Sei sulla pagina 1di 8

Composites Science and Technology 70 (2010) 901908

Contents lists available at ScienceDirect

Composites Science and Technology


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

Enhancement of delamination fatigue resistance in carbon nanotube reinforced


glass ber/polymer composites
Christopher S. Grimmer, C.K.H. Dharan *
Department of Mechanical Engineering, University of California, Berkeley, CA 94720-1740, USA

a r t i c l e

i n f o

Article history:
Received 15 September 2009
Received in revised form 27 January 2010
Accepted 2 February 2010
Available online 6 February 2010
Keywords:
A. Carbon nanotubes
C. Laminate
A. Nanocomposites
A. Polymermatrix composites (PMCs)
C. Delamination
B. Fatigue
B. Fracture

a b s t r a c t
We show that the addition of small volume fractions of multi-walled carbon nanotubes (CNTs) to the
matrix of glassber composites reduces cyclic delamination crack propagation rates signicantly. In
addition, both critical and sub-critical inter-laminar fracture toughness values are increased. These
results corroborate recent experimental evidence that the incorporation of CNTs improve fatigue life
by a factor of two to three in in-plane cyclic loading. We show that in both the critical and sub-critical
cases, the degree of delamination suppression is most pronounced at lower levels of applied cyclic strain
energy release rate, DG. High-resolution scanning electron microscopy of the fracture surfaces suggests
that the presence of the CNTs at the delamination crack front slows the propagation of the crack due
to crack bridging, nanotube fracture, and nanotube pull-out. Further examination of the sub-critical fracture surfaces shows that the relative proportion of CNT pull-out to CNT fracture is dependent on the
applied cyclic strain energy, with pull-out dominating as DG is reduced. The conditions for crack propagation via matrix cracking and nanotube pull-out and fracture are studied analytically using fracture
mechanics theory and the results compared with data from the experiments. It is believed that the shift
in the fracture behavior of the CNTs is responsible for the associated increase in the inter-laminar fracture
resistance that is observed at lower levels of DG relative to composites not containing CNTs.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
The formation and propagation of inter-laminar cracks in
composite materials leads to a signicant reduction in laminate
strength and stiffness, ultimately resulting in failure. The conditions
that favor delamination can range from out-of-plane tensile loads to
in-plane compressive loads, with crack initiation occurring at structural discontinuities such as edges, notches or voids. Delamination
can also be initiated by local transverse low-velocity impact, a
serious concern since the damage is sub-surface and difcult to
detect. Such defects can propagate under fatigue loading applied
transverse to the delamination plane resulting in cyclic delamination crack propagation leading to fast fracture and failure. Suppression of delamination is therefore of interest, particularly in primary
structures.
In this work, the effect of introducing CNTs into the matrix of
conventional ber-reinforced composites subjected to delamination
propagation under fatigue loading is studied. These hybrid composites were subjected to cyclic out-of-plane loading and both the critical and sub-critical strain energy release rates were measured and
compared to those measured in conventional composites. High-res* Corresponding author. Tel.: +1 510 642 4933; fax: +1 510 643 5599.
E-mail address: dharan@me.berkeley.edu (C.K.H. Dharan).
0266-3538/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compscitech.2010.02.001

olution scanning electron microscopy was used to establish the


interaction between the propagating inter-laminar crack and the
nano-scale reinforcement that it encounters. A mechanistic,
energy-based model is presented that takes into account observations of high-resolution scanning electron micrographs of the fracture surface. This model provides an explanation for the increase
in delamination fracture toughness observed in the CNThybrid
composites subjected to cyclic loading.
The mode I, or opening mode, delamination test has proven
very useful for quantifying the resistance of a particular matrix/ber combination to inter-laminar fracture and delamination. The
double cantilever beam (DCB) test has been established as the
standard method for determining the mode I energy release rate
[13], with the results generally indicative of the resistance of a
particular composite system to delamination crack propagation.
To date, the eld of nano-reinforced composites has been faced
with a series of tradeoffs and constraints regarding the quality of
nano-reinforcement, usually governed by CNT quantity and alignment, and the burden these factors place on manufacturability.
Typically these constraints have resulted in either easy-to-fabricate but marginally improved composite systems, or in composites
that display noticeably improved properties at the expense of
employing complicated and slow manufacturing processes. Attempts at using nano-scale reinforcements to improve the already

902

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

poor mechanical behavior of bulk polymers [410] have had varying success with the best-case examples managing to increase
properties such as stiffness and strength by a factor of two or so.
However, when these CNT-inltrated polymers are used as the matrix in conventional ber-reinforced composites, the overall
improvements in strength and stiffness are negligible [11]. While
integration of CNTs into the bulk matrix polymer of a laminate
has shown signicant improvement in inter-laminar shear
strength [12], the use of CNTs locally in the inter-laminar regions
of a laminate has shown limited [13,14] to no [15] improvement
in shear strength at the cost of added processing steps.
Micro-bers, or whiskers, such as silicon carbide have been grown
directly on the surface of continuous ber reinforcements with interlaminar mechanical properties improving signicantly, but at the expense of reduced in-plane properties due to weakening of the macroscale ber itself by the whisker growth [16]. More recently, CNTs
have been grown on the surface of continuous carbon ber; improvements in load transfer at the ber/matrix interface [17], inter-laminar shear strength [18,19], and delamination fracture toughness
[20] were obtained. Further efforts have also been directed at transfer-printing forests of aligned CNTs to the surface of prepreg [21].
While very different in their methods and results, these previous
efforts appear to be constrained by the trade-off mentioned between
the quality of the nano-scale reinforcement and manufacturability.
In the end, properties such as composite (monotonic) strength and
stiffness are almost completely dominated by the choice of the
macro-scale ber reinforcement. Furthermore, to obtain high volume fractions or a high degree of alignment in CNTs results in prohibitive manufacturing complications. Our efforts are instead
focused on evaluating the suppression of high-cycle fatigue damage
in conventional composite materials when small quantities of CNTs
are dispersed evenly in the matrix polymer using a simple and scalable manufacturing process.
Closed-form solutions and numerical models of classical delamination have been developed by several investigators [22,23]. Additionally, models of delamination in both random short ber
composites [24,25] and oriented through-thickness reinforced
composites [26] have been proposed. However, there have been
no models that take into consideration cyclic loading effects or
how nano-scale reinforcements affect cyclic delamination crack
growth. In this work, we demonstrate the effects that low volume
fraction randomly dispersed CNTs have on the propagation of inter-laminar fatigue cracks, and present an energy model to describe the behavior observed.
As discussed elsewhere [27], we have developed a simple manufacturing process that is scalable for use in industry, and yet is
able to achieve signicant improvements in the fatigue performance of the hybrid composite. By examining the behavior of
CNThybrid composites subjected to high-cycle fatigue loading,
we have brought to light an area where we believe nano-reinforcement can be of most use. While the macro-scale mechanical properties such as strength and stiffness are almost completely
dominated by the choice of ber and ber volume fraction used,
it is the micro- and nano-scale properties and failure mechanisms
that can be inuenced by nano-scale secondary reinforcements.

The average CNT dimensions, as given by the supplier, were 20 nm


in diameter and 6 lm in length. The current CNT loading level of
1% was selected as being adequate to inuence the mechanical
behavior of the composite [2729]. As observed in [30], higher
loadings resulted in an excessive increase in the viscosity of the
polymer making it difcult to impregnate the bers. The woven
glass ber reinforcement plain weave fabric, 0.28 mm thick designated Type 7500, was obtained from Hexcel (Fullerton, CA USA).
The nished composite samples were bonded on either sides to
3.4 mm thick 6061-T6 aluminum sheets to aid in testing.
2.2. Specimen processing
Both the CNT and non-CNT [0/90] ber-reinforced composites
were manufactured by wet lay-up with catalyzed and degassed resin
impregnated in eight plies of reinforcing fabric. A 50 mm wide,
0.05 mm thick polytetrauoroethylene (PTFE) sheet was embedded
at the mid-plane of the laminate to serve as a pre-crack from which
the delamination would initially propagate. Curing was done in a
heated platen press held at 80 C and 580 kPa pressure for one hour.
Acid digestion of the matrix of a sample of the cured composite gave
a ber volume fraction of 0.56. As noted previously [27], CNT penetration into the fabric reinforcement was good, with no observed
ltering of CNTs. The nal laminate thickness away from the PTFE
pre-crack was measured to be 1.5 mm. Constant-width and tapered-width delamination specimens were cut from the cured
sheets using a diamond blade wet saw. In all specimens, the fabric
warp direction was oriented parallel to the delamination propagation direction.
2.3. Experimental method
Monotonic tests using constant width delamination specimens
were conducted on an Instron (Norwood, MA) 5583 using displacement control at a crosshead speed of 5 mm/min sampled at 50 ms
per data point. The cyclic testing was conducted using tapered width
double cantilever beam (TWDCB) specimens on an MTS (Eden Prairie, Minnesota, USA) 20 kN servo-hydraulic testing machine retrotted with a variable ow hydraulic supply digitally controlled by an
Instron Labtronic 8400 controller. National Instruments (Austin,
TX) LabVIEW v7 was used to command the controller and perform
data acquisition. The sampling rate for load and displacement was
5 ms. Samples were rst loaded in a monotonic pop-in procedure
to advance the crack from the PTFE pre-crack, and the load required
for steady crack advancement was noted. Next, the crack tip was
sharpened by cycling the sample at 25% of this steady crack advancement load for 1000 cycles. Subsequent testing of the sample was
conducted in load control in zero-tension fatigue (stress ratio,
R = 0.05) at a frequency of 0.3 Hz with continuous monitoring of
crack front displacement. Following testing, sections of the fracture
surface were sputter-coated with a 2.5 nm platinum lm and imaged in a Hitachi S-5000 cold eld emission scanning electron microscope (SEM) with an accelerating voltage of 10 kV.
3. Results
3.1. Mode I delamination crack propagation

2. Materials, processing, and experimental methods


2.1. Materials
The epoxy resin and hardener used in this study was EPON 826
and Epikure 3234, respectively, obtained from Hexion Specialty
Chemicals, Inc. (Houston, Texas, USA). The EPON 826 resin was
blended with 1% by weight of multi-walled carbon nanotubes
using a high shear mixing process by Nanoledge (Clapiers, France).

Measurements of GIC, the critical strain energy release rate, were


obtained from the loaddisplacement proles using the relation:

H
GIC

Pdd
wDa

P is the applied load, d is the grip displacement, w is the sample


width and Da is the incremental change in the crack length [31].
The measured mean and standard deviation of GIC for the conven-

903

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

Fig. 1. Cyclic mode I delamination crack propagation data for glass ber-epoxy
laminates with and without the addition of 1% by weight of CNTs. Values shown are
the constants C, m in the equation: da/dN = C(DG)m.

tional and hybrid composites, respectively, were 790 [18.1] J/m2


and 816 [23.3] J/m2.

Fig. 2. Cyclic delamination fracture surface at low DG (220 J/m2). Note the
preponderance of pulled-out CNTs relative to the number of fractured CNTs.

3.2. Cyclic mode I delamination crack propagation


The testing of TWDCB specimens allowed for the stable measurement of crack propagation at sub-critical cyclic strain energy
release rates from which the constants C and m in the power law
relation given in Eq. (2) were determined [32].

da
CDGm
dN

da is the incremental crack advancement corresponding to an incremental load cycle, dN, at an applied cyclic strain energy release rate
of DG. For the TWDCB specimens, DG is given by [33]:
2

DG

12k
E1 h

P2MAX  P2MIN

where k is a measure of the angle of taper of the sample, E1 is the


elastic modulus of the laminated sample in the 1-direction, h is
the half-thickness of the sample, and PMAX and PMIN are the maximum and minimum applied loads, respectively.
Measured cyclic delamination crack propagation data are
shown in Fig. 1 with tted values for C and m. Vertical dashed lines
delineate the three regions of crack growth behavior; the nearthreshold region corresponding to Gth, the range described by Eq.
(2), and the critical crack growth or fast fracture region corresponding to GIC.
3.3. Fracture surface analysis
Specimens corresponding to different levels of the applied DG
were sectioned and imaged in the SEM. Examination of the fracture
surfaces revealed that CNTs at the surface were either pulled-out
or fractured. Fracture surfaces corresponding to low values of DG
(or low crack growth rates) were found to contain a signicantly
higher fraction of pulled-out rather than fractured CNTs. Conversely, those regions corresponding to higher levels of DG displayed an increasing fraction of fractured rather than pulled-out
CNTs. Figs. 2 and 3 show micrographs of surfaces fractured by a
cyclically propagating crack at DG values of 220 and 830 J/m2,
respectively. The micrograph shown in Fig. 2 is a fracture surface
created by a slow moving crack propagated by a low DG of 220 J/
m2 and shows a preponderance of pull-outs of the CNTs. The fracture surface created by a crack propagated by a high DG of 830 J/
m2 (Fig. 3) shows instead a much smaller number of pulled-out
CNTs compared to the fractured CNTs.

Fig. 3. CNT fracture behavior at high DG (830 J/m2). A much smaller number of
pulled-out CNTs than fractured CNTs are present at the fracture surface at high DG
values.

Table 1 lists measured fractions of pulled-out CNTs relative to


fractured CNTs at increasing levels of applied DG. In each case, the
measurements were performed over a 16 lm2 section of the fracture
surface.
4. Modeling
4.1. Background
Individual CNTs are assumed to be evenly distributed and randomly oriented within the epoxy matrix. The bond between the
Table 1
Observed CNT pull-out fraction, f, for varying DG.
Applied DG [J/m2]

223
430
830

0.56
0.39
0.26

904

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

nanotubes and the surrounding epoxy matrix is assumed to be van


der Waals [3436]. Single walled CNTs have been shown to possess
a lower transverse coefcient of thermal expansion (CTE) relative
to the polymer matrix [37]. Here we assume that a similar CTE mismatch occurs in the case of multi walled CNTs and that this CTE
mismatch following curing at an elevated temperature results in
a compressive radial pre-stress on the nanotubes at room temperature. This gives rise to a frictional shear stress, sP, that resists
nanotube pull-out. It is also believed that a certain degree of interlocking between the CNTs and the matrix may contribute to sP
[38]. Nanotubes are presumed to pull-out from the matrix unless
they cross the fracture plane at a steep enough angle, u, that will
cause fracture. From observations made of the fracture surface
micrographs and previous work which show a dependence of the
interfacial shear stress on pull-out velocity [5,6], the angle at which
the behavior transitions from pull-out to fracture, uc, is assumed to
be a function of the applied DG.
Relations describing the potential energy losses resulting from
matrix cracking in a brittle composite subjected to monotonic

loading have been developed previously by [39]. In the present


case of cyclic loading of a reinforced polymer, the behavior is analogous to monotonic behavior of a brittle system since the applied
cyclic strain levels are well within the elastic range, and crack
propagation is localized. For the hybrid composite, potential energy losses result from, in addition to matrix cracking, frictional
sliding at the CNT/matrix interface during CNT pull-out, and fracture of the CNTs. In the following, symbols corresponding to the
conventional composite will be natural (X, Y), whereas those corre~ Y).
~
sponding to the hybrid composite case will be accented (X,
Fig. 4 presents the three states corresponding to the two cases.
State (0) corresponds to a free body of volume V and surface ST, not
subjected to any external loads, and exhibiting no displacements
or strains, containing an internal stress distribution r0. State (1) corresponds to the same body, with applied tractions T, which result in a
stress distribution r1, displacements u1, and corresponding strains
e1. With no change in the surface tractions T, State (2) has undergone
some internal cracking resulting in crack surface SC, and new values
of stress distribution r2, displacements u2, and strains e2.

Fig. 4. Three states of a conventional and hybrid composite (adapted from Budiansky et al., 1986).

905

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

~
Similarly, for the case of the hybrid composite matrix, State (0)
~
~
corresponds to a free body of volume V and surface ST , not subjected
to any external loads, exhibiting no displacements or strains, con~ corresponds
~ 0 . State (1)
taining some internal stress distribution r
~ which result in a stress
to the same body, with applied tractions T,
~ 1 , and corresponding strains ~e1 .
~ 1 , displacements u
distribution r
~ has undergone
~ State (2)
With no change in the surface tractions T,
some internal cracking resulting in new crack surface ~
SC , frictional
sliding due to pull-out at the new CNT/matrix interfacial surface
~
SF , and new values of
SP , fracture of CNTs creating new CNT surface ~
~ 2 , displacements u
~ 2 , and strains ~e2 .
stress distribution r
The constitutive relations relating stress to strain in the two
cases are

e Mr  r0

for the conventional glassber composite and

~ r
~e M
~ r
~ 0

~ are the
for the CNT/glassber hybrid composite, where M and M
elastic compliances for the two cases.
For the case of the conventional composite, the potential energy
in each of the three states respectively is described by [39]:

Z
1
2 V
Z
1

2 V
Z
1

2 V

p0

r0 : Mr0 dV

p1

r1 : Mr1 dV 

p2

6
Z

T  u1 dV

r2 : Mr2 dV 

T  u2 dV

1
2

r1 r2 : Mr1  r2 dV 

ST

T  u1  u2 dS

r2 : Mr1  r2 dV

10

From 4, 9, and 10 the general form for potential energy loss due to
matrix cracking, in going from State (1) to State (2) is

1
2

r1  r2 : e1  e2 dV

11

Derivation of the potential energy loss for the case of an elastic


body containing frictional sliding (from nanotube pull-out) and ber breakage (from nanotube fracture), proceeds much the same as
before, up until the calculation of the virtual work. For completeness, the potential energy equation for the second case takes the
exact form as its predecessor, Eq. (6), but with accented notation

1
p~ 1  p~ 2
2

Z
~
V

~ r
~
~1 r
~ 2 : M
~1  r
~ 2 dV
r

Z
~S
T

~1  u
~2 d~S
T~  u

The difference in the derivation of the energy relations in the


hybrid matrix case is the two additional internal virtual work
terms corresponding to frictional sliding and ber breakage. The
frictional sliding term is

~S
P

~S
F

sP  dP d~S

Wdd~S

15

where W is dened as the cohesive traction corresponding to the inter-atomic forces binding two halves of a nanotube together per unit
nanotube cross-sectional area, is the distance between the two
halves, and 0 is the equilibrium carboncarbon bond length. Writing
Eq. (10) with accented notation and adding the terms found in (14)
and (15) gives the new virtual work equation for the hybrid composite

Z
~S
T

~1  u
~ 2 d~S
T~  u

Z
V~

~ r
~1  r
~ 2 dV~  sP
r~ 2 : M
Z

Z
~S
F

13

dd~S

~S
P

Wdd~S

16

Combining Eqs. (5), (12), and (16) gives the nal form for the potential energy loss in an elastic body due to matrix cracking, frictional
sliding and fracture of the nanotubes:

1
2

Z
~
V

~1  r
~ 2
r
Z
~S
P

dd~S

Z
~S
F

Wdd~S

17

The next step is to develop a means for making a direct comparison of the two potential energy loss equations previously developed, or more specically to make a direct comparison of the
~ While
potential energy difference between State (2) and State (2).
the two relations, Eq. (11) for the case of the conventional composite, and Eq. (17) for the case of the hybrid composite, have a similar
form, they are derived for two different materials and as such,
steps need to be taken before a direct comparison is valid. This crucial necessary step is to recognize that while the two materials are
in fact different, the outward, macro-scale mechanical behavior of
the two materials is argued here to be equivalent. As we have
shown earlier [27], the difference in the elastic modulus and the
yield strength of the CNT and non-CNT epoxy samples are small
and are well within the statistical variability and measurement
uncertainties that are to be expected in such tests. In light of this
established symmetry in material properties one can equate the
macro-scale compliances of the two matrix materials:

~
MM

18

~ are energetically equivRecognizing that State (1) and State (1)


~ 1 , setting the external tractions T and T~ and
alent, i.e., that p1 = p
the associated internal stresses to be equal, and making use of
Eq. (18), the difference between Eq. (17) and Eq. (11) is given by

p~ 2  p~ 2 sP
12

~ sP
: ~e1  ~e2 dV

ST

14

The ber breakage internal virtual work term is

ST

T  u1  u2 dS

p1  p2

dd~S

~S
P

Equating the external virtual work of the traction T over the displacement u1  u2 to the internal virtual work corresponding to
the internal stress, r2 over the strain M(r1  r2),

p~ 1  p~ 2

where for each state, the volume integral describes the stored elastic energy and the surface integral describes the work done at the
surface of the body by the applied tractions. Taking the compliance
to be a constant, r1 : Mr2 r2 : Mr1 ; the change in potential
energy of the body due to the internal cracking is

p1  p2

sP

ST

where sP is the shear traction applied by the polymer matrix on the


nanotube as it slides a distance dP during the pull-out process. Noting that the nature of a frictional shear traction implies that the
traction will always be operating in the direction opposite the sliding, we can remove the dot product, simplifying Eq. (13) to

Z
~S
P

dP d~S

Z
~S
F

Wdd~S

19

Eq. (19) represents the difference in lost potential energy between the
~ The right
two bodies required to reach the damaged States (2) and (2).
hand side of Eq. (19) is the amount of strain energy that is absorbed by
the two damage processes unique to the hybrid composite, during the
formation of a unit area of matrix crack. It is this absorption of strain
energy in a way that does not promote the growth and coalescence of
matrix microcracks that has been proposed here as the means by
which the hybrid composites are able to signicantly outperform
conventional composites in high-cycle fatigue.

906

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

In order to make use of the previous result for predicting the


behavior of the CNT/glassber hybrid composites in fatigue, it is
necessary to relate the terms it contains to measured values. While
the full nature of the properties of each individual CNT is difcult
to assess, their large quantity allows for typical or average values
to be established for use here. To this end, a simple model is presented that establishes typical values for CNT dimensions, distribution and orientation. In addition to this model of the average
nanotube, values for the pertinent physical constants in Eq. (19)
are derived from the previously described experiments as well as
from the literature. Lastly, the energy model will be converted into
a more useful energy rate form, which will allow for direct comparison to the results obtained experimentally.
4.2. Typical CNT characteristics
The details of the physical properties of the carbon nanotubes
used here were obtained from manufacturer. In addition, examination of the SEM micrographs conrmed the distribution of values
for nanotube diameter and pull-out length. This information is
used to establish an average mass density of the nanotubes distributed in the epoxy resin. This average density, found to be 1310 kg/
m3, along with an epoxy density of 1170 kg/m3 and the mass fraction, mf, of 0.01, gives a volume fraction, Vf, of the CNTs:

mf
Vf
0:009
mf 1  mf qf =qm

20

The number of randomly oriented tubes that will cross an arbitrary


unit fracture plane is given by [40]:

NR

Z p=2
Vf
Vf
cos u sin u du
pr2
2p r 2
0

21

where u is the angle between the normal to the fracture plane and
the axis of the nanotube as shown in Fig. 5. Substituting the pertinent values into Eq. (21)

NR 1:43  1013 m2


This number is in relative agreement with a series of nanotube
counts performed on micrographs taken of the fracture surface.
Recognizing that CNT involvement is not limited to the primary
fracture surface, but rather that CNTs can be expected to dissipate
energy by way of the previously mentioned mechanisms throughout the plastic zone wake formed during crack propagation, we
write the total number of CNTs involved as

N0R NR /

22

where / is proportional to the radius of the plastic zone, calculated as

1 EGIC
rp
2p S2y

!
23

where E is the Youngs modulus and Sy the yield strength whose values are 2 GPa and 40 MPa, respectively [27].

4.3. CNT pull-out fraction


As discussed previously, there is an observed relationship between the applied DG and the resulting pull-out fraction f. Table
1 contains the observed values of f at various levels of DG. This data
is used to predict the expected f over the range of applied DG in the
power law regime described by Eq. (2).
4.4. Determination of differential energy relation constants
In order to solve the right hand side of the differential energy
relation in Eq. (19) it is rst necessary to evaluate the various terms
that appear in the two integrals. Solving the rst of these two
terms, corresponding to the friction during pull-out of nanotubes,
requires the determination of the frictional shear stress, sP, that
acts on the surface of a single CNT as it is being pulled-out. Additionally, the total sliding distance, dP, and the area over which sliding takes place, ~
SP , are required. The second term, corresponding to
the energy required for CNT fracture, requires knowledge of the
cohesive force of the carboncarbon bonds as a function of separation distance, , and the area over which this force acts, giving us
SF , over which
the cohesive traction, W. Additionally, the area, ~
these forces are distributed are required. We note that

NCC  nCC 

Z
~S
F

WddS

24

where NCC is the number of carboncarbon bonds broken and nCC


is the carboncarbon bond energy. sP, the effective interfacial shear
stress, is obtained from the KellyTyson model [41], with a slight
modication for the case of hollow bers [35]. This modied model,
based on a simple force balance, relates the interfacial shear stress
to the CNT strength:

sP

rut do
4dt

 2 !
di
1
do

25

where rut is the ultimate tensile strength of a CNT, do and di are


respectively the outer and inner diameter of the CNT, and dt is the
pull-out length. Based on values of rut reported elsewhere ranging
from 50 GPa to 150 GPa [42,43], the calculated values for sP range
from 570 to 1700 MPa, the lower end of which is in close agreement
with [35]. Table 2 presents the conservative value of sP, the established value for nCC , and average values for the remaining constants.
Table 2
Constants used in differential energy relation.
Property

Symbol

Value

CNT-matrix shear stress


CNT sliding length
CNT sliding area

sp
dt
~P
S

5701700 MPa
250 nm
15,700 nm2

Number of CC bonds
CC bond energy

NCC
nCC

2330
348 kJ/mol

Fig. 5. Possible CNT congurations intersecting a fracture surface.

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

Fig. 6. Experimental data and predicted delamination behavior (shaded region) in


the da/dN = C(DG)m regime.

4.5. Formation of differential energy release rate model


The energy loss per unit area as a result of the inclusion of CNTs
in the polymer matrix is:

"
~2 P
~2
EP

s~P ~SP ~dt


2

!
f

#
NCC nCC  1  f N0R

26

~ 2 are p2 and p
~ 2 normalized per unit area of crack
where P2 and P
surface, respectively, the sum inside the square brackets corresponds to the potential energy loss through each of the two mechanisms, pull-out or fracture, scaled by the frequency by which each
mechanism occurs, and N 0R is the number of CNTs involved in the
fracture process per unit area of crack surface created, with / from
Eq. (23) conservatively estimated to be 2.
E, the differential energy release rate, is the difference in the rate
of potential energy dissipated by CNT pull-out and CNT fracture per
unit area of new crack surface created in a CNT/glassber hybrid
composite when compared to an equivalent delamination crack in
a conventional glassber composite. Eq. (26) allows one to estimate
the differential increase in energy dissipation resulting from the
incorporation of CNTs in glassber composites. Shifting the delamination response of the conventional composite (without CNTs) to
the right by E corresponds to the response of the composite containing CNTs. This was done using the experimentally measured values
of the pull-out fraction, f, as a function of DG (given in Table 1) for
the estimated range of values for sp discussed earlier. The resulting
range is depicted as a shaded zone in Fig. 6 and shows good correlation with the cyclic data. As DG increases, the effect of the CNTs in
delaying delamination crack propagation decreases due to a
decreasing fraction of pull-out relative to fracture. In other words,
the CNTs act to slow crack propagation when the applied DG is small,
i.e., in the high-cycle regime, when a larger fraction of them are being
pulled-out. At faster crack propagation rates corresponding to higher values of DG, the CNTs tend to fracture rather than pull-out making them ineffective in slowing crack propagation.
5. Summary
The delamination crack propagation behavior in the two composite types (glassber composites with and without the addition
of CNTs), both under critical and sub-critical levels of applied strain
energy, can be readily understood through examination of Fig. 1.
The critical strain energy release rate, GIC, in the conventional
glassber composites is found to be consistent with values reported elsewhere [1,44], with the corresponding value for the hybrid composites only slightly higher, in agreement with previous
studies [12]. The convergence of the response at increasing levels

907

of applied DG suggests that the mechanisms at work in the hybrid


composite inhibiting the crack propagation rate are, in part, diminished at higher levels of applied strain energy. This idea is backed
by visual evidence in micrographs of the fracture surface which
clearly show the replacement of the relatively high-energy dissipating pull-out behavior with the relatively low-energy dissipating
fracture behavior at higher applied DG.
The two threshold values in Fig. 1, noted as Gth, correspond to
the measured threshold in DG below which no crack propagation
was observed. At all levels of DG within the linear region, the rate
of crack propagation for the hybrid composites was found to be
signicantly lower than for the conventional composites, with
the difference ranging from 2.5 times as DG approached Gth, to
around 1.5 times as DG approached GIC.
An energy-based model has been developed that draws from previous work in fracture of ceramic composites [39] and integrates the
expected losses due to CNT pull-out and fracture. The model provides a range of behavior in the fracture of CNTreinforced hybrid
composites when compared to conventional matrix composites.
Acknowledgement
This work was supported in part by a Grant from Entropy Research Laboratories, San Francisco, California, USA.
References
[1] Devitt DF, Schapery RA, Bradley WL. A method for determining mode I
delamination fracture toughness of elastic and viscoelastic composite
materials. J Compos Mater 1980;14:27085.
[2] Whitney JM, Browning CE, Hoogsteden W. A double cantilever beam test for
characterizing mode I delamination of composite materials. J Reinf Plast
Compos 1982;1:297313.
[3] Saghizadeh H, Dharan CKH. Delamination fracture toughness of graphite and
aramid epoxy composites. J Eng Mater Technol 1986;108:2905.
[4] Barber AH, Cohen SR, Kenig S, Wagner HD. Interfacial fracture energy
measurements for multi-walled carbon nanotubes pulled from a polymer
matrix. Compos Sci Technol 2004;64:22839.
[5] Desai AV, Haque MA. Mechanics of the interface for carbon nanotubepolymer
composites. Thin-Wall Struct 2005;43:1787803.
[6] Frankland SJV, Harik VM. Analysis of carbon nanotube pull-out from a polymer
matrix. Surf Sci 2003;525:L1038.
[7] Frankland SJV, Harik VM. Simulation of carbon nanotube pull-out when
bonded to a polymer matrix. Mater Res Soc Symp Proc 2003;740:I12.1.1.6.
[8] Kao CC, Young RJ. Modeling the stress transfer between carbon nanotubes and
a polymer matrix during cyclic deformation. In: IUTAM symposium on
modelling nanomaterials and nanosystems, vol. 1; 2009. p. 21120.
[9] Tai NH, Yeh MK, Liu JH. Enhancement of the mechanical properties of carbon
nano tube/phenolic composites using a carbon nanotube network as the
reinforcement. Carbon 2004;42:27747.
[10] Zhang J, He C. A three-phase cylindrical shear-lag model for carbon nanotube
composites. Acta Mech 2008;196:3354.
[11] Qiu J, Zhang C, Wang B, Liang R. Carbon nanotube integrated multifunctional
multiscale composites. Nanotechnology 2007;18:275708.1275708.11.
[12] Wichmann MHG, Sumeth J, Gojny FH, Quaresimin M, Fiedler B, Schulte K.
Glassbre-reinforced composites with enhanced mechanical and electrical
properties benets and limitations of a nanoparticle modied matrix. Eng
Fract Mech 2006;73:234659.
[13] Zhu J, Imam A, Crane R, Lozano K, Khabashesku VN, Barrera EV. Processing a
glass ber reinforced vinyl ester composite with nanotube enhancement of
interlaminar shear strength. Compos Sci Technol 2007;67:150917.
[14] Thakre PR, Lagoudas DC, Zhu J, Barrera EV, Gates TS. Processing and
characterization of epoxy/SWCNT/woven fabric composites. In: Proceedings
AIAA structures, structural dynamics, and materials conference, vol. 47; AIAA
Paper 2006-1857:2006.
[15] Adhikari K, Hubert P, Simard B, Johnson A. Effect of the localized application of
SWNT modied epoxy on the interlaminar shear strength of carbon bre
laminates. In: Proceedings AIAA structures, structural dynamics, and materials
conference, vol. 47; AIAA Paper 2006-1855:2006.
[16] Kowbel W, Bruce C, Withers JC, Ransone PO. Effect of carbon fabric
whiskerization on mechanical properties of CC composites. Composites Part
A 1997;28A:9931000.
[17] Thostenson ET, Li WZ, Wang DZ, Ren ZF, Chou TW. Carbon nanotube/carbon
ber hybrid multiscale composites. J Appl Phys 2002;91:60347.
[18] Garcia EJ, Wardle BL, Hart AJ, Yamamoto N. Fabrication and multifunctional
properties of a hybrid laminate with aligned carbon nanotubes grown in situ.
Compos Sci Technol 2008;68:203441.

908

C.S. Grimmer, C.K.H. Dharan / Composites Science and Technology 70 (2010) 901908

[19] Bekyarova E, Thostenson ET, Yu A, Kim H, Gao J, Tank J, et al. Multiscale carbon
nanotubecarbon ber reinforcement for advanced epoxy composites.
Langmuir 2007;23:39704.
[20] Veedu VP, Cao A, Li X, Ma K, Soldano C, Kar S, et al. Multifunctional composites
using reinforced laminae with carbon-nanotube forests. Nat Mater
2006;5:45762.
[21] Garcia EJ, Wardle BL. Hart AJ Joining prepreg composite interfaces with aligned
carbon nanotubes. Composites Part A 2008;39:106570.
[22] Penado FE. A closed form solution for the energy-release rate of the double
cantilever beam specimen with an adhesive layer. J Compos Mater
1993;27:383407.
[23] Alfano G, Criseld MA. Finite element interface models for the delamination
analysis of laminated composites: mechanical and computational issues. Int J
Numer Meth Eng 2001;50:170136.
[24] Wang SS, Miyase A. Interlaminar fatigue crack growth in random short-ber
SMC composite. J Compos Mater 1986;20:43956.
[25] Wang SS, Suemasu H, Zahlan NM. Interlaminar fracture of random short-ber
SMC composite. J Compos Mater 1984;18:57494.
[26] Grassi M, Zhang X. Finite element analyses of mode I interlaminar
delamination in z-bre reinforced composite laminates. Compos Sci Technol
2003;63:181532.
[27] Grimmer CS, Dharan CKH. High-cycle fatigue of hybrid carbon nanotube/glass
ber/polymer composites. J Mater Sci 2008;43:448792.
[28] Ganguli S, Aglan H. Effect of loading and surface modication of MWCNTs on
the fracture behavior of epoxy nanocomposites. J Reinf Plast Compos
2006;25:17588.
[29] Gojny FH, Wichmann MHG, Fiedler B, Bauhofer W, Schulte K. Inuence of
nano-modication on the mechanical and electrical properties of conventional
bre-reinforced composites. Composites Part A 2005;36:152535.
[30] Kim JA, Seong DG, Kang TJ, Youn JR. Effects of surface modication on
rheological and mechanical properties of CNT/epoxy composites. Carbon
2006;44:1898905.

[31] Gurney C, Hunt J. Quasi-static crack propagation. Proc Roy Soc Lond, Sec A
1967;299:50824.
[32] Paris P, Erdogan F. Critical analysis of crack propagation laws. J Basic Eng, Trans
ASME 1963;85:52834.
[33] Hwang W, Han KS. Interlaminar fracture behavior and ber bridging of glassepoxy composite under mode I static and cyclic loading. J Compos Mater
1989;23:396430.
[34] Schadler LS, Giannaris SC, Ajayan PM. Load transfer in carbon nanotube epoxy
composites. Appl Phys Lett 1998;73:38424.
[35] Wagner HD, Lourie O, Feldman Y, Tenne R. Stress-induced fragmentation of
multiwall carbon nanotubes in a polymer matrix. Appl Phys Lett
1998;72:18890.
[36] Lordi V, Yao N. Molecular mechanics of binding in carbonnanotubepolymer
composites. J Mater Res 2000;15:27709.
[37] Jiang H, Liu B, Huang Y, Hwang KC. Thermal expansion of single wall carbon
nanotubes. J Eng Mater Technol 2004;126:26570.
[38] Wong M, Paramsothy M, Xu XJ, Ren Y, Li S, Liao K. Physical interactions at
carbon nanotubepolymer interface. Polymer 2003;44:775764.
[39] Budiansky B, Hutchinson JW, Evans AG. Matrix fracture in ber reinforced
ceramics. J Mech Phys Solids 1986;34:16789.
[40] Aveston J, Kelly A. Theory of multiple fracture of brous composites. J Mater
Sci 1973;8:35262.
[41] Kelly A, Tyson WR. Tensile properties of ber-reinforced metalscopper/
tungsten and copper/molybdenum. J Mech Phys Solids 1965;13:32950.
[42] Demczyk BG, Wang YM, Cummings J, Hetman M, Han W, Zettl A, et al. Direct
mechanical measurement of the tensile strength and elastic modulus of
multiwalled carbon nanotubes. Mater Sci Eng A 2002;334:1738.
[43] Yakobson BI, Smalley RE. Fullerene nanotubes: C-1000000 and beyond. Am Sci
1997;85:32434.
[44] Bazhenov SL. Interlaminar and intralaminar fracture modes in 0/90 cross-ply
glass epoxy laminate. Composites 1995;26:12533.

Potrebbero piacerti anche