relaxation times in concentrated systems.
Other types of linear viscoelastic experiments may be used. Dynamic shear
compliance measurements provide the storage and loss compliances J'(og) and
J"(o). An equation analogous to Eq.(3.12) is available for determining the initial
modulus from J"(og):
G O=
J"(og)d lnw
(3.29)
Stress build-up at the beginning of a steady shearing flow and stress decay at its
termination can also be employed. The analysis of results is somewhat different
than in the more usual experiments, but again, as long as the response obeys the
rules of linear viscoelasticity, the same fundamental information is obtained.
Flow birefringence measurements can also be used, subject to the additional
assumption that the stress and optical anisotropy tensors are linearly related.
A large body of experimental evidence supports this assumption (18). For
example, the recoverable compliance can be obtained from the extinction
angle X in steady shearing flow:
o_
l lim[~
d~- ~0~L
(3.30)
~'
]"
26
w.W. Graessley:
Finally, tensile deformations provide the same information as shear deformation

as long as the incompressibility assumption is not violated. In this case, the
tensile stress relaxation modulus E(t) is directly related to the shear modulus:
E(t) = 3 G(t), and all other relationships follow accordingly.
The experimental methods for measuring linear-viscoelastic properties are
reviewed elsewhere (15). Whether the behavior being measured does, in fact, lie
in the linear viscoelastic region must be judged separately in each system. Generally, tests are made at different levels of imposed stress, strain, or strain rate to
establish that the same material function is obtained. In polymer systems the
response is usually spread over many more decades of time or frequency than
can be covered in a single isothermal experiment. Typically, response is measured
over a convenient range at several temperatures and the results combined to form
a master curve according to the principle of time-temperature superposition. The
methods for forming a master response curve are described elsewhere (15).
4. Molecular Models in Polymer Rheology

4.1. Introduction
A variety of molecular models have been employed to represent polymer
chains and their interactions with surroundings. Equilibrium properties and
certain very simple transport properties depend mainly on the radial distribution
of segments about the center of gravity of the molecule. In one group of models
the segments are distributed in space according to their equilibrium positions,
fluctuations in segment density and the correlations in position between successive
segments are ignored. The particle cloud model (27) has been used to calculate
thermodynamic properties of random coils in dilute solution, such as the virial
coefficients and the expansion of the coil due to excluded volume. A similar porous
sphere model has been applied to viscosity and diffusion calculations (85). These
models are not well adapted for dealing with dynamic behavior, since mechanical
connection along the chain and the consequent coupling in the motions of
neighboring segments are crucial in such properties.
4.2. Necklace Models

Necklace models represent the chain as a connected sequence of segments,
preserving in some sense the correlation between the spatial relationships among
segments and their positions along the chain contour. Simplified versions laid the
basis for the kinetic theory of rubber elasticity and were used to evaluate configurational entropy in concentrated polymer solutions. A refined version, the
rotational isomeric model, is used to calculate the equilibrium configurational
27
properties of chains from local bond angle restrictions and the differences in
energy among the rotational isomeric states (22).
The pearl necklace model also provides the conceptual basis for theories of
chain dynamics and viscoelastic properties. As formulated by Kramers (86) and,
somewhat differently, by Kirkwood and co-workers (87), the configuration is
specified by the positions of n main chain atoms. The positions of each atom
relative to its nearest neighbors along the chain are subject only to bond length
and valence bond angle restrictions (excluded volume effects are neglected). For
simplicity, all allowed configurations are considered to have the same energy.
The configuration changes when chain elements move to new positions by bond
rotation. These changes are opposed by the viscous resistance of the medium.
As the change takes place, the effects are felt by other segments in the vicinity
(but perhaps remote along the chain contour) through the motions induced in the
fluid. Random collisions between chain elements and their surroundings provide
the impetus for configurational change in accordance with the theory of
Brownian motion (88). The large-scale motions of the chain, those of interest in
flow properties, arise out of the myriad of such primitive local bond rotations.
Necklace models can in principle deal with chain dynamics for all characteristic
frequencies which lie well below the collision rate in the liquid ( ~ 1012 sec-1),
and with response to external fluid motions of arbitrary magnitude. In practice,
they are seldom used because of the following problem. Given the local configurational restrictions, the subsequent movement of adjacent chain elements
must somehow be coordinated in order to satisfy the restrictions (preservation
of valence bond angles and lengths throughout the motion) and yet at the same
time not to require large portions of the chain to move each time a local
rotation occurs. It is not known how to determine the types of locally coordinated
motions and their relative frequencies from the chain structure. The relative
importance of the various possible motions probably are influenced by both the
viscosity of the medium and the internal activation barriers.
Attempts have been made to identify primitive motions from measurements
of mechanical and dielectric relaxatioja (89) and to model the short time end of
the relaxation spectrum (90). Methods have been developed recently for calculating the complete dynamical behavior of chains with idealized local structure
(91,92). An apparent internal chain viscosity has been observed at high frequencies
in dilute polymer solutions which is proportional to solvent viscosity (93) and
which presumably appears when the external driving frequency is comparable
to the frequency of the primitive rotations (94, 95). The beginnings of an analysis
of dynamics in the rotational isomeric model have been made (96). However,
no general solution applicable for all frequency ranges has been found for chains
with realistic local structure.
4.3. Spring-Bead Models
The bead-spring models are devices to circumvent the complications of the
local motion problem and still obtain information on the large-scale configurational relaxations which control viscoelastic behavior. Their utility lies in the
28
W.W. Graessley:
intuitive notion, borne out by calculations with specific models for the primitive
motions (91, 92), that the large-scale chain motions are largely independent of the
details of the primitive motions. Stated another way, the thermal motions of
long flexible molecules can be characterized by a set of relaxation times. The
nature of the short time end of the spectrum is very sensitive to the details of
local structure through the latter's influence on the primitive motions. However,
the relative spacings of the longest relaxation times is independent of the primitive
motions. Their net effect appears only as a featureless scaling factor, representable
in terms of a segmental mobility or its reciprocal, a segmental frictional
coefficient ~.
The Rouse-Bueche model (97, 98) replaces the real molecule of n main chain
atoms by a mechanical chain of N + 1 beads joined in sequence by N linear
springs. The frictional interactions with the medium, which are distributed
uniformly along the length of the real molecule to give a molecular frictional
coefficient n~o, are concentrated at regular intervals in the beads. The frictional
n
coefficient for each bead is taken to be ~
~0, so the model has the same total
frictional coefficient as the real molecule. The restraining effects of the n/N main
chain bonds which connect the sites of the beads are provided by linear springs
of spring constant 3NkT/(r2); (r 2) is the equilibrium mean-square end-to-end
distance of the entire molecule and kT is the Boltzmann factor. This expression
for the force is appropriate for Gaussian subchains undergoing random thermal
motion (98, 99), and it provides the model with the same equilibrium distribution
of large scale configurations as the real molecule. The actual value of N is
immaterial, provided that n/N is large enough for the Gaussian approximation
to be valid for the sub-chains, and that N itself is large enough for the frictional
sites to be distributed in a fairly uniform manner over the chain contour.
Calculated properties which turn out to be independent of N for N >> 1 should
be independent of local details of structure. Those which depend on N are, by
inference, sensitive to local structure and cannot be expected to be well represented
by spring-bead models.
The configuration of any chain is specified by the 3(N + I) coordinates of its
s dr,, is the probability
beads. In an ensemble of v chains per unit volume, ku//m=O
of finding a chain with a specified set of bead coordinates r,,(rn = 0, 1..... N),
where rr, is the position vector of bead m. Conservation of beads in the system
requires the probability distribution function 7j to satisfy the equation of
continuity:
0-~- +
I~m- ( i'~ ~u) = 0
(4.1)
ra=O
in which k,, is the velocity of bead m.

Inertial forces are neglected, and the chains are assumed to interact with the
medium independently. The flow is assumed to be locally homogeneous, i.e., the
velocity of the medium at each bead position is taken to be a linear function
The Entanglement Concept in Polymer Rheotogy
29
of the bead coordinates. A balance of the mean forces on each b e a d - - t h e spring

force, the frictional force due the systematic motion of the medium relative to the
bead, and the osmotic force or driving force for diffusion which arises from the
random thermal agitation of the s y s t e m - provides an expression for k in Eq. (4. t).
The result is a 3(N + 1) dimensional diffusion equation for-7~ (I00).,
The stress contributed by the chains is given bya:
Pjk=V ~ (F~Xk),.
(4.2)
m=O
in which Fj is the j component of frictional force exerted on the medium by the

bead, Xk is the k component of the position vector of the bead, and ( ),I means
the average for bead m over the distribution function ~. Calculation of these
averages requires a transformation to normal coordinates to separate the diffusion
equation into a sum of terms, each depending on only one of the coordinates.
Each term in the new coordinate system governs the response of one of 3(N + 1)
normal modes of the chain. Three modes are external, describing the diffusive
and convective displacements of the entire chain; the remaining 3 N (N for each
coordinate direction) are internal modes, describing the relative motions of the
beads.
The required transformation is that which diagonalizes the N + 1 x N + 1
matrix:
A=
1 -1
0
0
-1
2 -1
0
0 -1
2 -1
..,
...............
...........................
0 -1
2 -1
0
0 -1
2
0
0
0 -1
0
-1
1
(4.3)
For N large and i, N the eigenvalues of A are
2i =
i = 0, 1, 2 . . . . .
(4.4)
1 This equation has recently been derived in rather complete fashion for bead-spring
models by Lodge and Wu (101). Derivation and application to models other than springbead systems is given by Bird and co-workers (t02).
30
W.W. Graessley:
The associated relaxation times of the internal modes are
(r2)n(o 1
zi-- 6~z2kT ~-
i = 1, 2. . . . .
(4.5)
The internal modes are periodic functions of bead position along the chain; the
characteristic wavelength of internal mode i is N/i. Thus the lower modes govern
the large scale motions and have the longest relaxation times.
4.3.1. Configurational Relaxation

Configurational relaxation in the absence of flow is governed by the normal
modes and their corresponding relaxation times. The autocorrelation function for
the end-to-end vector is a measure of the configurational memory:
C(t)= (r(O). r(t))
(4.6)
in which r (0) = r. - ro is the end-to-end vector at some initial time, and r (t) is
the vector for the same chain at a later time t. For the spring-bead model one
can easily show"
8
C(t)= ~~z-
i~-e
- t[2 tl
(4.7)
i= 1,3,5,...
The correlation soon becomes dominated by the relaxation of the lowest mode:
C(t)= --~8 e-'/z~l.
(4.8)
The relaxation of (r 2) towards its equilibrium value (r2)e for chains with an
initial value (r2)o is given by an equation similar to Eq.(4.7), and is similarly
dominated by z 1 (103). Furthermore, for chains which are all stretched initially
to the same end separation ro and released, the relaxation of (r 2) to the
equilibrium value (r2)e is given by (103):
(r z) = r~ C2(t) + (r2)e [1 - C(2t)]
(4.9)
31
where C(t) is given by Eq.(4.7). Again, like the autocorrelation function, the
recovery to (r2)e is soon dominated by zl. Ingeneral, one might expect any
measure of the configurational memory time of the chains to be of the order
of the relaxation time of the lowest normal mode.
4.3.2. Response to Flow

The configurational response to flow depends upon which of the normal
modes interact frictionally with the flow field, tn simple shear the distribution
envelope in the flow direction alone is altered, and only the N normal modes
associated with the flow direction are active. The polymer contribution to the
shear relaxation modulus for a system with v chains per unit volume is:
N
G(t)=vkT ~ e-'m
(4.10)
i=1
in which z~ is the relaxation time associated with the ith internal mode. The
steady shear viscosity and recoverable shear compliance in dilute solutions
are therefore:
N
q-rh=vkT ~, zi,
(4.11)
i=1
N
1 (r/- q~)2
jo _
vkTtlZ()
~ zi
i=1
2z,t
(4.12)
where qs is the solvent viscosity. These expressions can be converted to macroscopic variables by replacing vkT by cRT/M.
The validity of Eqs. (4.10)-(4.12) probably extends well beyond the Rouse
model itself [characterized by the specific set of ~ values in Eq.(4.5)], and it seems
likely that they will apply, at least for small disturbances, whenever the elements
supporting the stress are joined by sufficiently flexible connectors and configurational relaxation is driven by simple Brownian diffusion. One might speculate
further that these same forms would apply even in concentrated systems, with
Eq.(4.10) expressed in a somewhat more general form because of intermolecular
interactions:
1
G(t)=
Nr
~-kT ~ e-'m.
i=1
(4.13)
32
W.W. Graessley:
In Eq. (4.13) Nv is the total number of internal degrees of freedom per unit volume
which relax by simple diffusion (ArT= 3 v N for dilute solutions), and zi is the
relaxation time of the ith normal mode (i = 1, 2, ..., 3 NT) for small disturbances.
Equation (4.13), together with a stipulation that all relaxation times have the
same temperature coefficient, provides, in fact, the molecular basis of timetemperature superposition in linear viscoelasticity. It also reduces to the expression
for the equilibrium shear modulus in the kinetic theory of rubber elasticity
when zi = oo for some of the modes.
Zimm (i00) has extended the Rouse model to allow for intramolecular hydrodynamic interaction, i.e., changes in medium velocity near each bead caused by
the flow disturbance from other beads on the same chain. The Oseen approximation, evaluated with the beads located at their mean equilibrium positions, was
used to estimate the velocity disturbances. The intensity of the disturbance
depends on the parameter h'
h=
n~0
(12n3)1/2 r/s(r 2) 1/2 "
(4.14)
Note that h is proportional to n ~/2 in 0-solvents, and thus to N 1/z. For h = 0 the
flow disturbance is zero, the chain is said to be free draining, and the original
Rouse model is recovered. For h >>1, flow in the coil interior is presumed to be
substantially reduced, the chain is frequently said to behave as an impenetrable
coil, and the Zimm model is obtained. Equations (4.10-4.12) continue to apply
for all values of h, although the distribution of relaxation times depends on h.
Some results for the two limiting cases and large N are:
h = 0 (Rouse model)
zi=
6
n2
tlo - qs
vkT
2
5
(4.t5)
~o(r 2) vn
,
36
q0 - t/s -
jo
1
i2 '
(4.16)
t (qo-qs )2
vkT~-
-qo
(4.17)
"
h>>l ( Z i m m m o d e l )
6 qo - rls 1
zl = ~-~--z- v k T
bi'
r/-r/s=
bi = 1.44,4.55, 8.60, 13.5.....

= 2.84 x 1023- (r2)3/2
t/s c,
M
)(36
jo=0.206
(q0-qs)
(4.18)
(4.19)
(4.20)
33
The Zimm model is most appropriate for long flexible chains in dilute
solution. Tschoegl (104) has calculated eigenvalues for intermediate values of h
and non-theta solvents. Fixman and Pyun (105) used a perturbation scheme to
allow the elements in the Oseen matrix to change as the coil deforms, obtaining
thereby slightly different eigenvalues for h >>1. Thurston and Morrison (106) have
evaluated the pre-averaged eigenvalues for unperturbed dimensions more precisely
and have examined the effect of the number of subchains on the results. Lodge
and Wu (107) recently extended the calculations to N = 300 for a range of
values of h* = h/N 1/2.
Constitutive equations for the Rouse and Zimm models have been derived,
and are found to be expressible in the form of Lodge's elastic liquid equation
[Eq.(6.15)], with memory function given by (101):
N e-(t- t')m
M ( t - t') = vkT ~
i= 1
(4.21)
Ti
In steady shear flow, the viscosity is independent of shear rate: r/= qo for all ~.
This property alone represents a serious qualitative failure of the conventional
bead-spring models. The normal stress functions are (108):
p l l _ P 2 2 = N 1 = 45 c RMT (t/o-t/~) 2 1;2 (h=0),
(4.22)
Pll --P22 =N1 =0.410 cR-~T-(r/o--?]s) 2 ]~2 (h = oo),
(4.23)
P22-P33=N2=O
(all values ofh).
(4.24)
The Rouse form (h = 0) is frequently applied to situations in which the solvent

contribution to the viscosity is negligible. For this case,
rio=
~o c((r2)/M) 1 M
36Na mo ] "
2 M
jo = 5 - cR~-T
(4.25)
(4.26)
in which mo is the molecular weight per main chain atom and N, is Avogadro's
number. In polydisperse systems (109):
qo =
[~o c((r2)/M) ] )14w,

36N~mo
jo = 2 M,~ M~ Mz+l
5 cRT
Mw-2
(4.27)
(4.28)
34
W.W. Graessley:
The various average molecular weights are defined as ratios of successive moments
ofthe weight distribution function W(M): Qk = ~ Mk W(M) dM and Mw = Q1/Qo,
0
~lz= Q2/Qx, and J~/z+ 1 = Q3/Qz. The steady state values of r/o and jo are independent of the number of elements in the bead-spring model as long as N is
sufficiently large. However, the initial modulus depends explicitly on N:
NcRT
Go = v N k T - ~
(4.29)
confirming the expectation that short time response is sensitive to structural

details of the model, and thus making spring-bead models inappropriate for
describing the very rapid relaxation processes.
The relaxation time distribution of the bead-spring models is discrete. The
spectrum is
N
H ( J = vk T ~ c5(~- zi)
(4.30)
i=1
in which 6( ) is the Dirac delta function. For the Rouse model with negligible
solvent contribution and N large:
i = h~ i2 ,
T1--
(4.31)
6 r/oM
cRT
(4.32)
7~2
The form of the spectrum is the same for Rouse chains fixed at one or both
ends, although h in these cases is no longer given by Eq. (4.32). This is also the
spectrum at long times for various models with explicit forms for the local motions
(91, 92). The Rouse spectrum can also be represented by a continuous function for
times which are small compared to h - For N ~ oe :
H(J:
v k T { z , t~n
--~\~-]
z<r~/a 2
H(z) = 0
~ > "ct/a 2.
(4.33)
The cut off parameter a is of order unity. Its value is somewhat arbitrary,
reflecting the inability of this continuous spectrum to represent the long time
behavior of the Rouse model precisely. Thus, with Eq. (4.33) and Eqs. (3.24) and
(3.25):
r/o =
1 cRT
-~a,
a M
jo_
3a c R T '
(4.34)
(4.35)
35
and no value of a will give the correct numerical coefficients in both equations
(rt2/6 and 2/5, respectively).
4.4. Properties of Alternative Models

Other models have also been used to represent polymer chains. The elastic
dumbbell concentrates all frictional resistance in two beads which are joined by
a linear spring, and is merely the Rouse model with N = 1. The spring constant
K is chosen to provide the same equilibrium radius of gyration as a polymer
molecule. The elastic dumbbell has many of the properties of the Rouse model,
except that it provides only a single relaxation time. Computational simplicity
makes it convenient for rough estimations. The rigid dumbbell also has two
frictional sites, but they are separated by a connector of fixed length which may
by chosen to match the radius of gyration of the polymer molecule. Unlike the
linear spring models, the viscosity of the rigid dumbbell depends on shear rate.
Reference (102) has a rather complete discussion of such models. Another model
which one might consider is a rigid spherical cloud of frictional sites with the same
mean spatial distribution of resistances as the undisturbed polymer molecule.
Curtiss et al. (102a) have recently developed general formulas for ~/o-qs
and jo for free-draining bead-connector models with arbitrary numbers of beads,
connecting arrangements and force-distance laws for the connectors. The expressions depend on averages over the unperturbed distribution of configurations
for the model:
r/ - qs -
'
(4.36)
jo = Y3{ (~) + (4~) + ( 4 z2 ) - +( ~2

4 1(+4 2Ca)z
) - ( 441 3 ) - (4243)
1
] (q0-rh)
vkT ~--~" (4.37)
in which
Ck = 1 ~ (Xk)~ k = 1, 2, 3.
(4.38)
/q i = 1
The number of beads in the model macromolecule is n, and ~ is the Stokes'

law friction coefficient of each bead. The Ck are to be evaluated for each macromolecule in its own internal coordinate system, with origin at the molecular
center of gravity and axes (k = 1, 2, 3) lying along the principal axes of the
macromolecule. The coordinates of the ith bead in this frame of reference are
(X1)/, (X2)i, and (x3)~. The averaging indicated by ( ) is performed over all
macromolecules in the system. Thus, (41 + 42 + 43) is simply S 2 for the macromolecules. The viscosity is therefore identical, for all free-draining models with
the same molecular frictional coefficient n~ and the same radius of gyration, to
the expression from the Rouse theory:
(nvS 2
q0 - q~ =
(4.39)
36
W.W. Graessley:
In rigid models each macromolecule has the same set of ~k, so (~k)~---~k
and ( ~ k ) = ~i~k" For flexible models the averages are calculated from the
equilibrium distribution of internal configurations (each configuration weighted
according to its energy through the Boltzmann equation). Thus, one sees
immediately that jo = 0 for rigid models with sufficient symmetry to yield
~1 = ~2 =~3, and hence for the rigid-spherical cloud model. Likewise in any
model with only two beads, ~2 = Ca = 0, and
jo _
(~1z) 01o -
rh) 2
(4.40)
5 (~1) 2 vkT~l~
For the rigid dumbbell, therefore, jo =3(q _ rh)Z/vk T rl~ ' while for the elastic
(linear spring connector) dumbbell it is easy to show that ( ~ ) / ( ~ 1 ) 2 = 5/3,
so jo = ( q o - q~)2/vkT q~. It is clear that all rigid models with the same set of Ck
and the same value of ~n give the same values of r/o - t h and jo, and that it is
molecular asymmetry in rigid models (differences among ~1, ~2 and ~3) which
governs the value of jo. The numerical factor can never exceed 3/5 in rigid
models. Molecular flexibility contributes to jo also. Thus, from the nature of the
averages in Eq. (4.37), one would expect jo always to be greater for a flexible
model than for a rigid model whose ~k are taken equal to (~k) (k = 1, 2, 3) of
the flexible model.
These results make it clear that the forms of t/0 - q s and jo are completely
independent of model details. Only the numerical coefficient of jo contains information on the properties of the model, and even then the result depends on
both molecular asymmetry and flexibility. Furthermore, polydispersity effects are
the same in all such free-draining models. The forms from the Rouse theory carry
over directly, so that t/0 - rh, translated to macroscopic terms, is proportional to
h4w and jo is proportional to the factor M~Mz+ 1/Mw. Unfortunately, no such
general analysis has been made for models with intramolecular hydrodynamic
interaction, and of course these results apply in principle only to cases where
intermolecular interactions are negligible.
Completely rigid models appear to provide rather peculiar short time response
The stress relaxation modulus for rigid dumbbells is (I 02):
G(t) = vk TE~ ~(t) +~e -'/']
(4.41)
while that for elastic dumbbells is

G(t) = vk T e-'/~.
(4.42)
In each the relaxation time z is (t/o -th)/vk T, and each satisfies the requirement
from linear viscoelasticity that
G(t) dt =
0
r/o - r h .
(4.43)
37
However, in performing this integration on experimental data it is found that

behavior at short times tends to be relatively unimportant, so that
G ( t ) d t ~ rl o - rh
(4.44)
over some range of sufficiently small ~. Elastic dumbbells behave this way for
~ ~, since
GO
S G(t) dt = (r/o - ~/~)e -~t~
(4.45)
but rigid dumbbells give

O9
f G(t) dt = ~(r/o - r/s) e -"/'
(4.46)
and acquire the extra ](r/0- r/s) only when e becomes literally zero. Thus, the
experimental stress relaxation behavior of a solution of rigid dumbbells would
appear to show large deviations from the laws of linear viscoelasticity. Similar
anomalies would appear in the dynamic loss modulus.
The delta function anomaly is probably present in the stress relaxation
modulus of all rigid models. It arises because the frictional sites in rigid models
must move relative to the solvent in any small but arbitrarily rapid deformation,
thereby generating a frictional force proportional to the deformation rate. In
models with flexibility the frictional sites can move instantaneously with the fluid.
The resisting force generated by the macromolecule is then simply that associated
with connector stretches, the latter remaining bounded even for arbitrarily rapid
deformation rates as long as the total deformation remained small. It seems clear
therefore that models must contain some degree of flexibility to simulate the
observed behavior of polymer systems.
4.5. Discussion
All calculations of visoelastic properties described here apply in principle
only to dilute solutions, since no allowance for intermolecular interactions has
been made. Nevertheless, the Rouse model in particular has been widely applied
to concentrated systems. There is probably no fundamental justification for such
an application. One simply assumes that each chain responds independently to
the systematic motions of a medium which is composed of other chains and
solvent, and which is taken to be a homogeneous Newtonian liquid (t09). The
contribution of the chains to the stress are taken to be additive.
The viewpoint parallels that of many other theories of condensed state behavior. The van der Waals theory develops an equation of state for dense gases
from the assumption that each molecule moves in an average field provided by
its neighbors and that the molecules contribute additively to the pressure. The
Flory-Huggins thermodynamic theory of concentrated polymer solutions proceeds similarly. Chains select configurations on a lattice partially occupied by
38
W.W. Graessley:
other chains and contribute additively to the configurational entropy. The cell
theory for flow in concentrated suspensions is developed along a similar line.
Each such theory seeks to capture the essential features of concentrated systems,
although without necessarily expecting agreement in detail. The Rouse theory
is qualitatively correct in many respects in concentrated systems. It therefore
provides a convenient framework for the empirical organization of experimental
data. It is also an appropriate starting point for theories which attempt to explain
linear viscoelasticity on a more quantitative molecular basis and to remedy its
failure to predict such non-linear properties as the shear-rate dependence of
viscosity.
Although both the Rouse and Zimm models provide for chain connectivity in
the sense that motions of individual beads are felt by adjacent beads, these models
do not rule out motions in which one section of chain literally crosses through
another. The success of the Zimm model in dilute solutions implies that
restrictions on the motion due to connectivity, i.e., self entanglement, are relatively unimportant in dilute solutions. The reason for this may be that hydrodynamic interaction already tends to rule out relaxation paths through the coil
interior, and it is just these paths for which self-entanglement effects would be
most prevalent. In Monte Carlo studies of lattice chains, Verdier (110) noted a
considerable slowing in configurational relaxation when the chain was programmed for excluded volume (non-simultaneous occupancy of lattice sites). In
this case volume exclusion disallows chain crossing as well, and so perhaps allows
the display of an effect caused by self-entanglement, unobscured by hydrodynamic
interaction but unobservable in systems of real chains in dilute solution.
5. Experimental Results on Linear Viscoelastic Behavior

5.1. Introduction
The effects attributed to entangling interactions, e.9., the plateau region in stress
relaxation, appear most prominently at high concentrations and in melts. It is
important, however, to distinguish this interaction from other types which are
present at lower polymer concentrations. To make the separation properly, it is
necessary to examine viscoelastic behavior at all levels of concentration, beginning at infinite dilution.
5.2. Very Dilute Solutions

The importance of intramolecular hydrodynamic interaction at low polymer
concentrations is well established. The impenetrable sphere form (20, 27):
[17] = lim r/o - r/~ _ ~
c~o r/sC
( r32 _) (2
M
(5.1)
39
relates intrinsic viscosity, molecular dimensions and molecular weight at high

molecular weights ( M > 106) for a wide variety of polymer-solvent systems (19,
111). In theta solvents the experimental values of [r/] M/(r2) 3/2 remain very nearly constant even down to rather low molecular weights (M ~ 104), while, broadly
speaking, in good solvents they already begin to decrease as the molecular weight
falls below the range M ~ 105. The observed value of 45o~agrees well with theoretical limiting values (h-* oo) from the original Zimm theory, qi~o= 2.84 x 1023,
or the Pyun-Fixman modification, q~ = 2.68 x 1023. The decrease of with
decreasing molecular weight is interpreted as a trend toward free-draining
behavior, and is expected theoretically since h [Eq. (4.14)] itself decreases with
decreasing molecular weight, being proportional to M/(r 2) lj2.
Information on dynamic behavior at infinite dilution has greatly increased in
recent years through the development of instruments for precise determination of
G'(co) and G"(o)) in dilute solutions (112, 113). Ferry and coworkers have
published extensively on the intrinsic dynamic moduli,
G' (69)
[G']o, = lim - - ,
c-*O
(5.2)
[G"]o, = lim G" (co) - ~/sco

c~O
(5.3)
Data on several polymers in both good solvents and theta solvents are now
available (114-116). With h* as an adjustable parameter, reduced plots of
[G']~ M/R T and [G"] ~ M/R T vs [q] rha)/RT can be placed in essentially quantitative agreement with moduli calculated from the Zimm eigenvalues (107, 116a)
for large numbers of submolecules (Fig. 5.1). Highly accurate calculations have
now been carried out for up to N = 200 submolecules. The deduced values of
h*00 vary somewhat with solvent power, but results for different polymers,
molecular weights and solvent powers are brought together by the correlation
(li6):
h~0o a, = 0.21 _ 0.02
(5.4)
in which a, is the viscometric coil expansion factor [[q]/[/'~]0] 1/3. T h i s form is

suggested by the (r2) lj2 term in the denominator of Eq. (4.14) in combination
with Eq. (5.1). Thus the inference is drawn that coil expansion, brought about by
excluded volume in good solvents, shifts the dynamic properties toward free
draining behavior, and that the net result can be understood simply in terms of a
uniformly expanded gaussian chain and a proportionate reduction of h.
Systematic departures from the Zimm moduli are observed at high frequencies
(93, 117). These deviations appear to stem from the expected inadequacies of
spring-bead models when the driving frequency approaches the frequency of the
primitive backbone motions. The effects are attributed to a local resistance to the
articulations of the chain which are required to bring about configurational
40
W. W. Graesstey:
Polystyrene in 2 Q-Solvents, M=860,000
I
'
--j
,%_:4
Z
3 /
oo,o
T
22"C
16"C.
~o
I
10g Wl]s
Fig. 5.1. Intrinsic moduli for narrow distribution polystyrene (M = 860000) in two theta
solvents (114).This comparison with theory is equivalent to that of reduced moduli described
in the text. [Reproduced from Polymer J. 1, 747 (1970).]
change. They are well correlated by an internal viscosity or frictional coefficient

operating on the normal modes (118). Experimental values for the internal
viscosity coefficient are insensitive to molecular weight as expected, but appear to
vary in direct proportion to the solvent viscosity. A review of these results has
recently appeared in this journal (118a).
5.3. Concentration Effects and Intermoleeular Interaction

Fundamental theories of transport properties for systems of finite concentration are still rather tentative (24). The difficulties are accentuated by the still
uncertain effects of concentration on equilibrium properties such as coil
dimensions and the distribution of molecular centers. Such problems are by no
means limited to polymer solutions however. Even for the supposedly simpler
case of hard sphere suspensions the theories of concentration dependence for the
viscosity are far from settled (I19, 120).
The Huggins constant k' characterizes the first effects of interaction on the zeroshear viscosity:
% = ~/~(1 + [~/] c + k'[r/] 2 c 2 + ...).
(5.5)
Experimentally k' is essentially independent of molecular weight for long chains,

with values of roughly 0.30-0.40 in good solvents and 0.50-0.80 in theta solvents.

101 - -
J '
i0
41
O
O
10-3
O
10-4
10-5
iff 2
10 "~
~~o
wM
PRT
I0 o
i0 ~
i0z
i O3
Fig. 5.2. Reduced dynamic moduli for undiluted narrow distribution polystyrene of low
molecular weight. Data for a sample of ~-1w=28900 were reduced to 160C, for which
e= 1.0 gm/ml and ~/0= 54500 poise (124). The solid lines were calculated from the Rouse
theory
Theories of k' for random coil molecules are very difficult and still somewhat
lacking in experimental confirmation (24, t21).
Fundamental theories of the relaxation spectrum at finite concentration have
yet to be developed. Experimentally, intermolecular contributions appear quickly
as the concentration increases. In the dynamic moduli they appear first as an increase in the longest relaxation time, the more rapid relaxations remaining relatively unaffected (114). Apparently the largest scale molecular motions are. slowed
as peripheral segments begin to impinge upon those of neighboring chains. At
somewhat higher concentrations, as the coils begin to overlap appreciably, the
faster relaxations also shift successively to longer times. The relative spacings at
long times increasingly resemble the Rouse spectrum (I 5,122,123). This behavior is
believed to reflect an increasing free-draining character in the local flow patterns,
caused by a gradual cancellation of intramolecular hydrodynamic effects as the
solution becomes uniformly filled with polymer segments.
The culmination of this trend is illustrated in Fig. 5.2 by dynamic data on
undiluted polystyrene of low molecular weight (124). Agreement with the Rouse
model here is by no means as good as that seen in Fig. 5.1 with the Zimm
model for a high molecular weight polystyrene at infinite dilution. Indeed, the
value of jo deduced from G'(o)) for the sample in Fig. 5.2 exceeds the value from
42
w.w. Graessley:
the Rouse theory by approximately a factor of two. Nevertheless, the agreement

is remarkable, considering that no arbitrary parameters are involved (aside from
the fact that the reduction scheme compels agreement with the observed value
of I/0) and that intermolecular interactions have not been considered at all.
Moreover, even the magnitude of q0 in this case is probably near that given by
the unmodified Rouse theory (see Part 5.4.1). Thus, although intermolecular interaction can be important even in dilute solutions, this importance does not
necessarily carry over into concentrated solutions and the melt state, at least for
short chains.
5.3.1. Parameters Measuring Intermolecular Interactions
The literature on viscosity at low and high concentrations of polymer
suggests very clearly that two fundamentally different types of intermolecular
interaction need to be considered. The viscosity behavior at low concentrations
resembles in many ways that of a suspension of discrete rigid particles (119,
I20). The molecules probably retain some vestige of non-draining character in
this regime, and intermolecular effects may be caused by the interaction between
the flow fields of the molecules, as in the case of rigid particles. The interaction
seems to depend upon the volume occupied by the molecules or, alternatively,
on the degree of overlap of the individual molecular domains. Frisch and Simha
(125) have pointed out that the parameter c It/] measures the degree of coil overlap
in a solution, if one neglects changes in coil dimensions with concentration. The
pervaded volume V of a molecule with radius of gyration S may be estimated
roughly as 4~ $3/3. In combination with Eq. (5.1) and with $2= (r2)/6 for
random coils,
4~ [q] M
3 q~o 63/2
(5.6)
The number of molecules per unit volume v is 6.02 x 1023 c/M. If the molecular
centers are distributed randomly, the product v V is the average number of
other molecules with centers lying within the pervaded volume of any one
molecule. Accordingly, v V is a measure of the potential degree of coil overlap, and
with q~oo= 2.68 x 1023 :
v V = 0.64 c[q].
(5.7)
Simha and Zakin (126), Onogi et al. (127), and Cornet (128) develop overlap
criteria of the same form but with different numerical coefficients. Accordingly,
flow properties which depend on concentration and molecular weight principally
through their effects on coil overlap should correlate through the Simha
parameter c[t/], or cM a, in which a is the Mark-Houwink viscosity exponent
(0.5 < a < 0.8). If coil shrinkage, caused by the loss of excluded volume in good
43
solvents, is itself a function of coil overlap, then c [r/] would remain an acceptable
correlating parameter even in good solvents.
A second source of intermolecular interactions , segment-segment contacts
between molecules, has been considered in concentrated solutions and melts where
free draining flow is assumed to dominate. This appears to be the mode of
interaction envisioned by Bueche in his discussion of chain entanglement (7). In
a system of uniform segment density the number of intermolecular contacts per
unit volume is proportional to c a. Since the number of polymer molecules per unit
volume is proportional to c/M, the number of intermolecular contacts per
molecule is proportional to cM. Thus, properties which depend on c and M
principally through their effects on the number of segment-segment contacts per
molecule should correlate in terms of the Bueche parameter cM. In free draining
systems the variation of (0 with polymer concentration must be accounted for
separately of course. Since tit/] and cM involve different combinations of c and
M, it should in principle be possible to distinguish between Simha and Bueche
interactions.
5.3.2. Viscosity Correlations
The effects of interaction on viscoelastic properties at low concentrations
depend on the Simha parameter. For example, Ferry has pointed out the
importance of c[q] for the transition from Zimm-like to Rouse-like behavior in
the dynamic properties and in the observed values of jo (15). The shear rate
dependence of viscosity undergoes a corresponding transition as a function of
e[t/] (see Part 8).
The usefulness of c[tl] as an approximate reducing parameter for the zero
shear viscosity at moderate concentrations is very well established. Master
equations of the form
t/0 = qs F(c[q])
(5.8)
are fairly successful in unifying data for different concentrations and molecular
weights in the same polymer-solvent system. One such expression is the Martin
equation (129):
~/o -
q~ = qs c [ q ]
e k' ct.~
(5.9)
Somewhat improved reductions can be obtained by using cM ~" as the independent variable, where a' is chosen empirically. In these cases a' is usually rather
close to the Mark-Houwink exponent for the system (129-133).
Cornet (128) points out that plots of logr/0 vs logc tend to be curved at low
concentrations but become straight lines at high concentrations. The concentration at the onset of straight line behavior, designated c*, was assumed by Cornet
to be the concentration at which a uniform segment density is attained in the sotu-
44
W.W. Graesstey:
tion. He calculates theta dimensions for several polymers from c*, finding reasonably good agreement with literature values. With Eq. (5.1), his relation between c*
and (rZ)o can be cast in the form:
c* = 6.14/[0]0
(5.10)
According to Cornet's criterion, the product c* M 1/2 should be a constant for a

given polymer-solvent system.
Onogi and coworkers (132, 133) have emphasized the importance of c[t/] (or
cM ~') in separating regions of weak and strong concentration dependence in t/o.
The transitions on plots of log qo vs log cM"' are gradual and the shapes of the
plots vary considerably from one polymer system to another. Figure 5.3 shows
the behavior of several polymer-solvent systems plotted as suggested by the Martin expression [Eq. (5.9)]. In this form of plot no obvious transition is observable.
It is therefore difficult to assign a critical value to (cM"'). Nevertheless, the
development of strong interactions, as judged by the viscosity, is clearly controlled by the overlap parameter c[q], or cM a' for a given polymer-solvent
system.
Williams and Gandhi (121, 134) have recently examined the effects of solvent
power and polymer polarity on viscosity. They find that No tends to increase
with concentration more rapidly in theta-solvents than in good solvents and
attribute this result to increased polymer-polymer association in thermodynamically poor solvents. Dreval and co-workers (135) and Ferry and co-workers
(I 36) had earlier noted similar differences between good and poor solvents and
drew similar inferences. On the other hand, Simha and Zakin (137) have pointed
out that the least part of the solvent power effect may be accounted for by the
expected decrease of coil dimensions in good solvents with increasing concentration. Thus in good solvents the coils are larger at low concentration due to
excluded volume, and the relative viscosity of their solutions tto/t/$ will be larger.
At higher concentrations the coil dimensions in good solvents shrink toward
their theta-solvent values. The relative viscosity might therefore be expected to
become the same at the same polymer concentration in different solvents.
Quadrat and Podnecka (I 38) have shown in a recent examination of literature
data that such ideas account very well for the behavior of polyisobutylene in a
variety of solvents. The results of Ferry and co-workers on polyisobutylene also
fall in line when relative viscosities are compared (139). However, in polyvinyl
acetate (136) the relative viscosities for a theta solvent cross over and became
somewhat higher than good solvent values at higher concentrations (> 35%
polymer). For polystyrene (121,137, 138, 140) theta solvents also give higher
relative viscosities at high concentration, with crossover concentrations in the
range 15-30% polymer. For polymethyl methacrylate (121,140a) the solvent
effect is dramatic, with a crossover concentration in the range of 5-7% polymer.
Some relationship between viscosity crossover in theta solvents and polymer
polarity is suggested by the results, supporting the idea of enhanced intermolecular
association in poor solvents. However, from the data on hand, one could also
infer a correlation with the glass transition temperature Tg of undiluted polymer,
PMMA-XYLENE
PS- DECALIN
PMMA- CHLOROBENZENE
45
6
c
TOLUENE
I0
[~]
Fig. 5.3. Viscosity at various concentrations and molecular weights in the low to moderate
concentration range. Polystyrene-decalin and polymethyl methacrylate-xylene are theta or
near-theta systems; the remainder are good solvent systems (121, 177). Note that the c[~]
reduction is somewhat better in theta solvents, and that the Martin equation [Eq. (5.9)],
which would give a straight line in the figure, is a somewhat better representation for
theta solvents
and hence with T8 for the polymer-solvent mixture. Vinogradov and co-workers
(141) have recently emphasized that such a dependence might develop as free
draining behavior is approached at the higher concentrations, since the viscosity
must lose its direct proportionality to solvent viscosity and become proportional
instead to a local frictional coefficient (o. The value of ~0 depends on the nature
of both solvent and polymer since both influence the free volume and glass
temperature of the mixture. The glass temperature of most solvents is well below
room temperature. In polyisobutylene solutions Tg probably remains relatively
constant and always well below room temperature because T~ for pure polyisobutylene is so low (Tg = - 70 C). For pure polyvinyl acetate Tg is 30 C, so Tg for
its solutions must ,rise somewhat with increasing concentrations. The effect for
polystyrene (Tg = 95 C) should be larger, while that for polymethyl methacrylate
46
W.W. Graessley:
(Tg= 100 C) should not only be large but might be unusual if the unusual
dependence of Tg with concentration in diethyl phthalate solutions (16, 142)
holds true in other solvents as well.
Variations in the temperature coefficient of viscosity with solvent, which have
also been presented as evidence of association in concentrated solutions (135,143),
could be similarly related to differences in Tg among the solutions. When free
draining behavior is a possibility, the relative viscosities in different solvents
should be compared at the same value of (o for the mixtures (that is, at constant
free volume rather than at constant temperature). In any case, it is clear that a
very well planned series of experiments is necessary in order to test for the
existence of additional specific effects such as association. These comments are
not meant to suggest that association can not occur at moderate concentrations.
Indeed, the existence of association in various forms of polymethyl methacrylate
seems welt established (144). The purpose is rather to advocate that less specific
causes be eliminated before association is inferred from viscosity measurements
alone.
5.3.3. Relationships between Low and High Concentration Regimes
An important and unanswered question is just how the low and high concentration regimes fit together (130). For long chains the viscosity changes from
dependence on c[q] at moderate concentrations to dependence on the product
of a local friction factor and a large-scale structure factor at high concentrations
(16). tn many of the studies already cited the reducing property of the parameter
c[q] persists to high concentrations (> 30% polymer), but this may be fortuitous.
Free draining behavior combined with a mild concentration dependence in the
friction factor could produce a similar correlation form [see Ref. (16), p. 303].
Perhaps the two flow regimes can be distinguished at least conceptually in the
following way. When steady motions are imposed on the boundaries of a fluid,
the flow pattern established in the interior is that which minimizes the rate of
energy dissipation per unit volume /~v (145). In steady shearing flow /~v= t/~2,
so the local flow pattern in a polymer solution will be that which yields a minimum
value for the viscosity. This principle then dictates the extent to which local lines
of flow are diverted around the individual pervaded volumes. Non-draining or
partial draining behavior is observed at infinite dilution because it results in a
lower dissipation rate (lower viscosity) than free draining behavior.
Let us suppose that the experimental contribution of polymer to solution
viscosity at low concentration, qo-q~, can be represented to a sufficient
approximation by the Martin equation [Eq.(5.9)]. An alternative which is
presumably always available to the solution is free draining behavior [Eq. (4.16)].
As concentration is increased, then, according to the minimum dissipation
principle, the behavior would be expected to change from the Martin form to the
free draining form in the range where the two expressions for t / o - q~ become
equal:
(o(r2)vn
36
= qsc[q] e k'4"~
(5.11)
47
If [r/] in the front factor on the right side of Eq.(5.11) is replaced by its value
according to the Zimm theory [Eq.(5.1)], and, if (o is replaced by the Stokes
formula for spheres, ~0 = 6 rc~/sR0, in which R0 is the effective Stokes radius of a
main chain atom, Eq.(5.11) becomes:
nRo
= e k"ct'l~
1.i (rZ)X/~
(5.12)
Thus, if Ro is approximately equal to the length of a main chain bond, the left
side of Eq.(5.12) is essentially the ratio of the contour length of a random coil
to its mean end-to-end distance, a quantity which depends on solvent power and
chain length, and which typically ranges from perhaps 10 to 100. Free draining
then begins to dominate in a critical range of c [~] values, given by Eq. (5.12):
1 In r~xtende~~5--15.
(c[~])* ,~ k-7
ron~,od
(5.13)
These values provide the upper limit for a subrange of solution concentrations,
beginning with the onset of coil overlap near c[r/] ~ 1 and ending with essentially
complete free draining character near c[q] ~ 10. This estimate of the upper limit
is very approximate of course, since the behavior must also depend somewhat on
solvent power and chain length. It is also subject to errors from the assumption
that (o is essentially constant through the subrange, and that the free draining
expression can be used without regard for other types of intermolecular interaction, such as chain entanglement.
5.3.4. Anomalous Flow Properties at Intermediate Concentrations

Time-dependent flow anomalies also appear in this region of moderate concentrations. Peterlin, for example, has observed that the viscosity of dilute
solutions of various polymers may begin to increase in the course of flow through
long capillaries (149). He attributes the effect to a shear-induced impingement
and temporary aggregation of coils. Lodge (150) has observed macroscopic
inhomogeneities in 2-3% solutions of polymethyl methacrylate and polystyrene
after prolonged shearing in a plate and cone viscometer. Philippoff (151) has
found that the flow birefringence of dilute polystyrene solutions is reduced by
filtration, although the original values are restored after the solutions are allowed
to stand for a time. Aggregation in dilute solutions which can be removed or
reduced by heating is well established in certain highly polar polymers such as
polyvinyl chloride. Not enough is known about such effects at the present time
to warrant any detailed consideration here.
48
W.W. Graessley:
5.3.5. Summary
The viscoelastic properties of polymers in solutions of moderate concentration,

roughly defined as the interval, 1 < clot] < 10, are not well understood in terms
of fundamental theory. The difficulties parallel in some ways those of unifying the
thermodynamic theories of very dilute and highly concentrated solutions. Broadly
speaking, the molecules appear to retain some of their dilute solution impenetrable
character. The choice of solvent influences viscosity both directly through its own
viscosity and indirectly through its thermodynamic interactions with the polymer.
Perhaps the most striking feature is the rather mild changes in relative spacings
of the relaxation time distribution (the shift from a Zimm-like to Rouse-like
spectrum) through the regime, even though the strength of the intermolecular
interactions, as judged by viscosity and mean relaxation time, increases
enormously.
It appears that one can develop a qualitative understanding of the simple
flow properties at moderate concentration without going beyond concepts which
are already inherent either in the dilute solution theory of polymers or in the
properties of particulate suspensions. The dependence of viscosity on c[t/] is
believed to reflect a particle-like or equivalent sphere (127) hydrodynamics in
solutions of low to moderate concentration. In particular, the experimental
facts do not force the consideration of effects which might arise from the
permanent connectedness of the polymer backbones. Situations conducive to the
entangling of molecules may be present, e.9., overlap of the coils, but either
entanglement contributions are small, or else they are overwhelmed by the
c[t/] interactions.
5.4. Concentrated Solutions and Melts
Portions of the literature on viscoelasticity in concentrated polymer systems

of narrow distribution have been reviewed recently (15, 16, 152, 153). The
following discussion concerns three principal characteristics, the viscositymolecular weight relation, the plateau modulus, and the steady-state compliance.
5.4.1. Zero-Shear Viscosity

A new set of flow characteristics gradually emerges as the concentration of
polymer becomes large. The solution viscosity loses its direct dependence on
solvent viscosity and comes to depend on the product of two parameters: a friction
factor ~ which is controlled solely by local features such as the free volume
(or alternatively the segmental jump frequency), and a structure factor F which
is controlled by the large scale structure and configuration of the chains (16):
r/o = ~F.
(5.14)
49
The Rouse expression for the viscosity [Eq.(4.25)] is of this form. The frictional
coefficient per main chain atom (0 is the friction factor in this case; the
remainder of the expression is the structure factor.
The friction factor depends upon the same features that govern the viscosity
of small-molecule liquids. At low temperatures (0 depends on T - Tg (Tg < T< Tg+
+ 100 C), and at higher temperatures it depends on an activation energy for
flow. The value of T~for a solution depends on the properties of both components
and their concentrations, but it is independent of the large scale structure of the
polymer as long as its molecular weight is large ( M , > 104 for most linear
polymers).
The structure factor, on the other hand, depends on the number of chains
per unit volume and on their molecular weight and dimensions. Its behavior is
largely the same from one polymer system to another, which is to say, independent
of the chemical structure of the repeating unit. The Rouse equation, in fact,
gives a good account of experimental results on linear chains if the molecular
weight is not too large. Thus, after correction to constant friction factor, the
viscosity for many undiluted polymers is nearly linear in molecular weight when
the chain length is less than 300--500 main chain atoms (t6). Moreover, the
numerical values of ~0 deduced from qo with the Rouse equation and known
values of (SZ/M)o are in quite reasonable accord with entirely independent
estimates based on the diffusion coefficients of small molecules in the same
system (t5). Thus, for polymers of low molecular weight the Rouse model, with
no adjustment for intermolecular interaction, gives not only the form but even
the correct order of magnitude of the viscosity.2
For longer chains the form of the structure factor changes, becoming a much
stronger function of molecular weight. At high molecular weights the well
known 3.4 power dependence results:
qo = K M
M < Mc
tto = K ' M 3"4
M > Me.
(5.16)
The change in M dependence occurs smoothly but over a relatively narrow

range in molecular weight. The characteristic value Mc is obtained from the
intersection of straight lines drawn through the two branches of the curve. For
many polymers Mc is independent of temperature and corresponds to a contour
length of about 300-700 main chain atoms (t6). In some systems the slope for
M > Mc appears to be somewhat higher than 3.4 (146). In others the slope begins
to increase beyond 3.4 at higher molecular weights (147, 148) although in the
case of polyvinyl acetate this may be caused by the presence of long chain branches.
On the whole, however, the 3.4 power is supported by data from a remarkably
wide range of linear polymers (I6).
2 It should be pointed out that this agreement cannot be used to support the Rouse
model per se, since any model with the same total molecular friction and value of S2 would
give the same expression for r/0 (Part 4).
50
W. W. Graessley:
109
r"
10 8
o,
106
>
iO1 /
A
cM w
Fig. 5.4. Viscosity vs the product c/~w for polystyrenes at concentrations between 25%
and 100%: O, undiluted at 217 C (154); d3, 0.55 gm/rnl in n-butyl benzene at 25 C (155);
O--, 0.415 gm/ml in di-octyl phthalate at 30 C (156); Q, 0.310 gm/ml in di-octyt phthatate
at 30 C (156); and -O, 0.255 gm/ml in n-butyl benzene at 25 C (155). Data at the various
concentrations have been shifted vertically to avoid overlap
Concentrated solutions (c > ca. 0.25 gm/ml) show similar behavior, although
with characteristic molecular weights that increase with concentration. In many
systems the transition remains fairly well defined and, within the accuracy of the
data, conforms to
(Me)so,,.= ( O_] M = Mc
\c/
(5.17)
where Me is the value for undiluted polymer, Q is its density, and tO is the volume
fraction of polymer in the solution. One can partially interpret this result by
noting that the transition takes place for different concentrations at the same
value of the segment contact parameter cM (see Part 5.3.1). Figure 5.4 shows
r/0 vs cM for polystyrene melts (154) and polystyrene solutions at several con-
51
Table5.1. Intersection points from Fig. 5.4 for ~/0 vs cM in polystyrenes of various
concentrations (visually estimated)
Concentration (gm/ml)
cM at intersection
1.00 (melt at 165 C)

0.55 (n-butyl benzene at 30 C)
0.415 (di-octyl phthalate at 25 C)
0.310 (di-octyl phthalate at 25 C)
0.255 (n-butyl benzene at 30 C)
32000
41000
26000
26000
38000
10 8
i
lo 5
io ~
I0 7
106
IO s
)r~
10 4
t.-
>
i0 3
iO 2
IOIto z
I0
i 3
I
lo 4
io 7
eM w
Fig. 5.5. Viscosity vs cM, for polystyrenes at concentrations between 25% and 100%.
The data in Fig. 5.5 have been shifted vertically to produce superposition at high
molecular weights
centrations (155, 156). The data have been shifted by arbitrary amounts vertically
(corresponding to arbitrary values of the friction factor ~o) to avoid overlap.
Table 5.1 shows visually estimated values of c M at the intersection for the different
concentrations. Figure 5.5 shows the same data, shifted vertically to achieve
52
W.W. Graessley:
oo
-id
9
-o
c~
0
6
oo-o
oO o~
o 6o
o-9
i0 ~
op
%
o-
io I
9
I0~10~ - - -
~2
I0
~
ILF
-- 0.68
cM w
~ -4- - _
I0
105
_~0.68 for polystyrenes at concentrations between 25% and 100%.

Fig. 5.6. Viscosity vs ~1~1.
The replotted data in Fig. 5.4 have been shifted vertically to produce superposition at
high molecular weights
superposition at high molecular weights. The asymptotic lines are drawn with
slopes 1.0 and 3.4.
Figure 5.6 shows the same data plotted as a function of cM 6s to test the
low concentration reduction scheme based on c[~/] with a typical value of the
Mark-Houwink exponent for good solvents. The data have been shifted vertically
to achieve superposition at high molecular weights. It is clear that the c M variable
produces a better superposition of data at all molecular weights and concentrations. The apparent variation in the values of c M at the intersections in
Fig. 5.4 (Table 5.1) is largely due to a lack of data to define the limiting behavior
at low molecular weights at some concentrations. The intersection on the superposed plot in the composite Fig. 5.5 is c M = 3 0 0 0 0 , giving M~=30600 for
undiluted polystyrene (0 = 0.98 at T = 217 C, in good agreement with the value
31200 reported by Berry and Fox (16).
53
i0? x
NN
>\
=~106
SLOPE
=~-I DN~ ~0 iD
,%
v
tOS
t0
Io-3
lo-'
lo-2
Jo
c (gm/ml)
Fig. 5.7. Characteristic molecular weight for the viscosity vs concentration for the polyisobutylene-toluene system. The open circles (O) were obtained from 7o vs c at constant
molecular weight (t59); the open squares (IS]) were obtained from ~/0 vs/~w at constant
concentration (159); the solid square (11) was obtained on undiluted polyisobutylene (9)
The thermodynamic properties of the diluent appear to be unimportant in

determining (M)so~,.. Polyvinyl acetate in diethyl phthalate and in cetyl alcohol
conform to Eq. (5.17), although the former is a good solvent and the latter a theta
solvent at the temperature of measurement (157).
Establishment of (Mc)so~,. becomes increasingly ambiguous at lower concentrations. The log-log slopes generally begin to differ from 1 and 3.4. In some
cases the transition becomes more diffuse and sensitive to the nature of the
solvent (t32). In others, generally in polymers of low Tg, it may remain relatively
sharp (158, 159) even at concentrations as low as 1-2% polymer. As noted earlier,
the product c[~/] controls viscosity at low to moderate concentrations, such that
one might expect behavior of the form (Mc)so~n.oc c- 1/a rather than (Me)sol,.oc c- 1.
The data of Schurz and Hochberger on polyisobutylene-toluene solutions (159)
are shown in Fig. 5.7. The values of (M~)sol,. at high concentrations obey Eq. (5.17),
while at low concentrations there does appear to be a trend in the direction
of c- 1/~,dependence.
This distinctive 70 vs M behavior in melts and highly concentrated solutions
is one of the features attributed to chain entanglement. The intensitivity of M~
to temperature, the simplicity and insensitivity to solvent power of the con-
54
W.W. Graessley:
centration dependence of (M)~o~n., and the fact that Me for different polymers
corresponds to roughly the same number of main chain bonds, all suggest the
onset of an intermolecular interaction based on geometrical rather than chemical
attributes. It must be noted that the transition itself is not intrinsically a sharp
one; indeed all sense of a transition is lost if the data are plotted in other ways,
such as logq0 vs M 1/2 (9). The curves in the vicinity of the transition appear to be
representable as the simple sum of high and low molecular weight forms (I 60):
{ (cM
%=KcM I+\oM~j
]"
(5.18)
The dashed line in Fig. 5.5 was calculated with this formula.
Rather than emphasizing the form of the transition region it is perhaps
preferable to recall that the Rouse free-draining theory, with (0 estimated
independently from the diffusion coefficient of small molecules in the system (15)
and without modification for interchain interactions, accounts almost quantitatively for the viscosity of concentrated systems at low molecular weights. For
molecular weights beyond the neighborhood of (Mc)so~,. the unmodified theory
is not adequate. An intermotecular interaction of some sort, differing fundamentally from the c[~/] interactions and dependent on the product cM, is required
to account for the behavior. The observed properties of Mc and (Mc)~o~n., cited
above, are consistent with the type of interaction which might arise from a
physical, rope-like coupling of chains, presuming that the number of couples
along a molecule is proportional to the number of its intermolecular contacts.
Porter and Johnson (14) surveyed the literature prior to 1965, assembling
Mc and (Mc)so~n.data for a large number of polymers. Berry and Fox (16) have
critically examined the values from this source and others, in many cases
reanalyzing the data upon which the reported values were based. Apart from
general questions about the accuracy of viscosity and molecular weight data, the
main problem in determining Mc is the need to correct isothermal viscosity
correlations to viscosities at constant segmental friction factor. Information is
required on the temperature dependence of viscosity as a function of molecular
weight, which is not always available. The corrections are largest for lower
molecular weights, and are especially important if the viscosities are measured at
temperatures within 50-100 C of Tr Without correction, the slope of the low
molecular weight branch of the curve is too large, and the deduced value of Mc
is too big. Values of ME for several well studied polymers are given in
Table 5.2 (15, 16).
5.4.2. Properties of the Plateau Modulus
Other viscoelastic properties also acquire new characteristics at high concentration. In undiluted systems the long time end of the relaxation time
distribution remains approximately Rouse-like for chains with molecular weight
below Me, as suggested by the agreement with reduced Rouse moduli in Fig. 5.2.
The Entanglement Concept in Polymer Rheolog3,
55
Table 5.2. Characteristic molecular weights for undiluted linear polymers
Polystyrene
Poly (or-methylstyrene)
1,4 Polybutadiene
Poly (vinyl acetate)
Poly (dimethyl siioxane)
Polyethylene
cis-Polyisoprene
Poly (methyl methacrylate)
Polyisobutylene
18100
13 500
1900 (2200)
12000
8 i00
-5 800 (4500)
5900 (10000)
8900
31200
28000
5900
24 500
24400
3 800
10000
27 500
15200
130000
104000
13800
86000
61000
(14400)
60000
> 150000
--
The values of M, [defined by Eq.(5.19)] are those reported by Ferry (15) and Berry
and Fox (16), except that for polystyrene (124) and the values in parentheses for poly-methyl
methacrylate (I46), polybutadiene (79, 197), and cis-polyisoprene (79, 197). The values of Mr
(defmed as described in 5.4.1) are from the same compilations, except that for poly (or-methyl
styrene) (/61). The values of M~were obtained as described in Section 5.4.4.
As molecular weight increases, the slow relaxations shift progressively to longer

times while the fast relaxations remain essentially unchanged.
This separation into two more or less distinct groups of relaxation processes
is evident in master curves of the stress relaxation modulus (162), the shear
compliance (15, 163), the dynamic moduli (15, 124), and the dynamic compliances
(15). Figure 5.8 and 5.9 illustrate the behavior with master curves of G'(~o) and
G"(~o) for undiluted polystyrene. At very high frequencies (not shown) G'(o~) is
approximately 101 dynes/cm 2 and comparatively constant (the glassy region).
As 09 decreases, G'(co) begins to decrease (the transition region) in a manner which
is independent of polymer molecular weight but which may differ considerably
from one substance to another (15). The moduli of polymers and monomeric
liquids (164) begin to differ qualitatively part way through the transition region
(G'(~o) ,-~ l0 s dynes/cm2). The former decay more slowly and in many polymers
appear to approach the form dictated by the Rouse spectrum, G'(og)occo1/2. The
high frequency slopes are approximately 1/2 in Figs. 5.8 and 5.9,
For long chains, the modulus at about 106-107 dynes/cm z passes into a
domain of much slower relaxation (the plateau region). The width of the plateau
depends strongly on molecular weight, while the characteristic modulus of the
plateau region G~ is independent of molecular weight. The presence of the
plateau confers rubber-like elastic properties over an intermediate range of
frequencies or times.
Although G'(og) appears to be relatively flat on log-log plots, especially for
very high molecular weights and narrow molecular weight distributions, the
plateau region is not entirely devoid of relaxation processes. This is most easily
seen in G"(og) (Fig. 5.10) which passes through a shallow minimum in the plateau
but clearly does not drop to zero. The plateau occupies the span of time between
the transition region and the terminal region, and those two groups of relaxations
always appear to overlap to some extent. In order to evaluate G~v,which may be
defined as the modulus associated with the terminal relaxation processes only,
56
W. W. Graessley:
9
PS
160"C
03
-
>,5
13
0
-
L1
"
2
LI2 )
1
-6
-5
-4
-3
-2
-1
Iogt~o T
, s e c -1
Fig, 5.8. Storage modulus vs frequency for narrow distribution polystyrene melts, reduced to
160C by temperature-frequency superposition. Molecular weight range from M'~,= 8900
(L9) to ~rw= 581000 (L 18) (124). [Reproduced from Macromolecules 3, i l l (1970).]
PS
e~
160C
"'
'
....
5-
Li'
--
L19
t27
y
2 --
LI4
"
L12
/
/
L9
I
16
-5
-4 '
-3
-2
-1
logt~a T
0
1
s e c -1
Fig. 5.9, Loss modulus vs frequency for narrow distribution polystyrene melts (124).
See Fig. 5.8. [Reproduced from Macromolecules 3, 112 (1970).]
57
i2,C
tO.C
A
~E
-2
8.(
4,0
/i
o
io -4
IC 3
1 .
\
io -z
IG ~
(SEC - i )
--.,
io
io'
Ice
Fig. 5.10. Loss modulus vs frequencyfor a narrow distribution polystyrenemelt reduced to

160C ()~w=215000) (124). The dashed line (---) is an approximate resolution of the
terminal relaxation peak. The construction line ( - - - ) is G" vs m calculated for a process
with a single relaxation time [G~)lG~ax = 269z/(1 + 092z:)]
one must separate the two contributions. Unfortunately, there is no unambiguous

way to make the separation. If a form is selected for the terminal relaxation
behavior, G} can be extracted directly (165). A common method is to choose the
value of the modulus at the inflection point of G(t). Creep compliance data can be
handled similarly (t66), the plateau modulus G} being then simply the reciprocal
of the plateau compliance J}. Other workers prefer to convert to relaxation
spectra, resolve the transition and terminal regions, and calculate J} or G} from
the area under the terminal peak (167).
Oser and Marvin (168) have developed a phenomenological theory involving
two groups of relaxations, a Rouse spectrum at short times and a second, shifted
Rouse spectrum at long times. They show that certain properties of G"(~0) and
J"(co), the frequencies of their minima in the plateau region, and the heights and
frequencies of their terminal loss peak maxima, are related in simple ways to
G} and ~/o. Ferry and co-workers determine G~vfrom the area under the terminal
peak in J"(m) through Eq. (3.29). The problem of resolving overlap in the terminal and transition peaks always exists, but for narrow distribution samples
of high molecular weight the separation can be handled in a reasonably
objective manner. Using J"(og), Ferry and co-workers have evaluated G} for a
large number of rubbery polymers (15) and have shown that the Marvin-Oser
equations, after slight adjustments in their numerical coefficients, are able to give
consistent estimates also. Onogi and co-workers have used the area under the
G"(og) loss peak to estimate G} in polystyrene (124) and polymethyl methacrylate
(t46).
58
W.W. Graessley:
The plateau region appears when the molecular weight exceeds Me [(M)soln.
for solutions], and is taken to be a direct indication of chain entanglement. Indeed
the presence of a plateau may be a more reliable criterion than t/0 vs M behavior,
especially in solutions of moderate concentration where viscosity may exhibit
quite complex concentration and molecular weight behavior. It is postulated
that when M greatly exceeds Me, a temporary network structure exists due to
rope-like interlooping of the chains. Rubber-like response to rapid deformations
is obtained because the strands between coupling points can adjust rapidly, while
considerably more time is required for entire molecules to slip around one
another's contours and allow flow or the completion of stress relaxation.
Plateau moduli are commonly expressed in terms of Me, the apparent average
molecular weight between coupling junctions. The modulus equation from the
(r2)
kinetic theory of rubber elasticity\ Eq. (7.2) with g (r-7~o = 1 and v = c/Me or
Q/Mefor undiluted polymers}/ is used, giving
(5.19)
M,- QRT
G~v
The number of junction points per molecule is therefore E = M / M e - 1. Values

of Me for a number of polymers have been tabulated recently by Ferry (15) and
Berry and Fox (141). Selected values are given in Table 5.2. The close parallel in
the values of M~ and Me is apparent.
In concentrated solutions, the value of (M&oin. is inversely proportional to
concentration, such that:
(G~)so,,. =
G~,
(5.20)
(Me)so,,. = (-0c-)M,.
(5.21)
or
This behavior, again paralleling that of (Mc)soln., is now clearly established for
a number of polymers (15, 142, 166, 167). Figure5.11 shows Eq.(5.21) to be
valid over approximately the same range of concentrations as the corresponding
equation for (Mo)sol,..
5.4,3. Properties of the Terminal Zone Spectrum
Relaxation time spacings in the terminal region are quite different than the
Rouse spacings: much more tightly grouped around the mean but with a tail
of faster relaxations which trails back into the plateau region. Tobotsky and
co-workers have found that the empirical form G(t) = A exp [ - (t/r) ~] fits experimental stress relaxation data in the terminal region (165). The parameter B correlates with molecular weight distribution with values of ca. 0.60 for typical narrow
distribution polystyrenes and an extrapolated value B =0,75 for hypothetical
to'
59
id
Z
SLOPE=2/
A
OZ
iO_ 2
iO_ ~
. . . . .
too
C(gm/ml)
Fig. 5.11. Plateau modulus vs. concentration for polymethyl methacrylate 0 {t42) and cispolyisoprene [] (167). The solid points are undiluted samples
monodisperse specimens. The dynamic moduli and the molecular weight

dependence of steady-state shear compliance (Section 5.4.4) demonstrate the
strikingly sharp character of the terminal spectrum (compared to the Rouse
spectrum) for narrow distribution polymers.
Terminal relaxations dominate the steady flow behavior. The zero-shear
viscosity has already been discussed. For long chains the steady-state compliance,
J~, and the characteristic relaxation time of the terminal region 17m (as determined
for example by the various procedures of Tobolsky and co-workers (165, 169) and
Ferry and co-workers (123) depend on molecular weight and concentration in a
manner quite different from that given by the corresponding Rouse forms (169,
173). Despite these differences, the observed orders of magnitude of Zm and J~
are surprisingly similar to values calculated from the Rouse theory (I I, 170).
For M >>(M)~oln.,the essential features are:
Observed form
Rouse form
rlo
"~mOC cbT
"t'm=
J~oc cb T
J~=
6 rloM
n2 cRT
2
cRT
60
W.W. Graessley:
The exponent b is a matter of some current dispute, being 2 in nearly all studies
but appearing to range as high as 3 in a few instances (Section 5.4.4). In any
case, the explicit molecular weight dependence is lost, and concomitantly the concentration dependence is increased (173). If all relaxation times had been scaled
up by the same factor (the relative spacings and intensities of the Rouse spectrum
being retained), the Rouse forms would have been preserved for M ~>(Mc)~oln.
5.4.4. The Steady State Compliance

Values of J2 are obtainable from several types of experiments (Table 5.3).
In principle, all the methods should agree, subject to very broad assumptions
(Section 3). However, each presents its own set of experimental problems and
sources of systematic error. Thus, values of J~ obtained from stress relaxation
are very sensitive to the shape of G(t) for long times, the region where the residual
stress itself is extremely small and difficult to measure accurately. Values obtained
from the intercept of the creep curve and from creep recovery are sensitive to
the attainment of steady state in the flow. Of the two, creep recovery is clearly
preferred because elastic recoil and hence J2 is measured directly. For all methods
based on linear viscoelasticity the stresses of course must be small enough to
place the response in the linear viscoelastic region. Values from the normal stress
function are usually determined from the total axial thrust developed in platecone viscometers. Deviations from a quadratic dependence on shear rate are
usually beginning to appear in the experimentally accessible region, and a somewhat uncertain extrap61ation to zero shear rate is required. Plots of N 1 vs crz are
usually much more linear, but still there remains the possibility of changes at
Table 5.3
Method for obtaining jo
Stress relaxation
Equation from phenomenological

viscoelasticity
lim j ~G(F,)d(
t ~
Recoverable shear after steady-state deformation

Intercept of steady-state creep curve
lim [7,/%]
lim {J(t) - t/rlo}
t~oo
Steady-state dynamic response

Normal stress measurements
Flow birefringence
- - tim
~o~ ~ o t - J - J
l olim~ N1}
1 tim[ cot2z ]
~/o ~t 3; J
61
lower shear stresses. Values from steady flow birefringence rely in addition on
the validity of the stress-optical law (18).
The data collected for review here were obtained on samples of narrow, but
still somewhat variable, molecular weight distribution. Steady-state compliance is
highly sensitive to polydispersity, so some disagreement among investigators is to
be expected, due simply to slight differences in the distributions of their samples.
The difficulties in all methods are probably somewhat reduced in systems of
low polydispersity because the distribution of relaxation times is narrow, and
relatively high stresses can be imposed before non-linearities appear. Nevertheless,
t h e presence of even small amounts of high molecular Weight species can
increase jo significantly (163). Non-linear response is difficult to avoid in such
circumstances and may cause apparently different values to result from the
various methods.
The steady-state compliance data have been analyzed in reduced form
~ -
t/o - t/$
"
(5.22)
This form for reduced compliance is suggested by the dilute solution molecular
theories, according to which J~R is governed by the dispersity of molecular
relaxation times [from Eq. (4.12)]:
Variations in JeR at finite concentration reflect the influence of intermolecular

interaction on the relative spacings of the long relaxation times. Through JaR
these effects may be examined separately from effects on the magnitude of the
viscosity (or the average magnitude of the long relaxation times). The ratio
E z2/(Z zi)2 is called $2/S 2 by Ferry and co-workers. The above expression for JeR
RT
jo
reduces to their JOR-- - - ~ T l i m

at zero concentration. It becomes
m[~/] c 0 c
jo cR T / M when r/o >>t/s, which is the definition of reduced compliance used by
Tschoegt (183) and Janeschitz-Kriegel (18). For the Rouse model JeR=0.4; the
Zimm model gives J~R= 0.206.
Figure 5.12 shows J~R obtained by different investigators on a single sample
of narrow distribution polystyrene [sample 6a, Mw = 860.000, Mw/Mn = 1.05 (174),
polymerized anionically by the Pressure Chemical Company, Pittsburgh, Pennsylvania]. A number of different solvents, temperatures, and methods are represented
(175-182). The values rise from below 0.4 at low concentrations, pass through
a broad maximum which somewhat exceeds 0.4, then decrease again at high
concentrations. Data from dynamic moduli, normal stress measurements, and
62
W. W. Graessley:
I.C
-O
f---~.,
o ~
oo
-o
-o ",,~
\~;o
O-
O-
o,
131
' o6,
c(gm/m~)
J.o
Fig. 5.12. Values of reduced compliance deR for solutions of a narrow distribution polystyrene (M,= 860000) according to several investigators and methods. Symbols: Q from
G'(o)) (175, 176), Ck from N1 (177), (3- from N 1 (178), 0 from flow birefringence (179),
- from flow birefringence (180), from G'(o)) and N1 (181), and -O from creep
recovery (182)
flow birefringence seem consistent with each other and generally follow the same
trend. Values from creep recovery differ (I82), exceeding those from other methods
by approximately a factor of two in the concentration range 0.15-0.25 grn/ml,
then decreasing rapidly towards them again at higher concentrations. Comparison
of literature data on two other Pressure Chemical samples [sample 3a,
1~,=411,000 in Fig. 5.13 (t77-179, 181, 182, 184, 185), and sample 14a,
1~,= 1,800,000] shows similar behavior. The reported creep recovery results
(182) exceed other values for jo at intermediate concentrations by factors as
large as three.
This discrepancy is rather disturbing because the creep recovery experiments
were obviously designed and tested with considerable care (186). Furthermore,
as pointed out earlier, creep recovery is the most direct method for measuring
J~. Most of the likely errors (such as inertial effects in the instrument or nonattainment of steady state) would tend to give values which are too small rather
than too large. Similar but smaller effects have been observed by other methods
in solutions of polyisoprene (67) and poly (a-methyl styrene) (187). In the latter

i
63
,?
l.C
'o d
o
/ / ~ o ~q~
-3,
oq
o
\\
Mw=41i,O00
-~QI
QO
' o,6
'
( gm/m,t )
Fig. 5.13. Values of J~R for solutions of a narrow distribution polystyrene (Mw=411000).
According to several investigators and methods. Symbols are the same as Fig. 5.12,
except "(3 from N 1 (184) and Cr from three methods (185)
a correspondence was noted with a jelly-like appearance of the solution. It is

entirely possible that creep recovery is detecting a real effect in the region where
reduced compliance is beginning to fall, one which is missed or seen only
imperfectly by the less delicate methods. Recognizing this possibility, we will
nevertheless assume that normal behavior corresponds to the general trend of
data in Figs. 5.12 and 5.13 and omit the data in Ref. (182) from later correlations.
Figure 5.14 shows J ~ for undiluted, narrow distribution polystyrenes from
a variety of investigators and methods (124, 163, 165, 180, 181, t88, 189, 191-193).
Values from one study using stress relaxation (188) lie below the general trend.
However, stress relaxation data on the same samples (the Dow S series polystyrenes), when first fitted to an analytical form for E(t), yield values in good
agreement with the other data in Fig. 5.14 (165). It seems clear that the stress
relaxation method is sensitive to extrapolation errors. Values from a study using
dynamic moduli appear to be consistently higher than the general trend for
molecular weights less than 105 (124). Values obtained from creep recovery (163),
normal stresses (18t, 193), and flow birefringence (180) are in fairly good
agreement; values from tensile creep (19I), the dynamic moduli (124, 192), and
64
W.W. Graessley:
::
= ::~--.-...--...~ _ ~
( M < IO0,O00)
REF 124
t0
P
q
q
6
o
d
d
k~ k.~.xo~ -o
-~,,9
ct
0.1
REF. 188
OC
'
'
','
. . . .
l'o5
~%~
. . . . .
. . . .
,o7
Mw
Fig. 5.14. Reduced compliance vs molecular weight for undiluted polystyrenes of narrow
molecular weight distributions. Symbols are: O from creep recovery (I63), (Y from G'(~o)
(192), O- from flow birefringence (180), CL from N1 (189), (~ from G'(~o) (M> 105 only)
(124), .(D extrapolated from steady state creep (191),-0 from stress relaxation (I65), and
-(3 from N 1 (193)
Table 5.4. Values of JoR for narrow distribution polystyrenes from Pressure Chemical
o
Company, Pittsburgh, Pa., USA; JeR= J~QRT/
,~
Mw
10200
19700
51000
97000
160000
411000
860000
Prest (181)
0.51
0.62
0.52
0.40
0.28
0.13
0.075
Mills and
Wales (190)
Nevin (192)
0.53
0.51
0.34
0.32
0.40
0.39
0.48
0.20
0.092
Mieras and
van Rijn (189)
0.77
0.77
0.47
0.40
0.49
0.20
0.11
Crawley (193)
0.47
0.24
some of the normal stress data (189) are slightly displaced. The values of Zosel
(194), after his corrections for polydispersity, are also in good agreement.
The scatter may perhaps be partially accounted for by differences in polydispersity a m o n g the samples. However, as Table 5.4 shows, even measurements
on the same polymer may result in values which differ appreciable among
investigators. On the whole, the overall behavior is fairly clear: JeR maintains
65
a value near but perhaps slightly above the Rouse value for 10* < M < 105, then
becomes approximately proportional to 1/M for higher molecular weights 3.
Figure 5.15 shows reduced compliance for polystyrene solutions from several
sources (175-179, 184, 185), combined and plotted as a function of cM. All are
narrow distribution samples. Selected melt data from Fig. 14 are also included
(163, 180, 181). The preponderance of data are in fair accord with the Rouse value
for 5000 ~ cM ~ t00000. As in the case of the melt data, they tend to rise slightly
above 0.4, influenced perhaps by residual polydispersity in all the samples. The
shallow maximum in this region may well be caused also by the fact that only
a few of the longest relaxation times are shifted at the smaller values of cM, as
suggested by Janeschitz-Kriegl (18). The reduced compliances tend to be smaller
and more scattered at still smaller values of cM, probably reflecting an approach
to Zimm-like behavior for high molecular weight samples at low concentrations,
and the loss of flexible coil character for low molecular weight samples at high
concentrations (15). Also, as noted earlier, c[~/] rather than cM is the appropriate
reducing variable for JoR at low and moderate concentrations. F o r values of cM
greater than about 130000, the reduced compliance acquires a dependence on
cM, becoming proportional to (cM)-1. No important difference in reduced
behavior between solutions and melts is apparent despite the 100-200C
difference in the temperatures of measurement.
Confirming an earlier conjecture (195), the data on several linear polymers
follow a similar pattern. Figures5.16 and 5.17 contain results for narrow
distribution samples of 1,4 polyisoprene (I66, 167, 196, 197) and poly (a-methyl
styrene) (161,176, 198-201). The onset of (cM)-1 behavior seems clearly related
to the same processes that produce the plateau region and the transition in
~/o vs M behavior. In almost all cases (cM)-1 dependence in JeR begins at a
3 The steady-state compliance is exceedingly sensitive to the high molecular weight tail
of the molecular weight distribution. Evaluation of J~ depends on a satisfactory handling of
two limiting processes. The stress or deformation must be small enough to achieve linear
viscoelastic response, and the time must be long enough (in the case of creep and creep
recovery) or the frequency low enough (in the case of dynamic experiments) to reach the
limiting response. These conditions merge in the normal stress and flow birefringence
methods, requiring low enough stress to achieve the limiting form, with no further
changes at still lower stresses. The contribution of trace amounts of high molecular weight
components in an otherwise narrow distribution sample is very sensitive to stress. It is
possible, although it is only a conjecture, that this contribution may be lost in methods
such as normal stress measurements which involve relatively high stresses and large
deformations. The same sample in low stress methods such as creep recovery may then yield
higher values of J~ unless the tails have been carefully removed by fractionation. From the
standpoint of obtaining the true value of J~ for a sample therefore, the creep recovery
is clearly preferred.
However, to obtain a value which represents the contribution of everything except
a trace tail, the normal stress meast~ement may in fact give acceptable results. It is
somewhat reassuring that .12values deduced from normal stress and flow-birefringence data
on undiluted narrow distribution polystyrenes agree with the Ptazek results (163),the latter
having been obtained by creep recovery with very carefully fractionated samples. Similarly,
the high values of J~ obtained by creep recovery on solutions of the Pressure Chemical
Samples (182)may be the result of trace tails at high molecular weight whose contribution
is missed by the other methods. It is obviously dangerous to place too much reliance on
such fortuitous behavior, however.
66
W. W. Graessley:
......
9q
q
.-
q ? o-c? %_ e_cOO-
O~
O0.-
0-
*co.-
Qt
O-
107
e~l w
Fig. 5.15. Reduced compliance vs the CAw product for solutions and undiluted samples of
narrow distribution polystyrenes. Symbols are: from flow birefringence (179), (U from
several methods (185), O- from N1 (178), (D.. from G'(~o) and N1 (177), ~ from G'(~o)
(I75, I76), .0 from N1 (184), from creep recovery (I63), 0 from G'(o)) and N l (181),
and 0 - from flow birefringence (180). The filled points are undiluted polystyrenes
characteristic value of cM which is roughly 3-4 times the value of 0 h - Figure 5.18
is a combined plot of reduced compliance vs cM/QM, including fragmentary
data on narrow distribution polyvinyl acetate (I76, I95), polybutadiene (202, 203),
and poly(dimethyl siloxane) (204). A rather tenuous estimate for polyethylene
(204) has also been included. Some published data on narrow distribution
polyvinyl acetate (205, 206) have been omitted because of possible complications
due to long chain branching.
The meaning of these observations on narrow distribution systems is the
following (153). In undiluted polymers the steady-state compliance is directly
proportional to molecular weight and approximately equal to the Rouse value
for molecular weights both large enough to display random coil behavior, but
smaller than a characteristic molecular weight M'. For molecular weights greater
than M~ the compliance approaches a constant value (Je)~. Values of (jo)~
estimated for the polymers in Fig. 5.18 are given in Table 5.5. Values for truly
monodisperse samples would undoubtedly be slightly smaller. The characteristic
molecular weight M~ is a small multiple of Me. The same molecular weight
behavior occurs in concentrated solutions and the characteristic value, (M~)sozn.,
has the same concentration dependence as (Me)~o~n.and (Mc)soln.,namely,
,
(M~)sol.. = -c- M~.
(5.23)
67
1.0
O
XII~
0/SLOPE= - I
\iN i~
\
\
\
\
\
0~0
cM w
Fig. 5.16. Reduced compliance vs c/~wfor solutions and undiluted samples of cis-polyisoprene.
Symbols are: undiluted samples from steady state creep (166, 196), 0 undiluted samples
from N1 (197), and O solutions extrapolated from steady state creep (196)
\'
'
\
\
I
9
%~.~o~
P9~o9
q)
,,,,.---REE
200
b~ _ o
,% ~ o ~ # oOO~O~OoX
o
@oO~o'o~D
fl=
O00(]i}
O.
O0
(ID
,o~
io~
CMw
Fig. 5.17. Reduced compliance vs c~rw for solutions and undiluted samples of poly (~-methyl
styrene). Symbols are: from NI (198, t99), ~ from G'(co) (176), Q from N 1 (20t),
and from stress relaxation (161)
68
W.W. Graessley:
O
O
99
0-
~,IF
\,,
~ . \ \ POLYSTYRENE
~
~POLYDiMETHYL
StLOXANE
i_._
__
0i
1.0
IC)l _
i I0|0
cMw/pM c
Fig. 5.18. Reduced compliance vs c~lw/QMcfor solutions and undiluted sample of linear, narrow distribution polymers. The lines for polyethylene and (polydimethyl siloxane) are based
on G'(co) data as reported by Mills (204). Points are selected values for polyvinyl acetate
solutions, (176) and Q (195), and undiluted 1,4polybutadiene, (202) and 0- (203)
The values of Mc were taken from Table 5.2
Table 5.5. Limiting compliance for undiluted linear polymers of narrow distribution
Polymer
Polyethylene
1,4 Polybutadiene
Poty(~-methyt styrene)
Poly(dimethyl siloxane)
Polyvinyl acetate
cis-l,4 Polyisoprene
Polystyrene
T C
190
25
200
20
30
- 30
200
(J)o~ (cm2/dyne)
G (dyne/cm 2)
G(J)o~
0.22 10-6
0.25
1.0
1.0
1.25
1.4
1.75
-11.5 x
3.2
3.0
2.65
3.9
2.0
-2.9
3.2
3.0
3.3
5.5
3.5
10 6
The steady-state compliance is, furthermore, inversely proportional to the

first power of concentration and nearly equal to the Rouse value for concentrations high enough for appreciable coil overlap to occur but less than
QM~/M'. At higher concentrations the compliance becomes inversely proportional to the square of concentration. Differences in the reduced compliance
due to temperature and solvent power are negligible, and differences from one
polymer system to another are greatly reduced if the polymers are compared
through the reduced variable cM/QMc. Values of M" were obtained from the
intersection of lines drawn through data at large and small values of cM and
are given in Table 5.2. When data were available only in the asymptotic region
for large values of M, as in the case of 1,4-polybutadiene, polyethylene and
69
poly(dimethylsiloxane), values of M" were estimated for the undiluted polymer

with the relation for the intersection with Rouse behavior: ME = (J)ooo RT/0.4.
Judged by the agreement of data on even high molecular weight fractions of
polymethyl methacrylate with the Rouse theory (t46), the value of M" for that
polymer exceeds 150000.
One early proposal for the form of the monodisperse compliance curve, based
primarily on data beyond (M')soln., was (I 78):
1
LR
= 0~1 "b O~2
cM
o M~
(5.24)
in which the parameters ~1 and ~z were taken to be independent of c, M, and

T, and were assumed to be the same for all linear polymers. Another proposal,
cast in a similar dimensionless form, is (207):
(~)2
2 ( cM t2
(5.25)
The subsequent comparisons of several authors (153, 199, 208) have shown that
the change in behavior near (M~)so~,. is more abrupt than the rather gradual
transition implied by Eq. (5.24). Equation (5.25) gives a behavior somewhat closer
to observations. The combined results in Fig. 5.18 can be summarized by the
expression (for r/0 ~>qs):
jo =
0.4 M/cRT
( cM ]2]~/2
(5.26)
1 +0"08 \QM ] j
Because of apparent residual variations in the ratio M~/M among the different
polymers, a more accurate expression can be formed with the reduced variable
cM/QM~:
jo =
0.4
M/cR T
(5.27)
cM 1~]~,2
Again the reader must be warned that a large proportion of the jo data in
the correlation, summarized in Eqs. (5.26) and (5.27) and Tables 5.2 and 5.5, are
based on normal stress measurements (total thrust in plate-cone rheometers) with
attendant uncertainties about whether limiting behavior was attained. Also, in
70
w.W. Graessley:
some polymers, such as polyethylene, polyvinyl acetate, polydimethyl siloxane,

and polybutadiene, they are based on the results of only a few investigators or
on a limited number of samples with narrow but uncertain polydispersity. Even
in the most thoroughly studied polymer, polystyrene, there are significant
unresolved questions on the concentration dependence of jo and even on the
value of jo for particular samples. Thus, although Eqs.(5.26) and (5.27) are
attractively simple, and probably describe overall behavior fairly well, one must
expect some variations from these forms as more complete and accurate
data appear.
5.4.5. Effects of Molecular Weight Distribution
The discussion to this point has dealt with polymers of narrow molecular
weight distribution. Distribution, in fact, affects all viscoelastic properties of the
plateau and terminal regions (15). As an molecular weight distribution broadens,
the relaxation time distribution broadens in a more or less corresponding
manner. Increasing amounts of relaxation occur throughout the plateau region,
making it progressively tess well defined. Two plateau regions may be discerned
in binary blends with a sufficiently large difference in the component molecular
weights (209). The initial plateau modulus G appears to be independent of
molecular weight distribution, subject of course to uncertainties due to the
increasing diffuseness of the plateau with increasing polydispersity.
Studies on both binary blends and whole polymers of moderate polydispersity have established that no at constant friction factor ~0 depends on the
weight-average molecular weight alone (15, 16, 181,209):
n0 = n0(Mw)
(5.28)
in which rio(M) is the viscosity for a monodisperse polymer of molecular

weight M. Thus, even above Mr, this characteristic of the Rouse model [Eq. (4.27)]
is borne out. Exceptions may exist in samples of very broad distribution, where
some data suggest an average lying between Mw and )~z (210, 210a).
The recoverable compliance jo is very sensitive to molecular weight
distribution, especially to the tail of the distribution at high molecular weights.
According to the Rouse model [Eq. (4.28)], when samples with the same M,. are
compared, their compliances should vary as MzMz+ 1/M~. Based on the success of
the Rouse model mixing law for viscosity, one might hope for correlations
of the form:
jo = J*(Mw) M~Mz+l
--2
M~
(5.29)
in which J*(M) is the compliance of a monodisperse polymer of molecular

weight M, obtainable from the results of the preceding section. In some cases
71
the polydispersity correction is fairly well represented by Eq. (5.29). For example,
jo values in concentrated solutions of samples with exponential distribution
(Mw/Mn=2, M,M~+l/Iffl2w= 3) are roughly three times the values for narrow
distribution samples with the same Mw (178, 195, 2tl). Values for blends at low
concentrations (198) are also in reasonable accord. However, there are indications
that jo rises more rapidly than the Rouse factor M~M,+ ~/ffl2 for slight amounts
of polydispersity (165). The Rouse factor again underestimates the polydispersity
effect in binary blends of polystyrenes which are not too different in molecular
weight (Mz/M1 ,,~2), but gives reasonable results for poly(dimethylsiloxane)
blends with M2/M1 ~ 10 (212). On the other hand, jo values for blends of polyisobutylene with M2/M~ ~ 8 are higher than predicted (213), while those for
polyvinyl acetate blends agree fairly well with the Rouse factor over a wide range
of M2/M1 and molecular weight levels (148).
Forms similar to Eq.(5.29) but involving different averages have been
suggested. Mills (204) has proposed the form:
(5.30)
based on data for high molecular weight polyethylene, polystyrene, and polydimethylsiloxane. Zosel (194) proposed the same form but with an exponent
2.5 rather than 3.7 on the polydispersity factor. Leaderman et al. (213) correlated
j o / j . with Mw/Mn, but the implied dependence on the properties of the low
molecular weight tail (through M,) must be incorrect in general.
Phenomenological blending relations for jo have been suggested, based on the
properties of t/o and jo for narrow distribution systems and the assumption
that r/0 always obeys Eq. (5.28) in blends. The relaxation spectrum for a binary
system according to the linear mixing rule is (214)
Hb(Z) = ~o1H1(~/21) + ~o2H2(~/22)
(5.31)
in which qh and P2 are the volume fractions of the two components in the blend,
Hi(r) and Hz(z) are their relaxation spectra before blending, and 21 and 22
represent the scaling factors by which their relaxation times are altered in the
blend. The factors 21 and 22 are chosen to make the resulting expression for t/0
agree with Eq. (5.28). If the pure component spectra are chosen to give q0 ocM34
and JocM (the shifted Rouse forms), then a polydispersity factor of
MzMz+ 1/fI~ follows directly. If t/o oc M 3"4 and jo oc M for the pure components
(the forms actually observed for M > ME), the form
s:= J*(;zw) (
\Mw:
(5.32)
is calculated, which is too weakly dependent on polydispersity to be correct.
72
W.W. Graessley:
Higher order blending laws ohave been proposed (215). The quadratic law,
Hb(z) = (o2H11(z/211) + 2 ~01~2H12('c/~-12)q- ~2 H22('t'/)-22)
(5.33)
with the added stipulation that the cross term has the same form as the pure
component terms, yields
Mz
(5.34)
This polydispersity dependence is probably still too weak, perhaps because of

the unjustified assumption about the form of //12. The quadratic law leads
directly to the experimentally observed behavior jo oc c- 2 if one of the components
is solvent; the linear law gives jo oc c- 1
There is still a great deal of disagreement about the systematics of polydispersity effects on jo. The effects themselves are clearly large in concentrated
systems, and the quadratic blending law appears promising if a satisfactory form
for the interaction spectrum Hlz can be found. Generalization of the quadratic
law to arbitrary distributions, perhaps with the help of some approximate theory
for Hx2, would help greatly as a framework for systematizing the experimental
observations.
6. Molecular Entanglement Theories of Linear Viscoelastic Behavior

The discussion on linear viscoelasticity of narrow distribution polymers has
centered mainly on three properties, the plateau modulus G , the steady state
viscosity r/o and the steady state compliance j o Each property is associated with
a characteristic molecular weight: Me, Me, and M~ respectively in the undiluted
state. Near Mc the viscosity--molecular weight relation changes from ~/0 ocM~to
r/o~M 3''~. For M~> Me the relaxation time distribution is separated into two
parts, slow relaxations which are highly sensitive to molecular weight and rapid
relaxations which are insensitive to molecular weight. The modulus contributed
by the slow relaxations is GO; G is independent of molecular weight for
M >>Me, and Me = QR T/G . Near M~, the steady state compliance changes from
direct proportionality to independence of molecular weight. The characteristic
molecular weights stand in the order Me < Me < M~, are practically independent
of temperature, are similar small multiples of each other for different polymers,
and (in the case of Me for example) correspond to roughly the same chain
length from one polymer to another. Their values in concentrated solutions are
accurately proportional to the reciprocal of polymer concentration and independent of solvent nature.
73
These properties, together with the order-of-magnitude agreement between

viscosities for undiluted polymers below Mc and calculated values from the Rouse
theory, suggest that behavior at high molecular weights is modified by an interaction between the chains. This interaction is a function of the product c M for
concentrated solutions and melts, and appears to be an intrinsic property of
concentrated long-chain systems. A rope-like entangling of chains satisfies these
qualitative requirements. The problem, of course, is to convert this idea into an
acceptable theory. From the standpoint of linear viscoelasticity this means relating
the relaxation time distribution in the terminal region to molecular weight and
concentration. Before proceeding to mechanistic explanations, it is useful to
consider what kinds of information are available on the terminal distribution
without appeal to specific rrtolecular arguments.
First, for M ~>M~ the terminal relaxations are fairly well separated from the
transition relaxations. In view of Eqs. (3.24) and (3.25), the values of ~/0 and jo
for M >>M~ are properties of the terminal relaxations alone. The mean relaxation
times of the terminal distribution are therefore given by Eqs. (3.26)-(3.28),
combined with the experimental results in Section 5.
r/o M~ O
"cn=~lo/G -
w = qo j o =
c 2 RT
0.4
(6.1)
'
r/0 ME Q
c z RT
(6.2)
'
and
Mt
-Zw
= J~oG ~o = 0 . 4 - -
c
Me
~o
= 0 . 4 (M~)so~..
(6.3)
(Mo)so~..
Thus, what might be termed the relaxation time polydispersity, "Cw/rn,tends to a

constant value for M ~>Mc (see Table 5.5). The difference between Mo and ME,
[or (M~)~o~,.and (M~)~o~,.]is a direct result of this polydispersity of relaxation times.
If the terminal spectrum were simply a shifted Rouse spectrum, rn would be
the same as that given in Eq. (6. I) but ~w and Zw/Vnwould be different:
z,,, -
0 . 4 r/0 M
cR T
~w
0.4 cM
~n
QM~
(Shifted Rouse spectrum),
(6.4)
(Shifted Rouse spectrum)
(6.5)
so the relaxation time polydispersity would increase with both concentration and
molecular weight, contrary to observation.
74
W.W. Graessley:
Any satisfactory rheological theory of entanglement for monodisperse chains

must not only account for the observed shift in the mean relaxation time zn with
concentration and molecular weight [Eq. (6.1)] but also for the relatively narrow
distribution of relaxation times in the terminal spectrum. It must also conform in
a natural way to the very sensitive relationship observed between relaxation time
polydispersity and molecular weight distribution. There are also questions from
the viewpoint of chemical structure. What is it that controls the value of Me for
a given polymer, and what precisely constitutes a chain entanglement? It is the
kind of localized, rope-like interaction that comes to mind from the analogy with
networks, or is it a more diffuse topological interaction that simply manifests
itself rheologically as network-like response at short times and enhanced friction
in steady flow?
This section covers the present status for both the rheological and structural
questions. The first section reviews the progress on formal molecular theories
which deal with the rheology of concentrated polymer systems without explicit
consideration of the mechanistics of intermolecular interaction. Successful and
truly predictive theories of that kind would of course remove much of the incentive
for a detailed mechanistic discussion of entanglement. At the very least, they
would serve to isolate particular characteristics of the interaction for examination.
Following this is a discussion of network theories which take explicit account of
interchain coupling by treating polymer liquids as relaxing network systems. This
is followed by a review of several theories which seek to explain the shift in
zn, actually the M 3'4 dependence of t/0 and the presence of a plateau modulus
which is independent of M, through an enhanced frictional resistance to large
scale motion. Next is a review of theories on the terminal relaxation spectrum,
with particular attention to their consistency with the observed behavior of jo.
Following this is a comparison among theories which deal with the characteristic
molecular weights. The last section is a brief review of theories dealing with the
effects of polydispersity.
6.1. Formal Molecular Theories
Chikahisa (26) and Williams (217-219) have examined flow behavior in
concentrated polymer systems without detailed consideration of the mechanism of
intermolecular interaction. Williams explicitly limits his discussion to unentangled
systems; Chikahisa uses an entanglement terminology, although not in a specific
way. Both approaches grow out of the formalism which was developed to deal
with transport properties in small-molecule liquids.
Williams begins with Fixman's equation (220) for the stress contributed by
intermolecular forces in flexible chain systems. The theory assumes that the
polymer concentration is high enough that intermolecutar interactions control the
stress. The shear stress contributed by polymer molecules in steady shear flow
is expressed in the form
1
~V
8V]
(6.6)
75
The intermolecular potential l/(x i, ~) is the energy associated with the interaction
of a pair of molecules whose centers are separated by vector distance r with
components x 1, x 2, x 3:
(6.7)
V = A S f ( r ' ) f ( r ' + r) d r ' ,
f ( r ' ) being the segment density distribution function of the molecules. In the
context of Fixman's theory, the potential has a thermodynamic origin. The
parameter A governs the magnitude of the pair interaction energy at equilibrium:
d 2 5"
A = M 2 kT
(6.8)
dc~-~-
and e* k T is the free energy of mixing per polymer segment. The potential also
depends on shear rate through the effect of the latter on the segment density
distribution.
The pair correlation function g describes the distribution of molecular
centers in the solution. In concentrated systems at rest, 9 ~ 1. Flow alters 9, and it
is this change which gives rise to the drag forces. For sufficiently slow shearing
flows,
xl x2
g(9, r) - 1 + C ~ 9
(6.9)
k T (x~ + x~ + x 2 ) s/2
in which ( is an intermolecular frictional coefficient and C is an undetermined

parameter with dimensions of (length) 5.
Viscosity at low shear rates is obtained by substituting the results of
Eqs. (6.8) and (6.9) into Eq. (6.6), and evaluating V from the segment density
distribution of undisturbed random coils.
d2
3 ]3j2
~/;:~ = tlo = 21rc 2 ( C --~c2 \ 4 - ~ /
(6.10)
Williams then estimates ( with Kirkwood's equation, employing the FixmanPeterson expression for 9(r) in chain molecules. He arrives at a variation with
molecular weight of the form ( GcM TM. Williams then estimates the chain length
dependence of C by a dimensional argument, concluding that C ocS 5 ocM 5/2.
The final result is of the form
rio 0 M 5/4 c 2 = (cM5/8) 2 .
(6.11)
76
W.W. Graes~ey:
Chikahisa (216) proceeds through a direct analogy with small molecule transport theory. At an intermediate stage of calculation, he arrives at an equation
which is formally similar to Eq. (6.6):
a = ) ~ PI(r) Pz(r) dr
(6.12)
in which ,1 is the radial distribution function for molecular centers, and P2 is a

pair-wise intermolecular force. The force is taken to be purely frictional in nature,
directly proportional to chain length, and proportional as well to the mean
shared volume of the pair. Using dimensional arguments and an approximate
expression for the radial distribution function, Chikahisa calculates an expression
for viscosity in highly concentrated systems. In the terms used in this review, the
result for E = M/Me - 1 >>1 is:
rlo =
(6g)3/2
270
(o v2
S 5 E 3/2 n
(6.13)
which can be rearranged to give:
rl= ( (v-S62n ) ( (6rO3/2vS3

(Rouse term)
(6.14)
(Entanglement factor)
or qo oc M 3 in undiluted systems.
It appears that the formal theories are not sufficiently sensitive to structure to
be of much help in dealing with linear viscoelastic response: Williams' analysis
is the most complete theory available, and yet even here a dimensional analysis
is required to find a form for the pair correlation function. Moreover, molecular
weight dependence in the resulting viscosity expression [Eq. (6.11)] is much too
weak to represent behavior even at moderate concentrations. Williams suggests
that the combination of variables in Eq. (6.11) may furnish theoretical support
correlations of the form r/0 = f(c[q]) at moderate concentrations (cf. Section 5).
However the weakness of the predicted dependence compared to experiment
and the somewhat arbitrary nature of the dimensional analysis makes the
suggestion rather questionable.
Unlike Williams, Chikahisa attempts to deal directly with entangling systems.
The result is a viscosity expression which is not far from that observed
experimentally, but the form is unfortunately more dependent on a series of
intermediate assumptions about the nature of the friction forces than on the
basic transport theory itself. Although not implausible, the assumptions are
nevertheless arbitrary and lacking in theoretical justification.
77
6.2. Network Theories

Network theories view a highly entangled polymer system as a network
structure with interconnections extending throughout the system. The entanglement junctions are the cross-links; the sections between junctions are the
network strands. At equilibrium the concentrations of strands and junctions are
constant, although individual elements of the network are transitory. Each is
created by the thermal motions of the molecules, and each is eventually lost by
the same process. Network theories employ the kinetic theory of rubber elasticity
to calculate the mechanical properties in flow. The ends of each strand are
assumed to move with the medium from the time the strand is created until it
becomes inactive through the loss of one of its junctions. The stress is calculated
as the sum of independent contributions from network strands of all ages.
The relaxing Gaussian network of Green and Tobolsky (4) is the earliest
version of this model. Lodge (I2) and Yamamoto (I3) independently derived
constitutive equations for similar systems, based on a stress-free state for each
newly created strand and a distribution of junction lifetimes which is independent
of flow history. For Gaussian strands in an incompressible system:
P~j = - Po 6ij + ~ M ( t - t') [Cij 1 (t', t) - ~ij] dt'.
(6.15)
--00
The pressure Po represents the arbitrary additive contribution to the normal

components of stress in an incompressible system, 6~j is the Kronecker delta,
C~] l(t',t) is the inverse of the Cauchy-Green strain tensor for the configuration
of material at t' with respect to the configuration at the current time t [a
description of the motion (221)], and M(t) is the junction age distribution or
memory function of the fluid.
Depending on the form chosen for the memory function, the Gaussian
network model can show many of the qualitative features of polymer flow behavior (78): linear viscoelastic behavior for sufficiently small or sufficiently slow
deformations, rubbery solid-like behavior for rapid strains, viscous liquid behavior for slow strain rates, and a positive first normal stress difference in
steady shearing flow. On the other hand, the model yields a shear viscosity
which is independent of shear rate, contrary to observed behavior. This defect
can be remedied by relaxing the restriction to Gaussian chains or the
assumption that the memory function is independent of flow history (222). In
either case there is a substantial increase in complexity of the constitutive
equation. Flory has criticized such theories (223) as being inconsistent with the
rigorous theory of composite Gaussian networks. According to the latter, new
strands which are created by cross-linking an already deformed network are not
stress-free, as judged by the retarded stress-relaxation of a deformed system in
which the junction concentration remains constant but individual junctions are
lost and new ones formed at random. Generalization of the elastic liquid model
to remove this inconsistency does not appear to have been made.
78
W.W. Graessley:
For our present purposes, the network theories suffer from an additional
defect. They supply no information on the form of the memory function. The
memory function must be obtained for each system by rheological experiments,
and there is no way at present to predict how it should vary with the molecular structure of the polymer. For example, M(t) can be obtained from the
stress relaxation modulus G(t):
dG
dt
(6.16)
k T ~ e-'m/,.
(6.17)
M(t) and using Eq. (4.10), for example:
M(t)=
i=1
IfM(t) is known, the behavior of the model in any flow situation can be calculated.
In network models the molecular arguments supply a form for the constitutive
equation, but do not provide the detailed connections to molecular structure. As
such, they provide a bridge between molecular theories which incorporate
specific structural information in rather specific flow situations and continuum
models which can generalize such information to arbitrary flows.
6.3. Entanglement Friction

Mutual obstructions to motion, such as would arise from a simple looping
of chains, must be present in any system of interpenetrating chain molecules. In
systems without permanent crosslinks these obstructions will only affect the
number of pathways available for configuration change and can not in themselves
affect the equilibrium distribution of configurations. It is important to distinguish
between two properties of mutually excluded volume of polymer chains. One is
the net excluded volume, a thermodynamic property which depends on the
mixing characteristics of the components and affects equilibrium properties such
as chain dimensions. The other is a physical exclusion of volume which always
prevents simultaneous occupancy and thus precludes the passage of one chain
directly through the backbone contour of another. Although the thermodynamic
excluded volume is probably small in concentrated systems, physical excluded
volume is always present, and chain connectedness will reduce the number of
topologically acceptable pathways between large-scale configurations, thereby
reducing the rates of configurational change.
If a steady systematic motion is imposed, those topological arrangements
which present barriers to configuration relaxation will come into play. As in
dilute solutions the chains are continually drawn into a disturbed distribution of
configurations by the external motions and respond by diffusing collectively
toward an equilibrium distribution. In concentrated systems each chain must
79
select paths which carry it around the topological impediments presented by

its surroundings. The result is a retardation of the collective response, appearing
macroscopically as a frictional resistance to steady motion. The effect must of
course be quite different in detail than simple Stokes friction for chains in a
continuous Newtonian medium. Theories of entanglement friction are based on the
hope [and some experimental evidence (33)] that these differences will cancel in
the large, at least for sufficiently slow or sufficiently small deformations, and that
entanglement effects can be represented by an enhancement in the molecular
frictional coefficient. All such theories ignore random thermal motion and adopt
a kind of"limp rope" picture for estimation of the topological contribution.
6.3.1. Theories of Entanglement Friction
The combined viscosity and diffusion measurements on concentrated systems
by Bueche et al. (33), were described in Section 2. The results suggest that
viscosity behavior above Mr can be explained simply in terms of an enhanced
Stokes friction for the molecules. Bueche (7, 224) suggested that since polymer
molecules in concentrated systems are looped through the coils of neighboring molecules, relative motion must be resisted by the need for these loops
to slip around one another. Each molecule is assumed to participate in an
average of E such temporary couples, creating (for E ~>1) a temporary network
vE
of Gaussian segments joining ~
couples per unit volume. The coupling frequency is taken to be proportional to molecular weight. For E >>1 the molecular
weight between coupling joints is M = M/E. In concentrated solutions, E occM,
assuming entanglement coupling to be simply a special kind of segment-segment
contact (Section 5).
A systematic motion imposed on the center of gravity of any molecule in
such a system induces additional systematic motions, that is, motions beyond
those which would occur if the chain were being pulled through a monomeric
liquid. Bueche separated these motions into two types, an induced motion
in the surroundings caused by the tendency of the molecule to drag along its E
coupling partner molecules, and a snaking circulatory motion of the molecule
itself as it moves around the E coupling points. (When the molecules move
simultaneously, as in shearing flows, each molecule in the system plays both
roles and thus presumably partakes in both types of additional motion.)
Induced Medium Motions
Circulatory Motions
80
W.W. Graessley:
The induced motions in the surroundings are estimated by examining the

effects of a systematic velocity v imposed on one of the molecules, called here
the central molecule. The couples of the central molecule lead to N1 other molecules which move in response to the motion of the central molecule. Likewise, the
NI first order partners are coupled to N2 second order partners, and so on
through a hierarchy of orders: order j contains Nj molecules, coupled for the
first time to the central molecule through a sequence of j couples. The induced
velocity of each order is assumed to be a constant fraction s of the velocity of
the preceding order. The formal expression for the force resisting the motion of the
central molecule becomes therefore:
j=l
in which n is the number of main chain atoms and ~0 is the frictional coefficient per main chain atom. The slip factor s ( 0 < s < 1) is left as an undetermined parameter. [An average value s = 4/9 is found for pairs of looped
ropes sliding past one another without contact friction in a viscous medium; s rises
to (4/9) 1/2 if viscous drag is proportional to n z rather than n.]
Redundancy (multiple coupling between molecules or between members of
different orders, coupling between molecules of the same order, etc.) reduces
N) below its maximum possible value, E ( E - 1)J- i. Bueche asserts that the number
of effective j-order couples is given by:
,zc
(6.19)
N j = ~ 4 n r 2 Qflr) e -Qj ~'~/"Ed r .

0
The product v E is the total number of coupling units per unit volume in the
system and Qj(r) is the concentration of potential j-order coupling sites at a
distance r from the central molecule:
3
oj(r) = E ( E - 1); - ~
]3/e
2rt $ 2 ( 2 j - 1) ]
[
exp[
3r2
2 SZ(2j - 1)
(6.20)
Bueche's result (for E >>1, and with some harmless approximations to simplify
integration) is:
Nj =
y E S 3 (2j - I) 3/2 .
(6.21)
Insertion in Eq. (6. t 8) leads to (again for E >> I):

F = [K(s) y E S 3] n~o v
(6.22)

in which
K(s) =
(37i
sJ(2 J - i) 3/2 .
81
(6.23)
j=l
For the circulation contribution, Bueche treats the coupling points as fixed
points around which the central molecule must move in order to proceed in the
direction of motion. The segments between successive coupling sites are labeled
1, 2, ..., El2 out from the center of the molecule. Bueche argues that the speed
of segment i relative to the medium, vi, compared to v, the speed of the center
of gravity, is given by:
vi=il/2v.
(6.24)
The contribution of circulation to the frictional coefficient is then obtained by

equating the rate of energy dissipation for the chain to the sum of the rates
for the individual segments:
F.v=2
e/2
2 (o n
i=1
~ nev2o -
E/2
v2 Z i
(6.25)
i=1
giving finally, for E > 1:

F =
n(o v.
(6.26)
The combination of enhancement terms for the molecular friction coefficient

from Eqs. (6.22) and (6.26) with the Rouse equation for viscosity yields
Rouse
term
Contribution
from induced
medium
motion
Contribution
from induced
central molecule motion.
When the molecular weight and concentration dependences of the structural

parameters S 2, E, n, and v are inserted (S2ocM, EoccM, nocM, vocc/M), an
expression for viscosity in highly entangled linear molecules is obtained.
/Io oC~onv2S5 E2 (o oc (oc4M 3'5 .
(6.28)
82
W.W. Graessley:
De Gennes (225) has recently examined the diffusion of chains in a medium

containing fixed, loop-like obstacles. In effect, the chain is confined within a
tunnel defined by the obstacles along its contour. It moves by an inching
displacement, called reptation, which proceeds by the diffusion of length defects
or loops along the chain. Only the end of the chain has a choice of direction
for subsequent movement as it moves out of the tunnel. The contour tunnel is
therefore created continuously as the chain moves among the obstacles. The
molecular friction coefficient of chains in this environment is proportional to ~o n2,
rather than ~on as in a viscous medium without obstacles. This increased power
of n corresponds to a circulation effect since only the central molecule moves.
As in the Bueche result, the circulation contribution is directly proportional
to chain length.
The De Gennes method can be applied directly to Bueche's model without
the defect diffusion formalism. Consider a chain with n main chain atoms, mean
square end-to-end distance (r2), and a large number of entanglement obstacles
along its contour. For simplicity, suppose the number of main chain atoms
between successive obstacles and the scalar distance between obstacles are
constants, ne = n/(E + 1) ~ n/E and re =
respectively. The obstacles are
labelled (i = 1, 2, 3, ... E) beginning at one end of the chain. The sections of chain
between successive obstacles are also labeled; the vector distance from the i to
the i + 1 obstacle is ai and the unit vector in that direction is ai/re.
vs
A force F = F~u is imposed on the chain (FIE on each segment between

obstacles), F~ being the magnitude of the force and u a unit vector in the
direction of the force. For any segment i, only the component of this force
along direction al will be effective in moving the chain through its tunnel. The
net force acting on the chain and moving it with a speed v through its tunnel
is therefore:
Fs ~, u . a i .
reE i= 1
(6.29)
This force must be equal to the total frictional force for movement of the chain
through the tunnel, (onv. Thus:
Fs
E
Fsrs
v = (onre-----Eu. i=1
~' ai= onreE
(6.30)
83
in which r~ = u. Z at is the component of the end-to-end vector of the chain

in the direction of the force. The component of the velocity in the same
direction for the segment i is:
vi = v
a~
E aj
re
rs
(6.31)
The motion of the center of gravity of the chain must always be in the
direction of its end-to-end vector E at. The velocity component in the direction
of the force is therefore:
Y.v,
vs = ~ -
= (onr2E 2 .
(6.32)
The average over all chains in the system (presenting all possible directions of
end-to-end vectors)yield (vs). With (r 2> = (rZ>/3 = ErZ~/3, Eq.(6.32) rearranges
after averaging to give
Fs = 3 n E ( o (vs}
(6.33)
which provides a circulation factor of 3 E, the same form as Bueche's expression

but with a different numerical coefficient.
Eyring et al. (226) examine the entanglement problem from a somewhat
different point of view, the activated complex theory of liquid viscosity. In a
monomeric liquid the molecules move by random jumps from one equilibrium
position to another. The jump frequency is controlled by an activation barrier
between neighboring sites. According to activated complex theory, a shear stress
lowers the barrier in the direction of the stress and raises it in the opposite
direction, producing a bias in jump frequency and a net flow of molecules in the
stress direction. For low stresses, the expression for viscosity in a monomeric
system is:
6kT
~m~--" Ko
22
~1~3,~2
(6.34)
in which Ko is the equilibrium jump frequency, 2 is the jump distance and

21, 22, and 23 are the distances separating layers of molecules in the three
coordinate directions.
If the molecules are now joined to form chains n units long, the jump
frequency per chain is Ko n, while the j ump distance (displacement in the direction
of the jump) of the center of gravity of the chain is 2/n. The zero shear viscosity
84
W.W. Graessley:
of a collection of independent chains, comprised of units which themselves move

independently, is therefore:
r/o = ~mn.
(6.35)
The viscosity is thus directly proportional to chain length, as in the Rouse

model. The authors argue however in favor of an additional factor proportional
to n 1/3 because the distance between molecular centers in the direction of the
velocity gradient, 22, increases by this factor as n increases. Their result is therefore
~lo oc n 4/3 oC M 4/3 = M 1"33
(6.36)
for non-interacting chains.

Eyring and co-workers extend the calculation to entangled systems, postulating
that E coupling points are distributed along the length of each chain. These
junctions divide the chain into E + 1 chain sections, each section containing
n~ units. They argue that the average displacement of the center of gravity per
jump is ;t/n(E + 1) rather than 2/n because only one of the E + 1 chain sections
contains the center of the molecule. Their expression for viscosity becomes
I7o = qmn E 2
(6.37)
or, employing the form which contains the extra n 1/3 factor,
rio oC n4/3 E 2 oc M 1/3 = M 3"33
(6.38)
for undiluted polymers.

Thus, in effect, Eyring and co-workers ascribe all effects due to entanglement
to additional motion of the central molecule, obtaining thereby a circulation
term proportional to E z. Graessley (227) on the other hand lumps all extra
motion in the molecules of the medium. Again, each central molecule is assumed
to participate in E coupling interactions. Each couple involves a different
molecule moving through the environs of the central molecule in simple shearing
flow. Each passing molecule is divided into two parts by its couple with the
central molecule: a longer section L containing nL main chain atoms, and a
shorter section S containing ns atoms (ns + nL = n). Likewise, there are Es entanglement junctions along S and EL along L, so Es + EL = E - 1. Each passing
chain undergoes additional motion as it slips around its junction with the central
molecule. This extra motion is assumed to be furnished mostly by the short
segment S, which as a consequence moves with speed Vs relative to the
medium.
85
If segment S were coupled to the central molecule only, its segments would
simply drift as a particle cloud towards the junction and the drag force on the
central chain would be approximately
F = 2 ~o nsvs.
(6.39)
Because S is coupled with other molecules, its segments must traverse a

circulating path in moving towards its junction with the central molecule. With
the circulation contribution included, the drag force on the central molecule
becomes
F = 2 (o nsEsvs.
(6.40)
Finally, Vs is itself a circulation velocity with respect to the motion of section L,

which is assumed to proceed steadily with the velocity of the medium at the center
of gravity of the passing chain v:
F = 2~onsEsE~/z v.
(6.41)
The average for all possible values of ns(0 < ns< n/2), with Es= ns/ne and
EL = nL/n~ = (n -- ns)/ne, becomes finally
16
(~=4 105
71 I n E3/2(o
21/2420
(6.42)
or
(.=0.131 n E 3 / 2 ( o .
(6.43)
The quantity (e is the mean frictional coefficient contributed by each entanglement junction, defined in terms of relative chain velocities.
Spring-bead models relate frictional force to the relative velocity of the
medium at the point of interaction. The entanglement friction coefficient above is
defined in terms of the relative velocity of the passing chain. Since the coupling
point lies, on the average, midway between the centers of the two molecules
involved, the macroscopic shear rate must be doubled when applying the result
to a spring-bead model. Substitution of 2(eE for (on in the Rouse expression
for viscosity yields
r/o =0.262 n ( S Z v E 5/2"
6
(6.44)
86
W.W. Graessley:
The usual concentration and molecular weight dependence of the structural

variables thus combine to give
/'/0 0(2 ~0 c3'5 M 3"5 .
(6.45)
Aharoni has recently proposed a theory of flow based on entanglements

between loops on the surface of collapsed coils (43, 228). The basic picture of
amorphous polymer structure is almost certainly incorrect (see Section 2), and
the derivation of viscosity is even more speculative than the others in this section.
6.3.2. Discussion of the Friction Theories
All the above theories attribute the enhanced friction in entangled systems to
extra motions. In a crude way they speak to the main issue, which is the toss of
relaxation pathways due to connectedness in the chain environment. However, all
are rather arbitrary: they select certain types of motion for examination and
exclude others, while in fact the motions must be cooperative and interdependent.
In addition, the theories have features which appear to be incorrect, or at least
inconsistent, even within the limited realm of motions examined.
Bueche's analysis of medium motion uses Eqs. (6.19) and (6.20) to evaluate N~,
the number of) generation chains coupled to the central chain. Equation (6.t9)
is itself rather arbitrary, and, moreover, Eq.(6.20) ignores attrition of lower order
couples due to multiple coupling. That is, ( E - 1)N~_ 1 should replace E ( E - 1)J
in Eq.(6.20) for the calculation of 0~(r), and the N~ population should be obtained
sequentially from the N~_ 1 population. Furthermore, even if the general form of
Bueche's weighting function [Eq. (6.19)] is accepted, the exponent should logically
be QJ r , rather than ~j/vE, since the probability of redundancy in coupling depends
on the ratio (density of potential j-order couples)/(density of molecules) rather
than (density of potential j-order couples)/(density of total couples). If this
substitution is made, a factor E is lost in the final expression for qo, giving
rlo oc M 5/2 instead of M 7/2.
A more direct argument for M 5/2 dependence as the outcome of Bueche's
theory can be made. The total number of coupling points on any central molecule
is directly proportional to molecular weight M. The number of polymer molecules
lying within its pervaded volume (potential first generation partners) is proportional to vS 3. For random coil molecules vS 3 is proportional to M ~/2. Thus,
for sufficiently large molecular weight, the number of couples on the central
molecule will exceed the number of molecules available for coupling. At this
point the effective number of first generation couples must become proportional
to vS3ocM 1/2. The volume pervaded by potential j-order couples is proportional
(2j--1)3/2S3, so the number of molecules available for j-order couples is
v S 3 [(2j - 1)3/2 _ (2j - 3)3/2]. Thus, in the limit of sufficiently long chains,
N j ~ vS 3 [(2j - 1)3/2 - (2j - 3)3/2].
(6.46)
F oc (on K(s)vS3 v .
(6.47)
With Eq.(6.18),
87
A factor of E has thus been lost from the frictional coefficient [compare
Eq.(6.22)] so the viscosity is only proportional to M 5/2.
The Eyring analysis presents some problems also. Suppose one accepts their
suggestion that only the movement of segments between one pair of couples, that
containing the chain center, serves to advance the chain. Then, to be consistent,
the correct modification of the viscosity equation for small molecules is to replace
Ko by n~ Ko and 2 by 2/n, rather than the original substitutions, K o - - * n K o and
2 ~ 2 / n ( E + 1). The result is
n
~lo = rlmn - -
= ~lmn E .
(6.48)
ne
Thus, with this modified analysis the circulation contribution is proportional

to E, in agreement with the Bueche, De Gennes and Graessley results. However,
now a power of molecular weight is lost in the viscosity,
~/0~ M2
(6.49)
11o oc M 7/3 = M 2"33
(6.50)
or with the n 1/3 term included 4,
Graessley's analysis neglects both redundancy in coupling and circulation of

the central molecule. The interaction in shearing flow takes place between an
unyielding central molecule and coupled passing molecules. The latter circulate
independently through their own couples with the medium, producing slip at the
junction with the central molecule. This distinction between central chain and
passing chain is somewhat artificial. Since all chains in the system are equivalent,
the additional motions generated by such pair-wise interactions must be
independent ofwich chain is chosen to be central. A later analysis (212) removes
4 The Eyring equation for the viscosity of long chains without entanglements turns out
to have the same form as the Rouse equation if 2 is taken to be the length of an
equivalent bond in the random coil molecule, n the number of such equivalent bonds, and
212223 = 23 the volume per chain unit comprising an equivalent bond. The self-diffusion
coefficient D for a freely jointed chain of n bonds is Ko22/6n, and its radius of gyration S
is (n22/6) 1/2. In an undiluted system, the number of chains per unit volume v is 1/23n. The
combination of these relations with Eqs.(6.34) and (6.35) and the Einstein equation,
D = k T/(o n, gives
7o = 6 n ( o S 2v.
(6.51)
Aside from the numerical factor, Eq.(6.51) is the Rouse formula. The fact that the Rouse
form works well for short chains at constant ~o casts doubt on the need for an extra
factor of n 1/3.
88
W.W. Graesstey:
part of this deficiency by examining the couples themselves, requiring the

additional motion to be assumed by whichever molecule of the pair contains
the shortest strand leading from the junction. However, the response due to each
entanglement is still calculated separately; the additional motions caused by
other entanglements along the same chain are ignored.
Thus, all present theories of the molecular frictional coefficient have rather
serious deficiencies. Physical and mathematical approximations are made
throughout, and there is really no independent way to justify or evaluate their
importance. Ziabicki and Takserman-Krozer (229-23 I) have pointed out a number
of difficulties in bringing dilute solution concepts to bear on concentrated
systems through lumping intermolecular effects into the medium by simple
averaging. Although their conclusions about such approaches are perhaps too
pessimistic, it is nevertheless true that the present theory of entanglement friction
is far from satisfactory.
It is not clear how improvements can be made without real progress on the
difficult fundamental problems of diffusion in media with obstacles and cooperation of large-scale motions between interpenetrating chains which do not
violate chain connectivity. The De Gennes reptation model (225) makes a
significant contribution to the first problem, although in an admittedly simplified
system. Rigorous calculations or computer simulations on well-defined models
which relate to the second problem would be extremely valuable, even if the models
themselves were not completely faithful representations of the assumed physical
situation. It is not obvious how even to pose solvable problems, simplified or not,
which relate to interchain cooperation.
6.4. Theories of the Terminal Relaxation Spectrum

Bueche's theory attributes the enhanced viscosity of entangled systems to an
increase in the molecular frictional coefficient. With the aid of the Rouse model,
Ferry et al. (1I) extended this idea to other viscoelastic properties. They
suggested that entanglement friction influences only the large-scale motions of
each chain, thereby dividing the relaxation times of the Rouse model into two
groups. Relaxation times associated with the higher normal modes, corresponding
to cooperative motions over distances smaller than the entanglement spacing,
are assumed to be unaffected by entanglement. Lower mode motions, governing
cooperation which extends beyond the entanglement spacing, are slowed and
their relaxation times shifted to larger values.
When E is large (M > iv/C) the relaxation times of the affected modes are
multiplied by the friction factor ratio \ q ] M ' q being the viscosity at Me. This
assumption accounts for many experimental observations. It automatically
builds in the correct molecular weight dependence of the viscosity. The gap in
the relaxation time distribution creates a plateau region, and the transition
relaxations comprise simply an unshifted Rouse spectrum. The plateau widens
with increasing molecular weight, and the modulus in the plateau region, being
dependent only on the chain length between entanglements, is independent of
89
molecular weight. However, the spacings of the terminal spectrum are incorrect.
As noted earlier, the experimental spectrum is much narrower than the shifted
Rouse spectrum which results from this procedure.
Ch6mpff and Duiser (232) analyzed the viscoelastic properties of an entanglement network somewhat similar to that envisioned by Ferry et al. Theirs
is the only molecular theory which predicts a spectrum for the plateau as well
as the transition and terminal regions. Earlier Duiser and Staverman (233) had
examined a system of four identical Rouse chains, each fixed in space at one
end and joined together at the other. They showed that the relaxation times of
this system are the same as if two of the chains were fixed in space at both ends
and the remaining two were joined to form a single chain with fixed ends of
twice the original size.
Ch6mpff and Duiser applied this tetrahedral model to entanglement networks. The suggest that the entanglement junction behaves as a site of large
frictional resistance for the motion of one chain relative to the other. They then
show that the relaxation times of an entanglement tetrahedron with equal strand
lengths are the same as if one chain possessed a large frictional resistance for
motion relative to the medium at the entanglement site and the other was un-
constrained. They use this result to transform a two dimensional network of

strands between entanglement junctions into a mechanically equivalent system of
independent primary chains with varying numbers of enhanced friction sites
(slow points) along their lengths. The relaxation spectrum of the network is
then formed from the spectra of independent chains in the equivalent system.
90
W.W. Graessley:
The network is decoupled by successively replacing junction tetrahedra by

two independent strands, each terminated by remaining junctions and one with
a slow point in the middle. The order of decoupling and assignment of slow points
must be performed in a rather special way: the symmetry of remaining junctions
must be preserved so that the same decoupling procedure can be applied to
them in turn.
The transition and plateau spectra resulting from this procedure are, for
E >>1, the same as those for a permanently crosslinked system with .crosslink
density equal to the entanglement density. The transition spectrum corresponds
to a set of identical Rouse strands with fixed ends. In the continuous representation [Eq.(4.33)], H(z)oc (zl/z) ~/2, z 1 being the longest relaxation time associated
with fixed-end strands as long as those of the original network. The plateau
spectrum arises mainly from cooperative motions extending beyond junction
points in the network. It is formed from the slow relaxations of long strands
between slow points in the equivalent system. Thus is depends directly on the
concentration of strands in the equivalent system which contain multiples of the
original network strands between slow points. For E ,> 1, Ch6mpff and Duiser
derive the plateau form, H(z)oczl/z, intersecting the transition spectrum near
z = z~, and extending from that point to longer times.
The terminal spectrum is furnished by cooperative motions which extend
beyond slow points on chain in the equivalent system. The modulus associated
with the terminal relaxations is rE k T, which is smaller by a factor of two than
the value from a shifted Rouse spectrum. It is consistent with a front factor 9 =
given by some recent theories of rubber elasticity (Part 7). The terminal spectrum
for E >>1 has the Rouse spacings for all practical purposes, shifted along the time
axis by an undetermined multiplying factor (essentially the slow point friction
coefficient). Thus, the model does not predict the terminal spectrum narrowing
which is observed experimentally.
The Ch6mpff-Duiser procedure of symmetry-preserved decoupling does not
appear to be applicable in an easy way to random networks in three dimensions.
Junctions which anchor strands of unequal length can be decoupled, but with
a considerable increase in difficulty (234). On the other hand, ChiSmpff and Prins
(235) have introduced a degree of randomness in the decoupling procedure (the
distribution of slow points within each uncoupled chain of the equivalent system
was taken to be random) without altering the results appreciably. Random
decoupling is much easier to implement, as shown below.
Consider the random decoupling of a system of v primary Rouse molecules,
each containing E entanglement points, with random assignment of the slow
points. In the equivalent system, the concentration of strands between successive
slow points which are made up of exactly i of the original vE network strands
is simply (for E >>1):
Ni = vE/2 i+ 1.
(6.52)
The transition and terminal spectra are then given by
H(z)= ~ Hi(z)
i=t
(6.53)
91
in which, using the continuous representation of the Rouse spectrum [Eq.(4.33)]:
H~(z) =
Nik T (zi/z)l/2
2
Hi(z) = 0
z < zi/a 2
z > zJa 2
(6.54)
"~i = Zl i2
and zx is the longest relaxation time of an origin network strand with fixed ends.
Combination of Eqs. (6.52)-(6.54) and summation yields
H(z)= v E k T ( ~ ) 1/2 I + 1
-2'
(6.55)
in which I is the first whole number whose magnitude exceeds a ('c/'~l)1/2. If I is

replaced by (z/zl) 1/2 to make an approximate continuous representation of the
spectrum,
t +
--
\zl/
e -~/~:1~1f21n2
(6.56)
This spectrum departs somewhat from the Ch6mpff-Duiser result in the plateau
region. The intensity remains larger in the vicinity of z = zl, but then falls below
the values for z >>zl. The terminal (shifted) spectrum for random decoupling is
unfortunately still Rouse-like.
The reptation model (225) also appears to produce a Rouse spectrum at long
times. In order to renew its configurations a chain must diffuse out of the tunnel
defined by the fixed obstacles along its length. De Gennes calculates the autocorrelation function for the end-separation vector, obtaining
c(o =
Z V
(6.57)
odd i
in which L is the configurational renewal time. Equation (6.57) is the same form
as C(t) for the Rouse model [Eq.(4.7)], with zr= 2"q. The renewal time is a
measure of the time required by a chain to diffuse out of its original tunnel.
De Gennes shows that z ~ M 3, a result which can be justified physically by the
following argument. To diffuse completely out of an original tunnel a chain must
92
W.W. Graessley:
move the length of the tunnel L, which is proportional to the molecular weight
of the chain. The diffusion coefficient D along this path is simply k T/~on. In order
of magnitude one must have D = LZ/zr, so
ZroC L 2 / D o c
M 3 .
(6.58)
The longest unshifted Rouse relaxation time is only proportional to M 2. The

extra power of M in the relaxation time is consistent with the circulation
contribution to qo which was described in the preceeding section. This modification
together with Eq.(6.57) suggests a simple Rouse spectrum shifted to longer
times.
Edwards and Grant (236, 237) have very recently proposed a new calculation
of self-diffusion and viscosity in highly entangled systems. They extend the
reptation picture of De Gennes by allowing the surroundings (the tunnel) to
diffuse as welt as the chain in the tunnel. They suggest that in fact the rate
limiting process may be the cooperative diffusion of tunnel and chain rather than
chain reptation through the tunnel, and obtain % oc M 3 for this case. A number
of intermediate assumptions are made in this calculation, and it is not easy to
judge their physical significance. The terminal spectrum obtained is of the form
"ri = T,i/i 4
i =" 1,
2. . . .
in which z 1 is the longest relaxation time. This spectrum has even wider spacings
than a shifted Rouse spectrum, at variance with the narrower terminal spectrum
demanded by the J o behavior.
Forsman and Grant (238) have recently proposed to handle the effects of
entanglement by adding an extra force to the Rouse analysis. The force on the
ith bead (i = 0, 1.... ) due to entanglement at the jth bead of a passing molecule
is expressed as the product of an unspecified coupling coefficient N~j and the
frictional force acting at thejth bead. Utilizing the balance of forces on each bead,
the frictional force on bead j is replaced by the sum of its osmotic and spring
forces. The result on the spectrum is the same as that obtained by replacing the
matrix A [Eq.(4.3)] of the Rouse theory by a new matrix A - M. The elements
of M are composed of combinations of the coupling matrix N.
The symmetry properties of M alone are sufficient to prove that only the odd
eigenvalues of A are affected by subtraction of M. Thus, according to this view,
only the relaxation times of odd order are affected by entanglement coupling.
Furthermore, the shifts diminish rapidly as mode number increases, so only the
longest odd relaxation times are affected.
This result is intriguing because the unshifted long modes could account for
the plateau relaxations. Also, the fact that only half the modes are shifted is
reminescent of the Ch6mpff-Duiser result, that the plateau modulus is only one
half the value given by the conventional kinetic theory of elasticity. Unfortunately
93
the coupling coefficients are unspecified, so the relative shifts and resulting spacing of the relaxations at long times are not given by the theory. An experimental
test was performed (239) in which G"(og) data on several concentrated polystyrene
solutions were fitted to determine the relaxation time distribution. The results
were consistent with shifts of the odd-order modes only, but other interpretations
seem possible also. Unless the shifts are such as to narrow the spacings, the
theoretical compliance jo will not match experimental behavior. If the same
shift factor applies to all odd relaxations the reduced compliance approaches
jOcRT/M= Z~oddii-4/(~oddi /-2)2 ~0.72 for M ~M. If the shift factor decreases
with increasing mode number, the spectrum broadens.
Thirion (239a) has suggested that the plateau and terminal regions are the
result of diffuse interchain interactions in a viscoelastic medium. He obtains a
modified Rouse spectrum by replacing the subchain frictional coefficient by a
time dependent "micro-memory" function. The theory is partly phenomenological
since the memory function is not specified. However, reasonable choices lead to
forms for G'(co)and G"(co)which are similar to those observed experimentally.
In view of the various suggestions on modifying the Rouse model to account
for terminal zone spectrum in entangled systems, it is worth examining what
effects can be produced by other modifications, particularly changes in the
matrix which controls the terminal relaxation time spacings. The terminal
relaxations of the Rouse model with E beads or slow points are given by the
non-zero eigenvalues of the E x E matrix A [Eq. (4.3)]. Ch6mpff and Duiser and
Ferry et al. obtained these same eigenvalues and therefore the shape of the terminal
is unchanged. Forsman and Grand require the eigenvalues of a difference matrix,
A - M. Depending on the choice of M, the spacing of the altered eigenvalues
could be either greater or less than the Rouse spacings. Hoffman (240) and
Ch6mpff and Prins (235) have speculated that the bead mobilities may depend
on the position of the bead along the chain. According to Graessley (212), the
mean entanglement frictional coefficients at the ends of the chain are small but
they grow to large values near the chain center, rising as approximately the
square of the contour distance from the chain end. The matrix governing the
eigenvalues in this case is B- A, in which B is a diagonal matrix containing the
mobilities (reciprocal of the frictional coefficients) of the successive beads.
Some narrowing of the spectrum does occur when there is a strong positional
dependence in the frictional coefficient. For example, the reduced compliance is:
JcRTtl~
J~R= ~ Z ~ ) z
~, i/2~
-- (E 1 / ~
(6.59)
where the 2~ are the non-zero eigenvalues of B.A. For the usual Rouse model,
the ~i are all equal and JoR = 0.4. The narrowest possible distribution would result
if all eigenvalues were the same, in which case JeR = t/E, which is the form observed
experimentally. If ~i grows as the square of the distance from the nearest chain
end, J~R is 0.31 for large E; for a cubic dependence, the value is 0.30 (241).
Vinogradov and co-workers (242) have reported some results by Pokrovsky in
94
W, W. Graessley:
which the frictional coefficients increase as the 2.4 power of distance measured
from the center of the chain. The reduced compliance is reported to be slightly
smaller, 0.315 in this case, but independent of E for E >>1.
In view of the comparative intensitivity of J~l~to the properties of B, it seems
unlikely that terminal spectra which are narrow enough to agree with experimental
compliance data (that is, to make JeR inversely proportional to E) can be produced
merely by introducing a distribution of frictional coefficients within the molecule.
The same holds true if the spring constants of the Rouse model are allowed to
vary with position (241).
A terminal spectrum of about the right shape is obtained if each entanglement
is treated as a separate bead-spring interaction between the molecule as a whole
and the medium (212).
The beads represent entanglement sites which are distributed uniformly along
the chain contour; the frictional coefficients increase rapidly with distance from
the chain ends. The spring, constant also depends on contour position, being
governed by the mean equilibrium distance of that position from the center of
gravity. The resulting spectrum is narrower than the Rouse spectrum, and
forE >>1:
1.798
(6.60)
jo = vEkT'
or
~O
j~oc E c R T
1
~c c 2 T
(6.61)
as observed experimentally. The tong relaxation times are also of the experimental form:
q0 M
ZwW E c R T
qo
oc c2--~ .
(6.62)
The failure of the Rouse theory was attributed to the pathological nature of
medium motions in entangled systems, and not any special defect in the Rouse
representation of the polymer chain itself. For Rouse chains in a deforming
continuous medium, the frictional force depends on the systematic velocity of the
bead relative to the medium. The frictional force on a bead is therefore a smootly
95
varying function of its internal coordinates. In entangled systems each interaction

site is well removed from the molecular centers, of both chains involved. Thus,
even if the molecular centers move affinely, the frictional forces at successive sites
along a chain will change abruptly both in direction and magnitude. Unfortunately, attempts to build this effect into the analysis while retaining the
Rouse model of the chain itself, for example by allowing the velocity gradient to
fluctuate from position to position along the chain, have not been successful. The
suggested loss of long-range correlations along the chain is therefore unconfirmed,
and the model modification that leads to Eqs. (6.60)-(6.62) must be regarded
as speculative.
Hayashi (243) has modified the Rouse model to account for both the plateau
and the observed forms of z I and jo. Following Ferry, he takes the medium to be
purely viscous for the high modes with a segmental frictional coefficient which is
independent of molecular weight. For correlated motions which extend beyond
the entanglement spacing he suggests that the medium is viscoelastic rather than
purely viscous as in the usual Rouse formulation. He uses (eoc M25( 0 and makes
the effective spring constants of the normal modes increase with decreasing mode
number. The frictional coefficient substitution is understandable, and amounts
to the original Ferry modification. The justification for modifying the spring
constants in the manner chosen is not clear. In any case, the modification should
take place before the transformation to normal coordinates, so that the physics
of the assumption can be seen. The result of the assumption is a narrowed
relaxation spectrum, with Ja~ oc 1/E, as observed experimentally.
Hayashi later suggests (244) that E is proportional to c at moderate concentrations and to c 2 at high concentrations, concluding therefore that jo progresses from c- 1 to c- 2 and finally c- 3 dependence as concentration approaches
the undiluted state. The prediction of c- 3 dependence seem to rest on the premise
that the number of junctions per unit volume Ev becomes proportional to c a at
high concentrations. The physical reasoning for any dependence other than c 2 is
not made clear in the paper, and the result should probably be rejected, at least
insofar as constituting a theoretical conclusion. As pointed out earlier, most
data support jo oc c- 2 and E oc c in highly concentrated systems.
Williams has derived the molecular weight and concentration dependence
of a viscoelastic time constant z 0 (actually the characteristic time governing the
onset of shear rate dependence in the viscosity) from his theory (217-219).
Employing a dimensional argument, he equates the parameters which control the
shear rate dependence of chain configuration and the intermolecular correlation
function. The result agrees with the observed form of characteristic relaxation
time in concentrated systems [Eq.(6.62)]:
z o oc qo/C 2 T.
(6.63)
The theory is not detailed enough to predict either numerical coefficients or the
relaxation time distribution. Examination of the dimensional argument suggests
that the form in Eq. (6.63) is probably not uniquely determined by the theory.
96
W.W. Graessley:
6.5. Theories of the Characteristic Molecular Weights
6.5.1. Relationships among Me, Me, and M~

Each of the viscoelastic parameters G, %, and jo has associated with it a
characteristic molecular weight which either measures an equivalent spacing of
entanglement couples along the chain (Iv/e, deduced from G with the kinetic
theory of rubber elasticity), or marks the onset of behavior attributed to the
presence of entanglements (Me and M~, deduced from qo and jo as functions of
molecular weight). Table 5.2 lists M , M , and M~ for several polymers. Aside from
certain difficulties in their evaluation, each is a rather direct and independent
reflection of experimental fact.
The characteristic molecular weights seem clearly related to the same underlying set of interactions, and their relative values provide useful tests for proposed
theories. In particular, if the interactions are strictly non-specific and topological,
then one might expect M J M e and M~/M to be the same for all linear polymers.
Differences among polymers would be reflected only in the differences of, for
example, their individual values of M . The latter differences might in turn be
expected to be related in some manner to the configurational and space-filling
properties of the chains of different polymers. Berry and Fox (16) point out
that M~ is approximately 2M~ for a variety of non-polar polymers. In mildly
polar polymers such as polydimethyl siloxane and polyvinyl acetate M appears
to exceed 2Me to some extent. In the case of polymethyl methacrylate M,/M~
may be as large as 58. Values of M~ appear to be approximately 3-4 M~, although
in polyisoprene the relationship appears to be more nearly M~ = 6M~.
Several of the entanglement friction theories predict values for M J M . The
ratio is obtained by setting equal the Rouse equation for % and the
corresponding theoretical expression for highly entangled systems. This is equivalent to evaluating M experimentally from the intersection of the low and
high molecular weight portions of % vs. M graphs.
K(s)
2(S2) 3/2Mx/2 = 1
Bueche(224):
~-4--o E~ - ~ -
Chikahisa(216):
(6zc)3/z
4~--5- 0 E3/2 (~_)3/2 M 3/2 = 1
Hayashi (243):
E = t.52
Graesstey (227):
0.262 ~5/z = 1
Graessley (212):
0.099 E~/2 = 1
(6.64)
in which Ec is M/M~, the number of entanglements per molecule at the

intersection.
The Bueche and Chikahisa relations reduce to roughly Mc/Mo=2 and
Mc/M e = 1 respectively, when typical values of S2/M and Q are inserted and
97
K(s) is taken to be approximately unity. Hayashi's result is Mc/Me= 1.52,

while Graessley's expressions give values of 1.71 and 2.52 respectively. Thus,
none of the friction factor theories is seriously inconsistent with the observed
magnitude of M J M e. The Bueche and Chikahisa formulas predict some residual
variation of M J M with (S2/M)3/2Q and the magnitude of Me itself, while the
Hayashi and Graessley results require M J M e to be the same for all linear polymers.
None of these results account satisfactorily for the detailed variations among
polymers noted earlier.
A theoretical value for M j M c can be calculated by equating Graessley's
expression for jo [Eq. (6-60)] with the Rouse expression:
0.4/vk T = 1.798/vE k T .
(6.65)
The result is E~ = 1.798/0.4, giving M~/Me = 4.5. With Mc/Me = 2.52 from the
same theory, M ' / M c = 1.8 is obtained. The observed values are about twice this
value; part of the difference is probably attributable to residual polydispersity in
the experimental samples.
Thus, the available theories are consistent with the observation that
M e < Me < M~. There is certainly no justification for the view sometime expressed
that the characteristic molecular weights differ because different types of
entanglements are responsible for the properties concerned.
It should be noted that the Bueche theory predicts a slightly different form
for the variation of (Mc)soln"with concentration than that inferred earlier on the
basis of experimental data:
c4tMctsola~7/2
_ . 4 MET/2
--
or
(Mc)soLn"=
M~ (Bueche theory).
Current experiments are not accurate enough to distinguish between this expression and Eq. (5.17). Finally, Bueche's theory is the only attempt to discuss
r/o vs M in the vicinity of Me. Berry and Fox (16) have concluded that the
predicted transition is too gradual. There is no theory for the behavior of
jo in the vicinity of ME.
6.5.2. Relationship between the Characteristic Molecular Weights and
Molecular Structure
Values of Mc (and Me and ME) are found to vary widely from one polymer
to another, although the variations are much smaller (ca _+ 50%) when comparisons are made in terms of n~, the number of chain bonds per molecule at
Mc. Fox and Allen (245) have pointed out that the variations are reduced still
98
W.W. Graessley:
further (ca __+30%) if values of S~Z/v are compared, S~ being the unperturbed
mean square radius of gyration of molecules with molecular weight Me, and v
the volume of polymer per main chain bond. The parameter S~/v comes directly
from the Rouse form and the empirical correlating expression:
Co X ( X ]"
~ = - 6 - \ x o /
(6.66)
in which a = 1 for X < Xc and a = 3.4 for X > Xc. This equation fits the experimental behavior of many linear polymers (16), X being the structure factor
of the Rouse theory, nS2v. For undiluted polymers, the parameter X reduces to
S2/v. Puzzling discrepancies remain however. For example, S2/v for polyisoprene
is more than twice as large as that for 1,4 polybutadiene, despite their similarity
in backbone architecture.
Hoffmann has attempted to estimate the magnitude of Mr directly from the
structure (246). He suggests that entanglements comprise a special class of intermolecular contacts in which one chain is looped tightly around another, and in
which the four strands leading away from this contact are oriented such that the
loop resists rapid deformations in the manner of a cross-link. His estimation of the
concentration of such arrangements in polyethylene goes as follows.
A tight intermolecular loop around any methylene group requires the
occupation of approximately 4 to 6 nearest neighbor positions by methylenes of
the partner chain. The possibility of a similar occurrence with another chain
at the same methylene group is negligible, and the possibilities on adjacent
methylene groups are greatly reduced. As a result of such packing considerations,
only about one methylene in three can be involved in a potential entanglement.
The fraction of these which are effective depends on the vectors drawn from the
junction to the next effective junction along each of the four strands.
According to Hoffman, these vectors must satisfy several conditions, each
condition having a certain probability p of fulfillment.
The angles AOB and COD should both be less than 90 (p = ~.

1 ~-). The loop
itself must lie on the correct side to engage when the chains are pulled in
opposite directions (p = ). The bisector of angle COD must make a fairly large
99
angle, say between 135 and 180, with the bisector of angle A O B ~ = (2 - l/~)).
The combined probability for all these events is therefore ( 2 - ]/3)/32 = 0.0084,
which is the fraction of potential loops which are effective. The number of
methylene groups per effective couple is therefore approximately 3/0.0084 g 350.
Since there is an average of two strands per couple, the value of ne is 175.
Experimental values for ne are in the range 100-250 for a variety of polymers.
The assumptions of the calculation are rather arbitrary, of course. The
result perhaps only illustrates that the observed order of magnitude of ne is not
grossly inconsistent with the idea of topological restrictions by loops. The
calculation gives no hint as to why the Fox-Allen parameter S2/v should
control the value of ne or Me however.
Other attempts to relate Me to structure have not been successful. Tonelli
(247) suggested that entanglement might be related to the probability of forming
closed intramolecular loops along the chain. As an example, he estimated the
probability p30-Tof forming at least one closed loop of greater than 30 main
chain atoms in a sequence of 700 main chain atoms. No correlation between
p70O
30 and the experimental values of Me or Mc was found in the six polymers
used in the calculation. Indeed, attempts to relate entanglement effects strictly to
the formation of transient intramolecular configurations or to local chain
kinkiness or roughness (248) seem bound to fail, since they imply a much greater
sensitivity to the details of local structure than is actually observed. Also, such
transient structures would not contribute to equilibrium stress in crosslinked
networks. The important contributions of entanglements to the equilibrium
modulus, indeed the deduction of similar values of Me in equilibrium and
transient experiments (see Part 7), seems to rule out such kinetically controlled
trapping mechanisms.
6.6. Effects of Polydispersity on r/o and jo
As noted earlier, qo in polydisperse systems appears to be a function of
M,, alone. One problem in entanglement theories is how to introduce the strong
molecular weight dependence of the entanglement friction factor without also
introducing a much higher molecular weight average than Mw in polydisperse
systems. Bueche (249) has applied his theory to a model polydisperse system,
using a separate averaging process for (e. He finds that viscosity depends on an
average molecular weight ~ t which lies between M~,. and M~, being closer to
Mw for small polydispersities and approaching /~z as polydispersity increases.
Graessley (227) arrives at an average molecular weight of the form:
)~rt
M 5/2 W(M) dM. S M W(M) dM
which also lies between/~w and Mz. For a most probable distribution Mw : Mt : Mz
as 1:1.23:1.50. Both results proceed essentially from the combination of a
separately averaged (communal) frictional coefficient with the ordinary Rouse
analysis for polydisperse systems.
100
W.W. Greassley:
In a later examination of the problem, Graessley (212) considered the drag

interactions between pairs of molecules in a uniformly effective medium, rather
than interactions between individual molecules and a uniformly effective medium.
The predicted viscosity in polydisperse systems was closer to )~rw dependence
than previously, although the sensitivity to high molecular weight components
was still too large. The fault may lie with the assumption that all communal
entanglement junctions, whether with large or small molecules, are equally effective
as obstacles for the motion of the interaction pair. As a result of the
assumption, pair interactions between the large molecules in a blend depend on
their surroundings only through the total entanglement density of the system;
the presumably reduced effectiveness as obstacles of their couples with small
molecules is ignored.
Very little guidance from molecular theory is available on polydispersity
effects in jo. This of course is not surprising in view of the unsettled state of the
theory of jo in monodisperse system. Recent theoretical calculations for entangled
polymer blends (212) were found to be in good agreement with experiments when
the blend components are not too different in molecular weight. However, as the
difference increases, the theory systematically overestimates jo (212, 250). The
same problem occurs as in the viscosity calculations: The contribution from pair
interactions between large molecules is overestimated when the difference in size
between the components is large.
7. Entanglements in Cross-linked Systems

Polymer systems can be converted to network-like structures by random
crosslinking. Network formation renders permanent some of the large scale
molecular arrangements which were present when the crosslinks were added.
Topological arrangements, which can influence only rate processes in a system
without crosslinks, become part of the permanent structure and may in principle
affect its equilibrium mechanical properties as well. In particular, some portion
of the entanglements which are present prior to crosslinking may be incorporated,
being trapped by the permanent connectedness of the network strands and loops.
If the system is subsequently deformed, the trapped entanglements may act as
additional restraints and increase the equilibrium modulus beyond that conferred
by chemical crosslinks alone.
Equilibrium moduli of networks can be used to distinguish to some extent
among various physical conceptions of the entanglement interaction. For example,
if entanglements were strictly transitory interactions with kinetically controlled
lifetimes (247, 248), one would expect no contribution to the equilibrium properties.
At the other extreme, the act of crosslinking could conceivably create as well as
trap topological constraints (251,252), resulting in an even higher entanglement
contribution in the network than in the same system without crosslinks.
Alternatively, the effects of crosslinking and pre-crosslinking entanglement might
101
simply be additive (253,255), with the equilibrium modulus controlled by the sum
of elastically effective network strands between chemical crosslinks and a
"trapped" fraction of the same entanglements that govern the plateau modulus.
Analysis of networks in terms of molecular structure relies heavily on the
kinetic theory of rubber elasticity. Altliough the theory is very well established in
broad outline, there remain some troublesome questions that plague its use in
quantitative applications of the kind required here. The following section reviews
these problems as they relate to the subject of entanglement.
7.1. Rubber-like Elasticity in Networks

The kinetic theory of rubber elasticity is so well known and exhaustively
discussed (17, 27, 256-257, 267) that the remarks here will be confined to
questions which relate only to its application in determining the concentration
of elastically effective strands. In principle, both network swelling properties and
elasticity measurements can provide information on network characteristics.
However, swelling measurements require the evaluation of an additional
parameter, the polymer-solvent interaction coefficient. They also involve
examining the network in two states, one of which differs from its as-formed
state. This raises some theoretical difficulties which will be discussed later.
Questions on local non-uniformity in swelling (17) also complicate the interpretation. The results described here will therefore concern elasticity measurements alone.
In the original kinetic theory the network is composed of Gaussian chains
or strands, joined at each end to junction points. Each strand is free to select any
configuration which is consistent with the instantaneous positions of its junction
points. The selection is made independently of the configurations of neighboring
network strands. If the network is deformed, the mean positions of the junction
points are assumed to move in an affine manner, i.e., in proportion to the
changes in macroscopic dimensions. The number of configurations available to
each strand is altered by the resulting change in end-to-end coordinates. The
result is a decrease in network entropy which gives rise to a mechanical
restoring force.
If no volume or internal energy changes accompany deformation, the network
is neo-Hookean (258), obeying the constitutive equation
Pij = G C:~ 1 _ Po 3ij.
(7.1)
The proportionality factor G is the equilibrium shear modulus of the network,

given by
(r 2)
G = g v k T (r~)
(7.2)
102
W.W. Graessley:
in which g is a parameter of order unity (which has continued to be the

subject of some dispute), v is the number of network strands per unit volume,
(r 2> is the mean-square end-to-end separation of the network strands in the
undeformed network, and (r2>o is the mean-square end separation of the same
strands with the junction-point constraints removed.
The "front factor" g as defined above 5 is unity in all the earlier theories
(17). Recently Duiser and Staverman (233) have obtained g = and Imai and
Gordon (259) g=0.54 with Rouse model theories which make no a priori
assumptions about the junction point locations after deformation. Edwards (260)
also arrives at g= and Freed (261) deduces that g = 1 is an upper bound
by similar approaches. The front factor usually assumed in the shifted
relaxation theory of the plateau modulus is g = 1, although Ch6mpff and Duiser
(232) obtain g--~ through their extension of the Duiser-Staverman result to
entanglement networks. The physical reasons for the different values of g in
different treatments are not clear at present.
James and Guth [see Ref. (17) for example] and more recently Eichinger
(262) obtain g = 1, but claim that the progressive addition of crosslinks to form a
network causes a reduction in chain dimensions through network shrinkage.
In a sample calculation James and Guth obtain (rZ>/(r2>o =, although the
implication of their suggestion is that the modulus should depend quite generally
on the rate of crosslinking compared to the rate of configurational relaxation,
and on the concentration of primary molecules in the crosslinking medium
(263). Other authors assume that network strands are incorporated with equilibrium configurations and maintain the same configurations throughout the
crosslinking process, (r2>/(r2>o being therefore unity for networks formed and
tested under the same set of conditions.
A concomitant of coil shrinkage would seem to be a macroscopic shrinkage of the gel structure, called syneresis, if the crosslinking were carried out in
solution. Syneresis is indeed observed at the later stages of crosslinking in highly
diluted systems [see Refs. (I 7) and (257) for examples]. Some evidence of localized
network inhomogeneities has also been adduced in polycondensation networks
(264) and in solution-cured systems (265), which have been attributed to a
localized syneresis. To this reviewer at least, appreciable coil shrinkage during
random crosslinking of preformed primary chains in concentrated solutions or
melts seems unlikely. It will be assumed that (r2>/(r2>o is unity in the lightly
crosslinked, as-formed networks discussed here.
Most testing of the theory has been done with tensile deformations. According to Eqs. (7.1) and (7.2) the nominal tensile stress f , the tensile force per
unit unstretched area, is related to ~, the ratio of stretched to unstretched length, by:
. ~ <r~> ( ~ _
s=o - 3Tt
7)
1
(7.3)
s Note that this definition of front factor differs from that given earlier by Tobolsky
(162), since the term (r2)/(r2>o has been separated out.
103
Some aspects of the theory have been examined in detail. It is well

established that the major part of the free energy change for small isothermal
deformations is entropic. The small internal energy contribution (,,~ 10% in the
case of natural rubber, for example) is well accounted for by the temperature
coefficient of (r2). Values of d l n ( r 2 ) / d T from elasticity measurements are in
generally good agreement with values obtained in dilute solution (267), showing
that intermolecular energy contributions must be rather small, at least in small
deformations.
Finite extensibility of the network chains and volume changes induced by
deformation also can influence the response. Quite understandably, the modulus
increases as the chains become fully extended, but in networks with long network
strands the upturn occurs only at rather large deformations (~ > 2). Small changes
in volume most certainly do occur in tensile deformation since elastomers have
roughly the same isothermal compressibility as small molecule liquids. However,
Poisson's ratio is usually very close to 0.5, and, except for large deformations and
in analyses to determine the internal energy contribution, the effect of volume
changes is probably quite small (266).
Even when the above complications are negligible or properly accounted for
and when strain-induced crystallization is absent, the stress-strain curves for
networks seldom conform to Eq. (7.3). The ratio f/(a - 1/a z) generally decreases
with elongation. An empirical extension of Eq. (7.1), the Mooney-Rivlin equation,
has been used extensively to correlate experimental results:
pij = c 1 ciy 1 - c2 ci~ - eo ,~j
(7.4)
in which C1 and C2 are experimentally determined constants 6, typically of

similar magnitude. For tensile deformations the stress based on initial area is
(7.5)
Even this form is probably only an approximation (267). More complicated

expressions appear to be necessary to fit data on the same sample for different
types of deformation (268). For sufficiently small deformations the tensile modulus
becomes
(7.6)
6 Many authors use 2Cl and 2C2, reflecting a definition of the constants in terms of the
elastic energy function. The factor of 2 is of course arbitrary and irrelevant to the discussion
here.
104
W.W. Graesstey:
Earlier the C2 term was thought to reflect merely a tack of thermodynamic

equilibrium in the measurements. Indeed, it is extremely difficult to achieve
equilibrium in unswelled rubber networks (269). Although the magnitude of C2
does appear to be sensitive to non-equilibrium effects and is reduced as
equilibrium is approached, it now seems clear that many systems display finite
C2values even at equilibrium (I 7). The temperature dependences of C1 and (2~ are
quite similar (267, 270), suggesting that both have essentially entropic origins.
On the other hand, diluents influence C1 and C2 rather differently. The C2 term
generally decreases rapidly with swelling. When compared at constant swelling
ratio, the values of C2 obtained are remarkably insensitive to the thermodynamic
nature of the solvent (271). The magnitude of C2 does depend on the type of
curing system used however (267). On the other hand, C1 changes in accord
with the predictions of kinetic theory. Thus, its value depends mainly on the
crosslink density and is altered by swelling in the manner required by kinetic
theory:
C1 ~ v ~r~)~- 0c
\-i/7/
=01/3
(7.7)
in which V0 is the volume of the sample when the crosslinks were introduced,
V is the swollen volume, and ~o is the volume fraction of polymer in the
swollen state Vo/V. In a recent study of peroxide-cured natural rubber (270)
the quantity C 1q~- 1/3 was independent of swelling ratio, while C z q~- 1/3, which
was approximately 0.4C 1 in the dry state, decreased linearly with q~ and
extrapolated to zero for q~~ 0.2. Sulfur cured vulcanizates of natural rubber and
various styrene-butadiene rubbers behaved in an almost identical manner (271).
A continuum explanation has been advanced for the rapid reduction of C2
with swelling (272), but Ftory and Mark (273) have pointed out the ambiguities
inherent in such an approach.
For the same polymer and curing procedure, networks formed in the
presence of various diluents (~o~ 0 . 1 5 - 0.40) give practically no Cz contribution
when tests are made in the dry state (274). Kinetic theory requires Cl(P -2/3
to be constant in this case (~o being the volume fraction of polymer during
solution cure), assuming comparable numbers of crosslinks per primary molecule
are introduced. Experimentally C 1 ~o- 2/3 is close to the values of C1 obtained for
dry state cures, although they do decrease somewhat in the more dilute curing
systems (q~< 0.2). Solution cured polydimethyl sitoxane networks show similar
behavior in their dry state moduli (275).
Gent and Rivlin (276) have suggested that C1 is in fact the kinetic theory
term, related directly to the concentration of effective network strands bounded
by chemical crosslinks. The molecular origin of the C z term is unknown,
although many possibilities have been suggested (17, 267, 2-77). The above
experiments seem to rule out both isolated chain explanations, such as intramolecular excluded volume, and suppositions based on local ordering of chains.
If either were important, it should logically contribute in dry state measurements,
regardless of whether the network had been formed originally in the dry state or
in solution. Non-Gaussian chain behavior seems an unlikely explanation also,
105
since C2 apparently is small in both the extended (swelled) and supercoiled (drystate after solution cure) states of the network strands.
Various authors have suggested that the source of deviations from neoHookean behavior lies in the large scale organization of the network itself
(257, 267, 274). An origin in network topology would explain both the sensitivity
of C2 to the conditions during network formation and the observations that
C2, like Cp is largely of entropic origin. Alfrey and Lloyd (252) have emphasized
that the conventional kinetic theory deals with an assembly of strands whose
configurations are statistically independent, in effect with "ghost" chains which
can pass freely through the backbone contours of neighboring chains. Chains in
real networks must certainly be subject to topological restrictions, and the configurations of neighboring chains must therefore be conditionally dependent. If
such dependency affects the number of configurations differently in the
deformed and undeformed states, then the entropy of deformation must contain
an extra contribution. This extra contribution, which one might call the topological
contribution to distinguish it from the independent strand term of the
conventional theory, could produce a response different than that of a neoHookean solid. In particular, it might contribute the C2 term and some portion
of C I.
If this view is correct, the measurements of the equilibrium modulus can
provide some information on network topology, provided the independent strand
contribution can be evaluated separately. Principal attention should focus on the
modulus for small deformations. Since the structure is essentially undisturbed
by such measurements, the small deformatibn modulus should contain the most
direct molecular information. Proceeding beyond this to an interpretation of large
deformation behavior would be expected to present a comparable increase in difficulty to that encountered in moving from molecular interpretation of q0 to molecular interpretation of q vs. 37behavior in melts and concentrated solutions. As a
corollary, the mechanical behavior of the network should be evaluated in the same
state of swell as that in which it was formed. Swelling or supercoiling subsequent
to network formation itself involves a finite deformation of the network and
under these circumstances even small deformation moduli may not be easily
interpretable in terms of topological contributions. Indeed, assuming that the term
represented by C2 contains an important part of the topological information, the
unexplained changes in C 2 with swelling or supercoiling suggest a relative change
in the independent strand and topological contributions. Of course, a complete
theory of topological effects must ultimately be able to account for swelling
and large deformation behavior, but until then, interpretation seems necessarily
limited to small deformation behavior in networks at the same state of swell as
when the crosslinks were added. If this is the undiluted state, as is many times the
case, then there is the corresponding, extremely difficult, experimental problem of
achieving stress equilibrium in the modulus measurement.
7.2. Tests of Theoretical Modulus Values--Model Networks
Quantitative tests of the absolute modulus value from kinetic theory [Eq.
(7.2)'1 are quite scarce; independent determination of the network strand
concentration is generally the difficulty. Schaefgen and Flory (278) prepared model
106
W.W. Graessley:
polyamide networks by coupling pre-formed star branched molecules. The

modulus was accurately proportional to the concentration of network strands,
but the values exceeded the theoretical moduli (calculated as G=vkT) by
approximately a factor of 2, implying 9 = 2 for these networks. The molecular
weights of the strands (8000-20000) probably exceed Me for polyamides to
some extent, so 9 may be too high because of additive entanglement contributions.
In two other studies (279, 280) model networks were formed by the endcoupling of pre-formed linear chains with multifunctional coupling agents. Strand
molecular weights in both cases were of the order of Me for the non-crosslinked
system; topological contributions to the modulus might be expected to be smaller
in these networks than in systems with longer strands, although not necessarily
absent. Kraus and Moczygemba (279) coupled carboxy-terminated polybutadiene
chains (M = 5500) to form a network in the dry state. With an excess of coupling
agent they observed modulus values approaching C~ = 4.8 x 106 dyne/cm2 and
C2 ~ 0 at 25 C. The value of vkT for this composition is 4.1 x 10 6, implying
9 = 1.2. The authors suggested that entanglement effects should be negligible in
these networks because the molecular weight of the strands is smaller than
M (M = 5900 for 1,4 polybutadiene). However, M is only 1900 for this polymer,
so it is not clear that entanglement contributions are absent.
Prins and co-workers (280) reported moduli on polystyrene networks which
had been formed by adding small amounts of divinyl benzene to preformed
polystyryl dianions. The gels were dried, reswelted to equilibrium in benzene and
tested in simple compression. The equation for G in this case becomes:
G = 9v*kT ~00
z/3 (~01/3
(7.8)
in which v* is the network chain concentration in the dry state, (% is the volume
fraction of polymer during network formation, and q~ is the volume fraction
during modulus measurement. These molecular weights are indeed much smaller
than (Me)so~n'for polystyrene at the curing and swelled-state concentrations. Gels
were prepared with network strand molecular weights from 7200 to 18700.
Deduced values of g ranged from 0.4 to 0.55; no information on C2 was
reported.
Allen and co-workers (280a) have recently prepared and studied a series of
model polystyrene networks. These networks were formed from narrow distribution polystyrenes (M = 70000-240000) containing small but known amounts of
secondary amine side groups. The amines were coupled in solution by reaction
with diisocyanates, resulting in networks with known values of v
(M* =7 000-17 000). Moduli were determined for the networks at the same diluent
concentrations as used to form the networks (~0= 0.05-0.25). After applying
approximate corrections for intramolecular (sterile) crosslinks and trapped
entanglement contributions, the authors concluded that a value g = 0.5 would
satisfy most of the data. Recent theoretical calculations by Tonelli and Helfand
(280b) suggest that sterile loops may play some role even for networks formed
107
in the undiluted state. On the other hand, when gel-sol studies are used to
evaluate crosslinking densities (see below), sterile loops remain uncounted, so there
is at least partial compensation for their reduced effectiveness in equilibrium
mechanical properties.
7.3. Tests of Theoretical Modulus Values--Statistical Networks

Networks formed by random multifunctional condensations or random crosslinking can be characterized by a combination of gelation studies and statistical
calculations of structure, based on the equal reactivity principle (281). There is
still some debate, however, about the correct counting procedure for calculating
the concentration of elastically effective network strands. Flory (3) noted that, in
a gel composed of N primary chains and C tetrafunctional crosslinks, N - 1 of
the crosslinks are required merely to join the primary chains together. Each
crosslink beyond this number links two parts of the structure together,
creating in the process two additional closed loops. According to Flory, each
such closed loop corresponds to an elastically effective strand. Thus, the total
number of active strands is
Flory criterion:
N A = 2 ( C - N + 1) ~ 2 ( C - N).
(7.9)
Alternatively, N A is the number of mers participating in crosslinks (2C) minus

the number of chain ends (2N). Tobolsky (t62) has criticized the Flory criterion,
showing that Eq. (7.9) is incorrect near the gel point. The Flory criterion has
been used in most of the network analyses, however, and for purposes of
comparison we will carry it along.
Scanlan has suggested another criterion (282). An effective network junction
point is a crosslink in which at least three of the four strands radiating from it
lead independently to the network. A crosslink with only two strands anchored
to the network simply continues an active strand; a crosslink with only one
anchored strand is part of a dangling end and can make no elastic contribution
at equilibrium. An elastically effective strand is therefore one which joins two
effective network junction points. Accordingly, the total number of active strands
is simply one half the number of get-anchored strands radiating from effective
junction points:
Scanlan criterion:
C
N a = ~-(3p 3 + 4p4)
(7.10)
in which P3 and P4 are the fractions of crosslinks with 3 and 4 gel-anchored

strands.
The two criteria give somewhat different results for the effective strand
concentration. Consider, for example, the random addition of C crosslinks to
108
W.W. Graessley:
vo monodisperse primary chains per unit volume (primary chain length = P

repeating units). The gel fraction Wg is related to crosslinking index 7 through
the equation (valid for P >>1) (27, 33):
1 - wg = e -~'w~
(7.11)
in which Vis 2 C/vo, v o is Q/M, ~ is the polymer density and M is the primary chain
molecular weight mo P. The crosslinking index of the gel is higher than the overall
crosslinking index (27):
~'g = (2 - Wg) ~,.
(7.12)
A simple calculation of the concentration of active strands, following Langley

(255), yields
Flory criterion:
v~ = v o (2 - Wg)In 1 - 1 wg
2Wg].
(7.13)
Scanlan's criterion has been used extensively by Gordon and co-workers. The
resulting equation for monodisperse chains, again for P >>1, is (283):
Scanlan criterion:
v~ -- Vo (2 - Wg)In 1 - w~
2wg
1
In 1 -w~-~
Wg "
(7.14)
Table 7.1 compares v J r o according to the two criteria. Values for primary
chains with exponential distribution ( M w / M ~ = 2) are also included. The number
of active strands according to the Scantan criterion is larger by a factor of 1.5
at the get point. The ratio decreases slowly with increasing 7 and finally approaches
a value of unity for large numbers of crosslinks per primary chain. Both
expressions approach the form v~ = %7 for sufficiently large values of 7The Flory criterion requires that each crosslink between units which are
already part of the get must increase the number of active strands by 2.
Examination of example structures shows that this number according to Scantan's criterion is either 1, 2, or 3, depending on the structures involved. The
difference between criteria is therefore a real one in physical terms. In the author's
opinion, the Scanlan criterion is clearly the correct one. Any other choice would
seem to depend ultimately on such questions as whether junctions with three
anchors in the gel are equivalent to 4-anchor junctions insofar as the assumption
of affine displacement is concerned. To the extent that the strands behave as linear
109
Table 7.1. Concentrations of elasticallyeffectivestrands according to the Flory and Scanlan

criteria for random crosslinkingof monodisperse primary chains
7
wg
vc/v o (Flory)
vc(Scanlan)/v~(Flory)
1
1.5
2.0
3.0
7.0
0 (gel point)
0.59
0.80
0.94
0.999
1
0
0.08
0.33
1.2
5.0
~
3/2
1.42
1.37
1.29
1.15
l
(3/2)a
(1.40)
(1.31)
(1.19)
(1.03)
1
a Values in parentheses were obtained at the same gel fraction wg by calculations for
primary chains with an exponentialdistribution of lengths.
springs (conform to the Gaussian distribution) junctions of any functionality

greater than two would seem to be equivalent.
Tobolsky and co-workers (283a) have determined 9 for a large number of
networks prepared by copolymerizing various acrylate and methacrylate monomers with small amounts of a tetrafunctional methacrylate monomer. Essentially
complete conversion was achieved in all cases; the amount of extractibles was
found to be negligible. Two effective strands were assumed to result from each
tetrafunctional mer in the system. The calculated molecular weights between
junction points ranged from 12000 to 50000 (v~ ~ 2 - 8 x 10- 5 moles/cc). This is
probably below Me for the chains with large pendant ester groups, but comparable
to or above that for the methyl and ethyl esters. Values of g from tensile moduli
varied systematically from 1.5-2.0 for the methyl and ethyl ester chains (consistent
with some entanglement contribution) to 0.24).3 for chains with bulky pendant
groups. The authors speculated that crowding at the junction points may have
influenced network structure in the latter case, resulting in fewer than the
theoretical number of effective strands and hence in g values which are too low.
It is difficult to draw clear conclusions from these results, aside from noting that
O~ 0.54).7 appears to satisfy most of the data from chains in which the pendant
groups are not too bulky, but for which the strand lengths are shorter than the
entanglement spacings.
Gordon et al. (284) have measured moduli on networks formed by condensing decamethylene glycol and benzene-1, 3, 5 triacetic acid. The concentration of active strands in each sample was calculated from gel point data
and the statistical theory of structure appropriate to this system, using Scanlan's
criterion. They found rather good agreement between the initial modulus and
that calculated using Eq.(7.2) with g = 1 for networks just beyond the gel point.
In this region the gel is quite open and theoretically (see following section) the
number of trapped entanglement is small relative to the number of strands
bounded by chemical crosslinks. At higher extents of reaction the moduti became
systematically higher than values predicted similarly.
To the author's knowledge no other evaluations of the g factor have been
made on statistical networks under circumstances in which the contribution of
trapped entanglements would be expected to be negligible. Aside from Gordon's
110
W.W. Graessley:
result of O ~ 1, the data on both model (280, 280a) and statistical (283a)networks
suggest very strongly that 9 < 1, and in fact that 9 probably lies in the vicinity
of 0.5 for many systems. Further estimates of 9 have been made in situations
where entanglements are important but separable, at least theoretically, in their
effects on the initial modulus (see following Section 7.4).
7.4. Determination of Entanglement Density in Networks

7.4. t. Theories of Entanglement Trapping
Mullins (253) and Bueche (254) suggest that entanglements which are trapped
in the system during crosslinking will simply add to the number of active strands
in the rubber elasticity modulus. An entanglement site on a primary chain is
considered to be trapped only if it occurs between two crosslinking sites on that
chain. For a network formed by adding C crosslinks to N primary chains of
(number-average) degree of polymerization P, the number of crosslinked units
in the system is 2 C. The probability that a repeating unit selected at random is
contained on a strand bounded at both ends by crosslinks (C ~>N)is "( 2C
k
2C+N
")~.
/
With C large compared to N, the gel fraction is practically unity, and for the network, 2 - ~ N
EN entanglement sites, the number

EN(1- N/C). With Flory's criterion, the total
~ 1 - -C" From a total of
which are trapped is therefore

number of active network strands according to Bueche and Mullins is
NT= 2C- 2N + EN (1- C).
(7.15)
When cast in terms of strand concentrations, the Bueche-Mullins equation

becomes
v=(v+ve)(1-2VO)v~
(7.16)
in which vc is v07, the concentration of repeating units which participate in

crosslinks. The primary chain concentration Vois Q/M,in which M is the (numberaverage) molecular weight of the primary chains. Finally ve is Q/Me,Me being the
molecular weight between entanglement points, a property which is assumed to
be independent of both primary chain length and crosslink density. The BuecheMullins trapping factor is therefore (1 - 2Vo/V).
Equation (7.16) rearranges to give
v=v+ve-(1 +vdvo) 2e
M
(7.17)
The EntanglementConcept in PolymerRheology
111
The active strand concentration v is obtained from measurements of the initial

modulus using Eq.(7.2) with g = 1. Values of v for samples at the same crosslink
density but with different primary molecular weights are extrapolated to I / M = O,
giving v + ve. Values of vc + ve obtained at two or more crosslink densities
provide ve, and hence Me, by extrapolation. An advantage of this method is that
only relative values of crosslink density must be known: absolute values of vc
are not required. If absolute values of v are known, the g factor can be
evaluated as well.
An intuitively more satisfying requirement for trapping was suggested by
Ferry and co-workers (285). In order for any entanglement to contribute to the
equilibrium modulus, all four strands leading from that entanglement must
terminate in crosslinks. To the same approximation as the Bueche-Mullins
relation (C>> N), and with the Flory active strand criterion, the strand concentration becomes
(7.18)
v = v,(1 - 2Vo/V,) + ve(1 - 2Vo/Vc) z .
The Ferry trapping factor is therefore ( 1 - 2Vo/V) 2. Unfortunately, the simple

extrapolation procedure of Kraus (286) can no longer be applied, but
numerical search techniques should still suffice to determine Me.
~~
~"
Bueche-Mtutrap
lins Ferrytrap
Gele~
~
G
Gel
Gel.
Langletrap
y
Both the Bueche-Mullins and the Ferry relations are attempts to allow forlthe
effect of finite primary chain length on active strand and entanglement strand
concentrations in well developed networks. In applying such relations to
determine Me in networks, it is dearly desirable both to begin with very long
primary chains and to measure moduli on networks which have several crosslinked units per primary chain. Thus, if M ~>M* >>Me, where M* is the average
molecular weight between crosslink junctions, the trapping factor is essentially
unity, and v ~ v, + ve ~ ve. The uncertainties due to the chain end correction to
the use of the Flory criterion for active strands, and to independently estimated
values of v~ (or extrapolation to v~= 0). will be small under these conditions.
Langley (255) has developed a refined definition of the Ferry trapping factor,
and he also accounts theoretically for effects of the primary chain distribution and
112
W.W. Graessley:
concomitant random chain scission during the crosslinking process. According

to Langley, an entangling interaction between chains contributes to the equilibrium modulus only if all four strands leading away from the entanglement
junction have their own separate connections to the network. It is assumed that
all repeating units in the system have the same probability of participating in an
entanglement. The fraction of trapped entanglement couples Te is then p2,
where p is the probability that each of the two strands leading away from a
randomly selected repeating unit in the system is anchored separately to the
network. For example, in a randomly crosslinked system of monodisperse primary
chains, the trapping factor is (255):
re=
2-wg-
2w~
[z
(7.19)
Langley employs the Flory criterion to evaluate the number of active strands
bounded by chemical crosslinks. The total concentration of active strands,
including strands bounded by trapped entanglement junctions, can be expressed (255)
(7.20)
v = vo~,wgT2/2 + veT.
Equation (7.20) is general, although the expression for Te itself depends on the
primary distribution, the gel fraction, and the relative rates of random crosslinking and chain scission.
With the Gordon equations (273) one can easily extend Langley's analysis to
the Scanlan criterion for active strands bounded by chemical crosslinks. With
the same generality as that of Eq.(7.20):
v= -2-v~' T2/2(3w~ _ T~/2) + voTo
(7.21)
For either Eq.(7.20) or (7.21) the trapping factor Te can be calculated from a
rearranged and simplified version of Eq. (17)in Ref. (255):
Te= 2 x - w g -
2x (1
YPn - F o
(7.22)
in which
Fo = ~ N(P) e -ye.
P=I
(7.23)
113
The initial number-average degree of polymerization of the primary chains is

Pn, N(P) is the fraction of primary chains with P repeating units, and x and y
have the meanings given by Langley (255):
x = qwg/(qwg + p),
(7.24)
y = qwg + p
(7.25)
where q and p are the respective fractions of repeating units which participate
directly in crosslinks and which have undergone scission.
Application of Eq. (7.20) or (7.21)to modulus measurements provides both
a value of M~ and a value of the front factor g. For example, the combination of
Eqs.(7.20) and (7.2) leads to:
~ ]
~-~ =
T~/2
~- Me
(7.26)
in which v0 is expressed in moles/volume. The modulus G for a series of

samples crosslinked to different extents is obtained (G= f / ( ~ - 1/0~2) at small
deformations), and the quantities y, wg, and T~ are evaluated for each network
from the gelation behavior of the system and its primary chain distribution. The
quantity (r2)/(r2)o is unity if the modulus is measured at the same state of
swell as is present during crosslinking. Thus g and Me are separately determinable
from the slope and intercept of a plot of G/T~ vs ywg/T~/2. Application of Eq.
(7.2L) proceeds similarly.
It should perhaps be emphasized that the trapping relations themselves can
be stated in more general terms than those implied by interpenetrating loop ideas.
The Langley relation is simply the fraction of intermolecular contacts between
mers in which both are parts of active strands after crosslinking. It is assumed
to measure the fraction of pair-wise topological arrangements in the system
which are rendered permanent when crosslinks are added, and therefore the
fraction of the total potential topological contribution which joins with the
chemical strand contribution to determine the equilibrium modulus. Viewed
thusly, the trapping analysis depends not at all on specific models of the
entanglement interaction. The values of Me deduced from small deformation
network elasticity have the same meaning in principle as those from the plateau
modulus: when described in terms of localized interactions, the topological
contribution is equivalent to that of a permanent network with crosslink spacing
Mo. Comparisons between values of M~ from network elasticity and from the
plateau modulus of uncrosslinked polymers provides therefore a clear and modelfree test of the topological entanglement concept.
114
W.W. Graessley:
7.4.2. Experimental Results
The Bueche-Mullins method has been applied in the separation of the

modulus contributions of crosslinks and entanglements in several elastomers.
A front factor of g = 1 was then used to determine M~. The Langley method has
also been applied in a few cases, resulting in values of both g and M~. These works
are summarized below; results are collected in Table 7.2.
Kraus (286) obtained M e = 6000 for peroxide-cured 1,4 polybutadiene networks
formed from samples of different primary molecular weights. Moduli were
determined indirectly from swelling behavior, previously correlated, however, with
elastic properties. Chain scission was assumed to be absent and the chemical
crosslinking densities were obtained indirectly. Van der Hoff and Buchler (277)
found Me = 4300 for 1,4 polybutadiene networks which were cured to different
extents by a quantitative crosslinking agent at room temperature. Both studies
used the Bueche-Mullins method. Pearson et al. (287) have applied Langley's
method [Eq.(7.26)] in the analysis of radiation-crosslinked 1,4polybutadiene,
obtaining g = 1.22 and Me = 5900 from gelation data and moduli of networks
which were crosslinked dry and subsequently swollen in toluene.
Vander Hoff and Buchler (277) found Mr= 13000 for natural rubber by
applying the Bueche-Mullins analysis to the data of Bristow (288). Examination
of the natural rubber results published some years ago by Flory et al. (289)
yields M ~ 15000. The data of Mullins (290) provides a value of 10000 for
1,4polyisoprene if the total initial modulus is used in the analysis (Mullins
reported Me = 16000 by examining the effects of primary molecular weight and
extent of cure on C1 alone; the initial modulus G is Ca + C2 and C2 ,~ 0.6 C1 in the
samples examined).
Bueche and co-workers (253, 291) obtained Me= 13000 in radiation-cured
poly(dimethyl siloxane) networks. They also determined absolute values of the
chemical crosslink density, using molecular weight measurements in the pre-gel
region to establish a crosslink density vs dose calibration. The front factor for the
chemical crosslink contribution was found to be g =0.65 (291). Langley and
Polmanteer (292) have recently applied Langley's method to three well characterized samples of poly(dimethyl siloxane). Gel fractions and moduli were
measured in radiation-cured networks. The front factor g was found to be
0.81, and the molecular weight between entanglement points (using g=0.81)
was Me = 8600 (Fig. 7.1). These workers have also analyzed the same experimental
data using the Scanlan active strand criterion [Eq.(7.21)], finding in this case
Me = 5600 and g = 0.60 (293). In an earlier study of poly(dimethyl siloxane)
Langley and Ferry (294) found that the fraction of entanglement strands which
are trapped in the structure is given with good accuracy by the calculated
trapping factor To. The number of active strands was determined by both modulus
and swelling measurements (the latter calibrated, however, in terms of modulus
values), and then calculated theoretically from gelation data and the primary
chain distribution, using the Langley equations.
Ferry and co-workers (295) have obtained a value Me = 13 000 for 1,2 polybutadiene by a new method which involves crosslinking the polymer in a strained
state. The initial extension of the uncrosslinked sample, the recovery after
0.65
0.81
0.60
13000
15 000 (est.)
10000
13000
8600
5600
5800
8100
3550
Polydimethyl siloxane
1,2 Polybutadiene
4100
1.22
1,4 Polyisoprene
and natural rubber
6000
4300
5900
1900
Value of
g factor
obtained
1,4 Polybutadiene
M, from
equilibrium
network modulus
M, from the
viscoelastic
plateau
Polymer
Bueche-Mullins
Langley
(Flory criterion)
Langley
(Scanlan criterion)
Kramer-Ferry
Bueche-Mullins
Buecbe-Mullins
Bueche-Mullins
Bueche-Mullins
Bueche-Mullins
Langley
(Flory criterion)
Method
Ferry and co-workers (295, 296)
Langley and Polmanteer (292)
Bueche and co-workers (254, 291)

Langley and Polmanteer (292)
Vander Hoff and Buchler (277)

Data of Flory et al. (289)
Data of Mullins (290)
Kraus (286)
Vander Hoff and Buchler (277)
Pearson et al. (287)
Investigators
Table 7.2. Values for the molecular weight between entanglements M~ from the equilibrium properties of crosslinked networks
O
O
g~
(3
u~
(IQ
'D
116
W.W. Graessley:
30(
2O(
r,,.
n
w
,,,.J
0
I-~' I o o
,6o
uo ) ' w g / T e It2
'
MOLE
300
26o
PER
M3
Fig. 7.1. Plot of modulus data on poly(dimethyl siloxane) networks to determine g and Me
by Langley's method [Eq.(7.26)] (292)
crosslinking, and the modulus of the polymer in its final rest state were measured,
and the entanglement density in the original state was calculated using Flory's
theory of composite Gaussian networks (223). The above value appears to have
been somewhat high because of entanglement slippage during the initial
deformation. Recent results, obtained with lower temperatures during deformation
and crosslinking, yield Me=4100 (296), in good agreement with the plateau
modulus value [Me = 3550 (15)].
Baldwin and Ver Strate (297) employed expressions equivalent to Eqs.(7.20)
and (7.21) to analyze modulus data on ethylene-propylene ethylidene norbornene
terpolymers swelled with a solvent. They fixed the crosslinking density by
postulating complete utilization of the peroxide curative added, this justified on
the basis of considerable ancillary work. The deduced molecular weight between
entanglements (assuming g = 1) on networks swollen to equilibrium was found
to depend on the chemical crosslink density, with values ranging from Me = 2300
at low crosslink densities to 1400 at higher values. The order of magnitude of
Me is probably correct. However, the systematic increase in Me with crosslinking
density may have been caused by the fact that the moduli were measured on
networks which swell to different extents at equlibrium. As pointed out earlier,
entanglement contributions to the modulus might be expected to decrease with
increasing swelling ratio, which would produce just the trend observed.
117
Table 7.2 shows that the reported values of Me from equilibrium moduli of
networks are consistently larger than those obtained from the plateau modulus
using 0 = 1. Superficially the results imply that the entanglement density
affecting equilibrium properties is somewhat less than that affecting dynamic
properties such as the plateau modulus. However, the choice of active strand
and trapping criteria is dearly important in the numerical value of Me and 9
obtained. In particular, Me deduced by the Bueche-Mullins method would be
expected to be systematically larger than that from the more accurate Langley
method because the former overestimates the number of trapped entanglements.
It may be significant that the results of Langley and Polmanteer (292) and
Langley and Ferry (294) with poly(dimethyl siloxane) are the most precise in
terms of characterization of primary chains and gelation behavior, and that both
provide equilibrium values of Me which agree best with values from the plateau
modulus. The analysis of 1,4 polybutadiene by Pearson et at. (287), is comparable
in precision and yields Me somewhat larger than that from the plateau modulus.
However, the mechanical properties in this latter study were determined in the
swollen state. By inference from the sensitivity of C2 to swelling noted earlier, the
effects of network topology on elasticity are reduced in swollen systems. The
Pearson study might therefore be expected to produce values of Me which are
too large.
It is clear that the application of Langley's method in other polymer systems
is essential to settle questions about Me and 9 in networks satisfactorily. The
Ferry composite network method (223, 296) appears to be broadly applicable as
well, although requiring special care to minimize slippage prior to introduction of
the permanent crosslinks. (One is also still faced with the difficult question of
whether 9 is the same for entanglements in crosslinked networks and in the
plateau region of dynamic response.) Based on the limited results of these two
methods in unsweUed systems, Me values deduced by equilibrium and dynamic
response appear to be practically the same.
7.5. Relaxation Processes in Networks
Stress relaxation in networks formed by random crosslinking tends to be

spread over many decades in the time scale. Ferry (15) attributes the slow
relaxation processes to dangling structures, composed perhaps of several primary
molecules which are linked to the main body of the network in one or only a few
places. The statistical theory of crosslinking predicts the existence of such
structures. They should be most prevalent in the vicinity of the gel point, but some
will be present even when there is an average of many cross-links per molecule.
Relaxation is much more rapid in model networks which have no dangling
structures and in networks swollen by solvent. Quantitative interpretation of
these slow relaxations awaits a dearer picture of relaxation in branched skeletal
structures. Recent work by Cohen and Tschoegl (298) on block copolymers with
dangling ends suggests that such materials may provide models for the systematic
study of slow relaxations in networks.
118
W.W. Graessley:
Ferry and co-workers (299) have recently reported linear viscoelastic studies
on polyisobutylene networks containing linear polyisobutylene molecules which
interpenetrate the network but are not chemically bound to it. Entanglement and
crosslink spacings are roughly the same. In this case the unbound molecules
relax in the network environment, in a manner presumed to be roughly
comparable to the reptation model of De Gennes (225). One surprising aspect
of this work is that the relaxation spectrum of the linear chain-network
composite is roughly independent of the concentration of unbound chains, up to
a weight fraction of 0.50. One might hope that in the future such systems will
provide insight on the large-scale motions of individual chains in entangled or
permanently connected structures.
7.6. Theoretical Analysis of Topological Constraints

7.6.1. Effects of Topological Classification
Edwards (300) considers the effect of entanglement in permanent networks to
arise from the topological classification imposed on the strands when the network
is formed. In the conventional theory of rubber elasticity, the network strands
are free to choose configurations without regard to the configurations of
neighboring strands. In real networks the strands cannot pass through one
another's backbone.contours. Thus, the configurations available to any strand
depend on its topological relationship with neighboring strands; certain
configurations become unavailable to each class which would be available to
independent strands. These additional constraints have no effect on the configurational entropy of the network in its as-formed state, since the cross-links
simply bind the chains permanently into what is already an equilibrium
distribution of configurations. If the network is deformed or swelled, however,
the additional constraints become important, as shown below, and the entropy
change will differ from that for a network of independent strands.
The simplest example of topological classification is the knotting of individual
strands, i.e., self-entanglement. However, the effects attributed to entanglement in
non-crosslinked polymers are clearly intermolecular. The simplest such case is
that of pair-wise classification without self-entanglement. Consider the following
three examples of strand pairs:
The Entanglement Concept in Polymer Rheoiogy
119
The equilibrium positions of the four junction points define the spatial
relationship between the strands of each pair. For simplicity, the mean relative
positions (internal coordinates) of the junction points of each pair are taken to
be the same. The junction points of each strand are separately anchored to the
network by at least two of their remaining strands, so each is an elastically
effective strand according to Scanlan's criterion. The network itself in effect
completes the loop for each strand, making the A, B, and C pairs as structurally
distinct as catenane molecules (301).
Each pair is one example of a myriad of permanent topological classes for
strand pairs whose spatial positions, described by the internal junction
coordinates only, are the same, but whose structures cannot be transformed from
one to the other without chain breakage. Members of each class have
configurations available which are not available to members of other classes.
Moreover, every distinguishable configuration of a strand pair must belong to
only one such class. Considering pair-wise relationships only, the same topological
classes will be available for any pair of strands in the network. However, the
probability (relative population) of each class will depend strongly on the internal
coordinates of their junction points, approaching a probability of unity for the
unlooped class in widely separated pairs. One would expect that the apportionment among classes for pairs with the same internal coordinates would depend on
the latent, i.e., equilibrium, topological arrangement of the primary chains at the
moment of crosslinking.
If the network is deformed, the mean internal coordinates of each strand
pair assume new values, and the number of configurations available to pairs
of each class changes. The number of configurations available to each class with
the new internal coordinates will be proportional to the population of that class
for pairs which form at equilibrium with this set of internal coordinates. The
change in entropy with deformation is calculated from the changes in the
number of distinguishable configurations of the strands, taking into account the
permanence of topological classification, through the Boltzmann equation. The
contribution of entanglements to the modulus is then obtainable from the difference between this value and the entropy of deformation for independent strands.
Edwards (300) was able to make some progress on the problem by showing
that topological classification can serve only to raise the modulus. As a simple
example, consider all strand pairs in a network which have, within some small
tolerance, a specified set of internal junction coordinates. Suppose there are B
such pairs, and that the strands of each pair, labeled 1 and 2, have 0)1 and co2
distinguishable configurations each as free strands, and fractions (O0o and (92)0
respectively which have the end-to-end distances specified by the equilibrium
junction coordinates. If the crosslinks were formed in the system at equilibrium,
then the total number of configurations for each strand of the pair is 0)1(90o
and 0)2(02)0 , and the number available to the pair is c010)2(O00(Oz)o.
Suppose that (f~)o is the fraction of the B pairs in topological class i from a
total of Q such classes. The number of configurations available to a pair in class
i will be 0)10)2(01)0(92)0(fi)0, and the total number of configurations in the
undeformed state for B pairs is
Q
t2o = 1-I [c1 ~2(91)o(g2)o(fi)o] t'r~)B"

i=1
(7.27)
120
W.W. Graessley:
The pairs will all have the same new internal coordinates in the deformed
state. Suppose that the fractions of configurations for free strands which have the
specified end-to-end distances of the deformed state are gl and g2, and that the
fractions in the various classes for pairs formed from equilibrium with these
coordinates are f~, The number of configurations available to a pair in class i
in the deformed state is therefore ~01092g lg2f~. However, the number of strands
in each class is always (fi)oB, so the total number of configurations in the
deformed state is:
12
Q= 1-I [l9zglgzfJ ~s'~B"
(7.28)
i=1
The entropy change contributed by the B pairs is given by the Boltzmann

equation:
AS = k In
f2
f20
(7.29)
which yields [using E(f~)0 = Ef~ = 1]
AS=kB[In gl +ln 92
(O0o
f~ ]
(02)0 + ,=~2(f~)o l n ~ o -
(7.30)
The first two terms on the right of Eq. (7.30) will be recognized as the
independent strand contribution to the entropy. The topological or entanglement
contribution is then
Q
f~
ASe = kB ~, (f~)o In - i= 1
( f i)o
(7.31)
The total pair-wise entanglement contribution is simply the sum of ASe for strand
pairs with all possible sets of internal coordinates. In a network of N strands
there will be N(N- 1)/2 such pairs. However, classification is only significant
for pairs that are relatively close. Pairs separated by more than a few radii of
gyration will belong to unentangled class exclusively. This assures that the total
entropy change will be proportional only to N.
Equation (7.31) is a familiar form in statistical mechanics. For small
displacements it is negative in sign and proportional to the square of displacement.
For tensile deformation a power series expansion in the elongation ratio cq and with
0 Efi = 0, yields
ASe= kB(o~-1)2 ~., 1 {

2
Of, t 2 + 0 ( ( ~ - t) 3)
i=1 (fi)o \ Oa 1~=1
(7.32)
121
or, to terms of order (~t - 1)2,

fi d l n f i
since all terms in the summation are positive, the topological contribution, A S~,
must be negative. Thus, in both the free energy of deformation AFo= - T AS~ and the
tensile stress f -
1 ~Fe
Vo
, topological effects must always increase the initial
modulus beyond the independent strand value. No general conclusions can be

drawn about magnitude, nor about mechanical response at large deformation,
other than to say that there is no reason to anticipate simple neo-Hookean
behavior.
7.6.2. Topological Models for Entangling Interactions
Evaluating of the apportionment of configurations among topological classes
is difficult even for simple models of non-crossing chains. Prager and Frisch
(302) have analyzed the topological classes of an infinite rod and a closed-loop
random coil. The random coil is allowed to pass freely through its own
contour, so self-entanglement is not considered. The classes correspond simply
to the number of loops the coil makes around the bar. Thus, Configurations A,
B, and C
belong to different topological classes, while Configuration D is entangled with the

bar only because of self-entanglement and therefore in this analysis is classified
with configuration A.
The apportionment among classes depends on r, the distance of one specified
point on the coil (the starting point of a self-closing random walk of n steps)
from the bar. An exact solution is obtained for the probability Po(r) that the coil
is entangled with the bar, i.e., that the coil is looped at least once around the
bar. The average displacement of the coil produced by an external force F was
then calculated from Pe(r), using the Boltzmann equation. The force-displacement
relation is non-linear, with an initial spring constant (initial slope of the forcedisplacement function) which is 1.66 times the value for a loop permanently
fastened to the bar at one point. For large displacements, the spring softens and
then again becomes linear in the force with approximately the same spring constant
122
W.W. Graessley:
as an attached coil. This result is especially interesting because molecular springs

which soften with deformation are exactly what is required to explain a positive
C2 term in the Mooney-Rivlin equation. The authors point out that a much simpler
probability model gives a very similar force-displacement relation: the probability
Po(r) that the starting point and midpoint of the coil lie on the opposite sides of
the plane normal to r and containing the rod.
Edwards (300)examined the classification problem for a random coil bound at
its end points and an infinite rod oriented normal to the end-to-end vector of the
coil:
The ends of each coil (contour length L, step length 1) are attached to points R 1
and R2 in a plane normal to the rod, which pierces the plane at R. The class
of a configuration depends on whether the contour passes above the rod without
encirclement (Configuration a), passes above with one turn around the bar
(Configuration b), etc. An integer defining the class of any configuration is
obtained from the angle swept out by the vector from R to a point on the path
of the configuration as that point moves from R1 to R2. Thus,
O=R1RR2+2nn ( - oo < n < ~ )
(7.34)
and, for example, n = 0 from Configuration a, n = - 1 for Configuration b, and

n = + 1 for Configuration c. Exact expressions for the fraction of configurations
in each class were obtained, using an analogy in electromagnetic theory for the
calculation of induced current in one conductor (the random coil) due to current
flowing in a second, nearby conductor (the rod). Perhaps not unexpectedly, the
actual structure of the computation turns out to be very similar to that of the
restricted diffusion calculation used by Prager and Frisch. Alexander-Katz and
Edwards (303) recently examined classification in a system comprising a random
coil and an object of solenoidal configuration.
In general, it appears that the fraction of configurations in the various
topological classes can be determined for models in which one of the elements
is a fixed curve and the other is a random coil. The detailed calculations are
intricate and difficult, however, and some simple generalizations are needed which
could be used as a step towards building classification effects into the network
theories. Classification for the case of two random coils and for self-entanglement
are unsolved problems at the present time.
123
7.6.3. Network Models with Some Non-Classical Character

The results of these model calculations can be applied in a rather crude way
to the network situation. One might consider each network strand individually,
and in a rough way represent its topological interaction with each neighboring
strand by a single probability Pe(ri) (from Prager and Frisch, for example), in which
ri is the mean distance between centers of gravity of the central strand and its
ith neighbor. Thus, only two topological classes, entangled and not entangled,
would be considered, with (f0o = Pe and (f2)o = 1 - P~ for the interaction with
each neighboring strand. The distances r~ and corresponding probabilities would
change with deformation, and, if the deformation were affine, f~ and f2 for each
neighbor could be calculated as a function of deformation. Assuming the total effect
is obtainable by summing over individual strands (neglecting three-chain and
higher order classification), the free energy and the modulus could be calculated.
Another approach to network modeling is suggested by the work of Edwards
and Freed (304), Collins and Wragg (305), and Eichinger (306) on the configurations available to a confined chain. The idea is to represent the effects of
environmental connectedness on the configurations of a network strand by a box.
The configurations of a strand with junction points centered within the box
could be classified according to whether or not the locus crosses the box surface
at any point. Deformation might be assumed to alter the positions of both the
junction points of the strand and the dimensions of the box in an affine manner.
Classification in the deformed state could be evaluated, leading ultimately to an
evaluation of AS~ by Eq. (7.31). Stress-strain behavior would presumably depend
on the ratio (r2)o/L 2.
'~/~.
"'~, "JF~ L0
.,." ,.,
/
~L0
$s S
One could also construct a network involving elements which allow some
adjustment in the location of their coupling points in response to deformation.
Consider a tetrahedral network element (3) in which two gaussian strands of
equal contour length are attached to the corners and joined together at their
midpoints by a crosslink:
L-1
3
124
W.W. Graessley:
If the network is given a tensile deformation in the 2 direction and the

corners move affinely, the four chains in the tetrahedron contribute independently to the modulus, with the well known result obtained from the rubber
elasticity theory:
F~ = 4 k T (~-~o
(7.35)
~-
in which Fc is the force, and the mean square end-to-end distances (r 2) and (re)o
refer to the four network strands in the unstressed network and as free chains
respectively.
If the crosslink is removed, the system reverts to two independent chains.
To preserve the symmetry of the system one needs to consider the average for
three tetrahedra operating together, one for each of the three possible pairings
of the strands. The result is again neo-Hookean behavior but with a lower
modulus contribution.
8kT(r2)(
F.c=T
(r2)0
1)
~- "
(7.36)
An intermediate situation is obtained if the central crosslink is replaced by

a slipping link such that the two chains are constrained to be in contact at one
shp-hnk /
r~
point, the contact point being allowed to lie anywhere along the chain contours.
Thus, only configurations in which the two chains are in contact are counted in
evaluating the change of entropy with deformation. Again, three tetrahedra are
required to preserve symmetry. For this case, numerical solutions are required;
some results (241) are summarized in Fig. 7.2 in the form of a Mooney-Rivlin
plot. Briefly, one sees deformation softening with the largest effects obtained in
super coiled systems ((rZ)/(r2) o < 1). For swelled systems ((rZ)/(r2) o > 1), the
ratio F
e-~-
is nearly constant with a value which is practically the
same as the system with no central link. In compression the slipping link in
supercoiled systems actually contributes slightly more than a permanent crosslink to the modulus.
125

<r2>
<~o=
1.0
-- CROSSLINK
I
v
kO
0,25
<r2>
<r2>0
0.8
<r=. >
< r = > =
I0
4.0
. . . .
Lt_
NO COUPLE
0.6
0.4
0.0
tendon
compression
0,5
1.0
15
2,0
I/Ct
Fig. 7.2. Mooney-Rivlinplot for the Slip-Link model
Some of the effects displayed by the model are observed experimentally:

F
e- ~-
is found to be a decreasing function of the extension ratio, and
non-Hookean behavior is reduced by swelling. However, it is also reduced by

supercoiling, which is not given by the model. Although probably not a good
model of entanglement interaction, the slipping link at least illustrates that even
simple configurational constraints can produce unexpected results in the stressstrain behavior. It also demonstrates that, as emphasized by Prager and Frisch, it
is not configurational restriction per se, but, rather, changes in configurational
restriction with deformation, which controls the modulus. Thus, swelling reduces
the slip-link contribution, while intuition suggests the reverse. Finally, the result
implies that entanglement (deduced from the plateau modulus in the system
before crosslinking) and permanent crosslinks may not necessarily contribute in an
additive fashion to equilibrium mechanical properties, even in the region of small
deformations.
8. Non-linear Viscoelastic Properties

8.1. Introduction
Only a few non-linear viscoelastic properties have been studied with polymers
of well-characterized structure. The most prominent of these is the shear-rate
dependence of viscosity. Considerable data have now been accumulated for
several polymers, extending over a wide range of molecular weights and concen-
126
W.W. Graessley:
trations. Several possible mechanisms for the shear-rate dependence of viscosity

have been suggested, some of them based on the supposition of flow-induced
changes in the entanglement density. One purpose of this section is to review the
data on q(~) in order to evaluate these theories. The steady state normal stress
functions N1 and N2 are also non-linear properties. A limited amount of data is
available on ~1(~). The coefficient at low shear rates, ~PI(0), has already been
discussed in Section 5 because of its theoretical relationship to linear
viscoelastic parameters. Data on ~2(~) are still scanty, which is also the case for
time-dependent flow properties such as the growth and relaxation of stress at the
onset and cessation of steady sheafing flow. A brief review of these latter properties
concludes with a discussion of the possible role of entanglements in non-linear
viscoelasticity generally. As in the examination of linear viscoelastic properties,
emphasis is placed on systems of narrow molecular weight distribution, with only
brief comments on the effects of polydispersity.
8.2. Shear Rate Dependence of Viscosity

The viscosity approaches a constant value qo at low shear rates; experimentally this region is reached in narrow distribution polymers without difficulty.
As the shear rate is increased the viscosity departs from q0, becoming in almost
all cases a decreasing function of shear rate, In many of these it attains an
approximate power-taw dependence at high shear rates: t/oc I~l-a. In concentrated
systems the transition from t/o to power-law behavior takes place over a relatively
narrow range of shear rates. Polydispersity, particularly the presence of a high
molecular weight tail, widens the transition and displaces the departure from r/o
to lower shear rates. In systems of very broad distribution, such as commercial
polyethylene, the shear rates required to reach t/0(~ g 10 -4 - 10-5 sec-1) may be
very difficult to reach in practice. At very high shear rates the viscosity probably
levels off again. However, except in dilute solution (307) or for low molecular
weight polymers (308), this range is seldom attainable because of mechanical
degradation or the appearance of various types of flow instability.
The onset of shear rate dependence can be specified by a characteristic shear
rate ~;0, chosen for example as the shear rate at which r/ has fallen to some
arbitrary fraction of t/0. Experimentally i0 is almost identical with co0, its
counterpart in the frequency dependence of the dynamic viscosity q'(co) (170, 309).
Review of the current literature reveals no polymer system which violates this
principle. Since the departure of ~/' from r/o is dictated by the longest
relaxation times in the linear viscoelastic spectrum, the onset of shear rate
dependence must be similarly controlled. This is not to say however that
t/(~;) and t/'(co) are mechanistically related [early suggestions along these lines were
made by Bueche (310) and Pao (311)], although their forms are sometimes
similar or even partially superimposeable by slight adjustments in the ~ or co
scales. Flow birefringence shows that chain configur~ttion is affected quite differently in the two types of deformation (307), so the resemblance can be no more
than a superficial one. Also, the forms of the two functions are in fact different,
q'(og) decreasing more rapidly than r/(~) and sometimes reaching a slope,

I
ioc
' i
127
I
o~
0
&
[3
[]
[]
I3
0
0
[]
0
0
E1
0
[3
o
16!it
io
__
td
io 2
[3
[ 0
Io 3
Fig. 8.1. Dynamic viscosity if(co) and steady state viscosity q(~) for undiluted narrow
distribution polystyrenes. The data are plotted in reduced form to facilitate comparison. The
dimensionless shear rate or frequency is %~Iw~/QRT_or %fflw~o/QRT. [See Eq.(8.3)].
The dynamic viscosities are for Mw= 215000 () and Mw= 581000 ([]) at 160C (312).
The steady shear viscosity is for Mw=411000 (A) at 176C (313). The shapes in the onset
region are similar for the three curves, but the apparent limiting slope for the dynamic
viscosities is about - 1:3
- d ( l o g q')/d(log o9), greater than unity in narrow distribution systems (Fig. 8.1).
Since the shear stress must be a monotonically increasing function of shear rate,
the slope, -d(log~l)/d(log~), can never exceed unity for real flows, and experimentally it seems to be no greater than about 0.85.
The absolute value of the complex viscosity,
t~t*(~)l = [G '2 + 6 " 2 ] m / o ~
(8.1)
has a slope which can never exceed unity, and often t~/*(e~)j agrees rather closely
with the form of q(~;) (314) (Fig. 8.2). The data in Fig. 8.3 show that the
correspondence is not a general one however. The objections to a fundamental
connection between q' and ~/must also apply to the case of t~/*1and ~/.
The Rouse and Zimm models provide little direct help in dealing with r/(~)
since each predicts a viscosity which is independent of shear rate. The principal
interest here is in concentrated systems where entanglement effects are prominent.
Nevertheless, shear rate can influence the viscosity of polymer systems at all levels
of concentration, including infinite dilution (307) and melts with M < M c (308,
3t5). It is therefore essential to identify the causes of shear rate dependence in
systems of isolated or weakly interactions molecules in order to separate intramole-
W. W. Graesstey:
128
I00
0
0
D ~A
A
D DA
o
0
[3A~ A
O&
o
o
o
E3
O
O
D
o
[3
o
D
o
n
O
[3
10 -~
__
__
I_~A_
I00
10 -q
i ____
102
I01
#
AND
o
O
i i
io
t03
B'
Fig. 8.2. Absolute complex viscosity Iq*(oo)j and steady state viscosity q0;) for undiluted
narrow distribution polystyrenes. Data are shown for the same samples as in Fig. 8.1.
Limiting slopes for both ~/(~) and I~*(o~)1are almost identical at ca. - 0 . 8
-o
g
io-
~o-I
._.,
~oo
io~
ond p'
Fig. 8.3. Dynamic viscosity, absolute complex viscosity, and steady state viscosity for
narrow distribution polystyrene. Data obtained at 25C on a 0.071 gm/ml solution of
polystyrene (Mw = 860 000) in Aroclor (316)
cular and intermolecular contributions at high concentration and molecular

weight.
8.2.1. Correlation of Experimental q(~) Behavior
The effects of molecular structure o n q(~) are m o s t conveniently discussed in
terms of reduced variables. The choice of form for reduced shear rate is guided
both by the observation that the long relaxation times govern the onset of shear
129
i0 3
M=3.30 I 0 e
0
000000
0000
1.82 10 6
,,", ~
.,,,, ,.', .ix /', A
1.14 x i0 ~
/,,A,~A^,
0
0~
0
~,~
oo
i01
OBOOOBo I~
4.9.5 x 10~
2.7'5 x io 5
i0
i0 -a
,
IO-'
i
t0
i
IO'
"':XlllI:IIlII
i
IO z
i .......
IOs
7;" ( $l~'tl
Fig. 8.4. Shear rate dependence of viscosity as a function of molecular weight at low
concentrations. Data were obtained at 25 C on narrow distribution poly(a-methyl styrene)
samples at a concentration of 0.02 gm/ml in kanechlor (I98)
rate dependence and by the form of the characteristic relaxation time in

spring-bead models:
q-q*
~o-~s
=f(fl),
(8.2)
and
fl=(%-~)M~/cRT.
(8.3)
The reduced shear rate fl retains a meaning over the entire range of concentration, b e c o m i n g f l = [~/] ~l~M~/RT at infinite dilution and fl=%M~/cRT
in concentrated solutions, where % >>t/s. The c o r r e s p o n d i n g reduced viscosities
are [t/]/[t/] 0 and q/t/0 7. R e d u c e d plots remove most of the observed variation
a m o n g systems (Figs. 8.4-8.7), and simplify the empirical examination of residual
7 Others have defined these variables slightly differently, using ( 1 - p)t/s in place of
% in Eqs.(8.2) and (8.3), where p is the volume fraction of polymer in the solution. However,
the practical effect is entirely negligible for all data examined in this review. The value of
% - t h increases so rapidly with polymer concentration that %-~/$ and % - ( t - P)~/s are
essentially indistinguishable at all levels of concentration.
G'M
Another alternative is to correlate data in terms of a reduced shear stress, c--c-~-' in
which a' is (t/- ~/~)~, or simply t/p = a, when t/>>r/$. The practical advantage of shear stress
as a correlating variable is that it emphasizes the sharpness of the down-turn in reduced
viscosity beyond the onset of non-Newtonian behavior. Also, the magnitudes of other nonNewtonian properties, such as the first normal stress difference; appear to be controlled by
the shear stress magnitude. For example, N1 vs ~r is practically independent of temperature
for a given sample, while N1 vs ~ is very temperature sensitive. For our purposes here the
choice is of course arbitrary.
130
W. W. Graessley:
lO5
C=0,23f gm/ml
00000000000000000
i04
O000000000CO O~ 0 192
"JO00000
0 0 0 0 0 0 0 0 0 0 0 ~ 0147
" 0 0 0 O00
t03
000o000o
00o 0101
'
00000 0
102
0 O00 0 0 0 0 0 0 0 0 00000676
O0 o
I01
000o00o000o0(~ 0503
i0I_____
iO -z
t0-t
t0 o
i0 i
t02
I03
Fig. 8.5. Shear rate dependence of viscosity as a function of concentration. Data were
obtained at 30 and 50C on one narrow distribution sample of poly(~-methyl styrene)
~t w= 1820000) in ct-chloronaphthalene (199)
iO s
-0,55 O o o O O O O O o O o o o o O o o o
o0
oo
1.
ooO oOooOooo
c:0.50
~10
c-O
oooo
Oooo
oo
00%00
45oooooooooo 0%0
OOoooO
O00
c-0./~0 o o o o o o o o o o o o o o o o o o o o o
o
00%
"
"u~lO S
C =0.35
~-.
1001-2
oooooooooooooooooooo
Oo o o o
00
c = 0 . 3 0 qm/ml ooooooooooooooooooooooo o
i
10 -t
i
100
~,Shear
i
10~
rate (sec)-~
i
10 ~
103
Fig. 8.6. Shear rate dependence of viscosity as a function of concentration. Data were
obtained on a single narrow distribution sample of polystyrene (/~w = 411000) in n-butyl
benzene (155) at 30 C. ~eproduced from Transactions of the Society of Rheologie, Volume
11, Fig. 2, p. 273, New York: Wiley & Sons.)
differences related to structure (Fig. 8.8). These differences are of t w o kinds: the
form of the reduced viscosity function f(fl), and the value of reduced shear rate
flo which tocates the onset of shear rate dependence. For the purposes of this
review the form of f(fl) is characterized by an estimated value of the power-law
131
R
o
_J
log ~' ( sec "1}
Fig. 8.7. Shear rate dependence of viscosity for undiluted polystyrenes of different molecular
weights. Weight-average molecular weights are O 48500; (3 117000; (3 179000; Q 217000;
Q 242000. All data are reduced to 183 C; the dashed line has a slope of -0.82 (324).
(Reproduced from Journal of colloid and interfache science, Vol. 22, Fig. 1, p. 520. New York:
Academic Press.)
I.C
0.~
O~
0'2 t
OI
i
io
t
iO ~
Fig. 8.8. Shear rate dependence of viscosity in reduced form as a function of concentration.
Data were obtained on one narrow distribution sample (5(tw= 1820000) of poly(~methyl
styrene) in two solvents (I98, 199). Symbols are: for 0.00552 gm/ml in ~t-chloronaphthalene
(CN), for 0.0231 in CN, A for 0,0676 in kanechlor (K), [] for 0.101 in K,<~ for 0.147
in K, and O for 0.192 in K. Note the progressive increase in slope with concentration at low
concentrations, followed by a progressive shift to large fl values without much shape
change at high concentrations
exponent d. The critical shear rate flo is chosen as the value o f reduced shear rate
at which r l - ~ls falls to 8 0 % of its value at zero shear rate. There is n o special
significance t o these definitions. M u c h information on the detailed shape of f ( f l )
is ignored of course, and other definitions of flo would also suffice. E a c h
132
W. W. Graessley:
0,8
O,B
'1o
IZ
1.1,1
Z
ql
'" 0.4
_1
I
W
0.2
o~ ?
;b
2b
c,7]
sb
40
Fig. 8.9. Power law exponent d as a function of the coil overlap parameter c[r/] at tow
concentrations. The filled circles are narrow distribution polystyrene solutions (177, 316,
318), the open circles are poly(~-methyl styrene) (198, 318). Solvents are chlorinated
di-phenyls except the intrinsic viscosity data which were obtained in toluene. Symbols are:
for polystyrene M = 13.6 x 10 6, 0 1.8 10 6, and O 0.86 106; for poly(~-methyl styrene)
O M = 7.5 x 106, ~ 3.3 X 10 6, CY" 1.82 10 6, O- 1.14 10 6, C)-. 0.694 x 106, and
Q 0.444 10 6
parameter can be evaluated with reasonable accuracy from the available d a t a

and can be examined without reference to any particular molecular viewpoint.
8.2.1.1. Form of the Reduced Viscosity Function. Endo e t al. have published
extensive data on ~/(~) in 1 - 1 0 % solutions of poly(~-methyl styrene) (198).
Ashare (I 77) and Harris (316) have covered a similar range of concentrations and
molecular weights for polystyrene. D a t a on the shear rate dependence of intrinsic
viscosity for the two polymers are also available (318-321). It is a p p a r e n t from
Fig. 8.8 that the viscosity-shear rate curves are not superimposable at these
m o d e r a t e concentrations, f(/~) changing with both concentration and molecular
weight t h r o u g h o u t the regime s . The shape, as characterized by the power-law
8 It is conceivable that a suitable reducing scheme could be found to unify the
t / - 2)at intermediate concentrations. For example, the viscosities in Fig. 8.4 tend to level off
at high shear rates. Aside from the two lowest molecular weights, the limiting viscosity
~/~- t/~ is approximately the same for all. The same data were replotted in the recently
suggested form (317), (r/-t/o~)/(t/o-t/J vs fl, in hopes of obtaining a master curve. A
complete reduction was not obtained, although this same method was successfully used with
similar data on a series of undiluted poly(dimethyl siloxane) samples (317). Possibly a
reduced form based on fl'=fl(rlo-rlo~)/(tlo-tl~) or some other variant would have been
more successful.

I
0.01
0.I
133
1.C
%
O. ~
O.01~.001
1.0
I0.0
.0
J'/ro
Fig. 8.10. Viscosity-shear rate master curve for concentrated polystyrene-n-butyl benzene
solutions. The data were obtained for molecular weights ranging from 160000 to 2400000;
concentrations from 0.255 to 0.55 gm/ml, and temperatures from 30 C to 60 C (155)
exponent, appears to be principally a function of the coil overlap parameter c[r/].

Figure 8.9 shows d vs c[r/]. The exponent rises from d~0.10 at infinite dilution
and approaches d ~ 0.8 for c[r/] > 20.
Although taking place over a somewhat wider range of c[t/], this transition
parallels the variation of JeR from Zimm-like to Rouse-like behavior at tow
concentrations (15). It also supports the contention (Section 5) that coil overlap
is the principal structural variable affecting viscoelastic behavior at low to
moderate concentrations.
Some variations in f(B) with solvent power and molecular weight are
observed in the intrinsic viscosity (3t8-32t). Significant shear rate dependence
in [r/] at convenient shear rates seems to require very high molecular weights, and
is enhanced in good solvents (318). The data available at low to moderate
concentrations were obtained only in good solvents, so the variation with
solvent power in this range is not known. The agreement between data for
polystyrene and poly(~-methyl styrene) in Fig. 8.9 suggests the possibility of a
universal correlation between the form of f(/~) and cEr/] for linear polymers of
narrow molecular weight distribution. However, these two polymers are probably
not sufficiently different to justify such a broad conclusion at this point. Also,
there is some indication in the data at low concentrations that (r/- ~/s)/(r/o- qs)
approaches a finite limit at high shear rates (Fig. 8.4). The source of this behavior
could be specific to each polymer although, as in the case of r/'(co) wliere internal
viscosity provides a lower limit at high frequencies, it probably has little
influence on f(/~) except at very high shear rates.
Judged by the superposability of viscosity-shear rate data on the same master
curve for a variety of polymers [polystyrene (155) (Fig. 8.10), poly(c~-methyl
134
W.W. Graessley:
I
0
O
~o
~o
cM w
Fig. 8.11. Dimensionless shear rate fl0 locating the onset of shear rate dependence for
viscosity in narrow distribution polystyrene systems. Symbols are: for solutions at 30C
in n-butyl benzene (155), ~ for solutions at 25C in arochlor (177), and 0 for undiluted
polymers at 159 and 183C (324). Values for the intrinsic viscosity (cM=0) lie in the
range/3o = t-2, varying somewhat with solvent-polymer interaction and molecular weight
(307, 318-321)
styrene) (199), polyvinyl acetate (195), polybutadiene (322), and polydimethyl

siloxane (323)], the form of f (fl) in concentrated solutions and melts appears to
become essentially a universal function for monodisperse linear chains of high

molecular weight. Based on Fig. 8.9 this limiting form applies for systems in which
c[~] > 30.
8.2.1.2. Onset of Shear Rate Dependence. At low concentrations fl0 is nearly
constant at 1.5-2.0, values not much different than those at infinite dilution.
At high concentrations and molecular weights the fl0 values increase systematically. As in the earlier correlations of steady state compliance, the departure
appears to be a function of the segment-segment interaction parameter cM.
Figure8.11 shows the behavior in polystyrene systems, including results for
melts (324) and for solutions in various solvents (155, 177, 316). Other studies
have shown that solvent power has very little influence on fin (53, 325). The
departure from constant fl0 begins somewhat beyond cM = QM~,and corresponds
roughly to the region where f(fl) ceases to vary with c and M. Figure 8.12 is a
similar plot for poly(~-methyl styrene). The apparent trend to larger values of fie
at small cM is believed to be an artifact of the definition of fin used here and the
flatness of f(fl) for dilute solutions of all but the very highest molecular weight
samples. Values of fin at zero concentration reported with the figure were obtained
from [it] vs ~ data on poly(~-methyl styrene) of much higher molecular weight
135
i0 ~
00
0 0
6o
66
,ao
6o-
tO4
IO5
I~
cM w
Fig. 8.12. Dimensionless shear rate 13o locating the onset of shear dependence in the viscosity
for narrow distribution poly(ct-methyt styrene) systems. Symbols are (t98, 199):
C) M=3.3x t0 ~, (~ M = l . 8 2 x 106,O-M=l.19x 106, and Q M=0.444x 106. Values for
intrinsic viscosity (cM =0) are similar to those for polystyrene (see caption of Fig. 8.11)
(M = 7 x 10 6) than the sample (M = 1.8 x 10 6) which shows the apparent rise in

/3o in dilute solution.
Following the procedure used with J~R (Section 5), flo data on these and other
polymers were correlated in terms of cM/QM, with results which are shown in
Fig. 8.13. The parallel in behavior between the/~o and J,R master correlations is
unmistakable. Even the relative positions of polymers on the master correlations
are similar; note for example the data on JoR and /~o for polybutadiene.
Published data on relatively narrow distribution polyethylene (210, 326) have not
been included in Fig. 8.13 because departures from ~/o were too small to define
fl0 with accuracy. However, estimates of/~o from the data provided suggest that
polyethylene may follow a different pattern than other polymers. Departures
from r/o seem to appear at anomalously low shear rates (326). Aside from ~/0 values,
viscoelastic data on well characterized crystallizable polymers in the melt state
are rather scarce. Although not especially anticipated, it is certainly conceivable
that crystallizability confers some unusual features to the flow behavior.
The parallel between/~0 and J~R has been noted elsewhere (208, 213, 328, 329),
and is not in fact fortuitous. It follows rather directly from the empirical
observation that departures of r/(3)) and r/'(~o) from r/0 are governed by the longest
relaxation times of the system, combined with slight extensions of a reduced
variables argument suggested by Markovitz for linear viscoelastic behavior
(329). Suppose one wants to compare the forms of the dynamic moduli on
136
W . W . Graessley:
,8o
IOIf
oPOLY
( ct -METHYL
STYRENE)
~
IIiiI
I0
"POLYSTYRENE
~ _ _
i
I01
A._
I
10 2
cMw/pM
Fig. 8.13. Dimensionless shear rate fl0 locating the onset of shear rate dependence in the
viscosity in narrow distribution systems of linear polymers vs cM/QM. Symbols for data
on additional polymers are: for undiluted 1,4polybutadiene (322), for undiluted
poly(dimethyl siloxane) (323), and O for solutions of polyvinyl acetate in diethyl phthalate
(195). The dotted lines indicate the ranges of fl0 for the intrinsic viscosity
samples having different values of q0 and jo. If solvent is present and G'(co) and
G"(co) - coth are the p o l y m e r contributions, then as usual
lim C,~_",co,- corh _ r/o - qs,
co--, 0
(8.4)
CO
G'(co)
lim
o---* 0
c o l - = r/o2 j o .
(8.5)
If the d y n a m i c moduli are expressed in the reduced forms,
G,_jol
R -- , k r/o -- rh }
Gi{ = g
G',
,7o /2 (G" -
\ r/o - rh /
corh),
(8.6)
(8.7)
,7 Jo
coR -
co
~/o - q~
(8.8)
137
it is then easy to show that

pf
lim GR = 1,
~ R "* 0
lim
taR--* 0
(8.9)
fO R
= 1.
(8.10)
fD 2
Thus, differences among samples due to differences in jo and r/0 are removed
in this reduced form. Since q0 and jo are sensitive characteristics of the long
relaxation time processes, and since the initial departure of r/~ - q s from q0 - ~ h
is governed by these same processes, one would espect significant departures
from ~/0 to occur near some roughly constant characteristic value of 09R with a
value of approximately unity. Finally, if the onset of shear rate dependence
depends on the same processes, then Eq.(8.8) also gives an appropriate reduced
form for ~;:
3~R--
~.
(8.11)
r/ -- qS
With the definitions of fl and J~g, Eq.(8.11) can be expressed:
7R = J~Rfl.
(8.12)
The onset of shear rate dependence therefore is governed by the product JeRfl
and should occur near JeRfl = 1. The observed value of this product at departure
will of course depend on the criterion chosen to define departure. It may also vary
somewhat if the criterion requires a large departure and the form of r/(~) differs
substantially from sample to sample.
Figure 8.14 shows the result of combining the correlations of flo (Fig. 8.13)
and JeR (Fig. 5.18). The product flOJ~Ris remarkably independent of concentration
and molecular weight from infinite dilution all the way to the undiluted melt
state and shows no substantial variation from one polymer to another. For the
particular definition of critical reduced shear rate ~o used here, the experimental
result for narrow distribution linear chains can be expressed as
flOJeR =
- = 0.6 _+0.2.
i'/o- r/~
(8.13)
138
W.W. Graessley:
2,0
1,0
0.8
Polybutodiene
Polystyrene
0.6
-3
cO..
0.4
0.2
0.1
i
o.i
lo.o
1.o
looo
cM/pMc
Fig. 8.14. Product of reduced compliance JeR and characteristic shear rate flo for narrow
distribution systems of linear polymers. The dashed lines indicate the range of floJR for
the intrinsic viscosity
Aside from minor differences this equation is the same as that suggested recently
by Prest et al. (208) 9.
There is considerable evidence to suggest that such correlations can be
extended to polydisperse linear polymers. The fact that jo tends to be unusually
large in branched polymers and that ~/0~;o is unusually small suggests that the
correlation may be applicable in branched systems also. If so, the result should
have considerable empirical utility. It must be pointed out, however, that the
definition of~0 used here is a rather arbitrary one, and that the result in any case,
like the Merz-Cox rule, has no foundation in fundamental continuum theories.
For example, although the implied relationship between elastic properties and
onset of shear rate dependence in the viscosity is acceptable within the framework of simple fluid theory, it is not a direct and unique consequence (330).
This is not surprising, of course, since both the generalized Newtonian liquid
and the Lodge elastic liquid are simple fluids, and yet the former has a shear
rate dependent viscosity without elasticity while the latter has elasticity but a
viscosity which is independent of shear rate.
9 This correlation brings together two definitions of characteristic time from continuum
rheology. Slattery (329a) notes that Bird has suggested a characteristic time vB for a viscoelastic fluid, defined in terms of the onset of shear rate dependence of the steady shear
viscosity. For the correlations here we may take vB= 1/Vo. He also notes that Truesdell had
suggested a characteristic time TT,defined in terms of the normal stress differences in steady
shear flow. For the correlations here we may take ~T= 7J1(0)/%. From the ColemanMarkovitz relation [Eq. (3.21)] ~ = 2 jo r/0,yielding, from Eq. (8.13):
z:r/rB= (1.2 +_0.4) % - ~/s
%
for the special case of narrow distribution polymers.
(8.13a)
139
8.2.2. Theories of Shear Rate Dependence of Viscosity

8.2.2.1. Intramolecular Contributions. Several explanations have been suggested
for shear rate dependence of viscosity in dilute solutions. Shearing flow tends to
extend the molecules in the flow direction, increasing the pervaded volume and
reducing the average concentration of segments within the coil (33t). Perturbation
of coil dimensions due to the effects of excluded volume is therefore smaller in
flow than at equilibrium. On the other hand, the surface of the pervaded volume
is greater in the flow-extended state, so more segments are exposed to the drag
forces of the external flow. The original Zimm analysis accounts for neither
effect. According to Peterlin (332), the two effects generally combine to produce
a net decrease in viscosity with increasing shear rate, beginning near fl = 1.
However, under some conditions, such as in the case of highly viscous good
solvents, the viscosity may actually increase with shear rate (333).
Although shear rate effects are more pronounced in good solvents, the
intrinsic viscosity decreases with shear rate even in 0-solvents, where excluded
volume is zero (317, 3I 8). The Zimm model employs the hydrodynamic interaction
coefficients in the mean equilibrium configuration for all shear rates, despite the
fact that the mean segment spacings change with coil deformation. Fixman has
allowed the interaction matrix to vary in an appropriate way with coil deformation (334). The initial departure from rq]0 was calculated by a perturbation
scheme, and a decrease with increasing shear rate in 0-systems was predicted to
take place in the vicinity of fl = 1.
In the above calculations the sub-molecule springs of the model are linear,
so the molecules are treated as though they were indefinitely extensible. Changes
in viscosity due to alterations in excluded volume or hydrodynamic interaction
are assumed to occur well before the mechanical non-linearity of real molecules
at high extensions becomes important. Light scattering measurements during
shearing flow show that the coils are indeed extended in the flow direction,
although perhaps not as much as predicted by the linear spring-bead models
(335-337). It turns out that finite extensibility can also provide a shear rate
dependent viscosity.
There is nothing to prevent one from formulating diffusion equations for
spring-bead models with non-linear springs. The difficulty comes in the calculation
of stresses from these equations. The many-bead diffusion equation is very
delicate. Seemingly simple modifications of the linear spring-bead model can
prevent the uncoupling of motions by transformation to normal coordinates,
thereby rendering the equations intractible. Inferences must then be drawn from
either approximate treatments of many element models or exact treatments of
low-order homologs for which the calculations can be carried to completion
without a separation of variables. The low-order homolog of the linear springbead models is the elastic dumbbell: two frictional beads connected by a linear
spring. Except for the distribution of relaxation times the elastic dumbbell has the
same qualitative features, including indefinite extensibility, as the multiple
bead Rouse model.
Bird and co-workers have recently examined a number of low-order models
which mimic finite extensibility (102, 338). The rigid dumbbell consists of two
140
W.W. Graessley:
frictional beads held at a constant separation distance. In the string-bead

dumbbell two beads are joined by a flexible string which transmits no force
unless fully extended and serves only to place a limitation on the maximum
separation of the beads. The finitely extendible non-linear elastic (FENE)
dumbbell consists of two beads connected by a spring which is linear for small
extensions but becomes non-linear at large extensions, having in fact an upper
limit, r = ro, on its extension (338):
Kr
F~p-1- (~o)2 (O<r<r)"
(8.14)
The bead separation vector is r, and r = Ir].

The equilibrium distribution of bead separation distances in the FENE
dumbbell is:
\-~0J }
(0<r<r)
(8.15)
in which b = Kr2/kT. The limit K--*0 at fixed r 0 corresponds to the string-bead

dumbbell, while the limit r0 ~ oo at fixed K yields the elastic dumbbell. The forceextension curve for F E N E dumbbells is in fact rather similar to the inverse
Langevin function which describes the behavior of freely jointed chains of finite
contour length. The flow-induced stresses contributed by dilute suspensions of
such species in a viscous medium have been calculated by Warner (338). The
viscosity is constant at low shear rates, then begins to decrease in a characteristic
region of shear rates, eventually approaching the power-law behavior
- -r]- ---o o~'/s

rio - r/~
t7t -113
(8.16)
Internal viscosity (Section 4) provides another possible source of shear-rate

dependence. For sufficiently rapid disturbances, a spring-bead model with internal
viscosity acts like a rigid body; for sufficiently slow disturbances it is flexible and
indefinitely extensible. The analytical difficulties for coupled, non-linear springbead systems are equally severe in linear spring-bead systems with internal
viscosity. Even the elastic dumbbell with internal viscosity has only been solved
exactly in the limit of small e (339), where e is the ratio of internal friction
coefficient to molecular (external) friction coefficient (0n. For this case, the
viscosity decreases with shear rate.
All these potential mechanisms for shear rate dependence--changes in
excluded volume, changes in hydrodynamic interaction, finite extensibility and
141
internal viscosity--can be made to predict viscosity-shear rate behavior which

imitates more or less faithfully the forms observed experimentally. The choice
among them must therefore be based on simultaneous comparisons with other
equilibrium and viscoelastic properties.
It seems possible to rule out finite extensibility as an explanation of shear
rate dependence in the viscosity, based simply on the equilibrium properties of
polymer chains and the parallel between r/ and r/' in their departures from q0.
Experimentally, the mean square end-to-end vector obeys Gaussian statistics in
0-solvents: (r2)/M= constant. If the spring constant K in FENE dumbbells is
chosen to match this condition, then K = 3kT/(r2). The parameter b is therefore given by
b= @2)
3r~
(8,17)
which must be a large number, of the order of the number of main chain atoms
in the molecule, in order for Gaussian statistics to be obeyed. The steady shear
viscosity departs from q o - rh at a dimensionless shear rate /~o- The dynamic
viscosity departs from ~/o - r/s at a dimensionless frequency/~;. The ratio of these
values, obtained from the coefficient of the quadratic terms in power series
expressions for viscosity in FENE dumbbells (338), is
90 = [ (2b + 11)(b + 5)] a/z

coo [ -2(4ff-+i~ -
(8.18)
which, for b ~>1, becomes
9o
COo
bl/2
-
(8.19)
Accordingly, given the necessity from equilibrium coil dimensions that b >>1,
the shear rate and frequency departures predicted by FENE dumbbells are
displaced from each other. Moreover, the displacement increases with chain
length. This is a clearly inconsistent with experimental behavior at all levels of
concentration, including infinite dilution. Thus, finite extensibility must fail as a
general model for the onset of nonlinear viscoelastic behavior in flexible polymer
systems. It could, of course, become important in some situations, such as in
elongational and shear flows at very high rates of deformation.
Similar arguments can be raised against internal viscosity as a primary
cause of shear rate dependence in the viscosity. A recent review (339a) has shown
that current many-bead theories predict differences in ~o and coo when e is small
and chosen to fit the oscillatory data. The approximate nature of these analyses
leaves serious doubts about their predictions in steady-state flow, however. A new
142
W.W. Graessley:
theory has been developed (339b) which addresses some of these problems, but its
predictions about oscillatory and steady-shear behavior appear to be very similar
to those of the earlier theories.
Both the many-bead and dumbbell models with internal friction predict
limiting viscosities at high frequencies, t/~, and high shear rates q+. The theories
predict that t/~ and t/~ are related+ such that
t
t/.
q+
t/~
r/0 - t/s
r/o -
q0-qs
qs
[many bead (339b)]
q+
t/o-t/s
[dumbbell (339)].
Experimentally, (t/" - tts)/ (q o - q+) 4. 1 (93, I 17), meaning that (qo~-qs)/(tlo -t/s)
should be only slightly less than unity, and shear rate dependence should
practically disappear if only internal viscosity is considered. Such behavior is not
of course observed.
A final piece of evidence against both finite extensibility and internal
viscosity is provided by flow birefringence studies. One would expect each to
produce variations in the stress optical coefficient with shear rate, beginning near
the onset of shear rate dependence in the viscosity (307). Experimentally, the
stress-optical coefficient remains constant well beyond the onset of shear rate
dependence in t/for all ranges of polymer concentration (18, 340).
It seems unlikely that these conclusions would be changed if the calculations
were extended rigorously to models containing many beads and springs. The
long (linear viscoeleastic) relaxation times of the undisturbed system clearly govern
the onset of non-Newtonian behavior. If the distribution of configurations at
equilibrium is Gaussian, then very few of the molecules will be greatly extended,
and their finite lengths can have very little bearing on the relaxation time
distribution. It is therefore difficult to see how finite extensibility could play a role
at the early stages of non-linear behavior. Likewise, internal viscosity is essentially
a local phenomenon, and seems quite unlikely to make its presence felt in the
long relaxation processes. Thurston and Peterlin's approximation calculation for
multiple bead systems with internal viscosity (118) contains this result, and the
linear viscoelastic data of Massa et al. (93), for polystyrene in Arochlor agree very
well with the resulting equations. It would therefore be very surprising to find that
internal viscosity, which has practically no influence on the long relaxation times,
could provide a mechanism for shear rate dependence in the viscosity, when
experiments show that the onset of shear rate dependence is controlled by those
same long relaxation times.
Changes in excluded volume and in intramolecular hydrodynamic interaction
appear at the present time to be the only acceptable explanations for the onset of
shear rate dependence in systems without appreciable intermolecular interactions.
It seems likely that both internal viscosity and finite extensibility would assume
importance only at much higher shear rates.
143
8.2.2.2. Intermoleeular Contributions. Increasing concentration reduces the

effects of excluded volume and intramolecular, hydrodynamic on viscoelastic
properties (Section 5). Internal viscosity and finite extensibilty have already been
eliminated as primary causes of shear rate dependence in the viscosity. Thus,
none of the intramolecular mechanisms, even abetted by an increased effective
viscosity in the molecular environment, can account for the increase in shear rate
dependence with concentration, e.9., the dependence of power-law exponent on
coil overlap c[~/] (Fig. 8.9). Changes in intermolecular interaction with
increased shear rate seems to be the only reasonable source of enhanced shear
rate dependence, at least with respect to the early dewiations from Newtonian
behavior and through a substantial portion of the power law regime.
In Williams' model the coil distortion produced by flow alters the intermolecular interaction and gives rise a shear rate dependence in the viscosity (2t7).
According to this analysis, the potential which governs the total force between
overlapping molecules decreases as the molecules are drawn out in the flow
direction. Williams calculates ~/(~) using the flow-dependent segment density
distribution for elastic dumbbells. A monotonically decreasing function of ~ is
obtained which is algebraically complicated but roughly obeys the form:
.i
7o
1+
11112
(8.20)
The time constant % is the relaxation time of the elastic dumbbell, which would
be ( % - q ~ ) M / c R T if the dumbbells were independent. Williams makes the
dimensional argument described earlier (Part 6) in favor of a different form, the
result being
t o = F(c) ~1o - tl,
(8.21)
in which F(c) depends only upon the equilibrium thermodynamic properties of the
solution. It is assumed to be independent of polymer molecular weight and at
most only a weak function of concentration.
This treatment assumes that the forces between molecules in relative motion
are related directly to the thermodynamic properties of the solution. The excluded volume does indeed exert an indirect effect on transport properties in
dilute solutions through its influence on chain dimensions. Also, there is probably
a close relationship between such thermodynamic properties as isothermal compressibility and the free volume parameters which control segmental friction.
However, there is no evidence to support a direct connection between solution
thermodynamics and the frictional forces associated with large scale molecular
structure at any level of polymer concentration.
144
W.W. Graessley:
The theory associates the onset of shear rate dependence with coil distortion,
and is thus consistent with the close relationship which is observed between r/vs
,~ and q' vs o. It does not predict any dependence of the f(fl) form on concentration, contrary to observations in the low to moderate concentration range.
Also, in this same range the characteristic relaxation time obeys the form suggested
by the Rouse theory, rather than (qo- q2)/czR T. The Williams theory, in fact,
describes highly concentrated solutions and melts somewhat better than the
moderately concentrated but unentangled solutions for which it was developed.
This may mean that the formalism provides a reasonable attack on flow properties at high concentrations, but without the need to invoke chain entanglement
directly.
Further examination of the Williams approach seems called for, both to
improve the method for estimating parameters such as the relaxation time, and to
clarify the relationship between the intermolecular potential form and nonthermodynamic frictional forces. The method might provide a fairly unified
description of non-linear flow porperties if a suitable "potential function" for large
scale molecular friction were found. Aside from the Williams work, there have been
no theoretical studies dealing with t/ vs. ?) at low to moderate concentrations.
The systematic changes in the master curve f(fl) with coil overlap c[q] are thus
without explanation at the present time.
Experimental viscosity-shear rate curves at high concentrations turns out to be
rather similar to an expression for non-Newtonian viscosity derived from the
Eyring's activated complex theory for the transport properties of liquids (341):
sinh- 1 ~ to
/10
)' l: 0
(8.22)
The derivation of this equation makes it seem more appropriate for small molecule
liquids or suspensions of hard spheres than for interpenetrating random coil
molecules however. Indeed, a somewhat modified version of the equation has
approximately the form observed for t/(,)) in concentrated suspensions of monodisperse spheres. In this case l/t0 turns out to be of the order 'of the rotational
qo a3
diffusion of the spheres, ~ - ~ (342, 343), the parameter a being the radius of
of the spheres.
The Eyring analysis does not explicity take chain structures into account, so
its molecular picture is not obviously applicable to polymer systems. It also does
not appear to predict normal stress differences in shear flow. Consequently, the
mechanism of shear-rate dependence and the physical interpretation of the
characteristic time zo are unclear, as are their relationships to molecular structure
and to cooperative configurational relaxation as reflected by the linear viscoelastic
behavior. At the present time it is uncertain whether the agreement with experiment
is simply fortuitous, or whether it signifies some kind of underlying unity in the
shear rate dependence of concentrated systems of identical particles, regardless of
their structure and the mechanism of interaction.
145
A master curve appropriate for concentrated systems has been derived from the
assumption that shear rate dependence is caused by a progressive decrease in the
steady-state entanglement density (227, 344). Linear viscoelastic behavior indicates
that each molecule at equilibrium participates in a number of potential coupling
arrangements with other molecules lying within its pervaded volume. Random
thermal motion continually changes the configuration of each chain, new potential
couples being formed while old ones are lost. If the average number of couples per
chain is large, the fluctuations in coupling density will be small and the number
of couples per chain will be roughly constant in time. This is the equilibrium
situation envisioned in most relaxing network theories.
Macroscopic deformation imposes a systematic relative motion on the
molecules. If each is entangled with a large number of other molecules, its
center of gravity moves with the macroscopic velocity at its location since the
effects of entanglement drag on its trajectory cancel. If the motion is steadily
continued, each molecule sees a constant stream of potential coupling partners
passing through its pervaded volume. If the flow is slow enough, random thermal
motion can establish new coupling arrangements with incoming partners as
previous partners leave. Thus, for sufficiently slow flows (or rapid but sufficiently
small deformations), the topological state of the system remains essentially the
same as that at equilibrium.
For higher deformation rates, the time that potential partners remain within
the pervaded volume of a molecule will decrease. The time of passage % is an
external time, governed by the macroscopic deformation rate and the trajectory
of each potential coupling partner. The time to form an entanglement z,, on the
other hand, is to be viewed as an internal time, governed by the rate of largescale configurational relaxation and of the order of the configurational memory
time (Section 4) for the chains. The coupling probability is assumed to depend on
the relative values of zp and % approaching the equilibrium probability when
z~ ~ zp and vanishing when -c~~>%. Accordingly, the steady-state entanglement
density of the system is expected to be a decreasing function of the deformation
rate.
A simplified model of this situation has been used to calculate viscosity-shear
rate behavior. The steady-state power dissipation per unit volume /~v was
calculated as the sum of the power dissipations by individual molecules, each
molecule being treated as if it interacted independently with the surrounding
medium. Entanglement probability was represented as a step function. Interactions with elements in the medium which pass through the pervaded volume
of the molecule (taken to be a sphere of unspecified Rs) in a time less than z~
were omitted, while those for which the passage time was greater than re were
given their full weight. Effects due to coil rotation and to deformation of the
pervaded volume by the frictional forces were ignored. EAccording to a recent
discussion (2t2), entanglement drag tends to align the sections of chain between
coupling points, but produces considerably tess overall distortion of the coil than
drag in a continuous medium of the same macroscopic viscosity.]
It was also assumed that internal times such as ze are proportional to the
macroscopic viscosity: ze = t//~o (Z~)o; (Ze)o or Zo is the mean entanglement time at
equilibrium. The viscosity was then calculated from the expression t/=/~v/~; 2,
146
W.W. Graessley:
which is an identity for simple shear flow at steady state. The result is an
equation for the viscosity master curve in monodisperse systems (227):
= 9 3/2 h
(8.23)
in which
g(0)= 2 [ c o t - 1 0 + l @ 0 Z t
(8.24)
h(O)= r~2 c o t - 1 0 + (1+02) 2 '
(8.25)
7~
and
0=
~/ ~Zo
~/o 2
(8.26)
The functions g and h represent two separate effects of disentanglement.

First, each molecule has fewer couples with the medium than at equilibrium,
couples being lost preferentially at the coil periphery in the direction of the
velocity gradient. The power dissipation is therefore less by the factor t - h than
would be the case if all couples available at equilibrium were acting. Second,
the medium itself has fewer couples than at equilibrium. Those couples remaining
between the molecule and the medium exert drag forces which are smaller by the
factor g 3/2 than would be the case if all couples present at equilibrium were acting.
The function 1 - g is the fractional reduction in the entanglement density
(g = E/Eo) under the existing condition of flow. Equations (8.23)-(8.26) define an
implicit expression for the master curve r//q0 vs ~Zo, the reduced viscosity q/q0
being also present in the arguments of the functions g and h. Appropriate
calculations yield a result which turns out to be practically indistinguishable
from Eyring's equation (34I).
Bueche (347) has analyzed a similar disentanglement model, evaluating in
somewhat greater detail the coupling probabilities and taking into account the expected differences in % for entanglements of different complexities. Again the form
is very similar to Eqs. (8.22) and (8.23).
Table 8.1 compares the asymptotic behavior at high rates of the various master
curves for monodisperse systems. Their limiting forms
Table 8.1. Behavior of theoretical master curves for viscosityat high shear rates
Investigator
Asymptotic
form of qltlo
Williams (217)
Eyring (341)
Graessley (227)
Bueche (347)
1/~Zo
In ~Zo/~Z~
1/(4/%)9/1'i
1/0; ZO)6/7
147

10 4
10 ~
g..
i0 ~
~'ld'
~o
rd
to~
"~
io~
Jo~
to~ "
~o~
(SHEAR RATE)
Fig. 8.15. Viscosity vs shear rate in concentrated solutions of narrow distribution polystyrene.
The solvent in n-butyl benzene, the concentration is 0.300 gm/ml and the temperature is
30 C. The symbols are O for M = 860000 and [] for M=411000 at low shear rates (155)
and at high shear rates (346). The solid line for M = 860000 is the master curve for
monodisperse systems from Graessley (227). The solid line for M=411000 is the master
curve from Ree-Eyring (341). Either master curve fits data for both molecular weights
are reached after a decrease of about one decade in q/qo" It is difficult to

distinguish a m o n g them experimentally, partly because data becomes less certain
at very high shear rates and partly because residual polydispersity would
probably obscure the relatively small differences anyway. In any case, the shape
of the curves between the Newtonian and high shear rate regions, where m a n y
data are available, is roughly the same in the four equations and in good
agreement with experiment (155, 346) (Fig. 8.15).
Theories based on the uniformly effective medium have the practical advantage
that they can be extended quite easily to polydisperse systems (227). Viscosity
master curves can be predicted from the molecular weight distribution, for
example. The only new assumption is that the entanglement time at equilibrium
for a chain of molecular weight M in a polydisperse system has the form
suggested by the Rouse theory (5):
t , ( M ) - - n2
= tr --
cRT
Mw
(8.27)
That is, at zero shear rate
(M)2
%(M) = t o -Mw
(8.28)
148
W, W. Graessley:
in which zo is now a communal time constant, evaluated [as suggested by

Eq. (8.27)] for a monodisperse system with the same molecular weight as Mw of the
polydisperse system.
The agreement with experimental master curves in polydisperse systems turns
out to be fairly good (345), although some significant differences may exist for
very broad molecular weight distributions (348). Values of z0 derived from systems
of different polydispersities also correlate well in terms of ~tw (345). One
hesitates to drawn firm conclusions about the validity of the underlying concept
of chain disentanglement from this agreement, however. For example, Middleman
(349) evaluates the master curve for a polydisperse system simply as a weighted
sum of monodisperse master curves. He uses the Rouse form for the weighting
factor [Eq. (8.27)] but introduces no explicit considerations of mechanism. The
predicted shapes are not in quite as good agreement with experiment as those
based on the disentanglement hypothesis, and the values of zo obtained correlate
less well in terms of Mw (345). However, the general features of the predicted
curves are very similar. What does seem to be clear is that entanglement is an
acceptable explanation of shear rate dependence and that the Rouse weighting
factor for individual contributions, (M//~w) 2, is roughly correct in polydisperse
systems.
Unfortunately, the energy dissipation method upon which these calculations
are based is only applicable to the evaluation of viscosity in steady
deformations. The method does not lend itself to an extension of the disentanglement model to other components of the stress or to other types of
deformation history.
8.3. First Normal Stress Function
The first normal stress function 7sl is defined as
~1(~)__- P l l - P 2 2
~2
N1
~z -
(8.29)
The properties of ~1 (0) for narrow distribution polymers have already been discussed in Section 5. The behavior of 7q(~) at higher shear rates has only been
determined for a few systems of well-characterized molecular structure. The
experimental problems are more difficult than in the case of q(?~), so the
conclusions here must be regarded as somewhat more tentative. Experimentally,
t/'l(;) ) and t/(,)) depart from their zero shear values within the same range of
shear rates (172). Shear rate sensitivity is much smaller when N1 is expressed as
a function of shear stress (350).
N1 = 2J(~) a 2 .
(8.30)
At the limit of low shear rates J(~) approaches jo according to the ColemanMarkovitz relation [Eq. (3.21)] and the experimental results in Section 5. In
149
many systems J(~;) is a decreasing function of shear rate (351). However,

Graessley and Segal (352) found that J(~) increased with ~ in concentrated
solutions of polystyrene with narrow molecular weight distribution, while
decreasing with ~ in broad distribution samples. Nagasawa and co-workers
(198, 199) have found similar behavior in poly(~-methyl styrene) systems, and
Uy and Graessley (195) in polyvinyl acetate, although in the latter study the shear
rate dependence of J(~) for narrow distributions was quite small in all cases.
Such behavior is qualitatively understandable in terms of partial disentanglement in steady shear flow. In highly entangled systems (cM > oM') jo is of the
form (Section 5):
M
jo~ _ _
cEoR T
(8.31)
in which Eo is M/(M~)so~n, the number of entanglement points per molecule at

equilibrium. According to the entanglement theory of ~();), the steady state
number of entanglements at shear rate ~ is reduced by the factor g [Eq.
(8.24)], so that
M
M
J(~)oc c E R T - cgEoRT
(8.32)
J(~) = J/g(~ro).
(8.33)
or
Since g is a monotonically decreasing function of shear rate, departing from

unity at the onset of shear rate dependence on ~/, J0;) should show positive
curvature (a montonic increase with shear rate) for systems of narrow molecular
weight distribution. The shear rate dependence shown by Nagasawa's data (Fig.
8.16) is qualitatively similar but actually somewhat greater than that predicted with
values of g from Eq. (8.24).
The effect of polydispersity on J(~) is also consistent with a disentanglement
mechanism. The largest molecules in the system contribute the longest relaxation
times to the linear viscoelastic spectrum, and entanglements are lost from large
molecules at lower shear rates. The long relaxation time processes are therefore
removed, progressively reducing the breadth of the relaxation time distribution
with increasing shear rate and tending thereby to reduce the apparent compliance
function J(~). If shear-induced narrowing of the relaxation time distribution
dominates in polydisperse systems, the function J(~) will decrease with shear
rate, as observed. Tanaka, Yamamoto and Takano (353) have developed a theory
for predicting non-linear behavior from flow-induced changes in the relaxation
spectrum. They concluded that the behavior of J(~) depends on the shape of the
terminal linear viscoelastic spectrum, decreasing with shear rate when the spectrum is broad and increasing when it is narrow.
Such qualitative agreement should not be interpreted too broadly in favor of
disentanglement however. Sakai et al. (199) have also observed that J(~) increases
with ~ even in narrow distribution systems of lower concentration, that is,
W. W. Graesstey:
150
"
W.......
10
t...
--Z._
Do
_
-w-
i',,vO--
- ~
. . . .
,,T~
v,--
I
10
16'
"i'~o 1
I
-w
t0
-1
I
~fTo
,I,
10
10 2
Fig. 8.16. Superimposed plots of q(7)#lo and j(~)/jo as functions of shear rate for
poly(ct-methyl styrene) solutions (199). Curve A is for data at low concentrations and
molecular weights; curve B is for data at high concentrations and molecular weights~
Data have been shifted along the shear rate axis to produre superposition. [Reproduced from
Macromolecules 5, 791 (1972).]
where jo still conforms to the Rouse expression; in this case the argument
leading to Eq. (8.33) is no longer valid. Also, certain dilute solution models, such
as F E N E dumbbells, predict an increase in J(~) with shear rate,
Rouse model:
J(~) = constant
(all 9),
(8.34)
Entanglement model:
j(~)~2/11
(';% >> 1),
(8.35)
(~% >> 1).
(8.36)
F E N E dumbbell model
(338): j(~)~z/3
Although the F E N E model has been shown earlier to be inappropriate for polymer
systems on other grounds, it nevertheless illustrates an instance in which J(~)
increases with shear rate naturally and without the need for postulating structural
changes in the system.
151
8.4. Second Normal Stress Function

Values of P22 - - P 3 3 --= N 2 appear to be negative and approximately 10-30%
of Nx in magnitude (82). The conventional bead-spring models yield N2--=0. Indeed,
N2 in steady shear flow is identically zero for all free draining models, regardless
of the force-distance law in the connectors (102a). Thus, finite extensibility and,
by inference at least, internal viscosity do not in themselves provide non-zero
N 2 values. Bird and Warner (354) have recently analyzed the rigid dumbbell model
with intramolecular hydrodynamic interaction, the latter represented by the Oseen
approximation. In this case N2 turns out to be non-zero but positive.
The bead-connector models have only been used in calculations for systems
at infinite dilution, and N2 has been measured in moderately concentrated
solutions. It is therefore possible that N2 is associated only with intermolecular
effects. The Lodge and Yamamoto network models (Section 6) introduce a kind
of intermolecular interaction, but they too yield N 2 - 0. The Williams model
(219) includes intermolecular effects in another manner. In this case N 2 is
negative (although changing sign at high shear rates), but much larger than
Nx, the latter approaching zero as ~;4, contrary to all observations. Thus, there
is no compelling reason, molecular or continuum, to expect N2 to have any
special magnitude or sign with respect to Nt.
It is interesting to examine the bead-spring models to see what flow-induced
configurational changes would be required in order to develop N2 values of the
proper magnitude and sign. In the Rouse model, the components of the stress
tensor are related directly to averages of the internal coordinates of the beads.
For the simplest case of the elastic dumbbell:
12vkT
Pij = -- PoOij +
(r2)
[(XiXj)-- (XiXi)o]
(8.37)
in which x1, x2, and x3 are the bead components relative to the center of
gravity of the dumbbell, v is the number of dumbbells per unit volume, (r 2) is
the equilibrium mean square bead separation distance, and (xixj) and (x~xj}o
are averages evaluated during flow and at equilibrium respectively.
(XiXj)
(xixj}o=
I~I xixj ~ d x l d x 2 d x 3 '
(8.38)
(r 2}
12
(8.39)
----0
(i=j),
(i ~:j).
152
W.W. Graessley:
The normal stress functions are
Xa-
t2vkT
(r2) [(xl2 ) - (x273,
(8.40)
N2=
12vkT
(rZ> [(x2z ) - (x2>].
(8.41)
The positive value of N1 thus reflects an expansion of the distribution of

configurations in the flow direction, relative to the distribution in the direction
of the velocity gradient. Negative N2 values imply an effective contraction of the
distribution along the direction of the velocity gradient relative to the 3 direction.
In the Rouse model the distribution of configurations in the 2 and 3
directions are independent of flow in the 1 direction, so N2 is identically zero.
Thus, one might speculate than N2 would b%negative if the frictional coefficient
of a bead depended upon xz such that ~ decreased as tx21 increased. Such an
effect could result from a systematic depletion of entanglements from segments
with large values of lx21. In Graessley's theory (227), the ratio hi9 is literally the
effective reduction in (xz2) for the frictional resistances of the coil due to a
systematic loss of entanglements at its periphery. Thus, with Eq.(8.41), one
might write
N2 = - vkT(1 - h/9)
(8.42)
and, for ~zo ~ 1, using Eqs.(8.24) and (8.25),
N 2 : - v k T [ 4 [ ' ~ z l 3] oc -[~]3.
2 IJ
(8.43)
The sign of N a is correct but the dependence on t~l3 at low shear rates is rather
contrary to expectations.
Wales and Philippoff (345a) have suggested recently that negative values of
N2 may be caused by differences in segmental diffusion coefficients parallel and
transverse to the chain direction. The beads in the spring-bead models are
conventionally treated as spheres, so no directional bias is imposed. If the chain
segments are non-spherical, and the parallel and transverse diffusion coefficients
of the beads, Dp and Dr,are calculated from the average spatial orientation of the
segments, finite values of N2 are obtained:
N2 = Dt-Dp
N1
5 D t + Dp
(8.44)
153
When the segments behave as long cylinders, Dp = 2D~, and:
N~
--=
N1
- 1/8.
(8.45)
Even considering this rather interesting suggestion of Wales and Philippoff
(354a), the molecular origins of N2 remain uncertain at the present time. One
might hope that experimental data on this property in well-characterized
systems, particularly on its behavior at low shear rates, will in the future
provide some guidance.
8.5. Non-linear Response in Time-dependent Simple Shear Flows

Studies have been made of the stresses produced in several non-steady flow
histories. These include the buildup to steady state of tr and Pl 1 -P22 at the onset
of steady shearing flow (355-358); relaxation of stresses from their steady state
values when the flow is suddenly stopped (356-360); stress relaxation after
suddenly imposed large deformations (361); recoil behavior when the shear stress
is suddenly removed after a steady state in the non-linear region has been
reached (362); and parallel or transverse oscillations superimposed on steady
shearing flow (363-367). Experimental problems caused by the inertia and
compliance of the experimental apparatus are much more severe than in steady
state measurements (368, 369). Quantitative interpretations must therefore still be
somewhat tentative. Nevertheless, the pattern of behavior emerging is suggestive
with respect to possible molecular flow mechanisms.
8.5.1. Stress Development at the Onset of Steady Shearing Flow (355-358)
At sufficiently low shear rates the viscosity is constant. If a constant shear
rate in this range is imposed, the shear stress should grow monotonically to its
steady state value tr(oo) according to the equation from linear viscoelasticity:
zH(z) [1 - e -'/'] d In
a(t) = cr(oo) - ~
(8.46)
vH(T) d In
where
~(t) = o
(t < O)
~(t) = ~
(t > o)
G(oo)= ~o~-
154
W.W. Graessley:
At higher shear rates the steady-state viscosity a(oo)/~} begins to depend on ~.

If a constant shear rate in this range is imposed, the shear stress rises beyond
its steady state value, passes through a maximum before eventually approaching
the steady state, a(oo) = r/(). The stress at the maximum a(tm) grows in relation
to a(oo) as ~ increases. The time to reach the maximum 6, is found to be inversely
proportional to shear rate:
tm~ = constant.
(8.47)
The total shear at the stress maximum ~)m is Ctm, SO for a given polymer 7m is
independent of ~, and has a value which is typically in the range of 2.0--3.0
shear units. The normal stress displays a similar overshoot behavior, although
7m and the time to reach steady state appear to be greater for P11-22 than
for a.
8.5.2. Stress Decay at the Termination of Steady Shearing Flow (356-360)
At sufficiently low shear rates the shear stress should decay at the termination
of steady state shear flow according to the equation from linear viscoelasticity:
zH(z)e_~/~ d l n z
--00
a(t) = a(O)
(8.48)
zH(z) d In z
--00
where
f,(t)= ~
(t < o)
~(t) = 0
(t > O)
a(O) = r/o~.
Accordingly. plots of a(t)/a(O) vs t from different shear rates should superimpose. Experimentally the curves do not superimpose when the shear rate is in
the non-Newtonian region, the initial rate of relaxation being increasingly more
rapid for higher shear rates. The normal stress decays more slowly than shear
stress, but behaves similarly with respect to the effect of previous shearing flow
in the non-Newtonian region.
8.5.3. Stress Relaxation from Sudden Strains of Large Amplitude (361)
For shear strains of sufficiently small amplitude, the response is linear and
the shear stress is governed by the stress relaxation modulus:
a(t) = "~oG(O .
(3.9)
155
For shear strains greater than approximately 2 the ratio a(t)/~ o for a
concentrated polystyrene solution was reduced at all observable times. For the
large strains, relaxation proceeded more rapidly at short times, but at longer times
the residual stress decayed with about the same time dependence as that in the
linear viscoelastic region.
8.5.4. Shear Recovery after Steady Shearing Flow (362)
The total amount of shear recoil after steady state shear flow at sufficiently
low shear rates is related to jo:
7r = j o ao
(3.17)
where
a(t) = a0
(t < 0)
,~(t) = o
(t > o)
ao= rto~
When the shear rate at steady state is in the non-Newtonian region for the
fluid, the recoil is smaller than Ja o for both broad and narrow molecular
weight distributions (362).
8.5.5. Small Oscillations Superimposed on Steady Shearing Flow (363-367)

The dynamic moduli for infinitesimal superimposed deformations parallel
and transverse to the flow direction in steady shearing flow should be unaffected
by flow if the shear rate is sufficiently small. According to the theory of simple
fluids, the superimposed dynamic moduli for shearing flows in the nonNewtonian region must change in order to conform with the relations (370, 371 :
lim G~(~,~n) _ ~/(~),

~-,0
(8.49)
(D
lim G',((~,co)
,(,[" [1 +
dlog~/]
(8.50t
The loss moduli at low frequencies are indeed reduced at high shear rates
(363-367), as implied by these equations. No conclusions about the behavior of
the storage moduli can be drawn from simple fluid theory. However, G'(~,o)) and
Gf~(~,o)) are also reduced at high shear rates, and Gf~(7,~o) may even become
negative at low frequencies (365).
156
W.W. Graessley:
8.5.6. Discussion
The slower rise and decay of normal stress transients compared to shear stress
arises quite simply and directly from the polydispersity of relaxation times (78),
and probably has no direct bearing on entanglement mechanisms per se.
Likewise, the depression of the superimposed moduli at low frequencies follows
from rather non-specific continuum models, the loss moduli by Eqs.(8.49)
and (8.50) from the simple fluid model, and the storage moduli from the
following properties of the more specific but still quite general BKZ model
(366, 372):
lim
2 Gi(7,~o)
N1
c2
- ~2 ,
';
N I [ 1 d logN1
lim 2G,' (.~o)
0)2
-- 2 [ "F dlog~
(8.51)
1] .
(8.52)
According to Eq.(8.52), negative values of GI, at low frequencies occur if N1

becomes proportional to a power of shear rate less than one, a condition which
should be attainable at moderate shear rates in concentrated solutions of polymers
with narrow molecular weight distribution. Thus, any molecular theory which
predicts q and N1 as functions of shear rate, and which is also consistent with
the BKZ and simple fluid theories should automatically yield superimposed
moduli which satisfy Eqs.(8.49)-(8.52) without special assumptions about entanglement mechanisms.
Stress overshoot at the onset of steady shearing flow and accelerated
relaxation at its termination are properties which are not automatic consequences
of general continuum theories. They can be interpreted as resulting from a
partial decoupling of the system which begins at a characteristic shear strain
at the onset of rapid shearing flow, and a history-dependent re-establishment of
the equilibrium degree of coupling after flow is stopped. In the integral continuum
models this behavior corresponds to a flow history dependence in the memory
function. The behavior appears to be inherently unsymmetrical in that the initial
decouplin9 depends explicitly on the extent of strain, while re-couplin9 depends only
on time. The initial relaxation of a sheared system is rapid because the system is
partially decoupled and therefore presents a reduced resistance to configurational
rearrangement. As relaxation proceeds, the structure returns to its equilibrium
state, the coupling builds up and the resistance to further relaxation increases.
One might expect the last traces of stress to fade at essentially the linear
viscoelastic rate for the system, although evidence on this point is still
unclear (373).
Lodge and Meissner have recently examined in detail the stresses during
start-up of steady etongational and shear flows (374). The Lodge equation
[Eq.(6.15)], with its flow-independent memory function, described the build-up
of stress rather well (even for rapid deformations) until a critical strain was
reached. Beyond the critical strain (which differed somewhat in shear and
157
elongational deformations), a history-dependent memory function was required.

Similar behavior has been noted in the rheologically more complicated property
of die-swell. Bagley and co-workers have studied the effect of capillary length on
the ratio of extrudate diameter to capillary diameter at constant flow rate (375).
For short capillaries the extrudate diameter is large, but it decreases with
increasing length and appears to approach a limiting value for long capillaries.
They demonstrated that the criterion for reaching the asymptotic diameter at
each shear rate was not the residence time in the capillary, but rather the total
amount of deformation experienced during passage through the capillary.
Stratton and Butcher have studied the effect of interrupted flow on response
(376). If a steady flow is stopped and then restarted quickly, the overshoot is
missing. For longer periods of rest the overshoot gradually reappears, showing
that during rest the equilibrium structure of the system is slowly re-established.
Great care is necessary in introducing a dependence of memory function
relaxation times on the previous flow history. For example, if the relaxation times
are expressed simply in terms of past strain rates alone, then accelerated
relaxation at the termination of steady-state flow may indeed be accommodated
satisfactorily. However, stress data at the onset of flow appear to be fundamentally
inconsistent with this kind of modification (377). Also the ability of the modified
constitutive equation to handle small but rapid oscillatory deformations [such as
are used to measure G'(o)) and G"(o))] may be seriously compromised. Strain rate
magnitudes may grow very large for high frequencies even though strain
amplitudes remain small. Information on the properties and limitations of
integral constitutive equations is accumulating rapidly (378, 379), but a detailed
summary of current work is beyond the scope of the present review.
Ziabicki (380) has pointed out that flow-history dependence of the stress
translates in molecular terms to a flow-history dependence of the distribution
of configurations, with current stress depending directly only on the current
configuration. Strain criteria for evoking a history dependence in the memory
function imply that deformation in itself can alter or even obliterate the
contribution of past states to the current configuration. [Among current
constitutive equations of fairly specific form, the network-rupture model of
Tanner (381) corresponds most closely to this idea.] Flow-induced changes in
entanglement coupling or topological state serve as a useful concept for discussing
and rationalizing many aspects of non-linear behavior (382). The incorporation
of such ideas into a quantitative constitutive theory is of course an eagerly
sought goal.
9. Conclusions
The conclusions of this review can be separated into those which concern
the experimental theology of networks and concentrated systems of narrow
distribution polymers and those related to the ability of molecular theories to
rationalize and predict these observations.
158
W.W. Graessley:
Rheotogical behavior at low to moderate concentrations parallels that for

particulate suspensions in its dependence on the Simha interaction parameter c[t/].
The relative viscosity r/0/th is primarily a function of c[r/]. The steady state
compliance jo lies in the range predicted by the conventional spring-bead
models, with residual variations depending on c [t/]. The shear rate sensitivity of
viscosity, judged by the power law exponent, increases rapidly through this region
and depends principally on c[r/]. There seems to be little necessity to postulate
specific interactions such as chain entanglement to rationalize most behavior
in this regime.
In concentrated solutions and melts the behavior is controlled by Bueche's
interaction parameter cM and the magnitude of the local friction factor (0. The
appearance of a plateau in the relaxation modulus, the onset of M 34 dependence
in t/o, and the change in form of jo from M / c R T to 1/c2RT all depend on cM.
The characteristic molecular weights associated with G, t/0 and jo in the undiluted state--M~, Mr, and ME respectively--are related to each other in
approximately the same way for a number of different polymers. The characteristic
molecular weights are all inversely proportional to concentration and virtually
independent of temperature and the choice of solvent. With some exceptions the
values of Mo, associated with the average molecular weight between entanglements,
correspond to roughly the same contour length for different polymers. For those
relatively few cases in which the chemical structure of crosslinked networks has
been characterized in detail, the values of M~ obtained from the equilibrium
modulus are similar to values obtained from the plateau modulus for the same
polymer without crosslinks. All these properties, which point clearly to a
universal physical interaction between chain molecules, are documented rather
completely in non-polar, non-crystallizable polymers. Information is much less
complete for systems of highly polar, stiff, and crystallizable chains.
The factors affecting the transition between low and high concentration
behavior, taking place roughly in the range 10-20% polymer but depending
strongly on chain length and the nature of both polymer and solvent, have been
studied very little and are not well understood.
A number of suggestions about the nature of the entanglement interaction
have been put forward. The only acceptable interpretation appears to be that it is
a direct consequence of the uncrossability of chain contours. Uncrossability imposes
topological restrictions which reduce the number of paths available for configurational relaxation in a system of real chains, compared to a hypothetical system
of crossable (ghost) chains. The long relaxation times, those governing large
scale configurational change, are therefore larger in the system of real chains.
The same effect reduces the number of available configurations in deformed
networks of real chains. The result of such permanent topological classification
has been shown invariably to increase the equilibrium modulus, making it larger
than that of an equivalent network of ghost chains.
The large scale rope-like interactions pictured by Bueche are crude representations, but certainly more realistic than interactions which depend on local
kinks, rings, or other special sources of contact friction or molecular roughness.
We are however a long way from being able to associate entanglement effects
with specific attributes of the topological arrangement. It would be premature,
159
for example, to associate on a one-to-one basis each "equivalent crosslink" of the

topological modulus contribution with one of Bueche's interlacing loops.
Unfortunately, theories on the shift of relaxation times caused by entanglement--essentiaUy explanations of the M ~4 dependence of ~/o--are still
quite arbitrary and inelegant. The early qualitative ideas of Treloar appear to be
basically correct, but quantitative theories have not progressed far beyond
Bueche's original analysis of 20 years ago. Not surprisingly, the theory of the
terminal spectrum in highly entangled systems is also less than satisfactory. Very
little progress has been made beyond Ferry's early suggestion of a shifted Rouse
spectrum and the later Ch~Smpff-Duiser supporting theory. However, the behavior
of jo with molecular weight makes it clear that the terminal spectrum for long
monodisperse chains is considerably narrower than the Rouse form. Explanations
have been suggested, based on a presumed difference in the local velocity fields
in entangled and nonentangled systems, but complete calculations starting from
such ideas have not been made.
There is essentially no theoretical guidance about the structural variables that
control the onset of entanglement effects, whether a critical contour length, a
critical radius of gyration, or some critical combination of space-filling and
mechanical properties of the chain. This problem presents many deep analytical
ditficutties. Computer simulation may provide some help here in developing
intuition and suggesting some simple models for topological classification.
The relationship between chain entanglement and non-linear viscoelastic
properties is somewhat equivocal. Shear rate dependence in viscosity is a rather
general characteristic of long chain systems at all levels of concentration,
including infinite dilution. It cannot therefore be solely an attribute of interchain
interactions. Viscosity changes seem invariably to commence at shear rates which
are of the order of the reciprocals of the longest relaxation times in the linear viscoelastic spectrum. The magnitude of these relaxation times is governed by intermolecular interactions in all but the most dilute range of concentrations. These
properties are sufficient to eliminate potential intramolecular sources, such as
internal viscosity and finite extensibility, as primary causes of shear rate
dependence in concerned systems. The rapid increase in shear rate dependence
with concentration suggests a close relationship with entanglement density in
concentrated systems, and supports the idea that the degree of intermolecular
coupling in such systems is reduced at high shear rates. The behavior of several
other non-linear properties is also qualitatively consistent with the uncoupling
hypothesis, but a fairly complete constitutive equation based on this idea and
having real physical content is still lacking.
I0.
Acknowledgments
The author is grateful for the help of a number of people who have read
and criticized various portions of this review. These include Dr. Gary Ver Strate,
Prof. Arthur Lodge, Prof. Robert Simha, Prof. Michael Williams, Dr. Edward
160
W.W. Graessley:
Bagley, Prof. S.F. Edwards, Dr. Alfred Ch6mpff, and Dr. Toshiro Masuda.
Unpublished data, pre-publication manuscripts and helpful comments were also
provided by Prof. Guy Berry, Prof. G. Vinogradov, Prof. Shigeharu Onogi, Prof.
Mitsuru Nagasawa, Dr. Neal Langley, Prof. Robert Bird, Dr. J.L.S. Wales,
Dr. U. Daum, Prof. G. Allen, Prof. Donald Plazek, Prof. M. Kurata, Prof. J. Schurz,
Dr. Nobu Nakajima, Prof. Robert Stratton, and Dr. Edward Collins. Prof. John
Ferry's invitation to prepare the review, his technical comments during the course
of the work, and his patience while awaiting its completion are much appreciated.
Considerable help in data reduction and figure preparation was furnished by
Ms. Kathryn Graessley. Finally, the author is grateful for the support of the
National Science Foundation through grant GK-34362 during the time the
review was being prepared.
11. Nomenclature
Matrix in the Rouse theory
Exponent in Mark-Houwink equation: [t/] = K M a.
a
Exponent in Fox-Allen qo - M correlation.
a
Parameter of order unity in the continuous form of the Rouse spectrum.
Concentration exponent in jo and Zmcorrelations.
b
3 ro2/(r2).
b
Total number of crosslinks (Part 7).
C
Auto-correlation function for end-to-end vector.
c(t)
Moduli in the Mooney-Rivlin elasticity equation.
C1, C2
Cis, C81 Components of the Cauchy-Green strain tensor and its inverse, with
the current configuration as the reference configuration.
Polymer concentration in gm/ml.

Cornet critical concentration.
c*
Simha interaction parameter.
c[~]
Bueche interaction parameter.
cM
Self-diffusion coefficient.
D
Power law exponent in ~/~ t~1-~
d
Average number of entanglement points per molecule; E = M / M c - 1.
E
Value of E at ~ = 0 (Part 8).
Eo
Viscosity structure factor.
F
Frictional force.
F
Change in free energy with deformation.
AF
Nominal tensile stress, force/initial cross sectional area.
f
Fraction of chain pairs belonging to topological class i.
f,
Reduced steady shear viscosity function (r/- t/s)/(q0 - r/s).
f(fl)
Equilibrium shear modulus.
G
GO
Instantaneous shear modulus in stress relaxation.
Shear modulus in the viscoelastic plateau region.
G
Shear stress relaxation modulus.
G(t)
Shear storage modulus.
G'(o~)
A
a
161
[6%
Shear loss modulus.

Intrinsic strage mdulus' lim
c-~0 (G'(; c) )"
[G"]
Intrinsic lss mdulus' c~o

lim ( G"(tn' c)c~ ~~/s ) "
g
g
Pair correlation function in molecular theory of liquids (Part 6).

Front factor in modulus equation from rubber elasticity theory (Part 7).
Fraction of configurations of free chains which are consistent with
specified end-to-end coordinates (Part 7).
Fractional reduction in entanglement density due to steady shear flow;
g(O)=E/Eo (Part 8).
Relaxation time distribution.
Intramolecular hydrodynamic interaction parameter.
G"(oJ)
gl,g2
g(O)
H(~)
h
h*
h(O)
J(O
JoR
J*(O
Jff)
K
K
K(s)
k
k'
M
M
M(O
Mo
Mo
M~
~W
J~Z
/~z+l
M~
N
N
N.
Nt
U~
h/N 1/z.
Fractional reduction in energy dissipation rate per molecule due to
dis-entanglement in steady shear flow (Part 8).
Shear creep compliance.
Steady state recoverable shear compliance.
Reduced steady state compliance; JeR -----JcRTq~/Mfflo - qs)2.
Steady state compliance for a monodisperse polymer of molecular
weight 4.
N d2 ~2.
Unspecified proportionality constant.
Spring constant in the bead-spring models.
Bueche entanglement slip function.
Boltzmann constant.
Huggins constant.
Forsman coupling matrix.
Molecular weight.
Memory function in Lodge theory.
Molecular weight between entanglements in undiluted polymer.
Characteristic molecular weight from ~/0 vs M behavior of undiluted
polymers.
Characteristic molecular weight from j o v s M behavior of undiluted
polymers.
Number-average molecular weight.
Weight-average molecular weight.
z-average molecular weight.
z + 1-average molecular weight.
General designation for characteristic molecular weights in the rheological behavior of undiluted polymers.
Number of sub-molecules in spring-bead models.
Number of primary molecules (Part 7).
Avogadro's number.
First normal stress function, Px1 -P22 at steady state in steady simple
shear flow.
Second normal stress function, Pz2 - Pa3 at steady state in steady simple
shear flow.
162
n
n~
h i , n2
Po
Pij
Rg
Ro
r
r0
(r2>
S
AS
s
To
u
V
v
u
Uo
wg
X
o~
t~0
~o
%
~o
o
q'
W. W. Graessley:
Number of main chain atoms.
Number of main chain atoms between entanglements.
Moles of solvent and polymer respectively in solution thermodynamics
(Part 2).
Degree of polymerization.
Unspecified isotropic pressure term in stress tensor p for incompressible
materials.
Component of stress tensor in rectangular coordinates.
Universal gas constant, kN~.
Gel point radiation dose (Part. 2).
Stokes radius.
Position vector.
Contour length of polymer chain.
Mean-square end-to-end distance of polymer chain.
Mean radius of gyration, (S 2) 1/2
Entropy change with deformation.
Bueche slip factor.
Absolute temperature.
Langley entanglement trapping factor.
Unit vector.
Pervaded volume of polymer coil.
Velocity vector.
Speed, velocity magnitude.
Volume per main chain atom in undiluted polymer.
Gel fraction.
Structure parameter S2/vo in Fox-Allen r/0 - M correlation.
Coil expansion ratio (Parts 2 and 5).
Extension ratio in tensile deformation (Part 7).
Reduced shear rate, (tl o - tls) M ~ / c R T.
Reduced frequency, (tl o - tl~) M o)/ cR T.
Characteristic reduced shear rate locating the onset of shear rate
dependence in the viscosity.
Extent of simple shear deformation from rest state.
Crosslink index, fraction of mers participating in crosslinks multiplied
by DPn of primary chains (Part 7).
Instantaneously imposed shear deformation.
Shear rate.
Characteristic shear rate locating the onset of shear rate dependence
in the viscosity.
Dirac delta function.
Kronecker delta function.
Parameter characterizing the internal viscosity of chain molecules
(Part 8).
Frictional coefficient.
Frictional coefficient per main chain atom.
Frictional coefficient associated with an entanglement junction.
Steady state shear viscosity, a(,~)/?;.
Dynamic viscosity, G'(o))/e).
~o
~rn
163
Absolute value of the complex viscosity [G'((D)2 + G"((D)z]I/2/(D.

Viscosity at zero shear rate.
Viscosity of solvent.
Viscosity of monomeric fluid in Eyring's theory.
Value of viscosity in undiluted polymer at M = M~.
Tn
Intrinsic viscosity, lim[ ~/- ~h ].

c--0[ ~hc 1
Intrinsic viscosity at zero shear rate (Part 8).
Theta condition for a polymer-solvent system.
Argument (~,%/2) (e/r/o) in 9(0) and h(O) functions of Graessley's theory
(Part 8).
Eigenvalues of transformation matrices.
Intermolecular distances in Eyring's viscosity theory (Part 6).
Chain concentration, molecules per unit volume.
Concentration of elastically effective strands in crosslinked network
(Part 7).
Twice the concentration of entanglement junctions in a system prior to
crosslinking (Part 7).
Concentration of primary molecules prior to crosslinking (Part 7).
Twice the concentration of crosslinks in a system (Part 7).
Polymer density, mass/volume.
Shear stress in simple shear flow.
Relaxation time.
Characteristic relaxation time associated with the onset of shear rate
dependence in the viscosity.
Characteristic "maximum" relaxation time determined from the
terminal region of the viscoelastic spectrum.
Number-average relaxation time of the terminal viscoelastic region,
Tw
Weight-average relaxation time of the terminal viscoelastic region,
E ]o
0
0
2i
2 1 , 2 2 , 23
V
"e
PO
Ve
Q
G
"C
~0
~m
Z
Z
Zo
%
%
f2
(D
(D
(DO
Volume fraction of polymer.

Flory constant, [rl]M/(r2) 3/2 for linear flexible chains at high
molecular weights.
Polymer-solvent interaction coefficient (Part 2).
Extinction angle in flow birefringence.
Effective polymer-solvent interaction coefficient in determining chain
dimensions in concentrated systems.
Probability density distribution function for bead positions in the
spring-bead molecular models.
First normal stress coefficient, NI/~ 2.
Second normal stress coefficient, N2/~ 2.
Number of distinguishable 'configurations.
Frequency (D, rad/sec.
Number of distinguishable configurations available to a free chain
(Part 7).
Characteristic frequency at which r/'(co) begins to depart from r/0.
164
W.W. Graessley:
12. References
1. Busse, W.F.: The physical properties of elastic colloids. J. Phys. Chem. 36, 2862-2879
(1932).
2. Treloar, L.R.G.: Elastic recovery and plastic flow in raw rubber. Trans. Faraday
Soc. 36, 538-549 (1940).
3. Flory, P.J.: Network structure and the elastic properties of vulcanized rubber. Chem.
Rev. 35, 51-75 (1944).
4. Green,M. S., Tobolsky, A.V.: A new approach to the theory of relaxing polymers media.
J. Chem. Phys. 14, 80-92 (1946).
5. Buchdahl,R.: Rheology of thermoplastic materials. I. Polystyrene. J. Colloid Sci. 3,
87-98 (19481.
6. Nielsen, L.E., Buchdahl,R.: Viscoelastic and photoelastic properties of polystyrene
above its softening temperature. J. Colloid Sci. 5, 282-294 (1950).
7. Buech6, F.: Viscosity, self-diffusion and allied effects in solid polymers. J. Chem. Phys.
20, 1959-1964 (1952).
8. Bueche, F.: Viscosity of polymers in concentrated solutions. J. Chem. Phys. 25,
599~500 (1956).
9. Fox, T.G., Flory, P.J.: Further studies on the melt viscosity of polyisobutylene. J.
Phys. Colloid Chem. 55, 221-234 (1951).
10. Fox, T.G., Loshaek, S.: Isothermal viscosity-molecular weight dependence for long
chains. J. Appl, Phys. 26, 1080-1082 (1955).
11. Ferry,J. D., Landel, R. F., Williams, M. L.: Extensions of the Rouse theory of viscoelastic
properties to undiluted linear polymers. J. Appl. Phys. 26, 359-362 (1955).
12. Lodge, A. S.: A network theory of flow birefringence and stress in concentrated polymer
solutions. Trans. Faraday Soc. 52, 120-130 (1956).
13. Yam amoto, M.: The visco-elastic properties of network structure. I. General formalism.
J. Phys. Soc. Japan 11,413-421 (1956).
13a. Lodge, A. S.: The isotropy of gaussian molecular networks and the stress-birefringence
relations for rubberlike materials cross-linked in stressed states. Kolloid-Z. 171,
46-51 (1960).
13b. Bagley, E. B., West, D.C.: Chain entanglement in non-newtonian flow. J. AppI. Phys.
29, 1511-15t2 (1958).
14. Porter,R. S., Johnson, J. F.: The entanglement concept in polymer systems. Chem. Rev.
66, 1-27 (1966).
15. Ferry, J. D.: Viscoelastic properties of polymers, 2nd ed. New York: Wiley 1970.
16. Berry, G.C., Fox, T.G.: The viscosity of polymers and their concentrated solutions.
Adv. Polymer Sci. 5, 261-357 (1968).
17. Dusek, K., Prins, W.: Structure and elasticity of non-crystalline polymer networks.
Advan. Polymer Sci. 6, 1-102 (1969).
18. Janeschitz-Kriegl, H.: Flow birefringence of elastico-viscous polymer systems. Advan.
Polymer Sci. 6, 170-318 (1969).
19. Berry, G.C., Casassa, E.F.: Thermodynamic and hydrodynamic behavior of dilute
polymer solutions. J. Polymer Sci. Pt. D: Macromol. Rev. 4, 1-66 (19701.
20. Fixman, M., Stockmayer, W. H.: Polymer conformation and dynamics in solution. Ann.
Rev. Phys. Chem. 21,407-428 (1970).
21. Orofino, T.A.: Dilution-solution properties of polystyrene in 0-solvent media. II. An
analysis of environmental effects. J. Chem. Phys. 45, 4310-4315 (1966).
22. Flory, P.J.: Statistical mechanics of chain molecules. New York: Wiley-Interscience
1969.
23. Fixman, M., Peterson, J. M.: Dimensions of polymer molecules in concentrated solutions.
J. Am. Chem. Soc. 86, 3524-3529 (1964).
24. Yamakawa, H.: Modern theory of polymer solutions, pp. 214-215. New York: Harper
and Row 1971.
25. Bluestone, S., Vold, M.J.: Monte-Carlo calculations of the dimensions of coiling type
polymers in solutions of finite concentration. J. Polymer Sci. Pt. A 2, 289-301 (1964).
26. Curro, J. G. : Computer simulation of semi-dilute polymeric systems. Paper presented at
1973 March Meeting of the American Physical Society, San Diego.
165
27. Flory, P.J. :Principles of polymer chemistry. Ithaca: Cornetl University Press, 1953.
28. Krigbaum,W.R., Godwin, R.W.: Direct measurement of molecular dimensions in bulk
polymers. J. Chem. Phys. 43, 4523-4524 (1965).
29. Cotton,J.P., Farnoux, G., Jannink, G., Mons,J., Picot, C.: Observation de la conformation de la chaine polymerique darts la solide anaorphe par diffusion de neutrons aux
petits angles. Compt. Rend. Set. C. 275, 175-179 (1972).
30. Ballard,D. G. H., Wignall, G.D., Schelten, J.: Measurement of molecular dimensions of
polystyrene chains in the bulk polymer by low angle neutron scattering. European
Polymer J. 9, 965-969 (1973).
31. Kirste, R.G., Kruse, W.A., Schelten, J.: Die Bestimmung des Tr~igheitsradius von Polymethyl-methacrylat im Glaszustand durch Neutronenbeugung. Makromol. Chem. 162,
299-303 (1972).
32. Schelten, J., Kruse, W.A., Kirste, R. G.: Uber die Gestalt von Polymethylmethacrylat im
Glaszustand. Kolloid-Z. u. Z. Polymere 251, 919-921 (t973); see also J. Appl.
Crystallography 7, 188 (1974) and Polymer (London) (in press) by the same authors.
33. Bueche, F., Cashin,W.M., Debye, P.: The measurement of self-diffusion in solid
34. Brandrup, J., Immergut, E. H. (Eds.): Polymer handbook. New York: Interscience 1966.
35. Bueche,F, Kinzig, B.J., Coven, C. J.: Polymer Chain Conformations in Bulk Polymers.
Polymer Letters 3, 399-402 (1965).
36. Smith,T. L., Frederick,J. E.: Ultimate tensile properties of elastomers. IV. Dependence
of the failure envelope, maximum extensibility, and equilibrium stress-strain curve on
network characteristics, J. Appl. Phys. 36, 2996-3005 (1965).
37. Hoffman, M.: Zur Konformation von Molektilen in unverdtinnten nichtkristallinen
Polymeren. Makromol. Chem. 144, 309-321 (1971).
38. Maron, S. H., Nakajirna, N., Krieger, I. M.: Study of entanglement of polymers in solution
by viscosity measurements. J. Polymer Sci. 37, 1-18 (1959).
39. Maron, S.H.: A theory of the thermodynamic behavior of non-electrolyte solutions.
J. Polymer Sci. 38, 329-342 (1959).
40. Marn' S"H" Nakajima'N': A thery f the thermdynamic behavir f nn electrlyte
solutions. II. Application to the system benzene-rubber. J. Polymer Sci. 40, 59-71
(1959).
41. Maron, S. H., Nakajima, N.: A theory of the thermodynamic behavior of nonelectrolyte
solutions. III. The osmotic pressure of polymer solutions. J. Polymer Sci. 42, 327-340
(1966).
42. Maron, S. H., Nakajima, N.: A theory of the thermodynamic behavior of nonelectrolyte
solutions. V. The scattering of light by polymer solutions. J. Polymer Sci. 47, 157-168
119601.
43. Aharoni, S. M.: Segmental density distribution of linear amorphous polymer molecules.
J. Macromol. Sci.-Phys. BT, 73-103 (1973).
44. Vollmert, B., Stutz, H.: Zur Struktur yon konzentrierten PolymerlSsungen und Gelen.
Angew. Makromol. Chem. 20, 71-10t (1971). A description in English of these and later
results by the same workers can be found on pp. 548-561 of Vollmert, B.: Polymer
chemistry. Berlin-Heidelberg-New York: Springer 1973.
45. Duportail, G., Froelich, D., Weill, G.: Fluorescence quenching as a tool in the study of
the distribution of local concentrations in polymer solutions. I. Polymers bearing
quenching side groups. European Polymer J. 7, 977-987 (t971).
46. Alberino, L. M., Graessley,W. W.: Cross-linking of polystyrene by high-energy radiation.
II. Molecular weight changes in the pre-gel region. J. Phys. Chem. 72, 4229-4235
(1968).
47. Budzot, M, Dole, M.: The radiation chemistry of polyethylene. XI. The molten state.
J. Phys. Chem. 75, 1671-1676 (1971).
48. Mandelkern,L.: Radiation chemistry of linear polyethylene, Vol. I. pp. 287-334.
In: Dole, M. (Ed.): The radiation chemistry of macromolecules. New York: Academic
Press 1972.
49. Flory, P.: Molecular configuration in bulk polymers. Pure Appl. Chem. (in press).
50. Kargin,V. A.: The role of structural phenomena in the formation of polymer properties.
Pure Appl. Chem. 12, 35-61 (1966).
166
W.W. Graessley:
51. Tager, A.A., Drevat,V.E., Lutsky, M.S., Vinogradov,G.V.: Rheotogicat behavior of

concentrated polystyrene solutions in good and poor solvents. J. Polymer Sci. Pt. C23,
181-192 (1968).
52. Tager, A.A., Dreval, V. E.: Non-newtonian viscosity of concentrated polymer solutions.
Rheol. Acta 9, 517-524 (1970).
53. Quadrat, O.: Viscometric properties of the concentrated polymer solutions. II.
Dependence of the non-newtonian viscosity on the thermodynamic quality of the
solvent and on its viscosity. Collection Czech. Chem. Commun. 37, 3039--3041 (1972).
54. Debye, P., Chu, B., Woermann, D.: Viscosity of Critical Mixtures. J. Polymer Sci. Part
A 1, 249-254 (1963).
55. VerStrate, G., Philippoff, W.: Phase separation in flowing polyiner systems. Paper
presented at 1973 March meeting of American Physical Society, San Diego.
56. Flory, P.J.: Statistical thermodynamics of semi-flexible chain molecules. Proc. Roy. Soc.
(London) A234, 60-73 (1956).
57. Robertson, R.E.: Polymer order and polymer density. J. Phys. Chem. 69, 1575-1578
(1965).
58. Yeh,G.~S.Y., GeiI, P.N,: Crystallization of polyethylene terephthalate from the glassy
amorphous state. J. Macromol. Sci. B1, 235-249 (1967).
59. Carr, S. H., Geil, P. H., Baer, E. :The development of spherulites from structural units in
glassy poly(bisphenol-A-carbonate). J. Macromol. Sci.-Phys. B2, 13-28 (1968).
60. Lin,W., Kramer, E.J.: Small-angle X-ray scattering from amorphous potycarbonate. J.
Appl. Phys. 44, 4288-4292 (1973).
61. Yeh,G. S. Y.: Morphology of amorphous polymers. CRC Critical Rev. Macromol. Sci.
1, 173-213 (1972).
62. Yeh,G. S. Y.: Order in amorphous polystyrenes as revealed by electron diffraction and
diffraction microscopy. J. Macromol. Sci.-Phys. B6, 451-464 (1972).
63. Yeh, G.S.Y.: A structural model for the amorphus state of polymers: folded-chain
fringed micellar grain model. J. Macromol. Sci. Phys. B6, 465-478 (1972).
64. Hoeve, C.A.J., O'Brien, M.K.: Specific diluent effects on polymer chain dimensions.
J. Polymer Sci. Part A 1, 1947-t954 (1963).
65. Gent, A. N.: Stress-induced birefringence of swollen polymer networks. Macromolecules
2, 262-265 (1969).
66. Tonelli, A. E.: On the existence of local segmental order in undiluted bulk polymers.
J. Chem. Phys. 53, 4339-4345 (1970).
67. Wellinghoff, S.T., Baer, E.: Crazing in ultra-thin films of atactic polystyrene. Unpublished manuscript.
68. Berens,A.R., Folt,V.L.: Resin particles as flow units in poly(vinyl chloride) melts.
Trans. Soc. Rheol. 11, 95-111 (1967).
68a. Collins, E. A., Krier, C. A.: Poly(vinyl chloride) melt theology and flow activation energy.
Trans. Soc. Rheol. 11, 225-242 (1967).
69. Schreiber, H. P., Rudin, A., Bagley, E. B.: Separation of elastic and viscous effects in polymer melt extrusion. J. Appl. Polymer Sci. 9, 887--892 (1965).
70. Ballman, R.L., Rademacher, L.E., Farnharn, W. H.: Visualization of polystyrene melt
flow. Presented at 1972 August U.S.-Japan joint seminar on polymer processing and
rheology, Knoxville, Tenn. See J. Appl. Polymer Syrup. 20 (1973).
71. Mooney, M., Wolstenholme, W.E.: The rheological unit in raw elastomers. J. Appl.
Phys. 25, 1098-1101 (1955).
72. Mooney, M.: Theory of the non-newtonian rheology of raw rubbers consisting of
supermolecular rheological units. J. Appl. Phys. 27, 691-696 (1956).
73. Prichard,J.H., Wissbrun, K.F.: Reversible melt flow rate increase in branched acetal
polymers. J. Appl. Polymer Sci. 13, 233-239 (1969).
74. Schreiber, H.P., Storey, S.H., Bagley, E.B.: Molecular fractionation in the flow of
polymeric fluids. Trans. Soc. Rheol. 10, 275-297 (1966).
75. Whitlock, L.R., Porter, R. S. : Experimental investigation of the concept of molecular
migration within sheared polystyrene. J. Polymer Sci., Pt. A-2 10, 877-886 (1972).
76. Col eman, B. D., Notl,W.: Recent results in the continuum theory of viscoelastic fluids.
Ann. N. Y. Acad. Sci. 89, 672-714 (1961).
t67
77. Coleman, B.D., Notl, W.: Simple fluids with fading memory, pp. 530--552. In: Proc. of
the International Symposium on Second-Order Effects in Elasticity, Plasticity, and Fluid
Dynamics, Haifa. New York: Pergamon Press 1962.
78. Lodge, A. S.: Elastic liquids. New York: Academic Press 1964.
79. Coleman, B.D., Markovitz, H., NolI,W.: Viscometric flows of non-newtonian fluids.
Berlin-Heidelberg-New York: Springer 1966.
80. Coleman,B.D., Noll,W.: Foundations of linear viscoelasticity. Rev. Mod. Phys. 33,
239-249 (1961).
81. Gross, B.: Mathematical structure of the theories of viscoelasticity. Paris: Hermann et
Cie. 1953.
82. Tanner, R.I.: A correlation of normal stress data for polydisobutylene solutions.
Trans. Soc. Rheol. 17, 365-373 (1973).
83. Coleman,B.D., Markovitz, H.: Normal stress effects in second-order fluids. J. Appl.
Phys. 35, 1-9 (1964).
84. Stratton, R.A., Butcher, A.F.: Experimental determination of the relationships among
various measures of fluid elasticity. J. Polymer Sci. Pt, A-2 9, 1703-1717 (1971).
85. Debye, P., Bueche, A.M.: Intrinsic viscosity, diffusion, and sedimentation rate of polymers in solution. J. Chem. Phys. 16, 573-579 (1948).
86. Kramers, H.A.:The behavior of macromolecules in inhomogeneous flow. J. Chem.
Phys. 14, 415-424 (1946).
87. Kirkwood, J.G.: Collected works Macromolecules, Auer, P.L. (Ed.). New York:
Gordon and Breach 1967.
88. Wang, M. C., Uhlenbeck, G. E.: On the theory of brownian motion. II. Rev. Mod. Phys.
17, 323 (1945). In: Wax,N. (Ed.): Selected papers in noise and stochastic process.
New York: Dover Publications.
89. Boyer,R.F.: The relation of transition temperature to chemical structure in high
polymers. Rubber Chem. Tech. 36, 1303-1421 (1963).
90. Clark, M.B., Zimm, B.H.: A linearized chain model for dielectric loss in polymers.
ACS Polymer Preprints 12, 116-120 (1971). See also: Tobolsky, A.V., DuPr6,D.B.:
Macromolecular relaxation in the damped torsional oscillator and statistical segment
models. Advan. Polymer Sci. 6, 103-127 (1969).
91. Iwata, I., Kurata, M.: Brownian motion of lattice-model polymer chains. J. Chem.
Phys. 50, 4008 4013(1969}.
92. Orwoll, R.A., Stockmayer, W. H.: Stochastic models for chain dynamics. Advan. Chem.
Phys. 15, 305-324 (1969).
93. Massa, D. J., Schrag, J. L., Ferry, J. D.: Dynamic viscoelastic properties of polystyrene in
high-viscosity solvents. Extrapolation to infinite dilution and high-frequency behavior.
Macromotecutes 4, 210-214 (1971).
9~ Peterlin,A.: Molecular model of internal viscosity. Rheol. Acta 12, 496-502 (1973).
95. Bazua, E.R., Williams, M.C.: A molecular formulation of the internal viscosity in
polymer dynamics, and stress symmetry. J. Chem. Phys. 59, 2858-2868 (1973).
96. Helfand, E.: Theory of the kinetics of conformational transitions in polymers. J. Chem.
Phys. 54, 4651-4661 (t971).
97. Rouse, P. E.: A theory of the linear viscoelastic properties of dilute solutions of coiling
polymers. J. Chem. Phys. 21, I272--1280 (I953).
98. ]3ueche, F. :The viscoelastic properties of plastics. J. Chem. Phys. 22, 603-609 (1954).
99. Chapter 10, Appendix B of Ref. (27).
100. Zimm, B.H.: Dynamics of polymer molecules in dilute solution: viscoelasticity, flow
birefringence and dielectric loss. J. Chem. Phys. 24, 269-278 (1956).
101. L odge, A. S., Wu,Y.-J.: Constitutive equations for polymer solutions derived from the
bead/spring model of Rouse and Zimm. Rheol. Acta 10, 539-553 (1971).
102. Bird, R.B., Warner, H.R.,Jr., Evans, D.C.: Kinetic theory and rheology of dumbbell
suspensions with brownian motion. Advan. Polymer Sci. 8, 1-90 (1971).
102a. Curtiss, C.F., Bird, R. B., Hassager0 O.: Kinetic theory and rheology of dilute solutions
of macromolecules. ACS Polymer Preprints 15, 22-27 (1974).
168
W.W. Graessley:
103. Verdier, P.H., Stockmayer, W.H,: Monte Carlo calculations on the dynamics of
polymers in dilute solution. J. Chem. Phys, 36, 227-235 (1962). See also:
Verdier, P.H.: Monte Carlo studies of lattice-model polymer chains. I. Correlation
functions in the statistical-bead model. J. Chem. Phys. 45, 2118-2121 (1966).
104. Tschoegl,N.W.: Influence of hydrodynamic interaction on the viscoelastic behavior of
dilute polymer solutions in good solvents. J. Chem. Phys. 40, 473-479 (1964),
105. Pyun, C.W., Fixman, M.: Intrinsic viscosity of polymer chains. J. Chem. Phys. 42,
3838-3844 (1965).
106. Thurston, G.B., Morrison, J.D.: Eigenvalues and the intrinsic viscosity of short
gaussian chains. Polymer (London) 10, 421-438 (1969).
107. Lodge, A. S., Wu, Y.-J.: Exact relaxation times and dynamic functions for dilute polymer
solutions from the bead/spring model of Rouse and Zimm, Report 16. Rheology
Research Center, University of Wisconsin (July, 1972).
108. Williams, M.C.: Normal stresses in polymer solutions with remarks on the Zimm
treatment. J. Chem. Phys. 42, 2988-2989 (1965); errata J. Chem. Phys. 43, 4542 (1965).
109. Ferry, J.D., Williams, M.L., Stern, D. M.: Stow relaxation mechanisms in concentrated
polymer solutions. J. Phys. Chem. 58, 987-992 (1954).
110. Verdier, P.H.: Monte Carlo studies of lattice model polymer chains. II. End-to-end
length. J. Chem. Phys. 45, 2t22-2128 (1966).
111. Berry, G.C.: Thermodynamic and conformational properties of polystyrene. II.
Intrinsic viscosity studies in dilute solutions of linear polystyrenes. J. Chem. Phys.
46, 1338-1352 (1967).
112. Tanaka, H., Sakanishi, A., Kaneko, M., Furiuchi,J.: Dynamic viscoelastic properties of
dilute polymer solutions. J. Polymer Sci. Pt. C 15, 31%330 (1966).
113. Schrag, J. L., Johnson, R. M.: Application of the Birnboim multiple lumped resonator
principle to viscoelastic measurements of dilute macromolecular solutions. Rev. Sci.
Inst. 42, 224-232 (1971).
114. Johnson, R. M., Schrag, J. L., Ferry, J. D.: Infinite-dilution viscoelastic properties of polystyrene in 0-solvents and good solvents. Polymer J. (Japan) 1, 742--749 (1970).
115. Osaki, K., Mitsuda, Y., Johnson,R. M., Schrag, J. L., Ferry, J. D.: Infinite-dilution viscoelastic properties of linear and star-shaped polybutadienes. Macromol. 5, 17-19 (1972).
116. Osaki, K., Schrag, J.L., Ferry, J. D. : Infinite-dilution viscoelastic properties of poly(7methylstyrene). Applications of Zimm theory with exact eigenvalues. Macromotecutes
5, 144-147 (1972).
116a. Osaki, K.: A revised version of the integrodifferential equation in the Zimm theory for
polymer solution dynamics. Macromolecules 5, 141-144 (1972).
117. Osaki, K., Schrag, J.L.: Viscoelastic properties of polymer solutions in high-viscosity
solvents and limiting high-frequency behavior. I. Polystyrene and poly(c~-methylstyrene). Polymer J. (Japan) 2, 541-549 (t971).
t18. Thurston, G.B., Peterlin, A.: Influence of finite numbers of chain segments, hydrodynamic interaction, and internal viscosity on intrinsic birefringence and viscosity of
polymer solutions in an oscillating laminar flow field. J. Chem. Phys. 46, 4881-4885
(1967).
118a. Osaki, K.: Viscoelastic properties of dilute polymer solutions. Advan. Polymer Sci. 12,
1~i4 (1973).
119. Simha~R., Somcynsky.T.: The viscosity of concentrated spherical suspensions. J. Colloid
Sci. 20, 278-281 (1965).
120. Batchelor, G. K., Green, J.T.: The determination of the bulk stress in a suspension of
spherical particles to order c z. J. Fluid Mech. 56, 401427 (1972).
I21. Gandhi, K.S., Williams, M.C.: Solvent effects on the viscosity of moderately concentrated polymer solutions. J. Polymer Sci. Pt. C 35, 211-234 (1971).
122. Frederick, J. E., Tschoegl, N.W., Ferry, J. D.: Dynamic mechanical properties of dilute
polystyrene solutions; dependence on molecular weight, concentration, and solvent.
J. Phys. Chem. 68, 1974-1982 (1964).
123. Holmes, L. A., Kusamizu, K., Osaki, K., Ferry, J. D. : Dynamic mechanical properties of
moderately concentrated polystyrene solutions. J. Polymer Sci. Pt. A-2 9, 2009-2021
(1971).
169
124. Onogi, S., Masuda, T., Kitagawa, K.: Rheological properties of anionic polystyrenes.
I. Dynamic viscoelasticity of narrow-distribution polystyrenes. Macromolecules 3,
109-116 (1970).
125. Frisch, H.L., Simha, R.: The viscosity of colloidal suspensions and macromolecular
solutions, Vol. 1, Chapter 14. In: Eirich,F.R. (Ed.): Rheology. New York: Academic
Press 1956.
126. Simha,R., Zakin, J.L.: Compression of flexible chain molecules in solution. J. Chem.
Phys. 33, 179t-1793 (1960).
127. Onogi, S., Kobayashi,Y., Kojima, Y., Taniguchi, Y.: Non-Newtonian flow of concentrated solutions of high polymers. J. Appl. Polymer Sci. 7, 847-859 (1963).
128. Cornet, C.F.: The determination of unperturbed dimensions of polymer molecules by
viscometry of moderately concentrated solutions. Polymer (London) 6, 373-384 (1965).
129. Utracki, L., Simha,R.: Corresponding states relations for the viscosity of moderately
concentrated polymer solutions. J. Polymer Sci. Pt. A 1, 1089-1098 (1963).
130. Simha, R., Utracki, L. A.: The viscosity of concentrated polymer solutions: corresponding states principles. Rheol. Acta 12, 455462 (1973).
131. Simha, R., Chan, F.S.: Corresponding states relations for the newtonian viscosity of
concentrated polymer solutions. Temperature dependence. J. Phys. Chem. 75, 256-267
(1971).
132. Onogi, S., Kimura, S., Kato, T., Masuda, T., Miyanaga,N.: Effects of molecular weight
and concentration on flow properties of concentrated polymer solutions. J. Polymer
Sci. Pt. C 15, 381-406 (1966).
133. Onogi, S., Masuda, T, Miyanaga, N., Kimura, Y.: Dependence of viscosity of concentrated polymer solutions upon molecular weight and concentration. J. Polymer Sci.
Pt. A-2 5, 899-913 (1967).
134. Gandhi, K.S., Williams, M.C.: Effect of solvent character on polymer entanglements.
J. Appt. Polymer Sci. 16, 2721-2725 (1972).
135. Dreval,V.Ye., Tager, A.A., Fomina, A.S.: Concentrated solutions of polymers. IV.
Viscosity of solutions of polystyrene in various solvents. Polymer Sci. USSR 5, 495-507
(1964) [Vysokomolekul. Soyedin. 5, 1404 (1963).
136. Ferry, J.D., Foster, E.L., Browning, G.V., Sawyer,W.M.: Viscosities of concentrated
polyvinyl acetate solutions in various solvents. J. Colloid Sci. 6, 377-388 (1951).
137. Simha, R., Zakin,J.L.: Solution viscosities of linear flexible high polymers. J. Colloid
Sci. 17, 270-287 (1962).
138. Quadrat, O., Podnecka,J.:Influence of the thermodynamic quality of a solvent upon
the viscosity of moderately concentrated polystyrene solutions. Collection Czech. Chem.
Commun. 37, 2402-2409 (1972).
139. Johnson, M. F., Evans,W. W., Jordan, I., Ferry, J. D.: Viscosities of concentrated polymer
solutions. II. Polyisobutylene. J. Colloid Sci. 7, 498-510 11952).
140. Ferry,J.D., Grandine, L.D.,Jr., Udy,D.C.: Viscosities of concentrated polymer solutions. III. Polystyrene and styrene-maleic acid copolymer. J. Colloid Sci. 8, 529-539
(1953).
140a. Toms, B.A., Strawbridge, D.J.: Elastic and viscous properties of dilute solutions of
polymethyl methacrylate in organic liquids. Trans. Faraday Soc. 49, 1225-1232 (1953).
141. Dreval,V.Y2., Malkin,A.Ya., Vinogradov, G.V.: Effect of the solvent nature on the
rheological properties of concentrated solutions of various polymers. European Polymer
J. 9, 85-99 (1973).
142. Masuda, T., Toda, N., Aoto, Y., Onogi, S.: Viscoelastic properties of concentrated solutions of poly(methyl methacrylate) in diethyl phthalate. Polymer J. (Japan) 3, 315-321
(1972).
143. Klein, J., Woernle, R.: Zum EinfluBder Polymer-L/Ssungsmittd-Wechselwirkung auf das
FlieBverhalten m~iBig konzentrierter Polymerl6sungen. Kolloid-Z. Z. Polymere 237,
209-219 (1970).
144. Borchard, W., Pyrlik, M., Rehage, G.: Association phenomena of PMMA in solution
and gels. Makromol. Chem. 145, 169-188 (1971).
145. Rotne, J., Prager, S.: Variational treatment of hydrodynamic interaction in polymers.
J. Chem. Phys. 50, 4831~4837 (1969).
t70
W.W. Graessley:
146. Masuda,T., Kitagawa, K., Onogi, S.: Viscoelastic properties of poly (methyl methacrylates) prepared by anionic polymerization. Polymer J. (Japan) 1, 418-424 (1970).
147. Onogi, S., Masuda, T., Ibaragi, T.: Rheological properties of polymethyl methacrylate
and potyvinyt acetate in the molten state. Kolloid-Z. Z. Polymere 222, 110-124
(1968).
t48. Ninomiya, K., Ferry,J.D., Oyanagi, Y.: Viscoelastic properties of potyvinyl acetates.
II. Creep studies of blends. J. Phys. Chem. 67, 2297-2308 (1963).
149. Peterlin, A.: Gradient and time dependence of viscosity of polymer solutions in very
viscous solvents. J. Lubrication Tech. 90, 571-576 (1968).
150. Lodge, A.S.: Rheological properties of concentrated polymer solutions. I. Growth of
pressure fluctuations during prolonged shear flow. Polymer (London) 2, 195-201 (1961).
151. Philippoff, W.: Studies of flow birefringence of polystyrene solutions. Trans. Soc. Rheot.
7, 45-59 (1963).
t52. Osaki, K., Einaga, Y.: Viscoelastic properties of concentrated polymer solutions. Prog.
Polymer Sci. Japan 1, 321-375 (1971).
153. Odani, H., Nemoto, N., Kurata, M.: The viscoelastic properties of undiluted linear
polymers of narrow molecular weight distribution in the terminal zone. Bull. Inst.
Chem. Res., Kyoto U. 50, 117-133 (1972).
154. Allen, V.R., Fox,T.G.: Viscosity-molecular weight dependence for short chain potystyrenes. J. Chem. Phys. 41, 337-343 (1964).
155. Graessley, W.W., Hazleton,R.L., Lindeman, L.R.: The shear-rate dependence of viscosity in concentrated solutions of narrow-distribution polystyrene. Trans. Soc.
Rheol. 11, 267-285 (1967).
156. Gupta, D., Forsman, W.C.: Newtonian viscosity-molecular weight relationship for
concentrated solutions of monodisperse polystyrene. Macromolecules 2, 304-306
(1969).
157. Berry, G. C. :The viscosity of polymer-diluent mixtures. J. Phys. Chem. 70, 1194-1198
(1966).
158. Fetters, L.J.: Determination of the intermolecular entanglement coupling spacings in
polyisoprene by viscosity measurements. J. Res. Natl. Bur. Std. 69A, 33-37 (1965).
159. Schurz, J., Hochberger, H.: Untersuchungen iJber den L6sungszustand yon m~il3ig
konzentrierten Polyisobutyten-L6sungen. Makromol. Chem. 96, 141-149 (1966).
160. Cross,M. M.: Viscosity, molecular weight, and chain entanglement. Polymer (London)
11, 238-244 (1970).
161. Fujimoto, T., Ozaki,N., Nagasawa, M.: Stress relaxation of monodisperse poly-~methylstyrene. J. Polymer Sci. Pt. A-2 6, 129-t40 (1968).
162. Tobotsky, A.V.: Properties and structure of polymers. New York: Wiley 1960.
i63. Plazek, D.J., O'Rourke, V.M.: Viscoelastic behavior of low molecular weight polystyrene. J. Polymer Sci. Pt. A-2 9, 209-243 (1971).
164. Barlow, A.J., Day, M., Harrison, G., Lamb,J., Subramanian, S.: Viscoelastic relaxation
in a series of polyethylacrylates and poty-n-butylacrylates. Proc. Roy. Soc. A 309,
497-525 (1%9).
165. Knoff, W.F., Hopkins, I.L., Tobolsky, A.V.: Studies on the stress relaxation of polystyrenes in the rubbery flow region. II. Macromolecules 4, 750-754 (1971).
166. Nemoto, N., Moriwaki, M., Odani, H., Kurata, M.: Shear creep studies of narrowdistribution poly(cis-isoprene). Macromolecules 4, 215-219 (1971).
167. Nemoto, N., Ogawa, T., Odani, H., Kurata, M.: Shear creep studies of narrow-distribution poly(cis-isoprene). III. Concentrated solutions. Macromolecules 5, 641-644
(1972).
168. Oser, H.. Marvin, R. S.: Effect of molecular weight on viscoelastic properties of polymers
as predicted by a molecular theory. J. Res. Natl. Bur. Std. 67B, 87-90 (1963).
169. Andrews, R.D., Tobotsky, A.V.: Elastoviscous properties of potyisobutylene. IV. Relaxation time spectrum and calculation of bulk viscosity. J. Polymer Sci. 7, 221-242
(1951).
170. De Witt, T.W., Markovitz, H., Padden, F.J., Jr., Zapas, L.J.: Concentration dependence of the rheological behavior of the polyisobutylene--decalin system. J. Colloid.
Sci. 10, 174-188 (t955).
171
171. Tobolsky, A.V., Aklonis, J.J., Akovali, G.: Viscoelastic properties of monodisperse
polystyrene. J. Chem. Phys. 42, 723-728 (1965).
172. Kotaka, T., Kurata, M., Tamura, M.: Non-Newtonian flow and normal stress phenomena in solutions of polystyrene in toluene. Rheol. Acta 2, 179-186 (1962).
173. Kotaka,T., Osaki, K.: Normal stresses, non-Newtonian flow, and dynamic mechanical
behavior of polymer solutions. J. Polymer Sci. Pt. C 15, 453-479 (1966).
174. Tung, L.H., Runyon,J.R.: Molecular weight distribution of standard polystyrene
samples by GPC and by sedimentation velocity analysis. J. Appl. Polymer Sci. 17,
1589-1596 (1973).
175. Holmes, L.A., Ferry, J.D.: Dependence of the steady-state compliance on concentration
and molecular weight in polymer solutions. J. Polymer Sci. Pt. C 23, 291-299 (1968).
176. Kusamizu, S., Holmes, L.A., Moore, A.A., Ferry, J.D.: The steady-state compliance of
polymer solutions. Trans, Soc. Rheol. 12, 559-571 (1968).
177. Ashare, E.: Rheological properties of monodisperse polystyrene solutions, Ph.D.
Thesis, University of Wisconsin (1968). Trans. Soc. Rheol. 12, 535-557 (1968).
178. Graessley,W.W,, Segal, L.: Flow behavior of polystyrene systems in steady shearing
flow. Macromolecules 2, 49-57 (1969).
179. Daum, U.: (Private communication), see Ref. (190).
180. Wales, J. L. S.: (Private communication), see Ref. (190).
181. Prest, W.M., Jr.: Viscoelastic properties of blends of entangled polymers, J. Polymer
Sci: Pt. A-2 8, 1897-1908 (1970).
182. Einaga, Y., Osaki, K., Kurata, M., Tamura, M.: Creep behavior of polymer solutions.
II. Steady-shear compliance of concentrated polystyrene solutions. Macromolecules
4, 87-92 (1971).
t83. Tschoegl, N. W.: Stress relaxation and creep in dilute polymer solutions. J. Chem. Phys.
44, 2331-2334 (1966).
184. Akers, L. C., Williams, M.C.: Oscillatory normal stresses in dilute polymer solutions.
J. Chem. Phys. 51, 3834-3841 (1969).
185. Stratton,R. A., Butcher, A. F.: Experimental determination of the relationships among
various measures of fluid elasticity. J. Polymer Sci.: Pt A-2 9, 1703-1717 (1971).
186. Osaki, K., Einaga, Y., Kurata, M., Tamura, M.: Creep behavior of polymer solutions.
I. A new kind of apparatus for creep and creep recovery. Macromolecules 4,
82-87 (1971).
187. Sakai, M., Fujimoto,T.,Nagasawa, M.: Steady flow properties of monodisperse polymer
solutions. Molecular weight and polymer concentration dependences of steady shear
compliances at zero and finite shear rates. Macromolecules 5, 786-792 (1972).
188. Akovali, G.: Viscoelastic properties of polystyrene. J. Polymer Sci. Pt.A-2 5,
875-889 (1967).
189. Mieras, H. J. M. A., van Rijn, C. F. H.: Elastic behavior of some polymer melts. Nature
218, 865-866 (1968).
190. Daum, U., Wales,J.L.S.: Critical concentrations in polystyrene solutions. Polymer
Letters 7, 459-462 (1969).
191. Nemoto, N.: Viscoelastic properties of narrow-distribution polymers. II. Tensile creep
studies of polystyrene. Polymer J. (Japan) 1, 485-492 (1970).
192. Mills, N.J., Nevin,A.: Oscillatory shear measurements on polystyrene melts in the
terminal region. J. Polymer Sci. Pt. A-2 9, 267-281 (1971).
193. Crawley, R.: Unpublished results in M. S. thesis. Northwestern University, 1973.
194. Zosel, A.: Der EinfluB von Molekulargewicht und Moleculargewichtsverteilung auf
die viskoelastischen Eigenschaften von Polystyrolschmelzen. Rheol. Acta 10, 215-224
(1971).
195. Uy,W.C., Graessley, W.W.: Viscosity and normal stresses in poly(vinyl acetate)
systems. Macromolecules 4, 458-463 (1971).
196. Nemoto, N., Odani, H., Kurata, M.: Shear creep studies of narrow distribution poly(cisisoprene). II. Extension to low molecular weights. Macromolecules 5, 531-535 (1972).
197. Vinogradov, G. V., Malkin, A. Ya., Yanovskii, Yu. G., Borisenkova, E. K., Yarlykov, B. V.,
Berezhnaya, G.V.: Viscoelastic properties and flow of narrow distribution polybutadienes and polyisoprenes. J. Polymer Sci. Pt. A-2 10, 106t-1084 (1972).
172
W.W. Graessley:
198. Endo, H., Fujimoto, T., Nagasawa, M.: Normal stress and shear stress in a viscoelastic
liquid under steady state flow: Effect of molecular weight heterogeneity. J. Polymer Sci.
Pt. A-2 9, 345-362 (1971).
199. Sakai, M., Fujimoto,T., Nagasawa, M.: Steady flow properties of monodisperse polymer
solutions. Molecular weight and polymer concentration dependences of steady shear
compliances at zero and finite shear rates. Macromolecules 5, 786-792 (1972).
200. Odani, H., Nemoto.N., Kitamura, S., Kurata, M.: Viscoelastic properties of monodisperse polymers. I. Tensile creep studies of poly(e-methylstyrene). Polymer J. (Japan)
1, 356-364 (1970).
201. Osaki, K., Sakato, K., Fukatsu, M., Kurata, M., Matusita, K., Tamura, M.: Normal
stress effect in dilute polymer solutions. III. Monodisperse poly-~-methylstyrene in
chlorinated biphenyl. J. Phys. Chem. 74, 1752-1756 (1970).
202. Valentine, R.H.: Viscoelastic properties of linear and cross-linked polybutadienes.
Ph.D. thesis, University of Wisconsin, 1967.
203. Kraus, G.: Private communication.
204. Mills, N.J.:The rheological properties and molecular weight distribution of polydimethylsiloxane. European Polymer J. 5, 675-695 (1969).
205. Ninomiya, K., Ferry, J.D.: Viscoelastic properties of polyvinyl acetates. I. Creep
studies of fractions. J. Phys. Chem. 76, 2292-2296 (1%3).
206. Oyanagi,Y., Ferry, J.D.: Viscoelastic properties of polyvinyl acetates. IV. Creep
studies of plasticized fractions. J. Colloid Sci. 21, 547-559 (t966).
207. Mieras, H.J.M.A.: Elastic or normal-stress behavior of monodisperse polystyrene
melts or solutions. Paper presented at the conference Advances in Rheology, Glasgow,
September 1969.
208. Prest, W.M., Jr., Porter, R.S., O'Reilly, M.M.: Non-Newtonian flow and the steadystate shear compliance. J. Appl. Polymer Sci. 14, 2697-2706 (1970).
209. Masuda, T., Kitagawa, K., Inoue,T., Onogi, S.: Rheological properties of anionic polystyrenes. II. Dynamic viscoelasticity of blends of narrow-distribution polystyrenes.
Macromolecules 3, 116-125 (1970).
210. Saeda, S., Yotsuyanagi, J., Yamaguchi, K.: Relation between melt flow properties and
molecular weight in polyethylene. J. Appl. Polymer Sci. 15, 277-297 (1971).
210a. Locati, G., Gargani, L.: Dependence of zero-shear viscosity on molecular weight
distribution. J. Polymer Sci., Polymer Letters 11, 95-101 (1973).
211. Pennline, H.W., Graesstey, W.W.: Flow properties of polymethyl methacrylate solutions.
Paper presented at American Physical Society Meeting, San Diego, March 1973.
2t2. Graessley, W.W.:Linear viscoelasticity in entangling polymer systems. J. Chem. Phys.
54, 5143-5157 (1971).
213. Leaderman, H., Smith,R.G., Williams,L.C.: Rheology of polyisobutytene. Iit. Elastic
recovery, non-Newtonian flow, and molecular weight distribution. J. Polymer Sci.
36, 233257 {1959).
214. Ninomiya, K., Ferry, J.D.: Phenomenological relation for viscoelastic properties of
polymer diluent systems; the relative dissipation index N21. J. Macromol. Sci.-Phys.
B 3, 237-258 (1969).
215. Bogue, D.C., Masuda, T., Einaga, Y., Onogi, S.: A constitutive model for molecular
weight and concentration effects in polymer blends. Polymer J. (Japan) 1,563-572 (1970).
216. Chikahisa, Y.: A theory on the relationship between viscosity and molecular weight in
bulk polymers. J. Phys. Soc. Japan 19, 92-100 (1964).
217. Williams, M.C.: Stresses in concentrated polymer solutions. Part I. Shear dependence
of viscosity. A.I.Ch.E.J. 12, 1064-1070 (1966).
218. Williams, M. C.: Concentrated polymer solutions. Part II. Dependence of viscosity and
relaxation time on concentration and molecular weight. A.I.Ch.E.J. 13, 534-539
(1967).
219. Williams, M.C.: Concentrated polymer solutions. Part III. Normal stresses in simple
shear flow. A.I.Ch.E.J. 13, 955-961 (1967).
220. Fixman, M.: Dynamics of polymer chains. J. Chem. Phys. 42, 3831-3837 (1965).
221. Middleman, S.: The flow of high polymers. New York: Interscience 1968.
173
222. Lodge, A.S.: Constitutive equations from molecular network theories for polymer
solutions. Rheol. Acta 7, 379-392 (1968).
223. Flory, P.J.: Elasticity of polymer networks cross-linked in state of strain. Trans.
Faraday Soc. 56, 722-743 (1960).
224. Bueche, F.: Viscosity of polymers in concentrated solution, J. Chem. Phys. 25, 599-600
(1956). See also: Bueche, F.: Viscosity of molten branched polymers and their concentrated solutions. J. Chem. Phys. 40, 484-487 (1964).
225. DeGennes, P.G.: Reptation of a polymer chain in the presence of fixed obstacles.
J. Chem. Phys. 55, 572-579 (1971).
226. Eyring, H., Ree, T., Hirai, N.: The viscosity of high polymers - - Random walk of a
group of connected segments. Proc. Natl. Acad. Sci. 44, 1213-1217 (1959).
227. Graessley, W.W.: Viscosity of entangling polydisperse polymers. J. Chem. Phys. 47,
1942-1953 (1%7).
228. Aharoni, S.M.: On some rheoiogical phenomena of amorphous polymers. J. Appl.
Polymer Sci. 17, 1507-1518 (1973).
229. Ziabicki, A., Takserman-Krozer, R.: General dynamic theory of macromolecular networks. I. Definitions and classification. J. Polymer Sci. Part A-2 7, 2005-2018 (1969).
230. Takserman-Krozer,R., Ziabicki, A.: General dynamic theory of macromolecular networks. II. Dynamics of network deformation. J. Polymer Sci. Part A-2 8, 321-332
(1970),
231. Ziabicki, A.: Molecular rheology of polymer systems. Pure Appl. Chem. 26, 481-497
(1971).
232. Ch6mpff, A.J., Duiser, J.A.: Viscoelasticity in networks consisting of crosslinked or
entangled macromolecules. I. Normal modes and mechanical spectra. J. Chem. Phys.
45, 1505-1514 (1966).
233. Duiser, J.A., Staverman, A.J.: On the theory of rubber elasticity. In: Prins,J.A. (Ed.):
Physics of non-crystalline solids, pp. 376-387. Amsterdam: North Holland Publ. 1965.
234. Ch6mpff, A.J.: Normal modes of branched polymers. I. Simple ring and star-shaped
molecules. J. Chem. Phys. 53, 1566-1576 (1970).
235. Ch6mpff, A.J., Prins, W.: Viscoelasticity of networks consisting of crosslinked or
entangled macromolecules. II. Verification of the theory for entanglement networks.
J. Chem. Phys. 48, 235-243 (1968).
236. Edwards, S. F., Grant, J.W.V.: Effect of entanglement on diffusion in a polymer melt.
J. Phys. A: Math. Nucl. Gen. 6, 1169-1185 (1973).
237. Edwards, S.F., Grant, J.W.V.: The effect of entanglement on the viscosity of a polymer
melt. J. Phys. A, Math. Nucl. Gen. 6, 1186-1195 (1973).
238. Forsman,W.C., Grand, H.S.: Theory of entanglement effects in linear viscoelastic
behavior of polymer solutions and melts. I. Symmetry considerations. Macromolecules
5, 289-293 (1972).
239. Wolkowicz,R.I., Forsman,W,C.: Entanglement in concentrated solutions of polystyrene with narrow distributions of molecular weight. Macromolecules 4, 184-192
(1971).
239a. Thirion, P.: Empirical extension of the molecular theory of Rouse to the viscoelasticity of bulk polymers. J. Polymer Sci., Polymer Letters Ed. 11, 673-678 (1973).
240. Hoffman, M.: Das Grundgesetz fiir die mechanische Relaxation und das FlieBen
unvernetzter Fadenmolekiile. Rheol. Acta 6, 92--100 (1967).
241. Graessley, W.W.: Unpublished calculations.
242. Vinogradov, G. V., Pokrovsky, V. H., Yanovsky, Yu. G.: Theory of viscoelastic behavior
of linear polymers in unimolecular approximation and its experimental verification.
Rheol. Acta 11, 258-274 (1972).
243. Hayashi, S.: Theory of viscoelasticity in temporarily crosslinked polymers. IV. Relaxation spectrum and molecular weight dependence of viscosity. J. Phys. Soc.
Japan 19, 2306-2312 (1964).
244. Hayashi, S.: Concentration dependence of rheological properties in concentrated
polymer solutions. In: Onogi, S. (Ed.): Proceedings of the Fifth International Congress
of R heology, Vol. 4, pp. 179-190. Baltimore: University Park Press 1969.
174
W.W. Graessley:
245. Fox, T.G., Allen, V.R.: Dependence of the zero-shear melt viscosity and the related
friction coefficient and critical chain length on measurable characteristics of chain
246. Hoffmann, M.: Der Einflug yon Verhakungen auf die mechanische Relaxation und das
Fliel3en der amorphen Polymeren. Rheol. Acta 6, 377-390 (1967).
247. ToneUi, A.E.: A molecular approach to chain entanglement in linear polymers.
J. Polymer Sci. Part A-2 8, 625-635 (1970).
248. Pechhold, W.: Molekiitbewegung in Polymeren. I. Teil: Konzept einer Festk6rperphysik
makromolekularer Stoffe. Koltoid-Z. Z. Polymere 228, 1-38 (1968).
249. Bueche, F.: Melt viscosity of polymers: effect of polydispersity. J. Polymer Sci. 43,
527-530 (1960).
250. Prest,W.M., Jr., Porter,R.S.: The effect of high-molecular-weight components on the
viscoelastic properties of polystyrene. Polymer J. (Japan) 4, t 54-t62 (1973).
251. Case, L.C., Wargin, R.V.: Elastomer behavior. IV. The loop structure of elastomer
networks. Makromol. Chem. 77, 172-184 (1964).
252. Alfrey, T.,Jr., Lloyd, W.G.: Network polymers. I. Theoretical remarks. J. Polymer Sci.
62, 159--165 (1962).
253. Moore, C. G., Watson, W. F.: Determination of degree of crosslinking in natural rubber
vulcanizates. Part II. J. Polymer Sci. 19, 237-254 (1956).
254. Bueche, A.M.: An investigation of the theory of rubber elasticity using irradiated
poly(dimethylsiloxanes). J. Polymer Sci. 19, 297-306 (1956).
255. Langley, N. R.: Elastically effective strand density in polymer networks. Macromolecules
1, 348-352 (1968).
256. Treloar, L.R.G.: The physics of rubber elasticity, 2nd ed. London: Oxford University
Press t958.
256a. Smith, K.J., Jr.: Theories of chain coiling, elasticity, and viscoelasticity. In: Jenkins,
A.D. (Ed.): Polymer science, Chapter 5. Amsterdam: North-Holland Publ. Co. 1972.
257. Mark, J.E.: An experimental comparison of the theories of elasticity of polymer
networks. J. Am. Chem. Soc. 92, 7252-7257 (1970).
258. Rivtin, R.S.: Large elastic deformations. In: Eirich, F.R. (Ed.): Rheology, Vol. 1,
pp. 351-385. New York: Academic Press 1956.
259. tmai, S., Gordon, M.: Rubber elasticity. J. Chem. Phys. 50, 3889-3903 (1969); also note
correction regarding 9 on p. 9 of Ref. (284).
260. Edwards, S.F.: The statistical mechanics of rubbers. In: Ch6mpff, A.J., Newman, S.
(Eds.): Polymer networks, structure, and mechanical properties, pp. 83-110. New York:
Plenum Press 1971.
261. Freed, K. F.: Statistical mechanics of systems with internal constraints: rubber elasticity.
J. Chem. Phys. 55, 5588-5599 (1971).
262. Eichinger, B.E.: Elasticity theory. I. Distribution functions for perfect phantom
networks. Macromolecules 5, 496-505 (1972).
263. Guth, E.:Recent developments in the statistical mechanics of polymers. J. Polymer Sci.
Pt. C 31, 267-274 (1970).
264. Labana, S. S., Newman, S., Chtimpff, A. J.: Chemical effects on the ultimate properties of
polymer networks in the glassy state, pp. 453477. In: Polymer networks, structure and
mechanical properties. See Ref. (260).
265. Dusek, K.: Inhomogeneities induced by crosslinking in the course of crosslinking
copolymerization, pp. 245-260. In: Polymer networks, structure, and mechanical
properties. See Ref. (260).
266. Goebel, J.C., Tobolsky, A.V.: Volume changes accompanying rubber extension.
Macromolecules 4, 208--209 (1971).
267. Krigbaum, W.R., Roe, R.-J.: Survey of theory of rubberlike elasticity. Rubber Chem.
Tech. 38, 1039-t067 (1965).
268. Valanis, K.C., Landel,R.F.: The strain-energy function of a hyperelastic material in
terms of extension ratios. J. Appl. Phys. 38, 2997-3002 (1967).
269. Thirion, P., Chasset, R.: Les rapports entre l'elasticit6 et la structure des r6seaux
macromol6cutaires: Esquisse d'une nouvette approche. R.G.C.P. 45, 859-866 (1968).
269a. Yu, C. U., Mark, J. E.: Evaluation of the theories of rubber-like elasticity using swollen
networks crosslinked at different degrees of dilution. Macromolecules 6, 751--754 (1973).
175
270. Allen, G., Kirkham, M.J., Padget, J., Price, C.: Thermodynamics of rubber elasticity at
constant volume. Trans Faraday Soc. 67, 1278-1292 (1971).
27 t. Gumbrell, S. M., Mullins, L., Rivlin, R. S.: Departures of the elastic behavior of rubbers
in simple extensions from the kinetic theory. Trans. Faraday Soc. 49, 1495-1506 (1953).
272. Van der Hoff, B. M.E.: The stress-strain relation of swollen rubbers. Polymer (London)
6, 397-399 (1965).
273. Mark, J.E., Flory, P.J.: Stress-strain isotherms for poly-(dimethylsiloxane) networks.
J. Appl. Phys. 37, 4635-4639 (1966).
274. Price, C., Allen, G., de Candia, F., Kirkham, M.C., Subramaniam, A.: Stress-strain
behavior of natural rubber vulcanized in the swollen state. Polymer (London) 11,
486-491 (1970).
275. Johnson, R.M., Mark, J. E.: Properties of poly(dimethylsiloxane) networks prepared in
solution, and their use in evaluating the theories of rubberlike elasticity. Macromolecules 5, 41-45 (1972).
276. Gent, A.N., Rivlin, R.S.: Experiments on the mechanics of rubber. II. The torsion,
inflation, and extension of a tube. Proc. Phys. Soc. (Pt. B) 65, 487-501 (1952).
277. Van der Hoff, B.M.E., Buckler, E.J.: Transient changes in topology and energy on
extension of polybutadiene networks. J. Macromol. Sci. (Chem.) A 1,747-788 (1967).
278. Schaefgen, J.R., Flory;P.J.: Multilinked polyamides. J. Am. Chem. Soc. 72, 689-701
(1950).
279. Kraus, G., Moczygemba, G.A.: Chain entanglements and elastic behavior in polybutadiene networks. J. Polymer Sci. Part A 2, 277-288 (1964).
280. Froelich, D., Crawford, D., Rozek, T., Prins, W.: Ideal network behavior of anlonically
prepared polystyrene gels. Macromolecules 5, 100-102 (1972).
280a. Allen, G., Holmes, P.A., Walsh, D.J.: To be published in Disc. Faraday Soc. See also:
Walsh, D. J., Allen, G., Ballard, G. Polymer (London) 15, 366072 (1974).
280b. Tonelli, A.E., Helfand, E.: Elastically ineffective crosslinks in rubbers. ASC Polymer
Preprints 15, 605-612 (1974).
281. Schultz, A.R.: Characterization of polymer networks, In: Mclntyre, D. (Ed.): Characterization of macromotecular structure, pub. 1573, pp. 389-405. Washington: Nat.
Acad. Sci. 1968.
282. Scanlan, J.: The effect of network flaws on the elastic properties of vulcanizates.
J. Polymer Sci. 43, 501-508 (1860).
283: Gordon, M., Kucharuk, S., Ward, T. C.: The statistics of elastically active network chains
and the efficiency of crosslinking in rubbers. Collection Czech. Chem. Commun.
35, 3252-3264 (1970).
283 a. Tobolsky, A.V., Carlson, D. W., Indictor, N.: Rubber elasticity and chain configuration.
J. Polymer Sci. 54, 175-192 (1961).
284. Gordon, M., Ward, T.C., Whitney, R.S.: Chemical and physical aspects of the three
stages in forming polymer networks. In: Polymer networks, structure, and mechanical
properties, pp. 1-23. See Ref. (260).
285. Mancke, R.G., Dickie, R.A., Ferry,J.D.: Dynamic mechanical properties of crosslinked rubbers. V. An approximate analysis of the modulus contributions of trapped
and untrapped entanglements. J. Polymer Sci., Part A-2 6, 1783-1789 (1968).
286. Kraus, G.: Quantitative characterization of polybutadiene networks. J. Appl. Polymer
Sci. 7, 1257-1263 (1963).
287. Pearson, D.S., Skutnik, B.J., Bohm, G.G.A.: Radiation crosslinking of elastomers.
I. Polybutadienes. J. Polymer Sci. Polymer Phys. Ed. 12, 925-939 (1974).
288. Bristow, G.M.: Relation between stress-strain behavior and equilibrium volume
swelling for peroxide vulcanizates of natural rubber and cis-l,4-polyisoprene. J. Appl.
Polymer Sci. 9, 1571-1578 (1965).
289. Flory, P.J., Rabjohn, N., Shaffer, M. C.: Dependence of elastic properties of vulcanized
rubber on the degree of crosslinking. J. Polymer Sci. 4, 225-245 (1949).
290. Mullins, L.: Determination of degree of crosslinking in natural rubber vulcanizate.
J. Appl. Polymer Sci. 2, 1-7 (1959).
291. St.Pierre, L.E., Dewhurst, H.A., Bueche, A.M.: Swelling and elasticity of irradiated
polydimethylsiloxanes. J. Polymer Sci. 36, 105-111 (1959).
176
W.W. Graessley:
292. Langley, N.R., Polmanteer, K. E.: The role of chain entanglements in rubber elasticity.
Am. Chem. Soc. Polymer Preprints 13, 235-240 (1972).
293. Langley, N.R,, Potmanteer, K.E.: The relation of elastic modulus to crosslink and
entanglement concentrations in rubber networks. J. Polymer Sci. Polymer Phys. Ed.
12, 1023-1034 (1974).
294. Langley, N.R., Ferry, J. D.: Dynamic mechanical properties of crosslinked rubbers. VI.
Poly(dimethylsiloxane) networks. Macromolecules 1, 353-358 (1968).
295. Kramer, O., Ty, V., Ferry, J. D.: Entanglement coupling in linear polymers demonstrated
by networks crosslinked in states of strain. Proc. Natl. Acad. Sci. 69, 2216-2218 (1972).
296. Kramer, O., Carpenter, R.L., Ty, V., Ferry, J.D.: Entanglement networks crosslinked in
states of strain, paper presented at Society of Rheology Meeting, Montreal, October
1973. Part of this work is described in Entanglement networks of 1,2-polybutadiene
crosslinked in states of strain. I. Cross-linking at 0 C. Macromolecules 7, 79---84 (1974),
by the same authors.
297. Baldwin, F. P., Ver Strate, G.: Polyolefin elastomers based on ethylene and propylene.
Rubber Chem. Techn. 45, 709-881 (1972).
298. Cohen, R.E., Tschoegl,N.W.: Dynamic mechanical properties of block copolymer
blends--a study of the effects of terminal chains in elastomeric materials. I. Torsion
pendulum measurements. Intern. J. Polymeric Mater. 2. 49-69 (t972); II. Forced
oscillation measurements, lbid 2, 205-223 (1973); III. A mechanical model for
entanglement slippage. 1bid (in press).
299. Kramer, O., Greco,R., Ferry, J.D.:Viscoelastic properties of butyl rubber networks
containing reptating polyisobutylene. Paper presented at the San Diego meeting of the
American Physical Society, March, 1973.
300. Edwards, S.F.: Statistical mechanics with topological constraints. I. Proc. Phys. Soc.
91, 513-519 (1967).
301. Schill, G.: Catenanes, rotaxanes and knots. New York: Academic Press 1971.
302. Prager, S., Frisch, H. L.: Statistical mechanics of a simple entanglement. J. Chem. Phys.
46, 1475-t483 (i%7).
303. Alexander-Katz, R., Edwards, S.F.: The statistical mechanics of entangled polymers.
J. Phys. A Gen. Phys. 5, 674-681 (1972).
304. Edwards, S.F., Freed, K.F.: The entropy of a confined polymer. I. J. Phys. A. Gen.
Phys. 2. 145-150 (1969).
305. Collins, R., Wragg, A.: The entropy of a confined polymer. It. J. Phys. A. Gen.
Phys. 2, 151--f56 (t969).
306. Eichinger, B.E.: Elastic theory. II. Matrix method for a confined circular chain. J.
Chem. Phys. 57, 1356-1357 (1972).
307. Peterlin, A.: Rheology of polymer systems. J. Elastoplastics 4, 112-130 (1972).
308. Ito, Y., Shishida, S.: Critical molecular weight for onset of non-newtonian flow and
upper newtonian viscosity of polydimethylsiloxane. J. Polymer Sci., Polymer Phys. Ed.
10, 2239-2278 (1972).
309. Philippoff, W.: Correlation of the elastic properties in steady-state flow and vibrational experiments. J. Appl. Phys. 36, 3033-3038 (1965).
310. Bueche, F.: Influence of rate of shear on the apparent viscosity of a---dilute polymer
solutions and b-bulk polymers. J. Chem. Phys. 22, 1570-1576 (1954).
311. Pao,Y.-H.: Dependence of intrinsic viscosity of dilute solutions of macromolecules on
velocity gradient. J. Chem. Phys. 25, 1294-t295 (1956).
312. Onogi, S., Masuda, T., Kitagawa, K.: Rheological properties of anionic polystyrenes. I.
Dynamic viscoelasticity of narrow-distribution polystyrenes. Macromolecules 3,
109-116 (1970).
313. Penwell, R. C., Graessley, W. W., Kovacs, A.: J. Polymer Sci. Polymer Phys. Ed. (in press).
314. Cox, W.P., Merz, E.H.: Correlation of dynamic and steady flow viscosities. J. Polymer
Sci. 28, 619-622 (1958).
315. Strattn'R" A': Nn'newtnian flw in plymer systems with n entanglement cupling'
Macromotecules 5, 304-310 (1972).
316. Harris, E. K., Jr.: Viscometric properties of polymer solutions and blends as functions
of concentration and molecular weight. Ph.D. thesis, University of Wisconsin, 1970.
177
317. Ito, Y., Shishido, S.: A modified Graessley theory for non-newtonian viscosity of polydimethylsiloxanes and their solutions. J. Polymer Sci. Polymer Phys. Ed. 12, 617-628
(1974).
318. Noda, I., Yamada, Y., Nagasawa, M.: The rate of shear dependence of the intrinsic
viscosity of monodisperse polymer. J. Phys. Chem. 72, 2890-2898 (1968).
319. Kotaka, T., Suzuki, H., Inagaki, H.: Shear-rate dependence of the intrinsic viscosity of
flexible linear macromolecules. J. Chem. Phys. 45, 2770-2773 (1966).
320. Suzuki,H., Kotaka, T., Inagaki, H.: Shear-rate dependence of the intrinsic viscosity of
flexible linear macromolecules. II. Solvent effect. J. Chem. Phys. 51, 1279-t285 (1969).
321. Yamaguchi, N.: Sugiura, Y., Okamo, K., Wada, E.: Non-newtonian viscosity and
excluded volume effect of dilute solutions of flexible high polymers. J. Phys. Chem.
75, 1141-1149 (1971).
322. Gruver, J. T., Kraus, G.: Rheological properties of polybutadienes prepared by n-butyllithium initiation. J. Polymer Sci. Pt. A 2, 797-810 (1964).
323. Lee, C.L., Polmanteer, K.E., King, E.G.: Flow behavior of narrow-distribution poly-"
dimethylsiloxane. J. Polymer Sci. Pt. A-2 9, t909-1916 (1970).
324. Stratton, R.A.: The dependence of non-newtonian viscosity on molecular weight for
monodisperse polystyrene. J. Colloid Sci. 22, 517-530 (1966).
325. PauI, D.R., St. Laurence, J. E., Troell, J. H.: Flow behavior of concentrated solutions of an
SBS block copolymer Polymer Eng. Sci. 10, 70-78 (1970).
326. Mendelson, R.A., Bowles, W.A., Finger, F.L.: Effect of molecular structure on polyethylene melt rheology. I. Low shear behavior. J. Polymer Sci. Pt. A-2 8, 105-126
(1970).
327. Kataoka, T.: Onset of non-newtonian flow: a correlation with an average molecular
weight. Polymer Letters 5, 1063-1068 (1967).
328. Vinogradov, G. V.: Flow and rubber elasticity ofpolymer systems. Pure and Appl. Chem.
26, 423-449 (1971).
329. Markovitz, H.: The reduction principle in linear viscoleasticity. J. Phys. Chem. 69, 671
(1965).
330. Markovitz, H.: (Private communication).
331. Subirana, J.A.: Solvent effects on the non-newtonian viscosity of dilute solutions of
flexible linear macromolecules. J. Chem. Phys. 41, 3852-3856 (1964).
332. Peterlin, A.: Gradient dependence of intrinsic viscosity of freely flexible linear
macromolecules. J. Chem. Phys. 33, 1799-1802 (1960).
333. Burow, S. P., Peterlin,A, Turner, D.T. : The upturn effect in the non-newtonian viscosity
of polymer solutions. Polymer (London) 6, 3547 (1965).
334. Fixman, M.: Polymer dynamics: non-newtonian intrinsic viscosity. J. Chem. Phys. 45,
793-803 (1866).
335. Cottrell, F.R., Merrill, E.W., Smith, K.A.: Conformation of potyisobutylene in dilute
solution subjected to a hydrodynamic flow field. J. Polymer Sci. Pt. A-2 7, 141%1434
(1969).
336. Cottrell, F.R., Merrill, E.W., Smith, K.A.: Intrinsic viscosity and axial extension ratio of
random-coiling macromolecules in a hydrodynamic flow field. J. Polymer Sci. Pt.
A-2 8, 289-294 (1970).
337. Champion, J.V., Davis, I.D.: Light scattering by solutions of flexible macromolecules
subjected to flow. J. Chem. Phys. 52, 381-385 (1970).
338. Warner, H.R.,Jr.: Kinetic theory and rheology ofdilute suspensions of finitely extendible
dumbbells. Ind. Eng. Chem. Fundam. 11, 379-387 (1972).
339. Booij, H. C.,van Wiechen, P. H.: Effect of internal viscosity on the deformation of a linear
macromolecule in a sheared solution. J. Chem. Phys. 52, 5056-5068 (1870).
339a. Zimmerman, R. D., Williams, M. C.: Evaluation of internal viscosity models. Trans. Soc.
Rheol. 17, 23-46 (1973).
339b. Bazfia, E. R., Williams, M. C.: Rheological properties of internal viscosity models with
stress symmetry. J. Polymer Sci. Polymer Phys. Ed. 12, 825-843 (1974).
340. Munk, P., Peterlin, A.: Streaming birefringence. IX. Invariant expressions for chain
molecules. Rheol. Acta 9, 288-293 (1970).
341. Ree, T., Eyring, H.: Theory of non-newtonian flow. I. Solid plastic system. J. Appl. Phys.
26, 793-800 (1955).
178
W.W. Graessley:
342. Woods, M. E., Krieger, I. M.: Rheological studies on dispersions of uniform colloidal
spheres. I. Aqueous dispersions in steady shear flow. J. Colloid Sci. 34, 91-99 (1970).
343. Papir, Y.S., Krieger, I.M.: Rheological studies on dispersions of uniform colloidal
spheres. II. Dispersions in non-aqueous media. J. Colloid Sci. 34, 126-130 (1970).
344. Graessley, W.M.: Molecular entanglement theory of flow behavior in amorphous
345. Graessley,W.W., Segal, L.: Effect of molecular weight distribution on the shear rate
dependence of viscosity in polymer systems. A.I.Ch.E.J. 16, 261-267 (1970).
346. Abdet-Alim, A. H., Balke, S. T., Hamielec, A. E. :Flow properties of polystyrene solutions
under high shear rates. J. Appl. Polymer Sci. 17, 1431-1442 (1973).
347. Bueche, F.: Viscosity of entangled polymers; theory of variation with shear rate. J.
Chem. Phys. 48, 4781-4784 (t968).
348. Shroff, R.N., Shida, M.: Correlation between steady state flow curve; and molecular
weight distribution for polyethylene melts. Polymer Eng. Sci. 11, 200-204 ( 1971).
349. Middleman, S.: Effect of molecular weight distribution on viscosity of polymeric
fluids. J. Appl. Polymer. Sci. 11,417-424 (1967).
350. Philippoff, W., Gaskins, F. H., Brodnyan, J. G.: Flow birefringence and stress. V. Correlation of recoverable shear strains with other rheological properties of polymer solutions.
J. Appl. Phys. 28, 1118-1123 (1957).
351. Spriggs,T.W., Huppler, J.D., Bird, R.B.: An experimental appraisal of viscoelastic
models. Trans. Soc. Rheol. 10, 191-213 (1966).
352. Graessley, W.W., Segal, L.: Flow behavior of polystyrene systems on steady shearing
flow. Macromolecules 2, 47-57 (1969).
353. Tanaka, T., Yamamoto, M., Takano, Y.: Non-newtonian flow in concentrated polymer
systems. J. Macromol. Sci. Pt. B 4, 931-946 (1970).
354. Bird, R.B., Warner, H.R.,Jr.: Hydrodynamic interaction effects in rigid dumbbell
suspensions. I. Kinetic theory. Trans. Soc. Rheol. 15, 741-750 (1971).
354a. Wales, J. L.S., Philippoff, W.: The anisotropy of simple shearing flow. Rheol. Acta 12,
25-34 (1973).
355. Vinogradov, G.V., Belkin,I.M.: Elastic strength, and viscous properties of polymer
(polyethylene and polystyrene) melts. J. Polymer Sci. Pt. A 3, 917--932 (1965).
356. Huppler, J.D., MacDonald, I.F., Ashare, E., Spriggs,T.W., Bird,R.B., Holmes, UA.:
Rheological properties of three solutions. Part II. Relaxation and growth of shear and
normal stresses. Trans. Soc. Rheol. 11, 181-204 (1967).
357. Zapas, L.J., Phillips, J.C.: Simple shearing flows of polyisobutylene solutions. J. Res.
Natl. Bur. St. 75A, 33-40 (1971).
358. Schurz, J., Mavrommatakos, A.: Die pr~istation~iren Anlaufkurven bei PolyvinylacetatL6sungen. Angew. Makromol. Chem. 15, 95-107 (1971).
359. Schremp, F.W., Ferry, J. D., Evans,W. W.: Mechanical properties of substances of high
molecular weight. IX. Non-Newtonian flow and stress relaxation in concentrated
polyisobutylene and polystyrene solutions. J. Appl. Phys. 22, 711-717 (1951).
360. Sakai, M., Fukaya, H., Nagasawa, M.: Time dependent viscoelastic properties of concentrated polymer solutions. Trans. Soc. Rheol. 16, 635-649 (1972).
361. Einaga, Y., Osaki, K., Kurata, M., Kimura, S., Tamura, M.: Stress relaxation of polymers
of polymers under large strain. Polymer J. (Japan) 2, 550-552 (1971). See also Polymer J.
(Japan) 5, 91-96 (1973).
362. Berry, G. C.: (Private communication).
363. Osaki, K., Tamura, M., Kurata, M., Kotaka, T.: Complex modulus of concentrated
polymer solutions in steady shear. J. Phys. Chem. 69, 4183-4191 (1965).
364. Booij, H.C.: Influence of superimposed steady shear flow on the dynamic properties
of non-Newtonian fluids. Rheol. Acta 5, 215-221 (1966).
365. Simmons,J. M.: Dynamic modulus of polyisobutylene solutions in superposed steady
shear flow. Rheol. Acta. 7, 184-188 (1968).
366. Tanner,R.I., Williams, G.: On the orthogonal superposition of simple shearing and
small-strain oscillatory motions. Rheol. Acta 10, 528-538 (1971).
367. SeI1,J.W., Forsman,W.C.: Loss moduli under steady shearing of concentrated polystyrene solutions. Macromolecules 5, 23-24 (1972).
179
368. Lee, K.H., Jones, L.G., Pandatai, K., Brodkey,R.S.: Modifications of an R-16
Weissenberg Rheogoniometer.Trans. Soc. Rheol. 14, 555-572 (1970).
369. Meissner, J.: Modification of the Weissenberg rheogoniometer for measurement of
transient rheological properties of molten polyethylene under shear. Comparison with
tensile data. J. Appl. Polymer Sci. 16, 2877-2899 (1972).
370. Pipkin, A.C.: Small displacements superposed on viscometric flow. Trans. Soc.
Rheology 12, 397-408 (1968).
371. Markovitz, H.: Small deformations superimposed-on steady viscometric flows. In:
Onogi, S. (Ed.): Proc. 5th Internat. Cong. Rheology, Vol. I, pp. 499-510. Maryland:
University Park Press 1970.
372. Bernstein, B., Fosdick, R.L.: On four rheological relations. Rheol. Acta 9, 186-193
(1970).
373. Penwell,R.C.: (Private communication).
374. Lodge, A.S., Meissner, J.: Comparison of network theory predictions with stress/time
data in shear and elongation for a low density polyethylene melt. Rheol. Acta 12,
41-47 (1973).
375. Bagley, E. B., Storey, S. H., West, D. C.: Post extrusion swelling in polyethylene. J. Appl.
Polymer Sci. 7, 1661-1672 (1963).
376. Stratton,R.A., Butcher, A.F.: Stress relaxation upon cessation of steady flow and the
overshoot effect of polymer solutions. J. Polymer Sci. Polymer Phys. Ed. 11, 1747-1758
(1973).
377. vanEs, H.E., Christensen,R.M.: A critical test for a class of non-linear constitutive
equations. Trans. Soc. Rheol. 17, 325-328 (1973).
378. Bogue, D. C., White,J. L.: Engineering analyses of non-newt onian fluids. AGARDograph
144 (N.A.T.O.). Springfield, Va.: National Technical Information Service, 1970.
379. Carreau,P.J.: Rheotogical equations from molecular network theories. Trans. Soc.
Rheol. 16, 99-127 (1972).
380. Ziabicki, A.: Structural theories in polymer rheology. In: Proc. 5th Internat. Cong.
Rheology, Vol. 3, pp. 235-253. Maryland: University Park Press 1970.
381. Tanner,R.I., Simmons, J.M.: Combined simple and sinusoidal shearing in elastic
liquids. Chem. Eng. Sci. 22, 1803-1815 (1967).
382. Ferry, J.D.: An example of a rheologicat conceptual scheme. In: Onogi, S. (Ed.): Proc.
5th Internat. Cong. of Rheology, Vot. 1, pp. 3-21. Maryland: University Park
Press 1970.
Received May 28, 1974

The Entanglement Concept in Polymer Rheolgy

Uploaded by

Copyright:

Available Formats

The Entanglement Concept in Polymer Rheolgy

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Entanglement Concept in Polymer Rheolgy

Uploaded by

Copyright:

Available Formats

The Entanglement Concept in

Material Properties of Viscoelastic Liquids

Molecular Models in Polymer Rheology . . . . . . . . . . . . . . . . .

Experimental Results on Linear Viscoelastic Behavior . . . . . . . . . . .

Chain Configuration in Amorphous Polymer Systems

Molecular Entanglement Theories of Linear Viscoelastic Behavior

Entanglement in Cross-linked Systems

Non-linear Viscoelastic Properties

It was argued that such structures would behave as additional network

The Entanglement Concept in Polymer Rheology

difficult because the configuration of the system is appreciably displaced from

2. Chain Configuration in Amorphous Polymer Systems

The Entanglement Concept in Polymer Rheology

2.2. Large Scale Configuration in Concentrated Systems

xl, xl+dXl; x2, x2+dx2;

and (r 2) is the mean square end-to-end distance:

For random coils, (r 2) is directly proportional to the contour length. If n is the

The Entanglement Concept in Polymer Rheology

Taken literally, however, this viewpoint leads to rather strange variations of

2.2.2. Evidence on Chain Dimensions from Physical Measurements

low molecular weight (M = 8000) deuterated polystyrene in ordinary polystyrene,

Table2.1. Chain dimensions in undiluted amorphous polymers obtained by scattering

X-ray scattering from

" Only/~rn = 87000 was reported.

The Entanglement Concept in Polymer Rheology

in which Na is Avogadro's number, n is the number of main chain atoms per

Elimination of the frictional coefficient between Eqs.(2.6) and (2.7) yields

For polystyrene fractions in diethyl phthalate solution ( 3 0 0 0 0 < M < 200000;

2.2.3. Collapsed Coil Theories

The EntanglementConcept in Polymer Rheology

This equation was then applied to a variety of experimental data on solvent

2.2.4. Evidence on Chain Dimensions from Chemical Properties

Table 2.3. Gel points in polystyrenecross-linked by gamma radiation

The total number of cross-links per gram is directly proportional to radiation

The Entanglement Concept in Polymer Rheology

2.2.5. Extended Configuration Theories

effects might reasonably be attributed to association, although not necessarily

2.3. Short-Range Organization

The Entanglement Concept in Polymer Rheology

these correlations on thermodynamic and mechanical properties appears to be

2.4. Structural Aspects of Flow and Flow History

be interesting to repeat such experiments in a plate-cone instrument where elastic

The Entanglement Concept in Polymer Rheology

3. Material Properties of Viscoelastic Liquids

LP31 P32 P33J

The Entanglement Concept in Polymer Rheology

shear deformation 7(t):

7(0 = ?'(0) + S f~(s)ds

i.e., the values of the shear deformation at all past times.

~r(t) = S G(s) ~(t - s) ds.

G'(co) = 09 ~ G(t) sinogt dr,

G"(~o) = ~o ~ G(t)cos lot dt.

The instantaneous modulus is given by:

The Entanglement Concept in Polymer Rheoiogy

7(t) = ~ J(s) gr(t -s) ds.

t = ~ G(t - s) J(s) ds.

If a sufficiently small constant shearing stress ao is imposed on the material at

r/o = ~im r/(~),

G(t) = S H(z) e -'/'dlnz.

GO= ~ H(z) din z,