Copyright © 2009 American Scientific Publishers Journal of

All rights reserved Computational and Theoretical Nanoscience

Printed in the United States of America Vol. 6, 2152–2155, 2009

Effect of Areal Density on the Growth

Morphology of Nanotube Forests
Haicheng Guo and Wei Lu∗
Department of Mechanical Engineering, University of Michigan, Ann Arbor, Michigan 48109, USA

Numerical simulations were performed to examine the effect of carbon nanotube areal density on
the growth morphology of multi-walled nanotube forests. Continuum modeling and finite element
method were adopted to model nanotube interaction. Results show that at a higher areal density,
nanotubes grow and form a vertically aligned morphology. However, at lower areal densities, the
morphology of nanotube forests can change from separate nanotube bundles to sparsely distributed
non-vertical arrangements.
Delivered by Ingenta to:
Keywords: Nanotube Forest,
RiceFinite
University, Fondren
Element Method, Library
Vertically Aligned Nanotubes.
IP : 176.14.23.89
Sun, 29 Jul 2012 19:21:05
1. INTRODUCTION shown that the areal density of the catalyst influences the
final morphology of nanotube forests.2 By uniformly vary-
Carbon nanotubes (CNTs) have been receiving much ing the areal density, the morphology of the CNT film
attention for their extraordinary thermal, electronic and can be manipulated from a tangled and sparse arrangement
mechanical properties. Recently, chemical vapor deposi- of individual CNTs, through a transition region of locally
tion (CVD) has emerged as a key technique to grow ver-
RESEARCH ARTICLE

bunched and self-aligned CNTs, to rapid growth of thick

tically aligned carbon nanotube (VACNT) arrays, where vertical CNT films.2
7 These observations call for a funda-
the CNTs self-orient perpendicular to the substrate surface mental understanding of how the growth of CNTs depends
due to initial crowding and continue to grow upward in on the spacing and distribution of catalysts, which is not
this direction.1
2 VACNT arrays are capable of utilizing the only scientifically intriguing but also technically critical
remarkable properties of individual nanotubes in macro-
for controlling the morphology of VACNT films.
scopic applications.2
3 Many experimental results show
Molecular dynamics simulations have been conduced to
that the superior properties of VACNTs depend largely
investigate the iron catalyzed nucleation and initial growth
on the good alignment. For instance, the enhanced ther-
of CNTs from catalysts.8
9 The effect of nanotube con-
mal conductivity for the CNT array composite film where
tact on their growth morphologies, however, has not been
CNTs were randomly dispersed in silicone elastomer was
addressed by a computational model and remains elusive.
only 4.6% of those enhanced by aligned CNTs.4 Further-
Limited by the number of atoms in the system,10 simulat-
more, it was found that electron transfer between the gold
ing the formation of nanotube forest using atomistic mod-
electrode and ferrocene moiety is 40 times slower through
eling alone is formidable. Elastic continuum modeling is
randomly dispersed nanotubes than through VACNTs,5
an effective way to address several nanotube problems,11
12
which makes aligned carbon nanotube arrays a bet-
especially for understanding the behavior of a large num-
ter choice for achieving more efficient electron transfer.
ber of nanotubes. For instance, finite element method has
Finally, the Young’s modulus of aligned CNT–polymer
been demonstrated to capture the growth of multiple iso-
nanocomposites along the CNT axis was found consid-
lated nanotubes under electric field.13 In this paper we
erably higher than that using randomly oriented CNTs
propose a continuum model for the growth of nanotube
embedded into a polymer matrix.6 Therefore, VACNTs are
considered a superior material for a wide range of appli- forest, and address the relation between nanotube spacing,
cations, including membranes, heat dissipation, electrical contact and the resulting morphology.
interconnects, and nanoelectronics.
In the CVD process, nanoparticle catalysts are formed 2. MODELING
on the substrate to yield nanotubes. Experiments have
We model the nanotube by two-node beam elements.
∗
Author to whom correspondence should be addressed. The shape of each element is described by a third order

2152 J. Comput. Theor. Nanosci. 2009, Vol. 6, No. 10 1546-1955/2009/6/2152/004 doi:10.1166/jctn.2009.1266

Guo and Lu Effect of Areal Density on the Growth Morphology of Nanotube Forests

polynomial. Each node has 6 degree of freedoms with the A new bottom element will be started when the element
generalized displacement given by u = u v w x y z . length reaches l. In each time step
t, we calculate K to
Here u, v, w are the displacements of the nodal point get the displacement and new nodal position xt+
t . Then
in the x, y, z coordinate directions while x , y , z are the program searches all contact pairs to exam whether
the rotations of the nodal tangent about the x, y, z axes. there is penetration indicated by gN < 0. If penetration
A nanotube seeds at the location of catalyst particles and happens, the program enters an iterative process. It calcu-
grows with a specific initial orientation. Based on this pic- lates KC and QC , and then uses Eq. (3) to get updated
ture we apply a fixed boundary condition at the bottom displacement and position xt+
t
1
. Note that in the new con-
end of each nanotube where it connects to the substrate. figuration the contact pairs and their positions will change,
In the simulation we continuously add new elements or leading to new KC and QC . So we repeat the process to
elongate elements at the bottom end of nanotubes in each get updated position xt+
t
2

  
xt+
t
n
. The iteration stops
time step. The inertia of nanotube is negligible, so at each
time step we solve an elastic equilibrium equation Ku = P
to obtain displacements. Here K and P are the standard (a)
10
stiffness matrix and nodal force vector when nanotubes are

(×10 nm)
not in contact. 5
To consider the contact between two nanotubes, we cal- 0
culate the relative position of their center lines. If two 0 10 20 30 40 50
nanotubes are detected to penetrate each other after apply- (×10 nm)
ing the displacement obtained from Ku = P,Delivered
we employby Ingenta
(b)
to:
Rice University,
a penalty method to enforce non-penetration and surface Fondren15
Library
IP : and
contact. To fix the idea, define one centerline as master 176.14.23.89
(×10 nm)
the other as slave. Starting from each node PSun,of the29slave,
Jul 2012 19:21:05
10

a straight line is drawn perpendicular to the master and 5

intersects it at point Q. The node P and point Q comprise 0
a contact pair. Denote the normal direction of the master 0 10 20 30 40 50

nQ , which points from Q to P . Add a penalty energy,

by (×10 nm)

C , to the potential energy of the system (c) 30

 

RESEARCH ARTICLE
 N 2
= a uP − uNQ + gN  dL (1) 25
C LC
20
(×10 nm)

where uNP and uNQ are the displacements of node P and

15
point Q in the normal direction nQ . The integration extends
over all intersecting elements. The term gN describes the 10
distance between the surfaces of two nanotubes,
5
 
gN = xP − xQ · nQ − D (2) 0
0 10 20 30 40 50

where xP , xQ are position vectors and D is the outer diam- (×10 nm)
eter of nanotubes. The coefficient  is a chosen large num- (d) 40
ber. Note that point Q is not necessarily a nodal point of
the master. We interpolate the displacement of Q by two
adjacent nodal point displacements using the beam ele- 30
ment shape function. The form of Eq. (1) ensures that after
(×10 nm)

several iterations surface contact is accurately captured.

Taking variation on the modified energy functional 20

that incorporates Eq. (1), we obtain a finite element

formulation 10
K + KC u = P + QC (3)
where KC = aNCT nQ nQT NC , QC = −aNCT nQ gN  and NC
0
is the shape function matrix. 0 10 20 30 40 50
A program is developed based on the above model. We (×10 nm)
define a maximum element length l to ensure resolution. Fig. 1. Simulation of nanotube growth with an areal density of
In each time step we add l/n (n = 10∼15) to the bottom 4 ×1010 cm−2 . Shown for time (a) 0.24 s, (b) 0.48 s, (c) 0.84 s, (d) 1.02 s.
element in the initial growth direction from the catalyst. Three separate nanotube bundles are formed.

J. Comput. Theor. Nanosci. 6, 2152–2155, 2009 2153

 Effect of Areal Density on the Growth Morphology of Nanotube Forests Guo and Lu

when xt+
t
n+1
− xt+
t
n
 < , where  is a small positive num-
taken to be 1 TPa and 0.4 TPa, respectively.11
12 In our
ber. This iteration results in an accurate surface contactcalculation these nanotubes are considered as cylindrical
configuration for time t +
t. tube beams with an outer diameter of 9 nm and thickness
of 1.7 nm.
Figure 1 shows the growth process of 10 nanotubes
3. RESULTS AND DISCUSSION with a spacing of 50 nm between seeds, which corre-
We take multi-walled carbon nanotube having 5 walls and sponds to a nanotube areal density of 4 × 1010 cm−2 .
an outer diameter of 9 nm as an example system.14 The The ratio of the nanotube diameter to the nanotube
mean distance between two adjacent walls is 0.34 nm. 11 spacing is 0.18. Each nanotube has its specific initially
Based on many experimental results and molecular growth orientation. Under these conditions the nanotubes
dynamic simulations, the Young’s and shear modulus are grow together and form three separate bundles shown in
Figure 1. Despite the random initial orientations, the over-
all growth demonstrates vertical alignment. This vertical
(a) 5 growth and bundling behavior is consistent with experi-
mental observations.2
(×10 nm)

To consider the effect of areal density, we next decrease

the nanotube spacing to 15 nm, which corresponds to
0
0 5 10 15
an areal density of 44 × 1011 cm−2 . The correspond-
(×10 nm) ing nanotube diameter-to-spacing ratio is 0.6. All other
Delivered by Ingenta
conditionsto:remain the same as the previous simulation.
(b) Rice University, Fondren
Figure 2 Library
shows that initially the nanotubes appear to form
5
IP : 176.14.23.89
three bundles due to their starting orientations. However,
(×10 nm)

Sun, 29 Jul 2012as 19:21:05

the nanotubes grow higher, some are pushed away
from their initial bundles and bridge other bundles due to
0 the small inter-bundles distance. These branching actions
0 5 10 15
finally eliminate the small bundles appear in the early stage
(×10 nm)
of growth, leading to a continuum vertically aligned nano-
(c) 10
tube forest. It was argued that the presence of nanotube
tangling on the top of the nanotube forest may help the
RESEARCH ARTICLE

growth of VACNTs.3
6 Here our simulations suggest that

(×10 nm)

growing VACNTs does not necessarily need to form a tan-

5
gling layer of nanotubes.

4. CONCLUSIONS
0
0 5 10 15 In conclusion, we have developed a finite element
(×10 nm)
approach to investigate the relation between nanotubes
(d) 15
areal density and vertically aligned nanotube forest for-
mation. Results show that vertically aligned nanotubes are
formed when the nanotubes are distributed densely. With
decreasing areal density, nanotube forest changes morphol-
10 ogy form vertically aligned growth to separate bundles and
(×10 nm)

finally to non-vertical growth.

Acknowledgments: The authors acknowledge financial

5 support from National Science Foundation Award No.
DMI-0700048.

0 References
0 5 10 15
(×10 nm) 1. M. Meyyappan, L. Delzeit, A. Cassell, and D. Hash, Plasma Sources
Sci. Technol. 12, 205 (2003).
Fig. 2. Simulation of nanotube growth with an areal density of 2. R. D. Bennett, A. J. Hart, and R. E. Cohen, Adv. Mater. 18, 2274
44×1011 cm−2 . Shown for time (a) 0.12 s, (b) 0.3 s, (c) 0.57 s, (d) 0.87 s. (2006).
The branching actions eliminate the small bundles appear in the early 3. H. Wang, Z. Xu, and G. Eres, Appl. Phys. Lett. 88 (2006).
stage of growth, leading to a continuum vertically aligned nanotube 4. H. Huang, C. H. Liu, Y. Wu, and S. S. Fan, Adv. Mater. 17, 1652
forest. (2005).

2154 J. Comput. Theor. Nanosci. 6, 2152–2155, 2009

Guo and Lu Effect of Areal Density on the Growth Morphology of Nanotube Forests

5. J. J. Gooding, A. Chou, J. Q. Liu, D. Losic, J. G. Shapter, and D. B.
Hibbert, Electrochem. Commun. 9, 1677 (2007).
6. E. J. Garcia, A. J. Hart, B. L. Wardle, and A. H. Slocum, Adv. Mater.
19, 2151 (2007).
7. L. Zhang, Z. R. Li, Y. Q. Tan, G. Lolli, N. Sakulchaicharoen, F. G.
Requejo, B. S. Mun, and D. E. Resasco, Chem. Mater. 18, 5624
(2006).
8. F. Ding, K. Bolton, and A. Rosen, J. Phys. Chem. B 108, 17369
(2004).
9. H. Duan, F. Ding, A. Rosen, A. Harutyunyan, T. Tokune,
S. Curtarolo, and K. Bolton, Eur. Phys. J. D 43, 185 (2007).
10. Q. Wang and V. K. Varadan, Smart Mater. Struct. 14, 1196 (2005).
11. R. C. Batra and A. Sears, Int. J. Solids Struct. 44, 7577 (2007).
12. C. Y. Li and T. W. Chou, Composites Sci. Technol. 63, 1517
(2003).
13. Q. L. Bao, H. Zhang, and C. X. Pan, Comput. Mater. Sci. 39, 616
(2007).
14. A. J. Hart and A. H. Slocum, Nano Lett. 6, 1254 (2006).

Received: 25 February 2009. Accepted: 11 March 2009.

Delivered by Ingenta to:

Rice University, Fondren Library
IP : 176.14.23.89
Sun, 29 Jul 2012 19:21:05

RESEARCH ARTICLE

J. Comput. Theor. Nanosci. 6, 2152–2155, 2009 2155