Toughening Performance of Glass Fibre Composites with Core-shell Rubber and Silica
Nanoparticle Modified Matrices
Shamsiah Awang Ngah1, Ambrose C Taylor*
Department of Mechanical Engineering, Imperial College London, South Kensington Campus,
London SW7 2AZ, UK.
1
Present address: WMG, University of Warwick, Coventry CV4 7AL, UK.
* Corresponding author: Email: a.c.taylor@imperial.ac.uk
Tel: +44 20 7594 7149
Abstract
The fracture energies of glass fibre composites with an anhydride-cured epoxy matrix
modified using core-shell rubber (CSR) particles and silica nanoparticles were investigated. The
quasi-isotropic laminates with a central 0°/0° ply interface were produced using resin infusion.
Mode I fracture tests were performed, and scanning electron microscopy of the fracture
surfaces was used to identify the toughening mechanisms.
The composite toughness at initiation increased approximately linearly with increasing
particle concentration, from 328 J/m2 for the control to 842 J/m 2 with 15 wt% of CSR particles.
All of the CSR particles cavitated, giving increased toughness by plastic void growth and shear
yielding. However, the toughness of the silica-modified epoxies is lower as the literature shows
that only 14% of the silica nanoparticles undergo debonding and void growth. The size of CSR
particles had no influence on the composite toughness. The propagation toughness was
dominated by the fibre toughening mechanisms, but the composites achieved full toughness
transfer from the bulk.
Keywords
A. Particle-reinforcement; A. Polymer-matrix composites (PMCs); B. Fracture toughness; A.
Core-shell rubber particles.
1
1
Introduction
The epoxy polymers used as the matrices of fibre composites are thermosets, and hence
are highly crosslinked. This makes them rather brittle, being susceptible to the initiation and
propagation of delaminations, so the epoxy matrix is toughened when the composites are used
as a structural material. Various approaches have been used, including the addition of rubber
[1-4], thermoplastic [5, 6] or ceramic [7, 8] tougheners. One of the most successful approaches
has been the addition of a rubber adduct such as carboxyl-terminated butadiene acrylonitrile
(CTBN) which will phase-separate into micron-sized particles during curing of the epoxy
polymer [1, 4]. Alternatively, the addition of pre-formed core-shell rubber (CSR) particles has
been shown to result in an impressive toughness improvement in bulk polymers [9-11]. These
particles typically comprise a soft rubbery core inside a shell of polymethylmethacrylate. The
use of CSR particles eliminates the incomplete phase separation that occurs when using liquid
rubber, which leads to a reduction in the modulus and the glass transition temperature of the
epoxy. For fast-curing epoxies (e.g. [12]), the formation of the epoxy network is much faster
than the phase-separation process, so CTBN rubber particles will not form properly. Further,
CTBN increases the viscosity of the epoxy significantly, and thus CTBN-modified resins are less
suitable for use with infusion processes than the lower viscosity CSR-modified resins. The
diameter of the CTBN particles is dependent on the cure conditions, so cannot be readily
controlled and will vary, especially through a thick composite component. The toughness of
CTBN-modified epoxies varies with particle size [13], so this can lead to a concomitant variation
in properties. Thus for many applications pre-formed CSR particles may be more suitable than
phase-separating rubbers such as CTBN. The toughness of fibre-reinforced composites (FRCs)
with CSR-modified matrices has been discussed by Day et al. [14] and Tsai et al. [15], but there
is relatively little work in the literature that discusses these materials.
The toughening mechanisms responsible for the increased toughness have been
reported as cavitation of the rubbery core followed by plastic void growth, and shear
deformation in the polymer matrix [9-11]. In these studies, the CSR particles used were in the
micron range. In other work [16, 17], cavitation was not observed but there was debonding at
the interface between the core and the shell. Both cavitation and debonding relieve the
constraint in the epoxy and allow plastic deformation, thus increasing the toughness. It has
been suggested the toughening effectiveness decreases when using particles less than 0.2 m
in diameter as they are difficult to cavitate [11]. However, several recent studies [18-20], have
2
demonstrated that nanoscale CSR particles (i.e. ~100 nm in diameter) can fully cavitate. This is
supported by modelling work which shows that more energy is required for the cavitation of
small particles, but does not show a size limit below which cavitation will not occur [21].
Recent developments in nanotechnology have seen the utilisation of rigid nanoparticles
to toughen epoxies. Nanoparticles such as silica, alumina and titanium oxide with a 5-50 nm
diameter range [22] have been used widely with epoxy resins. The toughening of polymers
using silica nanoparticles has been discussed extensively in the literature [23-29]. The
toughening mechanisms have been identified as debonding followed by plastic void growth, and
shear deformation in the polymer matrix [27, 30]. However, the increases in toughness are
generally relatively small compared to those from rubber particles, as only 1 in 7 silica
nanoparticles debond to form cavities [31], compared with all of the rubber particles (which
either debond or cavitate).
The use of silica nanoparticles in toughening of fibre-reinforced composites (FRCs) have
been reported by Kinloch and co-workers under both static [23-25, 32] and fatigue [33-35]
loading conditions. Among the advantages of using silica nanoparticles are that the particles do
not increase the viscosity of the resin significantly, do not exhibit thixotropic properties, have
good dispersion and also are free of agglomeration. These properties allow their application in
the resin infusion manufacturing method [36]. The suitability of this modified epoxy as a new
type of matrix for manufacturing FRCs using injection technology (or liquid composite moulding
techniques) has been further discussed by Mahrholz et al. [36]. These studies [23-25, 27, 32-35]
have reported that the silica nanoparticles increase the fracture energy and tensile modulus of
the epoxy and have no effect on the thermal properties. An improved fatigue life and reduced
fatigue crack growth have been observed [33-35]. In addition to the toughening mechanisms
described above, fibre bridging increases the toughness of FRCs. However, the addition of
silica nanoparticles may affect the level of fibre to matrix adhesion. Tsai et al. [15] attributed the
moderate toughness increase in their NS modified GF composites to the improved interfacial
bonding between the matrix and fibre. On the other hand, Wichmann et al. [37] found the
fibre/matrix interface of GF composite with a NS-modified matrix was very weak, leading to a
nearly entire interfacial failure.
In the present work, the matrix contribution towards the fracture energy for GF
composites is discussed, where composites at various CSR and NS concentrations are
compared with the unmodified composite (GF-Control). The effect of particle concentration and
particle size on the composite toughness is also studied. The toughness transfer from bulk to
3
composite is also discussed. In order to identify the toughening mechanisms responsible, the
fracture surfaces of the mode I double cantilever beam (DCB) specimens were examined using
scanning electron microscopy.
2
Experimental
2.1
Materials
An anhydride-cured epoxy was used, with a low viscosity and a long pot life which make it
suitable for resin transfer moulding and resin infusion [38]. The epoxy resin was a diglycidyl
ether of bisphenol A (DGEBA), LY556 from Huntsman Advanced Materials, UK, with an epoxide
equivalent weight (EEW) of 186 g/eq. The curing agent was methyl hexahydrophthalic
anhydride, Albidur HE 600 from Evonik Hanse, Germany, with an anhydride equivalent weight
(AEW) of 170 g/eq. A stoichiometric mixing ratio of 100:91.4 by weight was used.
The glass fibre was a multiaxial ±45° double layer non-crimp fabric (NCF) with an areal
weight of 450 g/m2, from SP Systems, UK. The fabric has two layers of unidirectional fibres with
a ±45° orientation relative to each other, stitched together for the ease of handling. NCF was
chosen because the resulting composites exhibit higher in-plane properties, superior
delamination resistance and better damage tolerance compared to conventional woven
laminates due to the absence of crimp in the fibre architecture [39].
Two types of core-shell rubber (CSR) particles were used, Kane Ace MX156 and MX960
from Kaneka, Belgium. These were supplied as concentrates comprising 25 wt% of CSR
particles dispersed in DGEBA epoxy resin [40]. Note that the use of such a masterbatch can
give a better dispersion than conventional CSRs, which are supplied in a powder form. The
MX196 and MX960 grades differ in their particle size and core material (see Table 1) where the
core material for MX156 has a lower Tg [41, 42].
The silica nanoparticles (NS) used were Nanopox F400 from Evonik Hanse, Germany.
They were supplied dispersed in DGEBA epoxy resin as a colloidal mixture containing 40 wt%
of silica (SiO2). The particles were supplied surface-modified to prevent agglomeration, and
have a narrow particle size distribution with an average particle diameter of 20 nm [43].
The particle-modified epoxy formulations were prepared by stirring the as-received CSR
and NS concentrates into the epoxy resin at various concentrations. A Heidolph RZR overhead
stirrer was used, fitted with a bladed impeller. The curing agent was added stoichiometrically to
4
these mixtures, and stirred again. The mixtures were degassed prior to infusion to remove any
air bubbles resulting from the stirring process.
2.2
Manufacturing of GF Composites
The CSR and NS particles were used at concentrations of 2, 5, 10 and 15 wt%. These
mixtures were infused into 8 ply glass fibre preforms, arranged as a quasi-isotropic lay-up of
[+45/-45/90/0]s to give laminates with a central 0°/0° ply interface. An additional 2 layers of
carbon fibre (CF) preforms were added either side of the GF preform stacks prior to infusion.
This increased the laminate stiffness to avoid large deflections of the arms during the mode I
fracture tests. A layer of non-perforated, 25 µm thick fluorinated ethylene propylene (FEP)
release film was inserted into the centre of the lay-up to form a 45-55 mm long starter crack.
Resin infusion under flexible tooling (RIFT) was used to prepare the 300 mm x 150 mm plates.
The RIFT equipment setup used is similar to that described by Donadon et al. [44, 45]. A full
vacuum was applied at all times during the process, and was measured to be approximately
0.98 bar below atmospheric. Once infusion was complete, the laminates were cured for 1 hour
at 90 °C, plus 2 hours at 160 °C.
The inclusion of the particles into the resin did not cause significant permeability issues
during infusion, and the particles were found to be fully infiltrated into the laminate.
Nonetheless, the resin flow speed did decrease with increase particle concentration and this
extended the infusion time. The use of polyethylene spiral tube (supplied by East Coast
Fibreglass Supplies, UK), placed around the laminates helped to regulate the resin flow into the
dry perform by creating a uniform flow front, thus distributing the resin evenly across the entire
laminate.
2.3
Mode I Interlaminar Fracture Test
Mode I interlaminar fracture toughness tests were carried out in accordance with ASTM
D5528 [46] and BS EN ISO15024:2001 [47]. Double cantilever beam (DCB) specimens, of 150
mm x 20 mm were cut from the 4-mm-thick plates. The ends of the DCBs containing the starterfilm were gritblasted and acetone-wiped, and aluminium endblocks were bonded on using a
two-component epoxy paste adhesive, Araldite 2011 from Huntsman, UK. The adhesive was
cured at room temperature overnight.
5
The DCBs were tested at a loading rate of 1 mm/min using an Instron 5584 universal
testing machine with a 5 kN load cell. The specimen was loaded in two stages. The initial
loading (pre-cracking) generated a natural crack approximately 10 mm beyond the film insert.
The specimen was unloaded, and then re-loaded until the crack propagated another 70-80 mm.
Load-displacement curves were recorded, and the crack growth was monitored using a
travelling microscope. The specimen was then unloaded, and subsequently broken open to
allow the fracture surfaces to be imaged.
The fracture energy, GIC, was calculated using the modified beam theory (MBT) method
using [48]:
𝑮𝑰𝑪 =
𝟑𝑷(𝜹/𝑵)𝑭
𝟐𝑩(𝒂 + ∆)
Equation 1
where P is the applied load,  is the displacement, and B is the specimen width. The
endblock correction factor, N, and the large displacement correction factor, F, were applied as
described in the Standard [47]. The crack length correction factor, , was determined
experimentally by plotting (C/N)1/3 as a function of crack length, where C is the compliance. The
resulting straight line intersects the crack length axis at -, and a value of  = 0 is used if a
positive intercept is obtained, as specified by the Standard [47].
The initiation value for GIC was determined using three approaches, i.e. visual observation
(VIS), deviation from non-linearity (NL) and the 5% offset or maximum load during the test (i.e.
after pre-cracking) [47]. The GIC initiation (GIC Init) value quoted is the non-linear value, as
accurate determination of the value from visual observation is always difficult and operator
dependant. The propagation values of the fracture energy (GIC Prop) were calculated by
averaging the values in the plateau region after the initiation point. Three replicate DCB
specimens were used in each case.
2.4
Microscopy Analysis
Microscopy of the fracture surfaces used a Hitachi S-3400N scanning electron
microscope (SEM), and a Leo Gemini 1525 field emission gun SEM (FEGSEM). A thin layer of
gold or chromium was sputtered onto the specimens prior to imaging. An accelerating voltage of
15 kV and a working distance of 5 to 7 mm were used. Images were taken from both the
initiation and propagation regions.
6
3
Results
3.1
Introduction
The mode I fracture energies, at initiation and propagation, for the GF composites with
CSR and NS particle concentrations ranging from 2 – 15 wt% are discussed. The fracture
behaviour and the toughening mechanisms of these composites are identified from electron
micrographs, to understand the correlation between the fracture energy, the type of crack
growth and detailed features of the associated fracture surfaces. However, firstly the relevant
bulk matrix properties are presented.
3.2
Bulk Polymer Tg and Young’s Modulus
The properties of the anhydride-cured epoxy used in the present work when modified with
CSR particles have been investigated by Giannakopoulos et al. [17] and with NS particles by
various authors [23-25]. The relevant properties are quoted in Table 2. A glass transition
temperature, Tg, of 143°C was measured for the unmodified epoxy using differential scanning
calorimetry. (Note that some authors measured the T g using dynamic mechanical analysis and a
higher value of 153°C was measured [23], as expected). The addition of these particles did not
significantly affect the Tg values, as they lie within the reported experimental error of ±3°C.
A Young’s modulus, E, of about 2.9 GPa was measured for the unmodified epoxy. For the
CSR modified epoxy, the Young’s modulus decreased approximately linearly with increasing
particle concentration, due to the low modulus of the rubber [17]. The effect of the particle
choice was clear as the E for the epoxy with the MX960 CSR particles was lower than that using
the MX156 CSR, and this was attributed to the larger volume of the soft core compared to the
total particle volume [17]. The addition of NS particles increased the Young’s modulus because
the modulus of the NS particles is much higher than that of the matrix [25].
3.3
3.3.1
GF Composites with Unmodified Matrices (GF-Control)
Mode I Fracture Energy
Fracture of the GF-Control occurred in a stable manner, and no stick-slip crack growth
was observed. However, the R-curves showed some small fluctuations in the propagation
region. An initiation fracture energy of GIC Init = 328 ± 9 J/m 2, and a propagation value of 819 ±
35 J/m2 was measured. These values indicate that a significant R-curve is present, where the
fracture energy increases with crack length.
7
3.3.2
Microscopy Analysis
The low magnification SEM images of the GF-Control fracture surfaces are shown in
Figure 1. The crisscross stitching pattern in Figure 1a is typical for NCF composites with a 0°/0°
ply interface. The fracture involved fibre debonding, which occurred predominantly at the
fibre/matrix interface, and pull-out. This resulted in a mixture of clean debonded and embedded
fibres (see Figure 1b). Ridges and riverlines were present in the intermatrix spacing (i.e. the
area between two embedded fibres) and in resin-rich areas (see Figure 1c). These features are
typical of fracture in a brittle matrix.
Crack branching or jumping was rarely observed as the crack stayed mainly in the same
plane. The effect of the similar ply interfaces and the brittle matrix causes the crack to initiate
easily and propagate in a stable manner, which explains the relatively low G IC Init and GIC Prop
values when compared to composites with different ply interfaces. However, the significant Rcurve is due to the poor interfacial adhesion between the fibres and the matrix, which results in
fibre debonding over a considerable length, followed by pull-out and bridging. These processes
absorb energy and result in GIC Prop being significantly higher than GIC Init.
3.4
3.4.1
GF Composites with MX156 CSR Toughened Matrices (GF/MX156)
Mode I Fracture Energy
Example R-curves for the GF/MX156 composite samples are shown in Figure 2. The
relatively constant plateau values of the R-curve indicate that stable crack growth occurred. The
calculated fracture energies are summarised in Table 3. Compared to the Control, the addition
of the CSR particles increased considerably the fracture energy at initiation and propagation.
The initiation fracture energies increased approximately linearly with the CSR content, to a
maximum of GIC Init = 829 ± 22 J/m2. These results are in broad agreement with the
corresponding bulk polymer properties reported by Giannakopoulos et al. [17], indicating that
the initiation values are dominated by the matrix toughness. However, the composite values are
slightly higher than the bulk due to the extra toughening mechanisms from the glass fibres.
The propagation values were significantly higher than the initiation fracture energies, due
to extra toughening from the fibres. However, it is noteworthy that the increase in toughness
with CSR content is relatively slight. Previous authors have suggested that a poor dispersion
quality of CSR particles in the epoxy matrix will reduce the toughening efficiency, as discussed
8
by Becu et al. [19] and Qian et al. [49]. However, microscopy showed that the CSR particles
were well-dispersed, so this does not apply here.
3.4.2
Microscopy Analysis
Figure 3 shows SEM images of the fracture surfaces of the GF/MX156 composites with 5
wt% and 15 wt% particle concentrations at a low magnification. The fracture was mainly
cohesive near the interface, where the crack propagated through the matrix layer leaving the
fibres mostly embedded in the matrix, indicating a strong fibre to matrix interfacial adhesion.
Loose and broken fibres were seen, plus fibre pull-out, but these were less common than for the
GF-Control discussed above. However, the increased interfacial adhesion and the increased
matrix toughness mean that each pulled-out fibre contributes more energy, and hence a similar
magnitude of R-curve is attained compared to the GF-Control.
The fracture surfaces of the 15 wt% CSR samples in Figure 3b were much rougher than
those of the 5 wt% CSR, with more river lines and ridges, indicating a more intense deformation
and higher toughness owing to the higher CSR particle concentration.
Figures 4 and 5 show high-magnification micrographs of the fracture surfaces, with 5 wt%
and 15 wt% CSR particle concentrations respectively. Voided CSR particles and matrix
deformation on the fracture surfaces are visible in these images. These voids are formed due to
cavitation of the cores of the particles, followed by void growth via plastic deformation of the
epoxy. Once cavitation has occurred, the reduction of constraint allows the epoxy to plastically
deform. These mechanisms are responsible for the increase in the fracture energy. Based on
the void distribution shown in these images, the CSR particles were dispersed homogeneously
across the entire matrix, with no agglomeration observed. Cavitation also indicates that there is
excellent adhesion between the epoxy matrix and the PMMA shell of the CSR particles. The
void diameters were measured using ImageJ software [50], and a mean diameter of 110 ± 27
nm was obtained. This is only slightly larger than the nominal core diameter of 100 nm quoted
by the manufacturer, see Table 1. However, Giannakopoulos et al. [17] measured the actual
core diameter using atomic force microscopy to be 58 ± 13 nm. (Note that some variation in the
measured value is expected due to the particles being sectioned at random plane during
microtome process). This large increase in the mean diameter, from 58 nm for the uncavitated
particles to 110 nm after cavitation and void growth, shows that the void growth mechanism is
very important and will be responsible for a significant proportion of the increase in toughness.
9
For 15 wt% of CSR particles, the image in Figure 5b shows the extensive deformation of the
epoxy matrix.
3.5
3.5.1
GF Composites with MX960 CSR Toughened Matrices (GF/MX960)
Mode I Fracture Energy
For the GF/MX960 composites, GIC Init increased approximately linearly with increasing
CSR concentration (see Table 3). Indeed, the measured values were not significantly different
from those for the GF/MX156 composites. The GF/MX960 composite with 15 wt% CSR showed
the highest initiation value, being over 2.5 times that of the Control.
As for the MX156 composites, the propagation fracture energies were somewhat higher
than the initiation values, due to extra toughening from the fibres. However, it is noteworthy that
the increase in toughness with CSR content is not significant, the values typically being around
1000 J/m2. This means that the GIC Prop values are only about 1.2 times that of the Control.
This suggests that increasing the particle concentration does not necessarily result in higher
composite propagation toughness, due to the changes in the fibre toughening mechanisms, as
discussed below.
3.5.2
Microscopy Analysis
Figures 6 and 7 show the SEM images of the fracture surfaces of the GF/MX960
composites with 5 wt% and 15 wt% of CSR particles respectively. Similar to the GF/MX156
composites, these show a cohesive fracture surface with high degree of surface roughness (see
Figure 6a and 7a) suggesting a strong fibre to matrix interfacial adhesion. Embedded fibres
were observed on the fracture surface as the crack propagated through the matrix layer. Loose
and broken fibres or fibre pull-out were present, but in smaller numbers than for the GF-Control.
The observed R-curve was caused by bridging and pull-out, but the higher interfacial adhesion
and matrix toughness meant that fewer fibres are required to contribute to the fibre toughening
mechanisms. This also means that the difference between the values of GIC Prop and GIC Init is
smaller than for the Control. Cavitation of the large MX960 particles is clearly observed even at
low magnifications (see Figure 6b and 7b).
At high magnification, the fracture surfaces are covered in voids caused by internal
cavitation of the rubbery cores, while the shells are still bonded to the matrix. These cavities are
well dispersed in the epoxy matrix. The cavities vary widely in size, with the diameter ranging
10
from 200 to 750 nm in Figure 6c, and from 150 to 600 nm in Figure 7c. Note that these cavities
are fractured at a random plane, resulting in the variation of size, and resulting in some cavities
being smaller than the original particles if they fractured close to one of the poles. As the
cavities are substantially larger than the particle diameter of 186 ± 100 nm [17] (see Table 1),
this indicates that the cavities continued to enlarge by plastic deformation of the epoxy after the
cores cavitated. Although the CSR particles were fully-cavitated and relieved the constraint at
the crack tip, the cavitation process is considered generally to absorb little energy compared to
the plastic deformation in the matrix.
3.6
3.6.1
GF Composites with Silica Nanoparticle Toughened Matrices (GF/NS)
Mode I Fracture Energy
All of the GF/NS composite samples showed rising R-curves, indicating that fibre bridging
occurs during crack propagation. The fracture energy results (see Table 3) were significantly
lower than those for the CSR particle-modified composites. Indeed, no significant toughness
improvement was measured for the 2 wt% and 5 wt% NS concentrations.
The lack of toughening efficiency of the composites with NS modified matrices has also
been observed by Hsieh et. al. [25] for their quasi-isotropic (QI) GF composites and Zeng et. al.
[20] for their woven CF composites. (It is worth noting that the composites in Hsieh et. al. [25]
were manufactured using the same epoxy-hardener combination, cured using the same curing
profile and had the same midply interface configuration as in the present work.) Zeng et. al. [20]
argued that the presence of much stiffer fibres may have constrained the matrix deformation,
and that a weakened fibre/matrix interface contributed to the poor fracture resistance in the
composite. However, in the present work the fibre to matrix adhesion is relatively good. This is
discussed further below.
3.6.2
Microscopy Analysis
The fracture surfaces of the composites containing 5 wt% and 15 wt% of silica
nanoparticles are shown in Figures 8 and 9 respectively. The fracture occurred cohesively in the
resin-rich layer in-between the midply. Loose and pulled-out fibres were present on the fracture
surfaces. The debonded fibres were covered with matrix indicating a good fibre to matrix
interfacial adhesion. The matrix fracture, on the other hand showed a relatively flat surface with
no large scale plastic deformation except for textured microflow and riverlines.
11
At a higher magnification for the samples with 15 wt% of silica nanoparticles (see Figure
9c), it is difficult to distinguish between the silica nanoparticles and the chromium coating used
to make the surface conductive for microscopy. Particle debonding and void growth in the matrix
are hard to identify, as the voids would be approximately 30 nm in diameter. These voids are
much smaller than those created by the cavitation of even the smaller CSR particles (which are
about 90 nm in diameter). It is thought that the coating has covered any voids present on the
fracture surface, as has been noted in previous work [31]. This is confirmed by examination of
Figure 8c, which shows the composite with 5 wt% NS, where voids are clearly visible on the
fracture surface. This confirms that debonding and void growth does occur, and hence the
toughness should increase. However, previous work has shown that only 14.3% of the silica
nanoparticles will undergo debonding and void growth, and hence the toughening effect is small
(and is lost within the experimental error at this low silica concentration).
4
Discussion
4.1
Effect of Particle Size and Particle Concentration
The relationships between the fracture energy and particle concentration for the bulk and
the GF composite (GIC Init) are shown in Figure 10. The bulk polymer data are taken from
previous research by Masania and co-workers [17, 24]. A general trend can be seen from these
figures whereby the addition of CSR and NS particles greatly enhanced the bulk and composite
toughness. The similarity of the bulk and composite data at initiation suggests a dominant matrix
influence. This is expected as GIC Init is often considered to be a material property of composite
materials and to be independent of fibre bridging effects [51]. It is also apparent that the CSR
particles provide a more significant toughening effect than the NS particles.
The CSR particle size had little influence on the toughness when the error bars are
considered. Kim et al. [11] have demonstrated that the toughness of such systems is more
dependent on particle content rather than size because of limited shear deformation in the
matrix. The cavitation resistance of rubber particles increases with decreasing particle size [52],
so the MX156 would be expected to show a greater cavitational resistance than MX960.
However, all of the particles cavitated, so the triaxial stresses which cause cavitation were
clearly high enough to overcome the cavitation resistance. As all of the particles cavitated then
the same toughening mechanisms were activated, hence the toughness of the CSR-modified
epoxies was approximately the same, and there was no size effect.
12
For toughening with NS particles, the increase in toughness is less steep with particle
concentration than for the CSR particles (see Figure 10). Only one particle size was used in the
present work, but Dittanet and Pearson [27] used NS particles 23, 74 and 170 nm in diameter
with bulk epoxy polymer samples. They concluded that the toughness increased with increasing
NS particle concentration, and that the NS particle size has a negligible effect on the fracture
toughness. The independence of bulk fracture toughness on the NS particle size has also been
reported by Liang and Pearson [30] and Bray et al. [31]. Indeed, Bray et al. [31] used modelling
to confirm that particle diameters from 23 nm to 170 nm had no effect on the toughness.
With respect to the toughening mechanisms, the silica nanoparticles are less effective
than the core shell rubber particles. Both types of particles would be expected to initiate shear
yielding. All of the core-shell rubber particles would be expected to cavitate, and the
micrographs shows that this does occur. However, previous work has shown that only 14.3% of
the silica nanoparticles will show debonding and void growth [31]. Examination of the
micrographs for 5 wt% NS indicates that the silica nanoparticles undergo debonding and void
growth. The results show that ΔGIC is similar for GIC Bulk and for GIC Init, which implies that the
toughening mechanisms would be expected to be the same. Hence, debonding and void growth
would be expected, but this was difficult to observe using microscopy for the 15 wt% NS
composite, as has been noted previously [31].
A different trend was observed for GIC Prop, whereby there was no clear relationship
between the fracture energy and the particle concentration (see Table 3), due to the influence of
the fibres. However, comparing the two types of CSR particles, the effect of the rubber particle
size was insignificant.
In CSR modified epoxy, previous work has shown that the degree of particle dispersion
strongly affected the fracture toughness whereby a uniform dispersion of CSR particles can lead
to a higher fracture energy [49]. From microscopy analysis, both types of CSR particles were
observed to be well dispersed in the present work. It should be noted that there was strong
interfacial adhesion between the shell of the CSR particles and the matrix, as evidenced by the
cavitated rubbery cores on the fracture surfaces. Otherwise, had the interfacial adhesion been
poor, debonding of CSR particles would have occurred.
4.2
Toughness Transfer
The efficiency of the transfer of toughness from the bulk to the GF composites can be
considered by plotting the values of GIC composite vs. GIC bulk. Figure 11 shows the
13
relationship for the CSR particles (MX156 and MX960), and Figure 12 for the silica
nanoparticles. The diagonal line represents the 1 to 1 relationship between the GIC values,
indicating that a full transfer of matrix toughness to the composite has been achieved. In all
cases, the composite fracture energy is greater than that of the bulk.
In general there is an increase in bulk toughness with increasing particle concentration,
and this gives a positive effect in improving the composite fracture resistance. The G IC Init
values tend to lie slightly above the 1:1 relationship, and the G IC Prop values further above due
to the additional fibre toughening mechanisms such as bridging [53]. However, the effect of fibre
bridging in the modified matrices is somewhat reduced compared to the Control due to the
better interfacial adhesion noted above, which reduces the length of fibre which pulls out and
bridges. The fibre to matrix adhesion is also better than that observed by Hsieh et al. [24] for a
CTBN-modified matrix.
The correlation between the matrix toughness (GIC bulk) and the composite toughness
(GIC composite) has been discussed in several studies [54-56]. These concluded that in brittle
matrix composites, the GIC composite is higher than GIC bulk indicating a full toughness transfer
from matrix to composite. The higher composite toughness was due to toughening mechanisms
which are not present in the bulk. In composites with tougher matrices there is only a partial
transfer of matrix to composite toughness due to the presence of the fibres, which restrict the
plastic zone size. Bradley [57] as cited by Siddiqui et al. [55] has established criteria for the level
of toughness transfer based on the bulk toughness as follows:
i).
For GIC bulk lower than 500 J/m 2, GIC composite is greater than GIC bulk and there is
an approximately linear correlation between these toughness values.
ii).
For GIC bulk between 500 J/m 2 and 2000 J/m 2, the toughening increment of GIC
composite is much smaller than GIC bulk.
iii).
For GIC bulk above 2000 J/m 2, there is little increase in GIC composite.
The GF composites at initiation for the CSR and the NS modified matrices are located
slightly above the diagonal line suggesting a slightly higher than 1:1 toughness transfer. The
toughness transfer was also found to increase linearly with increasing particle concentration and
seemed to follow the established trend (i) for GIC bulk lower than 500 J/m2.
In the propagation region, the GIC values for the composites with modified matrices fall
well above the diagonal line indicating the fracture toughness is largely affected by the fibre
toughening effect.
14
It has been argued that GF composites with tougher matrices (500 to 2000 J/m 2) can only
achieve partial toughness transfer as the GIC composite is lower than GIC bulk [55, 57].
However, this is not the case for the results obtained in this study as all G IC composite values
were higher than the GIC bulk. This discounted the conclusion that the stiff fibres restricted the
deformation of the plastic zone. The size of the plastic zone is calculated below to support this
argument. Indeed the data better support the conclusions of Hunston et al. [58] who proposed
that full toughness transfer occurs up to a fracture energy of 1000 J/m 2. Therefore, it can be
said that the increase in the matrix toughness for CSR and NS modified matrices is fully
transferred to the composites.
4.3
Plastic Zone Size
The plane strain dimension of the plastic zone can be calculated based on Irwin’s model,
assuming that the zone is circular and the crack occurs in the matrix region, using Equation 2
[59].
𝒓𝒚 =
𝟏 𝑲𝑰𝑪
(
)
𝟔𝝅 𝝈𝒚
𝟐
Equation 2
where KIC is the fracture toughness and σy is the yield stress of the bulk polymer. The
calculation uses the bulk properties for the CSR and NS modified epoxies from previous work
[17, 24], interpolating where necessary.
A plastic zone radius of 1.8 μm was calculated for the Control. The maximum plastic zone
size of 13.4 μm radius was calculated for the 15 wt% MX156. The other formulations fall
between these two limits. The plastic zones are substantially larger than the radius of the CSR
and NS particles, so the particles readily fit within the zone and hence can actively toughen the
matrix. When comparing the plastic zone size and the thickness of the matrix region between
the plies, even the largest plastic zone diameter of MX156 at 15 wt% concentration (i.e. ~27
m) is still smaller than the matrix thickness. Therefore it can be concluded that the presence of
fibres in the composite did not impose any significant restriction on the damage zone and so the
bulk toughness is fully transferred to the composite.
5
Conclusions
The fracture behaviour of glass fibre (GF) composites toughened with core-shell rubber
particles (CSR) or silica nanoparticles (NS) was investigated using mode I interlaminar fracture
tests. Two types of CSR particles with different particle size and core/shell materials were
15
employed in this study. The effects of particle concentration, particle size and interfacial
adhesion on the composite toughness were discussed.
The addition of CSR or NS particles increased the fracture toughness of the bulk polymer
and GF composites, and fracture toughness increased approximately linearly with increasing
particle concentration. The CSR particles provided a far larger toughening effect than the NS
particles. The particle size has no influence on the toughness of the CSR modified epoxy
composites. The composites achieved full toughness transfer as the GIC composite values are
much higher than GIC bulk. The effect of fibre restriction on the damage zone has been
discounted as the matrix thickness is much larger than the plastic zone size.
The toughening mechanisms were identified using scanning electron microscopy. The
particles were well-dispersed in the epoxy. All of the CSR particles cavitated, giving increased
toughness by plastic void growth and shear yielding. It appeared that the NS particles did not
debond, hence limiting the plastic deformation in the matrix and resulting in low fracture energy.
6
Acknowledgements
The authors gratefully acknowledge the Malaysian Government Agency, MARA for PhD funding
for Dr Shamsiah Awang Ngah, and additional funding from the Department of Mechanical
Engineering at Imperial College London. They would also like to thank Evonik Hanse and
Kaneka for the supply of materials and technical support. Some of the equipment used in this
study was provided by Dr. Ambrose Taylor’s Royal Society Mercer Junior Award for Innovation.
References
[1] Yan C, Xiao K, Ye L, Mai Y-W. Numerical and experimental studies on the fracture behavior of rubbertoughened epoxy in bulk specimen and laminated composites. Journal of Materials Science.
2002;37(5):921-7.
[2] Kumar P, Singh RK. Impact damage area and interlaminar toughness of modified FRP laminates.
Advanced Composite Materials. 2000;9(2):77-88.
[3] Sue HJ, Jones RE, Garcia-Meitin EI. Fracture behaviour of model toughened composites under Mode I
and Mode II delaminations. Journal of Materials Science. 1993;28(23):6381-91.
[4] Hunston DL, Moulton RJ, Johnston NJ, Bascom W. Matrix resin effects in composite delamination:
mode I fracture aspects. Toughened Composites, ASTM STP. 1987;937:74-94.
[5] Bonnaud L, Pascault JP, Sautereau H, Zhao JQ, Jia DM. Effect of reinforcing glass fibers on
morphology and properties of thermoplastic modified epoxy-aromatic diamine matrix. Polymer Composites.
2004;25(4):368-74.
[6] Mujika F, De Benito A, Fernández B, Vázquez A, Llano-Ponte R, Mondragon I. Mechanical properties
of carbon woven reinforced epoxy matrix composites. A study on the influence of matrix modification with
polysulfone. Polymer Composites. 2002;23(3):372-82.
16
[7] Wetzel B, Rosso P, Haupert F, Friedrich K. Epoxy nanocomposites – fracture and toughening
mechanisms. Engineering Fracture Mechanics. 2006;73(16):2375-98.
[8] Wetzel B, Haupert F, Qiu Zhang M. Epoxy nanocomposites with high mechanical and tribological
performance. Composites Science and Technology. 2003;63(14):2055-67.
[9] Pearson RA, Yee AF. Influence of particle size and particle size distribution on toughening mechanisms
in rubber-modified epoxies. Journal of Materials Science. 1991;26(14):3828-44.
[10] Becu L, Maazouz A, Sautereau H, Gerard JF. Fracture behavior of epoxy polymers modified with
core-shell rubber particles. Journal of Applied Polymer Science. 1997;65(12):2419-31.
[11] Kim DS, Cho K, Kim JK, Park CE. Effects of particle size and rubber content on fracture toughness in
rubber-modified epoxies. Polymer Engineering & Science. 1996;36(6):755-68.
[12] Keller A, Masania K, Taylor AC, Dransfeld C. Fast-curing epoxy polymers with silica nanoparticles:
Properties and rheo-kinetic modelling. Journal of Materials Science. 2016;1(1): in press.
[13] Bagheri R, Marouf BT, Pearson RA. Rubber-toughened epoxies: A critical review. Polymer Reviews.
2009;49(3):201-25.
[14] Day RJ, Lovell PA, Wazzan AA. Toughened carbon/epoxy composites made by using core/shell
particles. Composites Science and Technology. 2001;61(1):41-56.
[15] Tsai J-L, Huang B-H, Cheng Y-L. Enhancing fracture toughness of glass/epoxy composites by using
rubber particles together with silica nanoparticles. Journal of Composite Materials. 2009;43(25):3107-23.
[16] Brown HR, Schneider JA, Murphy TL. Experimental studies of the deformation mechanisms of coreshell rubber-modified diglycidyl ether of bisphenol-A epoxy at cryogenic temperatures. Journal of
Composite Materials. 2013.
[17] Giannakopoulos G, Masania K, Taylor AC. Toughening of epoxy using core–shell particles. Journal of
Materials Science. 2011;46(2):327-38.
[18] Liu J, Sue H-J, Thompson ZJ, Bates FS, Dettloff M, Jacob G, et al. Nanocavitation in self-assembled
amphiphilic block copolymer-modified epoxy. Macromolecules. 2008;41(20):7616-24.
[19] Liu H-Y, Wang G-T, Mai Y-W, Zeng Y. On fracture toughness of nano-particle modified epoxy.
Composites Part B: Engineering. 2011;42(8):2170-5.
[20] Zeng Y, Liu H-Y, Mai Y-W, Du X-S. Improving interlaminar fracture toughness of carbon fibre/epoxy
laminates by incorporation of nano-particles. Composites Part B: Engineering. 2012;43(1):90-4.
[21] Guild FJ, Kinloch AJ, Taylor AC. Particle cavitation in rubber toughened epoxies: the role of particle
size. Journal of Materials Science. 2010;45(14):3882-94.
[22] Kinloch AJ, Lee JH, Taylor AC, Sprenger S, Eger C, Egan D. Toughening structural adhesives via
nano- and micro-phase inclusions. The Journal of Adhesion. 2003;79(8):867 - 73.
[23] Johnsen BB, Kinloch AJ, Mohammed RD, Taylor AC, Sprenger S. Toughening mechanisms of
nanoparticle-modified epoxy polymers. Polymer. 2007;48(2):530-41.
[24] Hsieh TH, Kinloch AJ, Masania K, Sohn Lee J, Taylor AC, Sprenger S. The toughness of epoxy
polymers and fibre composites modified with rubber microparticles and silica nanoparticles. Journal of
Materials Science. 2010;45(5):1193-210.
[25] Hsieh TH, Kinloch AJ, Masania K, Taylor AC, Sprenger S. The mechanisms and mechanics of the
toughening of epoxy polymers modified with silica nanoparticles. Polymer. 2010;51(26):6284-94.
[26] Jumahat A, Soutis C, Jones F, Hodzic A. Effect of silica nanoparticles on compressive properties of an
epoxy polymer. Journal of Materials Science. 2010;45(21):5973-83.
[27] Dittanet P, Pearson RA. Effect of silica nanoparticle size on toughening mechanisms of filled epoxy.
Polymer. 2012;53(9):1890-905.
[28] Chen C, Justice RS, Schaefer DW, Baur JW. Highly dispersed nanosilica-epoxy resins with enhanced
mechanical properties. Polymer. 2008;49(17):3805-15.
17
[29] Ragosta G, Abbate M, Musto P, Scarinzi G, Mascia L. Epoxy-silica particulate nanocomposites:
Chemical interactions, reinforcement and fracture toughness. Polymer. 2005;46(23):10506-16.
[30] Liang YL, Pearson RA. Toughening mechanisms in epoxy–silica nanocomposites (ESNs). Polymer.
2009;50(20):4895-905.
[31] Bray DJ, Dittanet P, Guild FJ, Kinloch AJ, Masania K, Pearson RA, et al. The modelling of the
toughening of epoxy polymers via silica nanoparticles: The effects of volume fraction and particle size.
Polymer. 2013;54(26):7022-32.
[32] Kinloch A, Mohammed R, Taylor A, Sprenger S, Egan D. The interlaminar toughness of carbon-fibre
reinforced plastic composites using ‘hybrid-toughened’ matrices. Journal of Materials Science.
2006;41(15):5043-6.
[33] Sohn Lee J. The fatigue behaviour of nano-modified epoxy adhesives [PhD Thesis]. London: Imperial
College London; 2009.
[34] Manjunatha C, Taylor AC, Kinloch AJ, Sprenger S. The effect of rubber micro-particles and silica
nano-particles on the tensile fatigue behaviour of a glass-fibre epoxy composite. Journal of Materials
Science. 2009;44(1):342-5.
[35] Manjunatha C, Taylor A, Kinloch A, Sprenger S. The cyclic-fatigue behaviour of an epoxy polymer
modified with micron-rubber and nano-silica particles. Journal of Materials Science. 2009;44(16):4487-90.
[36] Mahrholz T, Stängle J, Sinapius M. Quantitation of the reinforcement effect of silica nanoparticles in
epoxy resins used in liquid composite moulding processes. Composites Part A: Applied Science and
Manufacturing. 2009;40(3):235-43.
[37] Wichmann MHG, Sumfleth J, Gojny FH, Quaresimin M, Fiedler B, Schulte K. Glass-fibre-reinforced
composites with enhanced mechanical and electrical properties - Benefits and limitations of a nanoparticle
modified matrix. Engineering Fracture Mechanics. 2006;73(16):2346-59.
[38] Hunstman. Araldite LY556. Technical Datasheet. Duxford, UK; October 15, 2010.
[39] Mattsson D. Mechanical performance of NCF composites [PhD Thesis Online]. Sweden: Luleå
University of Technology; 2005.
[40] Sober DJ. Kaneka Core-Shell Toughening Systems for Thermosetting Resins. Houston, Texas:
Kaneka Texas Corporation; n.d.
[41] Kaneka Texas Corporation. Kane Ace MX156. Technical Datasheet. Houston,Texas; n.d.
[42] Kaneka Texas Corporation. Kane Ace MX960. Technical Datasheet. Houston,Texas; n.d.
[43] Nanoresins AG. Nanopox F400. Technical Datasheet. Geesthacht, Germany; April 2006.
[44] Donadon MV, Falzon BG, Iannucci L, Hodgkinson JM. Measurement of fibre fracture toughness using
an alternative specimen geometry. 16th International Conference On Composite Materials. Kyoto, Japan:
Japan Society for Composite Materials; 2007. p. 7.
[45] Donadon M, Hodgkinson J, Falzon B, Iannucci L. The reliability of the resin infusion under flexible
tooling process for manufacturing composites aerostructures. Proceedings of the 13th SICOMP
conference. Gothenburg, Sweden: SICOMP; 2004.
[46] American Society for Testing and Materials. Standard test method for mode I interlaminar fracture
toughness of unidirectional fiber-reinforced polymer matrix composites. ASTM D 5528–01. West
Conshohocken, United States: ASTM International; 2001.
[47] British Standards Institute. Fibre-reinforced plastic composites - Determination of mode I interlaminar
fracture toughness, GIC, for unidirectionally reinforced materials. BS ISO 15024:2001. London: BSI; 2001.
[48] Hodgkinson JM. Mechanical testing of advanced fibre composites. Cambridge: Woodhead Publishing;
2000. p. xv, 362 p.
[49] Qian JY, Pearson RA, Dimonie VL, Shaffer OL, El-Aasser MS. The role of dispersed phase
morphology on toughening of epoxies. Polymer. 1997;38(1):21-30.
[50] Rasband WS. ImageJ. Bethesda, Maryland, USA: U.S. National Institutes of Health; 1997-2012.
18
[51] Kim K-Y, Ye L. Interlaminar fracture toughness of CF/PEI composites at elevated temperatures: roles
of matrix toughness and fibre/matrix adhesion. Composites Part A: Applied Science and Manufacturing.
2004;35(4):477-87.
[52] Bécu-Longuet L, Bonnet A, Pichot C, Sautereau H, Maazouz A. Epoxy networks toughened by core–
Shell particles: Influence of the particle structure and size on the rheological and mechanical properties.
Journal of Applied Polymer Science. 1999;72(6):849-58.
[53] Greenhalgh E, Hiley M. The assessment of novel materials and processes for the impact tolerant
design of stiffened composite aerospace structures. Composites Part A: Applied Science and
Manufacturing. 2003;34(2):151-61.
[54] Madhukar MS, Drzal LT. Fiber-matrix adhesion and its effect on composite mechanical properties: IV.
Mode I and mode II fracture toughness of graphite/epoxy composites. Journal of Composite Materials.
1992;26(7):936-68.
[55] Siddiqui NA, Woo RSC, Kim J-K, Leung CCK, Munir A. Mode I interlaminar fracture behavior and
mechanical properties of CFRPs with nanoclay-filled epoxy matrix. Composites Part A: Applied Science
and Manufacturing. 2007;38(2):449-60.
[56] Compston P, Jar PYB, Burchill PJ, Takahashi K. The transfer of matrix toughness to composite mode
I interlaminar fracture toughness in glass-fibre/vinyl ester composites. Applied Composite Materials.
2002;9(5):291-314.
[57] Bradley WL. Understanding the Translation of Neat Resin Toughness into Delamination Toughness in
Composites. Key Engineering Materials. 1989;37:161-98.
[58] Hunston DL, Moulton RJ, Johnston NJ, Bascom WD. Matrix resin effects in composite delamination:
Mode I fracture aspects. In: Johnston NJ, editor. Toughened composites. Philadelphia, USA: ASTM; 1987.
p. 74-94.
[59] Kinloch AJ, Young RJ. Fracture behaviour of polymers. London: Chapman & Hall; 1983.
[60] Kaneka Belgium N.V. Kane Ace® MX - The novel core shell rubber toughening system for Epoxy
Resin. Product Guide. Belgium: Kaneka Belgium NV; n.d.
19
Table 1 : Properties of the core-shell rubber particles [17, 41, 42, 60].
CSR
Material
Nominal Diameter
Measured Diameter
Type
Core
Shell
Core (nm)
Shell (nm)
Core (nm)
MX156
Polybutadiene
PMMA
100
85-115
58 ± 13
MX960
Siloxane
PMMA
300
250-350
186 ± 100
Table 2 : Bulk polymer properties from [17] and [23, 24].
CSR toughened epoxy [17]
Bulk Property
Control
Tg DSC (°C)
145
NS toughened epoxy [23, 24]
9 wt%
9 wt%
MX156
MX960
143
146
Tg DMA (°C)
E (GPa)
2.76 ± 0.08
2.33 ± 0.04
8 wt%
11 wt%
NS
NS
143
136
141
153
154
151
2.96
3.42
3.57
Control
2.21 ± 0.02
Table 3 : Mode I interlaminar fracture energy for GF composites with various CSR and NS concentrations.
Mode I Fracture Energy, GIC (J/m2)
Particle
Content,
wt%
MX156
MX960
NS
GIC Init
GIC Prop
GIC Init
GIC Prop
GIC Init
GIC Prop
0%
328 ± 9
819 ± 35
328 ± 9
819 ± 35
328 ± 9
819 ± 35
2%
429 ± 41
1059 ± 52
398 ± 12
798 ± 47
285 ± 37
514 ± 56
5%
455 ± 19
1012 ± 51
543 ± 25
1049 ± 51
297 ± 12
626 ± 46
10%
651 ± 39
1054 ± 40
626 ± 35
1045 ± 59
449 ± 45
849 ± 40
15%
829 ± 22
1475 ± 62
842 ± 23
961 ± 55
499 ± 100
810 ± 25
20
Crack direction
(a)
Crack direction
(b)
(c)
Crack direction
Intermatrix
spacing
Figure 1 : GF-Control – Fracture surfaces at initiation/propagation showing (a) Crisscross stitching pattern,
(b) Clean debonded fibres, and (c) Brittle matrix deformation at intermatrix spacing.
Fracture Energy, GIC J/m2
2000
1600
1200
800
2%
5%
400
10%
15%
0
20
40
60
80
100
120
Crack Length, a mm
Figure 2 : R-curves of fracture energy versus crack length for GF/MX156 composites at various
concentrations of core-shell rubber particles.
21
(a)
Crack direction
(b)
Crack direction
Embedded fibres
Figure 3 : Fracture surfaces of GF/MX156 composite at low magnification. (a) 5 wt% CSR (b) 15 wt% CSR
(b)
(a)
(c)
Figure 4: Fracture surfaces of GF/MX156 composite at high magnification – 5 wt% CSR.
22
(b)
(a)
Figure 5 : Fracture surfaces of GF/MX156 composite at high magnification – 15 wt% CSR.
(b)
(a)
(c)
Figure 6 : Fracture surfaces of GF/MX960 composite – 5 wt% MX960 CSR.
23
(b)
(a)
(c)
Figure 7: Fracture surfaces of GF/MX960 composite – 15 wt% MX960 CSR.
(a)
(b)
(c)
Figure 8: Fracture surfaces of GF/NS composite – 5 wt% NS.
24
(a)
(b)
(c)
Figure 9 : Fracture surfaces of GF/NS composite – 15 wt%NS.
25
(a)
(b)
Figure 10 : Fracture energy vs. particle concentration (a) Bulk G IC [17, 24], and (b) Composite GIC Init.
26
(a)
(b)
Figure 11 : Toughness transfer at crack initiation and propagation for (a) GF/MX156 composites (b)
GF/MX960 composites. Numbers in brackets represent the particle concentration.
27
Figure 12 : Toughness transfer at crack initiation and propagation for GF/NS composites. Numbers in
brackets represent the particle concentration.
28