Sunteți pe pagina 1din 14

Chapter 16

Fracture and Fatigue Behavior of


Amorphous (Co)polyesters as a
Function of Molecular and Network
Variables
J. Karger-Kocsis
1. Introduction
Thermoplastic molding operations of slowly crystallizing linear polyesters,
such as poly(ethylene terephthalate) (PET) or poly(ethylene naphthalate)
(PEN), usually result in amorphous products. The amorphous phase for-
mation is strongly favored by high cooling rates during processing. These
homopolyesters are characterized by good optical properties (high trans-
parency and gloss, low haze) and high ductility, but they are prone to
physical aging and crystallization. These result in material embrittlement,
accompanied by severe notch sensitivity. In addition, articles with thicker
walls are never fully amorphous when molded from homopolymers. The
need to circumvent the aforementioned undesirable effects triggered R&D
activities on amorphous copolyesters. These copolyesters are prepared when
a second diol or dicarboxylic acid (or their related precursors) is used in the
polycondensation reactions. The strategy of the production of amorphous
copolyesters is to reduce the chain flexibility by incorporating stiff moi-
eties (such as cyclohexylenedimethylene units, CHDM) and/or to disrupt
the chain regularity (e.g., partial replacement of terephthalic by isophthalic
acid) in order to diminish the tendency to crystallization. The beneficial
Handbook of Thermoplastic Polymers: Homopolymers, Copolymers, Blends, and Composites
Edited by Stoyko Fakirov
Copyright 2002 WILEY-VCH Verlag GmbH, Weinheim
ISBN: 3-527-30113-5
718 J. Karger-Kocsis
properties of such amorphous copolyesters are excellent transparency, high
toughness, small thermal shrinkage and good dimensional stability. Fur-
ther, the glass transition temperature, T
9
, and the related heat distortion
temperature can be tailored via the type and content of the additional
comonomer [ I ] . Commercially relevant copolyesters contain ethylene glycol,
terephthalic acid, and 1,4-dimethylol-cyclohexane (CHDM; e.g., Eastar
PETG 6763 and PCTG 5445 of Eastman Chemical Co.) or CHDM, tereph-
thalic, and isophthalic acid (e.g., Eastar A150 of Eastman Chemical Co.).
Though the crystallization of such copolymers is efficiently suppressed, it
may occur from the rubbery state (above T
9
and well below the melt-
ing temperature, termed cold crystallization) or from the melt (close to the
melting temperature). Figure 1 illustrates this behavior of some amorphous
homo-PET and copolyesters. In the differential scanning calorimetry (DSC)
traces, one can recognize the cold crystallization of PET and A150, whereas
PETG and PCTG do not show thermally induced crystallization, at least
under the DSC testing conditions set. Amorphous (glassy) copolyesters
are widely used in packaging (crystal-clear thermoformed or blow molded
containers; transparent injection molded parts; non-oriented, uniaxially or
biaxially oriented films), and in building and construction (e.g., glazing,
indoor and outdoor signage).
100 150 200
Temperature (
0
C)
250 300
Figure 1. Comparison of DSC heating (5C/min) traces for some amorphous
homo- and copolyesters
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 719
Note that toughness is a key property in all above listed application
fields. On the other hand, less information is available on the fracture and
failure behavior of amorphous polyesters. Further, the dependence of the
fracture response on molecular and network variables is not yet understood.
In order to shed light on the latter issue, first the inherent toughness of the
polymers, if any, should be determined. The working hypothesis of the au-
thor was that this could be realized by adopting suitable techniques of the
fracture mechanics. To deduce the supposed relationships between the frac-
ture and fatigue responses and molecular parameters, fracture mechanical
tests were performed on samples of various molecular weight and entan-
glement characteristics. Considering the aforementioned tendency of some
glassy polyesters to crystallize, this effect also is briefly surveyed. Major
emphasis is laid, however, on the relationship between the post-yield defor-
mation behavior and molecular and network variables. This is a key issue
if one aims at understanding the ductile and brittle failure processes and
the phenomenon of notch sensitivity. Further, knowledge of the molecular
and entanglement network dependence of the post-yield deformation should
contribute to a better understanding of large scale forming and deforma-
tion processes (e.g., cold drawing, blow molding, stretching) even under
high frequency conditions (i.e., fracture under impact loading conditions).
As mentioned above, this chapter is dedicated to the molecular dependence
of the fracture mechanical response of amorphous (co)polyesters. For the
state of knowledge of the physics of glassy polymers, the reader is directed
to the excellent book of Haward and Young [2].
2. Fracture toughness and fatigue crack growth behavior
Fracture mechanics aims at determining the response of a cracked mate-
rial to applied loads and at offering methods to measure the toughness.
Each fracture mechanical method defines, however, its own version of the
resistance of the material to fracture (called "fracture toughness") [3]. The
beauty of these approaches is that they give values which are either equiv-
alent (under given conditions) or can be converted into one another when
additional material properties are known. Thus, various fracture mechani-
cal results can be collated. This is a great advantage over the standardized
test methods, the results of which can be compared only if the specimen
preparation and testing conditions were exactly the same. For a detailed de-
scription of the various fracture mechanical methods, the interested reader
is addressed to valuable books [3-7].
2.1. Linear-elastic fracture mechanics
This approach works for brittle materials that fail by catastrophic crack
growth after reaching a threshold stress value. The related criteria rely on
either the stress field ahead of the crack tip (stress intensity factor or frac-
720 J. Karger-Kocsis
ture toughness, K
c
) or the energy release during crack extension (strain
energy release rate or fracture energy, G
0
). These two criteria find applica-
tion to amorphous and semicrystalline polyesters, showing limited ductility
in the related test. The linear-elastic fracture mechanics work particularly
well for the determination of the fatigue crack propagation (FCP) behav-
ior (see Section 5). Draft standards for the K
0
and G
0
determination of
polymers are available by the ESIS TC-4 and ASTM D-20 Committees.
2.2. Elastic-plastic and post-yield fracture mechanics
For the toughness assessment of tough polymers the J-integral, the crack
tip opening displacement (CTOD) and the essential work of fracture (EWF)
methods are mostly used.
2.2.1. J-I ntegral
The J-integral represents a path-independent integral around the crack
tip and thus considers also the plastic deformation at the crack tip. The
critical value of the J-integral (J
c
, its physical meaning being full crack
tip blunting prior to crack growth) can well be determined by the multi-
ple specimen technique (J^-curve where R stands for resistance). A draft
for the determination of J
0
was proposed by the ESIS TC-4 group. There
are, however, some problems with this technique. First, complete crack tip
blunting is an exceptional phenomenon (but can exist!) since the crack tip
of polymers may experience crazing or yielding. Therefore, no universal
equation holds for the blunting [8]. Some researchers follow the alterna-
tives given by ASTM E813 (valid for metals) in 1981 (J
c
determination by
linear extrapolation), and in 1987 and 1989 (description of the JR curve
by a power law function), respectively. Others determine J
0
as the direct
intercept with the y (J)-axis in order to overcome the ambiguity associated
with the blunting. Second, the crack advance due to partial loading of the
specimens (which yields the x-axis data of the JR curves) can hardly be
determined in polymers undergoing "work hardening" via various deforma-
tion processes. Note that after loading, the specimens should be unloaded
and broken up at appropriate conditions to detect the crack advance. This
procedure may "obscure" the real crack growth because of superimposed
molecular and microstructure-related crack tip shielding effects. Unfortu-
nately, the compliance-related crack growth determination techniques do
not always work for polymers. Third, the J-integral is surface-related. On
the other hand, the deformation zone in the crack tip region may deeply
penetrate also the bulk or cause a plane strain - plane stress transition,
associated with thickness reduction. Both mechanisms render difficult the
determination of valid J
c
values. Irrespective of these problems, the J-
integral technique has been successfully applied to numerous polymers,
including blends with linear polyesters [9-12].
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 721
2.2.2. Crack tip opening displacement
The crack tip opening displacement (CTOD) criterion is linked to the crack
opening prior to extension. Thus, CTOD also considers the plastic flow
capability of the polymer. The problem is again that many polymers tend
to craze or flow instead of exhibiting homogeneous plastic deformation. The
CTOD is a very straightforward technique reflecting the effects of the initial
structural order. Unfortunately, it is rather seldom adopted for polymers
(e.g., [13,14]) possibly because of instrumentation and crack inspection
problems.
2.2.3. Essential work of fracture
Recently, the essential work of fracture (EWF) approach is gaining accep-
tance to assess the toughness response of ductile polymers. As films are
preferred for this technique, the outcome is a plane stress toughness value.
Note that the EWF method is analogous to the J-integral technique since
both of them represent a type of resistance curve (-curve). Accordingly,
there is an agreement between the related parameters [15]. The greatest
advantage of EWF over the J-technique is that a clear distinction between
surface- (essential part) and volume-related (non-essential part) work is
made. The basic prerequisite of this method is that the ligament of the
specimen should be fully yielded before the crack propagation starts. This
is again rather seldom the case since in many of the polymers tested, liga-
ment yielding occurs simultaneously with a more or less pronounced crack
growth. On the other hand, it was found that amorphous (co)polyesters are
probably the best model polymers to perform this kind of test [16]. A draft
on how the EWF test should be performed (see Figure 2) was also elabo-
rated by the ESIS TC-4 group. As results achieved by this technique are
preferentially reported here, it is essential to introduce EWF accordingly.
The EWF theory, credited to Broberg (see [5,15-19] and references
there), splits the total energy required to fracture a precracked specimen
into two components: the essential, W
e
, and the non-essential or plastic
work of fracture, W
p
, respectively. The first term is needed to fracture the
polymer in the process zone and thus generate new surfaces. W
p
is the
actual work consumed in the outer plastic region where various energy
dissipation mechanisms take place. The total fracture energy, W/, calcu-
lated from the area of the force-elongation curves, can thus be expressed
by (Figure 2)
W/ = W
e
+ W
p
. (1)
Considering the surface (i.e., It) and volume dependence (i.e., I
2
t) of the
722 J. Karger-Kocsis
(W
P
)
B = 35 mm
Ligament yielding (onset of
necking via cold drawing)
20 mm
Elongation
w
p
= w
y
Thickness (i)
Figure 2. Determination of the essential, We, and non-essential or plastic work
of fracture, Wp, by using deeply double edge-notched tensile loaded (DDEN-T)
specimens of various ligament lengths. The energy partition (yielding, W
y
, and
necking and tearing, W
n
, respectively), recommended by the author's group, is
also given
constituent terms, Eq. (1) can be rewritten in the specific terms
Wf = wjt + w
p
l
2
t
Wf
Wf = - =w
e
+w
p
l
(2)
(3)
where / is the ligament length, t is the specimen thickness, and is a
factor related to the shape of the plastic zone. Based on Eq. (3), w
e
can be
estimated from the intercept of the linear regression of the plot of Wf vs. I
and the u7/-axis. w
e
or, more exactly, the critical value of w
e
under mode
I plane-strain conditions, w
e
,i? should represent a material parameter. On
the other hand, the slope, w
p
, of the Wf vs. I resistance curves or, more
exactly, w
p
, is a direct measure of the resistance to crack growth.
3. Molecular parameters and their determination
3.1. Molecular weight and its distribution
The mean molecular weight (MW) of linear polyesters is usually determined
by the Kuhn-Mark-Houwink-Sakurada equation
[ ] =
(4)
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 723
where [ ] is the limiting viscosity number or intrinsic viscosity, M
v
is the
viscosity average molecular mass, and and a are constants for a given
solvent and temperature. Let us recall that M
v
is between the number-,
M
n
, and weight-average molecular mass, M
117
, but closer to the latter. As
M
n
can be determined by end-group analysis and M
w
by light scattering
techniques in solution, similar relationships to those indicated in Eq. (4)
can be found for [77] vs. M
n
and [77] vs. M
w
, respectively [I]. [77] is defined
by (e.g., [2O])
(5)
c->O r]
s
c c->O C c->O C
where is the viscosity of the solution,
3
is the viscosity of the pure solvent,
c is the polymer concentration, and T7
sp
and ?7

are the so-called specific


and relative viscosities, respectively. According to Eq. (5), the intrinsic
viscosity is determined by the linear extrapolation of the respective values
achieved on a series of dilute solutions to c = O. Since the determination
of [77] is time consuming, determination of the inherent viscosity (?7inh) or
logarithmic viscosity number became more widespread
?7inh =
According to ASTM D 4603-96, 77
inh
of PET is determined at
0.5g/100ml concentration in a solution of 60/40 phenol/ 1,1, 2,2-
tetrachloroethane by means of a glass capillary viscometer. The
preferred solvents used to determine the mean molecular weights
according to Eq. (4) are 1,1,1,3,3,3-hexafluoroisopropanol, ra-cresol,
o- dichlor ophenol/ chlor ofor m , met hy lenedichlor ide/ hexafluor oisopr opanol ,
phenol/p-dichlorobenzene, etc. Some of them can be used as elution fluids
in low- and high-temperature size exclusion chromatography (SEC) exper-
iments [I]. The SEC curves are very useful since they provide information
about the MWdistribution, as well. Because of the variants of the r/j
n
h and
MW determinations, the reader should consult the cited work [1] in this
respect.
3.2. Entanglement network
Early investigations on uncured rubbers showed that their deformation is
recoverable after unloading, suggesting the existence of some temporary
network in which the j unctions are chain entanglements of a physical na-
ture ("physical network") rather than chemical crosslinks [21]. The term
"entanglement" refers to some constraint (mostly of topological origin) in
the glassy, rubbery, and molten states of polymers [22]. The model en-
tanglement network should consist of entangled strands linked at the en-
tanglement points. It is characterized by either the mean MW between
724 J. Karger-Kocsis
the adj acent temporary entanglements, M
6
, or the density of the entangle-
ment points, v
e
. M
e
is usually calculated by considering the position of the
plateau of the shear modulus, G^, by analogy with the rubber elasticity
(see, e.g., [23-25])
(7)
N
where p is the density, R is the gas constant and T is the temperature. G^
can be read from the master curves of log G' (shear storage modulus) vs.
log angular frequency, representing polymers of various MW characteris-
tics [26]. The determination of G^ is an uneasy task for (co) polyesters, as
they tend to crystallize in the melt. Further, to assess the plateau region
in the melt, samples in a broad MWrange are needed which can hardly be
produced by polycondensation reactions [27]. In addition, G^ can be calcu-
lated in various ways from experimental oscillation measurements [27,28].
Therefore, it is not surprising that M
e
values reported by various groups
for the same polymer may differ considerably from one another, although
M
e
should be a constant, at least above the critical MW, M
c
. Another
option to estimate M
6
is given by the Bueche theory (see, e.g., [21,24] and
references there)
M
0
w 2M
e
. (8)
M
0
can be read from the kink point of the zero shear melt viscosity plotted
against the MW. Let us recall that the melt viscosity increases linearly
with MWbelow M
c
and follows an exponential law (with an exponent of
3.4) above this M
0
threshold. The entanglement density, z/
e
, is given by
where / is the Avogadro number.
Since chain entanglement seems to control not only the rheological but
also the mechanical behavior of amorphous polymers, keen research interest
is focused on its modeling and calculation [27-3O].
4. Correlation between molecular and fracture parameters
4.1. Static fracture
To find a correlation between molecular and fracture mechanical parame-
ters is probably the easiest task in the case of amorphous polymers [31].
However, amorphous polymers that do not undergo crazing should be se-
lected since one has to distinguish between the effects of the initial structure
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 725
40
30
20
10
Shear band
formation
0 2 4 6 8
Strain (%)
Figure 3. Tensile stress-strain behavior of a PCTG dumbbell and the appearance
of the shear band formed
and that of the loading-induced one [31,32]. The loading-induced molecu-
lar and morphological changes may be associated with considerable work
hardening and thus do not reflect the initial structure any more.
It was shown by the author's group that amorphous (co)polyesters
which fail by shear banding (Figure 3) and diffuse shear yielding are most
suitable, since they satisfy all EWF prerequisites [16,33]. This means that
apart from the similarity of the force-elongation curves, the repeatedly
stressed, but often ignored, precondition of this approach is met, ie., com-
plete ligament yielding prior to the crack growth.
Figure 4 shows characteristic force-elongation curves of deeply dou-
ble edge-notched tensile loaded (DDEN-T) specimens (see Figure 2) of an
amorphous copolyester with various ligament lengths. Surprisingly, these
curves are similar to one another, ie., they can overlap by linear transfor-
mation ("self-similarity"). Obviously, this is the first indication that the ba-
sic requirement of the EWF theory is fulfilled. The force-elongation curves
in Figure 4 exhibit a load drop which indicates the completion of yielding
and the start of subsequent processes (necking and tearing). At this point,
the whole ligament yields suddenly. This can be evidenced by infrared ther-
mography, as yielding is a thermal softening process (Figure 5). The en-
ergy partition between yielding and necking (as sketched in Figure 2) is
the proper means to distinguish between the effects of the initial structure
and the strain-induced one. The results discussed below are mostly related
to experimental conditions and samples where full ligament yielding of the
specimens preceded the crack growth stage. For information regarding the
726 J. Karger-Kocsis
3000
2000 -
1000 -
0 2 4
Elongation (mm)
Figure 4. Comparison of the force-elongation curves of DDEN-T specimens of var-
ious ligament lengths for an amorphous copolyester conatining 68 mol% CHDM;
thickness 3.1 mm; inherent viscosity 0.705 dL/g; deformation rate 1 mm/min. For
A, B and C, see Figure 5 [36]
polymer samples, specimen preparation, EWF tests, and related data re-
duction (usually following the ESIS TC-4 recommendation), the interested
reader is referred to the papers cited below.
Another convenience of amorphous (co)polyesters is that the forma-
tion of the plastic zone in the DDEN-T specimens does not occur by true
plastic deformation, but via cold drawing, as will be demonstrated later.
Cold drawing is, however, controlled by the entanglement network char-
acteristics (entanglement density) [34,35]. This fact helps one to separate
the effects of molecular weight from those of the entanglement network
characteristics [33].
4.1.1. Molecular weight
In the study of the fracture toughness of amorphous copolyester sheets
of varying inherent viscosity (and MW) by the EWF, it was found that
w
e
is a composite term under plane-stress conditions: its constituents are
yielding, w
e
,
y
, and necking and tearing, w
e
^
n
. The force-elongation behav-
ior, displayed in Figures 2 and 4, allowed us to separate the specific work
of fracture required for yielding, w
y
, from that consumed by necking and
tearing, W
n
. As a consequence, the data reduction given by Eq. (3) can be
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 727
Figure 5. Serial infrared ther-
mographic frames taken in the
yielding and necking regions
of a DDEN-T specimen of an
amorphous copolyester containing
68mol% CHDM; taking positions
of the frames are indicated in the
load-displacement curves in Figure
4 [36]
10 ,
, ^-...--^...-... :: : : .-.-
10
represented as [16,33,36,37]
Wf = Wf^y + Wf
1
n = W
e
+ w
p
l (10)
Wf = w
e
+'w l (11)
Wf = w -\-"w l. (12)
Accordingly, the specific essential work of fracture, w
e
, is a combined
term under plane-stress conditions
w
e
(13)
728 J. Karger-Kocsis
50
40
30
20
D
O WJe,y WJ
6
,
DD
D
D
D
llllllQlllOlll
0.65 0.70
Inherent viscosity (dL/g)
0.75
Figure 6. Work of fracture parameters vs. inherent viscosity for an amorphous
copolyester containing 31 mol% CHDM[41]
This energy partition is in disagreement with the theoretical estimate of
the EWF parameters by Mai and Cotterell [19,38]. On the other hand, ex-
perimental results obtained with polymers other than (co)polyesters, such
as unplasticized poly(vinyl chloride) [39] support the validity of Eqs. (U)-
(13). It is worth noting that Ferrer-Balas et al have attempted a slightly
different energy partition for polypropylene, in order to consider the elas-
tically stored energy [40].
The plots of the w
e
partitioned terms (i.e., w
e
, w
e
^
y
, and w
e
^
n
) vs. the
inherent viscosity are displayed in Figure 6 [41]. It is seen that w
e
^
y
is
independent of rj mhj while w
e
and w
e
^
n
experience a considerable scatter.
This finding is very important since it suggests that the inherent viscos-
ity (and thus MW) affects solely the necking behavior. EWF results on
amorphous PEN (aPEN) of various MWs have shown, in fact, that the
effect of MW is reflected in the post-yield behavior (i.e., in the Wf^
n
vs.
I plots according to Eq. (12)). Figure 7 demonstrates the effect of ryinh on
the force-elongation behavior of DDEN-T specimens of aPENs [42]. aPEN
of low 77i
n
h shows unstable necking after the ligament yielding (Figure 7a).
On the contrary, aPEN of high ryinh undergoes stable necking after yielding
and thus the force-elongation curves are similar to amorphous copolyesters
(compare Figure 4 and Figure 7b). This is clear experimental evidence that
the mean MW affects only the necking behavior. In Figure 8, which dis-
plays the plots of W/ and its energy partitioned terms (i.e., Wf,
y
and Wf^
n
)
vs. the ligament length for aPEN of low rymh, the slope of the Wf^
n
vs. I
curve is negative. This fact seems to contradict the EWF theory since the
slope of the plots should never be negative (see Eqs. (3), (11), and (12)).
Fracture and Fatigue Behavior of Amorphous (Co)polyesters 729
800
600 -
I
1 2
Elongation (mm)
800
^ I 5 mm \ / ?i 10 mm\/ ca 15 mm
2 3 4
Elongation (mm)
Figure 7. Force-elongation behavior of DDEN-T specimens of amorphous PENs of
low (a) and high (b) inherent viscosity and thus of mean MW; room temperature,
deformation rate 1 mm/min [42]
This is due to the unstable necking caused by a ductile - brittle transition
in the failure mode. Note that an apparent negative slope suggests that the
preconditions of the EWF are violated. Changes in the stress state (from
plane stress to plane strain) and/or in the failure mode (from ductile to
brittle) during testing of the specimens are the usual causes of a negative
slope. The "embrittlement" observed in the necking stage of the low MW
aPEN should be attributed to the network disentanglement process (see
below) [42]. On the contrary, the necking of aPEN of high MWwas stable
730 J. Karger-Kocsis
150
120
60
30
O Wf
y = 67.54 + 2.53^ R
2
= 0.787
10 20
Ligament length (mm)
Figure 8. Total specific work of fracture, w/ , and its constituent works of yield-
ing, Wf,
y
, and of necking, w/ ,
n
vs. ligament length, Z, for DDEN-T specimens of
aPEN of low inherent viscosity (0.69dL/g); deformation rate Imm/min; y and
represent the related linear regression in the form y = B + Ac, along with the
correlation coefficient, R
2
[42]
and, as expected, all slopes of Eqs. (10)-(12) were positive [42]. In brief,
MWaffects the necking and tearing, but does not influence the yielding be-
havior. As a consequence, an inherent material toughness parameter, if any,
should be related to the yielding. Now let us focus on the failure behavior,
revealing the decisive role of the entanglement network.
4.1.2. Entanglement network
Bearing in mind that the temperature rise detected during static load-
ing of the DDEN-T specimens usually did not reach the T
9
of the related
polymer [36,37,42], the deformation of the glassy polymer should have oc-
curred by cold drawing rather than by true plastic deformation. There is
a simple way to check the onset of cold drawing. By keeping the broken
specimens j ust above T
9
for few minutes, the "plastic" zone should disap-
pear and the original shape of the specimen should be restored [34,35,42].
This is illustrated in Figure 9, on the example of an amorphous copolyester.
Cold drawing means stretching of the entanglement network without dis-
ruption (the latter occurs only in the process zone). The existence of an
"intact" but deformed network is the driving force for the observed "shape
memory" [34,35]. So, one can suggest that the entanglement network and
its characteristics control the fracture mechanical response. This hypothe-
sis is supported by the fact that M
e
is constant beyond a threshold MW
(= MC). Obviously, the mean MWwas higher than M
0
in the homo- and

S-ar putea să vă placă și