3

Entang lement Effects

in Polymer Solutions
WILLIAM w. GRAESSLEY

1. INTRODUCTION

Viscoelastic behavior in polymer solutions is a reflection of dynamical

processes at the molecular level. These processes are driven by Brownian
motion, and they govern diffusion and the rate of rearrangement of chain
orientation and conformation. Mechanical response and molecular relaxa-
tion go hand in hand. Properties which depend on large-scale chain motions
are strong functions of the molecular architecture (chain length, long chain
branching, etc.) at all levels of polymer concentration.
At high concentrations the response acquires some special character-
istics which are attributed to chain entanglement. Entanglement effects are
common to all polymer species and appear to derive simply from the
topological constraint of molecular uncrossability: each chain must move
in such a way as never to pass through the backbones of other chains.
Beyond a certain chain length, dependent on both species and concentration,
the uncrossability constraint slows the rate of large-scale conformational
rearrangement, but leaves the local chain motions essentially unaffected.
Entanglement interactions dominate all properties which depend primarily
on large-scale chain motions. They increase in importance with chain length

WILLIAM W. GRAESSLEY. Exxon Research and Engineering Co., Corporate Re-

search-Science Laboratories, Annandale, New Jersey 08801.

145
W. C. Forsman (ed.), Polymers in Solution
© Springer Science+Business Media New York 1986
146 CHAPTER 3

and concentration, and they appear to yield a set of quite general laws
relating chain architecture and properties.
The object of this chapter is to summarize the experimental observa-
tions for flexible linear and branched chains at high concentrations and to
describe recent theories about the entanglement effect. Linear viscoelastic
behavior is emphasized, although a few specific nonlinear properties are
also considered. Linear response lends itself most readily to molecular
interpretations because it reflects only the unperturbed dynamics of the
system. Nonlinear response depends also on the effect of a finite displace-
ment from equilibrium. The behavior of structurally uniform systems (model
polymers) is the primary focus; the complicated and still poorly under-
stood effects of molecular weight distribution will not be discussed. The
intermediate concentration regime is also omitted. The effect of entangle-
ments is less settled there, and a separate chapter would be required for
adequate coverage. Related material is discussed elsewhere in this volume,
i.e., chain conformation at equilibrium in concentrated solutions (Chapter
4), molecular models and chain dynamics in dilute solutions (Chapter 1),
and the fast dynamics of chains in concentrated solutions (Chapter 5). The
books of Ferryl and de Gennes 2 and reviews by Berry and Fox3 and Graess-
ley 4.5 provide more detailed discussions. The bibliography is by no means
exhaustive. Recent work is emphasized, but the references in those papers
lead back to the original contributions.

2. CHAIN ARRANGEMENT AND MOTION IN

CONCENTRATED SOLUTIONS

From results obtained by neutron scattering there is now a reasonably

good understanding about equilibrium chain conformation in concentrated
polymer liquids, at least on distance scales which are large compared to
the monomer size (see Chapter 4, for example). The pervaded volumes
of the chains interpenetrate extensively, and the intramolecular excluded
volume interactions are effectively screened. The chains assume their theta-
solvent dimensions. They obey random walk statistics, and the distribution
of large-scale conformations is governed only by molecular architecture
and C=, the characteristic ratio of the species. 6
Given these features, an undiluted polymer or concentrated solution
probably resembles a kind of random meshwork of chain contours, each
chain following some tortuous (random walk) trajectory through a tangled
mass of surrounding chains (Figure 1). Locally, i.e., at the monomer scale,
ENTANGLEMENT EFFECTS 147

FIGURE 1. A polymer chain and surroundings in an entangled liquid.

the physical structure and properties are almost certainly independent of

molecular architecture. This picture should apply over some range of con-
centrations, the spacings simply growing larger with dilution until a regime
of sufficiently weakened screening is reached, the semidilute region,2 where
chain dimensions then begin to increase. That transition is probably a
rather gradual one; estimates have placed it in the range of 10%-25%
polymer.7
The dynamics of chains at high concentration are rather more com-
plicated. As in dilute solution, the elementary motions of the chains are
probably cooperative bond rotations, with precise details and rates that are
highly individual to each species and local environment. 8 However they
occur, their cumulative effect with time is to relax the correlation with
past conformations over progressively larger distances along the chain.
Driven by these rapid elementary motions, each molecule gradually drifts
from one large-scale conformation to another. Like excluded volume, the
hydrodynamic interactions between units on the same chain are well
screened by the other chains.
Our main concern here is with the properties that depend on the large-
scale chain motions, i.e., the slow dynamics. The elementary motions are
still important in setting the time scale, but precise details tend to be aver-
aged out as the relaxation propagates to larger chain distances. The effect
of local motion on the slow dynamics can be represented by an empirical
parameter, the apparent monomeric mobility, or its reciprocal, the mono-
meric friction coefficient.l A concentrated polymer solution is assumed to
148 CHAPTER 3

behave locally as a Newtonian fluid. Each molecule moves like the cor-
responding Rouse chain (see Chapter I) would move if subjected to the
constraint that it cannot cross the contours of the other chains. Thus,
stripping the problem to its essentials, the basic picture for the slow dynam-
ics in concentrated solutions is that of a highly overlapping collection of
uncrossable Rouse chains. Each chain moves randomly in a locally viscous
medium which is permeated by a mesh of similarly moving, uncrossable
contours. The local motions are hardly affected by uncrossability since they
involve only short chain distances. The large-scale motions are strongly
affected because the chains must somehow diffuse around one another in
order to take up entirely new conformations.
This preliminary discussion of entanglement effects is based almost
entirely on inferences drawn from linear viscoelastic behavior and com-
parisons with Rouse model predictions for the fast dynamics of long chains
with the slow dynamics of chains which are below the entanglement thresh-
old. The following section treats linear viscoelasticity in the context of
concentrated polymer solutions.

3. LINEAR VISCOELASTICITY IN CONCENTRATED

SOLUTIONS

It is a remarkable fact, well substantiated experimentally, that all infor-

mation about the linear mechanical response of any liquid to volume-
conserving deformatio~s can be expressed in terms of a single time-de-
pendent function. 1 There are a number of equivalent functions, one being
the shear stress relaxation modulus G(t). That property describes the time
dependence of shear stress O'(t) following a small jump in shear strain Yo
(Figure 2). If Yo is small enough, the stress at any later time is directly
proportional to strain:
O'(t) = G(t)yo (I)

The initial modulus is (1(0), and G(oo) = 0 for any liquid. Stress and strain
for any history of sufficiently small or slow shear deformations are related

t <0 t >0
FIGURE 2. A simple shear step strain.
ENTANGLEMENT EFFECTS 149

through G(t) by the Boltzmann superposition formula. For a simple shear

history y(t),

a(t) = f~oo G(t - s) d~~s) ds (2)

At the molecular level this means that the disturbance from equilibrium is
always so small that the responses to past strain impulses are additive.
Storage and loss components of the complex dynamic modulus,
G*(w) = G'(w) + iG"(w), are the measured quantities at steady state in
the commonly used oscillatory shear experiment. These are connected with
G(t) through Eq. (2) [y(t) = Yo exp(iwt)]:

G*(w) = iw f~ G(t) expJ -iwt) dt (3)

where
G(O) = G'(O) = ~foo G"(w)dlnw (4)
11: °
Similarly, for the steady state response to a small and constant shear rate y,

'YJo = f°oo G(t) dt =

. G"(w)
hm ---'---'--
w-+o W
(5)

Jeo =
I
-2-
foo tG(t) dt = -2-
I . G'(w)
hm - - 2 - (6)
'YJo ° 'YJo w-+o W

where 'YJo is the zero shear viscosity, the ratio of steady state shear stress to
shear rate, and Jeo is the recoverable shear compliance, the ratio of total
recoil strain to shear stress at steady state. l Both 'YJo and Jeo are properties
which depend on the slow dynamics. Their product is a mean relaxation
time which characterizes the slow dynamics:

'l'm = 'YJoJeo = f~ tG(t) dt / f~ G(t) dt (7)

In entangled solutions of linear chains G(t) separates into two rather

distinct groups of relaxations, the transition region at short times and the
terminal region at long times (Figure 3). The separation begins to appear
at a certain chain length and then widens rapidly with increasing chain
length. The modulus at intermediate times, the plateau modulus GNo, is
independent of chain length but depends on polymer species and concen-
tration. The time scale of the terminal region is set by 'l'm [Eq. (7)]. In the
150 CHAPTER 3

. glassy

10"
plateau

I terminal

10"

10- 1• 10-1 10-1 10-4 10-2 10" 10-2

t(sec)
FIGURE 3. Representation of the master curve for shear stress relaxation modulus at
high polymer concentrations.

frequency domain (Figure 4) the low-frequency peak in G'(w), Gm" at fre-

quency wm , marks the location of the terminal region. In the plateau region
G'(w) is nearly constant, while GI/(w) passes through a shallow minimum.
Both rise again when the transition region is encountered. The plateau
modulus serves, in effect, as the initial modulus for the terminal region.
Its value can .be estimated either as the nearly constant value of G'(w) at
intermediate frequencies, or from the resolved area under the terminal
loss peak using Eq. (4).1
With few exceptions, a change in temperature merely shifts the time
scale in the terminal region. The form of the response curve remains the
same, and the modulus scale changes little, if at all. This property of time-

10"
,..-
Ol/(w)~

100
0: ___ ~_---7
~O'("')
0'(",)
and 100
01/(",)

dyne
cmr 10"

10" 10' 10'

'" sec- 1
FIGURE 4. Representation of the master curve for dynamic moduli at high polymer
concentrations.
ENTANGLEMENT EFFECTS 151

temperature superposition is used extensively to reduce data obtained at

different temperatures to a master curve at some chosen reference temper-
ature To.! Thus, for example,

(8)

where G(t) and Go(t) are relaxation moduli at temperatures T and To, the
temperature dependence is contained entirely in the time and modulus shift
factors aT and bT (aT = b1' = 1 at To), and b1' '" I within the accuracy of
most measurements. When Eq. (8) applies, the behavior is said to be
thermorheologically simple, and the temperature dependence of all linear
viscoelastic properties is set by aT and bT through Eq. (2).9 Since GNo,
Gm", and 'eo depend on b1' alone, they are quite insensitive to temperature.
Properties such as 'YJo, W m , and Tm depend strongly on temperature through
aT' Temperature coefficients are governed primarily by the local composi-

tion. They vary widely with polymer and solvent species and become inde-
pendent of chain length for long chains.! As a result, the relationships
between molecular architecture and properties in the terminal region have
forms which are independent of temperature. (Some exceptional behavior
in branched polymers of certain species is discussed in Section 6).

4. RELATIONSHIPS WITH MOLECULAR STRUCTURE

4.1. The Plateau Modulus

In undiluted polymers the plateau modulus G NO is essentially a charac-

teristic constant of the species, being independent of chain length and
insensitive to temperature. Well-established values for several species are
given in Table I. With the equation for modulus from the theory of rubber
elasticity, these numbers can be recast in terms of the entanglement molec-
ular weight Me, an apparent molecular weight of the temporary network
strands:
(9)

where (! is the density of undiluted polymer, ¢ is the volume fraction of

polymer, and R is the universal gas constant. The variation with species
is reduced somewhat when the data are expressed in terms of an entangle-
ment strand length L e , but even here the values range from about 15 to
60 nm.
152 CHAPTER 3

TABLE 1. Plateau Modulus for Selected Polymers

Polymer species TeC) GNo (MPa) Me

Polyethylene 190 2.2 1,300

Hydrogenated 1,4-polyisoprene 50 1. 25 1,800
1,4-Polybutadiene 25 1.15 1,900
1,2-Polybutadiene 25 0.62 3,600
1,4-Polyisoprene 25 0.35 6,300
Polyisobutylene 25 0.25 8,900
Poly(dimethyl siloxane) 25 0.24 10,000
Polystyrene 160 0.18 19,500

The variation of GNO with concentration has been determined in several

species. A remarkably general power law dependence on the volume frac-
tion of polymer is found (Figure 5):

(10)

The variation appears to be slightly stronger than quadratic (2.1 < d< 2.3,

N
.le
~ oJ

i-
OJ-
v:/'

~4~QD~4----~O~.I~--~~~2----~~~4~----~1=O~

•
FIGURE 5. Plateau modulus vs. concentration for solutions of 1,4-polybutadiene. The
various symbols represent different diluents, including low-molecular-weight polybuta-
diene. io The power law exponent (slope of the line) is 2.26 in this case.
ENTANGLEMENT EFFECTS 153

depending on species), insensitive to temperature, and independent of the

choice of diluent.IO,n (However, information on GNo in () solvents is still
very limited.) Equation (10) appears to apply even as low as 4> = 0.1,10
but resolution of the terminal and transition dispersions is difficult in that
range unless the chains are extremely long. Based on the idea that GNO is
proportional to the density of binary contacts one would expect d = 2.
The slightly larger power is still unexplained. A prediction that GNO ex: 4>2.25
for semidilute solutions 2 is close to the observation, but the theoretical
basis for this seems questionable at high concentrations.
Evaluation of plateau modulus in branched polymers is difficult be-
cause long branches broaden the terminal dispersion (see Section 6). How-
ever, the existing data suggest that GNO is changed little, if at all, by long
branches.12 Values of GNo have been estimated in several species from the
topological contribution to the equilibrium modulus of cross-linked net-
works. 13 These observations all support the idea that GNO is a fundamental
property of the species which transcends the various architectures, reflects
the uncrossability constraint, and varies universally with polymer concen-
tration.
Even the observed variation of GNo with polymer species can be
reconciled to some extent. n Consider the mechanical behavior of a hypo-
thetical liquid filled with uncrossable molecular "strings" which are very
thin, extremely long, and locally flexible. The plateau modulus of this
liquid should reflect only the uncrossability interaction between strings,
and, like the cross-link density in a rubber network, GN°/kT should be a
measure of the number of interactions per unit volume. This interaction
density certainly must depend on the total length of string per unit volume.
The contour length concentration is 'VLo, 'V being the number of strings
per unit volume and L o the length of each string. If these were the only
variables, the relationship between GN°/kT and 'VLo would be dictated by
dimensional considerations alone. The resulting form, however,
GNo
kT('VLo)3/2 = pure number (11)

gives the wrong concentration dependence [GNo ex: 4>3/2 according to Eq.
(11), instead of the experimental exponent, 2.1 < d < 2.3].
There must be at least one other variable in the problem. A natural
choice is the Kuhn step length I, characterizing the local flexibility of the
chain and thus the distribution of conformations. 14 Both interaction density
and contour length concentration can now be made nondimensional with
I, and their relationship can be chosen to match the observed power law
154 CHAPTER 3

0.6....----....-----,--r--,,,

0.4 ~Ei
HPIPj

PMMA • • PVAc;
0.2
• PcxMS
GO IS PBO(V)j P::
S
~
0.1
De
.06
PEBMA. I:.
PHMA • • PBO(c;)
·PIP
·PIB

.04
PDMS
•
2 4 6 e 10
ZlU.. 2
FIGURE 6. Dimensionless plateau modulus vs. contour length concentration for various
polymer species in the undiluted state. See Reference 11 for individual species designations.

dependence on concentration. The result,

GNOP = K(vL /2)d (12)

kT °
is compared with data for several undiluted polymer species in Figure 6.
There is considerable scatter, but the data are reasonably consistent with
a quadratic or slightly stronger dependence. For d = 2, the average value
K = 0.011 is obtained. The species dependence and concentration depen-
dence of GNo can thus be interpreted as separate manifestations of a more
general law relating entanglement interactions and uncrossability con-
straints.

4.2. The Viscosity and Recoverable Compliance

The Rouse model (Chapter 1) provides a useful framework for or-

ganizing data on 'Y/o and leo. The diffusion coefficient for noninteracting
Rouse chains is given by
D = kT (13)
r Con
ENTANGLEMENT EFFECTS 155

where Co is the monomeric friction coefficient, and n is the number of

monomers in the chain. The molecular friction coefficient is Con, so Dr
ex: M-l. The stress relaxation modulus for a solution of v noninteracting
linear Rouse chains per unit volume is given by

L exp ( - p2t)
Gr(t) = vkT Pm
p=l
-
Or
(14)

in which
Con<R2)
Or = 6n 2kT (15)

is the Rouse relaxation time, and <R2) is the mean square end-to-end
distance of the chains «R2) = Lo/). Both nand <R2) are proportional to
chain length, so 7:r ex: M2.
Each value of P specifies one of the normal modes of the chain. Quali-
tatively, 7:r/p 2 is the relaxation time for the equilibration of conformation
over a chain distance Lo/p. The upper limit, Pm' designates a species-
dependent cutoff at short times. Prior to t'" or/Pm2 , corresponding to
equilibration over distances smaller than Lo/Pm, the response is sensitive
to the polymer species. Although Eqs. (14) and (15) are exact for Rouse
chains only in the limit Pm ~ 00, they are sufficiently accurate for discussing
properties which depend primarily on the slow dynamics even for quite
modest Pm. The Rouse chain is a one-parameter model for the slow dy-
namics. Only the molecular friction coefficient nCo must be established by
separate dynamic experiments.
With Eqs. (5) and (6), and for Pm ~ 1, Eqs. (14), (15) give the follow-
ing expressions for viscosity and recoverable compliance:

vCon S2
('YIo)r = 6 ex: M (16)

0_21_2 M
(Je ), - 5" kT - 5" cRT (17)

where S2 = <R2)/6 is the mean-square radius of gyration of the chains.

The experimental observations on undiluted linear chains are as shown in
Figures 7 and 8. Below Me the viscosity behavior agrees fairly well with
the Rouse model both in magnitude (using values of Co estimated from
either the transition region response or the diffusion coefficients of small
molecules in the polymerl) and in chain length dependence. Below Me'
the recoverable compliance is in similarly good agreement. In this case no
estimation of Co is required to make the comparison. The characteristic
156 CHAPTER 3

14r-.-----.-----~----r-~

i
: 12~~----~·---_r___t

.[ I

~
... 10f----+--
(!)

~ 8~~-----r---rl~----~~

6 f----t-----r -<It --+-----t-------j

5 6
Mw
FIGURE 7. Viscosity vs. molecular weight for undiluted polyisoprene. The slope of the
line is 1.0 and 3.7, respectively. [See H. Odani, N. Nemoto, and M. Kurata, Bull. Inst.
Chern. Res. 50, 117 (1972).]

molecular weights Me and Me' mark the boundaries between short-chain

behavior [rough consistency with Eqs. (16) and (17)] and long-chain
behavior ("fJo ex: M3.4 and Jeo independent of M).
Values of Me, Me, and Me' for several species are given in Table 2.
They are insensitive of temperature, and they increase with dilution by
what appears to be the same simple law, i.e., they vary as cpl-d, d being
the dilution exponent for Gso [Eq. (10)]. The relative values, roughly
Me : Me : Me' :: I : 2 : 7, seem to be independent of species and dilution.
(There may be some exceptions; experimental uncertainties about the con-
centration dependence of Me preclude a definite statement at the present
time.)
Combining these observations, one can express the data on "fJo and
Jeo for solutions of linear, nearly monodisperse chains in the entanglement
region as follows:
MGNO )2.4
"fJo(cp, M) = ["fJo(Cp, M)]r ( 2et/J RT (18)

(19)

where Mel Me =2 and Me'l Me = 7 have been used, GNO is evaluated at

ENTANGLEMENT EFFECTS 157

CII
r:
>-
"0
J/
09'/0 0 0 0
0
N-' -6 0 /

E 0 /
u /

..,
/
CII /

/
/
/
(!)
o 0/ /

-'
/
9/
/

/
/

4 5 6
LOG Mw

FIGURE 8. Recoverable compliance vs. molecular weight for undiluted polyisoprene

(see caption of Figure 7). The dashed line is the Rouse model prediction.

concentration ¢, and precise agreement with Eqs. (16) and (17) at Me and
Me' is assumed. The monomeric friction factor is contained in ('f}o)r. It
depends on solvent and polymer species, concentration and temperature
but is supposed to be independent of chain length for long chains. Recov-
erable compliance is extremely sensitive to molecular weight distribution.!
Even the most carefully prepared samples are not monodisperse, of course,
and the numerical coefficient for truly mono disperse chains is probably
slightly smaller than that given in Eq. (19).
These universal laws for 'f}o and leO suggest the possibility that the form
of G(t) in the terminal region is also universal. That appears to be the

TABLE 2. Characteristic Molecular Weights for Selected Polymers

Polymer species

Polystyrene 19,500 35,000 130,000

Poly(a-methyl styrene) 13,500 28,000 104,000
1,4-Polybutadiene 1,900 5,000 13,800
Polyvinyl acetate 12,000 24,500 86,000
Poly(dimethyl siloxane) 10,000 24,400 61,000
Polyethylene 1,300 3,800 12,000
1,4-Polyisoprene 6,300 10,000 35,000
Polyisobutylene 8,900 15,200
158 CHAPTER 3

6
1.2
t7
6
1.0
6

0.8
d
.e
~
d
..... 0.6
~ d52

0.4

0.2

W1Wm
FIGURE 9. Loss modulus vs. frequency in reduced form for several species and chain
lengths (see Reference 10 for individual designations). The numbers indicate values
of M/Me • The dashed line is the curve for a single relaxation time: G"(w)/Gm " =
2(w/w m )f[1 + (w/w m)21.

case, judged by data on G*(w) at high entanglement densities.lO Thus,

reduced plots of G" /G m" vs. w/wm (Figure 9) for various species and con-
centrations appear to approach a common curve when M / Me becomes
large enough (M / Me"""" 100) to resolve fully the terminal and transition
regions.

5. REPTATION AND THE DOl-EDWARDS THEORY

In 1978 Doi and Edwards proposed an attractively simple and testable

theory for the slow dynamics of entangled, linear chain liquids. I5 - Is They
began with the assumption that the chains move from one large-scale con-
formation to another primarily by reptation, i.e., by diffusing in snakelike
fashion along their own contours. de Gennes had introduced this idea earlier
when considering the diffusion of unattached chains in a perrnanent net-
work. 19 Uncrossability constraints from the network tend to suppress
lateral motions beyond a certain distance, but motions of the chain as a
ENTANGLEMENT EFFECTS 159

whole along its own trajectory through the network are unimpeded. In
applying this to liquids Doi and Edwards assume that all chains move
independently, the role of network now played by neighboring chains, to
provide a "tube" for each chain of sufficiently long lifetime.
Before proceeding to the quantitative aspects it is useful to examine
some consequences of a purely reptational motion on the chain length
dependence of some simple dynamic properties. 19 In any diffusion process
the diffusion coefficient, distance and time are related by

D = K'X2/t (20)

where K' is a dimensionless constant of order unity. Consider first the

lifetime of a large-scale conformation. The chain must reptate a distance
of the order of its own length to abandon its current conformation, so
X2 ex M2. The diffusion coefficient for reptation is inversely proportional
to the molecular friction coefficient [see Eq. (13)], so D ex M-I. Thus,
from Eq. (20), the conformational lifetime should vary as M3. Now con-
sider the macroscopic diffusion coefficient for a collection of reptating
chains. Each time a chain escapes to a new conformation (t ex M3), its
center of gravity moves a distance of the order of the radius of gyration S.
For random coils S2 ex M, so in this case X2 ex M. Thus, from Eq. (20),
the macroscopic diffusion coefficient should vary as M-2. Note that the
diffusion coefficients governing reptation and center of gravity displace-
ment differ. The displacement of the center of gravity produced by a rep-
tational movement is dictated by the current end-to-end vector of the chain.
In the Doi-Edwards formulation the local physical structure of the
solution is defined by a mesh size d which depends only on polymer species
and concentration. The chains are random coils «R2) = IL o), and only
the highly entangled regime with local freedom of rearrangement, corre-
sponding to <R2)?> d 2?> 12, is considered. Each chain is regarded as a
connected sequence of random coil segments which is momentarily trapped
in an open-ended tube of constraints (Figure lO). It is assumed that the
constraint lifetime te is effectively infinite, as would be the case in a per-
manent network.
Local conformations are explored rapidly, and the average of these
explorations specifies a trajectory through the mesh, the primitive path of
the chain, which defines the current large-scale conformation. The primitive
path length L is proportional to chain length Lo but not as long because
the lateral excursions have been averaged out; both are much larger than
the mesh size: Lo > L ?> d. Fluctuations are suppressed: L is taken to be
160 CHAPTER 3

A B c
FIGURE 10. Development of the tube model: (A) A chain surrounded everywhere by
other uncrossable chains; (B) representation of the uncrossability restriction by a tube;
(C) representation of the tube as a random path of step length a with the same end-to-
end vector as the chain.

the same for all chains of the same size, and the length of chain per unit
length of path, the chain density Lo/ L, is constant along the path and has
some equilibrium value that depends only on the mesh size. The paths are
modeled as random walks with N independently directed steps of step
size a, where N ~ I, and a depends only on the mesh size. Thus,

Na=L (21)

and since the ends of chain and path coincide,

(22)

The chains are not permanently trapped because they can diffuse along
their tubes without violating uncrossability. The end of a chain can choose
any direction through the mesh as it emerges from its current tube. Primitive
path steps are created at the emerging end and abandoned at the other.
If the constraint lifetimes are long enough, the orientation of a path step
will remain constant from the moment it is created by an emerging chain
end until it is abandoned by being visited by either end of the same chain.
The dynamics of the chain in its tube are assumed to be governed by
the Rouse model. Thus the diffusion coefficient along the tube is Dr [Eq.
(13)], and the time to equilibrate any nonuniformity in the distribution of
chain density along the path, te , is of the order of Tr [Eq. (15)]. Solution
of the diffusion equation for chains moving along random walk paths
give equations for the macroscopic diffusion coefficient:

D = D,/3N (23)
ENTANGLEMENT EFFECTS 161

and the fraction of path steps with lifetimes longer than time t:

F(t) = --;- L -4 exp( -m t ) 2

(24)
:rt odd m Td
m

where Td is the disengagement time,

(25)

Note that D ex M-2 and Td ex M3, as anticipated by the arguments following

Eq. (20). Since Te""'" Tr ex M2, we can expect the chain density along the
path to equilibrate quickly compared with the time to reptate out of the
tube when the chains are very long. That property is important in the
interpretation of mechanical response.
An instantaneous deformation of the system carries the chains into a
nonequilibrium distribution of conformations. The resulting increase in
configurational free energy generates a restoring stress. The lengths and
directions of the primitive path steps are changed, the overall path lengths
are increased, and the local density of chain everywhere along the path is
displaced from its equilibrium value. If N is large the stress relaxes to zero
through a sequence of stages. In stage I the relaxation is dominated by local
chain motions. It proceeds rapidly and independently of the uncrossability
constraints up to a time of the order of to = Con d 2/6:rt 2kT, the Rouse
relaxation time for segments of chain with dimensions comparable to the
mesh siz~. Stage I corresponds to the glassy and transition regions (see
Figure 3).
In stage II (to ;:5 t ;:5 te) the chain segments in each path step act like
the strands of a random coil network, the stress relaxing slowly as the
density of chain equilibrates along the path, and the overall path lengths
return to the equilibrium value. At the end of stage II (the plateau region)
the orientations along the paths are still distributed according to the original
deformation, but the chains are no longer stretched. During stage III (the
terminal region) the stress decays from its plateau value to zero as the chains
abandon their deformed paths by reptation.
For small strains the stress relaxation modulus in the terminal region
is given by
(26)

where F(t) is given by Eq. (24). The relationship follows directly from the
property that reptation replaces the oriented path steps at the end of stage
162 CHAPTER 3

II by randomly ditected steps without altering the orientations of still-

occupied parts of the path. The fraction of the plateau stress still remaining
is then simply the fraction of steps still occupied. The number of path
steps per unit volume is 'liN ('II = chains/volume), so the shear modulus at
the beginning of stage II is 'IINkT from the theory of rubber elasticity. The
shear modulus for small strains at the end of stage II is the plateau modulus:

(27)

The decrease by 1/5 during stage II comes from the equilibration of local
stretches, leaving 'liN oriented but unstretched random coil segments per
unit volume. The combination of Eqs. (26), (27) with Eqs. (5), (6) provides
the expressions for viscosity and recoverable compliance:

(28)

3
JO=-'IINkT
e 2 (29)

The Doi-Edwards chain is a two-parameter model for the slow dy-

namics. The molecular friction coefficient nCo and one additional parameter,
characterizing the solution topology, must be supplied by dynamic experi-
ments. Several topological parameters are used (d, L, a, N), but the mesh
size d never appears in the equations, and the primitive path length L,
step size a, and number of steps N are related through Eqs. (21), (22).
Thus, for example, all properties can be expressed in terms of the primitive
path step length, which depends only on the polymer species and concen-
tration. From Eqs. (22) and (27),20

(30)

with GNo evaluated at the concentration of interest. Assuming the Rouse

model holds for solutions of short chains (M;:S Me at the concentration
of interest), we can extrapolate, using Eq. (16), to estimate the molecular
friction coefficient:

(31)

With these results the Doi-Edwards equations for diffusion coefficient,

viscosity, and recoverable compliance can be written entirely in terms of
ENTANGLEMENT EFFECTS 163

independently measured quantities:

(32)

(33)

(34)

where Me = e4>RT/G N °and [110(4), M)]r = 110(4), Me)(M/ Me) have been used.
Data on polymer diffusion coefficients in concentrated solutions are
rather limited, and espeCially so under conditions where all the quantities
in Eq. (32) can be estimated with confidence. The equation has been tested
with diffusion data for undiluted polyethylene,21 and the agreement in that
case is remarkably good. At 176°C the data can be expressed as D =
O.34/M2 (cm2/sec), and Eq. (32) predicts D = O.26/M2.20 This is well
within the limits of errors in the measurements (D, GNO, etc.), and certainly
better than would be expected from the simplicity of the model itself.
Extensive data on viscosity and recoverable compliance are available
for many systems. Comparing Eq. (33) with Eq. (18), expressing the experi-
mental results, one sees that the predicted exponent for viscosity is too
small (flo proportional to M3 instead of M3.4). However, the numerical
coefficient in Eq. (33) makes the predicted magnitude of viscosity too large
in the observable range. The situation is shown schematically in Figure 11;
the hypothetical crossing point would occur near M = 800Mc , well beyond
the range of current data. The recoverable compliance is predicted to be
independent of chain length, as observed, but its magnitude is too small
[compare Eqs. (18) and (34)], as indicated schematically in Figure II.
What can one conclude then about the assumption that reptation is
the dominant mode of motion in entangled solutions of linear chains?
Limited data on diffusion coefficients support the assumption, but results
for viscosity and recoverable compliance are less clear and imply, at the
very least, the existence of processes in addition to reptation. Competing
processes would of course result in a more rapid relaxation (smaller vis-
cosity) and a broadened terminal spectrum (larger recoverable compliance),
and that is the direction required to explain the observations. Two possible
candidates, finite lifetime of the tube constraints and fluctuations in the
primitive path length, have been considered. 5 ,20,22-24
The tube constraints are supplied by surrounding chains, and, in
entangled liquids, each chain is presumably reptating along its own primi-
tive path. It is natural therefore to identify the lifetime of constraints tc
164 CHAPTER 3

log 7]0

log M

/
/ observed
," /

log J~

log M
FIGURE 11. Comparison of the Doi-Edwards predictions for viscosity and recoverable
compliance with experiment.

in some sense with the mean lifetime of primitive path steps22:

to ~ f"
0 F(t) dt = IT La
;n;
(35)

Release of constraints allows the path locally to shift in position and

orientation, thus conferring a Rouse-like random jumping motion on the
primitive paths. The need to include constraint release is suggested by the
observation that unattached chains relax more slowly in a network environ-
ment than in the corresponding entangled liquid. 25 Omission of that con-
tribution may also account for the somewhat unrealistic predictions about
polydispersity effects in the Doi-Edwards theory.5,2o
A simple model has been used to examine the combined effect of
reptation and constraint release. 5 It retains the independent chain character
of Doi-Edwards; constraint release is treated as a communal property of
the solution, acting uniformly on the individual chains. Constraint release
ENTANGLEMENl EFFECTS 165

turns out to be much less effective than reptation in changing the spatial
positions of chains. 5• 22 In mixtures the diffusion coefficient is predicted to
be insensitive to matrix chain length, as observed experimentally.26a Depar-
tures from the pure reptation expression [Eq. (23)] should be negligible
except for short chains (very small N). The effect of constraint release is
predicted to be stronger in stress relaxation. The chain length dependence
of 17o and Jeo for large N are unchanged from pure reptation, but the nu-
merical coefficient is smaller for 17o and larger for Jeo. As N decreases,
constraint release becomes competitive with reptation more rapidly in
relaxation than in diffusion. This may account for the observation that the
diffusion law for reptation, D ex: M-2, appears rather quickly beyond the
coil overlap concentration,26b while much higher concentrations appear to
be necessary to evoke unambiguous entanglement effects in stress relaxation.
The effect of fluctuations in the primitive path length is omitted in
the Doi-Edwards theory. Simple reptation deals only with the zeroth mode
of motion (translation of the chain as a whole along its path), but the
higher modes, the first corresponding to "breathing" of the path, will also
result in abandonment of path steps.23.24 The breathing mode arises from
time-dependent fluctuations in chain density along the path. It can be
visualized in a crude way as the projection of loops into the surrounding
mesh, accompanied by retraction of chain ends toward the center and
abandonment of some portion of the path. 5 As the fluctuation subsides,
the chain ends then move out again but along random paths (Figure 12).

FIGURE 12. Two-dimensional representation of path length fluctuations. (A), (B), and
(C) correspond to the same chain at different times and with different instantaneous
path lengths.
166 CHAPTER 3

10"

R.pt.tlon only

10"
R.pt.tlon
Constraint R.I....
I
10"

IR.ptetlon
' ' - - - Constr.lnt R.I••••
Bra.thlng

10"

Rou••
100

FIGURE 13. Sketch of possible contributions to viscosity from various mechanisms for
relaxation.

For linear chains, the contribution of breathing falls off exponentially

with chain length,24 while the rate by reptation goes roughly as 7:tl-1 ex: M-3.
Reptation must then dominate for very long chains, but the contribution
from breathing may still be important at intermediate chain lengths. It has
been proposed that the experimental law for viscosity, 'fJo ex: MM, is not in
fact a power law at all but merely the result of a very slow loss of the
breathing contribution with increasing chain length, superimposed on rep-
tation. 24 The suggested effects of simple reptation, path length fluctuations
and constraint release on viscosity are sketched in Figure 13.
Fluctuations in path length may assume major importance in the
relaxation of chains with long branches. Simple reptation would tend to
be suppressed by the presence of branch points23 (see Section 6 follow-
ing).
ENTANGLEMENT EFFECTS 167

6. EFFECTS OF CHAIN BRANCHING

6.1. Experimental Observations

The preceding sections 4 and 5 deal mainly with entanglement effects

in solutions of linear chains. These effects are different for branched chains,
especially when the branches themselves are long enough to be entangled.
The most detailed studie~ have been made with near-monodisperse star
polymers, molecules with three or more arms of equal length joined at a
common junction. Each molecule has one branch point, and the branch
molecular weight Mb is M If, / being the number of arms or branch point
functionality. There are some data on near-monodisperse comb polymers.
Many varieties of randomly branched polymers are available, but even
fractionated samples are mixtures of structures, which complicates the
interpretation. The discussion here will focus on stars of low functionality
(f = 3, 4, 6) where crowding near the branch point is not too important.
Many of the effects, however, are seen for the other branching architectures
as well.
It is useful to discuss the effects of branching in relation to the behavior
of linear chains of the same species and at the same concentration. 3 ,27 In
this respect the behavior of branched polymer solutions seems to depend
on two structural features, the mean coil size and the length of the branches.
If the branches are not too long, the viscosities are rather similar for linear
and branched chains with the same radius of gyration. Thus, if 'YJL(</J, M)
is the viscosity for solutions of linear chains, then, approximately,

(36)

where, at the same molecular weight,

(37)

The size ratio g can be estimated either by dilute solution measurements

or by random walk calculations from the known branching architecture.
The latter give
3/-2
g= f2 (38)

for f-arm stars.

A dependence on coil size alone below the entanglement point is not
surprising. At high concentrations the Rouse model works fairly well in
that region, and Eq. (16) ('YJ oc: S2) is in fact valid for Rouse chains of any
168 CHAPTER 3

architecture. However, Eq. (36) continues to apply even in the entangle-

ment region [where 'YJL obeys Eq. (18)] as long as the molecular weight of
the branches, M b, is less than Me.
For longer branches the viscosity begins to exceed the predictions of
Eq. (36). The ratio 'YJn(1), M)/'YJL(1), gM) starts near unity but increases
extremely rapidly with both concentration and branch length. It can easily
reach values of 100 or more. The observed behavior of stars in relation to
linear chains is sketched in Figure 14. At high molecular weights the viscosity
is smaller for stars because the coil size effect is acting alone. At high mo-
lecular weights the viscosity becomes larger because the branch length
effect is also acting.
Departures from Eq. (36) are expressed in terms of a viscosity enhance-
ment factor,a.2?
(39)

The enhancement factor appears to be an exponential function of both

arm length and some power of the concentration:

(40)

where v is about 1.8. For stars the precise dependence on branch point
functionality is unsettled. At constant ¢ and Mb the enhancement clearly
increases withf,2? and this was attributed earlier to an increase in enhancing
"efficiency." However, r turns out to be about the same for different small
functionaIities (f = 3 and 4) at the same total molecular weight M = fM b •
Whether that behavior continues at still larger functionalities is not known.
Other complications enter the picture at larger f, e.g., chain dimensions are
increased, probably because of crowding near the branch point.28 There

log".

109 M 109M
FIGURE 14. Comparison of viscosity and recoverable compliance for liquids of linear
chains and star-branched chains.
ENTANGLEMENT EFFECTS 169

may also be some reduction of intermolecular entanglement density in the

coil interior. 29 At present, the viscosity behavior of stars can be summarized
by the equation
(41)

where Me is the characteristic molecular weight for the undiluted species,

and r is a number of order unity which is relatively insensitive to species
but varies somewhat with functionality. The rather limited data for combs
are also consistent with Eq. (41 ),30 g having been obtained by dilute solu-
tion measurements.
Like viscosity, the recoverable compliance for branched polymers at
low and moderate concentrations is smaller than for linear polymers at the
same concentration and molecular weight. For stars, the observed values
agree well with the predictions for Rouse stars,31

10 - 2 M (42)
e - g2S e¢RT
where
15/ - 14
g2 = (3/ _ 2)2 (43)

However, in contrast to linear chain behavior, Eq. (42) continues to de-

scribe the recoverable compliance for stars even at high concentrations and
molecular weights (Figure 14). Thus, above Me',

(linear chains) (44)

while
(stars) (45)

with coefficients remarkably close to those predicted by Eq. (42), (43) (see
Figure IS). Values of leO for highly entangled polymers with long branches
thus can greatly exceed those for their linear chain counterparts.
The terminal relaxation spectrum for entangled star solutions is broader
than the universal form for linear chains.1O Despite the surprising agree-
ment of leO with Eq. (42), the terminal region for long arms bears no
resemblance to the Rouse form for stars, being much broader and con-
tinuing to broaden indefinitely with increasing concentration and arm
length. 1O
One curious feature that accompanies viscosity enhancement, but only
in some species, is a change in the temperature coefficient of viscosity.
170 CHAPTER 3

0.4 ~--!-~ d_ --"'_-~-~i"-;--

0.2

a: 0.1
0...,
0.04

0.02

0.01 ~ __________- L____________ ~ __ ~~ ____ ~

10°

FIGURE 15. Recoverable compliance vs. entanglement density for solutions of linear and
star polybutadienes. The compliance values are expressed in reduced form: (J.O)R =
J.°e4>RT/g2M. For the Rouse model (J.O)R = 0.4 (dashed line). The open symbols repre-
sent linear polymers; the filled symbols represent branched polymers.

The most prominent example is polyethylene. Temperature effects in this

case were studied with model polyethylenes, made by hydrogenation of
linear and star polybutadienes. 32 The temperature dependence of viscosity
for this polymer is essentially Arrhenius, 'YJo(T) = A exp(Eal RT), where Ea
is the flow activation energy. For long linear chains, Ea 7 kcal for the
/'"'oJ

undiluted polymer, independent of chain length. For three-arm stars, Ea

increases in direct proportion to arm length, reaching 15 kcal for Mb =
30,000. The difference, 11E = (Ea)B - (Eah, decreases smoothly to zero
with dilution and generally seems to vary with 4> and Mb in the same way
as the enhancement factor exponent [Eq. (40)]. The shape of the terminal
region also changes with temperature when 11E,* O. Polyethylene stars are
thermorheologically complex: The short time response, corresponding to
small chain distances, has the same temperature coefficient for linear and
branched polyethylene, but the temperature coefficient for the longer time
response increases progressively in branched polyethylene. In the poly-
butadiene precursors, the temperature coefficients of viscosity are virtually
independent of chain architecture, and the behavior is thermorheologically
simple even when the viscosity enhancement is substantia1. 33 Other species
show similar effects: ethylene-propylene copolymer stars are thermo-
rheologically complex, for example, but polystyrene and polyisoprene stars
ENT ANGLEMENT EFFECTS 171

are thermorheologically simple. The linear chain versions of all these

species are thermorheologically simple. The effect appears to be related to
the temperature coefficient of chain dimensions 34 and will be considered
briefly in the following discussion of relaxation mechanisms in branched
polymers.

6.2. Molecular Theory

It seems reasonable to expect that entanglements will reduce the mo-

bility of chains with long branches relative to linear chains. A linear chain,
and perhaps even one with sufficiently short branches, can still move freely
along its tube of constraints. Interference should develop as the size of the
branch becomes comparable to the mesh size, and still longer branches
should tend to act as anchors. The tubes themselves are branched, and the
uncrossability constraints now resist displacement of the chain as a whole
in any direction. Simple reptation is suppressed; diffusion and relaxation
must then proceed by other mechanisms such as the fluctuations in primitive
path length discussed in Section 5.
The first investigation of fluctuation effects was made by de Gennes. 23
He examined the conformations of a long flexible chain, free at one end
but fixed at the other, in a lattice of uncrossable lines. The chain confor-
mation consists of the path through the lattice and nonentangled loops
projecting out from the path. The path length varies with the amount of
chain in the loops (Figure 16), and, to abandon a path entirely, the chain

"'.-H---I~HIf- - - - - - - - -
-fH'f---------- -

FIGURE 16. Progressive abandonment of an initial path by a chain tethered at one end.
The large-scale conformation of the branch relaxes as the result of progressively larger
fluctuations in path length.
 172 CHAPTER 3

must form itself into a completely nonentangled loop. de Gennes showed

that the probability of finding this conformation decreases exponentially
with chain length, implying a conformation lifetime of the form

l' = 1" exp({JN) (46)

where N is the average number of steps in the primitive path, {J is a con-

stant of order unity, and the prefactor 1" depends at most on some relatively
small power of N.
To test the effect of this on viscoelastic properties, consider the simple
example off-arm star chains (No = N If) in which the lifetimes of the path
steps vary exponentially with distance from the chain end t :

1'j = 1" exp({Jj) (47)

The fraction of initial steps which are still occupied at a later time t is
the fraction of steps with t < 1'j, or approximately

F(t) = - 1
No
L exp [ - -t ]
Nb
j=1 1'j
(4S)

and, with Eqs. (5), (6) and (24), (25),

4
'fJo =5 f1lkT L: 1'j (49)

o_ 5 L: 1',2
(50)
le - 4f1lkT (L: 1'j)2
For large N b , the example spectrum [Eq. (47)] gives

4 1IfkT1"
'fJo = 5 efJ _ 1 exp({JNo) (51)

5 efJ - 1
J = - - ---::-_.,- (52)
0
e 4f1lkT efJ +1

Thus, as observed experimentally, the viscosity grows exponentially with

arm length, and recoverable compliance has the Rouse form, leO ex: '11-1:

[ 25 efJ-l]2 M
leO = Sf efJ 1 5 + e¢RT (53)

t I am grateful to Dr. Robin Ball of Cambridge University for suggesting this illustra-
tion of the principle.
ENTANGLEMENT EFFECTS 173

Even the coefficient in brackets is similar in magnitude to the Rouse-Ham

g2 (Eq. (43)]: g2 = 0.63 for f= 3, and (with fJ = 1) the bracketed coeffi-
cient is 0.48.
The principle illustrated by these calculations is general even though
the example spectrum [Eq. (47)] is probably incorrect in detail. The terminal
spectrum for disengagement by fluctuations must be very broad compared
with that for simple reptation. The viscosity in such cases is dominated by
the longest relaxation time, and that will certainly be exponential in arm
length. Moreover, in the context of tube models generally, spectral broad-
ness will inevitably lead to a Rouse-like form for the recoverable com-
pliance. Equation (50) for leo is valid for tube models regardless of the
spectral form. For linear chains (simple reptatiQn) the step lifetimes are all
of comparable magnitude, so 2: T:/I(2: Tj)2 ex: NIN2 = N-l, and leo ex:
(VN)-I ex: (G N O)-l, the Doi-Edwards form and the observed behavior for
entangled linear chains [Eq. (19)]. For long arm stars the step lifetimes
increase rapidly with distance from the free end. The summations in such
cases are dominated by their largest terms, so 2: Tllef. Tj)2 is independent
of N, and leo ex: V-I ex: MI</J-I. A wide spacing of the longest relaxation
times is enough to produce the Rouse dependence on </J and M; no further
resemblance to the Rouse spectrum is required.
Disengag~ment by fluctuations in path length also provides a possible
explanation for thermorheological complexity in the branched polymers of
certain species. 34 The transition states for relaxation involve increasingly
more compact (highly looped) chain conformations for parts of the path
near the branch point, requiring an increasingly larger proportion of gauche
conformers and correspondingly fewer trans conformers. When these con-
formers have the same energy, the excess free energy of the transition state
will be purely entropic. When the gauche conformers have higher energy,
the chains must pass through states of higher energy as well as lower
entropy. The temperature dependence of relaxation in the latter case will
vary with distance from the branch point. Thus, the parameter fJ [Eq. (46)]
should depend on both temperature and location along the branch if LIe,
the energy difference between gauche and trans isomers, is greater than
zero for the species. Moreover, it should have the form

(54)

assuming that the excess of gauche conformers in each transition state is

proportional to the surplus of chain to be distributed as loops.
114 CHAPTER 3

In polymer species with LI e > 0, the spectrum of stars should narrow

with increasing temperature (thermorheologically complex behavior). The
viscosity, dominated by the longest relaxation time, should have a tempera-
ture dependence which exceeds that for linear chains by an Arrhenius term
with activation energy LIE = EtNb' The effect should therefore increase
with branch length and decrease with dilution, as observed. 32 In species
with LIe "-' 0, the viscosity temperature coefficient for linear and branched
chains should be the same and time-temperature superposition should be
obeyed.
The gauche-trans energy difference also controls the temperature de-
pendence of chain dimensions for the species: dlnCoo/dT = x, x being
negative when LIe is positive. 6 For polyethylene, in which stars exhibit
thermorheological complexity, x is indeed negative and fairly large com-
pared with most other species. For polybutadiene (precursors of the model
polyethylenes), in which stars are thermorheologically simple, the value of
x is nearly zero. These examples and others34 are consistent with the idea
that stars relax primarily by fluctuations in path length, and that differences
among species are related to conformational characteristics such as
dlnCoo/dT. However, some part of the effects may be caused by other
factors (see Reference 34). In addition, it is clear that constraint release
must also be considered in the relaxation of branched chains,25 so the
quantitative aspects are probably more complicated than suggested by
Eqs. (46) and (54).

7. NONLINEAR VISCOELASTICITY

The discussion thus far has only dealt with entanglement effects on
the slow dynamics of solutions which are essentially at equilibrium. For
sufficiently small or sufficiently slow deformations the response is linear
and governed by G(t) alone [Eq. (2), for example]. Combinations of large
strains and strain rates displace the physical structure of the solution from
equilibrium, and the response is no longer linear. Chain conformation and
entanglement determine the stress at each instant and these now vary,
depending on the prior sequence of strains (the deformation history). The
transient network theory of Lodge 35 accounts for many general features of
the nonlinear response in entangled solutions, but it requires certain func-
tions as input (the strand production and destruction rates) which must be
determined empirically. The BKZ equation 36 is another very useful form
but lacking in explicit molecular content. There are numerous continuum
ENTANGLEMENT EFFECTS 175

formulations and many mathematically acceptable ways to build in depar-

tures from linearity.37 Some fail on empirical grounds alone, but even the
remaining candidates suffer from the Lodge and BKZ model limitations.
It is not clear how to calculate the functions they require from theories
about the chain motions.
Flow instabilities and instrumental problems are more troublesome in
nonlinear experiments. 3s Much careful work has been done, but there is
still a serious lack of data on nonlinear response for polymers with well-
defined molecular structure. The only property which has been thoroughly
studied with near-monodisperse systems is the shear-rate dependence of
viscosity, i.e., the ratio of stress to shear rate at steady state in simple shear
flow: 'Y)(y) = a(y)/y. Another nonlinear property, the shear stress relaxa-
tion modulus for finite strains, G(t, y) = a(t, y)/y, has also received atten-
tion recently, and information on this property for well-characterized poly-
mers is accumulating rapidly. The discussion here will focus on these two
properties. Molecular effects for other components of the stress tensor at
steady state are well documented only at low shear rates. Data suitable
for molecular interpretation on stress transients are fragmentary for shear
flows and practically nonexistent for extensional flows.
The main features of shear rate dependence of viscosity in highly
entangled solutions are now well established. 4 Beyond some range of shear
rates, defined by a characteristic shear rate Yo, the viscosity begins to de-
crease from its zero shear rate value 'Y)o. Expressed in reduced form, 'Y)(Y)/'Y)o
vs. y/Yo, the behavior appears to be universal for highly entangled poly-
mers with narrow molecular weight distributions (Figure 17). The effects
of concentration, molecular weight, temperature, species, and chain archi-
tecture (linear or star39) are absorbed entirely in the values of 'Y)o and Yo.
Moreover, Yo appears to depend only on the product 'Y)oJ/, the mean
relaxation time '1:m of the slow dynamics [Eq. (6)]:

or (55)

where K' is a universal constant whose numerical value depends on the

particular method chosen to define Yo (K' ,...., 0.5 when Yo is the shear rate
at which the viscosity has fallen to 0.8'Y)o 4). Thus

(56)

where V(y'Y)ol/) is a universal function for highly entangled monodisperse

solutions. At shear rates well beyond Yo this reduced viscosity function
176 CHAPTER 3

0.1

o.01'::-:::,-'--'-............
0.001
u..:'-!:::-,-'--'-..........-'-'*'-:-,-'--'-L.L...\~'::---'-.....................""'-':';!:-=--'-......................I..Io'!
0.01 0.1 1.0 10.0 100.0

FIGURE 17. Viscosity-shear rate behavior of entangled, nearly monodisperse polymers,

expressed in reduced form. Solutions of polystyrene with various chain lengths, concen-
trations and temperature are represented.' Data for other entangled polymer systems,
linear and branched, obey the same form . The solid line was calculated from the theory
in Reference 40.

goes over to a power law

(57)

where the exponent p in the range 0.80-0.85. The behavior at still higher
shear rates is uncertain ; flow instabilities and viscous heating invariably
intervene at some point, although there is some evidence that the viscosity
approaches a limiting value.
A surprisingly accurate prediction of the reduced viscosity function
has been obtained from a rather simple model for the loss and replacement
of chain entanglements at steady state.40 The entanglements along a chain
are lost as the neighboring chains are carried out of its pervaded volume
by the flow. New entanglements are formed as new chains arrive, but it is
assumed that they are not formed instantly: An incoming chain must remain
in the pervaded volume longer than the time for complete rearrangement
of its conformation. If it is swept by more quickly, no entanglement occurs,
and the resistance to flow is correspondingly diminished. Thus, the steady
state entanglement density decreases with increasing shear rate, and the
viscosity decreases progressively from its zero shear value fJo . The confor-
mational relaxation time is assumed to vary with shear rate in the same
ENT ANGLEMENT EFFECTS 177

way as the viscosity: T(Y) = To'fJ(Y)/'fJo, where To is the relaxation time at

rest (y = 0). The resulting viscosity function for monodisperse chains,
'fJ(Y)/'fJo vs. yTo, fits data over a wide range of shear rates, and it goes finally
to a power law [Eq. (57)] for YTo'}> 1 with p = 9/II = 0.82, in good
agreement with observations. The interpretation of To as a relaxation time
for conformation rearrangement is of course consistent with Eq. (55) but
not unique. In general terms it means simply that departures from 'fJo begin
at shear rates of the order of the reciprocal of the longest relaxation time
in the linear viscoelastic spectrum.
As noted earlier, the observed visco'sity function is the same for highly
entangled solutions of linear and star chains. Their relaxation spectra are
entirely different, so it is clear that theories which postulate a direct and
universal connection between linear response curves and viscosity-shear
rate behavior must be wrong in principle.
The strain-dependent relaxation modulus G(/, y) reduces to G(/), the
relaxation modulus of linear viscoelasticity, for small strains. Departures
from G(t) develop as y increases. The ratio G(/, y)/G(t) decreases from
unity by a factor that in general depends both on strain and time. In en-
tangled solutions there are two time domains,41 separated by I k' a charac-
teristic time for the solution. Prior to tk the ratio depends on both strain
and time. Beyond I k> the ratio is a function of strain alone: The strain and
time dependences are approximately factorable (Figure 18). Thus

G(/, y) = h(y)G(t) (58)

The time Ik lies in the plateau region for the solution, so tk is small
compared with the longest relaxation times in entangled solutions, and Eq.
(58) describes the terminal response for finite strains. The strain-dependent
function h(y) appears to be relatively insensitive to polymer species, con-
centration and chain architecture at moderate entanglement densities,
roughly from M = 5Me(¢) to M = 50MeC¢). Below 5Me(¢) the value of
Ik becomes comparable to the longest relaxation time; the explorations
of behavior here are quite limited. Above "-' 50Me(¢) the factorability of
strain and time dependence becomes less precise, and the shifts with strain
become progressively larger and somewhat erratic. The reason for the
existence of an upper limit is unknown.
The Doi-Edwards theory provides a natural explanation for some of
these observations. I6 In the plateau and terminal regions, the relaxation
following large strains is expected to take place in two stages which are
well separated in time scale for long chains. The faster process is equilibra-
 10'
J I J

at:,., .
.. Cf.~'qr
10' I- Q

II •
: ... fI"fI'·ea
fI"" ~\O~
....
-
.'-
II • , , , ... fI ..

"-.0 " ""=~.:

"e .... II" ft, "ct'Q.":".'
A u
<;
10' I- ..... .. ppP f •• 0.'... f/"o (I
-
,.
c .... :":''':', ": ..".crt.
16-o.."PP'
"·0·
·0"·'
'G ..
:£
<) 10 ' I-
. . . b·: .
• ...... 0 ..

9,·~
-
....'.
'It .. - ' 0..(1
~ .... p,
.. .op' ...

0. -
~
10 l 1- .
~

10 J
10 ' 10 I 10 ' 10 ' 10 '
t(see)

10'

" ,
10'

"e
u
;;
,.
c 10 ' f&

B
:£
-
\
..
e
<) 10J

10 ' ~10~C------~-------L~------~------J,0 '

l(sec)

FIGURE 18. (A) Stress relaxation curves for various shear strains for a 20% solution of
high-molecular-weight polystyrene with narrow molecular weight distribution. 43 Up to
y = 1.87 the values of a(y, t) /y coincide (top curve); beyond that (up to y = 25.4) they
decrease progressively, but maintain the same shape beyond tk "-' 10 s. (8) Data in
Figure 18(A) shifted along the modulus axis to achieve superposition with the upper
(linear viscoelastic) curve. The shift factor obtained in this way is h(y).
ENTANGLEMENT EFFECTS 179

tion of stretches along the path and return of the path length to its value
at equilibrium (t "-' t e , the equilibration time). The slower process is re-
placement of Qriented path steps by randomly directed steps by reptation
(t "-' 'Td, the disengagement time). The characteristic time t k would cor-
respond with the equilibration time t e , and should therefore be of the order
of the Rouse relaxation time T, ex M2. Experimentally, tk varies approxi-
mately as M2, and its values are indeed of the order of estimations based
on Eq. (15).41
The strain-dependent function h(y) is accordingly the fraction of the
plateau modulus GNo remaining after equilibration. Part of the reduction
comes from the abandonment of path steps as the chain ends retract into
the tube and part from the equilibration of stretches in the remaining
steps.16 The former corresponds to a decrease in entanglement density, at
least in the sense that the concentration of path steps is smaller than at
equilibrium. However, the theory predicts there should be no effect on the
time dependence of subsequent disengagement: the path length, and hence
the reptation distance, is already at its equilibrium value. Therefore, the
time and strain dependences should be factorable beyond t e , as observed,
and the terminal region time dependence should be given by the linear
response function G(t), as in Eq. (58).
Doi and Edwards derived an equation for the strain-dependent function
assuming nothing more than affine displacements and return to the original
path length after equilibration. 16 The form of this function should be
universal for entangled linear and star chains. The expression obtained for
hey) is in excellent agreement with experimental data. 41 The theory of
course supplies no explanation for the observed departures from Eq. (58)
when M ~ 50Me¢.
The tube model has been generalized to arbitrary deformation his-
tories,17 but only by the introduction of additional assumptions. The result-
ing expression for n(r) does not agree with experiment. 18 The decrease
with shear rate is too rapid; indeed, the shear stress at steady state is pre-
dicted to pass through a maximum with shear rate, and this is clearly
inconsistent with the data. It is possible, however, that the problem is not
with the tube model itself, but with the simplifying assumption that equi-
libration is instantaneous. Measurements on other nonlinear properties
suggest that departures from the Doi-Edwards predictions are correlated
with the equilibration time of the solution. 42 A generalized theory which
includes both equilibration and reptation contributions will be required
to test that possibility.
180 CHAPTER 3

REFERENCES
1. J. D. Ferry, Viscoelastic Properties of Polymers, 3rd ed., John Wiley and Sons, New
York (1980).
2. P. G. de Gennes, Scaling Concepts in Polymer Physics, Cornell University Press,
Ithaca, New York (1979).
3. G. C. Berry and T. G. Fox, Adv. Polym. Sci. 5, 261 (1968).
4. W. W. Graessley, Adv. Polym. Sci. 16, 1 (1974).
5. W. W. Graessley, Adv. Polym. Sci. 47, 67 (1982).
6. P. J. Flory, Statistical Mechanics of Chain Molecules, Interscience, New York (1969).
7. G. C. Berry, H. Nakayasu, and T. G. Fox, J. Polym. Sci. Polym. Phys.17, 1825 (1979);
W. W. Graessley, Polymer 21, 258 (1980); B. L. Hager and G. C. Berry, J. Polym.
Sci. Polym: Phys. 20, 911 (1982).
8. W. H. Stockmayer, in Molecular Fluids, Les Houches 1973 (B. Balian and G. Weill,
eds.), pp. 107-144, Gordon and Breach, London (1976).
9. H. Markovitz, J. Polym. Sci. Polym. Symp. 50, 431 (1975); D. J. Plazek, E. Riande,
H. Markovitz, and N. Raghupathi, J. Polym. Sci. Polym. Phys. 17, 2189 (1979).
10. V. R. Raju, E. V. Menezes, G. Marin, W. W. Graessley, and L. J. Fetters, Macro-
molecules 14, 1668 (1981).
11. W. W. Graessley and S. F. Edwards, Polymer 22, 1329 (1981).
12. W. W. Graessley and J. Roovers, Macromolecules 12, 959 (1979).
13. S. Granick and J. D. Ferry, Macromolecules 16, 39 (1983), and references therein.
14. P. J. Flory, Principles of Polymer Chemistry, p. 412, Cornell University Press, Ithaca,
New York (1953).
15. M. Doi and S. F. Edwards, J. Chem. Soc. Faraday Soc. Trans. 274, 1789 (1978).
16. M. Doi and S. F. Edwards, J. Chem. Soc. Faraday Soc. Trans. 274, 1802 (1978).
17. M. Doi and S. F. Edwards, J. Chem. Soc. Faraday Soc. Trans. 274, 1818 (1978).
18. M. Doi and S. F. Edwards, J. Chem. Soc. Faraday Soc. Trans. 2 75, 38 (1979).
19. P. G. de Gennes, J. Chem. Phys. 55, 572 (1971).
20. W. W. Graessley, J. Polym. Sci. Polym. Phys. 18, 27 (1980).
21. J. Klein and B. J. Briscoe, Proc. R. Soc. (London) 365, 53 (1979).
22. J. Klein, Macromolecules 11, 852 (1978); M. Daoud and P. G. deGennes, J. Polym.
Sci. Polym. Phys. 17, 771 (1981).
23. P. G. de Gennes, J. Phys. 36, 1199 (1975).
24. M. Doi and N. Y. Kuzuu, J. Polym. Sci. Lett. 18, 775 (1980); M. Doi, J. Polym.
Sci. Lett. 19, 265 (1981).
25. H. C. Kan, J. D. Ferry, and L. J. Fetters, Macromolecules 13, 1571 (1980).
26a. J. Klein, Philos. Mag. A43, 771 (1981).
26b. L. Leger, H. Hervet, and F. Rondelez, Macromolecules 14, 1732 (1981).
27. W. W. Graessley, Acc. Chem. Res. 10, 332 (1977).
28. R. Roovers, N. Hadjichristidis, and L. J. Fetters, Macromolecules 16, 214 (1983);
M. Daoud and J. P. Cotton, J. Phys. 43, 531 (1982).
29. J. Carella, Ph. D. thesis, Northwestern University, Evanston (1983).
30. J. Roovers and W. W. Graessley, Macromolecules 14, 766 (1981).
31. J. S. Ham, J. Chem. Phys. 26, 625 (1957).
32. W. W. Graessley and V. R. Raju, J. Polym. Sci. Polym. Symp. 71, 77 (1984).
33. W. E. Rochefort, G. G. Smith, H. Rachapudy, V. R. Raju, and W. W. Graessley,
J. Polym. Sci. Polym. Phys. 17, 1197 (1979).
ENTANGLEMENT EFFECTS 181

34. W. W. Graessley, Macromolecules 15, 1164 (1982).

35. A. S. Lodge, Body Tensor Fields in Continuum Mechanics, pp. 142-143, Academic
Press, New York (1974); see also M. H. Wagner, Rheologica Acta 18, 33 (1979).
36. B. Bernstein, E. A. Kearsley, and L. J. Zapas, Trans. Soc. Rheology 7, 391 (1963);
see also B. Bernstein and R. L. Fosdick, Rheologica Acta 9, 186 (1970); L. J. Zapas
and J. C. Phillips, J. Res. Natl. Bur. Stand. A 75, 33 (1971).
37. R. B. Bird, R. C. Armstrong, and O. Hassager, Dynamics 0/ Polymeric Liquids, Vol.
I, John Wiley and Sons, New York (1977).
38. J. Meissner, Rheologica Acta 10, 230 (1971); J. Meissner, J. Appl. Polym. Sci. 16,
2877 (1972); R. C. Crawley and W. W. Graessley, Trans. Soc. Rheol. 21, 19 (1977).
39. W. W. Graessley, T. Masuda, J. E. L. Roovers, and N. Hadjichristidis, Macromole-
cules 9, 127 (1976).
40. W. W. Graessley, J. Chem. Phys. 47, 1942 (1967).
41. K. Osaki and M. Kurata, Macromolecules 13,671 (1980); C. M. Vrentas and W. W.
Graessley, J. Rheol. 26, 359 (1982).
42. E. V. Menezes and W. W. Graessley, J. Polym. Sci. Polym. Phys. 20, 1817 (1982).
43. Y. Einaga, K. Osaki, M. Kurata, S. Kimura, and M. Tamura, Polym. J. (Japan) 2,
550 (1971).