JOURNAL OF

COMPOSITE
Article M AT E R I A L S
Journal of Composite Materials
0(0) 1–13
! The Author(s) 2015
DGEBF epoxy blends for use in the resin Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
impregnation of extremely large DOI: 10.1177/0021998314568332
jcm.sagepub.com
composite parts

Madhu S Madhukar1 and Nicolai N Martovetsky2

Abstract
Large superconducting electromagnets used in fusion reactors utilize a large amount of glass/epoxy composite for
electrical insulation and mechanical and thermal strengths. The manufacture of these magnets involves wrapping each
superconducting cable bundle with dry glass cloth followed by the vacuum-assisted resin transfer molding of the entire
magnet. Due to their enormous size (more than 100 tons), it requires more than 40 h for resin impregnation and the
subsequent pressure cycles to ensure complete impregnation and removal of any trapped air pockets. Diglycidyl ether of
bisphenol F epoxy resin cross-linked with methyltetrahydrophthalic anhydride with an accelerator has been shown to be
a good candidate for use in composite parts requiring long impregnation cycles. Viscosity, gel time, and glass transition
temperature of four resin-blends of diglycidyl ether of bisphenol F resin system were monitored as a function of time and
temperature with an objective to find the blend that provides a working window longer than 40 h at low viscosity without
lowering its glass transition temperature. Based on the results, a resin-blend in the weight ratios of resin:hardener:accel-
erator ¼ 100:82:0.125 is shown to provide more than 60 h at low resin viscosity while maintaining the same glass
transition temperature as obtained with previously used resin-blends.

Keywords
Composites, thermosets, viscosity, glass transition, cure cycle

of the International Thermonuclear Experimental

Introduction
Reactor (ITER).1,14 The CS is a system of six large
The reactivity of blends of diglycidyl ether of bisphenol electromagnet coil modules, each module having an
F (DGEBF) epoxy resin can be altered by varying the outer radius of 2.1 m, inner radius of 1.2 m, height of
proportions of accelerator in relation to the resin and 2.2 m, and weighing about 112 tons each.14 The insula-
hardener.1 In addition, it has been shown that the blend tion material used in these coils is woven glass/epoxy
of epoxy containing fluorenyl structure in the backbone composite.8 Due to the extremely large size of the CS
improves the thermal stability, mechanical property, modules, the resin impregnation into the dry woven
and moisture resistance of the cured epoxy resins.2–7 glass fiber cloth and the subsequent milking cycles
Such resin-blends have been used as a matrix in com- (the term ‘‘milking’’ is used by some composite manu-
posites to study the their properties at cryogenic tem- facturers to describe a sequence of vacuum-pressure
peratures,8 thermal and electrical properties,9 damping cycles that are applied after the initial resin transfer
properties,10,11 mechanical behavior, and fracture into the dry fibers using the vacuum-assisted resin
toughness.12,13 One such blend of epoxy made of transfer to ensure that there is a complete wetting of
DGEBF (commercial designation – Ciba Geigy
1
GY282) cross-linked with slow-reacting anhydride The University of Tennessee, Knoxville, TN, USA
2
hardener methyltetrahydrophthalic anhydride US ITER, Oak Ridge National Laboratory, Oak Ridge, TN, USA
(MTHPA; commercial designation – Ciba Geigy HY
Corresponding author:
918) is being considered as a candidate for the insula- Madhu S Madhukar, The University of Tennessee, 316 Perkins Hall,
tion impregnation material as a woven glass/epoxy Knoxville, TN 37922, USA.
composite for the central solenoid (CS) coil modules Email: [email protected]

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
2 Journal of Composite Materials 0(0)

fibers and that any trapped air pockets are driven out of The four resin-blends studied in this investigation
the part) is expected to take about 40 h. are listed in Table 1.
This study was conducted to help develop most suit-
able resin-blend and resin transfer parameters (such as
viscosity, temperature, and duration of resin transfer) Experimental procedures
for the vacuum-assisted resin transfer molding
(VaRTM) of glass/epoxy composite insulation in the
Viscosity measurements
CS coil modules to ensure that resin viscosity remains Viscosity was measured using Hydramotion Viscolite
low during the duration (  40 h) of resin transfer and Model 700 viscometer, Figure 1(a). It utilizes a solid
the subsequent milking cycles. For a good resin impreg- stainless steel element that vibrates at high frequency
nation of extremely large parts (such as CS coil mod- while being immersed in a fluid. As the surface of the
ules), it is important for the resin-blend to have sufficient sensor shears through the liquid, energy is lost to the
pot life, and it should maintain low viscosity ( < 100 cP, fluid because of its viscosity. The dissipated energy is
cP ¼ centi poise ¼ 0.01 poise) throughout the resin trans- measured and is related to the fluid viscosity. The equip-
fer process. Viscosity control is also important to design ment is capable of measuring viscosity in the range of
the vacuum-pressure cycles during the cure to achieve 0–20,000 cP. However, it was not used beyond 800 cP
good part consolidation. Once a thermosetting resin is because as the polymer thickens, it tends to stick to the
mixed with a hardener, the cross-linking chemical reac- rod making it difficult to clean. Prior to using the visc-
tions begin and the resin viscosity begins to increase.15–22 ometer for the actual viscosity measurements, the accur-
The viscosity of the mixture can be lowered slightly by acy of the equipment was verified by measuring the
increasing the temperature of the mixture. However, the viscosity of water at room temperature (20  C).
cross-linking reactions also accelerate with the increase In the viscosity monitoring setup, the resin-blend
in the temperature resulting in a faster increase of the was heated in a metal cylindrical container (about
viscosity.15–22 Thus, while the resin transfer at higher 70 mm diameter and 100 mm long) by means of a flex-
temperatures may be desirable due to lower viscosity ible heat tape tightly wrapped around it. To prevent
of the resin-blend, the time window needed for the com- heat loss, fiber glass tape was wrapped around the
plete impregnation may not be enough. heat tape and secured in its place by again tightly wrap-
In this study, the viscosity, gel time, and glass tran- ping sticky Kapton tape over it. The metal cylindrical
sition temperature characteristics of blends of epoxy container was then inserted into another plastic cylin-
made of DGEBF cross-linked with slow-reacting der. This plastic cylinder was needed so that the visc-
anhydride hardener MTHPA in the presence of small ometer can be simply rested on it during the
amounts of the accelerator (designated as DY 073-1) experiment, Figure 1(b). To further minimize the heat
were investigated in order to find the most suitable loss, the plastic cylinder was inserted into a cylindrical
blend that will maintain low viscosity without lowering cavity of a foam insulation block. The photograph of
the glass transition temperature of cured samples. In an the complete setup is shown in Figure 1(c). The accur-
earlier study,1 the same resin system was characterized acy of the temperature measurement was  1  C. To
for different accelerator amounts – 0.25, 0.5, 1.0, and begin the experiment, the resin-blend was poured into
1.5 pbw (parts by weight). Based on the results, a resin- the metal cylindrical container such that when the visc-
blend containing 0.25 pbw of accelerator was recom- ometer is rested on the plastic cylinder, the metal probe
mended. However, such a resin-blend barely provides a at the end of the viscometer is completely immersed in
working window of 40 h for a resin transfer at 60  C. In
this study, the amount of the accelerator added to the
epoxy blend (100 pbw DGEBF, 82 pbw MTHPA) is Table 1. The four resin-blends studied in this study.
further reduced to values of 0.25 (control), 0.125,
Resin Hardener Accelerator
0.0625, and 0 pbw. The effects of the amount of the
DGEBF MTHPA DY 073-1 Designation of
accelerator on viscosity, gel time, and glass transition (pbw) (pbw) (pbw) the resin-blend
temperature are determined. From the results obtained,
optimum resin transfer parameters are proposed that 100 82 0.25 Standard
provide ample time for the resin impregnation into R:H:A ¼ 100:82:0.25
the dry woven glass cloth insulation while keeping the 100 82 0.125 R:H:A ¼ 100:82:0.125
resin viscosity below 100 cP. While these processing 100 82 0.0625 R:H:A ¼ 100:82:0.0625
parameters were determined for the manufacture of 100 82 0 R:H:A ¼ 100:82:0
CS coils, the findings of this research are applicable
DGEBF: diglycidyl ether of bisphenol F; MTHPA: methyltetrahydrophtha-
to the resin transfer molding of any composite part lic anhydride; R:H:A: resin:hardener:accelerator; pbw: parts by weight.
that requires extremely long resin transfer cycles.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 3

Figure 1. Viscosity measurement setup.

the resin-blend. The temperature of the blend was con-

trolled by a thermocouple placed inside the liquid and
positioned close to the probe. The temperature and vis-
cosity data were taken at regular intervals (once every
10 min) by a camera that was programmed to take pic-
tures of the readout windows of the viscometer and the
temperature controller at predefined time intervals. The
camera software stores these pictures as image files in a
computer and also sends an email attachment to the
operator. This way the experiment could be tracked
remotely, which was quite a useful feature since the
viscosity experiments were quite long (several days).
The temperature and viscosity values were later read
from the stored image files.

Gel time measurements

The gel point in a thermosetting resin is defined as a
point during its cure cycle when a three-dimensional
network of polymer chains is created. As a resin-
blend approaches gel point, its viscosity suddenly
increases and it loses its mobility. Equipment used for
gel time measurement was Sunshine Gel-Time Meter. It
contains a slowly rotating (about 2 rpm) glass spindle
immersed in a resin-blend, Figure 2. While the spindle
rotates, a digital counter records the time. As the resin-
blend approaches gel point, it thickens and the spindle
stops rotating triggering the time-counter to stop. Figure 2. Gel time tester set-up.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
4 Journal of Composite Materials 0(0)

Note that this instrument does not measure the actual much larger compared to that of the accelerator, slight
viscosity of the polymer. Instead it only tracks the time errors in their weight measurements would not impact
at which the viscosity begins to increase rapidly. the results significantly. The resin-hardener-accelerator
Some modifications were done to the recommended mixture was mixed at room temperature, stirring it by
procedure for the gelation experiments. The first was hand for about 20 min with a wooden tongue depressor
the use of a mineral oil (flash point ¼ 188  C) instead of followed by another stirring by a magnetic stirrer for
water as the heating medium. Although the equipment another 20 min. The mixed solution appeared homoge-
had a condenser to minimize the evaporation of water, neous and without any streaks. No degassing was done
for long duration experiments some water did evapor- of the mixed resin-blend.
ate requiring a periodic addition of water to the jar.
With the use of mineral oil, there was no risk of mineral
oil level dropping and no need to use the condenser. Results and discussion
The second modification was the use of a magnetic stir-
Viscosity of different resin-blends
rer to keep the temperature of the mineral oil uniform.
For the viscosity measurement, the resin-blend was
poured into the metal cylindrical container and the visc-
Glass transition temperature measurements ometer probe was lowered into the liquid. The blend was
The glass transition temperature of samples was mea- heated to the test temperature. When the blend reached
sured with TA Instruments Q-1000 DSC. Pieces of sam- the test temperature, the viscosity measurements began.
ples weighing 10–12 mg cut from cured (and partially In all cases, the blend reached the test temperature from
cured) discs (explained in Section ‘‘Glass transition the room temperature within about 10 min. Figure 3
temperature and cure cycle modification’’ below) shows the very first viscosity measurements (initial visc-
using clean razor blades were placed in a hermetic alu- osities) for the four resin-blends. Note that in this figure,
minum pans and heated to various test temperature at a it is not the same sample for each blend that is heated
heating rate of 10  C/min followed by a 5-min hold at from room temperature to all the way up to 70  C.
the test temperature. The temperature hold was neces- Instead, for each resin-blend, three separate samples
sary to equilibrate the sample at the test temperature. were heated to the three test temperatures (50  C, 60  C,
The glass transition temperature (Tg) was defined as the and 70  C) and their initial viscosities measured. The
middle point of the thermal transition regions of the change in the viscosity as a function of temperature (as
heat flux versus temperature plots. shown in Figure 3) includes both the thermal effect as well
as chemical effect. An increase in temperature will cause
the viscosity to decrease while at the same time the chem-
Resin-blend preparation method ical cross-linking reactions may cause the viscosity to
The resin-blends described in Table 1 were prepared as increase. However, for this slow reacting resin system,
follows. The accelerator was added to an empty and the amount of cross-linking will be insignificant consider-
clean glass beaker first. The weight of the accelerator ing that it took less than 10 min to heat the resin-blend to
was monitored by a laboratory balance scale (Radwag its desired temperatures (50  C, 60  C, and 70  C).
Intell-Lab Balance, Model XA 60/220) having a reso- The key observations from Figure 3 are (1) within
lution of 1 mg. Note that the total mixed amount of the the experimental scatter, there does not seem to be
blend needed for each of the viscosity experiment was much difference in the initial viscosities of the four
only about 200 g, but a larger amount (about 500– resin-blends and (2) the viscosity decreases significantly
600 g) was mixed to decrease the percentage weighing as the temperature is increased.
error of the constituents. Consequently, to prepare The lower the viscosity, the better is the resin
about 500–600 g of the blend, the amount of the accel- impregnation in fibrous composites; however, for this
erator needed for each of the four blends (shown in study and for comparison purposes, we set the viscosity
Table 1) determined. Note that the accelerator upper limit for the resin impregnation to be 100 cP (to
amount was very small (only 0.125 g for R:H:A ratio put this number in perspective, the viscosity of olive oil
of 100:82:0.125 and only 0.0625 g for R:H:A ratio of at room temperature is about 80 cP). Thus, while we
100:82:0.0625); therefore, even a small error in the only needed to know how long at a given temperature
accelerator content can alter the results. To minimize the resin viscosity would remain below 100 cP, the vis-
this error, the accelerator was weighed in a clean beaker cosity data were, however, collected until about 800 cP.
first. Knowing the exact weight of the accelerator, the To first verify the repeatability of the measurements,
resin and hardener were added in their correct weight viscosity measurements were made under identical con-
proportions according to the ratios given in Table 1. ditions on resin-blends prepared at different times.
Since the weight proportions of resin and hardener are Figure 4 shows some examples of data reproducibility.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 5

Figure 3. Initial viscosity of the standard resin mixture as a function of temperature.

R:H:A: resin:hardener:accelerator.

Figure 4. Viscosity development versus time for different isotherms and different resin-blends showing the repeatability of the
measurements. The beginning parts of the curves show good reproducibility, but as the time progresses the viscosity versus time
curves diverge.
R:H:A: resin:hardener:accelerator.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
6 Journal of Composite Materials 0(0)

Figure 5. Viscosity development at three isotherms for the standard resin-blend (R:H:A ¼ 100:82:0.25). Although the viscosity
increases slowly at lower isotherm, the initial viscosity at higher isotherm is still lower until about 40–50 h.
R:H:A: resin:hardener:accelerator.

In most cases, the data of beginning part (up to about accelerator amount slows the viscosity development
100 cP) are quite reproducible. However, as time pro- rate suggesting that the reactivity of the resin system
gresses, the results diverge. Considering the very long decreases with a decrease in the accelerator amount.
duration of these experiments, even a small difference in 2. As expected, the viscosity develops faster at higher
temperature (of the order of 1–2  C) from sample to temperature.
sample can cause some data divergence among samples. 3. While for the most part, the viscosity at a given time
Figures 5 to 8 show the viscosity changes for the four is higher at a higher temperature, the beginning parts
resin-blends at 50  C, 60  C, and 70  C isotherms. of these curve show that the viscosity at the higher
There are four key observations made from these temperature is lower than that at the lower tempera-
figures: tures and that it stays low for quite some time. This
could be an important information in support of the
1. The accelerator content has a strong effect on the resin transfer at a higher temperature if the transfer
rate of viscosity increase. A reduction in the can be completed within the low-viscosity time

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 7

Figure 6. Viscosity development at three isotherms for the resin-blend (R:H:A ¼ 100:82:0.125). Although the viscosity increases
slowly at lower isotherm, the initial viscosity at higher isotherm is still lower until about 50–70 h.
R:H:A: resin:hardener:accelerator.

window. For example, for the standard resin-blend data are summarized in Table 2. From these data, it
(R:H:A ¼ 100:82:0.25), a resin transfer at 70  C can be seen that except for R:H:A ¼ 100:82:0 (i.e. the
would provide about 25 h before the resin viscosity blend with no accelerator), all other blends provide
exceeds that at 50  C. Similarly, if resin- the longest working window at 60  C isotherm.
blend R:H:A ¼ 100:82:0 is used, the resin transfer
at 70  C would provide a much longer time
window (> 100 h).
4. If the viscosity of resin-blend is to be kept below 100
Gel time
cP while transferring the resin to a mold during a
vacuum pressure impregnation (VPI) process, it is During the cure of a polymer composite, the cure cycle
important to know how long it takes before the vis- should be chosen such that the gelation is made
cosity exceeds this value at the each isotherm. These to occur at a temperature no higher than necessary.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
8 Journal of Composite Materials 0(0)

Figure 7. Viscosity development at three isotherms for the resin-blend (R:H:A ¼ 100:82:0.0625). Although the viscosity increases
slowly at lower isotherm, the initial viscosity at higher isotherm is still lower until about 70–120 h.
R:H:A: resin:hardener:accelerator.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 9

Figure 8. Viscosity development at three isotherms for the resin-blend (R:H:A ¼ 100:82:0.0). Although the viscosity increases slowly
at lower isotherm, the initial viscosity at higher isotherm is still lower until about 160–200 h.
R:H:A: resin:hardener:accelerator.

Table 2. Number of hours before the viscosity of resin-blends reaches 100 cP.

Time to reach a viscosity of 100 cP (h)

Temperature R:H:A ¼ 100:82:0.25 R:H:A ¼ 100:82:0.125 R:H:A ¼ 100:82:0.0625 R:H:A ¼ 100:82:0

50  C 20 32 49 54
60  C 40 63 88 106
70  C 27 46 68 112
R:H:A: resin:hardener:accelerator.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
10 Journal of Composite Materials 0(0)

Figure 9. The cure cycle used for the standard resin-blend (R:H:A ¼ 100:82:0.25).

The reason is that gelation is accompanied by resin Time to gel (tgel)

shrinkage and a rapid decrease in polymer’s mobility.
90oC
Temperature

When used in in fibrous composites, the resin shrinkage

produces cure-induced stresses.23–27 If gelation is made 50oC for 2 hours Ramp Rate = 10oC/hr
to occur at lower temperatures, the gelation process is
slower, and as a result, a stress relaxation mechanism
removes majority of the cure-induced stresses.25
Time (hrs)
However, if the gelation is allowed to occur at higher
temperatures, the stresses may remain locked in the Figure 10. The 90  C hold period of the standard cure was
structure and may cause part distortion when it is extended until the gelation of a resin-blend occurred.
taken out of the mold.26,27
In the standard resin-blend (R:H:A ¼ 100:82:0.25),
the standard cure cycle (Figure 9) is designed so that
the gelation occurs at 90  C.1 For other resin-blends Table 3. Gel time measured for different resin-blends.
(shown in Table 1), the gelation temperature is still
Gel time, tgel (h)
chosen to be 90  C.
The cure cycles chosen for gel time measurement Resin-blend
experiments for all the four resin-blends were similar
to the one shown in Figure 9 except that the 90  C R:H:A ¼ R:H:A ¼ R:H:A ¼ R:H:A ¼
hold period was continued until the gelation (see Test # 100:82:0.25 100:82:0.125 100:82:0.0625 100:82:0
Figure 10), i.e. all resin-blends were held at 50  C for 1 7.1 11.7 26.1 202.3
2 h followed by 4-h ramp to 90  C followed by a hold at 2 7.1 12.2 23.6 200.4
90  C until the gelation. Table 3 shows the gel times 3 6.6 12.1 23.8 –
(tgel, as defined in Figure 10) for the four resin-blends.
Average 6.9 12 24.5 201.4
The data shown in Table 3 are plotted in Figure 11.
As expected, a decrease in resin-blend’s reactivity with a R:H:A: resin:hardener:accelerator.
decrease in accelerator content is again demonstrated
by an increase in the gel time. takes for the gelation may be impractically too long for
most applications. Therefore, the glass transition and
Glass transition temperature and cure cycle cure cycle modification studies were mostly focused on
the blends in the ratios of 100:82:0.125 and
modification
100:82:0.0625.
Based on the viscosity and gel time data on the four As we have shown earlier, a change in accelerator
resin-blends, it can be seen that the resin-blend with no content in the resin-blend alters the viscosity develop-
accelerator is a very slow reacting system and the time it ment characteristics and also shifts the time at which

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 11

Figure 12. The general cure cycle used for the curing samples
of different resin-blends.

Figure 11. The effect of accelerator content on the time to gel

at 90  C.
pbw: parts by weight.

gel occurs at 90  C. Thus, the cure cycle for each resin-

blend will be modified using the following three criteria:

1. The part of the cure cycle prior to the beginning of

the 90  C hold was kept the same for all the resin-
blends.
2. The duration of the 90  C hold for a given resin-
blend was the total time to achieve tgel (see
Table 3) plus some extra time te (see Figure 12)
added in the same proportion as was in the standard
resin-blend.
3. The duration of 128  C was extended until the glass
transition temperature reached the target tempera-
ture (112  C), which was the average glass transition
temperature obtained from the cured samples of
standard resin-blend cured according to the cure
schedule given in Figure 9.

The general cure cycle used for the different resin- Figure 13. Resin-blend samples cured in small aluminum pans
blends is shown in Figure 12. In order to determine and taken out at definite time intervals for Tg measurements.
the cure state at 128  C, resin-blend samples were
mixed (as explained earlier) and cured in several small
aluminum pans and placed in programmable ovens,
Figure 13. Once at 128  C, the samples were taken
out sequentially at regular time intervals (t1, t2, t3, . . .)
and the glass transition temperatures were determined.
Since the gel time for 100:82:0.125 resin-blend was
found to be 12 h (see Table 3), the total time at 90  C
hold was 12 h plus an additional time (about 3 h) that
was in the same proportion as that in the standard cure
cycle. After that, the temperature was raised to 128  C
and the samples were taken out at regular time intervals
(t1, t2, t3, . . ..tn in Figure 12) and their Tg measured.
Figure 14 shows how Tg develops as a function of soak
time at 128  C for the 100:82:0.125 resin-blend. It is
seen that for this particular blend, the Tg approached Figure 14. Change in the Tg with elapsed time at 128  C for
its highest value (111  C–112  C) within about 15–20 h resin-blend R:H:A ¼ 100:82:0.125.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
12 Journal of Composite Materials 0(0)

that the 90  C hold period was extended to account

for the longer gel time needed (see Table 3). As can
be seen that the Tg value of this resin-blend never
reached the target temperature of 112  C even after
being kept for more than 250 h at 128  C, Figure 16.
The development of the Tg as a function of soak at
128  C for the resin-blend 100:82:0 (i.e. no accelerator)
is shown in Figure 16. As can be seen that even after
about 1440 h (60 days) at 128  C, the Tg did not even
reach 90  C. It may be possible that additional hours
(or days) at 128  C will bring the sample Tg closer to the
target Tg ( ¼ 112  C); however, the tests were discontin-
ued because a cure cycle that is so long (more than 60
Figure 15. Change in the Tg with elapsed time at 128  C for days) would be of little practical use.
resin-blend R:H:A ¼ 100:82:0.0625. Thus, from the viscosity, gel time, and the glass tran-
sition temperature data, the resin-blend most suitable
for long duration resin transfer process is
R:H:A ¼ 100:82:0.125. This blend maintains the low
viscosity (< 100 cP) at 60  C for about 60 h. This pro-
posed cure cycle, Figure 17, meets the requirement of
yielding the glass transition temperature equal to that
of the standard resin-blend cured using the standard
cure cycle.
Also note that the suggestion made here to use the
R:H:A ¼ 100:82:0.125 resin-blend is based only on the
viscosity, gel time, and Tg measurements. There may be
other factors and properties such as resin density, stiff-
ness, thermal expansion coefficient, composite proper-
ties, etc. that may have to be determined to make sure
that all the requirements are met.
Figure 16. Change in the Tg with elapsed time at 128  C for
resin-blend R:H:A ¼ 100:82:0.0. The heating schedule prior to Summary and conclusions
reaching 128  C was as shown in Figure 12. The Tg stayed below
the target value even after spending 1440 h (60 days) at 128  C. Curing of extremely large composite parts requires the
resin-blend to maintain low viscosity during the entire
resin transfer process. One such resin-blend – DGEBF
cross-linked with MTHPA in presence of small amount
of and accelerator was studied for this purpose. The
effects of the weight percent of the accelerator on the
viscosity, gel time, and glass transition temperature of
the blends were studied.
The results show that the viscosity development of
the resin-blend is strongly influenced by the accelerator
content – as the accelerator weight percentage was
decreased from 0.25 to 0 pbw, the viscosity of the
resin-blend developed at a much slower rate and it
Figure 17. The recommended cure cycle to be used for the took significantly longer to gel. From the measured
resin-blend R:H:A ¼ 100:82:0.125. gel times, the cure cycle for each of the resin-blend
was modified so as to attain the same glass transition
of soak at 128  C. Additional soak at 128  C did not temperature (& 112  C) as that obtained for the stand-
produce any more increase in Tg suggesting that the ard resin-blend.
polymer blend is completely cured. For a better resin impregnation of a composite part
Similar results for the 100:82:0.0625 blend are shown during the resin transfer in a VaRTM process while
in Figure 15. The cure cycle used for this resin-blend simultaneously increasing the pot life, a resin-blend
was similar to the one used for the previous case except R:H:A ¼ 100:82:0.125 and the resin transfer at 60  C

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016
Madhukar and Martovetsky 13

(instead of currently used practice of resin transfer at nanocomposites as core materials. J Compos Mater
50  C) are found to be the most optimum. These con- 2011; 45: 2157–2164.
ditions offer low resin viscosity ( < 100 cP) and a longer 12. Sun L, Jia Y, Ma F, et al. Mechanical behavior and fail-
pot life (  60 h). From the gel time and the glass tran- ure mode of unidirectional fiber composites at low strain
rate level. J Compos Mater 2009; 43: 2623–2637.
sition temperature data, a new cure cycle is proposed.
13. Thakre PR, Lagoudas DC, Riddick JC, et al.
This cure cycle meets the requirement of yielding the Investigation of the effect of single wall carbon nanotubes
glass transition temperature equal to that of the stand- on interlaminar fracture toughness of woven carbon
ard resin-blend cured using the standard cure cycle. fiber—epoxy composites. J Compos Mater 2011; 45:
However, it should be noted that meeting the Tg 1091–1107.
requirement does not necessarily mean that all the 14. Schultz JH, Antaya T, Feng J, et al. The ITER central
other properties such as strength and stiffness will solenoid. In: Twenty-first IEEE/NPS symposium
remain unchanged. on fusion engineering. Knoxville, TN, September 2005,
pp. 1–4.
Conflict of interest 15. Brahmbhatt P, Unnikrishnan J, Sudha JD, et al. Cure
kinetics studies of cyanate ester and bisphenol-f epoxy
None declared.
blend. J Appl Polym Sci 2012; 125: 1068–1076.
16. Karkanas PI and Partridge IK. Cure modeling and moni-
Funding toring of epoxy/amine resin systems. ii. Network forma-
The financial support for this work was provided by US tion and chemoviscosity modeling. J Appl Polym Sci
ITER through grant # 4000090977. 2000; 77: 2178–2188.
17. Lane JW, Khattak RK and Dusi MR. Evaluation of
References dielectric cure models for an epoxy-amine resin system.
Polym Eng Sci 1989; 29: 339–346.
1. Reed RP, Evans D and Fabian PE. Development of a
18. Teyssandier F, Ivanković M and Love BJ. Modeling the
new resin system for the U.S. ITER central solenoid
effect of the curing conversion on the dynamic viscosity
model coil. Adv Cryog Eng (Mater) 2000; 46: 227–234.
of epoxy resins cured by an anhydride curing agent.
2. Chen CS, Bulkin BJ and Pearce EM. New epoxy resins.
J Appl Polym Sci 2010; 115: 1671–1674.
II. The preparation, characterization, and curing of
19. Roller MB. Rheology of curing thermosets: a review.
epoxy resins and their copolymers. J Appl Polym Sci
Polym Eng Sci 1986; 26: 432–440.
1982; 27: 3289–3312.
20. Mijovic J and Lee CH. A comparison of chemorheologi-
3. Chen CS and Pearce EM. The boron trifluoride mono-
cal models for thermoset cure. J Appl Polym Sci 1989; 38:
ethyl amine complex cured epoxy resins. J Appl Polym Sci
2155–2170.
1989; 37: 1105–1124.
21. Lee CL and Wei KH. Curing kinetics and viscosity
4. Liu W, Qiu Q, Wang J, et al. Curing kinetics and proper-
change of a two-part epoxy resin during mold filling in
ties of epoxy resin–fluorenyl diamine systems. Polymer
resin-transfer molding process. J Appl Polym Sci 2000;
2008; 49: 4399–4405.
5. Liu W, Qiu Q, Wang J, et al. Preparation and properties 77: 2139–2148.
of one epoxy system bearing fluorene moieties. J Appl 22. Tajima YA. Monitoring cure viscosity of epoxy compos-
Polym Sci 2009; 113: 1289–1297. ite. Polym Compos 1982; 3: 162–169.
6. Xiong Y, Liu H, Ou E, et al. Crystal structure, curing 23. Sham ML and Kim JK. Curing behavior and residual
kinetics, and thermal properties of bisphenol fluorene stresses in polymeric resins used for encapsulanting elec-
epoxy resin. J Appl Polym Sci 2010; 118: 827–833. tronic packages. J Appl Polym Sci 2005; 96: 175–182.
7. Dai Z, Li Y, Yang S, et al. Kinetics and thermal proper- 24. Genidy MS, Madhukar MS and Russell JD. An investi-
ties of epoxy resins based on bisphenol fluorene structure. gation of cure-induced stresses in low cure temperature
Eur Polym J 2009; 45: 1941–1948. thermoset polymer composites. J Reinf Plast Compos
8. Madhukar MS, Kidwell C, Reed D, et al. The effects of 1999; 18: 1304–1321.
glass weave on the low temperature flexure properties of 25. Genidy MS, Madhukar MS and Russell JD. A new
composite laminates. J Compos Mater 2013; 47: method to reduce cure-induced stresses in thermoset
1117–1127. polymer composites, part I: test method. J Compos
9. Brancato V, Visco AM, Pistone A, et al. Effect of func- Mater 2000; 34: 1882–1904.
tional groups of multi-walled carbon nanotubes on the 26. Madhukar MS, Genidy MS and Russell JD. A new
mechanical, thermal and electrical performance of epoxy method to reduce cure-induced stresses in thermoset
polymer composites, part II: closed loop feedback control
resin based nanocomposites. J Compos Mater 2013; 47:
3091–3103. system. J Compos Mater 2000; 34: 1905–1925.
10. Liu A, Wang KW and Bakis CE. Multiscale damping 27. Russell JD, Madhukar MS, Genidy MS, et al. A new
model for polymeric composites containing carbon nano- method to reduce cure-induced stresses in thermoset
tube ropes. J Compos Mater 2010; 44: 2301–2323. polymer composites, part III: correlating stress history
11. Tsai JL and Chang NR. Investigating damping properties to viscosity, degree of cure, and cure shrinkage.
of nanocomposites and sandwich structures with J Compos Mater 2000; 34: 1926–1947.

Downloaded from jcm.sagepub.com at OAK RIDGE NATIONAL LABORATORY on June 20, 2016