Vous êtes sur la page 1sur 149

ABSTRACT

ZHAO, HAIBO. Carbon Nanotubes and Carbon Nanotube Fiber Sensors: Growth,
Processing and Characterization. (Under the direction of Prof. F. G. Yuan and Prof. Y. T.
Zhu.)
With multiple outstanding properties, such as high Youngs modulus, high strength,
good thermal conductivity and electrical conductivity, carbon nanotube (CNT) has been
considered as a new generation of material that has many potential applications in many
fields. One obstacle that stands in the way of applying CNTs in the real world is the limited
growth length. Catalyst instability is one of the many factors that cause the stops of CNT
growth. In this research, intermetalic Fe-Zr catalyst was used to grow millimeter long CNT
arrays. The Fe-Zr particles enabled the growth of 1.7 millimeter long carbon nanotube arrays
in 45 minutes. A comparison with pure Fe catalyst indicated that adding Zr to iron can
stabilize the Fe catalyst at the CNT growth temperature and moderate its reactivity.
In future, when CNTs are largely used in industrial, mass production of CNTs at a low
cost is vital for market competition. In many current CNT growth methods, a process of
depositing a thin catalyst film on top of Al
2
O
3
film on a piece of silicon wafer is required.
Thus the size of CNT samples is limited by the size of the largest silicon wafer currently
available, which is 8 inch in diameter. In this study, FeCl
2
powders were used as the catalyst
to grow CNT arrays not only on traditional silicon substrates but also on quartz substrates
and carbon sheets. This unique method does not require the thin film deposition step, which
shortens the time used for each batch of CNT growth. The simplicity of this method allows
an easy scale-up for mass production of CNTs with a low cost. In order to improve this


method, HCl was used to assist the CNT growth. HCl was added via flowing a small amount
of C
2
H
2
thorough a bubbuler where HCl solution was contained. With the assistance of HCl,
CNT growth could be extended to 1 hour. 3 mm tall non-spinnable arrays and 2 mm tall
spinnable arrays were produced using this method.
With the increasing use of composite materials, real time health monitoring of
composite structures becomes vital for maintenance purpose as well as prevention of
catastrophic failure. In this research, a novel prototype of CNT fiber sensor with excellent
repeatability and stability was applied for in-situ structural health monitoring. The CNT fiber
was spun directly from CNT arrays, and its electrical resistance increases linearly with
tensile strain, which makes it an ideal strain sensor. Importantly, it shows consistent
piezoresistive behavior under repetitive straining and unloading, and good stability at
temperatures ranging from 77K to 373K. CNT fiber sensors can be easily embedded into
composite structures with minimal invasiveness and weight penalty due to the lightweight
and good mechanical properties of fibers. With multiple fibers aligned in the composite,
crack initiation and propagation could be monitored in situ.










Carbon Nanotubes and Carbon Nanotube Fiber Sensors: Growth, Processing and
Characterization

by
Haibo Zhao


A dissertation submitted to the Graduate Faculty of
North Carolina State University
in partial fulfillment of the
requirements for the Degree of
Doctor of Philosophy



Mechanical Engineering

Raleigh, North Carolina
2010

APPROVED BY:


_________________________ _________________________
Dr. Fuh-Gwo Yuan Dr. Yuantian Zhu
Committee Chair Committee Co-chair



_________________________ _________________________
Dr. Melur Ramasubramanian Dr. Yong Zhu

ii

DEDICATION
I dedicate this dissertation
to my wife, Qian,
for encouraging me, accompanying me in my study,
to our daughter, Grace
for bring me many joys,
to my parents
for unconditionally supporting me whenever I need.
iii

BIOGRAPHY
Haibo Zhao was born in Handan in Hebei Province of China on April 8
th
, 1983. He
attended Hebei Fengfeng Chunguang High School in 1997. During his high school years, he
often climbed Taihang Mountain with his friends. The beautiful view of the mountain
inspired his passion for love of nature. After graduation, Haibo decided to pursue a higher
education in the field of Chemistry at Harbin Institute of Technology (HIT). At the senior
year of his undergraduate study, Haibo followed Prof. J iang to conduct research in preparing
barium ferrite thin film via the spin coating method. Haibo received his bachelors degree in
2005 and continued to pursue a Master degree in Chemical Engineering at HIT. Under the
direction of Prof. Wu, he used a polyacrylamide gel method to prepare lithium
aluminosilicate glass-ceramic powders and studied the crystallization behavior under heat
treatment. After graduation, Haibo joined the Department of Mechanical and Aerospace
Engineering at North Carolina State University, where he started to work in the nano world.
At the first year of his PhD study, Haibo was supported by Honda Corporation to build a
nano device to reuse the waste heat. At the second year, he received a teaching assistant
position and changed his project to grow carbon nanotubes using chemical vapor deposition.
At the third year, he was supported by North Carolina Space Grant and worked on
fabrication of CNT fibers and applying them as strain sensors. In the summer of 2009, Haibo
was invited to Shizuoka University in J apan by Prof. Inoue, where he improved the chloride-
assisted CNT growth method that was invented by Prof. Inoue. After returning to North
Carolina State University, He continued improving this method at his fourth year of PhD
study.
iv

ACKNOWLEDGMENTS
First and foremost I would like to thank my advisors, Dr. Yuan and co-advisor Dr. Zhu
(MSE) for teaching and guiding me how to do research, for their generous support and
encouragement, for the freedom they offered to me to try different things in my research.
Without their support, I would not be where I am today. For this and so much more, I am
truly grateful.
I would also like to thank Dr. Ram and Dr. Zhu (MAE) for being my committee
members and sharing their knowledge.
I am very grateful to Dr. Inoue at Shizuoka University and his group, who offered me
an opportunity to learn their unique CNT growth method and to have a two-month life in
Japan, which has broaden my knowledge.
I would like to express thanks to Dr. Henderson at Department of Chemical
Engineering, Dr. Luo at Department of Material Science Engineering for allowing me to use
their instruments.
I would like to say thank you to everyone in Zhus group, especially Philip and Xin.
They are not only my kindest co-workers, but also good friends who share their joys and tell
jokes from time to time.
The most important people in my life are without a doubt my wife, Qian, our daughter,
Grace and my parents. They are always my anchor. Whenever and wherever I am upset or
anxious, their support always eases me and raises my spirit. Thank you and I love you with
all my heart.

v

TABLE OF CONTENTS
LIST OF TABLES........

LIST OF FIGURES..

CHAPTER 1: Introduction..
1.1 Carbon Nanotube..
1.1.1 Properties of Carbon Nanotube..
1.1.2 Applications of Carbon Nanotube.
1.2 Carbon Nanotube Growth.................
1.2.1 Influences of carbon source gases on carbon nanotube growth.
1.2.2 Influences of temperature on carbon nanotube growth..
1.2.3 Influences of catalyst composition on carbon nanotube growth............
1.2.4 Influences of buffer layers on nanotube growth............
1.2.5 Influences of Ar and H
2
on nanotube growth....
1.2.6 Influences of water vapor and air on nanotube growth..
1.3 Carbon Nanotube Array....
1.4 Carbon Nanotube Fiber.
1.4.1 Carbon Nanotube Fiber Fabrication...
1.4.2 Mechanical and Electrical Properties of Carbon Nanotube Fiber.
1.4.3 Diameter Measurement of Carbon Nanotube Fibers.
1.5 Carbon Nanotube Sensor..
1.5.1 Dispersing CNT into Polymer Matrix....
1.5.2 Advantages of CNT/polymer Composite Strain Sensors ..
1.5.3 Challenges in the Application of CNT/polymer Composite Strain Sensors .
1.5.4 Durability of CNT/polymer Composite Sensors....
REFERENCES....

CHAPTER 2: Experimental................
2.1 Materials...
2.2 Catalyst Thin Film Deposition using Magnetron Sputtering....
2.3 Using Thin Film Catalyst to Grow Carbon Nanotubes.
2.4 Using FeCl
2
Powders to Grow Carbon Nanotubes...
2.5 Measuring the Quality of Carbon Nanotube Arrays using Raman Spectra..
2.6 Carbon Nanotube Fiber Spinning.
2.7 Diameter Measurement of Fibers using Laser Beam Diffraction.
2.8 Tensile Testing of Carbon Nanotube Fibers.
2.9 Electrical Property of Carbon nanotube Fibers.
2.10 Embedding Carbon Nanotube Fibers in Resin....
2.11 Micro-structure Analyze ....


vii

viii

1
1
3
4
4
7
9
10
14
15
15
16
17
18
19
21
22
27
34
36
45
49

63
63
64
66
68
69
70
71
71
73
74
74




vi

CHAPTER 3: Growing Long Carbon Nanotube Arrays using Intermetalic Fe-Zr
Catalyst..........................................................................................
3.1 Background and Motivation.
3.2 Experimental.............
3.3 Results and Discussion.
3.4 Conclusions...
REFERENCES...

CHAPTER 4: Growing Carbon Nanotube Arrays using One-Step Chemical Vapor
Deposition Method....
4.1 Background and Motivation.
4.2 Experimental.
4.3 Results and Discussion.
4.4 Conclusions...
REFERENCES...

CHAPTER 5: Carbon Nanotube Fiber Strain Sensors.
5.1 Background and Motivation.
5.2 Experimental.....................
5.3 Results and Discussion.
5.4 Conclusions...
REFERENCES....

CHAPTER 6: Conclusions and Future Work....
6.1 Conclusions...
6.2 Future Work..










76
76
80
81
93
94


97
97
100
101
109
111

114
114
117
119
132
133

135
135
137
vii

LIST OF TABLES
Table 2.1 Materials used for growing nanotubes and building carbon nanotube fiber
sensors in composites..

Table 2.2 Sputtering conditions for Fe, Fe-Zr and Al
2
O
3
films....






















63

66

viii

LIST OF FIGURES
Figure 1.1 Carbon nanotube structures defined by the wrapping directions of graphite
sheet....

Figure 1.2 In-situ TEM images of nucleation and growth of a MWNT from a nano-
particle on a substrate showing that nano-particle could change shape very quickly during
the growth. The recording time is shown in images...

Figure 1.3 Electromechanical measurements set-up for a single-walled CNT with part of
the length suspended over a trench. a, Top view of the device; b, AFM image of the CNT;
c, side view of AFM pushing experiment...

Figure1.4 Schematic drawing of electrospinning process....

Figure 1.5 Illustration of the layer-by-layer deposition of CNT-polyelectrolyte thin film
using self-assembly process....

Figure 1.6 Hysteresis behavior of the resistance/strain response in cyclic loading...

Figure 2.1 Image of a custom-built radio frequency magnetron sputtering machine
consisting of major parts of a radio frequency power supply, a cathode, a 4-way cross
chamber, a quartz balance, a turbulent pump, a mechanical pump, a mass flow controller,
a thermocouple gauge, an ion gauge, a rotating stage and a chill water cooling
system.

Figure 2.2 Image of a Fe target (left) and Zr target (right). Fe target was cut into
turbulent blade shape..

Figure 2.3 Real image of the CVD system built for CNT array growth. Major parts of the
CVD system are mass flow controllers for Ar, C
2
H
4
, H
2
gases, a quartz tube reactor and a
furnace.

Figure 2.4 Schematic of the one-step CVD system built for CNT array growth...

Figure 2.5 Image of a fiber spinning machine...

Figure 2.6 Images of the set-up for measuring the diameter of CNT fibers using laser
diffraction. (a) Set-up of laser generator, sample stage and screen. (b) Screen with scale
printed out used to collect the diffraction pattern. (c) Image of diffraction pattern in dark
room...


2



13



26

27


30

32





65


66



67

69

70




72



ix

Figure 2.7 A CNT fiber was electrically wired and mounted for electromechanical
measurement...

Figure 3.1 AFM images show the evolution of catalysts. (a) Al
2
O
3
and Fe films became
rough and bumpy when Fe catalysts were heated to 750. (b) After Fe catalysts were
annealed under H
2
and Ar for 45 seconds, further coarsening resulted in very irregular
shaped particles and an irregular distribution. (c) Al
2
O
3
islands formed when Fe-Zr
catalysts were heated to 750. (d) Fe-Zr tiny particles appeared on top of the well
aligned Al
2
O
3
grains when Fe-Zr catalysts were annealed under H
2
and Ar for 45
seconds...

Figure 3.2 Cross section view of CNT arrays grown using catalyst (a) Fe for 15 minutes,
(b) Fe-Zr with 15 minutes growth (c) Fe, a lower magnification view of A, (d) Fe-Zr, a
lower magnification view of B, (e) Fe after 15 seconds growth. All the SEM images are
taken at the roots of the CNT arrays..

Figure 3.3 SEM images clearly show the height of the CNT arrays grown using Fe-Zr
catalysts in 45 minutes...

Figure 3.4 CNT array height for different growth time using Fe and Fe-Zr catalysts.

Figure 3.5 Raman shift spectrum for CNT arrays grown using iron and iron-zirconium
catalysts..

Figure 4.1 Array height at different growth time using HCl assisted growth...

Figure 4.2 A SEM image of CNT arrays with length of 3.0 mm..

Figure 4.3 Picture of 2 mm tall spinnable CNT array showing that CNT ribbons are
drawn continuously using a tweezer..

Figure 4.4 SEM images of CNTs grown in 60 minutes (a) with HCl assistance; (b)
without HCl assistance...

Figure 4.5 Comparison of CNT arrays growing with HCl assistance and without HCl
assistance in 60 minutes growth.

Figure 4.6 In-situ monitoring of partial pressure change of H
2
and HCl during CNT
growth using QMS..

Figure 5.1 Repeatitive testing of neat CNT fibers showing repeatable stress-strain and
resistance-strain behaviors up to 1% strain: a) tensile stress-strain curve, b) in-situ
resistance measuring during the tensile loading. Neat fibers underwent loading and
73







84




85


89

90


92

102

103


103


107


108


109








x

unloading sequentially to strains of 0.5%, 0.8%, 1% and break, respectively and the
resistance was measured in-situ. After each loading, the fibers were completely unloaded
before the next laoding...

Figure 5.2 CNT fiber spinning and deposition of Al electrodes on CNT fibers. a) SEM of
directly spinning CNT fibers from as-grown CNT array. b) low magnification SEM of an
as-spun fiber. c) schematic drawing of Al electrodes deposition on dry CNT fibers using
magnetron sputtering. d) low magnification SEM of a fiber end coated with Al...

Figure 5.3 Stability testing of fiber sensors under zero-load conditions showing that the
resistances were very stable at three different working temperatures: immersed in liquid
nitrogen, at room temperature and annealed at 393k in an oven. The measurement was
made in a continuous mode.

Figure 5.4 In-situ monitoring of strain and crack propagation in a composite structure. a)
Linear increase of resistance when the composite structure was under longitudinal tension
and neglectable increase under transverse tension; b) crack propagation was easily
monitored by building multiple CNT fibers in the composite structure.

Figure 5.5 Zoom-in view of the crack propagation causing the failure of the first CNT
fiber.

Figure 5.6 Zoom-in view of the crack propagation causing the failure of the second CNT
fiber.




120




121




126




130


131


132


1

Chapter 1: Introduction
Carbon nanotubes have attracted many researchers interests since CNTs were first
discovered by Dr.Iijima in 1991.
1
With their unique seamless tube structure, individual
CNTs possess many outstanding properties. Thus, extensive research has been done on
growing carbon nanotubes and attempting to apply these properties in real applications such
as high performance composites, electrodes for lithium batteries and super capacitors,
especially strain sensors.
2-23
CNTs have been regarded as a new type of materials for real-
time structural health monitoring due to their lightweight and good mechanical properties.
Thus, a series of work, including growing long and high quality CNT samples, processing
CNTs into sensors become important to utilize their sensing properties in real world. A
detailed literature survey of properties, growth, applications of carbon nanotubes will be
present in the following text.
1.1 Carbon Nanotube
Carbon Nanotube is one type of allotrops of carbon. It has a hollow cylinder structure.
As shown in Figure 1.1, Single-walled carbon nanotubes (SWNTs) can be considered as a
seamlessly rolled-up graphene sheet, while multi-walled carbon nanotubes (MWNTs) are
2

close tube structure with multiple layers. Three different types of CNT structures, zigzag,
chiral and armchair structures, can be formed based on the coordination of carbon atoms in
the crystal structure. Two integer indexes (n, m) and two basis vectors (a
1
, a
2
) for a graphite
sheet are used to define the wrapping, thereby the CNT structures, which is shown in Figure
1.1. a
1
and a
2
are two unit vectors with different directions. If the two indexes are identical
then the CNT has an armchair structure. If (m=0) then the CNT is considered as zig-zag
structure. Other wrapping cases produce CNTs with chiral structures.

Figure 1.1 Carbon nanotube structures defined by the wrapping directions of graphite sheet.
3

1.1.1 Properties of Carbon Nanotube
Nanotubes have many outstanding properties that are superior to other materials. For
example, the length-to-diameter ratio of nanotubes can be up to 1.3210
8
,
24
which is the
largest in currently available materials, and this ratio could be tuned based on the
requirements. The theoretical Yougs modulus of individual CNTs is around 1TPa and
tensile strength of single CNT is around 100 GPa.
25,26
The experimental value of Yougs
modulus and tensile strength of CNTs are around 0.8TPa and 60GPa,
27,28
which are lower
than theoretical value because in reality, defects exist in the CNT structure. Considering the
low weight density of CNTs, which is around 1.3g.cm
-3
,
29
the specific Yougs modulus and
strength are extremely higher than other materials. Therefore, scientists predict that CNT is
the only material among the currently available materials that wont fracture itself by its own
weight when used for building a space elevator from the outer space to the earth.
30
CNTs
also have a large electric current density as high as 410
9
A.cm
-2
and the conductivity of
CNTs range from 200 to 10
5
S.cm
-1
.
31,32
Thermal conductivity of nanotubes is around 3500
W.m
-1
K
-1
, which is much better than copper.
33
Nanotubes are also good piezoresistors. The
4

conductivity of CNTs changes dramatically when subjected to tension. The gauge factor
ranges from 200 to 1000 depending on the CNT structures.
34,35

1.1.2 Applications of Carbon Nanotube
Due to their unique properties described above, numerous potential applications of
CNTs have been proposed and studied in recent years. In combination of mechanical,
electrical and thermal properties, nanotubes could be applied in many fields, such as
transistors and emitters,
36-44
thermal acoustic instruments,
45
electrodes for lithium ion
batteries, fuel cells and supercapacitors,
9-12,46-50
composite materials for wind turbine (Bayer
Material Science), bullet proof vests, body armors (Block Textiles, Inc), atomic force
microscope tips,
51
chemical/biochemical sensors and strain sensors.
14-23,52-54

1.2 Carbon Nanotube Growth
As the technique for the real application of CNTs in industrial has already developed to
a new level, the cost of production of CNTs in sizable quantity becomes a major concern.
Today, tremendous efforts have been made on pushing CNTs from laboratory trials to
industrial mass production. Several companies, including Nanolab, Carbon
Nanotechnologies, Inc., Firstnano, Catalytic Materials, Hyperion, Nanocyl, SouthWest
5

Nanotechnology, have already supplied large amount of CNTs to the world. However, the
annual production of CNTs are only 36 500 metric tons in 2004, which is much lower than
carbon black at 9 600 000 metric tons.
55
Among the most popular methods used for growing
nanotubes, including arc discharge, laser ablation, and chemical vapor deposition (CVD),
CVD is now considered as the most efficient way for industrial yields of CNTs where low-
cost, large quantity and high quality could be achieved simultaneously. However, qualities of
CNTs produced by arc discharge or laser ablation are higher than CVD method. For
applications, such as single tube emitters, diodes, atomic force microscopy tips et al., where
CNTs with high quality are necessarily required but only a small amount of CNTs are
needed, arc discharge or laser ablation is preferred.
The principle of arc discharge or laser ablation is that carbon free atoms are generated
at a localized high temperature around the carbon electrodes or carbon targets by supplying a
large electric current or using a strong laser beam. The free carbon atoms will form CNTs
and other carbonaceous byproducts at the cool environment.
56
Due to the requirements of
large electric current, high vacuum, frequent changing of carbon electrodes and targets, arc
discharge and laser ablation are high-cost, difficult and expensive to scale up. Therefore
6

these two methods could not meet the needs of large quantities of CNTs in the growing
industrial market.
Chemical vapor deposition has shown the versatility and flexibility in growing carbon
nanotubes. Thousands of combination can be selected for growing CNTs from a large
variety of CVD designs, growth temperatures, carbon source gases and catalysts. Many types
of CVD have been successfully used to grow CNTs with relatively large quantities,
including fixed beds (catalysts are fixed on substrates),
57-60
fluidized beds (catalyst particles
are agitated by the flowing gas and filled with the whole space of the reactor),
61-63
aerosols
(catalysts are made into aerosols and are injected into the reactors),
64,65
floating catalysts
(catalysts are floating instead of being fixed on substrates).
66,67
Temperature used to grow
CNTs using CVD methods ranges from 450 to 1000C depending on the selection of other
parameters.
68-72
A variety of carbon source gases, including methane, carbon monoxide,
ethylene, acetylene, ethanol, propylene, ferrocence, benzene, xylene, propylene, n-pentane et
al., have been used to supply free carbon radicals for nanotube growth.
73-79
Transition metals,
metal oxides, and their alloys such as Fe, Co, Ni, Mo, Pt, Fe
2
O
3
, Fe-Co, Fe-Mo, LaFeO
3
,
CuNi, Co-V et al. have been successfully used for catalyzing the disassociation of carbon
7

source gases to grow CNTs.
80-90
Catalysts are often deposited on stable buffer layers on
silicon wafers including Al
2
O
3
, MgO and SiO
2
.
91-98
Ar and H
2
are often used to moderate the
CNT growth.
84,85
A small amount of water vapor and air are usually supplied to the CVD to
assist the growth.
99-103
With so many parameters available for choosing, it is important to
understand the influences of these variables on the CNT growth.
1.2.1 Influences of carbon source gases on carbon nanotube growth
Hundreds of CNT precursors have been successfully used for CNT growth. Different
carbon source gases require different temperature to break down C-H bonds and C-C bonds
to release free carbon radicals. A general rule is that triple C-C bonds are easier to break than
double C-C bonds and double C-C bonds are easier to break than single C-C bonds. Thus,
temperature used to grow CNTs for different CNT precursors is C
2
H
2
>C
2
H
4
>CH
4
.
104

However, the temperature might be different depending on other parameters such as type of
CVD systems, activity of catalysts. Generally speaking, the more thermally stable the CNT
precursors are, the higher temperature is required for CNT growth.
The morphology of CNTs could be affected by the molecular structure of CNT
precursors. CNTs produced from linear hydrocarbons such as methane, ethylene, acetylene,
8

are usually straighter than cyclic hydrocarbons such as benzene, xylene, fullerene and
camphor.
105
It is possible that the long zigzag carbon bone chains construct CNTs as a basic
building block without breaking into single free carbon radicals. Herein, the molecular
structure of CNT precursors is somewhat maintained in CNTs.
It has been proved that high quality CNTs could be produced from precursors
containing oxygen atoms such as ethanol.
106
CNTs are almost free of amorphous carbon due
to the etching effect of hydroxide groups.
106
However, ethanol is liquid at room temperature
and could not directly delivered to the reactor like other gases. Supplying ethanol to the
CVD system has been accomplished by flowing a carrying gas through the liquid ethanol to
pick up some vapors. It is believed that long CNT arrays could be grown using ethanol since
the CNTs are much cleaner and catalysts are expected to be active longer because of little
amorphous carbon produced during the growth. However, no paper has been reported on
long CNTs grown by ethanol.
Organometallic compounds such as ferrocene, cobaltocene, nickelocene have special
structures containing metal atoms in each molecular. Free metal atoms are released to form
nano-catalyst particles during the decomposition of these compounds. Therefore,
9

organometallic compounds are self-catalyzing and have a high efficiency in growing carbon
nanotubes. Nanocom has fabricated 3-D spider-web like carbon nanotube network using
ferrocene precursors and these CNT webs could be made into CNT sheets with nanotubes
aligned randomly. High Performance Composite Center in Florida State University has
produced high volume fraction carbon nnaotube composites from these CNT sheets.
107

Composites fabricated from these functionalized CNT sheets have both Youngs modulus
and tensile strength much higher than any other currently available carbon fiber reinforced
composites.
1.2.2 Influences of temperature on carbon nanotube growth
Temperature has a strong influence on the decomposition of hydrocarbon compounds.
The decomposition of hydrocarbon precursors often requires a high temperature. The higher
the growth temperature is, the faster the carbon sources gas decomposes. If the rate of
producing free carbon atoms is larger than the rate of carbon nanotube growth, extra carbon
atoms will precipitate into amorphous carbon. In other words, there is a rate limitation of
carbon radicals transferring into crystallized CNTs. Too much carbon supplying will
extinguish the growth by forming a amorphous carbon layer on the surface of catalysts.
10

Herein an optimized growth temperature is usually used for a certain type of CNT precursor
and catalyst to moderate the decomposition rate and growth rate.
108

Temperature also has a strong effect on the structure and morphology of carbon
nanotubes. It has been found that SWNT usually grows at high temperatures above 900C
and MWNT usually grow at a relatively lower temperature.
109,110
There is an argument on
the influence of temperature on CNT diameters. A couple of groups reported that the outer
diameter of CNTs increased as the increase of growth temperature.
110
However, Li and Zhu
observed the decrease in diameters as the growth temperature increases.
109
Zhu also found
that CNT grown at higher temperature have better quality than that grown at lower
temperature.
109
The contents of amorphous carbon and defects in CNTs decrease as the
growth temperature goes up.
1.2.3 Influences of catalyst composition on carbon nanotube growth
Many types of catalysts have been used for catalyzing the decomposition of carbon
precursors.
80-90
The purposes of using catalysts are decreasing the hydrocarbon
decomposition temperature and help arrange the decomposed carbon radicals to form tube
structures. By using catalysts, several groups have successfully grown CNTs at temperatures
11

as low as around 550C.
108
Among the tenth of hundredth of metal catalysts, Fe, Co and Ni
are the most widely used catalysts due to their high efficiency and low costs. The order of
catalyzing activity of these three transitional metals is Ni>Co>Fe.
111
Thus Ni is usually
chosen for catalyzing methane which is more thermally stable than ethylene and acetylene
and requires higher temperature to break down the molecular. And Fe is often used for
decomposing ethylene and acetylene.
Bimetallic catalysts consisting two or more elements have been developed for purposes
of moderating the catalysts activity and thermal stability. Alvarez
112
found that in a Co-Mo
catalyst system, single-walled CNTs are often produced with a high concentration of Mo in
the bimetallic catalyst, while multi-walled CNTs are grown using Co alone. Mo is inactive at
the growth temperature they used. Thus adding Mo to Co decreases the effective mass of
catalyst in the bimetallic system. The morphology of the CNTs grown has a relationship with
the catalyst activity in Co-Mo system. Zhu et al.
84,85
observed the composition of Co-Mo
particles under high resolution TEM. They found that Catalyst particles with 515 at.% Co
tend to produce long CNTs, those with 4045 at.% Co tend to produce short CNTs, and
those with 8598 at.% Co tend to produce onionated morphology.
12

It has been demonstrated that catalyst particles are very mobile and can change shape
quickly during the CNT growth.
113
The in-situ TEM observation of catalyst particles confirm
that metal particles are in liquid or quasi-liquid status. As shown in Figure 1.2, The Fe
particle was observed to keep changing shapes very quickly as CNTs grew longer.
Considering the mobility of metal particles at high temperature, it is not difficult to speculate
that the particles tend to grow larger and larger due to the Ostwald Ripening effect and
finally lost their catalyzing capabilities. It is expected that the bimetallic particles are more
thermally stable because the dragging interaction between different elemental atoms because
of the partial covalent bonding between them. Bimetallic (Fe-Pt) catalyst was observed to
remain static on the substrate as the CNTs went on growing up confirming the better thermal
stability.
82


13


Figure 1.2 In-situ TEM images of nucleation and growth of a MWNT from a nano-particle
on a substrate showing that nano-particle could change shape very quickly during the growth.
The recording time is shown in images (taken from Yoshida et al.
113
).

14

1.2.4 Influences of buffer layers on nanotube growth
It has been found that metal catalysts supported on a stable oxide layer have much
better performance in growing CNTs than catalysts on silicon substrates.
91-98
Al
2
O
3
, MgO
and SiO
2
have been considered as three types of stable oxides that are suitable for carbon
nanotube growth. Although the fundamental mechanism of this supporting effect has not
been established since no solid evidence has been found, a couple of assumption has been
made. One assumption is that the buffer layer could prevent the chemical reaction between
the metal catalyst and silicon substrates which are widely used as substrates in the fixed-bed
CVD.
114
Most metal catalyst, especially the most widely used Fe nanoparticles, could react
with Si substrates at the CNT growth temperature. The buffer layer between metal particles
and Si substrates physically separates them and prevents the chemical reactions. Another
speculation is that metal particles tend to appear in sphere shape on buffer layers which is
beneficial for CNT growth.
95,114
Metal catalyst has higher surface tension on oxide layer than
Si substrates. In order to reduce the surface energy, metal particles will have round shape
with a large contact angle with the oxide layer, which help the uniform distribution of
catalyst.
15

1.2.5 Influences of Ar and H
2
on nanotube growth
Argon and hydrogen gases are often used to grow CNTs together with carbon
precursors.
70,84,85
The function of argon is to dilute the carbon feedstock gases herein
moderating the decomposition rate of hydrocarbons. By decreasing the partial pressure of
carbon source gases in the CVD system, the probability of carbon precursors colliding with
catalyst particles decreases. With less hydrocarbon decomposed, less amorphous carbon is
expected with a sacrifice of growth rate. H
2
is often used to reduce the oxidized metal
catalysts before CNT growth starts. During the growth, the continuous purging of H
2
might
help to remove part of amorphous carbon. It is reported that a high hydrogen flow rate could
suppress carbon formation.
70
The evidence supported their assumption was that SWNTs
were usually grown with a higher hydrogen flow rate while growing MWNTs requires a
lower hydrogen flow rate.
1.2.6 Influences of water vapor and air on nanotube growth
Adding a small amount of water vapor or air into the CVD system is beneficial for CNT
growth. It is believed that water vapor and air are weak oxidizers that could transform
amorphous carbon into carbon oxide, which sustained a longer growth time since less
16

amorphous carbon is accumulated.
99-103
Accumulation of amorphous carbon on catalyst
particles is regarded as one reason that stops CNT growth. By supplying a finely-controlled
water vapor to the CVD system, Dr.Ijimas group grew 2.5 mm tall SWNT arrays in 10 min.
The fast growth rate and increased height of CNTs resulted from the water vapor that
dramatically enhanced the catalyst activity and lifetime. Following the water-assisted growth,
another group explored the role of water in super growth of single-walled carbon nanotube
carpets. They found that the present of oxygen and hydroxyl groups could reduce the
diffusion rate of catalyst atoms. Therefore Osawald ripening was inhibited and catalyst
particles could maintain their shape and size during the growth.
103
Ajayan et al.
100
produced
1.5 cm tall CNT bundles using air-assisted growth in 15 hours. Such a long catalyst active
time attributed to the ability of air to remove the amorphous carbon. With air injected in the
system, lengthening process continued and thickening process was inhibited.
100

1.3 Carbon Nanotube Array
Carbon nanotube array is also called carbon nanotube forest which is often vertically
grown on a piece of substrate using fixed-bed CVD method. As-grown CNTs in the array are
all aligned in the vertical direction. This alignment is beneficial for applications in
17

fabricating field emission devices. The alignment can be also transferred into horizontal
direction after shear pressing the array. The vertically grown array is a precursor for making
composites with CNTs all aligned in one direction rendering good electrical and mechanical
properties in the aligned direction.
The height of CNT array varies from a couple of micrometers to centimeters.
61-103
It is
very obvious that the taller the CNT array people can ever grown, the better the mechanical
and electrical properties of the composites made from these arrays will be. However, the
height of the array is still limited at the range of millimeters. The current record of longest
CNT array (18 mm) was set by a research group in Cincinnati University. Growing meters
long CNT array is still a challenging work that is under research by many groups.
1.4 Carbon Nanotube Fiber
A carbon nanotube fiber is a CNT assembly in the form of fibers consisting of solely of
nanotubes. There are two major methods to fabricate CNT fibers. One is winding the aerogel
like CNTs growing in the hot furnace.
115
The other method takes advantage of the unique
feature of special arrays in which CNTs are close to each other and have optimized
entanglement allowing them to be pulled out continuously one after another.
116,117
CNTs in
18

the fiber are buckled together and have good contact with each other which renders the fiber
with good mechanical and electrical properties.
1.4.1 Carbon Nanotube Fiber Fabrication
Li et al.
115
first developed the winding method by spinning CNT fibers from the CNT
aerogel in the hot zone of CVD system. The main advantage of this method is that the fiber
spinning process is continuous. Gas precursors are continuously flowing into the reactor
allowing a continuous CNT growth. CNTs are blown into a gas phase suspension and are
continuously wound up by a rotating needle. Zhang et al.
117
developed a method downsizing
the traditional textile technique for spinning fibers. CNTs are first drawn continuously from
as-grown arrays and then spun into fibers with diameters of several micrometers. A research
group in Los Alamos Lab produced millimeter long drawable arrays from which fibers can
be spun
116
. Fan et al
118
in Tsinghua University grew spinnable arrays on a large scale on a
piece of 4-inch diameter silicon wafer. Large sheet of CNTs could be drawn out of the array
and spun into fibers. Recently a group in Shizuoka University developed a one-step CVD
method for producing drawable CNT forests on quartz substrates.
119
This method does not
require the step of catalyst pre-deposition on substrates as most traditional CVD methods do.
19

Thus the production of such special arrays is faster and could meet the industrial needs if
scaled up.
It has been not achieved to an agreement on why CNTs could continuously drawn from
the vertically grown CNT array. Zhang et al.
117
claimed that CNTs are all connected at the
ends at the top and the bottom of the arrays by amorphous carbon layers. However, another
group showed that their spinnable arrays are free of amorphous carbon layers at the top and
bottom.
118
Instead, they assumed that the strong Van der Waals force between CNTs
allowing that when one CNT is pulled out, their closest neighbors are taken. Nevertheless, it
is quite challenging to produce such drawable arrays. A certain range of array height has to
be controlled for producing fibers. Even with the same growth parameters, if the array is too
short or too tall, no ribbons could be pulled out continuously any more.
1.4.2 Mechanical and Electrical Properties of Carbon Nanotube Fibers.
It is well known that CNTs are extremely strong, electrically conductive. Fibers
composed of pure CNTs are expected to be high strength, lightweight, thermally and
electrically conducting materials. However, fibers made from short CNTs are not as strong
as single nanotubes since there are millions of joints between CNTs. These joints are weak
20

and fracture usually starts from peeling off of CNTs from each other. SWNTs were
dispersed in sulfuric acid and extruded into fibers by Smalley.
120
The SWNT fibers have a
Youngs modulus of about 120 GPa and tensile strength of about 116 MPa. The electrical
resistivity of the fiber is around 0.2 milloohm.cm. Fibers would up from hot furnace by
Windle have a wide range of strength depending on process parameters, from 100 MPa to 1
GPa. The electric conductivity of the fiber is around 8.310
5

-1
m
-1
.
115

Fibers were directly spun from vertically grown long CNT arrays by Zhu et al.
116
They
found that the strength of fibers could increased by a post twisting due to the increased CNT
packing density. They also found that the strength of fibers increased as the length of CNT
elements increased and the diameter of fibers decreased. The highest strength of the fibers
achieved 1.91 GPa with Youngs modulus of 330 GPa. After a post-twisting the electric
conductivity increased from 1.710
4
S m
-1
to 4.110
4
S m
-1
. Another group spun fibers from
1-mm long CNT forests grown using water-assisted CVD. After densifying and post twisting
the fibers, a tensile strength of around 550 Mpa was achieved. The fibers have an electric
conductivity of 510
4
S m
-1
.
121
The effect of strength enhancement by solvent densifying
treatment and post-twisting were also proved by Fan et al. As-spun fibers were annealed in a
21

furnace after passed a solution. A further post-twisting was also performed. Fibers have a
tensile strength of 1.1 Gpa. Electric resistivity ranges from 1.210
-5
m

to 1.810
-5
m.
122

1.4.3 Diameter Measurement of Carbon Nanotube Fibers
Cross section area of a fiber is a major physical parameter since tensile strength,
modulus and electric impedance are all closely relate to it. An error in measuring the cross
section area will lead to a magnification of a factor of 2 in other properties, such as strength,
modulus, conductivity. Carbon nanotube fibers people ever made have a diameter ranging
from 4 m to 20 m.
115-122
At this scale, it is not accurate enough to use a micrometer to
measure the diameter. Although, SEM is a reliable tool to directly measure the diameter,
using SEM is time consuming and costly. Laser beam diffraction method has been
successfully used to measure the diameter of fibers and carbon fibers.
123
It is believed that
laser diffraction technique is not only fast but also capable of meeting the accuracy
requirements.
Laser diffraction technique is based on Fraunhofer diffraction and Babinets principle.
Simply, Fraunhofer diffraction is a wave diffraction phenomenon when parallel waves pass
through an aperture, they split into double number of waves and interfere with each other
22

leaving a image pattern if a screen is placed in its path. Babinets principle describes that the
diffraction patter obtained from an aperture is exactly the same to the one from a fiber that
has the same size of the aperture.
The set-up of the laser diffraction technique is quite simple. Usually a convergent laser
beam is generated from a beam generator. The beam passes through the fiber placed at the
focal point. A diffraction pattern will be collected on a screen. Based upon the laser
wavelength, the fiber-to-screen distance and spacing between dark and bright diffraction
spots, fiber diameters could be readily calculated.
1.5 Carbon Nanotube Sensors
Carbon nanotubes (CNTs) have been found to have excellent piezoresistive properties,
which may enable a new generation of sensors in nano or micro scales.
124-128
Piezoresistivity
is by its very nature a phenomenon defined by the changing electrical resistance of a material
due to applied mechanical stress. Unlike the piezoelectric effect, piezoresistive effect only
causes a change in resistance but not vice versa. The studies on the piezoresistance of
individual CNTs provide some insights in exploring the mechanisms why CNTs possess this
peizoresistive property.
23

Some challenging pioneering measurements have been conducted in the past few years
on the conductance change of individual CNTs under mechanical stress.
124-128
Tombler et
al.
124
prepared samples of individual single-walled CNTs with two ends covered by metal
electrodes on SiO
2
/Si substrates. CNT length of about 605 nm was suspended over SiO
2

trenches. (Figure 1.3) An AFM tip was used to manipulate the CNTs and the resistances
increased two orders of magnitude when CNTs were deformed. The changes of the
conductance were probably due to the local structure change of the CNTs. Grow et al.
125

measured the piezoresistance of CNTs adhered on deformable silicon nitride membranes and
they found that semiconducting and small-gap semiconducting CNTs had gauge factors of
up to 400 and 850, respectively. Three types of CNTs, metallic, quasimetallic and
semiconducting CNTs were systematically investigated on the electromechanical properties
under tensile strain by uniformly stretching the CNTs that are suspended on a
microelectromechanical machine.
126
In comparison of these three types of CNTs,
quasimetallic CNTs were found to have the largest piezoresistance gauge factors in the range
of 600 to 1000. Minot et al.
127
studied the changes of band gaps with strain by stretching
24

CNTs using an AFM tip. An equation was given describing the the total resistance of the
CNT (R
tot
). Where R
s
is the contact resistance, |t|
2
is the transmission probability of
2 2
1
1 exp
8
gap
tot s
E
h
R R
e kT
t
( | |
= + +
( |
\ .
(1)
electrons that cross the band gap barrier, T is the temperature, h, k, e are constants and E
gap
is
the band gap energy. The equation describes the changes of resistances when CNTs are
under stretching. The equation also points out the exponential decaying of the resistance as
the temperature increases. Single-walled CNT bundles were pressed in the radial direction to
study the piezoresistive properties.
128
More complicated than a single CNT, the resistances
of each CNT bundles changed differently when pressed, which was explained by the
different peizoresistive responses of each individual CNTs in the bundles.
Numerous methods have been applied to fabricate nano composites for sensing strain.
These methods can be categorized as three types based on the mechanism used for strain
sensing. For the first type, an electrospinning method was employed to fabricate CNT-based
piezoelectric strain sensors.
129
Polyvinylidene difluoride-trifluoroethylene (PVDF) was
dissolved in N, N-dimethyformamide (DMF) and mixed with predispersed CNTs in DMF
solution. The mixture was then transferred to a syringe and electro spun into non woven
25

fabric on a plate using an electrospinning machine which was schematically drawn in Figure
1.4. After a poling process, CNT/PVDF thin film piezoelectric strain sensors were prepared
and the strain-sensing ability of the resulting sensors was increased 35 times by adding 0.05
wt% of CNTs. Although the piezoelectric properties of the CNT/PVDF composites are
promising, the application of CNTs in piezoelectric sensing is very limited since the
properties are mainly come from the matrix but not the CNTs. Raman bands of CNTs were
observed to shift as the CNT incorporated composites were strained.
130,131
The second type
of CNT strain sensors takes advantages of the Raman-active properties of CNTs. However,
the inconvenience of using huge Raman spectra machine in examining the structural healthy
restricts the real application since it is difficult to realize in-situ healthy monitoring. The
third type of strain sensors takes advantages of the piezoresistive properties of CNTs. A
certain amount of CNTs were added into the polymer matrix to form a 3-D electric path way.
The force applied on the composites transfers to the CNTs that will be deformed under loads,
which leads to resistance change. By measuring the resistances of the composite sensors,
load conditions of the composites could be detected.


26


Figure 1.3 Electromechanical measurements set-up for a single-walled CNT with part of the
length suspended over a trench. a, Top view of the device; b, AFM image of the CNT; c,
side view of AFM pushing experiment (taken from Tombler et al.
124
).

27




Figure1.4 Schematic drawing of electrospinning process (taken from J ialili et al.
129
).

1.5.1 Dispersing Carbon Nanotubes into Polymer Matrix
Extensive research has been done recently on CNT/polymer strain sensors based on the
piezoresistive effect. Thin film sheet with carbon nanotube randomly oriented have been
fabricated and linear resistance changes in the sheet was found when subjected to tension or
compression.
21
However, the test was taken at very low-tension-compression cycles (the
maximum strain was only 0.04%), indicating the very limited sensing range. Multi-walled
carbon nanotube-based composites have been reported to be 3.5 times higher in sensitivity
28

than the traditional strain gage.
132
Kang et al.
15
dispersed the commercially obtained SWNTs
in dimethy formamide (DMF) solvent in a bath sonicator for 20 hours. The SWNTs
suspension was poured into a filter paper and dried in an oven for 12h. After that the bulky
paper was peeled off and characterized by electrochemical impedance spectroscopy. To
improve the consistence of piezoresistence, PMMA was added to the SWNTs suspension.
The CNT/PMMA strain sensor exhibited linear piezoresistivity up to strain of 1400 m/m.
They concluded that the CNT/PMMA composite is a low cost strain sensor for structure
healthy monitoring. Randomly tangled multi-walled carbon nanotube buckypapers were
tested under uniaxial tensile load-unload cycle for an evaluation of working as strain
sensors.
20
Thin copper layers were thermally deposited on both sides of the thin film. It was
found that the resistance change of the MWCNT film was proportional to the applied strain.
Results also showed that the resistance was temperature independent over the range 237-363
K. Drifting was observed when the samples were loaded and unloaded for many cycles. The
resistance-strain curves were not repeatable indicating the slipping of CNTs in the
buckypaper.
29

Multifunctional carbon naontube-polyelectrolyte thin films were prepared using a layer-
by-layer method.
14
The process is illustrated in Figure 1.5. A very long sonication process
was applied to disperse carbon nanotubes in a high-molecular-weight poly (sodium4-
styrene-sulfonate) polyanion solution. It is found that the strain sensitivity of the macro-scale
multilayer film was increased as the CNTs concentration increased. The CNT-
polyelectrolyte film displayed time-variant exponential decay of bulk film resistivity. A
higher current caused a faster resistance decaying. Loh et al.
133
studied a carbon nanotube
nanocomposite piezoresistive sensing skin fabricated by a layer-by-layer self-assemble
method. Two-dimensional strain and crack monitoring was achieved by combining with
electrical impedance topography conductivity mapping technique. The versatility of the
layer-by-layer thin film fabrication methodology allows film deposition on virtually any
surface and these low-cost 2-D thin film sensors are validated as sensing skins for
monitoring crack growth in cementitious composites and impact damage on thin aluminum
sheet. Electrical impedance topography spatial conductivity maps can identify crack growth
during an applied tensile-compressive cyclic loading by mounting the sensing skins to the
cementitious composites using epoxy. When the sensing skin was directly deposited on the
30

aluminum surface, the two-dimensional strain distribution of the aluminum plate was
captured.

Figure 1.5 Illustration of the layer-by-layer deposition of CNT-polyelectrolyte thin film
using self-assembly process (taken from Kenneth et al.
14
).
Multiwalled nanotubes were dispersed into the epoxy resin using a calendaring
approach to fabricate nanotube-epoxy-fiber composites.
16
The percolating network of
sensors was created and the cured composites laminates were machined into test samples for
sensing measuremnts. By combining micrometer sized glass fiber and nanometer scale
31

carbon nanotubes, the carbon nanotube can transfer the stress more effectively by filling the
areas between matrix and fibers. The initial stages of epoxy-dominated failure are detected in
situ by monitoring the strain of the composite and the resistance change at the same time.
Experiments are designed for different failure modes and the nature and progression of the
damage are identified. In another paper, Thostenson and Chou
18
presented the resistance
hysteresis for a specimen undergoing 25 cycles of strain loading prior to fracture. The
remarkable shift in sensing curve indicates the irreversible bonding damage between the
matrix and the carbon nanotubes (Figure 1.6). The carbon nanotube sensor network showed
some degree of failure after two cycles of loading, indicating the very limited accurate strain
sensing range. The permanent damage of the carbon nanotube network caused the permanent
resistance increase of the composites. The elimination of resistance hysteresis caused by the
slide of carbon nanotubes during the loading is a great challenge, which limits the real
application of the carbon nanotube-based strain sensor.
32


Figure 1.6 Hysteresis behavior of the resistance/strain response in cyclic loading (taken
from Thostenson et al.
18
).
Knite et al.
12
attempted to use the multi-wall carbon nanotube to develop a new
nanocomposite for sensing strain. CNTs were mixed in polyisoprene matrix and the mixture
was vulcanized under high pressure. The specimens were tested under small strain and large
strain. A relatively large hysteresis of resistivity was observed when the samples were 40%
stretched and released, while the resistivity was very reversible under strain of 4%. Pham et
al.
13
demonstrated a new type of thermoplastic/carbon nanotube films for strain sensing. The
33

surface resistivity of the thin films increased proportionally with increasing tensile strain,
which was explained by three mechanisms: firstly, the stretching caused a loss of contact
points and a widening of inter-nanotube distances leading to the decrease of current-carrying
ability of the conductive network formed in the films, herein the resistances increased;
secondly, contact resistance between CNTs was increased due to the distortion of the contact
region under tensile strain. Another possible contributing factor for the rise in resistance is
the decrease of conductance of the CNTs under stress. Loads applied on the films transferred
to the CNTs and caused mechanical deformation of the CNTs, which decreased the
conductivity of the CNTs. Thus, the total resistance of the films increased since the electric
path way in the films created by individual CNTs was distorted. The three mechanisms
proposed by Pham may be applicable to explain the resistance rise under strain for most of
the CNT/polymer composites strain sensors since they have three common features: (1) the
fabrication process involves the dispersing of CNTs into the polymer; (2) the electric path
way is formed by the 3-D network of CNTs; (3) the strain sensing ability originates from
piezoresistive properties of CNTs.

34

1.5.2 Advantages of CNT/polymer Composite Strain Sensors
Compared with the traditional commercial strain gauges, the CNT/polymer composite
strain sensors are superior in several aspects. (1) Versatility and flexibility: the commercial
available resistance-type strain gages lack versatility and flexibility because they can only
measure strains at the locations where these sensors can be bonded, while CNTs/polymer
sensors can be fabricated into a variety of shapes, sizes and even could be embedded into the
structures to enable the structure have self-sensing properties. (2) Tunable gage factors: CNT
composite sensors exhibit tunable resistance-strain sensitivity by simply adjusting the
fabrication recipes, such as, the amount of CNTs added into the polymer, degree of nanotube
dispersion and even types of polymers selected as the matrix. The gage factors can be
tailored to meet different requirements in industry. (3) In-situ healthy monitoring: real-time
structural healthy monitoring can be realized through embedded-in CNT network sensors.
With excellent mechanical and super lightweight properties, CNTs are non-invasive to the
structure and have little weight punishment when embedded in the structural components.
The 3-D electric path-way formed in the structural components provides the in-situ self-
sensing capability of the structure. The external loads the structures are undergoing could be
35

detected via the real-time resistance monitoring. (4) Detecting micro cracks: microcracking
in advanced fiber composites can occur at relatively low strains as compared to the ultimate
fracture strain of the composite laminate, and the development and progression of
microcracks have significant implications on the long-term durability and performance of the
structure. To detect micro cracks in the structure, philosophy logic is that accuracy could be
achieved only if nano sensing elements are sufficiently distributed in the structure. CNTs,
with diameters in the range of several to one hundred nanometers and length varying from
micrometers to millimeters, are capable of monitoring micro cracks when built in the
structure. (5) Monitoring the process of structure failure: in a resistance-strain testing cycle,
smooth and gradual increase of resistance indicating a slow increase of strain without
structure failure, while sharp increase in resistance reflects crack formation or delamination
of plies followed by crack propagation until the failure of the whole structure. Based on the
characteristics of resistance change, types of failure in the structure could be detected. (6)
Overall structural integrity evaluation: to achieve structural integrity in the whole aircraft or
large parts, inconvenience in using traditional strain sensors is that huge amount of strain
gages are needed to bond on the structure, which in addition will add much weight on the
36

plane. However, the ultra lightweight nanotubes sensors are readily built in widely used
structural composites, to access an overall health monitoring of the structure.
With the fast-growing use of fiber-reinforced composite materials in both military and
industry, there is an urgent need of a new generation of sensors that possess the above
discussed merits to realize real time structural health monitoring for maintenance purposes
as well as preventing catastrophic failure. CNT sensors, with unique mechanical, electrical
and lightweight properties, exhibit their sensing capabilities and the potentials to serve as
this required new type of sensors. Although extensive research has been done on developing
CNT/polymer strain sensors, challenges still exist and more work has yet to be done for a
mature technique in using CNT sensors.
1.5.3 Challenges in the Application of CNT/polymer Composite Strain Sensors
The results shown by the past research are very promising in utilizing CNT composites
for strain sensing. However, the past research also show several issues in attempting to
develop the new type of CNT strain sensors.
The electric conducting mechanism of CNT-based composites are so complicate that
challenges exist in fabricating CNT/polymer strain gages with consistent resistances. Due to
37

the extremely low electrical conductivity (in the order of 10
-10
-10
-15
S/m), polymeric or
ceramic, which is often used as matrix in a composite material, is usually regarded as a non-
conductive material. Conductive composites can be fabricated by dispersing a certain
amount of conductive materials into the non-conductive matrix. The volume fraction of the
conductive phase in the composite material determines its electrical conductivity. The
conductivity of the composites is dominated by the conductivity of the pure non-conductive
matrix when the conductive materials added into the matrix are of low volume fractions. The
conductivity of the composite increases drastically by many orders of magnitude after the
conductive material reaches a critical volume fraction. This phenomenon is termed as
electrical percolation. The size, structure, shape and purity of the conductive reinforcements
surrounded by the insulating matrix can greatly affect the electrical percolation threshold of
the composite. The percolation threshold can be lowered by embedding small size and large
aspect ratio conductive materials, such as carbon nanotubes, which have very large aspect
ratios of 100-10,000. Many tests
134-137
revealed that carbon nanotube composites have very
low electrical percolation thresholds.
38

Even carbon nanotube contents meet the electrical percolation threshold, the electrical
conductivity of carbon nanotube-based composites varies greatly, typically has a measured
value ranging from 10
-5
to 10
-2
S/m.
138,139
However, Ramasubramaniam
140
and Ahmad
141

reported that the electrical conductivity can be significantly increased by adding more
contents of naontubes in the matrix. The conductivity was found to increase by 5 orders of
the magnitude after increasing the content of CNTs by 2 orders of the magnitude.
141
The
large variation of the electrical conductivity is unlikely to be explained by the variation of
the volume fraction of carbon nanotubes.
Several research reports indicate that the contact resistance caused by the separation of
carbon nanotubes is a key factor that affects the overall conductivity of the composites rather
than the intrinsic conductivity of the carbon nanotube itself.
142-144
Measurements have been
done on different types of single-walled carbon nanotube junctions, specifically, metal/metal
SWNT crossed junctions, semiconducting/semiconducting SWNT junctions and
metal/semiconducting SWNT junctions.
145
Contact resistance for metal/semiconducting
carbon nanotube junctions is much higher than carbon nanotube junctions with the same type
of carbon nanotubes. Buldum et al.
146
calculated the contact resistance between carbon
39

nanotubes and found the value varying from 100 k to 3.4 M when structural relaxation of
the nanotubes, contact length, nanotube diameter and the atomic structures in the contact
region are altered. The situation becomes more complex once the carbon nanotubes are
mixed with the matrix where carbon nanotubes are isolated by the polymers.
The contact resistance may be extremely large if the nanotubes are far away from each
other; or increased markedly if a thin insulating layer is formed by the polymers between the
carbon nanotubes. If the insulating layer is thin enough, the electric tunneling effect
dominates the conductance of the carbon nanotubes. In the above two cases, the carbon
nanotubes do not really come in contact with each other. However, in the real composites,
the contact resistance is dependent on the density and dispersion of the nanotubes, the well
contacted carbon nanotubes and the ones are close enough for electric tunneling, which
caused a large range in values. The poor conductivity of the carbon nanotube/polymer
composites was suggested by Kilbride et al.
147
who found the resistance was much larger
than expected, and by Foygel et al.
148
who estimated the contact resistance between carbon
nanotubes in composites to be in the order of 10
13
.
40

Many literatures have reported the fabrication of carbon naontube composite materials.
One benefit obtained from the carbon nanotube is improved stiffness and strength of the
composite. Nanotube-reinforced composites have attracted great interest of scientists in the
past decade. One key procedure of fabricating CNTs-based composite is dispersing
nanotubes into the matrix. Extensive investigations have been done on the processing
techniques for carbon nanotube-based polymer composites and the development has been
reviewed and highlighted by Thostenson et al.
149
. The reinforcement effect depends on how
well the carbon nanotubes can disperse in the polymer matrix. However, it remains a
challenge to separate and disperse carbon nanotubes homogenously into the polymer matrix
since the Van der Waals interactions between small-diameter carbon nanotubes are so strong
that they are more prone to form bundles and aggregate into bundles. Even though entangled
carbon nanotubes grown by a CVD method can be dispersed by high energy sonication,
150

high speed mixing
151
and surfactant-assisted processing through formation of a colloidal
intermediate,
152
other problems may arise during the processing, such as the break down or
destruction of carbon nanotubes, reagglomeration and too much surfactant.
41

The complexity of the conductivity mechanism for carbon nanotube-based composites
results in difficulty of controlling consistent conductance of the composite. To maintain the
repeated response of the resistance with strain, the original resistance of the composite must
be consistent. However, due to the difficulty of manipulating carbon nanotubes, the
homogeneity of the carbon nanotube dispersion is a big issue. The separation of carbon
nanotubes bundles is inherently difficult because of the strong Van der Waals interactions
and those as-separated nanotubes have a great tendency to aggregate again due to the high
surface energy of carbon nanotubes. Furthermore, it is really hard to determine the
dispersion degree of carbon nanotubes in the polymer matrix. The extent of contacting
nanotubes, contact length and contact area may vary for each process because of the
uncertainty of the dispersion. Especially when the electrical tunneling dominates the
conductance, the precise control of the thickness of the matrix layer that surrounds the
carbon nanotubes seems unlikely. The maximum tunneling distance in carbon nanotube-
based polymer composites is about 1.8 nm, which was reported by Li et al.
144
. Their results
further demonstrated that the contact resistance plays a dominant role in carbon nanotube-
based composites in contrast to the intrinsic resistance of nanotubes. The insulating film
42

formed in between crossing carbon nanotubes caused the diverse electrical conductivity of
the composites. Although the weight percent of the CNTs used to fabricate the composites
can be determined, the orientation of CNTs in the composites is unknown. Based on the
above analysis, it is doubtful if the behavior of the carbon naontube/polymer composites that
fabricated by dispersing single carbon nanotubes into the matrix is repeatable.
Interface deterioration leads to a resistance-strain hysteresis in cycling tests. The
interface between CNTs and matrix degrades after several loading-unloading cycles. This
causes the stress transferred from the matrix to CNTs to be smaller than the real values.
When the composite is under the first cyclic loading, the strain and resistance change curve
represents the real sensing properties. However, the bonding between CNTs and matrix is
degraded to some degree. When the second cyclic loading comes, the strain of the composite
will be larger than the value of the first loading if the same resistance change is obtained. An
unrepeatable strain-resistance curve indicates that the sensing ability of the CNTs-based
composite is limited to a small strain and at this limited strain, no destruction happens at the
interface. This limitation greatly impedes the application of CNTs-based composites because
in most situations, the strain people want to monitor may be larger than the limited values.
43

Another issue that is of concern is the drifting problems when a current is applied.
Resistance fluctuating has been observed by several groups during the tests.
13,15,20
One
possible reason for the resistance drifting may be the temperature dependence of the
resistance. The CNT composite sensor itself is a resistor, which means heat will be produced
when a direct current passes through the CNT network. Increase of the temperature will
change the conductance of each individual CNT, thus the overall resistances will change.
Another reason causing the drifting may be the electrical short happed in the nanotube
networks. When the carbon naontube-based composites are under voltage, some nanotubes
may burn themselves because of weak contact between them. The current-carrying capability
of carbon nanotubes also affect the lifetime of nanodevices fabricated using carbon
naontubes. An electrical breakdown may occur if the voltage applied exceeds the allowable
value. Previous research shows that even the metallic nanotubes which can sustain current
densities up to 109-1010 A/cm
2
will fail at sufficiently high currents.
153,154
Simulation and
theoretical calculations have been conducted on the current-carrying ability of carbon
nanotubes in the past 20 years.
155-158
The charge induced failure of nanotubes with different
types of helix is simulated by gradually increasing the applied electric voltage until the
44

breaking of bonds occurred, which is defined as the critical status.
159
Experimental
investigations on the electric failure of nanotubes roughly are in agreement with the
theoretical studies. Bonard et al.
160
observed the irreversible failure of a multi-walled
nanotube emitter. The maximum current was around 0.2 mA and sustained for only a few
seconds before an emission failure occurred. On the other hand, the electrical failure of
carbon nanotubs can be utilized as self-healing sensing devices.
161
If an electric circuit short
occurs, the nanotubes burn themselves and disconnect the electric loop. As the nantubes are
embedded in the polymer matrix, the randomly aligned carbon naotubes may be weakly
connected, which lower the current-carrying ability. These weakly connected CNTs may
cause a circuit short when a voltage is applied.
The above issues have yet to be solved in future works and more mature techniques are
still under developing and will finally bring this new type of CNT/polymer strain sensors in
real applications. The CNT sensing elements can be built in composites for in-situ structural
health monitoring with neglectable weight penalty and little invasiveness to the structure.
Numerous potential military and industrial applications of the developed strain sensor are
available, overcoming the limitation of conventional strain gages. The sensor can be bonded
45

to a surface, such as aircraft skin, or embedded in the structure to monitor the microscopic
strain.
1.5.4 Durability of CNT/polymer Composite Sensors
The target of increase the durability of CNT/polymer composite sensors is the same
task in CNT reinforced composites since CNT composites have multifunction of reinforcing
and sensing. The interfacial shear strength of the matrix and CNTs determine the interface
detachment. To increase the strength reinforcement, strong interfacial bonding is desirable. It
is well known that the efficiency of load transfer from the matrix to the fiber is controlled by
the interfacial shear stress. The durability issue of the CNT based composites is relate to the
interfacial bonding, that is, the strength of the interface created by the matrix and fiber
reinforcement. The interfacial debonding is determined by the interfacial shear strength. The
larger the interfacial shear strength, the more difficult the interface to be fractured.
It is found that multi-wall CNTs have lower load transfer efficiency than single wall
CNTs due to the weak bondings between each graphite layers. For multiwall CNTs, carbon
atoms have covalent bonds within each concentric graphite layer. However, there is only van
der Waals interaction between each layer. The weak bonding results in the decrease of load
46

transfer.
162
The CNT-matrix interface fails when debonding occurs under a critical interfacial
shear stress followed by the CNT pullout. The energy dissipation by the CNT pullout
contributes to the improvement in the fracture toughness of the CNT reinforced composites.
Computational experiments conducted by Chen et al.
163
revealed that for a constant volume
fraction, the energy dissipated by the multiwall CNT is a function of the number of outer
layers that could carry load.
Curved-fiber pull out model was proposed by Chen et al.
164
to analyze three types of
interface microstructure, i.e. matrix with fiber reinforcement is well-bonded, debonding and
sliding. At the first stage, the matrix and fiber are well-bonded and an external force will
generally create shear stress at the interface. As this force increases to a critical point at
which the debonding initialized, part of the fibers start sliding within the interface. Finally,
the third stage, sliding of the fiber in the whole matrix begins. The geometry of the fiber
filler played a role that the curving of the fiber increases the pull-out force, hence increasing
the reinforcing effect.
To increase the durability of the CNT nanocomposites, three key rules should be
obeyed when designing the composites. From a structural point of view, the fiber spacing
47

should be smaller than the characteristic strength limiting flaw, and higher packing density
of fibers with smaller diameters are preferred to bridge the flaws and cracks. Uniform fiber
distribution is important to realize an even load transfer throughout the whole composite. If
fillers are ununiformly dispersed in the matrix, only part of the fibers are taking load and full
reinforcement could not achieved. Thirdly, the interfacial bonding is a key factor dominating
the reinforcing effect especially the fibers are extremely stronger than the matrix. Covalent
bond is the strongest type of the interfacial bonding between the matrix and fiber constitutes.
Thus, Chen et al.
165
functionalized MWNTs with active sites on the outer walls which
originally do not have strong adhesion with polymer matrix. The treated MWNTs covalently
bonded with the matrix and the interface bonding is enhanced. Tensile strength tests
indicated the increase of the strength via adding a certain amount of weight percent of CNTs.
Another improvement of interfacial adhesion in polystyrene-carbon nanotube
composites was achieved via grafting polystyrene chains onto the nanotube surface.
166
The
surface of nanotubes exhibits characteristics close to the neutral graphite causing a weak
interfacial adhesion between matrix and nanotubes when fabricated into composites. The
covalent bonds between the grafted polymer chains and CNT surfaces resulting in a stronger
48

fiber/matrix interface and reinforcing effect was increased on both the stiffness and
elongation-to-failure stress.
CNTs exhibit outstanding properties in nanoscale, attempts to transfer these mechanical,
thermal and electrical properties to a microscopic scale have been done extensively in recent
years to multifunctionlize CNTs with sensing, strength reinforcing and other functions.
Stable and reliable interface adhesion of CNTs with polymer matrix is an urgent task to put
forward the real application of CNT multifunctional composites. Near-term research should
focus on change the chemical activity of the CNT surface for strengthening the interface
bonding.










49

REFERENCES
1. Iijima, S., Nature 354, 6348-6356 (1991).

2. Zhao H.; Bradford P. D.; Wang X.; Liu W. Luo. T. J. M.; J ia Q. X.; Zhu Y. T.; Yuan
F. G., Material Letters 64, 1947-1950 (2010).

3. Chen X. H.; Wang J. F.; Zou J. G.; Wu X. L.; Chen X. J.; Xue F., Journal of Applied
Polymer Science 114, 3407-3413 (2009).

4. Jung Y. C.; Yoo J. H.; Kim Y. A.; Cho J. W.; Endo M., Carbon 48, 1598-1603
(2010).

5. Deka H.; Karak N.; Kalita R. D.; Buragohain A. K., Carbon 48, 2013-2022 (2010).

6. Smart S.; Fania D.; Milev A.; Kannangara G. S. K.; Lu M.; Martin D., Journal of
Applied Polymer Science 117, 24-32 (2010).

7. Khan U.; Blighe F. M.; Coleman J. N., Journal of Physical Chemistry C 114, 11401-
11408 (2010).

8. Abot J. L.; Schulz M. J.; Song Y.; Medikonda S.; Rooy N., Smart Materials and
Structures 19, 085007-085018 (2010).

9. Cui L. F.; Hu L. B.; Choi J. W.; Cui Y., ACS Nano 4, 3671-3678 (2010).

10. Sheem K. Y.; Sung M.; Lee Y. H., Electrochimica Acta 55, 5808-5812 (2010).

11. Zhou R. F.; Meng C. Z.; Zhu F.; Li Q. Q.; Liu C. H.; Fan S. S.; Jiang K. L.,
Nanotechnology 21, 345701-345707 (2010).

12. Knite M.; Tupurenia V.; Fuith A.; Zavickis J.; Teteris V., Materials Science and
Engineering C 27, 1125-1128 (2007).
50

13. Pham G. T.; Park Y. B.; Liang Z. Y.; Zhang C.; Wang B., Composites: Part B 39,
209-216 (2007).

14. Loh K. J.; Kim J.; Lynch J. P.; Kam, N. W. S; Kotov N. A., 2007 Smart Materials
and Structures 16, 429-438 (2007).

15. Kang I.; Schulz M. J.; Kim J. H.; Shanov V.; Shi D., Smart Materials and Structures
15, 737-748 (2006).

16. Thostenson E. T.; Chou T. W., Advanced Materials 18, 2837-2841 (2006).

17. Boger L.; Wichman M. H. G.; Meyer L. O.; Schulte K., Composites Science and
Technology 68, 1886-1894 (2008).

18. Thostenson E. T.; Chou T. W., Nanotechnology 19, 215713-215718 (2008).

19. Li C.; Thostenson E. T.; Chou T. W., Composites Science and Technology 68, 1445-
1452 (2008).

20. Li X.; Levy C.; Elaadil L., Nanotechnology 19, 045501-045507 (2008).

21. Dharap P.; Li Z.; Nagarajaiah S.; Barrera E. V., Nanotechnology 15 379-382 (2004).

22. Li Z.; Dharap P.; Nagarajaiah S.; Barrera E. V. Kim J. D., Advanced Materials 16,
640-643 (2004).

23. Zhao H.; Zhang Y.; Bradford P. D.; Zhou Q.; Jia Q. X.; Yuan F. G.; Zhu Y. T.,
Nanotechnology 21, 305502-305506 (2010).

24. Wang, X.; Li, Q.; Xie, J.; J in, Z.; Wang, J.; Li, Y.; J iang, K.; Fan, S. S., Nano Letters
9, 31373141 (2009).

25. Meo M.; Rossi M., Composites Science and Technology 66, 1597-1605 (2006).
51

26. Salvetat J. D.; Rubio A., Carbon 40, 1729-1734 (2002).

27. Yu M.; Files B. S.; Arepalli S.; Ruoff R. S., Physical Review Letters 84, 5552-5555
(2000).

28. Yu M.; Lourie O.; Dyer M. J.; Moloni K.; Kelly T. F.; Ruoff R. S., Science 287,
637640 (2000).

29. Gao G. H.; Cagin T.; Goddard W. A., Nanotechnology 9, 184-191 (1998).

30. Edwards B. C., Acta Astronautica 47, 735-744 (2000).

31. Hong S. H.; Myung S., Nature Nanotechnology 2, 207-208 (2007).

32. Dai H. J.; Wong E. W.; Lieber C. M., Science 272, 523-526 (1996).

33. Pop E.; Mann D.; Wang Q.; Goodson K.; Dai H. J., Nano Letters 6, 96-100 (2006).

34. Cao J.; Wang Q.; Dai H. J., Physical Review Letters 90, 157601-157604 (2003).

35. Grow R. J.; Wang Q.; Cao J.; Wang D. W.; Dai H. J., Applied Physics Letters 86,
093104-093106 (2005).

36. Tans S. J.; Verschueren A. R. M.; Dekker C., Nature 393, 49-52 (1998).

37. Javey A.; Kim H.; Brink M.; Wang Q.; Ural A.; Guo J.; McIntyre P.; McEuen P.;
Lundstrom M.; Dai H. J., Nature Materials 1, 241-246 (2002).

38. Javey A.; Guo J.; Wang Q.; Lundstrom M.; Dai H. J., Nature 424, 654-657 (2003).

39. Rosenblatt S.; Yaish Y.; Park J.; Gore J.; Sazonova V.; McEuen P. L., Nano Letters 2,
869-872 (2002).

52

40. Lu C.; Fu Q.; Huang S.; Liu J., Nano Letters 4, 623-627 (2004).

41. Huang S.; Woodson M.; Smalley R.; Siddons G. P.; Merchin D.; Back J. H.; Jeong J.
K.; Shim M., Nano Letters 4, 927-931 (2004).

42. Tan S. J.; Devoret M. H.; Dai H. J.; Thess A.; Smalley R. E.; Geerligs L. J.; Dekker
C., Nature 386, 474-477 (1997).

43. Bockrath M.; Cobden D. H.; McEuen P. L., Science 290, 1552-1555 (2000).

44. Han S.; Liu X.; Zhou C., Journal of American Chemical Society 127, 5294-5295
(2005).

45. Aliv A. E.; Lima M. D.; Fang S. L.; Baughman R. H., Nano Letters 10, 2374-2380
(2010).

46. Zhang H. X.; Chen F.; Zhai Y. C.; J iang K. L.; Li Q. Q.; Fan S. S, Advanced
Materials 21, 2299-2034 (2009).

47. An K. H.; Kim W. S.; Park Y. S.; Choi Y. C.; Lee S. M.; Chung D. C.; Bae D. J.;
Lim S. C.; Lee Y. H., Advanced Materials 13, 497-500 (2001).

48. An K. H.; Kim W. S.; Park Y. S.; Moon J. M.; Bae D. J.; Lim S. C.; Lee Y. S.; Lee Y.
H., Advanced Functional Materials 11, 387-392 (2001).

49. Jurewicz K.; Delpeux S.; Bertagna V.; Beguin F.; Frackowiak E., Chemical Physics
Letters 347, 36-40 (2001).

50. Chew S. Y.; Ng S. H.; Wang J. Z.; Novak P.; Krumeich F.; Chou S. L.; Chen J.; Liu
H. K., Carbon 47, 2976-2983 (2009).

51. Dai H. J.; Hafner J. H.; Rinzler A. G.; Colbert D. T.; Smalley R. E., Nature 384, 147-
150 (1996).
53

52. Collins P. G.; Bradley K.; Ishigami M.; Zettle A.; Science 287, 1801-1804 (2000).

53. Kong J.; Franklin N. R.; Zhou C.; Chapline M. G.; Peng S.; Cho K.; Dai H. J.,
Science 287, 622-625 (2000).

54. Star A.; Gabriel J. C. P.; Bradley K.; Gruner G.; Nano Letters 3, 459-463 (2003).

55. See C. H.; Harris A. T., Industrial and Engineering Chemical Research 46, 997-1012
(2007).

56. Guo T.; Nikolaev P.; Thess A.; Colbert D. T.; Smalley R. E., Chemical Physics
Letters 243, 49-54 (1995).

57. Peng X. S.; Koczkur K.; Chen A. C., Journal of Physics D: Applied Physics 41,
095409-095414 (2008).

58. Gui X. C.; Wei J. Q.; Wang K. L.; Cao A. Y.; Zhu H. W.; J ia Y.; Shu Q. K.; Wu D.
H., Advanced Materials 22, 617-621 (2010).

59. Wang D.; Song P. C.; Liu C. H.; Wu W.; Fan S. S., Nanotechnology 19, 075609-
075614 (2008).

60. Lee D. H.; Kim J. E.; Han T. H.; Hwang J. W.; Jeon S.; Choi S. Y.; Hong S. H.; Lee
W. J.; Ruoff R. S.; Kim S. O., Advanced Materials 22, 1247-1252 (2010).

61. Terranova M. L.; Sessa V.; Rossi M., Chemical Vapor Deposition 12, 315-325
(2006).

62. Couteau E.; Hernadi K.; Seo J. W.; Thien N. L.; Miko C.; Gaal R.; Forro L.,
Chemical Physic Letters 378, 9-17 (2003).

63. Wang Q.; Yu H.; Wei F.; Zhang Q.; Zhang Z., Carbon 40, 2961-2973 (2002).

54

64. Wang Z.; Chen X. H.; Nie H. G.; Zhang K.; Li W. H.; Yi B.; Xu L. S.,
Nanotechnology 19, 085606-085613 (2008).

65. Maghrebi M.; Khodadadi A. A.; Mortazavi Y.; Sane A.; Rahimi M.; Shirazi Y.;
Tsakadze Z.; Mhaisalkar S., Applied Physics A 97, 417-424 (2009).

66. Pint C. L.; Alvarez N. T.; Hauge R. H., Nano Research 2, 526-534 (2009).

67. Pint C. L.; Pheasant S. T.; Pasquali M.; Coulter K. E.; Schmidt H. K.; Hauge R. H.,
Nano Letters 8, 1879-1883 (2008).

68. Hernadi K.; Thien N. L.; Forro L., Journal of Physical Chemistry B 105, 12464-
12468 (2001).

69. Villoria R. G.; Figueredo S. L.; Hart A. J.; Steiner S. A.; Slocum A. H.; Wardle B. L.,
Nanotechnology 20, 405611-405618 (2009).

70. Zhang H.; Cao G. P.; Wang Z. Y.; Yang Y. S.; Shi Z. J.; Gu Z. N.;

71. Hahm M. G.; Kwon Y. K.; Lee E.; Ahn C. W.; Jung Y. J., Journal of Physical
Chemistry C 112, 17143-17147 (2008).

72. Mora E.; Harutyunyan A. R., Journal of Physical Chemistry C 112, 4805-4812
(2008).

73. Li Y. L.; Kinloch I. A.; Shaffer M. S. P.; Geng J.; Johnson B.; Windle A. H.,
Chemical Physics Letters 384, 98-102 (2004).

74. Wang Y.; Wei F.; Luo G.; Yu H.; Gu G., Chemical Physics Letters 303, 117-119
(1999).

75. Kumar M.; Ando Y.; Carbon 43, 533-540 (2005).

55

76. Lavrentiev V.; Abe H.; Yamamoto S.; Naramoto H.; Narumi K., Physica B 323, 303-
305 (2002).

77. Resasco D. E.; Alvarez W. E.; Pompeo F.; Balzano L.; Herrera J. E.; Kitiyanan B.;
Borgna A., Journal of Nanopartical Research 4, 131-136 (2002).

78. Cheng H. M.; Li F.; Su G.; Pan H. Y.; He L. L.; Sun X.; Dresselhaus M. S., Applied
Physical Letters 72, 3282-3284 (1998).

79. Maruyama S.; Kojima R.; Miyauchi Y.; Chiashi S.; Kohno M., Chemical Physics
Letters 374, 53-58 (2003).

80. Kamalakar G.; Hwang D. W.; Hwang L. P., Journal of Materials Chemistry 12,
1819-1823 (2002).

81. Rong C. B.; Poudyal N.; Chaubey G. S.; Nandwana V.; Liu Y. Z.; Wu Y. Q.; Kramer
M. J.; Kozlov M. E.; Baughman R. H.; Liu J. P., Journal of Applied Physics 103,
07E131-07E133 (2008).

82. Rong C. B.; Poudyal N.; Chaubey G. S.; Nandwana V., Journal of Applied Physics
102, 043913-043918 (2007).

83. Liao X. Z.; Serquis A.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Applied Physics Letters
82, 2694-2695 (2003).

84. Serquis A.; Liao X. Z.; Huang J. Y.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Carbon 41,
2635-2641 (2003).

85. Saito Y.; Tani Y.; Journal of Physical Chemistry B 104, 2495-2499 (2000).

86. Rong C. B.; Nandwana V.; Poudyal N.; Liu J. P.; Kozlov M. E.; Baughman R. H.;
Ding Y.; Wang Z. L., Journal of Applied Physics 102, 023908-023914 (2007).

56

87. Mauron P.; Emmenegger C.; Sudan C.; Wenger P.; Rentsch S.; Zuttel A., Diamond
and Related Materials 12, 780-785 (2004).

88. Herrera J. E.; Balzano L.; Borgna A.; Alvarez W. E.; Resasco D. E., Journal of
Catalysis 204, 129-145 (2001).

89. Jong W. J.; Lai S. H.; Hong K. H.; Lin H. N.; Shih H. C., Diamond and Related
Materials 11, 1019-1025 (2002).

90. Yuan D.; Ding L.; Chu H.; Feng Y.; McNicholas T. P.; Liu J., Nano Letters 6, 2642-
2645 (2006).

91. Andrews R.; Jacques D.; Rao A. M.; Derbyshire F.; Qian D.; Fan X.; Dickey E. C.;
Chen J., Chemical Physics Letters 303, 467-474 (1999).

92. Kumar M.; Ando Y., Chemical Physics Letters 374, 521-526 (2003).

93. Kumar M.; Kakamu K.; Okazaki T.; Ando Y., Chemical Physics Letters 385, 161-
165 (2004).

94. Cheung C. L.; Kurtz A.; Park H.; Lieber C. M., Journal of Physical Chemistry B 106,
2429-2433 (2002).

95. Hongo H.; Yudasaka M.; Ichihashi T.; Nihey F.; Iijiama S., Chemical Physics Letters
361, 349-352 (2002).

96. Colomer J. F.; Stephan C.; Lefrant S.; Van-Tendeloo G.; Willems I.; Konya Z.;
Fonseca A.; Laurent C.; Nagy J. B., Chemical Physics Letters 317, 83-89 (2000).

97. Ward J.; Wei B. Q.; Ajayan P. M., Chemical Physics Letters 376, 717-725 (2003).

98. Ago H.; Nakamura K.; Imamura S.; Tsuji M., Chemical Physics Letters 391, 308-313
(2004).
57

99. Patole S. P.; Alegaonkar P. S.; Shin H. C.; Yoo J. B., Journal of Physics D 41,
155311-155316 (2008).

100. Li X. S.; Zhang X. F.; Ci L. J.; Shah R.; Wolfe C.; Kar S.; Talapatra S.; Ajayan P. M.,
Nanotechnology 19, 455609-455615 (2008).

101. Stadermann M.; Sherlock S. P.; In J. B.; Fornasiero F.; Park H. G.; Artyukhin A. B.;
Wang Y. M.; Yoreo J. J. D.; Grigoropoulos C. P.; Bakajin O.; Chernov A. A.; Noy
A., Nano Letters 9, 738-744 (2009).

102. Chakrabarti S.; Nagasaka T.; Yoshikawa Y.; Pan L. J.; Nakayama Y., Japanese
Journal of Applied Physics 45, 720-722 (2006).

103. Amama P. B.; Pint C. L.; McJ ilton L.; Kim S. M.; Stach E. A.; Murray P. T.; Hauge
R. H.; Maruyama B., Nano Letters 9, 44-49 (2009).

104. See C. H.; Harris A. T., Industrial and Engineering Chemical Research 46, 997-1012
(2007).

105. Li Q.; Yan H.; Zhang J.; Liu Z., Carbon 42, 829-835 (2004).

106. Maruyama S.; Kojima R.; Miyauchi Y.; Chiashi S.; Kohno M., Chemical Physics
Letters 360, 229-234 (2002).

107. Cheng Q. F.; Wang B.; Zhang C.; Liang Z. Y., Small 6, 763-767 (2010).

108. Kumar M.; Ando Y., Journal of Nanoscience and Nanotechnology 10, 3739-3758
(2010).

109. Yao Y.; Li Q. W.; Zhang J.; Liu R.; J iao L. Y.; Zhu Y. T.; Liu Z. F., Nature Materials
6, 283-286 (2007).

110. Lee Y. T.; Park J., Journal of Physical Chemistry B 106, 7614-7618 (2002).
58

111. Lee C. J.; Park J.; Yu J. A., Chemical Physics Letters 360, 250-255 (2002).

112. Alvarez W. E.; Kitiyanan B.; Borgna A.; Resasco D. E., Carbon 39, 547-558 (2001).

113. Yoshida H.; Takeda S.; Uchiyama T.; Kohno H.; Homma Y., Nano Letters 8, 2082-
2086 (2008).

114. Nagaraju N.; Fonseca A.; Konya Z.; Nagy J. B., Journal of Molecular catalysis A:
Chemical 181, 57-62 (2002).

115. Li Y. L.; Kinloch I. A.; Windle A. H., Science 304, 276-278 (2004).

116. Zhang X. F; Li Q. W.; Tu Y.; Li Y.; Coulter J. Y.; Zheng L. X.; Zhao Y. H.; J ia Q.
X.; Peterson D. E.; Zhu Y. T., Small 2, 244-248 (2007).

117. Zhang M.; Atkinson K. R.; Baughman R. H., Science 306, 1358-1361 (2004).

118. Zhang X. B.; J iang K. L.; Feng C.; Liu P.; Zhang L. N.; Kong J.; Zhang T. H.; Li Q.
Q.; Fan S. S., Advanced Materials 18, 1505-1510 (2006).

119. Inoue Y.; Kakihata K.; Hirono Y.; Horie T.; Ishida A.; Mimura H., Applied Physics
Letters 92, 213113-213115 (2008).

120. Ericson L. M.; Fan H.; Peng H. Q.; Davis V. A.; Zhou W.; Sulpizio J.; Wang Y. H.;
Booker R.; Smalley R. E., Science 305, 1447-1450 (2004).

121. Zhang S.; Zhu L.; Minus M. L.; Chae H. G.; Jagannathan S.; Wong C. P.; Kowalik J.;
Rpberson L. B.; Kumar S., Journal of Materials Science 43, 4356-4362 (2008).

122. Liu K.; Sun Y.; Zhou R.; Zhu H.; Wang J.; Liu L.; Fan S.; J iang K., Nanotechnology
21, 045708-045714 (2010).

123. Li C. T.; Tietz J. V., Journal of Materials Science 25, 4694-4698 (1990).
59

124. Tombler T. W.; Zhou C.; Alexseyev L.; Kong J.; Dai H.; Liu L.; Jayanthi C. S.; Tang
M.; Wu S. Y., Nature 405, 769-772 (2000).

125. Grow R. J.; Wang Q.; Cao J.; Wang D.; Dai H., Applied Physics Letters 86, 093104-
093106 (2005).

126. Cao J.; Wang Q.; Dai H., Physical Review Letters 90, 157601-157604 (2003).

127. Minot E. D.; Yaish Y.; Sazonova V.; Park J.; Brink M.; McEuen P. L., Physical
Review Letters 90, 156401-156404 (2003).

128. Nishio T.; Miyato Y.; Kobayashi K.; Matsushige K.; Yamada H., Applied Physical
Letters 92, 063117-063119 (2008).

129. Laxminarayana K.; Jalili N., Textile Research Journal 9, 670-680 (2005).

130. Hadjiev V. G.; Lliev M. N.; Arepalli S.; Nikolaev P.; Files B. S., Applied Physics
Letters 78, 3193-3195 (2001).

131. Frogley M. D.; Zhao Q.; Wager H. D., Physical Review B 65, 113413-113416 (2002).

132. Zhang W.; Suhr J.; Koratkar N., Journal of Nanoscience and Nanotechnology 6, 960-
964 (2006).

133. Loh K. J.; Hou T-C.; and Lynch J. P., Proceedings of the 5th International Workshop
for Structural Control and Monitoring Dalian, China, June 5-6, (2008).

134. Sandler J. K. W.; Krik J. E.; Kinloch I. A.; Shaffer M. S. P.; Windle A. H., Polymer
44, 5893-5899 (2003).

135. Moisala A.; Li Q.; Kinloch I. A.; Windle A. H., Composite Science and Technology
66, 1285-1288 (2006).

60

136. Bryning M. B.; Islam M. F.; Kikkawa J. M.; Yodh A. G., Advanced Materials 17,
1186-1191 (2005).

137. Thostenson E. T.; Chou T. W., Carbon 44, 3022-3029 (2006).

138. Kilbride B. E.; Coleman, J. N.; Fraysse, J.; Fournet, P.; Cadek, M.; Drury, A.;
Hutzler, S.; Roth, S.; Blau, W. J., Journal of Applied Physics 92, 4024-4030 (2002).

139. McNally T.; Potschke P.; Halley P.; Murphy M.; Martin D.; Bell S. E. J.; Brennan G.
P.; Bein D.; Lemoine P.; Quinn J. P., Polymer 46, 8222-8232 (2005).

140. Ramasubramaniam R.; Chen J.; Liu H. Y., Applied Physics Letters 83, 2928-2930
(2003).

141. Ahmad K.; Pan W.; Shi S. L., Applied Physics Letters 89, 133122-133124 (2006).

142. Stadermann M.; Papadakis S. J.; Falvo M. R.; Novak J.; Snow E.; Fu Q.; Liu J.;
Fridman Y.; Boland J. J.; Superfine R.; Washburn S., Physical Review B 69, 201402-
201404 (2004).

143. Hecht D.; Hu L. B.; Gruner G., Applied Physics Letters 89, 133122-133124 (2006).

144. Li C. Y.; Thostenson E. T.; Chou T. W., Applied Physics Letters 91, 223114-223116
(2007).

145. Fuhrer M. S.; Nygard J.; Shih L.; Forero M.; Yoon Y. G.; Mazzoni M. S. C.; Choi H.
J.; Ihm J.; Louie S. G.; Zettl A.; McEuen P. L. Science 288, 494-497 (2000).

146. Buldum A.; Lu J. P., Physical Review B 63, 161403-161406 (2001).

147. Kilbride B. E.; Coleman, J. N.; Fraysse, J.; Fournet, P.; Cadek, M.; Drury, A.;
Hutzler, S.; Roth, S.; Blau, W. J., Journal of Applied Physics 92, 4024-4030 (2002).

61

148. Foygel M.; Morris R. D.; Anez D.; French S.; Sobolev V. L., Physical Review B 71,
104201-104208 (2005).

149. Thostenson E. T.; Li C. Y.; Chou T. W., Composite Science and Technology 65, 491-
516 (2005).

150. Miyagawa H.; Drzal L. T., Polymer 45, 5163-5170 (2004).

151. Sandler J.; Shaffera M. S. P.; Prasseb T.; Bauhoferb W.; Schulte K.; Windle A. H.,
Polymer 40, 5967-5971 (1999).

152. Pham G. T.; Park Y. B.; Wang B., ASME, Manufacturing Engineering Division 16,
987 (2005).

153. Umnov A. G.; Mordkovich V.Z., Applied Physics A 73, 301-304 (2001).

154. Collins P. G.; Arnold M. S.; Avouris P., Science 292, 706-709 (2001).

155. Lee Y. H. L.; Kima S. G.; Tomaneka D., Chemical Physics Letters 265, 667-672
(1997).

156. Deheer W. A.; Chatelain A.; Ugarte D., Science 270, 1179-1180 (1995).

157. Rinzler A. G.; Hafner J. H.; Nikolaev P.; Nordlander P.; Colbert D. T.; Smalley R. E.;
Lou L.; Kim S. G.; Tomanek D., Science 269, 1550-1553 (1995).

158. Raghavachari K,; Binkley S., Journal of Chemical Physics 87, 2197-2203 (1987).

159. Li C. Y.; Chou T. W., Carbon 45, 922-930 (2007).

160. Bonard, J. M.; Klinke C.; Dean K. A.; Coll B. F., Physical Review B 67, 115406-
115410 (2003).

62

161. Stadermann M.; Papadakis S. J.; Falvo M. R.; Novak J.; Snow E.; Fu Q.; Liu J.;
Fridman Y.; Boland J. J.; Superfine R.; Washburn S., Physical Review B 69, 201402-
201404 (2004).

162. Thostenson E. T.; Chou T. W., Journal of Physics D 36, 573-582 (2003).

163. Chen Y.; Balani K.; Agarwal A., Applied Physics Letters 92, 011916-011918 (2008).

164. Chen X. Y.; Beywelein I. J.; Brinson L. C., Mechanics of Materials, 41, 293-307
(2009).

165. Chen G. X.; Kim H. S.; Park B. H.; Yoon J. S., Polymer 47, 4760-4767 (2006).

166. Fragneaud B.; Varlot K. M.; Montiel A. G.; Terrones M.; Caville J. Y., Composites
Science and Technology 68, 3265-3271 (2008).









63

CHAPTER 2: Experimental
2.1Materials
A list of materials used to grow carbon nanotubes and prepare fiber sensors are listed in
Table 2.1.
Table 2.1 Materials used for growing nanotubes and building carbon nanotube fiber sensors
in composites.
Materials Vendor Description
Silicon wafer (100) University Wafer Company Thickness: 350 m
Al
2
O
3
target Kurt Lesker Company Purity: 99.99%
Fe target Kurt Lesker Company Purity: 99.99%
Zr target Kurt Lesker Company Purity: 99.9%
Al target Kurt Lesker Company Purity: 99.99%
Argon gas Machine and Welding Supply
Company
Purity: 99.999%
Hydrogen gas Machine and Welding Supply
Company
Purity: 99.999%
Ethylene gas Machine and Welding Supply
Company
Purity: 99.999%
Acetylene gas Machine and Welding Supply
Company
Purity: 99.9%
EPON
TM
826 Hexion Specialty Chemicals Bisphenol epoxy resin
EPIKURE
TM
9551 Hexion Specialty Chemicals Polyamine curing
agent
Heloxy
TM
62 Hexion Specialty Chemicals Modifier
Quartz plate and tube Technical Glass Products
FeCl
2
Fisher Scientific Purity: 99.5%
64

2.2 Catalyst Thin Film Deposition using Magnetron Sputtering
Alumina thin films with a thickness of 10 nm were sputtered on the polished side of a
piece of silicon wafer using a custom-built magnetron sputtering machine. Fe and Fe-Zr thin
films with thickness of 1.5 nm were deposited separately on the as-sputtered Al
2
O
3
layer.
Figure 2.1 is a real image of the magnetron sputtering machine. The diameter of the cathode
is 2.5 cm. The base pressure for each sputtering was 510
-7
Torr. The sample stage to the
cathode is about 7 cm. Fe-Zr film was prepared by co-sputtering a Fe target and Zr target.
Figure 2.2 shows the configuration of the Fe and Zr target. Fe target was cut into turbulent
blade shape and put underneath of Zr target to increase the uniformity of film composition.
To further obtain a uniform catalyst film, the sample stage was rotating during sputtering.
The thickness of the films was monitored by a quartz balance. The quartz balance was
calibrated by sputtering a thick film and measuring the thickness under SEM. The detailed
sputtering condition for each film is listed in Table 2.2.




65


Figure 2.1 Image of a custom-built radio frequency magnetron sputtering machine
consisting of major parts of a radio frequency power supply, a cathode, a 4-way cross
chamber, a quartz balance, a turbulent pump, a mechanical pump, a mass flow controller, a
thermocouple gauge, an ion gauge, a rotating stage and a chill water cooling system.



66


Figure 2.2 Image of a Fe target (left) and Zr target (right). Fe target was cut into turbulent
blade shape.

Table 2.2 Sputtering conditions for Fe, Fe-Zr and Al
2
O
3
films.
Films Fe Fe-Zr Al
2
O
3

Sputtering power (watts) 10 10 40
Sputtering rate (/s) 0.4-0.5 0.14 0.03-0.04
Thickness (nm) 1.5 1.5 10
Ar flow rate (sccm) 10 10 10
Sputtering pressure (torr) 110
-3
110
-3
110
-3


2.3 Using Thin Film Catalyst to Grow Carbon Nanotubes
As-sputtered catalyst films on silicon substrates were loaded in the middle of a quartz
tube reactor. CNTs arrays were grown at 750C using a CVD method in a 2.5 cm tube
furnace. Figure 2.3 shows the set up of the custom-built CVD system. The total gas flow rate
was 200 sccm consisting of 40 sccm H
2
, 30 sccm C
2
H
4
and 130 sccm Ar. The procedure was
67

as follows: (1) The tube was pumped to 25 Torr and then refilled with argon. The furnace
was then ramped to the growth temperature under a flow of argon at a flow rate of 160 sccm.
(2) When the furnace reached 735C the H
2
flow (40 sccm) was added to the argon. (3)
When the furnace reached 750C ethylene flow (30 sccm) was added to initiate growth. Ar
flow rate decreased to 130 sccm. Growth time was calculated from this point. All the CVD
growth parameters for the Fe and Fe-Zr catalysts were constant. .

Figure 2.3 Real image of the CVD system built for CNT array growth. Major parts of the
CVD system are mass flow controllers for Ar, C
2
H
4
, H
2
gases, a quartz tube reactor and a
furnace.

68

2.4 Using FeCl
2
Powders to Grow Carbon Nanotubes
FeCl
2
powders were loaded into a quartz boat and the quartz boat was inserted in the
middle of a liner quartz tube with diameter of around 1inch. Figure 2.4 shows the set-up of
one-step CVD system built for growing CNTs. The tube reactor was pumped down to at
least 110
-3
torr in one hour followed by a heating up process. The system was kept in
vacuum during heating. FeCl
2
started to evaporate at round 600C and fully became vapor at
the growth temperature of 820C. The hot zone was about 30 cm. FeCl
2
re-deposited at the
ends of the heated zone. After reached 820C, acetylene started to flow into the reactor.
CNTs began to grow. A 2.1mm-tall CNT array was grown in 20 minutes with a growth rate
of 100 m/min. A drawable CNT array could be grown in 7 minutes with a height of around
700 m. CNTs grew not only on the quartz substrates but also on the inner wall of the tube
reactor. To protect the main tube from growing CNTs, a liner tube was inserted in the main
tube.
69


Figure 2.4 Schematic of the one-step CVD system built for CNT array growth.
2.5 Measuring the Quality of Carbon Nanotube Arrays using Raman Spectra
Two types of Ramanscope were used to assess the quality of CNT arrays. Renishaw
Ramamscope was used to exam samples grown at atmosphere pressure. Wavelength of the
laser beam was 514 nm. At least 10 cycles of scan were performed on each sample to
increase the signal. Scans were completed in the Raman shift range of 1200-1700 cm
-1
.
Raman spectra were collected with a Jobin-Yvon LabRAM HR800 spectrometer for samples
grown using one-step CVD method. Spectra were excited with the 632.8 nm line of a HeNe
internal laser and the 90 angle measurement geometry was used to record the data. The
measurement time is 10 seconds with five accumulations.
70

2.6 Carbon Nanotube Fiber Spinning
Figure 2.5 shows a spinning machine built for making CNT fibers. CNT arrays
vertically grown on silicon substrates were attached on the stage of the spinning machine via
a piece of double-sided tape. CNT ribbons approximately 3 mm wide were pulled out from
an CNT array by hand using flat tweezers and then attached on a spindle using a piece of
tiny double-sided tape. A CNT fiber was spun by a rotating spindle that was pulling away
from the array at the same time. The rotating speed was 7000rpm and the fibers pulling
speed was about 2mm s
-1
. The diameters of the as-spun fibers were 3 to 30m and can be
controlled by the width of ribbons initially pulled out from the array.

Figure 2.5 Image of a fiber spinning machine.
71

2.7 Diameter Measurement of Carbon NanotubeFibers Using Laser Beam Diffraction
Fibers were mounted on a paper tab with a rectangular slot cut out. The paper tab then
was clamped with screws to a custom-built stainless steel tab. Figure 2.6 shows the
configuration of the fiber diameter measurement system. The laser beam has a wavelength of
543nm. The converged beam passed the sample and produced a diffraction pattern on the
screen. By measuring the sample to screen distance and spacing of bright spots, the fiber
diameter could be readily calculated using the following equation.


To increase the accuracy of the measurement, at least five different places of the fiber were
measured. Compared with the measurements under SEM, the accuracy of within 1% error
was achieved.
2.8 Tensile Testing of Carbon Nanotube Fibers
An EZ-S Shimazu tensile tester with a 2N load cell for pure dry fibers and a 100N load
cell for composite sensors was used for mechanical testing. Dry fibers were glued to the
sandpaper tab with a gauge length of 10 mm. The purpose of the sandpaper is to prevent
2 2 2
2
1
d L L
D
d d


| | +
= = +
|
\ .
72

slipping during loading. Samples were first screwed on the clamp. After zero the force, the
edge of the sandpaper tab was cut. A strain speed of 0.5mm/min was used for mechanical
tests. To avoid breakage of fibers at the ends, fibers have to be vertically aligned along the
loading direction. At least five samples made with the same conditions were tested to
increase the accuracy.

Figure 2.6 Images of the set-up for measuring the diameter of CNT fibers using laser
diffraction. (a) Set-up of laser generator, sample stage and screen. (b) Screen with scale
printed out used to collect the diffraction pattern. (c) Image of diffraction pattern in dark
room.
(a)
(b)
(c)
73

2.9 Electrical Property of Carbon nanotube Fibers
CNT fibers were collected on sandpapers with a rectangular slot cut out. Fibers in the
slot region were covered by a plastic film to protect from deposition and the exposed parts
were coated with 5m thick Al layers using magnetron sputtering in pure Ar at a pressure of
210
-3
Torr. The sputtering power was 40 watts. Silver paste was used to connect two thin
copper wires with fibers that were already coated with Al film for the resistance
measurement. An Agilent 34410A digital multimeter was used to measure the resistance of
the fibers. The sample gauge length for electromechanical testing was 1cm. The samples
were loaded under tension and the resistance change was recorded in real time. Figure 2.7
shows the electric wire connection and mechanical test configuration of CNT fibers.

Figure 2.7 A CNT fiber was electrically wired and mounted for electromechanical
measurement.
74

2.10 Embedding Carbon Nanotube Fibers in Resin
CNT fibers were attached on a piece of release plastic paper that is commonly used to
peel off composites from the mold. High conductive silver colloid is dropped on the ends of
the fibers to form electrodes. The resin system (59% EPON
TM
Resin826, 21%
EPIKURE
TM
9551 Curing Agent, 20% Heloxy
TM
Modifier62 in weight percent, all were
purchased from Hexion) was mixed using magnetic stirrer. Then bubbles trapped in the resin
were removed in a vacuum oven. Resin was warmed in the oven at around 60C to lower
the viscosity and then was cast on the fibers to form thin films and after curing at 373K for 3
hours, an integral structure of composites with built-in fiber sensing elements was produced.
The composite thin films were around 100m thick which was controlled by the thickness of
aluminum slot built for molding the resin. The film was cut into coupons using a razor blade.
Two copper wires were soldered on the silver electrodes for resistance tests.
2.11 Micro-structure Analyze
Micro-structure of CNTs and fibers were investigated using SEM and TEM. Two types
of SEM were used to study the morphology of CNTs. CNTs arrays grown using Fe and Fe-
Zr catalysts were observed using FE-SEM (Model: Hitachi SU-70). Fibers and CNTs grown
75

using FeCl
2
catalyst were observed using a JEOL 6400F at 5kV and a working distance of
15mm. Wall number, wall spacing and quality of CNTs were investigated using a JEOL
2010F transmission electron microscopy operating at 200kV. To prepare the TEM samples,
a small bundle of CNTs was put into ethanol, CNTs were dispersed in the solution under
sonicating for 30 minutes. A drop of the CNT suspension was placed on TEM copper grid
with lacy carbon film.
Morphology of catalysts was investigated using atomic force microscopy. As-sputtered
catalysts were cut into 12 pieces from an approximately one-inch square sample. Two key
morphology evolution stages of both Fe and Fe-Zr catalysts were studied by Veeco Caliber
atomic force microscope (Vecco Instruments Inc.) using silicon cantilevers operated under
tapping image mode. As-sputtered catalyst underwent a reduction process by purging H
2
gas
to the reactor starting at 735C. After for 45 seconds, the temperature raised to about 750C,
H
2
purging was stopped immediately and furnace was cooled down to room temperature.
Another process for preparing AFM samples was heating the catalyst to 750C under Ar
flow. Then samples were quickly cooled down to room temperature.


76

CHAPTER 3: Growing Long Carbon Nanotube Arrays using Intermetalic Fe-Zr
Catalyst
3.1 Background and Motivation
Metallic nanoparticles containing single and binary components have been known for
their catalytic properties to grow carbon nanotube (CNT) arrays. In this paper, an
intermetallic catalyst consisting of iron and zirconium was used to grow millimeter long,
well aligned arrays. The catalyst particles have well defined shape and were uniformly
distributed on the Al
2
O
3
buffer layer. The Fe-Zr particles enabled the growth of 1.7
millimeter-long carbon nanotube arrays in 45 minutes. Both the uniform catalyst particle
distribution and the long CNTs growth are attributed to the high thermal stability of Fe-Zr
catalysts. A comparison with pure iron catalyst indicated that adding Zr to iron can stabilize
the Fe catalyst at the CNT growth temperature and moderate its reactivity. AFM and SEM
images showed the different morphology evolutions and growth behaviors for Fe-Zr and Fe
catalysts. The long, uniform CNT arrays grown here have potential applications in many
advanced composites.
77

Carbon nanotubes (CNTs) have garnered extensive attention because of their excellent
thermal, mechanical and electrical properties.
1
CNTs can be grown using various techniques,
however, the production of ultra long (in the range of millimeters) CNTs is usually
accomplished by chemical vapor deposition (CVD) on nano-particle catalyst covered
substrates to produce aligned CNT arrays.
2-4
These arrays have been shown to be
advantageous when extremely large aspect ratio and alignment of CNTs are desirable.
Arrays have been the precursor for spinning high specific strength fibers,
5-8
which are
promising for future high strength composite materials.
6-8
The alignment inherent to the
arrays has shown the potential for producing composites with vertically aligned CNTs which
provide through-thickness multifunctionality to the composites.
9-11
Arrays have also been
used as a raw material for making buckypapers that contain much longer CNTs than are
normally achieved.
12
These CNT buckypapers could be used as reinforcement for advanced
composite materials. Achieving continuous growth of CNTs on a catalytic substrate is the
ultimate goal of many research groups and will not be realized until all of the issues facing
continuous growth are addressed.
78

An important issue hindering the continuous growth of CNT arrays is the thermal
stability of the catalyst particles. Growth of CNTs at higher temperature produces CNTs
with smaller diameter and fewer defects,
13,14
and defects-free CNTs with small diameters are
the strongest materials known.
15
However, at higher temperatures self-diffusion of metal
particles will enlarge the sizes of catalyst particles, which is well known as Ostwald
Ripening. In addition, small particles tend to agglomerate into larger, lower spatial-density
particles during the CNT growth.
16
Atomic-scale in-situ observation of carbon nanotube
growth reveals that catalysts are very mobile at the CNTs growth temperature.
17
For
example, iron particles are almost quasi-liquid and can change their shape dramatically in
seconds.
17

To overcome this challenge, many efforts have been taken to seek more stable
catalysts.
18-22
Numerous catalysts have been successfully used in the growth of carbon
nanotubes. These catalysts include single element metal and their binary alloys, such as Fe,
Co, Co-Mo and Ni.
18-22
However, none of these catalysts shows a high thermal stability, and
the CNT growth stops at a relatively short time. Attempts to grow longer nanotube arrays
79

without any other assistance, such as water vapor-assisted growth, have only reached limited
success thus far.
18-22

In this work, a new intermetallic catalyst, Fe-Zr, was used to grow 1.7 millimeter-long
CNTs array without water assistance in 45 minutes. To our best knowledge, pure zirconium
alone can not catalyze CNT growth. A previous study reported that the use of a Fe-Zr
catalyst only produced a 5 m tall CNT array.
23
The Fe-Zr catalysts made by magnetron
sputtering method have very high efficiency in growing CNTs and are very stable. AFM
images (Figure 3.1(d)) revealed the arrangement of small Fe-Zr particles on each Al
2
O
3

grain. Multiple motivations exist for making the intermetallic catalyst. First, the non-reactive
zirconium will decrease the activity of the catalyst, which might help to reduce the
amorphous carbon formation which usually is the main culprit for quenching carbon
nanotube growth.
24
After a short growth period, a catalyst substrate that supports CNT
growth will often appear clean when the CNT array is removed. After a long growth period,
the same substrate will be blackened by the buildup of amorphous carbon. Another function
of Zr is to form intermetallic bonds with Fe. These strong bonds may stabilize the particle
and thus reduce the catalysts susceptibility to coarsening. In addition, the formation of ZrC
80

can consume the amorphous carbon nucleated on the surface of iron catalysts.
23

Consequently the catalysts may remain active longer to increase the CNT array growth time
and ultimate height.
3.2 Experimental
Fe and Fe-Zr thin films were deposited under identical conditions using a custom built
magnetron sputtering system with a 2.5 cm diameter cathode. A silicon substrate was
cleaned with acetone for 5 minutes by super-sonication followed by rinsing in methanol.
Sputtering was carried at pressure of 110
-3
torr under argon gas with a 10 sccm (standard
cubic centimeter per minute) flow rate. Thickness of the films was monitored by an in-situ
quartz crystal microbalance. A 10 nm thick Al
2
O
3
buffer layer was first deposited with 40
watts of power. Sputtering power was 10 watts for Fe and Fe-Zr, producing a catalyst layer
of 1.5 nm thickness in 107 seconds. Energy dispersive X-ray spectroscopy (EDS) analysis of
the sputtered Fe-Zr metal films gave a composition of 87 atomic % iron and 13 atomic %
zirconium.
As-sputtered catalysts were cut into 12 pieces from an approximately 2.5 cm
2
sample.
Two key morphology evolution stages of both Fe and Fe-Zr catalysts were studied by Veeco
81

Caliber atomic force microscope (Vecco Instruments Inc.) using silicon cantilevers operated
under tapping image mode. The power of the tapping amplitude was adjusted to 4.5V with a
set point of 2.6V~2.8V for the sample engagement.
CNTs arrays were grown at 750 using a CVD method in a 2.5 cm tube furnace. The
total gas flow rate was 200 sccm consisting of 40 sccm H
2
, 30 sccm C
2
H
4
and 130 sccm Ar.
The procedure was as follows: (1) The tube was pumped to 25 Torr and then refilled with
argon. The furnace was then ramped to the growth temperature under a flow of argon. (2)
When the furnace reached 735 the H
2
flow was added to the argon. (3) When the furnace
reached 750 ethylene flow was added to initiate growth. Growth time was calculated from
this point. All the CVD growth parameters for the Fe and Fe-Zr catalysts were constant.
Morphology of the CNTs arrays was observed using FE-SEM (Model: Hitachi SU-70).
Array height was measured using FE-SEM (Model: JEOL 6400).
3.3 Results and Discussion
Catalyst particle formation, which is critical for the nucleation and growth of CNT
arrays, occurs after an annealing step in hydrogen atmosphere. For both of the catalyst
systems, catalyst particle formation ensued even when this step was not used. Figure 3.1(a)
82

and (c) show the particle distribution when the samples were heated to 750 under Ar
atmosphere and with a heating rate of 100/min. This result was different than another
groups report, where the surface of the catalyst film was quite smooth before the reduction
of the iron film by H
2
.
25
The iron film should dewet on the aluminum oxide buffer layer and
separate into nano-particles once hydrogen reduces the catalyst. The catalyst film has to be
reduced because the film is thin enough (usually around 1 nm) that it is oxidized
immediately by the ambient atmosphere once taken out of the sputtering chamber. This was
confirmed by the presence of iron oxides and no metal signal in the X-ray photoelectron
spectroscopy (XPS) spectrum. The particles observed in Figure 3.1(a) and Figure 3.1(c) are
Al
2
O
3
grains, not catalyst particles, because the 1.5 nm catalyst film could not form a
structure with such a high roughness. The grain size of the Al
2
O
3
buffer layer after annealing
is larger with the Fe-Zr catalyst layer than the pure Fe layer. The fact that the catalyst film
became bumpy before the reduction step by hydrogen could be linked to the morphology and
process for depositing the buffer layer.
The unique particle formation observed in this study probably resulted from the method
used to deposit Al
2
O
3
. In this case, the buffer layer was made by sputtering a pure Al
2
O
3

83

target using magnetron sputtering method under pure Ar atmosphere. Magnetron sputtered
Al
2
O
3
films have shown to be non-stoichometric and rich in aluminum when sputtered in an
argon environment.
26
Metallic aluminum in a non-stoichometric buffer layer could explain
the Al
2
O
3
morphology change and self reduction of the metal catalyst layer. When heated in
the inert atmosphere, the buffer layer, which was rich in aluminum, will react with the
oxidized catalyst layer. This is the common thermite reaction and is highly energetically
favorable. The reaction changed the morphology of the buffer layer from flat to islands and
self reduced the catalyst into metal for particle formation.



84


Figure 3.1 AFM images show the evolution of catalysts. (a) Al
2
O
3
and Fe films became
rough and bumpy when Fe catalysts were heated to 750. (b) After Fe catalysts were
annealed under H
2
and Ar for 45 seconds, further coarsening resulted in very irregular
shaped particles and an irregular distribution. (c) Al
2
O
3
islands formed when Fe-Zr catalysts
were heated to 750. (d) Fe-Zr tiny particles appeared on top of the well aligned Al
2
O
3

grains when Fe-Zr catalysts were annealed under H
2
and Ar for 45 seconds.




85



Figure 3.2 Cross section view of CNT arrays grown using catalyst (a) Fe for 15 minutes, (b)
Fe-Zr with 15 minutes growth (c) Fe, a lower magnification view of A, (d) Fe-Zr, a lower
magnification view of B, (e) Fe after 15 seconds growth. All the SEM images are taken at
the roots of the CNT arrays.

86

The H
2
pretreatment before CNTs growth coarsened the Fe samples but refines the Fe-
Zr samples. Figure 3.1(b) and (d) clearly showed the evolution of catalyst particles. For the
iron sample, the 45 seconds annealing under H
2
and Ar yielded irregular grain shapes.
Compared with Figure 3.1(a), the alumina particles grew much larger, and the shape became
more irregular. In order to minimize the surface energy, the non-uniform iron particles
possibly distributed in the valley formed by the larger bumpy alumina grains. We assume
the shape of the Fe particles could be very irregular considering the non-uniformity of the
alumina grains. In contrast, catalyst distribution on the buffer layer was much more uniform
for the Fe-Zr samples. In Figure 3.1(C), Fe-Zr particles were not quite visible. The clusters
were observed sitting on top of Al
2
O
3
grains after 45 seconds annealing under H
2
and Ar as
shown in Figure 3.1(d). The Fe-Zr particles and alumina grains were much more uniformly
distributed than that of Fe samples. The CNT growth produced by each type of sample will
be proposed and discussed later in the paper based on the buffer layer and catalyst particle
morphology seen here.
Intermetallic Fe-Zr has a higher melting point than that of pure iron because zirconium
can form ionic bonds with iron, which will greatly decrease the diffusion rate. The design of
87

Fe-Zr catalysts is to survive longer at CNTs growth temperature, and at 750, the Fe-Zr
particles exhibit this enhanced stability.
Both Fe and Fe-Zr catalysts were used to grow CNTs arrays with the same growth
conditions for direct comparison. The morphology of CNTs from the side view of the arrays
is shown in Figure 3.2(a)-(e). Clearly, CNTs grown with the two different catalysts systems
varied widely in terms of shape and the growth behavior. CNTs grown by Fe were coiled,
curled, twisted in columns; while CNTs grown by Fe-Zr were straight and well-spaced from
other growing CNTs. This can be explained by the morphology differences of the catalysts
particles.
As shown in Figure 3.2(a) and (c), CNTs grown using Fe catalysts were kinked, twisted
and grown in vertical columns. The column spacing was correlated to the spacing of alumina
grains indicating the Fe catalysts were separated by the bumpy Al
2
O
3
grains. Noticing that
both the size and the spacing of the CNT columns varied randomly, together with the
evidence of the irregular shape and non-uniform distribution of the Al
2
O
3
grains, we assume
that Fe particles have a very inhomogeneous distribution and non-sphere shape. With this
unique catalyst morphology in mind, some special features can be interpreted. For example,
88

CNTs grown by Fe catalysts had more defects and amorphous carbon, which was confirmed
by Raman shift spectrum. The outer shells of the CNTs were very rough and non-tubular. All
of these features are possibly corresponding to the irregular catalyst morphology.
Figure 3.2(e) is a sample grown by Fe for only 15 seconds. It clearly showed that the
CNTs coiled at the very beginning of the growth. This is in contrast to another study where it
was reported that the curling of the CNTs was time dependent: straight CNTs gradually
became curving during the growth because the increase of the CNT gravitational weight
caused the bending the CNTs.
18

Compared with Fe catalysts, CNTs grown by Fe-Zr were straighter and had fewer
defects and less amorphous carbon, as indicated by the SEM and Raman data. We propose
that this was due to more stable catalyst shape and uniform Fe-Zr particle distribution on
Al
2
O
3
grains. With the same catalyst preparation and CNT growth condition, Fe-Zr produced
higher quality CNT arrays. The as-sputtered Fe and Fe-Zr films experienced different
morphology evolution under the same heat treatment, and consequently grew CNT arrays
with different morphologies. The intermetallic Fe-Zr particles had a higher thermal stability
in catalyzing CNT growth than pure iron particles.
89

Fe-Zr catalysts also grew much longer CNT arrays than pure Fe catalysts. An
approximately 350 m array was grown by Fe in 15 minutes. Longer growth time only
yielded an array about 370 m in height (estimated using an optical microscope with scale
bar). Thus, the growth was assumed to stop at approximately 15 minutes. The Fe-Zr catalysts
produced arrays with a height of 0.7mm in 15 minutes, and a 1.7mm in 45 minutes, as shown
in figure 3. Array height versus growing time was plotted in Figure 3.4 for both Fe and Fe-Zr
catalysts. Fe-Zr catalysts showed a much longer survival time than pure Fe. The longer
growth time again confirmed the high thermal stability of the Fe-Zr particles.



Figure 3.3 SEM images clearly show the height of the CNT arrays grown using Fe-Zr
catalysts in 45 minutes.
90





Figure 3.4 CNT array height for different growth time using Fe and Fe-Zr catalysts.

Not only did the Fe-Zr particles produce a more aligned CNT array, but they also
produced cleaner nanotubes. Figure 3.5 showed the Raman spectrum for CNTs grown for 15
minutes using both iron and iron-zirconium catalysts. The quality of the CNTs can be
evaluated based on the intensity of D peaks and G peaks. G peak corresponds to the
stretching of carbon-carbon bonds in the grapheme plane. D peak corresponds to the defects
91

and amorphous carbon. The I
G
/I
D
ratio is much larger for the CNTs grown using the inter-
metallic catalyst, indicating less amorphous carbon. The quality of CNTs is closely related to
the activity of catalysts.
27,28
It was observed that Co-Mo bi-metallic catalyst particles with
higher Mo content grew longer, higher quality CNTs than those with lower Mo content. Mo
is a non-reactive element and Co was the active element that catalyzed the dissociation of
CO carbon source gas to grow CNTs. Higher Mo content leads to lower catalyst reactivity,
which consequently produced less amorphous carbon. At lower Mo content, the catalyst was
so reactive that it produced more free carbon atoms than can be used by the CNT growth.
The excessive free carbon atoms formed amorphous carbon on the catalyst surface and
eventually encapsulated the catalyst particle and stopped the CNT growth.
27,28
We believe
that similar situation occurred during the CNT growth in this study. In other words, with the
addition of Zr to Fe, the catalyst reactivity is reduced, which produced the longer, higher-
quality CNT arrays with less amorphous carbon, as compared with the pure Fe catalyst.
The authors acknowledge that the height of CNT arrays grown using pure iron is lower
than previously reported results.
2-4
This is most likely due to the high roughness of the buffer
layer created by sputtering and annealing. Nevertheless, the improvement in catalyst lifetime
92

provided by the inter-metallic Fe-Zr observed in this study provides a new approach for
growing long and high-quality CNT arrays, and should be studied further.



Figure 3.5 Raman shift spectrum for CNT arrays grown using iron and iron-zirconium
catalysts.



93

3.4 Conclusions
In summary, intermetallic Fe-Zr catalysts were co-sputtered using a magnetron
sputtering method. The Fe-Zr film formed uniformly distributed catalyst particles on Al
2
O
3

buffer layer. The Fe-Zr particles were more thermally stable than pure iron particles and
produced 1.7 mm long CNT arrays in 45 minutes. Pure iron catalysts prepared using the
same method and same experimental conditions had a lower performance in growing CNT
arrays. In combination with other techniques used to extend the lifetime of CNT growth of
CNT arrays, such as water assisted growth, new record breaking CNT array lengths might be
possible. These ultra long CNT arrays will be critical for providing the long nanotubes
needed in multifunctional, high performance composite materials.







94

REFERENCES
1. Zhang X.; Li Q.; Tu Y.; Li Y.; Coulter J. Y.; Zheng L.; Zhao Y.; J ia Q.; Peterson D.
E.; Zhu Y. T., Small 3, 244-248 (2007).

2. Chakrabarti S.; Gong K.; Dai L., Journal of Physical Chemistry C 112, 8136-8139
(2008).

3. Noda S.; Hasegawa K.; Sugime H.; Kakehi K., Japanese Journal of Applied Physics
46, 399-401 (2007).

4. Zhong G.; Iwasaki T.; Robertson J.; Kawarada H., Journal of Physical Chemistry B
111, 1907-1910 (2007).

5. Dalton A. B.; Collins S.; Munoz E.; Razal J. M.; Ebron V. H.; Ferraris J. P.;
Coleman J. N.; Kim B. G.; Baughman R. H., Nature 423, 703-704 (2003).

6. Zhang M.; Atkinson K. R.; Baughman R. H., Science 306, 1358-1361 (2004).

7. Koziol K.; Vilatela J.; Moisala A.; Motta M.; Cunniff P.; Sennett M.; Windle A.,
Science 318, 1892-1895 (2007).

8. Zhang X. F.; Li Q.; Holesinger T. G.; Arendt P. N.; Huang J. Kirven P. D.; Clapp T.
G.; DePanula R. F.; Liao X. Z.; Zhao Y. H.; Zheng L. X.; Peterson D. E.; Zhu Y. T.,
Advanced Materials 19, 4198-4201 (2007).

9. Ivanov I.; Puretzky A.; Eres G.; Wang D. H.; Pan Z.; Cui H.; J in R.; Howe J.;
Applied Physics Letters 89, 223110-223112 (2006).

10. Peng H.; Sun X., Chemical Physics Letters 471, 103-105 (2009).

11. Raravikar N. R.; Schadler L. S.; Vijayaraghavan A.; Zhao Y.; Wei B.; Ajayan P. M.,
Chemistry of Materials 17, 974-983 (2005).
95

12. Wang D.; Song P.; Liu C.; Wu W.; and Fan S., Nanotechnology 19, 075609-075614
(2008).

13. Cheung C. L.; Kurtz A.; Park H.; Lieber C. M., Journal of Physical Chemistry B 106,
2429-2433 (2002).

14. Yao Y.; Li Q.; Zhang J.; Liu R.; J iao L.; Zhu Y. T.; Liu J., Nature Materials 6, 283-
286 (2007).

15. Zhao Q.; Nardelli M. B., Physical Review B 65 144105-144106 (2002).

16. Placidus B. A.; Cary L. P.; Laura M.; Seung M. K.; Eric A. S.; P. T. M.; Robert H.
H.; Benji M., Nano Letters 9, 44-49 (2009).

17. Yoshida H.; Takeda S.; Uchiyama T.; Kohno H.; Homma Y., Nano Letters 8, 2082-
2086 (2008).

18. Wu J.; Huang Q.; Ma Y.; Huang Y.; Liu Z.; Yang X.; Chen Y.; Chen D., Colloids
Surface 13, 313-314 (2008).

19. Zhang H.; Cao G.; Wang Z.; Yang Y.; Shi Z.; Gu Z., Journal of Physical Chemistry
C 112, 4524-4530 (2008).

20. Jeong G. H.; Suzuki S.; Kobayashi Y.; Yamazaki A.; Yoshimura H.; Homma Y.,
Applied Physics Letters 90, 043108-043111 (2007).

21. Sugime H.; Noda S.; Maruyama S.; Yamaguchi Y., Carbon 47, 234-241 (2009).

22. Lee S.; Moon S.; Yoon H. S.; Wang X.; Kim D. W.; Yeo I. S.; Chung U. I.; Moon J.
T.; Chung J., Applied Physics Letters 93, 182106-182108 (2008).

23. Huang L.; Lau S. P.; Zhang Y. B.; Tay B. K.; Fu Y. Q., Nanotechnology 15, 663-666
(2004).
96

24. Stadermann M.; Sherlock S. P.; In J.; Fornasiero F.; Park H. G.; Artyukhin A. B.;
Wang Y.; Yoreo J. J.; Rigoropoulos C. P.; Bakajin O.; Chernov A. A.; Noy A.,
Nano Letters 9, 738-744 (2009).

25. Nessim G. D.; Hart A. J.; Kim J. S.; Acquaviva D.; Oh J.; Morgan C. D.; Seita M.;
Leib J. S.; Thompson C. V., Nano Letters 8, 3587-3593 (2008).

26. Vuoristo P.; Mantyla T.; Kettunen P.; Lappalainen R., Thin Solid Films 204, 297-302
(1991).

27. Liao X. Z.; Serquis A.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Applied Physics Letters
82, 2694-2695 (2003).

28. Serquis A.; Liao X. Z.; Huang J. Y.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Carbon 41,
2635-2641 (2003).










97

CHAPTER 4: Growing Carbon Nanotube Arrays using One-Step Chemical Vapor
Deposition Method
4.1 Background and Motivation
Vertically aligned CNT arrays are usually grown via CVD method using Fe/Al
2
O
3
/Si
substrates, which requires a pre-deposition of catalysts before transferring the substrates to
the CVD reactor.
1-9
There are basically two steps to complete the CNT growth: catalyst thin
film preparation and CNT growth. Although materials for catalyst film vary from Fe, to Co,
Ni, Pt etc
10-19
and thin film preparation technique varies from magnetron sputtering to e-
beam evaporation, the process still includes two steps.
The vertically grown CNT arrays could be shear-pressed to horizontally aligned CNT
papers for producing high volume fraction advanced CNT composites.
20
With finely
controlled growth, CNT sheets could be drawn from the vertically array continuously.
21

Ultra-strong, ultra-stiff CNT fibers are spun from the drawable array to use as sensors or
composite fillers.
22
Aligned CNT sheets are also used as composite perform for fabricating
CNT composites. CNT sheet drawn directly from vertically grown CNT arrays have shown
to be superior to CNT electrodes fabricated using filtration method because the alignment of
98

CNTs is maintained and transferred from vertical direction to the CNT drawing direction.
Therefore, the aligned CNT sheet electrodes have better electric conductivities than those
with randomly-oriented CNTs.
23

Both non-drawable and drawable CNT arrays show their potential applications.
However, traditional CNT array growth method requires two steps. The cost to fabricate
such vertically grown arrays is high because the quantity and size of arrays are restricted by
the size of silicon wafers where arrays are grown and difficulty in scaling up the high
vacuum catalyst deposition systems. What is more, there are only several limited groups in
the world that can produce such spinnable CNT arrays. Therefore, a new method could
overcome the limitation of the traditional growth method has to come up to meet the
requirements of mass production of such CNT arrays.
One-step chloride-assisted CNT growth was first reported by Dr.Inoue.
24
This
technique does not require the catalyst predeposition step and super spinnable arrays could
grow on large quartz plates. Several advantages of this method are: (1) The CNT growth is
simple, only one-step is required; (2) The growth rate is up to 100 m/min; (3) CNT array
are super drawable. Sheets are pulled out continuously up to 40 meters. The length is only
99

limited by the size of the sample; (4) This process is easy to be scaled up since it is simple;
(5) The efficiency in suing acetylene is as high as 25%.
However, the CNT growth time is limited by 20 minutes in the previous study.
24
CNTs
stop growing quickly at around 20 minutes. With the QMS data analysis, Dr.Inoue assumes
the CNT growth process with the following four reactions.
24
Firstly, FeCl
2
(in gas phase)
reacts with C
2
H
2
producing FeC
2
particles. FeC
2
particles colloid with each other and grow
larger. Finally FeC
2
nucleates and deposits on the quartz substrates. Carbon is saturated in
Fe and quickly nucleates out from the particles and grown into carbon nanotubes. HCl gas
trapped in or near the particles could react with Fe producing some FeCl
2
again. Reaction (4)
is repetitive of reaction (1).

Noticing that the amount of HCl is decreasing during the CVD process since the reactor
is under pumping. The above cyclic reactions will stop because of insufficient of HCl.
Hereby, in this paper, hydrogen chlorine gas was used to assist one-step chemical vapor
deposition to grow super spinnable long CNT arrays from which CNT sheets could be drawn
100

continuously. With the assistance of HCl gas, the ultimate array height was increased from
2.1mm to 3mm and the height of drawable CNT arrays was increased from 1mm to 2mm.
4.2 Experimental
Quartz tubes and quartz plates were first soaked in sulfuric acid (about 10 m%
concentration) for at least 24 hours to remove residuals on the surface of tubes and plates.
After rinsing with water and then ethanol, the quartz tubes and plates dried in air. Anhydrous
FeCl
2
powders directly purchased from chemical suppliers were loaded into a quartz boat
and the quartz boat was inserted in the middle of a liner quartz tube with diameter of around
1inch. The tube reactor was pumped down to at least 110
-3
torr in one hour followed by a
heating up process. The system was kept in vacuum during heating. FeCl
2
powders started to
evaporate at round 600C and fully became vapor at the growth temperature of 820C. The
hot zone of the quartz tube was about 30 cm. Part of FeCl
2
re-deposited at the ends of the
heated zone. After reached 820C, acetylene started to flow into the reactor. CNTs began to
grow. A 2.1mm-tall CNT array was grown in 20 minutes with a growth rate of 100 m/min.
A drawable CNT array could be grown in 7 minutes with a height of around 700 m. CNTs
101

grew not only on the quartz substrates but also on the inner wall of the tube reactor. To
protect the main tube from growing CNTs, a liner tube was inserted in the main tube.
HCl solution with concentration of 30% was put in a bubbler. A flow of carrying gas
produced some bubbles in the bubbler and pick up some HCl vapor to the reactor. A
Quadruple Mass Spectrometer was used to analyze the gas in-situ during CNT growth.
Quality of the array was examined using Raman spectra. Microstructure and morphology of
the CNTs were analyzed using SEM and TEM.
4.3 Results and Discussion
Figure 4.1 shows the CNT array height at different growth time. CNT stops growth at
around 20 minutes if no extra HCl gas flows into the CVD system. While with bubbling
additional HCl into the reactor, the growth time extends to about 60 minutes. At the first
twenty minutes growth, the growth rate of HCl gas assisted method is around 100 m/min,
which is pretty close to the growth rate without HCl. After 20 minutes, CNT grows slower
than before and finally stops after 1 hour.

102


Figure 4.1 Array height at different growth time using HCl assisted growth.
Figure 4.2 shows a 3.0 mm tall CNT array grown in 60 minutes using HCl assisted one-
step CVD method. The previous record using one-step CVD method is 2.1 mm. Figure 4.3
shows CNT ribbons being pulled out using a tweezer by hand. The spinnable array grown
with the assistance of HCl gas is about 2 mm tall, which is grown in 20 minutes. The
previous record of the length of spinnable array using one-step CVD method is 1 mm.
103


Figure 4.2 A SEM image of CNT arrays with length of 3.0 mm.


Figure 4.3 Picture of 2 mm tall spinnable CNT array showing that CNT ribbons are drawn
continuously using a tweezer.
104

It is interesting that spinnable arrays have to be grown at a certain time window. For
example, spinnable CNT array could be achieved if grown between 5 minutes and 10
minutes using one- step CVD method without extra HCl assistance. While the time window
could be expanded from 5 minutes to 20 minutes with HCl gas assisted growth. Why only
CNTs grown at certain time can be drawn from the forests?
Here we assume that entanglement, van der Waals force and amount of amorphous
carbon together attribute to the spinnability of CNT arrays. Several theories about the
spinnability of CNT arrays have been developed since the discovery of the continuous
drawing of ribbons. Zhang et al.
25
Claims that CNTs are bundled together similarly to the
connected loops due to the bundling in the disordered region at the top and bottom of the
forests. The entanglement allows the continuous pulling of CNTs. However, this explanation
seems to be weak in elucidating the spinnability of super-aligned arrays where CNTs are
highly ordered at the bottom.
26
The evidence that such arrays have clean surface may support
the idea that the interactions of neighboring CNTs due to the strong van der Waals force
connect the CNTs when pulled out from the array.
26
However, even there is no entanglement
at the top or bottom of the array, overall, CNTs still migrate between neighbors as they grow.
105

Thousands of SEM images show that CNTs do not grow in a perfect straight line. The
moderate entanglement helps CNT connect with each other when they are pulled out.
CNTs grown in a short time have a length of a few hundred micrometers. As van der
Waals force is a function of contact area, CNTs with very short length have a small van der
Waals force between them that is not strong enough to allow one CNT takes away its
neighbors when pulled out. As CNTs grow longer, contact surface between CNTs is larger.
As a consequence, van der Waals force is large enough to allow CNTs stick to each other.
However, if CNTs grow too long, a large CNT chunk is pulled out when trying to pull out
CNT ribbons probably because the amorphous carbon decreases the interaction between
CNTs. As CNT grows, more and more amorphous carbon deposits on the CNTs and
catalysts, which is also one of the factors that CNTs stop growing. Amorphous carbon
deposition causes the increase of roughness of CNT surface. Therefore interaction between
CNTs decreases.
Here we assume that less amorphous carbon is produced during the growth when HCl is
added to the system. According to the analysis in the previous paragraph, less amorphous
carbon accumulation is beneficial for growing long spinnable arrays. Figure 4.4 clearly
106

shows CNTs grown using HCl have a cleaner surface that the ones grown without HCl.
CNTs grown with HCl have a smooth surface. The diameters of the CNTs are around 45 nm.
Without HCl assistance, more amorphous carbon is accumulated on the CNT walls. The
diameters of the CNTs are much larger. A previous study also discovered that diameters of
CNTs increased as the growth time increased.
27
The author stopped CNT synthesis after 25
minutes growth by briefly introducing a flow of pure C
2
H
4
in the tube to smother the catalyst
with amorphous carbon and then cooling the furnace to room temperature under a flow of
pure argon. High resolution TEM revealed that additional graphitic layers grew in the tubes
after they stopped growing longer. In our study, CNTs stop growing vertically after 20
minutes if no HCl gas is used. Additional growth time only increases the CNT diameters.
CNTs grown in 60 minutes under HCl gas flow also have higher quality than without
HCl, which is revealed by the Raman spectra in Figure 4.5. G-peak/D-peak ratio of CNTs
grown with HCl is greater than 1 while the ratio reversed for CNTs grown without HCl
indicating more defects and amorphous carbon are in the tubes.



107

Figure 4.4 SEM images of CNTs grown in 60 minutes (a) with HCl assistance; (b) without
HCl assistance.
(a)
(b)
108


Figure 4.5 Comparison of CNT arrays growing with HCl assistance and without HCl
assistance in 60 minutes growth.
To explore the mechanism of HCl assisted CNT growth, partial pressure of H
2
and HCl
were monitored in situ during growth. As shown in Figure 4.6, there is an increase of H
2

produced in the CVD reactor when the supplying of HCl gas increases. The product of H
2

probably comes from the following reaction.
(1+z) FeC
y
+2HClFeCl
2
+Fe
z
C
y+z
+H
2

HCl gas probably reacts with the part of the Fe catalysts producing the initial catalyst
loaded in the chamber (FeCl
2
). FeCl
2
will vaporize at the growth temperature. We assume
that the vapor will quickly reacts with acetylene again around the Fe catalysts and redeposit
109

on the Fe catalysts.

This unique growth process probably keeps the catalyst active longer and
produced less amorphous carbon. There are more than 10 ingredients in the CVD reactor that
could react with each other. It is difficult to clearly understand how HCl gas helps CNTs to
grow longer. The mechanism needs to be further studied.

Figure 4.6 In-situ monitoring of partial pressure change of H
2
and HCl during CNT growth
using QMS.
4.4 Conclusions
One-step CVD method is superior to traditional two-step CVD method to grow CNT arrays.
It is economically preferable to grow CNTs using one-step CVD method. To further improve
110

this method, HCl gas is used to assist the one-step CVD method to grow long carbon
nanotube arrays. 3 mm tall CNT forests are grown in 60 minutes and 2 mm tall spinnable
arrays are grown in 20 minutes. HCl gas could reduce the amorphous carbon during the CNT
growth which extended the growth time.













111

REFERENCES
1. Chakrabarti S.; Gong K.; Dai L., Journal of Physical Chemistry C 112, 8136-8139
(2008).

2. Noda S.; Hasegawa K.; Sugime H.; Kakehi K., Japanese Journal of Applied Physics
46, 399-401 (2007).

3. Zhong G.; Iwasaki T.; Robertson J.; Kawarada H., Journal of Physical Chemistry B
111, 1907-1910 (2007).

4. Wu J.; Huang Q.; Ma Y.; Huang Y.; Liu Z.; Yang X.; Chen Y.; Chen D., Colloids
Surface 13, 313-314 (2008).

5. Zhang H.; Cao G.; Wang Z.; Yang Y.; Shi Z.; Gu Z., Journal of Physical Chemistry
C 112, 4524-4530 (2008).

6. Jeong G. H.; Suzuki S.; Kobayashi Y.; Yamazaki A.; Yoshimura H.; Homma Y.,
Applied Physics Letters 90, 043108-043111 (2007).

7. Sugime H.; Noda S.; Maruyama S.; Yamaguchi Y., Carbon 47, 234-241 (2009).

8. Lee S.; Moon S.; Yoon H. S.; Wang X.; Kim D. W.; Yeo I. S.; Chung U. I.; Moon J.
T.; Chung J., Applied Physics Letters 93, 182106-182108 (2008).

9. Kamalakar G.; Hwang D. W.; Hwang L. P., Journal of Materials Chemistry 12,
1819-1823 (2002).

10. Rong C. B.; Poudyal N.; Chaubey G. S.; Nandwana V.; Liu Y. Z.; Wu Y. Q.; Kramer
M. J.; Kozlov M. E.; Baughman R. H.; Liu J. P., Journal of Applied Physics 103,
07E131-07E133 (2008).

112

11. Rong C. B.; Poudyal N.; Chaubey G. S.; Nandwana V., Journal of Applied Physics
102, 043913-043918 (2007).

12. Liao X. Z.; Serquis A.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Applied Physics Letters
82, 2694-2695 (2003).

13. Serquis A.; Liao X. Z.; Huang J. Y.; J ia Q. X.; Peterson D. E.; Zhu Y. T., Carbon 41,
2635-2641 (2003).

14. Saito Y.; Tani Y.; Journal of Physical Chemistry B 104, 2495-2499 (2000).

15. Rong C. B.; Nandwana V.; Poudyal N.; Liu J. P.; Kozlov M. E.; Baughman R. H.;
Ding Y.; Wang Z. L., Journal of Applied Physics 102, 023908-023914 (2007).

16. Mauron P.; Emmenegger C.; Sudan C.; Wenger P.; Rentsch S.; Zuttel A., Diamond
and Related Materials 12, 780-785 (2004).

17. Herrera J. E.; Balzano L.; Borgna A.; Alvarez W. E.; Resasco D. E., Journal of
Catalysis 204, 129-145 (2001).

18. Jong W. J.; Lai S. H.; Hong K. H.; Lin H. N.; Shih H. C., Diamond and Related
Materials 11, 1019-1025 (2002).

19. Yuan D.; Ding L.; Chu H.; Feng Y.; McNicholas T. P.; Liu J., Nano Letters 6, 2642-
2645 (2006).

20. Bradford P. D; Wang X.; Zhao H.; Maria J. P.; J ia Q. X.; Zhu Y. T., Composite
Science and Technology in press (2010).

21. Zhang X. F.; Li Q.; Holesinger T. G.; Arendt P. N.; Huang J. Kirven P. D.; Clapp T.
G.; DePanula R. F.; Liao X. Z.; Zhao Y. H.; Zheng L. X.; Peterson D. E.; Zhu Y. T.,
Advanced Materials 19, 4198-4201 (2007).
113

22. Zhao H.; Zhang Y.; Bradford P. D.; Zhou Q.; J ia Q. X.; Yuan F. G.; Zhu Y. T.,
Nanotechnology 21, 305502-305506 (2010).

23. Zhang H. X.; Feng C.; Zhai Y. C.; J iang K. L.; Li Q. Q.; Fan S. F., Advanced
Materials 21, 2299-2304 (2009).

24. Inoue Y.; Kakihata K.; Hirono Y.; Horie T.; Ishida A.; Mimura H., Applied Physics
Letters 92, 213113-213115 (2008).

25. Zhang M.; Atkinson K. R.; Baughman R. H., Science 306, 1358-1361 (2004).

26. Zhang X. B.; J iang K. L.; Feng C.; Liu P.; Zhang L. N.; Kong J.; Zhang T. H.; Li Q.
Q.; Fan S. S., Advanced Materials 18, 1505-1510 (2006).

27. Bradford P. D., Doctoral Dissertation, North Carolina State University (2010).












114

CHAPTER 5: Carbon Nanotube Fiber Strain Sensors
5.1 Background and Motivation
Carbon nanotubes (CNT) based sensors are often fabricated by dispersing CNTs into
different types of polymer. In this paper, a prototype carbon nanotube (CNT) fiber strain
sensor with excellent repeatability and stability for in-situ structural health monitoring was
developed. The CNT fiber was spun directly from CNT arrays, and its electrical resistance
increased linearly with tensile strain, making it an ideal strain sensor. It showed consistent
piezoresistive behavior under repetitive straining and unloading, and good resistance
stability at temperatures ranging from 77K to 373K. The sensors can be easily embedded
into composite structures with minimal invasiveness and weight penalty. We have also
demonstrated their ability to monitor crack initiation and propagation.
The excellent piezoresistive property of individual carbon nanotubes (CNT) has been
demonstrated by several groups.
1-5
These studies generally involved measuring the
conductivity of individual CNTs under loading using an AFM tip or a micro-
electromechanical device. Although individual CNTs exhibit exciting properties for potential
application as sensors, in reality, integrated CNT-based nano-sensors have not been realized
115

due to the challenges associated with their integration into bulk structures. Assembling these
nano-sensors into microscopic sensors for easy integration into bulk structures is important
for practical applications. Some success has been reported on scaling up these nano-sensors
by dispersing them into polymers via different approaches .
6-12
These methods have at least
one feature in common: the sensor fabrication involves dispersing a small amount of CNTs
into a polymer matrix to form a 3-D CNT network for electron conduction. The addition of
CNTs dramatically increases the conductivity by forming electric pathways in the otherwise
insulating polymer.
13
As such, the composite itself can sense strain under loads by
monitoring the resistance change.
Recent research in developing CNT-based strain sensors has generated some
excitement, because CNTs, as compared to traditional strain sensors, offer the potential of
being embedded within composite structures for in-situ measurement. Due to their high
strength, small diameter and light weight, CNT based-sensors can overcome several
limitations of the existing conventional sensors, including limited monitoring locations,
fixed directions and separation from the structure that is being monitored. However, several
challenges still remain before CNT-based sensors are fully implemented into real structures.
116

Two significant issues that need to be addressed are repeatability and stability.
Repeatability refers to repeatable resistance-strain behaviors under cyclic service conditions
and stability refers to stable resistance. The resistance of CNT/polymer film sensors have
been observed to change over time even without an applied mechanical load,
8,9
which poses
a serious problem in practical applications. Resistance drifting may result from defects and
breakdown of the shell structure of CNTs,
14
as well as heating from electric current passing
through the CNT sensors.
8
In addition, CNT-composite strain sensors were found to exhibit
resistance hysteresis in a cyclic strain loading,
6,11,12
which indicates irreversible deterioration
of CNT/polymer interfaces. CNT buckypapers have also been used as strain gauges.
14-16

However, due to their very low strength and permanent deformation after a strain of
approximately 0.04%, they are not ideal for applications that require multi-functionality or
repeatability. CNT fibers have been reported as actuators and sensors when they were
immersed in electrolyte and charged at high voltage.
17
However, it is not convenient to
embed fiber sensors with electrolyte in composites. Here we report using neat CNT fibers
directly as peizoresistive strain sensors which could be readily incorporated in composites
for in-situ health monitoring. The fiberss do not require any chemical grafting or charging to
117

work as sensors. To our knowledge, this is the first report of using the piezoresistive
properties of pure CNT fibers to make strain sensors.
5.2 Experimental
CNT arrays were grown using a CVD method described before.
18
CNT ribbons
approximately 3 mm wide were pulled out from an CNT array, grown on silicon substrates,
by hand using flat tweezers and then attached on a spindle using a piece of double-side tape.
A CNT fiber was spun by a rotating spindle that was moved away from the array at the same
time. The rotating speed was 7000 rpm and the fibers take-up speed was about 2 mm s
-1
. The
diameters of the as-spun fibers were 3 to 30m and were controlled by the width of ribbons
initially pulled out from the array. The fiber diameters were measured using laser diffraction
method, whose accuracy was verified by diameter measurements in a SEM.
CNT fibers were collected on sandpapers with a rectangular slot cut out, as shown in
Figure 5.1c. Fibers in the slot region were covered by a plastic film to protect from
deposition and the exposed parts were coated with a 5m thick aluminum layer using
magnetron sputtering in pure Ar at a pressure of 210
-3
Torr. The sputtering power was 40
watts. Silver paste was used to connect two thin copper wires with fibers that were covered
118

by the sputtered aluminum contacts. The gauge length of the fibers was 10 mm for in-situ
resistance measurement during mechanical testing.
CNT fibers were attached on a piece of release film that is commonly used for vacuum
bagging of composites. High conductive silver colloid was dropped on the ends of the fibers
to form electrodes. The resin system (59% EPON
TM
Resin826, 21% EPIKURE
TM
9551
Curing Agent, 20% Heloxy
TM
Modifier62 in weight percent, purchased from Hexion) was
mixed using magnetic stirrer. Bubbles trapped in the resin were removed by degassing in a
vacuum oven. Resin was cast on the fibers to form thin films and after curing at 373K for 3
hours, an integral structure of composites with built-in fiber sensing elements was produced.
The composite thin films were around 100m thick and were cut into coupons using a razor
blade. Two copper wires were soldered on the silver electrodes for resistance tests.
SEM images of the CNT fibers were taken using a JEOL 6400F at 5 kV and a working
distance of 15 mm. An EZ-S tensile tester with a 2N load cell for neat fibers and a 100 N
load cell for composite sensors was used for mechanical testing. An Agilent 34410A digital
multimeter was used to measure the resistance across the fiber sensors. The sample gauge
119

length for electromechanical testing was 1cm. The samples were loaded under tension and
the resistance change was recorded in real time.
5.3 Results and Discussion
Neat CNT fibers can be utilized as a superior novel peizoresistive sensor with excellent
repeatability and stability for structural health monitoring. The electrical resistance as a
function of elastic strain of CNT fibers was measured in-situ during mechanical testing. Both
the resistance and the stress of the fiber have a linear behavior with strains up to 1%. The
resistance-strain behavior is repeatable with no hysteresis, as shown in Figure 5.1. The CNT
fiber itself is a macroscopic CNT assembly that offers an electric pathway along its
longitudinal direction, as shown in Figure 5.2. Carbon nanotubes within a fiber have good
contact with their neighbors due to the twisting process during the fiber spinning. Radial
inward forces provides close contact and load transfer between CNTs during tensile loading
of the fiber.

120


Figure 5.1 Repeatitive testing of neat CNT fibers showing repeatable stress-strain and
resistance-strain behaviors up to 1% strain: a) tensile stress-strain curve, b) in-situ resistance
measuring during the tensile loading. Neat fibers underwent loading and unloading
sequentially to strains of 0.5%, 0.8%, 1% and break, respectively and the resistance was
measured in-situ. After each loading, the fibers were completely unloaded before the next
laoding.
a)
b)
121




Figure 5.2 CNT fiber spinning and deposition of Al electrodes on CNT fibers. a) SEM of
directly spinning CNT fibers from as-grown CNT array. b) low magnification SEM of an as-
spun fiber. c) schematic drawing of Al electrodes deposition on dry CNT fibers using
magnetron sputtering. d) low magnification SEM of a fiber end coated with Al.

The resistance of the fibers monotonically increased with increasing strain. There are
two factors that affect the resistance of a fiber during the tensile strain: one is the strained
individual CNTs; the other is the improved contact between CNTs and their neighbors due to
the contracting radial force produced in fibers. The latter is obviously not the reason for the
resistance increase since improved contact should decrease the resistance. Due to this fact, it
appeared that the resistance increase with strain was caused by the strained individual CNTs.

122

It has been established by several pioneering experiments that the resistance of CNTs
changes when subjected to strain.
1-5
A generic equation provides useful insights for
understanding of resistivity of individual CNTs. Specifically, the total resistance (R
tot
) is:
2 2
1
1 exp
8
gap
tot s
E
h
R R
e kT
t
( | |
= + +
( |
\ .
(1)
where R
s
is the contact resistance between the electrodes and CNTs, which is negligible since
the electrodes have good contact with the fibers, |t|
2
is the transmission probability of
electrons across the band gap barrier, T is the temperature, h, k, e are constants and E
gap
is
the band energy gap. The validity of Equation 1 has been verified experimentally for
individual CNTs under loading.
1-5
The CNT fiber sample contains hundreds of thousands of
individual CNTs. Therefore, the resistance increase shown in Figure 5.1 is a collective
behavior of all CNTs in the sample. In other words, Figure 5.1 and equation (1) indicate that
the band gap energy of most individual CNTs in fibers increased under tensile strain, which
led to collective resistance increase across the fiber.
As discussed earlier, it appears that the improved contact between individual CNTs in
the fiber with strain had only a minor impact on the fiber resistance. Unlike dispersed CNT
composites where each CNT may have only a few contact points with other CNTs for
123

electric conduction, the CNTs in the fibers are packed closely and have thousands of contact
points with their neighbors along their whole length, which provides a large total contact
area for electron hopping between CNTs. Therefore, the increase in contact area within a
fiber under tensile loading will not provide much further improvement in electric conduction
because the contact area is already large.
The excellent repeatability of resistance-strain behavior under cyclic loading (see
Figure 5.1) was also related to the large contact area between CNTs. The close contact with
large contact area along the whole CNT length, the large van der Waals force between CNTs
and the radial inward force produced during loading provide good structural stability for the
fibers during the loading and unloading in the elastic region, which renders the fibers
excellent resistance-strain repeatability.
The sensitivity of the strain gauge is termed as gauge factor (GF), which is described by
equation (2):

/ /
/
R R R R
GF
L L

= =

(2)
The gauge factor of pure fibers was around 0.5. Single nanotubes have a gauge factor of
hundreds depending on the structure of the nanotube.
2
Why was the gauge factor of fibers
124

made of individual nanotubes much lower than single nanotubes? Fibers are made of billions
of short individual CNTs with length of around several hundred micrometers. Contact
resistances between nanotubes are much larger than resistance of nanotubes themselves.
Detailed discussion is given in the following paragraphs.
Total resistance across a fiber is described by equation (3).

(3)
The total resistance is derived from two parts, R
1
and R
2
. R
1
is the actual total electron
conducting resistance along the length of each nanotube. R
2
is the total resistance of electron
hopping from nanotubes to nanotubes. This is a result of the finite length of CNTs within the
CNT fibers. Usually, the conductivity of the CNT assembly is two orders of magnitude
smaller than individual CNTs.
19,20
With the above discussion, we are ready to further assess
the nature of the sensitivity of CNT fibers.
In the studied fibers, the resistance change is from R
1
. As discussed previously, no
improvement of electric conduction was observed during the loading since CNTs in the fiber
already have a large contact area. During loading, R
1
changes, while R
2
remains constant.
From equation (4), a combination of equations (2) and (3), it is clear that the fibers will have
2 1
R R R + =
125

a smaller gauge factor than that of the single CNTs since R
2
accounts for the majority of the
resistance measured in fibers.

2 1
2 1
R R
R R
L L
R R
GF
+
+
=

=
(4)

The resistance of the fiber sensors showed good stability during 100 minutes of
measurements at -196, 25 and 110, as shown in figure 3. Samples were immersed in
liquid nitrogen, kept in the oven and held at room temperature respectively for continuous
real-time resistance testing under zero load conditions. The observed stability holds the
promise for the sensors to function at a large range of temperatures.
The good stability of fibers sensors was a result of the ultra low heat producing power,
relatively large surface area for heat dissipation and steady contact and large contact area
between CNTs in the fiber. In the present experiments, the test current was 100A for the
10k range. As a result, the power of heat producing of the fiber resistor was around 10
-5
W.
Thermal stability of CNT fibers heated by electric current was shown in a recent paper and
no temperature increase was detected in the CNT fiber when a current of 30mA was used.
21


126


Figure 5.3 Stability testing of fiber sensors under zero-load conditions showing that the
resistances were very stable at three different working temperatures: immersed in liquid
nitrogen, at room temperature and annealed at 393k in an oven. The measurement was made
in a continuous mode.

Stable resistance shown by CNT fibers at a wide range of temperatures, is critical in the real
application as strain sensors. As shown in equation 1, resistance of CNT-based sensor can
change exponentially with the temperature. Unstable resistance under static testing was
reported in a CNT/polymer strain sensor fabricated using the dispersing method and
resistance was found to decay exponentially during testing.
8
The fibers not only have a stable
temperature when a testing voltage is applied, but also have a stable electric current
127

conducting path. The large contact area between CNTs prevents charge-induced contact
deterioration. Sensors fabricated by dispersing CNTs in polymer have few contact points,
where large current concentration may cause localized heating that may damage the contact
points. In contrast, CNTs in the fiber have large contact areas, which have less localized
heating and thus better long term resistance stability.
To demonstrate the concept of integral composite sensors, CNT fibers were embedded
into epoxy resin. The electric conduction mechanism in fiber composites was different from
traditional CNT composites fabricated by dispersing CNTs into polymers. Traditional CNT
composite strain sensors require adding a certain amount of CNTs to reach the electrical
percolation threshold. Electrons conducts through the 3-D CNT networks formed through
the whole composites. In contrast, in fiber composite sensors, electric current only passes
through the CNT fibers and the surrounding matrix remains electrically insulating. This is
analogous to inserting an electric wire in the composites that has a diameter of one tenth of a
human hair. Carbon nanotube fibers can be used in composite structures seamlessly since
they are minimally invasive to the composite structure and have negligible added weight due
to their low density.
22

128

As shown in Figure 5.4a, the resistance change was linear with strain as the composite
sample was loaded in the longitudinal direction, while very little resistance change was
observed when loaded in the transverse direction. The applied load was transferred to the
CNT fiber, which changed the conductivity of the fiber. The overall increase in resistance
under transverse loading of the fiber was attributed to the off axis CNT twist angle in the
fibers. The twist in the fibers ensured that the individual CNTs were subjected to some
tensile strain even in the transverse loading case.
The longitudinal GF of the strain sensor was about 0.38 and remained constant over the
whole strain range. The transverseGF ranged from 0.02 to 0.04. Compared to longitudinal
resistance change, the transverse resistance change could be neglected. The composite sensor
has a unidirectional sensing property in the fiber alignment direction. This observation could
help engineers to not only detect the load that the structure is undergoing but also determine
the direction of the external force applied by utilizing multiple CNT fiber sensors oriented in
different directions.
The initiation of micro-cracks and the crack propagation in the composites were
detected by aligning multiple fibers in the composites. As shown in Figure 5.4b, an initial
129

sharp crack was deliberately cut at the edge of the composite sample. The crack propagated
across the sample under loading. Tremendous resistance changes were observed when the
crack approached the fiber sensing elements. As the crack continued propagating, fibers
were broken sequentially. The breakage of the fibers caused a sharp increase in the total
resistance. The results provided here hold promise for the application of detecting cracks in
traditional composite structures using embedded CNT fiber sensors. The early detection of
small internal cracks could allow for early maintenance to prevent potential catastrophic
failure.
Figure 5.5 shows the total resistance change of fiber sensors when the crack is
propagating across the first fiber and breaks the first fiber. The initial resistance R
0
is around
110
3
Ohm. The resistance starts increasing obviously at the strain of 0.0185. The first CNT
fiber fails and there is a resistance jump from 110
3
Ohm to 1.510
3
Ohm. The strain
window of noticing the initiation of the failure of the first fiber is around 0.009.



130



Figure 5.4 In-situ monitoring of strain and crack propagation in a composite structure. a)
Linear increase of resistance when the composite structure was under longitudinal tension
and neglectable increase under transverse tension; b) crack propagation was easily monitored
by building multiple CNT fibers in the composite structure.
a)
b)
131


Figure 5.5 Zoom-in view of the crack propagation causing the failure of the first CNT fiber.
Figure 5.6 shows the total resistance change of fiber sensors when the crack is
propagating across the second fiber and breaks the second fiber. The resistance starts
increasing obviously at the strain of 0.026. The second CNT fiber fails and there is a
resistance jump from 1.510
3
Ohm to 2.810
3
Ohm. The strain window of noticing the
initiation of the failure of the second fiber is around 0.007.
132


Figure 5.6 Zoom-in view of the crack propagation causing the failure of the second CNT
fiber.

5.4 Conclusions
In conclusion, it has been demonstrated that CNT fibers, made by twisting individual
CNTs together, have outstanding repeatable and stable resistance-strain behaviors. Fiber
arrays can be permanently integrated within a composite structure conveniently during
fabrication with minimal invasiveness and weight punishment. The embedded fiber sensors
can be used to monitor the initiation and propagation of cracks in composite structures in
real time, as well as to work as a general strain gauge if top-mounted on the structural
components.
133

REFERENCES
1. Minot E. D.; Yaish Y.; Sazonova V.; Park J.; Brink M.; McEuen P. L., Physical
Review Letters 90, 156401-156404 (2003).

2. Cao J.; Wang Q.; Dai H., Physical Review Letters 90, 157601-157604 (2003).

3. Nishio T.; Miyato Y.; Kobayashi K.; Matsushige K.; Yamada H., Applied Physical
Letters 92, 063117-063119 (2008).

4. Grow R. J.; Wang Q.; Cao J.; Wang D.; Dai H., Applied Physics Letters 86, 093104-
093106 (2005).

5. Tombler T. W.; Zhou C.; Alexseyev L.; Kong J.; Dai H.; Liu L.; Jayanthi C. S.; Tang
M.; Wu S. Y., Nature 405, 769-772 (2000).

6. Knite M.; Tupurenia V.; Fuith A.; Zavickis J.; Teteris V., Materials Science and
Engineering C 27, 1125-1128 (2007).

7. Pham G. T.; Park Y. B.; Liang Z. Y.; Zhang C.; Wang B., Composites: Part B 39,
209-216 (2007).

8. Loh K. J.; Kim J.; Lynch J. P.; Kam, N. W. S; Kotov N. A., 2007 Smart Materials
and Structures 16, 429-438 (2007).

9. Kang I.; Schulz M. J.; Kim J. H.; Shanov V.; Shi D., Smart Materials and Structures
15, 737-748 (2006).

10. Thostenson E. T.; Chou T. W., Advanced Materials 18, 2837-2841 (2006).

11. Boger L.; Wichman M. H. G.; Meyer L. O.; Schulte K., Composites Science and
Technology 68, 1886-1894 (2008).

134

12. Thostenson E. T.; Chou T. W., Nanotechnology 19, 215713-215718 (2008).

13. Li C.; Thostenson E. T.; Chou T. W., Composites Science and Technology 68, 1445-
1452 (2008).

14. Li X.; Levy C.; Elaadil L., Nanotechnology 19, 045501-045507 (2008).

15. Dharap P.; Li Z.; Nagarajaiah S.; Barrera E. V., Nanotechnology 15 379-382 (2004).

16. Li Z.; Dharap P.; Nagarajaiah S.; Barrera E. V. Kim J. D., Advanced Materials 16,
640-643 (2004).

17. Mirfakhrai T et al. 2008 Advances in Science and Technology 61 65

18. Zhang Y. Y.; Zou G.; Doorn S. K.; Htoon H.; Stan L.; Hawley M. E.; Sheehan C. J.;
Zhu Y. T.; J ia Q. X., ACS Nano 3, 2157-2162 (2009).

19. Zhou W.; Wavro C.; Guthy C.; Winey K. I.; Fisher J. E.; Smalley R. E., Journal of
Applied Physics 95, 649-655 (2004).

20. Zhang M.; Atkinson K. R.; Baughman R. H., Science 306, 1358-1361 (2004).

21. Peng H. S.; Sun X. M.; Cai F.; Chen X.; Zhu Y.; Liao G.; Chen D.; Li Q.; Lu Y.; Zhu
Y. T.; J ia Q. X., Nature Nanotechnology 4, 738-741 (2009).

22. Zhang X. F.; Li Q.; Holesinger T. G.; Arendt P. N.; Huang J. Kirven P. D.; Clapp T.
G.; DePanula R. F.; Liao X. Z.; Zhao Y. H.; Zheng L. X.; Peterson D. E.; Zhu Y. T.,
Advanced Materials 19, 4198-4201 (2007).


135

CHAPTER 6: Conclusions and Future Work
6.1 Conclusions
Carbon nanotubes have many potential applications in future. New strategy of growing
CNTs has yet to be done to reduce the cost and produce long CNT arrays for thermal,
electrical and mechanical applications. The objectives of the research conducted in this
thesis were to grow tall and spinnable CNT arrays, spin fibers and apply the fibers as sensors.
The following are a collection of outcomes and conclusions made over the course of this
work:
A novel intermetallic Fe-Zr catalyst were designed and co-sputtered using a magnetron
sputtering method. Partial chemical bonding between Fe and Zr renders the good thermal
stability of this new catalyst allowing a uniform catalyst particles distribution on Al
2
O
3

buffer layer. 1.7 mm long CNT arrays were grown in 45 minutes using Fe-Zr catalyst. Pure
iron catalysts prepared using the same method and same experimental conditions had a lower
performance in growing CNT arrays. In combination with other techniques used to extend
the lifetime of CNT growth of CNT arrays, such as water assisted growth, new record
breaking CNT array lengths might be possible. These ultra long CNT arrays will be critical
136

for providing the long nanotubes needed in multifunctional, high performance composite
materials.
FeCl
2
powders were directly used as catalyst precursors for growing long carbon
nanotube arrays. With unique pressure control, CNT arrays grown could be spun into fibers
and winded into tethers. The characteristic of this method is that only one step is required for
CNT growth, which is superior to traditional two-step CVD method. HCl gas was used to
assist the one-step CVD method to grow long carbon nanotube arrays. 3 mm tall CNT
forests were grown in 60 minutes and 2 mm tall spinnable arrays were grown in 20 minutes.
HCl gas could reduce the amorphous carbon during the CNT growth which extended the
growth time.
CNT fibers composed of only individual CNTs were made by twisting CNTs ribbons
together. The neat fiber has outstanding repeatable and stable resistance-strain behaviors due
to the stable structure. Resistance of fibers increased linearly with strain indicating a reliable
strain sensor. Fiber arrays can be permanently integrated within a composite structure
conveniently during fabrication with minimal invasiveness and weight punishment. The
embedded fiber sensors can be used to monitor the initiation and propagation of cracks in
137

composite structures in real time, as well as to work as a general strain gauge if top-mounted
on the structural components.
6.2 Future Work
One-step CVD method has many advantages in growing carbon nanotubes, such as
simple, fast, reproducible, low cost et al. However, the post cleaning process is tedious since
CNTs not only grow on quartz substrates but also on the walls of quartz tubes. If some
techniques could be applied to prevent the growth on the walls, it will be very helpful to
shorten the trial time. Techniques such as coating a thin film on the quartz tube worthy
trying in future works.
Pressure control is critical for one step CVD method. Pressure keeps increasing during
the reaction. Therefore automatic pressure control valve could be installed in the system to
realize computer controlled pressure.
CNT fibers show very promising sensing properties especially for the structural health
monitoring of composite structure. Further study could be focused on two aspects: one is to
improve the sensitivity of CNT fiber sensors by functionalizing the CNTs; the other is to
step forward on the real application of these nano sensors in composite.