Polymer Degradation and Stability 109 (2014) 319e326

Contents lists available at ScienceDirect

Polymer Degradation and Stability

journal homepage: www.elsevier.com/locate/polydegstab

Characterizing the toughness of an epoxy resin after wet aging using

compact tension specimens with non-uniform moisture content
Gustavo Quino, Jalal El Yagoubi, Gilles Lubineau*
King Abdullah University of Science and Technology (KAUST), Division of Physical Sciences and Engineering, COHMAS Laboratory, Thuwal 23955-6900,
Saudi Arabia

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

Article history: Characterizing the change in toughness of polymers subjected to wet aging is challenging because of the
Received 13 May 2014 heterogeneity of the testing samples. Indeed, as wet aging is guided by a diffusion/reaction process,
Received in revised form compact tension samples (defined by the ASTM D5045 standard), which are relevant for toughness
18 July 2014
characterization but are somewhat thick, display a non-uniform moisture content over the bulk material.
Accepted 2 August 2014
Available online 13 August 2014
We define here a rigorous procedure to extract meaningful data from such tests. Our results showed that
the relation between the moisture uptake of the whole sample and the measured toughness was not a
meaningful material property. In fact, we found that the measured toughness depended on the locally
Keywords:
Epoxy
varying moisture uptake over the cracking path. Here, we propose a post-processing technique that relies
Toughness on a validated reaction/diffusion model to predict the three-dimensional moisture state of the epoxy.
Wet aging This makes identification of the variation in toughness with respect to the local moisture content
CT specimen possible. In addition, we analyze the fracture surface using micrography and roughness measurements.
Roughness The observed variations in toughness are correlated with the roughness in the vicinity of the crack tip.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction 1.2 MPa.m0.5 to 0.9 MPa.m0.5 after 1500 h at 70
C. Alessi et al. [7]
also studied an epoxy/anhydride system and observed a 62%
Composite materials used in advanced aeronautics and space ap- reduction in KIc after a week in distilled water at 70
C. Although the
plications are exposed to severe environmental conditions. One of the ASTM D5045 standard method produces useful results, a problem
most common environmental conditions that leads to degradation of in using it to identify toughness after hygrothermal aging is that the
such materials is hygrothermal aging, which takes place when the size of the samples, as defined by the standard, is massive. The
material is exposed to a wet environment at moderate or elevated moisture state is non-uniform in large samples, making it difficult
temperatures. Interactions between the water in the environment to draw conclusions about changes in the fracture toughness with
and the polymer backbone of the resin can result in various modifi- respect to the moisture content.
cations of the resin's properties [1e5]. Here, we focus on changes in To obtain more representative moisture/toughness relations
the fracture properties of the epoxy resin, and more specifically on its that could be used in three-dimensional (3D) simulations of
toughness, when it is exposed to a wet environment. moisture-induced failure, other sample configurations can be used,
Various experimental techniques can be used to characterize such as those introduced for the essential work of fracture (EWF)
changes in the toughness of epoxy resins. The ASTM D5045 stan- method. EWF uses double edge notched tension (DENT) thin
dard, which relies on compact tension (CT) testing, is a universally specimens in which the water content is much more uniform than
accepted method to measure fracture toughness. It has been used in CT samples. EWF has been used to measure plane stress fracture
previously in the study of wet aging. Alessi et al. [6] used CT testing toughness of aged polymers in previous studies [8,9] and is based
to study the influence of hydrothermal aging on the fracture on the assumption that dissipation occurs either by plastic defor-
toughness of radiation-cured bis(4-glycidyloxyphenyl) methane mation or by the fracture process. To distinguish between the two
(DGEBF) with polyethersulfone (PES) as the toughening agent. In processes, several configurations with different neck lengths are
their study, the critical stress intensity factor (KIc) decreased from needed. This method works well on ductile polymers, but it cannot
be applied to brittle epoxy resins, since it requires full yielding of
the neck before fracture. The samples considered in this study,
* Corresponding author. Tel.: þ966 1 2 8082983. based on anhydride cured epoxy, display brittle failure, which
E-mail address: [email protected] (G. Lubineau).

http://dx.doi.org/10.1016/j.polymdegradstab.2014.08.005
0141-3910/© 2014 Elsevier Ltd. All rights reserved.
320 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326

makes it difficult to utilize the EWF method. Chale
at et al. [8] made previously characterized [11e13]. Its dry elastic mechanical prop-
use of both ASTM D5045 and EWF techniques to measure the erties (Young's modulus and Poisson ratio) are E ¼ 3.1 ± 0.1 GPa and
fracture toughness of polymer samples that were in a brittle-to- n ¼ 0.3. We fabricated 48 CT samples (Fig. 1) by casting the liquid
ductile transition due to hygrothermal aging. In that study, EWF epoxy system in silicone rubber molds previously coated with an
was used to characterize the ductile samples, while ASTM D5045 anti-stick agent (MANN EasyRelease). The curing cycle was as rec-
was used to characterize the brittle ones. Barany et al. [9] used EWF ommended by the manufacturer: 6 h at 80
C and postcuring at
to study the influence of hygrothermal aging on the fracture 180
C for 8 h. Afterwards, the samples were sanded and polished
toughness of polyester sheets although the EWF method is not as with different grit papers (320, 500, 1000, 2000) until the desired
robust as the ASTM standard as reported by Clutton [10]. Although thickness was obtained. Finally, pre-cracks were introduced by
they were able to reduce some of the scattering in round robin test sliding a fresh razor blade into the root of notches to obtain sharp
results, they also observed that several major uncertainties are initial crack tips (details in Fig. 1).
inherent in the method.
Since the EWF method is not as robust as ASTM D5045, and
2.2. Aging conditions
cannot be used on our brittle samples, we chose to use the ASTM
D5045 method, but with full awareness of the non-uniformity of
We placed the samples in a climatic chamber (Tenney T2RC,
the moisture state. Our objective was to find a viable post-
Thermal Product Solutions) at a constant temperature of 70
C and
processing procedure to account for this problem.
a constant relative humidity of 90% R.H. Periodically, we extracted
Here, we used experimental results as well as an accurate
three samples from the chamber to perform fracture tests for that
simulation of the heterogeneous moisture content concurrently.
specific aging time. We made 15 extractions at intervals of 12 h
The CT samples were subjected to hygrothermal aging inside a
(extractions 1e6), 24 h (extractions 7, 8), and 48 h (extractions
climatic chamber and tested at various aging states. After tough-
9e15). The baseline properties before aging were obtained sepa-
ness testing, we analyzed the fracture surfaces of the samples by
rately by performing the fracture test on non-aged dry samples.
micrography and roughness measurements. To determine the wa-
ter distribution across the samples, we performed simulations
based on a validated diffusion-reaction model of water absorption. 2.3. Characterization
Understanding of the spatial distribution of the water content made
it possible to identify the relation between the observed toughness For each extraction, we characterized the toughness, the global
and the water content at the point where the cracks started to moisture uptake of the CT samples and the surface morphology of
propagate in an unstable manner. the fracture surface.
In the following section, we describe the material, samples, and The CT samples were tested following the ASTM D5045 [14] stan-
experimental method related to both fracture toughness identifi- dard for measuring plane strain fracture toughness in polymers. We
cation and surface analysis. In Section 3, we provide details on the utilized a crosshead speed of 0.3 mm/min (INSTRON 5882 Universal
simulations of water absorption and crack extension. In the fourth Testing Machine). We calculated KIc as a function of the geometrical
section, we present the experimental and simulation results. parameters and the maximum load (PQ) during the test (Eq. (1)):
Finally, we develop two main points in the discussion: (i) the ne-
cessity of considering a local approach for the correct identification PQ
KIc ¼ f ðxÞ; (1)
of intrinsic material parameters, and (ii) the correlations between BW 1=2
the toughness, roughness, and micrography.
where W ¼ 30 mm, B ¼ 7.5 mm and f(x) is defined by the standard
as:
2. Materials and methods
 
2þx
f ðxÞ ¼ 0:886 þ 4:64x  13:32x2 þ 14:72x3  5:6x4 ;
2.1. Material formulation and samples 1  x3=2
(2)
The material used in this study was EPOLAM 2063 (Axson
Technologies), a mixture of cycloaliphatic epoxy resin and a x is the ratio ao/W, and ao ¼ 15 mm is the initial distance between
diglycidyl ether of bisphenol-A (DGEBA). This formulation was the crack front and the pins (Fig. 1). We conducted the toughness

Fig. 1. Geometrical specifications of the standard CT sample used in the study and details of the initial crack tip.
 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326 321

evaluation on three samples for each extraction under lab condi-  Conservation equations:
tions (22 ± 1
C, 52 ± 5% R.H.).


We monitored the global moisture uptake by weighing the vw
samples before and after aging and averaging the measured prop- ¼ div j þ rw ; (3)
vt
erties at each extraction.
We analyzed the morphology of the fracture surfaces in two
vY
ways: (i) micrography with a Leica DM 2500M microscope, and (ii) ¼ rY ; (4)
roughness (Ra) measurement (Profilometer Veeco Dektak 150 Sty-
vt
lus). In both cases, the region close to the crack tip is analyzed
(Fig. 2). In the roughness analysis, the raw data are high-pass vR
¼ rR ; (5)
filtered with a cutoff, lc, of 1.25 mm. Ra is taken as the average of vt
the measure of three five-segment paths as shown in Fig. 2.
where w are the diffusing free water molecules, j is the mass flux, Y
denotes the product of a macroscopic reaction (R þ w ! Y) and R is
3. Modeling framework the reactive substrate. rw is the volumetric rate of water production
(which can be negative or positive) resulting from the reaction
To analyze the experimental results, we developed a numerical process.
model that accounts for possible non-homogeneous moisture
states. First, we implemented a diffusion/reaction model to provide  Constitutive equations (diffusion law):
the in-the-volume field of both free and bonded water based on our
previous work [11] (Section 3.1 below). Second, we calculated the

w
crack extension by assuming that the toughness is variable in space j ¼ DðYÞ Vw  Vws ðYÞ ; (6)
ws ðYÞ
depending on the local water content (Section 3.2 below). This is a
weakly coupled scheme in the sense that we do not recalculate the
moisture field after the initial propagation of the crack, which is DðYÞ ¼ D0 þ D1 ,Y; (7)
found to be close to unstable.
ws ðYÞ ¼ ðS0 þ S1 ,YÞae psat ; (8)
3.1. Modeling non-Fickian water absorption where ws is the maximum reachable local water concentration, S is
the local solubility, ae is the water activity in the environment, psat is
Based on our observation of the sorption/desorption behaviors the saturation water pressure and D is the diffusion coefficient.
of samples with various thicknesses [11], we found that the ab-
sorption process of this resin was strongly non-Fickian but that it  Constitutive equations (reaction kinetics):
could be efficiently modeled by a diffusion/reaction model. Then,
we could describe the water transport in the epoxy resin as a
rw ¼ kh ðTÞwR þ kr ðTÞY; (9)
competition between a diffusion process and a reactive one. We
found the aging mechanisms to be, in reality, multiple and complex
(the main reactive mechanisms being here the hydrolysis of a rR ¼ kh ðTÞwR þ kr ðTÞY; (10)
reactive substrate as was demonstrated by Fourier Transform
Infrared spectroscopy in Ref. [13]). Rather than trying to account for rY ¼ kh ðTÞwR  kr ðTÞY; (11)
each mechanism separately, our approach globalizes all of them in
a macroscopic diffusion/reaction scheme. It successfully reproduces where T is the temperature and kh and kr are rate constants.
two-stage sorption behaviors (sorption and desorption) in very This continuum model is then used for the specific configuration
different configurations [11]. A summary of the governing equa- of CT samples. The boundary conditions (Fig. 3) are detailed in
tions is given in Equations (3)e(11): Equations (12), (13): symmetry on boundaries Sy and Sz, and freely

Fig. 2. Targeted area for morphological analysis of the fracture surface.

322 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326

4. Increment the crack length to a new value aiþ1 (a goes from 15 to

30 mm) and repeat the process.

For the mechanical quasi-static FE model used in Step 3 (Fig. 4),

we assume a linear isotropic material model and plane strains. The
boundary conditions are defined in Equations (14)e(16) and Fig. 4
and are: symmetry on vU1, pure vertical displacement for point
Q1, and a vertical load, Pi, uniformly distributed on vU2.

u,n ¼ 0; cM2vU1 ; ux ¼ 0; for point Q1 ; (14)

s,n ¼ 0; cM2vU  fvU1 ∪vU2 g; (15)

Z
s,n ¼ pn; cM2vU2 such that s,n ¼ Pi y (16)
vU2

The resulting opening, ui, is defined by averaging the vertical

displacements on vU2.
Fig. 3. Boundary conditions for the moisture absorption simulation.

exposed to the environment on the rest of the boundaries. The 4. Results and discussion
aging conditions of the experimental campaign are 70
C and 90%
R.H. 4.1. Evolution of KIc with moisture content

j:n ¼ 0; cM2Sy ∪Sz ; (12) The “apparent” toughness of the material is strongly modified
   by hygrothermal aging (Fig. 5). The initial value of KIc
w ¼ Sae psat ; cM2 vU  Sy ∪Sz (13) (1.04 MPa.m0.5) for the dry material is consistent with classical
values reported in the literature for anhydride-cured epoxy resins
We implemented the model in COMSOL Multiphysics using [7]. Regarding the effect of moisture, Fig. 5 shows that KIc drops
tetrahedral elements. We solved it in a fully implicit manner using a quite rapidly with a 21% reduction after 48 h of aging (about 1% wt
staggered discretization scheme alternating between the diffusion of the global water uptake). For comparison, Alessi et al. [7] re-
and reaction equations. The constitutive parameters for this specific ported a KIc reduction up to 37% for an epoxy resin immersed in
resin were identified previously as: D0 ¼ 14  1012 m2 s1, distilled water at 70
C for 1 week (about 3% wt of global water
D1 ¼ 1.241014 m5 mol s1, S0 ¼ 5.7  102 mol m3 Pa1, uptake).
S1 ¼ 1.67  104 Pa1, R0 ¼ 2  103 mol m3, ae ¼ 0.9, psat ¼ 11.7 kPa, We show in Fig. 5(b) the evolution of KIc with respect to the
kh ¼ 8.53  1010 m3 mol1 s1, kr ¼ 5.6  106 s. The main output global water uptake. Evident is a first stage where a strong decrease
of this simulation is the 3D distribution of absorbed (both free e w in Ref. KIc is observed. A slow recovery of this quantity is then
e and bonded e Y e) water in the volume at a specific aging time, observed during a second stage. From these global observations,
making it possible to glean the history of water absorption in any the end of the first stage, which is defined by the point at which KIc
specified region. Two subregions (Fig. 3) were defined to monitor starts to recover, is reached at 0.8% wt of global water uptake. Yet,
the local water concentration: (i) a cylindrical subvolume of 0.5 mm this level of water content is not consistent with our previous ob-
radius with its axis in the initial crack tip (green (in web version) servations related to the evolution of the elastic modulus and the
volume, subregion 1), and (ii) a line parallel to the direction of the glass transition temperature and reported in Fig. 6(a) and (b) (see
growth of the crack in the middle of the fracture surface (red (in Ref. [12] for a comprehensive description of the testing campaign
web version) edge, subregion 2). from which these results have been extracted).
Results reported in Fig. 6(a) and (b) have been obtained on thin
3.2. Simulation of the crack propagation samples (1 mm) such that the water distribution is more homo-
geneous as compared to the CT samples. The first stage ended at
From the experimental results presented below in Section 4.1, 1.25% wt of global water uptake. The difference when compared to
we extract the relation between the local water uptake and the CT samples (1.25% wt versus 0.8% wt) can be attributed to the fact
local value of the critical stress intensity factor, KIc. Then, knowl-
edge of the 3D distribution of absorbed water (see Section 3.1) al-
lows us to define the 3D distribution of KIc that varies at each point.
For generating the simulated load/opening curves, we define a
simple and robust procedure that takes advantage of the standard
results for CT samples (Equation (1)):

1. Start with a crack length, ai.

2. Calculate the critical load, Pi, that triggers crack propagation for
this specific crack length using Eq. (1) (note that in Eq. (1),
KIc ¼ KIc(ai) is the local value for the studied location of the crack
tip).
3. Calculate the opening, ui, corresponding to Pi by running the
plane strain finite element computation defined below. A point
(ui, Pi) of the load/opening curve is then defined. Fig. 4. Mesh and boundary conditions for calculating the opening (step 3).
 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326 323

1.20 1.20
1.15 1.15 stage 1 stage 2
1.10 1.10
1.05 1.05

KIc (MPa.m 0.5)

KIc (MPa.m 0.5)

1.00 1.00
0.95 0.95
0.90 0.90
0.85 0.85
0.80 0.80
0.75 0.75
0.70 0.70
0 100 200 300 400 500 0.0 0.5 1.0 1.5 2.0 2.5
Time (hours) Global water uptake (%wt)

(a) (b)
Fig. 5. Evolution of KIc during aging (a): in terms of time; (b): in terms of the global water uptake (the error bars represent maximum and minimum values).

that CT samples display a very heterogeneous distribution of water 2.5

content. Then, the raw relation between KIc and the global water
uptake, as presented in Fig. 5, can not be considered as intrinsic to
2.0

Global Water uptake (%wt)

the material but should be understood as a structure-dependent
property. An important conclusion is that global water uptake is
poorly representative; we rather have to consider the water content 1.5
in the neighborhood of the crack tip. For this purpose, we look
again at the experimental results in light of the 3D numerical
simulations of the water absorption in the CT samples. 1.0
After complete aging (456 h), the samples display a global water
uptake of about 2% (Fig. 7). A first step is to validate that this change
0.5 Experimental
can be accurately captured by the diffusion/reaction model intro-
duced in Ref. [11] and briefly described in Section 3.1. Changes in Computational
the concentration fields (w, R and Y) were simulated over time. 0.0
Neglecting the swelling of the material, we derived the mass ratio 0 100 200 300 400 500
of the overall absorbed water from the average concentrations of Y Time (hours)
and w in U according to Equation (17):
Fig. 7. Measured and simulated global water uptake for the CT samples (the error bars
represent maximum and minimum values).
Z
MH2 O ðw þ YÞdV
Dm U Fig. 8 depicts the water content (including both bonded water, Y,
mnum
w ¼ ¼ ; (17)
m0 r0 V0 and diffused water, w) along the pre-defined path (subregion 2,
Fig. 3) for every extraction. It is observed that there is a strong
where r0 ¼ 1160 kg,m3 is the initial density of the epoxy and gradient in the water concentration close to the crack tip, especially
MH2 O ¼ 18$103 kg=mol is the molar mass of water. Results re- in the first extractions. Therefore, we extracted the change in water
ported in Fig. 7 confirm that the proposed model matches the content around the crack tip (subregion 1, Fig. 3) and plotted the
global gravimetric observation. evolution of KIc in relation to the local water uptake (Fig. 9), which

3.05 175

3.00 170
Young's modulus (GPa)

stage 1 stage 2
165
2.95
160
T g (°C)

2.90
155
2.85
150
2.80 145

2.75 140
0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Water uptake (%wt) Water uptake (%wt)

(a) (b)
Fig. 6. Evolution of the elastic modulus (a) and of the glass transition temperature (b) during wet aging (previous study [12]).
324 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326

1.8
1.6
1.4

Water uptake (%wt)

1.2
1.0
diffusing water (w)
0.8
bonded water (Y)
0.6
0.4
0.2
0.0
0 100 200 300 400 500
Fig. 8. The total water profile of subregion 2 at every extraction. Position 0 refers to the
crack tip. Time (hours)
Fig. 10. Change in diffused water (w) and bonded water (Y) in subregion 1 over time.
yields results that are consistent with the experimental data re-
ported in Fig. 6(a) and (b). A first induction stage can be clearly 4.2. Verification of the post-treatment procedure
identified for local water content ranging from 0 to 1.25% wt. This
first stage is characterized by the modification of KIc due to physical Since there exists a strong gradient of water uptake, it is to be
interactions between the diffused water and the polymer network. expected that KIc is changing along the crack path. The analysis in
Then, the change in KIc at higher water content could suggest a Section 4.1 relies on the assumption that the performed test is most
recovery process [15], although we proposed in Ref. [12] that likely unstable, even with a high moisture content and that it in
envisaging the effect of the heterogeneous distribution of water practice characterizes KIc close to the crack tip.
across the sample, as shown in Fig. 8, seems better. After this We verified this a posteriori using the simulation procedure
transient stage, the water distribution would become homoge- described in Section 3.2. By knowing the moisture-KIc relation
neous at the saturation level, and KIc would tend to a constant value (Fig. 9) and the moisture profiles along the cracking path (Fig. 8), it
during the equilibrium stage, as we reported in Ref. [12] for the was possible to map KIc along subregion 2. Then, assuming constant
elastic modulus and the glass transition temperature. KIc along the direction transverse to the crack propagation, quasi-
It is difficult, in this particular case, to distinguish the effect of static simulations were run to find the loadedisplacement curves
the reaction from the effect of the diffused water because the global of the heterogeneous samples following the procedure described in
process is diffusion-controlled. Indeed, the reaction characteristic Section 3.2. The loadedisplacement curves were obtained for
time (k ¼ 1=kh :R0 x105 s) is one order of magnitude smaller than samples with 0.51%, 0.98% and 2.03% global moisture content and
the diffusion characteristic time (tD ¼ L2 =D0 x106 s). The diffusion compared with the curve corresponding to an homogeneous
process here is the limiting phenomenon. We show in Fig. 10 the sample (0% moisture content) as shown in Fig. 11. Although from a
change in diffused water (w) and bonded water (Y) in subregion 1 theoretical point of view these curves characterize stable crack
over time. It is clear that the diffused water is the predominant extension under displacement conditions, the sharpness of the load
species during the first stage, leading to an important drop of the decrease when the crack starts to propagate indicates that these
fracture toughness by promoting the interchain motions. The up- tests are most likely unstable from an experimental point of view
take in bonded water during the second stage, mainly associated (in practice, a stable crack extension is very difficult to obtain in
with hydrolysis, does not lead to a major reduction in toughness. epoxy CT samples as any misalignment or material heterogeneity

1.10

1.05

1.00 stage 1 stage 2

KIc (MPa.m 0.5)

0.95

0.90

0.85

0.80

0.75

0.70
0.0 0.5 1.0 1.5 2.0 2.5
Total local water uptake (%wt)
Fig. 11. Simulated loadedisplacement curves for heterogeneous and homogeneous
Fig. 9. Identification of KIc with respect to the total local water uptake in subregion 1. samples.
 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326 325

Fig. 12. Conic-shaped patterns in the fracture surface.

will perturb these curves). We note that the heterogeneity due to The morphology of the fractured surface is also largely modified
moisture does not modify this conclusion and results in a very during wet aging as revealed by the correlation between KIc and the
sharp decrease. Moisture only tends to blunt the sharp peak at the roughness (Ra) in the vicinity of the crack tip (Fig. 13). Our obser-
maximum load but in a very minor fashion. vations are consistent with the results of Ravi-Chandar and Knauss
[19] and Araki et al. [20], from which it is deduced that there is a
4.3. Morphology of the cracked surface positive relationship between toughness and roughness.

It has been previously shown that the fractured surfaces of 5. Conclusion

polymers can have various patterns depending on the material and
system configuration [16,17,3,18]. In the fractography of our sam- We studied the effects of wet aging on the toughness of an
ples, some conic-shaped features can be seen (Fig. 12), as observed epoxy resin by using classical CT testing. Fracture toughness tests
also in other glassy polymers [16,5,4]. Lin and Chen [5] described and fracture surface analyses were performed by roughness mea-
the conic geometry in fractured surfaces as a competition between surement and micrography. We developed a numerical model to
the planar crack front producing shear yielding (a ductile mecha- gain a more detailed understanding of water absorption and its
nism) and secondary cracking producing crazing (a brittle mecha- consequences.
nism). This is characteristic of a crack growing at high-speed, As expected, wet aging leads to detrimental effects on the
supporting the conclusion in Section 4.2 that crack extension is in fracture toughness of an epoxy resin: there is a dramatic drop in the
practice unstable and is fully guided by the phenomenology close to value of KIc after 48 h of aging. Yet, we have shown that this change
the initial crack tip. in toughness can not be directly correlated with global water up-
take. Indeed, crack propagation is mainly guided by the local
1.10 moisture configuration close to the initial crack tip. The post pro-
cessing of data from such experiments to derive the material re-
1.05
lations between toughness and moisture uptake therefore has to
y = 0.0018x + 0.6682
1.00 account for the heterogeneity of the material.
KIc (MPa.m 0.5)

It has been shown that accurate simulation of the diffusion/re-

0.95 action process can be a way to analyze such experimental data
better, as such simulation leads to much more local information.
0.90 The measured toughness can be correlated to the water content in
aging time the neighborhood of the crack. This provides valuable and robust
0.85
information for the modeling of wet aging-induced failure in such
0.80 materials, as the obtained relation can be trusted for further 3D
simulation of wet aging/fracture coupling.
0.75
50 100 150 200 250
References
Roughness Ra (nm)
[1] Lai M, Botsis J, Cugnoni J, Coric D. An experimental-numerical study of
Fig. 13. Correlation between KIc and roughness (the error bars represent maximum moisture absorption in an epoxy. Compos Part A: Appl Sci Manuf 2012;43(7):
and minimum values). 1053e60.
326 G. Quino et al. / Polymer Degradation and Stability 109 (2014) 319e326

[2] Odegard GM, Bandyopadhyay A. Physical aging of epoxy polymers and their Second ESIS TC4 Conference on Fracture of Polymers, Composites and Adhe-
composites. J Polym Sci Part B: Polym Phys 2011;49(24):1695e716. sives. European Structural Integrity Society, vol. 27. Elsevier; 2000. p. 187e99.
[3] Chang TD, Brittain JO. Studies of epoxy resin systems: part D: fracture [11] Yagoubi JE, Lubineau G, Roger F, Verdu J. A fully coupled diffusion-reaction
toughness of an epoxy resin: a study of the effect of crosslinking and sub-Tg scheme for moisture sorption-desorption in an anhydride-cured epoxy
aging. Polym Eng Sci 1982;22(18):1228e36. resin. Polymer 2012;53(24):5582e95.
[4] Lin Y, Chen X. Moisture sorption-desorption-resorption characteristics and its [12] Yagoubi JE, Lubineau G, Saghir S, Verdu J, Askari A. Thermomechanical and
effect on the mechanical behavior of the epoxy system. Polymer 2005;46: hygroelastic properties of an epoxy system under humid and cold-warm
11994e2003. cycling conditions. Polym Degrad Stab 2014;99:146e55.
[5] Lin Y, Chen X. Investigation of the effect of hygrothermal conditions on epoxy [13] Yagoubi JE, Lubineau G, Verdu J, Traidia A. Hydrolysis reaction during mois-
system by fractography and computer simulation. Mater Lett ture sorption in an epoxy matrix: experimental evidence and numerical
2005;59(29e30):3831e6. simulations. Compos Part A 2014 [in press].
[6] Alessi S, Conduruta D, Pitarresi G, Dispenza C, Spadaro G. Hydrothermal [14] Standard test methods for plane-strain fracture toughness and strain energy
ageing of radiation cured epoxy resin-polyether sulfone blends as matrices for release rate of plastic materials. American Society of Testing of Materials;
structural composites. Polym Degrad Stab 2010;95(4):677e83. 2007D5045e99.
[7] Alessi S, Conduruta D, Pitarresi G, Dispenza C, Spadaro G. Accelerated ageing [15] Zhou J, Lucas J. Hygrothermal effects of epoxy resin. Part II: variations of glass
due to moisture absorption of thermally cured epoxy resin/polyethersulphone transition temperature. Polymer 1999;40:5513e22.
blends. thermal, mechanical and morphological behaviour. Polym Degrad Stab [16] Kusy R, Lee H, Turner D. Rib formation in the fracture of polymethyl meth-
2011;96(4):642e8. acrylate. J Mater Sci 1976;11(1):118e24.
[8] Chale
at C, Halley P, Truss R. Properties of a plasticised starch blend. Part 1: [17] Liu K, Piggott MR. Fracture failure processes in polymers. II: fractographic
influence of moisture content on fracture properties. Carbohydr Polym evidence. Polym Eng Sci 1998;38(1):69e78.
2008;71(4):535e43. [18] Zhang MJ, Zhi FX, Su XR. Fracture toughness and crack growth mechanism for
[9] Ba
ra

ny T, Fo
€ldes E, Cziga

ny T. Effect of thermal and hygrothermal aging on the multiphase polymers. Polym Eng Sci 1989;29(16):1142e6.
plane stress fracture toughness of poly(ethylene terephthalate) sheets. Ex- [19] Ravi-Chandar K, Knauss WG. An experimental investigation into dynamic
press Polym Lett 2007;1(3):180e7. fracture: II. Microstructural aspects. Int J Fract 1984;26(1):65e80.
[10] Clutton E. ESIS TC4 experience with the essential work of fracture method. In: [20] Araki W, Adachi T, Yamaji A, Gamou M. Fracture toughness of bisphenol A-
Williams J, Pavan A, editors. Fracture of Polymers, Composites and Adhesives type epoxy resin. J Appl Polym Sci 2002;86(9):2266e71.