View - Martin Kröger - ETH Zürich
View - Martin Kröger - ETH Zürich
View - Martin Kröger - ETH Zürich
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
A thermodynamically admissible reptation model<br />
for fast flows of entangled polymers. II. Model predictions<br />
for shear and extensional flows<br />
Jiannong Fang, <strong>Martin</strong> <strong>Kröger</strong>, and Hans Christian Öttinger a)<br />
Department of Materials, Institute of Polymers, <strong>ETH</strong> <strong>Zürich</strong> and Swiss F.I.T.<br />
Rheocenter, CH-8092 <strong>Zürich</strong>, Switzerland<br />
(Received 30 November 1999; final revision received 11 July 2000)<br />
Synopsis<br />
Numerical predictions of a previously proposed thermodynamically consistent reptation model for<br />
linear entangled polymers are presented for shear and extensional flows. Comparisons with<br />
experimental data and two alternative molecular-based models are given in detail. The model<br />
studied in this paper incorporates the essence of double reptation, convective constraint release, and<br />
chain stretching, and it avoids the independent alignment approximation. Here, no use is made of<br />
the ingredient of anisotropic tube cross sections of the previously proposed model. Simulation<br />
results reveal that the model at a highly simplified level with few structural variables, i.e., four<br />
degrees of freedom, is able to capture qualitatively all features of the available experimental<br />
observations and is highly competitive with recently proposed models in describing nonlinear<br />
rheological properties of linear entangled polymers. © 2000 The Society of Rheology.<br />
S0148-60550000406-5<br />
I. INTRODUCTION<br />
Stress–strain relationships for polymer melts are the main ingredient for the flow<br />
simulation of polymer processing such as injection molding, film blowing, and extrusion.<br />
The reliability and accuracy of these simulations depends crucially on the constitutive<br />
equations which describe the nonlinear viscoelastic properties of the underlying model<br />
polymer. Although the closed-form phenomenological models, such as K-BKZ equations,<br />
have been widely used in research and commercial codes, their degree of success is<br />
limited because of a lack of physical ingredient on the molecular level. For the purpose<br />
of realistic modeling, molecular-based models are uniquely suited. A molecular model<br />
was elaborated by Doi and Edwards 1978a, 1978b, 1978c, 1979 who extended the<br />
reptation idea introduced by de Gennes 1971 to a tube idea in order to describe the<br />
viscoelastic behavior of entangled polymers in the presence of ‘‘obstacles.’’ Within the<br />
tube and reptation pictures, the complex entanglement interaction between polymer<br />
chains has been treated in a rather direct approach, i.e., each chain in the polymer system<br />
is equivalent to a chain restricted to one dimensional motion so called ‘‘reptation’’ in a<br />
confining tube, except for its two ends which can move in any possible direction. In<br />
addition to the reptation mechanism, Doi and Edwards originally assumed instantaneous<br />
and complete chain retraction, affine tube deformation by the flow, and independent<br />
a Author to whom correspondence should be addressed.<br />
© 2000 by The Society of Rheology, Inc.<br />
J. Rheol. 446, November/December 2000 0148-6055/2000/446/1293/25/$20.00<br />
1293
1294 FANG, KRÖGER, AND ÖTTINGER<br />
alignment of tube segments. By doing so, they obtained a closed-form constitutive equation<br />
which only involves the second moment of the orientation vector for a tube segment.<br />
For highly entangled, linear polymers, the original Doi–Edwards DE model has been<br />
extended to incorporate chain contour length fluctuations Doi 1983; Ketzmerick and<br />
Öttinger 1989 and constraint release due to the motion of the surrounding chains so<br />
called ‘‘double reptation’’ Tsenoglou 1987; des Cloizeaux 1988. The combination<br />
of these two effects lead to a refined description of the linear viscoelastic properties<br />
O’Connor and Ball 1992, however, the model is much less successful for the nonlinear<br />
properties. The major experimental observations that the original DE theory fails to<br />
describe in the nonlinear regime are the following:<br />
1 There exist irreversible effects in double-step strain experiments with flow reversal<br />
Osaki and Kurata 1980; Osakia et al. 1981; Venerus and Kahvand 1994a, 1994b;<br />
Brown and Burghardt 1996.<br />
2 Over a wide range of shear rates ˙ above the inverse disentanglement time 1/ d ,<br />
the steady shear stress is nearly constant for very highly entangled ones. The first normal<br />
stress difference N 1 increases more rapidly with shear rate than does the shear stress over<br />
the same range of shear rates. The slope of N 1 vs ˙ increases as the molecular weight<br />
decreases Bercea et al. 1993; Kahvand 1995.<br />
3 The steady-state shear viscosity of different molecular weights merge into a single<br />
curve in the high shear rate, power-law regime Stratton 1966.<br />
4 The shear stress xy shows transient overshoots in the startup of steady shear flow<br />
at low shear rates. The strain p at which the maximum in the overshoot occurs increases<br />
with shear rate at high rates Pearson et al. 1989; Menezes and Graessley 1982;<br />
Kahvand 1995.<br />
5 The first normal stress difference exhibits transient overshoots in the startup of<br />
steady shear flow at moderate shear rates Pearson et al. 1989; Menezes and Graessley<br />
1982; Kahvand 1995.<br />
6 The rate of stress relaxation following cessation of steady shear flow is shear rate<br />
dependent Attane et al. 1985; Menezes and Graessley 1982; Kahvand 1995.<br />
7 The steady-state extinction angle decreases more gradually with shear rate than<br />
predicted by the DE model Mead and Larson 1990; Kahvand 1995.<br />
8 The transient extinction angle shows an undershoot at the startup of steady shear at<br />
high shear rates; it also shows an immediate undershoot when the shear rate is suddenly<br />
decreased after a steady state has been reached, finally it reaches a higher steady-state<br />
value Mead 1996; Oberhauser et al. 1998.<br />
9 Steady-state values of the dimensionless uniaxial extensional viscosity are nonmonotonic<br />
functions of extension rate Munstedt and Laun 1981; Ferguson et al.<br />
1997.<br />
In order to improve the situation, many attempts of modifying the original DE model<br />
have been made during the last years. Several physical effects have been found to be<br />
important for more realistic modeling of nonlinear properties of entangled polymers. A<br />
short summary of the important effects found so far is given in Secs. IA–IE.<br />
A. Avoiding independent alignment „IA…<br />
Recognizing that the large discrepancy between model prediction and experimental<br />
data in double step strain with flow reversal is caused by the IA approximation. Doi<br />
1980a, 1980b, and Doi and Edwards 1986 made a detailed analysis for this situation<br />
and tried to derive a constitutive equation without using the IA approximation. Again, the<br />
instantaneous-chain-retraction assumption was employed in their derivation. It was<br />
shown that the model is able to correct the previous discrepancy whenever the time
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1295<br />
interval between the two applied step strains is much larger than a characteristic time s ,<br />
called the retraction time. Marrucci 1986 and Marrucci and Grizzuti 1986 also showed<br />
that the model without IA predicts the Weissenberg effect correctly. In an attempt to find<br />
a thermodynamically admissible formulation, the reptation model without IA has been<br />
reformulated by Öttinger and Beris 1999 in the general equation for the nonequilibrium<br />
reversible-irreversible coupling GENERIC formalism of nonequilibrium thermodynamics<br />
Grmela and Öttinger 1997; Öttinger and Grmela 1997. For a comprehensive<br />
introduction see Öttinger 1999a. The model has been shown to be thermodynamically<br />
consistent after modifying the production term and introducing an additional term in the<br />
extra stress tensor. A consistent model with a uniform monomer distribution along the<br />
chain was proposed by Öttinger 1999b.<br />
B. Double reptation<br />
Tsenoglou 1987 and des Cloizeaux 1988 derived a successful mixing rule for<br />
polydisperse melts based on the idea of ‘‘double reptation,’’ which represents the relaxation<br />
mechanism for the tube that does not arise from motion of the probe chain, but<br />
rather motion from the surrounding chains. Öttinger 1994 has considered a reptation<br />
model in which this additional constraint release mechanism is mimicked through a noise<br />
term in the time-evolution equation for the orientation of inner chain segments. By doing<br />
so, he recovered the correct mixing rule in the linear regime and extended the idea to the<br />
nonlinear regime. But, the prescribed relaxation spectrum is not fully recovered due to<br />
the lack of contour-length fluctuations in the model. The model predicts an improved<br />
power-law index of 4/3 for the steady-state viscosity in shear flow instead of 3/2 for<br />
the DE model. Accordingly, the shear stress versus shear rate curve still exhibits a<br />
maximum when only double reptation is considered.<br />
C. Chain stretching<br />
Marrucci and Grizzuti 1988 extended the DE model to allow for chain stretching and<br />
predicted steady-state properties. An effect is predicted to result in elongational flows,<br />
giving rise to an expected upturn of the elongational viscosity, but surprisingly, there is<br />
no effect for steady shear flows, in particular, the extended model does not improve the<br />
quality of the model prediction for the power-law index of the steady-state shear viscosity.<br />
Later, Pearson et al. 1991 studied the same type of model with chain stretching in<br />
transient situations. They found that both the shear stress and the first normal stress<br />
difference overshoot in the startup of steady shearing flows and the times required to<br />
reach these maxima are independent of the shear rate. Mead and Leal 1995 and Mead<br />
et al. 1995 presented a comprehensive study of the DE model by including chain<br />
stretching and a nonlinear finitely extensible spring law. Numerical predictions were<br />
given for steady two-dimensional flows with a continuously varying degree of extensional<br />
and shear character. The results revealed that significant steady state stretch is<br />
predicted as the flow becomes increasingly extensional in character.<br />
Although chain stretching is important for correcting some of the failings such as an<br />
overshoot in N 1 , it doesn’t solve a long-standing problem in the DE model, namely, the<br />
excessive shear thinning of the viscosity at high shear rates associated with a maximum<br />
in shear stress followed by a region in which shear stress decreases with shear rate<br />
asymptotically as ˙ 0.5 , which leads to constitutive instabilities in shear flow. The<br />
reason for this problem is that, in fast shear flow, the model predicts that the tube<br />
segments become highly oriented in the flow direction and hence present a very slim
1296 FANG, KRÖGER, AND ÖTTINGER<br />
profile to the flow. As a result, the flow ‘‘loses its grip’’ on the molecules, leading to<br />
anomalously low friction and hence low viscosity.<br />
D. Convective constraint release „CCR…<br />
Marrucci 1996 and Ianniruberto and Marrucci 1996 proposed a CCR mechanism<br />
which removes the problem just mentioned above. They proposed a model for which,<br />
under flow conditions, relaxation of chain orientation occurs by two mechanisms. One of<br />
them is ordinary diffusion reptation and double reptation due to thermal motion, which<br />
of course also takes place in the absence of flow. The second mechanism is CCR, i.e., the<br />
topological obstacles on a probe chain are renewed through the relative motion among<br />
chains due to chain retraction. In fast flow situations, this mechanism leaves the chain<br />
much freer to relax than is possible only by the previously described mechanism, and<br />
hence prevents the tube segments from becoming highly oriented in the flow direction.<br />
E. Anisotropic tube cross sections<br />
Ianniruberto and Marrucci 1998 introduced the idea that during deformation an<br />
initial circular tube cross section may become elliptical. They derived the corresponding<br />
expression for the stress tensor, but did not present a time-evolution equation for the tube<br />
cross section in flow. It was shown that the idea of anisotropic tube cross section has an<br />
important influence mainly on the ratio of normal-stress difference in shear flow. In view<br />
of the well know intimate relation between the time evolution of the structural variables<br />
and the stress tensor expression implied by various approaches to nonequilibrium thermodynamics,<br />
Öttinger 2000 developed a thermodynamically admissible reptation<br />
model with anisotropic tube cross section and the constraint release mechanisms associated<br />
with double reptation and CCR. For that model, he proposed relationships between<br />
the ratio of normal-stress differences and the mean-square curvature of the tube cross<br />
section in shear flow.<br />
Very recently, reptation models incorporating all the well-established phenomena except<br />
for anisotropic tube cross sections have been formulated by two groups, based on a<br />
full-chain stochastic approach suitable for computer simulations by Hua and Schieber<br />
1998 and Hua et al. 1998, 1999 and on rather complicated coupled integraldifferential<br />
equations by Mead et al. 1998. It is encouraging that these reptation models<br />
can quite successfully reproduce the experimentally observed rheological behavior in a<br />
large number of flow situations. In a previous work by Öttinger 1999b, which is referred<br />
to as part I of the present work, a new reptation model including anisotropic tube<br />
cross sections, chain stretching, double reptation, and CCR, while avoiding the IA approximation,<br />
has been developed under the guidance of nonequilibrium thermodynamics.<br />
Two versions of the model, referred to as ‘‘uniform’’ and ‘‘tuned,’’ have been proposed.<br />
The purpose of this paper is to give a detailed model evaluation of the recommended<br />
‘‘uniform’’ model in shear and extensional flows by numerical simulations. Because the<br />
experimentally observed features listed in the beginning of Sec. I can be explained<br />
without considering anisotropic tube cross sections their implementation seems to be<br />
nontrivial to us, and they weakly affect the ratio of predicted normal stress differences for<br />
which experimental data is rarely available, the variable Q representing a tube cross<br />
section in the previously proposed model has been omitted here. On this simplified level,<br />
the model has only four degrees of freedom, which will be recalled in Sec. II. Note that<br />
our model takes into account only a single relaxation time reflecting the linear viscoelastic<br />
response of the material. This limitation can be released by considering
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1297<br />
a life span distribution for segments, being deduced from a measured relaxation modulus<br />
as described by Öttinger 2000. In this work, we focus entirely on the basic nonlinear<br />
behavior.<br />
II. MODEL DESCRIPTION<br />
The structural state variables chosen in Part I are the configurational distribution<br />
function f (u,s,r) and the stretching ratio of the chain contour length (r) L/L 0 ,<br />
where u is a unit vector describing the orientation of a tube segment, s is the position<br />
label of a tube segment in the interval 0,1 the values s 0 and s 1 correspond to<br />
the chain ends, r is the space position vector, and L 0 denotes the equilibrium contour<br />
length of the chain. After determining all the state variables the hydrodynamic variables<br />
and the above structural state variables, the model has been formulated in the GENERIC<br />
formalism by constructing its ‘‘building blocks’’ step by step. Here we summarize the<br />
final time-evolution equations for the structural variables.<br />
The equation for the chain contour length stretching and relaxation reads<br />
D<br />
Dt<br />
˙ tot ˙ convect ˙ dissip ,<br />
1<br />
where D/Dt is the material time derivative, and the total stretching rate is split into<br />
convective and dissipative contributions<br />
˙ convect :¯,<br />
2<br />
where is the transpose of the velocity gradient tensor, ¯ is the symmetric secondmoment<br />
orientation tensor defined by<br />
¯ <br />
0<br />
1<br />
ds d 3 ufu,s,ruu.<br />
3<br />
The dissipative contribution reads<br />
˙ dissip 1 s<br />
c<br />
3Z 1,<br />
4<br />
where Z the number of entanglement segments per chain which is given by M/M e M is<br />
the molecular weight of the chain and M e is the average molecular weight between<br />
entanglement points along one chain, s is the characteristic stretching time, and c() is<br />
the effective positive spring coefficient<br />
c 3Z 2<br />
max1<br />
2<br />
max 2 .<br />
5<br />
The spring coefficient is derivable from an entropy expression of the form<br />
3 2 Z ln 2 2<br />
max 1ln 2<br />
max 2<br />
2 ,<br />
max 1<br />
6<br />
such that the spring force F() (1)c() vanishes at 1, and diverges for<br />
→ 0 and → max , respectively, while for small extensions Hookean spring behavior,<br />
and for large extensions FENE type spring behavior is recovered.
1298 FANG, KRÖGER, AND ÖTTINGER<br />
TABLE I. Values for the maximum possible stretching ratio max via Eq.<br />
6, the entanglement molecular weight M e , the average end-to-end distance<br />
of the entanglement segment at equilibrium d t , and the length of a<br />
Kuhn step l K for three important melts at 140 °C.<br />
Melts M e gmol 1 d t (Å) l K (Å) max<br />
PS 13000 76 16.2 4.7<br />
HDPE 860 33 12.9 2.6<br />
PP atactic 5400 61 10.7 5.7<br />
In Part I, the spring coefficient was deduced from an entropy expression containing a<br />
term squared in the relative stretch (1), together with the constraint of positive<br />
extension, thus motivating two additional parameters, called c 1 ,c 2 . Equation 6 is compatible<br />
with Part I, e.g., for a rather particular choice for c 1 together with c 2 0. The<br />
two approaches for the suggested setting c 1 1, c 2 0 possess measurable differences<br />
in their prediction of rheological quantities at startup of shear flow slightly increased<br />
stress amplitude at overshoot and slightly decreased strain at overshoot—for the<br />
model based on Eq. 6, steady shear flow weak decrease of steady stresses, and extensional<br />
flow critical strain rate postponed by a factor of 1.3–1.5, where the current<br />
setting has overall advantages when compared to experimental data for more details see<br />
Sec. IV.<br />
The parameter max clearly is the maximum possible stretching ratio of the chain<br />
contour length, which is equal to the square root of the number of Kuhn steps per<br />
entanglement N KE ,<br />
max N KE d t<br />
l K<br />
,<br />
7<br />
where d t is the average end-to-end distance of the entanglement segment at equilibrium<br />
and l K is the length of the Kuhn step which is twice the persistence length. If we denote<br />
the molecular weight between entanglements for polymer melts as M e , then M e for<br />
polymer solutions can be estimated by using the relation M e M e / 1.2 , where is the<br />
volume fraction of polymer Ferry 1980. The values of d t and l K for some polymer<br />
melts can be calculated from the relevant experimental data tabulated in the literature<br />
Fetters et al. 1996, where the values of M e are also given. The value of l K is calculated<br />
from the characteristic ratio C by using the relationship<br />
C l K<br />
l 1,<br />
8<br />
where l is the bond length Flory 1988, p.111. Based on these available data, the<br />
values of max for several important melts are given in Table I.<br />
The diffusion equation for the configurational distribution function takes the form<br />
Df<br />
Dt <br />
u • 1 uu<br />
u 2 •–uf <br />
<br />
u •D 1 uu<br />
u 2 • <br />
u f ,<br />
˙ s ṡ tot f dissip<br />
<br />
f 1<br />
2 d<br />
2 f<br />
2 s<br />
9
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1299<br />
where<br />
ṡ tot 1 s 1 2˙ dissip .<br />
10<br />
This drift velocity for s means there is only a rescaling of the position label for the tube<br />
segment when the chain relaxes in the tube. The third term creation/destruction term on<br />
the right side of Eq. 9 compensates for configurations lost or gained at the boundaries.<br />
The terms involving second-order derivatives in Eq. 9 are of irreversible nature and<br />
express the erratic reptational motion along the chain contour second-order derivative<br />
with respect to s with the reptation time d and constraint release second-order derivative<br />
with respect to u with the orientational diffusion coefficient D, respectively. The<br />
form of D is<br />
D 1 6 1<br />
1<br />
˙ dissip<br />
<br />
2<br />
d <br />
H ˙ dissip<br />
<br />
, 11<br />
where H(x) is the Heaviside step function. The 1 term is interpreted as representing<br />
‘‘double reptation,’’ and the 2 term represents the CCR mechanism. The quantities 1/ d<br />
and ˙ dissip / determine the constraint release rate due to the loss of entanglements<br />
caused by reptation motion and chain retraction of side chains, respectively. The parameters<br />
1 and 2 determine the transfer rate from the constraint release rate to the relaxation<br />
rate of chain orientation, here we take 1 2 1/. This choice is motivated by<br />
the work of Mead et al. 1998 in connection with the appearance of their switch function.<br />
The argument is that the constraint release causes not only chain segments reorientation,<br />
but also contour length shortening this effect is not explicitly taken into account<br />
here. The role of the parameters 1 and 2 is to apportion the effects of constraint<br />
release between these two effects. When the chain is unstretched, the constraint release<br />
causes only chain segments reorientation; when the chain is highly stretched, the constraint<br />
release causes mainly chain contour length shortening. Hence, the parameters must<br />
be chosen in such a way that they approach unity when is near unity, and approach zero<br />
when is large.<br />
Conservation of the total probability implies the boundary conditions<br />
f˜ f˜<br />
0,<br />
ss 0<br />
ss 1<br />
12<br />
where f˜(s,r) f (u,s,r)d 3 u 1. At the chain ends, we assume random orientation by<br />
specifying the distribution<br />
fu,s,r 1<br />
u1, s 0,1, 13<br />
4<br />
which is common practice, but has been opened to discussion by <strong>Kröger</strong> and Hess 1993.<br />
The extra stress tensor consists of two contributions, 1 2 , namely, the original<br />
Doi–Edwards contribution<br />
1 r 3Zn p k B T<br />
0<br />
1<br />
uuf u,s,rd 3 uds,<br />
14
1300 FANG, KRÖGER, AND ÖTTINGER<br />
which reflects a proportionality between stress and alignment for fixed Z, and a contribution<br />
associated with the chain stretching,<br />
2 r c1n p k B T<br />
0<br />
1<br />
uuf u,s,rd 3 uds,<br />
15<br />
which may cause deviations from the stress-optic rule when stretching occurs. In these<br />
expressions, n p is the number density of polymers. The plateau modulus G N<br />
0 , to which 2<br />
does not contribute, is given by the Doi–Edwards result<br />
G N<br />
0 <br />
3<br />
5 Zn p k B T. 16<br />
When max approaches infinity, we obtain the simple expression<br />
5G N<br />
0 <br />
2<br />
0<br />
1<br />
uuf u,s,rd 3 uds,<br />
17<br />
which may be regarded as a manifestation of the stress-optic rule, see also Fuller 1995.<br />
The presented model accounts for double reptation, CCR, and chain stretching and avoids<br />
the IA approximation by the drift term and the creation/destruction term. It has been<br />
verified to possess the full structure of GENERIC, in particular, the time–structure invariance<br />
of reversible dynamics. The model has only four structural degrees of freedom,<br />
one from the position label s, two from the unit orientation vector u, and one from the<br />
chain stretching . In this paper, we assume d / s 3Z for our model. In addition to<br />
0<br />
the plateau modulus G N and the reptation time d , the model has four parameters<br />
(Z, 1 , 2 , max ) with the preferable values 1 2 1/ for melts as motivated in<br />
0<br />
Part I and above. In this work we regard G N and d as the key adjustable parameters,<br />
while Z and max are fixed by the chemistry of a particular polymer.<br />
III. SIMULATION ALGORITHM<br />
In this section, we describe the numerical simulation of the model. According to the<br />
theory of stochastic differential equations SDEs, the diffusion equation 9, when ignoring<br />
the creation/loss term discussed separately below, is equivalent to the following<br />
set of Itô SDEs for the stochastic processes u t , and s t :<br />
du t 1 u t u t<br />
u t 2 •–u t 2Du tdt2D 1 u t u t<br />
u t 2 •dW t ,<br />
ds t ṡ tot dt 1 2 d<br />
dW t ,<br />
18<br />
19<br />
where W t and all the three components of the vector W t are independent Wiener processes.<br />
The mentioned equivalence means that the average of an arbitrary function<br />
X(u,s), evaluated as an integral with the solution f of the diffusion equation at the time<br />
t, can be obtained as the expectation of the stochastic process X(u t ,s t )<br />
1<br />
<br />
0 Xu,sfu,s,td 3 uds Xu t ,s t .<br />
20
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1301<br />
Here we assume that we consider the problem in a coordinate system moving with an<br />
arbitrary, given fluid particle, and we hence suppress the position argument.<br />
One should note that Eq. 18 is independent of Eq. 19. However, there exists a<br />
coupling between the processes u t and s t resulting from the boundary condition Eq. 13<br />
and the creation/destruction term in Eq. 9, which need to be discussed for obtaining a<br />
full equivalence between diffusion equations and SDEs. Whenever s t reaches the boundaries<br />
0 or 1, u t must be replaced by a random unit vector. Actually, there is a nonzero<br />
drift through the boundaries, which is exactly compensated by the creation/destruction<br />
term in the diffusion equation. Therefore, and in view of Eq. 12, the configurations<br />
diffusing through the boundaries are reflected back into the range of allowed s values.<br />
In order to calculate the statistical error of the results, we perform simulations for a<br />
number of independent blocks (N block ). In each simulation block, a number of trajectories<br />
(N sample ) of the stochastic processes u t and s t are propagated as follows. First, the<br />
drift in Eq. 19 is treated by a deterministic method. Given an s value at time t, the<br />
intermediate value after this treatment is<br />
s sṡ tot tt.<br />
21<br />
If there is a net flux of configurations out of the interval 0, 1, the lost configurations are<br />
randomly replaced by existing ensemble members; if there is a net flux of configurations<br />
into the interval 0, 1, the s values of the gained configurations are set in equal distance<br />
from the two outest existing points to the corresponding chain ends and the u vectors are<br />
randomly oriented according to the boundary condition. In order to keep the ensemble<br />
size constant in the latter case, the same number of configurations are randomly selected<br />
from the existing ensemble members and discarded. This first step takes care of both the<br />
drift in Eq. 19 and the creation/destruction term in Eq. 9. In the next step, we construct<br />
the new configurations at time tt<br />
s new s 1 2 d<br />
W,<br />
22<br />
u u–ut2DW,<br />
u new u<br />
u .<br />
23<br />
24<br />
If s new leaves the interval 0, 1, it will be reflected back into it. Upon any reflection, a<br />
new random unit vector u new is chosen according to the boundary condition. The accuracy<br />
of this straightforward Euler discretization of the stochastic part in Eq. 19 is of<br />
order t which is due to unobserved reflections Öttinger 1989. In order to obtain a<br />
higher-order (t) algorithm, we use the improved scheme proposed by Öttinger 1989,<br />
in which the effect of the unobserved reflections is taken into account through the conditional<br />
probability for their occurrence. Finally, the chain stretching at time tt is<br />
obtained as<br />
new ˙ convect ˙ dissip t.<br />
25<br />
For the results in this paper, we choose N sample 100 000, N block 10, and the time<br />
step size such that the strain is a maximum of 0.02 per time step and the maximum of the<br />
time step is 0.01 for small deformation rates. In each block, transient values are obtained<br />
by ensemble average, and steady values are obtained by average over time after reaching<br />
steady state in addition to ensemble average. The relative statistical errors of the simu-
1302 FANG, KRÖGER, AND ÖTTINGER<br />
lation data are within 2%, except for those with very small absolute values smaller<br />
than 10 3 at the very beginning of the startup of steady flows. There is no detectable<br />
difference in the results by increasing ensemble size or decreasing time step size.<br />
IV. RESULTS<br />
The experimental data used for comparison is from Kahvand 1995 which were also<br />
used by Hua et al. 1999. The test fluid is a solution in tricresyl phosphate of nearly<br />
monodisperse polystyrene with a molecular weight M w of 1.910 6 polydispersity index<br />
of 1.2 at a polymer density of 0.135 g/cm 3 . This fluid has a zero-shear-rate viscosity 0<br />
of 6.810 3 Pa s and a zero-shear-rate first normal stress coefficient 1,0 of 2.0<br />
10 5 Pa s 2 . The longest relaxation time and the average number of entanglement of the<br />
fluid are estimated to be 15 s and 7, respectively. Correspondingly, for our model we will<br />
0<br />
use the set of parameter values Z 7, 21, d 15 s, and G N 1160 Pa beside<br />
1 2 1/, as stated above to compare the simulation results with the experimental<br />
data. The value of the finite extensibility parameter max corresponds to 441 Kuhn steps<br />
per entanglement segment, or 6 monomers per Kuhn step. This value of monomers per<br />
Kuhn step is consistent with the value calculated from the data on polystyrene PS melts<br />
in Table I. The results given here except for extensional flows are not sensitive to finite<br />
extensibility. This set of parameter values is used for most of the cases, otherwise to be<br />
mentioned. In addition to the comparisons with the experimental data, comparisons with<br />
the two other models are also presented. These two models are the full chain simulation<br />
FCS model by Hua and Schieber 1998, Hua et al. 1999, and the simplified S form<br />
of the theory by Mead et al. 1998. Here, we shall denote them as ‘‘FCS’’ model and<br />
‘‘MLDS’’ model, respectively. For the FCS model, the simulations were done by Neergaard<br />
and co-workers, with the parameter b 3N K /(N1) 150, where N K denotes<br />
the number of Kuhn steps per chain and N is the number of beads for the FCS model<br />
compare Table II. The chosen value is more reasonable than the value b 25 used by<br />
Hua et al. 1999, while the other parameters remain the same. The former value of b<br />
corresponds to 150 Kuhn steps per entanglement, or 17 monomers per Kuhn step. This<br />
value of monomers per Kuhn step is larger than the value we estimated above. However,<br />
as for our model, the results given here for the FCS model are not affected significantly<br />
by choosing an even larger value of b. For the MLDS model in which double reptation is<br />
not taken into account, one may regard d and s both as adjustable parameters. The<br />
0<br />
values d 9s, d / s 10, and G N 800 Pa are obtained by a best overall fit to the<br />
available experimental data, where the exponential switch function is used. For d / s<br />
3Z 21 the fit would deteriorate considerably. Prior to further discussion of the<br />
different models, we refer to Table II for a characterization of the three models: FCS,<br />
MLDS, and the one under study.<br />
A. Start-up of steady shearing<br />
Figure 1 shows the growth of the dimensionless shear viscosity for several dimensionless<br />
shear rates predicted by our model. The linear viscoelastic limit forms an upper<br />
envelope for all the shear rates, the transient overshoots occur at medium and high shear<br />
rates, and the undershoots occur only at high shear rates. These undershoots are also<br />
predicted by the FCS model and reported by the experimental results. In Fig. 2 we<br />
compare our model predictions to the experimental data for the transient growth of the<br />
shear viscosity for several shear rates. Actually the model is able to capture the features<br />
of the data quantitatively. The predictions of different models for the same property are<br />
plotted in Fig. 3 where only the curves for the highest shear rate are shown because all
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1303<br />
TABLE II. Characterization of the three models discussed in this work. Concerning the model predictions, the<br />
main discrepancies between FCS and our model are: larger stress overshoot predicted by FCS in the startup of<br />
steady shear flow Fig. 3 and larger second undershoot of the extinction angle predicted by FCS Fig. 12 due<br />
to segment connectivity only incorporated into FCS. The main discrepancies between predictions of MLDS and<br />
our model are: undershoots in transient stress and extinction angle only predicted by our model due to differences<br />
in the incorporation of CCR. For a detailed comparison see text.<br />
Parameters<br />
Model FCS MLDS This work<br />
Adjustable<br />
parameters<br />
The reptation time d<br />
and the number of<br />
beads per chain N fit<br />
from linear viscoelasticity<br />
alone<br />
The plateau modulus<br />
G N<br />
0 , the reptation<br />
time d , and the<br />
Rouse time s fit<br />
from nonlinear data<br />
The plateau modulus<br />
G N<br />
0 and the reptation<br />
time d fit from<br />
nonlinear data<br />
Physical<br />
effects<br />
incorporated<br />
and<br />
corresponding<br />
mathematical<br />
means<br />
Parameters<br />
fixed by the<br />
chemistry of<br />
the polymer<br />
or physical<br />
arguments<br />
Reptation<br />
The number of<br />
entanglements<br />
per chain Z,<br />
the Kuhn step length<br />
a K , and the number<br />
of Kuhn steps per<br />
chain N K<br />
None The number of<br />
entanglements<br />
and the chain<br />
Incorporated into a<br />
set of Langevin equations<br />
for the chain motion<br />
The second term in<br />
the equation for the<br />
tube survival probability,<br />
Eq. 10 of the<br />
reference for MLDS<br />
per chain Z,<br />
the constraint release<br />
parameters 1 , 2 ,<br />
stretchability<br />
parameter max<br />
The stochastic term<br />
in Eq. 19<br />
Avoiding IA<br />
approximation<br />
As for ‘‘reptation’’<br />
plus an equation for<br />
the tube motion<br />
Not considered<br />
The drift term in Eq.<br />
19 and the creation/<br />
destruction term in<br />
Eq. 9<br />
Chain<br />
stretching<br />
As for ‘‘avoiding<br />
IA approximation’’<br />
The equation for<br />
stretch, Eq. 12 of the<br />
reference<br />
The equation for<br />
stretch, Eq. 1<br />
Double<br />
reptation<br />
By a random,<br />
instantaneous constraint<br />
release algorithm<br />
Not considered<br />
The 1 -term in the<br />
orientational diffusion<br />
coefficient, Eq. 11<br />
Convective<br />
constraint<br />
release<br />
CCR<br />
As for ‘‘double<br />
reptation’’<br />
The term with the<br />
‘‘switch’’ function in<br />
Eq. 10 and the last<br />
term in Eq. 12 of the<br />
reference<br />
The 2 -term in the<br />
orientational diffusion<br />
coefficient, Eq. 11<br />
Chain-length<br />
breathing<br />
and segment<br />
connectivity<br />
Incorporated into a<br />
set of Langevin equations<br />
for the chain<br />
motion<br />
Not considered<br />
Not considered<br />
the models give similar results for the two lower shear rates. It turns out that the FCS<br />
model overpredicts the overshoot, our model predicts the magnitude precisely, and the<br />
MLDS model slightly overpredicts the overshoot and does not predict an undershoot.<br />
Comparisons between our model predictions and experimental data are also made for the
1304 FANG, KRÖGER, AND ÖTTINGER<br />
FIG. 1. Dimensionless shear viscosity as function of dimensionless time after startup of steady shearing at<br />
many dimensionless shear rates predicted by our model.<br />
first-normal-stress difference coefficient 1 in Fig. 4. The overshoot in 1 occurs at a<br />
higher shear rate than the shear viscosity does. Here, our model underpredicts the magnitude<br />
of the overshoot at the highest shear rate. Not shown are the predictions by FCS<br />
and MLDS models: they are very comparable with our model for the two lower shear<br />
rates. For the highest rate, again the FCS model overpredicts the overshoot and the<br />
MLDS model gives a better but underpredictive fit. In Fig. 5 we plot the magnitude of the<br />
strain p at which the maximum in the overshoot of the stress occurs as functions of<br />
shear rate predicted by the models and the experiment. It appears that the values of p<br />
stay nearly constant at low shear rates and increase with the increasing shear rate at high<br />
rates. Here, our model predictions are in good agreement with the experimental data,<br />
while the MLDS and FCS models overpredict. Not shown are the model predictions and<br />
FIG. 2. Transient growth of normalized viscosity as function of time under startup of steady shear flow at<br />
several shear rates predicted by our model lines and experiment symbols.
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1305<br />
FIG. 3. Transient growth of normalized viscosity as functions of time under startup of steady shear flow at<br />
˙ 10 s 1 predicted by the models and experiment.<br />
the experimental data of p for the first normal stress difference, which have a qualitative<br />
shape similar to that seen in Fig. 5 but differ by approximately a factor of 2.<br />
B. Cessation of steady shearing<br />
In Fig. 6 we plot the relaxation of the shear stress, normalized by its initial value, after<br />
cessation of steady shear flow for two shear rates. All the model predictions not shown<br />
are the predictions by MLDS which almost overlap with ours compare rather well with<br />
the experimental data. The stress relaxation rate increases with the increasing previous<br />
shear rate. Later in the relaxation process, at times much larger than s , the relaxation<br />
rate is governed only by the reptation process and constraint release due to double<br />
FIG. 4. Transient growth of normalized first-normal-stress difference coefficient as functions of time under<br />
startup of steady shear flow at several shear rates predicted by our model lines and experiment symbols.
1306 FANG, KRÖGER, AND ÖTTINGER<br />
FIG. 5. Magnitude of the strain p at which the maximum in the stress overshoot occurs as functions of shear<br />
rate for shear stress under startup of steady shear flow predicted by the models and experiment.<br />
reptation represented by a smaller value of d in the MLDS model. This behavior is<br />
also in accordance with the other experimental results Attane et al. 1985; Menezes and<br />
Graessley 1982.<br />
C. Steady shear flow<br />
In Fig. 7 the steady-state values of the shear stress xy and the first normal stress<br />
difference N 1 are plotted as functions of shear rate. Our model and the FCS model<br />
predict a slight decrease of the shear stress over a range of shear rates that extends from<br />
roughly ˙ 0.3 to ˙ 2s 1 , where the MLDS model predicts a slight increase which<br />
is consistent with the experimental data. This new observation of the MLDS model may<br />
be attributed to the fact that we use a relatively small ratio d / s 10 here. Over the<br />
same range of shear rates, all the model predictions for the first normal stress difference<br />
FIG. 6. Relaxation of the shear stress reduced by its value at steady state as a function of time after cessation<br />
of steady shear flow at two shear rates predicted by experiment and the models.
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1307<br />
FIG. 7. Steady-state values of the shear stress and first normal stress difference as functions of shear rate<br />
predicted by the models and experiment.<br />
N 1 increase gradually with shear rate, which are consistent with the experimental data.<br />
For the shear stress at rates after the range, our model predicts an increase, the FCS<br />
model shows a lesser increase, and the MLDS model shows a decrease. This decreasing<br />
behavior of MLDS is caused by the exponential switch function used and the additional<br />
term accounting for the CCR effect in the equation for stretch.<br />
We examine the behavior of the steady-state shear stress and its dependence on choice<br />
of our model parameters in more detail in order to understand how to avoid the prediction<br />
of an instability region as mentioned earlier. In Fig. 8, we see that the inclusion of chain<br />
FIG. 8. Steady-state values of the dimensionless shear stress as functions of dimensionless shear rate predicted<br />
by the DE model, the model without any constraint release effects 1 0, 2 0, the model with double<br />
reptation effect 1 1/, 2 0, the model with double reptation and convective constraint release effects<br />
( 1 2 1/), and the model with double reptation and enhanced convective constraint release effects<br />
( 1 1/, 2 2/. The other parameters of the model are set as Z 20, and max 10.
1308 FANG, KRÖGER, AND ÖTTINGER<br />
FIG. 9. Steady-state shear viscosity vs dimensionless shear rate which here is proportional to a physical shear<br />
rate rather than expressed in terms of a relaxation time which depends on molecular weight, for Z 20 and<br />
Z 40 with the other parameters 1 1/, 2 1/, and max 10.<br />
stretching alone yields the onset of an upturn of the shear stress at high shear rates, but<br />
the prediction of the power-law index is worse. When double reptation is considered, the<br />
situation is improved. After CCR is added and adjusted to proper intensity, the maximum<br />
in the steady-state shear stress disappears and the model predicts an almost constant<br />
stress over a wide range of shear rates, which is also observed in experiments of a highly<br />
entangled polymer solution Bercea et al. 1993. We thus confirm the previous conclusion<br />
that the CCR effect is critical for predicting the right shape of the shear stress curve.<br />
In Fig. 9 we plot the steady-state shear viscosity for Z 20 and Z 40 in order to<br />
resolve the effect of molecular weight. As a basis for comparison, we assume that the<br />
reptation time scales like d M 3 . It turns out that the steady-state viscosity of two<br />
different molecular weights coincides in the power-law region, in agreement with the<br />
behavior observed in experiments Stratton 1966 and, e.g., molecular dynamics simulations<br />
<strong>Kröger</strong> et al. 1993. Similar results are found by the FCS and MLDS models.<br />
The effect of molecular weight on N 1 is reported in Fig. 10, the model prediction is in<br />
agreement with the second part of observation 2 in Sec. I.<br />
D. Steady and transient extinction angle<br />
The extinction angle is given by (1/2)arctan(2 xy /N 1 ). In Fig. 11 we show the<br />
steady-state extinction angle as a function of shear rate. Obviously, in our model, CCR<br />
and double reptation prevent the chain segments from aligning with flow dramatically at<br />
high steady shear rates, and thus solve the overestimated steady shear orientation problem<br />
of the original DE model. The predictions of our model compare very favorably with the<br />
experimental results, while the predictions of the FCS and MLDS models go back to the<br />
DE model predictions at very high shear rates in steady flow. Not shown here are the<br />
predictions of our model with the parameters 1 2 exp((1)), which also coincide<br />
with the predictions of the DE model at high rates. It can be concluded from the<br />
above observations that the exponential switch function tunes down the constraint release<br />
effect on orientation too strongly, and hence is not appropriate. The reorientation algorithm<br />
in the FCS model for constraint release is successful at low and intermediate rates,
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1309<br />
FIG. 10. Steady-state first-normal-stress difference vs dimensionless shear rate for Z 20 and Z 40 with<br />
the other parameter values as for Fig. 9.<br />
but may profit from an adjustment if a plateau for the steady time-averaged extinction<br />
angle at high rates caused by unsteady rotation of chains would be experimentally confirmed.<br />
In Fig. 12 we show the transient extinction angle predicted in the startup of shear flow<br />
followed by a step down in shear rate. In both predictions by our model and the FCS<br />
model, the extinction angle shows an undershoot at the startup of the first shear rate and<br />
an immediate undershoot at the inception of the lower shear rate before reaching a<br />
steady-state value. These predictions are in agreement with the experimental data Mead<br />
1996; Oberhauser et al. 1998. However, the second undershoot predicted by our<br />
model is too small compared with both the experiment and the prediction of the FCS<br />
FIG. 11. Steady-state extinction angle as a function of shear rate predicted by the models and experiment.
1310 FANG, KRÖGER, AND ÖTTINGER<br />
FIG. 12. Time-dependent extinction angle as a function of dimensionless time after steady shearing and<br />
attainment of steady state followed by a step down in dimensionless shear rate predicted by our model and the<br />
FCS model.<br />
model. The MLDS model with the switch function either exp((1)) or 1/ does not<br />
possess these two observed undershoots. The undershoots predicted by our model is due<br />
to chain retraction and its effect on chain orientation through the noise term in Eq. 18.<br />
E. Extensional flows<br />
The steady-state uniaxial extensional viscosity ( 1 ( zz xx )/(3˙ )) as a function<br />
of extensional rate is considered next, see Fig. 13, where z is the direction of extension.<br />
One observes four regimes of extension rate: i a very low rate region, ˙ 0.1/ d ,<br />
where the extensional viscosity is a constant equal to one third of the Trouton value; ii<br />
FIG. 13. Steady-state values of the dimensionless uniaxial extensional viscosity as function of dimensionless<br />
extension rate for 1 2 1/, Z 20, and max 10. The figure inside is the amplification of the region<br />
of extension rate range from 0.01 to 2.
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1311<br />
FIG. 14. Dimensionless uniaxial extensional viscosity vs dimensionless time after startup of steady uniaxial<br />
elongation at many dimensionless elongational rates with the parameter values as for Fig. 13.<br />
a low rate region, 0.1/ d ˙ 1/ d , where the viscosity shows a weak extension<br />
hardening induced by orientation the maximum factor above the first constant region is<br />
around 1.05 here; iii an intermediate region, 1/ d ˙ 1/ s , in which the extensional<br />
viscosity is extension thinning, scaling as ˙ 1 ; and iv a high rate region, ˙<br />
1/ s , where the extensional viscosity increases with increasing strain rate and<br />
reaches another constant value finally. The same conclusion for regimes i, iii, and iv<br />
has been drawn from the MLDS model by Mead et al. 1998. Regime ii is caused by<br />
double reptation, because at low strain rates the other effects have a minor influence.<br />
Accordingly, regime ii is not predicted by the MLDS model and DE theory. Because of<br />
the lack of experimental data on monodisperse, linear entangled polymers, these unusual<br />
predictions have not yet been fully established. Two of the few works showing a pronounced<br />
regime ii were performed by Munstedt and Laun 1981 and Ferguson et al.<br />
1997. The first of these papers stated that two of the four HDPE investigated in their<br />
study do not show a maximum of the steady-state elongational viscosity; for the others it<br />
is less pronounced than in the case of LDPE.<br />
In Fig. 14 we plot the transient elongational viscosity as a function of time in the<br />
startup of uniaxial extensional flow. After the extensional rate exceeds 1/ s , the viscosity<br />
shows strain-hardening behavior and finally saturates because of the finite chain stretchability.<br />
To study the strain-hardening property of our model in detail, a strain-hardening<br />
parameter versus Hencky strain at the extensional rate of 100 is plotted in Fig. 15. The<br />
strain-hardening parameter is defined as the ratio of the strain rate-dependent nonlinear<br />
elongational viscosity to the strain rate-independent linear elongational viscosity at the<br />
same time Koyama and Ishizuka 1983. The figure reveals that the first departure from<br />
linearity occurs at a Hencky strain value around 0.5, fully in accord with the experimentally<br />
observed value by Takahashi et al. 1999. Figure 16 shows the influence of the<br />
maximum stretch max on the transient uniaxial extensional viscosity. An increase of<br />
max does not affect the occurrence of strain hardening, but postpones the saturation and<br />
increases the saturation value. Not shown are the predictions of the steady-state and<br />
transient viscosity in equibiaxial extension, which have qualitative shapes similar to those
1312 FANG, KRÖGER, AND ÖTTINGER<br />
FIG. 15. The strain-hardening parameter as a function of the Hencky strain at the dimensionless elongational<br />
rate ˙ d 100 with the parameter values as for Fig. 13.<br />
seen in the uniaxial extension, except that i the second regime of steady-state viscosity<br />
is absent and ii it is ‘‘softer’’ and less strain hardening than uniaxial extension Nishioka<br />
et al. 2000. Simulations of planar extension of our model are also performed. The<br />
steady-state planar viscosity p1 ( zz xx )/(4˙ ) direction of extension z and<br />
p2 ( yy xx )/(2˙ ) neutral direction y are plotted in Fig. 17. The viscosity p2<br />
decays much faster than the p1 does in the intermediate region, 1/ d ˙ 1/ s .<br />
Again, there is no second regime for both of them. In Fig. 18 we show the transient<br />
FIG. 16. Dimensionless uniaxial extensional viscosity vs dimensionless time after startup of steady uniaxial<br />
elongation at the dimensionless elongational rate ˙ d 200 for 1 2 1/, Z 20, and three different<br />
values of max .
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1313<br />
FIG. 17. Steady-state values of the dimensionless planar extensional viscosity p1 along the direction of<br />
extension z and p2 along the neutral direction y as functions of dimensionless extensional rate with the<br />
parameter values as for Fig. 13.<br />
<br />
viscosity p2 not shown is the<br />
p1 , which is similar to the transient viscosity in<br />
uniaxial extension. The model predicts overshoots at medium and high extensional rates.<br />
V. CONCLUSION<br />
A new thermodynamically admissible reptation model that includes chain stretching,<br />
double reptation, and convective constraint release, and that avoids IA approximation is<br />
numerically investigated for transient and steady properties in shear and extensional<br />
flows. Quantitative comparisons are made with experimental data of entangled PS solution<br />
in shear flows. We find that the model is able to capture qualitatively, or quantita-<br />
FIG. 18. Dimensionless planar extensional viscosity p2 along the neutral direction y versus dimensionless<br />
time after startup of steady planar elongation at many dimensionless elongational rates with the parameter<br />
values as for Fig. 13.
1314 FANG, KRÖGER, AND ÖTTINGER<br />
tively in most cases, all the nonlinear properties observed for shear flows summarized in<br />
Sec. I from item 2 to item 9 item 1 has already been checked in Part I. For extensional<br />
flows, the model exhibits unusual predictions. In order to confirm these, further measurements<br />
of the stress for ideal samples linear, monodisperse, entangled polymers in various<br />
extensional flows are required. There are difficulties in performing this kind of<br />
experiment: a monodisperse polymer is usually sold in powder form, which is often more<br />
difficult to shape into bubble free samples than granules; the polymer breaks at much<br />
lower Hencky strains and is less homogeneous than an industrial melt Schweizer<br />
1999.<br />
Comparisons with two recently formulated models, FCS Hua and Schieber 1998<br />
and MLDS Mead et al. 1998, respectively, in shear flows are also presented. It turns<br />
out that all three models are competitive with each other; in many cases, they show<br />
similar behaviors. Very detailed comparisons are given in the previous section, e.g., for<br />
the MLDS and our model it is shown that a weak switch function proportional to the<br />
inverse of relative stretch 1/ predicts more realistic results than obtained by applying<br />
a stronger one.<br />
In addition to the plateau modulus G N<br />
0 and the reptation time d , our model has two<br />
basic parameters, namely Z, the number of entanglements per chain which is proportional<br />
to the molecular weight and can be expressed in terms of the Rouse and reptation time,<br />
and the maximum stretch max , being equal to the square root of the number of Kuhn<br />
steps per entanglement segment. The remaining two model parameters 1 , 2 have been<br />
introduced but actually assigned the values 1 2 1/.<br />
A ‘‘stable’’ steady-state shear stress curve is obtained by setting 2 2/. Predictions<br />
of the steady shear stress at very high shear rates can be also improved by considering<br />
the stretching time s to be a decaying function in , to account for the effect of<br />
contour length shortening by constraint release in situations where highly stretched conformations<br />
occur.<br />
Notably, our model is a thermodynamically consistent single-segment theory which<br />
has only four degrees of freedom. In the sense of considering the time and memory<br />
requirements for the computation, it is very suitable to utilize the model to simulate<br />
complex flows by using the CONNFFESSIT idea Laso and Öttinger 1993. To our<br />
knowledge, variance reduction techniques Hulsen et al. 1997; Öttinger et al. 1997;<br />
van Heel et al. 1999; Gallez et al. 1999; and Bonvin and Picasso 1999 may not be<br />
directly applied to the model. The dynamics for the position label—and in particular its<br />
jump events upon touching the chain ends—depend on the local value for the macroscopic<br />
velocity gradient such that a synchronization of trajectories by cancellation of<br />
fluctuations is prevented at first glance.<br />
However, it should be a straightforward exercise to extend the presented model to<br />
account for polydispersity effects by a mean-field approach, such that the properties of<br />
surrounding chains enter into the dynamics of the test chain. The resulting model will<br />
necessarily locate beyond simple superposition models.<br />
ACKNOWLEDGMENTS<br />
The authors would like to thank Professor D. C. Venerus for providing the experimental<br />
data, Professor R. G. Larson for providing the simulation program of the MLDS<br />
model, J. Neergaard for providing calculations of the FCS model, and Professor J. D.<br />
Schieber for extremely helpful comments. J. Fang is grateful for the financial support<br />
from the European Community Program BRITE-EuRam.
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1315<br />
References<br />
Attané, P., M. Pierrand, and G. Turrel, ‘‘Steady and transient shear flows of polystyrene solutions I: Concentration<br />
and molecular weight dependence of non-dimensional viscometric functions,’’ J. Non-Newtonian<br />
Fluid Mech. 18, 295–317 1985.<br />
Bercea, M., C. Peiti, B. Dimionescu, and P. Navard, ‘‘Shear rheology of semidilute polymethylmethacrylate<br />
solutions,’’ Macromolecules 26, 7095–7096 1993.<br />
Bonvin, J. and M. Picasso, ‘‘Variance reduction methods for CONNFFESSIT-like simulations,’’ J. Non-<br />
Newtonian Fluid Mech. 84, 191–215 1999.<br />
Brown, E. F. and W. R. Burghardt, ‘‘First and second normal stress difference relaxation in reversing doublestep<br />
strain flows,’’ J. Rheol. 40, 37–54 1996.<br />
de Gennes, P. G., ‘‘Reptation of a polymer chain in the presence of fixed obstacles,’’ J. Chem. Phys. 55,<br />
572–579 1971.<br />
des Cloizeaux, J., ‘‘Double reptation vs simple reptation in polymer melts,’’ Europhys. Lett. 5, 437–442 1988.<br />
Doi, M., ‘‘Stress relaxation of polymeric liquids after double step strain,’’ J. Polym. Sci., Polym. Phys. Ed. 18,<br />
1891–1905 1980a.<br />
Doi, M., ‘‘A constitutive equation derived from the model of Doi and Edwards for concentrated polymer<br />
solutions and polymer melts,’’ J. Polym. Sci., Polym. Phys. Ed. 18, 2055–2067 1980b.<br />
Doi, M., ‘‘Explanation for the 3.4-power law for viscosity of polymeric liquids on the basis of the tube model,’’<br />
J. Polym. Sci., Polym. Phys. Ed. 21, 667–684 1983.<br />
Doi, M. and S. F. Edwards, ‘‘Dynamics of concentrated polymer systems. Part 1. Brownian motion in the<br />
equilibrium state,’’ J. Chem. Soc., Faraday Trans. 2 74, 1789–1801 1978a.<br />
Doi, M. and S. F. Edwards, ‘‘Dynamics of concentrated polymer systems. Part 2. Molecular motion under<br />
flow,’’ J. Chem. Soc., Faraday Trans. 2 74, 1802–1817 1978b.<br />
Doi, M. and S. F. Edwards, ‘‘Dynamics of concentrated polymer systems. Part 3. The constitutive equation,’’<br />
J. Chem. Soc., Faraday Trans. 2 74, 1818–1832 1978c.<br />
Doi, M. and S. F. Edwards, ‘‘Dynamics of concentrated polymer systems. Part 4. Rheological properties,’’ J.<br />
Chem. Soc., Faraday Trans. 2 75, 38–54 1979.<br />
Doi, M. and S. F. Edward, The Theory of Polymer Dynamics Clarendon, Oxford, 1986.<br />
Ferguson, J., N. E. Hudson, and M. A. Odriozola, ‘‘Interpretation of transient extensional viscosity data,’’ J.<br />
Non-Newtonian Fluid Mech. 68, 241–257 1997.<br />
Ferry, J. D., Viscoelastic Properties of Polymers Wiley, New York, 1980.<br />
Fetters, L. J., D. J. Lohse, and R. H. Colby, ‘‘Chain dimensions and entanglement spacings,’’ in Physical<br />
Properties of Polymers Handbook, edited by J. E. Mark AIP, New York, 1996.<br />
Flory, P. J., Statistical Mechanics of Chain Molecules Hanser, Munich, 1988.<br />
Fuller, G. G., Optical Rheometry of Complex Fluids Oxford University Press, Oxford, U.K., 1995<br />
Gallez, X., P. Halin, G. Lielens, R. Keunings, and V. Legat, ‘‘The adaptive Lagrangian particle method for<br />
macroscopic and micro-macro computations of time-dependent viscoelastic flows,’’ Comput. Methods<br />
Appl. Mech. Eng. 68, 345–364 1999.<br />
Grmela, M. and H. C. Öttinger, ‘‘Dynamics and thermodynamics of complex fluids. I. Development of a<br />
general formalism,’’ Phys. Rev. E 56, 6620–6632 1997.<br />
Hua, C. C. and J. D. Schieber, ‘‘Segment connectivity, chain-length breathing, segmental stretch, and constraint<br />
release in reptation models. I. Theory and single-step strain predictions,’’ J. Chem. Phys. 109, 10018–10027<br />
1998.<br />
Hua, C. C., J. D. Schieber, and D. C. Venerus, ‘‘Segment connectivity, chain-length breathing, segmental<br />
stretch, and constraint release in reptation models. II. Double-step strain predictions,’’ J. Chem. Phys. 109,<br />
10028–10032 1998.<br />
Hua, C. C., J. D. Schieber, and D. C. Venerus, ‘‘Segment connectivity, chain-length breathing, segmental<br />
stretch, and constraint release in reptation models. III. Shear flows,’’ J. Rheol. 43, 701–717 1999.<br />
Hulsen, M. A., A. P. G. van Heel, and B. H. A. A. van den Brule, ‘‘Simulation of viscoelastic flows using<br />
Brownian configuration fields,’’ J. Non-Newtonian Fluid Mech. 70, 79–101 1997.<br />
Ianniruberto, G., and G. Marrucci, ‘‘On compatibility of the Cox-Merz rule with the model of Doi and Edwards,’’<br />
J. Non-Newtonian Fluid Mech. 65, 241–246 1996.<br />
Ianniruberto, G. and G. Marrucci, ‘‘Stress tensor and stress-optical law in entangled polymers,’’ J. Non-<br />
Newtonian Fluid Mech. 79, 225–234 1998.<br />
Kahvand, H., ‘‘Strain Coupling Effects in Polymer Rheology,’’ Ph.D. thesis, Illinois Institute of Technology,<br />
1995.<br />
Ketzmerick, R. and H. C. Öttinger, ‘‘Simulation of a Non-Markovian process modelling contour length fluctuation<br />
in the Doi-Edwards model,’’ Continuum Mech. Thermodyn. 1, 113–124 1989.<br />
Koyama, K. and O. Ishizuka, ‘‘Nonlinearity in uniaxial elongational viscosity at a constant strain rate,’’ Polym.<br />
Proc. Eng. 1, 55–70 1983.<br />
<strong>Kröger</strong>, M. and S. Hess, ‘‘Viscoelasticity of polymeric melts and concentrated solutions. The effect of flowinduced<br />
alignment of chain ends,’’ Physica A 195, 336–353 1993.
1316 FANG, KRÖGER, AND ÖTTINGER<br />
<strong>Kröger</strong>, M., W. Loose, and S. Hess, ‘‘Rheology and structural changes of polymer melts via nonequilibrium<br />
molecular dynamics,’’ J. Rheol. 37, 1057–1079 1993.<br />
Laso, M. and H. C. Öttinger, ‘‘Calculation of viscoelastic flow using molecular models: the CONNFFESSIT<br />
approach,’’ J. Non-Newtonian Fluid Mech. 47, 1–20 1993.<br />
Marrucci, G., ‘‘The Doi-Edwards model without independent alignment,’’ J. Non-Newtonian Fluid Mech. 21,<br />
329–336 1986.<br />
Marrucci, G., ‘‘Dynamics of entanglements: A nonlinear model consistent with the Cox-Merz rule,’’ J. Non-<br />
Newtonian Fluid Mech. 62, 279–289 1996.<br />
Marrucci, G. and N. Grizzuti, ‘‘The Doi-Edwards model in slow flows. Predictions on the weissenberg effect,’’<br />
J. Non-Newtonian Fluid Mech. 21, 319–328 1986.<br />
Marrucci, G. and N. Grizzuti, ‘‘Fast flow of concentrated polymers: Predictions of the tube model on chain<br />
stretching,’’ Gazz. Chim. Ital. 118, 179–185 1988.<br />
Mead, D. W., ‘‘Orientation angle transients in step shear rate experiments on lightly entangled systems of linear<br />
flexible polymers,’’ In Proceeding of the 12th International Congress on Rheology, Aug. 18–23, 1996,<br />
edited by Aït-Kadi, A., J. M. Dealy, D. F. James, and M. C. Williams Canadian Rheology Group, Quebec<br />
City, Canada, 1996.<br />
Mead, D. W. and R. G. Larson, ‘‘Rheooptical study of isotropic solutions of stiff polymers,’’ Macromolecules<br />
23, 2524–2533 1990.<br />
Mead, D. W. and L. G. Leal, ‘‘The reptation model with segmental stretch I. Basic equations and general<br />
properties,’’ Rheol. Acta 34, 339–359 1995.<br />
Mead, D. W., D. Yavich, and L. G. Leal, ‘‘The reptation model with segmental stretch II. Steady flow<br />
properties,’’ Rheol. Acta 34, 360–383 1995.<br />
Mead, D. W., R. G. Larson, and M. Doi, ‘‘A molecular theory for fast flows of entangled polymers,’’ Macromolecules<br />
31, 7895–7914 1998.<br />
Menezes, E. V. and W. W. Graessley, ‘‘Nonlinear rheological behavior of polymer systems for several shear<br />
flow histories,’’ J. Polym. Sci., Polym. Phys. Ed. 20, 1817–1833 1982.<br />
Munstedt, H. and H. M. Laun, ‘‘Elongational properties and molecular structure of polyethylene melts,’’ Rheol.<br />
Acta 20, 211–221 1981.<br />
Nishioka, A., T. Takahashi, Y. Masubuchi, J.-I. Takimoto, and K. Koyama, ‘‘Description of uniaxial, biaxial,<br />
and planar elongational viscosities of polystyrene melt by the K-BKZ model,’’ J. Non-Newtonian Fluid<br />
Mech. 89, 287–301 2000.<br />
Oberhauser, J. P., L. G. Leal, and D. W. Mead, ‘‘The response of entangled polymer solutions to step changes<br />
of shear rate: signatures of segmental stretch?,’’ J. Polym. Sci., Part B: Polym. Phys. 36, 265–280 1998.<br />
O’Connor, N. P. T. and R. C. Ball, ‘‘Confirmation of the Doi-Edwards model,’’ Macromolecules 25, 5677–<br />
5682 1992.<br />
Osaki, K. and M. Kurata, ‘‘Experimental appraisal of the Doi-Edwards theory for polymer rheology based on<br />
the data for polystyrene solutions,’’ Macromolecules 13, 671–676 1980.<br />
Osaki, K., S. Kimura, and M. Kurata, ‘‘Relaxation of shear and normal stresses in double-step shear deformations<br />
for a polystyrene solution. A test of the Doi-Edwards theory for polymer rheology,’’ J. Rheol. 25,<br />
549–562 1981.<br />
Öttinger, H. C., ‘‘Computer simulation of reptation theories. I. Doi-Edwards and Curtiss-Bird models,’’ J.<br />
Chem. Phys. 91, 6455–6462 1989.<br />
Öttinger, H. C., ‘‘Modified reptation model,’’ Phys. Rev. E 50, 4891–4895 1994.<br />
Öttinger, H. C., ‘‘Nonequilibrium thermodynamics—A tool for applied rheologists,’’ Appl. Rheol. 9, 17–26<br />
1999a.<br />
Öttinger, H. C., ‘‘A thermodynamically admissible reptation model for fast flows of entangled polymers,’’ J.<br />
Rheol. 43, 1461–1493 1999b.<br />
Öttinger, H. C., ‘‘Thermodynamically admissible reptation models with anisotropic tube cross sections and<br />
convective constraint release,’’ J. Non-Newtonian Fluid Mech. 89, 165–185 2000.<br />
Öttinger, H. C. and A. N. Beris, ‘‘Thermodynamically consistent reptation model without independent alignment,’’<br />
J. Chem. Phys. 110, 6593–6596 1999.<br />
Öttinger, H. C. and M. Grmela, ‘‘Dynamics and thermodynamics of complex fluids. II. Illustrations of a general<br />
formalism,’’ Phys. Rev. E 56, 6633–6655 1997.<br />
Öttinger, H. C., B. H. A. A. van den Brule, and M. A. Hulsen, ‘‘Brownian configuration fields and variance<br />
reduced CONNFFESSIT,’’ J. Non-Newtonian Fluid Mech. 70, 255–261 1997.<br />
Pearson, D. S., A. D. Kiss, and L. J. Fetters, ‘‘Flow-induced birefringence of concentrated polyisoprene solutions,’’<br />
J. Rheol. 33, 517–535 1989.<br />
Pearson, D. S., E. A. Herbolzheimer, G. Marrucci, and N. Grizzuti, ‘‘Transient behavior of entangled polymers<br />
at high shear rates,’’ J. Polym. Sci., Polym. Phys. Ed. 29, 1589–1597 1991.<br />
Schweizer, T. personal communication, 1999.<br />
Stratton, R. A., ‘‘The dependence of non-Newtonian viscosity on molecular weight for ‘monodisperse’ polystyrene,’’<br />
J. Colloid Interface Sci. 22, 517–530 1966.
THERMODYNAMICALLY ADMISSIBLE REPTATION MODEL<br />
1317<br />
Takahashi, T., J. Takimoto, and K. Koyama, ‘‘Elongational viscosity for miscible and immiscible polymer<br />
blends. I. PMMA and AS with similar elongational viscosity,’’ J. Appl. Polym. Sci. 73, 757–766 1999.<br />
Tsenoglou, C., ‘‘Viscoelasticity of binary homopolymer blends,’’ ACS Polym. Preprints 28, 185–186 1987.<br />
van Heel, A. P. G., M. A. Hulsen, and B. H. A. A. van den Brule, ‘‘Simulation of the Doi–Edwards model in<br />
complex flow,’’ J. Rheol. 43, 1239–1260 1999.<br />
Venerus, D. C. and H. Kahvand, ‘‘Doi–Edwards theory evaluation in double-step strain flows,’’ J. Polym. Sci.,<br />
Part B: Polym. Phys. 32, 1531–1542 1994a.<br />
Venerus, D. C. and H. Kahvand, ‘‘Normal stress relaxation in reversing double-step strain flows,’’ J. Rheol. 38,<br />
1297–1315 1994b.