Internal stress in a cured e p o x y resin s y s t e m

C. Brahatheeswaran* and V. B. Gupta~

Department of Textile Technology, Indian Institute of Technology, Delhi,
New Delhi 110016, India
(Received 5 November 1991)

An epoxy resin system based on a diglycidyl ether of bisphenol-A type difunctional epoxy resin containing
different amounts of a tetrafunctional curing agent, namely metaphenylene diamine, was used to prepare
two types of specimen. First, an aluminium strip was coated with a thin layer of the epoxy-amine mixture
and then subjected to two curing-cooling cycles. Second, bulk specimens were cast in the form of sheets which
were cut into rectangles and their dynamic torsional spectra obtained from - 1 0 0 to 180°C. In the
resin-coated metal plate, the internal stress was measured by the bending beam technique and was found
to be highest in samples with high glass transition temperatures (Tgs). This is apparently because the internal
stress arises due to the difference in the thermal contraction coefficients between the resin and the metal
and develops as the sample cools from the Tg to room temperature. The rate of stress relaxation in these
samples is also high, apparently because they are the farthest from their equilibrium state and so have
higher free volume and a more open structure. In the case of cast sheets, the internal stresses are again
the highest in samples with high Tg but arise from non-uniform cooling. An intermediate relaxation peak,
the ~' peak, observed in these samples has been shown to be related to the frozen-in internal stresses which
relax on sub-Tg ageing due to free volume collapse.

(Keywords: epoxy resin; crosslink density; internal stress; thermal stress; free volume; glass transition temperature)

INTRODUCTION EXPERIMENTAL
It is known that when epoxy resin is cured, it can develop Sample preparation
internal stresses. These stresses reduce adhesive strength Materials used. The epoxy resin used was Araldite
and occasionally induce cracks in the casting materials. GY 250, which is based on diglycidyl ether of bisphenol-A
Thus it is useful to have an understanding of the origin and manufactured by Hindustan Ciba Geigy Ltd. It has
of these stresses in cured epoxy resin. With this aim, the an epoxy equivalent of 190 and molecular weight of 380.
internal stress which develops in a diglycidyl ether of The curing agent used was a tetrafunctional aromatic
bisphenol-A based epoxy resin cured with different diamine, metaphenylene diamine (mPDA), with a melt-
amounts of curing agent has been measured. The ing point of 69°C. The structures of the resin and curing
measurements were made on a resin-coated aluminium agent have been given in an earlier publication 1.
strip which was subjected to the following two curing
cycles which involve high temperature curing followed Resin-coated aluminium strips. Thin, flat aluminium
by cooling to room temperature (RT): (1) 75°C for 2h, strips (90 m m long, 15 m m wide, 0.2 m m thick) were used
followed by 125°C for 2 h (standard cure); and (2) the as the base plate. A degassed resin-amine mixture
standard-cure sample post-cured for 6h in nitrogen containing 10, 14.5, 20 and 2 5 p h r of m P D A was then
(post cure). The curvature acquired by the strip as it applied as a thin coating on one side of the strip using
cools after curing provides a measure of the internal a thin glass rod. The corresponding initial molar ratios
stress; in the present investigations it showed consider- of the reactive groups, P = [ N H ] / [ E ] , for these four
able time dependence due to stress relaxation. Some samples were 0.71, 1.0, 1.42 and 1.77, respectively. The
comments are also made on the origin of internal stress coated strip was placed between two parallel metal plates
and its relaxation. In addition, dynamic torsional studies with spacers to give a resin coating of 0.1 mm. A release
between - 100°C and + 180°C were conducted on cured, agent was applied to the surface of the top metal plate
and cured and then sub-glass transition temperature which was in contact with the resin. The strips were then
(Tg)-aged bulk samples. It was observed that the inter- subjected to standard and post-cure cycles as described
mediate relaxation, the ~' relaxation, occurring above in the next section.
RT, flattened with ageing and thus appeared to be related On completion of the cure cycle, the oven was switched
to the local motion of chains which were in a state of off and the top metal plate was removed. The samples
internal stress. were then allowed to come to RT in the oven.

*Present address: Central Institute of Plastics Engineering and Castin9 of sheets. Cast sheets (2.5mm thick) were
Technology, Guindy, Madras 600032, India prepared by first mixing 10, 14.5, 20 and 25 phr of molten
t To whom correspondence should be addressed m P D A and then curing the resin mixture between casting

0032-3861/93/020289~)6
~'~ 1993 Butterworth-Heinemann Ltd. POLYMER, 1993, Volume 34, Number 2 289
Stress in an epoxy resin." C. Brahatheeswaran and V. B. Gupta

plates. The casting plate assembly consisted of a Dynamic torsional measurements. Dynamic torsional
U-shaped poly(tetrafluoroethylene) spacer clamped measurements were carried out on small rectangular
between two polished steel plates using clamping metal specimens cut from the cast sheets using a free oscillating
strips. The following two curing cycles were used1: (1) inverted torsion pendulum 4. Damping curves were
Standard cure. The resin and the amine were taken in obtained in the temperature range - 1 0 0 to + 180°C,
two different beakers and heated separately to 75°C. The from which tan 3 was estimated and the storage torsional
resin and the molten amine were then mixed well and modulus, G', was calculated using standard procedures
degassed in the oven for 7 rain. The resin mixture was laid down in ASTM D-2236.
then poured in the space between the metal plates which
had been preheated to 75°C. The casting plate assembly
was then heated at 75°C for 2h followed by 125°C for RESULTS AND DISCUSSION
2 h. The oven was then switched off and the sample was
allowed to come to RT in the oven before being removed; Glass transition temperature
(2) Post cure. The standard-cure sample was post-cured The Tg values as obtained from tan 6 peaks in torsion
at 175°C for 6h in a vacuum oven in the presence of are presented in Figure 1. The Tg values show good
nitrogen, after which the oven was switched off and the correspondence with crosslink density; the higher the
sample was allowed to come to RT in the oven. crosslink density, the higher the Tr This has been
Rectangular samples were cut from these sheets for discussed in greater detail in a previous publication 1.
torsional studies.
Macrodensity
Sub T o agein9 of the samples. Only the rectangular The density data are shown in Figure 2. The data show
samples cut from the cast sheet containing a stoichio- a trend opposite to that shown by the T, data (Figure
metric amount of curing agent (14.5 phr) were subjected
to sub-Tg ageing treatment in an oven at 130°C for 10,
100, 1000 and 5000min, respectively, in a nitrogen 170
atmosphere. After the ageing treatment, the samples were
allowed to come to RT in the oven at a cooling rate of
0.2°C min- 1.
150 -
Measurements
Glass transition temperature. The Tgs of the samples
were obtained from the peak positions of the tan 6 curves
obtained using a torsion pendulum. O
v 130--

Macrodensity. The density values of the cured samples

cut from cast sheets were determined in a carbon tetra-
chloride/n-heptane density gradient column (Davenport,
I10 B
UK) containing standard floats. Four small pieces of
each sample were introduced to the column and allowed
to settle for 24 h before readings were taken. In each case,
at least two readings were taken. The average macro-
90 I I I I
density could thus be estimated to the third decimal place. 0 0.5 1.0 1,5 2.0
P
Internal stress. In the resin-coated metal strip, internal
stress is primarily generated during the cooling of the Figure l Glass transition temperatures as obtained from the tan 6
cured resin which is attached to a substrate with a peak in torsion, for standard-cure ( O ) and post-cure (1~) samples
different thermal expansion coefficient and is otherwise
constrained from normal thermal contraction 2. This 1.222
results in a bending of the coated strip. The curvature
of the strip (Pc) was measured using a spherometer with
a least count of 0.01 mm and related to the internal stress,
ai, by the expression3: I .218 -

7
_--x--
E1 h3 2 x [1
O'i-- 12h2 pcH 3\H2,]I
(1)
1.214-
where E x is the Young's modulus of the aluminium strip,
h 1 and h E a r e the thicknesses of the aluminium plate and
the epoxy coating, respectively, and H = h 1 +h a. The
initial bending in the middle of the strip could be as high 1.210
as 5 mm which gradually reduced with time. The beam
deflection could be measured accurately using this
method. I ,20E I I i I
In each case, measurements were made on at least four 0 0.5 1.0 1.5 2,0
coated strips. The first measurement was made after P
cooling for 12 h in the oven. Further measurements were Figure 2 R o o m temperature density data obtained from a density
made on three subsequent days at 24 h intervals. gradient column. Symbols as in Figure 1

290 POLYMER, 1993, Volume 34, Number 2

 Stress in an epoxy resin. C. Brahatheeswaran and V. B. Gupta

1); the most highly crosslinked samples now show the slightly lower in all the post-cure samples compared to
lowest density. This has been attributed to the trapping that in the corresponding standard-cure samples.
of relatively more free volume by samples having higher
crosslink density 1. The origin of internal stress. There has been consider-
able interest in understanding the origin and level of
Internal stress stresses that develop in cured epoxy resin systems,
Experimental data. The internal stress generated in the particularly in the coating systems as the coating is cured
resin-coated metal strip is principally due to the difference and cooled. These stresses can have different origins and
in the expansion coefficients between the resin and the the various terms used in the literature to describe them
metal plate 2. Provided that the resin adheres well to the have been considered by White 7.
aluminium plate, most of the volume shrinkage will be Shimbo et al. 8 have presented work on thin coatings
expected to occur as a decrease in thickness with only a of epoxy on aluminium strips which shows that above
minor lateral contraction. The epoxy-aluminium inter- the Tg of the epoxy there are no stresses present because
face was examined on a scanning electron microscope in the rubber-like state the relaxation processes alleviate
and was found to be intact implying good adhesion any stresses that develop. Upon cooling to temperatures
between the resin and the substrate. below the Tg, the stresses increase because in the glassy
Another factor which may also contribute to the state epoxy contracts at a rate greater than that of
generation of internal stress in the laminate is the kinetic aluminium. The calculated stress, acalc., achieved at, say,
nature of the Tg5'6. This will be related mainly to the rate RT where RT<< Tg, can be obtained fromZ:
of cooling of the sample, from the Tg downwards; the
higher the rate of cooling, the further the sample would
be from its equilibrium state and the greater the internal
stress.
O'calc" =
f?
T
Ee(fle-flal)dT (2)

The internal stress data for the standard- and post-cure where E¢ is the elastic modulus of the epoxy resin and
samples are shown in Figure 3 as a function of fie and flal are the linear expansion coefficients of the
measurement time. The following points are noteworthy. epoxy resin and aluminium, respectively. The latter was
(1) Standard-cure samples. The highest internal stress taken to be 2.30× 10-SK -1.
is generated in the stoichiometric sample which has the The main assumption made in deriving this relationship
highest Tg (Figure 1) and the highest crosslink density 1 is that the contraction of the resin in the glassy region
amongst the standard-cure samples. The stress generated is completely restricted and that it is totally converted
in the off-stoichiometric samples is relatively less. to internal stress.
The stress relaxation occurring in the standard-cure Ochi et al. 9 embedded a steel ring in the epoxy resin
samples is not very significant over the period studied, so that as the resin cured and cooled, the internal stresses,
namely 84 h. which developed because contraction of the resin was
(2) Post-cure samples. On post curing of the standard- constrained by the metal ring, deformed the ring. From
cure samples, the internal stress shows a significant measuring this deformation the stress could be measured.
increase in the case of the 10 and 14.5 phr samples, which When the measured internal stress was plotted against
have the higher T~s (Figure 1) and the higher crosslink the internal stress calculated from equation (2), they
densities ~ amongst the post cure samples. In other found that the data fell on the line given by a~,lc. = 3.8trobs,
samples, there is relatively less increase of internal stress. where trobs was the experimentally measured internal
The stress relaxation in the 10 and 14.5 phr post-cure stress. They concluded that all the shrinkage stress was
samples is very rapid. After 84h the internal stress is not converted to internal stress. Paipetis 1° has also
clearly established that shrinkage has no stress-producing
effect except in a constrained system.
For the present epoxy resin system, both the thermal
ok,, expansion data and the torsional modulus data were
\\
\\ obtained only for the post-cure samples because, as
IO x\
shown elsewhere 11, in these samples, the curing is
\"\\\\ complete and, therefore, the sample chemistry does not
9 \'D.\ \ undergo any change during measurement. The volu-
"B metric thermal expansion coefficient data have been
published in an earlier publication 1 and the linear
expansion coefficient was taken to be one-third of the
~ 7
"N volumetric coefficient. The torsional modulus data are
shown in Figure 4. The value of E e in equation (2) can
~6 be calculated from G' by using an assumed value of 0.33
c
for Poisson's ratio, which gives E e = 2.66G'.
The values of internal stress for the post-cure samples
calculated using equation (2) along with the measured
4 values on the basis of equation (1) after 12 and 84h of
stress relaxation are shown in Table 1. There are two
5 I I I I I I I noteworthy observations. First, the calculated values are
0 20 40 60 80
considerably higher than the measured values. And
Time ( h )
second, the measured stress values show clearly that there
Figure 3 Internal stress data for standard-cure ( - - ) and post-cure is considerable stress relaxation with time, particularly
( - . ) samples: (r-l) 10.0; (O) 14.5; (•) 20.0; (O) 25.0phr in the most highly crosslinked samples, as was also clear

POLYMER, 1993, Volume 34, Number 2 291

Stress in an epoxy resin: C. Brahatheeswaran and V. B. Gupta

3.98
2.5
" ~5~..~ ..................... .

1.0

"~ 0.5
tl.
£9
== 0 . 2 5

E 0.1
i
g 0.05
o
I- 0.025 - :.~
i
Figure 5 Forces arising from the interaction between the two layers
0.01 of epoxy-coatedaluminium
0.005 -
0.004 I I I I ] .... I
- I O0 --60 --20 20 60 I00 140 180
Temperoture (*C) M in each strip 13, as shown in Figure 5. For equilibrium
Figure 4 Torsional modulus data for the post-cure samples: (--) of the cross-section:
10; ( - - ) 14.5; (-.) 20; (-..) 25phr
M A + M E = Pt (3)
where t is the thickness of each layer. The tensile force
on the epoxy layer will tend to enhance molecular
Table 1 Internal stress in post-cure resin samples mobility and result in higher segmental motion in the
epoxy layer.
Internal stress (MPa)
Measured b
Sample
Structural factors affecting stress relaxation. The epoxy
(phr) Calculate& 12 h 84 h layer is non-linearly viscoelastic. With the passage of
time, therefore, the stress measured is expected to relax.
10 18.12 8.2 5.2 However, the overall equilibrium always holds in the
14.5 15.86 10.8 7.2 entire specimen, i.e. the redistribution of stress in
20 11.95 6.1 5.0
25 10.40 5.2 4.4 aluminium and epoxy occurs continuously.
The highest internal stress and stress relaxation rate
aFrom equation (2) are observed in samples with the highest crosslink
bFrom equation (1) density. A number of factors may be responsible for this.
First, the higher the internal stress, the higher the
viscoelasticity and greater the non-linearity. Second, it
has been clearly demonstrated in an earlier publication 1
from Figure 3. It needs to be emphasized that the first that samples with high crosslink density or high Tg have
measurement was recorded 12 h after switching the oven also the highest specific volume and free volume below
off. A considerable amount of stress relaxation would the Tgand at RT they are the farthest from the equilibrium
already have occurred during this period. Considering position. Also, they have the lowest density amongst the
this, the general trend predicted by equation (2) appears samples of different stoichiometries (Figure 2). The high
to be reasonably good, thus suggesting that the shrinkage free volume will allow greater molecular mobility.
in the glassy region seems to be directly converted to Finally, as shown in the previous section, the high
internal stress because the motion of network segments internal stress in samples of high crosslink density will
is restricted in this region. In the present investigation, result in high tensile force on the epoxy layer. All these
the most highly crosslinked samples develop the highest factors will be expected to enhance the relaxation rate.
internal stress though their elastic moduli are the lowest. Ochi et al. showed 9 that in a cured epoxy resin system
Thus it would appear that the Tg plays a predominant the internal stress was ~ 6 MPa at RT and decreased to
role in determining the magnitude of internal stress; the zero at the Tg. The RT value is reasonably close to the
higher the Tg, the higher the internal stress. value of internal stress in the present investigation.
However, significant stress relaxation is observed at RT
A mechanistic interpretation for the bending of the strip. in the present samples.
The aluminium-epoxy laminate can be considered as a
bimetallic strip lz as shown in Figure 5. As it cools from
above the Tg to RT the interaction between the two strips The ~' relaxation
(which have been considered to be of equal thickness in In a previous publication x4, a broad, weak relaxation
the present discussion) produces a non-uniform force. As called the ~' peak in the same epoxy system was shown
shown in Figure 5, the deformation of the common to occur during torsional measurements in the 50-100°C
surface occurs in such a manner that the aluminium layer range, having an activation energy of ~ 210 kJ mol-~. It
is compressed and the epoxy layer is extended. If the was also seen that this relaxation was more prominent
internal force is P, it gives rise to a direct load P at the in the stoichiometric post-cure sample than in the
centre of each section together with a bending moment standard-cure sample.

292 POLYMER, 1993, Volume 34, Number 2

 Stress in an epoxy resin." C. Brahatheeswaran and V. B. Gupta

Small relaxation regions intermediate between the results in reduced segmental mobility. Thus the c( peak
and // relaxations have also been reported for poly- appears to be related to free volume or to the frozen-in
carbonate 15 and atactic polystyreneX 6. These relaxations stress. It may be pointed out that in the torsion pendulum
reduce in intensity or disappear on annealing and have studies, the frequency is ~ 1 Hz. The stress relaxation
therefore been attributed to the presence of non- has, on the other hand, been studied from 12 to 84h. At
equilibrium stresses frozen in these samples during longer times or lower frequencies, the transition may shift
preparation• Recently, after this work had been com- to lower temperatures and could be partially effective in
pleted, we have seen the work of Sasuga and Udagawa introducing molecular mobility at RT also.
who have observed 17 similar relaxation in several epoxy It must be emphasized that the origin of internal stress
resins and have attributed it to internal stress. in the cast sheets, from which the torsional specimens
The tan6 data in torsion for the stoichiometric have been machined, is different to that for the resin-
post-cure sample before and after sub-Tg ageing are coated plate. In the samples subjected to ageing, differen-
shown in Figure 6. The corresponding torsional modulus tial shrinkage is not relevant. Instead, the internal stress
data are shown in Figure 7. It is interesting to note that would primarily result from non-uniform cooling of the
the ~' peak becomes progressively flatter as the ageing sheetS'6'18; the shell solidifying first, with the core shrink-
time is increased. Also, the aged samples show high ing next. The shell is thus under compressive stresses,
torsional modulus. while tensile stresses are induced inside• In addition, the
It is generally agreed that on sub-Tg ageing, non- non-equilibrium stresses can also be trapped due to the
equilibrium stresses can relax through free volume kinetic nature of the glass transition phenomenon 5'6.
collapse 6. The reduction in free volume consequently

CONCLUSIONS
1.0
! The internal stress that develops as an aluminium strip
coated with a thin layer of epoxy resin is cured and cooled
0.5- I has been measured at RT and shown to be high for
samples having a high Tg. This is shown to be pre-
0.25 - dominantly due to the difference in the thermal contrac-
tion coefficients between the resin and aluminium which
results in the generation of these stresses from the Tg to
oo 0.f
-

t
i RT due to the constraints to lateral shrinkage of the
t- adhesive layer. The rate of stress relaxation is also high
0.05 -'~-" : i~l in the high T8 samples because they are the farthest from
their equilibrium state and consequently show a greater
degree of non-linear viscoelasticity.
0.025 I~
In the case of cast sheets, internal stresses arise mainly
due to non-uniform cooling. The 7' peak in the 50-100°C
range in the dynamic torsional spectrum of the neat
o.o, i
0.00794 I I I I I I I I stoichiometric bulk sample cut from the cast sheet is
-I00 -60 -20 20 60 I00 140 180 relatively more pronounced. When this sample is sub-
Temperature (*C) jected to sub-Tg ageing, the peak flattens. Thus the ~'
Figure 6 Tanf curves for post-cure and sub-T~ aged samples: ( - - )
peak in these samples could originate from the frozen-in
14.5phr; ( - - ) 100min; ( - . ) 1000min; ( - - . ) 5000min internal stresses which relax on sub-Tg ageing due to free
volume collapse.

3.98
2.5 ACKNOWLEDGEMENT
The authors are grateful to Professor R. K. Mittal of the
1.0 Department of Applied Mechanics for valuable discus-
sions relating to the mechanics of the bending of epoxy
n
o 0•5 resin coated aluminium strips•

0.25
O
E REFERENCES
O
E 1 Gupta, V. B. and Brahatheeswaran, C. Polymer 1991,32, 1875
o 0.1
2 King, D. and Bell, J. P. J. Adhesion 1988, 26, 37
3 Nakamura, Y., Tabata, H., Suzuki, H., Iko, K., Okubo, M. and
# 0.05 Matsumoto, T. J. Appl. Polyrn. Sci. 1986, 32, 4865
[ 4 Pandit, S. N. and Gupta, V. B. Polymer Comp. 1981, 2, 121
0.025 5 Struik, L. C. E. Polym. Eng. Sci. 1978, 18, 799
6 Struik, L. C. E. 'Physical Ageing in Amorphous Polymers and
0.0158 I I I I I I Other Materials', Elsevier, New York, 1978, p. 7
-I00 -60 --20 20 60 I00 140 180 7 White, J. R. Polymer Testing 1984, 4, 165
T e m p e r o t u r e (*C) 8 Shimbo, M., Ochi, M., Inamura, T. and Inoue, M. J. Mater.
Sci. 1985, 20, 2965
Figure 7 Torsional modulus data for post-cure and sub-Tg aged 9 Ochi, M., Yamashita, K., Yoshizumi, M. and Shimbo, M.
samples. Symbols as in Figure 6 J. Appl. Polym. Sci. 1989, 38, 789

POLYMER, 1993, Volume 34, Number 2 293

Stress in an epoxy resin: C. Brahatheeswaran and V. B. Gupta

10 Paipetis, S. A. Colloid Polym. Sci. 1979, 257, 934 15 McCrum, N. G., Read, B. C. and Williams, G. 'Anelastic and
11 Schiering, D. W., Katon, J. E., Drzal, L. T. and Gupta, V. B. Dielectric Effects in Polymeric Solids', John Wiley & Sons,
J. Appl. Polym. Sci. 1987, 34, 2367 London, 1967, p. 536
12 Timoshenko, S. J. Opt. Soc. Am. 1925, 11,233 16 Petrie, S. E. B. in 'Physical Structure of the Amorphous State'
13 Ryder, G. H. 'Strength of Materials', B.I. Publications, Bombay, (Eds G. Allen and S. E. B. Petrie), Marcel Dekker Inc., New
1961, p. 106 York, 1976, p. 240
14 Gupta, V. B., Drzal, L. T., Lee, C. Y.-C. and Rich, M. J. 17 Sasuga, T. and Udagawa, A. Polymer 1991, 32, 402
J. Macromol. Sci. Phys. 1984-1985, B23, 435 18 Wintergrest, S. Kunststoffe 1973, 63, 636

294 POLYMER, 1993, Volume 34, Number 2