:w'mm mmmw

.‘Q

‘

”mam-rm

our. gnu“ «hwy-IN

u
m
.n
m
“H
n.
.
u
.
n
.
u
a
'

~Ngg‘lg ‘

 

 

I
:. gap“. In}
. : I
A! no It‘s!
tit... "W’s-fishy”
inlay 5‘5

x..." ;.
{Oi-tit...
11..Q‘£|l.:f

MM) #3..

as 3-...
59?. am“) éflkfis pin“... I

v v:
13”.... a).
1.5.9:... 31.:

 

 

THESIS

2,005}

This is to certify that the
dissertation entitled

CARBON NANOTUBE BASED INFRARED SENSORS —
DESIGN, FABRICATION, AND TESTING

presented by

JIANGBO ZHANG

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Electrical Engineering

 

 

Major Prrfessor’s Signature

12/1 /08

 

Date

MSU is an Affirmative Action/Equal Opportunity Employer

 

CARBON NANOTUBE BASED
INFRARED SENSORS - DESIGN,
FABRICATION, AND TESTING

By

J iangbo Zhang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Electrical Engineering

2008

ABSTRACT

CARBON NANOTUBE BASED
INFRARED SENSORS - DESIGN,
FABRICATION, AND TESTING

By
J iangbo Zhang

In this research, we focus on the design, fabrication, and testing of single carbon
nanotube (CNT) based infrared (IR) sensor, which consists of two microelectrodes
bridged by a single CN T. Sensors with symmetric structure and asymmetric structure
were designed respectively to study the dependence of the photocurrent on CNT-metal
contacts. In a symmetric structure (same metal for both sides), photogenerated elec-
trons and holes at one CNT-metal interface have to tunnel through another interface
to form photocurrent, and the photocurrents from the two reversely connected diodes
will cancel each other. While the asymmetric structure (different metals for both
sides) increases the measurable photocurrent by lowering one of the Schottky bar-
riers and enlarging the difference between the photocurrents generated at the two
CNT-metal interfaces.

Nano—manufacturing/assembly plays an extremely important role in studying
nanomaterials and nanodevices including the CNT nanodevices. To fabricate the
CNT based IR sensors, a nanoassembly process which is able to assemble a single
CNT/nanowire onto a pair of microelectrodes using the Atomic Force Microscope
(AFM) based nano-robotic manipulation system and the dielectrophoresis (DEP) de-
position system, was developed and experimentally tested. The DEP force attracts
one or more individual CNTs around the microelectrodes. Then the AF M based
nanomanipulation system manipulates one CNT to bridge the microelectrodes and

cleans the rest CNTs and particles away from the microelectrodes. This nanoassembly

process provides a controllable, reliable and efficient approach to assemble nanode—
vices.

With the developed nanoassembly process, single CNT based Schottky photodi—
odes using symmetric and asymmetric structures have been fabricated. Experimental
measurements on both types of devices have shown that the asymmetric structure can
improve the photodiode performance by increasing the signal—to-dark current ratio up
to three orders of magnitude higher than a symmetric device. It has also been verified
that the photocurrent is generated by the photovoltaic effect rather than heating or
photoconduction effect as previously reported. With the asymmetric structure, mid—
dle infrared (MIR) detection has been demonstrate using individual MWNTs, whose
bandgap was tuned by electrically burning out the outer layers of the MWNT.

Furthermore, multi-pixel CNT sensor arrays have been fabricated and tested in
this study. The fabrication of multi—pixel CNT sensor array demonstrated that the
developed nanoassembly process is able to fabricate single CNT/nanowire device ar-
rays, which is difficult for other fabrication methods. Experimental testing results on
CNT sensor arrays have shown that the adjacent pixels have small thermal crosstalk,
which is an important factor limiting the pixel pitch of IR focal plane arrays (IRF-
PAS). The small pixel pitch of the developed CNT detector array makes it possible
to fabricate high resolution IRFPAS. Moreover, a multicolor CN T IR sensor has also
been developed by assembling two CNTs with different diameter (bandgap) on two
adjacent microelectrode pairs respectively. The multicolor CNT IR sensor further
demonstrated the abilities of the developed nanoassembly process and showed po-
tential applications of CNTs in advanced IR sensor systems, which need multicolor
capability for better target discrimination, tracking, and identification, as well as

temperature determination.

To my parents, my wife Ying Zhao,

and to my daughter, for their love and support

iv

ACKNOWLEDGMENTS

I would like to express my sincere gratitude and appreciation to my advisor, Dr.
Ning Xi, for providing me with the unique opportunity to work in the research area
of nanomanufacturing and nanosensors, for his expert guidance and generous help,
and for his encouragement and support for my work. I am grateful to him for giving
me the support and confidence during the period of my dotoral studies.

I would like to thank all members of my guidance committee: Dr. Jes Asmussen,
Dr. Timothy Grotjohn, Dr. Hassan Khalil , Dr. Guowei Wei. At different stages
of my Ph.D. studies, they offered me timely help and unfailing support. Their as-
tute comments and suggestions have enhanced both the technical soundness and the
presentation of this dissertation.

I would like to express my sincere gratitude to all the colleagues from the Robotics
and Automation Laboratory especially to Dr. Guangyong Li for many hours of dis-
cussions and mentoring during my doctoral studies. Hongzhi Chen and Dr. King Wai
Chiu Lai have helped me fabricate electrode chips. The discussions with Dr. Yantao
Sheri, Dr. Quan Shi, Lianqing Liu, and Dr.Uchechukwu Wejinya have substantially
contributed to my work and broadened my knowledge.

No matter how much I acknowledge the support I received from all those men-
tioned above, my deepest gratitude goes to my parents, and my wife, Ying Zhao.
They have always been my cheerleaders and supporters. This dissertation would not
have been possible without their years of encouragement and continuous support.

With love and gratitude, I dedicate this dissertation to them.

TABLE OF CONTENTS

LIST OF TABLES viii
LIST OF FIGURES ix
1 Introduction 1
1.1 Overview .................................. 1
1.2 Literature Review ............................. 2
1.2.1 Carbon Nanotubes ........................ 2
1.2.2 Design of CNT Infrared Sensors ................. 5
1.2.3 Fabrication of CNT Devices ................... 7
1.2.4 Testing of CNT Infrared Sensors ................. 21
1.3 Objectives and Challenges ........................ 22
1.4 Organization of Dissertation ....................... 24
2 Design of Carbon Nanotube based Infrared Sensors 25
2.1 Manufacturing Issues in Design ..................... 25
2.2 Diode based CNT Sensor ......................... 26
2.3 Optimizing the CNT Sensor ....................... 27
2.4 Chapter Summary ............................ 29
3 Assembly of Carbon Nanotube based Infrared Sensors 30
3.1 Development of the Fabrication Process for Single CNT based N ano-
Devices .................................. 31
3.2 Development of the AFM based Nanomanipulation System ...... 33
3.2.1 Modeling of Adaptable End Effector .............. 34
3.2.2 Control of Adaptable End Effector ............... 45
3.2.3 Experimental Implementation .................. 50
3.2.4 Experimental Results on Manipulating Name-Objects ..... 56
3.3 Experimental Results on Fabricating Single CNT based Infrared Sensors 62
3.4 Chapter Summary ............................ 62
4 Testing of Single CNT based IR Sensors 66
4.1 Development of the Testing System ................... 66
4.2 Individual l\/Iulti—Walled Carbon Nanotube based IR Sensors ..... 68
4.2.1 Bandgap Engineering of MWNTs ................ 68
4.2.2 Experimental Testing of MWN T Sensors ............ 70
4.3 Individual Single-Walled Carbon N anotube based IR Sensors with
Symmetric Structure ........................... 73

vi

4.4 Individual Single-VValled Carbon Nanotube li)ased IR Sensors with

Asymmetric Structure ..........................
4.5 Sensing Middle IR Using C‘N T Based IR Sensors ............
4.6 Temperature Dependence of CNT Based IR Sensors ..........
4.7 Performance Evaluation and Analysis ..................
4.8 Chapter Summary ............................

5 Multi-Pixel CNT Sensor Array
5.1 Fabrication of Single CNT based IR Sensor Array ...........
5.2 Testing of Multi-Pixel CNT Sensor Array ................
5.3 CNT based Multi-color IR Sensors ...................
5.4 Chapter Summary ............................

6 Conclusions
6.1 Major Contribution ............................
6.2 Future Research ..............................

REFERENCES

BIBLIOGRAPHY

vii

81
86
88
96

97
98
99

114
115
116

119

124

3.1
3.2
3.3

4.1

LIST OF TABLES

Cantilever parameters

oooooooooooooooooooooooooo

Damping constants ............................

Modal frequencies (kHz) .........................

Quantum efficiency of different IR materials ..............

viii

51
52
52

95

1.1

1.2
1.3

1.4

1.5

1.6
1.7
1.8

1.9

1.10

1.11
1.12
1.13

LIST OF FIGURES

(a) Schematic honeycomb structure of a graphene sheet. (b) CNT
cylinders with different chiralities. ( [16]) ................

The structure of an ideal CNT diode. ( [14]) ..............

Diagram illustrating the displacement photocurrent measurement tech-
nique. ( [15]) ...............................

(a) Optical image of an array of CNT devices. (b) Optical image of
a single device with three CNTs bridging the electrodes. (c) Scanning
electron microscope (SEM) image of a single device. ( [35]) ......

The AF M image of two electrodes bridged by several SW NTs which
were deposited by DEP process. ( [37]) .................

Schematic diagram of a head scanning AF M. .............
Possible mechanical manipulation tasks using an AF M probe. ( [67])

Pushing a silver nanowire with an unpreloaded traditional tip
(k=0.57N/m). (a) Schematic illustration of the pushing process. (b)
The AF M image of a silver nanowire with length of 2.5pm and diame—
ter of 120nm. The scanning range is 10pm x 10pm.(c) The AP M image
by a new scan after pushing. ((1) Tip displacen'ient signal measured by
the photodiode sensor. ..........................

Pushing a silver nanowire with a preloaded traditional tip (k=1-5N/m).
(a) Schematic illustration of the pushing process. (b) The AF M image
of a Silver nanowire with length of 2.5,um and diameter of 120nm. The
scanning range is 10pm x 10mm. (c) The AFM image by a new scan
after pushing. ((1) Tip displacement signal measured by the photodiode
sensor. ..................................

Schematic illustration of pushing a nano—object using the adaptable
end effector .................................

Flexible beam ...............................
Temporal photocurrent response of a SVVNT film. ( [10]) .......

Gating characteristics of an arnl1)ipolar NT -FET with and without in-
frared illumination ( [13]) ........................

11
12
14

15

17

18
20
22

2.1

2.2

2.3

2.4

2.5

3.1

3.2
3.3

3.4

3.5

3.6
3.7
3.8
3.9
3.10

3.11

(a) Structure of the single CNT based infrared sensor. (b) Cross section
of the microelectrodes. .......................... 26

Illustrations of pushing CNT over the microelectrodes which are fab-
ricated by (a) etching and (1)) lift off process ............... 26

The energy band diagrams of the CNT-metal contact for (a) n-type
CNTS and (b) p—type CNTs ........................ 27

The single CNT based infrared sensors with synunetric contacts. (a)
Structure of the CN T sensors. (b) Energy band diagram of the CNT
sensors. .................................. 28

The single CN T based infrared sensors with asymmetric contacts. (a)
Structure of the CNT sensors. (b) Energy band diagram of the CNT
sensors. .................................. 29

(a) The CNT assembly process. (1)) The fabrication process of micro
electrodes .................................. 31

The CN T deposition process. ...................... 32

The setup of the AF M based nanomanipulation system using adaptable
end effector. ................................ 34

The control scheme of the AFM based nanon'ianipulation system us-
ing adaptable end effector. 1: Piezo tube. 2: Piezoceramic layer of
the adaptable end effector. 3: Adaptable end effector. 4: Substrate
surface. 5: Object. 6: Laser beam. 7: Quad-photodiode detector. 8:

Mirror. 9: Laser gun ............................ 35
Adaptable end effector with an integrated Zinc Oxide piezoelectric ac-

tuator. .................................. 36
Adaptable end effector configuration. .................. 37
Diagram of the cantilever under bending ................. 39
Controller block diagram. ........................ 50
Measured tip vibration of the cantilever without closed-loop control. . 53

Tracking response of (a) a sine wave and (b) a triangular wave. The
controller is LQR controller. ....................... 55

Tracking response of (a) a. sine wave and (b) a triangular wave. The
controller is a PD controller ........................ 56

3.12 (a) Tracking response of a square wave with PD controller. (b) Track-
ing response of a square wave with PD controller. (0) Zoom in around
0.016 sec in (b) ...............................

3.13 Tracking response of a square wave with the LQR controller and input
shaping ..................................

3.14 Pushing a silver nanowire with an adaptable end effector. (a) The
AFM image of a silver nanowire with length of 1.9,um and diameter
of 12011177.. The scanning range is 6mm X 6pm. (b) The AFM image
by a new scan after pushing. (c) The control voltage applied on the
adaptable end effector. ((1) Tip displacement signal measured by the
photodiode sensor. ............................

3.15 Successful rates of manipulations using different probe ........

3.16 Manipulating a CNT onto a pair of gold electrodes. (a) The AFM
image before manipulation. (b) The image from a new AF M scan
after manipulations .............................

3.17 Manipulating a CNT onto a pair of gold electrodes. 1 (a) The AFM
image before manipulation. (b) The image from a new AFM scan
after manipulations .............................

3.18 Four different results of pushing CNT onto gold electrodes .......

‘1

CI!

58

63

63
64

4.1 (a) The SVVNT IR detection system. (b) The structure of the CNT chip. 67

4.2 The cryogenic MIR testing system. ...................
4.3 I-V curves of a single MWNT device ...................
4.4 I-V curve of the single MW’NT device after three breakdowns.

4.5 AFM image of the MWNT IR sensor ...................
4.6 I-V curve of the MWN T IR sensor. ...................

4.7 Temporal photoresponse of the l\«IW'NT IR sei‘isor. The bias voltage
across the electrodes is —0.1 V. .....................

4.8 Temporal photoresponse of the MW NT IR sensor. The bias voltage
across the electrodes is 0.1 V. ......................

4.9 The plot of photocurrent vs. bias voltage. ...............

4.10 (a) AF M image of an individual SWNT based photodiode with sym-
metric Au electrodes. (1)) I—V curve of the CNT photodiode. .....

68
69
69
71

74

4.11

4.12

4.13

4.14

4.15

4.16
4.17
4.18
4.19

4.20
4.21

4.22

4.23

5.1
5.2

5.3
5.4
5.5

Temporal photocurrent response of the SVVNT photodiode with a zero
bias voltage. (3) IR irradiates the left electrode. (1.)) IR irradiates the
right electrode. ..............................

The plot of the photocurrent as a function of the IR spot position. . .

(a) AFM image of an individual CNT photodiode with asymmetric
Ti—Au electrodes. (b) I-V curve of the asymmetric CNT photodiode. .

(a) The photocurrent as a function of the IR spot position. (b) Tem-
poral photocurrent response of the photodiode with asymmetric elec-
trodes. The device was zero biased. The IR power was 30 mW.

I-V characteristic curves of the asymmetric CNT photodiode. PO-P3
represent different IR powers. ......................

I-V characteristic curve of a MWNT MIR sensor ............

75
77

78

79

82

I-V characteristic curve of the MWNT MIR sensor after first breakdown 83

Temporal photoresponse of the MWNT sensor to NIR .........

I-V characteristic curve of the MWNT MIR sensor after second break-
down ....................................

Temporal photoresponse of the MWNT sensor to MIR ........

I—V characteristic curves of a MW NT sensor at different temperatures.
The sensor was assembled through nanomanipulation ..........

I-V characteristic curves of a MWN T sensor at different temperatures.
The sensor was assembled without nanomanipulation ..........

Temporal NIR photocurrent response of an individual SW N T IR sensor
measured at high sampling rate .....................

Nano—manufacturing process for single CNT based nanodevices.

(a) The structure of a single pixel CNT IR sensor. (b) The 3D AFM
image of a three—pixel ON T IR sensor array. ..............

The AFM image of the three-pixel CNT IR sensor. ..........
I-V curves of the three pixels. ......................

Illustration of the photocurrent measurements. IR spot was on (a)
pixel 1, (b) pixel 2 and (0) pixel 3 respectively ..............

xii

84

85
86

87

88

90

102

102

5.6

5.7

5.8

5.9
5.10

5.11
5.12

5.13

5.14
5.15

Temporal photocurrent response of each pixels while the IR spot was
on pixel 1 (Fig. 5.5(a)). .........................

Temporal photocurrent response of each pixels while the IR spot was
on pixel 2 (Fig. 5.5(b)). .........................

Temporal photocurrent response of each pixels while the IR spot was
on pixel 3 (Fig. 5.5(c)) ...........................

Structure of the nn..11ticolor CNT IR sensor ...............

I-V characteristic curve of the SW NT pixel of the multicolor CN T IR
sensor ...................................

NIR response of the SWNT pixel of the multicolor CNT IR sensor . .

I-V characteristic curve of the MIN NT pixel of the multicolor CNT IR
sensor ...................................

I-V characteristic curve of the MWNT pixel after breaking down outer
layers ....................................

NIR response of the MWNT pixel of the multicolor CNT IR sensor
MIR response of the MWNT pixel of the multicolor CNT IR sensor .

xiii

103

103

104

105

106
106

108

109
111
112

CHAPTER 1

Introduction

1 . 1 Overview

The development of middle infrared (MIR) sensors that operate near ambient tem—
perature will allow significant simplifications of thermal imaging systems for appli-
cations such as remote sensing, environmental monitoring, medical diagnostics, as
well as defense and aerospace applications. Photonic MIR sensors have been made
with materials that have energy bandgaps sufficiently small to absorb MIR radiation.
However, the small bandgap makes such sensors susceptible to thermal noise. Hence
the traditional photonic MIR sensors have to work at ultra low temperature. As a
one-dimensional (l-D) nanostructural material, carbon nanotubes (CNTs) have po-
tential applications in solar collection and IR sensing due to their unique properties
such as direct bandgap, wide range of bandgaps, and reduced carrier scattering. With
the unique properties, CNT based MIR sensors have the potential to work at room
temperature or moderate low temperature with a high quantum efficiency. Hence
it is very important and necessary to study electronic and photonic properties of
CNTS. However, since CN T films or bundles lose the advantages brought by the 1-D

structure of CNTS, it becomes necessary to study the single CN T based nanodevices.

To study the single CN T based nanodevices, a manufacturing process is required
to make a CNT into nanodevice. This process is normally called nano-manufacturing
or nano—assembly. Nano-manufacturing plays an extremely important role in study-
ing nanomaterials and nanodevices including the CNT nanodevices. W'ith its ability
to characterize and modify the sample surface at molecular and atomic scales, Atomic
Force Microscope (AFM) provides such a promising way to build or modify nanode-
vices. With AFM, nanoobjects (such as DNAs, CNTs, nano-wires, etc.) can be moved
and positioned very accurately. Since AF M scans the sample surface by measuring
attractive or repulsive forces between a tip and the sample, it is impossible to image
the sample surface when the tip is used for manipulations. This means the object is
unobservable when it is being manipulated. To facilitate AFM based nanomanipula-
tions and improve the accuracy of nanomanipulations, it is very important to develop
a tele—operation system with a sensitive force feedback and accurately tip position

control.

1 .2 Literature Review

1.2.1 Carbon Nanotubes

As the traditional silicon based semiconductor industry faces increasing technological
and financial challenges on the way to further scaling down, new technologies and
concepts have to be developed and assessed. CNT is such a candidate and is likely
to create breakthroughs in the fields of nanoelectronics. Since the discovery of CN Ts
in 1991 [1], their mechanical and electrical properties have been intensively studied.
Due to their unique hollow cylindrical structure, carbon nanotubes have promising
potential applications in nano—electronics [2,3] and name-mechanics [4]. They are able
to work as fundamental building blocks of mane-electronic devices with their excellent

one—dimensional material properties [5]. They are also extremely strong materials and

Q. ,.

4:01.963
a. .

if
,.
z],

 

Figure 1.1. (a) Schematic honeycomb structure of a graphene sheet. (b) CNT cylin-
ders with different chiralities. ( [16])

have good thermal conductivity [6]. Moreover, carbon nanotube’s photobehavior has
attracted lots of attention due to their well—defined one-dimensional structure [7—15].

A single-walled carbon nanotube (SWNT) is formed by rolling a sheet of graphene
into a cylinder along a lattice vector (m,n) in the graphene plane. The (m,n) vector
determines the diameter and the particular roll orientation of a CNT, which is called
the chirality. The (m,n) vector is also called chiral vector. The angle between the
roll orientation and the axis of the tube is termed the chiral angel. The chirality
(or chiral angle) determines the electronic properties of a CNT. Depending on the
chirality, CNTs with same diameter can be either metallic or semiconducting. There
are three types of rolls: armchair, zigzag, and chiral. Armchair CNTs are formed when
m=n. The chiral angel of an armchair CNT is 30°, and the armchair CNTs display
metallic properties. Zigzag CNTS are formed when either n or m equals zero, Where
the chiral angel is 0°. All other lattice vectors result in nanotubes that are termed
chiral CNTS. Both zigzag and chiral CNTs are semiconducting. As shown in Fig.
1.1a, there are three lattice vectors (8,8), (8,0), (10,-2). Folding of the three vectors
leads to armchair, zigzag, and chiral tubes, as shown in Fig.1.1b—d respectively. The

diameter of a SVVNT is determined by the vector (m,n) with the equation [17]

 

 

a\/3(m2 + mm. + n2) _ C},

W W

(1.1)

dCNT =

where a is the carbon to carbon bond length (1.42 A), or the length between
consecutive carbon atoms. Ch is the length of the chiral vector (m,n).

A metallic CNT has a much higher conductivity than the best metals, and semi-
conducting CNTs have mobilities and transconductances that meet or exceed the best
semiconductors [5]. Hence the metallic CNTs have the potential to work as nanowires
for interconnection in nanodevices. And the semiconducting CNTs have the poten-
tial to be the future material for field effect transistor channels. The bandgap of a

semiconducting CNT is give by the equation [17—20]

4hvf

Ea =
gp 3dCNT

(1.2)

where v f is the ferini velocity, and h is the Dirac constant. The unique electrical
properties of SWNTS are caused by the confinement of electrons in the two direc-
tions and the requirements for energy and momentum conservation. These constrains
reduce the scattering processes of electrons and result in a ballistic transport.

As rolling a single graphene sheet into a SW’NT, a multi-walled carbon nanotube
(MWNT) can be formed by rolling multiple layers of graphene sheets into a cylinder
shape. Hence a MWNT consists of multiple tubes [17,21] with spacing between
consecutive tubes of 3.4 A [22]. Each layer of the MVVN T can have different chirality
and can be metallic or semiconducting. It has been reported that the concentric
tubes of a MWNT have poor electrical contact with each other [23]. Theoretical and

experimental studies have also shown that the outermost layer of 3 MW NT dominates

the electrical conduction of the MWNT [24]. Since the bandgap of a SWNT varies
inversely with its diameter and the outermost layer of MWNT normally has a. large
diameter exceeding 10 nm [22]. The bandgap of the outermost nanotube is so small
that MWNTS are metallic in nature.

With its outstanding properties, CNT has potential applications in a wide range
of fields. MWNTS have been studied as very-large-scale integration (VLSI) intercon-
nectors because of their high thermal stability, high thermal conductivity and high
current conductivity [25]. Semiconducting CNTs have been used as the channels of

field effect transistors (CNTFETs) because of their superior electrical characteris-

tics [26—31].

1.2.2 Design of CNT Infrared Sensors

As a 1-D nano—structural material, carbon nanotubes (CNTs) have potential applica-
tions in solar collection and IR sensing due to their unique properties such as direct
bandgap, wide range of bandgaps [5], and reduced carrier scattering [32]. Many re-
searchers are looking into CNT optoelectronics, and the photoconductivity of CNTs
has been studied in thin single-walled carbon nanotube (SWNT) films [9,10,33], multi-
walled carbon nanotube (MWNT) arrays [11], and single SWNT transistors [12,13].
It has also been demonstrated that both semiconducting and metallic SWNTs can
function as photodetectors over a wide spectral range using capacitive photocurrent
measurement [15]. Moreover, a single SW NT p-11 junction diode was built in [14] to
demonstrate its photovoltaic effect.

The ON T based photosensing devices can be classified into three categories based
on their structures. The first category is the basic structure with a pair of micro-
electrodes bridged by a CNT or CNT bundle/film [10, 11,33]. The second type of
the CNT photosensing device is featured with the gate structure. Gate electrodes

are employed to electrically control the photoconducting of the CNT channel [12,13].

 

Figure 1.2. The structure of an ideal CNT diode. ( [14])

As shown in Fig. 1.2, two bottom gate was used to electrically shift the fermi level
of the CNT at both ends to create a p—n junction in the middle of the CNT [14].
As shown in Fig. 1.3, the third type of the CNT photosensing device is a metal-
insulator-semiconductor photodetector [15]. Photons are detected by measuring the
displacement photocurrent on the semiconductor.

The first structure is simple and easy to implement. The other two structures
may leads to a better performance, but the complicated design makes it difficult to
fabricate, especially for the device arrays. In this research, we will first fabricate the
single CNT based IR sensors using the basic structure (two microelectrodes bridged
by a CNT) to study the photobehaviors of single CNTS. After that, new designs
enhanced by additional structures will be employed to improve the performance of

the single CNT based IR sensors.

Glass Dielectric

   
    

Copper
Ground Plate

Pulsed
Light
Source

 

DC Vout

Source Current

'7: Amplifier

Figure 1.3. Diagram illustrating the displacement photocurrent measurement tech-
nique. ( [15])

1.2.3 Fabrication of CNT Devices

The techniques for manufacturing nanodevices can be generally classified into bottom-
up and top-down methods. The fabrication process of CNT devices normally includes
both methods, where top-down methods are used to fabricate supporting structures
such as contacting electrodes, and bottom-up methods are used to assemble CNTS
with the supporting structures. Current available methods in fabricating single CNT
devices include directly growth of CNT across microelectrodes [14, 34—36], deposi—
tion of as-grown CNTS on electrodes by dielectrophoresis (DEP) [37—40], fabrica—
tion of electrodes on top of as-deposited CNTS either by electron beam lithography
(EBL) [41—43] or by shadow masks [44,45], and self-assembly by functionalizing CNTS

and electrodes with different chemicals or even DNA molecules [46,47]. To a certain
extent, these methods have their shortcomings in terms of repeatability, mass pro-

duction, and ability in elin‘iinating uncertainties.

Directly Growth

With the directly growth method, single CNT devices are fabricated by directly grow-
ing a single CNT between a pair of prefabricated microelectrodes to make connec-
tions [14,34—36]. The directly growth method is able to fabricate multiple single CN T
devices at one time. Thus it is good for making CNT based nanodevice arrays. But
limitation of this method is that the properties of the CNTS can not be effectively
controlled. Different CNTS may have different properties even they were produced
at one batch. Moreover, the production process may generate impurities around the
microelectrodes and CNTS, which will affect the electronic properties of the CNT de-
vice. The third limitation is that it is difficult to only grow a single CNT between the
microelectrodes. Since CNT bundles or films lose the unique properties brought by
the 1-D structure of CNTS, this limitation limits the performance of the ON T device.
As shown in Fig. 1.4, an array of CNT devices were fabricated by CVD growth of
SWNTS across the prefabricated electrodes pairs [35]. But as shown in Fig. 1.4(c),
there were more than one CNT between the electrodes pair. Since different CNTS
may have different properties, multiple CNTS bridging two electrodes will makes it

difficult to control and analyze the properties of the device.

EBL Fabrication

Another approach for CNT assembly is to grow CNTS on the substrate surface first
(or disperse the pre—grown and purified CNTS on the substrate surface), then use EBL
fabrication or shadow mask to form microelectrodes on the top of a single CNT which

was located by scanning electron microscope (SEM) [41—43]. This fabrication method

egg

9
s
§
§
E
§
§
a

II- I. in. I".
“H I' D“ F‘
C. -— fl‘ .‘
l. I” *‘ fi‘
- C. '0‘ .H‘
mm sun I"! M! I“
iiifllilit‘ “In"??? WW fill-

 

Figure 1.4. (3.) Optical image of an array of CNT devices. (b) Optical image of a single
device with three CNTS bridging the electrodes. (c) Scanning electron microscope
(SEM) image of a single device. ( [35])

provides a best contact condition between the CNT and I]11(3I‘O€l€Cl’.-l‘0(l(‘8 among these
fabrication methods. But due to the low efficiency of the EBL fabrication process, this
method is inefficient. More importantly, since the CNTS are randomly distributed on
the substrate surface, it is impossible to fabricate a CNT device array which aligns
in a desired pattern. Hence the EBL fabrication method is only good for prototyping

individual devices.

DEP Deposition

Instead of fabricating microelectrodes on the top of randomly distributed CNTS,
preproduced and purified CNTS are manipulated to bridge two preformed microelec-
trodes. There are two ways to manipulate CNTS, dielectrophoresis (DEP) manip—
ulation and nanoprobe based mechanical manipulation. With the DEP deposition
method, microelectrodes are first fabricated, then a droplet of the CNT suspension
is dropped between the microelectrodes and an AC voltage is applied across the two
microelectrodes. CNTS will be attracted by the dielectrophoretic force to the elec—
trodes to form connections [37—40]. Although the number of CNTS attracted to the
electrodes can be roughly controlled by varying the AC voltage and the concentration
of the CNT solution, it very difficult to deposit a single CNT with this method. As
shown in Fig. 1.5, several SWNTs were attracted to bridge the electrodes by DEP
force. Hence the DEP deposition method is normally used to fabricate devices with

CNT films or CNT bundles.

Nana—Manipulation Method

Another controllable assembly process for fabricating CNT devices is to mechanically
manipulate CNTS to bridge the microelectrodes. A single CNT attached at the tip
end of an AFM cantilever were manipulated using focused ion beam [48]. But the

nanotube has to be metal-coated for manipulation. Hence this technology is not

10

10.0

7.5

5.0

2.5

0

10.0
pm

 

2.5 5.0 7.5

Figure 1.5. The AF M image of two electrodes bridged by several SWNTs which were
deposited by DEP process. ( [37])
good for building nanoelectronic devices. Three-dimensional manipulation of carbon
nanotubes has been studied in [49]. But the manipulation has to be performed under
SEM, which limited the applications. A more general ways is to manipulate CNTS
using the AFM tip such that CNTS can be positioned on substrate surface in a bare
environment [50752]. But all these work limited to manipulating CNTS and none of
them built nanoelectronic devices through manipulating CNTS. In this research, we
will develop a hybrid fabrication process for building CNT based nanodevices using
AF M manipulation and DEP deposition systems.

The AFM, or scanning force microscope (SFM) was invented in 1986 by Binnig,
Quate and Gerber [53]. Fig. 1.6 shows the schematic diagram of a head scanning
AFM system. Like all other scanning probe microscopes, the AFM utilizes a sharp
probe moving over the surface of a sample in a raster scan. In the case of the AFM.

the probe is a tip at the end of a cantilever which bends in response to the force

11

 

 

Detector Piezotube_
Electronics controller

 

 

 

 

 

 

 

 

Photodiode @ Laser

   
 

[r Piezotube

 

Sample Surface pantilever 8. Tip

Figure 1.6. Schematic diagram of a head scanning AFM.

between the tip and the sample. The surface topography is acquired by recording the
bending of the cantilever at each sampling point.

AFM has already been used to study the supermicroscopic structures of objects
that could not be viewed with the Scanning Tunneling Microscope (STM). It is espe-
cially widely used to observe living cells, DNA molecules and some other biological
objects. Structures of DNA are observed and analyzed using AFM in [54—57]. Mem-
branes and proteins are inspected in [58—60]. The possibility of AFM to be operated

under physiological conditions, without submitting the sample to fixation or any other
chemical preparation, allowed its use on the observation of living cells [61,62]. Un-
der appropriate experimental conditions, these cells remained viable during extensive
periods of time, without damages caused by the scanning.
AFM has also been used to measure nanomechanical properties of biological sam—
pIes, such as elasticity, adhesion, rigidity, friction or viscosity. Some of these prop—
erties are measured using force vs. distance plots or force vs. time plots. In these
methodologies, the AF M does not carry out any kind of scanning, measuring only the
interactions between the tip and the surface of the sample. The improvement of the
studies using force vs. distance plots lead to the creation of a new experimental pos-

Sibility', named single-molecule force spectroscopy (SMFS) [63]. In order to measure

12

the interforce, the partners in the n'iolecular reaction have to be immobilized onto

the surface and the probe. So functionalized tips are used in order to bind easily to

the molecular under evaluation.The sample is analyzed following a. scanning patter

similar to tapping mode, until the formation of the bond. Then it becomes possi-

ble to measure and interpret the force necessary to break the bond or to modify the

conformation of a molecule simultaneously bound to the substrate and to the tip. Mi—

crobial cell surfaces can show important lateral variations in composition and physical

properties. Spatially resolved AFM force spectroscopy, combined with topographic

imaging, is emerging as a valuable technique to map such physical heterogeneities. For

instance, cell surface charges are mapped using chemically functionalized probes [64].

Through the optimization of the scanning conditions, the tip of the microscope

can be used, not only as a probe, but also as a tool for sample manipulation, allowing

its cut, dragging, dissection or conformational alteration. For example, S. Nishida, Y.

Funabashi and A. Ikai use the AF M tip to do the cell injection and verify the injection

results with a total internal reflection fluorescence microscope (TIRFM) [65]. The
AFM tip is also employed to dissect chromosomes in [66].

AFM [53] is primarily a tool for characterizing surface topography, but there is also

a strong interest in using AFM as a nanomanipulator to modify the sample surfaces

or manipulate nanostructures such as nanoparticles and nanowires on surfaces. Fig.

1 - 7 shows the possible mechanical manipulation tasks using an AFM probe [67].

Many research groups have been working on AFM based nanomanipulation

[68—80]. Some of them are trying to utilize the haptic devices to facilitate the nanoma-

nipulation [74, 77—79,81]. However, even with the assistance of the haptic devices, the

efficiency and accuracy of AFM based nanomanipulation is still a major issue due to

the nonlinearities and uncertainties in nanomanipulation operations. The compensa-

tion of drift, creep, hysteresis and some other nonlinearities generating large spatial

uncertainties has been studied by some research groups [80]. The compensation of

13

 

 

 

We»!

 

 

touching indenting/lithography

Figure 1.7. Possible mechanical manipulation tasks using an AFM probe. ( [67])

these uncertainties does help to improve the accuracy of nanomanipulation to a cer-

tain extent, but there are still some other important factors that affect the accuracy

of manipulation significantly. The deformation of the cantilever is one of the major
nonlinearities that affect the tip position accuracy during manipulation. It is diffi-
cult to control the tip position accurately due to the uncertainty of the deformation.
Therefore, it is necessary to take the deformation into consideration or find some way
to eliminate the deformation. Obviously, a rigid cantilever reduces the deformation
significantly if its spring constant is large enough. Unfortunately, as the deformation
is reduced, the feedback signal measured from the deformation is also reduced. Be-
cause the feedback signal is very important for closed-loop haptic nanomanipulation,
it is in a dilemma whether to use a soft cantilever or a rigid one for nanomanipulation.
Fortunately, the invention of the active AFM probe [82—84] provides a promising
way to solve this problem by actively changing the nominal rigidity of the cantilever.

The cantilever can be controlled to be rigid and maintain its straight shape, and

14

thus the deformation of the cantilever is eliminated during manipulation. At the
same time, the control signal is used to represent the interaction force, which can
not be measured by a rigid cantilever only. Hence, the active probe can be used
to improve the accuracy of nanomanipulation and the force sensitivity of the haptic
nanomanipulation system simultaneously. Since the cantilever keeps straight during
manipulation and is adaptable to different sized objects, it is called adaptable end

effector in this research.

—p

, 3 .

 

 

 

 

 

‘5 1w
‘5’
8 F 80 *
fi 5 40
.Z’. V
.0 0 - mm— 1
.9
I" -40 i . i . .
15 20 25 30 35 40 45
Time(sec)
((1)

Figure 1.8. Pushing a silver nanowire with an unpreloaded traditional tip
(k=0.57N/m). (a) Schematic illustration of the pushing process. (b) The AFM
image of a silver nanowire with length of 2.5pm and diameter of 120nm. The scan-
ning range is 10pm x 10pm.(c) The AFM image by a new scan after pushing. ((1)
Tip (Iisplacement signal measured by the photodiode sensor.

A soft cantilever can provide sensitive force feeling while pushing mane—objects,

but on the other hand, the tip is very easy to slip over the mane-objects due to the

15

flexibility of the cantilever and makes the nanomanipulation inefficient. Fig. 1.8.
(a) schematically shows the process of pushing a nano—object using an unpreloaded
soft traditional tip. Since there is no preload, the tip floats on the surface during
manipulation. As shown in the figure, the cantilever starts to bend after hitting the
object and the bending will increase with the tip moving ahead. Finally, the big
bending causes the tip to slip over the object. Experimental results from Fig.1.8. (b)
to (d) verified these analysis. Fig. 1.8. (b) shows an AFM image of a silver nanowire
with length of 2.5 pm and diameter of 120nm. A silicon nitride probe with a spring
constant of 0.57 N / m is used to push the nanowire following the trajectory indicated
by the arrow without preload. Fig. 1.8. (c) shows the image from a new AFM scan
after manipulation. It can been seen that the manipulation failed and the nanowire is
still in its original position. From Fig. 1.8. (d), which shows the photodiode output

(deflection signal), it can be seen that the tip slipped over the nanowire while pushing.

The traditional way to overcome the tip slipping over the nano—objects, is to apply

a preloaded normal force on the tip to keep the tip contacting the surface. Fig. 1.9. (a)
schematically shows the process of pushing a nano-object using a preloaded traditional
tip. As shown in the figure, the cantilever has a large bending even before touching
the object due to the preloaded force. This large bending makes the tip scratch over
the surface and thus avoid slipping over the object. However, since the preloaded
force is much stronger than the interaction force, the shape of the cantilever dose not
change much during pushing the object. Therefore, it is difficult to get sensitive force
information from the deflection of the cantilever. Experimental results from Fig. 1.9.
(b) to (d) verified the above analysis. The same tip used in the last experiment is used
here to push the same nanowire again. Fig. 1.9. (c) shows the image from a new AFM
scan after manipulation. It is observed that the nanowire is pushed away successfully.

Fig- 1 -9. (d) shows the deflection signal during manipulation. The silicon nitride

16

 

    

-4

(nm)
—- N
asset

Tip displacement

 

 

 

I
00
o

—‘

LII

20 25 3‘0 3'5 46 45
Time(sec)

Figure 1.9. Pushing a silver nanowire with a preloaded traditional tip (k=1-5N/m).

(a) Schematic illustration of the pushing process. (b) The AFM image of a silver
nanowire with length of 2.5,um and diameter of 120nm. The scanning range is 10pm x
lOum. (c) The AFM image by a new scan after pushing. (d) Tip displacement signal

measured by the photodiode sensor.

probe was preloaded and stuck on the surface at t : 24360, and then moved toward
the rod. At t : 33sec, the tip touched the nanowire and began to push it. Obviously,
the preloaded force almost saturates the photodiode output and the interaction force
is unobservable. It becomes very difficult to feel the actual tip-object interaction force
in this situation. Consequently, it also becomes very difficult to precisely control the
tip position in the lateral direction during manipulation because the preloaded force
not only causes the cantilever to bend in the normal direction but also causes the tip

to move in the lateral direction [78]. Furthermore, the preloaded force will wear out

the tip quickly and cause contamination easily.

17

 

 

 

 

Acliv/e Probe Object Control voltage
(a) (b) (C)

Figure 1.10. Schematic illustration of pushing a nano-object using the adaptable end
effector.

Fortunately, these problems can be solved using the active probe (spring constant

k = 1-5 N/ m) as an adaptable end effector for AFM based nanomanipulation. The
adaptable end effector can be controlled to behave like a rigid cantilever, at the same
time, the control signal, which is sensitive to the interaction force, can be used to
reconstruct the force information for better haptic force feedback. Fig. 1.10 shows
how an adaptable end effector works during pushing a nano—object. In Fig. 1.10. (a)
the adaptable end effector moves toward the object in straight shape; In Fig. 1.10.
(b), the effector touches the object and is bent by the interaction force; In Fig. 1.10.
(c), a control voltage is applied to the piezo layer of the end effector to eliminate the
bending caused by the interaction force. The cantilever backs to its straight shape
and the object is pushed away.

It has to be mentioned that AF M probe needs to push the nano—objects in any
random direction. When pushing direction is not in parallel with the direction of
cantilever, the manipulation force will not only cause the cantilever to bend in normal

direction, but also cause the cantilever to twist around the axis of the cantilever [85].
Since the cantilever is quite rigid in twisting mode, the twisting force will not be able to
twist the cantilever much and affect the accuracy of manipulation. Therefore, it is not
necessary to actively control the twisting. The only difficulty brought by the twisting

is that the twisting signal is very difficult to measure. For the nanomanipulation

118ng a soft traditional probe, a preload is needed to press the tip on the surface. The

18

preload causes huge friction force between tip and surface. Since the friction force
will submerge the twisting force and the deflection force caused by objects during
manipulation, the measured twisting signal is very noisy. In the nanomanipulation
system using active probe, the control signal represents the deflection force, and the
twisting signal from the photodiode represents the twisting force. Since the cantilever
is controlled to keep straight and stay above the surface, the friction force between
tip and sample surface is reduced significantly. Hence, the twisting force caused by
manipulation can be easily measured.

In this research, an active AF M probe is used as an adaptable end effector for
the AFM based nanomanipulation system. In order to actively control the flexible
adaptable end effector, the mathematical model of the probe is required. Since the
cantilever of the adaptable end effector is flexible and a distributed parameter system,
it is modeled using the Euler-Bernoulli beam model [86], which has been widely
studied to control flexible manipulators [87—91].

The Euler-Bernoulli equation governs the transverse vibration of a flexible beam.
Consider the beam shown in Fig. 1.11(a), which undergoes transverse vibrations.
since the width of the beam is much more significant than the thickness, the strain
along the width of the beam can be assumed to zero. Thus only transverse vibrations

of the beam will be analyzed and the torsional effects are ignored. To analyze the
dynamics of the beam, a small segment of this beam with length dar, at position :13,

is chopped out. The transverse deflection of the beam at position :1: and time t is
denoted by w(.r, t). The segment is subjected to a shearing force, V(.’1.‘), and a bending
moment, M (2:) On the opposite side of this segment, which corresponds to a position
1‘ + da, the shearing force is V(a: + dm) = V(:r) + BV/Brcdcr. Likewise the moment
force at the position :1? + (1.17 is fif(l‘ + (11‘) = M(.r) + 8111/ ardar. This is illustrated

in Fig. 1.11(b) where the term p.4d1'82w/ 012is an inertial force, and p is the mass

density.

19

we

 

 

V

 

 

 

 

 

JL

 

 

32w BM
Adx __
M<Vi (p at, l>M+axdx
3V

V+—dx
ax

 

 

 

 

7r
JL

(*3)

Figure 1.11. Flexible beam

The forces equilibrium condition in Fig. 1.11(b) gives,

8V 8210

The moment equilibrium condition gives,

8M
—-de + —d:1: = 0 (1.4)
8:1:
The bending moment It] is given as [86]
8220 r
M = Ell—83:2 (1.0)

20

Substituting (1. 4) and (1. 5) into (1 3) gives the Euler -Beinoulli equation [86];

1318411182112
E181;4 +pA— fit? 2 0 (1.6)

1.2.4 Testing of CNT Infrared Sensors

The photoconductivity of CNTS has been extensively studied and different optoelec-
tronic properties of CNTS has been tested and reported in the fields of Raman scat-
tering [7], CNT light absorption [92—95], photoinduced molecular desorption [8,96],
fluorescence [95,97], and electroluminescence [98]. The photoconductivity properties
of CNTS have also been studied in various laboratories. As shown in Fig. 1.12, the
temporal photocurrent response and the dependence of the photocurrent on the IR
light intensity were studied in [9—11, 13, 33]. Moreover, as shown in Fig.1.13, the
dependence of the photocurrent on the gate voltage has also been studied in [13].

So far most of the reported work focuses on studying the photoconductivity of
CNTS, but few work has been done to study the photovoltaic effect of CNTS. Since a
photovoltaic device is much more sensitive and has faster response speed than a pho—
toconducting device, it is very important to study the photovoltaic effect of CNTS.
More importantly, CNT based nanophotodiodes have the potential to be nanogen-
erators for driving nanodevices. As shown in Fig. 1.2, a p—n junction is created
in a single CNT to form a p-n junction photodiode [14], and the I-V characteristic
curves with different incident light intensity were measured. But since two bottom
gates are needed in the device, it brings difficulties in fabricating such device in
nanoscale. Hence it is necessary to further explore the photovoltaic effect of CNTS
and develop a simpler design. It has been theoretically and experimentally proven

that the CNT-metal contacts play an important role in the performance of CNT

devices [41,42,99—103]. But few researchers studied the dependence of CNT photo-

21

 

behavior on the CNT-metal contact conditions. Thus the CNT-metal contact will be

another study focus in this research.

8

 

 

8, 00 __L _I 7.

4 g l i l l
] 0a

988

age

2 1

I.

5

U .. a

 

 

 

Current Density, A/cm2
N

 

O

 

T

150 200 250
time, s

0 50 100

Figure 1.12. Temporal photocurrent response of a SWNT film. ( [10])

In this research, we will focus on studying both photovoltaic and photoconduct-

ing properties of single SWNTs and MWNTS. Experimental testings on the temporal
photoresponse and the I-V characteristic curves with different incident light intensity
will be performed. Furthermore, the dependence of CNT’s photobehavior on tem-
perature, and the dependence of photocurrent on CNT-metal contact will also be

studied. Multi-pixel CNT IR sensor array and multicolor CNT IR sensor will also be
fabricated and tested.

1.3 Objectives and Challenges

The objective of this research is to develop single CNT based IR sensors in-
Cluding single pixel sensors, multi—pixel sensor arrays and multicolor sensors. The

development involves not only the theoretical design of CNT IR sensors, but also

22

 

 

1E-7 ,
1E-8 ,

1E-9 1'

Current (A)

113-103

1

lE-ll 1‘

 

 

 

113-12 .....,
-6 —4

 

«-
q

-2 0
Gate voltage (v)

Figure 1.13. Gating characteristics of an ambipolar NT-FET with and without in-
frared illumination ( [13])

experimental manufacturing and testing of CNT IR sensors. Though some work has
been reported to study the photoconductivity of CNTS, the quantum efficiency is still
too low and the fabrication process is inefficient and unreliable. And the multi—pixel
CN T IR sensor and multicolor CNT IR sensor have not been demonstrated.

The main challenges involved in this research include:

o A well designed CNT IR sensor. Since the impedance of CNTS (especially
semiconducting CNTS) can be up to giga— or tera-Ohms in general, the current
signal will be very small. How to increase the photocurrent signal and effectively
read it out are the main difficulties to consider during the design of the CNT

88118018.

0 An efficient and reliable nanomanufacturing process. Since a single CNT is
desired to bridge the microelectrodes, the most diflicult work in this research

is how to deposit only one CNT onto the 111icroelectrodes in an efficient and

23

controllable manner. There are some methods available to fabricate a single
CNT based nanodevice. But there is no a solution to fabricate single CNT

based nanodevice arrays at present.

0 CNTS with appropriate bandgaps for MIR detection. One of the objectives of
this research is to detect MIR using CNT sensors. Since most SWNTS have very
thin diameters and large bandgaps, and most MWNTS have very large diameters
and are metallic, it will be a challenge to get a CNT with an appropriate

bandgap for MIR detection.

1.4 Organization of Dissertation

Design of the CN T based IR sensors is presented in Chapter 2, where the struc-
tures of the CNT sensors are discussed. Following the discussion of the design, the fab-
rication and assembly process for building single CNT based nanodevices is developed
in Chapter 3. First the fabrication process is addressed. Then the key technology for
assembling single CNT devices, the AFM based nanomanipulation system enhanced
by an adaptable end effector, is developed and experimentally verified. After that,
experiments are carried out to fabricate single CNT based IR sensors to show the
efficiency of the developed fabrication process. After describing the fabrication and
assembly process in Chapter 3, experimental testings of the fabricated CNT sensors
are discussed in Chapter 4, where the performance of the single CNT based IR sen-
sors are also evaluated and analyzed. Chapter 5 reports the fabrication and testing
of multi-pixel CNT sensor array and multicolor CNT sensor. Finally, a summary of

the developed methodologies and our contributions are provided in Chapters 6.

24

CHAPTER 2

Design of Carbon Nanotube based

Infrared Sensors

The challenges of designing single CN T based infrared sensors lie in two aspects, the
fabrication and the testing. Since the diameter of a CNT can be as small as sub-
nanometer, the design of the sensor structure is essential for the fabrication process.
An appropriate design can make it easier to assemble a single CNT onto the micro-
electrodes. Another challenge is how to improve the performance of the CNT sensors.
Since the impedance of CNTS (especially semiconducting CNTS) is very huge in gen—
eral, the current signal will be very small. How to increase the photocurrent signal
and effectively read it out are the main difficulties to consider during the design of

the CNT sensors.

2.1 Manufacturing Issues in Design

The structure of single CNT based IR sensor is shown in Fig. 2.1(a). It consists of two
microelectrodes separated by a small gap and bridged by a single CNT. Fig. 2.1(b)
Shows the cross section of the microelectrodes. Since the CNT will be pushed onto the

microelectrodes by AFM tip, a slope at the edge of the microelectrodes is necessary.

25

Fig. 2.2 illustrates pushing CNT onto the microelectrodes with different step profiles
which are generated by different fabrication processes. CNT can easily be pushed
onto the microelectrodes which are fabricated by lift off process rather than etching
process. The CNT may get stuck at the edge of the microelectrode and cannot be
pushed up onto the microelectrodes as shown in Fig. 2.2(a). Moreover, to make it
easier to push CNTS onto the microelectrodes, the thickness of the microelectrodes

has to be as thin as possible. Here the microelectrode is designed to be 30 nm thick.

/ Electrode \

[ Substrate ]

    

Electrode
7_ Substrate

(a) (b)

       

Electrode /
—/

 

 

Figure 2.1. (a) Structure of the single CNT based infrared sensor. (b) Cross section
of the microelectrodes.

 

 

 

 

 

 

CNT
Electrode A j (.215.
x 2 In
I l ' l
(a) (b)

Figure 2.2. Illustrations of pushing CN T over the microelectrodes which are fabricated
by (a) etching and (b) lift off process.

2.2 Diode based CNT Sensor

A semiconducting CNT can work as a photoconductor or a photodiode for photo

detection. Due to its high dark current and big noise, a plit’)toconductor normally has

26

 

a lower sensitivity than a photodiode. Therefore, a carbon nanotube Schottky diode
will be designed for the infrared detection. Fig. 2.3 shows the energy band diagrams
of the CNT-metal contact for n-type and p—type CNTS respectively. For n—type CNTS,
a Schottky barrier is formed at the CNT-metal interface when (15 M > X C NT1 where
d) M is the work function of the metal electrode, and XC' NT is the electron affinity
of CNT. For p-type CNTS, a Schottky barrier is formed at the CNT-metal interface
when ng < X C NT + Eg, where E9 is the bandgap of CNT. Since semiconducting
CNTS normally are p-type in air due to the oxygen doping, the case of Fig. 2.3(b) is
considered in this research. As shown in Fig. 2.3(b), there is a Schottky barrier at

the CNT-metal interface which blocks the transportation of holes.

50. __.

  

r-

__EC

L

 

.. __EFCNT
Ev

 

 

 

Figure 2.3. The energy band diagrams of the CNT-metal contact for (a) n-type CNTS
and (b) p—type CNTS. .

2.3 Optimizing the CNT Sensor

By choosing metals with different work functions as the microelectrodes, we can
adjust the Schottky barrier at the CNT-metal interface. Two different designs are
developed in this research to study the dependence of the photocurrent on the CNT—
metal contact conditions. Fig. 2.4 shows the structure and the energy band diagram
0f CNT sensors with symmetric contacts, where both of the microelectrodes are made

from a same material. As infrared photons hits the SWNT, the photons with energy

27

bigger than the band gap will excite electrons and holes inside the CNT to form
electron-hole pairs. The electron-hole pairs will be separated by an external electrical
field or the built-in field and contribute to the photocurrent. Since there is no built-in
field outside the depletion regions, the electron-hole pairs will recombine quickly and
do not contribute to the photocurrent. Only the electron-hole pairs generated at the
depletion regions can form photocurrent. Hence, the total photocurrent under zero
bias is calculated by Iv = 1111 + [1,2 = fhl — fh2' As shown in Fig. 2.4(b), the holes
generated at one Schottky barrier have to tunnel through another Schottky barrier
to form a current. This significantly limits the magnitude of the photocurrents [v1
and 1.02. Moreover, since the CNT—metal contacts are symmetric, 1,,1 is very close to
1.02 if the incident IR power intensity is same on the both electrodes. Thus the total

photocurrent will be very small and becomes difficult to measure.

 

 

 

 

 

 

Depletion region Depletion region
] [ Econduction i i .
. 1“ l 162
CNT hv>Eg
Au A” ' \ Efermi
.. E ih2 A”
Substrate Iv I . valence 4.— Iv2

 

 

 

(a) (b)

Figure 2.4. The single CNT based infrared sensors with symmetric contacts. (a)
Structure of the CNT sensors. (b) Energy band diagram of the CNT sensors.

Fig. 2.5 shows the structure and the energy band diagram of CNT sensors with
asymmetric contacts. By using different materials as the contact electrodes, asym-
metric CNT-metal contacts are created to maximize the total photocurrent Iv. By
lowering one of the Schottky barrier, the photogenerated holes at another barrier will
be able to tunnel through this barrier easier. Thus the total photocurrent will be

increased. Moreover, the total photocurrent will also be increased by enlarging the

28

Schottky contact
depletion region Ohmic contact
‘ I

l
f

Econduction

 

l

         

CNT hv>Eg

Efermi

 

 

Substrate

 

 

 

(a) (b)

Figure 2.5. The single CNT based infrared sensors with asymmetric contacts. (a)
Structure of the CNT sensors. (b) Energy band diagram of the CNT sensors.

difference between the photocurrents generated at the two barriers due to the asym—
metry of the CNT-metal contacts. For example, as shown in Fig. 2.5(b), since many
people have reported that Pd is the best materials to form an Ohmic CNT-metal
contact [41,104], when silver (Ag) and palladium (Pd) are used as the contact mate-
rials, a Schottky contact is formed at the CNT-Ag interface and an Ohmic contact is
formed at the CNT-Pd interface. In this way, the photocurrent 11,2 = 0 and the total
photocurrent Iv is maximized to 1.01. With the asymmetric CNT-metal contacts,
the sensitivity and other performance of individual SWNT based IR detectors are

expected to be further improved.

2.4 Chapter Summary

Design of the single CNT based infrared sensors is discussed in this chapter.
Single CNT based Schottky photodiodes have been designed for infrared sensing.
The shape of the microelectrode was specially designed for pushing CNT onto the
microelectrodes. Structures with symmetric and asymmetric contacts have also been
designed for studying the dependence of photocurrent on the CNT-metal contact

conditions.

29

CHAPTER 3

Assembly of Carbon Nanotube

based Infrared Sensors

Since a single CNT is desired to bridge the microelectrodes, the most difficult work
in this research is how to deposit only one CNT onto the microelectrodes. As dis-
cussed in Chapter 1, the directly growth method cannot guarantee only growing one
CNT between the microelectrodes, and the DEP deposition method normally brings
multiple CNTS to the electrodes. Although the EBL fabrication method is able to
form single CNT devices, the position of the CNT/device cannot be controlled, and
the efficiency is low. AFM manipulation can position CNTS on the substrate surface,
but the CNTS to be manipulated have to be close to the microelectrodes. Combining
the advantages of DEP deposition and AF M manipulation, a nanomanufacturing pro-
cess with which a single CNT can be manipulated to bridge a pair of microelectrodes
was developed. The DEP method is used to trap one or several individual CNTS
around the microelectrodes, then AFM is used to manipulate one of the CNTS to
make connections with the microelectrodes and clean the rest CNTS away from the

microelectrodes.

3O

3.1 Development of the Fabrication Process for
Single CNT based Nano—Devices

The structure of the individual CNT based IR sensor is shown in Fig. 3.1(a). A
CNT IR sensor is composed of a pair of electrodes and a single CNT. Sensors are
fabricated using standard photolithography process and AFM lithography process.
An array of gold micro electrodes is fabricated on a substrate with the process shown
in Fig. 3.1(b). After the connected electrodes are fabricated with the process of Fig.
3.1(b), as shown in Fig. 3.1(a), the AFM based nanomanipulation system is used to
create a small gap with distance of 0.5 — 2 pm across the two electrodes. With the
AF M system, the gap can be easily created and the width of gap can be controlled

to vary from several tens of nanometers to several micrometers.

 

r J Spin on photoresist on

Gold _ Gold / I l quartz substrate
_’ \

_
l | [El l l Pattern and develop PR

 

3. AFM cutting

- Deposit Titanium and

‘fl CNT deposition
l I Remove photoresist

(a) (b)

 

 

_ 2 GOId

Figure 3.1. (a) The CNT assembly process. (b) The fabrication process of micro
electrodes.

The MWN T in our experiments are obtained from Shenyang National Laboratory
(SYNL) for Materials Science, China. It takes several steps to deposit an individual
CNT onto the substrate. Fig. 3.2 shows the flow chart of the CN T deposition process.
Firstly, the powdery CNTS are put into acetone and dispersed by ultrasonicator for

10-15 min to form CNT suspension. The concentration of CNTS in acetone is experi-

31

2.8 ug/ml
Ultrasonic

CNT diSPe'Sing CNT M‘Cmp‘l’ene droPp‘mg CNTS between

owder ’l sus nsion . . ’ electrodes
p pe & dielectrophoreSIS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Manipulating using AFM based
Single CNT Nanomanipulation system

between electrodes ‘

 

 

 

 

 

 

Figure 3.2. The CNT deposition process.

mentally optimized to 2.8 ug/ ml. Then a drop of 5 — 10 pl of the CNT suspension is
dropped onto the substrate between the two electrodes. At the same time, a sine wave
ac signal with the frequency of 10 kHz and the peak to peak amplitude of 1.5 to 2 V
is applied between the two electrodes for several seconds. The dielectrophoresis force
traps CNTS to the electrodes. Finally, the acetone on the substrate is evaporated. In
the deposition process, the concentration of the CNT suspension, the frequency and
the amplitude of ac signal are experimentally optimized to get the highest possibility
to trap an individual CNT between the two electrodes. Even with those parameters
experimentally optimized, it can not be guaranteed that only one CNT is trapped
between the two electrodes and bridges them successfully. Normally, there are always
some impurities or more than one CNT trapped between the electrodes. Furthermore,
if only one CNT was successfully trapped between the electrodes, it might not connect
the electrodes very well. Therefore, it is necessary to take another step to clean up
the detector area and adjust the position of the CNT to make a connection. This final
step is very critical and is called CNT assembly. An AFM based nanomanipulation

system was developed to performe the CNT assembly.

32

3.2 Development of the AFM based Nanomanip-
ulation System

The AFM based nano-robotic manipulation system with the augmented reality inter-
face can provide the operator with real-time visual display and force feedback during
nanomanipulation [77-79]. It consists of three subsystems: the AFM system, the
augmented reality interface, and the real—time adaptable end effector controller.

As shown in Fig. 3.3, the AFM system includes the AFM head, signal access
module (SAM), AFM controller, and the main computer. The SAM provides inter-
faces among the AFM head, AFM controller, and other devices. The AFM controller
controls the AFM head through the SAM. It also connects to the main computer,
which is responsible for running main control program and providing an interface
for the imaging. The augmented reality interface consists of a haptic device and a
nanomanipulation program running on a Windows PC. The nanomanipulation pro-
gram provides an interactive interface for users. During manipulation, operators use
the haptic device to input the tip position command and feel the real-time interaction
force between the tip and the nano—objects. The real-time visual display is dynamic
AFM images locally updated in video frame rate. The augmented reality interface
also sends tip position command to the main computer. The real-time adaptable end
effector controller subsystem includes a real-time linux system and a DAQ card. The
adaptable end effector controller running in the real-time linux samples deflection
signal of the cantilever through DAQ card and then outputs control signal to the
adaptable end effector. At the same time, the control signal is also rendered to the
haptic device to display the interaction force. These three subsystems communicate
with each other through the Ethernet.

Fig. 3.4 shows the details of the control scheme of the nanomanipulation system

using the adaptable end effector. The operating environment sends position command

33

to the AF M controller which controls the tip position. An active probe is used as the
adaptable end effector and a control loop is introduced into the system to keep the
adaptable end effector rigid during manipulation. During pushing, the adaptable end
effector is controlled to keep straight by feeding back the deflection signal from the
photodiode detector. At the same time, the control signal is also sent to the haptic

device (PhantomTM) as a force signal through the operating environment.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

é Deflection signal
S ‘ . Adaptable
E control Signal DAQ End Effector
:2 y
3 I
a:
a a 53’
_ 3
E g E H AFM controller H SAM AFM head
in“ 8
visualization
E p. Interface
;
é Force signal
é ‘ Phantom
Position command

 

 

 

Figure 3.3. The setup of the AF M based nanomanipulation system using adaptable
end effector.

3.2.1 Modeling of Adaptable End Effector

As shown in Fig. 3.5, the active probe is a MEMS device, micromachined from bulk

silicon with a Zinc Oxide piezoelectric film patterned along a portion of a. micro—

34

Deflection signal y(t)

 

 

 

 

 

 

C(t) Adaptable End V(t)
Effector
+ Controller
(1
y (t)

 

 

 

 

 

 

 

Visualization:

. Command_

Operating AFM
Environment 4

Imaging
Phantom /

 

 

 

 

Controller

 

 

 

 

 

 

 

 

 

 

Figure 3.4. The control scheme of the AFM based nanomanipulation system using
adaptable end effector. 1: Piezo tube. 2: Piezoceramic layer of the adaptable end
effector. 3: Adaptable end effector. 4: Substrate surface. 5: Object. 6: Laser beam.
7: Quad—photodiode detector. 8: Mirror. 9: Laser gun.

cantilever [105]. The Active Probe was originally developed to improve the tapping
mode imaging rate by dynamically mediating the resonant vibration of the AFM
probe (Tapping Mode) while simultaneously positioning it vertically to track the sam-
ple topography [83, 84]. However, in this research, the active probe will be modelled

as an adaptable end effector of the AFM based nanomanipulators.

It can be observed from Fig. 3.5 that the foremost Si part does not have ZnO
layer bonded. Since the spring constant of a cantilever is in inverse proportion to
the cube of the length of the cantilever, the Si part is even a little bit more rigid
than the ZnO/ Si part. The calculated spring constants of the Si part and ZnO/ Si

part are 76.1N/7n and 71.3N/m respectively. Furthermore, when a force is applied

 

Figure 3.5. Adaptable end effector with an integrated Zinc Oxide piezoelectric actu—
ator.
on the free end of a flexible beam, the beam has large deformation at the fixed end
and little deformation near the free end. Therefore, the interaction force applied on
the tip will cause the 2110 / Si part to deform more and cause less deformation on the
Si part. Since the Si part is quite rigid, by controlling the shape of the ZnO/ Si part,
the stiffness of the cantilever can be controlled to change in the range of 1 — 76N / m.

A sketch of the adaptable end effector is shown in Fig. 3.6. Since the width of the
cantilever is much more significant than the thickness, the strain along the width of
the cantilever can be assumed to be zero. Thus only transverse vibrations of the beam
will be analyzed. The cantilever will be analyzed as two parts, Part I (0 S at g [1)
and Part II ([1 S :1: S L). For all variables in this research, a superscript 1 indicates
Part I, while a superscript II indicates Part II. Similarly, subscript b and 1) refers to
the silicon beam layer and the piezoceramic layer respectively. For example, (EI )1 is
the effective bending stiffness of Part I, and (EI)II is the effective bending stiffness
of Part II; Eb is the Young’s modulus of the silicon layer, and Ep is the Young’s
modulus of the piezoceramic layer.

As shown in Fig. 3.6(b), there is a layer of piezoceramic under the silicon can—
tilever. There are also another two very thin metal layer absent in the figure which

are used for electrical connection. The piezoceramic film is sandwiched between these

36

/\/]

 

Z
Y
Silicon layer
X
— ’4' _ Ir ‘

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

._A
,. L "'" —'
w‘4—— A
l
. A] 1 _, P.
\/ I920 ”11
(a)
A Z hb
///v::/// llv r W2 4
n? I 'A\\\ \\\\ luff /////// {2
I We Da

Cross section A-A

(b

)

Cross section 8-8

(0)

Figure 3.6. Adaptable end effector configuration.

two metal film. By applying a potential between these two metal film, the piezoce-

ramic film is driven to extend or contract along the length of the cantilever. Since

these layers are tightly bonded together and the structure is asymmetric in Z direc-

tion, extending or contracting the piezoceramic film will cause the whole cantilever

to bend up or down.

When an active voltage V(:r, t) is applied to the piezoceramic film, the strain 6a

induced in the piezoceramic film is given by [87]:

d
ea(:r, t) = V(a:,t) x i

hp

37

(3.1)

where ea is the induced strain in the piezoceramic film, V is the voltage applied on
the piezoceramic layer, (131 is the appropriate static piezoelectric constant that relates
the electric field applied across the piezoceramic layer to the resulted strain, and hp
is the thickness of the piezoceramic layer. Since the piezoceramic layer has uniform

geometry along its length, V(:r, t) is replaced by V(t) in the following analysis.
The location of the neutral axis at the cross section A-A, as shown in Fig. 3.6(b),

is given by:

D _ Ephgwp + EbhEWb + 2131,thth (3 2)
a — 2(Ebthb + Eprhp) .

 

where E is the Young’s modulus, h is the thickness of the layers and W is the
width of the layers.

Suppose a positive voltage is applied and the cantilever is driven to bend up as
shown in Fig. 3.7(a). To analyze the stress inside the cantilever, a small piece of the
cantilever with length dat is chopped out, as shown in Fig. 3.7(b). (19 is the bending

angle of this small piece and is given by

__ 82w(:z:, 25)

d6
8:132

(3.3)

where w(.1:,t) is the transverse deflection of the cantilever. In the piezoceramic

layer, the stress 0])(13, z, t) is given by [106]:

a,,(:1:,z,t) = Epo [5(33, 2,15) — s,,(:1:,t)] (3.4)

38

\ \\g\\\\

 

 

 

 

Ml(x,t)

 

 

..........

 

 

D
f op(x,z,t)

3 \

‘———————-—————>‘ \

Figure 3.7. Diagram of the cantilever under bending.

Then the strain 5(13, 2:, t) is

where E is the strain caused by the deflection of the beam and is different for each

2 position. In the silicon layer , as shown in Fig. 3.7(b), the stress 0b is given by:

0b(:c, 2,15) = E, - 5(1), 2:, t) (3.5)

At position z, the deformation of the cantilever in the a: direction is equal to d6 - z.

€($, z, t) 2 d6 - z/da: (3.6)

39

The stress 0b and 0p generate two moments Mb and Mp in the cross-section C-C,

which can be calculated as

Mb 2 0b(:r, 2,15) - dA . z = 0b(a:, z, t)debz (3 7)
Mp = 0,,(513, z, t) - dA - z = 019(3), 2, t)depz '
By integrating these two moments along the z direction, the overall moment

A1103, t) in the cross-section C-C can be derived as

MI(33, t) = :llifihp ap(:r, t)szdz + f:£l“:,:;p+hb 0;,(33, t)szdz
mrogxgh ma

Substituting (3.1) - (3.7) into (3.8) gives

M1(;c, t) = (EI)’ . 8213;? t) — Ca - V(t) (3.9)

 

where (EI )1 is the effective bending stiffness at the section A-A for part I, and
Ca, is a constant which depends on the geometry and materials of the beam. (E1)!

and Ca are given by

40

(131)1(37: t) :Epl/Vphp ' (1/3 * h?) + Di — hpDa)+

1
Ebthb[§h§ + hf, + D3, + hbh, — 2hpDa — hbDa]

hp — 2D,
2

 

C,=E,,.d31.wp.

On cross section B-B, as shown in Fig. 3.6.c, the moment of part II, [W I I , can be

[11

derived in the same way as A11. A is given by

11 . 82w(x, t)
(9:132

 

M”(:c, t) = (E1) (3.10)

where (EI )1 I is the effective bending stiffness at the section B-B and given by

(E1)” = EbW2h3/12

Considering part I (0 S .r S 11) and part II ([1 S :r S L) of the cantilever re-
spectively, and combining (3.9) and (3.10) with a. conventional Euler-Bernoulli beam
analysis yields the equations of motion for transverse vibrations 'w(.'r,t) of the can-

tilever. The governing equations are

41

8722(EI)1828(10(27,0 __W t)]+pIA182w§ -13-t)__0
for ngxgll

(E1)”—34:+§;’—>+ pHAUi‘gg—tl— _ 0 for 11 g a: g L

(3.11)
with the boundary conditions
r w(0, t) — 0,
w’(0, t) = 0,
I II _ = II II
, (E1) w (at) Cave) (E1) w (at). (3,2,

(EI Vii/”(11, 75) = (E1 Viv/”(11, t),
(EI)”w”(L, t) = O,
(EI)”w’”(L,t) = 0.

 

K

where

42

p1 = (ppwphp + awn/(mm + thb).

p” = pm

A1 = thp + thb,

A“ = Wgnb,

’LU’(CE, t) = 8w(:c,t)/8:17,
w”(a:, t) = 82w(a:, t)/8:1:2,
w”’(:c, t) = 83w(a:, t) /8:c3,

w””(:1:, t) = 84w(:r, t)/8$4.

and p is the density of the layers; A is the cross-sectional area of the layers.

Since the applied voltage V (t) is spatially uniform along the length of the can-
tilever, the spatial derivative of V, which is the input of the system described by
the first equation of (3.11), equals to zero. Now the applied voltage V(t) only has
effect on the third equation of the boundary conditions of (3.12). Thus, the system
is simplified to a linear distributed parameter system that is actuated only at the
interior boundary, a: = 11. These partial differential equations can not be directly
used for traditional controller design. In the following, the Lagrange method is used
to obtain a decoupled ordinary differential equation to describe the system in order
for controller design.

The kinetic energy of the. beam is given by

43

1 ‘1 1 L
Tz-é/ p1A1w2($,t)dx+§/ puAHw2(a:,t)d:c(3.13)
0 11

where if; is the derivative of it) with respect. to time t. The potential energy of the

beam is given by

 

 

P = g All 1 [(EI)182w(m’ t) + CaV(t)]2d:z: +

 

(EI)I 8x2
L 2
:- / (1:71)”? glides.» (3.14)
11 55

Using the assumed mode—summation method, the solution of (3.11) can be ex-

pressed by

wtnti = 23:. aim-(t) for o s a: s 11
w(:c,t) = 2:, q5,f1(a:)qi(t) for [1 g a: g L

where qi(t) is the modal coordinate, and gal-(cc) is the mode shape which has the

following form

¢f(:1:) = Afstnflf + chosflfa: + Cfsz'nhflfa; + Df'coshflfa:

where k = 1,11

44

The coefficients A,, Bin C17? Di are determined to satisfy the boundary conditions
given by (3.12).

Substituting (3.15) into (3.13) and (3.14), and then by using the Lagrange equa-
tion and augmenting proportional damping (damping coefficient (2), the governing
equations can be expressed by decoupled ordinary differential equation for each mode

of the cantilever as follows.

 

-- . ON 1 1 62(1),!
qi + QCiwiq'i + 013%: Z — 12'( ) ()1 de

for 2': 1,2, ...,00 (3.15)

where “’2'. is the eigenfrequency of each mode, and I,- is the generalized mass,

defined as follows.

11 L
1: / (aw/11m / (awn/111a
0 11

3.2.2 Control of Adaptable End Effector

In order to keep the adaptable end effector straight during manipulation, a LQR con-
troller is designed in this research to control the adaptable end effector. An advantage
of the LQR control method is the linearity of the control law, which leads to easy
analysis and practical implementation. Another advantage is its good disturbance
rejection and tracking performance and stability. This control algorithm is a state

feedback controller and requires a complete knowledge of the whole states for each

time instance. However, only the measurement of the tip displacement is available in
this system and no other states are measurable, a full order observer is then required
to estimate all of the states.

The goal of the controller is to eliminate the deformation caused by the interaction
force between the tip and the manipulated object. In other words, it is desired to
have w(L, t), the transverse deflection at the end of the cantilever, equal to zero. By

defining w(L, t) as the output y(t), the following equation can be derived from (3.15)

00

ytt) = th, t) = 2: ML) - (1.-(t) (3.16)

i=1

A state space model of the system can be obtained using the finite dimensional
approximation of the model. Since the higher frequency modes tend to damp out
faster and have less effect on the dynamics of the cantilever, only the first N modes

are considered in the controller. The state equation for the system is then given as:

$=Am+Bu

3N
y=C$+Du ( )

where

46

T
(I): ,7 . . .,'
[(11 C12 (in C11 (12 (In i1x2.v

0 N x N I N x N

A = we"? -2C1uJ1

—LU?V —2C1’VW‘N _

 

 

, (3.18)
B =[01XN — Ca/Il 0'182¢{/8£C2d$

— C, /IN [0’1 62¢{V/8w2dcrlT

C: [451(L) ¢N(L) 01le

D20, u=V(t)

Since the controller will be implemented on computer, the continuous model has
to be discretized with sampling and command updates at intervals T. In addition
we assume that a zero-order hold at the controller output will produce a piecewise
constant command u(t) = u(kT) for kT S t < (k + 1)T, k = 0, 1, ...00. The discrete-

time state space model is given by

513(k +1) 2 <I>a:(k) + Fu(k)

ye) = Cave) (3'19)

In state feedback, the control input takes the form

u(k) = —K:c(k') (3.20)

47

Therefore choosing K corresponds to choosing the closed loop system poles to
give the desired response. To help choosing appropriate values of K, several optimal
control techniques have been developed. The basic idea behind these techniques is
that a cost function is defined and then a controller is formed that minimizes this
cost function. In this research, the LQR method is used to optimize the value of K.

The discrete-time cost function is defined as [107]

00

k)T T
=2 :[93 k)kQ$( )+ uk( ) [221(k)] (321)
1:20
Here, Q = QT Z 0 and R 2 RT > 0 are weighting matrices which can be used
to define the importance of individual states and control inputs.
Since the modal coordinates and their derivatives are chosen as states and they

are unmeasurable in this system, a full order state-estimator is constructed as,

:i:(k +1|/<:) = (<1)— LC):i: (an — 1) + Ly(k) + Fu(k) (3.22)

where L is the observer gain matrix, and the notation .i:(k + llk) denotes that the
estimate of :r(k + 1) is made using measurements available at time k. By choosing L

appropriately, i7 approaches .1: exponentially since the estimation error is,

e(k+1|k)

517(19 +1) — :c(k +1lk)
2 (<1) — LC)e(k|/~c — 1) (3.23)

48

The dynamics of the observer are given by the poles of (<1) — L at C). Since the
dynamics of the estimator should be much faster than the system itself, the poles of
the observer have to be placed at least three to five times farther to the left than the
dominant poles of the system. With the state-estimator, we can obtain an output-

feedback controller using the estimated state it instead of the true state .17.

u(k) = —K:i‘:(k) (3.24)

Using this feedback, the closed loop system now becomes

:r(k+ 1) [ <1> —rK 512(k)
at +1) LLC <1> — LC — rK 23(k)
' BG

.25
+ BG 7“ (3 )

 

The structure of the controller is shown in Fig. 3.8. The term G is a constant

gain, and is included to achieve asymptotic tracking of a. step input.

The existence and stability of the steady—state LQR solution has been proven
in [108] and [109]. For a LQR problem with R > 0, and Q = CTC’, where the pair
((1), C) is detectable and the pair (QT) is stabilizable, it follows that a solution to
the steady-state LQR problem exists. In particular, there exists a unique positive

semidefinite solution W to the discrete algebraic Riccati equation

49

 

z-(£)>x(k+1)=<I>ir(k)+l"u(k) x(k) C‘ y(k)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

+| 5&(k+1)=<l>5c(k)+f‘u(k) 5&(k) 52(k)
+L(y(k)-9(k)) C ’

 

 

 

Figure 3.8. Controller block diagram.

W = Q + <1>TW<1> — <1>TWP[R + rTwrrerwe (3.26)

and if

K = (R + rTwr)-1rTW<1> (3.27)

then the closed loop system (3.25) is asymptotically stable.

3.2.3 Experimental Implementation

The experiments on the AFM based nanomanipulation system were performed at
stable room temperature. As introduced in Fig. 3.4, the nanomanipulation system
mainly consists of a NanoScope IV Atomic Force Microscope (Digital Instruments
Inc., Santa Barbara, CA) and some peripheral devices including an optical micro-
scope, a haptic device (Phantom), a Multifunction Data Acquisition (DAQ) card NI

PCI-6036E (National Instruments), and three computers.

50

System Identification

In the experiments, an active probe DMASP (Veeco Instruments), which has been
introduced in Section III, was used as the adaptable end effector. The tip displacement
is obtained from AFM’s photodiode sensor.

The parameters of the adaptable end effector are given in Table 3.1. These param-

eters are obtained from the manufacturer’s specification sheets or from the materials

 

 

 

 

 

 

handbook.
Table 3.1. Cantilever parameters

Parameters Values Parameters Values
11 374nm L 500nm
hp 35pm hb 411m
”(Up 250nm tub 250nm
2122 51 pm h? 4pm
Ep 1.2 x 1011N/m2 Eb 1.69><1011N/m2
pp 5.2 x 103Kg/m3 pb 2.33 x

103Kg/n13

d31 3.7 x 10—12m/V k 1 — 5N/m

 

The damping coefficients for each mode are identified experimentally. They are
adjusted to get a best match in the frequency response of the theoretical model and
the experimental data. The damping coeflicients for the first two modes are listed in
Table 3.2.

To experimentally verify the accuracy of the cantilever model, an excitation step
voltage of —5V is applied to the piezo actuator, and then the vibration of the can-
tilever tip is measured and analyzed. Fig. 3.9 shows the measured tip displacement
in the experiment. It is shown that the tip is driven to —100nm and then keeps oscil-

lating around that point until settling down. To identify the frequency components

51

Table 3.2. Damping constants

 

 

Mode Theoretical frequency (KHZ) Damping coefficients

 

1 49.2 0.005

 

 

2 216.3 0.007

 

of the vibration, the fast-Fourier transform (F F T) is taken to compute the power
spectral density, a measurement of the energy at various frequencies. The frequencies
of the vibration can be easily identified from the power spectral density plot as shown
in Fig. 3.9. It can also be observed from Fig. 3.9 that the first mode dominates the

response of the cantilever over other higher modes.

The natural frequencies of the cantilever can also been obtained numerically us-
ing the modal analysis method. The experimentally obtained modal frequencies are
compared with the theoretically determined modal frequencies in Table 3.3. It can be
seen that the theoretical values are in good agreement with the experimental values.
This confirms that the theoretical model is accurate enough to predict the modal

frequencies and can be used for controller design.

Table 3.3. Modal frequencies (kHz)

 

 

Mode Theoretical frequency Experimental frequency

 

I 49.2 49.8
2 216.3 216.0
3 374.4 372.5

 

 

 

52

 

 

 

 

 

 

 

 

 

 

8 I I I I I l U l I
50
7— 0 ‘ 4
E -50
5
6* ‘5-100- -
E
a150-
s. 5" a ‘
g g-zoo
o 9
a 4, H250» _
_300 1 1 1 1 L 4
0 1 2 3 4 5 6 7
3- 0'13 Time (sec) x10'3 _
2'- ..
0.005
1 I ___—1—._j_ —A
1 _ 2.15 2.2 3.7 3.72 3.74 _
0 J I 1 1(1‘ I 1 I ‘l \. 1 l

 

 

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Frequency (Hz) X10

Figure 3.9. Measured tip vibration of the cantilever without closed-loop control.

Implementation of the controller

The LQR control law has been implemented with the parameters in Table 3.1.
From Fig. 3.9, it can be seen that only the first modes dominates the response
of cantilever during free vibration. Therefore only the first mode is considered in
controller design. Thus the modes number N equals to 1. The weighting matrices
were chosen by the Bryson’s rule [110] and a trial-and error iterative design procedure.

The resulting weighting matrices are:

53

Q = diag(100 1.5) ,
R = 0.001

Then the gain K can be calculated:

(3.28)

K = [1.1830 x 106 5] (3.29)

The real-time implementation of this controller is performed using an x86 based
PC running Linux operating system. The RTAI [111] (Real-time Applications Inter-
face) patch is used to provide POSIX compliant, real-time functionality to the Linux
OS.

To illustrate the effectiveness of the controller, closed-loop trajectory tracking
experiments were carried out. The best effort has been made to tune the parameters
of the LQR controller to get a best tracking performance. Fig. 3.10 shows the tracking
responses of a sine wave and a triangular wave respectively. The results shows that
the tip can be controlled to follow the desired trajectory very well. The amplitude
of the error for sine wave tracking is 10 nm. The closed-loop trajectory tracking
experiments were also performed with a simple PD controller. The best effort has
also been made to tune the parameters of the PD controller to get a best tracking
performance. Fig. 3.11 shows the tracking responses with the PD controller. The
amplitude of the error for sine wave tracking is 20 nm. By comparing these two
experimental results. It can be seen that the tracking errors of the LQR controller

are smaller than those of the PD controller.

54

 

 

 

 

Tip displacement (m)

 

 

 

 

 

 

E
E
a)
E
o
O
S
O.
.9
'0
.9-
1..

0.0 0.1 0.2 0.3 0.4 0.5

(b) Time (sec)

Figure 3.10. Tracking response of (a) a sine wave and (b) a triangular wave. The
controller is LQR controller.

Experiments of tracking a square wave are also performed. Fig. 3.12. (a) and
(b) show the tracking responses of a square wave with the PD controller and the
LQR controller respectively. It can be observed that the settle down time of the LQR
controller (0.4 millisecond) is smaller than the settle down time of the PD controller
(0.7 millisecond). As shown in Fig. 3.12. (c), by zooming into 0.016 sec in Fig. 3.12.
(b), the settle down time of the closed-loop system with LQR controller is only 0.4
millisecond. It is much smaller than the settle down time of the open-loop system,
which can be estimated as 3 millisecond from the insert of Fig. 3.9. Hence, the LQR
controller does improve the dynamic performances of the cantilever.

It has to be mentioned that the noise of the measurement is quite large due to the
characteristics of the photodiode. From the insert of Fig. 3.9, it can be calculated
that the amplitude of the noise is more than seven percent of the signal. Undoubtedly,

the noise will affect the control effect. Another factor reducing the performance of the

55

i‘ ‘i i‘ i‘ 11‘ l l‘ ‘ l i‘i

 

‘11

 

A

 

Tip displacement (m)
o

 

1 5 error 1
' 0.0 0.1 0.2 0.3 0.4
1 x10” (a) Time (sec)

 

 

Tip displacement (m)

 

 

(b) . Time (sec)

Figure 3.11. Tracking response of (a) a sine wave and (b) a triangular wave. The
controller is a PD controller.

controllers is the relatively low sampling rate of the real-time linux system, which can
only reach 70 KHz. However, the frequency of the first mode is around 50 KHz. Since
only less than two control inputs can be applied on the cantilever during one vibration
period, it is really difficult to get good control effect. That is the reason that the LQR
controller, designed after a PD controller, does not improve the performance much.
The large overshoot of the step response is also due to the low sampling rate. However,
it can be reduced significantly by including input shaping [112] in the controller. The

step response of the LQR controller with zero vibration input shaping is shown in

Fig. 3.13.

3.2.4 Experimental Results on Manipulating Nano—Objects

In this section, nanomanipulation experiments using the adaptable end effector

were performed to verify the effectiveness of the controller and the efficiency of the

56

200

 

100
o.
-100

 

 

l
i
i
l
l

 

 

Tip displaciement
nm

200

0.0 0004 10.008 10.0112 10.0116 1 0.02

(a) Time (sec)

 

100
o.

Iacemenl

81m)

 

I I Y I I U
... _._ # J .4

 

 

Tip dis
8

0.0 10.0104 10.008 10.0112 0016 002

(b) Time (sec)

 

rim)
8
O

0
-100

 

A.-

 

 

Tip displacement

0.014

0.015

0.016 0.017 0.018

(c) Time (sec)

Figure 3.12. (a) Tracking response of a square wave with PD controller. (b) Tracking
response of a square wave with PD controller. (c) Zoom in around 0.016 sec in (b).

57

 

 

 

Tip displacement

l

 

 

-50 1 l 1
40.825 40.83 40.835 40.84 40.845 40.85

Time (sec)

Figure 3.13. Tracking response of a square wave with the LQR controller and input
shaping
system.

Fig. 3.14. (a) shows an AFM image of a silver nanowire with length of 1.9/1m
and diameter of 120nm. An adaptable end effector (spring constant k = 1-5 N / m)
was used to push this silver rod. Fig. 3.14. (b) shows the AF M image by a new
scan after pushing. It can be seen that the silver nanowire was pushed to the desired
position successfully. Fig. 3.14. (c) shows the control signal and the response of
tip. The AFM tip was driven to float on the top of the sample surface at t = 1986C.
After that, the controller started at t = 21886. The AFM tip also began to move
toward the nanowire and then hit the nanowire at t = 3088C. Then, at t = 37sec,
the silver nanowire was pushed to the desired position. Finally, the tip was lifted
up and the controller was turned off. From Fig. 3.14. (d), it can be seen that the
deflection signal is almost flat which means that the cantilever was controlled to keep
straight during manipulation. The two small peaks are due to turning on and off
the controller. It can also be observed from Fig. 3.14. (c) that there is substantial
increase of control signal after the tip touched the rod. Thus, the interaction force
can be easily reconstructed from the control signal. The force feeling from the haptic

device is much better than that using the traditional cantilever.

58

 

0
2.0 4.0 6.0

 

 

 

 

 

 

 

 

 

O
(a)
E 2
5 °' ' '" '
E '2' ‘
Te '4’ '
‘g -6 - _
0 '8 I J_ .1. 1 _L l
10 15 20 25 3O 35 40 45
E (C)
5 200 1 fii .7
E
E
m 100 -
0
L“
g- 0. WW 4
'O
D.
i: -100 I l l 1 | l
10 15 20 25 30 35 4O 45
(d) Time (sec)

Figure 3.14. Pushing a silver nanowire with an adaptable end effector. (a) The
AFM image of a silver nanowire with length of 1.5mm and diameter of 120nm. The
scanning range is 6pm x 6pm. (b) The AFM image by a new scan after pushing.
(c) The control voltage applied on the adaptable end effector. (d) Tip displacement
signal measured by the photodiode sensor.

59

To quantitively show that using adaptable end effector does improve the effi-
ciency of the nanomanipulations, a set of experiments were carried out and analyzed.
Manipulations using a soft probe (Model: OTR8-35, Veeco Instruments, k = 0.57
N /m) with/without preload, a rigid probe (Model: TESP, Veeco Instruments, k =
20-80N/m) with/without preload and an actively controlled adaptable end effector
(DMASP, Veeco Instruments, k = 1-5 N / m) are conducted for more than 10 times
respectively. Fig. 3.15 shows successful rates of manipulations using different probe
by counting the successful manipulations. It can be observed that the rigid probe has
higher successful rate than the soft probe when no preload applied. It consists with
the analysis in introduction. By applying a preload, both manipulations using the
soft probe and the rigid probe get higher successful rates. It is also shown that both
the manipulations using the adaptable end effector and the rigid probe (with preload)
reach the high successful rate of 90% . Therefore, the efficiency of nanomanipulation

is improved using an adaptable end effector.

To prove the high sensitivity of the adaptable end effector, another set of exper-
iments were performed and the force feedback signals are analyzed. Since the suc-
cessful rate of the manipulations using the soft probe without preload is too low, it is
meaningless to analyze the sensitivity of the soft probe without preload. Therefore,
soft tip without preload will not be discussed in the following discussion.

First, a probe is used to push a nanowire with length of ll. The change of the
signal (it can be the deflection of the traditional cantilever or the control signal of the
adaptable end effector) caused by the interaction force is denoted as A51. Then, the
same probe is used to push another longer nanowire with length of 12. The change of
the signal is denoted as AS2. Based on the assumption that nanowires with different

length will cause different manipulation force, the ratio of A82 over A31 should

60

Successful rate (%)

 

El — soft probe without preload E2 - soft probe with preload
E3 — rigid probe without preload E4 — rigid probe with preload
E5 — activelv controlled adaptable end effector

Figure 3.15. Successful rates of manipulations using different probe

represent the sensitivity of the probe. Hence, a metric S can be used to evaluate the

sensitivity.

A32

Two nanowires with length of 11 = 2pm and 12 = 4pm are pushed by different
probes for multiple times. Then for each manipulation, the value of S is calculated.
Except the adaptable end effector, the manipulations using other probes (soft probe
with preload, rigid probe with/ without preload) have the values of S around 1, which
means the cantilever is not sensitive to changes of the interaction force. At the same
time, A81 and A52 of these manipulations are also very small. This makes it difficult
to measure changes of the signal. For the manipulations using adaptable end effector,

much larger A31 and A52 are obtained from the control signal, and the value of S

61

is in the range of 1.2 — 1.8. Therefore, even manipulations using the adaptable end
effector and the rigid probe with preload have the same successful rate, the adaptable
end effector is prefered in the AFM based nanomanipulation because of its higher

force sensitivity.

3.3 Experimental Results on Fabricating Single

CNT based Infrared Sensors

The effectiveness and reliability of the AFM based nanomanipulation method using
adaptable end effector have also been proved by practical applications. This system
has been employed to manipulate the carbon nanotubes (CNT) onto some nano-
electrical devices to build nano—sensors. Fig. 3.16 shows the process of manipulating
a CNT onto a pair of gold electrodes to form a connection. Fig. 3.16. (a) shows
the AF M image before manipulation. There are two CNTS in the image, a longer
one and a shorter one. The longer one will be pushed onto the electrodes to form
the connection, and the shorter one will be pushed away. Fig. 3.16. (b) shows the
image from a new AFM scan after manipulation. Fig. 3.17 shows the process of
manipulating another CNT onto electrodes. With this system, the nanomanipulation
and nanoassembly is not difficult anymore. It becomes more deterministic and more

reliable. Fig. 3.18 shows another four assemble results of CNT based nano-sensors.

3.4 Chapter Summary

The fabrication and assembly process for building single CNT based nanodevices
is presented in this chapter. Firstly the fabrication process is addressed, where the
DEP deposition method is used to attract individual CNTS to the microelectrodes,

then the AFM manipulation method is employed to manipulate a single CNT to

62

     

O 2.0 4.0 6.0 8.0

 

(a)

Figure 3.16. Manipulating a CNT onto a pair of gold electrodes. (a) The AFM image
before manipulation. (b) The image from a new AFM scan after manipulations.

 

Figure 3.17. Manipulating a CNT onto a pair of gold electrodes. (a) The AFM image
before manipulation. (b) The image from a new AFM scan after manipulations.

63

 

Figure 3.18. Four different results of pushing CNT onto gold electrodes.

64

bridge the microelectrodes and clean the rest CNTS and particles away. Then the
key technology for assembling single ON T devices, the AFM based nanomanipulation
system enhanced by an adaptable end effector, is introduced and experiments for
verifying the manipulation system are discussed. Finally, experiments of fabricating
single CN T based infrared sensors are presented to show the efficiency of the developed

fabrication process.

65

CHAPTER 4

Testing of Single CNT based IR

Sensors

As a promising one dimensional nanomaterial, CN T has been extensively studied for
electronic and photonic applications. However, the reported results have shown a
big variance on the electronic properties of CNTS. It is necessary to have a reliable
testing procedure as well as testing environment for characterizing the CNT based
IR sensors. In this chapter, the development of a cryogenic IR sensor testing system
will first be presented. Then the experimental testing and characterization of CNT

IR sensors will be discussed.

4.1 Development of the Testing System

An IR sensor testing system is developed to test the sensor in different environments.
As shown in Fig. 4.1, the SWNT based IR sensor testing system includes four major
components: sensor chamber, lens, IR source and the signal measurement module.
The sensor chamber is electrically and thermally shielded to decrease the background
noise. The IR light from the IR source is focused onto the CNT sensor by the IR lens.

By using a NIR laser as IR source, the system can be used for NIR testing. The system

66

can also be used for MIR testing with a blackbody IR source. The photocurrent signal

is measured by the signal measurement module through the electrical feedthroughs.
IR source (”1"")

IR lens I”;
./ CNT

 

 

 

 

 

 

 

(a) (b)

Figure 4.1. (a) The SWNT IR detection system. (b) The structure of the CNT chip.

To test the MIR response of the CNT IR sensors, a cryogenic MIR testing system
has been developed as shown in Fig. 4.2. The MIR source is an Oriel 67030 blackbody
(Newport Corp.) with a temperature range of 50°C to 105000. Filters with different
spectrum are used to get IR light with different wavelengthes. The lens focuses IR
onto the CNT sensor, which is mounted inside a cryogenic cooling chamber. The
sensor temperature can vary from room temperature to 496°C. With the cryogenic
cooling system, the CNT sensor can be characterized at different temperature as well

as different gas environment.

67

Blackbody IR window

    
   
   

Cryogenic Cooling system

 

Filters Lens CNT Chip], Chip Holder

Figure 4.2. The cryogenic MIR testing system.

4.2 Individual Multi-Walled Carbon Nanotube

based IR Sensors

4.2.1 Bandgap Engineering of MWNTS

Since is difficult to guarantee the properties of each individual CNT from the synthetic
point of view. Therefore, the simplest way is to modify them using the electrical
shell breakdown process [113] after CNT based devices are fabricated by using AF M
nanomanipulation system. By applying a bias voltage on the MWNT and controlling
the current, the outer layers of the MWNT can be broken at certain current. Fig.
4.3 shows the first two measurements of the I—V curve of a MWNT device. It is
found that the currents suddenly dropped at a certain threshold voltage and increased
again. This was caused by the burning of the MWNT walls which contributed to the
conduction. Fig. 4.4 shows the I-V curve of the device after three breakdowns. It can

be seen that the metallic MWNT became semiconductive one after two breakdowns.

68

1 .2E-06

 

 

 

 

   

    
 
    

 

 

+ lsttim—efi
1.0E-06 +2nd time
Breakdown
g 8.0E'07 * pfints
E, 6 05-07 »
‘5
0 4.05-07
2.0E-07
0 OE+OO l ‘ ‘ ‘
0 5 1O 15 20 25
Voltage (V)

Figure 4.3. I-V curves of a single MWNT device.

 

 

 

 

 

 

2.0E-09
4— 4th timel I
1.6E-09 ~ +/
E 1.2E-09 ~
E
9
5 8.0E-10 ~ /
4.0E-10 ~ ‘
a
W“
0.0E+OO ¢¢::¢¢¢¢¢¢¢;¢¢;CCCM 1 L
0 5 1o 15 20 25
Voltage (V)

Figure 4.4. I-V curve of the single MW N T device after three breakdowns.

69

4.2.2 Experimental Testing of MWNT Sensors

The AF M image of the MWNT IR sensor is shown in Fig. 4.5. The diameter of the
MWNT on the chip is around 65 nm. Due to the large diameter, the MWNT have
a very small bandgap. Hence the MWNT-metal contacts are quasi-Ohmic contacts
with small barriers. The I-V curve of this device is shown in Fig. 4.6. Temporal
photocurrent response was tested with a N IR laser source, which has a wavelength of
830 nm and power of 30 mW. Fig. 4.7 shows the temporal IR response with a bias
voltage of —0.1 V while the IR source was switched on/ off periodically. The current
decreased as the IR source was switched on. Fig. 4.8 shows the temporal IR response
with a bias voltage of 0.1 V. The current increased as the IR source was switched on.
Fig. 4.9 shows the plot of the photocurrent (the current change AI when IR is on) as a
function of the bias voltage. It can be seen that the MWNT works as a photosensitive
resistor. This is caused by the photoinduced carrier concentration change in the CNT.
When the IR irradiates the CNT, photons will excite electrons and holes inside the
CNT and raise the carrier concentration. It can also be seen that the photoinduced
current change (1.4 uA at Vbias = 0.1 V) is quite big compared with the reported
photocurrent of SWNTs (0.25 nA at Vbias = —3 V [13]). Since a MWNT has bigger
diameter and multiple walls, it can absorb photons more effectively than SWNTs.
Hence MWNT sensors are expected to have a higher quantum efficiency than SWNT
sensors. The quantum efficiency of this device is calculated as N 30%, which is much
higher than the reported values (< 10%) of SWNT devices.

Individual MWNT based IR sensors were fabricated and tested in research. The
electrical shell breakdown technique was used to modify the electrical properties of the
MWNT after CNT devices were fabricated. Experiments were carried out to exam the
electrical properties and photoresponse of the single MWNT based IR sensors. The
results show that the MWNT based IR sensor has higher photocurrent and quantum

efficiency than the SW NT sensors. However, the detectivity of the MWNT IR sensor

70

 

 

0.0 1: Height 11.5 um

Figure 4.5. AFM image of the MWNT IR sensor.

 

 

 

 

 

 

8.0E-04
4.0E-04 -
E
E 0.0E+00 . r
S
0
-4.0E-04 «
-8.0E-04
-1 -0.5 O 0.5 1
Bias Voltage (V)

Figure 4.6. I-V curve of the MW NT IR sensor.

71

IR light ON OFF

 

 

 

 

 

 

 

 

 

 

 

 

3.321305 . .
3.363054% n m m fl
53.40305 —

25.

5 344B 05 ~
0" '

-3.48E-05 ~ Li M U U l:-
-3.52E-05

 

0 5 10 15 20 25
Time (s)

Figure 4.7. Temporal photoresponse of the MWNT IR sensor. The bias voltage across
the electrodes is —0.1 V.

IR light OFF ON

 

 

 

 

 

 

 

 

 

 

 

 

 

3.60E-05 U U
35213—05 fl

‘i n n n

E 344305 —

Ll
3.36E-05a U ‘hal u b
3.28E-05 T a . r .

o 5 10 15 20 25
Time (s)

Figure 4.8. Temporal photoresponse of the l\r'IWNT IR sensor. The bias voltage across
the electrodes is 0.1 V.

72

4.0E-06

 

ZOE-06

 

0.0E+OO

 

-2.0E-06

Photocurrent A I (A)

 

 

 

 

-4.0E-06
-0.4 -0.2 0 0.2 0.4
Bias voltage (V)

Figure 4.9. The plot of photocurrent vs. bias voltage.

is still smaller than traditional IR sensors due to the high noise and short absorption

depth caused by its small diameter.

4.3 Individual Single-Walled Carbon Nanotube
based IR Sensors with Symmetric Structure

The SWNTS used in our experiments were obtained from BuckyUSA, Inc. It takes
several steps to deposit an individual CNT onto the substrate. Firstly, the powdery
CNTS are put into acetone and ultrasonicated for 10-15 min to form CNT suspension.
Then a droplet of the CNT suspension is dropped onto the substrate between the two
microelectrodes. At the same time, a sine wave AC signal with frequency of 10 kHz
and peak to peak amplitude of 1.5 V is applied between the two microelectrodes for
3-5 seconds to generate the dielectrophoretic force to trap CNTS to the microelec-
trodes. During the deposition process, the concentration of the CNT suspension,
the frequency and the amplitude of the AC signal are experimentally optimized to
achieve optimal possibility to trap an individual CNT between two microelectrodes.
Even with those parameters experimentally optimized, it can not be guaranteed that

only one CNT is trapped between the two microelectrodes and bridges them success-

73

fully. Normally, there are some impurities or more than one CNT trapped between
the microelectrodes. Furthermore, if only one CNT is successfully trapped between
the microelectrodes, it might not bridge the microelectrodes very well. Therefore, it
is necessary to take another step to clean up the microelectrodes gap area and adjust
the position of the CNT to make the connection. This final step is very critical and is
termed CNT assembly. The AF M based nanomanipulation system [78] is employed

to perform the CNT assembly.

 

 

    

 

 

 

 

   

 

50.0 mV
, . E-O9
A lE-09
31
i=3
5
U
-lE-09
_ my". - , _ -3.0E-O9
' _ . ' -l.2 0.6 O 0.6 1.2
0.0 3. Amplitude 3.8 Voltage (V)
(a) (b)

Figure 4.10. (a) AFM image of an individual SWNT based photodiode with symmet-
ric Au electrodes. (b) I—V curve of the CNT photodiode.

We first fabricated individual SWNT based Schottky photodiodes with symmetric
gold (Au) contacts. The AFM image of a SWNT device is shown in Fig. 4.10(a).
The diameter of the SWNT on the chip is 1.7 nm. As shown in Fig. 4.10(b), its I-V
characteristic curve is symmetric due to the symmetry of the CNT—metal contacts.
Although Au has a. high work function of 5.2 eV, due to the work function reduction at
Au surface caused by the contaminants [114] and the low interaction energy between
CNTs and Au [115], the CNT—Au contact is a Schottky contact with a small barrier

rather than an Ohmic contact. Since the carriers can easily tunnel through the small

74

barriers, the current increases quickly even at a very low reverse bias voltage. Hence

the I—V curve in Fig. 4.10(b) did not show a very low reverse current.

IR OFF ON

    
  

 

 

 

 

 

 

 

 

A—lE-IO- A
g 5 213—10-
‘E-2E-10- E
E 2 113—10-
0 5

-3E-10- 01300

H U'l—l U'U U
‘4E'l'v‘ -lE-10
0 5 10 15 20 25 o 10 20 30
Time (s) Time (s)
(a) (b)

Figure 4.11. Temporal photocurrent response of the SWNT photodiode with a zero
bias voltage. (a) IR irradiates the left electrode. (b) IR irradiates the right electrode.

We tested the photovoltaic effect of the CNT based Schottky photodiode with a
zero bias voltage applied across the CNT. The current signal was measured with a
Keithley 6487 Picoammeter. The IR source was a near-IR laser (UH5—30G—830-PV,
World Star Tech.) with a wavelength of 830nm and a power of 30mW. A zero bias
voltage was applied on the CNT photodiode and the current was monitored while the
IR source was periodically switched on and off. Fig. 4.11(a) shows the temporal IR
response when the IR spot center was on the left electrode. The current decreased
when the IR source was on. Fig. 4.11(b) shows the temporal IR response when the
IR spot center was on the right electrode. The current increased when the IR source
was on. The negative dark current was caused by the output error of the voltage
source.

As shown in the top image of Fig. 4.12, the IR spot was moved from left to right

75

following the microelectrodes. The plot of the photocurrent (the current increment
when the IR is on) as a function of the IR spot center position is shown in Fig. 4.12. It
can be seen that the photocurrent changed from negative to positive continuously as
the IR spot moved from left to right. Since the device was zero biased, the photocur-
rent must be generated by the photovoltaic effect. If the photocurrent was caused
by the conductance change due to heating effect or photoconduction effect, it would
not be able to change from positive to negative when the IR spot moved from left to
right. The reason photocurrent changed from positive to negative was because the
CNT formed Schottky diodes at both electrodes and they were connected reversely.
When the IR spot center was on the left electrode, the diode at this electrode domi-
nated and the photocurrent was negative. When the IR spot center moved to another
electrode, the diode at that electrode dominated and the photocurrent became posi-
tive. Since the IR energy was normally distributed, the photocurrent was maximized
at the point where the difference between the light intensity at these two CNT-Au
interfaces was maximum rather than at the point where the spot center was at one of
the CNT-Au interface. Hence the distance between the two photocurrent peaks was
30 pm, though the distance between the two microelectrodes was only 2 pm. Since it
is impossible to only focus light onto one microelectrode during real applications, the
intensity of the incident light at these two CNT—electrode interfaces will be very close.
Hence the photocurrents generated at the two interfaces will cancel each other and
the measurable signal will be very weak, which limits the efficiency and sensitivity
of the symmetric CNT based photodiode. Furthermore, since there are two Schottky
diodes, the electrons and holes generated at each diode had to tunnel through another

Schottky barrier to form a current. This also limited the efficiency and sensitivity.

76

 

 

 

 

 

 

 

 

1
I f 1' ' «3"
"‘_________.__..___1—“ ““"'"‘ 7" ,' m“?
.-,
45-10 ~—
213-10
3
E o
E 0‘
U
8 0
O
J:
an
-2E-10
413-10
-400 -200 0 200 400
IR spot position (mm)

Figure 4.12. The plot of the photocurrent as a function of the IR spot position.

4.4 Individual Single-Walled Carbon Nanotube
based IR Sensors with Asymmetric Structure

To improve the performance of the SWNT photodiode, we designed an asymmetric
structure which used different materials as the contact electrodes to create Schottky
barrier at one electrode and ohmic contact at another electrode. Consequently the
photocurrent generated at the Schottky barrier will not be canceled, and it does not
need to tunnel through another Schottky barrier. For example, palladium (Pd) has
a work function as high as 5.1 eV and was reported to be the best metal to form an
Ohmic contact with CNT [41,99, 104], whereas titanium (Ti) has a lower work function
of 4.3 eV and forms Schottky barrier with CNT [41]. When Ti and Pd are used as the

contact materials, a Schottky contact is formed at the CNT-Ti interface and an Ohmic

77

contact is formed at the CNT—Ti interface. In this way. the photocurrent generated
at the CNT—Pd interface is zero and the total photocurrent is maximized to equal
the photocurrent generated at the CNT—Ti contact. Hence, with the asymmetric
contacts, the efficiency and sensitivity of individual SW NT based photodiodes can
further be improved. To compare with the experimental results of the symmetric Au-
Au structure, we used Ti and Au to make an asymmetric photodiode. Since the work
function of Au is quite high, the Schottky barrier at the CNT-Au interface is very
small and the photocurrent generated at the CNT-Au interface will not neutralize as

much the photocurrent generated at the CNT-Ti interface.

 

 

 

 

 

 

 

 

200
1.2E—9
6.0E—10
s“,
Eons-00 r—
:1
U
—6.0E—10
-1.2E-9
, . _ 1 -0.6 0 0.6
0.0 1. height 11f: Voltage (V)
(a) (b)

Figure 4.13. (a) AF M image of an individual CNT photodiode with asymmetric
Ti-Au electrodes. (b) I-V curve of the asymmetric CNT photodiode.

Fig. 4.13(a) shows a CNT photodiode with asymmetric Ti-Au electrodes. As
shown in Fig. 4.13(b), its I-V curve illustrates a Schottky diode behavior. Since the
Schottky barrier became higher as well as thinner with a reverse bias voltage and a low
reverse bias could make it thin enough for carriers to tunnel through, the saturation

current increased dramatically at a low reverse bias voltage. Fig. 4.1~I(a) shows the

78

plot of its photocurrent as a function of the IR spot position. It can be observed that
the photocurrent was only —2 x 10_12A when the IR spot was on CNT-Au interface.
and the photocurrent was 1.3 x 10_10A when the IR spot was on CNT-Ti interface.
There are two reasons for the asymmetry of the photocurrent. Firstly, since the build-
in potential at the CNT-Ti interface is stronger than the one at the CNT-Au interface,
there will be more electron-hole pairs generated at the CNT-Ti interface. Secondly,
since the barrier height at the CNT—Ti interface is higher than the one at the CNT—
Au interface, the electrons and holes generated at the CNT—Au interface are more
difficult to tunnel through the barrier at the CNT-Ti interface to form photocurrent.
But the electrons and holes generated at the CNT-Ti interface can tunnel through
the CNT—Au barrier relatively easier. Hence, a higher photocurrent will be generated
at the CNT-Ti interface when both microelectrodes are evenly illuminated. In this
way, the measurable signal was much stronger than the symmetric CNT photodiode.

Hence the efficiency and sensitivity were also improved.

 

 

 

 

 

 

 

 

 

 

 

A 1.3E-10
§ 1.0E-10
5 9.0E-ll :5
5 r _
g 5.01;“ E 5.0E-ll
3 5
°‘ 1.0E-11 , ODE-00 '— .—— _. _. __

-3.0E- l l amid ,,,,n-.,____ m ____,7_ —5.0E-l l

-600 -300 O 300 600 0 10 20 30
IR spot position (um) Time (s)
(a) (b)

Figure 4.14. (a) The photocurrent as a function of the IR spot position. (1)) Temporal
photocurrent response of the photodiode with asymmetric electrodes. The device was
zero biased. The IR power was 30 mW.

79

Fig. 4.14(b) shows the temporal photocurrent response of the asymn'ietric CNT
photodiode with a zero bias voltage when the IR spot was on Ti electrode. The
current increased from 1.2 x 10_13A to 1.3 x 10‘10A when the IR was switched on.
For the symmetric photodiode as shown in Fig. 4.11(b), the current increased from
—3.4 x 10"11/1 to 2.8 x 10_10A when the IR was switched on. we can see that the
symmetric CNT photodiode has a. higher photocurrent than the asymmetric one. This
is due to the variety in the properties of the CNTS for these two devices. Normally,
different CNTS from the same production batch may have a variation of 1 to 2 orders
in photocurrent. Although this symmetric device has a higher photocurrent, it also
has a higher dark current due to low barrier height at CNT-Au interface. Its signal to
dark current ratio (SDR) is only 9.2, which makes it difficult to detect low intensity
IR light. We also tried to fabricate symmetric CNT photodiode with Ti electrodes.
Larger barrier height at CNT-Ti interface leads to a much smaller dark current. Since
there are two CNT-T i Schottky barriers, the large barrier height also makes it difficult
for electrons and holes generated at one barrier to tunnel through another barrier,
which causes a very small or even zero photocurrent. Since an asymmetric CNT
photodiode has a low dark current and a big photocurrent at the same time, it has
much higher SDR than a symmetric CNT photodiode. The asymmetric photodiode
presented here has a SDR of 1.08 x 103, which is two orders of magnitude higher
than the SDR of the symmetric device. Based on our testings of more than fifty
similar devices, normally symmetric CNT photodiodes have a SDR in the range of
0.1 - 10, whereas the asymmetric CNT photodiodes have a SDR over 1000. As shown
in Fig. 4.15, we also measured the I-V characteristic curves of the asymmetric CNT
photodiode. The I-V curve shifted downwards as the IR power increased, which is a

typical behavior of the traditional photodiode.

80

 

2.0E-10

   

 

 

 

 

A 0.0E+00
it,
E
d)
g —~— P3 = 30 mW
U -2.0E-10 ”:20 mW
P1 :7 mW
+ P0 = o w
-4.0E-10
—0.3 —0.1 0.1 0.3
Voltage (V)

Figure 4.15. I-V characteristic curves of the asymmetric CNT photodiode. PO-P3
represent different IR powers.

4.5 Sensing Middle IR Using CNT Based IR Sen-
sors

A MIR sensing material has to have a bandgap smaller than the energy of MIR
photons. But the small bandgap causes a large thermal noise and the traditional
MIR sensing system has to work at ultra-low temperature to decrease the thermal
noise. Hence a cryogenic cooling system is normally required. However, cryogenic
cooling system severely limited the applications of MIR sensing systems and hindered
their advancement. As a. 1-D material, CNT has some unique properties such as wide
range of band gaps, and reduced carrier scattering, etc. As a result, CNT based IR
sensors have the potential to work at room temperature or moderate low temperature
with a high quantum efficiency, which is difficult for other MIR sensing materials to
implement. Therefore the MIR response of the CNT based IR sensors is investigated
in this section. However. since SWNT normally has a thin diameter, its bandgap

is too large to detect MIR. Fortunately, MWNTS may have much smaller bandgap

81

than SWNTS because of their large diameter. Hence it. is possible to detect MIR
using MWNTS. Moreover, as discussed in section 4.2.1, the advantage of bandgap

tunability makes MWNT a perfect material for building MIR sensors.

 

 

 

 

 

 

 

 

4.05-05
2.0E-05 /

g
E 0.0E+OO 4 .
t:
:3
U

4.01305 ,

4.01-3-05

-1 .05 0 0.5 1
Voltage (V)

Figure 4.16. I-V characteristic curve of a MWNT MIR sensor

The I—V characteristic curve of a MWNT device is shown in Fig. 4.16. It can
be seen that the I-V curve shows a metallic behavior. This MWNT device has been
tested with both NIR and MIR. However, none of them could be detected. The
reason is that the MWNT was too thick and its bandgap was almost zero. Hence
the MWNT was a metallic CNT which made it insensitive to IR light. To make the
MWNT device sensitive to MIR, its bandgap had to be adjusted such that the MWNT
became a semiconducting CNT. By using the technology developed in section 4.2.1,
the outer layers of the MWNT was electrically removed and the MWNT became a
semiconducting CNT. Fig. 4.17 shows the I-V curve of the MWNT device after first
breakdown. It can be observed that the I-V curve became curving. That was because

that the semiconducting MWNT formed barriers at. the MWNT-metal interfaces.

82

l .OE-O6

 

5.0E-07 . -——-—---—- - b, - - -,,,,,,,

0.0E+00 1 *

 

Current (A)

 

5.05-07 / W ~

 

 

 

 

-l.OE-O6
-l —O.5 0 0.5 1

Voltage (V)

Figure 4.17. I-V characteristic curve of the MW NT MIR sensor after first breakdown

After the MWN T was tuned to a semiconducting MWNT, the NIR photoresponse
was first tested. Fig. 4.18 shows the temporal photoresponse of the MWNT sensor
when a N IR laser was periodically switched on and off. It can be observed that pho~
tocurrent responsed to the IR illumination very quickly. During the measurement,
a zero bias voltage was applied on the MWNT pixel. It means that the detected
photocurrent was a short circuit current. Hence the MWN T should work as a photo-
voltaic device here. After taking measurements with the NIR, the MIR has also been
tested by using a blackbody as the MIR source. An IR filter with a spectral range
around 3 pm was used to block IR light with other wavelengthes. However, there
was no any signal detected when the MIR was switched on and off. Since the power
density of the blackbody emission was much lower than the NIR laser, it was possible
that the MIR signal had been submerged by noise.

To lower the noise and decrease dark current, the bandgap engineering process
was carried out again to adjust the MWNT bandgap. Fig. 4.19 shows the I—V

curve of the MW NT sensor after second breakdown. Obviously, the amplitude of the

83

NIR OFF ON

 

 

 

 

 

 

6.0E-O9

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4.0E-09 WW
3 M
E 2.0E-09
:3
U

0.0E+00 W 1+ 1 ‘ '

-2.0E-09

O 5 10 15 20 25

Time (See)

Figure 4.18. Temporal photoresponse of the MWNT sensor to NIR

current became much smaller and the I-V curve behaved like a diode. The reason is
that the bandgap of the MWNT became larger after the second breakdown process.
Hence the Schottky barriers at MWNT-metal interfaces also increased, which caused
smaller noise level and dark current. After lowering the noise level and dark current,
the MWNT was used to detect MIR again. Fig. 4.20 shows the temporal response
of the MWNT sensor when the MIR was periodically switched on and off by using
a mechanical chopper. It can be observed that the MIR response was quite different
with the NIR response. When then MIR was turned on/off, the current suddenly
increased/decreased and then gradually got back to the dark current level. This
is because the MWNT formed two reversely connected Schottky barrier at MWNT-
metal contacts. When the MWNT device was illuminated by IR, both Schottky diode
generated electrons and holes. However, since there were two reversely connected
Schottky diode, the photogenerated charges at one Schottky diode had to go through
another one to form a photocurrent. Since the incident power density of the NIR

was big enough to generate sufficient charges at each Schottky barrier and form a

84

 

large open circuit voltage at each Schottky diode, the photogenerated charges at one
diode were able to tunnel through another diode and formed a photocurrent. That
is why the photocurrent could keep its value after the NIR was switched on. But
for the MIR, the incident power density was much smaller and could not generate
enough charges at the Schottky diode to form a big open circuit voltage. Hence the
photogenerated charges at one diode could not get through another diode to form a
photocurrent. And the two Schottky barriers worked like two capacitors. When the
MIR was switched on, the capacitors got charged and there was a suddenly current
increase. But as time going, the charging current gradually reduced to zero. When
the MIR was switched off, the capacitors discharged and there was a suddenly current
decrease. And the discharging current gradually reduced to zero.

6.0E-12

 

4.0E-12 ~ —A —~— ——— “w ._- -fi m_- __

 

2.0E-12

Current (A)

0.0E+OO R g

 

 

 

 

-2.0E-12

 

-2 -l .5 -1 -O.5 0 0.5 1 1.5 2
Voltage (V)

Figure 4.19. I—V characteristic curve of the MW NT MIR sensor after second break-
down

85

MIR ON OFF

 

 

 

 

 

 

 

 

 

 

 

2.0E-12

1.0E-12 ~

0.0E+OO -

Current (A)

-l.OE-12 4

 

 

-2.0E-12 ,

 

Time (see)

Figure 4.20. Temporal photoresponse of the MW NT sensor to MIR

4.6 Temperature Dependence of CNT Based IR
Sensors

As a l—D material, CNT has some unique properties such as wide range of band
gaps, and reduced carrier scattering, etc. As a result, CNT based IR sensors have
the potential to work at room temperature or moderate low temperature with a high
quantum efficiency, which is difficult for other MIR sensing materials to implement.
Hence it is important to characterize the CNT sensors at different temperatures.
Fig. 4.21 shows the I-V characteristic curves of a MWNT sensor at different
temperatures. It can be observed that the dark current significantly dropped when
the temperature was lowered to 200 K. Since this device was assembled by pushing
the MWNT onto microelectrodes, the pushing process might induced defects on the
MWNT and the contact condition might be bad as well. Hence the significant cur-

rent drop upon temperature decrease might be because that the defects and contact

86

 

 

 

 

 

 

 

 

 

 

 

—T = 300 K

1.5E-11 ~ “T=25°K f
E
E, 1.0E-11
5
0

5.0E-12

0.0E+00

O 1 2 3 4
Voltage (V)

Figure 4.21. I-V characteristic curves of a MWNT sensor at different temperatures.
The sensor was assembled through nanomanipulation.

condition became worse when the temperature was decreased. To verify this analy-
sis, another MWNT sensor was assembled without nanomanipulation. The MWNT
was directly deposited onto the microelectrodes through DEP deposition. Since no
nanomanipulation was involved during the device fabrication, the defects on CNT
should be minimized and the contact condition should be improved. The I-V curves
of this MWNT sensor was also tested at different temperatures as shown in Fig. 4.22.
It can be seen that the I-V curve did not change much when the temperature was
lowered from 300 K to 180 K. Hence how to reduce the defects and improve the
contact condition is a important issue for the CNT based nanodevices. A CNT IR
sensor packaging process has been developed in [116]. After assembling a CNT onto
the microelectrodes, the CNT device will be packaged for keeping good contact be-

tween CNT and microelectrodes. First of all, the CNT chip will be annealed in a

87

vacuum oven at 300 CC to improve the contact between CNT and microelectrodes
right after the fabrication. The defects induced during nanomanipulation can also be
reduced through the annealing process. After that, a layer of parylene will be coated

on the top of CNT to fix it and isolate it from environmental uncertainties.

 

 

 

 

 

 

 

 

 

1.0505
6.0E-06 *
2 2.0E-06 ‘
i=2
t
:3
o -2.0E—06 T T ' SET
__.. T = 180 K
_._ T = 26 K
some “ O
T = 300 K
-1.0E-05
-1 es 0 0.5 1

Voltage (V)

Figure 4.22. I-V characteristic curves of a MWNT sensor at different temperatures.
The sensor was assembled without nanomanipulation.

4.7 Performance Evaluation and Analysis

IR response of the CNT based IR sensors have been experimentally tested in previ-
ous sections. Fig. 4.11 shows the temporal photoresponse of the symmetric CNT IR
sensor when IR spot was on different locations. Fig. 4.12 shows the photocurrent of

the symmetric CNT IR sensor as a function of the IR spot position. It can be ob-

88

served that the photocurrent changed from negative to positive continuously as the
IR spot moved from left to right. Since the device was zero biased, the photocurrent
must be generated by the photovoltaic effect. If the photocurrent was caused by the
conductance change due to heating effect or photoconduction effect, it would not be
able to change from positive to negative when the IR spot moved from left to right.
Fig. 4.15 further proved this analysis with the perfect photodiode I-V characteristic
curves. Based on this analysis, following sections evaluate some important proper—
ties of CNT based IR sensors, including response time, photocurrent, dark current,

quantum efficiency and etc.

Response Time

To test the response time, the temporal photocurrent response of a SWNT device
was measured with a sampling rate of 10 kHz, as shown in Fig. 4.23. Since it takes
time to do integral at low current measurement range, the sampling rate will be very
low in this situation. To reach the high sampling rate in this experiment, the current
was measured by the Agilent 4156C with a current measurement range of 100 nA.
But because the signal is at nano-ampere scale, the noise level became high in this
result. From the plot it can be seen that the current increased immediately once the
IR was switch on. The response time was at 100 as level. Since the sampling time

was 100 us, it is possible that the response time was less than 100 as.

Photocurrent and Dark Current

Since the CNT-metal contacts of the CNT based IR sensor are Schottky contacts,
the photocurrent of the CNT sensor is calculated with a Schottky diode model by the

following equation

89

3.00E-09

 

 

 

 

 

 

2.00E-09 -T
E
E
g 1.00E—09 ~
3
U

0.00E+00 ‘ ' ll L

-l.00E-09

0 10 20 30 40 50

Time (ms)

Figure 4.23. Temporal NIR photocurrent response of an individual SWNT IR sensor
measured at high sampling rate

Jph = (1 — R)qugn (4-1)

where R is the energy loss of photons due to reflection and absorption, q is the
charge of electron, th is the photon flux density, and n is the quantum efficiency
of the Schottky diode. The spatial optical power distribution of the laser (PV-UHS-
BOG-830, World Star Tech.) is a Gaussian distribution. Hence, by assuming the laser
spot center to be at one electrode, the optical power at the two electrodes can be
calculated by Gaussian distribution equation. Let the total power be P, the power

P1 at the electrode where the spot center is located at is given by:

P1 = 1/\/27TP = 0.3989P (4.2)

90

Assume that the laser spot size is 60 um and the optical power at the edge is
around 1 — 2% of the total power. Let :1: = 0 at the spot center, the position of the
spot edge (:1: = 30 am) can be normalized as :1: = 2.5. Since the distance between the
two electrodes is 3 mm, x is normalized as :rb = 0.25 at the second electrode. Hence

the optical power P2 at the second electrode is calculated as:

P2 = 1 V27resrp —a:2 2 P = 0.3867P (4.3)
b

With equation 4.1, the photocurrent of a CNT sensor with symmetric gold elec-

trodes (annotated by Au-Au) is calculated by:

Jpthu - AU) = (1 - R)Q(P177Au — P277Au)

(l — R)qP(O.01277) (4.4)

For a CNT sensor with asymmetric gold-titanium electrodes (annotated by Au-

Ti), the photocurrent is:

JP}Z(Ti — AU) 2 (1 _ R)q(P177Tl — P277Au)

= (1 — R)qP(O.398977T21 — 0.386777A,,)(4.5)

91

Where the quantum efficiency is given by

erp(—aWd)
Z 1 _ 4.6
77 l + aLp ( )

 

where a is the optical absorption coefficient, Lp is the diffusion length of holes,
W of is the depletion width of the Schottky barrier at CNT-metal interface. Here we

suppose "Tz' z 100 — 200%nAu, then

Jph(Ti _ All.) (1 _ R)q(Pl77Tz — P277Au)

= (l — R)qP77A,,(0.0122 ~ 0.4111) (4.7)

Hence the ratio of the photocurrent of asymmetric CNT sensors to the photocur—

rent of symmetric CNT sensors is

Jph(Ti — Au) _ (1 - R)anA,,(0.0122 ~ 0.4111) _ 1 33
Jph(Au — Au) — (l — R)qP77AuO.0122 —
(4.8)

 

 

Our experimental results show a value of Jph(Tz' — Au) \ Jph(Au — Au) around
10, which is in the range of the theoretically estimated values. The dark current of a

Schottky diode is calculated by

JS 2 A**T2€—q¢b/kT (4.9)

92

where A** is the effective Richardson’s constant, ob is the zero-bias barrier height,
and n. is the ideality factor. For a p—type CNT, the barrier height cub can be calculated

by

be : X + Eg _ (15771 (4.10)

Where X and E9 are the electron affinity and bandgap of ON T, respectively. gbm

is the work function of contact metal. Substituting equation 4.10 into 4.9, we can get

.1, = A**7112e‘1(¢m”HE-(1)”?T (4.11)

Equation 4.11 shows that a contact metal with smaller work function will leads
to a smaller dark current. It explains why the asymmetric Ti-Au CNT sensors have

smaller dark current than the symmetric Au-Au sensors.

Quantum Efficiency

The quantum efficiency is defined as

QE 2 3715 (4.12)

where n E is the number of collected electrons, and up is the number of absorbed

photons. The number of collected electrons can be calculated by

93

I
TIE = 131}- (4.13)
q

where Iph is the photocurrent and q is the electron charge. Using the photocurrent

measured in section 4.4, n E can be calculated as

n _ Iph _ 1.3 X 10-10
E _ q _ 1.6022 x 10-19

 

: 8.114 x 1085—1 (4.14)

The number of photons absorbed by CNT is given by

_ P/A x A,

E photon

 

up (4.15)

where P = 30mW is the total power of the IR source, A = 7r(100um)2 is the area
of the IR spot, A3 = (length of CNT) >< (diameter of CNT) is the sensing area of

the CNT sensor, Ephoton is the photon energy and is given by

Ephoton : h X g“ (4.16)

where h = 6.626 x 10—34J - S, is Plank’s constant. C = 3 x 108m - .S'_1 is the
light velocity. = 830nm is the wavelength of the IR light. Hence, the number of

absorbed photon is

94

30mW/[7T >< (lOOum)2] x (1.5mm X 2.53pm)

 

”1’ 2 6.626 x 10-34J . S x [3 x 108m - S‘1I/830nm

= 1.495 x 10105—1

Then the quantum efficiency can be calculated as

_ nE _ 8.114 X10815"1

 

QE—

This value is consistent with the previous reported results [13,117,118]. Table
4.1 shows the quantum efficiency of different IR materials. It can be seen that the
quantum efficiency of SWNT is still lower than other well developed IR materials.
The low quantum efficiency of the SWNT is due to its thin diameter. Most of the
photons which reach the SWNT surface will pass through the SW NT wall without

exciting the electron-hole pairs. Hence, further study on the structure of CNT based

n, ‘“ 1.495 x 10105--1

IR sensors is expected for improving the quantum efficiency.

= 5.43%

(4.17)

Table 4.1. Quantum efficiency of different IR materials

(4.18)

 

W'avelength (um)

Quantum efficiency

 

 

 

 

 

 

 

Quantum Well Infrared Photodetector 3-20 0.32
Quantum Dot Infrared Photodetector 4.7-5.2 0.0002
InSb 3—5 0.3-0.8

SW NT 2.2 0.054

 

 

95

hum

 

Fxfl.’.’b ‘-

4.8 Chapter Summary

The development of a cryogenic IR sensor testing system and testings of different
single CNT based IR sensors are discussed in this chapter. Both MWNT sensors
and SWNT sensors have been tested and evaluated. MWNT sensors showed a higher
quantum efficiency and bigger photocurrent than SWNT sensors. But the dark cur-
rent of MWN T sensors was also much higher than the dark current of SWNT sensors.
Hence the SWNT sensors were studied to get a higher signal-to—dark current ratio.
SW' NT sensors have been fabricated with symmetric structure and asymmetric struc-
ture respectively. Experimental testing results show that the asymmetric CNT-metal
contacts can improve the performance of the photodiodes by increasing the SDR up to
two orders of magnitude higher than the symmetric ones. Hence, SWNT has strong
potential for applications of NIR detection. With its bandgap tunability, MWNTS
have been studied for MIR detection. By electrically removing the outer layers of a
MWNT, the bandgap of the MWNT was adjusted and the IR response was tested.

Detection of MIR using MWNT IR sensor has been demonstrated in this chapter.

96

CHAPTER 5

Multi-Pixel CNT Sensor Array

Different technologies have been developed for IR detection, such as uncooled mi-
crobolometer [119], quantum dot photodetector [120], etc. IR focal plane arrays
(IRF PAS) have also been fabricated based on these technologies [121, 122]. How-
ever, the low resolution is still a major difficulty for IRFPAs. Due to the fabrication
difficulties and the thermal crosstalk between adjacent pixels, the pixel pitch of the
traditional IRF PAS is limited to be around several tens of micrometers [122,123].
Hence, it is necessary to explore the possibility of reaching smaller pixel pitch using
new nanotechnology and nanomaterials such as CNT. In this study, a CNT based
three-pixel IR sensor with 10 um pixel pitch will be fabricated and tested. With the
technologies to fabricate multi-pixel CNT sensor array, CNT based multi-color IR
sensor will also be fabricated by assembling CNTS with different diameter onto the

adjacent pixels.

97

5.1 Fabrication of Single CNT based IR Sensor
Array

Though much work has been done on the photoconductivity of CNTS, few researchers
successfully fabricated CNT based multi—pixel sensors, which is important for CNT’s
further application in IR sensing. As discussed in Chapter 1, current available meth-
ods have their shortcomings in terms of repeatability, mass production, and ability
in eliminating uncertainties. It is difficult to fabricate CNT based sensor array in a
reliable and controllable manner with those methods. In this study, an automated
nanomanufacturing process is developed for building single CNT based multi-pixel
IR sensors.

Fig. 5.1 shows the developed automated name—manufacturing process for single
CNT sensor arrays. Firstly, the microelectrodes with a submicrometer gap are fab-
ricated with the photolithography method. Then the automated DEP deposition
system is used to deposit individual CNTS around each pixel. After trapping one or
more individual CNTS around each pixel, the AFM based nanorobotic manipulation
will be performed to re-position the CNTS to make connections with microelectrodes
and push the rest CNTS away. Detailed DEP deposition process and AFM manipu-
lation process have been discussed in Chapter 3.

The structure of a single pixel CNT based IR sensor is shown in Fig. 5.2(a).
Fig. 5.2(b) shows the 3D AFM image of a three-pixel CNT IR sensor array withour
CNT deposited. The pixel pitch of each pixel is 10 pm. A single CNT will be
assembled onto each pair of microelectrodes to make a connection with the automated

nanomanufacturing process.

98

Micropipette Multiple CNT
’ de oosited

     
 

mm
manufacturing
process for single
CNT devices

IIII III}

 
     

llllflfl

Microelectrodes Single CNT
fabrication device

 

Figure 5.1. Nam-manufacturing process for single CNT based nanodevices.

5.2 Testing of Multi—Pixel CNT Sensor Array

The SWNT in our experiments was obtained from BuckyUSA, Inc., and synthesized
by the arc discharge method. The current signal from the sensor was measured with
an Agilent 4156C Precision Semiconductor Parameter Analyzer. The IR response of
the CNT sensor was measured with a near-IR laser (UH5-30G—830-PV, World Star

Tech) with a wavelength of 830 nm and power of 30 mW.

Fig. 5.3 shows the AFM image of a three-pixel CNT IR sensor array. Single-

walled carbon nanotubes (SWNTS) were assembled to each pixel using the developed

nanomanufacturing process.

99

CNT
1

I \
Electrode Electrode ]

 

 

 

(a) (b)

Figure 5.2. (a) The structure of a single pixel CNT IR sensor. (b) The 3D AFM
image of a three—pixel CNT IR sensor array.

First of all, the I—V characteristic curves of each pixel were measured as shown
in Fig. 5.4. It can be seen that the I-V behavior of each pixel varies because of the
ununiformity of CNTS. Therefor, it is necessary to normalize the photocurrent of
each pixel into a same scale for comparison. By scanning the IR spot around each
pixel, the maximum photocurrent 1127:? (where i=1,2,3 is the index of pixels) of that
pixel was measured when the IR spot was aligned to the pixel. For each pixel, the
photocurrent IphJ' can be normalized by the equation Iph_710rm_i = phi/[$1135
As shown in Fig. 5.5, the temporal photocurrent responses of each pixel were tested
while the IR spot was on each pixel respectively. Fig. 5.6 shows the photocurrent
responses while the IR spot was on pixel 1 and the IR source was periodically switched
on and off. The photocurrent of pixel 1 Iph_norm_1 was at its maximum value and
the photocurrent of pixel 2 and 3 (I h_norm_2 and Iph-norm_3) were getting smaller

P
as the pixels were getting away from the IR spot. Since all the pixels were zero biased

100

 

Figure 5.3. The AFM image of the three-pixel CNT IR sensor.

during IR measurements, the photocurrents were generated by the photovoltaic effect,
which is caused by the Schottky contact at CNT—metal interface. As shown in Fig.
5.7, Iph_norm_2 increased to around 1 and the photocurrent of pixel 1 and 3 became
smaller because the IR spot was on pixel 2. The reason that the photocurrent of pixel
1 and 3 did not became zero is that the diameter of the IR spot was big enough to
cover all three pixels. Fig. 5.8 shows the photocurrents while the IR spot center was
on pixel 3. Iph_n0rm_3 increased to around 1 while Iph_norm_1 and Iph_n0rm_2
getting smaller. These IR response measurements verified that the CNT IR sensor
array is able to detect the spatial variance of incident IR power density with a small

pixel pitch.

101

5.0E-O6

 

     

 

 

 

 

3.0E-06 P256”
g 1.0E-06 as“? We”
E ., 4...-
Q) I.
z:
:3
u -l.OE-06

-3.0E-06

-5.0E-O6

-1 .05 o 0.5 1

Voltage (V)

Figure 5.4. I-V curves of the three pixels.

5.3 CNT based Multi-color IR Sensors

Multicolor (multispectral) IR sensing is very important to advanced IR sensor systems
in the applications of remote sensing, environment monitoring, medical diagnostics,
as well as defense and aerospace applications. An IR sensor with multicolor capability
can offer a better target discrimination, tracking, and identification, as well as tem-

perature determination. Multicolor IR sensors are also very useful in industry for gas

pixell pixelZ pixel3
If
a"! /// .5"! " ,/ I I / if). 993‘"
4’} / I ‘. ‘ If . / "I? 1”. . / :-

(a) (b) (C)

Figure 5.5. Illustration of the photocurrent measurements. IR spot was on (a) pixel
1, (b) pixel 2 and (c) pixel 3 respectively.

102

IR ON OFF

 

Normallized Photocurrent

 

 

 

 

Time (Sec)
Figure 5.6. Temporal photocurrent response of each pixels while the IR spot was on
pixel 1 (Fig. 5.5(a)).
IR ON OFF

1.2

 

 

 

Normallized Photocurrent
O
A

 

 

 

Time (Sec)

Figure 5.7. Temporal photocurrent response of each pixels while the IR spot was on
pixel 2 (Fig. 5.5(l))).

IR ON OFF

 

Normallized Photocurrent

 

 

 

 

0 5 1 0 1 5 20 25 30
Time (Sec)

Figure 5.8. Temporal photocurrent response of each pixels while the IR spot was on
pixel 3 (Fig. 5.5(c)).

leakage detection, chemical analysis, and for environmental sensing and control. As
the IR technology continues to advance, there is a growing demand of multicolor IR
sensors for advanced IR sensor systems. However, the difficulties in fabricating mul-
ticolor IR sensors have hindered its development and applications. The traditional
multicolor IR sensors can be classified into two categories based on their structures.
The first type of multicolor IR sensor is fabricated by building multiple IR sensing
layers using different materials which are sensitive to IR with different wavelengthes.
Another type of multicolor IR sensor consists of multiple IR sensors which can detect
light with different lengthes. And a prism system is used to diffract light with de-
sired wavelength to the corresponding sensor. Obviously, the available multicolor IR
systems are complicated and difficult to make. As one of the promising nanomateri—
als that shows strong potential in IR detection, CNT has the advantage of bandgap

tunability, which makes it a perfect material for multicolor detection.

104

 

"[1 Ti T
SWNT
%
‘\ MWNT
/Au Au
I

Figure 5.9. Structure of the multicolor CNT IR sensor

 

As discussed in Chapter 4.2.1, the diameter of a MWNT can be adjusted by
electrically breaking down the outer layers. Since the bandgap of a semiconducting
MWNT is reversely proportional to its diameter, the bandgap of a MWNT can be
increased by breaking outer layers. Hence, a multicolor IR sensor can be fabricated
from MWNT arrays by electrically tuning the bandgap of certain pixels. Since this
fabrication process involves only one IR material, and the bandgap tuning process
can be performed at the device level, the fabrication of multicolor IR sensor becomes
much easier.

Fig. 5.9 shows the structure of the CNT based multicolor IR sensor. It consists of
two pairs of microelectrodes. Here the asymmetric structure is used to improve the
sensitivity as discussed in Chapter 2.3. One pair of the microelectrodes is bridged
by a semiconducting SWNT and another is bridged by a metallic MWNT. Since the
SWNT has a bandgap in the near IR band, it is used to detect IR. By electrically
removing the outer layer of the metallic MWNT, the MWNT can be tuned into a
semiconducting MWNT with small bandgap. By controlling the bandgap tuning
process, we can get an appropriate bandgap for detecting MIR. Obviously, we can
also put a MWNT on another pixel. Then we can have the bandgap of both pixels
tunable. It provides more flexibility for building multicolor IR sensors.

With the same procedure of fabricating multi-pixel CNT IR sensors, we have fabri—

cated prototype of the multicolor CNT IR sensor by depositing a SWNT on one pixel

 

4.0E-08

2.0E-O8

 

0.013+00 m

Current (A)

-2.0E-08 r—-*

 

 

 

 

 

-l -0.6 -0.2 0.2 0.6 1
Voltage (V)

Figure 5.10. I-V characteristic curve of the SW NT pixel of the multicolor CNT IR
sensor

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2.00E-10

1.6OE-10
S, 12013-10 ~ -~— ——- ~- —M~—
s
E 8.00E-11 . 4..-“.-- __ -_i __ _ -
U

4.00E-11 I—MEH—T n.____

0.00E+00 L L i L

O 5 10 15 20
Time (see)

Figure 5.11. NIR response of the SW NT pixel of the multicolor CNT IR sensor

106

and a MWNT on the adjacent pixel. Both the SWNT and the l\*IVV NT were obtained
from BuckyUSA, Inc. First of all, we tested the SWNT pixel. Since the semiconduct-
ing SWNT formed Schottky barriers at the CNT—metal interface. As shown in Fig.
5.10, the I-V curve of the SWNT pixel behaves like two reversely connected Schottky
diodes. The photobehavior of the SWNT pixel was tested with the same procedure
as discussed in Chapter 4. Fig. 5.11 shows the temporal photoresponse of the SW NT
pixel. A NIR laser (UH5-30G—830-PV, World Star Tech) with a wavelength of 830
nm and power of 30 mW was used as the IR source. The laser source was periodically
switched on and off, and no bias voltage was applied across the SWNT pixel. It can
be observed that the photocurrent immediately increased/ decreased when the NIR
was switched on/off. The SWNT pixel has also been tested for MIR band with a
blackbody (Oriel 67030). However, since the bandgap of the SWNT was too large to
detect MIR photons, there was no any signal detected.

The electronic and photonic properties of the MWNT pixel has also been tested.
Fig. 5.12 shows the I-V characteristic curve of the MWN T pixel. It can be seen that
the MWNT showed a metallic behavior and the resistance was quite small. Since
a metallic MWNT would not form Schottky barriers at the MWNT-metal contacts,
the MWNT pixel could not detect any IR light. Experiments has been carried out to
test the photoresponse of this MWNT pixel with both NIR and MIR. But there was
no any signal detected. To active this MWNT pixel for IR detection, the MWNT
bandgap engineering process discussed in Chapter 4.2.1 has been performed to adjust
the electronic and photonic properties of the MWNT. By electrically burning the
outer layers of the MWNT, the bandgap of the MWNT became bigger and its I-V
curve became more semiconducting. Fig. 5.13 shows the I—V characteristic curve of
the MWNT pixel after bandgap engineering. Since the MWN T became semiconduct-
ing and formed Schottky barrier at the MWNT-metal interfaces, the MW NT pixel

behaved like two reversely connected Schottky diode.

107

“dun—u

 

2.0E-O7

 

 

 

 

 

 

1 05-07 ~
E
E, 0.0E+OO . +
S
O

-1 05-07 ~

-2.0E-O7

-1 -0.5 o 0.5 1

Voltage (V)

Figure 5.12. I-V characteristic curve of the MWNT pixel of the multicolor CNT IR
sensor

108

8.0E-08

4.0E-08 /

0.0E+OO

 

 

 

Current (A)

-4.0E-08 ~-

 

 

 

 

 

-1 ~05 O 0.5 1
Voltage (V)

Figure 5.13. I-V characteristic curve of the MW NT pixel after breaking down outer
layers

109

After tuned the bandgap of the MW N T pixel, experiments was carried out to test
the photoresponse of the IVIVVNT pixel. First of all, N IR laser source was used for the
IR testing. Fig. 5.14 shows the temporal IR response of the MWN T pixel when the
NIR laser was periodically switch on and off. It can be observed that the photocurrent
responses to the IR illumination very quickly. During the measurement, a zero bias
voltage was applied on the MW NT pixel. It means that the detected photocurrent
was a short circuit current. Hence the MWNT should work as a photovoltaic device
here. After taking measurements with the NIR, the MIR has also been tested by
using a blackbody as the MIR source. An IR filter with a spectral range around 3 um
was used to block IR light with other wavelengthes. Fig. 5.15 shows the temporal
response of the MWNT pixel when the MIR was periodically switched on and off by
using a mechanical chopper. It can be observed that the MIR response was quite
different with the NIR response. When then MIR was turned on/off, the current
suddenly increased / decreased and then gradually got back to the dark current level.
As discussed in Chapter 4, this is because the MWN T formed Schottky barrier at the
contact of both end with the metal electrodes. The two Schottky diodes reversely
connected. When the MWNT device was illuminated by MIR, both Schottky diode
generated electrons and holes. However, since there were two reversely connected
Schottky diode, the photogenerated charges at one Schottky diode had to go through
another one to form a photocurrent. Since the incident power density of the NIR
was big enough to generate sufficient charges at each Schottky barrier and form a
big open circuit voltage at each Schottky diode, the photogenerated charges at one
diode were able to tunnel through another diode and formed a photocurrent. That
is why the photocurrent could keep its value after the NIR was switched on. But
for the MIR, the incident power density was much smaller and could not generate
enough charges at the Schottky diode to form a big open circuit voltage. Hence the

photogenerated charges at one diode could not. get through another diode to form a

110

IR OFF ON

 

 

 

 

 

 

 

 

 

 

1 .20E-1 1
8.00E-12
E
E 4.00E-12
5
O
0.00E+00
-4.00E-12
0 5 10 15 20 25 30
Time (See)

Figure 5.14. NIR response of the MW NT pixel of the multicolor CNT IR sensor

photocurrent. And the two Schottky barriers worked like two capacitors. When the
MIR was switched on, the capacitors got charged and there was a suddenly current
increase. But as time going, the charging current gradually reduced to zero. When
the MIR was switched off, the capacitors discharged and there was a suddenly current

decrease. And the discharging current gradually reduced to zero.

5.4 Chapter Summary

A three-pixel CNT based IR sensor array with 10 um pixel pitch has been fabricated
using the developed nanomanufacturing process. The experimental testing on the
electronic and photonic properties of each pixel shows that CN T formed a Schottky
junction with the contact metal. And the photovoltaic effect of each Schottky junction

was used to detect IR. It has also been shown that the CNT sensor array is able to

111

IR ON OFF

—LJ a. _

 

 

 

 

 

1.2E-12

 

 

8.0E-13

4.0E-13

0.0E+OO

 

Current (A)

 

 

-4.0E-13

 

 

 

-8.0E-13

 

O 5 1 0 1 5 20 25 30 35
Time (See)

Figure 5.15. MIR response of the MWNT pixel of the multicolor CNT IR sensor

detect spacial variance of incident IR power density and has small thermal cross talk
between adjacent pixels, which makes CNT a promising material for high resolution
IRFPAs. More importantly, the fabrication of the single CNT based IR sensor array
demonstrated the strong ability of the nanomanufacturing system. As the first CNT
based IR sensing array, it is also a significant advancement in the application of the
CNT based nanodevices.

With the technologies for fabricating multi-pixel CNT sensor array, CNT based -
multi-color IR sensors have also been demonstrated by having two pixels which can
detect IR with different wavelengthes. With its small diameter and relatively large
bandgap, a SWNT was used for NIR detection. Since the bandgap of CN T is reversely
proportional to its diameter, the bandgap of a MWNT can be tuned by electrically '
breaking its outer layers. With the bandgap tunability, a MWNT has been success-

fully engineered from a. metallic MW NT into a semiconducting MW NT for MIR de-

112

tection. This multicolor CNT IR sensor first demonstrated the application of CNTS
in multicolor detection. It further demonstrated that CNTS have strong potential
for applications in the area of remote sensing, environment. monitoring, defense and

aerospace applications, and etc.

113

CHAPTER 6

Conclusions

As the traditional silicon based semiconductor industry faces increasing technological
and financial challenges on the way to further scaling down, new technologies and
concepts have to be developed and assessed. Nanotechnology is such a candidate
and is likely to be the engine for the future advancement in the fields of electron-
ics, energy science, biomedical science, materials, and etc. Nanomanufacturing and
nanoelectronic devices are two of most important subdivisions of nanotechnology.

In this research, a nanomanufacturing process, which is able to fabricate single
CNT as well as other nanomaterials (such as nanowire, DNA, nanoparticle, etc.)
based nanodevices, was designed and experimentally tested. This technology will
greatly facilitate researcher’s exploration in the nano world by providing an efficient
and reliable manner to handle nanomaterials and build nanodevices. With this tech-
nology, single CNT based infrared sensors have been designed and implemented. CN T
sensors with different CNT-metal contacts have been fabricated and tested. The stud—
ies of CNT-metal contacts and the CNT based Schottky diode not only help improve
the performance of CNT based infrared sensors, but also are important for other ON T
based nanoelectronic devices which utilize the CN T Schottky diode.

With the asymmetric CNT—metal contact structure, SWNT devices have been

114

fabricated and characterized. SW NTs show a strong potential for applications of
NIR detection and solar collection. Since SWNTS are limited by their thin diameters
to detect NIR or lights with shorter wavelength, MWNTS have been studied for
MIR detection. A bandgap engineering technology has been developed to adjust the
bandgap of a MWNT by electrically removing outer layers of the MWNT. With this
technology, detection of MIR using MWNT sensors have been demonstrated in this
thesis. It is the first single CNT device which is able to detect MIR.

Moreover, ON T based multipixel IR sensor arrays have been fabricated and tested.
The CNT multipixel IR sensor array shows promising properties such as small pixel
pitch and low crosstalk. It is also the first time that a CNT device array can be fab-
ricated in a reliable and controllable manner, which means the number of CNTS on
each pixel, and the properties of each CNTS can be controlled during the fabrication
process. With the bandgap engineering technology and the ability to build CNT de-
vice array, a CNT based multicolor IR sensor has also been firstly demonstrated. The
CNT based multicolor IR sensor further expands CNT’s application into those fields
which need advanced IR sensor systems for a better target discrimination, tracking,

and identification, as well as temperature determination.

6.1 Major Contribution

In summary, this thesis has developed a nanomanufacturing system, which is a signif-
icant contribution in handling nanomaterials and building nanodevices. It provides
a reliable and efficient way for assemble nanodevices, especially CNT and nanowire
based nanodevices. This nanomanufacturing system enables the fabrication of single
CNT/nanowire based nanodevice arrays, which was difficult to achieve with other
nanofabrication methods.

Moreover, as another main contribution, CNTS have been thoroughly studied for

115

 

IR detection. The role of CNT-metal contacts in the CNT based IR sensors was firstly
studied for thoroughly understanding photodetection theory of CNT IR sensors and
improving the performance of CNT IR sensors. Based on the study of CNT-metal
contacts, the signal—to—dark current. ratio as well as quantum efficiency have been
significantly improved with an asymmetric structure.

The third major contribution is that a bandgap engineering process has been
developed to tune the bandgap of MWNTS for MIR detection. Quantum detection
of MIR using individual MWNT was first demonstrated in this dissertation. The
individual MWNT based MIR sensor has much higher sensitivity and faster response
speed than the previously reported MWNT film based MIR bolometer.

The last major contribution is that single CNT based multipixel IR sensor array
and multicolor IR sensor have been firstly demonstrated in the world. CNTS have
shown a strong potential of applications in the fields of solar collection, remote sens-
ing, environmental monitoring, medical diagnostics, as well as defense and aerospace

applications.

6.2 Future Research

The future research will focus on two directions. Firstly, it is necessary to further
improve the quantum efficiency of the CNT based IR sensors. As discussed in Chapter
4, though the developed CNT IR sensor dramatically outperformed the other reported
results, its quantum efficiency is still difficult to compete with traditional IR materials.
Further studies on the CNT itself and sensor structure are necessary to improve the
performance of CNT based IR sensors. Another future research direction will be
developing real IRF PAS. This dissertation has demonstrated the CNT based three-
pixel IR sensor arrays. But it is still far away from real IRF PA applications. Hence

it is necessary to further improve the nanomanufacturing technologies to make high

116

 

WM..- “‘—,

density IRFPAs.

117

REFERENCES

118

 

Author’s Publications

1. Jiangbo Zhang, Ning Xi, Hongzhi Chen, Wai Chiu King Lai, Guangyong Li,
U.C. Wejinya, Design, Manufacturing and Testing of Single Carbon N anotube
Based Infrared Sensors, IEEE Trans. Nanotechnology, accepted for publication,
2008.

2. Jiangbo Zhang, Ning Xi, Hongzhi Chen, Wai Chiu King Lai, Guangyong Li,
U.C. Wejinya, Photovoltaic Effect in Single Carbon Nanotube Based Schottky
Diodes, Int. J. of Nanoparticles, accepted for publication, 2008.

3. Jiangbo Zhang, Ning Xi, Lianqing Liu, Hongzhi Chen, King W.C. Lai, Guangy—
ong Li, Atomic Force Microscope Based Nano Robotic System for N anomanip—
ulation, IEEE Nanotechnology Magazine, accepted for publication, 2008.

4. U. C. Wejinya, Ning Xi, Yantao Shen, King W. C. Lai, Jiangbo Zhang, An Ef—
ficient Approach of Handling and Deposition of Micro and Nano Entities Using
Sensorized Microfiuidic End-Effector System, Sensor and Actuators, accepted
for publication, 2008.

5. Lianqing Liu, Yilun Luo, Ning Xi, Jiangbo Zhang, Guangyong Li, Sensor Ref-
erenced Real-Time Videolization of Atomic Force Microscopy for Nano Manip-
ulations, IEEE/ASME transactions on mechatronics, accepted for publication,
2008.

6. W. C. King Lai, Ning Xi, Jiangbo Zhang, Hongzhi Chen, Yilun Luo,
Uchechukwu C. Wejinya, Automated Nanomanufacturing System to Assemble
Carbon Nanotube Based Devices, International Journal of Robotics Research,
accepted for publication, 2008.

7. Jiangbo Zhang, Ning Xi, and King Lai, Single Carbon Nanotube Infrared De-
tectors, Nanotechnology, SPIE Newsroom (newsroomspieorg), Jan. 5, 2007.

8. Jiangbo Zhang, Ning Xi, Guangyong Li, Ho-Yin Chan, and U.C. Wejinya,
Adaptable End Effector for Atomic Force Microscopy Based Nanomanipula-
tion, IEEE Transactions on Nanotechnology, 5(6):628-642, 2006

119

 

 

10.

11.

12.

13.

14.

15.

16.

17.

18.

Jiangbo Zhang, Ning Xi, Hongzhi Chen, King W.C. Lai, Yilun Luo, Carbon
Nanotube based Infrared Detector Array, IEEE International Conference on
Nanotechnology, accepted, 2008

Jiangbo Zhang, King W.C. Lai, Jeffri Narendra, N ing Xi, Timothy Grotjohn,
Jes Asmussen, Computer Integrated Plasma Nano Manufacturing W'orkcell,
IEEE International Conference on Nanotechnology, accepted, 2008

Jiangbo Zhang, Ning Xi, Hongzhi Chen, and King W'. C. Lai, Fabrication
and Experimental Testing of Individual Multi-walled Carbon N anotube (CNT)
based Infrared Sensors, Proceeding of the 6th IEEE International Conference
on Sensors, Oct. 28 - 31, 2007, Atlanta, GA.

Lianqing Liu, Ning Xi, Yilun Luo, Jiangbo Zhang, and Guangyong Li, Sensor
Referenced Guidance and Control for Robotic Nanomanipulation, Proceeding
of the IEEE/RSJ 2007 International Conference on Intelligent Robots and Sys-
tems, Oct. 29 Nov. 2, 2007, San Diego, CA.

Lianqing Liu, Jiangbo Zhang, Guangyong Li, and Ning Xi, On-line Sensing and
Display in Atomic Force Microscope Based N anorobotic Manipulation, Proceed-
ing of the 2007 IEEE/ASME International Conference on Advanced Intelligent
Mechatronics, Sept. 4-7, 2007, ETH Zrich, Switzerland

Jiangbo Zhang, Ning Xi, King W. C. Lai, Hongzhi Chen, Yilun Luo, and
Guangyong Li, Single Carbon N anotube based Photodiodes for Infrared Detec-
tion, Proceeding of the 7th IEEE International Conference on Nanotechnology,
Aug. 2 - 5, 2007, Hong Kong.

Lianqing Liu, Ning Xi, Yilun Luo, J iangbo Zhang, and Guangyong Li, Real-time
Position Error Detecting in Nanomanipulation Using Kalman Filter, Proceeding
of the 7th IEEE International Conference on Nanotechnology, Aug. 2 - 5, 2007,
Hong Kong.

King W. C. Lai, Ning Xi, Jiangbo Zhang, Guangyong Li, and Hongzhi Chen,
Packaging Carbon Nanotube Based Infrared Detector, Proceeding of the 7th
IEEE International Conference on Nanotechnology, Aug. 2 - 5, 2007, Hong
Kong.

Guangyong Li, Lianqing Liu, Ning Xi, Jiangbo Zhang and King W. C. Lai,
Study of DNA Properties under Controlled Conditions Using AFM Based Nano-
Robotics, Proceeding of the 7th IEEE International Conference on Nanotech-
nology. August 2-5, 2007, Hong Kong.

Lina Hao, Jiangbo Zhang, and Ning Xi, A Controller for Adaptable End Ef-
fector of the AF M Based Nano-Robotic System, Proceeding of the 7th IEEE
International Conference on Nanotechnology, Aug. 2 - 5, 2007, Hong Kong.

120

,l

 

19.

20.

21.

22.

23.

24.

26.

27.

28.

29.

Lina Hao, J iangbo Zhang and Ning Xi, Optimal Periodic-Output—Feedback Con-
trol of Active AFM Probe for Nanomanipulation, Proceeding of the European
Control Conference 2007, July 2-5, 2007, Kos, Greece

Jiangbo Zhang, Ning Xi, Hongzhi Chen and King W .C. Lai, Efficient readout
for carbon nanotube (CNT)-based IR detectors, Proc. SPIE Int. Soc. Opt.
Eng. 6737, 673711 (2007).

Ning Xi, Jiangbo Zhang, Harold Szu, and Guangyong Li, Nanorobot asembly
of Carbon Nanotubes for Mid-IR sensor, Proc. SPIE Int. Soc. Opt. Eng. 6576,
65760K (2007)

J iangbo Zhang, N ing Xi and King W'.C. Lai, Performance Evaluation and Ana]-
ysis for Carbon Nanotube (CNT) based IR Detectors, Proc. SPIE Int. Soc.
Opt. Eng. 6542, 65421N (2007)

Lina Hao, Jiangbo Zhang, and Ning Xi, The Infinite Dimensional Control of
Flexible Cantilevers in AF M Based Nanomanipulation, Proceeding of the 2nd
IEEE International Conference on Nano-Micro Engineered and Molecular Sys-
tems, Jan. 16-19, 2007, Bangkok, Thailand

Jiangbo Zhang, Ning Xi, Hoyin Chan, and Guangyong Li, Single Carbon Nan-
otube based Infrared Sensor, Proc. SPIE Int. Soc. Opt. Eng. 6395, 63950A
(2006)

. Jiangbo Zhang, Ning Xi, Hoyin Chan, Guangyong Li, Single Carbon Nanotube

Based Ion Sensor for Gas Detection, Proceeding of the 6th IEEE International
Conference on Nanotechnology, June 17—20, 2006, Cincinnati, Ohio

Ho—Yin Chan, Ning Xi, Jiangbo Zhang, Guangyong Li, Nanoassembly and
Packaging of Single Carbon Nanotube Based Transistors, Proceeding of the 6th
IEEE International Conference on Nanotechnology, June 17—20, 2006, Cincin—
nati, Ohio

J iangbo Zhang, Ning Xi, Guangyong Li, Sensing in the Nano-environment based
on High Order Harmonic Modes of Flexible Arm, Proceeding of the 4th IFAC
Symposium on Mechatronic Systems, Sept. 12-14, 2006, Heidelberg, Germany

J iangbo Zhang, N ing Xi, Guangyong Li, Chanmin Su, Atomic Force Microscopy
Sensing Using Multiple Modes, Proceeding of the IEEE/RSJ 2006 International
Conference on Intelligent Robots and Systems, Oct. 9-15, 2006, Beijing, China

U.C. Wejinya, Yantao Shen, Ning Xi, Jiangbo Zhang, Microfluidic end effector
for manufacturing of nano devices, Proceeding of the 2006 IEEE International
Conference on Robotics and Automation, May 15-19, 2006, Orlando, Florida.

121

 

 

 

30.

31.

32.

33.

34.

Ho—Yin Chan, N ing Xi, Jiangbo Zhang, Guangyong Li, Tuning Serniconducting
Properties of Single Carbon Nanotube for Fabrication of Nano Devices, Pro-
ceeding of the Ist IEEE International Conference on Nana-Micro Engineered
and Molecular Systems, Jan. 18-21, 2006, Zhuhai, China

Jiangbo Zhang, Ning Xi, Guangyong Li, Optimal Control of Flexible End Ef—
fector in AF M based Nanomanipulation, Proceeding of the 2005 IEEE/RSJ
International Conference on Intelligent Robots and Systems, Aug. 2-6, 2005,
Edmonton, Canada.

Ho—Yin Chan, Ning Xi, Jiangbo Zhang, Guangyong Li, A deterministic process
for fabrication and assembly of single carbon nanotube based devices, Proceeding
of the 5th IEEE International Conference on Nanotechnology, July 11-15, 2005,
Nagoya, Japan

Jiangbo Zhang, Guangyong Li, Ning Xi, Modeling and Control of Active End
Effector for the AF M Based N ano Robotic Manipulators, Proceeding of the 2005
IEEE International Conference on Robotics and Automation, April 18-22, 2005,
Barcelona, Spain.

Heping Chen, Ning Xi, Guangyong Li, Jiangbo Zhang, Prokos, M., Planning
and Control for Automated Nanorobotic Assembly, Proceeding of the 2005
IEEE International Conference on Robotics and Automation, April 18-22, 2005,
Barcelona, Spain.

. Guangyong Li, Ning Xi, A. Saeed, Heping Chen, Jiangbo Zhang, W.J. Li,

C.K.M. Fung, R.H.M. Chan, Mingjun Zhang, szh-Jong Tarn, Experimental
studies of DNA electrical properties using AF M based nano-manipulator, Pro-
ceeding of the 4th IEEE Conference on Nanotechnology, August 17-19, 2004,
Munic, Germany

122

 

BIBLIOGRAPHY

 

123

BIBLIOGRAPHY

[1] Sumio Iijima. Helical microtubules of graphitic carbon. Nature, 354:56—58,
1991.

[2] Adrian Bachtold, Peter Hadley, Takeshi Nakanishi, and Cees Dekker. Logic
circuits with carbon nanotube transistors. Science, 294:1317-1320, 2001.

[3] Derycke V, Martel R, Appenzeller J, and Avouris P. Carbon nanotube inter-
and intramolecular logic gates. Nano Letters, 1:453—456, 2001.

[4] Dong Qian, Gregory J Wagner, Wing Kam Liu, Min-Feng Yu, and Rodney S
Ruoff. Mechanics of carbon nanotubes. Applied Mechanics Reviews, 55:495—533,
2002.

[5] Paul L. Mchen, Michael S. Fuhrer, and Hongkun Park. Single-walled carbon
nanotube electronics. IEEE Transactions on Nanotechnology, 1:78—85, 2002.

[6] Q Z Xue. Model for the effective thermal conductivity of carbon nanotube
composites. Nanotechnology, 17:1655—1660, 2006.

[7] MS. Dresselhaus, G. Dresselhaus, A. Jorio, A.G. Souza Filho, and R. Saito.
Raman spectroscopy on isolated single wall carbon nanotubes. Carbon, 40:2043—
2061, 2002.

[8] Robert J. Chen, Nathan R. Franklin, J ing Kong, J ien Cao, Thomas W. Tombler,
Yuegang Zhang, , and Hongjie Dai. Molecular photodesorption from single-
walled carbon nanotubes. Appl. Phys. Lett., 79:2258—2260, 2001.

[9] Akihiko Fujiwara, Yasuyuki Matsuoka, Hiroyoshi Suematsu, Naoki Ogawa, Ken-
jiro Miyano, Hiromichi Kataura, Yutaka Maniwa, Shinzo Suzuki, and Yohji

Achiba. Photoconductivity in semiconducting single—walled carbon nanotubes.
Jpn. J. Appl. Phys, 40:L1229—L1231, 2001.

[10] I. A. Levitsky and W. B. Euler. Photoconductivity of single—wall carbon nan-
otubes under continuous-wave near-infrared illumination. Appl. Phys. Lett,
83:1857-—1859, 2003.

124

 

[11] Liyue Liu and Yafei Zhang. h'lulti-wall carbon nanotube as a new infrared
detected material. Sensors and Actuators A, 1161394" 397, 2004.

[12] Kannan Balasubramanian, Yuwei Fan, Marko Burghard, Klaus Kern, .\‘Iarcel
Friedrich, Uli Wannek, and Alf Mews. Photoelectronic transport imaging of
individual semiconducting carbon nanotubes. A ppl. Phys. Lett., 84:2400-2402,
2004.

[13] M. Freitag, Y. Martin, J. A. hl'Iisewich, R. Martel, and Ph. Avouris. Photocon-
ductivity of single carbon nanotubes. Nano Letters, 321067—1071, 2003.

[14] Ji Ung Lee. Photovoltaic effect in ideal carbon nanotube diodes. Appl. Phys.
Lett., 872073101, 2005.

[15] A. Mohite, S. Chakraborty, P. Gopinath, G. U. Sumanasekera, and B. W.
Alphenaar. Displacement current detection of photoconduction in carbon nan-
otubes. Appl. Phys. Lett., 86206114, 2005.

[16] Hongjie Dai. Carbon nanotubes: Synthesis, integration, and properties. Acc.
Chem. Res, 35:1035—1044, 2002.

[17] M. Dresselhaus, G. Dresselhaus, and Phaedon Avouris (Eds). Carbon Nan-
otubes: Synthesis, Structure, Properties and Application. Springer, Berlin, 2001.

[18] J. W. Mintmire, B. I. Dunlap, and C. T. White. Are fullerene tubules metallic?
Phys. Rev. Lett., 681631 - 634, 1992.

[19] M. S. Dresselhaus, G. Dresselhaus, and Riichiro Saito. Carbon fibers based on
c60 and their symmetry. Phys. Rev. B, 4516234 — 6242, 1992.

[20] Phaedon Avouris. Carbon nanotube electronics. Chemical Physics, 2812429—
445, 202.

[21] Philip G. Collins and Phaedon Avouris. Carbon fibers based on c60 and their
symmetry. Scientific American, 283:62—29, 2000.

[22] H. Dai. Carbon nanotubes: opportunities and challenges. Surf. Sci, 500:218—
241, 2002.

[23] Markus Ahlskog, Pertti Hakonen, Mikko Paalanen, Leif Roschier, and Reeta
Tarkiainen. Multiwalled carbon nanotubes as building blocks in nanoelectronics.
Journal of Low Temperature Physics, 124:335—352, 2001.

[24] B. Q. Wei, R. Vajtai, and P. M. Ajayan. Reliability and current carrying
capacity of carbon nanotubes. Appl. Phys. Lett., 79211724174, 2001.

125

[25] N. Srivastava and K. Banerje. Performance analysis of carbon nanotube inter-

[261

127]

[281

[291

[301

[311

[321

1331

[341

[35]

connects for vlsi applications. In Proc. of the 2005 IEEE/A CM Int. Conf. on
Computer—aided design, pages 383—390, San Jose, CA, 2005.

S. J. Wind, J. Appenzeller, and Ph. Avouris. Lateral scaling in carbon-nanotube
field-effect transistors. Phys. Rev. Lett., 91(5):058301, Jul 2003.

S. Hasan, S. Salahuddin, M. Vaidyanathan, and M. A. Alam. High-frequency
performance projections for ballistic carbon-nanotube transistor. IEEE T rans-
actions on Nanotechnology, 5:14-22, 2006.

J. Appenzeller, Yu-Ming Lin, J. Knoch, Zhihong Chen, and P. Avouris. Com-
paring carbon nanotube transistors - the ideal choice: A novel tunneling device
design. IEEE Transactions on Electron Devices, 52:2568—2576, 2005.

S. J. Wind, J. Appenzeller, R. Martel, V. Derycke, and Ph. Avouris. Vertical
scaling of carbon nanotube field-effect transistors using top gate electrodes.
Applied Physics Letters, 80:2002, 3817-3819.

D. V. Singh, K. A. Jenkins, J. Appenzeller, D. Neumayer, A. Grill, and H.-S. P.
Wong. Frequency response of top-gated carbon nanotube field-effect transistors.
IEEE Transactions on Nanotechnology, 3:383—387, 2004.

Z. Chen, J. Appenzeller, Y.-M. Lin, J. Sippel-Oakley, A. G. Rinzler, J. Tang,
S. J. Wind, P. M. Solomon, and Ph. Avouris. An integrated logic circuit as-
sembled on a single carbon nanotube. Science, 311:1735, 2006.

Marcus Freitag, Vasili Perebeinos, Jia Chen, Aaron Stein, James C. Tsang,
James A. Misewich, Richard Martel, and Phaedon Avouris. Hot carrier elec-

troluminescence from a single carbon nanotube. Nano Letters, 4:1063 ~1066,
2004.

Shaoxin Lu and Balaji Panchapakesan. Photoconductivity in single wall carbon
nanotube sheets. Nanotechnology, 17:1843~1850, 2006.

Nathan R. Franklin, Qian Wang, Thomas W. Tombler, Ali Javey, Moonsub
Shim, and Hongjie Dai. Integration of suspended carbon nanotube arrays into

electronic devices and electromechanical systems. Appl. Phys. Lett., 81:913-915,
2002.

Pengfei Qi, Ophir Vermesh, Mihai Grecu, Ali Javey, Qian Wang, Hongjie Dai,
Shu Peng, and K. J. Cho. Toward large arrays of multiplex functionalized
carbon nanotube sensors for highly sensitive and selective molecular detection.
Nano Lett., 3:347—351, 2003.

126

 

[36]

[371

[381

[39]

[40]

[41]

[421

[431

[441

[451

[461

[471

Alireza Nojeh, Ant Ural, R. Fabian Pease, and Hongjie Dai. Electric-field-
directed growth of carbon nanotubes in two dimensions. J. Vac. Sci. Technol.
B, 22:3421—3425, 2004.

Jingqi Li, Qing Zhang, Dajiang Yang, and Jingze Tian. Fabrication of car-
bon nanotube field effect transistors by ac dielectrophoresis method. Carbon,
42:2263—2267, 2004.

Sebastian Taeger, Daniel Sickert, Petar Atanasov, Gerald Eckstein, and
Michael Mertig. Self—assembly of carbon nanotube field-effect transistors by
ac-dielectrophoresis. phys. stat. sol. (b), 243:3355—3358, 2006.

S. Banerjee, B. White, L. Huang, B.J. Rego, S. O’brien, and LP. Herman.
Precise positioning of carbon nanotubes by ac dielectrophoresis using floating
posts. Appl. Phys. A, 86:415—419, 2007.

P. Makaram, S. Selvarasah, X. Xiong, C.-L. Chen, A. Busnaina, N. Khanduja,
and M. R. Dokmeci. Three-dimensional assembly of single-walled carbon nan-
otube interconnects using dielectrophoresis. Nanotechnology, 182395204, 2007.

Zhihong Chen, Joerg Appenzeller, Joachim Knoch, Yu ming Lin, and Phaedon
Avouris. The role of metal-nanotube contact in the performance of carbon
nanotube field-effect transistors. Nano Letters, 521497-1502, 2005.

Chenguang Lu, Lei An, Qiang Fu, Jie Liu, Hong Zhang, and James Murduck.
Schottky diodes from asymmetric metal-nanotube contacts. Appl. Phys. Lett.,
88:133501, 2006.

David Perello, Moon J. Kim, DongKyu Cha, Gang Hee Han, Dong Jae Bae,
Seung Yol Jeong, Young Hee Lee, and Minhee Yun. Schottky barrier engineering
in carbon nanotube with various metal electrodes. In Proceedings of the 7th
IEEE International Conference on Nanotechnology, pages 189—193, Hong Kong,
Aug. 2007.

J. Lefebvre, M. Radosavljevic’, and A. T. Johnson. Fabrication of nanometer
size gaps in a metallic wire. Appl. Phys. Lett., 76:3828—3830, 2000.

Mandar M. Deshmukh, D. C. Ralph, M. Thomas, and J. Silcox. Nanofabrication
using a stencil mask. Appl. Phys. Lett., 75:1631—1633, 1999.

Saleem G. Rao, Ling Huang, Wahyu Setyawan, and Seunghun Hong. Large-
scale assembly of carbon nanotubes. Nature, 425:36—37, 2003.

Miron Hazani, Frank Hennrich, Manfred Kappes, Ron Naaman, Dana Peled,
Victor Sidorov, and Dmitry Shvarts. Dna—mediated self—assembly of carbon
nanotube—based electronic devices. Chem. Phys. Lett., 391:389—392, 2004.

127

[48]

[491

[501

[511

[53]

[541

[551

[561

[571

Zhifeng Deng, Erhan Yenilmez, Amy Reilein, Joshua Leu, Hongjie Dai, and
Kathryn A. Moler. Nanotube manipulation with focused ion beam. Appl.
Phys. Lett., 88:023119, 2006.

MinFeng Yu, Mark J Dyer, George D Skidmore, Henry W Rohrs, XueKun Lu,
Kevin D Ausman, James R Von Ehr, and Rodney S Ruoff. Three-dimensional
manipulation of carbon nanotubes under a scanning electron microscope. Nan-
otechnology, 10:244-252, 1999.

Tobias Hertel, Richard Martel, and Phaedon Avouris. lVIanipulation of individ—
ual carbon nanotubes and their interaction with surfaces. J. Phys. Chem. B,
102:910—915, 1998.

Ph. Avouris, T. Hertel, R. Martel, T. Schmidt, H.R. Shea, and RE. Walkup.
Carbon nanotubes: nanomechanics, manipulation, and electronic devices. Appl.
Surf. Sci, 141:201—209, 1999.

Henk W. C. Postma, Allard Sellmeijer, and Cees Dekker. Manipulation and
imaging of individual single-walled carbon nanotubes with an atomic force mi-
croscope. Adv. Mater., 12:1299-1302, 2000.

G. Binnig, C. F. Quate, and C. Gerber. Atomic force microscope. Phys. Rev.
Lett., 56(9):930—933, 1986.

Claudio Rivetti, Chip Walker, and Carlos Bustamante. Polymer chain statistics
and conformational analysis of dna molecules with bends or sections of different
flexibility. Journal of Molecular Biology, 280:41—59, 1998.

Masato Tanigawa, Masanori Gotoh, Masayuki Machida, Takao Okada, and Mi-
chio Oishi. Detection and mapping of mismatched base pairs in dna molecules
by atomic force microscopy. Nucleic Acids Research, 28:e38—e38, 2000.

S. Boichot, M. Fromm, S. Cunniffe, P. O’Neill, J. C. Labrune, A. Chambaudet,
E. Delain, and E. Le Cam. Investigation of radiation damage in dna by using
atomic force microscopy. Radiation Protection Dosimetry, 99:143—146, 2002.

Aiguo Wu, Lihua Yu, Zhuang Li, Huanming Yang, and Erkang Wang. Atomic
force microscope investigation of large-circle dna molecules. ANALYTICAL
BIOCHEMISTRY, 325:293—300, 2004.

Minoru Takeuchi, Hiroshi Miyamoto, Yasushi Sako, Hideo Komizu, and Akihiro
Kusumi. Structure of the erythrocyte membrane skeleton as observed by atomic
force microscopy. Biophysical Journal, 74:2171—2183, 1998.

128

[591

[601

[511

[ml

[631

[64]

[65]

[66]

[671

[681

[691

Nuno C. Santos, Evgeny Ter-Ovanesyan, Joseph A. Zasadzinski, Manuel Prieto,
and Miguel A. R. B. Castanho. Filipin-induced lesions in planar phospholipid
bilayers imaged by atomic force n'iicroscopy. Biophysical Journal, 75:1869-—1873,
1998.

F. VALLE, J. A. DEROSE, G. DIETLER, M. KAWE, A. PLUCKTHUN, and
G. SEMENZA. Imaging the native structure of the chaperone protein groel
without fixation using atomic force microscopy. Journal of Microscopy, 203: 195—
198, 2001.

FM Ohnesorge, JK Horber, W Haberle, CP Czerny, DP Smith, and G Binnig.
Afm review study on pox viruses and living cells. Biophysical Journal, 73:1997,
2183-2194.

Filip Braet, Ronald de Zanger, Carine Seynaeve, Marijke Baekeland, and Eddie
Wisse. A comparative atomic force microscopy study on living skin fibroblasts
and liver endothelial cells. Journal of Electron Microscopy, 50:283—290, 2001.

Jordanka Zlatanova, Stuart M. Lindsay, and Sanford H. Leuba. Single molecule
force spectroscopy in biology using the atomic force microscopy. Progress in
Biophysics 55 Molecular Biology, 74:37—61, 2000.

FranCois Ahimou, Frédéric A. Denis, Ahmed Touhami, and Yves F. Dufrée.
Probing microbial cell surface charges by atomic force microscopy. Langmuir,
18:9937—9941, 2002.

Shuhei Nishida, Yutaka Funabashi, and Atsushi Ikai. Combination of afm with
an objective-type total internal reflection fluorescence microscope (tirfm) for
nanomanipulation of single cells. Ultramicroscopy, 91:269—274, 2002.

Robert W. Stark, Ffancisco J. Rubio—Sierra, Stefan Thalhammer, and Wolf-
gang M. Heckl. Combined nanomanipulation by atomic force microscopy and

uv-laser ablation for chromosomal dissection. European Biophysics Journal,
32:33—39, 2003.

Sitti Metin. Teleoperated 2-D Micro/Nanomanipulation Using Atomic Force
Microscope. PhD thesis, University of Tokyo, 1999.

T. Junno, K. Deppert, L. Montelius, and L. Samuelson. Controlled manip-
ulation of nanoparticles with an atomic force microscope. Appl. Phys. Lett.,
66(26):3627—3629, June 1995.

D. M. Schaefer, R. Reifenberger, A. Patil, and R. P. Andres. Fabrication of two—
dimensional arrays of nanometer-size clusters with the atomic force microscope.
Appl. Phys. Lett., 66:1012—1014, Feb. 1995.

129

[70] R. W. Stark, S. Thalhammer, J. Wienberg, and W. M. Heckl. The afm as a tool
for chromosomal dissection-the influence of physical parameter. A ppl. Phys. A,
66:579—584, 1998.

[71] A. A. G. Requicha, C. Baur, A. Bugacov, B. C. Gazen, B. Koel, A. Madhukar,
T. R. Ramachandran, R. Resch, and P. Will. Nanorobotic assembly of two-
dimensional structures. In Proc. IEEE Int. Conf. Robotics and Automation,
pages 3368—3374, Leuven, Belgium, May 1998.

[72] L. T. Hansen, A. Kuhle, A. H. Sorensen, J. Bohr, and P. E. Lindelof. A technique
for positioning nanoparticles using an atomic force microscope. Nanotechnology,
9:337—342, 1998.

[73] M. Falvo, R. Superfine, S. Washburn, M. Finch, R.M. Taylor, V.L. Chi, and
RP. Brooks Jr. The nanomanipulator: A teleoperator for manipulating mate-
rials at the nanometer scale. In Proceedings of the International Symposium on

the Science and Technology of Atomically Engineered Materials, pages 579—586,
Richmond VA, Oct. 1995.

[74] M. Sitti and H. Hashimoto. Tele-nanorobotics using atomic force microscope.
In Proc. IEEE Int. Conf. Intelligent Robots and Systems, pages 1739—1746,
Victoria, B. O, Canada, October 1998.

[75] M. Sitti and H. Hashimoto. Controlled pushing of nanoparticles: modeling and
experiments. IEEE/ASME Trans. on Mechatronics, 5:199—211, 2000.

[76] G. Y. Li, N. Xi, M. Yu, and W. K. Ring. Modeling of 3—d interactive forces in
nanomanipulation. In Proc. IEEE Int. Conf. Intelligent Robots and Systems,
pages 2127—2132, Las Vegas, NV, Oct. 28—30 2003.

[77] Guangyong Li, Ning Xi, Mengmeng Yu, and Wai Keung Fung. Augmented
reality system for real-time nanomanipulation. In Proc. IEEE Int. Conf. Nan-
otechnology, San Francisco, CA, August 12-14 2003.

[78] G. Y. Li, N. Xi, and M. Yu. Development of augmented reality system for afm
based nanomanipulation. IEEE/ASME Trans. on Mechatronics, 9:358—365,
June 2004.

[79] Guangyong Li, Ning Xi, Heping Chen, Craig Pomeroy, and Mathew Prokos.
”videolized” atomic force microscopy for interactive nanomanipulation and
nanoassembly. IEEE Transactions on Nanotechnology, 4:605—615, 2005.

[80] B. Mokaberi and A. A. G. Requicha. Towards automatic nanomanipulation:
drift compensation in scanning probe microscopes. In Proc. IEEE Int. Conf.
Robotics and Automation), pages 416—421, New Orleans, April 2004.

130

[811

[82]

[831

[84]

[851

[86]

[871

[88]

[89]

[90]

[91]

[921

M. Guthold, M. R. Falvo, W. G. Matthews, S. Paulson, S. washburn, and D. A.
Erie. Controlled manipulation of molecular samples with the nanomanipulator.
IEEE/ASME Trans. on Mechatronics, 5(2):189—198, June 2000.

S.R.Manalis, S. C. Minne, and C. F. Quate. Atomic force microscopy for high
speed imaging using cantilevers with an integrated actuator and sensor. Applied
Physics Letters, 68:871—873, 1996.

T. Sulchek, S. C. Minne, J. D. Adams, D. A. Fletcher, A. Atalar, C. F. Quate,
and D. M. Adderton. Dual integrated actuators for extended range high speed
atomic force microscopy. Applied Physics Letters, 75:1637—1639, 1999.

T. Sulchek, R. Hsieh, J. D. Adams, G. G. Yaralioglu, S. C. Minne, C. F. Quate,
J. P. Cleveland, A. Atalar, and D. M. Adderton. High—speed tapping mode
imaging with active qcontrol for atomic force microscopy. Applied Physics Let-
ters, 76:1473—1475, 2000.

G. Y. Li, N. Xi, and M. Yu. Calibration of afm based nanomanipulation system.
In Proc. IEEE Int. Conf. Robotics and Automation, New Orleans, LA, April
26-May 1 2004.

E. Volterra and EC. Zachmanoglou. Dynamics of Vibrations. Charles E. Mer-
rill, Ohio, USA, 1965.

Thomas Bailey and James E. Hubbard Jr. Distributed piezoelectric-polymer ac-
tive vibration control of a cantilever beam. J. Guidance, Control and Dynamics,
8:605—611, 1985.

Chonan S., Jiang Z.W., and Koseki M. Soft-handling gripper driven by piezo-
ceramic bimorph strips. Smart Materials and Structures, 52407—414, 1996.

SB. Choi and CH. Lee. Force tracking control of a flexible gripper driven by
piezoceramic actuator. ASME Journal of Dynamic Systems, Measurement, and
Control, 119:439—446, 1997.

Dong Sun and James K. Mills. Combined pd feedback and distributed
piezoelectric-polymer vibration control of a single-link flexible manipulator. In
Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems,
pages 667—672, Victoria, BC, Canada, Oct. 1998.

AN. Singh and Y.M. Ram. Dynamic absorption in a vibrating beam. Journal
of Mechanical Engineering Science, 217:187—197, 2003.

H. Kataura, Y. Kamazava, Y. Maniwa, I. Umezu, S. Suzuki, Y. Ohtsuka, and
Y. Achiba. Optical properties of single-wall carbon nanotubes. Synth. Met,
103:2555—2558, 1999.

131

[931

[941

[951

1971

[981

[991

[100}

[101}

[102]

[103]

R. Jacquemin, S. Kazaoui, D. Yu, A. Hassanien, N. Minami, H. Kataura, ,
and Y. Achiba. Doping mechanism in single-wall carbon nanotubes studied by
optical absorption. Synth. Met, 115:283— 287, 2001.

L. Kavan, P. Rapta, L. Dunsch, M. J. Bronikowski, P. Willis, and RE. Smal-
ley. Electrochemical tuning of electronic structure of single-walled carbon nan-
otubes: in-sutu raman and vis-nir study. J. Phys. Chem. B, 105:10764—10771,
2001.

S. M. Bachilo, M. S. Strano, C. Kittrell, R. H. Hauge, R. E. Smalley, and R. B.
Weisman. Structure-assigned optical spectra of single-walled carbon nanotubes.
Science, 298:2361—2366, 2002.

Yoshiyuki Miyamoto, Noboru Jinbo, Hisashi Nakamura, Angel Rubio, and
David Tomanek. Photodesorption of oxygen from carbon nanotubes. Physi-
cal Review B, 70z233408, 2004.

M. J. O’Connel, S. M. Bachilo, C. B. Huffman, V. C. Moore, M. S. Strano, E. H.
Haroz, K. L. Rialon, P. J. Boul, W. H. Noon, C. Kittrell, J. Ma, R. H. Hauge,
R. B. Weisman, and R. E. Smalley. Band gap fluorescence from individual
single-walled carbon nanotubes. Science, 297:593—596, 2002.

J.A. Misevich, R. Martel, Ph. Avouris, J.S. Tsang, S. Heinze, and J. Ter-
soff. Electrically induced optical emission from carbon nanotube fet. Science,
300:783—786, 2003.

Woong Kim, Ali Javey, Ryan Tu, Jien Cao, Qian Wang, and Hongjie Dai.
Electrical contacts to carbon nanotubes down to 1 nm in diameter. Appl. Phys.
Lett., 87:173101, 2005.

Norbert Nemec, David Tomanek, and Gianaurelio Cuniberti. Contact depen-
dence of carrier injection in carbon nanotubes: An ab initio study. Phys. Rev.
Lett., 96:076802, 2006.

Wenguang Zhu and Efthimios Kaxiras. Schottky barrier formation at a carbon
nanotubemetal junction. Appl. Phys. Lett., 892243107, 2006.

M. H. Yang, K. B. K. Teo, and W. I. Milne. Carbon nanotube schottky diode
and directionally dependent field—effect transistor using asymmetrical contacts.
Appl. Phys. Lett., 872253116, 2005.

Harish M. Manohara, Eric W. Wong, Erich Schlecht, Brian D. Hunt, and Pe-
ter H. Siege]. Carbon nanotube schottky diodes using ti-schottky and pt-ohmic
contacts for high frequency applications. Nano Letters, 521469—1474, 2005.

132

[104] Ali Javey, Jing Guo, Qian Wang, Mark Lundstrom, and Hongjie Dai. Ballistic
carbon nanotube field-effect transistors. Nature, 424:654—657, 2003.

[105] [online]. available: https://www.veecoprobes.com.

[106] Edward F. Crawley and Javier de Luis. Use of piezoelectric actuators as ele-
ments of intelligent structures. AIAA, 25:1373—1385, 1987.

[107] Stewart John Moorehead. Position and force control of flexiblemanipulators.
Master’s thesis, University of Waterloo, 1996.

[108] William L. Brogan. Modern Control Theory. Prentice Hall, Englewood Cliffs,
N.J., third edition, 1991.

[109] Peter Dorato, Chaouki Abdallah, and Vito Cerone. Linear-Quadratic Control:
An Introduction. Prentice Hall, Englewood Cliffs, N.J., 1995.

[110] G. F. Franklin, J. D. Powell, and A. Emami-Naeini. Feedback Control of Dy-
namic systems. Prentice Hall, Upper Saddle River, NJ, fourth edition, 2002.

[111] [Online]. Available: http://www.rtai.org.

[112] William Singhose and Neil singer. Initial investigations into the effects of input
shaping on trajectory. In Proceedings of the 1994 American Control Conference,
pages 2526—2632, Baltimore, MD, test 1994.

[113] Ho—Yin Chan, Ning Xi, Jiangbo Zhang, and Guangyong Li. Tuning semicon-
ducting properties of single carbon nanotube for fabrication of nano devices.
In Proceedings of the Ist IEEE International Conference on Nana/Micro Engi-
neered and Molecular Systems, pages 1410—1413, Zhuhai, China, Jan. 2006.

[114] Xiaodong Cui, Marcus Freitag, Richard Martel, Louis Brus, and Phaedon
Avouris. Controlling energy-level alignments at carbon nanotube/au contacts.
Nano Letters, 32783—787, 2003.

[115] Amitesh Maiti and Alessandra Ricca. Metal-nanotube interactions - binding
energies and wetting properties. Chemical Physics Letters, 39527—11, 2004.

[116] King W.C. Lai, Ning Xi, Jiangbo Zhang, Guangyong Li, and Hongzhi Chen.
Packaging carbon nanotube based infrared detector. In Proc. IEEE Int. Conf.
Nanotechnology, pages 778 — 781, Hong Kong, August 2007.

[117] Derek A. Stewart and Francois Leonard. Energy conversion efficiency in nan-
otube optoelectronics. Nano Letters, 522005, 219—222.

[118] G. Y. Guo and K. C. Chu. Linear and nonlinear optical properties of carbon
nanotubes from first-principles calculations. Phys. Rev. B, 692205416, 2004.

133

[119]

[120]

[121]

[122]

[123]

D. P. Neikirk, D. B. Rutledge, and W. Lam. Far—infrared microbolometer de-
tectors. International Journal of Infrared and Millimeter Waves, 52245—277,
1984.

HQ Liu. Quantum dot infrared detector. OPTO-ELECTRONICS REVIEW,
1121—5, 2003.

Joseph N. Zalameda and andWilliam P.Winfree. Investigation of uncooled mi-
crobolometer focal plane array infrared camera for quantitative thermography.
Journal of Nondestructive Evaluation, 2421—9, 2005.

Sanjay Krishna, Sarath D. Gunapala, Sumith V. Bandara, Cory Hill, and
David Z. Ting. Quantum dot based infrared focal plane arrays. Proceedings
of the IEEE, 95:1838—1852, 2007.

J .L. Tissot. Ir detector with uncooled focal plane arrays, state-of—the art and
trends. OPTO-ELECTRONICS REVIEW, 122105—109, 2004.

134