Professional Documents
Culture Documents
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
1. Introduction
There has been much research activity on carbon nanotubes (CNTs) since their
discovery in 1991 by Iijima of the NEC Laboratory in Tsukuba, Japan (Iijima
1991). In such research, atomistic-based methods (Yakobson et al. 1996;
Hernandez et al. 1998; Sanchez-Portal et al. 1999) and continuum mechanics
(Govindjee & Sackman 1999; Harik 2002; Lau et al. 2004) are the two main
theoretical methods that are used to study the mechanical behaviour of CNTs.
However, the atomistic-based methods are currently limited by computing
capability. For example, in our molecular dynamics (MD) simulation (Liew et al.
2004a) of buckling behaviour, the calculation for a single-walled (10,10) CNT
with 2000 atoms required 36 h on a single CPU SGI origin 2000 system. The
computational time increases sharply with the increasing number of atoms, and
Nx
Nx RI
RO
van der
Waals forces
2. Model development
(a ) Matrix model
Previous studies by Wagner et al. (1998) on the stress transfer between a
CNT and a polymer matrix show that CNT–polymer adhesion is quite strong,
and that not only the normal stress but also the shear stress transfers from the
CNT to the polymer matrix. Thus, in the analysis of infinitesimal buckling, we
assume that the relation between the pressure and the deflection of the
outermost tube surface can be described by the Pasternak foundation model
(Pasternak 1954) i.e.
pN ðx; qÞ ZKKW wðx; qÞ C Gb V2 wðx; qÞ; ð2:1Þ
where the first parameter KW is the Winkler foundation modulus (Winkler
1867), the second parameter Gb is the stiffness of the shearing layer, N is the
number of layers of the outermost tube and V2 is the Laplace operator, which
is defined as
v2 1 v2
V2 Z C : ð2:2Þ
vx 2 RN vq2
This model assumes that the elastic foundation consists of a closed spaced and
independent springs, where the top ends of the springs are connected to an
incompressible layer that resists only transverse shear deformation. This model
can describe the interaction between the pressure and deflection of the
outermost tube and a shear interaction between the springs. When setting the
second parameter GbZ0, the Pasternak model is reduced to the Winkler
model, i.e.
pN ðx; qÞ ZKKW wðx; qÞ: ð2:3Þ
(b ) Basic formulas
Based on the classical thin shell theory (Timoshenko & Gere 1961), the basic
equations for the elastic buckling of a multi-walled CNT that is embedded in a
matrix can be derived as the N coupled equations, i.e.
9
L1 w1 Z V41 p1 ; >>
>
>
>
>
« >
>
=
4 ð2:4Þ
Li w i Z Vi p i ;
>
>
« >
>
>
>
>
>
4 ;
LN wN Z VN pN ;
v2 4 Nq v2 4 Eh v4
Li Z Di V8i KNx Vi K 2 2 Vi C 2 ; ð2:5Þ
vx 2 Ri vq Ri vx 4
in which x is the axial and q the circumferential coordinate, NxZsx h and NqZsqh
are the uniform forces per unit length in the axial and circumferential directions
of the i th tube prior to buckling with sx being the axial and sq being the
circumferential stress, Di is the bending stiffness of the i th tube, and
v2 1 v2
V2i Z C : ð2:6Þ
vx 2 R2i vq2
Due to the infinitesimal deflection between any two layers, the pressure at any
point of the tubes can be expressed as
X
N
pi ðx; yÞ Z p ij ðx; qÞ C Dpi ðx; qÞ; ð2:7Þ
jZ1
where p ij ðx; qÞ is the initial vdW pressure contribution to the i th layer from the
j th layer prior to buckling, N is the total number of layers of the multi-walled
CNT, and Dpi(x, q) is the pressure increment (after buckling) that is due to the
(c ) vdW interaction
It can be observed from equation (2.10) that all of the governing equations are
coupled with each other due to the vdW interaction, which is characterized by
the initial pressure p ij (before buckling) and the coefficient cij (after buckling).
Thus, the key issue for the buckling analysis is to develop an efficient approach
for the description of the vdW interaction. In our previous work (He et al. 2005),
we derived explicit formulas to describe the vdW interaction between any two
tubes of a multi-walled CNT. The vdW interaction can be characterized by
" ! ! #
20483s12 X5
ðK1Þk 5 12 10243s6 X2
ðK1Þk 2
p ij Z Eij K Eij6 Rj ; ð2:11Þ
9a4 kZ0
2k C 1 k 9a 4
kZ0
2k C 1 k
and
1001p3s12 13 1120p3s6 7
cij ZK Eij K Eij Rj ; ð2:12Þ
9a 4 9a 4
where aZ1.42 Å is the C–C bond length, Rj is the radius of the j th layer, the
subscripts i and j denote the i th and j th layers, respectively, and Eij6 , Eij7 , Eij12 and
Eij13 are the elliptical integrals.
The boundary conditions for the simply supported tubes are as follows.
v2 wk
Z 0; at x Z 0 and x Z L:
wk Z ð3:1Þ
vx 2
The deflection function that satisfies the boundary conditions equation (3.1) can
be approximated by
mpx
wk Z Ak sin sin nq; ð3:2Þ
L
where Ak (kZ1, 2, ., N ) are N unknown coefficients, L is the length of the multi-
walled CNT, and m and n are the axial half wavenumber and circumferential
wavenumber, respectively.
The substitution of equation (3.2) into equation (2.10) gives us
8
>
< mp 2 n 2 2 X N 2
ckj mp p k Rk n 2
C K C Nx K
>
: L Rk jZ1
D L2 D Rk
2 32 9
>
= XN
Eh 6 1 7 ckj
C 4 2 5 A k C A Z 0 ðk Z 1; 2; . ; N K1Þ;
DR2k 1 C Ln >
; jZ1
D j ð3:3Þ
mpRk
and
8
>
< mp 2 2 2 X N 2 p R n 2
n cNj mp
C K CNx K k k
>
: L Rk jZ1
D L2 D RN
2 32 9
2 >
=
Eh 6 1 7
KW Gb mp 2 n XN
cNj
2 4 5
C C 4C 2 C AN C A Z0;
DRN 1C Ln 2 L L L RN >
; jZ1
D j
mpRN
ð3:4Þ
where pkZKNq/Rk is the net pressure that is exerted on k th tube, which is
assumed to be inward, the dimensionless buckling load factor N *ZNxL2/D, the
dimensionless Winkler modulus factor KW ZKW L4 =D, and the shear modulus
factor Gb ZGb L2 =D. Equations (3.3) and (3.4) can be rewritten in matrix form as
2 c c c 3
b11 12 13 . 1n
6 D D D 7
2 38 9 6 6 c c c
78 9
7 A
1 0 0 . 0 > > A 1>> 6 21
b22 23
. 2n 7>> 1>>
6 >
7> >
> 6 D D D 7 >
> >
>
6 0 1 0 . 0 7> > A >
> 6 7 >
> A >
>
mp 2 6 7>
< > 2 = 6 6 c c c
7>
7 < 2 >
=
6 7 6 31 32
b 3n
7
6 0 0 1 . 0 7 A3 Z 6 D D . A3 ;
33
D 7
KNx
L2 6 7>
> >
> 6 7 >
> >
>
6 « « « « « 7> > « >
> 6 7 >
> « >
>
4 >
5> > 6 « > « « « « 7 >
> >
>
>
: ; 6 > >
7: > ;
0 0 0 . 1 An 6 7 An
6c 7
4 n1 cn2 cn3 5
. bnn
D D D
ð3:5Þ
or equivalently
8 9
> A1 >
> >
mp 2 >
>
<A >
>
=
2
KNx I N!N KC N!N Z 0; ð3:6Þ
L2 >
> « >
>
>
> >
>
: ;
AN
where
2 32
2 2 XN 2
mp 2 n ckj pk Rk n Eh 6 1 7
bkk Z C K K C 4 5
L Rk D D Rk DRk 1 C Ln 2
2
jZ1
jsk
mpRk
ðk Z 1; 2; .; N K1Þ; ð3:7Þ
and
2 32
2 2 X N 2
mp 2 n cNj pN RN n Eh 6 1 7
bNN Z C K K C 4 5
L RN D D RN DR2N 1 C Ln 2
jZ1
jsk
mpRN
2
K G mp 2 n
C W
4
C 2b C : ð3:8Þ
L L L RN
ð4:2Þ
The condition for the non-zero solution of A1 and A2 leads to a relation for the
buckling load of the double-walled CNT
2 2 2
mp mp c c
Nx 2
CðB1 CB2 Þ Nx 2
CB1 B2 K 12 221 Z0; ð4:3Þ
L L D
where
2 mp 2 32
2 2 2
mp 2 n c pR n Eh 6 L 7
B1 Z C K 12 K 1 I C 2 4 2 5 ; ð4:4Þ
L RI D D RI DRI mp 2 C n
L RI
and
2 mp 2 32
mp 2 2 2 2
n c pR n Eh 6 L 7
B2 Z K 21 K 2 O 4 5
DR2O mp 2 C n 2
C C
L RO D D RO
L RO
2
KW Gb mp 2 n
C 4C 2 C : ð4:5Þ
L L L RO
Note that we assume the inward pressure to be positive, and that any increase
(wiKwjO0, i, jZ1, 2) or decrease (wiKwj!0) in the space between the inner
and outer tubes would cause an attractive or repulsive vdW force, respectively.
Thus, from equation (2.9) we have c12!0 and c21!0, and therefore
c c c c
ðB1 CB2 Þ2 K4 B1 B2 K 12 221 ZðB1 KB2 Þ2 C4 12 221 O0: ð4:6Þ
D D
Thus, the solution to equation (4.3) is given by
2 1 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mp c c
KNx 2
Z B1 CB2 H ðB1 KB2 Þ2 C4 12 221 : ð4:7Þ
L 2 D
It can be seen that the axial buckling load factor that is determined by equation
(4.8) occurs on the inner tube, which is modelled as an individual cylindrical
shell. However, for a double-walled CNT with a small radius, such as (5,5) or
(10,10) nanotubes, the difference between radii is not small and the effect of
ROORI cannot be ignored. In this case, we have B1OB2, and the buckling load
factor is determined by
8 2 mp 2 32
2 2 >
< mp 2 2
L n 2 Eh 6 L 7
KNx Z C C 4 2 5
mp : > L RO 2
DRO mp C n 2
L RO
9 ð4:9Þ
2 > =
K G mp 2 n
C W C 2b C :
L4 L L RO >
;
The buckling load factor that is determined by equation (4.9) will occur on the
outer tube of the double-walled CNT.
when the interlayer separation is taken as 0.34 nm. Hence, we have B1B2O
c12c21/D2 and can easily arrive at
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c c
B1 C B2 O ðB1 KB2 Þ2 C 4 12 221 : ð4:10Þ
D
To ensure that the right-hand side of equation (4.7) is positive and lower, we take
the negative sign before the square root in equation (4.7) to obtain the solution to
the axial load that is given by
2 1 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mp c c
KNx 2
Z B1 C B2 K ðB1 KB2 Þ2 C 4 12 221 : ð4:11Þ
L 2 D
Normally, the ratio (ROKRI)/RI is very small and the terms that are related to
the ratio can be neglected. Thus, we have the relationships
2 2 2 2 2
mp 2 n mp 2 n n RO C R I RO KRI
C Z C C
L RI L RO RO RI RI
mp 2
2
n 2
z C ; ð4:12Þ
L RO
and
n n RO n n RO KRI n
Z Z C z : ð4:13Þ
RI RO R I RO RO RI RO
The substitution of equations (4.12) and (4.13) into equation (4.11) gives us an
approximate formula for the buckling load factor of a double-walled CNT that is
embedded in a8matrix
2 mp 2 32
2 2 < > 2 2
L mp 2 n Eh 6 L 7
KN Z C C 2 4 2 5
mp > : L R O DR O mp 2
C RnO
L
2
KW
Gb mp 2 n c Cc21
C 4C 2 C K 12
2L 2L L RO 2D
2 2
c12 Cc21 c Kc21 KW Gb mp 2 n p 2 RO n 2
K C 12 C C K
2D D 2L4 2L2 L RO D RO
9
2 2 ! 12 > =
KW Gb mp 2 n p2 RO n 2
C C C K :
2L4 2L2 L RO D RO >
;
ð4:14Þ
Hence, the critical buckling load factor for a double-walled CNT with a matrix
that is modelled as a Pasternak foundation can be determined by minimizing the
buckling loads that are obtained from equation (4.14). It can be observed from
equation (4.14) that the vdW force and the foundation are coupled with each
other. We now discuss the effect of the vdW interaction and the effect of both the
vdW interaction and the matrix on the buckling load factor of a double-walled
CNT that is embedded in a matrix.
ð4:21Þ
As the radius of multi-walled CNTs increases, the vdW coefficients cij and cji
approach the same constant (He et al. 2005), and thus the difference between c12
and c21 is very small compared to c12. Note that c12!0 and jc12j[jp2j (jc12j is
around 12 orders of magnitude higher than jp2j; He et al. 2005), it is easy to see
that if and only if
2 2
p2 RO n c12 Kc21 p2 RO n 2
2 K
Gb mp 2
D RO D D RO
KW n
C 2 C O ; ð4:22Þ
L 4 L L RO 2 c
p2 RO
D
n
RO K 12
D
then the presence of both the elastic matrix and the vdW interaction will raise
the buckling load that is determined by equation (4.21) or equation (4.14).
However, if we take GbZ0 and c12Zc21, then equation (4.14) reduces to the
result that Ru (2001b) obtained. In addition, when the vdW force is neglected, it
can be seen that equation (4.14) reduces to the classical equation for the buckling
load of the outer tube without a matrix
8 2 32 9
2 2 > mp 2 >
L < mp 2 n 2
2
Eh 6 =
L 7
KN Z C C 4 2 5 >: ð4:23Þ
mp > : L RO DR2O mp 2 n ;
L C RO
In this case, no influence on the buckling load comes from the matrix.
0.07
present model
0.06 Liew et al. (2004a)
0.04
0.03
0.02
0.01
1 2 3 4 5 6 7 8 9 10 11
number of layers of the multi-walled CNT
Figure 2. Comparison of the critical axial strains obtained by the present model and the results of
the MD simulation by Liew et al. (2004a) for multi-walled CNTs with an innermost radius
RIZ0.34 nm.
1.96
equation (4.11)
1.92 equation (4.14)
equation (4.9)
buckling load, N * (×10 5)
Ru (2001a)
1.88
n=8
1.84
1.80
1.76
1.72
70 75 80 85 90 95 100 105 110 115
axial half wavenumber, m
Figure 3. Comparison between the different solutions to the buckling loads of a double-walled CNT
with a Pasternak foundation (KW Z 1 !1010 and GbZ0).
vdW forces do not increase the critical axial strain for infinitesimal buckling of
double-walled CNTs, but it is obvious that he does not consider the effect of the
ignorance of the terms that are related to the ratio of (ROKRI)/RI on the
buckling load.
As the vdW force is coupled with the elastic foundation, we examine the effect
of both the vdW interaction and the elastic foundation by comparing equations
(4.11), (4.14) and (4.23). Figure 4 shows the buckling loads with respect to the
11.0
equation (4.11)
10.8 equation (4.21)
10.6 equation (4.23)
4.5
0.32
4.0 equation (4.11)
equation (4.14)
critical buckling load, N (N m –1) 0.30 equation (4.23)
3.5
3.0 0.28
2.5
0.26
2.0
0.24
1.5 50 52 54 56 58 60
1.0
0.5
0 10 20 30 40 50 60 70 80
inner radius, R (nm)
Figure 5. Critical buckling loads versus the inner radius for a double-walled CNT with a Pasternak
foundation (KW Z 1 !1010 and GbZ1!105).
1.28
n=0
1.26 n=1
n=2
buckling load, N * (×105)
1.24 n=3
n=4
1.22 n=5
n=6
1.20
1.18
1.16
1.14
1.12
50 55 60 65 70 75 80 85
axial half wavenumber, m
Figure 6. Dependence of the buckling loads on the wavenumbers (m, n) for a six-walled CNT
(KW Z 0 and Gb Z 0).
1.65
n =0
1.60 n =1
n =2
1.45
1.40
1.35
60 70 80 90 100 110
axial half wavenumber, m
Figure 7. Dependence of the buckling loads on the wavenumbers (m, n) for a six-walled CNT with
an elastic matrix modelled as a Winkler foundation (KW Z 1 !1010 and Gb Z 0).
1.85
n=0
1.80 n=1
buckling load, N * (×105 )
n=2
1.75 n=3
n=4
n=5
1.70 n=6
1.65
1.60
1.55
1.50
60 65 70 75 80 85 90 95 100 105 110 115
axial half wavenumber, m
Figure 8. Dependence of the buckling loads on the wavenumbers (m, n) for a six-walled CNT with
an elastic matrix modelled as a Pasternak foundation (KW Z 1 !1010 and Gb Z 1 !105 ).
shear modulus on the critical buckling load, figure 8 shows the dependence of the
buckling loads on the wavenumbers (m, n) for a six-walled CNT with KW Z1!
10 5
10 and Gb Z 1 !10 . The minimization of the buckling loads with respect to the
wavenumbers m and n gives a critical buckling load Ncr Z 151 711:4, which is
higher than the critical buckling load of the six-walled CNT with a Winkler
foundation. As expected in the discussion on equation (4.14), figures 6–8 show
that the critical buckling load increases with the increase of the Winkler modulus
and the shear modulus.
To illustrate the influence of the elastic matrix on the critical buckling load,
the critical buckling loads are obtained by minimizing the buckling loads with
respect to the wavenumbers m and n for various KW and Gb and are plotted in
2.0
1.9
Figure 9. Critical buckling loads of a six-walled CNT versus the elastic foundation modulus.
figure 9. It is very clear from figure 9 that the critical buckling loads increase as
KW and Gb increase. As can be seen, the critical buckling load rises most quickly
with the increase of KW for Gb Z 0. As Gb increases, the increase of the critical
buckling load slows gradually with the increase of KW . However, with any value
of Gb , the difference between any two critical buckling loads becomes smaller
with the increase of KW until all of the critical buckling loads approach the same
constant, as shown in figure 9.
Two cases are considered to examine the dependence of the critical buckling
load on the number of layers of a multi-walled CNT. The first case is that of a
multi-walled CNT with a fixed innermost radius of 8.5 nm. Table 1 presents the
critical buckling loads with any combination of KW Z 0, 1!1010, 2!1010 and
5 5
Gb Z 0, 1!10 , 2!10 for two- to ten-walled CNTs that are embedded in an
elastic matrix. Again, it can be seen that the critical buckling load increases with
the increase of KW and Gb for all of the CNTs. The critical buckling load
decreases as the layers increase from two to ten for any combination of KW Z 0,
10 10 5 5
1!10 , 2!10 and Gb Z 0, 1!10 , 2!10 . This is because the radius of the
outermost tube increases as the total number of layers increases, and the critical
buckling load lowers as the radius increases (Allen & Bulson 1980). Note that we
assume that the entire CNT buckles when one tube buckles, and that critical
buckling always occurs on the outermost tube, or the tube next to the outermost
tube when the outermost tube is bonded with the matrix.
In the second case, the critical buckling loads of a multi-walled CNT with a
fixed outermost radius of 11.56 nm are derived and presented in table 2.
The results can be compared with those in table 1 for multi-walled CNTs with a
fixed innermost radius of 8.5 nm, table 2. For multi-walled CNTs without a
matrix (KW Z 0 and Gb Z 0), the critical buckling load rises (or in other words,
the capability against buckling increases) as the number of layers increases.
However, it is interesting to note that the critical buckling load rises first
and then drops with the increase of the number of layers for multi-walled CNTs
Table 1. Critical buckling loads Nx (N mK1) for multi-walled CNTs with a fixed innermost radius
fixed as RIZ8.5 nm
Table 2. Critical buckling loads Nx (N mK1) for multi-walled CNTs with a fixed outermost radius
ROZ11.56 nm
that are embedded in a matrix, which means that due to the coupling
effects of the vdW force and the matrix, there is a given number of layers at
which the critical buckling load is lowest for a multi-walled CNT that is
embedded in a matrix. For example, the lowest critical buckling loads occur on
the eight-, ten-, seven-, nine-, ten-, nine-, ten-, and eleven-walled CNTs for the
matrix parameters KW Z 1 !1010 and Gb Z 0; KW
Z 2 !1010 and Gb Z 0; KW
Z0
10 10 10 10
a n d Gb Z 1 !10 ; KW Z 1 !10 a n d Gb Z 1 !10 ; KW Z 2 !10 a nd
Gb Z 1 !1010 ; KW
Z 0 and Gb Z 2 !1010 ; KW
Z 1 !1010 and Gb Z 2 !1010 ; and
10 10
KW Z 2 !10 and Gb Z 2 !10 , respectively. The corresponding lowest critical
buckling loads are 1.5397, 1.6547, 1.4854, 1.6347, 1.7116, 1.6037, 1.6984 and
1.7509 N mK1, respectively, as shown in table 2.
6. Conclusions
References
Allen, H. G. & Bulson, P. S. 1980 Background to buckling, ch. 7. UK: McGraw-Hill.
Bower, C., Rosen, R., Jin, L., Han, J. & Zhou, O. 1999 Appl. Phys. Lett. 74, 3317–3319. (doi:10.
1063/1.123330.)
Frankland, S. J. V. & Brenner, D. W. 2000 Mal. Rec. Symp. Proc. 593, 199–204.
Girifalco, L. A. & Lad, R. A. 1956 J. Chem. Phys. 25, 693–697. (doi:10.1063/1.1743030.)
Govindjee, S. & Sackman, J. L. 1999 Solid State Commun. 110, 227–230. (doi:10.1016/S0038-
1098(98)00626-7.)
Harik, V. M. 2002 Comp. Mater. Sci. 24, 328–342. (doi:10.1016/S0927-0256(01)00255-5.)
He, X. Q., Kitipornchai, S. & Liew, K. M. 2005 J. Mech. Phys. Solid 53, 303–326. (doi:10.1016/
j.jmps.2004.08.003.)
Hernandez, E., Goze, C., Bernier, P. & Rubio, A. 1998 Phys. Rev. Lett. 80, 4502–4505. (doi:10.
1103/PhysRevLett.80.4502.)