Está en la página 1de 14

Temperature Dependence of Polyolefin Melt


Quantum Chemical Corp., US1 Division

Allen Research Center
Cincinnati,Ohio 45249

The rheology of polymer melts depends strongly on temperature. Quantlfymg

this temperature dependence is very important for fundamental, as well as practi-
cal, reasons. The purpose of this paper is to present a unified framework for
handling the temperature dependence of rheological data. We considered the case
(by far the most common in polymer melts) where all relaxation times (in the
context of linear viscoelasticity) have the same temperature dependence (char-
acterized by a “horizontal shift activation energy”) and all relaxation moduli have
the same temperature dependence (characterized by a “vertical shift activation
energy”). The horizontal and vertical activation energies were extracted from loss
tangent vs. frequency and loss tangent vs. complex modulus data, respectively.
This is the recommended method of calculation, a s it allows independent estima-
tion of the two activation energies (statistically uncorrelated). It was shown theo-
retically, and demonstrated experimentally, that neglect of the vertical shift leads
to a stress (or modulus) dependent activation energy and necessitates different
activation energies for the superposition of loss and storage modulus data. The
long standing problem of a stress-dependent activation energy in long chain
branched LDPE was identified as originating from the neglect of the vertical shift.
The theory was applied successfully to many polyolefin melts, including HDPE,
LLDPE, PP, EVOH, LDPE, and EVA. Linear polymers (HDPE, LLDPE, PP) and
EVOH do not require a vertical shift, but long chain branched polymers do (LDPE,
EVA). Steady-shear viscosity data can be superimposed using activation energies
extracted from dynamic data.

INTRODUCTION Lem (1) and Harrel and Nakajima (2) discuss the

T emperature has a significant effect on the rheo-

logical properties of polymer melts. Quantification
of this dependence is important for several reasons.
connection between molecular structure and rheol-
ogy, and how the temperature effect is to be blocked
out in such a correlation.
First, it eliminates the need for rheological measure- The temperature dependence of rheology needs also
ments at multiple temperatures. Measuring rheologi- to be quantified and modeled appropriately for all
cal data at one temperature and knowing the temper- computer simulation work, including screw extrusion
ature dependence parameters (activation energies in simulations, coextrusion simulations, and general die
the nomenclature of this paper) is all we need to flow computations.
predict the rheological response at any other temper- Finally, the temperature dependence of rheology is
ature. An example drawn from industrial practice is of importance to fundamental studies on the mecha-
given in the Appendix. nism of flow of polymer melts. For example, it turns
A second major reason for establishing the temper- out that the flow activation energy of linear low den-
ature dependence of rheology relates to resin charac- sity polyethylene (LLDPE) with octene as a
terization. The rheology of polymer melts depends on comonomer is higher than that of a n LLDPE with
temperature and on the underlying molecular struc- butene as a comonomer (as will be shown and dis-
ture (molecular weight and molecular weight distri- cussed later in this paper). Long chain branched
bution, long chain branching). When characterizing LDPE shows a n even higher activation energy, plus
the rheology of a given resin we need to know how to some other anomalies (a stress-dependent activation
separate the various effects. For example, Han and energy when data are shifted at constant stress).

Temperature Dependence of PolyoleBnMelt Rheology

The purpose of this paper is to develop a unified ular structure on the flow activation energy of poly-
framework to treat both dynamic and steady rheologi- mer melts and concluded that long chain branching
cal data and apply it to polyolefin melts. in LDPE is responsible for the anomalous behavior of
activation energy of LDPE. Similar observations and
TEMPERATURE DEPENDENCE OF experimental confirmation of “thermorheological
RHEOLOGY complexity” in LDPE have been reported repeatedly
Background (9, 16-19).
Graessley and co-workers (10,20-22) found indica-
The temperature dependence of rheology has tions of thermorheological complexity” in model
mainly been discussed in the literature either in the branched polymer melts. Graessley (11, 12) exam-
context of linear viscoelasticity or in terms of steady- ined theroetically the effect of long branches on the
shear viscosity. In the latter case the shear stress (T, temperature dependence of rheology and suggested
the shear rate y, and the viscosity 77. are related two reasons for the observed.behavior. One has to do
through: with the temperature dependence of the entangle-
u = 77.j (1) ment density; the other is related to the differences in
the paths of conformational relaxation of linear and
Mendelson (3) showed how to extract the tempera- branched chains. Linear chains rearrange by repta-
ture dependence from steady shear viscosity data by tion. The presence of long branches blocks reptation;
plotting the shear stress (vertical) vs. the shear rate the long chain branches themselves rearrange by
(horizontal) data at different temperatures. passing through relatively compact conformational
A review of the literature shows that it was estab- states that may have different energies, depending on
lished early on that polymer melts are thermorheolog- the temperature coefficient of chain dimension in the
ically simple. According to Markovitz (41, the term species ( 12).
“thermorheological simplicity” was originated by Verser and Maxwell ( 19) examined the temperature
Schwarzl and Staverman (5) as the “assumption of dependence of linear viscoelastic data (frequency re-
uniform temperature dependence for all relaxation sponse). They found that when the classical treat-
elements.” The modulus is usually only weakly de- ment of time-temperature superposition (6) was a p
pendent on temperature (6). and thus thermorheo- plied to the data, then the resulting activation energy
logical simplicity has come to signify the ability to was modulus dependent (in fact their data show the
superpose data from different temperatures by a hori- activation energy to increase with complex modulus,
zontal shift on a shear stress vs. shear rate graph (for Fig. 19 in Ref. 19, an unusual result for LDPE). The
steady shear data) or on a modulus vs. frequency modulus dependence was removed when a modulus
graph (for linear viscoelastic data). The temperature shift was also applied to the data (“vertical” as well as
shift is expressed in terms of a flow activation energy “horizontal” shift). Verser and Maxwell (19) obtained
for polyolefin melts. When the flow activation energy the vertical shift from the temperature dependence of
shows a stress dependence (for stress vs. shear rate the steady-state compliance. However, most commer-
data) or a modulus dependence (for modulus vs. fre- cial polyethylenes are so broad that the terminal
quency data), then this is characterized as “thermo- region (i.e., the low frequency region where G‘ a w 2 ,
rheological complexity.” Long chain branched G” a w , and J,“ = G’/G” 2, is not accessible and there-
polyethylene (LDPE) exhibits thermorheological com- fore the steady-state compliance J,“ would have to be
plexity (in contrast to the behavior of linear polyeth- calculated through data extrapolation, which may
ylenes) as was pointed out repeatedly in the literature introduce large errors. I t is an objective of the present
(7- 13). paper to develop procedures for obtaining the vertical
I t will be shown in this paper that LDPE may still shift from data available over the experimentally ac-
be considered “thermorheologically simple” in the cessible range (thus not involving any extrapolation).
sense that all relaxation times have the same temper- In what follows we will deal mainly with the temper-
ature dependence and all relaxation moduli have the ature dependence of linear viscoelastic data for two
same temperature dependence (characterized by a reasons: (i) the steady shear data should have the
“vertical” or modulus temperature shift). However, same temperature dependence as the linear vis-
neglect of the “vertical” or modulus shift leads to coelastic data (41, and (ii) linear viscoelastic data
a modulus (or stress) dependent flow activation ener- include both the viscous and the elastic response of
gy, and thus to an apparent thermorheological the material and are therefore more complete. (Steady
complexity. shear data usually involve the viscous response only.
Early experimental observations that: (a)long chain The elastic response, i.e., normal stresses, is very
branched polyethylene (LDPE) has a higher flow acti- difficult to measure.)
vation energy than linear polyethylene (HDPE), and The temperature dependence of linear viscoelastic
(b) the flow activation energy of LDPE depends on data is usually discussed in terms of the relaxation
shear stress (for shear stress vs. shear rate data, spectrum H ( T ) ,where H ( T )is the relaxation strength
shifted horizontally) were reported by Sabia (7) and of the material at relaxation time T . All other linear
Mendelson (8). Porter and co-workers (9, 14, 15) ex- viscoelastic properties can be derived from the relax-
amined the effects of polymer composition and molec- ation spectrum. For completeness and ease of refer-

H. Mauridis and R. N . Shroff

ence we summarize the most common of them below: reference temperature. The horizontal shift factor a ,
reflects essentially the temperature dependence of
Storage Modulus: relaxation time :

while the vertical shift factor b, reflects the tempera-

ture dependence of modulus
Loss Modulus:

where G$ is the plateau modulus as defined in Eq

Complex Modulus: An early prediction for the temperature dependence
of relaxation time was provided by the WLF equation

Complex Viscosity: (7)

q*( 0 ) = - The WLF equation holds in the range of temperatures
from Tg to Tg + 100°C,where Tg is the glass transition
Dynamic Viscosity: temperature. For temperatures greater than Tg+
lOO"C, which is the processing range for polyolefin
G"(0) melts, the WLF equation is closely approximated by
v'( w ) = -
w an Arrhenius-type equation:

Loss Tangent:
G ( w )
tan 6 ( w ) = -
- ___-
(T+;73 To+ 273

G'( 0) where EH is the horizontal shift activation energy

Certain other rheological properties are also derivable and
from the above functions: R is the gas constant (1.987 cal/mole/"C). Equa-
tion 8 is the form employed in the present work for
Plateau Modulus: the horizontal shift factor, as it is the appropriate
form for polyolefin melts, and also because it only
involves one parameter-the activation energy EH.
The vertical shift factor will now be discussed in
some detail. Dilute solution theories (6) predict that
Zero-Shear Viscosity: the modulus is proportional to the product of density
+X and absolute temperature, which in turn suggests a
qo=/ H(T)TdhT= lim q * ( w ) (3.2) shift factor:
--ic 0'0

Po' (To + 273)

Compliance: bT= (9)
p * ( T +273)

where p is the density and T is the temperature in

degrees Celsius. The subscript "0" indicates refer-
ence conditions. The same result is predicted from
the theory of rubber elasticity, which gives the follow-
ing equation:
In the above w is the frequency in units of rad/s.
Let us now discuss the temperature dependence of
the linear viscoelastic spectrum. The fundamental
assumption is that relaxation spectra derived from where Me is the molecular weight between entangle-
data at different temperatures can be made to super- ments. On the assumption that the molecular weight
impose by vertical (along the relaxation strength axis) between entanglements is independent of tempera-
and horizontal (along the time axis) shifts, (4): ture, Eq 10 directly yields Eq 9.
However, Me may depend on temperature, and in-
deed this will be shown to be the case for long chain
where a , and b, are the horizontal and vertical shift branched polyethylene. In that case the result of Eq 9
factors respectively, T is the temperature, and To is a does not hold, and the molecular weight between

Temperature Dependence of PolyoleBn Melt Rheology

entanglements varies as (from Eqs 6 and 10): level of loss tangent provides equally acceptable scal-
ing parameters), i.e. plotting G'/G, or G / G , vs.
w / w , results in a temperature independent graph.
The temperature dependence of some commonly
referred to viscoelastic properties is given by:

As will be shown below, there is no reason to fuc Plateau Modulus:

a priori the vertical shift factor (as dictated above by
the theory of rubber elasticity). The appropriate verti-
cal shift factor can be extracted from the data along
with the horizontal shift factor. For that purpose we
will define a general form for the vertical shift factor,
analogous to that for the horizontal shift factor in- Zero-Shear Viscosity:
volving a "vertical activation energy":
-- _ -aT (16.2)
To( To) bT

where E, is the vertical activation energy. Compliance:

Temperature Shifts of Linear Viscoelastic Data
To summarize the results of the previous section, (16.3)
we assume that the relaxation spectra from different
temperatures can be made to superimpose according
to Eq 4, i.e. using a horizontal, a T , and a vertical, b,,
Calculation of Activation Energies
shift factor. The functional forms of the shift factors
are given in Eqs 8 and 12, and they involve two The calculation of the activation energies will be
parameters: the horizontal, E H , and the vertical, E,, based on the equations derived in the previous sec-
activation energies. tion. The important point to note is, from Eq 15.3 we
The storage and loss moduli shift with temperature observe that plotting data from different tempera-
as: tures as frequency vs. loss tangent (or vice versa)
should give parallel curves separated by a certain
b T G ' ( a T w , T )= G ' ( w , T o ) (13.1)
distance determined by the horizontal shift factor aT
) G(w,To)
b T G ( a T w , T= (13.2) only (the vertical shift factor does not appear in this
graph). Therefore. a frequency vs. loss tangent graph
From the above we see that the loss tangent shifts as:
is ideally suited for the estimation of the horizontal
tan G ( a , w , T ) = t a n 6 ( w , To) (14) activation energy E H , since the estimation can be
performed independently from that of the vertical
i.e., the loss tangent remains invariant under a tem- activation energy E,.
perature shift. Therefore, it is convenient to redefine Similarly, from Eq 15.4 we observe that plotting
the various viscoelastic properties in terms of the loss data from different temperatures as complex modu-
tangent and temperature: lus, G*, vs. loss tangent (or vice versa) should give
1 parallel curves separated by a certain distance deter-
G'(T, tan 6 ) = --'(To, tan 6 ) (15.1) mined by the vertical shift factor b, only (the hori-
bT zontal shift factor does not appear in this graph).
1 The above observations justify the selection of dy-
G " ( T ,tan 6 ) = - G ( T o , tan 6 ) (15.2) namic data for the determination of the temperature
bT dependence of rheology. Dynamic data allow indepen-
1 dent estimation of the vertical and horizontal activa-
w ( T , tan 6 ) = -@(To, tan 6 ) (15.3)
tion energies. Note that for other types of linear
viscoelastic data commonly employed (shear stress
G*(T,tan6)=-G*(To,tan6) (15.4) relaxation modulus data, at)vs. time t, or creep
b, compliance data, J ( t ) vs. time t ) , as well as for
steady-shear viscosity data, the two activation ener-
T*(T,tan6)=-q*(To,tan6) (15.5) gies cannot be computed independently.
bT The estimation is performed through nonlinear
Note that Eqs 15.1-3 indicate that the crossover least-squares fitting of the data, as explained below.
modulus G, (G,(T) = G'(T, tan 6 = 1)= G"(T,tan 6 =
Estimation of Horizontal Activation EH
1)) and crossover frequency o, ( o,(T) = w ( T , tan 6 =
1)) are convenient scaling parameters (selecting a Frequency vs. loss tangent data from different tem-
modulus and frequency values at another arbitrary peratures are least-squares fitted to the following

H. Mavridis and R. N . Shrofl

form: from:

where a, is given by Eq 8. rn is the order of the fitting

polynomial (usually 3) and ( ck, k = 0, rn) are the poly-
nomial coefficients. The nonlinear least-squares fit The vertical activation energy was computed from
involves ( rn + 2) parameters: ( rn + 1) polynomial coef- Eq 20 for two cases: first, with the density given by
ficients and the activation energy EH.The fit provides Eq 19 and, second, for a temperature independent
an estimate of the activation energy EH and its 95% density. The resulting vertical activation energy, when
confidence interval, EH A E H , where (EH- A EH. EH the density variation with temperature is taken into
+ A E H )is the 95% confidence interval (in a statistical account, is about 0.6 kcal/mole in the range 150 to
sense, (23)). 230°C. When the density variation is neglected, the
effective activation energy is about 0.92 kcal/mole,
Estimation of Vertical Activation Energy E, which is an order of magnitude smaller than that for
Complex modulus vs. loss tangent data from differ- the horizontal shift activation energy. The shift factor
ent temperatures are least-squares fitted to the fol- varies between 0.9 and 1.1 in the range 150 to 230°C,
lowing form: i.e. it varies by about rt 10%. a level of variation that
may fall within the limits of experimental error.
k = rn The vertical shift factor predicted by theory is very
log,,(tan6) = dk.[log,,(b,.G*)lk (18) close to unity and, therefore, it can be neglected for
k= 0 the example discussed above. When the vertical shift
factor is neglected for long chain branched LDPE,
where b, is given by Eq 12. rn is the order of the then anomalous results are observed: the computed
fitting polynomial (usually 3) and ( d k , k = 0,rn) are (horizontal) activation energy depends on stress (or
the polynomial coefficients. The nonlinear least- modulus) a s explained below.
squares fit involves ( rn + 2) parameters; ( rn + 1) poly-
nomial coefficients and the activation energy Ev. The Stress Dependent Activation Energy of LDPE:
fit provides a n estimate of the activation energy E , An Explanation
and its 95% confidence interval, E,+ A E,. We will show now that: shifting data horizontally
Once the horizontal and vertical activation energies only will generally introduce a stress (or modulus)
have been computed, the data from different temper- dependence of the (horizontal) activation energy,
atures can be shifted to the reference temperature To which only disappears in the limit of zero vertical
(using Eqs 13-15) for visual inspection of the good- activation energy. For this purpose we will use the
ness of superposition. schematic shown in Fig. 1 : frequency (or shear rate)
is plotted on the horizontal axis, while modulus (or
Comments on the Vertical Shift Factor stress) is plotted on the vertical axis. The symbol
As was mentioned in the Background discussion, "GSUP"may represent G', G or G*. We assume that
the vertical shift factor predicted by the theory of for the data to shift from temperature T to tempera-
rubber elasticity, as well a s by dilute solution theo- ture To a vertical, b,, and horizontal, a,. shifts are
ries, is given by Eq 9. This shift factor is usually very necessary, as shown in Fig. 1 . This shift is indicated
nearly unity and can often be neglected, i.e., the data by the path 1 -+ 2 -+ 3 in Fg. 1. Now let us consider a
can be shifted at constant stress (for shear stress vs. shift at constant modulus, as indicated by the path,
shear rate data) or constant modulus (for modulus 1 + 4, i.e., a horizontal shift only at constant level of
vs. time or frequency data). A shift factor of b,= 1 modulus. Let aT. be the shift factor associated with
corresponds to a vertical activation energy of E, = 0. this shift, and let us define a corresponding activa-

Such a shift factor is indeed appropriate for most tion energy Esup as:
linear polymers (as will be shown later).
To examine the magnitude of vertical shift factor
predicted by Eq 9, we select a HDPE, for which the
melt density was determined a s a function of temper-
[ (
~ 1 , , ~ = e x-
p ~-
T+;73 To+ 273

ature (24): From Fig. 1 we observe that the following relation-

ship holds:
p = 0.8696 - 5.62.lO-*.T (19)

where p is the density in g/cm3, and T is the temper-

ature in "C.
If the vertical shift factor is given by Eq 9, then the where a, and b, are given by Eqs 8 and 12 respec-
equivalent vertical activation energy can be computed tively. Taking the logarithm of both sides of Eq 22


Temperature Dependence of PolyoleJn Melt Rheology

25 shows that in that case it is impossible to

superimpose both the storage and loss modulus
data with the same activation energy. To prove
this, consider the variation with frequency in the
terminal region:
log GWP
d In[ G'( w ) ]
= 2 , for w - 0 (26.1)
d In w
d In[ G" ( w ) ]
= 1, for w + 0 (26.2)
d In w

log 0 and substituting in Eq 23:

Fg. 1 . Schematic of modulus us. frequency at two tem- EV
peratures, illustrating the two types of temperature shift; E'=EH--, forw-tO (27.1)
horizontal-only (1 + 4), and both vertical and horizontal shift 2
( 1 + 2-, 3). E =EH-Ev, forw-tO (27.2)
from Eq 2 7 we observe that in the terminal region
and differentiating with respect to the inverse of abse
the superposition of the storage modulus data gives
lute temperature we arrive at the following:
an activation energy E' which is different from the
activation energy E required for superposition of the
loss modulus data.
The conclusion, therefore, is that neglecting the
vertical shift not only makes the activation energy a
d In[ Gsup(a T w , T o ) ] function of modulus but also necessitates different
= -E,+ .EH* Esup
d In( a T w ) activation energies for the storage and loss modulus
It is now easily understandable why steady shear
viscosity data for LDPE yield a shear stress-depen-
dent activation energy: it is like the case of shifting
data at constant loss modulus examined above. In
fact if normal stress data were to also be shifted, then
a different activation energy would have been found
(much the same as shifting at constant storage mod-
ulus yields a n E' different from E" ). The above theo-
retical results will be further corroborated by the
experimental data to be presented in later sections.
To simplify the above equation for the purposes of
illustration, assume that over the range of frequen- Temperature Shifts of Steady-Shear Data
cies ( a ,G w , a T w ) the Gsup vs. w curve can be approx- The preceding discussion dealt with the tempera-
imated by a power-law form: ture shifts of linear viscoelastic data. It can be shown,
from fundamental considerations (41, that steady-
shear data must be superposable with the same shift
factors. If g ( j , T ) is the shear stress at shear rate y
where n is the power-law exponent. Substituting in and temperature T , then the superposition of the
Eq 23: steady-shear data can be obtained with the following
EsuP= EH - E,
From Eqs 22-25 we observe that: which is the equivalent of Eq 13 for steady-shear
(i) when E,= 0, then b,= 1, and = aT,which
data, where aT and b, are the horizontal and vertical
implies that P u p = EH, i.e., when a vertical shift shift factors respectively.
If the horizontal and vertical activation energies are
is not necessary then shifting the data at con-
stant modulus will yield a constant activation not available, they can still be extracted from the
energy, which is the horizontal activation energy. steady-shear data. The estimation can proceed simi-
(ii) when Ev* 0 then shifting the data at constant
larly to that outlined for dynamic data, i.e. by least-
modulus will yield an activation energy Esup that squares fitting the data to the form:
decreases with increasing modulus according to
Eq 25 (because the power-law exponent n de-
creases with increasing modulus). Moreover, Eq

H. Mavridis and R. N . Shroff

where ciT is given by Eq 8, b, is given by Eq 12, rn is Step 3: Using the computed activation energies, shift
the order of the fitting polynomial, and ( c k , k = 0, rn) the raw data to a reference temperature for
are the polynomial coefficients. The nonlinear least- visual inspection of the goodness of superpc-
squares fit would now involve (rn+ 3) parameters: sition.
( rn 1) polynomial coefficients, the horizontal activa-
tion energy E H ,and the vertical activation energy E,.
Note that in this case the estimates of two activation Dynamic data (frequency response in the linear
energies will be statistically correlated (23). This is to viscoelastic region) were obtained with the Rheomet-
be contrasted with the estimation obtained from rics Mechanical Spectrometer (RMS) and the Rhec-
dynamic data from which the two activation ener- metrics Dynamic Analyzer I1 (RDA2) operating in the
gies are obtained independently and therefore parallel plate mode, while steady-shear viscosity data
uncorrelated. were obtained with a gas extrusion rheometer (25).
In the present paper we extracted activation ener- All sample plaques for use in RMS and RDA2 were
gies from the dynamic data and then used these stabilized with additional antioxidants to avoid chain
activation energies to shift the steady-,shear data and extension and scission during measurement.
check whether steady-shear data superimpose. We Some of the data (HDPE-1 in Table I , and LDPE-1
could have done it another way; we could have ex- and LDPE-3 in Table 5) were obtained on solution
tracted activation energies from steady shear data (as dissolved samples. The polymer was dissolved in xyl-
discussed above) and compared them with those ex- ene, the xylene was allowed to evaporate, and the
tracted from dynamic data. We pursued the first a p final sample was dried in vacuum for 72 h. Polymer
proach because dynamic data are complete (include samples were solution dissolved in order to eliminate
both the viscous and elastic response) and are more effects of any shear prehistory (e.g., during pelletiz-
accurate than steady-shear data. ing). Shear prehistory is known to affect the rheologi-
cal behavior and its temperature dependence, espe
METHODOLOGY cially for LDPE ( 17).
The theory outlined in the previous section is a p
plied practically a s follows. HDPE
Tabulated results for HDPE are given in 'fable 1 for
Step 0: Collect dynamic data measured at different five different resins. In all cases the horizontal activa-
temperatures. The larger the number of tem- tion energy turns out to be about 6 kcal/mole, and
peratures at which data is available and the the vertical activation energy is zero. The value of 0.8
wider the temperature range, the better the kcal/mole for HDPE-1 must be treated with caution,
estimation of the activation energies will be. due to the fact that the material was xylene dissolved.
Step 1: Perform a preliminary screening of the data. Also the value of 0.14 kcal/mole for HDPE-5 is
The plot of log,,(tan 6 ) vs. log,,(G*) is partic- marginally significant. These results are in accord
ularly helpful. with those of the literature (3).
1.1 If the data from different temperatures in
the above plot of log,,(tan 8 ) vs. log,,(G*) LLDPE
superimpose (within experimental error),
Tabulated results for LLDPE are given in Table 2,
then a vertical shift is not required and
for three different resins. The horizontal activation
a n E,= 0 should be expected.
energy varies in the range 7 to 8 kcal/mole, which is
1.2 If the data from different temperatures in
the plot of log,,(tan 6 ) vs. log,,(G*) do
not superimpose but fall onto parallel Table 1. Results for HDPE.
curves, then a vertical shift is required
and a non-zero E , should be expected. Resin g/cm3 g/10 min kcal/mole kcal/mole
1.3 If the data from different temperatures in
HDPE-1* 0.96 1.o 6.60 & 0.44 0.80 & 0.33
the plot of log,,(tan 6 ) vs. loglo(G*) do HDPE-2 0.96 0.8 5.80 k 0.52 0.0
not superimpose and do not fall onto par- HDPE-3 0.96 0.8 5.94 f0.58 0.0
allel curves, then we cannot handle this HDPE-4 0.96 0.9 6.28 f 0.49 0.0
case within the present theoretical HDPE-5 0.96 0.2 6.10 f0.22 0.14 f 0.13
framework. Although rare, this case is * = Xylene dissolved sample.
"=190'C12.16 Kg.
indicative of relaxation times (or moduli)
having non-uniform temperature depen-
dence, and does occur in practice (e.g., Table 2. Results for LLDPE.
certain multiphase systems). MI
Step 2: Calculate the horizontal and vertical activa- 9/10 P EH E"
tion energies, as described above. Examine Resin Comonomer min g/cm3 kcal/mole kcal/mole
the confidence intervals of the vertical activa- LLDPE-1 Butene 1.0 0.918 7.27f0.19 0.61 ko.14
tion energy and discard if found to be statisti- LLDPE-2 Butene 1.0 0.918 7.07 & 0.14 0.0
cally insignificant (23). LLDPE-3 Octene 1.O 0.920 7.77 k 0.24 0.25 k 0.18


Temperature Dependence of Polyolejn Melt Rheology

in agreement with literature results. The highest acti- rial, and are plotted in Fig. 5 (three temperatures) as
vation energy (7.77 kcal/mole) is for LLDPE-3, a n shear stress vs. shear rate. Shifting these data (as
octene based LLDPE (as opposed to LLDPE-1 and explained earlier) gives the graph of Fig. 6, which
LLDPE-2, which are butene based LLDPEs). This re- shows good superposition, i.e. steady-shear data were
sult was expected since the activation energy in- superposed with the activation energies determined
creases with the molar volume of the repeating unit from linear viscoelastic data.
of the polymer (1 5).
The vertical activation energy is nearly zero for
LLDPE-2 and LLDPE-3. For LLDPE-1 it is 0.61 Tabulated results for PP are given in Table 3 for
kcal/mole, which represents a vertical shift factor of seven different resins. Averaging out the horizontal
only 10% between 150 and 2 10°C. activation energies gives a value of 9.4 kcal/mole,
As a typical example consider the LLDPE- 1; a graph again close to values reported in the literature (9- 10
of log,,(tan 6) vs. log,,(G*) is given in I?g. 2. Raw kcal/mole). The vertical activation energy turns out
data of storage modulus and dynamic viscosity vs. to be zero, or very small and marginally significant.
frequency at different temperatures are plotted in
Fig. 3. These same data upon shifting give the graph
of Fig. 4. The good superposition is a verification that Results for EVOH (ethylene-vinyl alcohol copoly-
the activation energies used for the shifting (listed in mer) are given in Table 4 for two different resins. The
Table 2, and in Fig. 4 ) are adequate. average horizontal activation energy is 14.6kcal/mole
Steady-shear data were also available for this mate- and the vertical activation energy is zero.
It is not clear why the vertical activation energy
turns out to be zero, given that EVOH is produced

Shi'!ed t o 190°C
t,=7,270 cal/rnole
Ev- 610 cal/rnole
a" lI

?210 2 - 4

10' 105
Complex Modulus, d y n j c m
2 I I ,,q
I I I llllil
it 10 1 - 10
I Shifted Frequency
Fg. 2. Loss tangent us. complex modulus data a t differ- I/)

ent temperatures superimpose for a linear polyethylene Rg. 4 . Shijted data of storage modulus and dynamic uiscos
( L D P E -1). ity us. frequency, showing superposition (LLDPE-1).


4 T= 170°C
T= 1 9 0 ° C

-r 1 0 5 4

t +
w i LLDPE- 1 i


T= 170°C
Capillary Data
Bagley & Rablnowitch

- T=190°C
corrections applied

FTg. 3. Storage modulus and dynamic uiscosity us. frequency FTg. 5. Steady-shear data (shear stress us. shear rate) a t
a t different temperatures (LLDPE-1). different temperatures (LLDPE-1).

H. Mavridis and R. N . Shroff

through the hydrolysis of VAE (Vinyl acetate-Ethyl- EVOH may have a value of d In( r z > / d T , such that
ene copolymer). VAE, in turn, does have long chain the effect of long chain branches on the temperature
branching and does require a vertical shift. Part of dependence of rheology is minimal. More work on
the reason may be that all branches through the this matter is in progress.
acetate group in VAE are lost after hydrolysis and,
therefore, EVOH is expected to have substantially LDPE
less long chain branching than its predecessor VAE Results for LDPE (long chain branched) are tabu-
(26).A second possible reason relates to the tempera- lated in Table 5: invariably, the vertical activation
ture coefficient of unperturbed chain dimensions energy is non-zero and around 2 kcal/mole (with the
d In( r:)/dT. As was pointed out by Graessley (12). exception of the model star-HPBs).
"species where long branches have little effect on the Let us consider in some detail a typical LDPE-the
temperature coefficient of viscosity also have small LDPE-4. Raw data from different temperatures are
temperature coefficients of chain dimensions." For plotted a s log,,(tan 8 ) vs. loglo(G*)in Fig. 7. Obvi-
polyethylene this coefficient is d In( r z > / d T = ously the data do not superimpose, but they fall onto
- 1 . 1 * l o p 3" C - (12, 27, 28). Stars of polystyrene, parallel curves, an indication that a vertical shift is
polybutadiene, and polyisoprene have positive and the necessary and sufficient step to superpose the
small temperature coefficients of chain dimensions data. Indeed, the effect of vertical superposition is
and do not show the anomalous behavior due to long shown in Fig. 8. Now the data superimpose and they
chain branching as polyethylene does (12). Therefore, form a master curve. It is important to note that the
two shifts (horizontal and vertical) are sufficient to
superpose all viscoelastic properties. As is shown in
Shifted to
EH=7.270 cal/rnole
E,,=610 c a l / r n o l e *B
db % 1 Fig. 11, both the storage modulus vs. frequency data
(Fig. 9 ) and loss modulus vs. frequency data (Fig.10)
superimpose after a horizontal and vertical shift are
applied to the data.
The importance of the vertical shift will be further
illustrated by examining the problems resulting from
a horizontal-only shift. Traditionally, the data are
shifted horizontally only. As was shown for HDPE,
LLDPE, PP, and EVOH (and for a wide variety of other
systems) a horizontal-only shift is adequate for linear
m 1 Bagley '& Rabinowitch materials. When we attempt to apply a horizontal-only

2 4
I corrections applied
shift for long chain branched materials we see the
following: Fig. 12 shows a horizontal-only shift for
+I g o

10 --e--7-YTr
1 1 I I
6 7 8 9 ' 2
I 5
p-7 storage modulus vs. frequency data. The superposi-
m 1 10 10 tion is indeed good for the storage modulus data. This
Shifted Sbeor Rate ( s e c - ) superposition was made possible by allowing the acti-
Fig. 6. The data of FUJ. 5, shijted at 190°C (LLDPE-1). vation energy to become a function of storage modu-
lus. Specifically, the requirement of a horizontal-only
shift produces a set of temperature shift factors at
Table 3. Resultsfor Polypropylene. every level of storage modulus, from which set of shift
MFR* EH EV factors we compute a n activation energy at every level
Resin Comments g/10 min kcal/mole kcal/mole of storage modulus.
PP-1 Homopolymer 37 8.95 k 0.23 0.29 & 0.17 When the temperature shift resulting from Fig. 12
PP-2 Homopolymer 26 9.94 k 0.32 0.24 k 0.19 is applied to the loss modulus data we get Fig. 13,
PP-3 Random 2 8.74 &- 0.22 0.36 &- 0.16 which shows that the data do not superimpose very
Copolymer well. (One must be very critical and demanding on
PP-4 - 0.33 10.74 f 0.22 0.0
PP-5 Homopolymer 42 8.79 k 0.26 0.0 the goodness of superposition. If the non-superposi-
PP-6 Copolymer 0.2 9.90 k 0.55 0.0 tion of the data in Fig. 1 3 is dismissed as simple
PP-7 Copolymer 13.4 8.51 f 0.66 0.0 scatter, then one would totally miss the problem. In
25.9% CF fact this is what has happened in the literature, and
* = 2 3 O C / 2 . 1 6 Kg it was pointed out earlier by Verser and Maxwell (19).
Moreover, the activation energy required for shifting
the data in Fig. 12 is now a function of storage
Table 4. Resultsfor EVOH.
modulus, a s shown in Fig. 14 (solid line), decreasing
MI En EV with increasing modulus.
Resin Comments g/10 min kcal/mole kcal/mole Similarly, when we apply a horizontal-only shift to
EVOH-I 32% mole C; 1.3 15.45 k 0.96 0.0 the loss modulus data (by allowing the activation
I.19 g/cm3 energy to become a function of loss modulus), we get
EVOH-2 44% mole C; 5.5 13.78 k 0.60 0.0 good superposition for loss modulus a s shown in
1. I 4 g/cm3
Fig. 15. But applying this temperature shift to the


Temperature Dependence of Polyolejln Melt Rheology
Table 5. Results for LDPE.

Mi P EH E"
Resin Comments g/10 min glcm3 kcailmole kcal/mole
LDPE-1 Xylene - 0.931 14.67 k 0.30 2.31 f 0.26
(NBS 1476*) Dissolved
LDPE-2 Pellet 1.2 0.931 16.02 ?C 0.38 2.47 + 0.31
(NBS 1476)
LDPE-3n Xylene -
0.919 15.27 f 0.28 2.44 f 0.17
LDPE-4n Pellets 1 .I 0.919 16.01 f 0.15 2.28 f 0.09
LDPE-5 Pellets 2.9 0.926 15.78 +_ 0.20 1.45 0.13
LDPE-6 Pellets 1.8 0.923 14.62 k 0.17 1.77 f 0.13
3-Star HPB M, = 98,000 - - 15.31 f 0.31 1.01 f 0.32
Data from (10)
4-Star HPB M, = 133,000 - - 16.47 ?C 1.43 0.97 f 0.37
Data from (10)
*=Shift at constant modulus performed.
=Same LDPE.

-- T= 130°C

LD P E -. 4

0 % .

T- 1 7 0 ° C
T- ' 90°C
I \ i
- * a .
r 1-

Shifted t o T=15OoC i
€,=I 6,010 cal/mole
E,= 2,280 c al / mo l e

Fg. 7. Loss tangent us. complex modulus data at dlfferent R g . 8. The data of Fg. 7, superimpose after a "vertical"
temperatures fall onto parallel curves for a long chain (modulus) shift has been applied to the data (LDPE-4).
branched polyethylene (LDPE-4), illustrating the need for a
modulus shift with temperature ("vertical" shifi.

storage modulus data does not produce good super-

position (Fig. 16). Also, the activation energy be-
comes a function of loss modulus (dashed line in Fig.
Steady-shear data were also available for LDPE-4
at different temperatures, shown in Fig. 17. Applying
a horizontal as well as a vertical shift to these data
with activation energies determined from the dy-
namic data produces relatively good superposition,
as shown in Fig. 18.
It should also be noted that comparison of results
between LDPE-2 and LDPE-1 (pellet form and solu-
tion dissolved sample) and between LDPE-4 and
LDPE-3 (pellet form and solution dissolved sample)
shows only marginal differences (with regards to the Fg. 9. Storage modulus us. frequency at different tempera-
temperature dependence). A more systematic study tures (LDPE-4).
of the effect of shear prehistory on the temperature
dependence of LDPE rheology is in progress and will
be reported in a future paper. which one would expect to be related to long chain
As Table 5 shows, all LDPEs require a vertical shift. branching (LCB) parameters. Figure 19 shows appar-
There is a certain variation in the vertical activation ent (not corrected for branching) molecular weight
energy E , (e.g., compare values for LDPE-4,5,6), distributions for three LDPEs from Table 5. For the


H. Mavridis and R.N . Shroff

Shifted t o 1=150"C
a t c o n s t a n t G'
with EH=EH(C)

Ftg. 10. Loss modulus us. frequency at different tempera Ftg. 12. The storage modulus data of Fig. 9 (LDPEE-~),
tures (LDPE-4). horizontally only, with an activation energy as afunction of
storage modulus.

LDPE-4 * 0 4

0 0 :

Shifted t o T=15O0C
E,=16,010 col/rnole 3
Ev= 2,280 col/mole

a0 I
:t> ~
m r T T - -7-1, , -7--m----

10 r,
-7 -+
'0 '0
STIfteo F r e q J e T c j
Ftg. 1 I. The storage and loss modulus data of Fqs. 9 and 10
after horizontal and vertical shifting, showing good superpe Ftg. 13. The temperature shift from Fig. 12 applied to the
sitwn (LDPE-4). loss modulus data (LDPE-4).

LDPEs in Fig. 19 we expect a n ordering in the level of Naturally, the question arises as to what makes
LCB a s LDPE-G/LDPE-5/LDPE-4, with LDPE-4 hav- long chain branched materials to require a vertical
ing the highest level of LCB. The E, values of Table 5 shift. One potential answer can be provided with
do not follow this order. One might expect a better reference to Eq I I and the concept of molecular
correlation of E , with other LCB parameters such as weight between entanglements (suggested also by
branch length and distribution. Unfortunately, this Graessley in his discussions of the effect of long
type of information was not available for the LDPEs chain branches on the rheology and polymer melts
studied in this work. I t appears that the EH and E, and its temperature sensitivity (11 , 12)). For a verti-
values within the LDPE group of resins do not corre- cal activation energy of 1.0 kcal/mole and greater,
late with the level of LCB present in these resins. Eq 1 1 predicts that the molecular weight between
It is interesting that even Graessley's model star- entanglements decreases with increasing tempera-
branched polybutadienes (lo) show the necessity of a ture. This means that the entanglement density in-
vertical shift (HPB in Table 5). In fact, Ref. 10 dis- creases with increasing temperature, although it is
cussed the anomaly observed when shifting the data not clear why would the entanglement density in-
horizontally only (i.e., the modulus dependence of crease (with increasing temperature) for long chain
activation energy, see Fig. 9 in Ref. 10). As was branched and not for linear polymers.
mentioned repeatedly above, a graph of log,,(tan 8 ) As noted by Graessley (12). the temperature depen-
vs. log,,( G*) can immediately identify the necessity dence of entanglement spacing is one of two con-
(or absence thereof) of a vertical shift. The graph for tributing factors. The second relates to the suppres-
the 3-star HPB in Fig. 20 clearly shows parallel sion of reptation in branched polyethylenes (12).
curves that require vertical shift. Linear chains rearrange by reptation; long chain


Temperature Dependence of Polyole$n Melt Rheology

O f


~ ~ Shift a: Coqs ta nt G',E,=E,(G')

Shift a: Conskon: G ', E.,-E,(G")

Ftg. 14. Activation energy as a function qf storage modulus Fig. 16. The temperature shlft from F ~ J . 15 applied to the
Cfor horizontal-onlyshijt at constant G') and as afunction of storage modulus data (LDPE-4).
loss modulus (for horizontal-onlyshtft at constant G").

r TT ~ 1 m r . .~


5h'f:ed t o T = ' 5 0 " C -

c t constant G'
P rwth E,-E,(G")


' ,./
Fig. 17. Steady-shear data (shear stress us. shear rate) at
Frg. 15. The loss modulus data of Fig. 10 (LDPE-4),shtfted di'erent temperatures (LDPE-4).
horizontally only, with an activation energy as a function of
loss modulus.
dence (characterized by a "horizontal activation en-
ergy") and all relaxation moduli have the same tem-
branches block reptation and rearrange by a process perature dependence (characterized by a "vertical
of "path breathing" (12). Both mechanisms have an activation energy"). Procedures were outlined for ex-
impact on the temperature dependence of rheology, tracting these activation energies for rheological data.
but it is not clear how to separate the two effects. The recommended (and adopted) approach was to
EVA extract the temperature dependence from linear vis-
coelastic data: in this case the two activation energies
Results for EVA (ethylene-vinyl acetate copoly- can be estimated independently (and statistically un-
mers) are tabulated in Table 6 for two resins. The correlated).
horizontal activation energy for EVA- 1 is higher than I t was also shown theoretically (and demonstrated
that for EVA-2, presumably due to its higher VA experimentally) that neglect of the vertical shift leads
content. As expected, a vertical activation energy of to a stress (or modulus) dependent activation energy
similar level as that of LDPE was found, thus further and makes impossible the superposition of both loss
reconfirming that the necessity of the vertical shift is and storage modulus data. The long standing prob
associated with long chain branching. lem of a stress-dependent activation energy in long
chain branched LDPE was identified as originating
from the neglect of the vertical shift.
The fundamentals of the temperature dependence The theory was applied successfully to many poly-
of rheology were reviewed. We considered the case (by olefin melts, including HDPE, LLDPE, PP, EVOH,
far the most common in polymer melts) where all LDPE, and EVA. Linear polymers (HDPE, LLDPE, PP)
relaxation times have the same temperature depen- and EVOH do not require a vertical shift, but long

H. Mavridis a n d R. N . Shroff

3-Star HPB, l\llw=98,C100


;;A' T
2 : ; ; ; ; ; - : m B
105 10
FKJ. 18. The data of Fg. 17 shLJed at 150°C. using the Co m p 1 e x M o d u I u s ( d y n / c rri *)
activation energies computed from dynamic data (LDPE-4). Fg. 20. Rheological data f o r 3-Star HPB (datafrom (10)).

i Table 6. Resultsfor EVA.
-.. -
MI Weight E, E"
1 .
- . P
Resin g/cm3 g/10 rnin %VA kcal/mole kcal/rnole
EVA-1 0.98 3.5 51 17.63 f 0.30 1.65 k 0.15
EVA-2 0.932 28 14.8 14.30 0.53 2.04 f 0.36

below, shows that a pre-established temperature de-

pendence of rheology allows the prediction of rheolog-
ical properties at one temperature given these proper-
ties at another temperature.
The Melt Index shift with temperature can be
worked out by considering the temperature depen-
dence of steady-shear data and the definition of the
Melt Index. The Melt Index (MI) can be approximately
computed from steady-shear data according to the
Fg. 19. Apparent Molecular Weight Distributions (uncor- equation:
rected for long chain branching)for three LDPEs.
MI = __ at cr= 1.95. lo5 dyn/cm2 (A.l)
chain branched polymers do (LDPE, EVA). Steady- 2.4
shear viscosity data can be superimposed using acti-
vation energies extracted from dynamic data. Note that the load specified in the MI (2.16 Kg)
determines the stress U .The temperature affects the
APPENDIX: MELT INDEX SHIFT FOR EVA shear rate, through the temperature dependence of
As a n example drawn from industrial practice, con-
The temperature dependence of steady-shear data
sider the Melt Index (MI) of a low %VA (vinyl acetate)
was examined in 92.6, and it was quantified in Eq 29.
EVA (ethylene-vinyl acetate) copolymer (e.g., 5% VA);
For the purpose of the derivation to follow, we will
it is measured at 2.16 Kg load at 190°C.Suppose we
assume that the shear stress vs. shear rate can be
want to compare this resin with a high %VA EVA
approximated by a power-law relationship, in the
(e.g., 30%VA), which is characterized by a Melt Flow
neighborhood of u = 1.95*lo5 dyn/cm2:
Rate (MFR) measured at 2.16 Kg load at 125°C (high
%VA EVA degrades at high temperatures). We need ff = cl. y" (A.2)
to know what is the equivalent MFR of the low %VA
EVA, or the equivalent MI of the high %VA EVA. I t where c1 is a proportionality constant, known a s the
turns out that we can estimate the equivalent MI of consistency index, and n is the power-law-index. To
the high %VA EVA, with the relation: include the temperature dependence, Eq A.2 must
be written as:
MFR @ (2.16 Kg, 125°C)
= 0.116 * MI @ (2.16 Kg, 190°C). b,. CT = c1.( 01,. i / ) '' "4.3)
for high %VA EVA
where a, and b, are the horizontal and vertical shift
The above equation, whose derivation is considered factors defined in Eqs 8 and 12 respectively. Combi-

Temperature D e p e n d e n c e of P o l y o l e j h Melt Rheology
nation of E q s A. 1 and A . 3 gives:

Taking the reference temperature To that appears

in the definition of a, and b, to be 190°C. and using
the fact that the stress (T is kept constant during
melt index measurements at different temperatures, /' i

we derive the following from E q A.4:

MI( T ) b+/"
MI(190"C) QT
0 E~x p .0 Data
0 ~
E x p . Data'
A a a a
Equation A . 5 can now be used to shift melt index
__ Theory
with temperature. Application to high EVAs is consid- * : Stabilized d u r i n g
ered below: from Table 6, we use E , = 14.9 kcal/mole Mi m e a s u r e m e n t
and E , = 1.97 kcal/mole (estimated averages for the
range 13 to 30%VA). Then the shift factors from 125 1 10 !O
to 190°C are (substituting in E q s 8, 12, T = 125"C, MI ( I S C ' C , 2 . ; 6Kg)
To= 190°C): a T = 14.08 and bT= 1.42. For the power- Fig. A. 1 . Comparison of theory and experiment for Melt In-
law-index n we will use the value l / n = 1.4 at ( T = d e x shift with temperature f o r EVA (EthyleneVinyl Acetate
1.95* lo5 dyn/cm2, determined from viscosity data copolymers).
on a series on EVA resins (the value l / n = 1.4 repre-
sents a n average). Substituting the above shift fac-
tors and power-law-index into Eq A. 5 we get: 9. R. S. Porter, J . R. Knox, a n d J . F. Johnson, Trans. SOC.
RheoL, 12,409 (1968).
MI( 125°C) 10. V. R. Raju, H. Rachapudy, and W. W. Graessley, J .
=0.116 Polyrn Sci.: Polyrn Phys. Ed., 17,1223 (1979).
MI( 190°C) 11. W. W. Graessley, Acc. Chern Res., 10,332 (1977).
12. W. W. Graessley, Macromolecules, 15,1164 (1982).
E q u a t i o n 6 is the equation used earlier in the intro-
13. A. Santamaria, Mater. Chern Phys., 12,l(1985).
duction of this appendix. Comparison with experi- 14. R. S. Porter a n d J . F. Johnson, J . Polyrn Sci., Part C ,
mental data is given in Fig. A. 1 ; the line is Eq A . 6 15,365 (1966).
and the symbols are experimental data: the compari- 15. R. S. Porter a n d J. F. Johnson, J . Polyrn Sci., Part C,
son shows reasonable agreement. 15.373 (1966).
16. M. S . Jacovic, D. Pollock, a n d R. S. Porter, J . Appl.
ACKNOWLEDGMENT Polyrn Sci., 23,517 (1979).
17. M. Rokudai a n d T. Fujiki, J . Appl. Polyrn Sci., 26,1343
The authors are indebted to Quantum Chemical (1981).
Corp., US1 Division for permission to publish this 18. H.M. Laun, Prog. Coll Polyrn Sci., 75,111 (1987).
paper. Helpful comments by Dr. L. Wild are also 19. D. W. Verser a n d B. Maxwell, Polyym Eng. Sci., 10,122
acknowledged. 20. J . T. Gotro a n d W. W. Graessley, Macromolecules, 17,
2767 (1984).
REFERENCES 21. J. T. Gotro and W. W. Graessley, and L. J . Fetters,
1. C. D. Han and K.-W. Lem, Polyrn Eng. Reu., 2, 135 Macromolecules, 17,2775 (1984).
(1982). 22. J . M. Carella, J. T. Gotro, a n d W. W. Graessley, Macro
2. E.R. Harrell and N. Nakajima, J . Appl. Polyrn Sci., 29, molecules, 19,659 (1986).
995 ( 1984). 23. G. E. P. Box, W. G. Hunter, a n d J. S. Hunter, Statistics
3. R. A. Mendelson, Polyrn Eng. Sci., 8 , 235 (1968). for Experimenters, J . Wiley, New York (1978).
4. H. Markovitz, J . Polyrn Sci.: Symp. No. 50,431 (1975). 24. B. W. Terry a n d K. Yang, SPE J., 20 (6). 1 (1964).
5. A. J. Staverman a n d F. Schwarzl, in Die Physik der 25. R. N. Shroff and M. Shida, J . Appl. Polyrn Sci., 26,
Hochpolymeren Vol. 4, Ch. 1, H. A. Stuart, ed., Springer 1847 (1981).
Verlag, Berlin (1956). 26. S. H. Agarwal, R. F. Jenkins, a n d R. S. Porter, J . Appl.
6. J . D. Ferry, Viscoelastic Properties of Polymers, Third Polyrn Sci.. 27,113 (1982).
Edition, J . Wiley, New York (1980). 27. J. E. Mark, Rubber C h e m Technol, 46,593 (1973).
7. R. Sabia, J . Appl Polyrn Sci., 8 , 1651 (1964). 28. P. J. Flory, Statistical Mechanics of Chain Molecules,
8. R. A. Mendelson, SPE Trans., 5,34 (1964). Interscience, New York (1969).