Available online at www.sciencedirect.

com

Composites: Part A 39 (2008) 540554

www.elsevier.com/locate/compositesa

Interlaminar shear strength of glass ber reinforced epoxy

composites enhanced with multi-walled carbon nanotubes
Zhihang Fan, Michael H. Santare, Suresh G. Advani *
Department of Mechanical Engineering and Center for Composite Materials, University of Delaware, Newark, DE 19716, USA

Received 8 May 2007; received in revised form 20 November 2007; accepted 23 November 2007

Abstract

In this study, we examine the interlaminar shear strength (ILSS) of traditional glass ber reinforced epoxy composites enhanced by
strategically injecting multi-walled carbon nanotube (MWNT)epoxy suspensions into stationary glass ber mats. The suspensions were
prepared by combining the techniques of high-speed mechanical stirring, ultrasonic agitation and acid oxidation. Two types of process
designs were introduced to fabricate the hybrid MWNT/glass/epoxy composites and their relative merits were discussed. Short beam
shear (SBS) and compression shear tests (CST) were conducted on the manufactured components to characterize the inuence of the
process and the weight percentage of nanotubes on the ILSS. The results show that the introduction of MWNT into the composite
increased the ILSS by up to 33%. The preferential orientation of the MWNTs in the thickness direction was found to contribute to
the increase in the interlaminar shear properties.
2007 Elsevier Ltd. All rights reserved.

Keywords: A. Preform; A. Nano structures; B. Interface/interphase; B. Mechanical properties; Capillary ow

1. Introduction Many researcher have explored other ways to improve

the ILSS by using nanoparticles and carbon nanotubes
While laminated ber reinforced polymermatrix (CNT) [917]. Zhou et al. [10] found that adding 2 wt%
composite materials have been used successfully in many carbon nanobers (CNF) into glass ber reinforced
industries such as aerospace, automobile, marine, military, epoxy composites increased the ILSS by 22.3%. In 2005,
etc., their interlaminar shear strength (ILSS) is usually a Gojny et al. [16] gained 20% improvement in ILSS by add-
limiting design characteristic. This is because conventional ing 0.3 wt% double-walled carbon nanotube (DWNT) into
manufacturing techniques do not produce reinforcing ber reinforced epoxy composites. Wichmann et al. [11]
bers oriented in the thickness direction to sustain trans- also got similar enhancement of ILSS, as well as electrical
verse load [1]. Improving ILSS has long been an important conductivity by using the DWNT. Yokozeki et al. [12]
goal in the ber reinforced composites eld, and to this found that carbon ber reinforced composites can benet
end, dierent approaches have been tried. Some of the from dispersion of cup-stacked carbon nanotubes
more eective attempts have been weaving bers in the (CSCNTs) between ber mats and that these can delay
thickness direction or using Z-pins to connect the laminae the onset of matrix cracking. Siddiqui et al. [13] found that
[27]. However, these techniques are labor intensive and adding organoclay nanoparticles increases both crack
require additional manufacturing processes that can growth resistance and fracture toughness in carbon ber
greatly increase the cost of the resulting components [8]. reinforced polymer composites. Zhu et al. [18] proposed
an innovative way to create CNT/ber/polymer compos-
ites by coating SWNTs on the surface of glass bers before
*
Corresponding author. lling the mold with vinyl ester and got 35% enhancement
E-mail address: advani@udel.edu (S.G. Advani). in ILSS. Veedu et al. extended this method by growing

1359-835X/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compositesa.2007.11.013
 Z. Fan et al. / Composites: Part A 39 (2008) 540554 541

Fig. 1. Schematic of CNTs distributed between glass ber layers which could enhance the interlaminar shear strength (ILSS).

MWNTs on the SiC woven fabrics to make 3D hybrid single-walled carbon nanotubes (SWNT are still very expen-
composites and achieved remarkable improvement in inter- sive). The MWNT were made by CVD (>95%) and pur-
laminar fracture toughness (424%), hardness, delamination chased from the ILJIN nanotech Company, Korea. The
resistance, and thermal and electrical conductivities [17]. diameters of the nanotubes were of the order of 1020 nm
All these results have shown an encouraging path toward and the length was between 0.5 and 15 lm. The woven glass
CNT application in ber reinforced structure composites ber mats were supplied by Saint-Goban Vetrotex Com-
and suggest that distributing a carbon nanotube (CNT)/ pany and had a ber density of 800 g/m2. The resin used
epoxy suspension between layers, as shown in Fig. 1A was epoxy SC-15 from Applied Polymetric Inc.
may improve the ILSS of ber reinforced laminated com-
posites. Further, if the CNT suspension is guided to ow
in the thickness direction, the CNT in the suspension might 2.2. Preparation of MWNT suspensions
orient in the thickness direction due to shear [19] as shown
schematically in Fig. 1B. The preferentially orientated CNT Much research has been conducted to develop methods
between glass ber layers in the thickness direction could to separate MWNT and disperse them in various polymer
enhance the ILSS more eectively if the load from the melts [18,2030]. In this study, we used a technique com-
matrix can be transferred eectively to the nanotubes. bining high-speed mechanical stirring, ultrasonic agitation
To test this hypothesis, we fabricated glass/epoxy compos- and acid oxidation to create a well dispersed MWNT
ites with a small quantity of oxidized multi-walled carbon suspension.
nanotubes (OMWNT) dispersed in the epoxy. The First, approximately 1.2 g of MWNT was oxidized by
OMWNT/epoxy suspension was injected into the glass pre- reuxing in 48 ml of a 3:1 mixture of sulfuric and nitric acid
forms using the vacuum assisted resin transfer molding for 30 min at 140 C and then ltering with distilled water
(VARTM) method and a newly developed injection and double (3000 ml) to wash away any acidic residue from the nano-
vacuum assisted resin transfer molding (IDVARTM) method. tubes surfaces. Then the oxidized MWNTs (OMWNTs)
The interlaminar shear strength (ILSS) of the composite was were placed in acetone and agitated in an ultrasonic water
measured using the short beam shear (SBS) and compression bath for 1 h. A predetermined amount of epoxy (depending
shear test (CST) methods. Control samples were manufactured on desired OMWNT concentration) was added and the
in the same manner but using neat epoxy to compare the ILSS mixture was agitated for another 1 h. Finally, the mixture
of unenhanced composites with composites containing of MWNT, epoxy and acetone was placed into a water
OMWNT at 0.5, 1 and 2 wt%. Statistical analysis was carried bath at 80 C and continuously stirred by a high-speed
out to determine the level of statistical signicance of the results. magnetic stirrer for 4 h to fully evaporate the acetone. This
method generated a nearly uniformly separated MWNT/
2. Experiments epoxy suspension as shown in Fig. 2. Although there were
still a number of 0.51 lm MWNT aggregates in the sus-
2.1. Materials pension, the fraction of separated, single MWNT bers
was signicant. However, the oxidizing process results in
In an attempt to keep the material costs relatively low, the attrition of MWNT (OMWNT) length to an average
multi-walled carbon nanotubes (MWNT) were chosen over of about 500 nm from the original MWNT length of about
542 Z. Fan et al. / Composites: Part A 39 (2008) 540554

15 lm. due to the strong acid reaction with the open molec- The vacuum-induced compression reduced the total thick-
ular chains on the MWNT surfaces [20,21,31]. ness of the eight glass ber mats from 8.4 to 4.4 mm, result-
ing in a higher ber volume fraction and consequently a
2.3. Manufacturing process reduced spacing between the bers. The reduced spacing
between glass bers was observed to be of the order of
2.3.1. Vacuum assisted resin transfer molding (VARTM) 5 nm to 10 lm [19]. This intra-ber spacing is small enough
VARTM is a well-established method for fabricating to lter OMWNT from the suspension during the infusion.
glass or carbon ber epoxy composites with 5055% ber Therefore, the concentration of OMWNT increases on the
volume fractions. Distribution-media-assisted VARTM upstream side as the suspension ows through the ber
has an additional advantage in that the suspension will mat. This increases the viscosity of the suspension and
ow across the ber mat in the thickness direction. Because reduces the ow into the fabric signicantly [21,28,33,34].
OMWNT typically align in the shearing direction [19], it is Thus, the driving pressure becomes insucient to over-
possible that the resin ow in the thickness direction will come this ltering and viscous resistance eect and the
help to create preferential orientation of OMWNT in the OMWNT suspension fails to penetrate the glass ber mats
thickness direction. Thus, we rst attempted the conven- at concentrations higher than 0.5%.
tional distribution-media-assisted VARTM method to For suspensions with higher concentrations, the
manufacture the hybrid OMWNT/glass/epoxy composites. OMWNT blocked the ow at the top few ber layers
Eight layers of the glass ber fabric were stacked and the suspension was not able to wet the fabric com-
together and a distribution media was placed on the top pletely. Gojny et al. [16] did not notice such ltering eect,
as shown in Fig. 3. A vacuum was applied to the stack to which might be due to the larger spacing between bers in
provide the pressure gradient needed to inltrate the sus- their RTM method as compared to our VARTM method.
pension into the preform. Another consequence of the But they also found that it would be very dicult to man-
applied vacuum is that the stacked ber mats compress ufacture a sample containing more than 0.5 wt% CNT in
due to the external atmospheric pressure on the sealed the polymer because of the much higher viscosity of the
bag, increasing the ber volume fraction. The OMWNT/ suspension.
epoxy suspension was drawn through a tube into the mold. Transmission electron microscopy (TEM) was used to
Because the distribution media had much higher perme- characterize the OMWNT orientations between the ber
ability than the glass ber mats [32], the OMWNT suspen- mats for the 0.5% suspension. Since there will be many
sion preferentially lls the distribution media rst before bers or nanotubes at a selected location as shown in
owing across the ber layers in the thickness direction. Fig. 4, it will be important to use a description that can
As soon as the suspension arrives at the vent located at embody many dierent directions. Thus, we chose a second
the bottom, the injection is discontinued but the vacuum order orientation tensor method using image analysis [19].
is maintained. After the suspension cures, the vacuum is An ellipse derived from the orientation tensor was drawn
discontinued and the hybrid OMWNT/glass/epoxy com- for each image to indicate the orientation state of all
posite part is de-molded. MWNTs in that image. An elongated ellipse signies a high
One disadvantage of this VARTM process is that the degree of preferential orientation in the direction of the
permeability of the compressed ber preforms is an order major axis as shown in Fig. 4, whereas a circle will signify
of magnitude lower than the uncompressed ber performs no particular preference of orientation. More details of this
(Kalong_compress = 7.9  1012 m2 along ber direction and method can be found in the reference by Fan and Advani
Kacross_compress = 1.39  1013 m2 across ber direction). [19]. Due to the existence of the glass bers in the matrix,

Fig. 2. Transmission electron microscope (TEM) image of OMWNTs in suspension.

 Z. Fan et al. / Composites: Part A 39 (2008) 540554 543

Fig. 3. Schematic of OMWNT/glass/epoxy composite being processed by the VARTM method to create preferentially oriented OMWNTs in the
thickness direction.

Fig. 4. TEM images of OMWNT orientation state (A) preferential orientation in the thickness direction between glass ber mats and (B) preferential
orientation in ow direction between two bers. The major axis of the ellipse signies the strength of the orientation state of the OMWNT while its
direction signies the direction of preferential orientation [19].

it is dicult to achieve successful sections for TEM to the elongated ellipse in Fig. 4A. Examining the ow
observe the OMWNT orientation states between the glass through the gaps between the bers, OMWNT were also
ber mats. But we did observe preferentially orientated found to be preferentially orientated in the ow direction
patterns of MWNTs in the thickness direction from all as shown in Fig. 4B.
the microtome sections taken. Under TEM, we observed Although the ow speed across the thickness direction is
appreciable preferential orientation of the OMWNT low, the spaces between glass bers in a compacted
between the mats in the thickness direction, as shown by VARTM mold is much smaller than the spaces between
544 Z. Fan et al. / Composites: Part A 39 (2008) 540554

mats, on the order of 5 nm to 10 lm. Therefore, the shear Step (1) Eight layers of 5 cm  5 cm woven glass ber mat
rate and the high viscosity can generate sucient shear are sealed with a vacuum bag and the air is evac-
stress to aect the orientation state of the OMWNTs, as uated by applying the rst vacuum, which is con-
we observed in Fig. 4 and in Ref. [19]. As discussed in that nected to the outlet of the mold; the thickness of
paper, Brownian motion was found to have little eect on ber mats shrinks from 8.4 mm down to 4.4 mm
the MWNT orientation state during the curing process due to the compression from the external atmo-
[19]. Fig. 4 shows that the orientation state of OMWNTs spheric pressure.
is essentially the same after 24 h of curing. Step (2) The outlet is clamped to discontinue the rst vac-
Due to the limitations of the VARTM process discussed uum and the OMWNT/epoxy suspension is
above, processing of hybrid OMWNT/glass/epoxy com- injected into the closed mold; excess suspension
posites with OMWNT concentrations higher than 0.5% is injected to over saturate the space between the
requires the development of a modied process design. bag and the preform.
Step (3) The inlet is clamped and the entire closed mold is
placed into a second vacuum chamber for 30 min.
2.3.2. Injection and double vacuum assisted resin transfer This releases the external atmosphere pressure
molding (IDVARTM) method induced by the rst vacuum and allows the eight
To increase the concentration of OMWNT, we devel- compressed ber mats to recover some of their
oped a new variation of the VARTM method, which we original thickness, increasing the thickness back
term: injection and double vacuum assisted resin transfer to 6.8 mm. The recovery expands the gaps
molding (IDVARTM). The IDVARTM method is similar between the glass mats making it easier for the
to the conventional distribution-media-assisted VARTM OMWNT/epoxy suspension to infuse into the lar-
method, but takes advantage of the capillary action of ger channels between the glass ber mats by
the ber tows for enhanced infusion. With this method intrinsic capillary pressure.
we were able to inject suspensions of epoxy containing up Step (4) After the OMWNT/epoxy suspension lls the
to 2 wt% concentration of OMWNT into the ber mats. gaps between the glass ber mats, it starts to
As depicted in Fig. 5, the procedure can be described in infuse across into each glass ber mat by capillary
the following steps: pressure.

Fig. 5. Schematic of injection and double vacuum assisted resin transfer molding (IDVARTM) manufacturing process.
 Z. Fan et al. / Composites: Part A 39 (2008) 540554 545

Step (5) The second vacuum is released to reapply atmo- posite manufactured using the traditional distribution-
sphere pressure to the mold for 30 min to further media-assisted VARTM method in which, due to ltering,
squeeze the OMWNT/epoxy suspension into the the OMWNT did not penetrate through all the layers.
ber mat. Although some OMWNTs are still Fig. 6B shows a cross-section of a composite made using
being ltered at this stage, this turns out to be the IDVARTM method. Notice that in Fig. 6B, the
an advantage because the higher concentration OMWNT appear to be uniformly distributed between the
OMWMT/epoxy suspension will remain between glass ber mats, as opposed to Fig. 6A where they do
the ber mats where it is most needed to improve not. To observe the OMWNTs in the hybrid composites
the ILSS. made from IDVARTM, we examined a fracture surface
Step (6) After the OMWNT suspension lls the empty of the sample under SEM. In Fig. 6C, OMWNTs are
regions within and between the ber mats, the rst clearly seen embedded in the matrix surrounding the glass
vacuum is reapplied to draw o the excess ber surface, which indicates that OMWNTs had success-
OMWNT/epoxy suspension and recompress the fully infused into the glass ber mats.
mold to obtain the desired higher ber volume
fraction. 2.4. Interlaminar shear strength (ILSS) characterization

With the IDVARTM method, we successfully manufac- When the transverse shear load experienced by a lami-
tured OMWNT/glass/epoxy composites with up to 2 wt% nated composite exceeds the interlaminar shear strength
OMWNT concentration in the suspension without creation (ILSS), a delamination failure will occur between the layers
of any dry regions or voids. Subsequent microscopy of reinforcing bers. To measure the ILSS of a composite
showed the presence of nanotubes in the regions in the directly, a pattern of pure shear stress should be generated
interlaminar spaces between the ber mats. Fig. 6A shows between laminae to induce an interlaminar shear failure.
the cross-section of a hybrid OMWNT/glass/epoxy com- Any normal loads introduced perpendicular to ber layers

Fig. 6. Cross-section of (A) OMWNT composite manufactured with conventional distribution-media-assisted VARTM process in which OMWNTs were
ltered and (B) OMWNT composite manufactured using IDVARTM in which the OMWNTs are distributed evenly between glass ber layers. (C) SEM
images of OMWNTs embedded in matrix on the glass ber surface after fracture of hybrid composite sample made from IDVARTM.
546 Z. Fan et al. / Composites: Part A 39 (2008) 540554

can alter the failure process, making it impossible to deter- load along the laminate interface and force the specimen to
mine an accurate value for the ILSS. break apart in pure shear (Fig. 7E). The advantages and
A number of dierent tests (Fig. 7) have been developed disadvantages of each method have been reviewed and dis-
for the purpose of characterizing the ILSS. These experi- cussed in the previous literature [1,3639] and we will not
ments are designed to load the specimen so that it fails in attempt to reproduce that discussion here. In the present
interlaminar shear and the shear stress at failure is taken work, we selected the short beam shear (SBS) and compres-
as the ILSS. Comparing the available methods, the short sion shear test (CST) methods to measure the ILSS of our
beam shear (SBS) method is the simplest and therefore samples of glass ber reinforced composites made by the
most commonly used in practice. Based on classical beam IDVARM method.
theory, it causes transverse shear failure through three-
point bending as shown in Fig. 7A. The main drawback 2.4.1. Short beam shear (SBS) test
of this test is that not all specimens fail in transverse shear. The short beam shear (SBS) test is designed to generate
Some fail by crushing at the load point or exure due to the interlaminar shear indirectly through bending and is the
bending, or by a combination of these with transverse most popular method to characterize the apparent inter-
shear. Several test methods (Fig. 7B and C) use axial tensile laminar shear strength of unidirectional, ber-reinforced
or compressive loading to fail the sample in shear. In these composites (ASTM D 2344). As shown in Fig. 8, the spec-
methods, the expected failure plane is actually loaded in imen is placed on two cylindrical supports and a cylindrical
tension or compression simultaneously with the shear. head is moved down to apply a force at the center and gen-
Therefore, the failure is a mixed-mode type of failure and erate an increasing transverse load until the rst failure is
does not necessarily correspond to the ILSS. The curved- recorded. The load at failure is then used to determine
double-cantilever-beam (CDCB) method characterizes the apparent interlaminar shear strength of the composite.
ILSS by recording the strength to tear the laminates apart
following an initial pre-crack between laminates (Fig. 7D). 2.4.2. Compression shear test (CST)
The interlaminar loading pattern in this experiment starts We built a compression shear test (CST) device similar
out as tension and gradually becomes shear making it ques- to the one described in the paper by Schneider et al. [36],
tionable as a measure of the static shear strength. The com- with some minor modications. As shown in Fig. 9, in
pression shear test (CST) is designed to create direct shear the CST test, a specimen is placed between two clamps.

Fig. 7. Geometry of the dierent experimental interlaminar shear tests [35,36].

 Z. Fan et al. / Composites: Part A 39 (2008) 540554 547

Fig. 8. Picture of short beam shear (SBS) (ASTM D 2344).

Fig. 9. Picture of compression shear test (CST).

The left clamp contacts the left half of the top surface, by 30% or more immediately after the peak load is reached,
which is perpendicular to the laminate planes. Symmetri- it is assumed that the specimen failed in laminar shear and
cally, the right clamp contacts the right half of the bottom the peak load is then used to determine the apparent ILSS.
surface of the specimen. Two vertical guides support the The main advantage of the SBS is its simplicity. The
left and right sides of the specimen to prevent it from rotat- specimens are relatively easy to prepare and the test itself
ing or moving horizontally. The right clamp is xed in posi- is simple to conduct and requires little xturing. However,
tion, and the left clamp moves down to generate direct the SBS gives an accurate measure of the ILSS value only if
interlaminar shear along the central plane of the specimen pure interlaminar shear failure takes place. Many doubts
until failure. The load on the specimen at failure is divided have been raised regarding the accuracy and reliability of
by the area of the failure plane to determine the interlam- the SBS. The most common criticism is that in the SBS,
inar shear strength (ILSS) of the specimen. the specimen usually encounters other types of damage
before or concurrent with interlaminar shear failure, caus-
3. Results and discussion ing the specimen to fail by other than pure interlaminar
shear [8,40,41].
3.1. Short beam shear (SBS) test In Fig. 10A, the shear stress eld in half of a curved SBS
specimen is shown as calculated by nite element modeling
In the SBS test, the determination of ILSS is based on [1]. We can see from the gure, that the maximum shear
classical (BernoulliEuler) beam theory. For a beam of stress does not occur in the middle of the section; instead,
rectangular cross-section loaded in three-point bending, it occurs closer to the surface where the load is applied.
the maximum interlaminar shear stress occurs at the mid- Additionally, a highly concentrated compressive stress
thickness of the beam between the center and end supports can be seen near the loading point. This compression is
and is calculated to be in the transverse direction where the strength is not as high
as it is in the ber direction and therefore can cause signif-
3 PB
s  1 icant damage before the specimen fails by shear.
4 wt In the present study, our tests were further complicated
where PB is the maximum load and w and t are the width due to the fact that we used woven glass ber reinforced
and the thickness of the beam, respectively. We selected a epoxy composites. These materials have a relatively tough
span to thickness ratio for the samples of w/t = 4 as recom- matrix and non-planar (undulating) interlaminar regions
mended by the ASTM standards (ASTM D 2344). During and therefore can experience a variety of dierent failure
bending in SBS, the load increases proportionately with and damage modes before interlaminar shear failure occurs
deformation, until a peak load is reached. If the load drops under SBS loading. Fig. 11 shows a number of experimen-
548 Z. Fan et al. / Composites: Part A 39 (2008) 540554

Fig. 10. Contours of shear stress in specimen of (A) short beam shear and (B) compression shear test [1].

Fig. 11. Two typical curves of load vs. deection for a short beam shear specimen (A) and (B).

tal curves of SBS load vs. deformation for several speci- loaddeection behaviors in the gure. In Fig. 11A, the
mens cut from the same sample. We see two typical curve rises gradually and then drops suddenly showing a
 Z. Fan et al. / Composites: Part A 39 (2008) 540554 549

Fig. 12. SEM images of (A) interlaminar shear failure mode; (B) resin microcracks caused by compression; (C) indentation failure mode in SBS.

distinct failure load developed in the sample, while in Table 1

Fig. 11B, the curve rises, attens for some time and then Interlaminar shear strength (ILSS) of OMWNT/glass/epoxy composites
measured by short beam shear (SBS)
shows a small drop.
We found that in the case of the specimen generating a Specimen number ILSS (MPa)
curve such as Fig. 11A, we were able to observe a clear 0 wt% 0.5 wt% 1 wt%
interlaminar shear failure in the sample, as shown in 1 30.17 47.95 42.09
Fig. 12A. However, in case of Fig. 11B, we observed micro- 2 32.89 28.77 34.80
cracking of the resin inside the glass ber tows as shown in 3 41.21 29.09 35.31
4 31.58 31.55 41.07
Fig. 12B. It appears that this damage was caused by com- 5 31.03 26.31 35.50
pression in the glass ber tow. Some indentation failure, as 6 27.83 37.79 45.28
shown in Fig. 12C, was also observed for the specimens 7 31.37 39.57 36.09
that resulted in the type of curves shown in Fig. 11B. 8 33.86 27.22 37.03
Therefore, for specimens exhibiting shear load curves such Average 32.49 33.53 38.40
as Fig. 11B, we assumed that some other damage modes Standard deviation (%) 12.2 22.6 10.1
contributed to the failure. The calculated ILSS therefore p-Value 0.09
represented ultimate failure under a combination of dier-
ent failure modes in the tested specimen, for example, resin
microcracks, indentation and interlaminar shear. This type
of combined failure was found in 90% of the tested speci- analysis. From a statistical viewpoint, if the p-value is less
mens resulting in a large standard deviation in the measure- than 0.05, the data from dierent groups can be considered
ment data as shown in Table 1 and Fig. 13. Although, the statistically distinct, otherwise they are assumed to belong
average ILSS as measure by SBS, of the composites to the same population. The smaller the p-value, the more
improved 3.2% by adding 0.5 wt% OMWNT in epoxy, signicant the dierences among dierent groups. The
and 18.2% for 1 wt% OMWNT, the standard deviation p-value for the data from the SBS test was 0.09 (higher
of each group of specimens was 1022.6%, as shown by than 0.05). Therefore, the ILSS values determined by the
error bars in Fig. 13. Because the deviation among speci- SBS test for samples containing dierent OMWNT concen-
men groups is not much dierent from the deviation within trations were not statistically dierent. Put another way,
each group, we applied the analysis of variance (ANOVA) based on this statistical analysis, we could not distinguish
method to statistically analyze the dierence. For the a dierence in the interlaminar shear strength of dierent
details of ANOVA, see for example, [42]. Only the single hybrid MWNT/glass/epoxy composites using the short
p-value derived from all the data was calculated for this beam shear (SBS) test.
550 Z. Fan et al. / Composites: Part A 39 (2008) 540554

Fig. 13. Chart of ILSS of the hybrid OMWNT/glass/epoxy composites at dierent OMWNT concentrations in epoxy resin measured using the SBS
method.

3.2. Compression shear test (CST) with deection until it reaches a peak followed by a sharp
drop of at least 50%. This type of sharp drop is taken as an
The advantage of the CST method is that it can create indication that a pure interlaminar shear failure has
pure interlaminar shear stress between ber planes and occurred in the CST test. The calculated values for the
therefore minimizes the inuence of other possible failure ILSS determined from this test did not dier by more that
modes, such as compression, indentation and exure. 10% for specimens cut from a single manufactured part.
Fig. 10B shows the shear stress eld in a specimen loaded Considering the variations in the sample size and the mate-
in CST as calculated by the nite element method [39]. rial and the local heterogeneities in the process, this varia-
One can easily identify the maximum interlaminar shear tion is rather small.
stress along the middle plane, which will cause interlaminar Fig. 15 shows an SEM image of the failure plane from a
shear failure. Similar to the stress contours for the SBS test, specimen tested in CST. Interlaminar shear failure can
we can still observe stress concentration at the surface where clearly be observed along the axis of the specimen.
the load is applied. However, this stress concentration is
directed along the ber direction and because the strength 3.3. Eect of OMWNTs on ILSS of composites
is higher in this direction, the incidence of compressive
failure will be minimized. Thus, the CST is seen to be a more 3.3.1. ILSS of glass ber reinforced composites fabricated by
reliable method to characterize the interlaminar shear IDVARTM
strength (ILSS) [39]. Further analysis and more detailed As stated earlier, using the injection and double vacuum
discussion have been reported in the literature [1,36,39,43]. assisted resin transfer molding (IDVARTM) method, we
Fig. 14 shows typical loaddeection curves recorded fabricated four types of hybrid composite samples by infus-
during CST testing. The load increases proportionately ing eight layers of 5 cm  5 cm woven glass ber mats with

Fig. 14. Deection load curves recorded in compression shear test (CST).
 Z. Fan et al. / Composites: Part A 39 (2008) 540554 551

Fig. 15. SEM image of interlaminar shear failure mole in compression shear test (CST).

Fig. 16. Interlaminar shear strength (ILSS) of hybrid OMWNT/glass/epoxy composites made by injection and double vacuum assisted resin transfer
molding (IDVARTM) method.

0, 0.5, 1 and 2 wt% OMWNT/epoxy suspensions, respec- Table 2

tively. We found that adding OMWNT to the resin Interlaminar shear strength (ILSS) of hybrid OMWNT/glass/epoxy
increased the ILSS of these hybrid OMWNT/glass/epoxy composites made by IDVARTM and measured by CST method
composites. Fig. 16 shows the average ILSS for the various Specimen number ILSS (MPa)
OMWNT percentages as determined by the CST device. 0 wt% 0.5 wt% 1 wt% 2 wt%
The results show that 0.5, 1, 2 wt% OMWNT in epoxy 1 25.79 28.05 28.01 31.01
increased the ILSS by 9.7%, 20.5% and 33.1%, respectively, 2 25.52 29.34 30.21 31.46
with respect to the neat epoxy (0 wt%). Table 2 shows the 3 26.75 27.18 31.37 35.41
actual ILSS test data of each group of OMWNT/glass/ 4 25.83 26.76 32.88 36.78
5 24.77 29.96 31.23 33.59
epoxy specimens as well as a single p-value derived for this 6 24.36 26.22 30.92 32.55
group taken as a whole. The very small p-value, 2.35  7 25.43 27.90 30.21 34.00
1011 conrms that adding OMWNT aects the ILSS of 8 24.90 27.77 30.34 35.87
glass ber reinforced composites in a statistically signicant Average 25.42 27.90 30.65 33.83
way. Standard deviation (%) 2.9 4.5 4.5 6.2
p-Value 2.35  1011
3.3.2. Eect of preferentially orientated OMWNT on ILSS
of glass ber reinforced composites
Using the conventional distribution-media-assisted OMWNT in the thickness direction between glass ber
VARTM method, we were able to make hybrid OMWNT/ layers was observed. As described in the earlier section of
glass/epoxy composites that contained 0.5 wt% OMWNT the paper, the 0.5 wt% OMWNT/epoxy suspension was also
in epoxy. As shown in Fig. 4, preferential orientation of used to make a hybrid composite by the IDVARTM
552 Z. Fan et al. / Composites: Part A 39 (2008) 540554

Fig. 18 shows the ILSS improvement of glass ber rein-

forced epoxy composites using the 0.5 wt% OMWNT/
epoxy suspension relative to unenhanced specimens: one
sample was made by IDVARTM method resulting in no
preferential OMWNT orientation in the thickness direction
and the other sample was manufactured by distribution-
media-assisted VARTM resulting in preferentially orien-
tated OMWNTs in the thickness direction. The sample
containing the preferentially orientated OMWNTs had a
12.2% higher ILSS as compared to the control sample
without nanotubes made by VARTM while the sample
containing randomly oriented OMWNTs fabricated by
IDVARTM showed an improvement of 9.7% over the con-
trol sample made using the same IDVARTM process.
Because the OMWNTs are preferentially orientated in
the thickness direction during VARTM, it is highly likely
that these oriented OMWNTs have a slightly greater eect
in improving the ILSS as compared to the non-preferential
orientation resulting from the IDVARTM. Since the
improvement was measured by comparing VARTM
(without nanotubes) to VARTM (with nanotubes) and
Fig. 17. Random OMWNT orientation state in a sample made by IDVARTM (without nanotubes) to IDVARTM (with
IDVARTM. nanotubes) samples, the improvement dierences of 9.7%
and 12.2% cannot be attributed to the thickness or volume
method. Because the IDVARTM method does not generate fraction variations between VARTM and IDVARTM.
shear ow in the thickness direction, OMWNTs did not However, it is possible that other factors during processing,
show any preferential alignment in any particular direction, for example, compaction and ow path, may have also
as shown in Fig. 17. Thus, we have two types of 0.5 wt% contributed to the dierences in the magnitude of orienta-
OMWNT/glass/epoxy composites: one that contains prefer- tion state observed. The 9.7% and 12.2% improvements are
entially orientated OMWNTs in the thickness direction and average results obtained from testing eight samples from
one in which the MWNTs exhibit random orientation state. dierent laminates fabricated using both VARTM and
Comparing the ILSS of these two composites should help us IDVARTM processes. For higher than 0.5% MWNT con-
determine the eect of the OMWNT orientation on the inter- tent in the epoxy, the ltering eect was prominent in the
laminar shear properties. VARTM method, hence it was not possible to investigate

Fig. 18. Percentage improvement in interlaminar shear strength (ILSS) of two types of glass ber reinforced epoxy composites containing 0.5 wt%
OMWNT/epoxy, as compared to the unenhanced composite: one sample contains no preferentially orientated OMWNT in the thickness direction and the
other sample contains preferentially orientated OMWNTs in the thickness direction.
 Z. Fan et al. / Composites: Part A 39 (2008) 540554 553

the inuence of preferential MWNT orientation in the Acknowledgements

thickness direction on the ILSS of the composite.
The authors gratefully acknowledge the support pro-
4. Conclusions vided by the National Science Foundation under Grant
DMI- 0115127. Additional appreciation goes to Dr. Lauke
This study was conducted to determine if we could for generously providing information on design of CST de-
improve the ILSS of glass ber reinforced composites by vice and Dr. Chao-Ying Ni for his TEM expertise. Thanks
reinforcing them with carbon nanotubes in the regions also to Dr. D.J. Pochan and Mr. Frank Friss for micro-
between the fabric layers. When the concentration of tome sectioning the suspensions.
OMWNT in the Epoxy was lower than 0.5 wt%, the tradi-
tional distribution-media-assisted VARTM method could References
be used to make hybrid OMWNT/glass/epoxy composites
with preferential orientation of OMWNT in the thickness [1] Rosselli F, Santare MH. Comparison of the short beam shear (SBS)
and interlaminar shear device (ISD) tests. Compos Pt A: Appl Sci
direction between the glass ber layers. However, in case
Manuf 1997;28(6):58794.
of OMWNT suspensions with concentrations higher than [2] Byrd LW, Birman V. Eectiveness of z-pins in preventing delamina-
0.5 wt%, ltering and low permeability across the ber lay- tion of co-cured composite joints on the example of a double
ers, caused this method to break down. The low permeabil- cantilever test. Compos Pt B: Eng 2006;37(45):36578.
ity and resulting ltering were caused, at least partly, by the [3] Jain LK, Mai YW. In the eect of stitching on mode-I delamination
compaction of the preform under the application of the toughness of laminated composites. Compos Sci Technol
1994;51(3):33145.
vacuum. A novel injection and double vacuum assisted [4] Partridge IK, Cartie DDR. Delamination resistant laminates by Z-
resin transfer molding (IDVARTM) method was designed ber (R) pinning: part I manufacture and fracture performance.
to reduce the eects of compaction. This method allows Compos Pt A: Appl Sci Manuf 2005;36(1):5564.
capillary pressure to assist the infusion. Using this method, [5] Sharma SK, Sankar BV. Eect of stitching on impact and interlam-
hybrid OMWNT/glass/epoxy composites were made suc- inar properties of graphite/epoxy laminates. J Thermoplast Compos
Mater 1997;10(3):24153.
cessfully with a relatively uniform distribution of OMWNT [6] Zhang X, Hounslow L, Grassi M. Improvement of low-velocity
between the layers, using suspensions of up to 2 wt% impact and compression-after-impact performance by z-bre pinning.
OMWNT epoxy. Compos Sci Technol 2006;66(15):278594.
The short beam shear test was found to not be reliable in [7] Mouritz AP, Leong KH, Herszberg I. A review of the eect of
stitching on the in-plane mechanical properties of bre-reinforced
characterizing the interlaminar shear strength (ILSS) of the
polymer composites. Compos Pt A: Appl Sci Manuf
hybrid composites because failures occur by a combination 1997;28(12):97991.
of dierent modes: compression, indentation, exure and [8] Abali F, Pora A, Shivakumar K. Modied short beam shear test for
laminar shear. The compression shear test (CST) minimizes measurement of interlaminar shear strength of composites. J Compos
the inuence of other failure modes, generating a nearly Mater 2003;37(5):45364.
pure interlaminar shear failure and allowing for consistent [9] Hsiao KT, Alms J, Advani SG. Use of epoxy/multiwalled carbon
nanotubes as adhesives to join graphite bre reinforced polymer
evaluation of the ILSS of hybrid OMWNT/glass/epoxy composites. Nanotechnology 2003;14(7):7913.
composites. The results show that the addition of OMWNT [10] Zhou Y, Pervin F, Rangari VK, Jeelani S. Fabrication and evaluation
improved the ILSS in these composites made by both the of carbon nanober lled carbon/epoxy composite. Mater Sci Eng A
distribution-media-assisted VARTM and IDVARTM 2006;426:2218.
methods. For the 0.5 wt% samples, the improvement seen [11] Wichmann MHG, Sumeth J, Gojny FH, Quaresimin M, Fiedler B,
Schulte K, et al. Eng Fract Mech 2006;73:234659.
using the distribution-media-assisted VARTM process [12] Yokozeki T, Iwahori Y, Ishiwata S. Matrix cracking behaviors in
was greater than that using IDVARTM process, which we carbon ber/epoxy laminates lled with cup-stacked carbon nano-
suspect is due to the preferentially orientated OMWNT in tubes (CSCNTs). Compos Pt A: Appl Sci Manuf 2006;38:91724.
the former case. However, we were not able to manufacture [13] Siddiqui NA, Woo RSC, Kim JK, Leung CCK, Munir A. Mode I
composites with more than 0.5 wt% OMWNT using the dis- interlaminar fracture behavior and mechanical properties of CFRPs
with nanoclay-lled epoxy matrix. Compos Pt A: Appl Sci Manuf
tribution-media-assisted VARTM process. The IDV- 2006;38:44960.
ARTM allowed us to fabricate composites with 0.5, 1 and [14] Zhu J, Imam A, Crane R, Lozano K, Khabashesku V, Barrera EV.
2 wt% OMWNT/epoxy resin composites in which the ILSS Processing a glass ber reinforced vinyl ester composite with
increased by 9.7%, 20.5% and 33.1%, respectively, with nanotube enhancement of interlaminar shear strength. Compos Sci
respect to the unreinforced resin. Considering that the Technol 2006;67:150917.
[15] Shahid N, Villate RG, Barron AR. Chemically functionalized
weight fraction of resin in the composites is about 50%, alumina nanoparticle eect on carbon ber/epoxy composites.
the OMWNT weight fraction in the hybrid composites as Compos Sci Technol 2005;65:22508.
a whole, is much lower, ranging from 0.25 to 1 wt%. This [16] Gojny FH, Wichmann MHG, Fiedler B, Bauhofer W, Schulte K.
shows that a very small amount of MWNT can enhance Inuence of nano-modication on the mechanical and electrical
properties of conventional bre-reinforced composites. Compos Pt A:
the ILSS eectively by using IDVARTM processing
Appl Sci Manuf 2005;36:152535.
method. Suspensions containing more than 2% OMWNT [17] Veedu VP, Cao A, Li X, Ma K, Soldano C, Kar S, et al.
were too viscous to penetrate the ber mats using either Multifunctional composites using reinforced laminae with carbon-
manufacturing technique. nanotube forests. Nat Mater 2006;5(6):45762.
554 Z. Fan et al. / Composites: Part A 39 (2008) 540554

[18] Zhu J, Khabashesku V, Imam A, Crane R, Lozano K, Barrera E. [31] Xu JH, Chatterjee S, Koelling KW, Wang YR, Bechtel SE. Shear and
Processing and properties of polymer composites reinforced by extensional rheology of carbon nanober suspensions. Rheol Acta
functionalized SWNTs. In: Pricm 5: the fth pacic rim international 2005;44(6):53762.
conference on advanced materials and processing, Pts 15, vols. 475 [32] Estrada G, Vieux-Pernon C, Advani SG. Experimental characteriza-
479; 2005. p. 105962. tion of the inuence of tackier material on preform permeability. J
[19] Fan Z, Advani SG. Characterization of orientation state of carbon Compos Mater 2002;36(19):2297310.
nanotubes in shear ows. Polymer 2005;46(14):523240. [33] Kinloch IA, Roberts SA, Windle AH. A rheological study of
[20] Esumi K, Ishigami M, Nakajima A, Sawada K, Honda H. Chemical concentrated aqueous nanotube dispersions. Polymer 2002;43(26):
treatment of carbon nanotubes. Carbon 1996;34(2):27981. 748391.
[21] Shaer MSP, Fan X, Windle AH. Dispersion and packing of carbon [34] Abdel-Goad M, Potschke P. Rheological characterization of melt
nanotubes. Carbon 1998;36(11):160312. processed polycarbonate-multiwalled carbon nanotube composites. J
[22] Tang BZ, Xu HY. Preparation, alignment, and optical properties of Non-Newtonian Fluid Mech 2005;128(1):26.
soluble poly(phenylacetylene)-wrapped carbon nanotubes. Macro- [35] Rosselli F. Interlaminar strength of thermoplastic composites rings
molecules 1999;32(8):256976. manufactured using laser assisted on-line consolidation. Masters
[23] Gong XY, Liu J, Baskaran S, Voise RD, Young JS. Surfactant- thesis, University of Delaware; 1995 [Master].
assisted processing of carbon nanotube/polymer composites. Chem [36] Schneider K, Lauke B, Beckert W. Compression shear test (CST)
Mater 2000;12(4):104952. a convenient apparatus for the estimation of apparent sheer
[24] Eitan A, Jiang KY, Dukes D, Andrews R, Schadler LS. Surface strength of composite materials. Appl Compos Mater 2001;8(1):
modication of multiwalled carbon nanotubes: toward the tailoring of 4362.
the interface in polymer composites. Chem Mater 2003;15(16): 3198201. [37] Adams DF, Lewis EQ. Current status of composite material shear
[25] Fan Z, Hsiao KT, Advani SG. Experimental investigation of test methods. SAMPE J 1994;31(6):3241.
dispersion during ow of multi-walled carbon nanotube/polymer [38] Kinloch AJ. Adhesion and adhesives. Chapman and Hall; 1994.
suspension in brous porous media. Carbon 2004;42:8716. [39] Rosselli F, Santare MH, Guceri SI. Eects of processing on laser
[26] Zhu J, Peng HQ, Rodriguez-Macias F, Margrave JL, Khabashesku assisted thermoplastic tape consolidation. Compos Pt A: Appl Sci
VN, Imam AM, et al. Reinforcing epoxy polymer composites Manuf 1997;28(12):102333.
through covalent integration of functionalized nanotubes. Adv Funct [40] Shindo Y, Wang R, Hogriguchi K. Analytical and experimental
Mater 2004;14(7):6438. studies of short-beam interlaminar shear strength of G-10CR glass-
[27] Sabba Y, Thomas EL. High-concentration dispersion of single-wall cloth/epoxy laminates at cryogenic temperatures. Trans ASME
carbon nanotubes (vol. 37, p. 4815, 2004). Macromolecules 2004; 2001;123:1128.
37(17):6662. [41] Cui WC, Wisnom MR. Contact nite element analysis of three- and
[28] Du FM, Scogna RC, Zhou W, Brand S, Fischer JE, Winey KI. four-point short-beam bending of unidirectional composites. Compos
Nanotube networks in polymer nanocomposites: rheology and Sci Technol 1992;45:32334.
electrical conductivity. Macromolecules 2004;37(24):904855. [42] Rutherford A. Introducing ANOVA and ANCOVA: a GLM
[29] Wang M, Pramoda KP, Goh SH. Enhancement of the mechanical approach (introducing statistical methods series). Sage Publications
properties of poly(styrene-co-acrylonitrile) with poly(methyl methacrylate)- Ltd.; 2001.
grafted multiwalled carbon nanotubes. Polymer 2005;46(25): 115106. [43] Ramanathan T, Liu H, Brinson LC. Functionalized SWNT/polymer
[30] Moniruzzaman M, Winey KI. Polymer nanocomposites containing nanocomposites for dramatic property improvement. J Polym Sci Pt
carbon nanotubes. Macromolecules 2006;39(16):5194205. B: Polym Phys 2005;43(17):226979.