Polymer 52 (2011) 1811e1819

Contents lists available at ScienceDirect

Polymer
journal homepage: www.elsevier.com/locate/polymer

Criteria to predict the embrittlement of polycarbonate

Tom A.P. Engels a, b, c, Lambert C.A. van Breemen a, b, Leon E. Govaert a, b, *, Han E.H. Meijer a, b
a
Materials Technology (MaTe), Eindhoven University of Technology, P.O. Box 513, NL-5600 MB, Eindhoven, The Netherlands
b
Dutch Polymer Institute (DPI), P.O. Box 902, NL-5600 AX, Eindhoven, The Netherlands
c
DSM Ahead, Materials Sciences R&D, P.O. Box 18, NL-6160 MD, Geleen, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: A direct correlation is found between the time evolution of the yield stress in unnotched tensile bars and
Received 30 December 2010 that of the impact energy measured using notched tensile bars. In both cases a master curve can be
Received in revised form constructed with an Arrhenius type of shift function, using the same activation energy. Combining these
16 February 2011
experimental findings with numerical simulations lead to a maximum hydrostatic stress as a criterion to
Accepted 16 February 2011
Available online 24 February 2011
predict the onset of failure, shc . Where aging kinetics is not dependent on the polymer’s molecular
weight, embrittlement is, and for higher molecular weights a higher shc is found. Moreover, for high
polymers we observe a more stable craze extension and crack propagation after cavitation, causing the
Keywords:
Polymer glasses
computed impact energies to underestimate the experimental ones for all but the lowest molecular
Criteria for embrittlement weight polymers. Because the maximum in hydrostatic stress under the notch, sh, increases with the
Physical aging polymer’s age, defined by the value of the state parameter Sa, and given the proportionality between
Sa and the yield stress, sy, an alternative failure criterion is proposed which is a molecular weight
dependent critical value of the evolving yield stress, sy,c. It predicts a product’s increased sensitivity to
damage under the influence of progressive aging as enhanced by temperature.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction experimentally by gradually shifting the deformation mode of

Embrittlement poses a serious threat in the application of poly-
mers in structural applications. Materials that initially show
a ductile failure mode, can, with time, fail in a brittle manner. This
change in failure mode is accelerated by annealing, as was demon-
strated by LeGrand [1]. Annealing causes the yield stress of a poly-
mer glass to increase: a phenomenon known as physical aging [2,3].
Aging on one hand improves the life-time under static and dynamic
loading conditions [4e8], but on the other hand simultaneously
increases strain softening, caused by the increase in yield stress,
which results in a stronger tendency to strain localization [9], ulti-
mately leading to complete brittle failure [1,10e13]. Aging thus
results in enhanced long-term properties, but diminished impact
properties. In the presence of a notch embrittlement is attributed to
the buildup of a critical positive hydrostatic stress, on some distance
underneath the surface, notch sensitivity [14,15]. The preferred
mode of deformation of polymers is shear yielding. But, when
a positive hydrostatic stress is present, the only option for the
polymer material to relieve stress, and deform by flow, is by void-
ing, which is the start of crazing. This was convincingly shown
* Corresponding author. Materials Technology (MaTe), Eindhoven University of
Technology, P.O. Box 513, NL-5600 MB, Eindhoven, The Netherlands. Tel.: þ31 40
247 2838; fax: þ31 40 244 7355.
E-mail address: L.E.Govaert@tue.nl (L.E. Govaert).
a notched bar from mode I to mode II bending, reducing the
hydrostatic stress contribution, finally to zero, and enforcing
a transition from brittle to ductile failure [16]. The critical hydro-
static stress at which voiding, followed by crazing, sets in is material
dependent and related to the network density. For low entangled
polystyrene a value of around 40 MPa is found [17,18], whereas for
highly entangled polycarbonate the values range between 80 and
95 MPa [19e21]. The critical hydrostatic stress can not be measured
directly, and is determined using a combined experimental-
numerical approach, for PC e.g. using numerical approaches like the
slip-line field theory [22] and a 3D elasto-plastic (finite element)
analysis [21]. Although the material models used in these
approaches do take yield into account, albeit not time dependent,
they neglect the most characteristic polymer behavior which is
strain softening followed by strain hardening and, even more
important, the change in intrinsic behavior caused by the evolution
of the yield stress with temperature and time. Other influences that
determine the embrittlement of polymers include strain rate
[11,23,24], molecular weight [11,12,23,25], thermal history
[1,10,11,26e29], temperature [11], notch radius [11,12,27,30], sample
thickness (in notched specimen) [10,27,28] and the presence of
solvents, an effect usually referred to as Environmental Stress
Cracking (ESC) [31,32]. Molecular weight, thermal history and ESC
influence the polymers intrinsic behavior, whereas the other
parameters change the experimental conditions, which, once the

0032-3861/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.polymer.2011.02.027
1812 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819
polymer’s intrinsic behavior is known, and taken into account by
using a proper constitutive equation, accurately follows from
predictive numerical modeling. All influences on a polymer’s brit-
tleness can consequently be captured by either the material’s
intrinsic behavior, or by the loading geometry and loading
conditions.
In our group a full 3D large-strain visco-plastic constitutive
model (the so-called Eindhoven Glassy Polymer, EGP, model) has
been developed over the years. It accurately describes the large-
strain deformation behavior of glassy polymers over a broad range
of loading conditions (and in multi-mode versions also the small
strain deformation) and it is based on the evolution of the yield
stress in time, as accelerated by temperature and stress [33,34].
Consequently, the model is able to quantitatively predict long-term
properties, including their enhancement by an increase in yield
stress [4]. In this paper, the model is applied to analyze changes in
macroscopic mechanical behavior, and local stress fields, that occur
upon aging of notched tensile bars. The intention is to identify
a critical hydrostatic stress criterion, related to the embrittlement of,
in this case, polycarbonate. In contrast to yield and evolution of yield
[33], embrittlement is molecular weight dependent [11,12,23,25].
Therefore, any criterion should also take this into account. Once
identified, a proper criterion would provide a valuable tool in
polymer design and engineering, since it can be used as a safety limit
to ensure at least a ductile deformation mode, once the product fails.
2. Experimental
2.1. Materials
Materials used are injection molding grades of polycarbonate
(PC): Lexan 121R, 141R and 101R, supplied as granules by Sabic
Innovative Plastics (Bergen op Zoom, the Netherlands). The molec-
ular weights, and molecular weight distributions, of the grades are
listed in Table 1: PC-LM (121R), PC-MM (141R) and PC-HM (101R).
Following the analysis of Bersted and Anderson, which was
elegantly applied by Tervoort et al. to relate wear in polymers
directly to the relevant molecular parameters, Table 1 also lists the
number-average molecular weight corrected for the fraction of
short chains that do not contribute to the connectivity in the
material and act as a diluent, multiplied by the volume fraction of
*
‘effective’ polymer, fMn [35,36].
Tensile bars are injection molded according to ISO 527 (cross-
sectional area 4  10 mm2) using an Arburg 320S 500-150 All-
rounder. Processing conditions are kept the same for all grades.
Notched specimens are obtained by machining the tensile bars at
room temperature. The radius of the notch is 0.4 mm, the included
angle 35 , and the depth is 4 mm, resulting in a cross-sectional area
of the notched tensile bars of 4  6 mm2. Annealing treatments are
performed using hot air circulating ovens at various temperatures.
Prior to testing, the samples are allowed to slowly cool to room
temperature for at least 15 min.
2.2. Methods
Tensile tests are performed on a Zwick Z010 universal tensile
tester at ambient temperature (23  C). Experiments are performed
Table 1
Polycarbonate molecular weights.
*
grade M n ½kg=mol M w ½kg=mol PDI fMn ½kg=mol
PC-LM Lexan 121R 9.8 23.4 2.38 13.6
PC-MM Lexan 141R 9.2 25.8 2.82 14.9
PC-HM Lexan 101R 16.1 30.7 1.91 16.6
by applying constant engineering strain rates or stresses. Unless
indicated otherwise, a standard constant rate of 103 s1 is used. In
case of applied constant stresses, the time-to-failure is taken as the
time-to-onset of localization. Impact tests are performed on a servo-
hydraulic MTS 831 Elastomer Testing System at ambient tempera-
ture (23  C) and a constant displacement rate of 100 mm s1. 1A
notched tensile test bar geometry is chosen, rather than performing
an Izod test, since tensile testing allows recording of the complete
load-displacement curves. The notch radius used (0.4 mm) is chosen
somewhat larger than those normally used in Izod impact tests (ISO
108A: 0.25 mm) to facilitate the numerical investigations (smaller
notch radii require extremely more refined meshes, thus severely
more computational effort). All tensile yield stresses listed in the
Results Section are engineering yield stresses. The impact results are
taken as the mean value of 5 experiments; 68.3% confidence inter-
vals are calculated.
3. Numerical
Numerical simulations are performed using the EGP-model,
implemented as a user-subroutine in MSC.Marc [34,37]. The model
decomposes the Cauchy stress into a hardening stress and a driving
stress, as originally proposed in the pioneering work of Haward and
Thackray [38]. The driving stress is modeled as a combination of
multiple parallel linked Maxwell elements [34,37,39], whose
viscosities are described by an Eyring flow rule, modified to take
pressure dependence and intrinsic strain softening into account
[33,40]. The viscosities are defined with respect to a rejuvenated
reference state [33,41,42] which is the “virgin” state where all prior
thermal and mechanical history is erased. The current state is
subsequently defined by the evolution in time of the state param-
eter, Sa, of which a detailed definition can be found in [33]. The
current yield stress is directly proportional to the present value of
Sa. The model is capable to quantitative describe various defor-
mation modes, once the initial thermodynamic state of a material,
Sa, is known [4,33,34]. Its value is predicted based on processing
history [43,44] but if that is unknown, it can be obtained by a single
mechanical test. The model even accounts for aging during the
experimental investigation [4]. Parameters used here are adopted
from [34,37].
4. Experimental results
4.1. Deformation kinetics and yield-stress evolution
Fig. 1 displays the deformation kinetics of the three different
molecular weight polycarbonates, showing the yield stress as
a function of the strain rate applied (left), and the time-to-failure as
a function of the stress applied (right). The slopes of the curves
reveal the kinetics that is identical for all three polycarbonates,
illustrating that kinetics is independent of molecular weight. Not
only the slopes, but also the absolute values of the yield stresses are
identical, as a result of the fact that all samples of all molecular
weights received exactly the same thermal history during pro-
cessing (Differences in thermal history are expressed in an increase
in the absolute level of the yield stresses, whereas the slopes stay
unaffected [33]).
1
At high deformation/strain rates, the secondary relaxation process in PC
contributes to the stress. The deformation rate was chosen such that the maximum
strain rates in the specimen are not in this high strain rate regime. Typically the
secondary process starts to contribute at strain rates of 10 0e101 s1 for PC, and the
transition to the high strain rate regime occurs gradually over a number of decades.
The macroscopic strain rate chosen here is 100 s1, which will be amplified under
the notch by a factor of w6, which is still below the high strain rate regime.
 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819
Fig. 1. Deformation kinetics of polycarbonate. Left: yield stress versus strain rate. Right: Time-to-failure versus applied stress.
Results of the evolution of yield stress with time and temperature
for all molecular weights are presented in Fig. 2. Yield stresses
increase with time for all temperatures investigated, and no level-
ing off is observed. Therefore no equilibrium is reached, as might
happen at high temperatures, just below Tg, for long annealing times
[44,45].
The master curve on the right-hand-side is constructed by Time-
Temperature-Superposition (TTS), under the assumption of thermo-
rheological simple behavior, using an Arrhenius type of temperature
dependence resulting in a shift factor, aT, given by [33]:
!!
DU 1 1
aT ðTÞ ¼ exp  (1)
R T Tref
where DU is the activation energy, R the universal gas constant,
T the annealing temperature, and Tref a reference temperature (here
taken as room temperature: 23  C). The yield stress evolution with
time is described by Equation (2), using the parameters from
Table 2, and is shown as the solid drawn line in Fig. 2 (right):
 
teff ðt; TÞ þ ta
sy ðt; TÞ ¼ sy;0 þ c$log (2)
t0
where the two constants sy,0 and c give the intersection of the curve
with the vertical axis at t ¼ 1 s, and the slope of yield stress with
(the logarithm of) time in the linear relation, respectively, while
ta is the initial age of the material. The scaling constant t0 ¼ 1 s, and
the effective time, teff , is defined as:
Fig. 2. Yield stress versus annealing time for different annealing temperatures, and the master curve constructed thereof. For dashed lines, see text below.
1813
Zt
teff ðt; TÞ ¼ a1 0
T ðTðt ÞÞdt
0
(3)
0
with aT the shift function, defined in Equation (1).
Master curves for all three molecular weights coincide within
experimental error. Therefore also the evolution of the yield stress
is, as expected, molecular weight independent. The activation
energy related to the kinetics of aging (DU ¼ 205 kJ/mol), corre-
sponds well to the value found in a previous publication [33] that
uses another molecular weight polycarbonate.
4.2. Embrittlement upon aging
With the increase in yield stress embrittlement occurs as
a result of an enhanced tendency to strain localization [9]. This is
illustrated in Fig. 3 showing the notched impact energy versus
annealing time for the three molecular weights used (LM, MM and
HM) at five different annealing temperatures, 90e130  C, with their
master curves shown at the right-hand-side of the figure. Embrit-
tlement occurs faster with increasing annealing time and temper-
ature, and occurs slower for higher molecular weights, which is in
agreement with the observations of LeGrand [1]. Remarkably, if
Time-Temperature-Superposition (TTS) is applied, using the same
activation energy as found for the yield stress evolution of Fig. 2,
(205 kJ/mol), smooth master curves can be constructed, see Fig. 3
(right). This indicates that the evolution of the yield stress is the
determining factor for the observed evolution of embrittlement.
1814 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819

Table 2 Table 3
Master curve parameters. Master curve parameters.

sy,0 [MPa] c [MPa] ta [s] PC-LM PC-MM PC-HM

PCa 22.0 3.60
a sy,ca [MPa]
Taken from [33]; corrected for strain rate and engineering stress.
The value of the activation energy found in this study is somewhat
higher than the one found by LeGrand (which was: 150 kJ/mol).
The vertical dashed lines in Fig. 3 indicate the time at which half
of the initial impact energy for each material remains, and is taken as
the representative time-to-embrittlement at room temperature, and
it is different for the three different molecular weights, see Table 3.
Where the aging kinetics, that causes the yield stress to increase,
is not dependent on molecular weight, see Fig. 2, the time-to-
embrittlement is, see Fig. 3. The corresponding critical value of the
yield stress, sy,c, increases with molecular weight, see the dashed
lines in Fig. 2, that start at the times-to-embrittlement for the three
polymers, and the values in Table 3.
It can, moreover, be concluded from Fig. 3 that not only the
initial value of the notched impact energy increases with molecular
weight, but also the final value in impact energy (after embrittle-
ment caused by aging), with even a factor two for the highest
molecular weight material.
5. Numerical results
5.1. Investigation of notch sensitivity
In the experiments we concluded that the change in failure
mode is initiated at a critical value of the yield stress, and the cause
5.0$1010 tembrittlement [1012 s] 2.4 4.2 18.0
66.1 67.0 69.1
Sa [e] 35.8 36.7 38.6
a
Eng. yield stress at strain rate 103s1.
of this is investigated now by performing detailed numerical
analyses. Rather than using an aging dependent yield stress, we use
the state parameter Sa as input for the simulations, since that
parameter is, in contrast to the yield stress, independent of loading
conditions like strain rate, temperature and notch geometry. The
state parameter can be directly determined from the yield stress
when taking into account the relevant testing conditions, see the
appendix of Klompen et al. [33].
First the results of simulations based on an arbitrarily chosen
value of Sa ¼ 40 (corresponding to a tensile engineering yield stress
of 70 MPa at a strain rate of 103 s1) are discussed, see Fig. 4. The
force increases with displacement until a maximum is reached. At
this point we find plastic deformation over the complete cross-
section of the test bar. A detailed analysis of all numerical data
reveals the value and position of the maximum in hydrostatic stress
somewhere under the notch in the sample (The hydrostatic stresses
are stored for each integration point per displacement increment).
A maximum in hydrostatic stress (top-right) is found earlier than
the maximum in force occurs (top-left). Before this maximum
(stages 1e3), plastic deformation is restricted to a small area
directly underneath the notch; after this maximum, spreading of
plastic deformation over the cross-section of the bar starts. The

Fig. 3. Notched impact energy versus annealing time, for different annealing temperatures and the master curves constructed thereof, for all three molecular weights used, from top
to bottom PC-LM, PC-MM and PC-HM. In shifting of the curves the same activation energy is used as in Fig. 2.
 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819
Fig. 4. Top left: Force versus displacement for a notched tensile bar. Top right: Maximum hydrostatic stress underneath the notch (dashed line represents the hydrostatic stress
close to the surface of the notch.) Bottom left: Initial geometry of the notched bar, and, right, after deformation (gray-scale indicates plastic deformation). Note: because of
symmetry only a quarter of the actual geometry is used.
maximum value of the hydrostatic stress starts close to the surface
of the notch; then moves inwards, to finally return toward the
surface again at large deformations (indicated by the dashed line in
Fig. 4).
Fig. 5 shows the details of plastic deformation (left) and
hydrostatic stress (right) at each loading step, indicated with the
corresponding numbers in Fig. 4. The maximum hydrostatic stress
builds up close to the surface underneath the notch until plastic
deformation starts to evolve (stages 1e2). The area of maximum
hydrostatic stress is then pushed in front of the plastic deformation
zone, moving toward a position away from the notch (stages 3e5).
When plastic deformation starts to evolve over the whole cross-
section of the bar, the maximum returns to the surface of the notch
(6e8). The results agree well, at least qualitatively, with those of
other numerical investigations [19,21]. Since the constitutive model
does not incorporate a failure criterion, deformation in the simu-
lation continues indefinitely. Experimentally, however, complete
failure of the notched bar would long be observed.
Next we investigate the influence of physical aging. To do so we
change the thermomechanical state of the material by choosing
different values of Sa. Varying its value over a large range, from
Sa ¼ 20, 30, 40 to 50 (corresponding to yield stresses of about
sy z 50, 60, 70 and 80 MPa, respectively), we find the characteristic
force-displacement curves and the evolution of the maximum
hydrostatic stress, see Fig. 6, both significantly increasing with
increasing Sa.
Upon increasing Sa, the plastic deformation zone is stronger
localized for longer loading stages. This causes an increase in
hydrostatic stress, with plateau values ranging from sh,max z 67 MPa
(Sa ¼ 20) to sh,max z 100 MPa (Sa ¼ 50). During prolonged aging
the state parameter Sa and the yield stress sy increase, as well as the
hydrostatic stress under the notch that eventually surpasses the
critical value for PC, triggering a change in failure mode.
1815
6. Combining experimental and numerical results
6.1. A critical hydrostatic stress criterion
We try to find a hydrostatic stress criterion that predicts the
embrittlement of notched samples and first start to investigate the
macroscopic responses. Fig. 7 (left) shows stress-strain results of
unnotched tensile bars that neck, and Fig. 7 (right) gives the force
versus displacement curves for notched tensile bars (for clarity the
annealed curve is shifted horizontally). The as-received and the
annealed samples (1000 h at 120  C), give different Sa values
(independently measured as Sa ¼ 29 and Sa ¼ 49, respectively) that
rather accurately describe all stress-strain and force-displacement
responses, including necking. In the notched tests, the as-received
sample fails in a ductile mode, whereas the annealed sample fails in
a brittle manner. Of course the simulations continue, where
experimentally samples fail, since no failure criterion is imple-
mented yet.
Next we perform local analyses at the time-to-embrittlement of
the three molecular weights used. Their Sa values differ, because
their time-to-embrittlement is different, see Fig. 3. Since the higher
molecular weight polymers fail later at longer timescales, higher
values for Sa (and therefore also sy) result, see Table 3 and Fig. 10.
With these three Sa values for the three polymers, we simulate the
tensile tests at the time-of-embrittlement on the notched spec-
imen, Fig. 8.
Apart from the force versus displacement curves (solid lines, left
vertical axis), also the maximum hydrostatic stress, sh, versus
displacement curves (dashed lines, right vertical axis) are shown. We
know that the maximum in hydrostatic stress increases with Sa, see
Fig. 6. From Fig. 8 we learn that the maximum, indicated here as the
critical stress, shc , therefore also increases with molecular weight,
although the effect is not that large: shc ¼ 86:5; 87:6 and 90:0 Mpa,
1816 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819

Fig. 5. Left: Evolution of the plastic deformation underneath the notch. Right: Evolution of the hydrostatic stress underneath the notch. Numbers correspond to the positions as
indicated in Fig. 4 (planes in view are sectional planes based on symmetry).

for fMn* ¼ 13:6; 14:9 and 16:6 ½kg=mol, respectively, see Fig. 8,2
and 10, and using the definition of the corrected molecular weights
of Table 1. These values compare well with those mentioned in
literature [19e21], which are between 80 < shc < 95 MPa.
2
We did not push mesh refinement to the limit of full convergence, i.e. no
further change in calculated values upon further mesh refinement. Reason for this
is that the 3D computations, required to evaluate the evolution of local stress states,
are very cost intensive. Our expectation is that further mesh refinement would lead
to only a minor increase in the critical hydrostatic stresses found.
From the predicted force-displacement curves, impact energies
can be calculated using the maximum in hydrostatic stress as
criterion to determine the point of failure where the computations
are stopped, see Table 4. The computed values underestimate the
experimentally determined ones with 13%, 40% and 55% for PC-LM,
PC-MM and PC-HM, respectively.
The markers (+) on the load-displacement curves in Fig. 8
indicate where experimentally the maximum in force is observed
and failure occurs. Only for the lowest molecular weight this point
coincides with the position of shc . For the higher molecular weights,
macroscopic failure occurs at a force and displacement that is
 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819 1817

Fig. 6. Influence of physical aging on the response of a notched tensile bar. Left: Force versus displacement. Right: Maximum hydrostatic stress versus displacement (note: Sa is
a measure for the thermodynamic state of the material and larger values indicate an increase in age).

Fig. 7. Experimental versus numerical results for as-received samples and annealed samples. Left: Stress versus strain for unnotched bars. Right: Load versus displacement for
notched bars.

increasingly larger than the force and displacement that corre-
spond to the position of the maximum in hydrostatic stress. This is
caused by that after crack initiation by voiding, craze extension and
crack propagation are more stabilized by higher molecular weights.
The experimental images of Fig. 9 show signs of this extra absorbed
energy, and reveal different shapes and sizes of the crazes for-
med underneath the notch [19], that eventually evolve into
a propagating crack in the middle of the test bar. For PC-LM, a craze
is initiated just below the surface of the notch, that almost imme-
diately transforms into an unstable propagating crack, witnessed
by the rough fracture surface. For the PC-MM and PC-HM extended,
stable, crazes are observed before unstable propagation of the crack
occurs through these crazed surfaces [46,47]. For PC-MM the crazed
surface covers one third of the fracture surface, adjacent to the

Fig. 8. Calculated force and maximum hydrostatic stress versus displacement for three molecular weights at their respective times-to-embrittlement.
1818 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819

Fig. 9. Images of fracture surfaces: crack propagates from top to bottom.

6.2. A critical yield stress criterion

Above we demonstrated that initiation of failure can be pre-

Fig. 10. Molecular weight dependence of the critical hydrostatic (,) and yield stress
(B).
initiation region, whereas for PC-HM it covers almost the entire
fracture surface (and only at the very last can a small truly brittle
fracture surface be observed). Since a craze is load bearing, also in
polycarbonate [48], the experimental force can still increase after
craze initiation at a critical hydrostatic stress, see the differences
in force, Fb, at maximum shc and those at final fracture (+) observed
in Fig. 8 for the PC-MM and PC-HM, and the differences found in
impact energies in Fig. 3 and Table 4, which is in complete corre-
spondence to the results in [19].
A molecular weight dependent critical value of the hydrostatic
stress, shc , is a useful indicator for the initiation of failure in poly-
mers differing in molecular weight, and it properly describes the
experimental time-to-embrittlement observed in specimen con-
taining a well defined notch. In predicting the impact energy at this
point of failure it behaves poor for all but the lowest molecular
weight polymer, since the propagation of crazes and cracks up to
final failure are absent in the model and should be included.
Table 4
Impact energies calculated from predicted and experimental force-displacement
curves.
Impact energy [kJ/m2]
dicted via a numerical analysis of the stresses developing under
a notch. This procedure can in principle be applied to any loading
geometry and temperature history. However, this always requires
a detailed FEM simulation with exact geometrical details (size,
depth, radius of notch etc.). An alternative failure criterion is
a molecular weight dependent critical value of the evolving yield
stress, sy,c. This is demonstrated by the data in Figs. 2 and 3, and
motivated in Fig. 6 which shows that the hydrostatic stress under
the notch, sh, increases with the state parameter Sa, which is
directly proportional to the yield stress, sy.
The values of the critical molecular weight dependent yield
stress, sy,c, are derived in Fig. 2 (and are based on the time-to-
embrittlement data of Fig. 3 and Table 3) and in Fig. 10 they are
compared to the molecular weight dependent critical hydrostatic
stress, shc , using again the definition of the corrected molecular
weights of Table 1.
The time at which the critical yield stress is reached can be
determined for any use-temperature and load, using Equation
(1e3). Load can be taken into account, since Klompen et al. [33]
showed that, similar to temperature annealing, also stress anneal-
ing is quantitatively captured by the EGP-model.
Alternatively the maximum allowable Continuous Use
Temperature (CUT) can be defined for a desired life-time (tCUT), by
combining Equation (1e3):
 !1
1 R t0 ta
CUT ¼  ln 10ððsy;c sy;0 Þ=cÞ  (4)
Tref DU tCUT tCUT
As an example, if a life-time of 50 years is required, we can plot
a vertical line in Fig. 2 at w109 s and apply TTS to find the maximum
allowable application temperatures, that range from 49  C for the
lowest molecular weight, to 57  C for the highest molecular weight
PC. Like in the case of the critical hydrostatic stress, these limits do
not precisely predict actual embrittlement by predicting correct
impact energies, apart from applications to relatively low molecular
weight polymers, but they rather tell you when a product will fail
brittle.

Calulated at the Measured at Error 7. Conclusions

position of ðshc Þ time-to-embrittlement %
PC-LM 42.7 48.2 13 The embrittlement of notched polycarbonate samples upon aging
PC-MM 43.7 61.5 40
PC-HM 45.2 70.1 55
is investigated. In unnotched bars the evolution of the yield stress
with annealing time and temperature is found to be independent of
 T.A.P. Engels et al. / Polymer 52 (2011) 1811e1819 1819

the molecular weight. In contrast, in notched bars both the impact and properties of polymers, volume 12. New York: VCH Publishers Inc.; 1993.
p. 699e765.
energy and time-to-embrittlement depend on the molecular weight.
[18] van Melick HGH, Bressers OFJT, den Toonder JMJ, Govaert LE, Meijer HEH.
Evolution of the yield stress in unnotched samples and that of the Polymer 2003;44:2481e91.
impact energy in notched samples can be shifted to master curves, [19] Ishikawa M, Narisawa I, Ogawa H. Journal of Polymer Science: Polymer
using the same activation energy. This indicates that the yield stress Physics Edition 1977;15:1791e804.
[20] Kambour RP, Farraye EA. Polymer Communications 1984;25(12):357e60.
determines the transition to embrittlement. A molecular weight [21] Nimmer RP, Woods JT. Polymer Engineering and Science 1992;32
dependent critical hydrostatic stress is found to be a good criterion to (16):1126e37.
predict via the onset of voiding, the initiation of failure, but it appears [22] Hill R. The mathematical theory of plasticity. Oxford: Oxford University Press;
1950.
to be insufficient for a quantitative description of embrittlement, [23] Golden JH, Hammant BL, Hazell EA. Journal of Applied Polymer Science
mainly since craze widening and crack propagation after cavitation 1968;12:557e69.
contribute (up to in the order of 50%) to the load bearing capacity, [24] Gaymans RJ, Hamberg MJJ, Inberg JPF. Polymer Engineering and Science
2000;40(1):256e62.
when high molecular weight polymers are used. An alternative [25] Golden JH, Hazell EA. Journal of Polymer Science: Part A 1963;1:1671e86.
failure criterion is a critical molecular weight dependent value of the [26] Golden JH, Hammant BL, Hazell EA. Journal of Applied Polymer Science
developing yield stress. Application of this criterion does not require 1967;11:1571e9.
[27] Allen G, Morley DCW, Williams T. Journal of Materials Science 1973;8:1449e52.
extensive numerical modeling and nevertheless indicates when [28] Broutman LJ, Krishnakumar SM. Polymer Engineering and Science 1976;16
a material with a given age, used at a given temperature, will fail in (2):74e81.
a brittle manner would it be scratched or tested under notched [29] Cheng TW, Keskkula H, Paul DR. Journal of Applied Polymer Science
1992;45:531e51.
impact conditions.
[30] Plati E, Williams JG. Polymer 1975;16:915e20.
[31] Jacques CHM, Wyzgoski MG. Journal of Applied Polymer Science 1979;
23:1153e66.
References [32] Arnold JC. Journal of Materials Science 1995;30:655e60.
[33] Klompen ETJ, Engels TAP, Govaert LE, Meijer HEH. Macromolecules 2005;38
[1] LeGrand DG. Journal of Applied Polymer Science 1969;13:2129e47. (16):6997e7008.
[2] Struik LCE. Physical aging of amorphous polymers and other materials. [34] van Breemen LCA, Engels TAP, Pelletier CGN, Govaert LE, den Toonder JMJ.
Amsterdam: Elsevier; 1978. Philosophical Magazine 2009;89(8):677e96.
[3] Hutchinson JM. Progress in Polymer Science 1995;20:703e60. [35] Bersted BH, Anderson TG. Journal of Applied Polymer Science 1990;39:
[4] Klompen ETJ, Engels TAP, van Breemen LCA, Schreurs PJG, Govaert LE, 499e514.
Meijer HEH. Macromolecules 2005;38(16):7009e17. [36] Tervoort TA, Visjager J, Smith P. Macromolecules 2002;35:8467e71.
[5] Janssen RPM, de Kanter D, Govaert LE, Meijer HEH. Macromolecules [37] van Breemen LCA. Contact mechanics in glassy polymers. Ph.D. thesis. Eind-
2008;41:2520e30. hoven, The Netherlands: Eindhoven University of Technology; 2009.
[6] Narisawa I, Ishikawa M, Ogawa H. Journal of Polymer Science: Polymer [38] Haward RN, Thackray G. Proceedings of the Royal Society of London A
Physics Edition 1978;16:1459e70. 1967;302(1471):453e72.
[7] Crissman JM, McKenna GB. Journal of Polymer Science: Part B: Polymer [39] Tervoort TA, Klompen ETJ. Journal of Rheology 1996;40(5):779e97.
Physics 1987;25:1667e77. [40] Govaert LE, Timmermans PHM, Brekelmans WAM. Journal of Engineering
[8] Crissman JM, McKenna GB. Journal of Polymer Science: Part B: Polymer Materials and Technology 2000;122(2):177e85. Transactions of the ASME.
Physics 1990;28:1463e73. [41] Hasan OA, Boyce MC, Li XS, Berko S. Journal of Polymer SCience: Part B:
[9] van Melick HGH, Govaert LE, Meijer HEH. Polymer 2003;44:3579e91. Polymer Physics 1993;31:185e97.
[10] Adam GA, Cross A, Haward RN. Journal of Materials Science 1975;10:1582e90. [42] van Melick HGH, Govaert LE, Raas B, Nauta WJ, Meijer HEH. Polymer
[11] Ryan JT. Polymer Engineering and Science 1978;18(4):264e7. 2003;44:1171e9.
[12] Pitman GL, Ward IM, Duckett RA. Journal of Materials Science 1978;13:2092e104. [43] Govaert LE, Engels TAP, Klompen ETJ, Peters GWM, Meijer HEH. International
[13] Bubeck RA, Bales SE, Lee HD. Polymer Engineering and Science 1984;24 Polymer Processing 2005;(2):170e7.
(10):1142e8. [44] Engels TAP, Govaert LE, Peters GWM, Meijer HEH. Journal of Polymer Science:
[14] Takano M, Nielsen LE. Journal of Applied Polymer Science 1976;20:2193e207. Part B: Polymer Physics 2006;44(8):1212e25.
[15] Fraser RAW, Ward IM. Journal of Materials Science 1977;12:459e68. [45] Bauwens-Crowet C, Bauwens JC. Polymer 1985;27:709e13.
[16] Kuriyama Y, Narisawa I. Annual technical conference, 3. Society of Plastics [46] Hull D, Owen TW. Journal of Polymer Science: Polymer Physics Edition
Engineers; 1994. 3235e3237. 1973;11:2039e55.
[17] Narisawa I, Yee AF. Materials science and technology: a comprehensive [47] Agrawal CM, Pearsall GW. Journal of Materials Science 1991;26:1919e30.
treatment [chapter 15]. In: Cahn RW, Haasen P, Kramer EJ, editors. Structure [48] Kambour RP, Kopp RW. Journal of Polymer Science: Part A-2 1969;7:183e200.