Polymer Degradation and Stability 111 (2015) 55e63

Contents lists available at ScienceDirect

Polymer Degradation and Stability

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

Absorption kinetics and swelling stresses in hydrothermally aged

epoxies investigated by photoelastic image analysis
Giuseppe Pitarresi a, *, Michele Scafidi a, Sabina Alessi a, Maria Di Filippo a,
Claude Billaud b, Giuseppe Spadaro a
a
University of Palermo, Dipartimento di Ingegneria Chimica, Gestionale, Informatica, Meccanica (DICGIM), Viale delle Scienze, 90128 Palermo, Italy
b
Cytec, Wilton Centre R420, Redcar TS10 4RF, UK

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

Article history: The present work proposes an experimental optical methodology able to measure the transient swelling
Received 31 July 2014 stresses induced by the water uptake ageing of polymers. In particular, the work describes the imple-
Received in revised form mentation of a Photoelastic technique to quantify internal stresses arising during the hydrothermal
21 October 2014
ageing of cast epoxy samples. The material investigated is a model DGEBA/DDS epoxy system. Curing and
Accepted 24 October 2014
Available online 31 October 2014
post-curing cycles have been optimised in order to obtain a fully cured, high Tg, and completely stress
free initial condition. Rectangular beam samples were then left in a hydrothermal bath at 90
C, and
regularly monitored by gravimetric and photoelastic analyses. The quantitative evolution of stresses
Keywords:
Hydrothermal ageing
induced by water ingress was obtained by a Photoelastic Stress Analysis technique applied during water
Swelling stresses absorption, and successive desorption in controlled conditions. The evolution of stresses is correlated
Photoelastic stress analysis with gravimetric data, allowing to gain a new insight to investigate the complex swelling and diffusion
Thermosetting resin kinetics of water ingress into thermoset polymer networks.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction The influence of water ingress in epoxy networks has received

growing attention in the last years, and a number of studies are
Epoxy based thermoset polymers are used in many structural now available, generally based on material characterisation after
applications for their combination of good physical and mechanical accelerated ageing conditioning in either hot humid environments
performances, chemical resistance and synthesis processability. (hygrothermal) or hot water immersion (hydrothermal).
Typical applications include matrices for advanced composites, The diffusion kinetics has been generally investigated by means
structural adhesives, anticorrosive pipe coatings, electronic pack- of gravimetric analyses, and for most neat epoxy systems a Fickian
aging and plastic cases. Many of such applications involve the diffusion behaviour has been generally observed. It has been also
presence of some hybrid material assemblies and interfaces, where reported that insufficient curing, dispersion of nano-fillers or sec-
the epoxy network is usually coupled to materials with severe ondary phases, microcavities or crazing may all lead to some de-
mismatches in properties. Environmental stability is then a very grees of deviations from the Fickian behaviour in the near
important requirement in order to preserve the optimised me- saturation stages [2e8].
chanical cooperation of such couplings. Glassy epoxies are gener- From gravimetric analyses the initial diffusion coefficient
ally prone to solvent penetration, in particular water uptake by (diffusivity), the activation energy and the water mass uptake at
exposure to humid environments, which represents the principal equilibrium have been correlated to physical, chemical and me-
concern in terms of environmental ageing for such materials. This chanical parameters. Physical parameters have included the free
generally can produce both degradation of the network structure volume and the amount of free water [9,10], the glass transition
and of the mechanical properties, with material swelling likely to temperature and the dynamic thermal mechanical behaviour
lead to severe strains and stresses in a constrained assembly [1]. [11e16]. Chemical parameters have comprised the degree of for-
mation of bonded water by hydrogen bonds [17], the functionality
of the epoxy monomer [18], the nature of curing agent and its
* Corresponding author. Tel.: þ39 320 9242242; fax: þ39 091 23860671.
stoichiometric ratio [3,19], the curing cycle [12,19,20]. Mechanical
E-mail addresses: giuseppe.pitarresi@unipa.it (G. Pitarresi), michele.scafidi@
unipa.it (M. Scafidi), sabina.alessi@unipa.it (S. Alessi), claude.Billaud@cytec.com parameters have included: fracture toughness [13,14], static tensile
(C. Billaud), giuseppe.spadaro@unipa.it (G. Spadaro). behaviour [21], the Coefficient of Hygroscopic Swelling [6,22,23]

http://dx.doi.org/10.1016/j.polymdegradstab.2014.10.019
0141-3910/© 2014 Elsevier Ltd. All rights reserved.
56 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63

and the swelling induced deformation field (investigated by both fundamental for the implementation of transmission Photo-
experimental evaluations [6,22e27] and numerical simulations elasticity (see Section 3). The curing and post-curing specifications
[28e31]). have been optimised after preliminary DSC and DMTA analyses,
One difficulty associated with a full comprehension of ageing by here not reported, of the curing degree under various conditions.
solvent uptake is the coexistence of different phenomena, which The chosen thermal conditioning resulted in two curing steps. The
are closely interconnected and which require a multi-physics first one consisted of a temperature gradient of 2
C/min from room
approach to be effectively correlated. One example is the forma- temperature to 180
C, resting at this temperature for 2 h, and
tion of free and bonded water, which strongly depends on the cooling at the same rate to room temperature. The post-curing step
amount of free volume made available by the network structure was performed at 200
C for 2 h, followed by a slow cooling to room
[9,20,32], and the presence and accessibility of hydrophilic and temperature in 24 h. This last treatment was applied on beam
polar groups to form hydrogen bonds with water molecules samples (nominal dimensions of about 30  10  3 mm3) cut from
[33e35]. Many works have suggested that only bonded water is the cured plaques by using a high speed milling cutter.
responsible for swelling [32,36], through a mechanism that The glass transition temperature, Tg, of the unaged material has
weakens the inter-chain connections by breaking the hydrophilic been determined by Dynamical Mechanical Thermal Analysis
groups and bonding them to water. Very few works though have (DMTA) by means of a Rheometrics DMTA V in a single cantilever
fully clarified these mechanisms, and their mutual consequences beam arrangement at a heating rate of di 5
C/min, frequency of
on the changing network structure and further diffusion behaviour 1.8 Hz and elongation of 0.02%. In these experimental conditions
(diffusion-reaction dynamics). the value of Tg, derived as the temperature corresponding to tan d
From a mechanical characterisation perspective, the contribu- peak, was about 250
C.
tion has been even poorer, due to the difficulty to implement
sensitive and real time full field stress and strain measurement 2.2. Hydrothermal conditioning and gravimetric analysis
techniques. Few works are available, attempting to measure
swelling strain fields by either direct local [37] or full-field [24] Gravimetric and Photoelastic data were collected throughout
deformation measurements. Many of such works have usually the entire conditioning history, at shorter intervals during the
measured only bulk strain components on fully saturated samples. initial stages of absorption and desorption. Hydrothermal ageing
Very little information is then available on the local and transitory was performed in a temperature controlled bath of distilled water,
evolution of stresses and strains induced by swelling. More which was maintained at 90
C. Desorption was carried out in
recently, Digital Image Correlation (DIC) has emerged and proposed controlled room and moisture conditions, keeping the samples in a
to acquire the full field strain evolution [25e27]. The technique sealed container with a calcium chloride salt to obtain a desiccated
though has some limitations in sensitivity and is not fully effective low moisture airborne. In both cases the samples were occasionally
for measuring water uptake induced stresses/strains. taken out of the conditioning environment, for the weighting and
The present paper proposes the implementation of a Photoe- acquiring of the photoelastic images. In particular all samples were
lastic Stress Analysis (PSA) technique which has allowed the weighted on a 0.01 mg resolution electronic balance. During water
quantitative and full field measurement of stresses induced by sorption the sample surface was gently wiped out of superficial
swelling during the transitory hydrothermal sorption and desorp- water before testing.
tion of rectangular beam samples, made of a model DGEBA/4,40 DDS
system. The hygro/hydro-thermal ageing of this chosen combina- 2.3. Photoelastic acquisitions
tion of monomer and curing agent have been widely investigated in
the literature [9,17], and this has allowed to evaluate the repro- Photoelasticity is a well known full field stress analysis tech-
ducibility of some typical features of the absorption behaviour. The nique based on the optical properties of transparent birefringent
optical transparency of samples, achievable by opportune materials [38]. In this work Transmission Photoelasticity is imple-
manufacturing, has allowed the implementation of transmission mented, with Photoelastic maps obtained by placing samples in a
photoelasticity [38]. The material has shown a marked sensitivity to circular polariscope with a monochromatic 589 nm light source
this technique, allowing the quantitative evaluation of the small (see Fig. 1 left). In particular four optical filters composed the
residual internal stresses arising during the various ageing stages. polariscope used in this work: two polarised plates called Polariser
It is believed that the quality and nature of the information (P) and Analyser (A), and two quarter wave plates, called RP and RA.
measured by the proposed technique, has the potential to allow a The orientation of these optical plates can be changed (see Fig. 1
new insight into the investigation of the complex swelling and right), allowing to obtain different information on the stress field
diffusion phenomena associated to water ingress into ageing glassy of the photoelastic component [38]. Due to the analogy with the
thermoset polymers. residual stress analysis of the glasses, the water absorption stresses
can be evaluated by the techniques proposed in Refs. [39e47]. In
2. Experimental procedure particular a simplified phase-shifting method, based on the set-up
of the Tardy Compensator (briefly called Tardy Phase-Shifting
2.1. Material and sample preparation Method), was used in this work (see Section 3.2) [43].
A RGB camera connected to a PC with a digital acquisition board
A model epoxy system has been selected, obtained by a mono- was used to acquire bitmapped images. The optimised optical setup
mer base 2,2-bis[4-(glycidyloxy)phenyl]propane (DGEBA), epoxide of the camera was achieved after placing a reference beam sample
equivalent weight 172e176, and a 4,40 diamino-diphenyl sulfone at a certain distance from the camera lens, and was used
(DDS) curing agent, both supplied by Sigma Aldrich. The two com- throughout the ageing monitoring by placing samples always in the
ponents where blended by mixing a stoichiometric amount of DDS same reference position.
to the DGEBA monomer at 80
C, and fully dissolved after me-
chanical stirring for 30 min at 130
C. The resin blend was cast into 3. Photoelastic stress analysis
an aluminium open mould flat plate, having a smooth finish. This
allowed the manufacturing of small rectangular plaques with a good Stresses in a component may arise by externally applied loads or
thickness control, smooth surfaces and adequate transparency, by internal loads (e.g. residual stresses due to physical-mechanical
 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63 57

Fig. 1. (Left) Lay-out of the polariscope: S: light source, P: polariser, A: analyser, RP: polariser quarter wave plate, RA: analyser quarter wave plate, C: photoelastic component, O:
observer; (right) general orientation of plates axes and principal stresses directions.

treatments, or stresses induced by constrained thermal or hygro- usually called dark field circular polariscope, and the emerging in-
scopic deformation). In the case studied in this paper, the load is tensity (eq. (2)) becomes:
due to the non-uniform swelling induced by the water adsorbed by
the material in the transitory preceding saturation. This stress I0
I ¼ If þ ð1  cos 2pdÞ; (3)
mechanism is somewhat similar to the case of thermal tempering 2
treatment of glasses, where a non-uniform distribution of residual
where the emerging intensity I is now uniquely related to the
stresses is generated, induced by the reciprocal restrain of defor-
relative retardation d.
mation between the external parts of the component (in
The sample shown in Fig. 2 is a rectangular beam of dimensions
compression) and the internal parts (in tension) [39]. In the case of
75  15  3.2 mm3, made of the same material system studied in
water absorption, the external part is initially absorbing a higher
this work, and loaded in four point bending with the purpose to
amount of water, which determines a volume swelling restrained
measure the Photoelastic constant of the material (see also Section
by the inner part, and thus generating a non-uniform compression-
3.3). The dark lines of Fig. 2 are called isochromatic fringes and the
tension stress distribution from the outer to the inner material.
points of this lines present constant and entire values of retardation
d.
3.1. The photoelastic effect

Photoelastic Stress Analysis techniques are in general able to 3.2. Implementation of TPSM e Tardy Phase-Shifting Method
evaluate two parameters of the stress field: the relative retardation
d, and the orientation a of the principal stress s1 with respect the In this work the Tardy Phase-Shifting Method (TPSM) is proposed
horizontal axis. Usually a is determined by unwrapping the isoclinic for the quantitative stress analysis [43]. This method requires the
parameter q, directly determined by photoelastic methods, that is acquisition of three images with three different values of bA. During
the orientation of the closer principal stress to the x horizontal axis the time required for the images acquisition, the sample is removed
(see. Fig. 1 right). from the thermal bath only for few minutes. During this short time
The relative retardation d is related to the principal stress the stress field can be considered as frozen, and the sample sorp-
quantities s1 and s2 by the well-known relationship: tion/desorption kinetics little affected.
The rectangular shape of the samples and the full-immersion in
Cd water determines a symmetric stress field distribution, were the
d¼ ðs  s2 Þ (1)
l 1 sample geometric axes are also the axes of symmetry for stresses
(see Fig. 3 and Section 4.2). By considering the points along the
where C is the photoelastic constant of the material, d is the symmetry x and y axes (with x aligned with the longer length of the
thickness of the bi-dimensional photoelastic component, and l is
the wavelength of the monochromatic light source.
Both quantities d and q can be determined by opportunely
combining the light intensities emerging from different set-ups of
the polariscope. For example the emerging light intensity I from the
polariscope configuration described by Fig. 1 right, called Tardy
compensator setup, using the Jones matrix calculus [48], is given
by:

I0
I ¼ If þ ð1  cos 2pd cos 2bA þ sin 2pd sin 2bA cos 2 aÞ (2)
2

being If the background intensity, I0 the fringe pattern intensity and

bA the orientation of the analyser plate with respect to the x axis.
In Fig. 2 a typical photoelastic image is shown. The image was Fig. 2. Calibration sample loaded in four point bending and observed in dark-field
obtained by setting bA ¼ 0
in the Tardy polariscope: this setup is circular polariscope.
58 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63

Fig. 3. (Up) Full field isochromatic map acquired in white light circular polariscope after 48 h of immersion in the thermal bath; (down) Schematic representation of the beam
rectangular sample, with qualitative interpretation of principal stresses along the symmetry axes.

beam sample, and with the polariscope reference axis), the isoclinic 3.3. Determination of stresses from the photoelastic data
parameter q is null (no tangential stresses arise along the symmetry
axes, which are then principal stress directions). The quantitative Once the relative retardation d has been determined by eq. (7),
analysis by the TPSM method is in particular conducted only along the stress term (s1s2) can be calculated by eq. (1). Specifically,
the y-axis of symmetry. Here q ¼ 0
and a is either 0
or 90
, application of eq. (1) requires the use of the photoelastic constant C.
depending if the higher principal stress s1 is aligned with x or y- Because of the peculiar characteristics of the used epoxy-resin
axis. Hence, equation (2) simplifies and three unknowns If, I0, and system, the photoelastic C constant cannot be reliably taken from
d are present, even if the only one quantity of interest is d. To the literature. In the present work C is obtained by a calibration
determine the unknowns, at least three images have to be acquired. procedure, by using the 4-point bending loaded sample shown in
The TPSM allows the user to determine the retardation d by Fig. 2.
acquiring three different images, setting bA1 ¼ 0
, bA2 ¼ 45
and In particular the beam sample for calibration was cut from the
bA3 ¼ 90
[43]. From equation (2), the light intensities emerging same plaque used to obtain the aged samples, in order to guar-
from the polariscope are: antee the same curing conditions. Then the calibration beam was
post-cured together with the samples destined to ageing, and
tested immediately after extraction from the oven, to ensure no
I0
I1 ¼ If þ ð1  cos 2pdÞ; (4) significant air humidity uptake. The initial photoelastic image of
2
the unloaded sample confirmed the absence of any residual
stresses at this stage. The image in Fig. 2 in particular has been
I0 acquired by using a dark-field circular polariscope using a mono-
I2 ¼ If þ ð1±sin 2pdÞ; (5) chromatic light source provided by yellow sodium lamps with
2
l ¼ 589 nm wavelength. The sample section height was
h ¼ 14.96 mm high and had a constant thickness d of 3.2 mm. The
I0 applied bending moment was B ¼ 0.4655 Nm. From equation (1),
I3 ¼ If þ ð1 þ cos 2pdÞ; (6)
2 the photoelastic constant C has been determined by the following
relationship:
where the upper sign in eq. (5) is for a ¼ 0
and the lower sign is for
a ¼ 90
.
The retardation d is obtained by the following relationship: lh3 jDNj
C¼ (8)
12B jDyj
I1 þ I3 ±2I2
tan 2pd ¼ ; (7) being jDyj the vertical distance between two points selected along
I3  I1
the symmetry y-axis and jDNj the absolute fringe-order difference
where the same sign convention of eq. (5) applies. between these points (considering for the calculation negative
However, due to the periodicity of the tangent function, eq. (7) fringe order for the points selected in the tensile zone, and vice
provides the fractional retardation df, which is defined in the versa). It is noticed that, using this calibration procedure, the
range ±0.5 fringe orders. Unwrapping procedure can be success- photoelastic constant C does not depend on the thickness d of the
fully applied to determine the retardation d, provided that the sample.
proper retardation value in one point is supplied by the user. The photoelastic constant C was:
The presented TPSM data reduction scheme is not the only
applicable photoelastic stress analysis technique [43e47], but for .
the present case it is believed to provide fewer errors than other C ¼ 76:95 Br≡7:695$105 mm2 N : (9)
digital photoelastic methods, and is fast and easy in terms of both
experimental implementation and image analysis [43,47].
 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63 59

4. Results and discussion

4.1. Water uptake

Weight measurements of five nominally identical samples have

been taken throughout absorption (up to 842 h) and desorption (up
to 576 h). Results are plotted in terms of relative masspffiffi change
versus time, t, and of normalised water content versus t =h, where
h is the sample thickness (see Fig. 4), according with the following
definitions:

Wt  Wi
relative mass change :  100 (10)
Wi

Mt Wt  Wi
normalised water content : ¼ (11)
Minf Winf  Wi

where Wt is the weight at time t, and subscripts i and inf indicates

the initial and saturated conditions, respectively. In the above
definitions it is implicitly assumed that no water content is present
at time i, when the sample is taken out of the oven after slow post-
curing and placed in the thermal bath.
A one dimensional Fick diffusion kinetics is assumed, which
allows to model the initial water uptake by the following simple
linear relation (valid for values of Mt up to about 0.5$Minf):
pffiffiffiffiffiffi
Mt 4 Dt
¼ pffiffiffiffi (12)
Minf p h

where h is the thickness of the sample and D is the Fickian diffusion

coefficient. Linear interpolation of the initial points in Fig. 4b and c
have then allowed to measure D ¼ 69.5  109 cm2/s
(25  103 mm2/h) at absorption, and D ¼ 2.54  109 cm2/s
(0.91  103 mm2/h) at room temperature desorption.
Several studies in the literature are available where D and Minf
are reported for glassy polymers using similar monomers and
various curing agents, curing cycles, and ageing conditions. In this
work a comparison is proposed in particular with the works of [9]
and [17], based on the fact that both works use a fully cured DGE-
BAþ4,40 DDS system, and ageing by water immersion at some
controlled temperature. Furthermore, a similar Tg (about 250
C) is
achieved for the system cured in this work and that of [9] (while no
data on Tg are available from Ref. [17]). A direct comparison of
diffusivity values shows that D increases with the ageing temper-
ature, T, also by an order of magnitude, moving from T ¼ 25
C to
T ¼ 90
C. In order to account for the different ageing temperatures,
values of diffusivity D are plotted according to the following
Arrhenius equation [16,49]:
 
E 1
D ¼ Do exp  d (13)
R T

where Ed is the activation energy of diffusion, R the universal gas

constant and T the absolute temperature. By taking the natural
logarithms of both members of eq. (13), this yields a linear rela- Fig. 4. Relative mass uptake (a) and normalised water content during sorption (b) and
tionship. Fig. 5 plots values of ln(D) versus 1/T of various data from desorption (c) versus time.
the literature. In particular the comparison between the point ob-
tained in this work, with the point from work [9] and the points bulky samples prepared in this work. Three further data sets are
from Ref. [17] shows a very good linear trend, with a resulting added in Fig. 5: one point from Ref. [15], one point from Ref. [18],
activation energy Ed ¼ 47 kJ/mol, obtained by a linear regression of and three points from Ref. [16]. All these mentioned works uses
all these points. This result arises even if the samples dimensions in DGEBA monomers cured with amine, and perform ageing by water
this work and in Refs. [9,17] have slightly different aspect ratios immersion. Some differences though exist in the achieved values of
between planar and thickness dimensions. This can be assumed as Tg (about 148
C for [15], 178
C for [18] and 99.1
C for the neat
an indirect proof that the one dimensional Fick law is still suitable epoxy system of [16]), due to differences in the exact formulation of
to model the initial diffusion kinetics even with the rather thick and the epoxy blend (monomer and curing agent) and curing cycles. In
60 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63

4.2. Photoelastic stress analysis: discussion of qualitative result

Fig. 3 up shows a typical isochromatic map from the analysed

samples after immersion in water for a given time. The rectangular
shape of the samples and the full-immersion in water determines a
symmetric stress field distribution, were the sample geometric axes
are also the axes of symmetry for stresses.
Fig. 6 shows some isochromatic maps acquired throughout the
ageing conditioning from a circular dark field polariscope in
monochromatic light. The first image at time zero (i.e. beginning of
the hydrothermal ageing) shows the sample in a completely stress
free condition, as a result of the slow post-curing cycle, which had
both an annealing and drying effect on the material. The images
taken in the first 50 h of permanence in the thermal bath are
showing a peculiar symmetric distribution of isochromatic fringes,
due to the internal stresses induced by the non-uniform swelling.
Stresses in this initial transitory changes at a fast rate, loading the
sample inner parts in traction and the outer parts in compression. A
slower stress relaxation follows as soon as absorption of water, and
hence swelling, reaches the inner parts of the sample. The last
image after 380 h shows an almost completely restored stress free
condition, due to the homogenisation of swelling on the whole
volume of material.
Fig. 5. Comparison of literature results on the dependence of diffusion on temperature
for DGEBA-amine systems.
The axisymmetric distribution of photoelastic fringes is due to
the homogeneous nature and isotropic behaviour of the material,
which determines the same diffusion coefficient in all directions
Ref. [18] in particular the epoxy system uses an isomer of 4,40 DDS, and the same water diffusion kinetics on all exposed faces of the
that is 3,30 DDS. In all cases, the values of D are well aligned (e.g. beam sample. The non-uniform swelling in the initial transitory is a
data from Ref. [16]) or slightly below (e.g. data from Refs. [15,18]), consequence of the higher amount of water absorbed along the cut
the regression line obtained by the DGEBAþ4,40 DDS systems. The edges of the sample, where a wider surface is exposed to water. This
lower crosslink density achieved by the lower Tg systems might determines a higher rate of swelling on the outer material, and the
determine an increased chain packing that probably is the main mutual restrain between outer and inner material determines the
cause of the lower diffusivity, even though the effect is tiny and all internal stresses detected by the photoelastic images.
compared systems have remarkably close diffusion behaviour. In order to provide a physical interpretation of the arising stress
It is also observed that the value of activation energy obtained field during water uptake, it is helpful to consider a simplified two-
from the comparison of the DGEBAþ4,40 DDS systems (47 kJ/mol) is dimensional schematic representation of the rectangular beam
just slightly lower than that reported in Ref. [16] (51 kJ/mol), even if sample as composed by an outer zone and an inner zone (see Fig. 3
the DGEBA þ amine system in Ref. [16] achieves a much lower Tg of down). The outer zone undergoes a hydrostatic volume expansion
100
C. (swelling), while the inner zone is unperturbed. The two zones are
It is also interesting to report that the value of Minf reached in though bound together, and this internal constraint is the cause for
this work, i.e. about 3.92%, is similar to those measured in Refs. the arising internal self-equilibrated stresses. The map in Fig. 3 up
[9,17] for the equivalent systems, and for the post-cured DGEBA- was acquired by using a white light source in the circular polari-
4,40 DDS system studied in Refs. [19], this despite the different scope in dark field. This allows to interpret the dark zones (black
ageing temperatures in these mentioned works. Already Merdas fringes) as the loci of points where (s1s2) ¼ 0. These dark fringe
et al. [50] have found that water solubility in polymers of medium zones provide a good direct indication of the boundary between
to high polarity is almost temperature independent. Lower values these inner and outer zones, at least in the initial stages of water
of equilibrium water uptake Minf are instead reported for the neat uptake. By considering points along the y-axis of symmetry, the
DGEBA systems in Refs. [15,18] and the not post-cured DGEBA- main stress component will be sx. The constrained swelling is such
4,40 DDS system in Ref. [19]. to give a negative sx in the outer zone (compression), and positive
The above results are then in partial agreement with some in the inner zone (traction). Along the y-axis there will be also a sy
recent suggestions (e.g. Ref. [17]), that the free volume has a pre-
dominant influence on the diffusion coefficient D and activation
energy Ed. While the equilibrium water content Minf is more sen-
sitive to the quantity of bonded water, which depends primarily on
the chemical nature of the monomer and curing agents.
As shown in Fig. 4a, water desorption was also monitored, while
leaving the saturated samples in controlled room conditions. It is
noteworthy how the decreasing water content seems to reach a
pseudo horizontal plateau well above complete water rejection. A
significant amount of absorbed water (about 2.4% in terms of
relative mass uptake) was still present in the samples after about
600 h, soon after the onset of a steep reduction in diffusivity (see
Fig. 4c). It is also noteworthy to observe that the initial desorption
kinetics seems to follow a linear behaviour when plotted versus the Fig. 6. Isochromatic maps from the monochromatic light circular polariscope, acquired
square root of time (Fickian behaviour). on the same sample throughout water absorption.
 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63 61

component growing from zero at the boundary to a positive

(traction) value at the mid-centre (intersection with the x-axis of
symmetry). This sy component arises due to the swelling action of
the outer left and right ends of the sample. Due to the rectangular
aspect ratio of the sample, the contribute of sy in the term (s1s2)
along the y-axis of symmetry is small. By assuming sy z 0 along the
y-axis of symmetry, it is possible to consider:
8
< s1 ¼ sy ¼ 0 and s2 ¼ sx < 0 and a ¼ 90
ðouter zoneÞ
s ¼ sx > 0 and s2 ¼ sy ¼ 0 and a ¼ 0
ðinner zoneÞ
: 1
s1 ¼ s2 ¼ 0 ðdark fringeÞ
(14)
A similar interpretation, with opportune adaptations, can be
made for the stress components along the x-axis of symmetry. In
this case the sy component near the borders will be negative
(compression), turning positive in the inner zone, and becoming
Fig. 8. Stresses versus absorption ageing time at different positions along the sym-
smaller towards the sample centre. The sx component will be al-
metry y-axis.
ways positive, going from zero on the left and right borders, and
growing towards the sample centre.
Since the quantitative TPSM analysis, discussed in the next reconstructed, than the reported values have to be considered as
section, determines (sxsy) along the symmetry y-axis, no attempt qualitative results. It is then not possible to evaluate the exact
is made here to separate the stress components, but the relations maximum compression stress sx, while for the maximum positive
given in eq. (14) can be assumed in first approximation to ease the (sxsy) in the inner sample this has been found to reach a value of
interpretation of data. 1.87 MPa. From Fig. 7 it is also possible to notice that the neutral
axis (sxsy) ¼ 0 moves towards the centre by very little (less than
half mm), and the part of the sample in traction is about two
4.3. Photoelastic stress analysis: discussion of quantitative result
thirds of that in compression, which will then reach higher
compressive values. The higher peak in compression measured in
Fig. 7 plots the value of (sxsy) z sx along the vertical y axis of
this work is about 11 MPa, which is a slight underestimation of
symmetry. The horizontal axis in this plot is normalised by the
the real maximum compression peak.
sample width (H ¼ 8 mm), and plots are arrested at mid sample
Fig. 8 plots the evolution of (sxsy) at four specific locations: y/
width, such that the value y ¼ 0 corresponds to the outer border
H ¼ 0.1, 0.15, 0.25, 0.5. It is observed that points near the border
and y ¼ H/2 to the sample centre. Several curves of sx are plotted
experience a negative stress during all the absorption time, up to
at different ageing times. It is in particular seen that sx near the
saturation, reaching a maximum in the first stages of absorption,
outer border is compressive and becomes positive by moving to-
and then gradually relaxing to zero (this is the case for point y/
wards the sample centre. The point at sx z (sxsy) ¼ 0 (neutral
H ¼ 0.1). Moving towards the centre of the sample there are a
axis) moves from the outer border towards the centre with ageing
number of points that experience both compression (first) and
time. Also (sxsy) reach a negative and positive peak at some time
traction (after). Moving further towards the centre all remaining
after absorption initiation, and then gradually relax up until an
points experience only traction, reaching a maximum tensile stress
almost complete, restored stress free condition. It is noticed that
and then relaxing. It is observed that the compressive and tensile
the digital photoelastic analysis is not able to correctly quantify
peaks are reached at about the same time, after 30 h, which is a very
stresses very near the outer border, in particular when high fringe
early time compared to the time needed to the full onset of satu-
gradient are present (more than 0.1 fringe order/pixel) [51]. This
ration (about 700 h). The photoelastic quantitative analysis was not
condition is presented in the zone near the edge (between 0 < y/
able to resolve stress differences after about 400 h of ageing time,
H < 0.1). In these zones, the stress profile was manually
due to the very small residual stresses at this stage.
The desorption behaviour at room conditions was also moni-
tored by Photoelastic Stress Analysis. In fact, even if the photoe-
lastic maps at saturation revealed an almost ideal stress free
condition of samples, this are though deformed due to swelling.
The fact that no residual internal stresses are detected is due to the
hydrostatic nature of swelling and its uniform distribution over the
whole volume.

Fig. 7. Stresses along the symmetry y-axis during absorption (position normalised by Fig. 9. Full field isochromatic map acquired in white light circular polariscope after
the sample height H). 338 h from initiation of desorption at controlled room conditions.
62 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63

5. Conclusion

The present work has proposed the implementation of a Pho-

toelastic Stress Analysis technique for the qualitative and quanti-
tative evaluation of swelling stresses arising during the absorption
and desorption kinetics of water diffusion into a model DGEBA/DDS
epoxy system. While the only requisite of the technique is the
optical transparency of the cured cast polymer, transmission pho-
toelasticity has proven to be very sensitive in revealing the state of
small internal stresses induced by non-uniform swelling. The
experimental equipment, comprising a polariscope and a camera
for digital image acquisition and processing, is simple, cheap and
easy to set-up. The acquisition of photoelastic images for full
quantitative post-processing analysis is also fast enough to be
performed at regular intervals throughout the ageing conditioning,
without affecting the measured stresses and the absorption/
desorption kinetics.
Fig. 10. Stresses along the symmetry y-axis during desorption.
In this work a quantitative Tardy Photoelastic Phase Shifting
method has been proposed in particular, which was able to quantify
Fig. 9 shows an isochromatic map acquired after 338 h from the stress field in small rectangular beams made of fully cured
beginning of desorption. It is seen that a stress field arise, in DGEBA þ DDS system, aged by resting in a water bath at 90
C. The
particular along the sample borders. Figs. 10 and 11 shows plots of photoelastic stress analysis has evidenced the rise of a peculiar,
the (sxsy) on the vertical axis of symmetry versus point location self-equilibrated and symmetric stress field, which was rapidly
(Fig. 10) or versus time (Fig. 11). It is interesting to observe that the evolving during the first 100 h of absorption, within the initial
stress distribution is now somewhat opposite to what observed linear Fickian absorption behaviour of the material. The samples
during absorption, with the outer regions being progressively reached an almost stress free condition at saturation. Desorption at
stressed in traction, and the inner regions in compression. The room conditions was also monitored by Photoelasticity, revealing a
mechanism behind the rise of these internal stresses is due to the tendency to develop a quasi-permanent traction stress along the
mutual constraint between outer parts, which will un-swell faster sample boundaries, and compression in the inner core of the
due to water desorption through the sample faces, and the inner sample.
parts which tend to stay swelled. The un-swelling region will It is believed that the quality and nature of the information
compress the inner region, while the inner region will prevent the measured by the proposed Photoelastic technique, has the poten-
outer region from complete relaxation. tial to allow a new insight into the investigation of the complex
Fig. 10 shows the evolution of (sxsy) z sx along the y axis, with swelling and diffusion phenomena associated to water ingress into
the progressive formation of a peak in compression of glassy thermoset polymers. Further work is in due course to exploit
about 2 MPa, located nearby the border (between 0.1 < y/H < 0.2), the potentialities of the technique to analyse epoxy systems with
and tensile stresses higher than 6 MPa on the border. different network structures, in order to fully correlate the nature of
From Fig. 11 it is observed that outer points such like y/H ¼ 0.1 the measured swelling stresses to the material, sample geometry
are first stressed in compression and then remain loaded in trac- and water diffusion/reaction behaviour. Another promising appli-
tion, and inner points such like y/H ¼ 0.15, 0.25 remain constantly cation of the technique is the investigation of the influence of
stressed in compression. It is interesting to notice that after about ageing on the stress field at crack sites or biomaterial de-
500 h of monitoring of desorption the samples did not develop a laminations. A first example is provided in Refs. [52], where the
stress free condition, but rather tended to slow down the stress evolution of the fracture toughness of ageing bulk resin cracked
gradients, leaving the sample loaded in traction at the boundary, samples is interpreted by correlation with the photoelastic data.
and in prevalent compression in the centre.
Ackowledgments

The authors would like to thank Cytec Engineering Materials for

supporting this work through a grant to the research program
titled: “Structureeproperties relationships of epoxy matrices for
carbon fibre composites, the effect of the network structure on the
materials behaviour towards solvents exposure”.

References

[1] Fan XJ, Suhir E. Moisture sensitivity of plastic packages of IC devices. US:
Springer; 2010.
[2] Wong TC, Broutman LJ. Moisture diffusion in epoxy resins. Part 1. non-Fickian
sorption processes. Polym Eng Sci 1985;25(9):521e8.
[3] VanLandingham MR, Eduljee RF, Gillespie Jr JW. Moisture diffusion in epoxy
systems. J Appl Polym Sci 1999;71(5):787e98.
[4] Ivanova KI, Pethrick RA, Affrossman S. Investigation of hydrothermal ageing of
a filled rubber toughened epoxy resin using dynamic mechanical thermal
analysis and dielectric spectroscopy. Polymer 2000;41(18):6787e96.
[5] Popineau S, Rondeau-Mouro C, Sulpice-Gaillet C, Shanahan MER. Free/bound
water absorption in an epoxy adhesive. Polymer 2005;46(24):10733e40.
Fig. 11. Stresses versus desorption ageing time at different positions along the sym- [6] He Y. Moisture absorption and hygroscopic swelling behavior of an underfill
metry y-axis. material. Thermochim Acta 2012;546:143e52.
 G. Pitarresi et al. / Polymer Degradation and Stability 111 (2015) 55e63 63

[7] El Yagoubi EJ, Lubineau G, Roger F, Verdu J. A fully coupled diffusion-reaction Simulation and Experiments in Micro-Electronics and Micro-Systems, Euro-
scheme for moisture sorption-desorption in an anhydride-cured epoxy resin. SimE; 2006.
Polymer (United Kingdom) 2012;53(24):5582e95. [29] Zhou J. Transient analysis on hygroscopic swelling characterization using
[8] Coniglio N, Nguyen K, Kurji R, Gamboa E. Characterizing water sorption in sequentially coupled moisture diffusion and hygroscopic stress modeling
100% solids epoxy coatings. Prog Org Coat 2013;76(9):1168e77. method. Microelectron Reliab 2008;48(6):805e10.
[9] Jackson M, Kaushik M, Nazarenko S, Ward S, Maskell R, Wiggins J. Effect of free [30] Yoon S, Han B, Cho S, Jang C-. Non-linear finite element analysis for electronic
volume hole-size on fluid ingress of glassy epoxy networks. Polymer packages subjected to combined hygroscopic and thermo-mechanical
2011;52(20):4528e35. stresses. In: Proceedings of 7th Electronics Packaging Technology Confer-
[10] Wang J, Gong J, Gong Z, Yan X, Wang B, Wu Q, et al. Effect of curing agent ence, EPTC; 2005.
polarity on water absorption and free volume in epoxy resin studied by PALS. [31] Haghighi-Yazdi M, Lee-Sullivan P. Modeling of structural mechanics, moisture
Nucl Instrum Methods Phys Res Sect B: Beam Interact Mater Atoms diffusion, and heat conduction coupled with physical aging in thin plastic
2010;268(14):2355e61. plates. Acta Mech 2013:1e22.
[11] Zhou J, Lucas JP. Hygrothermal effects of epoxy resin. Part II: variations of glass [32] Xiao GZ, Shanahan MER. Swelling of DGEBA/DDA epoxy resin during hygro-
transition temperature. Polymer 1999;40(20):5513e22. thermal ageing. Polymer 1998;39(14):3253e60.
[12] Perrin FX, Nguyen MH, Vernet JL. Water transport in epoxy-aliphatic amine [33] Xiao GZ, Delamar M, Shanahan MER. Irreversible interactions between water
networks e influence of curing cycles. Eur Polym J 2009;45(5):1524e34. and DGEBA/DDA epoxy resin during hygrothermal aging. J Appl Polym Sci
[13] Alessi S, Conduruta D, Pitarresi G, Dispenza C, Spadaro G. Hydrothermal 1997;65(3):449e58.
ageing of radiation cured epoxy resin-polyether sulfone blends as matrices for [34] Apicella A, Tessieri R, de Cataldis C. Sorption modes of water in glassy epoxies.
structural composites. Polym Degrad Stab 2010;95(4):677e83. J Membr Sci 1984;18(C):211e25.
[14] Alessi S, Conduruta D, Pitarresi G, Dispenza C, Spadaro G. Accelerated ageing [35] Zhou J, Lucas JP. Hygrothermal effects of epoxy resin. part I: the nature of
due to moisture absorption of thermally cured epoxy resin/polyethersulphone water in epoxy. Polymer 1999;40(20):5505e12.
blends. thermal, mechanical and morphological behaviour. Polym Degrad Stab [36] Shirangi MH, Fan XJ, Michel B. Mechanism of moisture diffusion, hygroscopic
2011;96(4):642e8. swelling and adhesion degradation in epoxy molding compounds. In: Pro-
[15] Starkova O, Chandrasekaran S, Prado LASA, To € lle F, Mülhaupt R, Schulte K. ceedings e 2008 international symposium on microelectronics. IMAPS; 2008.
Hydrothermally resistant thermally reduced graphene oxide and multi-wall [37] Lin YC. Investigation of the moisture-desorption characteristics of epoxy resin.
carbon nanotube based epoxy nanocomposites. Polym Degrad Stab J Polym Res 2006;13(5):369e74.
2013;98(2):519e26. [38] Ramesh K. Photoelasticity. In: Sharpe WN, editor. Springer handbook of
[16] Starkova O, Buschhorn ST, Mannov E, Schulte K, Aniskevich A. Water transport experimental solid mechanics. New York, New York: Springer; 2008.
in epoxy/MWCNT composites. Eur Polym J 2013;49(8):2138e48. p. 701e42.
[17] Li L, Yu Y, Wu Q, Zhan G, Li S. Effect of chemical structure on the water [39] Aben H, Guillemet C. Photoelasticity of glass. Berlin: Springer; 1993.
sorption of amine-cured epoxy resins. Corros Sci 2009;51(12):3000e6. [40] McKenzie HW, Hand RJ. Basic optical stress measurement in glass. Sheffield:
[18] Frank K, Childers C, Dutta D, Gidley D, Jackson M, Ward S, et al. Fluid uptake Society of Glass Technology; 1999.
behavior of multifunctional epoxy blends. Polymer (United Kingdom) [41] Aben H, Anton J, Errapart A. Modern photoelasticity for residual stress mea-
2013;54(1):403e10. surement in glass. Strain 2008;44:40e8.
[19] Frank K, Wiggins J. Effect of stoichiometry and cure prescription on fluid [42] Redner AS, Voloshin AS. Surface and face stress in tempered glass. In: Pro-
ingress in epoxy networks. J Appl Polym Sci 2013;130(1):264e76. ceedings of 9th international conference on experimental mechanics, vol. 2;
[20] Nogueira P, Ramírez C, Torres A, Abad MJ, Cano J, Lo pez J, et al. Effect of water 1990. p. 884e91. Copenhagen.
sorption on the structure and mechanical properties of an epoxy resin system. [43] Ajovalasit A, Petrucci G, Scafidi M. Measurement of edge residual stresses in
J Appl Polym Sci 2001;80(1):71e80. glass by the phase-shifting method. Opt Lasers Eng 2011;49(5):652e7.
[21] Lin YC, Chen X. Moisture sorption-desorption-resorption characteristics and [44] Ajovalasit A, Petrucci G, Scafidi M. Photoelastic analysis of edge residual
its effect on the mechanical behavior of the epoxy system. Polymer stresses in glass by automated “test fringes” methods. Exp Mech 2012;52(8):
2005;46(25):11994e2003. 1057e66.
[22] Karalekas D, Cugnoni J, Botsis J. Monitoring of hygrothermal ageing effects in [45] Ajovalasit A, Petrucci G, Scafidi M. RGB photoelasticity applied to the analysis
an epoxy resin using FBG sensor: a methodological study. Compos Sci Technol of membrane residual stress in glass. Meas Sci Technol 2012;23(2).
2009;69(3e4):507e14. [46] Ajovalasit A, Petrucci G, Scafidi M. Photoelastic analysis of edge residual
[23] Lai M, Botsis J, Cugnoni J, Coric D. An experimental-numerical study of stresses in glass by the automated tint plate method. Exp Tech 2013. http://
moisture absorption in an epoxy. Compos Part A: Appl Sci Manuf 2012;43(7): dx.doi.org/10.1111/ext.12017. in press.
1053e60. [47] Ajovalasit A, Petrucci G, Scafidi M. A critical assessment of automatic pho-
[24] Stellrecht E, Han B, Pecht MG. Characterization of hygroscopic swelling toelastic methods for the analysis of edge residual stresses in glass. J Stress
behavior of mold compounds and plastic packages. IEEE Trans Comp Packag Anal Eng Des 2014;(5):361e75.
Technol 2004;27(3):499e506. [48] Theocaris PS, Gdoutos EE. Matrix theory of photoelasticity. Berlin: Springer;
[25] Park S, Zhang H, Zhang X, Ng SL, Lee HC. Temperature dependency of coef- 1979.
ficient of hygroscopic swelling of molding compound. In: Proceedings e [49] Bird RB, Stewart WE, Lightfoot EN. Transport phenomena. revised 2nd ed.
Electronic Components and Technology Conference; 2009. John Wiley & Sons; 2002.
[26] Jang C, Yoon S, Han B. Measurement of the hygroscopic swelling coefficient of [50] Merdas I, Tcharkhtchi A, Thominette F, Verdu J, Dean K, Cook W. Water ab-
thin film polymers used in semiconductor packaging. IEEE Trans Comp Packag sorption by uncrosslinked polymers, networks and IPNs having medium to
Technol 2010;33(2):340e6. high polarity. Polymer 2002;43(17):4619e25.
[27] Jackson MB, Heinz SR, Wiggins JS. Fluid ingress strain analysis of glassy [51] Ajovalasit A, Barone S, Petrucci G. A review of automated methods for the
polymer networks using digital image correlation. Polym Test 2012;31(8): collection and analysis of photoelastic data. J Strain Anal 1998;33(2):75e91.
1131e9. [52] Pitarresi G, Toscano A, Scafidi M, Di Filippo M, Alessi S, Spadaro G. Photoelastic
[28] Zhou J. Sequentially-coupled finite element transient analysis with hygro- stress analysis assisted evaluation of fracture toughness in hydrothermally
scopic swelling. In: 7th Int. Conf. on Thermal, Mechanical and Multiphysics aged epoxies. Frattura ed Integrita  Strutturale 2014;30:127e37.