Вы находитесь на странице: 1из 41

Role of slip and fracture in the oscillating flow of HDPE in a capillary

Savvas G. Hatzikiriakos and , and John M. Dealy

Citation: Journal of Rheology 36, 845 (1992); doi: 10.1122/1.550320


View online: http://dx.doi.org/10.1122/1.550320
View Table of Contents: http://sor.scitation.org/toc/jor/36/5
Published by the The Society of Rheology

Articles you may be interested in


Wall slip of molten high density polyethylenes. II. Capillary rheometer studies
Journal of Rheology 36, 703 (1998); 10.1122/1.550313

Wall Slip and Extrudate Distortion in Linear Low-Density Polyethylene


Journal of Rheology 31, 815 (2000); 10.1122/1.549942

Wall slip of molten high density polyethylene. I. Sliding plate rheometer studies
Journal of Rheology 35, 497 (1998); 10.1122/1.550178

Explicit Formulas for Slip and Fluidity


Journal of Rheology 2, 210 (2005); 10.1122/1.2116364

Surface fractionation effects on slip of polydisperse polymer melts


Physics of Fluids 28, 093101 (2016); 10.1063/1.4962564

Wall slip and extrudate distortion of three polymer melts


Journal of Rheology 47, 683 (2003); 10.1122/1.1562156
Role of slip and fracture in the oscillating flow
of HDPE in a capillary

Savvas G. Hatzikiriakosa) and John M. Dealy


Department of Chemical Engineering, McGill University,
Montreal H3A 2A 7, Canada

(Received 16 September 1991; accepted 9 March 1992)

Synopsis

Certain polymers exhibit two distinct branches in their capillary flow curves
(wall shear stress versus apparent wall shear rate). This gives rise to oscillatory
flow in constant-piston-speed rheometers and to flow curve hysteresis in
controlled-pressure rheometers. These curious phenomena have attracted con-
siderable interest over a period of many years, but their basic mechanisms are
still the subject of debate. Building on previous work we have developed a model
that predicts all the essential features of the curves of pressure and flow rate
versus time in the oscillatory flow regime. Fluid compressibility and the second
branch of the flow curve are necessary features of the model, but fluid elasticity
is found not to be an essential element. While our macroscopic measurements do
not prove it conclusively, our data lead us to believe that on the high-flow-rate
branch of the flow curve there is slip along a cylindrical fracture surface near the
wall. The jump to the high-flow branch occurs when this fracture occurs, at an
upper critical value of the shear stress, while the jump back to the low-flow
branch occurs when adhesion is established at the fracture surface at a lower
critical shear stress.

CAPILLARY FLOW OF MOLTEN POLYMERS

The flow curve


The flow curve of linear polymers, i.e., the plot of wall shear stress
versus apparent shear rate that is obtained by means of a capillary
rheometer, contains several interesting features that cannot be explained
strictly on the basis of a viscosity that decreases with shear rate. These

a)Current address: Department of Chemical Engineering, University of British Columbia,


Vancouver, BC V6T IZ4, Canada.

@ 1992 by The Society of Rheology, Inc.


J. Rheol. 36(5), July 1992 0148-6055/92/050845-40$04.00 845
846 HATZIKIRIAKOS AND DEALY

features of the flow curve are often associated with other observations
such as extrudate distortion and fluctuations in the pressure and the
flow rate. Explanations of these phenomena have been proposed on the
basis of various combinations of melt compressibility, melt fracture in
the reservoir or in the capillary, wall slip, melt viscoelasticity, a maxi-
mum in the shear stress-shear rate curve, and viscous heating. How-
ever, this field is still a controversial one, and we will introduce the
subject here by describing first empirical observations without specula-
tion about cause and effect. Then we will review briefly the various
explanations that have been proposed.
First we look at the flow curve itself, which is a plot of wall shear
stress a w versus apparent wall shear rate rA' The wall shear stress can
be calculated using Eq. (1):

a w = (P d-Pen)D/4L, (1)

where a w is the magnitude of the wall shear stress, Pd is the driving


pressure determined from the force on the piston, and Pen is the end
correction for the pressure drop. Equation (1) is based on the assump-
tion that the shear stress is uniform along the capillary. However, we
have argued (Hatzikiriakos and Dealy, 1992a) that this assumption is
not valid when slip is occurring. We will therefore refer to the quantity
calculated in this way as the "nominal shear stress." Flow curves are
often prepared without making the end correction, and in this case it is
an "apparent" wall shear stress that is plotted:

aA=PdD/ 4L. (2)


The apparent wall shear rate is calculated from the volumetric flow rate
Q as follows:
3 (3)
rA=32Q/(1TD ) .

(This would be the true wall shear rate if the fluid were Newtonian and
there were no slip.)
To illustrate this discussion it will be convenient to refer to Fig. 1,
which shows data obtained in this study. The procedures used to obtain
the curve and a detailed discussion of its features will be presented in a
later section. Over a certain shear rate range there is, for most polymers,
a power law regime in which the curve of log(aw) vs log ( rAj is a
straight line. For most linear polymers that have been studied, there is
a critical value of the wall shear stress, acl> usually within the power law
regime, at which there is a change of slope, and data obtained using
SLIP, FRACTURE, AND OSCILLATORY FLOWS 847

0.60 RESIN A
D=0.0254 em and L/D=60 m
T=160C
[!]
A OSCIlLATING FLOW
0.40
m STEADY FLOW

-. m
0
0-
2 A B
gJ
o
3
0.20
..ti" ~
L!:.
L!:.

D C
[1]

[!]
0.10 ~

.1 .1 .1
0.08
10 100 1000 10000 100000
1
'YA (5- )

FIG. 1. The apparent flow curve of resin "A" showing two distinct branches. No steady
flow is possible for apparent shear rates between those corresponding to points A and C.

capillaries of different diameters start to diverge. There is strong evi-


dence that this is due to wall slip (Lupton and Regester, 1965; Rama-
murthy, 1986; Kalika and Denn, 1987; Hatzikiriakos and Dealy,
1992a).
When the wall shear stress reaches an "upper critical" value, a c2'
(point A in Fig. 1) the flow rate suddenly jumps from A to B, the
so-called "spurt" effect. What happens if the stress is further increased
depends on the mode of operation of the rheometer. If it is the pressure
that is controlled, the flow curve jumps to a second, high-flow-rate
branch, and continues along it. If the pressure is now decreased, the
flow rate decreases continuously until it reaches a "lower critical
stress," ac3 (point C in Fig. 1) at which the flow rate suddenly de-
creases, jumping from C to D, and the system jumps back to the low-
flow rate branch of the flow curve (Bagley et al., 1958; Vinogradov et
ai., 1972). This effect we will refer to as "flow curve hysteresis."
848 HATZIKIRIAKOS AND DEALY

If it is the piston speed that is controlled, the case considered in detail


here, for the range of piston speeds between those associated with steady
flow with shear stresses of Uc2 and Uc3, i.e., between the values corre-
sponding to points A and C in Fig. 1, the pressure and the extrudate
flow rate oscillate between limiting values. The frequency of the oscil-
lations increases as the volume of melt in the reservoir decreases. As the
piston speed is increased, oscillations continue until the apparent shear
rate exceeds the value corresponding to the lower critical shear stress,
Uc3, (point C in Fig. 1) at which point the oscillations cease, and the
curve progresses along the high-flow-rate branch of the flow curve, in
the same way as in a controlled pressure rheometer. If the piston speed
is now reduced again, the shear stress will continue to decrease in a
regular way along the high-flow-rate branch of the flow curve until it
reaches the lower critical value, when oscillations will again be ob-
served. This behavior continues with further reductions in piston speed
until the apparent shear rate reaches the value corresponding to the
upper critical stress, u c2' (point A in Fig. 1) on the low-flow-rate
branch of the flow curve, when oscillations cease and the data follow the
low-flow-rate branch with further decreases in piston speed.
In the oscillatory flow regime of constant piston speed operation,
when the frequency is sufficiently low, the extrudate flow rate can be
determined as a function of time, and it is found that the system is
jumping between the two branches of the flow curve in a very regular
manner. The variation of extrudate flow rate, in spite of the constant
piston speed, arises from the compressibility of the melt in the reservoir,
as was first noted by Lupton and Regester (1965). Thus, melt com-
pressibility is a necessary condition for oscillatory flow. Utracki (1984),
however, has pointed out that it is not a sufficient condition, as LDPE
is as compressible as HDPE but does not exhibit oscillatory flow.
One more curious feature of the behavior of certain linear polymers
in a constant piston speed capillary rheometer is the appearance of a
second range of oscillatory flow at yet higher piston speeds (Li et al.,
1986). We have also observed this phenomenon for a LLDPE that we
studied.

Extrudate appearance
The appearance of the extrudate varies considerably as the flow rate
is increased. The extrudate is smooth at the lowest shear rates but begins
to exhibit small-scale roughness, often called sharkskin, as the flow rate
increases. Ramamurthy (1986) reported that the visual onset of this
SLIP, FRACTURE, AND OSCILLATORY FLOWS 849

roughness was associated with the change of the slope of the low-flow-
rate branch of the flow curve at aet. However, Beaufils et al. (1989)
used a profilometer to show that some roughness appears at lower flow
rates. This roughness increases in intensity as the shear rate increases,
until the upper critical stress is reached. Sometimes a "wavy" surface is
seen at the upper end of the lower branch of the flow curve (Becker et
al., 1991). Some polymers exhibit a very regular "screw thread" type of
surface defect at the upper end of the low-flow branch, and Bergem
(1976) has shown by a tracer technique that, like sharkskin, this arises
from a tearing of the melt at the exit of the capillary.
When the pressure is controlled so that the system jumps to the
high-flow-rate branch and continues along it, it is found that the extru-
date can exhibit a wide range of shapes. It can be smooth, it can exhibit
regular helical shape, or it can exhibit the severe, random roughness
often called gross melt fracture. If there is a region of smooth flow, it is
usually observed at the lower end of the upper branch of the flow curve
(Uhland, 1979).
In the case of oscillatory flow at constant piston speed, the extrudate
appearance varies in a cyclical manner in accordance with the period-
icity of the reservoir pressure. The extrudate obtained during one cycle
in the oscillatory extrusion of linear polyethylenes consists of two seg-
ments of distinctly different appearance. Lupton and Regester (1965)
and Ramamurthy (1986) report a relatively smooth segment exhibiting
sharkskin that is associated with the ascending portion of the pressure
waveform, and a rougher segment that is associated with the descending
portion of the pressure waveform. However, Kalika and Denn (1987)
reported that for a linear low density polyethylene the smooth portion
of the extrudate was obtained in the descending portion of the pressure
waveform.
We wish to point out that one must be very careful in describing the
appearance of this extrudate in order to avoid confusion. For example,
when the system is following the low-flow-rate branch of the flow curve,
i.e., when the pressure is rising, the extrudate generally exhibits shark-
skin. However, the extrudate that appears after the jump to the high-
flow-rate branch may be much rougher, leading to a description of the
extrudate in the rising-pressure portion of the cycle as (relatively)
"smooth." On the other hand, if in the high-flow-rate part of the cycle
the extrudate was smooth, the low-flow-rate material may be described
as (relatively) "rough," because of the presence of sharkskin.
Weill (1980) described his extrudates as exhibiting sharkskin as the
pressure was rising and as being "rough" during the stage of decreasing
850 HATZIKIRIAKOS AND DEALY

pressure. He also observed a third, very smooth segment of each cycle,


which he called a "plug," a phenomenon also observed by Bergem
(1976). This consists of material that was in the capillary at the instant
when the stress reached the upper critical value. It corresponds to ma-
terial that entered the capillary at a relatively low flow rate but was then
extruded very rapidly at the instant after the jump to the high-flow-rate
branch. Thus, this material has escaped both the extreme reservoir flow
irregularities leading to large-scale roughness and also the exit fracture
that leads to sharkskin. Of course the behavior varies with the molecular
structure of the polymer. Bergem (1976) found that in the case of 1,5
polypentenamer the material extruded during decreasing pressure was
grossly distorted, except for the "plug." However, for a HOPE he found
that the material extruded in the high-flow rate phase of a cycle was
locally very smooth with a larger scale, regular, helical shape.
Another complicating factor is that in oscillatory flow, the flow rate,
and shear stress, are changing continuously so that the extrudate ap-
pearance will also change during each stage of a cycle. When the stress
is increasing and reaches U c2' it will jump to the high-flow branch, while
the stress only decreases gradually. Thus, the system jumps initially
onto this branch at a point where the extrudate may be rough. But as
the stress decreases along that branch, it may move back into a region
of smooth flow (Lupton and Regester, 1965) so that by the time the
lower critical stress is reached, the jump to the low-flow branch may
cause an increase in roughness because of the reappearance of sharkskin.
Becker et al. (1991) reported the average swell ratio of solidified
extrudate over the entire flow curve for HOPE. The swell first increases
with flow rate, but at a stress somewhat below U c2 it decreases slightly.
At stresses above U c2, i.e., above the oscillatory flow range of flow rates,
the swell is much increased compared to that just below U c2 when the
system is in the low-flow branch of the flow curve. At the same time the
extrudate is smooth. However, when gross melt fracture begins at some-
what higher stresses, the swell falls sharply. This implies that for this
case there was no instability in the reservoir flow at stresses somewhat
above U c2' leading to a high degree of extensional-flow-induced orien-
tation, which did not have time to relax during its rapid transit through
the capillary.

Flow pattern in the reservoir


It is important to examine the role of flow irregularities in the region
just upstream of the entrance to the capillary. Utracki (1984) and El
SLIP, FRACTURE, AND OSCILLATORY FLOWS 851

Kissy and Piau (1990) have argued that jumps from one branch of the
flow curve to the other are triggered by the onset of such irregularities.
At low flow rates the flow here is steady and axisymmetric, and if the
entrance angle is 90, there is often a nearly stagnant zone of polymer in
the corner of the reservoir. As the flow rate is increased, several types of
complexity can arise. Tordella ( 1956) found that the flow could lose its
symmetry with the onset of a spiralling motion, which was accompanied
by a helical shape of the extrudate. Bergem (1976) used a tracer tech-
nique to show clearly that the spiralling entrance flow is the cause of the
helical shape of the extrudate. Furthermore, Piau et al: (1990) showed
that spiralling flow leads to helical extrudate even when there is no
capillary, i.e., for an orifice die.
Another phenomenon that can occur is a highly erratic, unsteady
pattern that clearly involves the generation of fracture surfaces within
the melt at irregular time intervals. This fracture phenomenon, which is
associated with gross distortion of the extrudate, has been studied ex-
tensively by Bagley and Schreiber (1961), who showed that moving to
a tapered die changes the flow pattern and the extrudate appearance.
As has been explained above the flow in the lower part of the high-
flow branch of the flow curve can be accompanied by all of these types
of reservoir flow patterns, viz., symmetric and steady, spiralling, and
fracture, depending on the circumstances. Therefore, it seems unlikely
that a change of flow type in the reservoir triggers the jump from one
branch to the other of the flow curve. Another observation that supports
this point ofview is that changing the entrance geometry has little effect
on U c2 (Uhland, 1979; Becker et al., 1991), even though it is known to
have a strong effect on the extrudate appearance. Finally, we note that
while LOPE exhibits pronounced die entry fracture and gross extrudate
distortion (Bagley and Schreiber, 1961) it does not exhibit a second
branch in its flow curve.

Role of Interfacial phenomena


Lupton and Regester (1965) used the Mooney technique that is
described later to infer slip velocities from their flow curve data. They
found clear evidence of slip in the low-flow-rate branch of the flow
curve, and this has been confirmed by the measurements of Ramamur-
thy (1986), Kalika and Denn (1987), and Hatzikiriakos and Dealy
(1992a). Ramamurthy noted that slip began at about the same shear
stress as that at which sharkskin first appeared and concluded that slip
was directly related to the appearance of this surface defect, although
852 HATZIKIRIAKOS AND DEALY

this point of view has been disputed (Beaufils et al., 1989). Lupton and
Regester (1965) also found evidence of slip on the high-flow-rate
branch of the flow curve, with slip velocities more than ten times those
on the low-flow branch. The slip velocities in the high-flow branch
calculated using the Mooney technique were found to be a substantial
fraction of the average velocity in the capillary.
Lupton and Regester found that viscous heating could result in a
significant increase in temperature in the high-flow branch because of
the high shear stress and a reduction in the heat flux due to the occur-
rence of slip. Indeed, Lim and Schowalter (1987) used changes in the
heat flux at the wall as evidence of slip. However, Lupton and Regester
concluded that viscous heating could not explain the discontinuity in
the flow curve.
All of the observations described above can be described in terms of
wall slip, i.e., adhesive failure, in the upper part of the low-flow branch,
and cohesive failure in the high-flow branch. Bergem (1976) machined
screw threads into the wall of a capillary and used a tracer technique to
show that during oscillatory flow material is trapped in the screw
threads with the mainstream flowing past it. This suggests slippage
along an interface formed by cohesive fracture. Furthermore, Uhland
(1979) showed that a processing additive increased slip in the low-flow
branch but had no effect on the upper branch.
It is difficult, however, to explain the presence of a second region of
oscillatory flow (Li et al., 1986). One hypothesis is that the first jump
involves adhesive failure and the second, cohesive failure. Thus, for
some reason there is a range of flow rates in which neither no-slip flow
nor wall slip provide a stable mechanism for transport of the melt, and
this gives rise to a new mechanism for the spurt phenomenon.
Leonov (1984) pointed out that slip does not necessarily lead to
instability, and this point is supported by observations that slip can
occur on the low-flow branch of the flow curve. He proposed an expla-
nation for oscillatory flow based entirely on the presence of a maximum
in the curve of shear stress versus slip velocity. In his model of the
process, he related the shape of the pressure versus time curve to fluid
viscoelasticity rather than to melt compressibility.
The major remaining mystery in the phenomena of hysteresis and
oscillatory flow is the molecular nature of the adhesive and cohesive
failures that cause them. A quantitative model for these processes is the
missing link necessary for a complete understanding of these phenom-
ena.
SLIP, FRACTURE, AND OSCILLATORY FLOWS 853

Rheological explanations of oscillatory flow


There have been many attempts to relate the flow discontinuities that
occur in capillary flow to an instability arising from the rheological
properties of the melt. Pearson and Petrie (1968), Schowalter (1988),
and Renardy (1990) examined the stability of simple shear of a vis-
coelastic liquid with wall slip, but they did not find an instability of a
type that could explain oscillatory shear. Vinogradov et al. (1984) pro-
posed that at the upper critical stress the melt is in a "forced high elastic
state" and cannot sustain further deformation. This rheological trans-
formation, he argued, explains the jump to a high-slip, high-flow regime.
Another approach has been to use purely rheological explanations
that do not involve slip at all. Huseby (1966) proposed that if the curve
of shear stress versus shear rate in steady simple shear has a maximum
followed by a minimum, this would imply that a given shear stress could
be supported by two different shear rates. This opens the door to the
possibility that in pressure flow there could be a discontinuity in the
shear rate profile y(r) in the capillary, with a layer of higher shear rate
nearer the wall. This would have an effect on the flow rate that would
be very similar in appearance to that of wall slip. The jump from a flow
with a monotonic variation of shear rate with r to one with a disconti-
nuity would produce a sudden increase in the flow rate that could
explain the spurt effect (Weill, 1980; Uhland, 1919; Hunter and Slern-
rod, 1983).
This concept has been more fully explored by a number of workers.
Lin (1985) noted that the Dol-Edwards theory predicts a shear stress
curve with a maximum in it and found that a( y) curves for narrow
MWD polystyrenes seemed to have such a maximum. McLeish and Ball
(1986) and McLeish (1981) used a modified form of the Dol-Edwards
theory to predict flow curve hysteresis. They concluded that the high
shear rate layer propagates as far as possible into the capillary once the
wall shear rate reaches the point where the shear stress is at its maxi-
mum. The question of how the high shear rate layer develops was
examined in detail by Kolkka et al. (1988) and by Malkus et al. (1990)
using the Johnson-Segalman constitutive equation modified by the ad-
dition of a small Newtonian viscosity. Marrucci and Grizzuty (1985)
and Morrison and Larson (1992) have proposed that anomalous stress
relaxation curves at large strains could be explained by strain inhomo-
geneities, and Morrison and Larson speculated that this phenomenon
could be related to flow curve hysteresis.
We note that the unusual rheological properties that give rise to
854 HATZIKIRIAKOS AND DEALY

TABLE I. Molecular parameters of the polymers used.

Resin %56B Mv Mw Mz MZ+1 I==MulMv


A 100 19000 177 800 1062700 1875500 9.4
B 80 18600 153400 996500 1 859700 8.2
C 60 18400 129100 906800 1829 100 7.0
D 40 18100 104 700 775300 1771 300 5.8
E 20 17800 80400 564 200 1622300 4.5
F 0 17500 56000 169800 351200 3.2

discontinuities in the shear rate or strain that are invoked in these


explanations of flow curve hysteresis have only been observed in poly-
mers having a very high molecular weight and a very narrow molecular
weight distribution. The commercial polymers used in nearly all of the
studies of oscillatory flow and flow curve hysteresis have been of mod-
erate molecular weight with a broad molecular weight distribution. No
evidence of a maximum in the curve of shear stress versus shear rate has
ever been seen for such materials. Therefore, we do not believe that
rheological considerations alone can explain the phenomena in question.
Objectives of this work
The mechanisms underlying the various phenomena described above
are still the subject of debate in spite of many studies over a period of
several decades. Conflicting points of view have been put forward re-
garding cause and effect, but the data available never seem to be suffi-
cient to prove one or the other conclusively. The problem is that several
phenomena occur simultaneously, and it is almost impossible to sort out
cause and effect. The way to break out of this pattern is to establish a
model of the process that predicts all the essential features of oscillatory
flow and flow curve hysteresis, in which each component of the model
can be verified independently. It was our objective to develop such a
model, as far as is possible at the present time, and to compare its
predictions with experimental data.

MATERIALS STUDIED
The two base polymers used were high-density polyethylenes (Sclair
56B and Sclair 2910) manufactured by Dupont Canada. These resins,
"A" (Sclair 56B) and "F" (Sclair 2910) have significantly different
values of M w (Table I) and contain only a stabilizer (no processing aids
or stearates). The melt indexes of these two base resins are 0.35 for "A"
SLIP, FRACTURE, AND OSCILLATORY FLOWS 855

TABLE II. Dimensions of dies used.

Diameter (em) LID


0.1320 5, 10, 20, 40, 60
0.0762 10, 20,40, 80, 100
0.0254 10,40,60, 80, 100

and 16.0 for "F." Four blends, containing 20, 40, 60, and 80% resin A,
were prepared using a twin-screw extruder. Some average molecular
weights for all the resins used are listed in Table I. The molecular
weight distributions of resins "A" and "F" were determined using size
exclusion chromatography while those for the blends were calculated
from their composition. These same resins were used by Hatzikiriakos
and Dealy (1992a) in their studies of wall slip below the flow rate at
which oscillations begin.
The density and compressibility of the resins were measured as func-
tions of temperature and pressure using a previously described tech-
nique (Hatzikiriakos and Dealy 1992a). Using a linear regression
method, the isothermal compressibility (f3) for all temperatures and
resins used was found to be 9.923X 1O-42% MPa- l . At constant
temperature, the density can be represented to a reasonable approxima-
tion as

p(P) =Po(l +f3P). (4)


While we have found it necessary to take into account the compress-
ibility of the resin in our interpretation of capillary flow data for these
high-density polyethylenes, we assumed that the viscosity is indepen-
dent of pressure. We have presented a detailed justification for this
assumption in a previous paper (Hatzikiriakos and Dealy, 1992a).

EXPERIMENTAL APPARATUS
The experiments were carried out in an Instron, piston-driven,
constant-speed, capillary rheometer. The piston speed was not contin-
uously variable, and only those speeds generated by the gear box could
be selected. A number of circular dies having the diameters and LID
ratios given in Table II were used. All were made from 420 stainless
steel and had an entrance angle of 45. For the small diameter dies the
cycle times in the oscillatory flow regime were sufficiently long that it
856 HATZIKIRIAKOS AND DEALY

RESIN A
D=0.0254 em, L/D=60 T=322 s
6 T=160 C 4000
3rd CYCLE
4 3500

2 3000
--.
III
<,
tlII
Z
--
S6
~4
4000

3500 ~
--
~2
ll::
3000 ~
~
0
...:I
~

6 4000

4 3500

2 3000

0 2500
0 0.2 0.4 0.6 0.8 1
NORMALIZED TIME

FIG. 2. Pressure (continuous curve) and mass flow rate (points) data for three cycles in
the oscillating flow region at 160'C for resin "A." YA=742 S-I.

was possible to cut samples at constant time intervals in order to esti-


mate the mass flow rate as a function of time.

Calculation of wall shear stress and apparent shear rate


The shear stress was corrected for end effects by use of the Bagley
method and calculated using Eq. (l). As will be explained later, once
slip occurs the shear stress varies along the capillary, and the value
given by Eq. (I) is thus valid only at some particular value of z (the
axial coordinate, measured from the entrance to the capillary), which
we call Z00 We will call this value of U w the nominal wall shear stress.
The apparent shear rate at the wall was calculated using Eq. (3). To
determine the true wall shear rate, it is necessary to take into account
both slip, when it occurs, and the variation of viscosity with shear rate.
This will be explained in detail in the section on "slip velocity." Because
SLIP, FRACTURE, AND OSCILLATORY FLOWS 857

8 4500
T=160oC
7 =0.0254 em and L/D=60
6 3.... CYCLE
(') 6 th CYCLE 4000
[T] 9'"' CYCLE ~

// 3500 ~
c:=
o
~

// 3000

o IS1 2500
o 0.2 0.4 0.6 0.8 1
NORMALIZED TIME tiT

FIG. 3. Pressures (continuous curve) and mass flow rates (points) of Fig. 3 replotted as
functions of normalized time t/T, where T is the period of oscillation for the particular
cycle.

the piston speeds are preset and cannot be varied continuously, only the
points indicated by square symbols in Fig. 1 could be obtained in steady
flow experiments. The triangles were determined from transient flow
data in the oscillatory flow regime.

EXPERIMENTAL OBSERVATIONS
Pressure and flow rate oscillations
Once the nominal wall shear stress surpasses the upper critical value,
and within a certain range of piston speeds, the driving pressure and the
flow rate oscillate between two limiting values. Figure 2 illustrates this
phenomenon for resin "A" at 160 DC and an apparent wall shear rate of
742 s- I. The force indicated by the load cell (continuous curve) and
the corresponding mass flow rate (points) are plotted as functions of the
normalized time t/T for the third, sixth, and ninth cycles from the start
of the run, where T is the period of the corresponding cycle.
It can be seen that as the force increases, the mass flow rate also
increases. However, when the force reaches its maximum value, the
858 HATZIKIRIAKOS AND DEALY

350
RESIN A
D=0.0254- em
300 T=160C
<t L/0=10

-
250 (!) L/0=4-0
['] L/0=60
III
'-' 200
10
o
......
~ 150
r.:l
ll<

100

50

o
0.00 0.05 0.10 0.15 0.20 0.25
BARREL LENGTH [rn)

FIG. 4. The period of oscillation as a function of the volume of the polymer in the barrel
for three LID ratios and constant die diameter. YA=742 s-I (Curves are model predic-
tions).

mass flow rate suddenly increases by a factor of about 8. Subsequently,


the force starts to decrease, and when it reaches a minimum value, the
mass flow rate suddenly decreases while the force starts to increase
again. The period of oscillation T decreases steadily from one cycle to
the next.
The average mass flow rate, calculated from the piston speed assum-
ing incompressibility, is shown by the horizontal lines in Fig. 2. It is
clear that material is compressed during the ascending portion of the
force curve and that there is a corresponding decompression and sudden
"spurt" at the beginning of the descending portion of the force curve.
This interpretation is verified by the observation that the net area be-
tween the mass flow rate curve and the horizontal line corresponding to
the time-averaged flow rate is nearly zero.
In Fig. 3 the data of Fig. 2 are plotted on a single graph to show that
the force and mass flow rate curves are invariant when represented as
functions of the normalized time tiT. The horizontal line again corre-
sponds to the flow rate calculated assuming incompressibility.
SLIP, FRACTURE, AND OSCILLATORY FLOWS 859

At a given piston speed, the period of the oscillations was found to


scale very well with the volume of polymer in the reservoir, which is in
accord with the observations of Lupton and Regester (1965) and Ka-
lika and Denn (1987). Figure 4 shows that a linear relationship exists
between the period of the oscillations and the length of the barrel oc-
cupied by the polymer. The lines in the figure are predictions of a model
that is presented in a later section of this paper. It can also be observed
that the period of the oscillations increases with the LID ratio. This is
not surprising since the use of longer dies causes an increase of the
pressure in the reservoir. Thus, more material is accumulated in the
reservoir, which in turn takes a longer time to flow out.

The flow curve


Figure I was prepared using the quantities calculated using Eqs. (1)
and (3) for resin A at 160C using a die with LID=60. Because the
piston speed could not be varied continuously, the values of the appar-
ent shear rate for the data points shown as squares were governed by the
preset piston speeds. The points shown as triangles were determined by
analyzing transient data in the oscillatory flow regime. One can calcu-
late the instantaneous nominal shear stress and apparent shear rate
corresponding to each point on the curves of force and mass flow rate of
Fig. 3. First, normalizing the force with the cross-sectional area of the
reservoir and using the Bagley correction (see Hatzikiriakos and Dealy,
1992a), the instantaneous nominal shear stress can be calculated using
Eq. (l). Then, making use of the density correlation given by Eq. (4)
the instantaneous apparent shear rate can be obtained from the follow-
ing modified form of Eq. (3):

YA(t) = 32M(t)/[ 1Tp(t)D3 ] , (5)


where M is the instantaneous mass flow rate.
The two distinct branches of the flow curve are apparent, and in the
range of apparent shear rates between points A and C, only oscillatory
flow is observed. In the range of oscillating flow, the shear stress and
apparent shear rate move in a clockwise direction around the closed
loop defined by points ABCD, the jumps from point A to point Band
from C to D occurring nearly instantaneously.

Extrudate distortion in the oscillating flow regime


Figure 5 is a photograph of an extrudate obtained in the oscillatory
flow regime by extruding resin "A" through a capillary die of diameter
860 HATZIKIRIAKOS AND DEALY

FIG. 5. Extrudate appearance of resin A in the oscillating flow regime. fA= 1482 sm .

0.0254 cm and L/D equal to 40 at 180 C. The total length of the


extrudate produced during one cycle is approximately 4 m. Figure 5
shows only the portions of the extrudate extruded just before and after
the jumps between the branches of the flow curve. The corresponding
pressure waveform is also shown in Fig. 5, and the times at which the
two segments were produced are indicated on this graph.
The portion shown at the top of the photo indicates the change that
occurs at the maximum pressure, and it can be observed that the tran-
SLIP, FRACTURE, AND OSCILLATORY FLOWS 861

sition from a relatively smooth extrudate (ascending pressure) to a


grossly distorted extrudate (descending pressure) occurs abruptly. The
lower portion shows the change that occurs at the minimum pressure.
Again the transition from a grossly distorted extrudate (descending
pressure) to a smooth extrudate (ascending pressure) occurs discontin-
uously.
However, note that as the pressure increases the extrudate becomes
gradually rougher. To see this compare the left end of the upper portion
with the right end of the lower portion; both these belong to the as-
cending part of the pressure waveform. On the other hand, as the pres-
sure decreases the extrudate gradually becomes smoother. To see this
compare the right end of the upper sample to the left end of the lower
sample; both of these belong to the descending part of the pressure
waveform. This is consistent with the previous observations summa-
rized above. The increase in roughness as the pressure increases reflects
the increased severity of the exit fracture on the low-flow branch of the
flow curve. The jump to the high-flow branch at the maximum pressure
takes the system into a region of the flow curve where there is entrance
fracture and gross extrudate distortion. As the pressure now falls, how-
ever, the entrance fracture becomes less severe and the extrudate be-
comes smoother.

THE SLIP VELOCITY


Modified Mooney technique for determining the slip velocity
The basis for the traditional Mooney technique is the following equa-
tion:

YA,s=YA-8u/D, (6)

where YA=32QI1rD 3, YA,s is the apparent wall shear rate corrected for
slip, which is assumed to be a function only of the wall shear stress, Us
is the slip velocity, and D is the capillary diameter. This equation is
based on the assumption that the wall shear stress, slip velocity, and
pressure gradient are constant along the length of the capillary. When
this is true, a plot of Y A vs liD, using data at a constant wall shear
stress, will give a straight line whose slope is equal to Sus and whose
intercept is YA,s' We have also made use here of the fact that 'VA,s is a
function only of (T w'
In a previous paper (Hatzikiriakos and Dealy 1992a) we studied the
sharkskin flow region for the same resins as were used in the present
work, and our results indicated that the slip velocity depends on the
862 HATZIKIRIAKOS AND DEALY

local pressure (actually the local wall normal stress). This means that in
capillary flow with slip neither the slip velocity nor the shear stress are
constant along the die, and the traditional Mooney technique is thus no
longer valid. We developed a modified Mooney technique to estimate
the slip velocity as a function of both wall shear stress and wall normal
stress. We noted that at some axial distance Zo the true local value of the
wall shear stress is expected to be equal to what we have called above
the "nominal" wall shear stress, i.e., (Pd- Pen)D/4L. Thus:
aw(zo) = (P d-Pen)DI4L. (7)
We then rewrote Eq. (6) in terms oflocal quantities as follows:
YAizo) =YA-8u s(zo)ID [aw(zo) =constantl. (8)

Now from a plot of YA vs liD holding LID and aw(zo) constant [which
assumes that the local value of the slip velocity us(zo) is a function of
aw(zo) and LIDl, the slip velocity can be calculated as a function of
wall shear stress and LID ratio. Then we derived an expression to relate
the LID ratio to the pressure P(zo).
lfthe melt viscosity follows a power law, the shear stress at the wall
is

aw=Ky~, (9)
and the wall shear rate Yw is related to the apparent shear rate corrected
for slip yA,s by Eq, (10):

Yw=[(3n+l)/4nlYA,s' (10)
Thus, the shear stress can be written as
aw=K[ (3n + 1)/4nl n( YA,s)n. (11)

For a power-law fluid, one can then eliminate YA,sCZO) , from Eq. (8) to
obtain an equation for the slip velocity at zo, which we call the nominal
slip velocity:
8u s(zo)ID=YA - [4nl(3n+ 1) 1[aw(zo)/K llln. (12)

This equation can be used to calculate the nominal slip velocity corre-
sponding to a single point on the flow curve, without using the more
laborious method described in the previous paragraph.
In order to verify the validity of the modified Mooney technique
described above, we developed a detailed capillary flow model in which
the slip velocity and wall shear stress were allowed to vary with z
SLIP, FRACTURE, AND OSCILLATORY FLOWS 863

TABLE III. Constants of Eq. (13).

Constant Value

m 3.23
So 0.1078 mls
Cl 2.552
Cd 83.61 K
C2 0.987
c) 15.5 cal/g mole
E 1238 cal/g mole

(Hatzikiriakos and Dealy 1992a). It was found that the above described
method gives a good estimate of the slip velocity.

Slip velocity on low-flow-rate branch of flow curve


In a previous paper (Hatzikiriakos and Dealy, 1992a) we determined
the slip velocity for the low-flow-rate branch of the flow curve (ael < a w
< a c2) for the resins used in the present work as a function of wall shear
stress, temperature, wall normal stress, and molecular parameters by
using the modified Mooney technique described above. Using an acti-
vation rate theory first proposed by Lau and Schowalter (1986), we
derived an expression for the slip velocity valid in the low-flow-rate
branch of the flow curve:

2!
us = 50 [ c't(T-To) ] ( ~
w a )m{ l - c2 tanh [- 1 ( E+c3- an)]}
2+(T-To) ucl] RT Uw

(ael < U w < u c2)' (13)

where Us is the slip velocity, 50 is an empirical constant, Un is the


compressive stress acting normal to the interface, uel is the critical shear
stress for the onset of slip, I is the polydispersity, a w is the wall shear
stress, m is a constant that depends only on the type of polymer, cr
and
c1 are empirical constants that are fitted to experimental data, T is the
absolute temperature, To is a reference temperature (we used To
= 140C, the melting point of resin "A"), c2 and c3 are empirical
constants, E is the activation energy, and R is the gas constant. Values
of the constants for the resins of interest are listed in Table III. The
values of 50, cr, c1,
C2' c3, and E were found to be the same for all the
high-density polyethylenes studied. For U w < ael we assume that the slip
864 HATZIKIRIAKOS AND DEALY

velocity is zero. However, we do not believe that the slip velocity is


discontinuous at ad but in fact increases continuously from zero to the
value given by Eq. (13) over a narrow range of shear stresses around
the critical value.

Slip velocity on the high-flow-rate branch of the flow curve


The modified Mooney technique was also used to determine the slip
velocity as a function of wall shear stress on the high-flow-rate branch
of the flow curve. However, it is difficult to infer directly from the data
the true wall shear rate, because most of the flow in this region occurs
by slip and relatively little by shear flow. In fact, in some cases the
Mooney method gives slip velocities that are greater than the plug flow
velocity, which is obviously unrealistic. An additional complicating fac-
tor is that there is significant viscous heating in the high-flow branch of
the flow curve (Lupton and Regester, 1965). Therefore, the best esti-
mate of the slip velocity on the high-flow branch is obtained by neglect-
ing the fluid deformation and assuming plug flow. The equation for the
approximate slip velocity is thus

2 .
us=QhrR =rAR/4. (14)

The same phenomenon, i.e., the fact that the portion of Q related to YA,s
is very small, means that viscosity values for wall shear stresses greater
than the second critical value a c2 could not be obtained.
We also calculated the slip velocity as a function of time in the
oscillating flow region. Using the oscillating flow data plotted in Fig. 1
the slip velocities were calculated for both branches of the flow curve,
using Eq. (12) for the data in branch DA and Eq. (14) for the data
in branch Be. A typical result is shown in Fig. 6. It can be observed
that the slip velocity waveform resembles that of the mass flow rate
of Fig. 3.
Figure 7 is a plot of the slip velocity versus the wall shear stress for
several temperatures for both branches of the flow curve, including the
maximum and minimum values calculated from the oscillating flow
data. The slip velocities for the low-flow branch are those determined by
Hatzikiriakos and Dealy (1992a) using long capillaries for which the
effect of varying the L/D ratio on the slip velocity is small. For the
high-flow branch, the slip velocities calculated from the Mooney tech-
nique (open symbols) have been plotted along with the slip velocities
calculated assuming plug flow (solid symbols). As can be seen the two
SLIP, FRACTURE, AND OSCILLATORY FLOWS 865

12
RESIN A
T=160C CJ
";i' 10 D=0.0254 em and L/D=60
S'
..::, 8
m
m
~
U 6
o
~ CJ[]
~ 4 m
P.. r::J[
~
UJ 2

o I ill mill [l] [T) fTlrl m m m


0.0 0.2 0.4 0.6 0.8 1.0
NORMALIZED TIME tiT

FIG. 6. The slip velocity of resin "A" during one cycle corresponding to conditions of
Fig. 1.

values are very close, implying that the portion of the total flow result-
ing from viscous flow is relatively small.
The jumps from the lower branch to the upper and from the upper to
the lower occur discontinuously in the oscillatory flow region, as shown
by the dashed lines in Fig. 7 for a temperature of 200 C.
We fitted the estimated slip velocities on the high-flow branch of the
flow curve to the following empirical power law model:
,m
us=a a w ' (15)

Table IV lists the values of the two parameters of this equation that
were determined using a linear regression method.

TRUE MELT VISCOSITY ON THE LOW-FLOW-RATE BRANCH


OF THE FLOW CURVE
To calculate the viscosity of the polyethylenes at high shear stresses
from experimental data, one has to take into account the slip velocity.
866 HATZIKIRIAKOS AND DEALY

1000.0
RESIN A Q)-'
A 160 e 0 ~
~
<, 100.0
C)
[']
IBOoC
200 0 e
Q)J;!
sc i-
>-
Eo-< 10.0
......
U
0
...:l
~
:> 1.0
c,
....
...:l
tr: Q)
0.1
Q) C)
A
(f

0.1 0.5 1.0


SHEAR STRESS (MPa)

FIG. 7. The slip velocity of resin "A" as a function of wall shear stress at three temper-
atures. Open symbols are slip velocities calculated from the modified Mooney analysis.
and closed symbols are calculated assuming plug flow. YA=742 s-l.

We explained above how to determine the slip velocity from capillary


rheometer data and how to determine the apparent shear rate corrected
r
for slip, A,s'

TABLE IV. Constants for the slip model in the fracture region.

Resin T rei a'(MPa-mm/s) m Correlation factor

A 160 7.60 2.63 0.991


180 10.0 2.86 0.987
200 13.2 2.71 0.992
B 180 15.7 3.85 0.995
C 180 17.6 3.80 0.988
D 180 21.8 4.28 0.990
E 180 19.5 4.32 0.982
F 180 14.6 4.08 0.985
SLIP, FRACTURE, AND OSCILLATORY FLOWS 867

T=1800C RESIN
)'\ A
)'\):!(
~ B
0.0100
<!> ~ '" DC
(')
..-
'" 6 ~
Lh
)l:
~ )!()!(
[:::J E
fIl ~ [I F
@ <!><!>

~
C\1 Lh
0... ~
~ m (':Jet )():!(

~~~
......... IS
0.0010 ~~

~""I'I\
,::-
[]I
ill []I [I
~ II I
0.0001
1 10 100 1000
-; (S-l)

FIG. 8. The viscosity of all resins used at 180 'C. Shear rates correspond to the low-flow-
rate branch of the flow curve. Data are corrected for slip where necessary.

Using these techniques, and using the Bagley method to determine


the end correction for the shear stress, the viscosities of all the high-
density polyethylenes were determined and are plotted in Fig. 8. All the
data for apparent shear rates greater than a certain value fall on a single
straight line for each resin. Thus, the wall shear stress-wall shear rate
relationship can be represented by a power-law expression [Eq. (9)].
This justifies the use of the power law to calculate the slip velocity in the
low-flow branch. It was not possible to determine the rheological be-
havior on the high-flow branch, because the very high contribution of
slip to the total flow reduced the relative viscous contribution to a level
within experimental error. However, there is no evidence that the
power-law is not valid in the high-flow region.
The power-law constants for all materials used in this study are listed
in Table V.

THE CRITICAL SHEAR STRESSES


Our previous work (Hatzikiriakos and Dealy, 1992a) indicated that
the first critical shear stress ad is independent of D, LID, and temper-
868 HATZIKIRIAKOS AND DEALY

TABLE V. Power-law constants and exponents.

Resin T(T) n K (MPas")

A 145 0.44 0.0222


160 0.44 0.0203
180 0.44 0.0178
200 0.44 0.0160
B 180 0.47 0.0119
C 180 0.52 0.0080
D 180 0.57 0.0049
E 180 0.65 0.0025
F 180 0.72 0.0013

ature. Ramamurthy (1986) studied several polyethylenes (high density,


linear low density, and branched) and determined the critical wall shear
stresses, ad and a c2' He found ad to be independent of temperature
and to be in the range of 0.1-0.147 MPa. These findings are consistent
with our observations that ad is in the range of 0.09-0.18 MPa. How-
ever, the value of 0.18 MPa was found for our resin "F," and based on
viscosity levels, we believe that no resin with such a low molecular
weight was included in Ramamurthy's study. A value of 0.26 MPa was
reported by Kalika and Denn (1987) for the LLDPE, which is in sharp
contrast with the values reported by Kurtz (1984) and by Ramamurthy
( 1986) and with those of the present work.
The values of the upper critical wall shear stress, a c2' determined in
the present work, are given in Table VI. Values were found to be in the
range of 0.22-0.50 MPa, which is in agreement with the value of 0.43
MPa reported by both Ramamurthy (1986) and Kalika and Denn
(1987) for LLDPEs. Values of ac2 in the range 0.1-1 MPa were re-
ported by Vinogradov and co-workers (1984) for polybutadienes of
various molecular weights and polydispersities.
A range of oscillating flow was observed for all polymers except
resins "E" and "F," which had the lowest molecular weights. The pe-
riod of oscillation increases with molecular weight in spite of the fact
that the compressibility is approximately the same for all the polymers.
This is mainly due to the fact that the stress difference (acrac3) in-
creases with molecular weight (see Table VII), and oscillating flow
occurs at a higher apparent shear rate. The expected value of this stress
difference for resin "E" should be close to zero, which implies that there
is no oscillating flow region.
SLIP, FRACTURE, AND OSCILLATORY FLOWS 869

TABLE VI. Values of I7c2' 17c3 and the difference I7c2-17c3 for resin A.

rrci D(em) LID rA(s-l) I7c 2(M Pa ) 17c3(MPa) I7c2 -17c3( MPa)

160 0.0762 20 550 0.243 0.18 0.063


160 0.0762 40 550 0.251 0.17 0.081
160 0.0254 10 1484 0,210 0.18 0.03
160 0,0254 40 742 0,215 0.175 0.04
160 0.0254 60 742 0.223 0.16 0.063
180 0.0762 80 550 0.270 0.19 0.08
180 0,0762 100 550 0.273 0.197 0.076
180 0.0254 40 1484 0.255 0.195 0.06
180 0.0254 60 1484 0.260 0.19 0.07
200 0.0762 20 1374 0.290 0.20 0.09
200 0.0762 40 1374 0.303 0.20 0.103
200 0.0762 80 1374 0.302 0.195 0.107
200 0.0254 40 1484 0.275 0.18 0.095
200 0.0254 40 3710 0.275 0.18 0.095
200 0.0254 60 3710 0.285 0.18 0.108

Our results indicate that (Te2 increases with both LID and tempera-
ture. The apparent dependence of (Te2 on D and LID may result from a
dependence on pressure. Thus, one finds higher values of upper critical
shear stress in a longer capillary, where the pressure is higher. In order
to test the hypothesis that the apparent dependence of the upper critical
stress on geometric factors actually reflects a dependence on pressure, it
was desirable to obtain slip data at atmospheric pressure. We had pre-
viously used a sliding plate rheometer for this purpose (Hatzikiriakos
and Dealy, 199Ib). However, because of the shear rate limitations of
our sliding plate rheometer, this effect could only be demonstrated at a
relatively low temperature of 145C, which is about 5C above the
melting point.

TABLE VII. Values of I7c2. 17c3 and the difference I7c2-ac3 for the high-density polyeth-
ylenes at 180C with a die of diameter, D=0.0254 em, and LID ratio, 40.

Resin %A a c2(MPa) ac3(MPa) a c2- 17cJ(MPa)

A 100 0.255 0.195 0.06


B 80 0.33 0.285 0.055
C 60 0.357 0.315 0.042
D 40 0.364 0.360 0.004
870 HATZIKIRIAKOS AND DEALY

TABLE VIII. Values of second critical stress for three values of L/D and as determined
in the sliding plate rheometer.

LID

60 0.214
40 0.200
20 0.182
Sliding plate 0.17

In constant speed experiments in the sliding plate rheometer at


145 C it was found that for resin A, the stress increased with the speed
of the moving plate up to a maximum value of 0.17 MPa. Above this
speed a steady shear stress was not generated, and the stress fluctuated
and then dropped well below this value. This decrease in the wall shear
stress is believed to reflect the same phenomenon as occurs in the cap-
illary rheometer at the upper critical stress, a c2' so that the value of 0.17
MPa is thought to be the value of the upper critical shear stress at
atmospheric pressure.
This value of U c2' as well those determined using capillaries having
three LID ratios, all at 145 C, are listed in Table VIII. We note a steady
progression in values starting with the sliding plate value and going
toward higher values of LID. This is consistent with the hypothesis that
the upper critical stress is dependent on pressure, although it does not
prove the point conclusively.

A MODEL FOR OSCILLATING FLOW


Development of the model
In this section we develop a capillary flow model to predict the
oscillating flow that is observed for capillary flow driven by a piston
moving at constant speed. The objective was to see if all the features of
the pressure and flow rate waveforms could be modeled using only
equilibrium slip laws, melt compressibility, and a power-law viscosity
model. We do not attempt to model the flow in the entrance region and
assume that this flow is not a causative factor in the phenomenon of
interest here. However, the entrance pressure loss is taken into account
by using experimentally determined Bagley end corrections.
Consider a capillary rheometer having a reservoir (or "barrel") of
radius R b , filled with polymer up to a length LbO at time (=0, at which
moment the piston starts to move at velocity Vp to force the melt
SLIP, FRACTURE, AND OSCILLATORY FLOWS 871

through a cylindrical die of radius R and length L. We let L bO=0.25 m


in our simulations, as this was typical of the initial depth of melt in the
reservoir in our experiments. A mass balance around the reservoir is as
follows:
M=dMldt= - [mass flow rate out] = -QrPb' (16)
where dM I dt is the rate of mass accumulation in the reservoir, Qo is the
volumetric flow rate at the inlet to the die, and Pb is the density of the
material in the reservoir. The mass of the polymer in the reservoir is
related to Lb(t) by
M(t) =rrif/-b(t)Pb(t). (17)
The instantaneous length of the barrel occupied by polymer Lb is a
function of time t and is given by Eq. (18).

Lb(t) =LbO - Vi (18)

Substituting Eqs. ~), (17), and (18) into Eq. (16), Eq. (19) is ob-
tained:
dPb Qo(l+I3Pb)
I3(LbO - Vpl)-d - (l +I3Pb ) Vp= - 2 (19)
t rrRb
Now turning to the capillary, we use the lubrication approximation
(see Pearson and Petrie, 1968 and Hatzikiriakos and Dealy, 1992a for a
justification) and allow for slip at the wall to obtain the velocity profile
at a given time and distance from the inlet as follows:
uz(r,z,t) =us(t,z) + [aw(t,z)/K] lin [Rn/( I +n)]

X [1- (r/R)(l+n)/n]. (20)


This can be integrated across the capillary to obtain the volume flux,
which, multiplied by the local density, must be equal to the mass flow
rate out of the reservoir, as shown by Eq. (21):
Mlp(t,z) = [rrR 3n/(3n+ 1)] [uw(t,z)IK] lin +rrR2uS<t,z). (21)
For creeping flow in a tube, the inertial terms in the momentum
equation can be neglected. Integrating across the capillary and neglect-
ing the gravitational force we have:

-d
dz
iR
0
azJ.rrrdr-2rrRaw=0, (22)
872 HATZIKIRIAKOS AND DEALY

where a w = -1" rz(R).


We now use the following definitions:

P= -1I3(azz+arr+aee),

NJ=azz-a,.,.,

N 2=arr-ae80
to write the axial normal stress as follows:

a zz= -P + t (2N J +N2 ) (23)

Next we make the approximation that N 2 = -N J/4 (Ramachandran


and Christiansen, 1983). Furthermore, it has been suggested that the
first normal stress difference Nl> for high-density polyethylenes at shear
stresses greater than 0.1 MPa, is proportional to the shear stress (Han,
1976), with a proportionality constant of about 10, i.e., N J = lOaw' Now
the axial normal stress becomes
azz(r,z) = -P(z) - (70/12)a(r,z). (24)
Consistent with the use of the lubrication approximation, we assume
that the shear stress is linear with r at constant z, i.e., a( r,z)
= -aw(z)rIR. Now we can evaluate the integral in Eq. (22). Solving
for a w we obtain:

a = - ( -R ) -dP + (35R)daw
-18 - . (25)
w 2 dz dz
We note that the equation for the slip velocity [Eq. (13)] involves the
compressive stress acting normal to the interface, which is an [defined
here as -arr(R)], which can be written in terms of the pressure and
first and second normal stress differences, N J and N 2, as follows:
an=~arr(R)=P+(NJ-N2)RI3. (26)
Making the same approximations as were used to obtain Eq. (24),
namely, that N 2= -N J/4 and N J = lOaw' an can be written as
a n=P+50aw/12. (27)
A Galerkin finite element program using linear basis functions was
written to solve Eqs. (13), (IS), (19), (21), (25), and (27) for the wall
shear stress a w, pressure P, and slip velocity as functions of both time
and axial distance along the capillary. The number of elements was 50,
SLIP, FRACTURE, AND OSCILLATORY FLOWS 873

120
RESIN A
... r =
1
- - "A 74.2 8-1
L/D=60. D=O.0254 em
A =1484 8-
- -r A =7420 8-
1 T=180 DC
ttl
~
::g 80

~
0::
c
00
00
~
40
0::
c,

0
0 1000 2000 3000 4000 5000
t (s)

FIG. 9. Three numerical simulations illustrating the three characteristic flow regions of
the flow curve.

and an adaptive time step algorithm was used in all the calculations. To
simulate the two flow rate jumps (A-B and C-D of Fig. I) we imposed
the following two conditions:

if Pmin<P<Pmax and dPldt> 0, then Us given by Eq. (13);


(28a)

if Pmin<P<Pmax and dP/dt<O, then Us given by Eq. (15),


(28b)
where Pmin and Pmax are the pressures at which the jumps from C-D
and A-B, respectively, occur (Fig. I). The values of these quantities
used in the simulations were those obtained experimentally. The corre-
sponding values of a c2 and ad are listed in Table V.
Figure 9 shows numerical simulations for three apparent wall shear
rates. The lowest apparent shear rate belongs to the subcritical region,
and the pressure increases gradually until it approaches a steady value.
The second case belongs to the oscillating flow region. Initially the
pressure increases with time until it approaches the critical value Pmax'
At this point the mass flow rate jumps because the slip law changes.
874 HATZIKIRIAKOS AND DEALY

Then, because there is a net flow out of the reservoir, the pressure
decreases until it reaches Pmin' At this point the system jumps back to
the low-flowbranch due to condition (28a). It can also be observed that
the period of the oscillation decreases with time due to the decreasing
volume of polymer in the reservoir.
The third apparent shear rate (top curve of Fig. 9) corresponds to a
shear stress greater than the upper critical value, and the pressure in-
creases sharply until it reaches a level corresponding to this stress.
Note that as the apparent shear rate decreases, the pressure rises
more slowly during the initial transient. This phenomenon can cause
problems in viscosity measurement when long dies of small die diameter
are used. We have analyzed this situation in detail and have proposed a
method to reduce the time taken to determine the viscosity using such
dies (Hatzikiriakos and Dealy, 1992b).

Comparison with experimental data


Figure 4 shows the period of oscillation as a function of the length of
the barrel occupied by polymer for three dies of the same diameter but
different lengths. The continuous curves show the model predictions. It
can be seen that the predictions are good in all cases. Note that the
period scales with the LID ratio for L/D=40 and 60 due to the fact that
the oscillations in these two cases were obtained at the same apparent
shear rate. However for L/D= 10 the oscillations were obtained at an
apparent shear rate double that for cases L/D=40 and 60.
An increase in the diameter of the die causes a dramatic decrease in
the period of the oscillations as is shown in Fig. 10. For example, the
case L/D=40 of Fig. 4 can be compared with the case L/D=4O of Fig.
10. The dramatic difference between the periods is due to the fact that
the same apparent flow rate corresponds to different piston speeds, since
YA=4QhrR 3 The piston speed in the case L/D=4O of Fig. 10 was 20
times greater than the speed in the case L/D=40 of Fig. 4. Thus, the
polymer in the reservoir is compressed in a shorter period of time in the
first case, and due to the larger diameter of the die it also flows out of
the reservoir faster.
An interesting phenomenon was observed for the case L/D = 40, and
this is also illustrated in Fig. 10. After the first cycle, the oscillations
stopped and then started again at the occupied length of the reservoir
indicated in Fig. 10. The experiment was repeated three times, and the
oscillations always started again at the same occupied length. This is
probably due to the fact that the apparent shear rate was close to one of
SLIP, FRACTURE, AND OSCILLATORY FLOWS 875

25
RESIN A
D=0.076~ em
T=160 C
20 NO OSCI nONS
[T] L!D=20
6 L/D=40
,..-.
~ 15
I::::l
o
....
~ 10
p.,

o
0.00 0.05 0.10 0.15 0.20 0.25
BARREL LENGTH (m)
FIG. 10. The period of oscillation as a function of the length of the barrel occupied by
polymer for two L/D ratios. Curves are model predictions, (No oscillations were observed
for occupied barrel lengths between the values indicated by the arrows on the upper
curve.)

the limits of the range of apparent rates where oscillating flow occurs.
Unfortunately, due to the fixed crosshead speeds, it was not possible to
increase or decrease slightly the apparent shear rate in order to see if the
same phenomenon could be observed.
The period of the oscillations decreases with an increase in the ap-
parent shear rate in the range where oscillating flow occurs. This fact is
illustrated in Fig. 11, where the period is plotted as a function of barrel
length occupied by polymer for two apparent shear rates. This was the
only case where oscillating flow was obtained at two different crosshead
speeds, due to the limitation of operating at fixed crosshead speeds. It
can be seen from Fig. 11 that the period decreases with an increase in
apparent shear rate. This is due to the fact that the polymer is com-
pressed and decompressed in shorter periods of time as the speed of the
plunger increases. We note the good agreement between the experimen-
tal data and the model predictions.
It is interesting to see how the form of the shear stress waveform
varies with apparent shear rate. Figures 12(b) and l3(b) show curves
876 HATZIKIRIAKOS AND DEALY

400
RESIN A
D=0.0254 em, L/D=40
T=200C
2J -y,=14B4 s-
300 6 1,=3710 Sl

---rn

200
iia
l"..:l
p.,

100

o
0.00 0.05 0.10 0.15 0.20 0.25
BARREL LENGTH (m)

FIG. 11. The effect of the apparent shear rate on the period of oscillation for one die and
temperature. Curves are model predictions, and points are experimental data.

of pressure versus normalized time t/T corresponding to the apparent


shear rates of Fig. II. In the first case [Fig. (12b)] the time for the
ascending portion of the pressure drop is longer than the time for the
descending portion. Therefore, as the apparent shear rate increases, the
normalized time corresponding to the ascending portion decreases while
the normalized time corresponding to the descending portion increases
[Fig. l3(b)]. The calculated pressure drop cycles are also plotted in
Figs. l2(b) and l3(b). It can be observed that the agreement is excel-
lent in Fig. 12(b) and fair in Fig. l3(b).
Kalika and Denn (1987) also studied the oscillating flow of a linear
low density polyethylene and reported the pressure drop waveforms for
several apparent shear rates, although the reported periods of oscillation
do not correspond to the same occupied length of the barrel, so that a
quantitative comparison of the period of oscillations is not possible.
However, they did observe that the duration of the pressure-rising seg-
ment of the cycle increases, relative to that of the pressure-falling seg-
ment, which is in agreement with our findings. They also found that the
SLIP, FRACTURE, AND OSCILLATORY FLOWS 877

_ 20 r-------------------.
G'l fa)
<,
SO 15
--
5

fbI
D=0.0254 em, L/D=40
T=200'C
... Calculated
- Experimental

25
\
0.0 0.2 0.4 0.6 0.8 1.0
NORMAUZED TIME

FIG. 12. (a) Calculated and experimental mass flow rates as functions of the normalized
time for one cycle. (b) Calculated and experimental pressures as functions of the nor-
malized time II T for one cycle.

pressure always oscillates between the same extreme values for a given
run, so that the pressure difference, Pmax - Pmin' is a constant for the
cycles of a given run.
Figures 12(a) and 13(a) are plots of the mass flow rate vs normal-
ized time t/T, The agreement between experiment and model prediction
is generally very good except in the descending portion of the pressure
drop waveform. This may be due to experimental error resulting from
the fact that the flow rate in this part of the cycle decreases very sharply
with time and it is not possible to cut samples for very short time
intervals.
In Fig. 14 we compare predicted cycle periods (curves) with data
(points) for the four resins that exhibited oscillating flow for one die
and temperature. The agreement is excellent for resins A and B and fair
for resins C and D. The period increases with molecular weight, as a
878 HATZIKIRIAKOS AND DEALY

_20
rn la)
<,
b/l 15
S
'-"

(bl RESIN A
"';;;'50 D=0.0254 em, L/D=40
~ T=200C
645

~ ~
//./ -.~:cula~d ~~
40
;::J
~ 35
~
g: 30 - Experimental....

25
0.0 0.2 0.4 0.6 0.8 1.0
NORMALIZED TIME

FIG. 13. Calculated (continuous curve) and experimental (points) mass flow rates as
functions of the normalized time for one cycle. (b) Calculated and experimental pressures
as functions of the normalized time t/T for one cycle.

result of the increase of the stress difference, ac2-ac3. with molecular


weight, which is shown in Table VII.
Pressure waveforms for several of the resins studied are plotted in
Fig. 15 as functions of the normalized time t/T along with the model
predictions. It can be seen that the forms of the shear stress waveforms
are accurately predicted.

Comments on the oscillating flow model


We summarize below the major assumptions made in our model: (I)
The viscosity is independent of pressure. (2) The flow is isothermal in
the low-flow branch of the flow curve. (3) Fluid elasticity is not taken
into account. (4) Creeping flow is assumed, and the lubrication approx-
imation is used. (5) The velocity distribution on the high-flow branch is
SLIP, FRACTURE, AND OSCILLATORY FLOWS 879

200
T=lBOOC D=0.0254 em L/D=40
RESIN
160 ~ A
A B
(') C
[!] D
~ 120
l:::
o
~
l::il 80
0..

40

o
0.00 0.05 0.10 0.15 0.20 0.25
BARREL LENGTH (m)

FIG. 14. The effect of molecular weight on the period of oscillation for one die and
temperature. Curves are model predictions, and points are experimental data.

flat (plug flow). (6) Empirical slip models are used, and the slip veloc-
ity is assumed to depend only on the current value of the stress and not
on the stress history. (7) The flow in the reservoir is not a causative
factor for the oscillatory flow modeled. (8) Empirical equations for
N 2(NI) and N 1(aw ) are assumed.
It is our opinion that the accuracy of the predictions might be im-
proved by including melt viscoelasticity and including a detailed mod-
eling of the flow in the high-flow branch, including viscous heating. We
believe that the other factors ignored play no significant role in the
phenomena of interest. The missing elements in the model, from the
point of view of a complete understanding of the phenomena, are mech-
anistic, predictive theories of the slip processes that occur on the two
branches of the flow curve.

CONCLUSIONS
The agreement between the model predictions and experimental ob-
servations implies that the essential factors that give rise to oscillatory
flow are a flow curve with two distinct branches and melt compressibil-
ity. Melt elasticity is not an essential factor, although it must play some
880 HATZIKIRIAKOS AND DEALY

75
RESIN
T=180C
-A 0=0.0254 em, L/0=40
---8
65 C
....... . D
S
p.,

~55
l":"l
P::
;:J
en
en 45
l":"l
P::
p.,
35

25
0.2 0.4 0.6 0.8 1
0
NORMAUZED TIME, tiT

FIG. 15. Model predictions of one cycle of the pressure waveform as functions of the
normalized time, t/T for four of the resins used. The bottom curve is for resin A.

role, and its inclusion may improve somewhat the accuracy of the model
predictions. Our macroscopic observations do not allow us to make a
conclusive statement about the ultimate mechanism that gives rise to the
second branch of the flow curve. However, our observations and those
of others provide some support for the hypothesis that the high flow
branch occurs as the result of a shear fracture in the die near the wall at
the upper critical shear stress. The jump back to the low-flow branch
then occurs when adhesion across the fracture surface occurs at the
lower critical shear stress. As the shear stress and flow rate now increase
again along the low-flow branch, new polymer flows into the capillary
and fractured polymer exits, so that the fracture process occurs in ex-
actly the same way as before when the upper critical stress is again
reached.
Because the commercial polymers used in our studies have moderate
molecular weights and broad molecular weight distributions, we do not
believe that they have the unusual rheological properties that have been
invoked by several theoreticians to explain flow curve hysteresis.
SLIP, FRACTURE, AND OSCILLATORY FLOWS 881

Finally, we believe that changes in the flow pattern in the reservoir


are results and not causes of the jump to the high-flow branch of the
flow curve. This point of view is supported by previous observations and
by the success of our model in predicting actual flow phenomena.

ACKNOWLEDGMENTS
This work was supported by the Natural Sciences and Engineering
Council of Canada. The authors also wish to thank Dupont Canada for
supplying the resins used. The SEC measurements were carried out by
P. C. Wong and R. W. Murray at the Industrial Materials Institute
(Nat. Res. Counc. Canada) in Boucherville, Quebec.

NOMENCLATURE

a' Slip coefficient for the slip model [Eq. (15)]


c1 Constant in WLF equation
~ Constant in WLF equation
D Diameter of the die
E Activation energy
I Polydispersity
K Proportionality constant in the power law model
L Length of the die
Lb Instantaneous length of the barrel occupied by polymer
Lw Initial length of the barrel occupied by polymer
m Power law exponent for the slip model
M Mass of polymer in the barrel
M Mass flow rate in the capillary rheometer
Mv Number-average molecular weight
Mw Weight-average molecular weight
Mz z-average molecular weight
M z+ \ z+ l-average molecular weight
n Power-law exponent
N\ First normal shear stress difference
N2 Second normal shear stress difference
P Pressure
Pd Driving pressure
882 HATZIKIRIAKOS AND DEALY

P(O) Pressure at the entry of the capillary calculated from the


driving pressure using the Bagley correction
P(ZO) Pressure at the axial position Zo
Pen Entrance pressure loss
Q Volumetric flow rate
R Radius of the capillary die
Rb Radius of the barrel
T Temperature
TO Reference temperature of 14OC
US Slip velocity
us(zo) Slip velocity at the axial length Zo
Uz Axial velocity in a cylindrical tube
Zo The axial position at which the wall shear stress assumes
the value aw(zo)

Greek letters
f3 Compressibility
y Shear rate
YA Apparent shear rate in the capillary rheometer
fA,s Apparent shear rate in the capillary rheometer corrected
for slip
Yw Wall shear rate
Po Density at pressure = 0
P Density at pressure P
Ph Density of melt in the reservoir (barrel)
a Shear stress (a12 or an)
aA Apparent wall shear stress-a positive number [Eq. (2)]
an Compressive stress acting normal to the wall-a positive
number defined as -arr(R)
aw Magnitude of the nominal wall shear stress-a positive
number [see Eq. (1)]
ac\ Critical shear stress for the onset of slip
a c2 Upper critical shear stress
a c3 Lower critical shear stress
azz Normal stress in z direction
arr Normal stress in radial direction
S Constant that depends on the molecular characteristics
of polymer
So Constant
SLIP, FRACTURE, AND OSCILLATORY FLOWS 883

References
Bagley E. B., I. M. Cabott, and D. C. West, "Discontinuity in the Flow Curve of Poly-
ethylene," 1. Appl. Phys. 29, 109 (1958).
Bagley, E. B. and H. P. Schreiber, "Effect of die entry geometry on polymer melt fracture
and extrudate distortion," Trans. Soc. Rheol. 5, 341 (1961).
Beaufils, P., B. Vergnes, and 1. F. Agassant, "Characterization of the sharkskin defect and
its development with flow conditions," Intern. Polym. Proc. IV, 78 (1989).
Becker, J., P. Bengtsson, C. Klason, J. Kubat, and P. Saha, "Pressure oscillations during
capillary extrusion of high density polyethylene," Intern. Polym, Proc, VI, 318
(l99\).
Bergem, N., "Visualization studies of polymer melt flow anomalies in extrusion," Proc.
7th intern. Congr. Rheol. (Swedish Society of Rheology, Gothenberg, 1976), p. 50.
EI Kissi, N. and J. M. Piau, "The different capillary flow regimes of entangled PDMS
polymers: Macroscopic slip at the wall, hysteresis and cork flow," J. Non-Newt. Fluid
Mech. 37, 55 (1990).
Han, C. D., Rheology in Polymer Processing (Academic, New York, 1976).
Hatzikiriakos, G. S. and J. M. Dealy, "Wall Slip of Molten High Density Polyethylenes
I. Sliding Plate Rheometer Studies," J. Rheol. 35, 497 (1991).
Hatzikiriakos, G. S. and J. M. Dealy, "Wall Slip of Molten High Density Polyethylenes
II. Capillary Rheometer Studies," J. Rheol. 36, 703 (1992a).
Hatzikiriakos, G. S. and J. M. Dealy, "Start-up Pressure Transients in a Capillary Rhe-
ometer," Soc. Plastics Eng. ANTEC, Tech. Papers 38, 1743 (l992b).
Huseby, T. W., "Hypothesis on a certain flow instability in polymer melts," Trans. Soc.
Rheol. 10, 181 (1966).
Hunter, J. K. and M. Slemrod, "Viscoelastic Fluid Flow Exhibiting Hysteretic Phase
Changes," Phys. Fluids 26,2345 (1983).
Kalika, D. S. and M. M. Denn, "Wall Slip and Extrudate Distortion in Linear Low-
Density Polyethylene," J. Rheol. 31, 815 (1987).
Kolkka, R. W., D. S, Malkus, M. G. Hensen, G. R. lerley, and R. A. Worthing, "Spurt
Phenomena of the Johnson-Segalman Fluid and Related Models," J. Non-Newt. Fluid
Mech. 29,303 (1988).
Kurtz, S. J., "Die Geometry Solutions to Sharkskin Melt Fracture," in B. Mena, A,
Garcia-Rejon and C. Rangel Nafaile, Advances in Rheology (Universidad Autonoma
de Mexico, Mexico City, 1984), Vol. 3, p. 399.
Lau, H. C. and W. R. Schowalter, "A model for Adhesive Failure of Viscoelastic Fluids
during Flow," J. Rheo!. 3D, 193 (1986),
Leonov, A. I., "A linear model of the stick-slip phenomena in polymer flow in rheome-
ters," Rhea!. Acta 23,591 (1984).
Li, H., H. P. Hiirlimann, and J. Meissner, "Two separate ranges for shear flow instabilities
with pressure oscillations in capillary extrusion of HDPE and LLDPE," Polymer Bull.
15, 83 (1986).
Lim, F. J. and W. R. Schowalter, J. Non-Newt. Fluid Mech, 26, 135 (1987).
Lin, Y. H., "Explanation for Slip-Stick Melt Fracture in Terms of Molecular Dynamics in
Polymer Melts," J. Rheal. 29, 605 (1985).
Lupton,1. M, and J. W. Regester, "Melt Flow of Polyethylene at High Rates," Polym.
Eng. Sci. 5, 235 (1965).
Malkus, D. S., J. D. Nohel, and B. J. Plohr, "Dynamics of shear flow of a non-Newtonian
fluid," J. Compo Phys. 87, 464 (1990).
McLeish, T. C. 8., "Stability of the interface between two dynamic phases in capillary
flow of linear polyethylene melts," J. Polyrn. Sci" B: Polym. Phys. 25, 2253 (1987).
884 HATZIKIRIAKOS AND DEALY

McLeish, T. C. B. and R. C. Ball, "A Molecular Approach to the Spurt Effect in Polymer
Melt Flow," J. Polym. Sci. B: Polym. Phys. 24, 1735 (1986).
Morrison, F. A. and R. G. Larson, "A study of shear stress relaxation anomalies in binary
mixtures of monodisperse polystyrenes," J. Polym. Sci., B: Polym. Phys. (in press,
1992).
Pearson, J. R. A. and C. J. S. Petrie, "On melt flow instability of extruded polymer," in
Polymer Systems, Deformation and Flow, edited by R. E. Wetton and R. W. Whorlow
(Macmillan, London, 1968),
Piau, J. M., N. El Kissi, and B. Trembley, "Influence of upstream stabilities and wall slip
on melt-fracture and sharkskin phenomena during silicones extrusion through orifice
dies," J. Non-Newt. Fluid Mech. 34, 145 (1990).
Ramachandran, S. and E. B. Christiansen, "The Dependence of Viscoelastic Flow Prop-
erties on the Structure for Styrene-Butadiene Colpolymers," J. Non-Newt. Fluid
Mech. 13, 21 (1983).
Ramamurthy, A. V., "Wall slip in Viscous Fluids and Influence of Materials of Construc-
tion," J. Rhea!. 30, 337 (1986).
Renardy, M., "Short wave instabilities resulting from memory slip," J. Non-Newt. Fluid
Mech. 35, 73 (1990).
Schowalter, W. R., "The behavior of complex fluids at solid boundaries," J. Non-Newt.
Fluid Mech. 29, 25 (1988).
Tordella, J. P., "Fracture in the extrusion of amorphous polymers through capillaries," J.
App!. Phys. 27, 454 (1956).
Uhland, E., "Das anomale Fleissverhalten von Polyaethylen hoher Dichte," Rheol. Acta
18, I (1979).
Utracki, L. A., "Pressure oscillations during extrusion of polyethylenes. II," J. Rheo!. 28,
601 (1984).
Vinogradov, G. V., A. Ya. Malkin, Yu, G. Yanovskii, E. K. Borisenkova, B. V. Yarlykov,
and G. V. Berezhnaya, J. Polym. Sci. A-2. 10, 1061 (1972).
Vinogradov, G. V., V. P. Protasov, and V. E. Dreval, "The Rheological Behavior of
Flexible-Chain Polymers in the Region of High Shear Rates and Stresses, the Critical
Process of Spurting, and Supercritical Conditions of their Movement at T> T!!,"
Rheo!. Acta 23, 46 (1984).
Weill, A., "Capillary flow of linear polyethylene melt: Sudden increase of flow rate," J.
Non-Newt. Fluid Mech. 7,303 (1980).

Вам также может понравиться