Composites Science and Technology 105 (2014) 86–95

Contents lists available at ScienceDirect

Composites Science and Technology

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

Carbon fiber-reinforced composites using an epoxy resin matrix

modified with reactive liquid rubber and silica nanoparticles
Stephan Sprenger a,⇑, Martin Heinz Kothmann b, Volker Altstaedt b
a
Evonik Hanse GmbH, Charlottenburger Str. 9, 21502 Geesthacht, Germany
b
University of Bayreuth, Department of Polymer Engineering, Universitaetsstr. 30, 95440 Bayreuth, Germany

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

Article history: Carbon fiber-reinforced composites are gaining importance and are about to play a very prominent role in
Received 8 July 2014 automotive applications. In structural applications mainly epoxy resins are used as matrix materials. The
Received in revised form 6 September 2014 properties of epoxy resins and laminates made thereof can be improved by tougheners like reactive liquid
Accepted 3 October 2014
rubbers. Further improvements can be achieved by adding surface-modified silica nanoparticles, with
Available online 13 October 2014
20 nm in size and a very narrow particle size distribution. In this study carbon fiber-reinforced laminates
made from epoxy resins modified with reactive liquid rubber and silica nanoparticles have been prepared
Keywords:
and investigated. A very fast amine cure of 15 min has been chosen to match industrial needs. Mechanical
A. Laminate
A. Nanoparticles
properties for bulk resins and laminates are compared and the mechanisms responsible for the property
A. Carbon fibers improvements are discussed. Structure property relationships between the neat resin fracture toughness
B. Fracture toughness and the interlaminar GIc and GIIc of the reactive liquid rubber and silica nanoparticles modified resins
B. Impact behaviour were established. Tough laminates could be prepared. However, CAI performance of the hybrid laminates
was slightly inferior as compared to the rubber-toughened laminates, most probably due to agglomerates
of nanoparticles found in the cured resin systems. The reason for the nanoparticle aggregation was
detected and their influence on laminate performance discussed.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction under investigation. Thermoplastic or thermosetting materials

Environmental requirements like reduced fuel consumption,
lower carbon dioxide emission and an improved carbon footprint
are changing the automotive industry. Especially in Europe, where
carbon dioxide emissions are subject to state regulations [1]. Light-
weight construction becomes the key technology to solve these
issues which currently drives a trend towards the increased use
of fiber-reinforced composites in automotive construction [2].
Automotive could be the third large industrialization of com-
posites technologies, after aerospace and wind energy. In other
areas of application like railway or shipbuilding more and more
fiber-reinforced materials are used as well. Of course automotive
manufacturing has its own rules, one is a very short part produc-
tion cycle, like the stamping of metal parts. Thus composites man-
ufacturing technologies and the materials used have to be adapted
accordingly [3].
Glass or carbon fibers, woven fabrics and unidirectional or mul-
tiaxial nonwovens are used as reinforcements; natural fibers are
⇑ Corresponding author.
E-mail address: stephan.sprenger@evonik.com (S. Sprenger).
are used as matrices. Besides unsaturated polyester resins mainly
epoxy resins are used as thermosetting matrices for automotive
components. The diglycidylether of bisphenol A (DGEBA) is the
most commonly used epoxy resin in automotive industry. Very
important is low viscosity of the resin as the preferred manufactur-
ing methods for automotive parts currently are injection or infu-
sion processes – which enable short cycle times [3].
The combination of epoxy resin and hardener defines the matrix
material and the various possible combinations enable to formulate
according to all possible property requirements [4]. Hare described
the variety of epoxy curing agents a few years ago [5,6]. Most pop-
ular for automotive part manufacturing are aliphatic or cycloali-
phatic amines like isophorone diamine (IPD) due to their low
viscosity and fast cure at elevated temperatures. Various accelera-
tors and additives are typically part of the hardener formulation.
However, epoxy resins are inherently brittle. Therefore many
epoxy resin formulations are toughened. In the seventies of last cen-
tury carboxy-terminated nitrile butadienes (CTBNs) were intro-
duced as tougheners. They are immiscible with epoxies but
become miscible after a pre-reaction with an excess of epoxy resin.
These soluble adducts phase separate again upon cure and form
rubber domains in the epoxy resin. The mechanisms of rubber

http://dx.doi.org/10.1016/j.compscitech.2014.10.003
0266-3538/Ó 2014 Elsevier Ltd. All rights reserved.
 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
toughening are well known [7,8]. Rubber toughening has several
drawbacks like lowering the modulus and reducing the glass tran-
sition temperature, due to some rubber molecules not participat-
ing in the phase separation, but crosslinking randomly into the
matrix. Another issue is the relatively high viscosity of epoxy res-
ins containing reactive liquid rubbers, which can be critical for
infusion processes. Nevertheless they are used in composites
manufacturing.
In 2002/2003 surface modified silica nanoparticles became
available in industrial quantities. They are manufactured in situ
directly in the epoxy resin by a modified sol–gel process and have
an average particle diameter of 20 nm as well as a very narrow par-
ticle size distribution [9]. These spherical particles are transparent,
monodisperse and just slightly increase the resin viscosity up to
quite high addition levels. Resins containing such nanoparticles
exhibit no thixotropic properties but behave like Newtonian liq-
uids. The nanosilica does not agglomerate upon cure [9].
Properties like strength, modulus, toughness and especially fati-
gue performance of cured bulk epoxy resins can be improved by a
modification with silica nanoparticles. In 2006 Zhang et al. found
significant improvements for an anhydride-cured cycloaliphatic
epoxy which was modified with silica nanoparticles [10]. GIc
increased with increasing nanosilica addition level, whereas the
impact strength determined by an un notched Charpy test showed
a maximum in impact strength at approximately 5 wt% addition
level. They studied the correlations between the interparticle dis-
tance and the corresponding property improvements and found
the improvements for both stiffness and toughness to be larger
for addition level where the innerparticle distance became smaller
than the particle diameter.
In 1992 Kinloch and Huang postulated that the toughness of
modified epoxy resins is the toughness of the unmodified resin
plus the contribution of the toughening mechanisms defined by
the toughener used [11]. They proposed one of the first models
for rubber toughening.
The property improvements achievable with a silica nanoparti-
cle modification and the mechanisms behind have been subject of
intensive research in the last ten years [12]. Several toughening
mechanisms have been suggested to explain the complex behav-
iour of polymers with nanofillers, like crack deflection, crack
branching, crack pinning/bowing, particle bridging, particle deb-
onding, microcracking or crazing of the matrix and elastic defor-
mation of the matrix. Comparing models with experimental
results, Zhang et al. found only a minor contribution to the tough-
ening by crack deflection, but a major contribution by an enhanced
local deformability around the crack tip, very probably due to void
formation at the crack tip [13].
Today, the mechanisms responsible for the property improve-
ments are rather well understood [14]. Several extensive modelling
studies were conducted. Lauke developed a micromechanical
model for the toughening of polymers with particles taking into
account particle debonding, void formation and subsequent void
growth [15]. He concluded that particle size and particle size distri-
bution will both have a very strong influence on toughness
improvements achievable with fillers. Williams investigated the
toughening from plastic void growth around debonded or cavitated
micro- and nanoparticles in epoxy resins [16]. He predicted that
smaller particles (<50 nm) will increase toughness more than lar-
ger ones but that the improvements only occur when a good par-
ticle dispersion can be achieved.
Zappalorto et al. analyzed the debonding-induced toughness
improvement from nanoparticle debonding from a polymer using
a multiscale model [17]. He concluded that the contribution of
matrix debonding to the toughness increase is only a minor one
(less than 5%) although necessary for void formation and subse-
quent plastic yielding.
87
Kinloch, Taylor et al. studied various epoxy resins cured with
different hardeners and different addition levels of nanosilica
[18]. They found a good agreement between experimentally mea-
sured and predicted values of the fracture energy (GIc). Their model
was based on two major toughening mechanisms: localized plastic
shear bands and debonding of silica nanoparticles followed by sub-
sequent plastic void growth.
A multiscale model for toughness increases due to the plastic
yielding of voids formed by debonded nanoparticles was developed
by Quaresimin et al. [19]. They confirmed that the void growth
mechanism provides a major contribution to the toughness
increase. Furthermore they were able to show that the toughness
improvements depend strongly on nanoparticle size as well. In
continuation of this work, Quaresimin et al. investigated a multi-
mechanism modelling approach to predict toughness increases of
nanoparticle-modified polymers [20]. The major contributions to
fracture toughness improvements were described by separate mod-
els and then combined: particle debonding, plastic void growth of
nanovoids and localized shear banding. The multi-mechanism
model shows reasonable agreement with the experimental data.
Due to their size and the only slight increase in viscosity of the
resin, nanosilica particles can be used for injection or infusion
methods for composite manufacturing where they can penetrate
even close meshed fabrics easily. It is well known that properties
like modulus, fracture toughness, compressive strength and espe-
cially fatigue performance upon cyclic loading can be improved
by modifying the epoxy resin matrix with silica nanoparticles.
Some properties, e.g. high speed impact performance, were not
improved. An extensive review regarding the potential of silica
nanoparticles to toughen fiber-reinforced composites was pub-
lished recently [21].
The combination of reactive liquid rubbers or core shell elasto-
mers and the use of silica nanoparticles as additional modifier in
epoxy resin systems yields synergistic property improvements in
cured bulk epoxy resins [22]. It is possible to formulate tough
and stiff materials. These resin systems and their synergies in var-
ious applications, including fiber-reinforced composites were pat-
ented consequently at a very early stage [23].
Today such hybrid systems are used in many industrial epoxy
formulations. Fig. 1 shows the transmission electron microscopy
(TEM) picture of a cured bulk hybrid epoxy resin. Obviously both
the rubber particles which have formed upon cure from the reac-
tive liquid rubber and the silica nanoparticles are well dispersed.
Pearson et al. investigated not only the toughening mechanisms
but also tried to develop a suitable model to describe them [25].
Based on their results they supposed that the toughening
Fig. 1. Transmission electron microscopy (TEM) image of amine-cured hybrid
epoxy resin containing reactive liquid rubber (CTBN) and silica nanoparticles [24].
88 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
mechanisms for rubber particles and silica nanoparticles remained
the same when both epoxy modifications were used together. Nev-
ertheless they raised the question why the combination of reactive
liquid rubber (CTBN) and silica nanoparticles increased the tough-
ness more than the addition of both mechanisms would explain.
Zhang et al., using a carboxyl-terminated polyurethane-co-poly-
ether block copolymer with silica nanoparticles as hybrid resin sys-
tem confirmed such an additive behaviour of toughening
mechanisms [26]. They described a model for modulus calculations
but did not provide a model for the toughening mechanisms of
hybrid epoxy systems.
The toughening mechanisms of such hybrid systems are well
understood [22] and described recently in a review article regard-
ing their use in fiber-reinforced composites [27]. A realistic model
for their toughening behaviour which shows reasonable agreement
with all the experimental results available has not been developed
so far, however. When looking into fiber-reinforced composites,
significant property improvements reported for hybrid neat resins
are transferred only partially into improvements in the corre-
sponding composites [27].
In this experimental study epoxy resin formulations similar to
industrial resin systems and modifications by reactive liquid rub-
ber and silica nanoparticles were studied on carbon fiber-rein-
forced composite made by resin transfer moulding (RTM). Beside
neat resin data we intended to obtain an extended property profile
based on interlaminar fracture toughness (GIc and GIIc) data, com-
pression after impact (CAI) data and interlaminar shear strength
(ILSS) data as well. Neat resin and corresponding laminates were
investigated to achieve relationships between neat resin and com-
posite properties and to probe the conversion factor as proposed
recently by Sprenger [27].
The results published so far have been obtained with resin sys-
tems cured for very long times. Amine cured systems with cure
cycles of 4–6 h; sometimes followed by a postcure. Anhydride
cured systems cured for 3–12 h at temperatures above 100 °C. Thus
a hardener system with a relatively fast cure of 15 min, more real-
istic for industrial applications, was selected.
2. Experimental
2.1. Materials
A diglyciylether of bisphenol A (DGEBA) (AralditeÒ GY 250,
Huntsman, Switzerland) was used. To lower the viscosity the tri-
glycidylether of trimethylolpropane (TGETMP) (GrilonitÒ V51-3,
EMS, Switzerland) was added as reactive diluent. A reactive liquid
rubber was added as a 40 wt% concentrated adduct with DGEBA
(AlbipoxÒ 1000, Evonik Hanse GmbH, Germany). The silica nano-
particles were introduced by using a 40% concentrate in DGEBA
(NanopoxÒ F 400, Evonik Hanse GmbH, Germany).
As hardener a blend of 90 wt% isophorone diamine (Vestamin
IPD) and 10 wt% trimethyl hexamethylene diamine (Vestamin
TMD) was used (Evonik Industries, Germany). The active hydrogen
of the hardener blend was calculated to be 42.3 T
The control was a mixture of 90 parts DGEBA with 10 parts
TGETMP. The ratio epoxy resin/reactive diluent was kept constant
for all other resin systems. The epoxy equivalent weight of the dif-
ferent epoxy resin systems was calculated and they were cured
with the stoichiometric amount of hardener.
Nomenclature: system 10R contains 10 wt% of reactive liquid
rubber, and system 10R10 N contains 10 wt% reactive liquid rubber
as well as 10 wt% silica nanoparticles.
Carbon fiber Tenax-E HTS40 F13 12 K was arranged into a 45°/
45° nonwoven with 254 g/m2. This material was kindly provided
by Saertex GmbH, Saerbeck, Germany.
2.2. Bulk sample preparation
Prior to mixing, the DGEBA as well as the nanosilica and CTBN
concentrate were preheated to 60 °C in a convection oven. First
the reactive diluent was added to the DGEBA (DGEBA: TMPTGE
90:10) and homogenized. A conventional dissolver stirrer (diame-
ter 40 mm) with a rotational speed of 500 rpm was used for all
mixing procedures. Subsequently the required amounts of the
reactive liquid rubber concentrate and the nanosilica concentrate
were added. After mixing the nonreactive resin mixture was condi-
tioned at 80 °C and degassed at 10 mbar. Finally the resin mixture
was carefully mixed with the stoichiometric amount of the pre-
mixed IPD/TMD curing agent. The final mixture was again
degassed at 10 mbars. The reactive liquid was now poured in pre-
heated (70 °C) release agent coated (Loctite Frekote 770-NC, Hen-
kel, Germany) aluminum mold. The resin systems were cured for
15 min at 120 °C followed by 2 h postcuring at 160 °C. Nanocom-
posite plates with a defined thickness of 4 mm respectively
2 mm were obtained.
2.3. Laminate preparation
Carbon fiber-reinforced laminates were produced using a resin
transfer molding (RTM) machine (Isojet 200915, Isojet, France)
designed for 1 and 2-part vacuum-assisted RTM. Stainless steel
frames were put in the aluminum mold, cavity size of approx.
40 cm  40 cm, to obtain laminates with a defined thickness.
Before placing the dry carbon fibers, the aluminum mold and the
steel frame were preconditioned with a release agent (Loctite Frek-
ote 770-NC, Henkel, Germany) and thoroughly dried.
For the determination of the interlaminar fracture toughness
Mode I and II (GIc, GIIC), 3 mm thick laminates with [(90/0)]5s layup
and 48 vol% fibers were prepared. A 3.5 lm Teflon foil was inserted
in the mirror plane to create the preinduced crack.
Laminates for compression after impact (CAI) testing were pro-
duced with a [(90/0)/( 45/+45)]4s layup and 54 vol% fibers at a
laminate thickness of 4.2 mm. The closed aluminum mold was
put into a hot press (LZT 110 L, Maschinenfabrik Langzauner
GmbH, Austria) and evacuated (10 mbar). An injection tempera-
ture of 70 °C was chosen for the resin systems in order to have a
reasonably fast filling of the mould. The injection pressure was
set to 8 bars. After 15 min curing at 120 °C the laminates were
demoulded and postcured in an oven for 2 h at 160 °C.
2.4. Bulk sample testing
All test specimens were prepared using a circular saw (Diadisc
6200, MUTRONIC Präzisionsgerätebau GmbH & Co. KG, Germany)
equipped with a diamond saw blade.
Glass transition temperature of the cured bulk resin was deter-
mined by dynamic mechanical thermal analysis (DMTA) according
to DIN EN ISO 6721-7 using an Advanced Rheometric Expansion
System (ARES I, Rheometric Scientific, Germany) in torsional mode.
A sinusoidal deformation of 0.1% with a frequency of 1 Hz was
applied from 25 °C to 200 °C with a heating rate of 3 K/min. The
glass transition temperature (Tg) was determined as maximum of
the loss factor, tan d.
The critical stress intensity factor (KIc) and the critical energy
release rate (GIc) were determined according to ISO 13586 using
compact tension (CT) test specimens. The test specimen width
was w = 33 mm, the thickness d = 4 mm. For each sample, a sharp
crack was generated in the V-notch by tapping a new razor blade.
The tests were performed using a universal tensile testing machine
(Zwick BZ 2.5/TN1S, Zwick Roell, Germany) with a test speed of
10 mm/min. The crack opening displacement was measured using
 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
a clip extensometer (632.29F-30, MTS, Germany). The modulus of
the bulk resin samples was calculated according to the method
of Saxena and Hudak [28].
Transmission electron microscopy (TEM) characterizations were
carried out using a Zeiss LEO 902 A EFTEM (Carl Zeiss AG, Germany)
applying an acceleration voltage of 200 kV. Thin sections of 50 nm
were cut on a Leica Ultracut microtome (Leica Biosystems GmbH,
Germany) equipped with a glass knife. The Cryo-TEM was carried
out by freezing the resin/hardener blend after mixing in liquid nitro-
gen followed by specimen cutting over a bath of liquid nitrogen.
2.5. Laminate testing
All test specimens were prepared from the particular laminate
using a circular saw (Diadisc 6200, MUTRONIC Präzisionsgerätebau
GmbH & Co. KG, Germany) equipped with a diamond saw blade.
The glass transition temperature (Tg) of the laminates was mea-
sured as described in Section 2.4.
Interlaminar fracture toughness Mode I (GIc) was determined in
accordance to AITM1-0005, whereas aluminum blocks were used
instead of hinges – as described in ASTM D 5528-94a, applying a
test speed of 10 mm/min. Interlaminar fracture toughness Mode
II (GIIc) was determined using the tested GIc-specimen as described
in DIN EN 6034 applying a test speed of 1 mm/min. Both tests were
performed on a universal tensile testing machine (Zwick BZ 2.5/
TN1S, Zwick Roell, Germany).
The interlaminar shear strength (ILSS) was measured according
to DIN EN 2563 using test specimen with the dimension
25  10  3 mm prepared from the laminates made for the GIc test-
ing. The tests were performed on a universal tensile testing machine
(Zwick 1475, Zwick Roell, Germany) with a test speed of 1 mm/min.
Compression after impact (CAI) data was gathered in accor-
dance to DIN 65561 using 3 samples with dimensions of
150  100 mm. The specimens were impacted with an energy of
30 J using a drop-weight impactor. Subsequently the delaminated
area was measured using an ultrasonic inspection method (AirTech
HFUS 2400, Ingenieurbüro Dr. Hilger Ultraschall-Prüftechnik, Ger-
many). The compression after impact strength was measured on a
universal tensile testing machine (Zwick 1485, Zwick Roell, Ger-
many) with a test speed of 0.5 mm/min.
Scanning electron microscopy (SEM) analysis was carried out
using a Zeiss 1530 (Carl Zeiss AG, Germany) possessing a field
emission cathode for high-resolution micrographs (acceleration
voltage 1.5 kV).
3. Results and discussion
All other studies cited used the same commercially available
epoxy resins with 40 wt% silica nanoparticles with a diameter of
20 nm. The reactive liquid rubber used was always introduced into
the formulations as epoxy resin adduct; typically 40 wt% CTBN
rubber with an acrylonitrile content of 18 wt% had been pre-
reacted with 60 wt% DGEBA.
3.1. Bulk resin properties
As expected, the modification with 10 wt% reactive liquid rub-
ber (system 10R) lowered both modulus and Tg by approx. 10%
compared to the unmodified control; as can be seen in Table 1.
When epoxy resins containing reactive liquid rubbers are cured,
the rubber phase separates and forms a secondary phase, the rub-
ber domains. It is well known that some of the long rubber mole-
cules do not participate in the phase separation upon cure but
crosslink randomly into the matrix. This lowers the network den-
sity, which affects of course modulus and Tg.
89
The addition of silica nanoparticles as a filler reinforces the
matrix and thus increases the modulus again. The epoxy resin sys-
tem with 10 wt% rubber and 10 wt% nanosilica has the same mod-
ulus as the unmodified control. An increase in modulus of 0.35 GPa
for 10 wt% silica nanoparticle addition is similar to what was found
for other amine-cured DGEBA systems [12].
The glass transition temperature, however, is not increased by
the addition of nanosilica – as the silica nanoparticles have no
influence on the network density. Instead, the Tg is lowered slightly
at addition levels above 7.5 wt% nanosilica. Most probably higher
addition levels of nanosilica hinder the formation of rubber
domains upon cure. Thus the crosslink density is further reduced
by the higher number of randomly crosslinked rubber molecules
[22].
Xu et al. studied piperidine-cured DGEBA [29]. They reported a
huge increase in both KIc and GIc for the addition of CTBN which
reaches a plateau at 10 wt% addition level. Additional modification
of a 5 wt% rubber containing system with silica nanoparticles did
only slightly increase KIc. GIc was not further improved. Though
they used an 18 h cure cycle, our system cured for only 15 min per-
forms quite similar.
As can be seen in Fig. 2, the increase in toughness by the rubber
addition is significant; the further addition of silica nanoparticles
does not contribute to toughening markedly. At 10 wt% nanosilica
however, the toughness is further increased – beyond the increase
deriving from the rubber modification. An improvement in GIc of
approx. 250% is not untypical. In a recently published review
improvements in the range of 111–302% were reported for
amine-cured, CTBN-modified DGEBA resins [22].
Fig. 3 shows that the hybrid system with 10R10N is a tough and
stiff matrix resin, clearly outperforming the unmodified control.
The Tg with 140 °C is still acceptable for structural automotive
applications. In Table 1 an overview of all bulk resin properties is
given.
3.2. Laminate properties
It is well known that very often property improvements of
matrix resins do not translate into improvements of fiber-rein-
forced composites made thereof. Especially carbon fibers can be a
very dominating reinforcement. The recently published review
from Ye et al. gives an excellent overview regarding the improve-
ments of interlaminar fracture toughness and compression after
impact strength by using nanoparticles in epoxy resin matrices
for fiber-reinforced composites [30]. They distinguished two differ-
ent relationships between GIc of the matrix and GIc of the laminate
– depending if the matrix is rather brittle or tough. Furthermore
they concluded that a positive correlation between the GIc of the
matrix and the GIIc of the composite or the CAI strength is not clear.
Therefore they voiced the opinion that the research ‘‘. . . is at an
immature stage. . .’’. Adding an additional component, elastomeric
tougheners like reactive liquid rubbers or core–shell elastomers
makes the understanding of such systems even more complex.
In a review regarding composites made from hybrid epoxy res-
ins, Sprenger recently suggested the existence of a conversion fac-
tor for hybrid epoxy resins [27]. He found for both glass and carbon
fiber reinforced materials that the percentual improvement of the
GIc of the cured bulk epoxy resin converted into 0.18 times the per-
centual improvement of the GIc of a laminate made from this
matrix resin.
The first laminate properties investigated in this study were GIc
and GIIc. The results are given in Fig. 4.
As can be seen in Fig. 4, despite the higher modulus of the
hybrid matrix systems, the laminate toughness was improved by
approx. 50% for the hybrids and even more by the rubber-only
modified system.
90 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95

Table 1
Overview of cured bulk epoxy resin properties and carbon fiber-reinforced laminate properties for the different epoxy resin systems.

Control System 10R System 10R5N System 10R7.5N System 10R10N

Tg (°c) 154 ± 0.1 145 ± 1.4 145 ± 0.7 145 ± 0.1 140 ± 0.6
Bulk modulus (GPa) 2.63 ± 0.16 2.29 ± 0.02 2.51 ± 0.12 2.48 ± 0.11 2.64 ± 0.05
Bulk KIc (MPa m1/2) 0.61 ± 0.02 1.03 ± 0.02 1.02 ± 0.04 1.04 ± 0.03 1.16 ± 0.03
Bulk GIc (J/m2) 126 ± 6 406 ± 17 364 ± 41 383 ± 14 447 ± 22
Laminate ILSS (MPa) 56.4 ± 2.5 51.0 ± 2.0 51.6 ± 1.3 50.0 ± 1.6 48.8 ± 2.6
Laminate GIc (J/m2) 433 ± 7 688 ± 22 613 ± 19 555 ± 57 621 ± 28
Laminate GIIc (J/m2) 1299 ± 132 1124 ± 44 1034 ± 45 1031 ± 72 1080 ± 139
Delam. area* (mm2) 2689 ± 456 2265 ± 337 2709 ± 159 3108 ± 93 3427 ± 336
CAI* (MPa) 161.3 ± 2.9 180.0 ± 2.8 164.0 ± 2.0 167.3 ± 1.1 147.5 ± 0.7
*
After 30 J impact.

Fig. 2. KIc and GIc of the cured bulk epoxy resin systems.

However, other findings have been published as well: Sprenger

Fig. 3. Modulus versus GIc for cured bulk epoxy resin systems.
For carbon fiber-reinforced materials much less literature data
are available than for glass fiber-reinforced composites made from
hybrid epoxy resins containing both the reactive liquid rubber and
the silica nanoparticles [27]. Nevertheless in most studies with
CFRC published until now, the fiber-reinforced composites based
on a hybrid system were found to have the highest GIc. Kinloch
et al. reported about a CFRC made from anhydride cured DGEBA
[31,32]. The control with 439 J/m2 was increased to 1050 J/m2 by
a modification with 9 wt% reactive liquid rubber. Further addition
of 10.5 wt% silica nanoparticles increased the GIc of the laminate
further to 1320 J/m2. Ye et al. investigated piperidine-cured DGEBA
[33]. The GIc could be increased from 741 J/m2 to 1256 J/m2 by a
modification with 10 wt% rubber. The addition of 10 wt% nanosili-
ca further increased the GIc to 1323 J/m2.
et al. reported about a CFRC based on tetrafunctional epoxy resin
cured with aromatic amines; a system similar to the aerospace
industry benchmark HexflowÒ RTM6 [34]. Though the cured
hybrid bulk system had the highest GIc, the hybrid laminate was
lower in GIc than the system containing only reactive liquid rubber.
Nevertheless the modulus of the hybrid was higher than for the
rubber-only modification.
Comparing the relative improvement of the bulk GIc with the
relative improvement of the laminate GIc, the average of laminate
improvement versus bulk improvement was 0.2 ± 0.07 – which is
very close to the conversion factor of 0.18 ± 0.02 recently discov-
ered by Sprenger [27].
Not much GIIc data has been published for hybrid systems so far.
The data shown in Fig. 4 indicates that once a crack has formed, the
delamination under mode II loading continues at a slightly lower
force. How this is related to the hybrid modification is unclear. Fur-
ther clarification of the fracture mechanisms is expected from the
scanning electron micrographs of the fracture surfaces of GIc and
GIIc test specimen.
The ILSS was tested and is slightly lower for all modified sys-
tems, but still on a very high overall level. Having a look at the
error bars, there cannot be seen real differences between the rub-
ber-only modification and the hybrids with different nanosilica
concentrations. The very small differences in ILSS for all samples
tested indicate an excellent laminate quality (compare Table 1).
Quaresimin et al. studied CFRC made from amine-cured DGEBA
[35]. The modification with 7.3 wt% reactive liquid rubber and
3.7 wt% silica nanoparticles increased the bulk resin KIc from
0.91 MPa m1/2 to 2.11 MPa m1/2 or by 130%. The mode I interlami-
nar fracture toughness was increased by 74%. However the inter-
laminar shear strength was reduced from 55 MPa to 50 MPa –
very similar to our results.
 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
Fig. 4. GIc and GIIc of carbon fiber-reinforced laminates based on various epoxy resins systems.
Falling dart test and the CAI strength were investigated next.
The results are given in Fig. 5. As can be seen, for higher nanosilica
contents the hybrid systems show more delamination and the
compressive strength is reduced. This is unexpected, as for a lam-
inate made from a tougher epoxy resin matrix one would expect an
increase in CAI.
Kuehn et al. investigated carbon fiber reinforced laminates
made from anhydride cured DGEBA [36]. A modification with
10 wt% reactive liquid rubber reduced the damage area (at 30 J)
from 38 to 24 cm2 – the compressive strength after impact was
increased from 165 MPa to 205 MPa. A modification with 25 wt%
nanosilica did not reduce the delaminated area at all but increased
the CAI to 175 MPa. Obviously the higher modulus of the systems
modified with nanosilica is improving the residual compressive
strength. Unfortunately they did not investigate hybrid systems.
Tate et al. reported results of a glass fiber-reinforced laminate
based on DGEBA and cured with a commercial amine blend [37].
They compared the unmodified control with a hybrid containing
6.5 wt% reactive liquid rubber and 8.1 wt% of nanosilica. For all dif-
ferent impact energies tested the delamination was always larger
for the laminate made from the hybrid epoxy resin system and
the post impact compressive strength was lower. They concluded
that the hybrid is capable to dissipate the impact energy over a lar-
ger area, which would explain the larger delaminated area and the
lower CAI values. They noticed that small agglomerates had
formed which seemed not to affect the performance of the nanosil-
ica modification.
Fig. 5. Delaminated area and CAI strength after impact of carbon fiber-reinforced laminates based on various epoxy resins systems (after 30 J impact).
91
The comparison of the delaminated area between the rubber-
modified system 10R and the hybrid 10R5N which contains addi-
tionally 5% of nanosilica is very interesting: they exhibit the same
delaminated area, but the hybrid shows a lower compression after
impact strength. GIc and GIIc of this hybrid are lower as well. This
indicates a more brittle matrix behaviour, which is confirmed by
the bulk resin results for 10R and 10R5N.
Nevertheless, the damage tolerance, given by the compression
after impact strength and the delaminated area after impact, is
higher for 10R5N and very similar for 10R7.5N to the unmodified
control. Table 1 gives an overview of all laminate properties for a
quick glance.
In a future paper the fatigue properties will be studied, as they
have been reported best for hybrid systems [27].
3.3. Microscopical investigations
3.3.1. Laminate fracture surfaces
In order to understand the mechanisms of failure the fracture
surfaces were investigated. Fig. 6 shows the fracture surfaces from
the GIc test specimen and Fig. 7 gives the fracture surfaces from GIIc
test specimen.
Fig. 6 reveals some information regarding the fracture mecha-
nism. The fracture surface of the unmodified control shows less
residual epoxy resin than the rubber-modified epoxy and the
hybrids, which indicates a more interfacial failure between fiber
layer and matrix. A similar effect was found by Ye et al. [33].
92 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95

Fig. 6. SEM images of fracture surfaces from GIc test specimens. The white arrows indicate the direction of crack propagation.

The difference becomes more evident on the fracture surfaces
from GIIc test specimens at a similar magnification in Fig. 7. The
matrix resin between the carbon fibers of the control shows the
typical angular structures of brittle materials; designated ‘‘hack-
les’’. In the picture of the matrix containing 10 wt% reactive liquid
rubber the rubber domains, which have formed upon cure, are
clearly visible. The matrix shows no angular structures any more
– it looks rather peeled or teared off. This indicates a more ductile
behavior. The hybrid matrix with 10 wt% rubber and 5 wt% nano-
silica looks similar; however at 10 wt% nanosilica the matrix does
not look ductile anymore. Apparently the nanosilica addition
reduces the ductility gained by the rubber modification. This
would explain the increase in delaminated area and subsequent
decrease in compressive strength after impact with increasing sil-
ica nanoparticle content.
A major aspect comes into focus when looking closer. In Fig. 8
the matrix resin of the hybrid 10R5N with 10 wt% rubber and
5 wt% nanosilica can be seen between the surfaces of the carbon
fibers. The rubber domains with their typical size of 0.5–0.7 lm
are clearly visible, and, unexpected, agglomerates of silica
nanoparticles are found. This triggered an additional microscopical
investigation of the cured bulk resins in order to understand these
systems.
3.3.2. Bulk resin cross sections
Despite the rather fast cure the rubber domains have formed per-
fectly and are evenly distributed, as becomes evident in Fig. 9 for the
system 10R. All hybrids show agglomerates of silica nanoparticles.
They increase in size with increasing addition level in nanosilica.
At 5 wt% the agglomerates seem to be approximately 100 nm large;
at 7.5 wt% 200–300 nm and at 10 wt% they seem to exceed 300 nm
in size. Eventually the agglomerates have formed due to an incom-
patibility of the surface-modified silica nanoparticles with the hard-
ener system. Another possibility would be the very fast cure. This
will be discussed in 3.4.
3.4. Formation of nanosilica agglomerates
In most studies regarding hybrid epoxy systems, no agglomer-
ates were detected [22]. Very few researchers reported the presence

Fig. 7. SEM images of fracture surfaces from GIIc test specimens. The white arrows indicate the direction of crack propagation.
 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
Fig. 8. SEM image of hybrid 10R5N (fracture surface of GIIc test specimen). The
white arrow indicates the direction of crack propagation.
of agglomerates in their cured resin systems. Manjunatha et al.
investigated the fatigue performance of glass fiber reinforced lami-
nates based on anhydride cured DGEBA [38]. They observed
agglomerates of 400–800 nm in the hybrid with 9 wt% reactive
liquid rubber and 10 wt% nanosilica. The significant improvement
in fatigue performance seemed not to be affected by their presence.
Hsieh et al. prepared laminates with carbon fibers and glass
fibers using different hybrid systems [39]. Scanning electron
microscopy of the anhydride-cured bulk DGEBA revealed the for-
mation of agglomerates as well. The nanoparticles appeared to
cluster in necklace-like structures and increased in size with
increasing nanosilica content, however the GIc of the bulk resin
and the laminates was increased significantly – and was always
higher for the hybrid than for the rubber-only modified epoxy
resin. For the 10R10N system they reported agglomerate sizes of
up to 2000 nm in length and 100 nm in width. Therefore it was
concluded that the formation of agglomerates does not affect
performance.
We started to examine the different components of our epoxy
resin system and found no incompatibility between silica nanopar-
ticles and the curing system consisting of IPD and TMD. If our sys-
tem is cured without the presence of reactive liquid rubber but
with the same hardener/accelerator and the same fast cure, no
Fig. 9. TEM images of cross sections of cured bulk epoxy resins systems.
93
agglomeration occurs – see Fig. 10. The SiO2-nanoparticles are well
dispersed in such a cured epoxy resin.
Next we investigated the blending process with reactive liquid
rubber adducts. When epoxy resins containing reactive liquid rub-
ber adducts and epoxy resins containing silica nanoparticles are
blended, these blends are transparent and no agglomeration takes
place, even if such a resin blend is stored over weeks and months.
Now being mixed with the hardener and accelerator at room tem-
perature, the resin system remains transparent – still no agglomer-
ation occurs. We froze the system 10R5N immediately after
blending with the hardener using liquid nitrogen and examined
using TEM under cryogenic conditions (Fig. 11).
Despite the presence of hardener and accelerator and their
intensive mixing with the resin part no agglomeration has
occurred – the silica nanoparticles are well dispersed. As the sys-
tem is uncured, the rubber phase separation has not occurred
and no domains have formed.
Eventually the relatively fast cure cycle of 15 min, inducing a
rather quick rubber phase separation of the CTBN, is the cause
and forces the nanoparticles to agglomerate. This would explain
why other researchers using unaccelerated isophoronediamine
curing did not report the formation of agglomerates [22].
In order to confirm this assumption the hybrid system 10R5N
was cured at room temperature for 48 h. The microscopical inves-
tigation was evident, as can be seen in Fig. 12.
Though the room temperature cure was almost completed after
2 h we prepared the microscopy test specimen after 48 h of RT
cure. As can be seen, even during this somewhat slower cure
agglomerates were formed.
Curing 10R5N without the accelerator TMD did not impede
nanosilica agglomeration either.
It can be concluded that the agglomeration of the silica nanopar-
ticles is linked to the relatively fast rubber phase separation of the
reactive liquid rubber due to the fast increase in crosslink density
defined by the fast curing system of IPD/TMD. Eventually the rubber
still present in the epoxy network and randomly crosslinked into
the matrix causes the nanosilica to agglomerate upon cure.
If the agglomerates affect the composite performance when
increasing in size could not be clarified finally. It seems not to be
the case for properties like GIc (bulk and laminate), GIIc (laminate)
or ILSS. However the compressive strength after impact (laminate)
is the lowest for the system with the largest agglomerates (10R10N).
94
Fig. 10. TEM image of cross section of cured epoxy resin with 5 wt% nanosilica (0R5N) at different magnifications.
Fig. 11. Cryo–TEM image of uncured epoxy resin system with 10 wt.% CTBN and
5 wt% nanosilica (10R5N).
Fig. 12. TEM image of cross section of RT-cured epoxy resin with 10 wt% CTBN and
5 wt% nanosilica (10R5N) cured 48 h at RT.
S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
In the literature Pearson et al. reported about bulk hybrid sys-
tems containing large agglomerates as well [25]. They claimed that
the formation of agglomerates occurs for higher concentrations of
reactive liquid rubber; probably above 12 vol%. The improvements
in fracture toughness by nanosilica addition became smaller for
systems with agglomerates present. This confirms the predictions
of Williams modelling work [16] and would explain our findings
for bulk GIc.
The question if the performance of our laminates would even be
better with no agglomerates present could not be answered as the
rubber-phase separation of our system, probably due to the rela-
tively fast curing system, leads to the formulation of agglomerates.
Eventually another hardener system might behave different. How-
ever, this will be the subject of future work. We tried to develop a
suitable model to predict at least the bulk resin properties. How-
ever, our system is quite complex. Besides that we have to deal with
the fact that very probably an unknown amount of rubber molecules
did not participate in phase separation but crosslink randomly into
the molecular network, nanosilica agglomerates have never been
modelled. As additional data will be necessary to approach this sys-
tem in a proper way, this will be the subject of a future study.
Another approach for future investigations will be to use core
shell rubber particles instead of reactive liquid rubber; therefore
no phase separation during cure will occur. Thus, theoretically no
agglomerates should be formed during cure. This should provide
systems easier to model and predict.
4. Conclusions and outlook
Using a fast curing epoxy resin system based on DGEBA/IPD/
TMD carbon fiber-reinforced laminates suitable for automotive
applications were made in a reasonable short manufacturing cycle.
The modification of the matrix with both reactive liquid rubber and
silica nanoparticles resulted in tough laminates. A couple of con-
clusions can be drawn from the investigations:
 The bulk resin modified with the reactive liquid rubber
shows the expected increase in toughness alongside the
loss in modulus and Tg.
 The hybrid bulk resins modified additionally with silica
nanoparticles show a reduced loss of the modulus.
 The hybrid bulk resin with the highest addition level of
nanosilica performs best; exhibiting the highest toughness
and modulus.
 The laminate made from the rubber-only resin shows an
increase in GIc, a slight reduction in GIIc and ILSS as well
as a reduction of the delaminated area in impact testing
alongside with an increase in CAI.
 S. Sprenger et al. / Composites Science and Technology 105 (2014) 86–95
 The hybrid laminates showed no further increase in tough-
ness compared to the rubber-toughened system, but rather
a slight reduction of both GIc and GIIc. ILSS was on a similar
level; CAI was reduced further.
 The reason for this unexpected lack of further increase in
toughness could be the formation of agglomerates of nano-
silica during the cure.
 The agglomeration found is caused by the presence of the
reactive liquid rubber. Very probably the fast phase separa-
tion of the rubber, pushed by the fast cure, is the driving
force of the agglomeration.
 The conversion factor for the relative improvement in GIc
versus the relative improvement of GIc of the laminate
was found to be 0.2 – in excellent agreement with earlier
findings.
The systematic investigation of other fast curing hardeners
regarding agglomerate formation in presence of reactive liquid
rubber will be the subject of future work. Furthermore, of course,
hybrid systems based on core–shell rubbers, which do not phase
separate upon cure will be investigated. Eventually they are the
material of choice to formulate fast curing hybrid epoxy systems.
Acknowledgements
The authors would like to thank Saertex GmbH, Saerbeck, Ger-
many, namely Lars Ischtschuk and Joerg Buenker for providing the
textile material for this study. Furthermore they would like to
thank B.Sc. Georg Tauer, University of Bayreuth, Department of
Polymer Engineering, Bayreuth, Germany for supporting the test
specimen preparation.
References
[1] Regulation (EC) No 443/2009 of the European parliament and of the council of
23 April 2009 setting emission performance standards for new passenger cars
as part of the Communitýs integrated approach to reduce CO2 emissions from
light-duty vehicles, Official Journal of the European Union, 05.06.2009, L 140/1.
[2] Estin & Co., Main dynamics of the composite industry for automotive
applications 2010–2015. Paris, France: Publ. by JEC Composites; 2011.
[3] Hillermeier R, Hasson T, Friedrich L, Ball C. Advanced thermosetting resin
matrix technology for next generation high volume manufacture of
automotive composite structures. SAE Technical Paper 2013-01-1176; 2013.
http://dx.doi.org/10.4271/2013-01-1176.
[4] Ellis B, editor. Chemistry and technology of epoxy resins. Springer
Science+Business Media Dordrecht; 1993. p. 1–142.
[5] Hare C. Amine curing agents for epoxies. J Prot Coat Lin 1994;9:77–103.
[6] Hare C. Epoxy curing agents II. J Prot Coat Lin 1994;10:197–213.
[7] Kinloch AJ, Shaw SJ, Tod DA. Deformation and fracture behavior of a rubber-
toughened epoxy: 1. Microstructure and fracture studies. Polymer
1983;24:1341–54.
[8] Bagheri R, Marouf BT, Pearson RJ. Rubber-toughened epoxies: a critical review.
Macromol Sci Part C: Polym Rev 2009;49:201–25.
[9] Sprenger S, Eger C, Kinloch AJ, Taylor AC. Nanotoughening of epoxies. In:
Proceedings of stick! Conference April 9th 2003. Nürnberg, Germany: Vincentz
Verlag; 2003.
[10] Zhang H, Zhang Z, Friedrich K, Eger C. Property improvements of in situ epoxy
nanocomposites with reduced interparticle distance at high nanosilica
content. Acta Mater 2006;54:1833–42.
[11] Huang Y, Kinloch AJ. Modelling of the toughening mechanisms in rubber-
modified epoxy polymers. J Mater Sci 1992;27:2763–9.
[12] Sprenger S. Epoxy resin composites with surface-modified silicon dioxide
nanoparticles: a review. J Appl Polym Sci 2013;130:1421–8.
95
[13] Zhang H, Tang LC, Zhang Z, Friedrich K, Sprenger S. Fracture behaviours of in
situ nanoparticle-filled epoxy at different temperatures. Polymer
2008;49:3816–25.
[14] Johnsen BB, Kinloch AJ, Mohammed RD, Taylor AC, Sprenger S. Toughening
mechanisms of nanoparticle-modified epoxy polymers. Polymer
2007;48:530–41.
[15] Lauke B. On the effect of particle size on fracture toughness of polymer
composites. Comp Sci Technol 2008;68:3365–72.
[16] Williams JG. Particle toughening of polymers by plastic void growth. Comp Sci
Technol 2010;70:885–91.
[17] Zappalorto M, Salviato M, Quaresimin M. Assessment of debonding-induced
toughening in nanocomposites. Proc Eng 2011;10:2073–978.
[18] 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;51:6284–94.
[19] Zappalorto M, Salviato M, Quaresimin M. A multiscale model to describe
nanocomposite fracture toughness enhancement by the plastic yielding of
nanovoids. Comp Sci Technol 2012;72:1683–91.
[20] Quaresimin M, Salviato M, Zappalorto M. A multi-scale and multi-mechanism
approach for the fracture toughness assessment of polymer nanocomposites.
Comp Sci Technol 2014;91:16–21.
[21] Sprenger S. Improving mechanical properties of fiber-reinforced composites
based on epoxy resins containing industrial surface-modified silica
nanoparticles: review and outlook. J Comp Mater 2014. http://dx.doi.org/
10.1177/0021998313514260.
[22] Sprenger S. Epoxy resins modified with elastomers and surface-modified silica
nanoparticles. Polymer 2013;54:4790–7.
[23] EP 1457509, WO 2004081076 2003.
[24] Picture courtesy of Army Research Laboratory.
[25] Liang YL, Pearson RA. The toughening mechanism in hybrid epoxy-silica-
rubber nanocomposites (HESRNs). Polymer 2010;51:4880–90.
[26] Tang LC, Zhang H, Sprenger S, Ye L, Zhang Z. Fracture mechanisms of epoxy-
based ternary composites filled with rigid-soft particles. Comp Sci Technol
2012;72:558–65.
[27] Sprenger S. Fiber-reinforced composites based on epoxy resins modified with
elastomers and surface-modified silica nanoparticles. J Mater Sci
2014;49:2391–402.
[28] Saxena A, Hudak SJ. Review and extension of compliance information for
common crack growth specimen. Int J Fract 1978;14:453–68.
[29] Xu SA, Wang GT, Mai YW. Effect of hybridization of liquid rubber and
nanosilica particles on the morphology, mechanical properties and fracture
toughness of epoxy composites. J Mater Sci 2013;48:3546–56.
[30] Tang Y, Ye L, Zhang Z, Friedrich K. Interlaminar fracture toughness and CAI
strength of fibre-reinforced composites with nanoparticles – a review. Comp
Sci Technol 2013;86:26–37.
[31] Sprenger S, Kinloch AJ, Taylor AC, Mohammed RD, Eger C. Rubber-toughened
FRCs optimised by nano-particles. JEC Compos Mag 2005;19:73–6.
[32] Kinloch AJ, Mohammed RD, Taylor AC, Sprenger S, Egan D. The interlaminar
toughness of carbon-fibre reinforced plastic composites using ‘‘hybrid-
toughened’’ matrices. J Mater Sci 2006;41:5043–6.
[33] Zhang J, Deng S, Ye L, Zhang Z. Interlaminar fracture toughness and fatigue
delamination growth of CF/EP composites with matrices modified by
nanosilica and CTBN rubber. In: Proceedings of the 13th international
conference on fracture, Beijing, China; June 16–21, 2013.
[34] Sprenger S, Kinloch AJ, Taylor AC, Mohammed RD. Rubber-toughened CFRCs
optimized by nanoparticles – Part III. JEC Compos Mag 2007;30:54–7.
[35] Caccavale V, Wichmann MHG, Quaresimin M, Schulte K. Nanoparticle/rubber
modified epoxy matrix systems: mechanical performance in CFRPs. In:
Proceedings of AIAS XXXVI Convegno Nazionale 2007, Ischia, Naples, Italy;
4–8 September 2007.
[36] Kuehn A, Mahrholz T, Mosch. Matrix optimization of CFRP parts concerning
fire and impact properties with process acceleration. J CEAS Aeronaut J
2013;4:191–201.
[37] Tate J, Trevino E, Gaikward S, Sprenger S, Rosas I, Andrews MJ. Low velocity
impact on epoxy glass composites modified with rubber micro-particles and
silica nano-particles. J Nanosci Nanoeng Appl 2013;3(1). ISSN: 2231–1777.
[38] Manjunatha CM, Taylor AC, Kinloch AJ, Sprenger S. The tensile fatigue behavior
of a glass–fiber reinforced plastic composite using a hybrid-toughened epoxy
matrix. J Compos Mater 2010;44(17):2095–109.
[39] 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. J Mater Sci 2010;45:1193–210.