Computational Materials Science 29 (2004) 1–11

www.elsevier.com/locate/commatsci

Square representative volume elements for evaluating

the effective material properties of carbon
nanotube-based composites
X.L. Chen, Y.J. Liu *

Computer-Aided Engineering Research Laboratory, Department of Mechanical Engineering, University of Cincinnati, P.O. Box 210072,
Cincinnati, OH 45221-0072, USA

Abstract

Carbon nanotubes (CNTs) demonstrate unusually high stiffness, strength and resilience, and may become an ideal
reinforcing material for new nanocomposites. However, much work has to be done before the potentials of CNT-based
composites can be fully realized. Evaluating the effective material properties of such nanoscale materials is one of many
difficult tasks. Simulations using molecular dynamics and continuum mechanics models can play significant roles in this
development. Currently, the continuum approach seems to be the only feasible approach for such large scale analysis.
In this paper, effective mechanical properties of CNT-based composites are evaluated using a square representative
volume element (RVE) based on the continuum mechanics and with the finite element method (FEM). Formulas to
extract the effective material constants from solutions for the square RVEs under two load cases are derived based on
the elasticity theory. Numerical results using the FEM show that the load carrying capacities of the CNTs in a matrix
are significant. For example, with the addition of CNTs in a matrix at a volume fraction of 3.6%, the stiffness of the
composite can increase as much 33% in the axial direction with long CNTs. These simulation results are consistent with
the experimental results reported in the literature and the earlier results using cylindrical RVEs.

2003 Elsevier B.V. All rights reserved.

Keywords: Carbon nanotubes; Nanocomposites; Effective properties; Finite element method

1. Introduction reinforcing materials for the development of a new

class of nanocomposites (see, e.g., recent compre-
Carbon nanotubes (CNTs) possess exception- hensive reviews [1,2]). It has been demonstrated
ally high stiffness, strength and resilience, as well that with only 1% (by weight) of CNTs added in a
as superior electrical and thermal properties. matrix material, the stiffness of a resulting com-
Many believe that CNTs may provide the ultimate posite film can increase between 36% and 42% and
the tensile strength by 25% [3]. The mechanical-
load carrying capacities of CNTs in nanocom-
*
Corresponding author. Tel.: +1-513-556-4607; fax: +1-513- posites have also been demonstrated in some
556-3390. experimental work [3–6] and numerical simula-
E-mail address: [email protected] (Y.J. Liu). tions [7,8]. All these investigations show that the

0927-0256/$ - see front matter
2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0927-0256(03)00090-9
2 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11

load-carrying capacities of CNTs in a matrix is YoungÕs modulus and PoissonÕs ratios, and buck-
significant and the CNT-based composites have ling modes of CNTs have been successfully pre-
the potential to provide extremely strong and dicted by using these continuum approaches.
ultralight new materials. However, cautions should be excised in applying the
However, much work still need to be done be- continuum approach, as discussed in Ref. [7]. Em-
fore the potentials of the CNT-based composites phasis should be placed on the overall responses of
can be fully realized in real engineering applica- CNTs or CNT-based composites, rather than on
tions. Evaluating the effective material properties the local detailed phenomena, such as interfacial
of such nanoscale materials is one of the chal- stresses or debonding, where the nanoscale MD
lenging tasks for the development of nanocom- approach should be employed. Evaluating the ef-
posites. Computational approaches, based on the fective materials properties of the CNT-reinforced
molecular dynamics (MD) approach (for smaller composites deals with the overall mechanical re-
scales) and continuum mechanics approach (for sponses of the RVEs, as presented in Ref. [8] and
larger scales), can play significant roles in the areas this paper. Thus, the continuum approach seems
of characterizing CNT-based composites. adequate for this study, although further devel-
The MD approach has provided abundant simu- opment and validation are needed.
lation results for understanding the behaviors of The modeling considerations in characterizing
individual and bundled CNTs [9–17]. However, MD CNT-based composites using the continuum ap-
simulations of CNTs are currently limited to very proach are discussed in Ref. [7]. It is proposed that
small length and time scales and cannot deal with the 3-D elasticity models, instead of beam or shell
the larger length scales in studying nanocomposites. models, should be employed for modeling the
Nanocomposites for engineering applications must CNTs embedded in a matrix, in order to ensure the
expand from nano to micro, and eventually to accuracy and compatibility between the models for
macrolength scales. Therefore, continuum mechan- the CNTs and matrix. A method based on the
ics models can be applied initially for simulating the elasticity theory for evaluating effective material
mechanical responses of the CNTs in a matrix, as properties of CNT-based composites using the
has been done in [7,8] for studying the overall re- representative volume elements (RVEs) (Fig. 1) is
sponses of CNT composites, before efficient large established and cylindrical RVEs (Fig. 1(a)) are
multiscale models are established. investigated in Ref. [8]. Formulas to extract the
The continuum mechanics approach has been effective material properties from numerical solu-
employed for quite some time in the study of in- tions for the cylindrical RVEs under three loading
dividual CNTs or CNT bundles to investigate their cases are derived. Analytical results (extended rule
mechanical properties. The validity of the contin- of mixtures) based on the strength of materials
uum approach to modeling of CNTs is still not theory to estimate the effective YoungÕs modulus
fully established and the practice will continue to in the axial direction, which can help validate the
be questioned for some time to come. However, it numerical solutions, are also derived for both long
seems to be the only feasible approach at present and short CNT cases in [8]. Numerical results
to obtain preliminary results for characterizing using the finite element method (FEM) for the
CNT-based composites using modeling and simu- cylindrical RVEs show significant increases of the
lations. The best argument for using this contin- stiffness in the CNT direction of the nanocom-
uum approach for now is simply the fact that it has posites under various combinations of the CNT
been applied successfully for studying single or and matrix material properties [8]. However, al-
bundled CNTs, as given in Refs. [18–25]. In these though cylindrical RVEs are easy to use, for which
studies, the CNTs are considered as homogeneous analytical solutions can be derived and efficient
and isotropic materials using continuum beam, 2-D axisymmetric FEM models can be applied,
shell, as well as 3-D solid models in the analyses of they are the most primitive models and can lead to
the deformation, buckling and dynamics responses errors due to ignoring materials not covered by the
of CNTs. Material properties such as equivalent cylindrical cells.
 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11 3

(a) (b) (c)

Fig. 1. Three possible RVEs for the analysis of CNT-based nanocomposites [7]. (a) Cylindrical RVE; (b) square RVE; and (c) hex-
agonal RVE.

In this paper, the work initiated in [8] is extended

to square RVEs (Fig. 1(b)) for the evaluations of
effective material properties of the CNT-based
composites. New formulas based on the elasticity
theory for extracting the effective material proper-
ties from solutions of the square RVEs are derived
and numerical studies using the FEM are con-
ducted. Square 2-D RVE models containing mul-
tiple CNTs are also investigated for evaluating
the effective material constants in the transverse
directions. The numerical results from the square
RVEs are compared with those obtained ear-
lier using the cylindrical RVEs in [8]. It is found
that the cylindrical RVEs tend to overestimate
the effective YoungÕs moduli of the CNT-based Fig. 2. A square RVE containing a short CNT shown in a cut-
composites, and the square RVEs may be the through view.
preferred models for obtaining more accurate re-
sults. Fig. 2 for the orientation of the coordinates). The
fifth independent material constant––the shear
modulus Gxz ð¼ Gyz Þ can be obtained using a
simple (torsion) load case and will not be consid-
2. Formulas for extracting the effective material
ered in this paper. The general 3-D strain–stress
constants
relation relating the normal stresses (rx ; ry ; rz ) and
strains (ex ; ey ; ez ) for a transversely isotropic mate-
To derive the formulas for extracting the
rial can be written as (see, e.g., [26]):
equivalent material constants, a homogenized
8 9 2 1 m
38 9
elasticity model of the square RVE (Fig. 2) is < ex = Ex
 Exyx  mEzxz < rx =
considered. The geometry of the elasticity model is 6 m 1
 mEzxz 7
e ¼ 4  Exyx 5 ry : ð1Þ
corresponding to a solid square RVE with length L : y; m
Ex
m 1 : ;
ez E zx
E zx
E
rz
and cross-sectional area 2a
2a (Fig. 3). Elasticity z z z

solutions can be obtained under certain load cases. To determine the four unknown material con-
The elasticity model is filled with a single, trans- stants (Ex , Ez , mxy and mzx ), four equations will be
versely isotropic material that has five independent needed. Two loading cases (Fig. 3) have been de-
material constants. The four effective material vised to provide four such equations based on the
constants (YoungÕs moduli Ex and Ez , and Pois- elasticity theory, as illustrated below (for com-
sonÕs ratios mxy and mzx , relating the normal stress parison, three load cases are needed for the cylin-
and strain components) will be determined (see drical RVEs [8]).
4 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11

Fig. 3. Two loading cases for the square RVE used to evaluate the effective material properties of the CNT-based composites. (a)
Under axial stretch DL; and (b) under lateral uniform load p.

2.1. Square RVE under an axial stretch DL (Fig. 3 with A being the area of the end surface. The value
(a)) of rave is evaluated for the RVE using the FEM
results.
In this load case (Fig. 3(a)), the stress and strain Using the first (or second) equation in Eq. (1)
components on the lateral surface are: and the result (2), one has along x ¼ a:
mzx DL Da
DL Da ex ¼  rz ¼ mzx ¼ :
rx ¼ ry ¼ 0; ez ¼ ; ex ¼ along Ez L a
L a
Da Thus, one obtains an expression for the PoissonÕs
x ¼ a; and ey ¼ along y ¼ a;
a ratio:



Da . DL
where Da is the change of dimension a of the cross- mzx ¼  : ð3Þ
section under the stretch DL in the z-direction a L
(Da < 0, if DL > 0). Integrating and averaging the Eqs. (2) and (3) can be applied to estimate the
third equation in (1) on the plane z ¼ L=2, one has effective YoungÕs modulus Ez and PoissonÕs ratio
immediately: mzx ð¼ mzy Þ, once the contraction Da and the stress
rave in case (a) are obtained.
rave L
Ez ¼ ¼ rave ; ð2Þ
ez DL 2.2. Square RVE under a lateral uniform load p
where the averaged value of stress rz is given by: (Fig. 3(b))
Z
1 In this load case (Fig. 3(b)), the square RVE is
rave ¼ rz ðx; y; L=2Þ dx dy;
A A loaded with a uniformly distributed load (negative
 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11 5

pressure) p in a lateral direction, for example, the 3. Rules of mixtures based on the strength of
y-direction. The RVE is constrained in the z- materials theory
direction so that the plane strain condition is main-
tained, in order to simulate the interactions of the Similar to the cylindrical RVE cases [8], simple
RVE with surrounding materials in the z-direction. rules of mixtures can be established based on the
Since ez ¼ 0, rz ¼ mzx ðrx þ ry Þ for plane strain strength of materials theory. These rules of mix-
cases, the 3-D stress–strain relation (1) (for normal tures can be applied to verify the numerical results
components) is reduced to: for the effective YoungÕs moduli in the CNT axial
  " 1 m2zx m 2 #  direction. More general theories and extended re-
ex Ex
 Ez  Exyx  mEzxz rx sults, in the context of fiber-reinforced composites,
¼ 2 2 : ð4Þ
ey m
 Exyx  mEzxz 1
 mEzxz ry can be found in Refs. [26,27].
Ex

For the corresponding elasticity model (Fig. 3(b)), 3.1. CNT through the length of the RVE (Fig.
one has the following results for the normal stress 4(a))
and strain components at a point on the lateral
surfaces: This is the case when the CNT is relatively long
Dx (with large aspect ratio) and therefore a segment
rx ¼ 0; ry ¼ p; ex ¼ along
a can be modeled using an RVE. For the square
Dy RVE, the volume fraction [26] of the CNT (a tube,
x ¼ a; and ey ¼ along y ¼ a; Fig. 4(a)) is defined by:
a
where Dx (<0) and Dy (>0) are the changes of di- pðro2  ri2 Þ
mensions in the x- and y-direction, respectively, in Vt ¼ : ð7Þ
4a2  pri2
this load case. Applying the first equation in (4) for
points along x ¼ a and the second equation in (4) Following Ref. [8], the effective YoungÕs modulus
for points along y ¼ a, one has: Ez in the axial direction is found to be:

 Ez ¼ Et V t þ Em ð1  V t Þ; ð8Þ
mxy m2zx Dx
ex ¼  þ p¼ ; and
Ex Ez a where Et is the YoungÕs modulus of the CNT and


1 m2zx Dy Em that of the matrix.
ey ¼  p¼ ;
Ex E z a
3.2. CNT inside the RVE (Fig. 4(b))
respectively. By solving these two equations, one
obtains the effective YoungÕs modulus and Pois- In this case (Fig. 4(b)), the square RVE is di-
sonÕs ratio in the transverse direction (xy plane, vided into two segments: one segment accounting
Fig. 2): for the two ends with a total length of Le and
1 YoungÕs modulus Em ; and another segment ac-
Ex ¼ Ey ¼ Dy 2 ; ð5Þ counting for the center part with a length of Lc and
pa
þ mEzxz
an effective YoungÕs modulus Ec . Note that the two

,
 hemispherical end caps of the CNT have been ig-
Dx m2zx Dy m2zx nored in this simple strength of materials model.
mxy ¼  þ þ ; ð6Þ
pa Ez pa Ez For the center part, which is a special case of the
one shown in Fig. 4(a), its effective YoungÕs
in which Ez and mzx have been determined from modulus is found to be:
Eqs. (2) and (3) from load case (a). Once the
Ec ¼ Et V t þ Em ð1  V t Þ; ð9Þ
changes in dimensions, Dx and Dy, are determined
for the square RVE from, for example, a finite using Eq. (8), in which the volume fraction of the
element analysis, Ex ð¼ Ey Þ and mxy can be com- CNT V t given by Eq. (7) is computed based on the
puted from Eqs. (5) and (6), respectively. center part of the RVE (with length Lc ) only.
6 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11

Fig. 4. Simplified strength of materials models based on the square RVEs for estimating the effective YoungÕs modulus Ez in the CNT
direction. (a) CNT through the length of the RVE; and (b) CNT inside the RVE (L ¼ Le þ Lc ).

Again, following Ref. [8], the effective YoungÕs described in Section 2. The FEM results are then
modulus Ez is found to be: processed, and Eqs. (2), (3), (5) and (6) are applied
1 to extract the effective YoungÕs moduli and Pois-
Ez ¼  
1 Le 1 Lc
  A  ; ð10Þ sonÕs ratios for the CNT-based composite. Two
m þ c
E L E L Ac numerical examples are studied, one on RVEs with
which is derived based on the strength of materials long CNTs and the other on an RVE with a short
theory, in which A ¼ 4a2 , Ac ¼ 4a2  pri2 (Fig. CNT. In all the cases, quadratic solid (brick) ele-
4(b)). This is an extended rule of mixtures com- ments are employed for the 3-D models and qua-
pared to that given in Eq. (8). Eq. (10) can be dratic 8-node elements are used for 2-D plane
employed to estimate the effective YoungÕs mod- strain models, both of which offer higher accuracy
ulus for the case shown in Fig. 4(b) when the CNT in FEM stress analysis.
is relatively short and thus inside the square RVE.
In this paper, Eqs. (8) and (10) will be applied to 4.1. Long CNTs through the length of a square RVE
compare the FEM estimates of the effective
YoungÕs moduli in the axial direction in the fol- First, an RVE for a long CNT all the way
lowing section. through the RVE length, similar to the one shown
in Fig. 4(a), is studied. The dimensions are: for the
matrix, length L ¼ 100 nm, a ¼ 10 nm; for the
4. Numerical examples CNT, length L ¼ 100 nm, outer radius ro ¼ 5 nm,
inner radius ri ¼ 4:6 nm (effective thickness ¼ 0.4
Several square RVE models for single-walled nm, which is close to the theoretical value of 0.34
CNTs in a matrix material are studied using the nm for CNT thickness). The YoungÕs moduli and
FEM in this section, in order to evaluate the ef- PoissonÕs ratios used for the CNT and matrix are:
fective material constants of the CNT-based
nanocomposite. The deformations and stresses are CNT : Et ¼ 1000 nN=nm2 ðGPaÞ; mt ¼ 0:3;
computed first for the two loading cases (Fig. 3) as Matrix : Em ¼ 100 nN=nm2 ðGPaÞ; mm ¼ 0:3:
 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11 7

These values of the dimensions and material con- strength of materials solutions is also listed in
stants are chosen for illustration purposes only, Table 1 for comparison. The strength of materials
which are within the wide ranges of those for solution (Eq. (8)) is identical to that using the
CNTs reported in the literature [12,15,16,18,28– FEM, due to the simple geometry and load con-
34]. These parameters can be modified readily for a dition in this case. The results reveal that the in-
specific case in future simulations. crease of the stiffness of the composite can be
First, a full 3-D finite element model containing significant in the CNT axial direction. With a
one CNT is used as shown in Fig. 5. One layer of volume fraction of the CNT being only about
elements are used for the CNT in this mesh, which 3.6%, the stiffness of the composite in the axial
have been found to be good enough to obtain direction ðEz Þ can increase by about 33% com-
converged FEM results. The 3-D FEM model is pared with that of the matrix, when Et =Em ¼ 10.
applied with the two load cases shown in Fig. 3. Next, two plane strain (2-D) FEM models,
From the FEM results, the four material constants containing 5 · 5 and 10 · 10 CNTs, respectively,
are extracted using Eqs. (2), (3), (5) and (6). are studied under the lateral loading (Fig. 3(b)).
The results for the effective material constants The FEM mesh for the 10 · 10 CNT model is
of the CNT-based composite from the 3-D FEM shown in Fig. 6 and the stress plot is given in Fig.
model are given in Table 1 (first row of data). The 7, which shows the typical distributions of the
effective YoungÕs modulus Ez estimated by the stresses around a CNT. The results of the effective
YoungÕs modulus and PoissonÕs ratio in the trans-
verse direction (xy plane, Fig. 2) are listed in Table
1 (second and third rows of data), for which the
YoungÕs modulus Ez and PoissonÕs ratio mzx needed
in Eqs. (5) and (6) are from the above 3-D RVE
results. The results from the three models (3-D
single CNT, 2-D 5 · 5 and 10 · 10 CNT models)
are also almost identical as shown in Table 1. This
suggests that a single CNT model, either 2-D or
3-D, may be sufficient in determining the effec-
tive material constants in the transverse direction,
as has been the case in studying conventional fiber-
reinforced composites, if the reinforcing fibers
(CNTs here) are distributed uniformly in a square
pattern in the transverse direction.
For comparison, the effective material constants
Fig. 5. A 3-D FEM model for the square RVE with a long obtained using the cylindrical RVE from Ref. [8],
CNT (CNT thickness ¼ 0.4 nm, with one layer of elements). which is of the same size (same length L and the

Table 1
Computed effective material constants for case (a): long CNT(s) through the RVE
FEM model Ez =Em FEM
FEM RM mzx , mzy Ex =Em , Ey =Em mxy
3-D, single CNT 1.3255 1.3255 0.3000 0.8492 0.3799
2-D, 5 · 5 CNTs – – – 0.8561 0.3745
2-D, 10 · 10 CNTs – – – 0.8534 0.3745
Cylindrical RVE [8] 1.4384 1.4384 0.3000 1.3336 0.4855
Note: Modulus ratio Et =Em ¼ 10, CNT thickness ¼ 0.4 nm, volume fraction ¼ 3.617%; RM ¼ rule of mixtures (Eq. (8) for square
RVEs; Ref. [8] for cylindrical RVEs).
8 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11

4.2. A short CNT inside the square RVE

In this example, a square RVE for a short CNT

in a matrix, as shown in Fig. 2, is studied. 3-D
FEM models has to be employed for this case,
even for the lateral loading situation. The dimen-
sions for the RVE are the same as in the previous
example, except for the total length of the CNT,
which is 50 nm (including the two end hemi-
spherical caps). The material constants used for
the CNT and matrix are the same as in the pre-
vious example. The finite element mesh used for a
quarter symmetry model is shown in Fig. 8. Again,
one layer of elements (quadratic bricks) are used
through the thickness of the CNT. Coupled DOF
constraints are imposed on the four lateral sur-
faces under both axial and lateral loading (Fig.
3(a) and (b)), so that all points on the lateral sur-
Fig. 6. A 2-D (plane strain) FEM model of the RVE contain- faces will move the same amount in the normal
ing 10
10 long CNTs (CNT thickness ¼ 0.4 nm, with one direction to simulate the constraints from the
layer of elements). surrounding material.
Stress contour plots of the first principal stresses
in the quarter model of the RVE are shown in Fig.
9 for the axial stretch case and in Fig. 10 for the
lateral loading case. The load-carrying capacities
of the CNT are obvious from these two stress
plots, which are consistent with the earlier simu-
lation results using cylindrical RVEs [7,8]. The
computed four effective material constants using
the FEM results are shown in Table 2, along with
the strength of materials solution (Eq. (10)) for the
stiffness in the axial direction ðEz Þ. The increase of

Fig. 7. Plot of the first principal stresses (
p) for the 10
10

RVE model under lateral loading (zoomed in view; Et =
Em ¼ 10).

diameter of the cylindrical RVE ¼ 2a of the square

RVE), are also listed in Table 1. It is seen that the
cylindrical RVE overestimates the YoungÕs mo-
duli. This may be explained by the fact that a cy-
lindrical RVE overestimates the volume fraction of
the CNT due to the negligence of the small amount Fig. 8. A 3-D quarter-symmetry FEM model for the square
of matrix material (at the four corners of the RVE with a short CNT (CNT thickness ¼ 0.4 nm, with one
square RVE) in the cylindrical RVE. layer of elements).
 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11 9

Fig. 9. Plot of the first principal stresses (
DL) for the 3-D RVE under the axial stretch DL ðEt =Em ¼ 10Þ.

Fig. 10. Plot of the first principal stresses (
p) for the 3-D RVE under the lateral load p ðEt =Em ¼ 10Þ.

the stiffness in the axial direction is moderate for overestimates the YoungÕs moduli compared with
Et =Em ¼ 10, due to the small volume fraction of the current square RVE model.
the CNT (about 1.6%). These results suggest that The strength of materials solution for the stiff-
short CNTs in a matrix may not be as effective as ness in the axial direction ðEz Þ, using the extended
long CNTs in reinforcing a composite. Again, for rule of mixtures (Fig. 4(b) and Eq. (10)), is quite
comparison, the effective material constants ob- close to the FEM solution which is based on 3-D
tained using the cylindrical RVE in Ref. [8] are elasticity, with a difference of only about 1%.
also listed in Table 2. The cylindrical RVE again Therefore, the extended rule of mixtures (Eq. (10))
10 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11

Table 2
Computed effective material constants for case (b): a short CNT inside the RVE
FEM model Ez =Em FEM
FEM ERM mzx , mzy Ex =Em , Ey =Em mxy
3-D, single CNT 1.0391 1.0500 0.3009 0.9342 0.3217
Cylindrical RVE [8] 1.0491 1.0628 0.3014 1.0033 0.2614
Note: Modulus ratio Et =Em ¼ 10, CNT thickness ¼ 0.4 nm, volume fraction ¼ 1.620%; ERM ¼ extended rule of mixtures (Eq. (10) for
square RVEs; Ref. [8] for cylindrical RVEs).

may serve as a quick tool to estimate the stiffness nally, the rules of mixtures, for both long and
of the CNT-based composites in the axial direction short CNT cases, are found to be quite accurate in
when the CNTs are relatively short, while the estimating the effective YoungÕs moduli in the
conventional rule of mixtures (Eq. (8)) can con- CNT axial direction. This may suggest that 3-D
tinue to serve in cases when the CNTs are rela- FEM modeling may not be necessary in obtaining
tively long. the effective material constants in the CNT direc-
tion, as in the studies of the conventional fiber-
reinforced composites.
Efforts in comparing the results presented in
5. Conclusion this paper using the continuum approach directly
with the MD simulations are underway. This is
The effective mechanical properties of CNT- feasible now only for a smaller RVE of one CNT
based composites are evaluated using square embedded in a matrix. In future research, the MD
RVEs based on 3-D elasticity theory and solved by and continuum approach should be integrated in a
the FEM. Formulas to extract the effective mate- multiscale modeling and simulation environment
rial constants from solutions for the square RVEs for analyzing the CNT-based composites. More
under two loading cases are established based on efficient models of the CNTs in a matrix also need
elasticity. Square RVEs with multiple CNTs are to be developed, so that a large number of CNTs,
also investigated in evaluating the YoungÕs mod- in different shapes and forms (curved or twisted),
ulus and PoissonÕs ratios in the transverse plane. or randomly distributed in a matrix, can be mod-
Numerical examples using the FEM are presented, eled. The ultimate validation of the simulation
which demonstrate that the load-carrying capa- results should be done with the nanoscale or mi-
bilities of the CNTs in a matrix are significant. croscale experiments on the CNT reinforced
With the addition of only about 3.6% volume composites.
fraction of the CNTs in a matrix, the stiffness of
the composite in the CNT axial direction can in-
crease as much as 33% for the case of long CNT Acknowledgements
fibers. These simulation results are consistent with
both the experimental ones reported in the litera- The financial support to the first author
ture (see, e.g., [3–6]) and the earlier numerical ones (X.L.C.) from the University Research Council of
using the cylindrical RVEs [8]. It is also found that the University of Cincinnati is acknowledged.
cylindrical RVEs tend to overestimate the effective
YoungÕs moduli due to the fact that they overes-
timate the volume fractions of the CNTs in a
matrix. The square RVEs, although more de- References
manding in modeling and computing, may be the
[1] E.T. Thostenson, Z.F. Ren, T.-W. Chou, Advances in the
preferred model in future simulations for estimat- science and technology of carbon nanotubes and their
ing the effective material constants, especially composites: a review, Composites Science and Technology
when multiple CNTs need to be considered. Fi- 61 (2001) 1899–1912.
 X.L. Chen, Y.J. Liu / Computational Materials Science 29 (2004) 1–11 11

[2] D. Qian, G.J. Wagner, W.K. Liu, M.-F. Yu, R.S. Ruoff, [18] E.W. Wong, P.E. Sheehan, C.M. Lieber, Nanobeam
Mechanics of carbon nanotubes, Applied Mechanical mechanics: Elasticity, strength, and toughness of nanorods
Review 55 (2002) 495–533. and nanotubes, Science 277 (1997) 1971–1975.
[3] D. Qian, E.C. Dickey, R. Andrews, T. Rantell, Load [19] K. Sohlberg, B.G. Sumpter, R.E. Tuzun, D.W. Noid,
transfer and deformation mechanisms in carbon nanotube– Continuum methods of mechanics as a simplified approach
polystyrene composites, Applied Physics Letters 76 (2000) to structural engineering of nanostructures, Nanotechnol-
2868–2870. ogy 9 (1998) 30–36.
[4] H.D. Wagner, O. Lourie, Y. Feldman, R. Tenne, Stress- [20] S. Govindjee, J.L. Sackman, On the use of continuum
induced fragmentation of multiwall carbon nanotubes in a mechanics to estimate the properties of nanotubes, Solid
polymer matrix, Applied Physics Letters 72 (1998) 188– State Communications 110 (1999) 227–230.
190. [21] C.Q. Ru, Column buckling of multiwalled carbon nanotu-
[5] L.S. Schadler, S.C. Giannaris, P.M. Ajayan, Load transfer bes with interlayer radial displacements, Physical Review B
in carbon nanotube epoxy composites, Applied Physics 62 (2000) 16962–16967.
Letters 73 (1998) 3842–3844. [22] C.Q. Ru, Axially compressed buckling of a doublewalled
[6] C. Bower, R. Rosen, L. Jin, J. Han, O. Zhou, Deformation carbon nanotube embedded in an elastic medium, Journal
of carbon nanotubes in nanotube–polymer composites, of the Mechanics and Physics of Solids 49 (2001) 1265–
Applied Physics Letters 74 (1999) 3317–3319. 1279.
[7] Y.J. Liu, X.L. Chen, Continuum models of carbon [23] D. Qian, W.K. Liu, R.S. Ruoff, Mechanics of C60 in
nanotube-based composites by the BEM, Electronic Jour- nanotubes, Journal of Physical Chemistry B 105 (2001)
nal of Boundary Elements, in press (2003). 10753–10758.
[8] Y.J. Liu, X.L. Chen, Evaluations of the effective materials [24] D. Qian, W.K. Liu, R.S. Ruoff, Load transfer mechanism
properties of carbon nanotube-based composites using a in nanoropes, Composite Science and Technology, in press
nanoscale representative volume element, Mechanics of (2003).
Materials 35 (2003) 69–81. [25] P.S. Das, L.T. Wille, Atomistic and continuum studies of
[9] D. Srivastava, M. Menon, K. Cho, Computational nano- carbon nanotubes under pressure, Computational Materi-
technology with carbon nanotubes and fullerenes, Com- als Science 24 (2002) 159–162.
puting in Science and Engineering 3 (2001) 42–55. [26] M.W. Hyer, Stress Analysis of Fiber-Reinforced Compos-
[10] M. Macucci, G. Iannaccone, J. Greer, J. Martorell, D.W.L. ite Materials, first ed., McGraw-Hill, Boston, 1998.
Sprung, A. Schenk, I.I. Yakimenko, K.-F. Berggren, K. [27] S. Nemat-Nasser, M. Hori, Micromechanics: Overall
Stokbro, N. Gippius, Status and perspectives of nanoscale Properties of Heterogeneous Materials, second ed., Else-
device modelling, Nanotechnology 12 (2001) 136–142. vier, Amsterdam, 1999.
[11] J. Han, A. Globus, R. Jaffe, G. Deardorff, Molecular [28] R.S. Ruoff, D.C. Lorents, Mechanical and thermal prop-
dynamics simulations of carbon nanotube-based gears, erties of carbon nanotubes, Carbon 33 (1995) 925–930.
Nanotechnology 8 (1997) 95–102. [29] M.M.J. Treacy, T.W. Ebbesen, J.M. Gibson, Exceptionally
[12] C.F. Cornwell, L.T. Wille, Elastic properties of single- high YoungÕs modulus observed for individual carbon
walled carbon nanotubes in compression, Solid State nanotubes, Nature 381 (1996) 678–680.
Communications 101 (1997) 555–558. [30] J.P. Lu, Elastic properties of single and multilayered
[13] S.B. Sinnott, O.A. Shenderova, C.T. White, D.W. Brenner, nanotubes, Journal of Physics and Chemistry of Solids 58
Mechanical properties of nanotubule fibers and composites (1997) 1649–1652.
determined from theoretical calculations and simulations, [31] A. Krishnan, E. Dujardin, T.W. Ebbesen, P.N. Yianilos,
Carbon 36 (1998) 1–9. M.M.J. Treacy, YoungÕs modulus of single-walled nanotu-
[14] T. Halicioglu, Stress calculations for carbon nanotubes, bes, Physical Review B––Condensed Matter 58 (1998)
Thin Solid Films 312 (1998) 11–14. 14013–14019.
[15] G.H. Gao, T. Cagin, W.A. Goddard, Energetics, structure, [32] N. Yao, V. Lordi, YoungÕs modulus of single-walled
mechanical and vibrational properties of single-walled carbon nanotubes, Journal of Applied Physics 84 (1998)
carbon nanotubes, Nanotechnology 9 (1998) 184–191. 1939–1943.
[16] M. Buongiorno Nardelli, J.-L. Fattebert, D. Orlikowski, C. [33] C. Goze, P. Bernier, L. Henrard, L. Vaccarini, E. Hernan-
Roland, Q. Zhao, J. Bernholc, Mechanical properties, dez, A. Rubio, Elastic and mechanical properties of carbon
defects and electronic behavior of carbon nanotubes, nanotubes, Synthetic Metals 103 (1999) 2500–2501.
Carbon 38 (2000) 1703–1711. [34] J.P. Salvetat, J.M. Bonard, N.H. Thomson, A.J. Kulik, L.
[17] J.-W. Kang, H.-J. Hwang, Mechanical deformation study Forro, W. Benoit, L. Zuppiroli, Mechanical properties of
of copper nanowire using atomistic simulation, Nanotech- carbon nanotubes, Applied Physics A––Materials Science
nology 12 (2001) 295–300. and Processing 69 (1999) 255–260.