INFRARED SENSING SYSTEMS USING CARBON
NANOTUBE BASED PHOTODETECTORS
By
Hongzhi Chen

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Electrical Engineering
2012

ABSTRACT

INFRARED SENSING SYSTEMS USING CARBON
NANOTUBE BASED PHOTODETECTORS
By
Hongzhi Chen
Infrared (IR) sensing systems have versatile applications; however, low performance and
high cost of the conventional photodetectors have prevented their widespread utilization in
various fields. Carbon nanotube (CNT), a promising nano-material with excellent electrical
and optical properties, has potential to develop high performance IR detectors compared to
their conventional counterparts. However, there are three major difficulties that impede the
application of CNT based photodetectors for imaging systems. Firstly, there are challenges
in design, fabrication, and testing of CNT photodetectors because of their nano-scale size
and unique geometry. Secondly, the small diameter of the CNTs results in low fill-factor
(absorption area). Thirdly, it is difficult to fabricate large scale of photodetector array for
high resolution focal plane due to the limitations on the efficiency and cost of the manufacturing. The issues related to the design and fabrication of CNT based photodetectors were
addressed by configuring the device in field effect transistors with different metal and gate
structures. In addition, the theoretical foundations as well as the implementation schemes for
the development of nano-structure lens to improve absorption efficiency of IR detectors were
developed. The topics include the optical antennas that confine light into a sub-wavelength
volume, as well as photonic crystals to increase the effective absorption area. Furthermore,
a novel CNT based IR sensing system was developed. The experimental results showed
that the new IR sensing system can achieve the superb performance enabled by CNT based
photodetectors, and, at same time, to obtain high resolution and efficient imaging.

To my parents, my wife Chenyi Wang,
and to my son, for their love and support

iii

ACKNOWLEDGMENTS

I would like to express my most sincere appreciation to my advisor, Dr. Ning Xi, for his
expert guidance, generous encouragement and support for my research.
In addition, I would like to thank all my PhD committees: Dr. Timothy Grotjohn, Dr.
Fathi Salem, Dr. Tim Hogan, Dr. Stuart Tessmer. They offered me timely help and unfailing
support that improve the technical soundness and the presentation of this dissertation.
Furthermore, I would like to express my gratitude to all my colleagues, especially to Dr.
Jiangbo Zhang, Dr. King Wai Chiu Lai, Bo Song, Liangliang Chen, Ruiguo Yang, Jianguo
Zhao, and Yunyi Jia, for their support in the experiments and discussion.
Last but not the least, I want to thank my parents and my wife, Chenyi Wang. 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.

iv

TABLE OF CONTENTS

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Background and Literature Review . . . . . . . . . . . . . . . . .
1.2.1 Blackbody Radiation . . . . . . . . . . . . . . . . . . . . .
1.2.2 Infrared Detectors . . . . . . . . . . . . . . . . . . . . . .
1.2.3 Image Acquisition in Scanning and Staring Configuration .
1.2.4 Advantages of One Dimensional Semiconductors as Sensing
1.2.4.1 Reduced Light Reflection and Transmission . . .
1.2.4.2 Bandgap Engineering . . . . . . . . . . . . . . . .
1.2.4.3 Multiple Electron and Hole Pairs Generation . .
1.2.4.4 Reduced Scattering . . . . . . . . . . . . . . . . .
1.2.4.5 Higher Defect Tolerance . . . . . . . . . . . . . .
1.2.5 Carbon Nanotube Based Photodetectors . . . . . . . . . .
1.3 Objectives and Challenges . . . . . . . . . . . . . . . . . . . . . .
1.4 Organization of the Study . . . . . . . . . . . . . . . . . . . . . .

. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Element
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
1
2
2
5
7
9
10
10
13
14
14
14
19
21

Chapter 2 Carbon Nanotube Schottky Photodiodes
2.1 Design of the CNT Schottky Photodiode . . . . . . .
2.2 Symmetric Schottky Photodiodes . . . . . . . . . . .
2.3 Asymmetric Schottky Photodiodes . . . . . . . . . .
2.4 Chapter Summary . . . . . . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

23
23
26
36
38

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Chapter 3 Carbon Nanotube Field Effect Transistor based Photodetectors
3.1 Back-gate Au-CNT-Au Transistor . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Back-gate Ag-CNT-Ag Transistor . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Back-Gate Au-CNT-Ag Transistor . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Middle-Gate Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 Multi-Gate Transistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Detector Array using CNT based Transistors . . . . . . . . . . . . . . . . . .
3.7 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

42
46
51
55
59
65
76
78

Chapter 4 Spectral Responses of CNT Photodetectors . . . . . . . . . . . .
4.1 Spectral Response of SWCNT based Photodetectors . . . . . . . . . . . . . .

82
82

v

4.2
4.3
4.4

Electrical Breakdown Control . . . . . . . . . . . . . . . . . . . . . . . . . .
Spectral Response of MWCNT based Photodetectors . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

88
92
95

Chapter 5 Plasmonic Lens using Optical Antenna . . . . . .
5.1 Introduction of Optical Antenna . . . . . . . . . . . . . . . .
5.2 Design of Bowtie Antennas . . . . . . . . . . . . . . . . . . .
5.2.1 Insulator Thickness Dependent E-field Enhancement
5.2.2 Tip Angle Dependent E-field Enhancement Spectra .
5.3 Fabrication and Experimental Results . . . . . . . . . . . . .
5.3.1 Fabrication Process . . . . . . . . . . . . . . . . . . .
5.3.2 Experimental Results . . . . . . . . . . . . . . . . . .
5.4 Chapter Summary . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

97
97
101
102
106
109
109
111
116

Chapter 6 Photonic Crystal Cavity . . . .
6.1 Introduction of Photonic Crystal Cavity
6.2 Design and Fabrication . . . . . . . . . .
6.3 Experimental Results . . . . . . . . . . .
6.4 Chapter Summary . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

117
117
119
121
127

Nano-photodetector
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .
. . . . . . . . . . . . . .

128
129
133
137
140
142
143

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Chapter 7 Infrared Sensing Systems Using a Single
7.1 Compressive Sensing Theory . . . . . . . . . . . . .
7.2 Sensing System Architecture . . . . . . . . . . . . .
7.3 Experimental Results . . . . . . . . . . . . . . . . .
7.3.1 Dynamic Observation . . . . . . . . . . . . .
7.3.2 Zooming . . . . . . . . . . . . . . . . . . . .
7.4 Chapter Summary . . . . . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Chapter 8 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . 146
8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
8.2 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Bibliography . . . . . . . . . . . . . . .

vi

. . . . . . . . . . . . . . . . . . . 151

LIST OF TABLES

Table 1.1

Performance Comparison of Photon and Thermal Detectors

. . . .

7

Table 2.1

Performance of Four Detectors with Different Metals . . . . . . . . .

35

Table 4.1

Dark Current and Noise at Different Gate Voltages

84

vii

. . . . . . . . .

LIST OF FIGURES

Figure 1.1

Energy distribution of a blackbody at three temperatures. The dot
line indicates the Wien’s law. (For interpretation of the references
to color in this and all other figures, the reader is referred to the
electronic version of this dissertation.) . . . . . . . . . . . . . . . . .

4

Figure 1.2

The classification of IR detectors. . . . . . . . . . . . . . . . . . . .

6

Figure 1.3

Staring configuration with a 2D focal plane array. . . . . . . . . . .

8

Figure 1.4

(a) Raster scanning system. (b) Parallel scanning system.

. . . . .

9

Figure 1.5

Two-diameter Ge nanowire array for maximized absorption. (a) The
structure of the two-diameter nanowire array (called double nanopillar (DNPL)). (b) Average absorption efficiency over λ=300-900 nm
for single-diameter nanowire arrays as a function of diameter along
with that of the double-diameter nanowire array. [3] . . . . . . . . .

11

Figure 1.6

The structure of a CNT thin film photodetector. . . . . . . . . . . .

15

Figure 1.7

The structure of a silicon-CNT heterojunction photodetector with
nanoporous anodic aluminum oxide (AAO) template to separate the
vertical CNT array. . . . . . . . . . . . . . . . . . . . . . . . . . . .

17

Figure 1.8

The structure of a single CNT based p-n junction photodiode. . . .

18

Figure 1.9

The structure of a single-CNT-based Schottky photodiode. . . . . .

20

Figure 2.1

(a) Energy alignment of Au and CNT. (b) Band diagram of a symmetric CNT photodetector. . . . . . . . . . . . . . . . . . . . . . . .

25

(a) Energy alignment of metals with high/low workfunctions and
CNT. (b) Band diagram of the asymmetric CNT photodetector. . .

27

An AFM image of a symmetric CNT detector with Au electrodes. .

28

Figure 2.2

Figure 2.3

viii

Figure 2.4

(a)I-V characteristics of CNT Schottky diodes with four different metals (Pd, Au, Cu and Al). (b)The I-V characteristics shown in logarithm scale. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

30

Figure 2.5

I-V characteristics of a Au-CNT-Au diode at different temperatures.

31

Figure 2.6

The IR responses of a Au-CNT-Au Schottky photodiode by radiating
on (a) left electrode, (b) right electrode. . . . . . . . . . . . . . . . .

32

The IR responses of two CNT Schottky photodiodes with metal structures of (a) Pd-CNT-Pd, (b) Au-CNT-Au. . . . . . . . . . . . . . .

33

The IR responses of two CNT Schottky photodiodes with metal structures of (a) Cu-CNT-Cu, (b) Al-CNT-Al. . . . . . . . . . . . . . . .

34

The IR responses of CNT Schottky photodiodes with bias (a) CuCNT-Cu, (b) Au-CNT-Au. . . . . . . . . . . . . . . . . . . . . . . .

37

(a-c) Fabrication process of the Au-CNT-Au and Al-CNT-Au photodiodes using a single CNT. (d) Their I-V characteristics. . . . . . . .

39

Photo-response of the Au-CNT-Au and Al-CNT-Au photodiodes using a single CNT. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

40

Figure 2.7

Figure 2.8

Figure 2.9

Figure 2.10

Figure 2.11

Figure 3.1

An AFM image of the structure of CNTFET with a silicon back-gate. 45

Figure 3.2

Structure of the back-gate Au-CNT-Au CNTFET. . . . . . . . . . .

47

Figure 3.3

(a)The I-V characteristics of the CNTFET. (b) Transfer characteristics of the CNTFET at Vds=50 mV. . . . . . . . . . . . . . . . . .

48

(a) Transfer characteristics of the Au-CNT-Au CNTFET with IR
source off (IR=0) and on (IR=50mW) at zero bias. (b) The relationship between gate voltage and photocurrent and corresponding band
diagrams at negative and positive gate voltages. . . . . . . . . . . .

49

Figure 3.5

The energy levels of Au and CNT. . . . . . . . . . . . . . . . . . . .

50

Figure 3.6

The bias-dependent measurement of the Au-CNT-Au CNTFET with
IR=0 and IR=50 mW at (a) Vg=-0.5V and (b)Vg=2V. . . . . . . .

52

The energy levels of Ag and CNT. . . . . . . . . . . . . . . . . . . .

53

Figure 3.4

Figure 3.7

ix

Figure 3.8

(a) I-V characteristics of the CNTFET. (b) Transfer characteristics
of the CNTFET at Vds=50 mV. . . . . . . . . . . . . . . . . . . . .

54

(a) The relation between gate voltage and photocurrent of the AgCNT-Ag CNTFET. (b) Bias dependent measurement with IR at four
different output powers at Vg=3 V. . . . . . . . . . . . . . . . . . .

56

Figure 3.10

Structure of the back-gate Au-CNT-Ag CNTFET. . . . . . . . . . .

57

Figure 3.11

(a) Transfer characteristic of the asymmetric CNTFET with and
without IR illumination. (b) Temporal photo-response with three
different gate voltages at IR=50 mW. . . . . . . . . . . . . . . . . .

58

Bias dependent measurement with IR at four different output powers
at Vg=5 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

59

(a) Structure of the middle-gate CNTFET. (b) AFM image of the
middle-gate CNTFET. (c) Band diagram of the working status of
the middle-gate CNTFET. . . . . . . . . . . . . . . . . . . . . . . .

61

(a) I-V characteristics of the middle-gate CNTFET at three Vg in
log-scale. (b) Transfer characteristics of the CNTFET with Vds=10
mV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

62

(a) Transfer characteristics of the middle-gate CNTFET with IR on
and Off at zero bias. (b) Temporal photo-response measurement by
switching IR on and off at 0.2 HZ. . . . . . . . . . . . . . . . . . . .

64

Bias dependent measurement with IR at two output powers at Vg=2
V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65

AFM image of the multi-gate asymmetric CNTFET with a back-gate
and three pairs of side-gates that can locally modulate source (Gate1),
SWCNT channel (Gate2), and drain (Gate3). . . . . . . . . . . . . .

67

The I-V characteristics of the CNTFET with symmetric (Au-CNTAu) and asymmetric (Al-CNT-Au) metal structures. The insets shows
the cross-section of the CNTFET. . . . . . . . . . . . . . . . . . . .

69

Figure 3.9

Figure 3.12

Figure 3.13

Figure 3.14

Figure 3.15

Figure 3.16

Figure 3.17

Figure 3.18

x

Figure 3.19

Transfer characteristics versus the back-gate bias at zero bias of the
symmetric CNTFET with absence and presence of IR laser illumination (wavelength of 830 nm and output power of 50 mW) by sweeping
the gate bias from -3 V to 3V. Two band diagrams show the Schottky
barriers at negative and positive gate biases. . . . . . . . . . . . . .

71

Transfer characteristics at zero bias of the asymmetric CNTFETs
with absence and presence of IR illumination. The inset shows the
transfer characteristics in logarithm scale. The band diagram shows
the energy alignment of the device. . . . . . . . . . . . . . . . . . .

72

Photocurrent transfer characteristics of the asymmetric CNTFET at
zero drain to source bias. The band diagrams show the barrier width
(w) at negative and positive gate biases. . . . . . . . . . . . . . . . .

73

Temporal photocurrent measurement of the asymmetric CNTFET
with three different gate biases (-3V, -1V, and 0V) by switching IR
OFF and ON at 0.2 Hz. . . . . . . . . . . . . . . . . . . . . . . . . .

74

(a) Photocurrent transfer characteristics by applying gate biases to
back-gate (Vgs), source (Vg1), SWCNT channel (Vg2), and drain
(Vg3). (b)A 3D geometry of the multi-gate CNTFET and the charge
distribution of the SWCNT, showing the charge tail on the SWCNT.

77

Figure 3.24

Asymmetric CNTFET with a simplified gate structure. . . . . . . .

78

Figure 3.25

The image of a 3x3 detector array and the AFM image of the Pixel9.

79

Figure 3.26

The images obtained from the detector array, and color lightness
shows the magnitude of photocurrent. The laser spot center is on
(a)Pixel5, (b)Pixel1. . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

Figure 4.1

The blackbody testing system. . . . . . . . . . . . . . . . . . . . . .

83

Figure 4.2

(a) The I-V characteristics of the transistor with Vg from -3V to 3V.
(b) The relationship between Vg and current, showing conductivity
decreases with increasing positive gate voltages. (c) The temporal
dark current measurement with three different gate voltages. . . . .

85

The temporal IR response by radiating blackbody source to the detector at 0.5 Hz with (a) Vg=0 V, (b) 1 V , (c) 2 V and (d) 3 V. . .

86

Figure 3.20

Figure 3.21

Figure 3.22

Figure 3.23

Figure 4.3

xi

Figure 4.4

The spectral response of the SWCNT based photodetector. . . . . .

87

Figure 4.5

Breakdown (BD) control curve. . . . . . . . . . . . . . . . . . . . . .

90

Figure 4.6

The I-V characteristics of the device before and after BD. . . . . . .

91

Figure 4.7

(a) Illustration showing a single MWCNT connected to three gold
microelectrodes. The I-V characteristics of the MWCNT at different
segments: (b) Segment AB, (c) segment BC, and (d) segment AC. .

93

The I-V characteristics of a MWCNT photodetector before and after
consecutive breakdowns (BD). . . . . . . . . . . . . . . . . . . . . .

94

Figure 4.9

The spectral response of the MWCNT based photodetector. . . . . .

96

Figure 5.1

(a) A 3D geometry shows the relative position between the CNT
based photodetector and the bowtie antenna. The incident light with
electric field polarized in the x direction propagates in the z direction.
(b) Top view of the device, and parameters of the bowtie antenna:
L=400 nm is the length of the antenna, g=30 nm is the gap of two
pieces of the antenna, w=30 nm is the apex width, and is the tip angle.
(c) Side view of the device, showing the CNT based photodetector
and bowtie antenna sandwiching the Si3 N4 layer that has a thickness
of t. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

Figure 5.2

(a) The E-field distribution of the resonance frequency (845 nm) of
the bowtie antenna with θ=60 degrees, and t=10 nm. The x-y axis
show the coordinate for simulation. (b) The t dependent peak field
intensity Max(| E |2 ), blue curve is the simulation values, and red
curve is the exponential fit. The inset shows field intensity enhancement | E |2 at wavelength of 830 nm. . . . . . . . . . . . . . . . . . 105

Figure 5.3

E-field intensity enhancement | E |2 spectra of bowtie antennas with
tip angles ranging from 0 degrees to 120 degrees. The inset shows
a scanning electron microscopy (SEM) image of a bowtie antenna,
denoting the tip angle θ. . . . . . . . . . . . . . . . . . . . . . . . . 107

Figure 5.4

(a-e) Fabrication process of integrating a bowtie antenna to a CNT
based photodetector. The inset of (c) is an AFM image of a SWCNT
bridging a Ti and an Au electrode. (f) A 3D AFM image of the device.112

Figure 5.5

Photo-responses of the CNT based photodetector without and with
bowtie antenna by switching IR laser on and off at 0.2 Hz. . . . . . 113

Figure 4.8

xii

Figure 5.6

I-V characteristics of the CNT photodetector without and with the
bowtie antenna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

Figure 6.1

(a) AFM image of a parylene photonic crystal with a HfO2 defect on
the top of a CNT detector. (b) Cross section of the device. (c) The
working principle of the photonic cyrstal cavity by confining the light
radiating on photonic crystal into the defect. . . . . . . . . . . . . . 120

Figure 6.2

(a)The simulation geometry of the photonic cavity with parylene slab
and a HfO2 defect. (b) Electric field distribution of a 1064 nm IR light
in the photonic cavity, most of the electric field is confined within the
point defect. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Figure 6.3

Fabrication process of the testing device (a) CNT Schottky photodiode, (b) CNT photodetector with a HfO2 defect,(c) Photodetector
with photonic crystal and a HfO2 defect. . . . . . . . . . . . . . . . 123

Figure 6.4

Experimental setup to obtain scanning photocurrent images. . . . . 125

Figure 6.5

Scanning photocurrent images of the CNT photodetectors without
and with photonic crystal cavity. The insets are the AFM images of
detectors without and with photonic crystal cavity. . . . . . . . . . 126

Figure 7.1

(a) Scanning electron microcopy (SEM) image of an array of micromirros in the DMD. (b) Three status of the micro-mirrors: “on”,
“off”, and power-off states. . . . . . . . . . . . . . . . . . . . . . . . 134

Figure 7.2

The setup of compressive sensing based imaging system using a CNT
detector with photonic cavity and DMD. . . . . . . . . . . . . . . . 136

Figure 7.3

(a) The position and orientation of the rectangular light bar on the
DMD. (b) The image of 50×50 pixels recovered by our imaging system, highlighted by a red rectangle. . . . . . . . . . . . . . . . . . . 139

Figure 7.4

Six images of a laser spot captured by our imaging system, showing
the movement of the spot.(a) The infrared image shows the initial
position of the laser spot. (b)-(e) The laser spot moved downwards
towards the bottom of the image. (f) The laser spot moved upwards
after it approached the bottom of the DMD. . . . . . . . . . . . . . 141

Figure 7.5

×1, ×2, ×8, ×12 magnification IR images captured by our imaging
system, and their corresponding sections of DMD used to capture the
images were highlighted by blue rectangles. . . . . . . . . . . . . . . 144
xiii

Chapter 1
Introduction

1.1

Overview

Infrared (IR) sensing systems, including IR cameras and IR detection systems, were widely
employed in various areas, including military, manufacturing quality control, medical diagnosis, weather forecasting, environmental monitoring, and so on. The core component of
IR sensing systems is the photodetector, which detects heat distribution, rather than visible
light, according to the blackbody radiation theory. IR spectrum, especially middle-wave IR
(MWIR), is of particular importance, since electromagnetic emission from humans and the
atmosphere are within this spectrum.
The photodetectors mainly determine the performance of IR cameras. There are two
categories of traditional photodetectors, including thermal detectors and photon detectors.
Both detectors have their own limitations: thermal detectors have slow response and low
detectivity; while photon detectors have narrow and selective spectral detection, and need
to operate at cryogenic temperature. Therefore, an IR sensing system with fast response,
low detectivity, multi-spectral detection, and room temperature operating is demanding.
With the development of nano-technologies, one-dimensional (1D) materials, including
nanowires and nanotubes, emerged as important components in photonics and optoelectronics. Compared to traditional semiconductors, 1D materials possess a number of unique
properties arose from their geometries, making them potential candidates to outperform
1

traditional photodetectors. In particular, carbon nanotubes (CNTs) are of great interest
for exploring as functional elements for 1D photodetectors due to their perfect nano-hollow
cylinder structure. Assembling a CNT thin film has been demonstrated, but the interaction
between CNTs in thin film resulted in the loss of their unique optical properties. Therefore,
one single CNT is more favorable to fabricate high performance photodetectors.
However, there is no reliable manufacturing technique to fabricate large scale CNT based
photodetectors, preventing the integration of CNT photodetectors into traditional sensing
systems, particularly the focal plane array imaging system. An alternative sensing platform
is required to utilize small number or even a single CNT photodetector. Aiming to taking
advantages of the outstanding optoelectronic properties of CNTs, we developed an IR sensing
system that integrated a CNT photodetector into a novel camera structure that can use a
few or even a single photodetector. The sensing system can achieve the superb performance
enabled by CNT based IR detectors, and, at same time, to obtain high resolution and efficient
imaging.

1.2
1.2.1

Background and Literature Review
Blackbody Radiation

All matter emits electromagnetic (EM) waves at temperature above absolute zero (0 K).
This is known as thermal radiation. Thermal radiation is also called IR radiation, which
covers a wide EM spectrum ranged from 0.77 µm to 1000 µm.
Blackbody is an idealized entity that absorbs all radiation falling on it, but reflects none.
Blackbody radiation is the emission of a blackbody at unity efficiency. Therefore, blackbody
emission has a temperature-dependent characteristic frequency distribution when it is at
2

a uniform temperature. Blackbody radiation provides a convenient theoretical model to
predict emitting properties of objects.
The Planck’s law describes blackbody’s energy distribution of different wavelengths at a
given temperature:

Wλ =

2πhc2
hc

λ5(e λkT

− 1)

(1.1)

h is the Planck constant, c is the speed of light, λ is the wavelength, k is the Boltzmann
constant, and T is the absolute temperature of the blackbody.
The Wien’s displacement law predicts the wavelength (λmax ) that delivers the highest
intensity of radiation at a given temperature:

λmax =

b
T

(1.2)

b is the Wien’s displacement constant with a value of 2.898×10−3 m·K. The λmax decreases with increasing temperature.

The Stefan’s law estimates the total energy radiated in an given area per unit time at a

certain temperature:

W = σAT 4
3

(1.3)

10

Spectral Exitance (J/m2/m)

Spectral Exitance (J/m /m)
Spectral Exitance
Spectral Exitance
Spectral Exitance (J/m (J/m
Spectral Exitance (J/m2 (J/m

33

10
10

T=300 K
T=300 K
T=600 K
T=600 K
T=900 K
T=900 K

22

10

1
10

10

1

00

10
10

-1
-1

10
10

-2
-2

10
10

-3
-3

10
10

10

-6

-6

10

-5

10-5

-4

10-4

10
10 10
Wavelength (m)

10
10

Wavelength (m)

Wavelength (m)
Wavelength
Waveleng
Figure 1.1: Energy distribution of a blackbody at three temperatures. The dot line indicates
the Wien’s law. (For interpretation of the references to color in this and all other figures,
the reader is referred to the electronic version of this dissertation.)

σ is the Stefan’s constant with a value of 5.67×10−3 Jm−2 K −4 , A is the surface area of
the body. Total radiation energy is proportional to T 4 , thus a minor increase of blackbody
temperature will involve a relative major change of the total energy.
Figure 1.1 shows the blackbody radiation at three temperatures according to the Planck’s
law. The dot line reflects the Wien’s law. The peak radiation wavelength at 300 K is at 10
µm, which cannot be seen by human eyes. The peak radiation wavelength blueshift with
increasing temperature, gradually approaching visible light. The radiation of a real object
can be calculated according to these three laws, adjusted by the emissivity ε. ε is the ratio
of energy radiated by a real object to that of the blackbody radiation.

4

1.2.2

Infrared Detectors

The core component of the IR sensing systems is the IR detector. IR detectors are photodetectors that convert electromagnetic signals within IR spectrum into other forms of signals
(current or voltage) for detection and quantification.
The IR detectors can be mainly categorized into two classes: thermal detectors and photon detectors. Photon detectors generate electrical outputs through the direct interaction
between photons and electrons, photo-exciting bound electrons of semiconductors from valence band to conduction band. Thermal detectors change the temperature of photosensitive
materials by absorbing incident radiation, resulting in some physical properties change that
converts into electrical outputs.
Depending on the working principles, photon and thermal detectors can be sub-divided
into different types, as shown in Figure 1.2. Three main types of photon detectors are
photovoltaic, photoconductive, and photoemissive detectors. Photovoltaic detectors utilize
their internal built-in potential to separate photo-excited electron and hole pairs in order
to generate current or voltage for detection, for example the p-n junction. Photoconductive
detectors detect the change of conductivity caused by the increase of photo-excited electrical
carriers, and external bias are required to operate. Photoemissive detectors collect electrons
escaped from a photosensitive surface, which absorbs IR radiation and delivers enough kinetic
energy for electron escaping.
Three main types of thermal detectors are thermistors, pyroelectric detectors, and thermovoltaic detectors. Thermistors sense the conductivity change of the functional element
due to the change of temperature upon the absorption of IR photons, and bolometer is one of
thermistors. Pyroelectric detectors measure the electric field generated by the temperature-

5

IR Detectors

Photon Detectors

Thermal Detectors

Photovoltaic

Thermister

Photoconductive

Pyroelectric

Photoemissive

Thermovoltaic

Figure 1.2: The classification of IR detectors.
dependent dipole moment change of the responsive elements. Thermovoltaic detectors are
also called thermocouples or thermopiles. Upon IR illumination, the temperature change of
the junction of two dissimilar metal produces a voltage for detection due to the difference of
Seebeck coefficients of two metals.
Table 1.1 compares the performance of photon and thermal detectors. Thermal detectors
require heating the detection element in order to detect IR signals, therefore they have
an inherently slow response and relatively low detectivity compared to photon detectors.
However, thermal detectors can be operated at room temperature and have a wide spectral
response, while photon detectors have narrow and selective spectral response depending on
their bandgap energies and most of them need to be operated in cryogenic temperature. As
a result, the manufacturing of photon detectors are more expensive than thermal detectors.
Our research will focus on developing an IR sensing system using CNT based photon
detectors, because photon detectors using nano-materials may address the problems of the

6

Table 1.1: Performance Comparison of Photon and Thermal Detectors

Photon Detectors
Speed
Operation Temperature
Detectivity
Spectral Response
Cost

Thermal Detectors

Fast
Cryogenic Cooled
High
Narrow
Expensive

Slow
Room Temperature
Low
Wide
Economical

narrow spectral response and requirement of cryogenic cooling.

1.2.3

Image Acquisition in Scanning and Staring Configuration

In IR sensing systems, in particular IR cameras, the photodetectors must be distributed over
the focal plane (image plane) in order to capture the spatial IR images. The photodetector
distribution across the entire image can be accomplished by systems in scanning or staring
configurations [1]. Scanning configuration scans the image across a single photodetector or a
linear detector array. Staring configuration has a two-dimensional (2D) photodetector array
distributing in the focal plane (focal plane array), as shown in Figure 1.3.
Scanning configuration can be sub-divided into raster and parallel scanning systems. In
the raster scanning system, a scanner moves the scene over a single or a few fixed photodetectors in the focal plane, as shown in Figure 1.4(a). The photodetectors are aligned
horizontally and scanned sequentially in a 2D raster pattern to form an image. The parallel
scan system comprises a linear photodetector array that is perpendicular to the scanning
direction, as shown in Figure 1.4(b). It should be noted that the linear photodetector array
can have more than one columns of photodetectors. Figure 1.4(b) only shows one column
of photodetectors for making the schematic clearer. The scanner horizontally scans one line

7

lens

2D focal plane array

Figure 1.3: Staring configuration with a 2D focal plane array.
across the linear array, and then a slight tilt of the scanner in the vertical direction is applied
and the image is scanned across the photodetector array in the opposite direction to cover
the image plane in order to obtain a complete image. There are a wide variety of scanners
to implement the scanning, and a mechanical mirror is the most efficient and well developed
one.
Imaging systems with the staring configuration can result in enhanced performance and
lighter system. The scanning system can be used in applications where large array of photodetectors are very expensive or difficult to manufacture.

8

scanning pattern

scanning pattern

image size
photodetector

(a)

(b)

Figure 1.4: (a) Raster scanning system. (b) Parallel scanning system.

1.2.4

Advantages of One Dimensional Semiconductors as Sensing
Element

Nanowires and nanotubes are 1D materials, since electrons in these nano-size cylinders was
freely to move in axial direction, while electrons are quantum confined in radial directions
when their diameters are smaller than Bohr radii. 1D semiconductors were proposed to be
the functional element of high performance photodetectors and solar cells, since they have
some unique properties stemming from their geometry and quantum confinement [2].
It should be noted that our goal is developing an IR sensing system with high performance
photon detectors. In photon detectors, there are many sources of losses in the process of
converting IR signals into electrical signals: (1) photons are reflected and transmitted instead
of being absorbed; (2) photons with energy that is lower and higher than bandgap energy
causes some energy losses; (3) photo-excited electron and hole pairs recombine; (4) the
electrical carriers were scattered during transportation. These losses can be reduced by using
1D semiconductors as the sensing element.The potential efficiency benefits mainly come from

9

higher efficiency in photon absorption and trapping; improvement in photo-generated carrier
separation, transportation and collection; lowering the purity standard needed to obtain high
efficiencies; enhancing the defect tolerance.

1.2.4.1

Reduced Light Reflection and Transmission

It was reported that both light transmission and reflection were reduced by using Ge nanowires
compared to Ge thin film. As a result, the absorption was highly improved [3]. A double
diameter platform was designed to optimize absorption, with small diameter tip and large
diameter base, as shown in Figure 1.5(a). Small diameter tips were used to minimize reflectance, while large diameter bases were utilized to suppress transmittance.
The broad band absorption enhancement for the double-diameter nanowire array (double
nanopillar (DNPL)) is depicted in Figure 1.5(b). For single-diameter nanowire array, a
strong diameter dependency is observed, with a peak absorption 94% when diameter =80
nm. Double-diameter nanowire array with D1= 60 nm and D2=130 nm exhibits absorption
of 99%, clearly demonstrating the advantage of this novel structure. Notably, both singlediameter and double-diameter nanowire array drastically outperform a 2 µm-thick planar
Ge thin film (TF) absorption 53%.

1.2.4.2

Bandgap Engineering

When the diameter of a 1D semiconductor is smaller than its Bohr radii, bandgap energy
is inversely proportional to its diameter due to the quantum confinement. In other words,
the band gap energy of 1D materials can be modulated by controlling their diameters rather
than using different semiconductors.
When an electron approaches a hole, they may form a Wannier exciton, approximately

10

(a)

(b)
Figure 1.5: Two-diameter Ge nanowire array for maximized absorption. (a) The structure of
the two-diameter nanowire array (called double nanopillar (DNPL)). (b) Average absorption
efficiency over λ=300-900 nm for single-diameter nanowire arrays as a function of diameter
along with that of the double-diameter nanowire array. [3]

11

described by a hydrogenic Hamiltonian [4]:

H=

− ∗ ▽2
2mh h

− ∗ ▽2
2me e

q2
−
ε | re − rh |

(1.4)

is the Planck constant. m∗ and m∗ are the effective masses of hole and electron. ε is the
e
h
dielectric constant. By solving the equation using this model, the relation of the diameter
of the 1D semiconductors and band gap energy can be expressed as [5]:

1
1.8q 2
π 1
(
+
)−
Egn = Egp +
2d2 m∗ m∗
εd
e
h
2 2

(1.5)

Egp is the bandgap energy of planar material. Egn is the bandgap energy of the quantum
confined semiconductor. d is the diameter of the 1D semiconductor. The bandgap of a
semiconducting CNT is give by the equation [6–9]

Egc =

where vf is the fermi velocity, and

4 vf
3dCN T

(1.6)

is the Dirac constant. It shows that bandgap energy

is inversely proportional to the diameter.
One type of CNT has a multi-wall structure, comprising of multiple hollow cylinders with

12

different diameters, called multi-wall CNT (MWCNT). The diameter of a MWCNT can be
easily tailored by electrical breakdown. It has been demonstrated that the peak responsivity
of the MWCNT photodetectors can be tuned to the desired wavelength by controlling the
diameter of MWCNTs through electrical breakdown [10].

1.2.4.3

Multiple Electron and Hole Pairs Generation

Absorption of a single photon typically creates only one electron and hole pair, resulting in
energy loss because excess energy from photons with energies greater than the semiconductor
bandgap is loss through phonon emission. However, absorption of a single photon with high
energy can potentially create more than one pair of electron and hole, since the excess energy
of a hot electron can excite other electros across their bandgaps.
The creation of multiple electron and hole pairs via absorption of a single photon is a wellknown phenomenon in planar semiconductors [11]. However, the efficiencies are insignificant
for bulk materials. It has been proposed that generation of multiple electron and hole pairs is
more efficient in quantum confined materials due to the well-separated discrete energy bands
in the quantum confined materials that should slow phonon-assisted carrier cooling (phonon
bottleneck). This makes multiple electron and hole pairs generation more competitive with
other relaxation pathways.
An efficient multiple electron and hole pair generation process was observed in a CNT
p-n junction photodiode [12]. In the CNT p-n junction, a single photon with energy larger
than E22 (energy of two times of band gap energy) is converted into multiple electron and
hole pairs, leading to enhanced photocurrent and increased efficiency. This process could
potentially result in higher photon-to-current conversion efficiencies.

13

1.2.4.4

Reduced Scattering

Once the photo-excited carriers were separated by built-in potential in the depletion regions,
they need to transport to the electrodes in order to be collected, and thus contribute to
photocurrent generation. The efficiency of photodetectors was reduced by the scattering of
carriers during the transportation to the electrodes. The efficiency can be increased if the
carrier scattering reduced.
It was demonstrated that scattering was reduced in a 1D material based device: ballistic transport (no scattering) was observed in a CNT based field effect transistor (FET) at
room temperature [13]. Contacting semiconducting CNT by Pd greatly reduced or eliminated the barriers for transport through the valence band of nanotubes. The ’ON’ states of
FET can behave like ohmic contacted ballistic metallic tubes, exhibiting room temperature
conductance near the ballistic transport limit of 4q 2 /h and high current-carrying capability.

1.2.4.5

Higher Defect Tolerance

Another advantage of the 1D photodetectors is the suppressed sensitivity to defects. Simulation results predicted that 1D materials with radial p-n junctions show only a slight efficiency
degradation with decreasing lifetime until the minority carrier diffusion length approaches
the nanowire radius, whereas planar p-n junction shows severe degradation once the diffusion
length drops below the film thickness [14].

1.2.5

Carbon Nanotube Based Photodetectors

The basic theory of planar semiconductors cannot directly apply to the 1D structures due to
the quantum effects [15], thus the behavior of CNT based photodetectors may be different

14

Light radiation
CNT film

Electrodes
A

Figure 1.6: The structure of a CNT thin film photodetector.
from traditional photon detectors. The photoconductivity of CNTs has been extensively
studied, and photodetectors with a number of different structures were reported. A review
of different photodetectors will be given as follow.
By depositing CNT thin film between two pre-patterned electrodes with insulating gap,
infrared detection was first realized by observing the conductance change with infrared radiation on the device [16], as shown in Figure 1.6. However, not only the mixture of metallic
and semiconductor CNTs in the film severely degraded the response of the thin film photodiode, but also the adsorbed oxygen in the film decreased the photo-sensitivity. In order
to improve the performance of the thin film detector, CNT was embedded into the polycarbonate nanocomposite [17]. This insulating polymer matrix dramatically enhanced the
photo-response, since the nano-composite changed the detection mechanism from a thermal
effect into a photo effect.

15

Instead of connecting CNT film to the metal electrodes, detectors can also be fabricated
by stacking CNT film onto the bulk silicon substrate [18]. In this structure, the contact
between CNTs and silicon form heterojunctions, which are responsible for the photocurrent
generation. Nano-optoelectronic devices using films or sheets will lose some particular properties and advantages due to the interaction between nanotubes. By growing CNTs on a
nanoporous anodic aluminum oxide (AAO) template, the interaction of CNTs can be eliminated [19], as shown in Figure 1.7. This synthesis method will form a periodic dense array
of CNTs that are parallel to each other and perpendicular to the silicon substrate. In this
heterojunction system, several mechanisms may contribute to the photo-response. It is very
difficult to fully understand its underlying operation mechanism, since the two materials are
very dissimilar in bandgap, workfunction, and dimensionality, and the synthesized CNTs
may have varied properties.
A single CNT is a perfect 1D system, therefore photodetectors using a single CNT is
more favorable, and easier controlled for manufacturing. Studying the single-CNT-based
photodetectors can help us understand the real operation principle of the 1D detectors,
which will lead to the fabrication of photodetectors with optimized performance.
Photoconductivity of a single-CNT-based p-n junction photodiode was observed by fabricating two split gates underneath both ends of a semiconductor CNT [20], as illustrated in
Figure 1.8. By applying opposite gate voltages with high amplitude to the split gates, the
CNT at both ends will be electrostatically doped into p and n type semiconductor, resulting
in the formation of a p-n junction in the CNT. The fabrication begins with a thick thermal
oxide grown on a heavily doped Si wafer. Split gates were formed using Mo. Using the split
gate metal as an etch mask, part of oxide was dry etched. Next, a thin layer of silicon oxide
was deposited, forming the gate dielectric for the split gate. The source and drain contacts
16

Light radiation

Semi-transparent Au electrode

CNT array

A
Silicon

Figure 1.7: The structure of a silicon-CNT heterojunction photodetector with nanoporous
anodic aluminum oxide (AAO) template to separate the vertical CNT array.

17

Light radiation
Single CNT
Source

Drain

SiO2

SiO2

Gate1(++)

Gate2 (--)
SiO2

A

Figure 1.8: The structure of a single CNT based p-n junction photodiode.
were then lithographically defined. Finally, CNTs were grown on top of the source and drain
metals. In the center of the device, where the built-in field is greatest, photo-generated
electron and hole pairs were separated, causing currents to flow. This photodiode provided
some insights to the CNT detector, and showing some promising results [20].
However, the fabrication process of the device is too complicated, which will hinder its
applications in wide realm. Therefore, a CNT detector with simpler structure is needed.
CNT Schottky photodiode emerged as an alternative.
When metal and CNT come into contact, a Schottky barrier may be formed depending
on the Fermi energy alignment. The Schottky barrier determines the electrical and optical properties of the CNT devices. However, the importance of the Schottky barrier was
neglected at early stage of developing CNT based devices, when researches attributed the
electrical properties of the device to the intrinsic properties of the CNT [21] rather than the

18

contacts. With the development of CNT field effect transistors (CNTFETs), increasing evidence suggested that the Schottky barrier played a key role in determining the characteristics
of the nano-devices [22].
It was demonstrated that Schottky diodes between an individual SWCNT and metals
were responsible for the photocurrent generation under continuous infrared illumination [23].
Figure 1.9 shows the structure of Schottky photodiode that has a much simpler configuration
than the p-n junction photodiode.
The main reason of generating photocurrent in the CNT Schottky barrier photodiode is
the depletion regions of the barrier. Photons with energy higher than the band gap of CNT
can produce electron and hole pairs or excitons in the entire CNT, whereas the current flow
is only distributed to the depletion region, since the built-in potential in the depletion region
(and within diffusion length) can separate the photo-generated electron and hole pairs, and
thus produce photocurrent, while the electron and hole pairs outside the depletion region
will recombine shortly.
The Schottky photodiodes have shown superb performance and the simplest structure.
However, there was no systematic study about optimizing the design of CNT based Schottky
photodiodes. The performance of CNT based photodetectors is extremely important to the
IR sensing systems. We will focus on investigating the Schottky barrier based photodetectors
of single CNTs, and optimizing the structure of the photodetectors.

1.3

Objectives and Challenges

The objective of this research is to develop a high performance IR sensing system that
utilizes a CNT based photodetector as its core component. The system development involves

19

Light radiation
A single CNT

Electrodes
A
Figure 1.9: The structure of a single-CNT-based Schottky photodiode.
three main parts: (1) develop a CNT based photodetector with high performance and simple
fabrication process; (2) explore nano-structured light confinement components that integrate
with CNT based photodetectors, due to the sub-wavelength size of the photodetectors; (3)
develop a novel sensing system structure that can use a few or even a single phodetector.
In order to accomplish this work, not only the theoretical designs of photodetectors, light
confinement components, and camera structure were involved, but also fabrication and experimental testing were performed.
The main challenges involved in this research include:
• Design and fabrication of high performance photodetectors using a single CNT. Due
to nanometer-scale diameter of the CNTs, reliable and high-yield manufacturing was
difficult to accomplish in order to fabricate CNT based photodetectors. In addition,
the optimized photodetector design may be different from conventional devices due to

20

the unique quantum confined geometry of CNTs.
• CNTs with appropriate bandgap energies for MWIR detection. There was not a reliable
synthesis technique that can grow CNTs with uniform electrical and optical properties.
In particular, it is difficult to control the bandgap energy of CNT to detect MWIR.
• Light confinement to the CNT based photodetectors with sub-wavelength size. A
single CNT has a diameter of as small as sub-nanometer, while detection wavelength
is much larger than the diameter of the CNT, resulting in small fill factor (absorption
area). Traditional lens cannot address this problem due to the light diffraction. It is a
challenge to develop a light confinement component that can trap light onto the CNT
based photodetectors.
• Proper sensing system architecture to integrate with CNT based photodetectors. It is
difficult to fabricate large focal plane array of CNT based photodetectors. Conventional
camera structure addressed this problem using scanning configuration to scan each
pixel per cycle, making the sensing system slow and inefficient.

1.4

Organization of the Study

Design, fabrication and testing of CNT Schottky photodiodes are presented in Chapter2
to investigate the importance of metal structure. It is followed by developing CNT field effect
transistor (CNTFET) based photodetectors in Chapter3 to study the dependence of photoresponses on electrostatic doping. In Chapter4, spectral responses of SWCNT and MWCNT
based photodetectors are compared, and breakdown process is introduced to engineer the
bandgap energy of MWCNT. In Chapter5, bowtie antenna is described to improve the fill

21

factor of the CNT photodetectors. In Chapter6, photonic crystal cavity was utilized to
further increase the fill factor of the CNT photodetectors. Chapter7 introduces the novel
sensing system architecture that enables capture high resolution images using even a single
photodetector. Chapter8 presents the conclusions and future works.

22

Chapter 2
Carbon Nanotube Schottky
Photodiodes
Understanding the 1D Schottky barrier within CNT is important. The traditional Schottky
barrier theory for planar contact between metal and bulk semiconductor is well established,
and can lead to optimized photodiode design. However, the conventional understanding of
the Schottky barrier cannot be applied to 1D Schottky barriers between metal and nanotubes,
since the planar contacts are dominated by the metal-induced gap states (MIGS), whereas
the MIGS induced Fermi level pining is trivial for the 1D structure [15] due to quantum
confinement. Therefore the physics and fundamentals of 1D Schottky barriers within a CNT
need to be reexamined. In the following sections, I will introduce the design, fabrication,
testing and analysis of the CNT Schottky photodiodes so as to understand their working
principle.

2.1

Design of the CNT Schottky Photodiode

After photons were absorbed by semiconductor CNTs, photo-generated electrons will be
separated by the electric field within the Schottky barrier, and need to tunnel though the
barrier before contributing to the photocurrent. The Schottky barrier width determined
the effective absorption area, while the built-in potential within the barrier determined the
23

speed of separated carriers, and both barrier width and height defined the carrier tunneling
probability. The basic structure of a CNT Schottky photodiode is shown in Figure 1.9, in
which a single semiconductor CNT bridge two metal electrodes to form the Schottky barriers.
Figure 2.1(a) shows the energy alignment between a metal (take Au as example) and a
CNT. The CNT was reported as a p-type material in the air with Fermi energy of around
5 eV [24] due to oxygen doping or some accidental doping during the synthesis [25]. The
workfunction (Fermi energy) of Au is also close to 5 eV, and it can be slightly smaller or
higher than 5 eV due to the environmental differences [26] (in the Figure 2.1(a), the Fermi
energy of Au is slightly smaller than CNT). The majority carriers of the semiconductor CNT
are holes, therefore the energy difference between the Au workfunction to the CNT valence
band (Ev) is indicated as the barrier height. When CNT connect two Au electrodes at both
ends, denoted as symmetric photodetector, two symmetric Schottky barriers will be formed
at the contacts, as shown in Figure 2.1(b).
This structure results a photodiode that consists of two Schottky barriers inversely connecting to each other. When the photons with energy higher than the CNT bandgap strike
the CNT, electron and hole pairs will be generated and separated in the depletion region,
as indicated in Figure 2.1(b). Hence, the total photocurrent under zero bias is calculated by
Iphoto = Ilef t + Iright = ib1 − ib2 . These symmetric barriers significantly reduce the sensitivity of the photodetector, since the separated carriers need to tunnel through another barrier
before being collected. What is more, this structure detects the signal intensity difference
between two contacts, resulting in poor performance.
Asymmetric photodetectors, using metals with high/low workfunctions to connect two
ends of a single CNT, can be used to improve the performance of the detectors. The energy
alignment and resulting band diagram are shown in Figure 2.2. On the left contact, the
24

Metal

CNT

Ec

Ei
Au (Ef)
Barrier height

Ef
Ev

(a)

Photons
Schottky barriers

e-

e-

Au

Au

h+

h+

(b)
Figure 2.1: (a) Energy alignment of Au and CNT. (b) Band diagram of a symmetric CNT
photodetector.

25

metal has a workfunction that is smaller than the CNT and forms a Schottky barrier, while
on the right contact, the metal workfunction is higher than the CNT and forms an ohmic
contact. The electrical and optical properties of this device will be determined by one
Schottky barrier. Thus the total photocurrent will only depend on one Schottky barrier
rather than the difference between the photocurrent of two barriers as in the symmetric
structure. Moreover, by transferring one of the Schottky barriers into ohmic contact, the
photo-generated electron holes at the Schottky barrier will be able to transport through the
contact easily. As a result, the photo-response of the CNT photodetector is expected to be
highly enhanced in the asymmetric structure.

2.2

Symmetric Schottky Photodiodes

An AFM image of a symmetric CNT Schottky photodiode with Au electrodes is shown in
Figure 2.3. An individual single wall CNT (SWCNT) connects two Au electrodes on top of
a quartz substrate. The fabrication process begins with making two Au electrodes with an
insulating gap of approximately 1 µm by photolithography, thermal evaporation, and liftoff. After that, a droplet of SWCNT suspension (ethanol) was dropped to the gap between
electrodes that connects to the dielectrophoresis (DEP) deposition system [27]. An individual
SWCNT was deposited between the electrodes by optimizing the suspension concentration,
and assisted by the atomic force microscope (AFM) manipulation system [28] [29] [30]. A
more detailed fabrication process was discussed in [31] and [32].
The properties of the Schottky barriers are determined by the energy alignment between
the metals and SWCNT; thus Schottky photodiodes made of different metals were investigated to understand the importance of the metal workfunctions.

26

Metal1
Low Workfunction

CNT

Metal2
High Workfunction

Ef1

Ef

Ef2

(a)
Ohmic
Contact

Photons

e-

Ef1

Ef2

h+

(b)
Figure 2.2: (a) Energy alignment of metals with high/low workfunctions and CNT. (b) Band
diagram of the asymmetric CNT photodetector.

27

Au

CNT

Au

Figure 2.3: An AFM image of a symmetric CNT detector with Au electrodes.

28

The I-V characteristics of symmetric Schottky photodiodes with four different metals
(Pd, Au, Cu and Al)were studied. Their typical I-V curves are shown in Figure 2.4, which
based on more than 100 devices tested in our laboratory.
Figure 2.4 shows that the I-V characteristics for positive and negative biases are symmetric, indicating the symmetric barriers at both ends, while the conductance of these devices varied significantly. The Pd-CNT-Pd diode shows linear I-V characteristics, implying
an ohmic contact or very weak Schottky barrier, which has been verified in [13]. A quasimetallic I-V characteristic of the Au-CNT-Au (weak parabolic) was observed, which resulted
from two small Schottky barriers reversely connecting between the source and drain, since Au
and SWCNT have similar workfunctions. While Cu and Al have much smaller workfunctions
than SWCNT; the contacts between these metals with CNT form larger Schottky barriers,
which is illustrated by the band diagram shown in Figure 2.1. Consequently, the Cu-CNTCu and Al-CNT-Al photodiodes have much smaller conductance. The highest conductance
magnitude difference for these 1D photodiodes was approximately four orders, indicating
the importance of the metal workfunctions. In conventional planar contacts, metal-induced
gap states (MIGS) dominate the contacts, which weakens the role of metal workfunction in
determining the properties of 3D Schottky barriers. That is to say, the 1D Schottky contacts
have a much larger flexibility to tune the performance of the detectors by selecting proper
metals.
In a conventional bulk Schottky diode, carrier injection is dominated by thermionic emission, while the thermal assisted tunneling dominates the carrier injection in the 1D Schottky barrier [33]. Temperature-dependent I-V characteristics of a Au-CNT-Au photodiode is
shown in Figure 2.5. The conductance was reduced significantly as the temperature decreased
from 300 K to 80 K, resulting in dark current suppression, since the decrease temperature
29

Current (A)

1.00E-06
Pd-CNT-Pd
Au-CNT-Au
Cu-CNT-Cu
Al-CNT-Al

5.00E-07

0.00E+00

-5.00E-07

-1.00E-06
-1

-0.5

0

0.5

1

0.5

1

Bias (V)

（a）
1.00E-05
1.00E-06

Current (A)

1.00E-07
1.00E-08
1.00E-09
1.00E-10
Pd-CNT-Pd
Au-CNT-Au
Cu-CNT-Cu
Al-CNT-Al

1.00E-11
1.00E-12
1.00E-13
1.00E-14
-1

-0.5

0

Bias (V)

（b）
Figure 2.4: (a)I-V characteristics of CNT Schottky diodes with four different metals (Pd,
Au, Cu and Al). (b)The I-V characteristics shown in logarithm scale.

30

1.00E-06

1.00E-07

Current (A)

1.00E-08

1.00E-09

1.00E-10

300 K
266 K
236 K
206 K
176 K
106 K
80 K

1.00E-11

1.00E-12

1.00E-13
0

0.2

0.4

0.6

0.8

1

Bias (V)
Figure 2.5: I-V characteristics of a Au-CNT-Au diode at different temperatures.
depresses the tunneling probability of carriers through the barriers.
The photo-responses of the Schottky photodiodes were also investigated. The testing
system consisted of a 830 nm IR laser, radiating on a CNT detector housed in a chamber
after passing through an IR window, which was schematically depicted in [34]. The current
signal was measured using Agilent 4156C semiconductor analyzer by periodically switching
the IR source ON and OFF. All measurements were conducted at zero bias. Figure 2.6 shows
the IR response of a SWCNT Au-CNT-Au photodiode by radiating IR photons on its left
and right contacts. One distinct characteristic is that the photocurrent changed direction but
kept similar magnitude (1.5×10−11 A) when the laser illuminated at two different contacts.

31

0.00E+00
0.00E+00
-5.00E-12
-5.00E-12

Current (A)

Current (A)

OFF

OFF

OFF

(a)

-1.00E-11
-1.00E-11
ON
ON
-1.50E-11
-1.50E-11
-2.00E-11
-2.00E-11
0
5
10
15
0
15
5 10 10
5 (S)
Time
Time (S)
TimeTime (S)
(S)
(a)
2.00E-11
2.00E-11

ON

20
20

ON

1.50E-11
1.50E-11

Current (A)

IR Laser

5.00E-12
5.00E-12

Current (A)

IR Laser

1.00E-11
1.00E-11
5.00E-12
5.00E-12

OFF
0.00E+00
0.00E+00
0
0

OFF

5
10
5 (S)
5 10
Time

OFF
15
15
10

Time (S)
TimeTime (S)
(S)

(b)

20
20

Figure 2.6: The IR responses of a Au-CNT-Au Schottky photodiode by radiating on (a) left
electrode, (b) right electrode.
This is caused by two symmetric Schottky diodes between Au and CNT inversely connecting
to each other, as indicated in the band diagram in Figure 2.1(b), a more detailed discussion
can be found in [23].
The typical IR responses of four CNT Schottky photodiodes with different metals are
shown in Figure 2.7 and Figure 2.8, in which IR laser was radiated at one contact.
The dark current and photocurrent (Iphoto) (current difference between IR ON and OFF)
of these devices are summarized in Table 2.1. Pd-CNT-Pd and Au-CNT-Au photodiodes
32

-2.50E-10
-2.50E-10

Current (A)

Current (A)

0.00E+00
0.00E+00

-5.00E-10
-5.00E-10

Pd-CNT-Pd
Pd-CNT-Pd

-7.50E-10
-7.50E-10

-1.00E-09
-1.00E-09
0
0

5

5

10
10
10

15
20
15 20

Time (S)

Time (S)

(a)

-1.00E-10
-1.00E-10

Current (A)

Current (A)

0.00E+00
0.00E+00

-2.00E-10
-2.00E-10

Au-CNT-Au
Au-CNT-Au

-3.00E-10
-3.00E-10

-4.00E-10
-4.00E-10
0
0

5

5

10
10
10

15
20
15 20

Time (S)

Time (S)

(b)
Figure 2.7: The IR responses of two CNT Schottky photodiodes with metal structures of (a)
Pd-CNT-Pd, (b) Au-CNT-Au.

33

-7.50E-11
-7.50E-11

Current (A)

Current (A)

0.00E+00
0.00E+00

-1.50E-10
-1.50E-10

Cu-CNT-Cu
Cu-CNT-Cu

-2.25E-10
-2.25E-10

-3.00E-10
-3.00E-10
0
0

5

10
10
10

5

15
20
15 20

Time (S)

Time (S)

(a)

-1.50E-11
-1.50E-11

Current (A)

Current (A)

0.00E+00
0.00E+00

-3.00E-11
-3.00E-11

Al-CNT-Al

Al-CNT-Al

-4.50E-11
-4.50E-11

-6.00E-11
-6.00E-11
0
0

5

5

10
10
10

20
15
15 20

Time (S)

Time (S)

(b)
Figure 2.8: The IR responses of two CNT Schottky photodiodes with metal structures of (a)
Cu-CNT-Cu, (b) Al-CNT-Al.

34

had relatively high dark current and photocurrent. This can be explained that both injected
and photo-generated carriers are easily tunneling through the weak barriers. In contrast,
the Al-CNT-Al photodiode had two strong barriers, making the carriers difficult to tunnel
through. As a result, it has relatively low dark current and photocurrent. The Cu-CNT-Cu
photodiode had moderate Schottky barriers, resulting in the highest signal-to-noise ratio
among them.
Table 2.1: Performance of Four Detectors with Different Metals

SWCNT IR Detectors

Dark Current
(A)

Iphoto
(A)

Pd-CNT-Pd
Au-CNT-Au
Cu-CNT-Cu
Al-CNT-Al

2.0×10−10
1.0×10−10
2.0×10−12
9.0×10−13

7.0×10−10
2.0×10−10
2.3×10−10
4.5×10−11

The importance of the signal-to-noise ratio of nano-devices needs to be emphasized here.
Nano-photodetector is difficult to have photocurrent comparable to bulk photodiodes due to
the small fill factor. However, the open circuit voltage (Voc) is determined by the ON/OFF
ratio (the ratio between IR ON and OFF), which can be high even absorbing small amount of
photons, because the dark current can be extremely low due to its nano-scale size and phonon
scattering suppression in 1D structure. The typical IR responses of Cu-CNT-Cu and AuCNT-Au with applied bias are shown in Figure 2.9. The bias-dependent photo-response of
Cu-CNT-Cu shows photovoltaic characteristics: negative bias increases the photocurrent by
widening the barrier, while positive bias reduces the photocurrent due to the shrinkage of the
barrier. A Voc of 100 mV was observed when injected carriers were equal to photo-generated
carriers, resulting in zero net current. The short circuit current of this device is only -

35

5×10−11 A, which is negligible compared to conventional photodiodes, but it corresponds to
100 mV Voc, which is on the same order as other traditional photodiodes. The photocurrent
of Au-CNT-Au with bias is difficult to observe since the dark current dominates the current
flow.
In this section, the performance and characteristics of symmetric photodiodes using different metal electrodes were introduced. Pd-CNT-Pd, Au-CNT-Au, and Al-CNT-Al detectors
have low signal-to-noise ratio, since they have either too small or large Schottky barriers
at both ends. The Cu-CNT-Cu detector has moderate Schottky barriers, generating a high
signal-to-noise ratio and Voc.

2.3

Asymmetric Schottky Photodiodes

The symmetric structure of two reversely connecting Schottky diodes will significantly suppress photocurrent generation, since the two diodes are close to each other, and the photocurrents generated at the two contacts will cancel each other. In addition, the photo-generated
carriers need to tunnel through another barrier before being collected. In order to address
these problems, asymmetric Schottky diodes using metals with high/low workfunctions at
two ends were designed. The band diagram is shown in Figure 2.2.
In order to verify the concept of this asymmetric infrared detector, a fabrication process
was designed to measure the electrical properties and IR reponses of both Au-CNT-Au and
Al-CNT-Au photodiodes using a single CNT. The process began with fabricating two Au
electrodes with a gap of 1.2 µm using ebeam lithography on top of a quartz substrate,
as shown in Figure 2.10(a). It was followed by depositing an individual SWCNT bridging
these two electrodes using the DEP and AFM manipulation system (Figure 2.10(b)). The

36

Current (A)

5.00E-11

-5.00E-11

IR OFF
IR ON

-1.50E-10

-2.50E-10
-0.3

-0.2

-0.1

0

0.1

0.2

Bias (V)

（a）

Current (A)

1.50E-06

5.00E-07

IR OFF
IR ON

-5.00E-07

-1.50E-06
-0.2

-0.1

0

0.1

0.2

Bias (V)

（b）
Figure 2.9: The IR responses of CNT Schottky photodiodes with bias (a) Cu-CNT-Cu, (b)
Au-CNT-Au.

37

I-V characteristics were measured, denoted as Au-CNT-Au in Figure 2.10(d). After that,
another ebeam lithography was used to pattern an Al electrode that covered one of the Au
electrodes and extended 0.2 µm into the gap, making the device structured as Al-CNT-Au
with a gap of 1.0 µm, as shown in Figure 2.10(c). Its electrical characteristics were measured
and as shown in Figure 2.10(d), denoted as Al-CNT-Au. The Au-CNT-Au showed symmetric
and quasi-metallic electrical characteristics (weak parabolic I-V characteristics with higher
bias), while typical diode electrical characteristics were observed in the Al-CNT-Au structure.
The electrical properties of the device changed significantly due to the replacement of an
Au electrode. Therefore, there must be a higher barrier between Al and nanotube that
dominates the optoelectronics properties of the detector [35].
The photo-responses of the Au-CNT-Au and Al-CNT-Au are also compared, as shown
in Figure 2.11. The dark current decreased from 8.8×10−10 A to 1.1×10−11 A, while the
photocurrent increased from 2.1×10−10 A to 2.4×10−9 A. The photocurrent enhancement
increased the responsivity (photocurrent/infrared power) and quantum efficiency (number of generated electrons/number of photons), and both dark current suppression and
photocurrent enhancement contributed to the improvement of the noise equivalent power
(NEP)(noise/responsivity) and noise equivalent temperature difference (NETD). In other
words, the asymmetric structure greatly increased the sensitivity and efficiency of the CNT
based IR detector.

2.4

Chapter Summary

In summary, CNT based Schottky photodiodes using different metals and structures were
fabricated and investigated. It was found that the workfunctions of metals play an impor-

38

Au

Au

(a)
Au

Au

(b)
Al

Au

(c)
1.00E-07

Current (A)

Au-CNT-Au
Al-CNT-Au
5.00E-08

0.00E+00

-5.00E-08

-1.00E-07
-1

-0.5

0

0.5

1

Bias (V)
(d)
Figure 2.10: (a-c) Fabrication process of the Au-CNT-Au and Al-CNT-Au photodiodes using
a single CNT. (d) Their I-V characteristics.

39

1.00E-08

Current (A)

1.00E-09

Au-CNT-Au
Al-CNT-Au

1.00E-10

1.00E-11

1.00E-12
0

5

10

15

20

Time (S)
Figure 2.11: Photo-response of the Au-CNT-Au and Al-CNT-Au photodiodes using a single
CNT.

40

tant role in determining the properties of the photodiodes. Symmetric Schottky photodiodes
using different metals have varied dark current and photocurrent in several orders in magnitude. Low built-in potentials resulted in large dark current and photocurrent, while high
built-in potentials caused smaller dark current and photocurrent. This can be explained
by two Schottky barriers reversely connecting each other in the symmetric metal structure,
which was verified by position-dependent photocurrent measurements. In order to improve
the performance of CNT-based Schottky photodiodes, asymmetric structure photodiodes
using metals with high/low workfunctions at two contacts were fabricated. Compared to
the symmetric photodiodes, asymmetric photodiodes had lower dark current and higher
photocurrent.

41

Chapter 3
Carbon Nanotube Field Effect
Transistor based Photodetectors
In the previous chapter, we demonstrated that 1D Schottky photodiodes between an
individual CNT and metals were responsible for the photocurrent (Iphoto) generation under
continuous IR illumination. The optoelectronic characteristics of CNT based IR detectors
are determined by the Schottky barriers, which depend on the Fermi energy (workfunction)
alignment between the CNTs and metals, therefore photo-response can be improved by
selecting proper metals. However, the choices of metals are limited, especially those that
are both compatible with current technology (e.g., CMOS) and have desired workfunctions.
In addition, the performance of CNT Schottky photodiodes is constrained by workfunction
variation caused by technical difficulties in controlling CNT synthesis [36]. Therefore, it is
more effective to control the Fermi energies of CNTs in order to optimize the performance
of CNT based photodetectors, instead of changing workfunctions by selecting proper metals.
CNT field effect transistors (CNTFETs) can be considered as CNT Schottky diodes with an
external terminal, called a gate, capacitively coupled to the CNT channel to electrostatically
dope CNT by applying gate voltages (Vg). In other words, the Fermi energies of CNTs can
be controlled by CNTFETs through their gates.
To begin with, a brief history of CNTFETs will be given. Conventional metal-oxide-

42

semiconductor field-effect transistor (MOSFET) technologies have been improved dramatically in the past decades by down scaling, which results in better performance. Down
scaling of the MOSFET has recently led to a device size below 20 nm, but the ultimate
limit will be reached in the near future due to current leakage. Therefore there is intense
interest in finding alternative materials and devices. CNT is very promising in this respect,
because its exceptional electrical properties, such as high current carrying capability and
excellent carrier mobility, make it a potential candidate for replacing traditional semiconductors. CNTFETs have been demonstrated to operate in the ballistic transport regime at
room temperature [13]. In addition, CNT essentially conducts current on its surface, where
all the chemical bonds are saturated and stable. Thus it is compatible with high-k gate
dielectrics, and there is no need to passivate the interface between the CNT channel and the
gate dielectric.
The first CNTFET was reported in 1998 [21] [37]. The researches fabricated the CNTFET
by dispersing CNTs in a solvent, separating bundles in an ultrasonic bath, and depositing
the CNTs onto a Si/SiO2 surface with prefabricated metal electrodes. A few of the semiconducting CNTs would bridge two electrodes and define a conducting channel between them.
Modulation of the transistor conduction was achieved by applying voltages to the silicon
back-gate. During the first few years after the report of CNTFETs, their working principle
was not well understood. They were analyzed either like MOSFETs, barrier injection transit
time (BARITT) diodes, or double 1D Schottky barrier transistors. The dominant Schottky
barriers were experimentally confirmed in CNT cross-junctions made of one semiconducting
and one metallic CNT in 2000 [38]. Since then, the performance of CNTFETs has been
greatly improved by decreasing the thickness of dielectric layer [39], using high-k materials
as gate the dielectrics [40], and eliminating barriers by forming ohmic contacts [13].
43

Although CNTFETs were extensively developed, CNTFET based IR detectors are not
well understood, since high performance CNTFETs functioning as switches may not perform
well as IR detectors. They require different characteristics: CNTFETs working as switches
need high switching speed, low subthreshold slope, and high transconductance, while IR
detectors require low noise, fast response, and large effective absorption area. In addition,
although the theory of planar contact between metals and traditional semiconductors was
fully developed and understood, it cannot be applied to 1D structures due to their unique
characteristics, which arise from quantum confinement. That is to say, the behavior of CNT
based photodetectors may differ from traditional photovoltaic devices even thought they
share similar structures. Therefore, IR detectors using CNTFETs need to be investigated to
understand their working principles so as to optimize their performance.
CNTFET based IR detectors have been fabricated to investigate their photoconductivity
by modulating the Fermi energy of CNT though silicon back-gates [41] [42], as illustrated
in Figure 3.1. In this configuration, a semiconductor CNT contacts two metal electrodes
on a heavily doped Si/SiO2 substrate. The metals serve as source and drain electrodes,
while the conducting substrate acts as a gate terminal, which is used to control the Schottky
barriers by modulating the Fermi level of the CNT. However, these CNTFETs focused on
exploring the physics of the CNTs, rather than concentrating on optimizing performance of
CNT photon-detectors. In this chapter, optimizing the design and performance of CNTFET
based IR detectors will be discussed.
We have already studied the performance of CNT based Schottky photodiodes in the
previous chapter. However, in order to optimize the design and performance of CNTFET
based detectors, three questions remain to be answered: 1) how the Vg affects the detector
performance; 2) whether an asymmetric metal structure is suitable for CNTFETs; 3) what
44

Metal1
(Source)

Metal2
(Drain)
CNT

SiO2
Silicon Back Gate

Figure 3.1: An AFM image of the structure of CNTFET with a silicon back-gate.
is the optimal gate structure. Five different CNTFETs will be introduced in the following
sections to find the answers to these questions. A back-gate CNTFET with Au-CNT-Au
structure is introduced first, and its photo-responses are compared with back-gate CNTFETs
with Ag-CNT-Ag configuration, so as to understand the roles of metal workfunctions and Vg
in determining the detector performance. Asymmetric back-gate CNTFET with Ag-CNTAu structure was then investigated, and both high photocurrent and open circuit voltage
(Voc) were obtained. The back-gate structure, albeit simple, provides valuable information
to guide the design of CNTFET based IR detectors. However, the back-gate structure cannot be applied to a large scale detector array, since its component devices will be modulated
simultaneously. Individual gate is required for each single detector. A middle-gate CNTFET
consisting of an Al/Al2 O3 gate between source and drain was examined, showing the capability to control detectors independently. Eventually, an asymmetric multi-gate CNTFET
was investigated to find the optimal gate structure. The multi-gate CNTFET comprise four

45

different gates: back-gate, gates for source and drain, and middle-gates for the CNT channel.

3.1

Back-gate Au-CNT-Au Transistor

Figure 3.2 shows the configuration of a back-gate Au-CNT-Au CNTFET. The fabrication
process of this device began with making two Au electrodes with gap of around 1 µm using
photolithography, thermal evaporation, and lift-off on top of a heavily doped Si substrate
covered with 200 nm SiO2 . It was followed by dropping a droplet of CNT suspension between the electrodes that connect to the DEP system [27]. An individual CNT can be
deposited between the electrodes by adjusting the suspension concentration, assisted by an
AFM manipulation system [28] [29].
Conductance of the CNTFET was varied by applying Vg to the back-gate, which changes
the tunneling probability through the Schottky barriers. The I-V characteristic of a CNTFET using CNT with three different Vg are shown in Figure 3.3(a). A weak parabolic I-V
characteristic was observed without Vg, because of two small Schottky barriers reversely
connecting between the source and drain. When no Vg was applied, the Fermi energy of Au
is close to that of CNT (∼5.0 eV) [43]. By applying different Vg to the gate, p-type transistor characteristics were observed: positive Vg reduced the conductance, while negative Vg
increased the conductance, as shown in Figure 3.3(b). An exceptional characteristic is that
negative Vg caused linear I-V relationship, while positive Vg resulted in strong parabolic
I-V characteristics. In 1D barriers, thermally assist tunneling dominates the carrier injection [44], therefore the linear I-V characteristics indicated an extreme thin or zero barrier,
while strong parabolic I-V characteristics were caused by a strong barrier.
The photo-responses of the CNTFET were measured by focusing an IR laser with 830 nm

46

CNT
Au
Source

Au
Drain
SiO2

Silicon Back Gate
Figure 3.2: Structure of the back-gate Au-CNT-Au CNTFET.
wavelength and a maximum power of 50 mW, on one barrier. A more detailed introduction
to the testing system can be found in [31]. The transfer characteristics of the device at zero
bias with IR laser off and on (50 mW) are shown in Figure 3.4(a). When the IR source
was off, the conductance has similar correspondence to Vg as in Figure 3.4(b). However,
extremely different transfer characteristics were observed when the IR laser was turned on.
This indicated that excess carriers were generated by IR illumination, and contributed to
current flow. The current difference between IR on and off status was denoted as photocurrent (Iphoto) in the following discussion. The relationship between Vg and Iphoto is shown
in Figure 3.4(b). The Iphoto at zero Vg was approximately -17.5 pA. The Iphoto was positive and enhanced by increasing positive Vg, while it became negative when negative Vg was
applied. This can be explained by the band diagram shown in Figure 3.4(b), and the energy
level alignment shown in Figure 3.5 (CNT is p type material in the air [45]). When positive
Vg was applied, the Fermi energy of the CNT became lower than that of Au. In contrast,
sufficient negative Vg increased the Fermi energy of the CNT beyond that of of Au. As a
result, the direction of the Schottky barrier changed, which inversed the sign of Iphoto.
The magnitude of Iphoto was enhanced by increasing the magnitude of both positive and
47

2.00E-08

Ids (A)

1.00E-08

0.00E+00
Vg=-2V
-1.00E-08

Vg=0V
Vg=2V

-2.00E-08
-0.5

-0.25

0

0.25

0.5

Vds (V)
(a)
1.00E-07
1.00E-08
Vds=50mV

Ids (A)

1.00E-09
1.00E-10
1.00E-11
1.00E-12
1.00E-13
-3

-2

-1

0

Vg (V)

1

2

3

(b)
Figure 3.3: (a)The I-V characteristics of the CNTFET. (b) Transfer characteristics of the
CNTFET at Vds=50 mV.

48

2.00E-10

Ids (A)

Vds=0
0.00E+00

-2.00E-10
IR=0
IR=50mW
-4.00E-10
-3

-1.5

0

1.5

3

Vg (V)

(a)

Iphoto (A)

2.00E-10

EF (Au) >
EF (CNT)

0.00E+00

EF (Au) <
-2.00E-10

EF (CNT)

-4.00E-10
-3

-1.5

0

1.5

3

Vg (V)

(b)
Figure 3.4: (a) Transfer characteristics of the Au-CNT-Au CNTFET with IR source off
(IR=0) and on (IR=50mW) at zero bias. (b) The relationship between gate voltage and
photocurrent and corresponding band diagrams at negative and positive gate voltages.

49

Vacuum
Au

CNT
Ec

Ei
EF (Au)

EF
Ev

Figure 3.5: The energy levels of Au and CNT.
negative Vg. However, the Iphoto to dark current ratio (denoted as ON/OFF ratio in the
following discussion) at positive Vg was much higher than that at negative Vg, as explained
in the following discussion.
For negative Vg, increasing Vg caused higher built-in potential and p-doping levels (assume Au and CNT have the same Fermi energy). The additional p-doping may turn the
CNT into a degenerated p-doped semiconductor. This is supported by the linear I-V characteristics at negative Vg (Figure 3.3). The bias-dependent Iphoto values at Vg=-0.5 V are
shown in Figure 3.6(a). One particular feature is that the current decreased with IR illumination, which was caused by the contradiction of the separated photo-generated carriers
and injection carriers. Higher bias decreased carrier transit time, leading to photoconductive
gain, resulting in higher Iphoto.
For positive Vg, larger bias resulted in higher built-in potential but lower p-doping level.

50

Note that the ON/OFF ratio at positive Vg was greatly improved due to the suppression
of dark current and enhancement of Iphoto. The bias dependent measurement at Vg=2
V (Figure 3.6(b)), which clearly showed the photovoltaic characteristics. At Vds=0.12 V,
Iphoto was equal to the current produced by the bias, resulting in zero net current. This
bias is the open circuit voltage, denoted as Voc. Simulation has shown that depletion width
varies exponentially with inverse doping [15] and built-in potential [46]. Therefore, the Iphoto
at positive Vg was increased by the widened depletion width, which enlarged the effective
absorption area.

3.2

Back-gate Ag-CNT-Ag Transistor

It was demonstrated that applying proper voltages to the back-gate can improve the performance of CNT detectors in Au-CNT-Au structure. In order to achieve that, both high
built-in potential and low doping level are desired. However, the built-in potential in the
Au-CNT-Au structure is limited by the workfunction of Au, which is close to that of CNTs.
In addition, the barrier height using Au is small because the majority is holes, as shown in
Figure 3.5. In 3D bulk material, the Schottky barrier properties are dominated by the Fermi
level pining effect, which makes metal workfunction insignificant in determining Schottky
barriers, but this effect is trivial in 1D structures [47]. Therefore, using a metal with lower
workfunction (Ag=4.2-4.5 eV) [48] as electrode can form a higher built-in potential and
barrier height, as shown in Figure 3.7.
A back-gate CNTFET using Ag electrodes was fabricated to investigate the role of metal
workfunction. The fabrication process was the same as for the Au-CNT-Au CNTFET, but
depositing Ag instead of Au. Figure 3.8(a) shows the electrical characteristics of the Ag-

51

1.00E-08

5.00E-09

Ids (A)

Vg=-0.5V
0.00E+00

-5.00E-09
IR=0
IR=50mW
-1.00E-08
-0.15

-0.05

0.05

0.15

Vds (V)

(a)

Ids (A)

0.00E+00

-1.00E-10

-2.00E-10

IR=0
IR=50mW

Vg=2V
-3.00E-10
-0.15

-0.05

0.05

0.15

Vds (V)

(b)
Figure 3.6: The bias-dependent measurement of the Au-CNT-Au CNTFET with IR=0 and
IR=50 mW at (a) Vg=-0.5V and (b)Vg=2V.

52

Vacuum
Ag

CNT
Ec

EF (Ag)
Barrier height
Built-in potential

Ei
EF
Ev

Figure 3.7: The energy levels of Ag and CNT.
CNT-Ag CNTFET with an individual CNT of 1.4 nm diameter. P-type transistor transfer
characteristics were also observed in this device. In addition, negative Vg caused linear I-V
characteristics, while positive Vg resulted in current rectifying. However, one distinct feature
was that the conductance was ∼1-2 orders of magnitude smaller than in the Au-CNT-Au
CNTFET, resulting in smaller dark current. This can be explained by a decreased tunneling
probability through stronger Schottky barriers between Ag and CNT. The dark current of
this detector was changed corresponding to the conductance change: increased by negative
Vg, but decreased by positive Vg, as shown in Figure 3.8(b).
The relation between Vg and Iphoto is shown in Figure 3.9(a). Positive Vg not only
reduced dark current, but also increased Iphoto, resulting in enhanced ON/OFF ratio. The
Voc at zero Vg was small; however, a Voc of 0.3 V was observed when a Vg of 3 V was applied

53

5.00E-10

Ids (A)

2.50E-10

0.00E+00
Vg=-2V
Vg=0
Vg=2V

-2.50E-10

-5.00E-10
-0.5

-0.25

0

0.25

0.5

Vds (V)
(a)
1.00E-08

Ids (A)

1.00E-09
Vds=50mV

1.00E-10
1.00E-11
1.00E-12

1.00E-13
-3

-2

-1

0

1

2

3

Vg (V)

(b)
Figure 3.8: (a) I-V characteristics of the CNTFET. (b) Transfer characteristics of the CNTFET at Vds=50 mV.

54

to the back-gate with 50 mW IR illumination (Figure 3.9(b)). Bias-dependent measurement
with four different power outputs was conducted. Typical photovoltaic characteristics were
measured: Iphoto was linearly proportional to the output power, while Voc was exponentially
proportional to the output power.
One particular feature is that the highest Iphoto of the Ag-CNT-Ag CNTFET at IR=50
mW (5×10−11 A) was smaller than that of the Au-CNT-Au CNTFET (1×10−10 A) in the
bias-dependent measurements, but Voc was much higher. This can be explained by the
symmetric metal structure: a higher built-in potential in Ag-CNT contacts resulted in wider
absorption width (depletion width) and lower dark current, but the separated electrons and
holes at one barrier needed to tunnel through a wider barrier in order to contribute to Iphoto
generation. The comparison between these two CNTFETs indicates that metal selection
plays a vital role in determining detector performance. Ag has lower work function than Au,
resulting in a higher built-in potential and barrier height in the metal-CNT contacts. As a
result, a higher Voc was obtained in the back-gate Ag-CNT-Ag CNTFET.

3.3

Back-Gate Au-CNT-Ag Transistor

Electrostatic doping in two symmetric CNTFETs has been demonstrated in previous sections as an effective method to improve detector performance by depressing dark current
and enhancing Iphoto with widened Schottky barriers. However, the symmetric structure
depressed photocurrent generation, since the separated electrons and holes need to tunnel
through a wide barrier before contributing to current flow. Motivated by this problem, a
back-gate Au-CNT-Ag CNTFET was fabricated, as shown in Figure 3.10. CNT and Au form
a quasi-ohmic contact, while CNT and Ag form a Schottky barrier. As a result, this detec-

55

Iphoto (A)

2.00E-11

0.00E+00

-2.00E-11

-4.00E-11

-6.00E-11
-5

-3

-1

1

3

5

Vg (V)
(a)
8.00E-11

Ids (A)

4.00E-11

IR=0
IR=15mW
IR=25mW
IR=50mW

Vg=3V

0.00E+00

-4.00E-11

-8.00E-11
-0.4

-0.2

0

0.2

0.4

Vds (V)
(b)
Figure 3.9: (a) The relation between gate voltage and photocurrent of the Ag-CNT-Ag
CNTFET. (b) Bias dependent measurement with IR at four different output powers at
Vg=3 V.

56

CNT
Ag
Source

Au
Drain
SiO2

Silicon Back Gate
Figure 3.10: Structure of the back-gate Au-CNT-Ag CNTFET.
tor becomes a Schottky diode. By irradiating the device with IR photons, photo-response
can be improved, since the separated electrons and holes do not need to pass through as a
strong barrier (Ag-CNT) as in the symmetric structure. The fabrication process began with
making an Au electrode, followed by fabricating an Ag electrode with a gap of 1 µm to the
Au electrode after fine alignment.
Figure 3.11(a) shows the transfer characteristics of the Au-CNT-Ag CNTFET with and
without IR illumination. The dark current showed p-type transistor characteristics. Iphoto
was highly improved with positive Vg, although no obvious improvement is apparent in
Figure 3.11(b) due to the hysteresis effect [49]. By measuring the temporal photo-response,
it is shown that the Iphoto was increased from 2×10−11 (Vg=0) to 1.2×10−9 A (Vg=8V)
(Figure 3.11(b)). The highest Iphoto value, around 1 nA, was an order in magnitude higher
than that of symmetric detectors.
A high ON/OFF ratio was obtained with positive Vg, thus the Voc was expected to
improve. The bias dependent measurement with four different output powers at Vg=5 V is
shown in Figure 3.12. This is a typical photovoltaic photo-response: negative bias enlarged
the Schottky barrier that increased the Iphoto, while positive bias shrank the depletion
region that suppressed the Iphoto. In addition, the Iphoto is proportional to the output
57

1.00E-08

Ids (A)

1.00E-09
1.00E-10
1.00E-11
1.00E-12
IR=0
IR=50mW

1.00E-13
1.00E-14
-10

-6

-2

2

6

10

Vg (V)

(a)
ON
IR

OFF

1.6E-09

Ids (A)

1.2E-09

8E-10
Vg=8V
Vg=0V
Vg=-8V

4E-10

0
0

5

10

15

20

Time (S)

(b)
Figure 3.11: (a) Transfer characteristic of the asymmetric CNTFET with and without IR
illumination. (b) Temporal photo-response with three different gate voltages at IR=50 mW.

58

Ids (A)

0.00E+00
-4.00E-10
-8.00E-10
IR=0
IR=12.5mV
IR=25mV
IR=50mV

-1.20E-09
-1.60E-09
-0.5

-0.25

0

0.25

0.5

Vds (V)
Figure 3.12: Bias dependent measurement with IR at four different output powers at Vg=5
V.
power, while the Voc is exponentially proportional to the output power. A Voc around 0.45
V was observed with 50 mW IR illumination, which is the highest Voc reported to date for an
individual CNT based photo-detector. In contrast to the Iphoto, which is linearly dependent
on the absorbed photons, the Voc is determined by the ON/OFF ratio. Therefore, although
the detector absorbs small amounts of photons, Voc can still have a high value because the
dark current can be extremely low in 1D structures due to the phonon scattering suppression
and its nano-scale size.

3.4

Middle-Gate Transistor

Previous CNTFETs showed promising improvement by using a back-gate to modulate the
Schottky barriers between metals and a CNT. However, back-gates cannot be used in large
59

scale detector arrays, since all the devices will be modulated simultaneously. In this section, a middle-gate CNTFET, which has the capability to control each individual detector
independently, will be introduced.
The structure of the middle-gate CNTFET and its AFM image are shown in Figure 3.13.
A CNT bridges two Au electrodes as source and drain, and an Al/Al2 O3 middle-gate is made
underneath the CNT and partially couples it. The fabrication process begins with making
two Au electrodes with an insulating gap of 2 µm. A 1-µm-wide Al electrode was fabricated
in between the source and the drain by e-beam lithography, thermal evaporation, and lift-off.
The sample was then bombarded by oxygen plasma, and a thin layer of Al2 O3 grown on the
Al as the insulator for the gate. The last step is to deposit a CNT to connect the two Au
electrodes using the DEP deposition system and AFM manipulation system. Two inversely
connecting p-n junctions (p-n-p) will be formed by applying proper Vg to the middle-gate,
as shown by the band diagram in Figure 3.13(c).
The I-V characteristics of the middle-gate CNTFET with three different Vg are shown
in Figure 3.14(a). P-type transistor characteristics were observed. Its conductance slightly
increased with negative Vg, but decreased significantly with positive Vg. Current was linearly
dependent on the bias without Vg, implying that ohmic contact is most probably between
the CNT and Au. This was further verified by the no photo-response at Vg=0 V. The
conductance increased with negative Vg may have resulted from the fringing field effect to
the contacts. Positive Vg converts the middle of CNT into n-type material, forming a p-n-p
structure, which can be considered as two inversely connected p-n junctions. Thus, diodelike I-V characteristics are shown at Vg= 2 V in Figure 3.14(a). The transfer characteristics
at Vds=10 mV are shown in Figure 3.14(b) with a 200 mV/dec subthreshold slope.
The transfer characteristics, at Vds= 0 V with IR on and off, are shown in Figure 3.15(a),
60

Source
(Au)

Gate
(Al/Al2O3)
CNT

Drain
(Au)
Au

Substrate
(a)

Source
(Au)

Drain
(Au)

CNT

Gate
(Al/Al2O3)
(b)

p

p
n

(c)
Figure 3.13: (a) Structure of the middle-gate CNTFET. (b) AFM image of the middle-gate
CNTFET. (c) Band diagram of the working status of the middle-gate CNTFET.

61

1.00E-07

Ids (A)

1.00E-09

1.00E-11
Vg=-2
Vg=0
Vg=2

1.00E-13

1.00E-15
-0.5

-0.3

-0.1

0.1

0.3

0.5

Vds (V)

(a)
1.00E-07

Ids (A)

1.00E-08
1.00E-09
1.00E-10
1.00E-11
1.00E-12

Vds=10mV

1.00E-13
1.00E-14
-2

-1

0

1

2

Vg (V)

(b)
Figure 3.14: (a) I-V characteristics of the middle-gate CNTFET at three Vg in log-scale.
(b) Transfer characteristics of the CNTFET with Vds=10 mV.

62

by sweeping the Vg from -2 V to 2 V. Dark current of the CNTFET decreased from 5×10−11
A(zero Vg) to 5×10−14 A with increasing positive Vg, since the electrons needed to tunnel
through two p-n junctions in the CNT in order to contribute to the dark current. When 50
mW IR illuminated the CNT, similar transfer characteristics were observed with negative
and small positive Vg, indicating no Iphoto was produced. Iphoto was observed with Vg
values that are larger than 0.5 V, as shown in Figure 3.15(a). There was a sudden drop of
the current signal at Vg of around -1 V, which may have been caused by defects in the CNT.
The hysteresis phenomenon was observed in this device, and that is the reason why the
sudden drops of the current with and without IR di not cause at the same Vg. In order to
eliminate the effect of the hysteresis, we measured the temporal photo-response at fixed Vg
by switching IR on and off at 0.2 Hz, as shown in Figure 3.15(b). No Iphoto was generated
at Vg=0 V and Vg=-2 V, and the dark currents were large compared to those with positive
Vg. When a 2 V Vg is applied, Iphoto with value of 1×10−11 A was generated, and an
ON/OFF ratio of around two orders was observed.
Although the Iphoto is small, the Voc is determined by the ON/OFF ratio and operating
temperature. We investigated the Voc by measuring bias dependent current at Vg=2 V, as
shown in Figure 3.16. A typical photovoltaic photo-response was observed, and a 0.1 V Voc
was obtained.
By introducing a middle-gate to the CNTFET, each individual detector in a large scale
array can be controlled separately. However, this structure has some limitations that prevent
optimal performance. First of all, the Iphoto generation is based on two inversely connecting
p-n junctions, which is similar to the CNTFET with symmetric Schottky barriers. The
detected Iphoto reflects the signal intensity difference between two junctions that are only
1 µm away from each other. What is more, although ohmic contact can be formed between
63

1.00E-08
1.00E-09

Ids(A)

1.00E-10
1.00E-11
1.00E-12
IR=0
IR=50mW

1.00E-13
1.00E-14
1.00E-15
-2

-1

0

Vg (V)

1

2

15

20

(a)
IR OFF

ON

1.00E-08
1.00E-09

Ids (A)

1.00E-10
1.00E-11
1.00E-12
1.00E-13
1.00E-14
Vg=-2V
Vg=0
Vg=2V

1.00E-15
1.00E-16
0

5

10

Time (s)

(b)
Figure 3.15: (a) Transfer characteristics of the middle-gate CNTFET with IR on and Off at
zero bias. (b) Temporal photo-response measurement by switching IR on and off at 0.2 HZ.

64

1.50E-10

Ids (A)

1.00E-10
5.00E-11

IR=0
IR=50mW

0.00E+00
-5.00E-11
-1.00E-10
-1.50E-10
-0.5

-0.3

-0.1

0.1

Vds (V)
Figure 3.16: Bias dependent measurement with IR at two output powers at Vg=2 V.
a CNT and metals by selecting a proper metal, the fringe field from the middle-gate may
produce a barrier that suppresses the generation of Iphoto.

3.5

Multi-Gate Transistor

In previous sections, we have demonstrated that both an asymmetric metal structure and
electrostatic doping using a back-gate can improve the performance of the CNT detectors by
manipulating the energy alignment between metals and a CNT, but a back-gate modulates
all the detectors simultaneously. In order to control detectors independently, a middlegate CNTFET was fabricated and investigated. However, all these CNTFETs have their
limitations to perform optimally.
In order to optimize the design of the CNT detectors, an asymmetric multi-gate CNTFET

65

based IR detector was fabricated, integrating an asymmetric metal structure(Au-CNT-Al)
with multiple gates, which allow controlling the Fermi energies at source, drain, and CNT
channel independently. Examining gate structure dependent photo-responses may lead to
an improved understanding and design of the photodetectors.
A multi-gate asymmetric CNTFET was fabricated with a back-gate and three pairs
of side-gates that can locally dope the source, SWCNT channel, and drain, as shown in
Figure 3.17. This device was an extension from the asymmetric (Al-CNT-Au) CNFEET
with three pairs of Al side-gates located 400 nm to the SWCNT, denoted as Gate1, Gate2,
and Gate3. Gate1 and Gate3 are used to modulate the Fermi level of source and drain
independently. Gate2 is not only used to modulate the Fermi level of the SWCNT channel,
but also blocks the fringing electrical field of Gate1 from Gate3 to guarantee the local doping
capability.
Previous comparisons between symmetric and asymmetric CNTFET are based on different CNTs. A fabrication process was developed to compare electrical and optical characteristics of symmetric and asymmetric CNTFETs connecting to the same CNT. The fabrication
process started with a heavily doped p-type single-crystal Si substrate, which was cleaned
and coated with 200 nm of thermal SiO2 . The source-drain electrodes made of Au were
fabricated by EBL, thermal evaporation, and liftoff, with a separation of 1000 nm, width of
500 nm, and thickness of 40 nm. SWCNT suspension was prepared by dispersing SWCNT
powder in alcohol medium, followed by 10-15 minutes ultrasonication. Metallic SWCNTs
were filtered out by a micro-dielectrophoresis chamber [32]. A droplet of SWCNT suspension
was deposited between the source-drain electrodes, and dielectrophorisis force was generated
by applying a sine wave (frequency of 10 kHZ and Vpp of 1 V) to trap a semiconductor
SWCNT between the electrodes. An atomic force microscopy (AFM) manipulation system
66

Gate1

Source
(Au)

200 nm

Gate1

Gate2

Gate3

Drain
(Al)

SWCNT

Gate2

Gate3

Figure 5 AFM image of the multi-gate AM-FEP with a back-gate and three pairs of side-gates that can
Figure 3.17: AFM image of the multi-gate asymmetric CNTFET with a back-gate and three
pairs of side-gates that can locally modulate source (Gate1), SWCNT channel (Gate2), and
drain (Gate3).

67

was utilized to guarantee the deposition of an individual SWCNT for each device [30]. Then
the symmetric CNTFET was formed, and its I-V characteristics and photo-responses were
measured. To convert this device into an asymmetric CNTFET, an Al electrode was used to
cover one of the Au electrodes. The Al electrode was patterned by EBL and liftoff using 50
nm Al. To guarantee the formation of the asymmetric metal structure, the Al electrode was
200 nm extending towards the gap of two Au electrodes, as shown in the insets of Figure 3.18.
Meanwhile, three pairs of side-gates are fabricated. The electrical and optical characteristics
of the asymmetric CNTFET were recorded using the same experimental setup as in the
symmetric CNTFET.
Electrical measurements were performed on the devices with source electrode maintained
at ground potential using Agilent 4156C semiconductor parameter analyzer. Figure 3.18
shows the I-V characteristics of the symmetric and asymmetric CNTFETs at zero gate bias.
The symmetric CNTFET forms two reversely connecting Schottky barriers at two contacts
(inset of Figure 3.18), resulting in symmetric I-V characteristics. The I-V characteristics of
asymmetric CNTFET exhibited current rectifying, with a high conductivity when forward
biased and a low conductivity when reverse biased, due to the formation of two Schottky
barriers in the same direction at two contacts.
Before the measurement of all gates, we firstly test the back-gate (Vgs) dependent photoresponses. An IR laser with wavelength of 830 nm and an output power of 50 mW was used
to investigate the photo-responses of the CNT CNTFETs. The IR light is linearly polarized
along the axis of the SWCNT. The laser was mounted on a high precision triple divide linear
translation stage to align the laser spot onto the CNTFETs. Figure 3.19 shows the transfer
characteristics of the symmetric CNTFET at zero bias with and without IR illuminating.
The symmetric CNTFET exhibited typical unipolar p-type transfer characteristics at dark
68

0.10

(Drain)
Al

(Source)
SWCNT

0.05

Au

SiO2

Ids ( A)

(Gate)
p-Si

0.00
(Drain)
(Source)
SWCNT Au
Au

-0.05

SiO2
(Gate)
p-Si

-0.10
-1.0

-0.5

0.0

0.5

1.0

Vds (V)
Figure 3.18: The I-V characteristics of the CNTFET with symmetric (Au-CNT-Au) and
asymmetric (Al-CNT-Au) metal structures. The insets shows the cross-section of the CNTFET.

69

(denoted as ”IR OFF” in Figure 3.19): the dark current decreased from 60 pA (Vgs=-3 V) to
0.8 pA (Vgs =3 V). Under IR illumination, denoted as ”IR ON” in Figure 3.19, a different
transfer characteristic was measured (blue curve), implying photo-generated electron and
hole pairs were generated and separated by the Schottky barriers, and consequently contributed to the photocurrent. Photocurrent was positive with positive Vgs, while it changed
polar with negative Vgs, as indicated by the arrows in Figure 3.19. This implied that Vgs
changed the direction of Schottky barriers through electrostatic doping (band diagrams in
Figure 3.19), because the workfunction of Au is only slightly bigger than that of SWCNT.
The asymmetric CNTFET at dark converted to an n-type transistor (Figure 3.20). The
lowest dark current was obtained at Vgs = -3 V with a value of 0.02 pA. Upon IR illumination, the current signal exhibited p-type transfer characteristics: Figure 3.20 clearly shows
current increased towards decreasing Vgs with IR ON. As a result, the photocurrent transfer
curve of asymmetric CNTFET showed p-type characteristics, as shown in Figure 3.21. With
regard to a transistor, this device can induce n- to p-type conversion by illuminating light
onto the device due to change of the barrier width (w): negative gate bias increased the
built-in potential between Al and SWCNT that enlarged the effective absorption region (w),
as illustrated in the band diagrams of Figure 3.21. With regard to a photodetector, the
sensitivity was enhanced 140 times according to the definition of sensitivity as the ratio of
photocurrent to dark current. Because the lowest dark current of asymmetric CNTFET was
40 times smaller than symmetric CNTFET, and highest photocurrent was 3.5 times larger
than symmetric CNTFET. In addition, photocurrent of the asymmetric CNTFET remained
the same polar with Vgs in the range between -3 V to 3 V due to the large workfunction
difference between Al and CNT.
Temporal photocurrent measurement was performed with different gate biases by switch70

60
IR OFF
IR ON

50
40

Au

Au

Isd (pA)

30

Au

Au

20
10
0
-10
-20
-3

-2

-1

0

1

2

3

Vgs (V)

(a)

Figure 3.19: Transfer characteristics versus the back-gate bias at zero bias of the symmetric
CNTFET with absence and presence of IR laser illumination (wavelength of 830 nm and
output power of 50 mW) by sweeping the gate bias from -3 V to 3V. Two band diagrams
show the Schottky barriers at negative and positive gate biases.

71

80

2

Ids (pA)

70
60

Isd (pA)

50
40

Au

Al

30
20

10
1
10
0
10
-1
10
-2
10
-3
10
-3 -2 -1 0 1 2 3
Vgs (V)

IR OFF
IR ON

10
0
-3

-2

-1

0

1

2

3

Vgs (V)
Figure 3.20: Transfer characteristics at zero bias of the asymmetric CNTFETs with absence
and presence of IR illumination. The inset shows the transfer characteristics in logarithm
scale. The band diagram shows the energy alignment of the device.

72

Photocurrent (pA)

80

60

40

Al

Al

w

w
20

0
-3

-2

-1

0

1

2

3

Vgs (V)
Figure 3.21: Photocurrent transfer characteristics of the asymmetric CNTFET at zero drain
to source bias. The band diagrams show the barrier width (w) at negative and positive gate
biases.
ing IR on and off at 0.2 Hz. Figure 3.22 shows that the dark current was reduced with
decreasing Vgs at the absence of IR illumination, and the current was increased significantly
upon light illumination, especially at large negative gate bias. The current dropped to the
dark current when the IR laser was turned off. The reproducibility of the photo-response
was excellent, and the response time was short. The maximum sensitivity was obtained at
Vgs =-3 V with lowest dark current and highest photocurrent.
After the testing of back-gate dependent photo-responses, we compared the transfer characteristics of all gates. Considering the differences of distance and dielectric constant of the
73

ON
IR

OFF

100
Vgs=-3V
Vgs=-1V

80

Vgs=0V

Ids(pA)

60
40
20
0
0

2

4

6

8

10

12

14

Time (s)
(b)
Figure 3.22: Temporal photocurrent measurement of the asymmetric CNTFET with three
different gate biases (-3V, -1V, and 0V) by switching IR OFF and ON at 0.2 Hz.

74

back-gate and side-gates that capacitively-couple to the SWCNT, we normalized the gate
voltages in transfer characteristics measurement. Figure 3.23(a) shows the normalized photocurrent transfer characteristics by sweeping the gate biases of back-gate (Vgs), source
(Vg1), SWCNT channel (Vg2), and drain (Vg3) to independently dope different portions of
the SWCNT. Unexpectedly, the trend of photocurrent transfer characteristics was the same,
independent of the gate location. Although the Gate1 and Gate3 locally doped Au-CNT and
Al-CNT contacts, all photocurrent transfer characteristics showed p-type characteristics that
were the unique signature of Al-CNT contact (photocurrent decreased with increasing gate
biases). The magnitude of the gate voltage to obtain maximum photocurrent was smallest
for the back-gate (Vg=-2.3 V), while the largest gate voltage was required for the side-gates
of drain (Vg=-3.5 V). These results were verified by temporal photocurrent measurement in
order to avoid the influence of hysteresis [49]. We have fabricated five devices, all of which
showed similar results.
These photocurrent transfer characteristics were completely different from conventional
planar devices. We attribute this unique property to the charge distribution in 1D materials: there is a long tail extending from the junction into the undoped region, and this
charge tail can be several micrometers long [15], as shown in Figure 3.23(b). This implies
that the depletion region from the Al-CNT junction may extend into the whole SWCNT
channel, and even approach the Au-CNT junction. In other words, the SWCNT channel
became a partially ”full” depletion region, resulting in an enlarged effective absorption area.
Traditional p-n photodiode increases its effective absorption area by introducing an intrinsic
semiconductor between p-type and n-type semiconductors to form the p-i-n junction, while
1D semiconductors can naturally enlarge their depletion region. The charge along the charge
tail decreased from the Al-CNT depletion region towards to the Au-CNT contact, resulting
75

in the difference of magnitude of gate voltages required to obtain maximum photocurrent
in Figure 3.23(a). In addition, this implies that the gate structure of CNTFETs can be
simplified into a single gate structure, which can significantly facilitate large scale device
manufacturing, and thus CNTFETs may become the building blocks for nano-photodetector
arrays.

3.6

Detector Array using CNT based Transistors

In previous sections, we demonstrated that photocurrent improvement has weak dependence
on the gate position. Thus multiple gates can be replaced by one pair of gates, as shown
in Figure 3.24. This simplification can not only make the device smaller, but also avoid the
electric field influence from neighboring devices.
A 3×3 CNT detector array was fabricated, as shown in Figure 3.25, which also shows the
AFM image of the structure of one of its pixels. For this array, two Au electrodes of each
detector were fabricated with a gap of 1 µm using EBL, thermal evaporation, and lift-off.
A single CNT was then deposited between the electrodes using DEP deposition system and
AFM manipulation system, followed by fabricating two gates and replacing one of the Au
electrodes with Al. At the end, the whole device was packaged by a parylene-C layer in
order to protect the device from environmental contaminations.
The pitch size of this array is 15 µm. Each individual detector in the array can be
controlled independently using their own gates. We first measured the photo-response of
each detector separately, and then calibrated proper values of Vg for each individual detector
in order to make them generate the highest photocurrent.
By irradiating the detector array with IR photons from a laser, photocurrent of all detec-

76

Photocurrent (A. U)

1.0
0.8

Vgs
Vg1

0.6

Vg2
Vg3

0.4
0.2
0.0
-4

-3

-2

-1

0

Vg (V)
(a)

Gate1
Gate2
charge tail

Gate3
charge
Drain

Source
SWCN

Gate1

Gate2

Gate3

Back Gate

(b)
Figure 3.23: (a) Photocurrent transfer characteristics by applying gate biases to back-gate
(Vgs), source (Vg1), SWCNT channel (Vg2), and drain (Vg3). (b)A 3D geometry of the
multi-gate CNTFET and the charge distribution of the SWCNT, showing the charge tail on
the SWCNT.

77

Al
(Source)
CNT

Al
(Gate)

Al
(Gate)

Au
(Drain)
SiO2
Silicon Si
Substrate

Figure 3.24: Asymmetric CNTFET with a simplified gate structure.
tors was measured. The images obtained from the measurement are shown in Figure 3.25.
The color lightness shows the magnitude of photocurrent, which was normalized. In the
images, the darker the color, the higher the Iphoto. Figure 3.25(a) shows that the highest
light intensity of the laser is radiating on Pixel5. Other pixels have smaller Iphoto, and
the value of Iphoto decreased as the distance between the detector and laser spot increased.
When the laser was moved towards top-left, Figure 3.25(b) shows that the highest intensity
point has moved to Pixel1. This detector array can be considered a small focal plane array
in an IR imaging system. If the CNT detector array had a larger number of detectors, a
higher resolution IR image could be obtained.

3.7

Chapter Summary

In summary, five types of CNTFETs were fabricated to investigate and improve the performance of the CNT based IR detectors. In the back-gate Au-CNT-Au CNTFET, the detector
performance was improved by applying voltages to the back-gate, resulting in smaller dark

78

3x3 Detector Array
1

2

3

4

5

6

7

8

9

Pixel9 Detector
Al
(Source)
Al
(Gate)

Al
(Gate)

CNT

Au
(Drain)
Figure 3.25: The image of a 3x3 detector array and the AFM image of the Pixel9.

79

Laser Spot Center (Pixel5)

(a)

Laser Spot Center (Pixel1)

(b)
Figure 3.26: The images obtained from the detector array, and color lightness shows the
magnitude of photocurrent. The laser spot center is on (a)Pixel5, (b)Pixel1.

80

current and higher Iphoto. Back-gate is capacitive-coupled to the CNT, thus applying Vg
can electrostatically modulate the Fermi energy of CNT so as to control the Schottky barriers
that are responsible for Iphoto generation. The back-gate Ag-CNT-Ag CNTFET produced
higher Voc under IR illumination, since the built-in potential between Ag and CNT is higher
than that of Au and CNT, showing the significance of metal workfunction. The previous
two CNTFETs are with symmetric metal structures, which constrained the performance of
the detectors, since separated electrons and holes from one Schottky barrier need to tunnel through another barrier. In the back-gate Au-CNT-Ag CNTFET, an asymmetric metal
structure was used, and both Iphoto and Voc were highly improved. The reason is that the
electrons and holes from the Ag-CNT interface only needed to pass through a small barrier
between Au and CNT before being collected. CNTFET with back-gate structure is easy to
fabricate and provide valuable information for improving detector performance. However,
the back-gate structure is not suitable for larger scale detector arrays, because the back-gate
will modulate all detectors simultaneously. A middle-gate CNTFET was fabricated in order
to control each detector independently. But CNT detector with middle-gate structure and
symmetric metal structure could not deliver optimal performance. In order to find the optimal detector design, an asymmetric multi-gate CNTFET was introduced, with asymmetric
metal structure and multiple gates, including back-gate, gates for source and drain, and middle gates. A multi-gate CNTFET was developed and tested, revealing an unconventional
photocurrent transfer characteristic, which was attributed to the unique charge distribution
of 1D materials. Therefore, CNT detectors can be simplified into a single gate structure. A
3×3 focal plane array using the asymmetric CNT detector with one pair of side-gates was
fabricated, effectively detecting the laser intensity distribution.

81

Chapter 4
Spectral Responses of CNT
Photodetectors
The bandgap energy of a CNT is inversely proportional to its diameter, therefore precise
control of CNT diameter will find a wide range of applications in electronics and photonics.
However, there are some technique difficulties to synthesize CNTs with an uniform diameter
and chirality. In this chapter, we will firstly measure the spectral response of a SWCNT
based photodetectors. Electrical breakdown process was introduced to tailor the bandgap
energy of MWCNTs. After a number of consecutive breakdowns, it was found that CNT
photodetectors can detect MWIR at room temperature.

4.1

Spectral Response of SWCNT based Photodetectors

In previous chapters, we have demonstrated that both asymmetric metal structure and
electrostatic doping through a gate can significantly improve the photo-responses of CNT
photodtectors. However, all previous testings used IR laser with wavelength of 830 nm as
light source. Therefore, the spectral responses of the CNT photodetectors are not clear.
In order to understand the spectral responses of CNT photodetectors, a blackbody test-

82

Cryogenic Cooling System Chamber
Detector
Blackbody

Lens

Electrical

Filter

Feed through

Semiconductor

IR source
Analyzer

IR window

Figure 4.1: The blackbody testing system.
ing system has been developed as shown in Figure 4.1. The IR source with wide range of
wavelengths radiates from an Oriel 67030 blackbody (Newport Corp.) with a temperature
range of 50o C to 1050o C. To measure the spectral responses of the photodetectors at a specific wavelength, different bandpass filters were used and positioned between the blackbody
and the CNT photodetectors. The lens focuses IR signal onto the CNT photodetector that
is mounted outside the cryogenic cooling chamber. The sensor temperature can vary from
room temperature to -196◦ C by feeding liquid nitrogen. With the blackbody testing system,
the CNT photodetectors can be characterized at different temperatures as well as different
gas environments.
We fabricated an asymmetric (Ti-CNT-Au) CNTFET based photodetector with a backgate to provide electrostatic doping. The fabrication started with fabricating Au and Ti
electrodes with a gap of 1 µm on top of a heavily doped Si substrate with a 200 nm-thick SiO2
layer. An individual SWCNT was deposited between the electrodes using dielectrophorisis
deposition system and AFM manipulation system. The I-V curves of the device at different
gate voltages are shown in Figure 4.2(a). Current rectifying of these I-V curves revealed the
asymmetric Schottky barriers at two contacts. The transfer characteristics (Figure 4.2(b))

83

of the transistor show that conductivity decreases with increasing gate voltage, a typical
p-type transistor characteristic.
The temporal current at different gate voltages without IR illumination was shown in
Figure 4.2(c). The dark current and noise obtained from Figure 4.2(c) are summarized in
Table 4.1. The dark current is the average current; noise is the standard derivation of the
current. Positive gate voltages can significantly reduce both dark current and noise due to
the increase of built-in potential and the consequently enlarged Schottky barrier width.
Table 4.1: Dark Current and Noise at Different Gate Voltages

Vg (V)

Dark Current (pA)

Noise (pA)

-3
0
3

-52.3
-13.9
-2.65

14.5
1.29
0.14

The photo-responses of the detector were measured by using the blackbody as the IR
source. There was no photo-response when zero gate voltage was applied (Figure 4.3(a))by
switching the shutter on and off at 0.5 Hz. When the magnitude of gate voltage was increased,
the photo-responses became clearer to be detected due to the suppression of dark current
and noise, and enhancement of photocurrent with proper electrostatic doping.
We also measured the spectral responses by using different bandpass filters in front of
the detector. This photodetector can sense IR signals with wavelengthes of shorter than
1.55 µm, as shown in Figure 4.4. The diameter of the SWCNT was 1.2 nm, identified by
the height information from an AFM. Diameter of 1.2 nm corresponds to bandgap energy of
0.7 eV, which equal to wavelength of 1.6 µm. There are two peaks in the spectral response:
one at 1.55 µm, the other at 0.83 µm. Since the CNT is 1D material, its density of states is
not continuous, but discrete. The two peaks can be explained by the Van Hove singularities
84

(a)

(b)

(c)

Figure 4.2: (a) The I-V characteristics of the transistor with Vg from -3V to 3V. (b) The
relationship between Vg and current, showing conductivity decreases with increasing positive
gate voltages. (c) The temporal dark current measurement with three different gate voltages.

85

ON

ON

IR

IR

OFF

OFF

(a)

(b)

ON

ON

IR

IR
OFF

OFF

(d)

(c)

(d)

Figure 4.3: The temporal IR response by radiating blackbody source to the detector at 0.5
Hz with (a) Vg=0 V, (b) 1 V , (c) 2 V and (d) 3 V.

86

Normalized Photocurrent (A. U)

1.2

1

0.8

0.6

0.4

0.2

0
0.6

1.1

1.6

2.1

Wavelength (um)
Figure 4.4: The spectral response of the SWCNT based photodetector.
occurs at E11 and E22 . The band structure of CNTs that have certain chirality can be
calculated through well-established models [7,50]. A Kataura plot was proposed by Hiromichi
Kataura to relate the diameter and CNTs and their bandgap energies in order to rationalize
experimental findings [51]. It was found that E22 and E11 of 1.2 nm-diameter CNT resides
at 0.83 µm and 1.55 µm.

87

4.2

Electrical Breakdown Control

It was demonstrated that SWCNT with 1.2 nm diameter can detect IR signal with wavelength
shorter than 1.55 µm. In order to detect MWIR signal, the diameter of CNTs should be
larger than 3 nm. Current synthesis techniques are difficult to grow uniform SWCNT with
this large diameter.
MWCNTs are CNTs with a number of concentric nano-cylinders with varied diameters.
MWCNTs may have much smaller bandgap than SWNCTs due to their large diameters.
However, selecting MWCNTs with desired bandgap energies remains a major challenge.
The as-synthesized MWCNTs are typically mixed with different bandgaps. In this section,
we will introduce how to control MWCNT bandgap energy using an electrical breakdown
system.
The electrical breakdown process was proposed by Collins et al. to peel off the outer
shell of MWCNTs with applied voltages [24]. The shells of a MWCNT can only sustained
certain power, therefore outer shells can be electrically removed when applied voltage beyond
the breakdown limits. However, this bandgap engineering process remains uncontrolled that
may result in cascade breakdown and cause the damage of MWCNTs.
In order to electrical breakdown the outer shell of MWCNTs in a control manner, we designed a controller to control the breakdown process in order to prevent cascade breakdown.
The breakdown voltage was increased gradually according to the optimal estimate from the
extended Kalman filter, which was designed based on the quantum electron transport model.
Utilizing the statistical information of the current data as the feedback, the modeling error
and random measurement noise are optimally minimized by the filter without sacrificing
the response speed. The current signal caused by breakdown process can be detected by

88

fault diagnosis technique, which generates fault residual that stops the breakdown process.
The controller is capable to generate stable breakdown voltage even with noisy and unstable
current. The random current jitters will not yield large fault residual to affect the fault
decision and cause a false alarm. When the fault residual exceeded fault threshold, a real
breakdown was detected. The controller withdraw the applied voltage immediately to prevent the cascade breakdown. The fault threshold is experimentally optimized to minimize
the false alarm and maintain quick response.
Figure 4.5 and Figure 4.6 show the breakdown process to control the MWCNT shell peel
off using the controller. The breakdown process started with a bias voltage of 0.5 V, an abrupt
current change was observed when the voltage was increased to 2.23 V (Figure 4.5). The
control voltage was immediately withdraw to stop the cascade breakdown. The conductivity
was decreased from 2.20×10−4 S to 1.52×10−4 S after one breakdown process due to the
removal or the defect generation on the outmost shell (Figure 4.6).
In addition, an experiment was conducted to verify the CNT breakdown phenomenon.
The results indicated the CNT breakdown happened at a particular region on the CNT. Single MWCNT was placed on three pairs of Au microelectrodes as illustrated in Figure 4.7(a).
I-V characteristics of different segments (AB, BC, and AC) of the CNT were measured and
studied before and after the breakdown process (see Figure 4.7). When a voltage was applied to the segment AB, the first CNT breakdown occurred between the segment AB. We
observed a significant change in the I-V curve of the segment AB while no obvious change
in the I-V curve of the segment BC occurred. Since the segment AC included both the
segments AB and BC, the I-V curve of the segment AC was also affected. Similarly, when
the voltage was applied to the segment BC, the second CNT breakdown occurred between
the segment BC, and the I-V curve between the segment AB remained unchanged because
89

6.00E-04

5.00E-04

Current (A)

4.00E-04

3.00E-04

2.00E-04

1.00E-04

0.00E+00
0

0.5

1

1.5

Bias (V)
Figure 4.5: Breakdown (BD) control curve.

90

2

2.5

2.50E-04
Before BD
AfterBD

2.00E-04
1.50E-04

Current (A)

1.00E-04
5.00E-05
0.00E+00
-5.00E-05
-1.00E-04
-1.50E-04
-2.00E-04
-2.50E-04
-1

-0.5

0

0.5

Bias (V)
Figure 4.6: The I-V characteristics of the device before and after BD.

91

1

the breakdown occurred on the segment BC only. This experimental result clearly indicated
that the breakdown position of a CNT can be adjusted by applying the electrical current at
specific locations.

4.3

Spectral Response of MWCNT based Photodetectors

In order to detect MWIR, the functional materials need to have a bangap energy that is
smaller than the energy of the MWIR photons. Bandgap energy of a MWCNT is proportional
to its diameter, but most of MWCNTs have a over-size diameter, resulting in a zero bandgap
energy. In order to control the bandgap energy to be a desired value, we used the breakdown
control system to consecutively tailor the outer shells of MWCNT.
The fabrication of this device is similar to that of the SWCNT photodetector, but deposited MWCNT instead of SWCNT using the dielectrophorisis deposition system. Figure 4.8 shows the I-V curves of the MWCNT photodetector before and with 6 breakdowns.
A linear I-V characteristic before breakdown indicates ohmic contact at both ends even
though both metals have different work functions, reflecting the zero bandgap or metallic
property of the as-prepared MWCNT. After serval consecutive breakdowns, current rectifying (curving I-V characteristic) became clearer, indicating the metallic MWCNT converts
to a semiconductor MWCNT. The MWCNT photodetector cannot sense MWIR before the
breakdown due to the zero bandgap energy of MWCNT with large diameter. After a number
of breakdowns, we found that the photodetector can detect the radiation from the blackbody.
In order to understand its spectral responses, we measured the photocurrent by placing bandpass filters in front of the photodetector. It was found that the photodetector can
92

Figure 4.7: (a) Illustration showing a single MWCNT connected to three gold microelectrodes. The I-V characteristics of the MWCNT at different segments: (b) Segment AB, (c)
segment BC, and (d) segment AC.

93

2.50E-04
Before BD
1st BD
2nd BD
3rd BD
4th BD
5th BD
6th BD

2.00E-04
1.50E-04

Current (A)

1.00E-04
5.00E-05
0.00E+00
-5.00E-05
-1.00E-04
-1.50E-04
-2.00E-04
-2.50E-04
-1

-0.5

0

0.5

1

Bias (V)
Figure 4.8: The I-V characteristics of a MWCNT photodetector before and after consecutive
breakdowns (BD).

94

measure a wide spectrum ranging from 1 µm up to 8.8 µm, as shown in Figure 4.9. The measurement was performed at room temperature. In other words, this photodetector can sense
MWIR and IR signal with longer wavelengthes at room temperature. In addition, we found
that the detection spectrum can be tuned by controlling the times of breakdown. It should
be noted that the detection spectrum of the MWCNT photodetector is much wider than
that of traditional photodetectors. We attributed this to that multiple layers of MWCNT
are contributing to the photon absorption and photocurrent generation. Conventional photodetectors need to stack materials that have different bandgaps to widen the detection
spectrum, highly increase the fabrication complexity and cost [52–55]. The MWCNTs with
controlled breakdown can be naturally a wide spectral material.

4.4

Chapter Summary

In summary, we measured the spectral responses of a SWCNT based photodetector, which
can detect IR signal with wavelength less than 1.5 µm using a SWCNT with diameter of 1.2
nm. In order to measure MWIR, CNTs with larger diameters are required. An electrical
breakdown system was developed to conquer the technical difficulties in synthesis CNT
with desired diameters and bandgap energies. The electrical breakdown system precisely
controlled the applied voltage to the MWCNT in order to prevent cascade breakdown. In
other words, the outermost shell of the MWCNT can be breakdown one by one. After a
number of consecutive breakdowns, the MWCNT based photodetector can detect MWIR at
room temperature. The MWCNT showed wide spectral responses, which was attributed to
the photon absorption at multiple layers of the MWCNTs.

95

Normalized Photocurrent (A. U)

1.2

1

0.8

0.6

0.4

0.2

0
0

1.5

3

4.5

6

7.5

Wavelength (um)
Figure 4.9: The spectral response of the MWCNT based photodetector.

96

9

Chapter 5
Plasmonic Lens using Optical
Antenna
CNT based photodetectors have shown outstanding performance by using CNT as sensing
element. However, detection wavelength is much larger than the diameter of the CNT,
resulting in small fill factor (absorption area). Bowtie antenna can confine light into a subwavelength volume based on plasmonic resonance, acting as a plasmonic lens. Therefore, integrating a bowtie antenna to CNT based photodetectors can highly improve photo-response
of the detectors. The electric field enhancement of bowtie antennas was calculated using the
device geometry by considering fabrication difficulties and photodetector structure. It is
shown that the electric field intensity enhancement increased exponentially with distance
reduction between the CNT based photodetector to the antenna. A redshift of the peak
resonance wavelength is predicted due to the increase of tip angles of the bowtie antennas.
Experimental results showed that photocurrent enhancement agreed well with theoretical
calculations. Bowtie antennas may find wide applications in nano-scale photonic sensors.

5.1

Introduction of Optical Antenna

We have demonstrated that CNT can be utilized as a functional element for photodetectors.
However, a critical challenge for obtaining high performance 1D photodetectors is the di97

mension incompatibility between photodetectors and optical wavelengths. The radial sizes
of CNTs are much smaller than detection wavelengths, resulting in low fill factor and responsivity. Motivated by this challenge, optical antennas were designed to integrate with
CNT based photodetectors so as to bridge the gap between nano-size sensing elements and
micro-size wavelengths. Due to plasmonic resonance, optical antennas can confine strong
optical near fields in a volume far below a cubic wavelength of excitation light. This can
greatly enhance photon absorption owing to the increase of coupling between CNT and vibrating electrical fields of photons. By integrating CNT based photodetectors with optical
antennas, the nano-scale photodetectors can be made with outstanding performance and
photon absorption efficiency.
In our design, the optical antenna performs as a plasmonic lens that funnels light into
the CNT based photodetector. This is analogous to the microlenses on top of conventional
charge coupled device (CCD) array to increase detector fill factor [56]. If the CNT based
photodetector occupied the same area as the antenna/CNT photodetector, they have same
fill factor. However, technical difficulties prevent the assembling of a large array of uniform
CNTs. Assembling a CNT thin film has been demonstrated, but the interaction between
CNTs in thin film resulted in the loss of their unique optical properties. Therefore, one
single CNT is more favorable and easier to manipulate in order to fabricate high performance photodetectors. But the challenge of one individual CNT as the core material for
photodetectors is the small diameter of CNT. The optical antenna can strongly concentrate
light into sub-wavelength volume that couples to the CNT in order to improve fill factor.
Optical antennas have invoked considerable interests, because of their ability to generate
large and localized electromagnetic fields. The basic principle of optical antennas is based
on the resonant behavior between excitation light and metal structure. When light inter98

acts with optical antennas that have proper dimensions, surface plasmons are generated due
to resonant coupling between the electromagnetic surface waves and oscillations of free electrons in the metal. Light irradiating on the optical antenna at frequencies below the resonant
frequency is reflected, since it induces motion in the charge carriers that act to screen out
the incident field [57]; when the frequency of light is above the resonant frequency, light
is transmitted because the charges are unable to respond quickly enough to screen out the
incident field; only at the resonant frequency, light is resonant coupling to free charges of
metals, and the incident waves will be absorbed and reradiated. By properly orienting the
reradiated waves, intensely localized near-fields can be generated. Confining or localizing
light into a small volume has found plenty of applications in many research fields, including
spectroscopy, microscopy, lithography, bio-sensing, chemical studies, and so on [58, 59]. In
recent years, optical antennas used to improve the performance of planar photovoltaic devices became a hot field [60]. However, very little research was conducted to improve the
performance of 1D photodetectors using optical antenna, although 1D materials become an
emerging platform in photovoltaic devices [61].
There are a variety of novel shaped optical antennas, such as nano-sphere, nano-rod and
nano-triangle monomers, nano-sphere dimmers, dipole antennas, bowtie antennas, Yagi-Uda
antennas, cross antennas, and so on [59]. Among all these optical antennas, dipole and
bowtie antennas are particularly suitable for integrating with 1D photodetectors because
of their geometries. Dipole antennas are nano-rod dimmers, while bowtie antennas are
nano-triangle dimmers, therefore both dipole and bowtie antennas can be considered as
two opposing tip-to-tip nano-triangles (non-zero apex width) with tip angles ranging from
0 degree (dipole antennas) to wider angles. The coupling between two pieces of metals in
these antennas concentrates the incident light into a nano-scale gap, which locates above a
99

CNT photodetector in order to enhance the interaction between the localized near-fields and
CNTs.
There were a number of approaches to improve a photodetector’s performance using plasmonic resonance: photocurrent enhancement was demonstrated in a Si photodiode via the excitation of plasmon resonances in spherical Au nanoparticles deposited on the Si surface [62];
a middle-wave IR photodetector, comprised of a metallic grating filled with HgCdTe, was
proposed to improve the signal-to-noise ratio by plasmonic resonance [63]; an optical antenna was utilized as a metal electrode for a Si Schottky diode to boost photo-detection [64].
Optical antennas were embedded in these photodetectors, acting as part of the photodetectors. In our design, the bowtie antenna performs as a plasmonic lens, which can be easily
integrated into all types of high performance 1D photodetectors [20, 23, 65]. In particular,
the bowtie antenna perfectly matches with the CNT photodetector due to the compatibility
of the diameter of CNT and gap distance of the bowtie antenna, because the electric field
(E-field) enhancement exponentially increases with decreasing gap distance [66].
The properties of free-standing optical bowtie antennas have been studied extensively
[67], and our group has demonstrated that dipole antenna can boost photo-response of 1D
IR detectors [68]. However, the design of bowtie antennas for 1D photodetectors was not
systematically studied. In this chapter we will present the design of bowtie antennas that
can be integrated with CNT photodetectors to highly improve the detection of IR signal
with wavelength of 830 nm. E-field enhancement with bowtie antennas was calculated to
seek the optimized tip angle and insulator thickness. Experimental results showed that
the photocurrent enhancement was 12.67 times by adding a bowtie antenna with precise
alignment and dimension control in fabrication. Calculation of relative energy difference
on the CNT photodetector at absence and presence of the bowtie antenna quantitatively
100

explained the photocurrent enhancement.

5.2

Design of Bowtie Antennas

The bowtie antenna consists of two tip-to-tip nano-triangles that can confine strong electromagnetic field between two triangles at resonance frequency. Light illuminating on the
bowtie antennas will generate instantaneous current, which oscillates at the frequency along
the polarization of the incident light. The instantaneous current is maximum near the center
of the nano-triangle, and decreases toward the edges of the antenna. This leads to a high
density of surface charge at the tip of the nano-triangles. Displacement current flows into the
small air gap of the bowtie antennas, reradiating the absorbed energy. The strong coupling
of these two opposing tip-to-tip triangles brings the center of the reradiating source to the
center of the gap [69]. The maximum intensity enhancement can be obtained at resonance
frequency, which is mainly determined by the dimension and materials of antennas.
One thing should be noted is that the depletion region of the CNT photodetector may
extend into the whole tube of the CNT, since the charge distribution of the 1D systems has
a long-range tail, extending over the entire CNT [15]. In other words, the entire CNT can
be considered as depletion region. Therefore, the optical antennas can not only be located
on the contacts, but they also can be placed on top of the CNT body. The relative position
of the CNT photodetector and bowtie antenna is shown in Figure 5.1. A thin insulator
(Si3 N4 ) is sandwiched between the bowtie antenna and the photodetector (Figure 5.1(c)),
with the gap of bowtie antenna aligning to the center of CNT channel. The geometrical
parameters are also denoted in Figure 5.1: where d is the gap distance, w is the apex width,
L is the length of the bowtie antenna, θ is the tip angle, and t is the thickness of the insulator

101

between antenna and CNT.
This CNT photodetector is designed for detecting IR signal at 830 nm, thus the bowtie
antenna should resonate at or close to this wavelength. Resonance wavelength is very sensitive to the dimensions of bowtie antennas, but several parameters are fixed in simulation
considering the fabrication constraints, and some parameters have been extensively studied [70]: the gap distance d is set to be 30 nm for easier alignment of the bowtie antenna to
the photodetector; apex width is 30 nm to leave some tolerance for EBL; antenna length L
is 400 nm because it is close to half-wavelength of the IR signal; the bowtie antenna is made
of 3 nm Ti and 20 nm Au. The width of both Au and Ti electrodes is 400 nm, and the gap
between two electrodes is 1000 nm. The following simulation was based on the geometry
including both electrodes and bowtie antennas, as shown in Figure 5.1.
The E-field enhancement spectra of bowtie antennas will be simulated with varying insulator thickness t and tip angle θ, not only since these two parameters can be easily tuned
in real device fabrication, but they also have significant impacts to the E-field intensity
enhancement.
Figure 5.1(a) also shows the direction of incident light polarization and the coordinate
system for all numerical simulations using ANSYS HFSS: a plane wave is normally incident
on the structure from the top with the electric field being polarized along the x axis, which is
parallel to the axis between two triangles, with electric field component amplitudes of Ex =1
V/m, Ey =Ez =0.

5.2.1

Insulator Thickness Dependent E-field Enhancement

The plasmonic resonance of the bowtie antennas was first investigated by varying the insulator thickness t between the antenna to the CNT photodetector. This is a critical parameter,
102

Ex
Hy

Au

antenna

kz

Si3N4

CNT
Ti
quartz substrate
(a)

w
g

Ti

CNT
Au

L

(b)

antenna
Si3N4

CNT
t

Ti

Au

(c)
Figure 2 (a) A 3D geometry shows the relative position between the CNT photodetector and
Figure 5.1: (a) A 3D geometry shows the relative position between the CNT based photodetector and the bowtie antenna. The incident light with electric field polarized in the
x direction propagates in the z direction. (b) Top view of the device, and parameters of
the bowtie antenna: L=400 nm is the length of the antenna, g=30 nm is the gap of two
pieces of the antenna, w=30 nm is the apex width, and is the tip angle. (c) Side view of the
device, showing the CNT based photodetector and bowtie antenna sandwiching the Si3 N4
layer that has a thickness of t.

103

since the photodetector needs to be placed in close proximity to the optical antennas in
order to fully take advantage of near fields [71]. The insulator thickness dependent E-field
distribution at the substrate plane (the plane between Si3 N4 and quartz substrate), where
CNT located, was calculated based on equilateral Au triangles ( θ=60 degrees) with d=30
nm, w=30 nm, L= 400 nm. The CNT photodetector and optical antenna are on the top of a
quartz substrate, with a plane-wave light polarized in the x direction normally illuminating
the device from the top.
The resonance wavelength for this antenna is at 845 nm with t=10 nm. Figure 5.2(a)
shows the E-field distribution between two electrodes at the substrate plane for the device
with t = 10 nm at its resonance wavelength. The E-field is confined into the gap of the
bowtie antenna with the maximum amplitude located at the center of the gap due to the
strong coupling between two nano-triangles. The E-field distribution is in the shape of a
horizontal bowtie, and the amplitude of the E-field decreases as the distance to the center
of the gap increases in the horizontal direction. The illuminating light had E=1 V/m, while
the peak amplitude at substrate plane became E=14.608 V/m at resonance after adding the
bowtie antenna, with an enhancement of 14.608 fold.
The peak E-field intensity enhancement Max(| E |2 ) as a function of insulator thickness
at resonance wavelength was shown in Figure 5.2(b). The E-field distributions with thickness
of Si3 N4 ranging from 10 nm to 50 nm were investigated. The blue curve shows the peak
intensity enhancement at the substrate plane: the enhancement decreased from 213.1 to
26.7 when the thickness of the Si3 N4 increased from 10 nm to 50 nm. The red curve
in Figure 5.2(b) is the exponential fit of the data, which matches very well with the blue
curve. The resonance wavelength of the antenna redshifted from 845 nm to 1034 nm with the
thickness increasing of Si3 N4 layer from t=10 nm to t=50 nm [72]. The inset of Figure 5.2(b)
104

antenna
x

2

Peak Intensity Enhancment Max(|Ε| )

Ti

y

Au

(a)

250

250
200

200

150
100

150

50
0
10

100

50

20

30

40

50

Simulation Data
Exponential Fit

0
10

20

30

40

50

Thickness of Insulator t (nm)
(b)

Figure 5.2: (a) The E-field distribution of the resonance frequency (845 nm) of the bowtie
antenna with θ=60 degrees, and t=10 nm. The x-y axis show the coordinate for simulation.
(b) The t dependent peak field intensity Max(| E |2 ), blue curve is the simulation values,
and red curve is the exponential fit. The inset shows field intensity enhancement | E |2 at
wavelength of 830 nm.
105

shows that the E-field enhancement at wavelength of 830 nm, which corresponds to the
wavelength of our laser source, decreases dramatically by increasing the insulator thickness.
The simulation results indicate that the field intensity enhancement decreases significantly
with the increasing thickness of the insulator, because evanescent waves decay exponentially
from the antenna. Therefore, in order to fully use the power of the antenna, the photodetector
needs to be fabricated close to the bowtie antenna.

5.2.2

Tip Angle Dependent E-field Enhancement Spectra

Not only is the tip angle of the bowtie antennas easily modified in real device fabrication, but
it also has critical impact on both resonance wavelength and E-field intensity enhancement.
The E-field spectra as a function of the tip angle were calculated with d=30 nm, w=30 nm,
L= 400 nm, and t=10 nm. Figure 5.3 shows the E-field spectra of bowtie antenna with tip
angles of 0 degrees (dipole antenna), 30 degrees, 60 degrees, 90 degrees, and 120 degrees.
The spectra vary significantly with different tip angles. It is shown that bowtie antennas
cover wider spectrum with larger tip angles. Bowtie antennas can be considered as a twodimensional analogue of biconical antenna, therefore wider tip angles are expected to possess
broader bandwidth. The bowtie antenna with tip angle of 120 degrees shows field intensity
enhancement in a broadest spectrum. What is more, two resonance wavelengths was found
with a fundamental resonance wavelength at 1.13 µm and a high-order resonance at 0.77
µm. High-order resonance is easier to be excited for large angle Bowtie antennas, since
larger angle bowtie antennas are prone to overcome the symmetry limitation of octupolar
resonance, while octupolar resonance is forbidden in the 0 degree bowtie antenna under
normal incidence [73].
The peak E-field intensity enhancement increases with wider tip angles when the tip angle
106

240
0Degree
30Degree
60Degree
90Degree
120Degree

Intensity Enhancement (

2

)

200

160

120

80

40

0
0.6

0.8

1.0

1.2

1.4

1.6

Wavelength ( m)
Figure 5.3: E-field intensity enhancement | E |2 spectra of bowtie antennas with tip angles
ranging from 0 degrees to 120 degrees. The inset shows a scanning electron microscopy
(SEM) image of a bowtie antenna, denoting the tip angle θ.

107

is smaller than 60 degrees, while it reduces when the tip angle is larger than 90 degrees. In
addition, the resonance wavelengths, where the peak intensity enhancements are obtained,
redshift with increasing tip angles. Resonance wavelength is proportional to the length of
the triangle edges of bowtie antennas, and thus resonance wavelength redshifts as tip angle
increases [74]. This can be explained as follows: the two opposite Au/air interfaces in one
cross section vertical to the antenna axis are far away from each other in the y direction, thus
the local surface plasmon polariton is confined at one Au/air interface and not affected by the
opposite one, especially when the cross section is close to the extremity edges. As a result,
the bowtie antennas are dominated by isolated surface plasmon polaritons that propagate
along the triangle edges; consequently, a standing wave along the triangle edges is generated,
with resonance wavelength scales with the length of the triangle edges, therefore redshifts as
the tip angle increases [73]. The length of triangle edges of the bowties increase from 185 nm
to 272 nm when the tip angle is widened from 0 degrees to 90 degrees, with an increment
of only 87 nm. As a result, the cross section along y axis from the tip to extremity edges
changes gradually, and thus resonance wavelength differs slightly. The length of triangle
edge increases to 383 nm when the tip angle becomes 120 degrees. This major length change
causes a rapid change of local cross section, resulting in a major shift of the resonance
wavelength.
The simulation results show that larger angle bowtie antennas can cover broader spectrum
together with multiple resonances, which are suitable for photodetectors with wide spectrum
detection. In our experiment, we were detecting IR signals with single wavelength of 830
nm. Bowtie antennas with a 60 degree tip angle were fabricated to verify the field intensity
enhancement of the antennas, since the highest peak field intensity enhancement is obtained
at this tip angle, and whose resonance wavelength is close to 830 nm.
108

5.3

Fabrication and Experimental Results

In order to investigate the performance of bowtie antennas, Au bowties were fabricated on top
of a CNT photodetector as depicted in Figure 5.1(a). Photocurrent of CNT photodetector
with absence and presence of bowtie antennas was compared so as to examine the field
intensity enhancement of the antennas.

5.3.1

Fabrication Process

The fabrication process starts with EBL an Au electrode on top of a quartz substrate. Prior
to EBL, the quartz substrate was cleaned in acetone and bombarded with oxygen plasma for
5 minutes to remove residues. A 360 nm-thick layer of poly-methyl-methacrylate copolymer
resist (PMMA/MMA) (9% in Ethyl Lactate, from MicroChem Corp.) was spun at 4000
rpm on the substrate, and the sample was baked in a hot plate at 175o C for 5 minutes to
harden the resist and remove solvent. It was followed by spinning a layer of poly-methylmethacrylate (PMMA) (29% in Chlorobenzene, from MicroChem Corp.) at 4000 rpm with a
thickness of 130 nm to form a bilayer resist. After that, the sample was baked in a hot plate
at 175o C for 30 minutes to remove solvent residue. A thin layer of Au with thickness of 10 nm
was deposited on top of the resist to avoid charging effect on the sample during EBL writing.
The resist was exposed with the patterned of an Au electrode with width of 400 nm and
some alignment mark at an acceleration voltage of 30 kV and an area dose of 290 µC/cm2 .
After exposure, the thin Au layer on the resist was stripped using gold etching for 10 seconds,
and then the resist was developed using a 1:3 methyl-iso-butyl-ketone (MIBK) in isopropyl
alcohol (IPA) solution for 29 seconds. Following development, the sample was rinsed with DI
water, and bombarded by oxygen plasma for 90 seconds to remove residual resist. A 3 nm-

109

thick Ti adhesion layer and 40 nm-thick Au layer were deposited using a thermal evaporator.
The Au electrode and alignment mark were transferred to the substrate via a lift-off process
using acetone (Figure 5.4(a)). Following the same processes with precise alignment, another
electrode made by Ti was fabricated on the quartz substrate, with a gap of 1 µm to the Au
electrode (Figure 5.4(b)). A droplet of CNT suspension in ethanol alcohol was dispersed at
the vicinity to the gap of electrodes, and an AC voltage of 1 Vpp and 10 kHz frequency was
applied between the electrodes through the DEP deposition system. With the assistance
of an atomic force microscope (AFM) manipulation system, an individual single wall CNT
(SWCNT), with diameter of 1-2 nm, was assembled between the electrodes, as shown in the
inset of Figure 5.4(c), which is the AFM image of the CNT based photodetector. A detailed
deposition process was introduced in [32]. After that, a layer of Si3 N4 with thickness of
10 nm was deposited on top of the device using plasma-enhanced chemical vapor deposition
(PECVD). The PECVD is a uniform deposition process, but Figure 5.4(d) shows the Si3 N4
only in the region between two electrodes for making the schematic clearer. Eventually, a
bowtie antenna with two tip-to-tip equilateral Au triangles was fabricated on top of the
Si3 N4 layer by aligning the gap of bowtie antenna to the SWCNT between two electrodes.
The fabrication of the bowtie antenna started with spinning a 100 nm-thick layer of PMMA
at 6000 rpm on the device, and the sample was baked on a hot plate at 175o C for 30 minutes.
After that, an Au layer of 10 nm was deposited on top of the device to prevent charging
effect. In order to precisely align the antenna to the detector, AFM was used to record the
position of the SWCNT for designing antenna layout. The bowtie antenna was layout using
single pass lines to reduce proximity effects in order to increase resolution. During EBL, the
resist was exposed with a line dose of 0.75 nC/cm to form the pattern of bowtie antenna.
After EBL, the thin Au layer was stripped, and the resist was developed using a 1:3 MIBK
110

in IPA solution for 29 seconds. Following development, a 2 nm-thick Ti adhesion layer and
20 nm Au layer were then deposited using a thermal evaporator. The antenna pattern was
transferred to the substrate via a lift-off process Figure 5.4(e).
Figure 5.4(f) shows a 3D AFM image of one of the devices integrating bowtie antenna
to a CNT based photodetector. The SWCNT was not located at the center of the axis of
two electrodes, thus the position of bowtie antenna was adjusted according to the position
of the CNT. AFM images cannot show accurate dimensions in x-y plane due to the image
broadening, which is caused by the large radius of curvature of the apex of AFM tips,
therefore two triangles of the bowties in the AFM image seems connecting. SEM was utilized
to measure the real dimensions of the bowtie antennas. The gap of bowtie antennas (g) was
30±10 nm, the apex width (w) of the triangles was 30±10 nm, length of the bowtie antenna
(L) was 400±30 nm.

5.3.2

Experimental Results

Photocurrent measurements were carried out on CNT photodetectors with absence and
presence of bowtie antennas so as to investigate the field enhancement of the antennas. An
IR laser, which had 830 nm wavelength and polarization in x direction (the axis of bowtie
antenna), illuminated on the CNT photodetector that was housed in a testing chamber, and
electrical signals were measured using Agilent semiconductor analyzer 4156C.
Photo-response of the CNT photodetector without bowtie antenna was firstly recorded
at room temperature by switching the IR laser on and off at 0.2 Hz for several cycles (blue
curve of Figure 5.5). Photo-response of the CNT photodetector was measured again after
the fabrication of the bowtie antenna (red curve of Figure 5.5). The comparison of photoresponses was based on the amplitude change of photocurrent. Figure 5.5 shows that the
111

Au

Ti

(a)

(b)

Au
CNT
Si3N4

CNT

Ti

(d)

(c)

Au

antenna

CNT
Ti

(e)

(f)

Figure 5.4: (a-e) Fabrication process of integrating a bowtie antenna to a CNT based photodetector. The inset of (c) is an AFM image of a SWCNT bridging a Ti and an Au electrode.
(f) A 3D AFM image of the device.

112

1.00E-09

ON
IR

OFF

Current (A)

1.00E-10
without antenna
with antenna
1.00E-11

1.00E-12

1.00E-13
0

5

10

15

20

25

Time (S)
Figure 5.5: Photo-responses of the CNT based photodetector without and with bowtie antenna by switching IR laser on and off at 0.2 Hz.
dark current was 2×10−12 A at the absence of IR illumination, and the current was increased
significantly upon light illumination because of the photons generated electron and hole pairs
within the CNT, which were separated in the Schottky barrier to produce photocurrent. The
current dropped to the dark current when the IR laser was turned off. The reproducibility of
the photo-response was excellent. The photocurrent for CNT photodetector without antenna
was 1.5×10−11 A, and it was increased to 1.9×10−10 A after the fabrication of the bowtie
antenna, with a photocurrent enhancement of 12.67 times.
The bowtie antenna functions as a plasmonic lens to funnel light into the CNT, which

113

has a size much smaller than the gap of bowtie antenna. Hence, the area of detector material
(physical cross section) and absorption probability did not change with the absence or presence of the antenna. With the presence of bowtie antenna, the photon energy density on the
detector material was enhanced due to sub-wavelength near field confinement. It has been
defined that the absorption cross section of the detector is the total light power absorbed
by detector material normalized by the incident photon flux [63]. Therefore, calculation of
relative energy intensity difference on the CNT body at absence and presence of antenna
will directly reflect the photocurrent enhancement.
The field intensity increment on the CNT can be calculated by integrating the E-field
∫
intensity along the CNT by I = | E(y) |2 dy (Figure 5.2(a)), since the depletion region of
the asymmetric CNT photodetector extended to the whole CNT. In other words, the photocurrent enhancement was caused by the E-field intensity increment on the CNT. Without
the nano-antenna, the total field intensity on the CNT equal to 10−6 V because Ex(y)=1
V/m and y=1 µm. The field intensity distribution with the antenna was calculated by using
the parameters that were the same as fabricated device and experimental setup (d=30 nm,
w=30 nm, L= 400 nm, t=10 nm, θ=60 degrees, and radiation wavelength=830 nm). By integrating the field intensity along the y axis between two electrodes (where the CNT located
in Figure 5.2(a)), the field intensity was enhanced to 1.48×10−5 V, with a field enhancement
of 14.87 fold. Therefore, the photocurrent enhancement is in good agreement with the simulation. Differences are attributed to geometry differences between the actual antennas and
the simulation model, and some loss were occurred in the antennas due to edge and surface
roughness.
In order to validate that the photocurrent enhancement was caused by the light confinement of the nano-antenna, rather than the contact condition change during fabrication, I-V
114

2.00E-06

Current (A)

1.50E-06
without antenna
with antenna

1.00E-06

5.00E-07

0.00E+00

-5.00E-07

-1.00E-06
-1

-0.5

0

0.5

1

Bias (V)
Figure 5.6: I-V characteristics of the CNT photodetector without and with the bowtie
antenna.
characteristics of the photodetector without and with antenna were measured and shown in
Figure 5.6. The I-V characteristics showed current rectifying at low voltage, implying the
existence of Schottky barriers. In addition, the bias-dependent current of the CNT photodetector with absence and presence of nano-antenna were almost identical, which indicated that
the contact condition did not change with the fabrication of the nano-antenna, since electrical signal was extremely sensitive to contact condition. This verifies that the photocurrent
increment was caused by field intensity enhancement from the bowtie antenna.

115

5.4

Chapter Summary

The bowtie antenna functioned as a plasmonic lens that funnel confined light onto a CNT
based photodetector. The design of bowtie antennas integrating to the CNT based photodetectors was studied. The E-field intensity enhancement of bowtie antennas with a small gap
located on top of the asymmetric CNT based photodetector was calculated. It was shown
that the peak field intensity of the antenna was exponentially decreased with the distance
between the antenna to the CNT based photodetector. The resonance wavelength of the
bowtie antenna redshifts with increasing tip angles. A fabrication process was developed to
integrate the bowtie antenna with the CNT detector. It should be noted that the 1D CNT
detector is perfectly matched with the bowtie antenna, since the diameter of CNTs is in
the same scale as the gap of the bowtie antenna. The experimental results agreed well with
the simulation by measuring photocurrent enhancement. The nano-antenna will become an
important component to the nano-scale photodetectors to overcome the diffraction limit.

116

Chapter 6
Photonic Crystal Cavity
It has been demonstrated that optical antenna can confine light into sub-wavelength volume to improve the fill factor of CNT based photodtectors. However, the fill factor of
antenna/CNT photodetector is still small, limited by the small absorption area of optical
antennas. In order to address this problem, a photonic crystal cavity, using photonic crystal
with a point defect made of HfO2 , was fabricated on top of the CNT photodetectors in order
to enhance the local photon density. By drilling holes, with size of the scale of light wavelengths, in the parylene slab, the periodic change of reflective index forms photonic crystal.
Replacing one of the holes with HfO2 as a point defect, most of the light will be trapped
in this defect. A photonic crystal was designed to concentrate the infrared light with 1064
nm wavelength, and 91 % of the light can be confined into the HfO2 defect. The fabrication
process of this photonic crystal cavity is introduced, and photocurrents of a CNT based
photodetector are compared under the circumstance of without and with photonic crystal.
It is found that the photocurrent is enhanced about 5 times after the fabrication of photonic
crystal cavity.

6.1

Introduction of Photonic Crystal Cavity

The outstanding performance of CNT photodetectors, inherited from the nano-scale size of
CNTs, which, however, cause low absorption due to the small absorption area of the CNT

117

photodetectors, since diameters of CNTs (including SWCNTs and MWCNTs) are normally
less than 100 nm. In other words, the CNT detectors have low fill factor. In order to
increase the fill factor, light needs to focused into a small region that has size comparable to
the detectors. Traditional lens is difficult to focus light into this miniature size due to light
diffraction and lens imperfection. Motivated by this challenge, a parylene photonic crystal
slab with a point defect made of HfO2 was designed to integrate with a CNT photodetector.
Photonic crystals are artificial optical materials with periodically structured dielectrics
of alternative reflective indexes to control the propagation of light. Photonic crystals has
periodic structure that is analog to solid state crystals, but with dimensions of the order of
optical wavelengthes. As a result, photonic crystal can guide light with frequency bands and
frequency bandgap, in the same fashion as atomic lattices conducting electrons with energy
bands and bandgap. If the optical frequency falls in frequency bands, a photonic crystal
becomes a perfect optical conductor, which can be employed to guide the light propagation.
On the other hand, when the photon frequency falls in a photonic bandgap, a photonic
crystal acts like an optical insulator, which can be used to build micro-scale optical mirrors,
resonant cavities, and waveguides [75].
When the symmetric arrangement of the photonic crystal was broken, it operates in
defect modes. Two most basic defects are point defects [76], which remove a single hole
of the photonic crystal, and line defects [77], which remove a row of dielectric holes in
the symmetric structure. Photonic crystal with a point defect, which was termed photonic
crystal cavity, can confine light in the defect region, functioning like a lens [78]. Fill factor of
CNT photodetector can be highly improved if the defect of photonic crystal cavity aligned
to the absorption area of the CNT photodetectors.

118

6.2

Design and Fabrication

An AFM image of a photonic crystal cavity integrating with a CNT photodetector is shown
in Figure 6.1. A 2D photonic crystal slab made by parylene/air resides on top of a CNT
photodetector with a HfO2 defect located on one of the contacts between CNT and electrode.
The photonic crystal cavity we used is photonic crystal based quantum resonance cavity.
When the resonance wavelength of the photonic crystal is close to the wavelength of light,
light will be trapped into the defect, as shown in Figure 6.1(c).
3D photonic crystals have periodic structures in three coordinate axis, similar to the real
solid state crystal structures [79]. However, the fabrication of 3D photonic crystal involves
complicated fabrication processes, in particular, to integrate with CNT photodetectors.
2D photonic cavity can accomplish same function as a micro-lens to confine light in the
point defect, while with simpler structure and fabrication process. The 2D photonic crystal
cavity consists of a parylene slab with air holes for total internal reflection of the light in
the vertical direction, and a 2D photonic crystal is to provide lateral localization within the
point defect. Traditional optical cavity is created by removing an air hole (point defect) in
the photonic crystal, while a HfO2 was utilized as point defect in this design. The refractive
index of HfO2 (1.90 [80] ) is higher than parylene (1.62 [81] ), which is higher than air (1.00),
thereby total internal reflection will be more likely to occur between the HfO2 and parylene,
resulting in more light confining in the defect region.
As mentioned above, the photonic crystal acts as optical insulator when optical frequency
falls in the bandgap frequency. If the optical frequency is within the bandgap frequency of
the 2D photonic cavity, light illuminating on the photonic crystal has nowhere to go except
the point defect, functioning as confining all the light into a small region. Therefore, proper

119

Photonic Crystal
CNT Photodetector
Ti

Au
CNT

HfO2 Defect

(a)
Parylene

Air

Photonic Crystal
Ti

CNT

HfO2
Au

Substrate
(b)

Ti

Au

high photon density in the defect
(c)
Figure 6.1: (a) AFM image of a parylene photonic crystal with a HfO2 defect on the top of
a CNT detector. (b) Cross section of the device. (c) The working principle of the photonic
cyrstal cavity by confining the light radiating on photonic crystal into the defect.

120

dimensions of the photonic crystal need to be found in order to direct light into the defect.
A laser with 1064 nm wavelength, corresponding to 282 THz, was used to test the CNT
photodetectors. By simulating the dispersion relationship of the photonic crystal, it was
found that the middle of bandgap frequency equal to 282 THz when the lattice constant
(the distance between air holes) a= 509 nm, and diameter of air hole d= 407 nm.
The electric field distribution of the photonic crystal cavity with 9×9 air holes was simulated. When the defect material of photonic crystal cavity is parylene, 31 % light focuses
into the defect. The energy inside the defect increase to 91 % when the defect became HfO2 .
Figure 6.2(b) shows the electric field distribution of the photonic crystal cavity with a HfO2
defect: most of the electric field concentrates in the defect area.
The fabrication process began with fabricating Ti and Au electrodes with a insulating
gap of 0.5 µm on top of a quartz substrate by EBL, thermal evaporation, and lift-off. After
that, a single MWCNT was deposited between the electrodes using the dielectrophoresis
(DEP) deposition system assisted by the AFM manipulation system [27]. A single MWCNT
connecting to two electrodes is shown in Figure 6.3(a). It is followed by fabricating an HfO2
cylinder with 0.5 µm-diameter on top of a contact between Ti and CNT by EBL, e-beam
evaporation, and lift-off, as shown in Figure 6.3(b). A layer of parylene will be deposited on
the device, and air holes will be drilled by plasma etching after EBL to define the dimensions
of the air holes, as shown in Figure 6.3(c).

6.3

Experimental Results

The testing system consists of a chamber that houses the CNT photodetector, an IR laser
of 1064 nm wavelength, and Agilent 4156C semiconductor analyzer.The photo-response in

121

HfO2 Defect

(a)

HfO2 Defect

(b)
Figure 6.2: (a)The simulation geometry of the photonic cavity with parylene slab and a
HfO2 defect. (b) Electric field distribution of a 1064 nm IR light in the photonic cavity,
most of the electric field is confined within the point defect.

122

Ti

HfO2 defect

CNT

Au

(a)

(b)

Photonic crystal
HfO2 defect

(c)
Figure 6.3: Fabrication process of the testing device (a) CNT Schottky photodiode, (b)
CNT photodetector with a HfO2 defect,(c) Photodetector with photonic crystal and a HfO2
defect.

123

terms of photocurrent with absence of photonic crystal was firstly measured using the semiconductor analyzer by focusing the IR laser on the CNT photodetector. Temporal photoresponses of CNT detector with absence and presence of photonic crystal cavity were measured, showing the photocurrent enhancement of about 5 times after adding photonic crystal
cavity. However, the photocurrent enhancement may caused by the misalignment between
laser and CNT photodetector in two measurements, since the size of laser spot used in our
experiments is larger than the CNT detector, and photocurrents were recorded when highest
photocurrent was obtained after aligning laser spot with CNT detector. In order to have a
better understanding of the photonic crystal cavity as a micro-lens, scanning photocurrent
images of the CNT detector with and without photonic crystal cavity were studied.
Figure 6.4 illustrates the setup to obtain scanning photocurrent images. Laser spot scans
around the CNT photodetector along the scanning path as shown in Figure 6.4. The scanning
photocurrent image of a CNT photodetector was firstly obtained without a photonic crystal
cavity. After that, a photonic cryatal cavity made from parylene slab with a HfO2 defect was
fabricated by aligning the point defect on top of the detector. The photocurrent distribution
of the CNT detector with photonic cavity was recorded by scanning the laser around the
detector.
A 3D mapping of position dependent photocurrent was obtained, as shown in Figure 6.5.
Photocurrent was less than 1 nA without the photonic crystal cavity in a 20×20 µm2 area.
After the fabrication of the photonic crystal cavity, photocurrent was enhanced to around 5
nA in the same area, indicating that the photonic crystal cavity increases the light intensity
approximately 5 times within the area of the detector. This result shows that photonic cavity
functioned as an effective micro-lens to increase the interacting of CNT photodetector and
light.
124

Laser spot

Scanning path

CNT detector

Figure 6.4: Experimental setup to obtain scanning photocurrent images.

125

x 10

-9

x 10
5

Photocurrent (A)

6

4

4

3

With Photonic Crystal
Without Photonic Crystal

2

2

0
20

-9

1

10

0 0

10

0
20

X (um)

Y (um)

Figure 6.5: Scanning photocurrent images of the CNT photodetectors without and with
photonic crystal cavity. The insets are the AFM images of detectors without and with
photonic crystal cavity.

126

6.4

Chapter Summary

In summary, a photonic crystal cavity with a HfO2 defect was demonstrated to improve the
performance of the CNT photodetectors. Optical antenna has demonstrated as a plasmonic
lens to focus light into a sub-wavelength volume, resulting in improved fill factor. However,
the absorption area of antennas is limited. In order to further improve the fill factor, a
photonic crystal cavity was utilized as a micro-lens to focus light into CNT photodetector.
The photonic crystal cavity comprise of a parylene slab with air holes and a HfO2 point
defect, which is placed on top of a CNT photodetectors. The simulation results show that
about 91 % of light in a area of 9×9 air holes can be confined into the defect, and as a result,
photocurrent was increased 5 folds with the presence of the photonic crystal cavity.

127

Chapter 7
Infrared Sensing Systems Using a
Single Nano-photodetector
We have demonstrated a 3×3 array of CNT photodetectors. However, it is difficult to
manufacture a megapixel count focal plane array as in conventional camera structure, because
IR detectors were assembled by bottom-up approaches.
In this chapter, we will introduce a novel IR sensing system, which enables utilizing the
promising performance of 1D-photodetectors, and overcoming the difficulties to obtain high
resolution and efficient IR images. The sensing system is constructed based on an emerging
compression theory: compressive sensing. A given image can be compressed if the image
is sparse or can be sparsely represented in another basis, for example piecewise smooth
images are sparse in wavelet basis. Traditional compression algorithms perform compression
after sampling, while compressive sensing allows image compression during sampling so as
to reduce the number of measurements and data storage capacity [82] [83].
Our sensing system mainly consisted of a digital micromirror device (DMD), a CNT photodetector, a high speed readout system, and an image reconstruction computer. Instead of
projecting IR images onto the focal plane array of image sensors, IR images were projected
onto a DMD, which generates measurement matrices according to compressive sensing algorithms [84]. The images are compressively sampled using a single CNT photodetector via

128

recording the linear projection of IR images onto the measurement matrix. IR images are
recovered through a compressive sensing reconstruction algorithm based on the measurement results. The experimental results show that the new imaging system can achieve the
superb performance enabled by CNT based IR detectors, and, at the same time, overcame
its difficulties to obtain high resolution and efficient imaging.

7.1

Compressive Sensing Theory

The core of our IR sensing system is compressive sensing theory. In this section, the basics of
compressive sensing will be briefly reviewed. A recent complete introduction of this topic can
be found in [85]. Compressive sensing is a new paradigm for signal sampling, which has found
a wide spectrum of applications in imaging [86], control and robotics [87] [88], geophysical
data analysis [89], communication [90], and etc. Traditional sampling approaches require
the sampling rate to be at least twice the maximum frequency of the signals according to
the Nyquist-Shannon sampling theorem. Nevertheless, compressive sensing shows that the
sampling rate can be significantly reduced if the signal satisfies general sparse conditions.
Specifically, this new sensing paradigm directly samples the signal in compressed form, and
then recovers the original signals by advanced optimization or greedy algorithms [91].

The general compressive sensing theory can be summarized as follows. Assume an unknown high dimensional signal x ∈ Rn is projected to lower dimensional measurements
y ∈ Rm :

y = Φx
129

(7.1)

with Φ ∈ Rm×n (m < n) being the measurement matrix. Then the unknown x can be
recovered from y, if x and Φ satisfy certain conditions. In what follows, the conditions of
both x and Φ for successful recovery will be briefly described.
One condition for recovery is that the original signal is sparse (or approximately sparse)
or has a sparse representation in some basis. A vector x ∈ Rn is S-sparse if the support,
i.e., the number of nonzero elements, of x is at most S. Denote the set of all the S-sparse
vectors by ΩS . Formally,
ΩS = {x ∈ Rn : ||x||0 ≤ S}
where ||x||0 denotes the support of x.
In some cases, x ∈ Rn may not be sparse in its original form but sparse in some orthonormal basis. Let Ψ be such a basis. In this case, x can be expanded in Ψ as: x = Ψ˜ with
x
x ∈ ΩS . The compressive sensing can be rewritten as:
˜

y = Φx = A˜
x

(7.2)

with A = ΦΨ being the new measurement matrix. Without loss of generality, we will
assume the original signal is sparse (or approximately sparse) and use x to represent it in
the following discussion.
In general, the exact S-sparse cannot be achieved in real-world signals, but they can
be well approximated by an S-sparse signal. The best S-sparse approximation for a vector
x ∈ Rn is a vector xS ∈ Rn by retaining the largest S elements in x and setting the other
elements to zero.
130

Another condition for recovery is that the measurement matrix needs to satisfy the restricted isometry property (RIP), which is defined as follows:

Definition 1. A matrix A ∈ Rm×n with m < n satisfies the RIP of order S if there exists
a constant σS ∈ (0, 1) such that for any x ∈ ΩS , we have

(1 − σS )||x||2 ≤ ||Ax||2 ≤ (1 + σS )||x||2
2
2
2

(7.3)

Note that measurement matrix can follow other conditions, such as spark [92], null space
property [93], or coherence [92], but the RIP is the most powerful one because it can deal
with the case when the measurements are corrupted by noise.

With the above two conditions, we can discuss the original signal recovery. To find the

sparsest solution, the recovery problem can be stated as:

min ||ˆ||0 subject to Aˆ = y
x
x

(7.4)

x is the recovered signal. The above problem, however, is NP-hard. As an alternative, the
ˆ

l1 minimization can be used:

min ||ˆ||1 subject to Aˆ = y
x
x

131

(7.5)

Using the l1 minimization, the recoverability is guaranteed by the following theorem:

Theorem 1 (Cand´s [94]). Suppose the matrix A satisfies the RIP of order 2S with constant
e

σ2S <

√

2 − 1, then the solution to Eq. (7.5) satisfies:

√
||ˆ − x||2 ≤ C||x − xS ||1/ S
x

(7.6)

with C a constant related to σ2S .
The theorem suggests that the error between the recovered signal and the original signal
is bounded by the error between the original signal and the best S-sparse approximation.
Note that the noise measurement can also be handled under the RIP condition [95], which
will not be discussed here.

With the above theorem for recoverability, the next question is how to design the measure-

ment matrix satisfying the RIP condition. The goal is to recover the original signal with the
least measurements. In other words, we need to design the measurement matrix A ∈ Rm×n
such that m is smallest for fixed n. Although it was difficult to design deterministic matrix

for RIP, various random matrices were shown to satisfy the RIP with high probability. If the

entries in the matrices are chosen according to Gaussian [96], Bernoulli [95], or sub Gaussian

132

distribution [97], then the matrices satisfy the RIP with very high probability if:

m ≥ D · S log(n/S)

(7.7)

where D is also a constant [97]. Note that the random matrices discussed above follows the
so-called universality. In fact, if a random matrix Φ satisfies the RIP, then the matrix ΦΨ
will also satisfy the RIP with high probability. In other words, we can design the random
measurement matrices Φ without considering the sparsity of the original signal [95].
The final piece of compressive sensing is the recovery algorithm, i.e., the algorithm to
solve Eq. (7.5). There are many algorithms to solve the problem, which can be classified into
convex optimization, greedy algorithm, or combinatorial algorithms. A detailed summary of
those algorithms can be found in [98].

7.2

Sensing System Architecture

Prior to the discussion of the sensing system, I will give a brief introduction of the DMD.
DMD in our system is one of the most important components to accomplish compressive
sensing algorithm. Traditional DMD is the essential component of digital light processing
(DLP) projection technology, controlling light propagation by a large array of electrostatically actuated micromirrors. Figure 7.1(a) shows a scanning electron microscopy (SEM)
image of nine micromirrors in a DMD. Each micromirror is square in shape, and stands
closely to adjoining micromirrors with a small gap, implying a high fill factor.
In our imaging system, DMD from Texas Instruments (TI) was used to generate a measurement matrix according to compressive sensing. Each individual micromirror can be

133

(a)
Light

Light

photodetector

“on” state

“off” state
power off state
(b)

Figure 7.1: (a) Scanning electron microcopy (SEM) image of an array of micromirros in the
DMD. (b) Three status of the micro-mirrors: “on”, “off”, and power-off states.

134

poised into two different angles (-10 degree and +10 degree), which define the “on” and
“off” states of the micromirrors. In the “on” state, light irradiated on a micromirror is oriented to the CNT photodetector (the left mirror of Figure 7.1(b)). In the “off” state, the
light that fell onto a micromirror is reflected elsewhere (the right mirror of Figure 7.1(b)).
When the power for DMD is off, the micromirror stays horizontal without a rotation (the
middle mirror of Figure 7.1(b)). By controlling the states of each micromirror, the DMD
forms a specific measurement matrix.
The structure of the IR sensing system is built upon the compressive sensing theory
introduced in the previous section. Figure 7.2 depicts the system setup using a single pixel
CNT photodetector as its image sensor.
IR images were directed onto a DMD through a set of lenses. The IR images on the
DMD represents the original signal x in Eq. (7.1) (x ∈ Rn , n is the number of image pixels).
The DMD generated patterns according to a measurement matrix Φ ((Φ ∈ Rm×n )) so as
to compress the IR images. Each pattern on the DMD comprised n pixels, thus m different
patterns were generated to construct the measurement matrix Φ. The compressed IR images
were reflected to another lens, which focused all the light to a CNT photodetector. The IR
signal arrived at the CNT photodetector represents the linear projection of the images onto
the measurement matrix, which can be considered as the inner product of IR images x
and each row vector in the measurement matrix Φ. The photocurrent generated in the
photodetector was recorded by a fast readout system integrating with a charge integrator,
an analog-to-digital converter (ADC), and a data acquisition (DAQ) card [99] [100]. The
amplitudes of photocurrent represent the values of y in Eq. (7.1) (y ∈ Rm , m is the number
of measurements). The IR images were recovered by a personal computer. It should be
noted that compressive sensing makes m < n; therefore, the IR images x were compressively
135

Lens
DMD

IR Source

Φ

x

Lens
Readout System

CNT Photodetector

y=Φx
Image Recovery

x
IR Image
Figure 7.2: The setup of compressive sensing based imaging system using a CNT detector
with photonic cavity and DMD.

136

sampled into the y. Based on the measurement results y and the designed measurement
matrix Φ, the original signal x can be recovered using an image reconstruction algorithm.

7.3

Experimental Results

For the imaging experiment, the laser source was projected onto the DMD to form an IR

image x. DMD generated a measurement matrix Φ according to random Bernoulli distribu-

tion, which had high probability of satisfying the RIP condition [95]. Each random pattern

is a row of the measurement matrix, and m Bernoulli random patterns were used to capture

an IR image. As a result, each measurement result from the CNT photodetector obtains the

inner product between an IR image and a row of the measurement matrix. After m times of

measurements, one image x is compressively sampled into the measurement results y. The

image was recovered via minimizing the total variation norm (TV-norm). Minimizing such

a norm can make the intensity difference between neighboring pixels sparse, thus smoothing

the images. Formally, let xij be the pixel intensity in the i-th row and j-th column in the
h
v
image x. The intensity differences in the horizontal and vertical directions (Dij and Dij )

137

can be defined respectively as follows:

h
Dij = xi+1,j − xij ,

v
Dij = xi,j+1 − xij

(7.8)

The total variation for the image x is then defined as

∑√ 2
h
v
Dij + Dij 2
T V (x) =

(7.9)

i,j

Therefore, the image recovery problem can be defined as:

min ||T V (ˆ)||1 subject to Aˆ = y
x
x

(7.10)

Note that for consistence, the above x can be considered as the vector form of the image
ˆ
matrix x.
The DMD has 1064×768 micromirrors. Each micromirror is square in shape with a
dimension of 12 µm × 12 µm. We first obtained IR images with a low resolution of 30×30
pixels to adjust the alignment and focus of the system. Traditional cameras require 900
measurements in order to obtain a 30×30 image, while our imaging system can recover the
image with only 500 measurements [84]. Images of 30×30 pixels correspond to 1064×768
micromirrors, thus each pixel contains 908 micromirrors. Note that the DMD is not square
in shape, thus each pixel in the image is a rectangular shape.
The imaging system was used to capture a higher resolution image of a rectangular IR

138

light bar

(a)

(b)
Figure 7.3: (a) The position and orientation of the rectangular light bar on the DMD. (b)
The image of 50×50 pixels recovered by our imaging system, highlighted by a red rectangle.
light bar. The position, orientation, and geometry of the light bar on the DMD are shown in
Figure 7.3(a). The recovered image is shown in Figure 7.3(b). The location and orientation of
the recovered rectangular light bar, highlighted by a red rectangle, are similar to that shown
in Figure 7.3(a). The image has a resolution of 50×50 pixels. This image was recovered by
only 800 measurements, which is less than 35% of the total pixel numbers (2500).

139

7.3.1

Dynamic Observation

The imaging system was utilized to observe the dynamic movement of a laser spot. The
laser spot was initially located at the lower left of the DMD, as shown in Figure 7.4(a). We
designed a trajectory of the laser spot movement: it was first moved downwards, and then
moved upwards when the laser approached the bottom edge of the DMD. Six frames of the
IR images captured by the imaging system were shown in Figure 7.4. The resolution of these
images is 50×50 pixels, and 800 measurements were performed to recover the images using
minimization TV-norm reconstruction algorithm. The images clearly show the movement
of the laser spot following the designed path: the laser spot moved towards the bottom of
the DMD in the first five frames (Figure 7.4(a-e)), and it bounced upwards in the last frame
(Figure 7.4(f)).
However, it is difficult to identify the actual size and geometry of the laser spot from these
images due to their low resolution. Figure 7.4(b) shows the laser spot covered one pixel, while
it covered three pixels in Figure 7.4(c) and Figure 7.4(e), and the rest of frames show the
laser spot occupied two pixels. The spot size variation may be caused by the difference of
the location of the laser spot: the laser spot covers one pixel if the spot resides at the center
of a pixel, while it covers more than one pixel if the laser spot located at the boundary of a
few pixels; the difference, however, may possibly stem from the measurement noise. In order
to identify the actual size of the laser spot, we can increase the image resolution. However,
this approach will reduce the speed of the sensing system due to the increase of sampling
and recovery time. In next section, we will introduce the zooming technique to investigate
the laser spot by keeping the same image resolution.

140

laser spot

(a)

(b)

(c)

(d)

(e)

(f)

Figure 7.4: Six images of a laser spot captured by our imaging system, showing the movement
of the spot.(a) The infrared image shows the initial position of the laser spot. (b)-(e) The
laser spot moved downwards towards the bottom of the image. (f) The laser spot moved
upwards after it approached the bottom of the DMD.

141

7.3.2

Zooming

In previous experiments, all 1024×768 micromirrors in the DMD were used to capture the
IR images with a resolution of 50×50 pixels. In such a setup, 315 micromirrors in unison
represent one pixel, corresponding to an area of 0.045 mm2 on the DMD (144 µm2 per
micromirror). Therefore this setup is difficult to identify the features of an object with
similar or smaller size than that area. We developed a zooming technique to investigate the
details of small objects without degrading the performance of the system.
The imaging system can perform zooming in and out by varying the number of micromirrors for each pixel. When the imaging system worked in zooming-in mode, only a portion
of the DMD was used as the measurement matrix generator. As a result, each pixel included fewer micromirrors. This mode is supported by the DMD because each individual
micromirror can be controlled independently.
In order to verify the zooming concept, we designed an experiment to study the geometry and size of the IR laser spot shown in Figure 7.4. In the zooming experiment, each
IR image (50×50 pixels) was recovered using the same measurement matrix, number of
measurements, and recovery algorithm as previous experiments. The only difference is the
number of micromirrors for each pixel.
We first recovered an image (×1 magnification image in Figure 7.5) by using the whole
DMD, which shows the laser spot covering two pixels, corresponding to 0.090 mm2 in area
on the DMD. The intensity of each pixel for images in Figure 7.5 is represented by 8 bits
grey-scale: each pixel has a brightness from 0 to 255, where 0 is black and 255 is white.
A typical OTSU method [101] is used to convert gray-scale images into binary images in
order to identify the actual area of the laser spot. When we used 521×384 micromirrors

142

(highlighted by the blue rectangle in ×1 magnification image) to generate the measurement
matrix, a ×2 magnification image was obtained. It is termed ×2 magnification because the
micromirrors only cover half of the length and width of the DMD, while the recovered image
had same pixel number as the ×1 magnification image. The ×2 magnification image shows
that a laser spot covered six pixels, reducing to 0.068 mm2 in area. We further sectioned
the measurement matrix into 130×96 micromirrors. The recovered ×8 magnification image
shows that the laser spot covered forty one pixels, corresponding to 0.029 mm2 in area. A
×12 magnification image was obtained when we used 85×64 micromirrors to generate the
measurement matrix. A clear outline of the laser spot was exhibited in elliptical shape,
which is similar to the laser spot geometry captured by a CMOS camera. The laser spot
area was reduced to 0.028 mm2 in the ×12 magnification image. The experimental results
showed that the shape and area of the laser spot are progressively refined to its actual
geometry by zooming in with our sensing system. It was also observed that images with
larger magnification associated with lower image quality, since each pixel contained fewer
micromirrors, which reflected less light to the photodetector.
This sensing system is universal and adaptable. It is universal because we can use the
same measurement matrix for all experiments [95]. The system is adaptable because the
resolution of images and sections of DMD used for zooming can be adaptively adjusted
depending on the features of objects.

7.4

Chapter Summary

A novel sensing system using CNT based IR detector as core component was demonstrated.
A compressive sensing algorithm based IR sensing system was developed by generating mea-

143

1 magnificati

image

2 magnificati

image

8 magnificati

image

2 magnificati

image

Figure 7.5: ×1, ×2, ×8, ×12 magnification IR images captured by our imaging system, and
their corresponding sections of DMD used to capture the images were highlighted by blue
rectangles.

144

surement matrices using DMD, which compressed the original images. The experimental
results showed that this imaging system can recover original images by decoding the electrical signals, with measurement times much less than original dimension. This camera
architecture provide an novel platform to effectively capture high resolution IR images.

145

Chapter 8
Conclusions and Future Work

8.1

Conclusion

IR sensing systems have versatile applications, but their performance was limited by the their
core component, photodetectors. The low performance of conventional photodetectors stems
from the optical and physical properties of bulk semiconductors. With the development
of nanotechnology, these problems may be addressed by developing nano-materials based
photodetectors. In particular, CNTs have promising optoelectronic and optical properties
arose from its perfect 1D hollow-cylinder geometry. In this research, an IR sensing system
was built using a CNT based photodetector.
Based on the results and discussion presented in previous chapters, the following conclusion can be made:
• The CNT based Schottky photodiodes have been thoroughly studied for IR detection.
The role of CNT-metal contacts in determining the performance of IR detection was
systematically investigated. The performance of CNT photodetectors with different
metals, symmetric and asymmetric metal structures, and at different temperatures
was studied. This study improved our understanding of the underlying physics of the
photo-detection, and photo-response of CNT based Schottky photodiodes was highly
enhanced by using asymmetric metal structure.

146

• The importance of electrostatic doping in determining IR detection of the CNT photodetectors was demonstrated using CNTFET based photodetectors. The built-in
potential between metals and a CNT in a CNTFET was responsible for the separation of photo-excited electron and hole pair generation, thus electrostatic doping can
control the built-in potential so as to improve photon detection. It was found that
asymmetric metal structure and electrostatic doping can reduce dark current, improve
photocurrent, and enhance open circuit voltage. In addition, experimental results of
a multi-gate CNTFET revealed an unconventional charge distribution in the CNT,
which can naturally enlarge the absorption area of CNT photodetectors.
• A 3×3 focal plane array of CNTFET based photodetectors was fabricated. This small
focal plane array demonstrated the feasibility of imaging using CNT based photodetectors. Meanwhile, it showed the difficulties of manufacturing megapixel count pixels
using nano-materials.
• A breakdown process was developed to engineer the bandgap energy of MWCNT.
MWCNT based photodetectors were demonstrated to detect MWIR at room temperature by tuning the bandgap energy. Comparing with the spectral responses of
SWCNT based photodetectors, MWCNT based photodetectors exhibited wide spectral responses due to the stacking of layers with different bandgap energies.
• A bowtie antenna was developed to improve the fill factor of CNT based photodetectors by confining electromagnetic wave into a sub-wavelength volume. Wavelength
of IR is much larger than the size of CNT, resulting in small fill factor. Optical antenna functioned as a plasmonic lens to compress light onto CNT in order to improve
photocurrent generation.
147

• Photonic crystal cavity was fabricate to further improve the fill factor of CNT based
photodetectors. Introducing a defect in a photonic crystal with a periodic dielectric/air
structure, most of light was trapped into the defect region. By aligning the defect onto
the CNT photodetector, a photocurrent enhancement was demonstrated.
• A compressive sensing based sensing system was developed using a single pixel CNT
photodetector. In order to fully use the outstanding performance of CNT photodetecotrs and overcome the difficulties of manufacturing large detector array, an IR sensing
system was developed using compressive sensing algorithm. This camera system requires only a single photodetector. Experimental results showed that the number of
measurement using this camera system was much less than the size of the original
images.

8.2

Future Research

Future research will focus on the following three directions:
• Both the optical antenna and photonic crystal cavity have demonstrated to improve the
fill factor of CNT photodetectors, but the design of these devices was not optimized.
These unconventional light confinement devices are worthy to further develop in order
to improve the photo-responses of the detectors.
• Further studies are needed to improve the performance of the sensing system. We
have demonstrated the feasibility of using compressive sensing algorithm to construct
a sensing system. However, the speed of the system is still low compared to commercial
IR cameras.

148

• This sensing system architecture can be applied to other types of photodetectors.
A CNT photodetector is just one example of nano-material based photodetectors.
Graphene, nanowires, quantum dots, and quantum wells are also promising optoelectronic materials. This imaging system may accelerate the development of IR cameras
that utilize other high performance nano-photodetectors.

149

BIBLIOGRAPHY

150

BIBLIOGRAPHY
[1] Antoni Rogalski. Infrared detectors: an overview. Infrared Physics and Technology,
43(3-5):187 – 210, 2002.
[2] R. Agarwal and C.M. Lieber. Semiconductor nanowires: optics and optoelectronics.
Applied Physics A: Materials Science and Processing, 85:209–215, 2006.
[3] Zhiyong Fan, Rehan Kapadia, Paul W. Leu, Xiaobo Zhang, Yu-Lun Chueh, Kuniharu
Takei, Kyoungsik Yu, Arash Jamshidi, Asghar A. Rathore, Daniel J. Ruebusch, Ming
Wu, and Ali Javey. Ordered arrays of dual-diameter nanopillars for maximized optical
absorption. Nano Letters, 10(10):3823–3827, 2010.
[4] L. E. Brus. Electron-electron and electron-hole interactions in small semiconductor
crystallites: The size dependence of the lowest excited electronic state. The Journal
of Chemical Physics, 80(9):4403–4409, 1984.
[5] R and Reisfeld. Nanosized semiconductor particles in glasses prepared by the solcgel
method: their optical properties and potential uses. Journal of Alloys and Compounds,
341(1-2):56 – 61, 2002.
[6] J. W. Mintmire, B. I. Dunlap, and C. T. White. Are fullerene tubules metallic? Phys.
Rev. Lett., 68:631 – 634, 1992.
[7] M. S. Dresselhaus, G. Dresselhaus, and Riichiro Saito. Carbon fibers based on c60 and
their symmetry. Phys. Rev. B, 45:6234 – 6242, 1992.
[8] M. Dresselhaus, G. Dresselhaus, and Phaedon Avouris (Eds.). Carbon Nanotubes:
Synthesis, Structure, Properties and Application. Springer, Berlin, 2001.
[9] Phaedon Avouris. Carbon nanotube electronics. Chemical Physics, 281:429C445, 202.
[10] King Wai Chiu Lai, Ning Xi, Carmen Kar Man Fung, Hongzhi Chen, and Tzyh-Jong
Tarn. Engineering the band gap of carbon nanotube for infrared sensors. Applied
Physics Letters, 95(22):221107, 2009.

151

[11] M. Wolf, R. Brendel, J. H. Werner, and H. J. Queisser. Solar cell efficiency and carrier
multiplication in si[sub 1 - x]ge[sub x] alloys. Journal of Applied Physics, 83(8):4213–
4221, 1998.
[12] Nathaniel M. Gabor, Zhaohui Zhong, Ken Bosnick, Jiwoong Park, and Paul L.
McEuen. Extremely Efficient Multiple Electron-Hole Pair Generation in Carbon Nanotube Photodiodes. Science, 325(5946):1367–1371, 2009.
[13] Ali Javey, Jing Guo, Qian Wang, Mark Lundstrom, and Hongjie Dai. Ballistic carbon
nanotube field-effect transistors. Nature, 424:654–657, 2003.
[14] Brendan M. Kayes, Harry A. Atwater, and Nathan S. Lewis. Comparison of the device
physics principles of planar and radial p-n junction nanorod solar cells. Journal of
Applied Physics, 97(11):114302, 2005.
[15] Fran ¸ois L´onard and J. Tersoff. Novel length scales in nanotube devices. Phys. Rev.
c
e
Lett., 83(24):5174–5177, 1999.
[16] I. A. Levitsky and W. B. Euler. Photoconductivity of single-wall carbon nanotubes
under continuous-wave near-infrared illumination. Appl. Phys. Lett., 83:1857–1859,
2003.
[17] Basudev Pradhan, Kristina Setyowati, Haiying Liu, David H. Waldeck, and Jian Chen.
Carbon nanotube?polymer nanocomposite infrared sensor. Nano Letters, 8(4):1142–
1146, 2008.
[18] William B. Euler Pang-Leen Ong and Igor A. Levitsky. Carbon nanotube-si diode as
a detector of mid-infrared illumination. Applied Physics Letters, 96(3):033106, 2010.
[19] Marian B. Tzolov, Teng-Fang Kuo, Daniel A. Straus, Aijun Yin, and Xu. Carbon nanotube?silicon heterojunction arrays and infrared photocurrent responses. The Journal
of Physical Chemistry C, 111(15):5800–5804, 2007.
[20] Ji Ung Lee. Photovoltaic effect in ideal carbon nanotube diodes. Appl. Phys. Lett.,
87:073101, 2005.
[21] R. Martel, T. Schmidt, H. R. Shea, T. Hertel, and Ph. Avouris. Single- and multi-wall
carbon nanotube field-effect transistors. Applied Physics Letters, 73(17):2447–2449,
1998.
[22] S. Heinze, J. Tersoff, R. Martel, V. Derycke, J. Appenzeller, and Ph. Avouris. Carbon
nanotubes as schottky barrier transistors. Phys. Rev. Lett., 89(10):106801, 2002.
152

[23] Jiangbo Zhang, Ning Xi, Hongzhi Chen, King W. C. Lai, Guangyong Li, and
Uchechukwu Wejinya. Photovoltaic effect in single carbon nanotube-based schottky
diodes. International Journal of Nanoparticles, 1(2):108–118, 2008.
[24] Philip G. Collins, Michael S. Arnold, and Phaedon Avouris. Engineering Carbon Nanotubes and Nanotube Circuits Using Electrical Breakdown. Science, 292(5517):706–
709, 2001.
[25] Jijun Zhao, Jie Han, and Jian Ping Lu. Work functions of pristine and alkali-metal
intercalated carbon nanotubes and bundles. Phys. Rev. B, 65(19):193401, 2002.
[26] XD Cui, M Freitag, R Martel, L Brus, and P Avouris. Controlling energy-level alignments at carbon nanotube/Au contacts. Nano Letters, 3(6):783–787, 2003.
[27] King W. C. Lai, Ning Xi, and Uchechukwu C. Wejinya. Automated process for selection of carbon nanotube by electronic property using dielectrophoretic manipulation.
Journal of Micro - Nano Mechatronics, 4(1-2):37–48, 2008.
[28] 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, 2004.
[29] J. Zhang, N. Xi, G. Li, H. Y. Chan, , and U. C. Weijinya. Adaptable end effector for
atomic force microscopy based nanomanipulation. IEEE Trans. on Nanotechnology,
5:628–642, 2006.
[30] Jiangbo Zhang, Ning Xi, Lianqing Liu, Hongzhi Chen, King Lai, and Guangyong
Li. Atomic force yields a master nanomanipulator. Nanotechnology Magazine, IEEE,
2(2):13 –17, june 2008.
[31] Jiangbo Zhang, Ning Xi, Hongzhi Chen, King Lai, Guangyong Li, and U.C. Wejinya.
Design, manufacturing, and testing of single-carbon-nanotube-based infrared sensors.
Nanotechnology, IEEE Transactions on, 8(2):245–251, 2009.
[32] King Wai Chui Lai, Ning Xi, Carmen Kar Man Fung, Jiangbo Zhang, Hongzhi Chen,
Yilun Luo, and Uchechukwu C. Wejinya. Automated Nanomanufacturing System to
Assemble Carbon Nanotube Based Devices. International Journal of Robotics Research, 28(4):523–536, 2009.
[33] J. Appenzeller, M. Radosavljevi´, J. Knoch, and Ph. Avouris. Tunneling verc
sus thermionic emission in one-dimensional semiconductors. Phys. Rev. Lett.,
92(4):048301, Jan 2004.
153

[34] J. Zhang, N. Xi, H. Chan, and G. Li. Single carbon nanotube based infrared sensor.
Electro-Optical and Infrared Systems: Technology and Applications III, SPIE Proc.,
6395:63950A, 2006.
[35] 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., 87:253116, 2005.
[36] Hongjie Dai. Carbon nanotubes: Synthesis, integration, and properties. Accounts of
Chemical Research, 35:1035–1044, 2002.
[37] SJ Tans, ARM Verschueren, and C Dekker. Room-temperature transistor based on a
single carbon nanotube. NATURE, 393(6680):49–52, 1998.
[38] MS Fuhrer, J Nygard, L Shih, M Forero, YG Yoon, MSC Mazzoni, HJ Choi, J Ihm,
SG Louie, A Zettl, and PL McEuen. Crossed nanotube junctions. SCIENCE,
288(5465):494–497, 2000.
[39] 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, 80:3817 – 3819, 2002.
[40] A. Javey, J. Guo, D.B. Farmer, Q. Wang, D. Wang, R.G. Gordon, M. Lundstrom, and
H. Dai. Carbon nanotube field-effect transistors with integrated ohmic contacts and
high-k gate dielectrics. Nano Letters, 4(3):447–450, 2004.
[41] M. Freitag, Y. Martin, J. A. Misewich, R. Martel, and Ph. Avouris. Photoconductivity
of single carbon nanotubes. Nano Letters, 3:1067–1071, 2003.
[42] Xiaohui Qiu, Marcus Freitag, Vasili Perebeinos, and Phaedon Avouris. Photoconductivity spectra of single-carbon nanotubes: Implications on the nature of their excited
states. Nano Letters, 5:749–752, 2005.
[43] Masashi Shiraishi and Masafumi Ata. Work function of carbon nanotubes. Carbon,
39(12):1913 – 1917, 2001.
[44] J. Appenzeller, M. Radosavljevi´, J. Knoch, and Ph. Avouris. Tunneling verc
sus thermionic emission in one-dimensional semiconductors. Phys. Rev. Lett.,
92(4):048301, 2004.

154

[45] Donghun Kang, Noejung Park, Ju hye Ko, Eunju Bae, and Wanjun Park. Oxygeninduced p-type doping of a long individual single-walled carbon nanotube. Nanotechnology, 16(8):1048–1052, 2005.
[46] Changxin Chen, Wei Zhang, Eric Siu-Wai Kong, and Yafei Zhang. Carbon nanotube photovoltaic device with asymmetrical contacts. Applied Physics Letters,
94(26):263501, 2009.
[47] Fran ¸ois L´onard and J. Tersoff. Role of fermi-level pinning in nanotube schottky
c
e
diodes. Phys. Rev. Lett., 84(20):4693–4696, 2000.
[48] M Chelvayohan and C H B Mee. Work function measurements on (110), (100) and
(111) surfaces of silver. Journal of Physics C: Solid State Physics, 15(10):2305–2312,
1982.
[49] Woong Kim, Ali Javey, Ophir Vermesh, Qian Wang, Yiming Li, and Hongjie Dai.
Hysteresis caused by water molecules in carbon nanotube field-effect transistors. Nano
Lett., 3:193–198, 2003.
[50] A. Phaedon. Carbon nanotube electronics. Chem. Phys., 281:429–445, 2002.
[51] 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.
[52] Stephen W. Kennerly Samuel C. H. Wang Mani Sundaram Parvez N. Uppal Michael L.
Winn Gregory L. Milne Arnold C. Goldberg, Theodore Fischer and Mark A. Stevens.
Dual-band qwip mwir/lwir focal plane array test results. In Proc. SPIE, volume 4028,
page 276, 2000.
[53] J. Caufleld. Simultaneous integrating staring two color irfpa. In Proc. the 1997 IRISDetector Specialty Group Meeting on Detectors, page 351, 1998.
[54] M. Sundaram. Advances in qwip fpa (2-color and 1-color) technology. In Proc. the
1998 IRIS Specialty Group Meeting on Detectors, page 219, 1999.
[55] P. Warren M. Satyshur M. Kruer D. Scribner, J. Schuler. Dual-band qwip mwir/lwir
focal plane array test results. In Proc. SPIE, volume 3379, page 2, 1998.
[56] E.R. Fossum. Cmos image sensors: electronic camera-on-a-chip. Electron Devices,
IEEE Transactions on, 44(10):1689 –1698, oct 1997.
155

[57] W. A. Murray and W. L. Barnes.
19(22):3771–3782, 2007.

Plasmonic materials.

Advanced Materials,

[58] Jon. A. Schuller, Edward S. Barnard, Wenshan Cai, Young Chul Jun, Justin S. White,
and Mark L. Brongersma. Plasmonics for extreme light concentration and manipulation
(vol 9, pg 193, 2010). NATURE MATERIALS, 9(4), 2010.
[59] Palash Bharadwaj, Bradley Deutsch, and Lukas Novotny. Optical antennas. Adv. Opt.
Photon., 1(3):438–483, Nov 2009.
[60] Harry A. Atwater and Albert Polman. Plasmonics for improved photovoltaic devices.
Nature Materials, 9(10):865, 2010.
[61] Zhiyong Fan, Daniel Ruebusch, Asghar Rathore, Rehan Kapadia, Onur Ergen, Paul
Leu, and Ali Javey. Challenges and prospects of nanopillar-based solar cells. Nano
Research, 2:829–843, 2009.
[62] D. M. Schaadt, B. Feng, and E. T. Yu. Enhanced semiconductor optical absorption via surface plasmon excitation in metal nanoparticles. Applied Physics Letters,
86(6):063106, 2005.
[63] Zongfu Yu, Georgios Veronis, Shanhui Fan, and Mark L. Brongersma. Design of midinfrared photodetectors enhanced by surface plasmons on grating structures. Applied
Physics Letters, 89(15):151116, 2006.
[64] Mark W. Knight, Heidar Sobhani, Peter Nordlander, and Naomi J. Halas. Photodetection with Active Optical Antennas. SCIENCE, 332(6030):702–704, MAY 6 2011.
[65] Hongzhi Chen, Ning Xi, K.W.C. Lai, C.K.M. Fung, and Ruiguo Yang. Development
of infrared detectors using single carbon-nanotube-based field-effect transistors. Nanotechnology, IEEE Transactions on, 9(5):582 –589, 2010.
[66] P. J. Schuck, D. P. Fromm, A. Sundaramurthy, G. S. Kino, and W. E. Moerner. Improving the mismatch between light and nanoscale objects with gold bowtie nanoantennas.
Phys. Rev. Lett., 94:017402, Jan 2005.
[67] DP Fromm, A Sundaramurthy, PJ Schuck, G Kino, and WE Moerner. Gap-dependent
optical coupling of single Bowtie nanoantennas resonant in the visible. Nano Letters,
4(5):957–961, 2004.

156

[68] Carmen Kar Man Fung, Ning Xi, Balasubramaniam Shanker, and King Wai Chiu Lai.
Nanoresonant signal boosters for carbon nanotube based infrared detectors. Nanotechnology, 20(18), 2009.
[69] A Sundaramurthy, KB Crozier, GS Kino, DP Fromm, PJ Schuck, and WE Moerner.
Field enhancement and gap-dependent resonance in a system of two opposing tip-to-tip
Au nanotriangles. Physical Review B, 72(16), OCT 2005.
[70] KB Crozier, A Sundaramurthy, GS Kino, and CF Quate. Optical antennas: Resonators
for local field enhancement. Journal of Applied Physics, 94(12):7950, 2003.
[71] E Betzig, JK Trautman, TD Harris, JS Weiner, and RL Kostelak. Breaking the
diffraction barrier- optical microscopy on a nanometric scale. Science, 251(5000):1468–
1470, MAR 22 1991.
[72] Ragip A. Pala, Justin White, Edward Barnard, John Liu, and Mark L. Brongersma.
Design of Plasmonic Thin-Film Solar Cells with Broadband Absorption Enhancements.
ADVANCED MATERIALS, 21(34):3504+, SEP 11 2009.
[73] Wei Ding, Renaud Bachelot, Sergei Kostcheev, Pascal Royer, and Roch Espiau
de Lamaestre. Surface plasmon resonances in silver bowtie nanoantennas with varied bow angles. 108(12):124314, 2010.
[74] Kevin L. Shuford, Mark A. Ratner, and George C. Schatz. Multipolar excitation in
triangular nanoprisms. 123(11):114713, 2005.
[75] Tomoyuki Yoshie, Jelena Vuckovic, Axel Scherer, Hao Chen, and Dennis Deppe.
High quality two-dimensional photonic crystal slab cavities. Applied Physics Letters,
79(26):4289–4291, 2001.
[76] Toshihiko Baba, Daisuke Sano, Kengo Nozaki, Kyoji Inoshita, Yusuke Kuroki, and
Fumio Koyama. Observation of fast spontaneous emission decay in gainasp photonic crystal point defect nanocavity at room temperature. Applied Physics Letters,
85(18):3989–3991, 2004.
[77] M. Notomi, K. Yamada, A. Shinya, J. Takahashi, C. Takahashi, and I. Yokohama.
Extremely large group-velocity dispersion of line-defect waveguides in photonic crystal
slabs. Phys. Rev. Lett., 87(25):253902, 2001.
[78] Yoshihiro Akahane, Takashi Asano, Bong-Shik Song, and Susumu Noda. High-q photonic nanocavity in a two-dimensional photonic crystal. Nature, 425(6961):944–947,
2003.
157

[79] Mehmet Bayindir and E. Ozbay. Heavy photons at coupled-cavity waveguide band
edges in a three-dimensional photonic crystal. Phys. Rev. B, 62(4):R2247–R2250, 2000.
[80] M Jerman, ZH Qiao, and D Mergel. Refractive index of thin films of SiO2, ZrO2, and
HfO2 as a function of the films’ mass density. APPLIED OPTICS, 44(15):3006–3012,
2005.
[81] H. W. Hubers, J. Schubert, A. Krabbe, M. Birk, G. Wagner, A. Semenov, and G. Goltsman. Parylene anti-reflection coating of a quasi-optical hot-electron-bolometric mixer
at terahertz frequencies. Infrared Physics and Technology, 42(1):41 – 47, 2001.
[82] J. Haupt and R. Nowak. Compressive sampling vs. conventional imaging. In Image
Processing, 2006 IEEE International Conference on, pages 1269 –1272, oct. 2006.
[83] M. F. Duarte, M. A. Davenport, D. Takhar, J. N. Laska, T. Sun, K. F. Kelly, and R. G.
Baraniuk. Single-pixel imaging via compressive sampling. IEEE Signal Processing
Magazine, 25(2):83–91, 2008.
[84] Hongzhi Chen, Ning Xi, Bo Song, LiangLiang Chen, and K.W.C. Lai. Single pixel
infrared camera using a carbon nanotube photodetector. In Sensors, 2011 IEEE,
pages 1362 –1366, oct. 2011.
[85] M. A. Davenport, M. F. Duarte, Y.C. Eldar, and G. Kutyniok. Introduction to Compressed Sensing, chapter One. Cambridge University Press, 2012.
[86] Michael Lustig, David Donoho, and John M. Pauly. Sparse mri: The application of
compressed sensing for rapid mr imaging. Magnetic Resonance in Medicine, 58(6),
2007.
[87] Jianguo Zhao, Bo Song, Ning Xi, King Wai Chiu Lai, Hongzhi Chen, and Chengeng Qu.
Compressive feedback based non-vector space control. In American Control Conference
(ACC), 2012, pages 4090 –4095, june 2012.
[88] Bo Song, Jianguo Zhao, Ning Xi, King Wai Chiu Lai, Ruiguo Yang, Hongzhi Chen, and
Chengeng Qu. Non-vector space control for nanomanipulations based on compressive
feedbacks. In Robotics and Automation (ICRA), 2012 IEEE International Conference
on, pages 2767 –2772, may 2012.
[89] Tim T. Y. Lin and Felix J. Herrmann. Compressed wavefield extrapolation. Geophysics,
72(5):SM77–SM93, 2007.

158

[90] S.F. Cotter and B.D. Rao. Sparse channel estimation via matching pursuit with application to equalization. Communications, IEEE Transactions on, 50(3):374 –377, mar
2002.
[91] E. Cand`s and M. Wakin. An introduction to compressive sampling. IEEE Signal
e
Processing Magazine, 25(2):21–30, 2008.
[92] David L. Donoho and Michael Elad. Optimally sparse representation in general
(nonorthogonal) dictionaries via l1 minimization. Proceedings of the National Academy
of Sciences, 100(5):2197–2202, 2003.
[93] Albert Cohen, Wolfgang Dahmen, and Ronald Devore. Compressed sensing and best
k -term approximation. Journal of the American Mathematical Society, 22(1):211–231,
2009.
[94] E. Cand`s. The restricted isometry property and its implications for compressed sense
ing. Comptes Rendus Mathematique, 346(9-10):589–592, 2008.
[95] E. J. Candes and T. Tao. Near optimal signal recovery from random projections: universal encoding strategies? IEEE Transactions on Information Theory, 52(12):5406–
5425, 2006.
[96] E.J. Candes and T. Tao. Decoding by linear programming. IEEE Transactions on
Information Theory, 51(12):4203–4215, 2005.
[97] Richard Baraniuk, Mark Davenport, Ronald Devore, and Michael Wakin. A simple
proof of the restricted isometry property for random matrices. Constructive Approximation, 28(3):253–263, 2008.
[98] Joel A. Tropp and Stephen J. Wright. Computational methods for sparse solution of
linear inverse problems. Proceedings of the IEEE, 98(6):948–958, 2010.
[99] Liangliang Chen, Ning Xi, Hongzhi Chen, and K.W.C. Lai. Development of readout
system for carbon nanotube based infrared detector. In Nanotechnology Materials and
Devices Conference (NMDC), 2010 IEEE, pages 230–234, 2010.
[100] Liangliang Chen, Ning Xi, Hongzhi Chen, and King Wai Chiu Lai. High gain current
readout method for mwcnt infrared sensor. In Nanotechnology (IEEE-NANO), 2012
12th IEEE Conference on, pages 1–4, 2012.
[101] Nobuyuki Otsu. A threshold selection method from gray-level histograms. Systems,
Man and Cybernetics, IEEE Transactions on, 9(1):62 –66, jan. 1979.
159