Downloaded from rspa.royalsocietypublishing.

org

Prediction of elastic properties of carbon nanotube

reinforced composites
N Hu, H Fukunaga, C Lu, M Kameyama and B Yan
Proc. R. Soc. A 2005 461, 1685-1710
doi: 10.1098/rspa.2004.1422

References This article cites 31 articles, 1 of which can be accessed free

http://rspa.royalsocietypublishing.org/content/461/2058/1685.full.
html#ref-list-1

Article cited in:

http://rspa.royalsocietypublishing.org/content/461/2058/1685
.full.html#related-urls
Email alerting service Receive free email alerts when new articles cite this article - sign up in
the box at the top right-hand corner of the article or click here

To subscribe to Proc. R. Soc. A go to: http://rspa.royalsocietypublishing.org/subscriptions

This journal is © 2005 The Royal Society

 Downloaded from rspa.royalsocietypublishing.org

Proc. R. Soc. A (2005) 461, 1685–1710

doi:10.1098/rspa.2004.1422
Published online 10 May 2005

Prediction of elastic properties of carbon

nanotube reinforced composites
B Y N. H U 1 , H. F UKUNAGA 1 , C. L U 2 , M. K AMEYAMA 1 AND B. Y AN 1
1
Department of Aeronautics and Space Engineering, Tohoku University,
Aobayama 01, Aoba-ku, Sendai 980-8579, Japan
(hu@ssl.mech.tohoku.ac.jp)
2
Department of Chemical Engineering, Tohoku University, Aobayama 01,
Aoba-ku, Sendai 980-8579, Japan

In this paper, the macroscopic elastic properties of carbon nanotube reinforced

composites are evaluated through analysing the elastic deformation of a representative
volume element (RVE) under various loading conditions. This RVE contains three
components, i.e. a carbon nanotube, a transition layer between the nanotube and
polymer matrix and an outer polymer matrix body. First, based on the force field theory
of molecular mechanics and computational structural mechanics, an equivalent beam
model is constructed to model the carbon nanotube effectively. The explicit relationships
between the material properties of the equivalent beam element and the force constants
have been set-up. Second, to describe the interaction between the nanotube and the
outer polymer matrix at the level of atoms, the molecular mechanics and molecular
dynamics computations have been performed to obtain the thickness and material
properties of the transition layer. Moreover, an efficient three-dimensional eight-noded
brick finite element is employed to model the transition layer and the outer polymer
matrix. The macroscopic behaviours of the RVE can then be evaluated through the
traditional finite element method. In the numerical simulations, the influences of various
important factors, such as the stiffness of transition layer and geometry of RVE, on the
final macroscopic material properties of composites have been investigated in detail.
Keywords: carbon nanotube; elastic property; composites;
representative volume element; molecular mechanics; finite element method

1. Introduction

The exceptional mechanical, thermal and electrical properties plus their low
density and high aspect ratio, exhibited by carbon nanotubes, make them an
ideal candidate for composite reinforcement (Tostenson et al. 2001). Although
there is a lot of research (Qian et al. 2002) dealing with the elastic properties of
the carbon nanotube through various means, the investigations of the mechanical
properties of carbon nanotube reinforced composites have been rarely reported
(Tostenson et al. 2002; Liu & Chen 2003). To understand the elastic properties of
nanotube reinforced composites, a fundamental challenge exists in the
characterization and modelling of these materials at the nanoscale. From
the traditional theoretical frame for evaluating the macroscopic elastic properties
Received 26 April 2004 1685 q 2005 The Royal Society
Accepted 22 September 2004
 Downloaded from rspa.royalsocietypublishing.org

1686 N. Hu and others

Figure 1. A representative volume element of nanotube reinforced composite and its corresponding
FEM macroscopic model.

of composites, a possible approach is to build up a representative volume element

(RVE) constituted by a cylindrical body of matrix with an embedded nanotube.
However, it is almost impossible to model this RVE using the molecular
dynamics directly if the matrix materials (e.g. polymer) and the nanotube are all
considered. A continuum mechanics approach (Tostenson et al. 2002; Liu & Chen
2003) can be employed to tackle this problem, although the reflection of some
important characteristics of components in composites at the level of atoms
should be required.
In this paper, as shown in figure 1, a RVE of cylindrical matrix containing a
carbon nanotube is set-up. This volume element consists of three components, i.e.
a nanotube, a transition layer between the nanotube and the polymer matrix, and
an outer polymer matrix body. These three portions need to be modelled
effectively and independently. First, before entering the stage of prediction of the
mechanical properties of composites, there is an urgent need to set-up a reliable
structural mechanics model for analysing the nanotube, which can reflect the
interaction of atoms on the nanotube at the nanoscale. Here, we construct a
structural mechanics 3D beam model based on the second generation molecular
force field in molecular mechanics and the computational structural mechanics.
This equivalent beam frame model is employed to describe the deformation
behaviour of nanotubes under various mechanical loads. To uncover the
connections between the force constants and the material properties of
the equivalent beam element, a graphite sheet is considered first. The ties between
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1687

the force constants and the macroscopic elastic properties of graphite sheet have
been identified explicitly. Then, by evaluating the macroscopic elastic properties
of a graphite sheet through the equivalent beam model, the explicit relationships
between the material properties of the equivalent beam element and the molecular
force constants are set-up. The correctness of these relationships has been further
verified through a proposed optimum model. The predicted Young’s and shear
moduli of nanotube are accurate using this equivalent frame model. Second, to
describe the interaction between the nanotube and the outer polymer matrix at the
level of atoms, the molecular mechanics computations have been performed to
obtain the thickness of the transition layer, i.e. the equilibrium distance between
the nanotube and the polymer matrix. The material properties of this layer have
been further identified through molecular dynamics computations. An efficient 3D
eight-noded brick element proposed by the present authors (Cao et al. 2002) is
employed to model this layer and the outer polymer matrix. Therefore, the elastic
deformation of the RVE under various loading conditions can be predicted using
the traditional finite element approach. Finally, the macroscopic elastic properties
of nanotube reinforced composites have been investigated using this RVE.

2. Prediction of material properties of nanotube

(a) Relationships between force constants and macroscopic material properties

of graphite sheet
The molecular mechanics force field or potential functions among atoms can be
expressed as follows (Cornell et al. 1995):
X X X X X
Epot ¼ Vs þ Va þ Vt þ Vv þ Ve ; (2.1)
bonds angles dihedrals atoms atoms
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
bonded interactions non-bonded interactions

where Vs is the bond stretch potential summed over all bonds, Va the bond angle
bending potential summed over all angles, Vt the torsion potential owing to
dihedrals torsion, Vv the van der Waals potential summed over all atoms and
Ve the electrostatic interactions, respectively.
For the covalent bond stretching, the bond stretch potential between two atoms
can be estimated as: Vs Z ð1=2ÞKs ðR 0 K RÞ2 Z ð1=2ÞKs DR2 in the form of harmonic
potential, e.g. in AMBER (assisted model building with energy refinement), in
which R is the equilibrium bond distance, which is 0.142 nm for C–C bonds, and
Ks the force constant of bond stretching. Similarly, the bond angle bending
potential can be computed as: Va Z ð1=2ÞKq ðq 0 K qÞ2 Z ð1=2ÞKq Dq2 , where q
denotes the equilibrium angle in radians, which is equal to 1208 for C–C bonds,
and Kq is the angle bending force constant.
To evaluate the elastic properties of carbon nanotube, first, we start from a
graphite sheet, which is generally considered to be homogeneous in plane. Based on
this prerequisite, for a unit cell in graphite as shown in figure 2, in appendix A
we obtain the relationships between the force constants and macroscopic
material properties of graphite. Therefore, the macroscopic material properties
of graphite, i.e. E, n and G, can be calculated from Ks and Kq in equations (A 7a–c).
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1688 N. Hu and others

Figure 2. An arbitrary deformation state for a hexagonal unit cell.

In the currently widely existing references, there are many force constants. Using
the different kinds of force constants, can lead to different macroscopic material
properties of the graphite sheet. We discuss them as follows. Like many other
authors, the thickness of graphite sheet ts in equation (A 5) is taken as 0.34 nm,
i.e. the spatial extent of the p orbitals. Usually, in most references of molecular
mechanics, the units of the force constants Ks and Kq are kcal (mol Å2)K1 and kcal
(mol rad2)K1, respectively. For the convenience of computation, we exchange
them into nN nmK1 and nN nm radK2, respectively. For the force constants
(Cornell et al. 1995) KsZ652.2 nN nmK1 and KqZ0.876 nN nm radK2, we get
nZ0.2502, GZ0.332 124 1 TPa and EZ0.830 426 5 TPa. For the constants
used in the second Tersoff–Brenner potential (Tersoff 1988; Brenner 1990),
K s Z848 nN nm K 1 and K q Z1.211 34 nN nm rad K 2 , we get nZ0.245,
GZ0.4364 TPa and EZ1.087 TPa. For the constants (Jin & Yuan 2002)
K s Z973.33 nN nm K 1 and K q Z1.3905 nN nm rad K 2 , we get nZ0.245,
GZ0.501 TPa and EZ1.2475 TPa. For the parameters identified by Chang &
Gao (2003), KsZ742.0 nN nmK1 and KqZ1.42 nN nm radK2, we get nZ0.217 905 7,
GZ0.404 56 TPa and EZ0.985 425 TPa. For the force constants used by Walther
et al. (2001), KsZ795.765 nN nmK1 and KqZ0.9342 nN nm radK2, we get
nZ0.2601, GZ0.396 74 TPa and EZ0.999 845 TPa. Generally, at the present
stage, the widely reported Young’s modulus E and Poisson’s ratio n of a graphite
sheet ranges from 1.02 to 1.06 TPa, and from 0.16 to 0.29, respectively. In this
paper, we take EZ1.06 TPa and nZ0.225. Therefore, from equation (A 7a–c),
we can identify KsZ805.5 nN nmK1 and KqZ1.438 nN nm radK2, which are
consistent with the above values in various references.
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1689

Figure 3. A hexagonal unit cell and equivalent truss models.

(b) Relationships between force constants and material properties

of equivalent beam element
To predict the Young’s modulus of nanotube, an equivalent computational
structural model is preferred, which can be easily incorporated into the
computational model of RVE. The first candidate considered is a truss model,
which can easily connect to the outer 3D brick elements of the transition layer.
However, we have found that it is impossible to set-up an appropriate truss
model to represent the graphite sheet. For instance, for the equivalent truss
model, i.e. model 1 in figure 3, two kinds of truss members, type a and type b, are
considered (Odegard et al. 2002). In this truss model, the strain energy for the
deformation state shown in figure 2 can be expressed as
 pffiffiffi   pffiffiffi 
9 9 3 2 3 3 3
GZ RE1 A1 C RE2 A2 3x C RE1 A1 C RE2 A2 3x 3y
8 8 4 4
 pffiffiffi   pffiffiffi 
9 9 3 2 3 3 3
C RE1 A1 C RE2 A2 3y C RE1 A1 C RE2 A2 g2xy ; (2.2)
8 8 8 8
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1690 N. Hu and others

where E1A1 is the stiffness constant of truss members of type a, and E2A2 is the
stiffness constant of truss members of type b.
By comparing the both sides of equations (A 5) and (2.2), three equations
can be set-up:
pffiffiffi pffiffiffi
9 9 3 3 3 E
RE1 A1 C RE2 A2 Z R2 t; (2.3a)
8 8 2 1 K n2
pffiffiffi
3 3 3 pffiffiffi En
RE1 A1 C RE2 A2 Z 3 3 2
R2 t; (2.3b)
4 4 1 Kn
pffiffiffi pffiffiffi
3 3 3 3 3
RE1 A1 C RE2 A2 Z GR2 t: (2.3c)
8 8 2
The above three equations can be satisfied simultaneously only when 3nZ1.
This means that this equivalent truss model can only represent a continuum
homogeneous material of Poisson’s ratio of nZ1/3. However, it is well known
that the Poisson’s ratio of graphite sheet ranges from 0.16–0.29, which is
obviously lower. In fact, no matter what kind of topology of an isotropic
hexagonal equivalent truss model is chosen (e.g. model 2 in figure 3), this truss
model can also only describe a homogeneous material of Poisson’s ratio of nZ1/3.
To overcome this difficulty, a beam model is created. For the graphite sheet
shown in figure 4, as derived in appendix B, the material properties of the beam,
i.e. EA and EI, can be determined from equation (B 14) using the force constants
Ks and Kq. Also, GJ is estimated by equation (B 18). The derivation procedure in
appendix B for EA and EI has been verified through the finite-element method
(FEM) analysis, combined with an optimum model for the graphite sheet in
figure 4, under the same simple compressive deformation state. The Timoshinko’s
beam element is employed. By using x1, x2 and x3 in equation (B 3) as the design
parameters, the following optimization problem can be constructed:

~ K EÞ2 C ð~
minðE n K nÞ2 s:t: E~ O 0 n~O 0; (2.4)

where E~ and n~ are the Young’s modulus and Poisson’s ratio numerically
calculated from the beam model using the FEM, and E and n are the preset
Young’s modulus and Poisson’s ratio of graphite sheet, which are 1.06 TPa and
0.225, respectively.
The solution of the above optimization problem for the case of nZ3 in figure 4
using the complex optimum method implemented with constraint conditions,
leads to the following identified results: x1Z0.861 132, x2Z0.405 722 and
x3Z0.998 278!106. Using the relationships in equation (B 3), the shear constant
in the beam element therefore is: fZ12EI/GAsR2Z5.268!10K6. It means that
the influence of transverse shear deformation is very small. By comparing the
numerically obtained x1 and x2 with the theoretical ones in equation (B 14), there
is a little difference between two kinds of results. The reason for this deviation
is the small number of repeated cells, i.e. nZ3. To increase the number of the
cells along the y-axis direction, we can get the converged numerical results,
which are close to the theoretical ones as shown in figure 5. The above
verification procedure justifies our theoretical results.
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1691

Figure 4. A graphite sheet under simple compression and its representative cell.

(c) Numerical verification of material properties of nanotube

After obtaining the stiffness parameters of beam elements, we can employ the
traditional FEM to solve the displacements of the equivalent beam structure
under the various loading conditions. The capability and efficiency of this
structural mechanics model is verified for single walled carbon nanotubes with a
fixed end and the applied tension and torsion loads at another free end.
Nanotubes of armchair and zigzag configurations are considered, which are of
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1692 N. Hu and others

Figure 5. Comparison between numerical and theoretical results.

length of 4.9225 and 4.9735 nm, respectively. A typical beam model of a zigzag
nanotube (15, 0) is shown in figure 6. The Young’s modulus and shear modulus of
nanotube can be estimated as

FLn TLn
EZ and G Z ; (2.5)
dST uJT

where F and T are the applied tension force and torque, d and u are the
corresponding displacement and rotation angle, Ln is the length of the nanotube,
and ST and JT are the section area and the polar moment of inertia of the annular
cross-sectional area of single walled nanotubes of thickness of 0.34 nm,
respectively. In this 3D space problem, we set EIyZEIzZEI, where EI is
connected with force constants as shown in equation (B 14). Also, the shear
constant of beam element is set to be fZ5.268!10K6.
The results of the Young’s modulus and shear modulus are demonstrated in
figure 7. From this, it can be seen that the Young’s modulus of both kinds of
nanotubes is very close to the experimental value for graphite sheet, i.e.
1.06 TPa. The effect of the nanotube diameter on the Young’s modulus is not
obvious. The shear modulus of nanotubes also gradually tends to be close to the
value of graphite sheet, i.e. 0.433 TPa. The diameter of the nanotubes has more
significant effect on the shear modulus. The present Young’s modulus and shear
modulus of nanotube are both in good agreement with the many previous
theoretical predictions and experimental results (Lu 1997; Wong et al. 1997;
Hernandez et al. 1998; Chopra & Zettl 1998; Krishnan et al. 1998; Salvetat et al.
1999; Popov et al. 2000; Jin & Yuan 2002; Chang & Gao 2003). Note that the
currently existing theoretical and experimental values of the Young’s and shear
moduli are quite scattered. The Young’s modulus of nanotube ranges from 0.9 to
1.3 TPa, and shear modulus of nanotube is from 0.4 to 0.46 TPa from the
different references. Conclusively, the mechanical properties of carbon nanotubes
can be accurately predicted using the equivalent beam model. The obtained
Young’s and shear moduli of single walled nanotubes are very close to those of
graphite sheet. The present structural mechanics beam model can be easily
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1693

Figure 6. A beam model of a zigzag carbon nanotube (15,0).

Figure 7. Young’s and shear moduli of nanotubes using the equivalent beam model.

adopted into an FEM model for the prediction of the mechanical properties of
nanotube reinforced composites.

3. Analysis of transition layer

To obtain the global mechanical behaviour of the RVE efficiently, one key task is
to understand the interactions between the nanotube and the polymer matrix,
which is described by a transition layer here. The various characteristics of the
transition layer, such as the thickness of the transition layer, i.e. the equilibrium
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1694 N. Hu and others

distance between the polymer chain and nanotube, and mechanical properties of
this layer, are discussed in this section. At present, there is a great deal of debate
on this topic. The interfacial characteristics seem to be unclear. According to
some previous studies, which are mostly focused on the interfacial bond
categories and strength, so far there are three considerations of possible
connections between these two atomic structures, i.e. the nanotube hexagonal
lattice structures and polymer chain structures.
In the first consideration, some researchers advocated that there is no chemical
bond between the nanotube and polymer. For instance, using Raman scattering
and X-ray diffraction, Chang et al. (2000) showed that there is no chemical bond
between carbon nanotube and the polypyrrole matrix. Therefore, in some studies
(Lordi & Yao 2000; McCarthy et al. 2000; Liao & Li 2001), only non-bonded
potentials from electrostatic and van der Waal’s forces are considered. To avoid
weak interfacial strength, Lordi & Yao (2000) conjectured that the chains of
polymer, wrap the nanotube. The key factor in forming a strong bond at the
interface is having a helical conformation of the polymer around the nanotube.
McCarthy et al. (2000) suggests a van der Waals interaction, analogous to
J-aggregate stacking of aromatic p-systems, between the benzene rings of the
polymer and C–C hexagonal lattice structure. They experimentally observed
that the polymer strands are in fact coiling around each other to form ropes,
which in turn surround the nanotubes in a regular, structured way.
In the second consideration, it is claimed that there are strong chemical bonds
between the tube and polymer. For example, from infrared transmission spectra,
Jia et al. (1999) concluded that C–C bond exists between carbon nanotubes and
poly (methylmethacrylate) system. Wagner et al. (1998) and Cooper et al. (2002)
suggested that high interfacial shear stress determined from a carbon nanotube/
polyurethane composite system is attributed to a ‘2C2’ cycloaddition reaction
between the tube and the polymer. Therefore, the strong chemical links between
the nanotube and polymer are supposed to be formed.
In the third consideration, Frankland et al. (2002), investigated how covalent
cross-links between a polymer and nanotubes affect the mechanical properties of
the composites. However, only a very small amount of cross-links are assumed,
which is consistent with the finding of Chang et al. (2000). A newer form of the
Brenner potential (Brenner et al. 2002) was used to describe the intramolecular
interactions in the nanotubes, polymer chains and cross-links.
The contribution of the present study in this field originates from the following
two aspects: (i) to reveal the chirality configuration of polymer chain around the
nanotube; (ii) to study the properties of transition layer, especially the average
mechanical properties of transition layer, which have not been reported
previously.
To fulfil the above objectives, we select the chain of polystyrene (PS)

½ðKCH2 CHC6 H6KÞn
or

with 24 benzene rings and a (6, 6) armchair nanotube with length of 4.9225 nm.
A many-body bond-order potential with cvff-950-1.01 force field (consistent valence
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1695

Figure 8. Molecular mechanics model of nanotube and polymer chain. (a) Side view; (b) top view.

force field) implemented in molecular mechanics software CERIUS2 v. 3.8 was used to
describe the intramolecular interactions between the nanotubes and polymer
chains. This potential allows the formation of possible chemical bonds between the
atoms on the tube and polymer. After putting the polymer chain closely to the
nanotube, the chain automatically starts to surround the nanotube in a regular,
structured and stable way after several hundred iterations. The convergence of
system energy is reached after around 400 iterations. The equilibrium configur-
ations after convergence are shown in figure 8. The equilibrium distance between H
atoms on polymer and C atoms on nanotube ranges from 0.2851 to 0.5445 nm.
Finally, we take the average value of the range as the thickness of the transition
layer, th, i.e. 0.415 nm. In our computations, no chemical bonds between the
nanotube and the polymer have been constructed, although there may be possible
covalent cross-links (Frankland et al. 2002).
To further investigate the properties of transition layer, the molecular
dynamics computation has been performed to simulate the nanotube pull-out
process under 300 K with the time-step of 0.1 fs. A total of 1000 time-steps have
been carried out in the computation. As shown in figure 8, we apply for the
tension force at the top of nanotube and pull it out in the direction of length.
The polymer is fixed in the computation. Furthermore, all non-bonded
interactions within the polyethylene matrix, i.e. CH, CH2, CH3 and between
the nanotube and the matrix, i.e. C and H, are modelled with Lennard–Jones
6–12 potentials.
To get the average mechanical property of the transition layer, we consider the
nanotube as a rigid body. The motion of nanotube is controlled by,
F C GT AT gðtÞ Z maðtÞ; (3.1)
where F is the constant applied tension force at the top section of nanotube, GT
the shear modulus of transition layer, AT the side surface area of the transition
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1696 N. Hu and others

layer, i.e. a cylindrical shell in figure 1, g(t)Zu(t)/th the average shear strain
defined by the tangential displacement u(t) in the force direction and thickness of
transition layer th, m the effective mass of a nanotube, and a(t) the acceleration.
A (6,6) armchair nanotube of 492 C atoms possesses the total effective mass of
9.81!10K21 g.
Note that it is very approximate to take the whole side surface area to define
AT since by observing figure 8, only part of the side surface of the nanotube
is directly surrounded by a single polymer chain. Because there may be other
chains near the nanotube simultaneously, the above definition may lead to the
lower shear modulus. Here, by applying the central difference scheme to equation
(3.1), the shear modulus GT can be predicted using the displacements of
nanotube at different time points obtained by molecular dynamics pull-out
simulation. For the total external force of 1.0 nN molK1, figure 9 demonstrates
the variation of shear modulus versus the shear strain g, which shows that the
shear modulus is shear strain dependent. In the initial stage, the shear modulus
decreases significantly in a strong nonlinear pattern as the shear strain increases.
This is not peculiar since the Lennard–Jones 6–12 potential is a strong nonlinear
form of
X K a K b 
EN Z K 6 :
r 12 r
This potential has an attractive tail at large r, but it is strongly repulsive at
small r. The term w1/r12, dominating at short distance, models the repulsion
between atoms to avoid the full overlapping of them when they are brought very
close to each other. Its physical origin is related to the Pauli principle: when
the electronic clouds surrounding the atoms start to overlap, the energy of the
system increases abruptly. The term w1/r6, dominating at large distance,
constitutes the attractive part, originated by van der Waals dispersion forces.
In figure 9, after gZ0.0025 at 10 fs, where the shear modulus is around
50 GPa, the result tends to be stable. The shear modulus converges to around
4.2 GPa at the maximum shear strain of 0.018 at 100 fs. Since the polymer is
fixed in computation, the shear stress in the transition layer is almost stable no
matter what level of strain is reached. It is around 79.4 MPa, which falls into a
reasonable range, e.g. molecular mechanics simulations predict maximum
frictional stress from 18 to 135 MPa for sliding (10, 10) nanotubes within
various types of single polymer chains (Lordi & Yao 2000). It should be noted
that all polymer systems may act differently. It has been shown (Barber et al.
2003) that polymers that should not chemically bond with a nanotube do form
strong interfaces with the nanotube. Therefore, the different choice of polymer
may lead to the different shear modulus of transition layer.

4. Prediction of material properties of composites

After determining the properties of nanotube and interaction between the

nanotube and polymer matrix, the material properties of the nanotube reinforced
composites can be predicted. Our numerical simulation is concentrated on the
case of polymer matrix reinforced with many unidirectional short nanotubes,
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1697

Figure 9. Variation of shear modulus of transition layer versus shear strain.

which may be the ultimate goal for reinforcement. The (6, 6) armchair single wall
nanotubes of length 4.9225 nm are considered. In this case, the RVE of our
FEM model is schematically shown in figure 1. In this model, the Young’s
modulus of transition layer is considered to vary from 3.0 to 400.0 GPa, and the
Poisson’s ratio of the transition layer is set to be 0.33. Taking the varying
Young’s modulus of transition layer from the following consideration: when we
determine the shear modulus of transition layer above, only the non-bonding
potentials are considered. However, there may be some possible chemical bonds
between the nanotube and polymer chains (Frankland et al. 2002). If so, the
bonds should be stronger. Also, in figure 8, only one polymer chain is considered.
In practical cases, there may be other chains, which turn around the nanotube to
enhance the shear modulus. For the PS polymer matrix, its Poisson ratio nm is
chosen as 0.33. The Young’s modulus of PS polymer Em is quite scattered,
depending on the various structures. Here, we take it as 3.0 GPa in our
computation.
For the model in figure 1, the macroscopic composites can be thought of as
transversely isotropic materials with four material constants, i.e. Ez, Ex (ZEy),
nxy (Znyx) and nzx (Znzy). The stress and strain relationship can be expressed as

12 nxy nzx 3
8 9 6 E K
Ex
K
Ez 78 s x 9
< 3x = 6 nx 1 nzx 7< =
6 xy 7
3 Z 6K K 7 sy : (4.1)
: y ; 6 Ex Ex Ez 7: ;
3z 4 n n 1 5 sz
K zx K zx
Ez Ez Ez

For the above four material constants, three kinds of loading cases shown in
figure 10 should be applied. Under the tension load, we can have
 
DRa
L Rm
Ez Z save ; nzx Z K  DL  : (4.2)
DL L

Proc. R. Soc. A (2005)

 Downloaded from rspa.royalsocietypublishing.org

1698 N. Hu and others

By applying the uniform pressure and torsion loads, one can get (Liu &
Chen 2003),
4pR3m Ez Gxy
E x Z Ey Z ; (4.3aÞ
pR3m Ez C 2Ez Gxy R2m DRb C 4pR3m n2zx Gxy

2pR3m Ez
nxy Z K1 C ; (4.3bÞ
pR3m Ez C 2Ez Gxy R2m DRb C 4pR3m n2zx Gxy

TL
Gxy Z : (4.3cÞ
aJ
At first, for the models in figure 10, we set LmZ5 nm, in this case, it is implied
that the short carbon nanotubes are distributed evenly in both longitudinal and
lateral directions in a matrix, so that the RVE is the same when considering any
nanotube. The material constants of transition layer are first assumed to be the
same as the polymer matrix, i.e. ETZ3 GPa and nTZ0.33. Under the axial load,
some forces along the radial direction are imposed on the nodes of lateral surface
and adjusted through an iterative procedure so that all points on the lateral
surface move the same amount in the radial direction to simulate the periodicity
conditions. The various moduli normalized by those of polymer versus volume
fraction of nanotube are illustrated in figure 11. From it, we can see that along
the nanotube direction, the Young’s modulus Ez increases significantly. When
the volume fraction varies from 0.4 to 5.0%, the increase in Ez ranges from 10 to
40%. However, in the transverse plane, only for high volume fractions, Ex and
Gxy are larger than the Young’s modulus of polymer, which shows the effective
reinforcement of polymer. The normalized Poisson’s ratios are plotted in
figure 12. From this, it can be seen that nzx decreases as the volume fraction
increases, which is lower than the Poisson’s ratio of polymer. However, nxy varies
insignificantly, which is higher than the Poisson’s ratio of polymer.
To investigate the influence of stiffness of transition layer parametrically, we
adjust its Young’s modulus from 3 to 400 GPa. The results of normalized Ez and
Gxy are shown in figure 13. Observation of this figure reveals that the influence of
stiffness of transition layer is remarkable when ET ranges from 3 to 100 GPa.
However, when ET is higher than 100 GPa, its influence is quite limited.
The reason is that for the simple tension case, only two ends of transition layer
play the main role of the load transferring capacity. The middle portion of
transition layer does not work so much. To uncover the internal stress
distribution, a section of RVE as shown in figure 14 is employed.
The distributions of stresses szz and tzx on this section of transition layer and
matrix are shown in figures 15a,b under the simple tension load. From these
figures, it can be found that there are high stress concentrations at two ends of
transition layer. In the middle portion of transition layer and matrix, the stress
level is quite low. The high shear stresses are transferred through the transition
layer to the nanotube at two ends, and then produce the axial force in the
nanotube beam structure, which works as stiffeners connected to the inner side of
transition layer. The axial force of the nanotube beam structure is calculated and
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1699

Figure 10. Three loading cases for a cylindrical RVE. (a) Under axial strength; (b) under lateral
uniform pressure; (c) under torsional load.

its distribution along the nanotube is shown in figure 15c. It can be observed that
the axial force is very low at two ends, but increases rapidly and reaches the
highest at the middle point of the nanotube. This result is similar to that of
an embedded fibre in matrix predicted by the Cox shear lag theory. Also, the
distribution of stress sxx of transition layer and matrix under the lateral uniform
pressure is shown in figure 16. This figure demonstrates that the stress
is uniformly distributed in the length direction under this loading condition.
The geometry of RVE has influences on the results too. Generally, like the
fibre-reinforced composites, with the increase of nanotube length, the reinforce-
ment effect along the direction of nanotube will be more significant (Liu & Chen
2003). To investigate this effect, we change the length of polymer, i.e. Lm in
figure 10. Obviously, when Lm/0, it is the limit case of long nanotube reinforced
composite. The Young’s modulus of transition layer is fixed to be 50 GPa here.
The influence of Lm on Ez and nzx are shown in figure 17. Figure 17a reveals that
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1700 N. Hu and others

Figure 11. Various normalized Young’s and shear moduli.

Figure 12. Various normalized Poisson’s ratios.

with the decrease of Lm, Ez increases remarkably. When Lm is very low, the
increase tendency of Ez becomes more obvious. For instance, when LmZ1 nm,
the increase of Ez ranges from 37 to 75%, and for volume fraction, from 0.48 to
2.75%. Till now, there have been some experimental results for the polymer
matrix reinforced with randomly oriented nanotubes. For instance, Qian et al.
(2000) have found that with only the addition of 1.0% multi-wall nanotubes
by weight (0.487% by volume fraction), the elastic stiffness achieved between
a 36 and a 42% increase. Also, the experimental results have been reported by
Schadler et al. (1998) for multi-wall nanotubes in epoxy, Andrews et al. (2002)
for multi-wall nanotubes in propylene, and Santare et al. (2003) for multi-wall
nanotubes in high-density polyethylene. These published experimental results
are also illustrated in figure 17a, which shows that our results are consistent with
these experimental ones. Some experimental results are lower than our results.
This may be caused by: (i) the case of randomly aligned reinforcement in
experiments is used, but the unidirectionally reinforced case is considered in our
numerical simulations; (ii) the defects in nanotubes walls of practical composites
may result in the lower results. Furthermore, compared with a single-wall
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1701

Figure 13. Influence of stiffness of transition layer on Ez/Em and Gxy/Gm.

y x

transition
layer F

Figure 14. Schematic of section of RVE for illustration of stresses on transition layer and matrix,
and axial force on nanotube.

nanotube in numerical simulation, as reported by Lourie & Wanger (1998), the

multi-wall nanotube possesses the lower Young’s modulus. Also, for multi-wall
nanotube composites, Qian et al. (2000) have shown that multi-wall nanotubes
fails via a sword-and-sheath mechanism, which implies that, even if load is
transferred to the outmost nanotube in a multi-wall nanotube, the inner tubes
slide within the concentric cylinder and do not carry much load. All these facts
may imply that the effectiveness of Young’s modulus of multi-wall nanotube
composites may not be higher than that of single-wall nanotube composites in
our simulations, although for single-wall nanotube composites, the single-wall
nanotubes are usually in the form of a bundle, and individual tubes may slip
within the bundle.
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1702 N. Hu and others

Figure 15. Various normalized stress distributions on transition layer and matrix, and normalized
axial force distribution on nanotube for simple tension. (a) szz of transition layer and matrix; (b) tzx
of transition layer and matrix; (c) normalized axial force distribution on nanotube.

When considering the tube as a solid shell body, from the classical elastic
theory and rule of mixture, for the case of nanotube through the length of the
RVE (LmZ0 in figure 10), there is

Ez Z EVn C Em ð1 K Vn Þ; (4.4)

where E is the Young’s modulus of carbon nanotube, i.e. 1.06 TPa, and Vn is the
volume fraction of the nanotube. For the case of the nanotube embedded inside
the RVE in figure 10, with the assumption of perfect bonding between the
nanotube and matrix, Liu & Chen (2003) gave out the expression of the effective
Young’s modulus as follows:
    
K1
1 Lm 1 Ln A
Ez Z C ; (4.5)
Em L E L An

where A and An are defined as: AZ pR2m and An Z pðR2m K R2n Þ.

Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1703

Figure 16. Distribution of normalized sxx on transition layer and matrix for lateral uniform
pressure.

For the (6,6) armchair single wall nanotube, its diameter is 0.8142 nm, and the
thickness of shell is 0.34 nm. Figure 17a shows the effective Young’s moduli
predicted by equations (4.4) and (4.5), which are much higher than those of the
present model and experimental results.
Figure 17b shows that nzx increases obviously as Lm decreases. The reason for
this is that to keep the same volume fraction, the thickness of polymer layer
increases when Lm decreases. Therefore, the higher Poisson’s ratio of polymer
will dominate nzx of composites.

5. Conclusions

The elastic properties of carbon nanotube-reinforced composites are predicted in

this paper through analysing the elastic deformation of a RVE under various
loading cases. To construct this RVE, consisting of the carbon nanotube, the
transition layer between the nanotube and the polymer matrix and the outer
polymer matrix, first, an equivalent beam model for carbon nanotube is built up.
The explicit relationships between the material properties of the equivalent beam
element and the force constants have been derived. Second, to describe the
interaction between the nanotube and the outer polymer matrix at the level of
atoms, the molecular mechanics and molecular dynamics computations have
been performed to obtain the thickness and material properties of the transition
layer. Moreover, an efficient 3D eight-noded brick finite element is employed to
model the transition layer and the outer polymer matrix. Consequently, the
macroscopic behaviours of the RVE can be evaluated using the traditional FEM.
From the predicted results, the following conclusions can be made.

(i) When using Lennard–Jones type potentials, the shear modulus of transition
layer strongly depends on the deformation in a form of nonlinearity.
This shear modulus decreases as the relative shear deformation increases.
The stable value ranges from 4.2 to 50 GPa.
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1704 N. Hu and others

Figure 17. Influence of RVE geometry on various material constants. (a) Ez/Em; (b) nzx/nm

(ii) When the Young’s modulus of transition layer ranges from 3 to 100 GPa, a
stronger transition layer can result in significantly higher Ez. However, when ET
is higher than 100 GPa, the effect of stiffness of transition layer becomes weak.
(iii) Increase in nanotube length leads to higher reinforcement in the nanotube length
direction. When the volume fraction ranges from 0.48 to 2.75%, depending
on the stiffness of transition layer and nanotube length, the reinforcement along
the nanotube length direction can be achieved from 10 to 70%.

Appendix A

Consider a plane hexagonal unit cell in a graphite sheet, an arbitrary

deformation state (3x, 3y, gxy) is assumed as shown in figure 2. For this in-plane
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1705

deformation, the stretching lengths of bonds 1, 2 and 3 are calculated as follows

by neglecting the higher order terms:
pffiffiffi
3 1 3
DR1 Z R3x C R3y K Rgxy ; (A 1aÞ
4 4 4

DR2 Z R3y ; (A 1bÞ

pffiffiffi
3 1 3
DR3 Z R3x C R3y C Rgxy : (A 1cÞ
4 4 4
The variation of angles q1, q2 and q3 can be evaluated as follows:
pffiffiffi pffiffiffi
3 3 3
Dq1 Z K 3x C 3y K gxy ; (A 2aÞ
4 4 4
pffiffiffi pffiffiffi
3 3 3
Dq2 Z K 3 C 3 C g ; (A 2bÞ
4 x 4 y 4 xy
pffiffiffi pffiffiffi
3 3
Dq3 Z 3x K 3 : (A 2cÞ
2 2 y
Finally, by neglecting the torsion potential, van der Waals potential and the
electrostatic interactions, the total potential of the unit chemical hexagonal cell
can be cast into
1
Epot Z ð2Ks DR21 C 2Ks DR22 C 2Ks DR23 C 2Kq Dq21 C 2Kq Dq22 C 2Kq Dq23 Þ: (A 3)
2
Substitution of equations (A 1a–c) and (A 2a–c) into equation (A 3) leads to
   
9 2 9 2 3 2 9
Epot Z K R C Kq 3x C K R K Kq 3x 3y
8 s 8 4 s 4
   
9 2 9 2 3 9
C K R C Kq 3 y C K R C Kq g2xy :
2
(A 4)
8 s 8 8 s 8
On the other hand, by considering this unit cell as a continuum homogeneous
material, the strain energy stored in it under the assumed deformation can be
estimated as

pffiffiffi
E 2 2 2 3 3 2
UZ 2
ð3x C 3y C 2n3x 3y Þ C Gðgxy Þ R ts ; (A 5)
1 Kn 2
where E is the Young’s modulus, n the Poisson’s ratio, and ts the thickness
of sheet.
By equalling Epot to U and comparing the corresponding terms of both sides,
we can get the following three equations:
pffiffiffi
3 3 E 9 9
2
R2 ts Z Ks R2 C Kq ; (A 6aÞ
2 1 Kn 8 8
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1706 N. Hu and others

pffiffiffi En 3 9
3 3 2
R2 t s Z K s R2 K K q ; (A 6bÞ
1 Kn 4 4
pffiffiffi
3 3 3 9
GR2 ts Z Ks R2 C Kq : (A 6cÞ
2 8 8
Finally, the relationships between the elastic properties of the unit graphite
cell and the force constants can be set-up as follows:
3
Ks R2 K 98 Kq
n Z 89 2 9
; (A 7aÞ
8 Ks R C 8 Kq

3
Ks R2 C 94 Kq
GZ4 pffiffiffi ; (A 7bÞ
3 3R2 ts

9
Ks R2 C 94 Kq
EZ4 pffiffiffi 2 ð1 K n2 Þ: (A 7cÞ
3 3R ts
With the above solutions, three equations (A 6a–c) can be satisfied
simultaneously. The relationship of GZE/2(1Cn) can be maintained also.

Appendix B

Without consideration of torsion deformation, the strain energy of a

Timoshinko’s beam element of length R can be expressed as
ðR ðR ðR
1 2 1 2 1
GZ EA3 dx C EI k dx C GAs g2 dx; (B 1)
0 2 0 2 0 2

where R, 3, k and g are the length, stretching strain, curvature and transverse
shear strain of the beam, respectively. For a two-noded beam element, they can
be expressed as
DR q K q2 ðw K w1 Þ 1
3Z ; kZ 1 ; gZ 2 K ðq1 C q2 Þ: (B 2)
R R R 2
To have the correct dimensions, by comparing the strain energy in equation
(B 1), and potential of atoms model in equation (A 3), the following relationships
can be set-up:
EA Z x1 Ks R; EI Z x2 Kq R; GAs Z x3 Kq =R; (B 3)
where x1, x2 and x3 are some non-dimensional parameters that need to be
identified.
To get the relationships between the material parameters of beam and force
constants of molecular dynamics, we consider a simple compressive deformation
state of a graphite sheet under unit distributed compressive loads, applied along
the x-axis shown in figure 4. To simplify the derivation, the influence of shear
deformation is neglected in equation (B 1). Under this simple deformation state,
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1707

according to the symmetrical conditions, there is no rotation at each node. Also,

there is no deformation and stress in the vertical beams along the y-axis direction
since the graphite sheet can freely stretch or contract in this direction. For one
representative unit cell, after removing the rigid body motion, the deformation
configuration is displayed in figure 4. For a typical beam element, i.e. beam 1 in
this unit cell, under unit compressive loads, the concentrated loads and moments
at two nodes are illustrated. From the analysis of structural mechanics, the
displacements at two nodes of beam 1 can be obtained as follows
pffiffiffi
ðEAR2 C 36EI ÞR 3ðEAR2 K 12EI ÞR
uZ ; v ZK : (B 4)
48EAEI 48EAEI

To consider many repeated cells in the y-axis direction, the stress under the
distributed unit compressive loads in this graphite sheet can be evaluated as

2n
sx Z : (B 5)
ts ð3nR K RÞ

Also, the strain in the x-axis direction for a half unit cell can be expressed as
pffiffiffi
u 3ðEAR2 C 36EI Þ
3x Z pffiffi Z : (B 6)
3 72EAEI
2 R

Then, the equivalent Young’s modulus of the graphite sheet can be evaluated
as EZsx/3x, and finally we can get
pffiffiffi
48 3nEAEI
EZ : (B 7)
ts ð3nR K RÞðEAR2 C 36EI Þ

The Poisson’s ratio is evaluated as nZ3y/3x, where

pffiffiffi
2nðKvÞ 3nðEAR2 K 12EI Þ
3y Z Z : (B 8)
ð3nR K RÞ 24EAEI ð3n K 1Þ

Finally, we can get

3nðEAR2 K 12EI Þ
nZ : (B 9)
ð3n K 1ÞðEAR2 C 36EI Þ

From the Young’s modulus and Poisson’s ratio, the shear modulus can
be obtained as
pffiffiffi
24 3nEAEI
GZ : (B 10)
ts Rð6nEAR2 C 72nEI K EAR2 K 36EI Þ
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

1708 N. Hu and others

When n is sufficiently large, e.g. n/N, there are the following simple
relationships:
ðEAR2 K 12EI Þ
nZ ; (B 11aÞ
ðEAR2 C 36EI Þ
pffiffiffi
4 3EAEI
GZ ; (B 11bÞ
ts RðEAR2 C 12EI Þ
pffiffiffi
16 3EAEI
EZ : (B 11cÞ
ts RðEAR2 C 36EI Þ
The above relationships connect the macroscopic material properties of
graphite sheet to the material properties of local beam elements constituting this
sheet. Then, by comparing the Poisson’s ratio and shear modulus listed in
equations (A 7a,b), (B 11a,b), we can get the following two equations:
3
Ks R2 K 98 Kq ðEAR2 K 12EI Þ
n Z 89 9
Z ; (B 12aÞ
2
8 Ks R C 8 Kq
ðEAR2 C 36EI Þ
pffiffiffi
Ks R2 C 94 Kq
3
4 4 3EAEI
GZ pffiffiffi Z : (B 12bÞ
3 3R2 ts ts RðEAR2 C 12EI Þ
By employing the relationships, i.e. EAZx1KsR, EIZx2KqR in equation (B 3),
from equations (B 12a,b), we can finally identify that
Ks R2 ðKs R2 C 3Kq Þ
x1 Z 1 and x2 Z : (B 13)
36Kq ðKs R2 K Kq Þ
For the force constants suggested previously, i.e. KsZ805.5 nN nmK1,
KqZ1.438 nN nm radK2, we can obtain x1Z1, x2Z0.436, respectively. Therefore,
there are
EA Z Ks R; EI Z 0:436Kq R: (B 14)
In the above derivation, there is no torsion deformation. Therefore, the torsion
rigidity should be determined independently from any other aspect. From the
expression of Cornell et al. (1995), the potential owing to dihedrals torsion can be
expressed as
X 1
Vt Z V ð1 C cosðjni jui K FÞÞ: (B 15)
all torsions
2 ui
where Vui is the torsional barrier (kcal molK1), jnij the periodicity and ui is the
torsion angle. For the C–C bonds, F is 1808 and ni is 2, respectively.
By considering the infinite small torsion deformation, uiZFCDu, we can
expand cos(ui) into Taylor series, and get the potential for one torsional bond as
Vt Z Vu Du2 : (B 16)
Proc. R. Soc. A (2005)
 Downloaded from rspa.royalsocietypublishing.org

Evaluating nanotube reinforced composites 1709

On the other hand, for a beam element under pure torsion, the strain energy is
ðR
1
GZ GJ Du2 dx: (B 17)
0 2
By comparing equation (B 16) with equation (B 17), we can obtain
2Vu
GJ Z : (B 18)
R
By taking the value provided by Cornell et al. (1995), VuZ14.5 kcal molK1,
we can finally arrive at GJZ0.201 618 1 nN nm RK1.

References
Andrews, R., Jacques, D., Minot, M. & Rantell, T. 2002 Fabrication of carbon multiwall
nanotube/polymer composites by shear mixing. Macromol. Mater. Eng. 287, 395–403.
Barber, A. H., Cohen, S. R. & Wagner, H. D. 2003 Measurement of carbon nanotube–polymer
interfacial strength. Appl. Phys. Lett. 82, 4140–4142.
Brenner, D. W. 1990 Empirical potential for hydrocarbons for use in simulating the chemical
vapor-deposition of diamond films. Phys. Rev. B 42, 9458–9471.
Brenner, D. W., Shenderova, O. A., Harrison, J. A., Stuart, S. J., Ni, B. & Sinnott, S. B. 2002
A second-generation reactive empirical bond order (REBO) potential energy expression for
hydrocarbons. J. Phys.: Condens. Matter 14, 783–802.
Cao, Y. P., Hu, N., Lu, J., Fukunaga, H. & Yao, Z. H. 2002 A 3D brick element based
on Hu–Washizu variational principle for mesh distortion. Int. J. Numer. Methods Eng.
53, 2529–2548.
Chang, T. C. & Gao, H. J. 2003 Size-dependent elastic properties of a single-walled carbon
nanotube via a molecular mechanics model. J. Mech. Phys. Solids 51, 1059–1074.
Chang, B. H. et al. 2000 Conductivity and magnetic susceptibility of nanotube/polypyrrole
nanocomposites. J. Low Temp. Phys. 119, 41–48.
Chopra, N. G. & Zettl, A. 1998 Measurement of the elastic modulus of a multi-wall boron nitride
nanotube. Solid State Commun. 105, 297–300.
Cooper, C. A., Cohen, S. R., Barber, A. H. & Wagner, H. D. 2002 Detachment of nanotubes from
a polymer matrix. Appl. Phys. Lett. 81, 3873–3875.
Cornell, W. D. et al. 1995 A second generation force field for the simulation of proteins, nucleic
acids, and organic molecules. J. Am. Chem. Soc. 117, 5179–5197.
Frankland, S. J. V., Caglar, A., Brenner, D. W. & Griebel, M. 2002 Molecular simulation of the
influence of chemical cross-links on the shear strength of carbon nanotube–polymer interfaces.
J. Phys. Chem. B 106, 3046–3048.
Hernandez, H., Goze, C., Bernier, P. & Rubio, A. 1998 Elastic properties of C and BxCyNz
composite nanotubes. Phys. Rev. Lett. 80, 4502–4505.
Jia, Z. J., Wang, Z. Y., Xu, C. L., Liang, J., Wei, B. Q., Wu, D. H. & Zhang, Z. M. 1999 Study on
poly (methyl methacrylate)/carbon nanotube composites. Mater. Sci. Eng. A 271, 395–400.
Jin, Y. & Yuan, F. G. 2002 Elastic properties of single-walled carbon nanotubes. In 43rd
AIA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference,
Denver, Colorado, April 22–25, 2002, article AIAA-2002-1430. Reston, VA: American Institute
of Aeronautics and Astronautics.
Krishnan, A., Dujardin, E., Ebbesen, T. W., Yianilos, P. N. & Treacy, M. M. J. 1998 Young’s
modulus of single-walled nanotubes. Phys. Rev. B 58, 14 013–14 019.

Proc. R. Soc. A (2005)

 Downloaded from rspa.royalsocietypublishing.org

1710 N. Hu and others

Liao, K. & Li, S. 2001 Interfacial characteristics of a carbon nanotube–polystyrene composite

system. Appl. Phys. Lett. 79, 4225–4227.
Liu, Y. J. & Chen, X. L. 2003 Evaluations of the effective material properties of carbon nanotube-
based composites using a nanoscale representative volume element. Mech. Mater. 35, 69–81.
Lordi, V. & Yao, N. 2000 Molecular mechanics of binding in carbon-nanotube–polymer composites.
J. Mater. Res. 15, 2770–2779.
Lourie, O. & Wanger, H. D. 1998 Evaluation of Young’s modulus of carbon nanotubes by micro-
Raman spectroscopy. J. Mater. Res. 13, 2418–2422.
Lu, J. P. 1997 The elastic properties of single and multilayered carbon nanotubes. J. Phys. Chem.
Solids 58, 1649–1652.
McCarthy, B., Coleman, J. N., Curran, S. A., Dalton, A. B., Davey, A. P., Konya, Z., Fonseca, A.,
Nagy, J. B. & Blau, W. J. 2000 Observation of site selective binding in a polymer nanotube
composite. J. Mater. Sci. Lett. 19, 2239–2241.
Odegard, G. M., Gates, T. S., Nicholson, L. M. & Wise, K. E. 2002 Equivalent-continuum
modeling of nano-structured materials. Compos. Sci. Tech. 62, 1869–1880.
Popov, V. N., Van Doran, V. E. & Balkanski, M. 2000 Elastic properties of single-walled carbon
nanotubes. Phys. Rev. B 61, 3078–3084.
Qian, D., Dickey, E. C., Andews, R. & Rantell, T. 2000 Load transfer and deformation mechanisms
in carbon nanotube–polystyrene composites. Appl. Phys. Lett. 76, 2868–2870.
Qian, D., Wagner, G. J., Liu, W. K., Yu, M. F. & Rouff, R. S. 2002 Mechanics of carbon
nanotubes. Appl. Mech. Rev. 22, 495–533.
Salvetat, J. P., Bonard, J. M., Thomson, N. H., Kulik, A. J., Forro, L., Benoit, W. & Zuppiroli, L.
1999 Mechanical properties of carbon nanotubes. Appl. Phys. A 69, 255–260.
Santare, M. H., Tang, W., Novotny, J. E. & Advani, S. G. 2003 Mechanical characterization of
a nanotube–polyethylene composite material. In Proceedings of IMECE’03, paper 43351.
Washington, DC: ASME.
Schadler, L. S., Giannaris, S. C. & Ajayan, P. M. 1998 Load transfer in carbon nanotube epoxy
composites. Appl. Phys. Lett. 73, 3842–3844.
Tersoff, J. 1988 Empirical interatomic potential for carbon, with applications to amorphous-
carbon. Phys. Rev. Lett. 61, 2872–2879.
Tostenson, E. T., Ren, Z. F. & Chou, T. W. 2001 Advances in the science and technology of carbon
nanotubes and their composites: a review. Compos. Sci. Tech. 61, 1899–1912.
Tostenson, E. T., Li, C. & Chou, T. W. 2002 Carbon nanotube-reinforced composites: processing,
modeling and property characterization. In Proceedings of the tenth US–Japan conference
on composite materials, Stanford, 16–18 September 2002, pp. 641–647.
Wagner, H. D., Lourie, O., Feldman, Y. & Tenne, R. 1998 Stress-induced fragmentation
of multiwall carbon nanotubes in a polymer matrix. Appl. Phys. Lett. 72, 188–190.
Walther, J. H., Jaffe, R., Halicioglu, T. & Koumoutsakos, P. K. 2001 Carbon nanotubes in water:
structural characteristics and energies. J. Phys. Chem. B 105, 9980–9987.
Wong, E. W., Sheehan, P. E. & Lieber, C. M. 1997 Nanobeam mechanics: elasticity, strength and
toughness of nanorods and nanotubes. Science 277, 1971–1975.

Proc. R. Soc. A (2005)