University of Southern California Dissertations and Theses

The growth and characterization of III-V semiconductor nanowire arrays by nanoscale selective area metalorganic chemical vapor deposition

(USC Thesis Other)

The growth and characterization of III-V semiconductor nanowire arrays by nanoscale selective area metalorganic chemical vapor deposition

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
THE GROWTH AND CHARACTERIZATION OF
III-V SEMICONDUCTOR NANOWIRE ARRAYS
BY NANOSCALE SELECTIVE AREA METALORGANIC
CHEMICAL VAPOR DEPOSITION

by

Hyung-Joon Chu

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)

August 2010

Copyright 2010 Hyung-Joon Chu
ii
Dedication

To My Parents,
Taejong Joo and Hyewon Lee

iii
Acknowledgments

If I tried to describe all my thanks here, this part might be longer than the
main article, and the biggest part of the article would be my sincere appreciation to
my thesis advisor, Professor P. Daniel Dapkus. It was very lucky for me to meet him
when I came to Los Angeles and started my studies at the University of Southern
California. Without his invaluable guidance and care, I would not be able to reach
this point. He has transformed me from a naïve student into an independent
researcher.
I would like to thank Professor William H. Steier and Aiichiro Nakano.
Their guidance as my dissertation committee makes this work more complete. Also,
I want to express thanks to John O’Brien and Chongwu Zhou, who both guided me
on my qualifying committee.
This work would have not been possible without my colleagues in the
compound semiconductor lab (CSL). Especially, I thank Larry Stewart for his
support with my cleanroom work. My MOCVD colleague, Tingwei Yeh took all
TEM pictures in this work and I also commend his devotion in helping with reactor
maintenance. I thank all of the other CSL members: Suzana Sburlan, Yunchu Li,
Chunyung Chi, and Yenting Lin. I wish them success in their future research
endeavors.
I would like to share the completion of my work with all of the CSL alumni
who worked with me. Seung-June Choi, who first taught me about MOCVD in
iv
Korea, spent his time here in the early stage of my life in the CSL and will share time
in Intel PTD. I also thank Yuanming Deng and Euihyun Hwang for their support
with MOCVD and Zhen Peng and Qi Yang for their helps in the cleanroom.
I should mention the devoted support from my family. My father, Taejong
Joo (He likes ‘Joo’ rather than ‘Chu’) and my mother Hyewon Lee receive my
sincere gratitude. I thank my brother, Hyung-Kwon Chu and his family. I also thank
all my relatives in Korea for their support in my studies.
I would like to thank Hanseok Chae, Jongseok Chae, and their family who
have supported my life in Los Angeles.
I have luxury of having genius friends in Seoul National University. I would
like to thank Taeseop Choi, and Ikjoon Chang for their hearty support. Hocheol Shin
and his family deserve my gratitude, too. He has supported me and helped me find
my future life with Intel.
The turtle family of the Korea Military Academy also deserves my
appreciation. Especially, I would like to thank Jiseok Kim and his family for their
kind support.
I should mention the tremendous helps throughout my graduate student life
from my friends, Alexander Chang, Hyunsuk Song, Sungeun Cho, Sungho Cho,
Inwan Lee, Hyunseop Lim, Dongjoon Choi and Youngbeom Kim.

v
Table of Contents
Dedication ii

Acknowledgments iii

List of Tables viii

List of Figures ix

Abstract xx

Chapter 1: Introduction 1
1.1 Semiconductor Nanowires 1
1.2 Metalorganic Chemical Vapor Deposition (MOCVD) 3
1.3 Selective Area MOCVD 7
1.4 Semiconductor Nanowire Growth Technique 9
1.4.1 Vapor-Liquid-Solid Assisted MOCVD 9
1.4.2 Nanoscale Selective Area MOCVD (NS-SAG) 11
1.5 Thesis Overview 15
Chapter 1 Endnotes 18

Chapter 2: Electron Beam Lithography and Processes for Nanoscale Selective
Area MOCVD Templates 21
2.1 Introduction 21
2.2 Background of Electron Beam Lithography 22
2.2.1 Electron Guns 23
2.2.2 Electron Optics 26
2.2.3 Pattern Generation 30
2.2.4 Coordinate Systems 33
2.2.5 Resists 42
2.3 Optimization of Electron Beam Lithography 45
2.3.1 Acceleration Voltage 46
2.3.2 Resist Material and Process 51
2.3.3 Electron Dose 52
2.3.4 Consideration in Layouts for Nano Opening Arrays 55
2.3.5 Working Distance 59
2.3.6 Beam Current 60
2.3.7 Write Field Size 61
2.4 Processes for NS-SAG Template 61
2.4.1 Template Layouts 62
2.4.2 Process Flow of Template Fabrication 64
2.4.3 Metal Contacts for Individual Nanowires 70
vi
2.5 Summary 75
Chapter 2 Endnotes 77

Chapter 3: Growth of Phosphide Nanowire Arrays 78
3.1 Introduction 78
3.2. Growth of InP Nanowire Arrays 83
3.2.1 Growth of InP Nanowire Arrays on (111)A Substrates 84
3.2.2 Growth of InP Nanowire Arrays on (111)B Substrates 95
3.3 Growth of GaP Nanowire Arrays 99
3.4 Summary 103
Chapter 3 Endnotes 106

Chapter 4: Growth of Arsenide and Heterostructure Nanowire Arrays 108
4.1 Introduction 108
4.2. Growth of GaAs Nanowire Arrays 112
4.2.1 Growth of GaAs Nanowire Arrays on (111)A Substrates 113
4.2.2 Growth of GaAs Nanowire Arrays on (111)B Substrates 114
4.3. Growth of InAs Nanowire Arrays 123
4.3.1 Growth of InAs Nanowire Arrays on (111)A Substrates 123
4.3.2 Growth of InAs Nanowire Arrays on (111)B Substrates 124
4.4 Growth of Ternary Compound Nanowire Arrays 128
4.5 Growth of InP/InAs heterostructure Nanowire Arrays 132
4.6 Summary 134
Chapter 4 Endnotes 137

Chapter 5: Growth Models of NS-SAG and Electron Beam Induced Bundling 140
5.1 Introduction 140
5.1.1 Models of Selective Area Growth 140
5.1.2 Surface Reconstructions 145
5.1.3 Scanning Electron Microscopy 149
5.1.4 Chapter Overview 151
5.2 Diffusion Model for NS-SAG 153
5.2.1 Experiments 153
5.2.2 Theoretical Model 155
5.2.3 Modeling Results 162
5.3 Dependence of Nanowire Growth on the Polarity of Substrates 164
5.4 Electron Beam Induced Bundling of Nanowires 172
5.5 Summary 180
Chapter 5 Endnotes 184

Chapter 6: Conclusions and Future Work 186
6.1 Nanoscale Selective Area MOCVD 186
6.2 Electron Beam Lithography 186
vii
6.3 Growth of Phosphide Nanowire Arrays 188
6.4 Growth of Arsenide and Heterostructure Nanowire Arrays 191
6.5 Growth Models of NS-SAG and Electron Beam Induced Bundling 194
6.6 Future Work 198
Chapter 6 Endnotes 200

Comprehensive Bibliography 201

viii

List of Tables
Table 2-1 Beam currents corresponding aperture diameter of e-Line system. 29
Table 2-2 Flow of two-step deposition process. 66
Table 2-3 Flow of one-step deposition process. 69

ix
List of Figures
Figure 1-1 Experimental nanowire (NW) critical diameters (CDs) and
theoretical critical diameters (CD
MDF
) as a function of lattice
mismatch. Three CD data points were experimentally obtained
and then fitted[8]. 2
Figure 1-2 Process flow of VLS assisted MOCVD. (a) Metal particles are
deposited on a (111) substrate. (b) At eutectic temperature, metal
dots and substrate form eutectic alloys. (c) Metal dots are
supersaturated by supplying precursors. (d) Crystal nanowires
are precipitated under the metal dots. 10
Figure 1-3 Schematic of the (111) crystal plane and nearest low index
planes. 11
Figure 1-4 Process flow of InP nanowire NS-SAG. (a) 30 nm thick SiNX is
deposited on a InP (111)A substrate. (b) Nano opening array is
formed on SiN
X
layer by electron beam lithography and dry etch.
(c) Nanowire arrays are grown by MOCVD. 14
Figure 2-1 Cross-section drawing of a typical electron beam column along
with a raytrace of the electrons as they pass through the various
electron optical components [1]. 23
Figure 2-2 Schematic of thermal field emission gun. 26
Figure 2-3 Cross section schematic of Gemini column designed by Carl
Zeiss, Ltd [4]. 27
Figure 2-4 Emission image of the aperture turret of Raith e-Line system.
The bottom empty space is for blocking the beam when the
blanker is on. 29
Figure 2-5 Two scanning strategies of EBL, (a) raster scan, and (b) vector
scan. 31
Figure 2-6 A write field is a pixel array. The number of bit of the D/A
converter in the pattern generator determines the size of the array.
The pixel spacing is determined by the D/A converter and the
field size. 32
x
Figure 2-7 Finite curvature of focal plane (a) makes astigmatism (b) at the
edge of a write field. The finite curvature also makes pincushion
distortion (c) in the write field. 33
Figure 2-8 Definition of gthe lobal coordinate system (GCS) and the local
coordinate system (LCS). The GCS describes coordinate of
entire sample and the LCS describes coordinate of localized
chips. 35
Figure 2-9 The procedure for UV coordinate definition by using angle
detection. 36
Figure 2-10 The procedure for UV coordinate definition by using 3 point
mark detection. 37
Figure 2-11 Parameters for BCS definition, (a) zoom factors, (b) shifts and (c)
rotations. X and Y are the axes of SCS. X
B
and Y
B
are the axes
of the BCS. 39
Figure 2-12 The procedure for manual write field calibration. The calibration
starts from (a) when the particle image is at the center of the
image. (b) The stage moves in a direction in the SCS. The stage
displacement is set by the user. Then the system makes same
displacement in the BCS by deflecting the beam. If the SCS and
BCS have a mismatch, the image of the particle is not at the
center. An error vector (red arrow) is created. (c), (d) repeat this
in two different displacements. (e) With 3 measured error vectors,
the system calculates the BCS in terms of the SCS. (f) A real
particle image after the displacement. Because of the mismatch,
the particle center is not on the green cross. (g) Users locate the
center of the particle (blue cross). The system calculates the error
vector by mismatch between green and blue cross. 40

xi
Figure 2-13 The illustration of automatic write field calibration. (a) The
system deflects the beam toward the center of the predefined
marks and locates the center of the marks by line scan (red lines).
From the line scan in horizontal and vertical direction, the system
figures out the center of the mark and creates error vectors to
calculate the BCS. (b) The line scan result of a 1.5 m deep etch
pit mark. The system detects the edge of the mark by detecting
falling or rising edge in the scan profile. The graph shows falling
edge detection. Two green horizontal lines are upper and lower
threshold of edge detection. If the profile crosses the threshold
from upper to lower, it detects the falling edge. By detecting
falling edge in two sides, it tries to find the edge from left to right
in the left half of the scan profile and from right to left in the
right half of the profile. From this two edge information, the
system figure out the location of the center (yellow line). 41
Figure 2-14 Expired ZEP 520A cracks after development. The SEM
micrograph shows that the cracks mainly start from the corner of
rectangles. 44
Figure 2-15 Monte Carlo simulation of electron scattering in resist on a
silicon substrate at (a) 10 kV and (b) 20 kV [6]. 47
Figure 2-16 SEM micrographs of 60 nm center-to-center spacing, rectangular
nano opening arrays. The arrays have several defects which
indicate the electron dose condition is very close to the clearing
dose. The V
ACC
and electron dose of the writings are (a) 2.5 kV,
0.3 fC per opening, (b) 10 kV, 0.5 fC and (c) 25 kV, 1.4 fC,
respectively. 49
Figure 2-17 SEM micrographs of the Faraday cup on universal sample holder
of Raith e-Line system. The micrographs are taken with (a) 7.5
m and (b) 30 m limiting apertures. 50
Figure 2-18 Spin curve of 950k PMMA casted in Chlorobenzene [8]. 51
Figure 2-19 3 different types of writing. (a) area writing, (b) single pixel line
writing, and (c) single pixel dot writing. The red dots are
exposed pixels. 52

xii
Figure 2-20 200 nm center-to-center spacing rectangular nano opening arrays
written by e-Line system. All arrays are written at 10 kV. The
smallest, 7.5 m aperture is used, which delivers 7.0 pA beam
current. The micrograph shows surface morphology after the
nano patterns are transferred on 30 nm thick SiN
X
layer by
plasma enhanced reactive ion etch. Three arrays are written by (a)
1.0 fC, (b) 2.4 fC and (c) 5.2 fC electron dose. 54
Figure 2-21 Opening diameter of nano opening in arrays with respect to the
dot dose. This result comes from 200 nm center-to-center
spacing rectangular nano opening arrays written by e-Line
system. All arrays are written at 10 kV. The smallest, 7.5 m
aperture is used, which delivers 7.0 pA beam current. 55
Figure 2-22 D/A converter output and the beam position with respect to time.
The beam position is smoothed out for simplicity. If the
maximum beam deflection speed is slower than that of the D/A
converter sweeping. It creates dynamic effect and distorts the
writing. 58
Figure 2-23 The mechanism of dynamic effect compensation in e-Line
system. Prescan and postscan compensate lagging of beam
deflection. 59
Figure 2-24 The layout of the standard template. (a) the templates has 14 by
14 SiN
X
pad array (cyan boxes), mix-and-match alignment marks
(Green cross is defined by EBL. Pink inverted cross is defined by
photolithography), EBL alignment marks (orange cross), and 100
m write field calibration marks (dark yellow small cross). The
red box shows details of an EBL alignment mark and 4 write
field calibration marks. (b) Illustration of shaped pads. A 50 m
× 50 m SiN
X
pad is enclosed by macroscopic opening array. 63
Figure 2-25 The opening diameter of single deposition process is affected by
the back scattering of pad writing and becomes larger than that of
double deposition process at same electron dose. The diameter
ratio between two processes is approximately 2. 70
Figure 2-26 Schematic of metal contact formation on a vertical nanowire
array, (a) first, the gap between nanowires is filled with polymer.
(b) The top surface is flattened by chemical mechanical etching
(CMP). (c) Contact metal is deposited on the top of the array. 71
xiii
Figure 2-27 The design and alignment scheme of the metal contact pad.
Alignment is achieved by putting the center of layout at the
center of the nanowire. Rotation of group 2 layout (contact arms)
is calculated by the angle between nanowire and u axis. 72
Figure 2-28 PMMA bilayer process is using sensitivity difference between
layers. (a) More sensitive layer (PMMA 495k) is located under
the less sensitive layer (PMMA 950k). (b) After the exposure and
development we can get overhang structure because of
sensitivity difference. (c) SEM micrograph of trenches formed by
the bilayer process. 74
Figure 2-29 SEM micrographs of the metal contact. (a) overview, and (b)
alignment. If alloying is done in hydrogen or nitrogen
environment, (c) nanowires are damaged by thermal etching. 75
Figure 3-1 Cross-section and top view of stacking of tetrahedral
semiconductors. (a) zincblende stacking, and (b) wurtzite
stacking. Red spheres are cations and blue spheres are anions. 80
Figure 3-2 SEM micrographs of InP nanowire array grown on InP (111)A
substrates. All samples are grown at different NPC. The NPC of
each sample is (a) 4.0, (b) 5.6 (c) 10.2, and (d) 24.0, respectively. 86
Figure 3-3 Surface morphology parameters with respect to NPC. (a)
Tapering parameter, 
T
and (b) lateral growth parameter, 
L
. 87
Figure 3-4 60  tilted SEM micrograph of an InP nanowire has sidewall
bumps 88
Figure 3-5 Average numbers of bumps per nanowire with respect to the
normalized precursor concentration. 89
Figure 3-6 TEM micrographs of [111]A direction InP nanowires. Upward
direction is the [111]A direction. (a) Lattice image of an InP
nanowire taken in the [11 ¯ 00] zone axis, (b) image of a sidewall
bump and SFs taken in the [112 ¯ 0] zone axis, and (c) selective
area electron diffraction pattern in the [11 ¯ 00] zone axis. 90
Figure 3-7 Cross sectional schematic of the sidewall bump. Cross section
plane is (112 ¯ 0). It is also marked red solid line in the top view of
nanowire. Red dotted line is the projection of the (11 ¯ 01)A semi-
polar plane. Blue dotted line is the projection of {111}A polar
plane. ZB style SFs are marked by green band. 92
xiv
Figure 3-8 Surface morphologies of doped InP nanowire arrays. N-doping is
done by Si
2
H
6
at (a) 4.00 and (b) 15.87 IV/III ratio. Circled
nanowires show weakened sidewall growth suppression. P-
doping is done by DEZ at (c) 0.21 and (d) 1.70 II/III ratio. 94
Figure 3-9 Surface morphologies of InP nanowire NS-SAG on a GaAs
(111)A substrate. 94
Figure 3-10 Surface morphologies of InP nanowire arrays grown on InP
(111)B substrates. Two InP nanowire arrays are grown under
6.0×10
-7
atm of TMI partial pressure and 1.4×10
-4
atm of PH
3

partial pressure at (a) 625 C and (b) 650 C, respectively. Other
arrays are grown at 650 C with different precursor partial
pressures. (c) An InP nanowire array is grown at 3.0×10
-7
atm of
TMI partial pressure and 1.4×10
-4
atm of PH
3
partial pressure. (d)
The other array is grown at 6.0×10
-7
atm of TMI partial pressure
and 1.8×10
-4
atm of PH
3
partial pressure. All templates are
prepared by 1 step deposition process and treated by diluted HCl
before the growth. 96
Figure 3-11 Surface morphology change of InP nanowire arrays grown on
InP (111)B substrates with respect to SAG template preparations.
(a) InP nanowires grown on a template prepared by 2 step
deposition process, and (b) tripod structures occurred on a
template treated by H
2
SO
4
before the growth. 97
Figure 3-12 HR-TEM observations of InP nanowire grown in [111]B
direction. (a) The lattice image of the [111]B direction InP
nanowire. Upward direction is the [111]B direction. (b) SAED
pattern of a [111]B direction InP nanowire. 98
Figure 3-13 SEM micrograph of an InP nanowire array grown on GaAs
(111)B 99
Figure 3-14 Surface morphology of an GaP nanowire array grown at 700 C.
The substrate is InP (111)B. (a) GaP nanowire array grown at
2.36×10
-6
atm of TMG partial pressure and 2.14×10
-4
atm of PH
3

partial pressure, (b) GaP nanowire array grown at 1.18×10
-6
atm
of TMG partial pressure and 2.14×10
-4
atm of PH
3
partial
pressure, and (c) GaP nanowire array grown at 1.18×10
-6
atm of
TMG partial pressure and 1.42×10
-4
atm of PH
3
partial pressure. 100

xv
Figure 3-15 Schematic of growth window. P
PH3
is partial pressure of PH
3
. P
V

and P
L
are respectively the minimum partial pressure of PH
3
for
growth on polar (111) surface and the maximum PH
3
partial
pressure to suppress growth on non-polar sidewalls. The
overlapping region is the growth window. 101
Figure 3-16 SEM micrographs of GaP nanowire arrays grown at 650 C.
Partial pressures of TMG and PH
3
are 1.18×10
-6
atm and
1.42×10
-4
atm, respectively. The opening diameters are (a) 30 nm
and (b) 50 nm. 102
Figure 3-17 SEM micrographs of GaP nanowire arrays grown on InP (111)A
substrate. 103
Figure 4-1 The top view of GaAs NS-SAG on a GaAs (111)A substrate. 114
Figure 4-2 SEM micrographs of GaAs nanowires grown on GaAs (111)B
substrates. The center-to-center spacing of the nanowire array is
200 nm and opening diameter is 50 nm. Growth temperature and
AsH
3
partial pressure changes from (a) 650 C, and 1.00×10
-4

atm, (b) 700 C and 2.00×10
-4
atm, and (c) 750 C and 2.00×10
-4

atm. Circled region shows abnormal growth. 116
Figure 4-3 High magnification SEM micrographs of localized growth
suppression in GaAs nanowire arrays. Opening diameter shown
here are (a) 80 nm and (b) 40 nm, respectively. 117
Figure 4-4 Growth behavior difference of GaAs nanowire NS-SAG caused
by array packing density. Both templates are grown at 650 C.
Center-to-center spacing of two templates is (a) 80 nm and (b)
200 nm, respectively. 118
Figure 4-5 AsH
3
flow modulation growth for uniform array growth. (a)
AsH
3
flow is modulated stepwise. The center-to-center spacing
of array is (b) 80 nm and (c) 200 nm, respectively. 120
Figure 4-6 HR-TEM observation of GaAs nanowire in the [111]B direction.
(a) A lattice image. The [111]B direction is upward direction of
image. (b) SAED pattern of grown nanowire. 121

xvi
Figure 4-7 SF density change of 3 GaAs nanowires in the [111]B direction
grown in different growth conditions. The nanowires are grown
under different temperature and TMG partial pressure. Tried
temperature and TMG partial pressure are, (a) 750 C and
3.79×10
-7
atm (b) 650 C and 3.79×10
-7
atm, and (a) 750 C and
7.58×10
-7
atm. 122
Figure 4-8 InAs nanowire NS-SAG on a (111)A substrate. 124
Figure 4-9 InAs nanowire arrays on GaAs (111)B substrates. The center-to-
center spacing is fixed at 200 nm. The opening diameter is also
fixed at 50 nm. These arrays are grown at (a) 550 C, (b) 600 C,
and (c) 650 C, respectively. 126
Figure 4-10 Growth rate fluctuation with respect to the opening diameter.
The center-to-center spacing is fixed at 200 nm. (a) 50 nm, and
(b) 85 nm are chosen to contrast the effect of surface diffusion
term. 127
Figure 4-11 Bright field HR-TEM images of an InAs nanowire grown in the
[111]B direction. (a) Low magnification over view looks similar
to GaAs nanowires. Upward direction is the [111]B direction. (b)
Lattice image shows rapid SF generation. Measured SF period is
2.4 MLs. (c) SAED pattern shows streaks due to dense SFs. 128
Figure 4-12 SEM bird view (45  tilt angle) micrographs of InAsP nanowires
grown at 1.17 of PH
3
/AsH
3
ratio. The center-to-center spacing is
fixed at 200 nm. The diameters of openings are respectively (a)
50 nm, (b) 60 nm, (c) 80 nm, and (d) 160 nm. 130
Figure 4-13 SEM bird view (45  tilt angle) micrographs of InAsP nanowires
grown at 9.00 of PH
3
/AsH
3
ratio. The center-to-center spacing is
fixed at 200 nm. The diameters of openings are respectively (a)
50 nm, (b) 60 nm, (c) 80 nm, and (d) 160 nm. 131
Figure 4-14 InP/InAs heterostructured nanowire arrays grown by NS-SAG. (a)
InP nanowire array are grown for 10 minute for bottom part of
heterostructured nanowires. (b) After InAs growth, the nanowires
are bent by non-uniform InAs growth on InP nanowire sidewalls.
(c) Vertical InP/InAs heterostructures are also formed. InAs part
are tilted toward the nearest <111>B direction. (d) An InAs
tripod structure is grown on the top of an InP nanowire. 133
xvii
Figure 5-1 Schematic diagram of two-dimensional simulation domain and
boundary conditions[5]. 142
Figure 5-2 Schematic diagrams of surface reconstructions, (a) GaAs(111)A-
(2×2), (b) GaAs(111)B-(2×2), and (c) InP(111)A-( 3× 3)R30.
Red dotted lines are surface unit cell boundaries. 148
Figure 5-3 Origin and information depth of secondary electrons (SE),
backscattered electrons (BSE), Auger electrons (AE) and X-ray
quanta (X) in the diffused cloud of electron range R for normal
incidence of the primary electrons (PE)[13]. 151
Figure 5-4 SEM images of an InP nanowire array with 350 m wide skirt
grown by NS-SAG with respect to the location in the array. Scale
bars are 200 nm. 154
Figure 5-5 Surface morphology of InP nanowires at the edge of array. The
array has no skirt. Because of diffusion toward bare
semiconductor region, The growth is suppressed. 155
Figure 5-6 Schematic diagrams of VPD simulation, (a) full VPD model
without any approximation, and (b) VPD model with average
adsorption approximation. 157
Figure 5-7 Simulation result of VPD model with and without average
adsorption approximation. 160
Figure 5-8 Simulation result of approximated VPD model. The size of array
is 1 mm × 1 mm, with 350 m wide skirt. The diameter and
center-to-center spacing is 90 nm and 200 nm, respectively. (a)
2D concentration profile is expressed in color. (b) 1D
concentration profile of AB section clearly shows VPD from the
skirt regions. 161
Figure 5-9 Estimation of nanowire length by the VPD based model. The
model only considers VPD mechanism. The size of the tested
array is 1 mm × 1 mm, with 350 m wide skirt. The diameter and
center-to-center spacing is 90 nm and 200 nm, respectively. 162
Figure 5-10 Estimation of nanowire length by the VPD based model. The
model only considers VPD mechanism. The size of the tested
array is 1 mm × 1 mm, with a 350 m wide skirt. The diameter
and center-to-center spacing is 90 nm and 200 nm, respectively. 164
xviii
Figure 5-11 SF density of InP nanowires with respect to TMI partial pressure.
The nanowires are grown in the [111]A direction. The average
diameter of nanowires is 50 nm. 166
Figure 5-12 Schematic diagram of Adatom adsorption on GaAs (111)B
surface with high As trimer coverage. Red arrows represent
strong repulsion and orange arrows represent weak repulsion. 168
Figure 5-13 Schematic diagram of Adatom adsorption on (a) GaAs (111)A
surface and (b) InP (111)A surface. The GaAs surface forms
(2×2) reconstructed surface and the InP surface forms
( 3× 3)R30  reconstruction. 171
Figure 5-14 Top view of 200 nm thick InP nanowire array. Black stains on
the top are pits on the top surface. 172
Figure 5-15 Typical artifact in SEM measurement of thin nanowires whose
diameter is smaller than 60 nm. The artifact looks like surface
oscillation in (a) fast scan view and smeared ghost image in (b)
slow scan view. All micrographs are taken by Hitachi S4800. 173
Figure 5-16 Time evolution of EBIB. The snap shots are taken at (a) the
beginning of EBIB event and (b) 2 scans, (c) 4 scans and (d) 6
scans after the event. 174
Figure 5-17 EBIB experiment in Raith e-Line system. Acceleration voltage
and beam current are 5 kV and 180 pA, respectively. As in
Hitachi S-4800, (a) ghost images appear before the EBIB event.
After 2 minutes of single spot electron beam projection, (b) the
target and neighboring nanowires are damaged by electron beam.
(c) EBIB propagates along the beam direction. 176
Figure 5-18 Schematic of opposite charge building in nanowires. If electron
beam passes through a nanowire, positive charges are built by
secondary electron emission. On the contrary, if the electron
beam stops in the middle of a nanowire, the nanowires are
negatively charged. Due to strong depletion, these charges
cannot be dissipated effectively. Then, strong Coulomb attraction
moves the nanowires. 176

xix
Figure 5-19 Calculation setup of nanowire bending force. Center-to-center
spacing is 2d. The displacement of nanowire, x, by bending is
equal on both sides. It is assumed that the nanowires have
prefect hexagonal cross section. 177
Figure 5-20 The concept and experiment result of of controlled EBIB. (a)
Incident angle must be adjusted to project electron beam only on
the first nanowire. Electron beam pass should not touch the third
nanowire. By controlled EBIB, pair-wise bundling is possible in
wanted direction. (b) The result of coltrolled EBIB is circles in
the micrograph. 179

xx
Abstract
The interesting properties and potential applications of semiconductor
nanowires have received significant attention. Nanoscale selective area growth using
MOCVD (NS-SAG) has been demonstrated as an attractive growth technique for
compound semiconductor nanowires. With this technique, the diameter and location
of wires can be controlled, and no unwanted metal incorporation occurs. This
technique is also suitable for large scale uniform nanowire arrays.
Electron beam lithography conditions are optimized to define nano opening
arrays for NS-SAG. From this optimization, 25 nm minimum opening diameter and
80 nm center to center spacing features are achieved. InP, GaP, GaAs, and InP
nanowire arrays are grown on nano patterned InP (111) and GaAs (111) substrates
using NS-SAG. Vertical and uniform InP nanowire arrays are demonstrated on InP
(111)A substrates; however, uniform nanowire arrays of GaAs and InAs are
achievable only on GaAs and InP (111)B substrates. The surface morphology of InP
nanowires is strongly affected by the effective precursor concentration. Enhanced
growth in non-polar directions under high precursor concentration disturbs InP
nanowire formation. GaAs and InAs NS-SAGs are affected by heterogeneous
pyrolysis of arsine and surface diffusion on the sidewalls. The excessive As supply
by heterogeneous pyrolysis causes epitaxial burial of nanowires in dense GaAs
nanowire arrays and growth rate fluctuations in InAs nanowire arrays. InP nanowires
grown in the [111]A direction are in the wurtzite structure at diameters greater than
xxi
the thermodynamic estimation of the structural transition in diameter. GaAs, InAs,
and InP nanowires grown in [111]B direction are in the zincblende structure with
rapid stacking fault generation. InAsP and InP/InAs heterostructure nanowire NS-
SAG shows the conflict in the preferred growth direction for each binary compound.
A diffusion theory based NS-SAG model is proposed. To reduce the huge
computational cost of a vapor phase diffusion simulation of NS-SAG, an average
adsorption approximation is proposed. The surface diffusion term is also included to
describe the strong diffusion effect induced by the sidewalls. The model predicts the
growth rate well, with an average error of 9%, and also predicts a strong surface
diffusion effect in GaAs NS-SAG which is consistent with experimental results. The
difference of the preferred growth direction, crystal structure transition and stacking
fault generation are explained by surface reconstruction of {111} surfaces and the
bond strength of III-V and V-V bonds. The presence of the mask and phosphorus
trimers, enhances wurtzite stacking of InP nanowires on InP (111)A substrates. GaAs
and InAs are prone to grow in [111]B direction because As trimers can be easily
dissociated due to weak As-As bonds. Stacking fault generation is also explained by
the interaction between surface trimers and adatoms.
Strong artifacts and induced bundling are observed in scanning electron
microscopy. By single spot electron beam projection, it is found that the electron
beam can pass through the nanowire easily and bundling can propagate along the
electron beam path. This induced bundling is explained by attraction between
xxii
positive charges generated by secondary electron emission and negative charges
from the electron beam. From this observation, controlled bundling is demonstrated.

1
Chapter 1: Introduction
1.1 Semiconductor Nanowires
Semiconductor nanostructures have attracted much attention because of their
unique electrical and optical characteristics. Among the nanostructures, quantum
wells have been widely used for semiconductor laser and LED applications[1-2].
Recently, lower dimension structures, such as nanowires and quantum dots have
been actively studied. By virtue of their stronger quantum confinement, the devices
have more energetically discrete natures and many interesting properties. Especially,
nanowires are proposed as a good candidates for next generation electronic
devices[3-4] and photonic devices[5].
Nanowires have many interesting properties. Vertical nanowires can be used
as a resonant cavity for quantum dot devices[6]. If a quantum dot is embedded in a
nanowire, spontaneous emission of the quantum dot is enhanced by the Purcell effect.
This increases the emission efficiency of a quantum dot single photon emitter and
suppresses exciton dephasing[7]; therefore, more indistinguishable photons are
generated, which essential for quantum optic photon sources. Also increase of
emission efficiency reduces multi photon emission from the source.

2

Figure 1-1 Experimental nanowire (NW) critical diameters (CDs) and theoretical
critical diameters (CDMDF) as a function of lattice mismatch. Three CD
data points were experimentally obtained and then fitted[8].
All nanowires have limited lateral dimension; however, freestanding
nanowires have no inherent limitation in their lateral expansion. In heteroepitaxy,
nanowires have far less accumulation of strain due to their small radial dimension.
This enables the growth of heavily mismatched heterostructures, often known as
‘nanoheteroepitaxy (NHE)’. The possibility of NHE was first proposed by Luryi and
Suhir[9]. Their original idea is not for NHE. They proved that if we have free space
in the lateral direction, the strain of the heterostructure is relaxed and critical
thickness is enhanced. Ertekin et al applied this idea to nanowires and proved that
the critical thickness criterion can be alleviated in heterostructure nanowires[10].
Chuang et al experimentally proved his idea[8]. They reported that 450 nm long,
misfit dislocation free nanowires can be achieved at 11% lattice mismatch if the
nanowire diameter is smaller than 45 nm as shown in Figure 1-1. This capability of
defect free, highly mismatched structures is extremely important for photovoltaic
applications. To capture and convert the wide solar spectrum from IR to UV without
3
losing conversion efficiency, we should use many different materials with various
bandgap. Spectrolab’s triple junction tandem solar cell achieved the state-of-the-art
performance[11]; however, since their technology is based on conventional film
epitaxy technology, they must chose their layers to be lattice matched to prevent
misfit dislocation problem. Nanowires are a promising candidate for multijunction
photovoltaic devices.
Nanowires have been considered as a promising structure for transistor
channels. As the conventional silicon (Si) based planar integrated circuit (IC)
technology approaches its dead end, pursuing Moore’s law has become increasingly
challenging. Many researchers believe that compound and Si nanowires could be the
breakthrough for this challenge. Tanaka demonstrated vertical field effect transistors
(FETs)[12], which enable three dimensional ICs. For high speed devices, indium
arsenide (InAs) nanowires are considered as a current channel for a transistor. The
superior electron mobility of InAs could boost the performance of n-channel FETs.
Many successful demonstrations have been published[12-13].
1.2 Metalorganic Chemical Vapor Deposition (MOCVD)
In order to achieve single crystal nanowires with well defined crystal plane
sidewalls, growing nanowires using crystal growth technology is preferable to
etching techniques. It is also challenging to achieve perfectly clean sidewalls and
high aspect ratio structures at the same time using etching techniques. Moreover, as
4
we reach nanometer level feature sizes, control of interfaces and features close to the
single crystal lattice level is required, which can only be practically achieved using
epitaxial technologies. Metalorganic chemical vapor deposition (MOCVD) is the
central epitaxial technologies and is the leading candidate for the fabrication of the
sophisticated heterojunction devices and nanostructures.
MOCVD is a process for crystal film growth that utilizes vapor phase
precursors, which carry needed elements for the target materials being grown such as
gallium arsenide (GaAs) or indium phosphide (InP). In a MOCVD reactor chamber,
vapor phase precursors undergo phase transformation from vapor to solid and form
crystal layer on substrates. Many source metals, especially group III elements in III-
V semiconductors, which are essential components of compound semiconductor, are
not volatile in useful temperature and pressure ranges. Metalorganic compounds are
used as vapor phase precursors. Many metalorganics are relatively volatile and can
generate diluted vapor phase precursor stream by passing a carrier gas (usually
hydrogen or nitrogen) through or over the metalorganics contained in bubbler vessels
at controlled temperatures and pressures. This process is called ‘bubbling’. The
carrier gas transports the equilibrium vapor pressure of the metalorganic into a gas
mixing manifold where it is then diluted with a carrier gas and mixed with the other
reactants. In MOCVD growth of III-V compound semiconductor, group V elements
for III-V semiconductors are usually organic-group V element compounds or
hydrides, such as arsine (AsH
3
) or phosphine (PH
3
). These are also transported into
the gas mixing manifold by the carrier gas. In the reactor, the transported precursors
5
are thermally cracked and metallic (group III in III-V) and nonmetallic (group V in
III-V) elements are released. The released reactants interact on a substrate and form
crystals. This process can be expressed by simple chemical reaction formulae. In
the growth of the compound semiconductors of the type AB[14]:

nRH AB BH A R
H
n n
  
2
(1-1)
' or ' '
2
RR B A H nR nRH AB B R A R
H
n n
     (1-2)
, where R and R’ represents an organic radical of unspecified form. However, the
real reactions in the reactor are much more complicated than Equation (1-1) and (1-
2). The pyrolysis process of precursors is the key to this understanding. We can
categorize the pyrolysis process into two groups. One is homogeneous pyrolysis,
which occurs entirely in the vapor phase. Heterogeneous pyrolysis is a pyrolysis
process occurring at a solid surface.
Growth temperature, reactor pressure and partial pressures of precursors are
important parameters to control the reaction. The precursor partial pressures are
controlled by adjusting the ratio of the flow rate between precursors and the carrier
gas. If a precursor flow is not diluted, the partial pressure of the precursor is

tot
pre
r p
I
I
P P  (1-3)
Here, P
p
, P
r
, I
pre
, and I
tot
are reactor pressure, flow rate of precursor and total flow
rate of gases into the reactor, respectively. We must perform more analysis to find
the partial pressure of the metalorganic precursors, because they form an equilibrium
6
vapor pressure in the carrier gas. Therefore the partial pressure of metalorganic can
be expressed as

tot
pre
b
v
r p
I
I
P
P
P P  (1-4)
, where P
v
and P
b
are vapor pressure of the metalorganic at a bubbler vessel
temperature and bubbler pressure.
In the reactor, the precursors are hydrodynamically transported from the gas
inlet to the boundary layer. The boundary layer is the region of the reactor tube cross
section over which the relatively uniform flow rate of the laminar gas flow
approaches zero velocity at the substrate surface or the reactor walls. In the boundary
layer, precursors are transported by diffusion. In the diffusion path, they cross the
region in which there exists strong temperature gradient. Most of the metalorganics
are decomposed in this region. The hydrides may be decomposed in this region, but
because of their superior thermal stability, most of them are heterogeneously
decomposed with assistance of the surface reaction[15].
In the mass transport limited temperature regime, the growth rate will be
relatively independent of temperature owing to the small temperature dependence of
the diffusivity of most gaseous species. In this regime, the growth rate of, for
example, InP, is controlled by the partial pressure of trimethylindium (TMI). Since
the surface reactions are much faster than the transport of the reactants through the
boundary layer in this temperature regime, the growth rate was found to be linearly
dependent upon the concentration of the reactant in the inlet[14]. This is the case for
7
most of the MOCVD processes. Outside of the mass transport regime, the growth
rate will be a sublinear function of the partial pressure of the metalorganic but also
substrate orientation, and / or hydride partial pressure, and / or temperature
dependent[16-17].
1.3 Selective Area MOCVD
If we want to grow an epi-structure across a limited region of a substrate, the
crystal growth reaction must be limited outside of that region. This can be achieved
by depositing a patterned dielectric mask on the substrate. Because precursors
cannot form crystals on the dielectric mask, growth occurs only in openings of the
dielectric mask. This selective area MOCVD or simply, selective area growth (SAG)
has been widely used to demonstrate photonic integrated circuits such as laser
diodes[1], DBR lasers[18], mode profile converters[19], and optical amplifiers[20].
In SAG, the growth rate and composition of ternary and quaternary alloys
changes due to complex diffusion mechanisms[21]. Because there is no precursor
consumption on the mask region, the spatial precursor concentration profile is
modulated on a substrate is modulated by the patterning of the dielectric mask.
There are three different diffusions involved in the growth of the openings[21];
vertical vapor phase diffusion across the boundary layer, lateral vapor phase
diffusion driven by the lateral precursor concentration difference caused by the
8
dielectric mask, and surface diffusion from the mask region, which is diffusion of
precursors falling on the mask region that cannot form crystals on the mask region.
SAG has been modeled using surface diffusion theory and vapor phase
diffusion theory. In the surface migration model, precursor molecules falling on the
mask region diffuse into the opening, as they cannot react for growth on the mask.
The growth is locally enhanced at the mask edge region because of this diffusion.
Kayser showed that the vapor phase and surface stoichiometry and the V/III ratio
play an important role in controlling the morphology of the growth[22]. According
to the vapor phase diffusion theory, vapor phase precursors that are crossing through
the boundary layer to the surface accumulate on the dielectric mask regions because
there is no precursor consumption on the mask region. This modulates the lateral
precursor concentration profile and causes diffusion. The local growth enhancement
also occurs at the edge of mask region due to this diffusion. Coronell and Jenson
used both vapor phase diffusion and surface diffusion together on masks to model
SAG[23]. Choi introduced a vapor phase diffusion based SAG model[24]. In his
model, binary compounds were treated as a single diffusive molecule to consider the
diffusion length difference of group III elements with respect to the group V
precursors.

9
1.4 Semiconductor Nanowire Growth Technique
1.4.1 Vapor-Liquid-Solid Assisted MOCVD
Wagner and Ellis reported a growth mechanism for a Si pillar structure in
1964. This growth technique utilizes precipitation of single crystal under precursor
supersaturated metal dots. This technique has been called vapor-liquid-solid assisted
MOCVD or simply the VLS technique, because there are vapor phase precursors,
liquid phase precursors dissolved in metal dots, and solid growth. Figure 1-2 shows
the process step for VLS growth. Nano size metal dots are used as localized solvents
for precursors. By an annealing process, the dots form a eutectic mixture with the
semiconductor material. Continuous precursor flows causes the dots to become
supersaturated. Then the supersaturated dots precipitate single crystal nanowires,
with the growth occurring at the dot-substrate interface. This process is simple and
has a wider growth window for nanowires, because it has a physical limitation for
lateral growth, the diameter of the eutectic mixture. The growth temperature is
around the eutectic temperature of III-metal mixture, which approximately ranges
from 400 C to 500 C.
10
Ga Ga As
(a) (b)
Ga Ga As
(c) (d)

Figure 1-2 Process flow of VLS assisted MOCVD. (a) Metal particles are deposited
on a (111) substrate. (b) At eutectic temperature, metal dots and substrate
form eutectic alloys. (c) Metal dots are supersaturated by supplying
precursors. (d) Crystal nanowires are precipitated under the metal dots.
In this process, unwanted metal incorporation in nanowires cannot be avoided,
and it is also difficult to control the nanowire location. Because the metal
semiconductor surface is unstable in the useful growth temperature range, the dots
can drift on the surface before they form nanowires. To prevent this, a capturing
layer is required to trap the dots. In VLS growth, the growth reaction happens at
liquid-solid-vapor interface. Therefore, surface tension among a metal dot,
semiconductor surface and newly grown material seriously affects the growth
behavior. It has been reported that heterostructure nanowires grown by VLS
techniques have bending or kinks at the heterointerfaces due to the surface tension
differences between the materials[25]. According to this work, only a few
heterostructure combinations allow for vertical stacking. This would be a big
obstacle for fabricating heterostructure nanowires. Forming abrupt heterojunctions
is another potential difficulty. Since the growth mechanism relies on the
11
precipitation of nano size metal solvents, the growth cannot be abruptly stopped.
Specifically, fabricating group III heterojunctions, such as InAs/GaAs heterojunction
is challenging, which may severely degrade the performance of possible
heterojunction devices.
1.4.2 Nanoscale Selective Area MOCVD (NS-SAG)
According to section 1.3, we can restrict the growth, region on a substrate by
utilizing a patterned substrate. If a nano opening array is patterned on a dielectric
mask, the mask may develop nanometer size clusters. This lateral growth restriction
exposes crystal facets which do not usually appear in typical film growth. If the
growth condition does not allow growth on these planes, growth occurs only on the
top of the cluster, which is the same as the substrate plane, and we can get
nanoclusters with high aspect ratios. This is the key idea of nanoscale selective are
MOCVD growth (NS-SAG).
(111)
(111) (111)
(111)
(011)
(101)
(001) (100)
(010)
(011) (110)
(110)
(110)
(101)
(011)
(101)

Figure 1-3 Schematic of the (111) crystal plane and nearest low index planes.
12
The (111) plane of zincblende (ZB) semiconductors can make this story real.
In compound semiconductors, the {111} planes are perfect polar planes. One of the
tetrahedral bonds is perpendicular to {111} planes. The dipole moment of the cation
and anion is along this <111> directions and the polarization is maximized in this
direction. In the ideal case, the plane surface is covered by one kind of atoms, group
III or group V. There are a total of 8 {111} planes. They can be divided into two
groups with respect to the direction of polarization. If the surface normal vector is
along cation (group III atom) to anion (group V atom), the planes are {111}A, which
are (111), (1 ¯ 1 ¯ 1), (1 ¯ 11 ¯ ), and (11 ¯ 1 ¯ ). If the surface normal vector is along anion to
cation, the plane is {111}B, which are (1 ¯ 1 ¯ 1 ¯ ), (1 ¯ 11), (11 ¯ 1), and (111 ¯ ). Figure 1-3 is
the (111) plane and neighboring low index planes. The schematic depicts that the 6
of {11 ¯ 0} planes enclose the (111) plane, and {11 ¯ 0} planes and the (111) plane are
perpendicular. These {11 ¯ 0} planes are non-polar planes, of which the net
polarization is zero in the surface normal direction. The {11 ¯ 0} planes consist of an
equal number of group III atoms and group V atoms.
The growth behaviors on {111} planes and {11 ¯ 0} planes are significantly
different due to the difference in the composition of atoms on the surfaces and the
surface reconstruction. Therefore, it is possible to find a growth condition that
allows growth on {111} planes and, at the same time, restricts the growth on the
{11 ¯ 0} planes. NS-SAG utilizes this growth behavior difference. Figure 4 is the
schematic diagram of the InP nanowire NS-SAG process flow. The process flow of
NS-SAG for other materials such as GaAs or InAs is exactly same for InP nanowire
13
NS-SAG except for the precursors and substrates. First a thin dielectric film (silicon
dioxide (SiO
2
) or silicon nitride (SiN
X
, Si
3
N
4
in stoichiometric conditions)) is
deposited on a (111) substrate using plasma enhanced CVD. Nano opening arrays
are defined by a nano lithography technique such as electron beam lithography. The
defined nano pattern is transferred on the dielectric film by wet etching[26-27] or dry
etching. Wet etching is advantageous to minimize surface damage that can be
induced by ion bombardment during the dry etching process; however, the resolution
is limited by isotropic nature of wet etching. In the MOCVD process, the mask
restricts the lateral growth of the (111) plane and causes {11 ¯ 0} planes to appear. In
a proper growth condition, the growth on {110} planes is suppressed. The growth
only occurs on the top of the nanowires, the (111) plane. As a result, we can get
hexagonal cross section nanowire arrays.
This technique makes no unwanted metal incorporation. This is extremely
important for integration on Si electronics. The most widely used metal solvent in
VLS technique is gold, which causes critical performance degradation in Si
electronic devices. Fukui has demonstrated InAs nanowire arrays on Si (111) wafers
using NS-SAG[28]. NS-SAG is also suitable for uniformly distributed dense
nanowire arrays. The size and location of nanowires are strictly regulated by the
nano pattern. In a proper growth condition, the critical dimension and positioning
accuracy can be identical to the accuracy of current nano lithography technologies.
The technique is advantageous in fabricating abrupt heterojunctions. In typical
MOCVD, the control of layer thickness and heterojunction interface reaches down to
14
the atomic monolayer. Theoretically, the abruptness of heterojunctions grown by
NS-SAG is identical to generic MOCVD. Fukui and his colleagues have made many
advances in nanowire growth using different materials[26-27, 29], and device
fabrication[12, 30] using this technique .
SiN
X
InP(111)A substrate
Electron beam lithography
CF
4
dry etch
MOCVD growth
(TMI/PH
3
)
(a)
(b)
(c)

Figure 1-4 Process flow of InP nanowire NS-SAG. (a) 30 nm thick SiNX is deposited
on a InP (111)A substrate. (b) Nano opening array is formed on SiNX
layer by electron beam lithography and dry etch. (c) Nanowire arrays are
grown by MOCVD.
However, the complex diffusion processes in the growth makes the growth
estimation challenging. Because this technique relies on the growth behavior
difference between crystal planes, the allowable growth window is narrower than
15
with the VLS technique. Therefore, more precise control of growth conditions and a
good growth model, which can estimate growth behavior accurately, are required.
1.5 Thesis Overview
III-V semiconductor nanowires are emerging materials as good candidates for
next generation electronic and photonic devices as well as photovoltaic and light
emitting devices. Since the metalorganic chemical vapor phase deposition technique
was first introduced in 1968[31], it has been widely used for epitaxial device
fabrications. Single crystal nanowires can be achieved using MOCVD with
assistance of nano patterned dielectric masks. For the study of this dissertation,
nanoscale selective are MOCVD growth technique, which is using metal organic
precursors as sources for group III elements and hydrides gases for group V elements,
is used to grow InP, GaAs, InAs and GaP nanowire arrays.
For NS-SAG, preparing nano patterned dielectric masks is important. Size,
location and density of nanowires as well as growth rate and behavior are governed
by the patterning parameters of the nano patterned dielectric masks. The electron
beam lithography technique is suitable to generate nano pattern with high flexibility.
This lithography technique is extremely useful if patterns need to be changed
frequently. While it gives high flexibility in pattern generation, the process is very
slow and there are many parameters which should be optimized for generating
uniform nano opening array patterns and reaching ultimate resolution limit of the
16
system. In this work, a Raith e-Line electron beam writer is used to define the nano
opening arrays. In chapter 2, the parameters for electron beam lithography and basic
principles of operation are discussed. Based on these principles, extensive
optimization work and its subsequent results are presented. An overview of the NS-
SAG template process is also explained in chapter 2.
NS-SAG condition optimization and characteristic growth behavior of the
phosphide materials, InP and GaP, are discussed in chapter 3. Growth behavior
changes of InP nanowire NS-SAG with respect to substrate polarity, growth
conditions such as growth temperature and precursor partial pressure, patterning
parameters such as opening diameter and center-to-center spacing of openings are
discussed. Extensive TEM research reveals the crystal structure of grown InP
nanowires. It has been known that, in nanowire form, these materials show
transitions in their crystal structures due to large energy contribution from the crystal
surface formation. The observations of structural transition of InP nanowire are also
discussed in chapter 3.
Chapter 4 explains the growth results of arsenide based nanowire arrays
using NS-SAG. Again, studies of the growth behavior changes with respect to
growth parameters and patterning parameters are presented. The results of ternary
indium arsenide phosphide (InAsP) nanowire growth and InP/InAs heterostructure
nanowire growth are explained in chapter 4.
In chapter 5, a growth model of for NS-SAG is proposed. The model is
based on the vapor phase diffusion theory of SAG. It also considers surface
17
diffusion term to model the strong surface diffusion component on the nanowire
sidewalls. This chapter will provide a model for the surface kinetics in NS-SAG and
explain the dependence of the preferred growth direction on the polarity of substrate,
stacking fault generation and structural transitions of nanowires using this proposed
model. In the last section of this chapter, observations and a possible model for
electron beam induced bundling are discussed. InP and GaAs nanowires are bundled
under focused electron beam projection. This process can be understood by charging
induced by the electron beam. The results of extensive scanning electron
microscopy and bundling experiments are presented in this chapter. Based on these
observations, a model based on secondary electron emission is proposed which can
explain why an electron beam causes positive charge accumulation.
In Chapter 6, all of the results of this thesis will be summarized.

18
Chapter 1 Endnotes
[1] J. S. Osinski, Y. Zou, P. Grodzinski, A. Mathur, and P. D. Dapkus, IEEE
Photonics Technol. Lett. 4 10-13 (1992).
[2] N. C. Frateschi, P. D. Dapkus, S. S. Ou, J. J. Yang, and M. Jansen, IEEE
Photonics Technol. Lett. 5 741-743 (1993).
[3] Y. Cui, Z. Zhong, D. Wang, W. U. Wang, and C. M. Lieber, Nano Lett. 3
149-152 (2003).
[4] V. Schmidt, H. Riel, S. Senz, S. Karg, W. Riess, and U. Gösele, Small 2 85-
88 (2006).
[5] M. A. Zimmler, T. Voss, C. Ronning, and F. Capasso, Appl. Phys. Lett. 94
241120 (2009).
[6] A. Tribu, G. Sallen, T. Aichele, R. g. André, J.-P. Poizat, C. Bougerol, S.
Tatarenko, and K. Kheng, Nano Lett. 8 4326-4329 (2008).
[7] J. Vuckovic, D. Fattal, C. Santori, G. S. Solomon, and Y. Yamamoto, Appl.
Phys. Lett. 82 3596-3598 (2003).
[8] L. C. Chuang, M. Moewe, C. Chase, N. P. Kobayashi, and C. Chang-Hasnain,
Appl. Phys. Lett. 90 043115 (2007).
[9] S. Luryi and E. Suhir, Appl. Phys. Lett. 49 140-142 (1986).
[10] E. Ertekin, P. A. Greaney, D. C. Chrzan, and T. D. Sands, J. Appl. Phys. 97
114325 (2005).
[11] R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H.
Yoon, R. A. Sherif, and N. H. Karam, Appl. Phys. Lett. 90 183516 (2007).
19
[12] T. Tanaka, K. Tomioka, S. Hara, J. Motohisa, E. Sano, and T. Fukui, Appl.
Phys. Express 3 025003 (2010).
[13] B. Tomas, L.-E. Wernersson, T. Lowgren, and L. Samuelson,
Nanotechnology 17 S227 (2006).
[14] P. D. Dapkus and J. J. Coleman, "Metalorganic Chemical Vapor Deposition,"
in III-V Semiconductor Materials and Devices, R. J. Malik, Ed., 1st ed
Amsterdam: North-Holland, 1989.
[15] G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, 2nd ed. San Diego:
Academic Press, 1999.
[16] J. P. Duchemin, M. Bonnet, F. Koelsch, and D. Huyghe, J. Cryst. Growth 45
181-186 (1978).
[17] D. H. Reep and S. K. Ghandhi, J. Electrochem. Soc. 130 675-680 (1983).
[18] L. Silvestre, A. Ougazzaden, D. Delprat, A. Ramdane, C. Daguet, and G.
Patriarche, J. Cryst. Growth 170 639-644 (1997).
[19] K. Takemasa, M. Kubota, and H. Wada, IEEE Photonics Technol. Lett. 12
471-473 (2000).
[20] K. Djordjev, S.-J. Choi, W.-J. Choi, S.-J. Choi, I. Kim, and P. D. Dapkus,
IEEE Photonics Technol. Lett. 14 603-605 (2002).
[21] T.-W. Lee, Y.-B. Moon, J.-H. Choi, D.-Y. Kim, and E. Yoon, J. Cryst.
Growth 182 299-308 (1997).
[22] O. Kayser, J. Cryst. Growth 107 989-998 (1991).
[23] D. G. Coronell and K. F. Jensen, J. Cryst. Growth 114 581-592 (1991).
20
[24] S.-J. Choi, "Modeling of Selective Area Growth of InP-Related Alloys by
Metal Organic Chemical Vapor Deposition," Ph.D. Ph.D. Dissertation, Dept.
Material Science, University of Southern California, Los Angeles, CA, 2004.
[25] K. A. Dick, S. Kodambaka, M. C. Reuter, K. Deppert, L. Samuelson, W.
Seifert, L. R. Wallenberg, and F. M. Ross, Nano Lett. 7 1817-1822 (2007).
[26] M. Akabori, J. Takeda, J. Motohisa, and T. Fukui, Nanotechnology 14 1071-
1074 (2003).
[27] M. Inari, J. Takeda, J. Motohisa, and T. Fukui, Physica E 21 620-624 (2004).
[28] K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Nano Lett. 8 3475-3480
(2008).
[29] J. Noborisaka, J. Motohisa, J. Takeda, M. Inari, Y. Miyoshi, N. Ooike, and T.
Fukui, "Growth of GaAs and InGaAs Nanowires by Utilizing Selective Area
MOVPE," in International Conference on Indium Phosphide and Related
Materials, Kagoshima, Japan, 2004, pp. 647-650.
[30] B. Hua, J. Motohisa, Y. Kobayashi, S. Hara, and T. Fukui, Nano Lett. 9 112-
116 (2008).
[31] H. M. Manasevit, Appl. Phys. Lett. 12 156-159 (1968).
21
Chapter 2: Electron Beam Lithography and Processes for
Nanoscale Selective Area MOCVD Templates
2.1 Introduction
In nanoscale selective area MOCVD (NS-SAG), the diameter of the
nanowires is determined by the diameter of the nano openings and degree of lateral
growth on the sidewalls. A uniform and reproducible lithography technique is
essential for NS-SAG. For the nano opening array fabrication, many different types
of nano lithography techniques including electron beam lithography (EBL) [1],
diblock copolymer lithography [2], anodic aluminum oxidation [3], and diffraction
lithography have been used. Among these, EBL has been used most frequently and
widely because of its high flexibility and uniformity. It is one of the most mature
nano lithography techniques and is capable of delivering superior reproducibility
compared to other techniques.
However, even with the EBL technique, fabricating uniform nano opening
arrays is always challenging and all of the process parameters must be controlled
precisely to get the ultimate resolution which is guaranteed in the specification sheet
from the manufacture of the EBL system. To maximize the resolution, appropriate
electron beam sensitive resists and EBL parameters such as acceleration voltage,
writing field size, working distance, electron dose, beam current, and writing speed
22
must be chosen. In this chapter, the basics of EBL systems are discussed. The
system description is mainly related to the Raith e-Line EBL system which has been
installed at USC, but the discussion is not limited to the specific system. The
lithography condition optimization for the Raith e-Line system for nano opening
array fabrication will also be discussed. This includes dose tests for electron dose
optimization, the feature size and uniformity with respect to acceleration voltage and
considerations for nano pattern transfer. This chapter also describes a bilayer
process for metal lift-off, which is necessary to the contact pad formation for electric
conduction measurements. Further discussions are available in ref. [1].
2.2 Background of Electron Beam Lithography
Basically, EBL is the process of writing patterns on an electron sensitive
layer by a focused electron beam. To generate and deflect a focused electron beam,
an EBL system has an electron column which is a complex of an electron source and
electron optics to focus and deflect the generated electron beam. EBL systems are
actually very similar to conventional scanning electron microscope (SEM) systems,
and it is even possible to modify a SEM system for EBL applications. As shown in
Figure 2-1, a typical EBL electron column consists of an electron source, electron
accelerator, electron lens system, beam blanker, deflection coils, and stigmation
control. All those parts are controlled by one or more external computer systems.
23
The column must be kept under high vacuum for electron generation and beam
formation.

Figure 2-1 Cross-section drawing of a typical electron beam column along with a
raytrace of the electrons as they pass through the various electron optical
components [1].
2.2.1 Electron Guns
The electron gun generates electrons. The oldest and most conventional type
of electron source is the thermionic electron gun. Lanthanum hexaboride (LaB
6
) and
tungsten filaments have been used as the thermionic electron source. The high
24
operating temperature (2700 K for tungsten and 1800 K for LaB
6
) in the vacuum
environment and large beam diameter make the system complex and degrade the
achievable resolution. A cold field emission gun (FEG) uses a very high electric
field at a sharp tip, typically less than 1 m, to extract electrons. A cold FEG can
deliver superior results because it generates a beam with a smaller diameter.
Because its emission is easily affected by contamination of the tip, it requires very
high vacuum (~10
-10
torr), and the long term stability of beam current does not meet
the requirements for EBL. Since the small beam diameter is very advantageous for
imaging, Cold FEG’s have been used for electron columns of high resolution SEM
systems. For example, the Hitachi S4800 field emission SEM which is used in this
study has this type of electron gun.
The thermal FEG is a hybrid of a thermionic gun and a FEG, and it is often
called a Schottky FEG. The tip is formed by a sharp tungsten tip coated with
zirconium oxide (ZrO) to reduce the work function and the extraction voltage. The
typical tip diameter is less than 0.5 m. For easy electron extraction and emission
stability, the tip is heated to 1800 K by supplying a filament current. Electrons are
extracted from the tip by applying a positive voltage on an extractor grid in front of
the tip. Figure 2-2 is the schematic diagram of thermal FEG. According to Figure 2-
2, we have two parameters for emission control. We can increase the emission by
increasing the filament current (I
FIL
) and extractor voltage (V
EXT
). In order to get
optimal performance, these two parameters need to be balanced. This balance must
be checked by measuring the extractor current (I
EXT
), which is the electron current
25
falling on the extractor grid. The I
EXT
is a strong function of the acceleration voltage
(V
ACC
), because the voltage between the tip and extractor grid can be affected by
V
EXT
as well as V
ACC
. If I
FIL
is too low and V
EXT
is too high, the difference of I
EXT

between the on and off state of V
ACC
is small because the hot spot on the tip is too
small or the extraction is almost governed by the large V
EXT
. If I
FIL
is too high and
V
EXT
is too low, the difference becomes large because the hot spot is too large or the
voltage between the tip and extractor grid is strongly affected by V
ACC
due to the
small V
EXT
. For the Raith e-Line system, the rule of thumb is to keep the difference
in between 10% and 30%. The change in the I
EXT
can also be an effective monitor
tool for the tip degradation. Throughout the tip lifespan, the I
EXT
increases, because
the emission angle becomes larger. Typical values for I
FIL
and V
EXT
is 2.430 A and
5.95 kV, respectfully. The typical value for I
EXT
are 240 A with 0 kV V
ACC
and 203
A with 10 kV V
ACC
. These parameters show small changes with respect to the
installed tip and the target beam current.
26

Figure 2-2 Schematic of thermal field emission gun.
2.2.2 Electron Optics
The other parts of column are used for electron acceleration, deflection,
blanking, beam current control, and focusing. The Raith e-Line system has a special
column whose design is optimized for low V
ACC
applications. Originally, the column
of the e-Line system was designed for the Carl Zeiss Supra 40 field emission SEM.
Carl Zeiss designed a special column, called Zemini, which has no crossover point of
the beam in order to maximize the beam current at low acceleration voltage. Figure
2-3 is the schematic of the Zemini column. The Zemini column uses its maximum
V
ACC
throughout the column and decelerates the electron beam at the end of the
objective lens to match the beam energy to the target value. This reduces the
27
aberration of the column by increasing directivity of electron beam for low V
ACC

settings.

Figure 2-3 Cross section schematic of Gemini column designed by Carl Zeiss, Ltd [4].
The beam deflector steers the electron beam to shine the beam to the target
location. The beam deflector is controlled by the scan controller in an SEM and the
pattern generator in an EBL system. The pattern generator converts a layout file to a
voltage waveform for the deflector to steer the beam to follow layout. The speed of
the pattern generator and the deflection mechanism governs the maximum beam
speed. If the writing speed is too fast, the beam deflection cannot follow the writing,
and distortion in pattern writing occurs. This phenomenon is called the dynamic
effect. To suppress the dynamic effect, the beam scanning speed must be limited.
This will be discussed further in section 2.3.
28
In the electron column, there are apertures to control the electron beam. A
blanking aperture is located right after the beam blanker coil and used to turn on the
beam on and off. It blocks the electron beam when the beam blanker coil deflects
the beam out of the column axis. By this, we can turn off the beam output without
turning off the electron gun. A beam limiting aperture, located after the zoom lens,
limits the beam current. In a conventional electron column, there are several limiting
apertures which determine the coarse range of the beam current. Within the coarse
range, the beam is tuned by a zoom lens and the beam current can be adjusted
continuously. A beam limiting aperture is normally set in an X-Y stage to allow it to
be centered, or aligned, with respect to the optical axis.
The Raith e-Line system has a turret with 6 apertures on a thin film menbrane.
One of the apertures is located at the column optical axis and the other 5 are located
around the perimeter of the center aperture as shown in Figure 2-4. The center one is
called standard aperture whose diameter is 30 m. The diameters of the other 5
apertures are 7.5 m, 10 m, 20 m, 60 m, and 120 m. Because the e-Line
system has no zoom lens in its column, we only have 6 beam current settings and
further tuning is impossible. Table 2-1 is the typical beam current value with respect
the aperture. This aperture turret also acts as a blanking aperture. In Figure 2-4, we
can see a section of the turret that has an empty blocked space. The beam is
deflected and lands at this location when the blanker is on.

29
Table 2-1 Beam currents corresponding aperture diameter of e-Line system.
Aperture ( m) Beam current (pA) Aperture ( m) Beam current (pA)
7.5 12.5 30 230
10 25.5 60 920
20 102.2 120 3682

Figure 2-4 Emission image of the aperture turret of Raith e-Line system. The bottom
empty space is for blocking the beam when the blanker is on.
To achieve the best column performance, the beam must pass through the
center of the aperture, so aperture alignment is a very important step in writing.
From the emission mode imaging (Figure 2-4), we can see the actual image of the
apertures and align them. This is coarse alignment done by a system superuser. The
alignment must be fine tuned by users whenever a new exposure is set up for the best
performance. To accomplish this, a small particle is imaged. Usually the particle
diameter is around 20~40 nm and the magnification is more than 100k. By wobbling
the focus, if the alignment is perfect, the image stays at the same place and blinks
30
regardless of wobbling; however, if the alignment is not good, the image shows drift
by wobbling. By adjusting DC bias of deflection coil, we can stabilize the image
with wobbling and get a perfect column alignment.
2.2.3 Pattern Generation
Conventional EBL systems for nano lithography, including the Raith e-Line
system, are Gaussian beam, vector scan, and moving beam and fixed stage (MBFS)
systems. A Gaussian beam EBL system uses focused electron beam whose electron
current density profile is Gaussian. The typical beam diameter ranges are of the
nanometer order. This focused Gaussian electron beam is scanned by the pattern
generator and deflector coil to reproduce user patterns. Figure 2-5 explains two
different types of scans. In vector scan, scanning occurs only inside of patterns and
the beam jumps from one shape to the next via a direct vector. In raster scan, the
entire space is scanned, but the beam is enabled only inside of the patterns.
31

(a) (b)
Figure 2-5 Two scanning strategies of EBL, (a) raster scan, and (b) vector scan.
Because the beam deflection angle is limited, we cannot usually cover all the
writing area without moving the stage. The term, write field or, simply, field,
describes the maximum writing area that can be achieved by beam deflection without
stage movement. The size of a write field is determined by the magnification of the
column. If the size of a layout is bigger than a write field, the layout is divided into
multiple write fields. The system writes the layout field-by-field and it moves the
stage between fields. This writing scheme is called MBFS, because the stage is
always stationary when the beam is on. In this writing scheme, errors caused by
stitching between fields are inevitable.
The write field is discretized by the digital to analog converters (D/A
converters) in the pattern generator. For example, 16 bit D/A converter can
32
discretize the write field into a 65536 (2
16
) × 65536 pixel array. The pixel is often
called an exel (electron beam pixel). If the size of the write field is 100 m × 100
m, the pitch between neighboring pixels is 1.52 nm (100 m/65536). This pixel
pitch is the ultimate limit of spatial resolution of the system. Figure 2-6 is a
schematic of composition of a write field in a conventional Gaussian beam EBL
system. Therefore, scanning happens pixel by pixel, not line by line. Dwell time is
the exposure time for each pixel in the field.

Figure 2-6 A write field is a pixel array. The number of bit of the D/A converter in
the pattern generator determines the size of the array. The pixel spacing is
determined by the D/A converter and the field size.
The size of the write field affects astigmatism and distortion in pattern
writing. Because the curvature of the focus plane is finite (Figure 2-7 (a)), if the
write field is too large, the focus at the edge of the field cannot stay in the same plain
and a large confusion ellipse appears at the edge of the write field. Figure 2-7 (c)
shows pincushion distortion that occurs in a large write field. Smaller write field
requires many stage movements to cover a large layout area. However, stage
33
movement is slow due to the settling time required for stage stabilization, and also
generates many stitching errors between write fields.

(a)

(b) (c)
Figure 2-7 Finite curvature of focal plane (a) makes astigmatism (b) at the edge of a
write field. The finite curvature also makes pincushion distortion (c) in
the write field.
2.2.4 Coordinate Systems
To write a pattern at a specific location, a coordinate system must be defined.
The Raith e-Line system has 5-independent coordinates to describe the location with
respect to the stage, sample, layout, and beam. The stage coordinate system (SCS) is
34
the absolute coordinate system of the stage, and user cannot redefine it. This
coordinate system is the standard coordinate system and all other coordinate systems
are defined in the relation to this coordination system. The calibration of the SCS
determines the overall alignment and placement accuracy of the system. The
calibration of the X and Y axis is done by a laser interferometer. The X-Y
movement of the stage is tuned by piezoelectric motors with 2 nm accuracy. The Z
axis has no high precision movement system, but the Z axis is never changed during
the writing, and it only determines the working distance (WD).
To describe locations on a sample, the e-Line system offers 2 independent
coordinate systems, the global coordinate system (GCS) and the local coordinate
system (LCS). To distinguish from the SCS, the horizontal and vertical axis of GCS
is expressed by U and V, respectively. The LCS is expressed by lower case, u and v.
These coordinate systems are typically known as the UV coordinate systems. Figure
2-8 illustrates the use of the two coordinate systems. The GCS is useful to describe
the entire sample space. If it is defined, it will not be changed unless the GCS is
redefined, and it is not affected by switching between the GCS and LCS. The LCS is
useful to describe the coordinate system of a local die or chip. A sample or wafer
can have many chips or dies. The LCS allows us to use same coordinate system
between different dies on one sample. If we only have a GCS, the coordinates of
each die must be recalculated in the GCS, which would be a very cumbersome
process. The LCS is not permanent, and if a different place in the GCS is selected,
the previous LCS setting is no longer valid. With two coordinate systems, step-and-
35
repeat exposures can be performed without complex calculations for coordinate
conversion.
#1. Define the GCS.
#2. Locate the origin of the first chip in the GCS.
#3. Define the LCS at that location.
#4. Write chip patterns in the LCS.
#5. Switch back to the GCS and locate the origin of the next chip.
#6. Repeat 3 to 5 until all of the chips are written.

Figure 2-8 Definition of gthe lobal coordinate system (GCS) and the local coordinate
system (LCS). The GCS describes coordinate of entire sample and the
LCS describes coordinate of localized chips.
There are two ways to define UV coordinate systems. One method is angle
detection. In this scheme, two different on the same horizontal axis are defined. The
horizontal wafer flat or cleaving edge of a sample can be used as the reference of this
36
two point selection. From this reference, the system calculates the angle of rotation
between UV and XY. The origin can be assigned anywhere, and the system will
calculate the shift between UV and XY referenced to the origin. This procedure is
illustrated in Figure 2-9. The other method is 3 point detection. Given 3 known
points in the UV system, the system can determine the corresponding XY
coordinates, because the system always has XY information whenever it allocates a
certain place. This method is quite useful if we have predefined marks, known as
alignment marks. Figure 2-10 illustrates this scheme.
The design coordinate system (DCS) is the coordinate system for layouts. If
the DCS is intentionally matched to the LCS or GCS, no additional coordinate
translation for the DCS is needed. Especially, in a step-and-repeat exposure, it is
very useful to match the LCS and DCS.

Figure 2-9 The procedure for UV coordinate definition by using angle detection.
1. Select 2 points, Label 1 and 2
which will be on same U axis
2. The system calculates the
angle  and set V axis
perpendicular to the defined
U axis.
3. Set the origin at desired point.
4. The system assumes zoom
factor is 1, which means
distance between two
arbitrary point is same in the
two coordinates.
37
X
Y
(U,V)=(0,0)
(U,V)=(10,0)
(U,V)=(0,10)

Figure 2-10 The procedure for UV coordinate definition by using 3 point mark
detection.
The beam coordinate system (BCS) is the coordinate system for beam
deflection. Because the electron column and stage movement system are completely
independent, the BCS should be related to the SCS (or UV coordinates such as the
GCS or LCS) in order to write a pattern at a desired location. In this case, the
coordinate numbers in the BCS do net need to be explicitly listed. If the relation
between the BCS and SCS is correct, the beam can be located without numbers in
the BCS. The procedure relating the BCS to the SCS is known as write field
calibration. In this calibration, the relation between two coordinates is expressed by
zoom factors, shifts and rotations. The zoom factor is the scaling factor between the
SCS and BCS. If the factor is 1, the measured distance between two arbitrary points
in the SCS or BCS should be the same regardless of the coordinate system used to
measure the distance. Shift is the lateral displacement between the SCS and BCS.
Rotation is the angle between the axes of the two coordinate systems. If the beam
5. Assign coordinates to 3
predefined marks whose
relative position is already
known.
6. According to the assigned
coordinate, the system
calculates the relation
between XY and UV (Those
3 marks do not need to be
on same U or V axis. The
system automatically
calculate the axis)
38
coordinate system is distorted and the horizontal and vertical axes are not
perpendicular to each other, there will be 2 rotation values. Figure 2-11 shows the
definition of the 6 parameters. The location of the beam can be determined from
SEM images. The center of the image is the center of the write field in the beam
coordinate system. From this, the calibration procedure can be made. If there are no
predefined marks, a small particle on a sample can be used for calibration. This
procedure must be carried out manually and it is known as manual write field
calibration. The procedure is described in Figure 2-12.
#1. Locate the particle at the center of the image.
#2. Move the stage away in the SCS by a predefined distance.
#3. Deflect the beam with the BCS to locate the particle at the center
of the image.
#4. Move the stage to another place and measure the error vector
#5. Measure the error vectors 3 times and calculate the correction
parameters.
If there are predefined marks, whose location in UV or the SCS is known, by
measuring their location in the BCS, the error vectors can be determined. In this
scheme, the location of the predefined mark can be found by edge detection of a
single line scan. The e-Line system software offers automated write field calibration
mark detection and calibration, known as automatic write field calibration, which is
described in Figure 2-13.

39

(a) (b)
X
Y
X
B
Y
B
Rotation X
Rotation Y

(c)
Figure 2-11 Parameters for BCS definition, (a) zoom factors, (b) shifts and (c)
rotations. X and Y are the axes of SCS. X
B
and Y
B
are the axes of the
BCS.

40

(a) (b) (c)

(d) (e)

(f) (g)
Figure 2-12 The procedure for manual write field calibration. The calibration starts
from (a) when the particle image is at the center of the image. (b) The
stage moves in a direction in the SCS. The stage displacement is set by
the user. Then the system makes same displacement in the BCS by
deflecting the beam. If the SCS and BCS have a mismatch, the image of
the particle is not at the center. An error vector (red arrow) is created. (c),
(d) repeat this in two different displacements. (e) With 3 measured error
vectors, the system calculates the BCS in terms of the SCS. (f) A real
particle image after the displacement. Because of the mismatch, the
particle center is not on the green cross. (g) Users locate the center of the
particle (blue cross). The system calculates the error vector by mismatch
between green and blue cross.
41

(a)

(b)
Figure 2-13 The illustration of automatic write field calibration. (a) The system
deflects the beam toward the center of the predefined marks and locates
the center of the marks by line scan (red lines). From the line scan in
horizontal and vertical direction, the system figures out the center of the
mark and creates error vectors to calculate the BCS. (b) The line scan
result of a 1.5 m deep etch pit mark. The system detects the edge of the
mark by detecting falling or rising edge in the scan profile. The graph
shows falling edge detection. Two green horizontal lines are upper and
lower threshold of edge detection. If the profile crosses the threshold from
upper to lower, it detects the falling edge. By detecting falling edge in two
sides, it tries to find the edge from left to right in the left half of the scan
profile and from right to left in the right half of the profile. From this two
edge information, the system figure out the location of the center (yellow
line).
42
2.2.5 Resists
The resist for EBL is not a part of the EBL system, but it is an equally
important consideration to achieve maximum resolution. As with photoresists, there
are two types of EBL resists, positive and negative tone resists. In positive resists,
the electron beam cuts the polymers or activated acid contents in the exposed region
and after development, the exposed area is removed. In negative resists, the
incidence electron bean cross-links the polymer and increases resistivity against
developers. So, after the development only the exposed region remains.
Poly-meta-methyl-acrylate (PMMA) is the most famous and widely used
positive tone EBL resist. It is inexpensive and capable of 10 nm resolution [1].
Typically, PMMA is available in two molecular weight forms, 495 k or 950k.
Heavier molecular weight has lower sensitivity, because more electrons are required
to break down the material into smaller fragments of methyl acrylate. PMMA is sold
in two different casting solvents, chlorobenzene and anisole. By adjusting
concentration, the thickness of PMMA film after spin coating may be adjusted. The
adhesion of PMMA is quite good on most of semiconductor material surfaces and
dielectrics, such as silicon dioxide (SiO
2
) or silicon nitride (SiN
X
), so adhesion
promoter is not essential when PMMA is coated on those surfaces. After the spin
coating, PMMA must be baked on a hot plate or in a convection oven. The baking
temperature typically ranges from 170 C to 200 C. On a hot plate, the time ranges
from 1 minute to 1 hour or even longer. In a convection oven, the baking takes 30
43
minutes to several hours. Within these temperature and time ranges, the exposure
sensitivity does not change dramatically, but longer. Baking makes the film harder
and make it adhere to the surface more strongly.
Methyl iso-butyl ketone (MIBK) is a widely used developer for PMMA.
Concentrated MIBK is too strong and can dissolve unexposed PMMA, so the
developer is usually diluted by mixing it with isopropyl alcohol (IPA). A mixture of
1 part MIBK to 3 parts IPA produces very high contrast and abrupt transitions[1]
between exposed and unexposed areas, but has low sensitivity, in other words, the
dose for clearing the resist becomes high. By making the developer more
concentrated, say, 1:1 MIBK:IPA, the sensitivity is improved significantly with the
tradeoff of reduced contrast.
When exposed to more than 10 times the optimal positive dose, PMMA will
again crosslink, forming a negative resist. This effect can be plainly seen after
exposing one spot for an extended time (for instance, when focusing on a mark). The
center of the spot will become crosslinked, leaving resist on the substrate, while the
surrounding area is exposed positively and is washed away during development.
PMMA is one of an excellent electron sensitive resist, but it has poor etching
selectivity in plasma etching conditions. PMMA is especially weak against O
2

plasma and in other plasma etching, such as CF
4
or BCl
3
etching, it also attacked
easily. For example, in the standard CF
4
SiN
X
etching condition of the reactive ion
etching (RIE) system in the USC CPT cleanroom, which is 100 W of RF power, and
100 mT of process pressure, the etching selectivity of PMMA 950k is about 0.7:1.
44
This poor selectivity requires thicker resist films for deep plasma etching, which then
degrades the ultimate resolution because of its thickness. For this research, PMMA
is still a suitable resist, because the dielectric mask layer is thin.
ZEP 520A is a positive EBL resist made by the Nippon Zeon Company of
Japan. It delivers very high resolution comparable to PMMA. Its contrast is better
and is far more sensitive than PMMA. With the same thickness, the sensitivity of
ZEP 520A is almost 10 times higher than that of PMMA. This indicates we can
reduce the writing time dramatically, but at the same time, we need far more accurate
electron dose control to achieve the best results. One of the problems with ZEP
520A is its limited shelf-life of 1 year and types of failures, where expired resist
often develops cracks in the developed patterns as shown in Figure 2-14. This resist
also has high etching resistance against plasmas, comparable to generic photoresists
[5]. If fresh ZEP 520A is available, it is a promising candidate for large scale SAG
pattern generation.

Figure 2-14 Expired ZEP 520A cracks after development. The SEM micrograph
shows that the cracks mainly start from the corner of rectangles.
45
Negative resists are not as useful for NS-SAG, because nano opening arrays
are needed. However, they may be used for making macroscopic dielectric pads or
shaped dielectric islands where the nano opening arrays are located. UVN 30 is a
negative tone EBL resist made by Shipley (now Rohm and Haas). From its name,
UVN 30 was originally designed as a negative tone deep UV resist. The sensitivity
is comparable to ZEP 520A and it also delivers high plasma etching selectivity.
However, fluorination during etching in long fluorine based plasma etching turns the
film into a Teflon-like film which is very hard to remove even with strong O
2
plasma
ashing. So, users must pay great attention when they use a UVN 30 layer as an etch
mask in fluorine based plasmas.
For further discussions of different types of EBL resists, refer to ref. [1].
2.3 Optimization of Electron Beam Lithography
Many parameters can affect the EBL writing. Some parameters such as tip
parameters, (I
FIL
, V
EXT
, and I
EXT
) are not adjustable by general users. Focus,
stigmation control, aperture alignment and coordinate calibration are user-adjustable,
but they are not inter-related. There is only one optimal point for each parameter.
V
ACC
, beam current, working distance, step size, and electron dose are user-
adjustable and they are dependent on each other. There are trade-offs in these
settings and is no single optimal point for all the parameters. The best combination
of the parameter values must be found whenever a new layout or pattern is tried.
46
This section describes how to find the optimal combination of parameters for nano
opening arrays.
2.3.1 Acceleration Voltage
Acceleration voltage (V
ACC
), often referred as high tension voltage or beam
energy (if this parameter is called beam energy, the unit becomes keV) plays a
central role in EBL. It affects almost all parts of EBL, such as effective dose, spot
size of the beam, beam current, the degree of charging effect, the electron beam
induced damage, imaging performance, forward and backward scattering and so on.
Among those, the scattering must be considered first, especially for dense opening
array exposures. Electron scattering governs the ultimate resolution of each nano
opening, the highest packing density, and effective dose setting.
The electron beam shined on the resist surface penetrates the resist and
creates both mechanical and chemical interaction. Within the resist, the incident
beam is scattered by the interaction. Forward scattering is the scattering in the resist
layer and the scattering direction is the same as the incident beam. The size of
confusion circle created by this type of scattering is a strong function of resist type,
resist thickness, and V
ACC
. If V
ACC
is high, the confusion circle caused by scattering
becomes smaller, because the interaction cannot change the momentum of electron,
if electrons have high kinetic energy and momentum.
47
Backward scattering occurs in high V
ACC
. High energy electrons which pass
through the resist layer penetrate into the substrate, and some of the electrons
undergo scattering events and are reflected back to the surface. This effect is known
as backward scattering. In this scattering, the direction of scattering is opposite to
the incident direction, which is why it is known as ‘backward’. Backward scattering
is a strong function of substrate material and, of course, V
ACC
. If we V
ACC
is
increased, more electrons penetrate into the substrate and more electrons reflect back.
Figure 2-15 illustrates a Monte Carlo simulation of electron beam scattering with
respect to V
ACC
. From the picture, the relation between forward scattering and
backward scattering can clearly be seen. In high V
ACC
, backward scattering occurs
more and forward scattering occurs less and vice versa for low V
ACC
conditions.

Figure 2-15 Monte Carlo simulation of electron scattering in resist on a silicon
substrate at (a) 10 kV and (b) 20 kV [6].
48
Figure 2-16 is a nano opening array pattern written by the e-Line system in
different V
ACC
setting ranges from 2.5 kV to 25 kV. In Figure 2-16, the fidelity of
the pattern increases as V
ACC
increases. This indicates that forward scattering
dominates nano pattern writing. The confusion circle of forward scattering is smaller
than that of backward scattering, but, in the confusion circle, the electron dose is
much higher than that of backward scattering. Because the pitch of nano opening
array ranges few hundred nanometers, the definition of the opening and gap is totally
governed by forward scattering. Therefore, for smaller openings and higher packing
density, writing must be done in higher V
ACC
setting. This also reflects the relation
between V
ACC
and the beam diameter. Even though the column of the e-Line system
is optimized for low V
ACC
application, the minimum beam diameter is only available
from 25 kV upward.
If V
ACC
increases, the effective dose decreases. The effective dose refers to
the quantity of incident electrons which lose their energy in the resist layer. This
indicates, in high V
ACC
, the most of incident electrons do not stay in the resist layer
and instead lose their energy in the substrate. For 100 nm thick PMMA, 2.5 kV is
enough to expose the entire film [7]. If the effective dose decreases, a higher
electron dose setting is required, which elongates the writing time. From Figure 2-16,
a clearing dose in 25 kV V
ACC
is 280 % higher than that of 10 kV V
ACC
setting, which
is a problem when writing large sized arrays. Writing an opening array at 25 kV
takes 3.8 times longer than writing at 10 kV.

49

(a) (b)

(c)
Figure 2-16 SEM micrographs of 60 nm center-to-center spacing, rectangular nano
opening arrays. The arrays have several defects which indicate the
electron dose condition is very close to the clearing dose. The VACC and
electron dose of the writings are (a) 2.5 kV, 0.3 fC per opening, (b) 10 kV,
0.5 fC and (c) 25 kV, 1.4 fC, respectively.
High V
ACC
causes other problems. In high V
ACC
setting, it becomes more
difficult to get high fidelity surface morphology images, because electrons penetrate
relatively deeply into the substrate. This causes focusing problems. In high V
ACC
,
very small particles, especially when made of a dielectric material, cannot be seen
clearly by SEM imaging. In high V
ACC
, the center of SEM images shows less
contrast and brightness (shown in Figure 2-17), because of the characteristics of the
50
in-lens detector. The detector should not block the incident electron beam path, so it
is shifted from the optic axis and located at the perimeter of the optic axis. The e-
Line system has an aperture turret, so the in-lens detector must be further away from
the optic axis of the column compared to other systems, which causes collection
efficiency problem at the center of the image. This effect becomes worse in higher
V
ACC
, because high V
ACC
reduces the scattering angle and drops the collection
efficiency more.

(a) (b)
Figure 2-17 SEM micrographs of the Faraday cup on universal sample holder of
Raith e-Line system. The micrographs are taken with (a) 7.5 m and (b)
30 m limiting apertures.
Regarding these, in this research, the standard V
ACC
is set at 10 kV, if the
center-to-center spacing of the nano opening array is farther than 200 nm. At 10 kV,
25 nm opening arrays with 200 nm center-to-center spacing on 100 nm thick 950k
PMMA can be formed. If the layer thickness is reduced to 50 nm, the minimum
diameter achieved is 15 nm. For dense array whose center-to-center spacing is
shorter than 200 nm, the patterns are written at 20 kV. At 20 kV, no noticeable
51
resolution improvement in large center-to-center spacing arrays is observed, but the
definition of the gap between openings when the center-to-center spacing reaches 80
nm improves. 60 nm center-to-center spacing is possible at 25 kV, but is
irreproducible and too sensitive to the process condition variations.
2.3.2 Resist Material and Process
In this research, PMMA 950k for the nano patterning was chosen. As was
shown in section 2.2, PMMA is inexpensive, readily available, and suitable for high
resolution lithography. 2% casting solution in chlorobenzene was chosen. By 6
kRPM spin coating, the thickness of PMMA layer reaches 100 nm (the spin curve is
Figure 2-18), which is thick enough to be a dry etch mask for a 30 nm SiN
X
layer.
After the spin coating, coated samples are baked on a hot plate at 180 C. The soft
bake time is 90 seconds.

Figure 2-18 Spin curve of 950k PMMA casted in Chlorobenzene [8].
52
2.3.3 Electron Dose
The electron dose is one of the most important parameters. For a positive
EBL resist, certain amount of dose to clear an exposed region without residual layer
is needed. The threshold dose of a positive resist indicates the minimum electron
dose required to clear the exposed region after development without leaving a
residual layer. Though the resist manufacture gives us the number for the threshold
dose, it must be recalibrated as the dose is dependent on the resist thickness, baking
condition, V
ACC
, and shape and density of the pattern being written.

d
S
d
S

(a) (b) (c)
Figure 2-19 3 different types of writing. (a) area writing, (b) single pixel line writing,
and (c) single pixel dot writing. The red dots are exposed pixels.
There are 3 different types of dose: area dose, line dose and dot dose. The e-
Line system offers another type of dose known as curved shape dose, but this can be
considered as a sort of area dose. Area dose setting is used to write polygons or
circles, which have finite area as shown in Figure 2-19 (a). If the beam current (I
B
),
area dwell time (t
DA
) which is exposure time of each pixel, and step size (d
S
) which is
53
distance between pixels are known, the area dose (D
A
) is expressed by Equation (2-1).
The unit of area dose is C/cm
2
.

2
s
DA B
A
d
t I
D  (2-1)
Line dose is dose setting for single pixel lines. A single pixel line is
constructed by single row or column of pixels. Line dose (D
L
) is also calculated by
I
B
, line dwell time (t
DL
) and d
S
. The unit of line dose is C/cm

S
DL B
L
d
t I
D  (2-2)
Dot dose is dose setting for single pixel dots. Because a single pixel dot is
one shot exposure of the electron beam, the electron dose of the single pixel dot is
simply I
B
multiplied by dot dwell time (t
DD
). The unit of dot dose is pC.

DD B D
t I D  (2-3)
The e-Line system offers a parameter called dose factor to adjust dose for
each object in a layout. Each object in a layout has its dose factor. The dose factor
is actually multiplier of a dose setting and its default value is 1. Before an exposure
is started, the dose condition must be set. In dose setting, the dwell time is calculated
by Equation (2-1), (2-2) or (2-3), with desired dose level and d
S
, and measured I
B
.
This setting is called the standard dose setting and becomes the default dose setting
afterward; however, the actual dose of objects in a layout is determined by standard
dose times the dose factor. For example, if an object has dose factor 2, the object
will be exposed with twice as large of a dose as the standard dose. We have another
54
dose factor for entire layout, which is layout dose factor, which affects the entire
exposure. The layout dose factor is useful to change the dose, exposure-by-exposure,
in batch writing. The actual dose of an object becomes.
dose = standard dose × (object) dose factor × layout dose factor (2-4)
The nano opening arrays are a collection of single pixel dots. It is
convenient to discuss the electron dose with the dot dose. Figure 2-20 is SEM
micrographs of nano opening arrays whose center-to-center spacing is 200 nm with
various dot doses. The diameter of openings becomes larger with respect to the dot
dose. Figure 2-21 shows the opening diameter is proportional to the dose. By
changing the dose, the diameter of openings can be tuned from 22 nm to 125 nm. To
get smoother edges, circular objects are better than very high dose single pixel dots
for openings larger than 125 nm in diameter. Also, if the dose is too high, the PMMA
at the center of opening turns into negative tone polymer due to crosslinking by
excessive electron dose.

(a) (b) (c)
Figure 2-20 200 nm center-to-center spacing rectangular nano opening arrays written
by e-Line system. All arrays are written at 10 kV. The smallest, 7.5 m
aperture is used, which delivers 7.0 pA beam current. The micrograph
shows surface morphology after the nano patterns are transferred on 30
nm thick SiNX layer by plasma enhanced reactive ion etch. Three arrays
are written by (a) 1.0 fC, (b) 2.4 fC and (c) 5.2 fC electron dose.
55
0 1 23 4 5 67 8 9
0
25
50
75
100
125

Opening diameter (nm)
dot dose (fC)

Figure 2-21 Opening diameter of nano opening in arrays with respect to the dot dose.
This result comes from 200 nm center-to-center spacing rectangular nano
opening arrays written by e-Line system. All arrays are written at 10 kV.
The smallest, 7.5 m aperture is used, which delivers 7.0 pA beam
current.
2.3.4 Consideration in Layouts for Nano Opening Arrays
The layout of a nano opening arrays may be just a simple collection of single
pixel dots in a matrix form. This construction is the most intuitive and direct way to
construct the nano opening array. However, if we construct the opening array by
single pixel dot array, several problems may arise. First, the layout becomes very
large. A single pixel dot is an object like a polygon or circle. So, the computer and
the pattern generator must index every dot in the layout. Also, all single pixel dots
have dose factor information. If a 1 mm × 1 mm size opening array with 200 nm
center-to-center spacing is chosen, the array is 5000 × 5000, which contains
56
25,000,000 dots, or in other terms, objects. Indexing this large of a number of
objects often creates a buffer overflow error in the control program, which will then
causes an abnormal termination of writing.
When writing a single pixel dot array, part of the writing time is wasted,
because each dot is treated as an object. When jumping between dots, the system
follows the same sequence as when it makes jumps between two large polygons.
Before the jump, the system enables the blanker to turn off the beam and then moves
the stage or deflects the beam to the desired location. It then spends some time
waiting before unblinking the beam and starting to write the next object. This time is
known as the settling time, the time to stabilize deflector to avoid location error
induced by the slow transient response of the deflector. This sequence always occurs
in between the exposure of each dot. The spacing between openings is usually on
the sub micron scale and the exposure caused by sweeping between dots can be
negligible. This is why D
A
can be calculated by Equation (2-1). If jumping between
pixels causes significant exposure, it should be included in the calculation of D
A
.
Instead, an opening array is created with a simple rectangular box, and, in the
exposure setting, the pixel step size is set to the desired spacing between openings.
t
DD
is calculated for the required D
D
and this number is used for t
DA
. By this method,
an opening array is written with desired spacing and dose by exposing a single object.
In this case, regardless of the size of the array, the array is treated as an object. This
saves tremendous computation power for indexing and job file generation. However,
when opening arrays with rectangular objects are written, the beam speed must be
57
considered to prevent unwanted distortion of array writing. The beam speed
indicates the speed of scanning to make an exposure of a scan line. If the time for
sweeping between pixels is negligible, the beam speed, v is

DA
S
t
d
v  (2-5)
By combining Equation (2-1) and (2-5), v can be expressed using I
B
, d
S
and
D
A
.

A S
B
D d
I
v  (2-6)
If the beam speed is too fast, the deflector cannot follow the voltage input
from the pattern generator and the actual beam position lags. This is known as the
dynamic effect. As shown in Figure 2-22, the dynamic effect causes overexposure at
the beginning of the writing due to the slow turning on of the deflection coil and
lagging in writing. Both effects distort the pattern writing. The e-Line system has a
dynamic effect compensation algorithm. Figure 2-23 illustrates the algorithm. To
remove the overexposure and strong distortion at the beginning of the writing, the
pattern generator starts scanning away from the desired start point. During this
prescan, the deflection coil turns on completely and starts scanning in full speed.
Because still the sweeping of D/A converter is faster than deflection speed, the
starting point in D/A converter sweeping will be passed and wait for t
delay
to locate
the beam at the starting point. After the beam reaches the desired starting point, the
beam is unblanked and the writing starts. In fast writing regime whose scanning
58
speed is faster than that of deflection coil, the D/A converter sweeping reaches the
end point of scanning before the beam arrives that point. To continue the writing to
the end point, the D/A converter must go further to keep deflection. This postscan
must be continued until the beam reaches the end point. t
post
is the needed time for
the beam deflection to arrive the end point after D/A converter sweeping pass the
end point. However, this compensation is not perfect and cannot remove the effect
completely. The rule of thumb is to restrict the beam speed slower than 10 mm/s
even with the compensation. When writing faster than 10 mm/s, the system cannot
guarantee distortion-free operation.

Figure 2-22 D/A converter output and the beam position with respect to time. The
beam position is smoothed out for simplicity. If the maximum beam
deflection speed is slower than that of the D/A converter sweeping. It
creates dynamic effect and distorts the writing.

59
Time
Position
Output of D/A converter
Actual beam position
Prescan
Postscan
Target
writing
t
pre
t
post
t
delay
t
wr
Beam on

Figure 2-23 The mechanism of dynamic effect compensation in e-Line system.
Prescan and postscan compensate lagging of beam deflection.
2.3.5 Working Distance
The working distance (WD) is equivalent to the effective focal length in
optics. In a complex lens optics system, the effective focal length is the distance
between the 2
nd
principal point of the optic system and focus point. This length is
variable in variable focal length systems, or in other words, zoom lenses. The e-Line
system is also a zoom lens system and its WD ranges from 2.8 mm to 30 mm. The
focus point must be put on the surface of the samples. The WD is actually
determined by the vertical location of the sample, Z height in the SCS. In the system,
60
Z ranges from 0 mm to 30 mm. At 29.3 mm in Z, focus is achieved by setting a
7mm WD. So, useful range of WD is from 7 mm to 37 mm. In principle, shorter
WD improves the resolution of writing. At the same time, the depth of focus (DoF)
becomes shallower. DoF is vertical depth enclosing the exact focus point, in which
the electron beam stays in-focus. Deeper DoF is desirable to overcome the height
difference of the sample surface. In this work, 10 mm WD setting was chosen,
which delivers comparable resolution to 7 mm but gives a deeper DoF.
2.3.6 Beam Current
According to section 2.2, the beam current can be selected by choosing the
size of the beam limiting aperture. The size of the limiting aperture is able to affect
the beam current as well as the resolution, imaging quality and DoF. First, the
smaller aperture delivers better resolution. A smaller aperture blocks the outer edge
of the electron beam and utilizes the well collimated core of the beam. This limited
beam is controlled by the very center of the optic axis which also shows better
performance than the edge. This increases the resolution of writing. At the same
time, a smaller aperture makes the confusion circle smaller as in generic optic
systems, which increases the DoF. A smaller aperture makes the absolute value of
the beam current smaller, which decreases the signal to noise ratio of SEM imaging.
This makes focusing and stigmation control harder. A smaller beam current means
longer writing time. In this work, the smallest, 7.5 m aperture, was used, because
61
the maximum resolution of the system for nano openings was needed. The nominal
current value of 7.5 m aperture is 7~8 pA
2.3.7 Write Field Size
As discussed in section 2.2, the size of write field affects the writing time,
stitching and pincushion distortion. The best solution is the smallest write field
which is capable of enclosing a single array to avoid array disorder by stitching. In
the standard template used in this work, which will be introduced in section 2.4, the
biggest opening array is 100 m × 100 m, so the write field size of the standard
template is 100 m × 100 m. For large scale opening arrays such as 2 mm × 2 mm
array, a 1 mm × 1mm write field was used. Up to a 1 mm × 1mm write field, no
noticeable pincushion distortion was observed.
2.4 Processes for NS-SAG Template
Before NS-SAG, nano opening array mask must be fabricated by a dielectric
film. This requires the initial EBL and then a pattern transfer step. The simplest
way to prepare templates is making nano opening array on the substrates covered by
dielectric film, but the dielectric film must be tailored to a in certain size or shape.
Then marks for lithography steps, such as photolithography or EBL, are needed.
62
This section discusses the layout of the sample and process flow for template
preparation.
2.4.1 Template Layouts
In this work, two different templates are used. One of them is shown in
Figure 2-24 (a), named the standard template. This template has 14 by 14 array of
SiN
X
pads. Those pads are defined by photolithography and RIE etching. 50 m ×
50 m, 80 m × 80 m, 120 m × 120 m, and 200 m × 200 m pad array
photomasks were prepared. Among them, a 120 m × 120 m pad array is the most
frequently used one. Two types of shaped pad arrays have also been used. As
shown Figure 2-24 (b), those pads have macroscopic opening arrays in the perimeter
of 50 m × 50 m SiN
X
pads. The openings are 1 m × 1 m. One of them has a
macroscopic opening array with 3 m spacing and the other has 4 m spacing. The
size of a template is 8 mm × 8 mm.

63

(a) (b)
Figure 2-24 The layout of the standard template. (a) the templates has 14 by 14 SiNX
pad array (cyan boxes), mix-and-match alignment marks (Green cross is
defined by EBL. Pink inverted cross is defined by photolithography),
EBL alignment marks (orange cross), and 100 m write field calibration
marks (dark yellow small cross). The red box shows details of an EBL
alignment mark and 4 write field calibration marks. (b) Illustration of
shaped pads. A 50 m × 50 m SiN
X
pad is enclosed by macroscopic
opening array.
The layout has several marks for mix-and-match lithography and automatic
write field calibration on layout. As discussed in section 2.2, the size of write field
affects writing time, stitching and pincushion distortion. The best solution is the
smallest write field which is capable of enclosing a single array to avoid array
disorder by stitching. In the standard samples, which are introduced in section 2.4,
the biggest opening array is 100 m × 100 m. The write field size of the standard
samples is 100 m × 100 m. For large scale opening arrays such as 2 mm × 2 mm
array, a 1 mm × 1mm write field is used. For up to a 1 mm × 1mm write field, no
noticeable pincushion distortion is observed. The templates only have automatic
64
write field calibration for the 100 m × 100 m write field. Manual write field
calibration is used for larger size write fields.
2.4.2 Process Flow of Template Fabrication
Two different template fabrication processes are developed. One is known as
the two-step deposition process, because there are two SiN
X
deposition processes.
The other one is a one-step process that has one SiN
X
deposition processes. The
two-step process is a mix-and-match process, designed to reduce writing time of
EBL by defining macro patterns using conventional photolithography techniques.
This process consists of 3 steps, alignment mark formation, pad generation, and EBL.
In alignment mark formation step, the alignment marks and write field calibration
marks are defined. The most important consideration is that they must be visible
through the imaging capability of the e-Line system. Metal marks are the best
alignment and calibration marks because of their superior reflectivity. However, the
samples will be loaded in the MOCVD chamber nanowires be grown on them, so
metal marks might cause unintentional metal doping, which must be avoided in NS-
SAG. The second best solution uses etch-pit marks. In this case, the etch depth and
abruptness of the sidewall are important. Because the automatic mark detection of
the e-Line system locates the marks by detecting the rising or falling edges of mark
images, blurred edge images by sidewall slope make large detection errors. If the
etch depth is shallow, the contrast of mark image is not good enough for mark
65
detection, especially in high V
ACC
setting, in which the penetration depth of electrons
is deep. With the etch condition of the inductively coupled plasma (ICP) etcher in
table 3, the etch depth is approximately 1.5 m, which can be seen clearly in the e-
Line system.
PMMA stripping is one of the most important steps in the process. If the
cleaning is not perfect, the growth is ruined and the growth chamber is contaminated.
An O
2
plasma ashing step at the end of the process is for this reason, but this also
oxidizes the growth surface and may alter the growth behavior. If this O
2
ashing step
is skipped, no PMMA residue should be visible. The effect of O
2
plasma treatment
is discussed in chapter 3. Table 3 contains detail of the template preparation process.
Throughout this process, 4 templates per samples are created because the photomask
has the templates as 2×2 arrays.

66
Table 2-2 Flow of two-step deposition process.
Job
No.
Job Equipment Condition Check
point
A. Alignment mark definition
A01 SiN
X
deposition PECVD SiH
4
:NH
3
:N
2
=40:20:60
275 C/30W/440mT/20min
Yellow-
orange color
A02 PMMA Coating Spinner/
Hotplate
PMMA 495k C5
Slow/main spin coating
0.25/3krpm, 3/60s
Clean the backside
Baking on the hotplate
180 C/90s
Magenta
color after
baking
A03 EBL e-Line V
ACC
=10kV
WD=10mm
Aperture=30 m
Area dose=180 C/cm2
Gold colloids for focusing
Pattern: Alignment_mark

A04 Development MIBK:IPA=1:3 60s
Rince in IPA for 30s
Dry N
2
blowing

A05 Check Optical
microscope
Check mark
definition
A06 SiN
X
etch
RIE
Old RIE CF
4
/100W/100mT/2.5min Use
slow pump
A07 Stripping Acetone/MeOH/DI bath
5 mins each.
Dry N2 blowing follows
Acetone bath must be with
ultrasonic agitation

A08 Check Optical
microscope
Check
residue
A09 O
2
ashing Old O
2

asher
Sufficiently wait for the base
pressure
120W/150mT/5min

A10 InP/GaAs etch
ICP
STS ICP BCl
3
MQW recipe
165 C/2.5 min
Green color
after etching
B. SiNX pad generation
B01 SiN
X
etch
Wet
Dipping in HF 5min
Rinsing in flowing DI 5min
Dry N2 blowing
Check
complete
removal
B02 SiN
X
deposition PECVD SiH
4
:NH
3
:N
2
=40:20:60
275 C/30W/440mT/2.5min

67
Table 2-2 Continued
Job
No.
Job Equipment Condition Check
point
B03 PR coating Spinner/
Hotplate
AP405 adhesion promoter
Slow/main spin coating
0.6/5krpm, 3/30s
Shipley S1813
Slow/main spin coating
0.6/5krpm, 3/30s
Clean the backside and edge
bead
Baking on the hotplate
110 C/60s

B04 Photolithography MJB-3 Use 405 nm probe for cal.
150 mJ/cm
2

B05 Development MF321 (diluted TMAH)
30+s (dipping until the pad
definition is clear)

B06 Check Optical
microscope
Check pad
definition
B07 SiN
X
etch
RIE
Old RIE CF
4
/100W/100mT/0.12min Use
slow pump
B08 Stripping Acetone/MeOH/DI bath
5 mins each.
Dry N2 blowing follows
Acetone bath must be with
ultrasonic agitation

B09 Check Optical
microscope
Check
residue
B10 O
2
ashing Old O
2

asher
Sufficiently wait for the base
pressure
120W/150mT/5min

C. Nano lithography
C01 Cleaving Scriber
C02 PMMA Coating Spinner/
Hotplate
PMMA 950k C2
Slow/main spin coating
0.25/6krpm, 3/45s
Clean the backside
Baking on the hotplate
180 C/90s
Magenta
color after
baking
C03 EBL e-Line V
ACC
=10kV
WD=10mm
Aperture=7.5 m
Dot dose=1.0 fC
Pattern: QD_PTN_200_500

68
Table 2-2 Continued
Job
No.
Job Equipment Condition Check
point
C04 Development MIBK:IPA=1:3 60s
Rince in IPA for 30s
Dry N
2
blowing

C05 Check Optical
microscope
Check mark
definition
C06 SiN
X
etch
RIE
Old RIE CF
4
/100W/100mT/0.12 min
(Just etch : 0.08 min)
Use
slow pump
C07 Stripping Acetone/MeOH/DI bath
5 mins each.
Dry N
2
blowing follows
Acetone bath must be with
ultrasonic agitation

C08 Check Optical
microscope
Check
residue
C09 O
2
ashing Old O
2

asher
Sufficiently wait for the base
pressure
120W/150mT/5min

Two-step deposition process can reduce EBL writing time, but in this process,
the growth surface is exposed in plasma two times for the two different SiN
X
layer
depositions. The growth surface is also exposed in HF solution for removal of SiN
X

hard mask of alignment mark etch. These processes can alter the surface conditions
which are very critical to NS-SAG. To minimize this effect, the SiN
X
pads and nano
pattern should be written at the same time, which is known as the one-step deposition
process. Because PMMA is a positive tone resist, the area outside of the pads must
be written to make then SiN
X
pads. This is very time consuming process because of
its large writing area. The writing time of the pads is approximately 32 hours with
the largest beam current, 3.7 nA in e-Line system. This big pad writing also affects
the following nano pattern writing by back scattering. Though the pad areas are not
69
directly exposed, they are partially exposed by back scattered electrons. This lowers
the threshold dose and dose to opening diameter relation is also drifted. Figure 2-25
shows the drift of opening diameter with respect to the beam dose. The figure said
that the required dose for certain diameter becomes a half of its original value.
Table 2-3 Flow of one-step deposition process.
Job
No.
Job Equipment Condition Check
point
T01 SiN
X
deposition PECVD SiH
4
:NH
3
:N
2
=40:20:60
275 C/30W/440mT/2.5min

T02 PMMA Coating Spinner/
Hotplate
PMMA 950k C2
Slow/main spin coating
0.25/6krpm, 3/45s
Clean the backside
Baking on the hotplate
180 C/90s
Dark blue
color after
baking
T03 EBL (Pads) e-Line V
ACC
=10kV
WD=10mm
Aperture=120 m
Area dose= 180 C/cm
2

Pattern: Inverse_Pads

T04 EBL (Dots) e-Line V
ACC
=10kV
WD=10mm
Aperture=7.5 m
Dot dose=0.5 fC
Pattern: QD_PTN_200_500

T05 Development MIBK:IPA=1:3 60s
Rince in IPA for 30s
Dry N
2
blowing

T06 Check Optical
microscope
Check mark
definition
T07 SiN
X
etch
RIE
Old RIE CF
4
/100W/100mT/0.12 min
(Just etch : 0.08 min)
Use
slow pump
T08 Stripping Acetone/MeOH/DI bath
5 mins each.
Dry N
2
blowing follows
Acetone bath must be with
ultrasonic agitation

T09 Check Optical
microscope
Check
residue

70
Table 2-3 Continued
Job
No.
Job Equipment Condition Check
point
T10 O
2
ashing Old O
2

asher
Sufficiently wait for the base
pressure
120W/150mT/5min

12 345 6 7
25
50
75
100
125
150
175
One-step deposition process
Two-step deposition process

Opening diameter (nm)
dot dose (fC)
1.0
1.5
2.0
2.5
Diameter ratio

Figure 2-25 The opening diameter of single deposition process is affected by the back
scattering of pad writing and becomes larger than that of double
deposition process at same electron dose. The diameter ratio between two
processes is approximately 2.
2.4.3 Metal Contacts for Individual Nanowires
To measure electric conduction property of a nanowire, 2 metal contacts
should be formed on the two ends of the nanowire. However, forming metal
contacts on a nanowire is very challenging, due to their small size. For vertical and
uniform nanowire arrays, the contact can be formed on the top of nanowires array, as
71
illustrated in Figure 2-26. However, this approach only available for nanowire
arrays and it is impossible to measure the property of individual nanowires.
Substrate
Polymer

Substrate
CMP Milling

(a) (b)
Substrate
Contact metal

(c)
Figure 2-26 Schematic of metal contact formation on a vertical nanowire array, (a)
first, the gap between nanowires is filled with polymer. (b) The top
surface is flattened by chemical mechanical etching (CMP). (c) Contact
metal is deposited on the top of the array.
Using a suspension solution of nanowires is another method. Individual
nanowires can be selected by covering a substrate with the suspension solution and
then drying the casting solvent. On the sample, individual nanowires must be
located before forming a metal contact because the location of nanowires is fully
72
random. Figure 2-27 is the design of metal contact and probe pads and alignment.
Photolithography is not a good choice to make this type of metal pattern because a
unique mask would be needed for each sample. To use EBL for this contact
formation, the substrate must have alignment marks because it is also one kind of
mix-and-match lithography. The position of two ends of a nanowire must be
determined in the UV coordinate system. Then the calculated center of nanowire
will be the center of the pad layout. The angle between nanowire and U axis gives
the rotation angle of contact arms to make them perpendicular to the wire. The
location of the pad layout and rotation and location of contact arms can be
automatically calculated.










 

1 2
1 2 1
tan 90
x x
y y


 




  

2
,
2
,
2 1 2 1
y y x x
y x
C C

Figure 2-27 The design and alignment scheme of the metal contact pad. Alignment is
achieved by putting the center of layout at the center of the nanowire.
Rotation of group 2 layout (contact arms) is calculated by the angle
between nanowire and u axis.
73
For the lift off process of metal patterns, a dove tail like resist profile is
needed, which can be achieved by a bilayer PMMA process. Figure 2-28 illustrates
the PMMA bilayer process. As discussed in section 2.2.5, the sensitivity of PMMA
resist is inversely proportional to the molecular weight of PMMA. If a 950k PMMA
under 495k PMMA, wider area of 495k PMMA reaches its threshold dose during an
exposure, and an artificial overhang structure like Figure 2-28 (c) is created. The
process is
#1. Spin coating 1: PMMA 495k 5% in chlorobenzene, 4 kRPM, 45s
#2. Prebaking 1: on a hot plate, 180 C, 10 mins
#3. Spin coating 2: PMMA 495k 5% in chlorobenzene, 6 kRPM, 45s.
#4. Prebaking 2: on a hot plate, 180 C, 10 mins
#5. EBL: @ 10kV, 250 C/cm
2

#6. Development: MIBK:IPA=1:2 or 1:3, 60s
The threshold dose of the bilayer process is 220 C/cm
2
, which is 30% higher
dose than that of 100 nm thick 950k PMMA single layer process. Figure 2-29 is the
SEM micrographs of formed metal contact and its alignment. From fig, if alloying
step is needed to form ohmic contact, it must be done in phosphine or arsine
environment to prevent damage on the nanowire due to thermal etching of III-V
material.

74

PMMA 950k
Substrate
Electron Beam
PMMA 495k

(a) (b)

(c)
Figure 2-28 PMMA bilayer process is using sensitivity difference between layers. (a)
More sensitive layer (PMMA 495k) is located under the less sensitive
layer (PMMA 950k). (b) After the exposure and development we can get
overhang structure because of sensitivity difference. (c) SEM micrograph
of trenches formed by the bilayer process.

75

(a) (b)

(c)
Figure 2-29 SEM micrographs of the metal contact. (a) overview, and (b) alignment.
If alloying is done in hydrogen or nitrogen environment, (c) nanowires are
damaged by thermal etching.
2.5 Summary
In this chapter, EBL parameters are optimized to achieve minimum opening
diameter and center-to-center spacing as small as possible with the Raith e-Line
system at USC. PMMA 950k C2 is used for EBL resist. A resist thickness of 100
nm is chosen to get enough thickness in dry etch. Acceleration voltage, beam
current, electron dose, working distance, and write field size are optimized for nano
76
opening array fabrication. The acceleration voltage is set to 10 kV, which balances
forward and backward scattering. Electron dose varies with respect to center-to-
center spacing because of scattering. At 200 nm center-to-center spacing, 1 fC dot
dose is the slight overexposure condition for 100 nm thick 950k PMMA. The
working distance is set to 10 mm to achieve high resolution and deep depth of focus
in the writing. The write field is 100 m × 100 m, because typical pattern area is
smaller than 100 m × 100 m. From this optimization, 25 nm diameter and 80 nm
of center-to-center spacing are achieved.
To prepare nano patterned templates, dry etch and other process are designed
for SAG template preparation. There are two different approaches developed to
generate templates. Two-step deposition process utilizes mix-and-match technique
to reduce the EBL writing time. One-step deposition process is developed to
minimize damage on the growth surface. Bilayer lift-off process is also developed
metal contact patterns.

77
Chapter 2 Endnotes
[1] M. A. McCord and M. J. Rooks, "Electron beam lithography," in Handbook
of Microlithography, Micromachining, and Microfabrication. vol. 1, P. Rai-
Choudhury, Ed., ed Bellingham, WA: SPIE Optical Engineering Press, 1997.
[2] R. Li, "III-V compound semiconductor nanostructures by selective area
growth using block copolymer lithography," Ph.D., Material Sciences,
University of Southern California, Los Angeles, 2003.
[3] J. Liang, H. Luo, R. Beresford, and J. Xu, Appl. Phys. Lett. 85 5974-5976
(2004).
[4] M. Fujii, H. Iwanaga, and N. Shibata, J. Cryst. Growth 99 179-182 (1990).
[5] "Technical Report: ZEP520A," ZEONREX Electronic Chemicals 2003.
[6] D. F. Kyser and N. S. Viswanathan, J. Vac. Sci. Technol. 12 1305-1308
(1975).
[7] P. A. Peterson, Z. J. Radzimski, S. A. Schwalm, and P. E. Russell, J. Vac. Sci.
Technol. B 10 3088-3093 (1992).
[8] "Nano PMMA and Copolymer Datasheet," Microchem Co. 2001.

78
Chapter 3: Growth of Phosphide Nanowire Arrays
3.1 Introduction
InP and related materials have played important roles in optoelectronics and
optical telecommunications[1-3]. These materials have superior gain and low loss
properties in the fiber telecommunication band, 1550 nm wavelength range. InP
based materials are also good candidates for photovoltaic applications. This material
system provides the capability to engineer a direct bandgap over a wide range near
1.344 eV. As opposed to GaAs based material system, InP shows fewer surface
states with native oxides. Fewer surface states are extremely important to
nanostructure devices such as nanowires and quantum dots because their surface to
volume ratio is huge. High surface state density can easily deplete carriers in a
nanostructure, preventing it from interacting with photons. GaP has a large bandgap
2.27 eV in a heterojunction system. Though GaP has a low photon-valence electron
interaction due to its indirect bandgap, it is a seed material for InGaP, which has a
direct bandgap. InGaP can be used to make strain free structures on GaAs at 49% of
In composition with a 1.9eV bandgap. The tandem 3 junction solar cells
demonstrated by Spectrolab are based on InGaP-GaAs-Ge material system[4], which
have given the state-of-the-art conversion efficiency.
InP is an interesting material, because its ionicity is located in between GaN
and GaAs. The ionicity of a bond in a crystal determines the interaction strength
79
between bonds and ions in the crystal. Figure 3-1 is two different stacking of
tetrahedron in a III-V semiconductor. Figure 3-1 (a) is called zincblende (ZB)
structure. ZB structure is the common structure of the bulk form of InP, GaP, GaAs,
and InAs, and is equivalent to the diamond structure of Si or Ge. In the ZB structure,
all tetrahedron rotate by 120  when stacked. This is a sign of the strong interaction
between bonds, and occurs when the bond is less ionic. If a bond is less ionic, the
electron distribution is concentrated at the center of bond and ions are more neutral.
120 rotation of the tetrahedrons in ZB structure is caused by repulsion between
bonds. ZB lattice sites are the location where the distance between bonds is
maximized. On the other hand, Figure 3-1 (b) illustrates in-phase stacking known as
wurtzite (WZ) structure. This structure is the common structure of GaN. In phase
stacking of WZ structure indicates a strong interaction between ions. If the bond is
more ionic, electrons are moved toward the anion. The repulsion between bonds is
reduced, and the attraction between ions becomes stronger. At WZ lattice sites, the
distance between ions is minimized and the interaction is maximized. In the Philips
scale[5], the ionicities of GaN and GaAs are 0.500 and 0.310, respectively[6]. For
reference, the Philips ionicity of NaCl is 0.935[7], and InP is 0.420[6]. This
indicates that the ionicity of InP is not large enough to form a bulk WZ structure, but
the structural stability is lower than GaAs.
80

(a)
B
A
B
A
B
A
<1100>
<1120>
<0001>
<0001>

(b)
Figure 3-1 Cross-section and top view of stacking of tetrahedral semiconductors. (a)
zincblende stacking, and (b) wurtzite stacking. Red spheres are cations
and blue spheres are anions.
81
Structural stability of the surface is very important in nanostructures because
of their high surface to volume ratio. There are several suggestions for this stability
calculation[8-10]. A simple approach is to apply the change in bulk cohesive
energy[8]. Bulk cohesive energy is the structure formation energy of the basic
building blocks of structures, such as tetrahedron in GaAs and InP. Equation (3-1) is
the total formation energy of a structure. E
0
is the bulk cohesive energy part in the
total formation energy. Equation (3-2) and (3-3) show the calculation of the bulk
cohesive energy.
E E E   
0
(3-1)



j i
ij
V E
,
0
(3-2)
    









     



  
i
ij ij i ij ij
Z
G
r B r R r A V
0
exp exp (3-3)









   
ii
i
i
bb
b
i
r
Z
f
r
Z
f K E
2 2
1
2
3
(3-4)
Here V
ij
is the Khor-Das-Sarma type empirical interatomic potential with the second-
nearest neighbor[11-13]. Z
i

is the effective coordinate number for atom i. R
i
is the
minimum distance between neighbors. G( ) is the bond bending term for
tetrahedrally bound atom pairs. Z
b
, Z
i
, r
bb
, and r
ii
are the effective charge in a bond
and ion, distance between the nearest bonds and distance between the nearest ions,
respectively. The potential parameters A, B
0
, , , , ,  and  are determined using
82
cohesive energy, elastic moduli and relative stability among various structures
obtained by first principles calculations [8]. K is 8.7 meV/atom[14].
In the bulk phase, bulk cohesive energies of WZ and ZB are the same,
because they are based on a tetrahedral structure. The relative structural stability is
determined by stacking. The effect of stacking in total is E in equation (3-2). As
expressed in Equation (3-4), E is a function of ionicity. The relation of ionicity and
stacking is expressed quantitatively in Equation (3-4). If a nanostructure is formed,
truncation of the tetrahedron structure occurs at the surface. Truncation of
tetrahedron is different in different crystal planes and changes the bulk cohesive
energy. Akiyama et al reported this truncation is beneficial to form WZ structures in
nanowires, because the bulk cohesive energy rise of the ZB structure induced by
truncation is larger than that of the WZ structure[8]. To make structural transitions
in nanowires, the diameter of the nanowires should be smaller than a certain critical
diameter. The critical diameter is the diameter, where the bulk formation energy
difference between two structures is compensated by the surface formation energy
difference. Therefore the critical diameter is inversely proportional to the formation
energy difference of two bulk structures, in other words, less ionic semiconductors
require smaller diameters for structural transitions due to the strong stability of the
ZB structure in bulk form. According to reference[8], the critical diameters of GaAs
and InP nanowires are 15 nm and 32 nm, respectively. The critical diameter of InP
nanowires can be achieved by EBL, and this causes structural transitions in InP
nanowires grown by NS-SAG.
83
In this chapter, growth condition of InP and GaP nanowire NS-SAG will be
explained. SEM and TEM observation results will be also provided. Characteristic
growth behaviors of InP nanowire array growth and their relation to growth
conditions and patterning parameters is also explained.
3.2. Growth of InP Nanowire Arrays
Growth behavior changes in InP nanowire NS-SAG with respect to the
growth parameters such as partial pressure of precursors or growth temperature have
been studied[15-16]. In prior works, InP nanowire growth has been performed in
very low V/III ratio ranges of 10 to 50. Most research has ben undertaken with
Tertiarybuthylphosphine (TBP) as the group V precursor which shows better
cracking efficiency even at low temperature. Growth behavior with phosphine (PH
3
)
has not been intensively researched because of its narrow parameter window for
successful growth. PH
3
is very stable and homogeneous pyrolysis of PH
3
is very
challenging in the range of typical InP MOCVD growth. Also, the cracking
efficiency of PH
3
is a very strong function of growth temperature in the temperature
range of InP MOCVD growth. Changing the growth temperature also changes the
effective supply of group V element and the V/III ratio, which complicates growth
optimization. However, PH
3
is less expensive and more accessible material than
TBP. So, developing InP nanowire array NS-SAG conditions and the growth
parameter optimized for use with PH
3
is important.
84
The group III and group V precursors are trimethylindium (TMI) and PH
3
,
respectively. The TMI is solution based, developed by Epichem, whose vapor
pressure is 3.78 psi at 30.1 C. The pressure of the TMI bubbler is set at 900 psi.
The standard growth condition (SGC) of InP nanowire array is 30 sccm of TMI flow
and 10 sccm of PH
3
flow at 650 C. The partial pressure of TMI and PH
3
is 1.8×10
-6

atm and 1.43×10
-4
atm, respectively. The V/III ratio is 79.4. For reference, standard
InP film growth condition is 180 sccm of TMI flow and 130 sccm of PH
3
flow at 650
C, which delivers 172 of V/III ratio. The templates are prepared by EBL as
described in chapter 2. Before loading the templates, they are treated by diluted
hydrochloric acid (HCl) for 20 seconds to remove native oxide and damaged top
surface layer. The dilution is 1 part of HCl in 3 parts of deionized (DI) water.
3.2.1 Growth of InP Nanowire Arrays on (111)A Substrates
As reported by Inari et al [16], InP (111)A templates are able to yield vertical
and uniform nanowire arrays. The SGC is optimal condition for nanowire growth on
InP (111)A substrates. Figure 3-2 is the result of nanowire growth on InP (111)A
with respect to normalized precursor concentration (NPC). NPC is the precursor
concentration which is normalized by concentration on the bare semiconductor
region without mask effects. The detailed information of NPC is described in
chapter 5. Two shape parameters, 
T
and 
L
are defined to discuss the shape of
grown InP nanowires quantitatively. Equation (3-5) is tapering parameter 
T
.
85

L
r r
top btm
T

  (3-5)
, where r
top
and r
btm
is the radius of the top and bottom of a nanowire. L is the length
of the nanowire. Lateral growth parameter 
L
is defined as

o
btm
L
r
r
  (3-6)
Here, r
0
is the diameter of the nano opening on the mask. Figure 3-3 is the graphs of

T
and 
L
with respect to NPC. Both parameters show that the reaction at the surface
increases as NPC increases. 
T
starts to rolls off at 6.0. As shown in Figure 3-2 (c),
nanowires longer than 1 m have a kink in the middle and above the kink, the degree
of tapering is reduced. The roll off depicts this tapering reduction in long nanowires.
Enhanced sidewall growth and aspect ratio degradation in NS-SAG was also
reported by Kitauchi et al [10]. The sidewalls of InP nanowires are non-polar planes.
So, the growth enhancement on the sidewall implicates the growth enhancement on
the non-polar planes. In NS-SAG, growth on the non-polar plane is only restricted
by growth conditions which us the major difference from VLS growth. In VLS
growth, the nanowire growth is physically restricted by the size of the metal catalysts.
Also, the growth temperature of VLS is too low for typical semiconductor film
growth. Even in VLS growth, unusual sidewall morphologies such as branching[17],
tapering[18], and bump generation[19] have been reported. Sidewall growth in NS-
SAG is much stronger and very sensitive to the growth conditions compared to VLS
86
technique. The relation among 
T
, 
L
, and NPC illustrates sensitivity of sidewall
morphology to growth condition.

(a) (b)

(c) (d)
Figure 3-2 SEM micrographs of InP nanowire array grown on InP (111)A substrates.
All samples are grown at different NPC. The NPC of each sample is (a)
4.0, (b) 5.6 (c) 10.2, and (d) 24.0, respectively.

87

3 456 789 10 11
0.00
0.01
0.02
0.03
0.04
0.05

Tapering parameter ( 


Normalized precursor concentration

(a)
3 456 789 10 11
0.9
1.0
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2.0

Lateral growth parameter ( 
L
)
Normalized precursor concentration

(b)
Figure 3-3 Surface morphology parameters with respect to NPC. (a) Tapering
parameter, 
T
and (b) lateral growth parameter, 
L
.

88
The tapering parameter and lateral growth parameter can tell the degree of
sidewall growth but are easily affected by sidewall bumps. Figure 3-4 is the typical
morphology of sidewall bumps. If bumps appear, it is hard to define the nanowire
diameter, and at the same time, the location of the bump affects 
T
and 
L
. However,
bumps can describe the degree of sidewall growth. Figure 3-5 illustrates the density
of bump on the sidewalls is proportional to NPC. From the bump density, growth on
the sidewalls becomes very active at 5.5 of NPC or above. In high NPC conditions,
the growth on non-polar planes and other semi-polar planes become comparable to
the growth on (111)A. In this condition, wire formation is stopped and clusters
appear. Figure 3-2 (c) and (d) illustrate this tendency. In the SGC, the formation of
clusters starts where NPC is greater than 8.5 and if NPC is greater than 14, cluster
formation is much stronger than wire formation and the occupation ratio of
nanowires drops below 10%. This suggests that the precursor concentration must be
lower for better surface morphology of nanowires and array uniformity.

Figure 3-4 60  tilted SEM micrograph of an InP nanowire has sidewall bumps.
89
3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4

Average number of bump per nanowire
Normalized precursor concentration

Figure 3-5 Average numbers of bumps per nanowire with respect to the normalized
precursor concentration.
Extensive TEM observations reveal the crystal structure of InP nanowire in
the [111]A direction. Figure 3-6 confirms that the grown InP nanowire in [111]A are
WZ crystal up to 125 nm in diameter, which is significantly larger than the
thermodynamic estimation of 32 nm. The structure is not free from stacking faults
(SFs). The SF density is proportional to TMI partial pressure and ranges from 10
m
-1
to 40 m
-1
. The model for this enhanced WZ structure formation and SF
generation will be discussed in chapter 5. Figure 3-6 (b) shows the TEM
observations in the <112 ¯ 0> zone axis, which delivers better contrast at SFs and
shows the cross section of the sidewall bumps. From the figure, it is clear that the
bumps are lateral growth in the <11 ¯ 00> direction and the inclined facets appearing at
the bump is always the {11 ¯ 01} semi polar crystal planes. It is of interest to note that
90
a bump always starts from a SF. There is no bump generated without a SF from the
TEM observations. The bump continues until the (11 ¯ 00) plane disappears and a
sharp edge is formed unless there is no SF formed in the middle of the bump. If the
bump encounters another SF, the bump generation is terminated and the nanowire
continues growing with a deformed hexagonal cross section which has a truncated
triangular bump on a facet. This indicates that the SFs make great contribution in
initializing and terminating of the sidewall bump generation mechanism. The
density of sidewall bumps is proportional to the density of SFs in the nanowire.
Therefore, enhanced sidewall bump generation can be explained by growth
enhancement in the non-polar direction and increased SF density.

(a) (b) (c)
Figure 3-6 TEM micrographs of [111]A direction InP nanowires. Upward direction is
the [111]A direction. (a) Lattice image of an InP nanowire taken in the
[11 ¯ 00] zone axis, (b) image of a sidewall bump and SFs taken in the [112 ¯
0] zone axis, and (c) selective area electron diffraction pattern in the
[11 ¯ 00] zone axis.
91
The observations upon stacking fault generation and weakened growth
suppression on the non-polar plain give an explanation of the mechanism for
sidewall bump generation. Figure 3-7 is the schematic of the atomic layout of a WZ
InP nanowire grown in the [111]A direction with a sidewall bump. If a SF appears
in a WZ crystal, the SF exposes polar {111}A planes in the sidewalls. Therefore, the
SF introduces a polar plane in the non-polar sidewall planes. Though growth on the
non-polar direction is strongly suppressed by the growth conditions, the growth on
the exposed polar {111}A plane occurs relatively easily, because the growth on the
polar plane is promoted by the growth conditions. This can generate a small bump at
the SF and supplies extra dangling bonds which act as adsorption sites for precursors.
If the growth suppression in the non-polar, lateral direction is not strong enough, the
bump formation initiates to stabilize the extra dangling bonds from the bump seed
from the stacking fault. During growth, the bump growth forms semi polar (11 ¯ 01)
plane. The appearance of semi-polar planes such as (11 ¯ 01) or (112 ¯ 2) in SAG of WZ
semiconductor crystal has been reported in GaN SAG[20-22]. In prior works, the
strong growth suppression on the (11 ¯ 01) semi-polar plane is explained by surface
reconstruction due to the excessive N adatoms in high V/III ratio conditions. If the
V/III ratio is far greater than 1, the surface dangling bonds in the semi-polar plane
are stabilized by forming bonds with excessive N atoms. The stabilized planes
become inactive and show strong growth suppression. However, there has been no
direct observation of surface reconstruction studies or growth kinetics simulations on
these semi-polar planes. This bump generation can be stopped when another SF is
92
generated in the bump. Introducing another SF in a bump disturbs the growth in the
non-polar direction and forms the (11 ¯ 01) semi-polar plane which is very important to
continue the bump growth.
<1100> <1120>
[0001]
A
P Atom
B
In Atom
P Atom
In Atom
Hexagon facet
surface
Bump
[0001]
A
B
SF
SF
Top View

Figure 3-7 Cross sectional schematic of the sidewall bump. Cross section plane is
(112 ¯ 0). It is also marked red solid line in the top view of nanowire. Red
dotted line is the projection of the (11 ¯ 01)A semi-polar plane. Blue dotted
line is the projection of {111}A polar plane. ZB style SFs are marked by
green band.
93
Introducing dopants in InP NS-SAG alters the growth behavior. Figure 3-8
shows the micrographs of doped InP nanowire arrays. The n-type and p-type
dopants are disilene (Si
2
H
6
) and diethylzinc (DEZ), respectively. Both of these
dopants affect growth, especially sidewall growth, but the impact of DEZ is stronger
than that of Si
2
H
6
. The surface morphology changes in n-doped nanowire are
insignificant. The circled region is a thick nanowire induced by n-doping. The
occurrence of thick nanowires stays constant in the 0.25 sccm to 2 sccm Si
2
H
6
flow
range, which leads to a IV/III ratio from 1.98 to 15.87. For reference, a 1.32 IV/III
ratio in n-InP film growth causes 1.2×10
18
cm
-3
doping. The tested II/III ratio range
is from 0.21 to 1.70. For p doping of planar InGaAs, a 0.60 II/III ratio gives
2.0×10
18
cm
-3
doping. DEZ doping changes the growth behavior further. Random
growth suppression and strong lateral growth happen in high DEZ flow. Figure 3-8
(c) shows the degradation of array uniformity at DEZ flow of 1.70. Both of the
dopants weaken growth suppression on the sidewalls. The effect of DEZ is stronger
than Si
2
H
6
.
InP nanowire growth on (111)A is very sensitive to the incorporation of other
elements such as Ga or As. This makes the heteroepitaxial nanowire growth of InP
related materials challenging. Figure 3-9 shows InP nanowires grown on a GaAs
(111)A subsrtate. Interestingly, the InP nanowires grow in <111>B or <112>B
directions. Most of the openings have small clusters, and only a few openings have
tilted nanowires. The diameter of nanowires in <111>B directions is larger than 100
nm.
94

(a) (b)

(c) (d)
Figure 3-8 Surface morphologies of doped InP nanowire arrays. N-doping is done by
Si
2
H
6
at (a) 4.00 and (b) 15.87 IV/III ratio. Circled nanowires show
weakened sidewall growth suppression. P-doping is done by DEZ at (c)
0.21 and (d) 1.70 II/III ratio.

Figure 3-9 Surface morphologies of InP nanowire NS-SAG on a GaAs (111)A
substrate.
95
3.2.2 Growth of InP Nanowire Arrays on (111)B Substrates
Vertical uniform nanowire arrays on (111)B substrates is very important for
heterostructures with arsenide based nanowires or GaP nanowires whose preferred
growth direction is [111]B. For tandem heterosturcture nanowires, uniform
nanowire array growth conditions must be found. The SGC of InP nanowire NS-
SAG is not optimal for InP (111)B templates. Because the TMI flow in the SGC is
too large for NS-SAG on InP (111)B, many big clusters are generated as shown in
Figure 3-10 (a). For better result, the TMI flow must be reduced, but it is still
challenging to achieve uniform and vertical nanowire arrays on InP (111)B. Figure
3-10 shows the change in growth with respect to the growth parameter change. If the
growth temperature is higher than 650 C, the growth rate becomes very slow, which
indicates the desorption of phosphorus (P) atoms on the (111)B surface is too fast
and the PH
3
overpressure is insufficient to compensate for the desorption. If the
growth temperature is lower than 650 C, large InP clusters develop, which shows
enhanced growth on the sidewall and other high index planes. Slight changes in the
flow rate of TMI and PH
3
do not improve growth uniformity. The growth behavior
is strongly affected by the preparation process of the template. The templates
prepared by the 1 step deposition process shows better results in terms of density of
nanowires. Figure 3-11 (a) is the surface morphology of nanowire grown on a
template prepared by 2 step deposition process described in chapter 2. This indicates
that the surface condition of template plays an important role in the growth behavior.
96

(a) (b)

(c) (d)
Figure 3-10 Surface morphologies of InP nanowire arrays grown on InP (111)B
substrates. Two InP nanowire arrays are grown under 6.0×10
-7
atm of
TMI partial pressure and 1.4×10
-4
atm of PH
3
partial pressure at (a) 625
C and (b) 650 C, respectively. Other arrays are grown at 650 C with
different precursor partial pressures. (c) An InP nanowire array is grown
at 3.0×10
-7
atm of TMI partial pressure and 1.4×10
-4
atm of PH
3
partial
pressure. (d) The other array is grown at 6.0×10
-7
atm of TMI partial
pressure and 1.8×10
-4
atm of PH
3
partial pressure. All templates are
prepared by 1 step deposition process and treated by diluted HCl before
the growth.

97

(a) (b)
Figure 3-11 Surface morphology change of InP nanowire arrays grown on InP
(111)B substrates with respect to SAG template preparations. (a) InP
nanowires grown on a template prepared by 2 step deposition process, and
(b) tripod structures occurred on a template treated by H
2
SO
4
before the
growth.
The pre-surface treatment affects the growth. Figure 3-11 (b) is the surface
morphology of InP nanowire array growth with a template treated by diluted H
2
SO
4
.
From Figure 3-10 and Figure 3-11 (b), the templates treated by diluted HCl in DI
(dilution ratio is 1:3 and treatment time is 10s) before the growth have vertical
structures after the growth. However, the templates treated by diluted H
2
SO
4
in DI
(dilution ratio is 1:1 and treatment time is 30s) show tripod structures whose legs are
grown toward the nearest <111>A directions.
Figure 3-12 is TEM observations of InP nanowires in the [111]B direction.
The grown nanowires are in the ZB structure. Because of strong growth on the
sidewalls, the minimum diameter achieved is 60 nm. Here, the estimation of critical
diameter is valid down to 60 nm. No enhanced WZ structure formation occurs in the
[111]A direction InP nanowires. SAED pattern is streaked out because of rapid
98
twining in the structure. The SF density in the [111]B direction InP nanowires is
almost an order higher than of the [111]A direction InP nanowires. The average SF
density ranges from 100 m
-1
to 200 m
-1
. The sidewalls shows no visible bumps,
but in the atomic scale, the actual sidewall is not non-polar {11 ¯ 0} planes, but semi-
polar {112} planes or higher index planes. Because of rapid twining, the {112}
planes alternate between A and B polarities. It is unknown if the appearance of
semi-polar planes is caused by the growth or ex-situ oxidation after growth.

(a) (b)
Figure 3-12 HR-TEM observations of InP nanowire grown in the [111]B direction. (a)
The lattice image of the [111]B direction InP nanowire. Upward direction
is the [111]B direction. (b) SAED pattern of a [111]B direction InP
nanowire.
As opposed to InP nanowire growth on (111)A, InP nanowire growth on
(111)B is not sensitive to the incorporation of other elements such as Ga or As.
Growth on GaAs (111)B is suppressed of the growth and large clusters show {111}B
facets or growth in the <111>B directions. Figure 3-13 shows there is no major
99
difference in the growth behavior between InP nanowire growth on InP (111)B and
GaAs (111)B.

Figure 3-13 SEM micrograph of an InP nanowire array grown on GaAs (111)B.
3.3 Growth of GaP Nanowire Arrays
The available window of growth parameter combination for GaP NS-SAG is
very narrow and because of the trade-offs between parameters, it is challenging to
achieve uniform nanowire array condition with high aspect ratio and small wire
diameters. Typically, the growth temperature of planar epitaxial GaP film is higher
than 700 C. As shown in Figure 3-14, a cluster formation is dominant at 700 C.
Reducing precursor flow decreases cluster generation, but at the same time, the
growth itself is also suppressed. To maintain the growth, a high PH
3
partial pressure
is essential. This increases the growth on the sidewalls and other planes. Noticeable
nanowire formation occurs at 2.36×10
-6
atm and 2.14×10
-4
atm PH
3
partial pressure,
in which many islands are grown as shown in Figure 3-14 (a).
100

(a) (b)

(c)
Figure 3-14 Surface morphology of an GaP nanowire array grown at 700 C. The
substrate is InP (111)B. (a) GaP nanowire array grown at 2.36×10
-6
atm of
TMG partial pressure and 2.14×10
-4
atm of PH
3
partial pressure, (b) GaP
nanowire array grown at 1.18×10
-6
atm of TMG partial pressure and
2.14×10
-4
atm of PH
3
partial pressure, and (c) GaP nanowire array grown
at 1.18×10
-6
atm of TMG partial pressure and 1.42×10
-4
atm of PH
3
partial
pressure.

101
This can be explained by the narrow growth window of the PH
3
partial
pressure. Figure 3-15 illustrates growth window. At a certain growth temperature,
the PH
3
partial pressure must be greater than a certain pressure to overcome P
desorption from the surface. In Figure 3-15, P
V
is the minimum PH
3
partial pressure
which can overcome P desorption. P
V
increases as the growth temperature increases,
because the high growth temperature accelerates dissociation of In-P bonds on the
surface, and as a result, P desorption increases. At the same time, there is another
partial pressure point, P
L
, which is the maximum PH
3
partial pressure that can
suppress growth on non-polar or semi-polar planes. To fabricate nanowires with
high aspect ratios, growth on the polar (111) plane must be active and growth on the
non-polar sidewall planes must be suppressed. The overlapping region in Figure 3-
15 is the growth window which enables nanowire growth. If P
V
is near P
L
, it is
challenging to keep the right growth conditions for nanowire synthesis. GaP
nanowire NS-SAG at 700 C shows this tight growth window.
P
V
P
PH3
P
L

Figure 3-15 Schematic of growth window. P
PH3
is partial pressure of PH
3
. P
V
and P
L

are respectively the minimum partial pressure of PH
3
for growth on polar
(111) surface and the maximum PH
3
partial pressure to suppress growth
on non-polar sidewalls. The overlapping region is the growth window.
The span of the growth window can be tuned by the growth temperature,
because the growth temperature governs surface P desorption and the surface
102
diffusion length of Ga on the sidewalls. At 750 C, the growth is completely
suppressed and occupancy of the opening drops to less than 1%. At 650 C, the
occurrence of clusters can be reduced, and vertical nanowire arrays can be formed;
however, as shown in Figure 3-16, the sidewall growth is still noticeable. The
average diameter of the grown GaP nanowires in 30 nm openings is 85 nm, or 2.8
times larger than the original openings. Interestingly, cluster generation increases in
larger diameter openings. This tendency is independent to the growth temperature.
This can be explained by large lattice mismatch between GaP and InP, 11%.
Normally, heterostructure film growth with a 11% lattice mistmatch shows Volmer-
Weber growth mode. Larger openings may have more strain in the structure, and the
strain forms islands.

(a) (b)
Figure 3-16 SEM micrographs of GaP nanowire arrays grown at 650 C. Partial
pressures of TMG and PH
3
are 1.18×10
-6
atm and 1.42×10
-4
atm,
respectively. The opening diameters are (a) 30 nm and (b) 50 nm.
The results of GaP nanowire NS-SAG on InP (111)A are shown in Figure 3-
17. The growth shows many islands. But, we can find that all of the short pillar
103
structures are heading toward the <111>B or <112> B directions. This indicates that
the preferred growth direction of GaP nanowire is [111]B direction, which is
opposite to the other phosphide, InP.
<110>
<112>

Figure 3-17 SEM micrographs of GaP nanowire arrays grown on InP (111)A
substrate.
3.4 Summary
Optimal InP NS-SAG conditions are found by extensive growth experiments.
The optimized growth condition for InP nanowires is 1.8×10
-6
atm of TMI partial
pressure, 1.43×10
-4
atm of PH
3
partial pressure at 650 C. The V/III ratio for this
condition is almost 7 times higher than previous work which uses TBP as a group V
precursor. The preferred growth direction of InP nanowire is [111]A direction. The
effective precursor concentration of the nano opening array determines the surface
morphology of the nanowire, because the precursor concentration controls the degree
of reaction on the sidewalls and other non-polar and semi-polar planes. In high
precursor partial pressure tapering and lateral growth increase. Bumps also appear
104
on sidewalls in high precursor concentration, especially when NPC is greater than
5.5. If NPC is greater than 8.5, cluster growth starts to appear. The appearance of
clusters increases as NPC increases, and if the NPC exceeds 14, nanowire formation
is completely stopped due to strong growth in the non-polar direction other than
nanowire growth direction, [111]A.
[111]A direction InP nanowires have crystal structure transitions from ZB to
WZ. The critical diameter of this structural transition is larger than 125 nm. This is
larger than thermodynamic estimation, 32 nm. Average SF density is 20 m
-1
and is
proportional to TMI partial pressure. The sidewalls of WZ InP nanowires are the
{1 1 ¯ 00} planes. HR-TEM observations also illustrate the shape and origin of
sidewall bumps. {111}A planes are exposed on the sidewall by a SF and in high
precursor concentration, this small section of {111}A can be a seed for sidewall
bump generation. Dopants are able to affect the behavior of InP NS-SAG. Si
2
H
6

and DEZ are used to test surface morphology changes by dopant incorporation. Both
affect growth, especially sidewall growth, but the impact of DEZ is stronger than that
of Si
2
H
6
. InP NS-SAG on (111)A is very sensitive to incorporation of other
elements such as Ga or As.
InP NS-SAG on InP (111)B is not uniform. Most of the openings have no
growth and only a few openings have vertical nanowires in the [111]B direction.
This growth is sensitive to pre-treatment of template. If the template is treated by
diluted sulfuric acid, vertical nanowires disappear and tripod structures appear.
Through HR-TEM observations, it is found that the [111]B direction InP nanowires
105
are in the ZB structure with dense SFs. Average SF density is from 100 m
-1
to 200
m
-1
.
GaP NS-SAG is tried on InP (111) substrates. The preferred growth
direction of GaP nanowire is the [111]B direction. The growth parameter window of
GaP NS-SAG is very narrow. It is challenging to get a reasonable nanowire growth
rate and sidewall growth suppression. Noticeable nanowire formation occurs at
2.36×10
-6
atm and 2.14×10
-4
at 700 C. At 750 C, the growth is totally suppressed.
At 650 C, cluster generation is reduced but the diameter of grown nanowires is 2.8
times larger than that of the opening. The cluster generation is stronger in larger
opening. This indicates the growth behavior is affected by the large lattice mismatch
between InP substrate and grown GaP clusters.

106
Chapter 3 Endnotes
[1] S. J. Choi, P. Zhen, Y. Qi, H. Eui Hyun, and P. D. Dapkus, IEEE Photonics
Technol. Lett., 17 2101-2103 (2005).
[2] T. Sadagopan, S. J. Choi, C. Sang Jun, K. Djordjev, and P. D. Dapkus, IEEE
Photonics Technol. Lett. 17 414-416 (2005).
[3] S. J. Choi, K. Djordjev, C. Sang Jun, P. D. Dapkus, W. Lin, G. Griffel, R.
Menna, and J. Connolly, IEEE Photonics Technol. Lett. 16 828-830 (2004).
[4] R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H.
Yoon, R. A. Sherif, and N. H. Karam, Appl. Phys. Lett. 90 183516 (2007).
[5] J. C. Phillips, Rev. Mod. Phys. 42 317 (1970).
[6] S. Adachi, Properties of group-IV, III-V and II-VI semiconductors. West
Sussex: John Wiley & Sons Ltd., 2005.
[7] M. A. Wahab, Solid state physics: structure and properties of materials, 2nd
ed. Middlesex: Alpha Science International Ltd., 2005.
[8] T. Akiyama, K. Sano, K. Nakamura, and T. Ito, Jpn. J. Appl. Phys. 45 L275
(2006).
[9] E. Hilner, U. Håkanson, L. E. Fröberg, M. Karlsson, P. Kratzer, E. Lundgren,
L. Samuelson, and A. Mikkelsen, Nano Lett. 8 3978-3982 (2008).
[10] Y. Kitauchi, Y. Kobayashi, K. Tomioka, S. Hara, K. Hiruma, T. Fukui, and J.
Motohisa, Nano Lett. 10 1699-1703 (2010).
[11] K. E. Khor and S. Das Sarma, Phys. Rev. B 38 3318 (1988).
[12] T. Ito, K. E. Khor, and S. Das Sarma, Phys. Rev. B 40 9715 (1989).
107
[13] T. Ito, J. Appl. Phys. 77 4845-4886 (1995).
[14] T. Ito, Jpn. J. Appl. Phys. 37 L1217 (1998).
[15] P. Mohan, J. Motohisa, and T. Fukui, Nanotechnology 16 2903-2907 (2005).
[16] M. Inari, J. Takeda, J. Motohisa, and T. Fukui, Physica E 21 620-624 (2004).
[17] Z. H. Wu, X. Mei, D. Kim, M. Blumin, H. E. Ruda, J. Q. Liu, and K. L.
Kavanagh, Appl. Phys. Lett. 83 3368-3370 (2003).
[18] Y. Wang, V. Schmidt, S. Senz, and U. Gösele, Nature Nanotech. 1 186-189
(2006).
[19] J. Zou, M. Paladugu, H. Wang, G. J. Auchterlonie, Y.-N. Guo, Y. Kim, Q.
Gao, H. J. Joyce, H. Tan, and C. Jagadish, Small 3 389-393 (2007).
[20] K. Hiramatsu, K. Nishiyama, A. Motogaito, H. Miyake, Y. Iyechika, and T.
Maeda, Phys. Stat. Sol. (a) 176 535-543 (1999).
[21] D. Ren, W. Zhou, and P. D. Dapkus, Appl. Phys. Lett. 86 111901 (2005).
[22] T. W. Kim, Y. J. Hong, G.-C. Yi, J.-H. Kwon, M. Kim, H. N. Han, D. H.
Kim, K. H. Oh, K.-J. Kong, and Y.-K. Kwon, J. Phys. D: Appl. Phys. 41
015406 (2008).

108
Chapter 4: Growth of Arsenide and Heterostructure Nanowire
Arrays
4.1 Introduction
GaAs and related materials are the most widely used material system in the
compound semiconductor industry. Their superior electron mobility enables fast
transistors and other active devices for radio frequency or millimeter wave
applications[1]. The large band gap arsenide, AlAs has a very similar lattice
constant to GaAs and GaAs/AlAs heterojunction systems have virtually no lattice
mismatch problems. This allows for high quality quantum confinement systems,
such as quantum wells without mistfit dislocation problems. GaAs based high
electron mobility transistors (HEMT) and doped channel field effect transistors
utilize this heterojunction structure. GaAs has a direct band gap as InP does, so
GaAs is a good candidate for photovoltaic devices. Moreover, the lattice constant of
GaAs is also matched to germanium, Ge. The tandem 3 junction solar cells from
Spectrolab are based on the InGaP-GaAs-Ge material system[2], which gives the
state-of-the-art conversion efficiency.
InAs is an attractive material due to its superior electron mobility, and its
small band gap is suitable for far infra red optoelectronic devices. However, the
lattice mismatch to GaAs, InP, or Si is too large to achieve high quality crystal film
109
on these substrates. Native InAs substrates are available, but because the substrate is
highly fragile, only small size substrates are available on the market and they are
very expensive. A native substrate is suitable for high quality single crystal epitaxial
InAs films, but making dislocation-free heterojunctions with other materials is still
challenging. Because of this restriction, InAs has been studied for nanostructure
applications, instead of two dimensional epitaxial layer applications, such as
quantum dots or nanowires[3-4]. As opposed to GaAs, the surface Fermi level of
InAs is pinned in the conduction band[5-7], which causes electron accumulation at
the surface. Even with high surface state density, InAs is still a good conducting
material. In GaAs and InP, surface states deplete the carriers in the nanowire and
those materials become optically and electronically ‘dead’ without proper surface
passivation[8]. Effective and reproducible surface passivation is still challenging.
On the other hand, InAs nanowire devices have fewer requirements for surface
passivation. Many researchers have noted this property of InAs nanowires and have
also made InAs nanowire transistors[9].
InAs is also a seed material for InAlAs and InGaAs ternary materials.
InAlAs has been used as a high bandgap barrier of InP based HEMT devices which
are capable of operating over 77GHz or GaAs based metamorphic HEMT
devices[10-11]. InGaAs can be lattice matched to InP and provides a lower band gap.
InGaAs can be a good metal contact capping layers for InP based material
systems[12] and is a very important material in the photo detector world[13-14]. It
110
also plays important role as a high electron mobility channel in HEMT devices
[10-11].
The InAsP ternary compound semiconductor consists of InP and InAs.
Because the preferred growth direction and growth behavior of InP and InAs NS-
SAG are very different, InAsP can show the growth behavior change induced by
growth behavior conflict of binary components in a ternary compound.
To take full advantage of nanoheteroepitaxy discussed in chapter 1, tandem
heterostructure nanowire growth must be possible. Because there is no physical
restriction of sidewall growth, fabrication of tandem heterostructure nanowire using
NS-SAG is challenging. Though Fukui et al have reported several heterostructure
devices using NS-SAG[15-16], they are core-shell structures, which are on the
sidewalls of nanowires. Moreover, if the preferred growth directions of the two
different layers are different, the tandem nanowires may be kinked or bent.
Therefore, it is important to test the possibility of vertical tandem heterostructure
nanowire fabrication using NS-SAG.
For MOCVD of arsenic based materials, AsH
3
has been widely used for the
group V precursor. Tertiarybutylarsine (TBA) is another choice. But, the pyrolysis
efficiency of AsH
3
is far better than PH
3
in the growth temperature range of most III-
V materials, growth temperature limit of arsenide based material MOCVD is less
than for phosphide based material MOCVD[17]. There are two different pyrolysis
mechanisms of precursors in MOCVD. Pyrolysis that occurs entirely in the vapor
phase is known as homogeneous pyrolysis and pyrolysis occurring at a solid surface
111
is known as heterogeneous pyrolysis[18]. Pyrolysis can be unimolecular, which is
undergone by an excited reaction species alone, or bimolecular, which requires the
collision of two species. At the growth temperature of most III-V semiconductors,
heterogeneous cracking of AsH
3
is the dominant cracking mechanism. Stringfellow
reported that homogeneous AsH
3
pyrolysis is 50% complete at 600 C, while
heterogeneous AsH
3
decomposition on GaAs surface is 50% complete at 476 C[18].
Decomposition of PH
3
is also affected by heterogeneous pyrolysis, but the absolute
pyrolysis efficiency is much lower than AsH
3
because of strong bonds between P and
H atoms. Therefore, the effective V/III ratio of PH
3
related growth is a strong
function of growth temperature, which causes complexity and trade-offs in the
growth condition. Challenges in growth temperature control are the motivation for
the use of TBP as a precursor, which decomposes at relatively low temperatures.
Since AsH
3
can be efficiently decomposed at the growth temperature, there is
more freedom to change growth temperature without changing the effective V/III
ratio. This is why AsH
3
has been widely used in NS-SAG[3, 19-20]. In NS-SAG
with AsH
3
, the dominant mechanism of AsH
3
pyrolysis is heterogeneous
decomposition. In NS-SAG, the surface area is much larger than that of planar film
growth due to the contribution of the sidewalls. The contribution of the sidewalls is
a function of time. As nanowires grow, the sidewall area increases, which also
enhances the heterogeneous decomposition of AsH
3
. The As species supply
increases during growth.
112
In this chapter, growth conditions of GaAs and InAs nanowire NS-SAG will
be discussed. SEM and TEM observations confirm that the GaAs and InAs
nanowire are prone to grow in the [111]B direction, which is opposite to InP
nanowires. All grown nanowires are in the ZB crystal structure with dense SFs.
Characteristic growth behaviors of GaAs and InAs nanowire array are observed and
explained by heterogeneous pyrolysis of AsH
3
on the sidewalls. In order to see the
influence of the conflict between the preferred growth directions of arsenides and
InP, vertical GaAs/InP heterostructure nanowire growth and InAsP nanowire growth
are tested.
4.2. Growth of GaAs Nanowire Arrays
GaAs nanowire NS-SAG is done in higher growth temperature and lower
partial pressure conditions than InP nanowire NS-SAG. The growth temperature of
GaAs nanowire NS-SAG ranges from 650 C to 750 C. Trimethylgallium (TMG)
and AsH
3
are used for the group III and V precursors. The bubbler temperature and
pressure is set to -10 C and 1200 psi, respectively. The TMG flow is fixed at 0.8
sccm, which is equivalent to 3.79 ×10
-7
atm partial pressure. This TMG flow is the
minimum flow that shows no growth suppression even with the lowest AsH
3
flow.
The AsH
3
flow can be set between 5 sccm and 17 sccm. The SGC for GaAs
nanowire arrays is 0.8 sccm of TMG flow and 17 sccm of AsH
3
flow at 700 C. The
partial pressure of AsH
3
in SGC is 2.42×10
-4
atm. The V/III ratio is 638.5. Before
113
loading, the templates are treated by diluted HCl to remove any native oxide and
damaged top surface. The dilution ratio is 1:1 HCl : DI water. The treatment time is
30 seconds for GaAs substrates.
4.2.1 Growth of GaAs Nanowire Arrays on (111)A Substrates
It has been reported that the preferred growth direction of GaAs nanowires
grown by NS-SAG is the [111]B direction[21]. As shown in chapter 3, NS-SAG on
opposite polarity substrates causes non-uniform growth in arrays. Figure 4-1 is the
GaAs nanowire arrays grown using NS-SAG on GaAs (111)A substrates, but the
surface morphology of GaAs nanowire grown by NS-SAG on InP (111)A substrate
is quite similar. It also develops non-uniform nanowire array growth, but the surface
morphology is different from the InP nanowires grown on InP (111)B substrates.
First, there is no growth suppression in GaAs NS-SAG. Almost all of the openings
develop crystals. The occupation is greater than 95%. Only a few openings near
large clusters have no growth, because of precursor diffusion. The nano opening is
usually smaller than the lateral precursor diffusion length. The precursors can
diffuse out from the openings and diffuse into a GaAs cluster where the growth
favorable plane is located. Because of this diffusion, the openings near big clusters
have no growth. The other thing to note is the direction of nanowires, some
openings form nanowires, but the direction of the nanowire growth is the nearest
114
<111>B directions. This indicates that the preferred growth polarity of GaAs NS-
SAG is B polar. The mechanism of substrate selection will be discussed in chapter 5.

Figure 4-1 The top view of GaAs NS-SAG on a GaAs (111)A substrate.
4.2.2 Growth of GaAs Nanowire Arrays on (111)B Substrates
The [111]B direction is the preferred growth direction for GaAs NS-SAG.
Though, in this work, GaAs nanowires are grown on GaAs substrates, growth on InP
(111)B substrates shows similar results. Figure 4-2 are the results of a GaAs
nanowire array with 200 nm center-to-center spacing on GaAs (111)B with respect
to growth temperature and AsH
3
partial pressure. Tapering and lateral growth are
not noticeable by SEM measurement and are not relevant to growth parameters such
as the AsH
3
partial pressure or growth temperature; however, the array uniformity is
affected by the growth temperature and AsH
3
partial pressure. In Figure 4-2, we can
find that the array uniformity is degraded by localized growth suppression. This
growth suppression becomes larger as growth temperature increases or as the AsH
3

partial pressure decreases. An interesting fact to note is that the growth rate is
115
affected by the AsH
3
partial pressure. In planar film growth on (001) substrates, the
growth rate is governed by the group III partial pressure if the V/III ratio is greater
than 1, while growth of GaAs nanowires is affected by the AsH
3
partial pressure.
The circled region in Figure 4-2 (a) shows a growth disorder in the other direction.
This abnormal growth is caused by excessive net As supply at the top of the
nanowire, which indicates that the supply and adsorption of As are far greater than
the thermal desorption of As due to the low growth temperature. Low AsH
3
supply
or high growth temperature removes this growth behavior.
Figure 4-3 is a high magnification view of the localized growth suppression
with respect to the opening diameters. This micrograph clearly represents the
localized growth suppression. At the early stage of GaAs nanowire growth, the
overall As supply is very limited in low AsH
3
partial pressure conditions due to the
small area of bare semiconductor surface in the nano opening array. In the early
stage of growth, openings are the only locations for heterogeneous AsH
3
pyrolysis.
If there is a localized As deficiency at a site, the nucleation and forming wires at that
site is retarded. Also, Ga atoms also diffuse out from the site and migrate where
there is more As supply. This surface diffusion further retards the growth. These
two effects cause localized growth suppression. This is why the localized growth
suppression easily happens at high growth temperature and low AsH
3
partial pressure.
The growth suppression is stronger if the opening diameter is smaller. Because the
AsH
3
deficiency is inversely proportional to the area of opening, it is hard for smaller
openings to develop nucleation sites at the center. The edge driven nucleation
116
exposes crystal planes which are not favorable for nanowire growth. Figure 4-3 (b)
shows that the growth seed in the opening expose higher index planes rather than
{111}B. This abnormal nucleation in the small opening suppresses the growth
further.

(a) (b)

(c)
Figure 4-2 SEM micrographs of GaAs nanowires grown on GaAs (111)B substrates.
The center-to-center spacing of the nanowire array is 200 nm and opening
diameter is 50 nm. Growth temperature and AsH
3
partial pressure changes
from (a) 650 C, and 1.00×10
-4
atm, (b) 700 C and 2.00×10
-4
atm, and (c)
750 C and 2.00×10
-4
atm. Circled region shows abnormal growth.
117

(a) (b)
Figure 4-3 High magnification SEM micrographs of localized growth suppression in
GaAs nanowire arrays. Opening diameter shown here are (a) 80 nm and
(b) 40 nm, respectively.
Reducing the growth temperature and increasing AsH
3
partial pressure causes
another disorder in array formation. Figure 4-4 is a GaAs nanowire array grown at
650 C and 1.00×10
-4
atm of AsH
3
partial pressure. In Figure 4-4 (a), there is strong
lateral growth over multiple openings and the lateral growth buries some nanowires.
This epitaxial burial occurs easily in densely packed GaAs nanowire arrays. The
arrays shown in Figure 4-4 (a) and (b) are grown under the same growth conditions.
The difference between two is packing density, or in other words, center-to-center
spacing. Epitaxial burial only occurs in the dense array and the array with large
spacing shows uniform nanowire growth.
118

(a) (b)
Figure 4-4 Growth behavior difference of GaAs nanowire NS-SAG caused by array
packing density. Both templates are grown at 650 C. Center-to-center
spacing of two templates is (a) 80 nm and (b) 200 nm, respectively.
According to vapor phase diffusion theory of SAG[22-24], dense arrays have
lower the precursor concentrations. This seems to contradict the high AsH
3

precursor concentration for the epitaxial burial. Actually, this is not contradictory,
because the lower precursor concentration of dense arrays is the result of the higher
adsorption rate of precursors on the pattern region. The dominant decomposition
mechanism of AsH
3
is heterogeneous pyrolysis, so the actual As supply in a dense
opening array is higher than in a sparse array. The As supply is increasing as
nanowires grow. Though growth on the sidewalls is suppressed by the growth
condition, the sidewalls are still a good place for AsH
3
pyrolysis. Because of the
large semiconductor surface area of dense nanowire arrays, the effective As supply
in the arrays is higher than estimation by vapor phase diffusion. This excessive As
supply weakens the growth suppression and lateral growth occurs.
119
To fabricate uniform GaAs nanowire arrays, the localized growth suppression
and epitaxial burial must be eliminated. Apparently the two cannot be achieved at
the same time, because the growth condition suppressing one effect enhances the
other. Localized As deficiency affects the growth at the beginning and epitaxial
burial appears after nanowires grow sufficiently. If the AsH
3
flow decreases
gradually during the growth, both array formation disorders can be reduced. Figure
4-5 (a) is the tested graded AsH
3
flow profile and Figure 4-5 (b) and (c) are the
results. The two effects are effectively suppressed at the same time.
HR-TEM studies reveal the grown GaAs nanowires in the [111]B direction
are ZB crystal. As shown in Figure 4-6 (a), the grown GaAs nanowires have many
SFs in the structure. Rapid SF generation streaks out the SAED pattern (Figure 4-6
(b)). SF density ranges from 200 m
-1
to 400 m
-1
. Figure 4-7 shows that the SF
density is a function of TMG partial pressure and temperature. GaAs nanowires
grown at 750 C and 3.79×10
-7
atm of TMG partial pressure (Figure 4-7 (a)) have the
least SF density. The average SF repetition period is 12.7 monolayers (MLs). If the
growth temperature drops to 650 C in same TMG partial pressure, average SF
repetition period becomes 4.2 MLs as shown Figure 4-7 (b). High TMG flow also
increases SF generation. According to Figure 4-7 (c), the average SF repetition
period becomes 5.4 MLs if TMG partial pressure is doubled from the Figure 4-7 (a)
case.

120

0 300 600 900 1200 1500 1800
8
9
10
11
12
13
14
15
16
17
18
19
20

AsH
3
flow (sccm)
Growth time (s)
8.0x10
-5
1.0x10
-4
1.2x10
-4
1.4x10
-4
1.6x10
-4
1.8x10
-4
2.0x10
-4
2.2x10
-4
AsH
3
partial pressure (atm)

(a)

(b) (c)
Figure 4-5 AsH
3
flow modulation growth for uniform array growth. (a) AsH
3
flow is
modulated stepwise. The center-to-center spacing of array is (b) 80 nm
and (c) 200 nm, respectively.

121

(a)

(b)
Figure 4-6 HR-TEM observation of GaAs nanowire in the [111]B direction. (a) A
lattice image. The [111]B direction is upward direction of image. (b)
SAED pattern of grown nanowire.
122

(a) (b)

(c)
Figure 4-7 SF density change of 3 GaAs nanowires in the [111]B direction grown in
different growth conditions. The nanowires are grown under different
temperature and TMG partial pressure. Tried temperature and TMG
partial pressure are, (a) 750 C and 3.79×10
-7
atm (b) 650 C and
3.79×10
-7
atm, and (a) 750 C and 7.58×10
-7
atm.

123
4.3. Growth of InAs Nanowire Arrays
InAs nanowire NS-SAG shows similar growth kinetics as GaAs nanowire
NS-SAG, but the diffusion length of In is much shorter than that of Ga. Because of
the short diffusion length, nanowire growth is less affected by the diffusion of
precursors. Similar to planar InAs epitaxial growth, the growth temperature of InAs
nanowire NS-SAG is lower than that of GaAs or InP nanowire growth. The growth
temperature ranges from 550 C to 600 C. Solution type TMI made by Epichem
(now SAFC Hightec) and AsH
3
are used for the group III and V precursors. The
bubbler temperature and pressure is set to 30.1 C and 900 psi, respectively. The
TMI flow is fixed at 5 sccm, which is equivalent to 3.00 ×10
-7
atm. The AsH
3
flow
can be set between 3.5 sccm and 7 sccm. The SGC of InAs nanowire array is 5 sccm
of TMI flow and 7 sccm of AsH
3
flow at 550 C. The partial pressure of AsH
3
in the
SGC is 1.00×10
-4
atm. The V/III ratio is 333.3. In this work, InAs nanowire growth
are attempted on non-native substrates, such as GaAs and InP. The pre-surface
treatment conditions are the same as GaAs or InP nanowire NS-SAG.
4.3.1 Growth of InAs Nanowire Arrays on (111)A Substrates
It has been reported that the preferred growth direction of InAs nanowires
grown by NS-SAG is [111]B direction[3]. The growth behavior of InAs nanowires
is similar to GaAs nanowire NS-SAG. Figure 4-8 is the GaAs nanowire NS-SAG
124
on a GaAs (111)A substrate and InP (111)A substrate. There are no major
differences between two different substrates. Non-uniform nanowire array growth
occurs. The surface morphology is similar to GaAs nanowires on GaAs (111)B
substrates. Almost all of the openings develop crystals. The occupation is even
better than GaAs, because of the short surface diffusion length of In. In adatoms
tend to stay in the openings and participate in the growth rather than diffusing out.
Nanowires are tilted toward the nearest <111>B or <112>B directions, which
indicates that the preferred growth polarity of InAs NS-SAG is B polar. The
mechanism of substrate selection will be discussed in chapter 5.

Figure 4-8 InAs nanowire NS-SAG on a (111)A substrate.
4.3.2 Growth of InAs Nanowire Arrays on (111)B Substrates
As GaAs nanowire NS-SAG, the preferred growth direction of InAs NS-SAG
is also the [111]B direction. Figure 4-9 is the surface morphology of InAs nanowire
arrays with 200 nm center-to-center spacing on GaAs (111)B with respect to growth
temperature. From SEM observations, there is no major tapering or widening on the
125
bottom side of the nanowires. The growth rate is much faster than for GaAs or InP
nanowires. This shows high incorporation of In atoms in the opening. An InAs
nanowire array with smaller openings has strong growth rate fluctuations. Figure 4-
10 illustrates the relation between growth rate fluctuation and opening diameter,
which depicts the effect of surface diffusion (SD). The growth rate equation of
nanowire NS-SAG is expressed as

   
 y x u
r
t y x rL
g
dt
t y x dL
,
, , 2
1
, ,
2 0






 


 (4-1)
Here, L, g
0
, , r, and u are respectively length of nanowire at the location (x,y) and
time t, growth rate at the bare semiconductor region, incorporation ratio of surface
diffusion, opening diameter and precursor concentration at (x,y). They are defined in
detail at chapter 5. The first term on the right side is the contribution of VPD and the
second term is the contribution of SD. The surface diffusion term is proportional to
the area of sidewalls and inversely proportional to the area of the top surface. The
simplified equation is

 
  y x u t y x L
r
g
dt
t y x dL
, , ,
2
1
, ,
0






 

(4-2)
From equation (4-2), it is clear that the SD term is inversely proportional to the
diameter of opening. Due to the SD term, the growth rate of the nanowire becomes
an exponential function. If there is small non-uniformity in the nucleation step at the
early stage of nanowire formation, it is amplified by the SD term and causes large
fluctuations of individual nanowire lengths in an array. This is actually the same
126
mechanism as localized growth suppression of GaAs nanowire arrays. Since the
diffusion length of In is shorter than that of Ga and In incorporation is stronger,
strong suppression does not occur in InAs nanowires, but growth rate fluctuation
occurs.

(a) (b)

(c)
Figure 4-9 InAs nanowire arrays on GaAs (111)B substrates. The center-to-center
spacing is fixed at 200 nm. The opening diameter is also fixed at 50 nm.
These arrays are grown at (a) 550 C, (b) 600 C, and (c) 650 C,
respectively.
127

(a) (b)
Figure 4-10 Growth rate fluctuation with respect to the opening diameter. The center-
to-center spacing is fixed at 200 nm. (a) 50 nm, and (b) 85 nm are chosen
to contrast the effect of surface diffusion term.
HR-TEM observations indicate that the grown InAs nanowires in the [111]B
direction are in the ZB crystal form with dense SFs. Figure 4-11 shows the bright
field TEM image of a grown InAs nanowire. The SF density is 400 m
-1
or higher,
which means SF repetition period is approximately 3 ML per SF. This fast SF
generation in InAs nanowires fabricated by NS-SAG has been reported
elsewhere[25]. Compared to GaAs nanowires, the SF density of InAs nanowires
cannot be steered by growth conditions, such as growth temperature and AsH
3
partial
pressure. The growth temperature of InAs is too low to modify the surface
reconstruction of (111)B surface which generates SFs. This mechanism will be
discussed in chapter 5.

128

(a) (b)

(c)
Figure 4-11 Bright field HR-TEM images of an InAs nanowire grown in the [111]B
direction. (a) Low magnification over view looks similar to GaAs
nanowires. Upward direction is the [111]B direction. (b) Lattice image
shows rapid SF generation. Measured SF period is 2.4 MLs. (c) SAED
pattern shows streaks due to dense SFs.
4.4 Growth of Ternary Compound Nanowire Arrays
As discussed in previous chapters, the preferred growth direction of InP
nanowires is opposite to GaAs, InAs or GaP nanowires. If InP is mixed in a ternary
129
compound, transition of the preferred growth direction may occur. Other ternary
materials such as InGaAs nanowires have been demonstrated without the transition
of the preferred growth direction and growth suppression[26]. In order the see the
growth behavior change caused by the conflict in the preferred growth direction of
ternary components, InAsP is selected. InAsP can be used to clearly contrast the
growth behavior between InAs and InP. For InAsP nanowire growth, the growth
temperature and TMI partial pressure is fixed at 650 C and 4.20×10
-7
atm,
respectively. Two different PH3/AsH
3
ratios, 1.17 and 9, are used to change the
composition of the InAsP. The growth on InP (111)A substrate gives clusters
regardless of growth parameters. InP nanowire growth is easily effected by the
incorporation of other elements, and As incorporation entirely disordered the growth
behavior. On the other hand, growth trials on InP (111)B substrates generate
nanowires. Figure 4-12 and Figure 4-13 illustrate the effect of the preferred growth
direction conflict on InP (111)B. As expected, as the composition of InP increases,
growth suppression appears, which is exactly same behavior as InP growth on InP
(111)B.
This growth suppression is also related to the opening diameter. As shown in
Figure 4-12 (a) and Figure 4-13 (a), smaller opening diameters show stronger growth
suppression. On InP (111)B polar substrates, InAsP nanowire growth must be driven
by InAs growth whose preferred growth direction is same as the polarity of the
substrate. In the array of small openings, As deficiency is caused by high
temperature and low AsH
3
partial pressure. This mechanism is same as the localized
130
growth suppression of GaAs nanowires. It is challenging to achieve high P
composition InAsP if the fill factor of the pattern is too low.
Figure 4-12 (d) and Figure 4-13 (d), depict growth suppression in the
nucleation sites with WZ facet orientation. It may be possible that these locations
have higher P composition which causes growth suppression, but further study is
required.

(a) (b)

(c) (d)
Figure 4-12 SEM bird view (45  tilt angle) micrographs of InAsP nanowires grown
at 1.17 of PH
3
/AsH
3
ratio. The center-to-center spacing is fixed at 200 nm.
The diameters of openings are respectively (a) 50 nm, (b) 60 nm, (c) 80
nm, and (d) 160 nm.
131

(a) (b)

(c) (d)
Figure 4-13 SEM bird view (45  tilt angle) micrographs of InAsP nanowires grown
at 9.00 of PH
3
/AsH
3
ratio. The center-to-center spacing is fixed at 200 nm.
The diameters of openings are respectively (a) 50 nm, (b) 60 nm, (c) 80
nm, and (d) 160 nm.

132
4.5 Growth of InP/InAs heterostructure Nanowire Arrays
As discussed in the previous section, the preferred growth direction mismatch
can be a problem in ternary nanowire synthesis. This also causes many challenges in
InP containing heterostructure nanowire. As mentioned in section 4.1, to take full
advantage of reduced misfit dislocation generation in nanoheteroepitaxy, vertical
tandem structures must be possible. In addition, suppressing growth on the sidewall
is challenging in NS-SAG. Therefore, the possibility to grow InP containing
heterostructures must be tested. In this work, InP/InAs heterostructures are tested.
Because InP nanowire NS-SAG can be affected by the incorporation of other
elements, InAs nanowires are grown on the top of InP nanowire arrays grown on InP
(111)A substrates. The growth conditions of InP and InAs nanowire are the SGC of
those materials. Figure 14 (a) is an InP nanowire array which is grown at the same
growth condition and growth time as the bottom side of the heterostructure nanowire
and figure 14 (b) is the surface morphology of InP/InAs heterostructure nanowires.
From the diameter of the nanowires, sidewall growth occurs. There are many bent
nanowires, which indicate non-uniform InAs sidewall growth. Because of strain
between InP and InAs, if the growth on the sidewall is not uniform, the nanowire is
bended to minimize the strain energy accumulation. This bending has been reported
by Fukui[20]. Figure 14 (c) illustrates InAs nanowire grown on the top of InP
nanowires, but only a few nanowires have this type of heterostructure. The top
GaAs nanowires are grown in the preferred growth direction, the nearest <111>B
133
direction. As shown in figure 14 (d), tripod structures develop on the top of InP
nanowires. Based on these observations, matching the preferred growth direction is
essential for vertical tandem heterostructure nanowire growth using NS-SAG.

(a) (b)

(c) (d)
Figure 4-14 InP/InAs heterostructured nanowire arrays grown by NS-SAG. (a) InP
nanowire array are grown for 10 minute for bottom part of
heterostructured nanowires. (b) After InAs growth, the nanowires are bent
by non-uniform InAs growth on InP nanowire sidewalls. (c) Vertical
InP/InAs heterostructures are also formed. InAs part are tilted toward the
nearest <111>B direction. (d) An InAs tripod structure is grown on the
top of an InP nanowire.
134
4.6 Summary
GaAs NS-SAG conditions are optimized to achieve uniform and vertical
nanowire arrays. The SGC temperature of GaAs NS-SAG is 700 C. The SGC
precursor partial pressure is 3.79 ×10
-7
atm for TMG and 2.42×10
-4
atm forAsH
3
.
GaAs NS-SAG on GaAs (111)A is not uniform and develops clusters, and many
nanowires are tilted toward the nearest <111>B directions. The occupation ratio of
growth cluster is greater than 75%. GaAs NS-SAG on GaAs (111)B is uniform and
vertical, which indicates that the [111]B direction is the preferred growth direction of
GaAs NS-SAG. In high growth temperature and low AsH
3
partial pressure,
localized growth suppression appears. This is caused by localized As deficiency at
the early stages of growth, because of insufficient bare semiconductor region for
heterogeneous AsH
3
pyrolysis. At low growth temperature and high AsH
3
partial
pressure, epitaxial burial of nanowires appears especially in dense GaAs nanowire
array. This is caused by excessive As supply caused by heterogeneous AsH
3

pyrolysis on the sidewalls. To eliminate both disorders, AsH
3
flow modulation is
introduced in the growth. This technique can suppress both disorders effectively.
According to HR-TEM observation results, GaAs nanowires in the [111]B direction
are in ZB crystal phase with rapid SF generation. The average SF repetition period
is a function of growth temperature and TMG partial pressure. 3.79×10
-7
atm of
TMG partial pressure condition can increase the SF repetition period up to 12.7
monolayers (MLs) at 750 C. If the growth temperature decreases down to 650 C
135
or if the TMG partial pressure doubles, the average SF repetition period becomes 4
to 5 MLs.
The SGC of InAs NS-SAG is 3.00 ×10
-7
atm of TMI partial pressure and
1.00×10
-4
atm of AsH
3
partial pressure at 550 C. The preferred growth direction is
the [111]B direction. InAs nanowire arrays with smaller opening have strong growth
rate fluctuations, which illustrates the effect of SD on the sidewalls. Because the SD
effect is inversely proportional to the radius of nano openings, larger opening arrays
show better growth rate uniformity. HR-TEM observations indicates that the grown
InAs nanowires in [111]B direction are in the ZB crystal phase with dense SFs. The
SF density is 400 m
-1
or higher, which corresponds to a SF repetition period of
approximately 3 ML per SF. The SF density of InAs nanowires cannot be steered by
growth conditions due to its low growth temperature.
InAsP NS-SAG is attempted to test the conflict between the preferred growth
directions in ternary alloys. For InAsP nanowire growth, the growth temperature and
TMI partial pressure is fixed at 650 C, and 4.20×10
-7
atm, respectively. Two
different PH3/AsH
3
ratios, 1.17 and 9 are used to change composition of InAsP in a
ternary compound. On InP (111)B substrates, as the composition of InP increases,
growth suppression appears. This growth suppression is also related to opening
diameter, which is caused by an AsH
3
deficiency in the early stage of growth.
InP/InAs heterostructure nanowire arrays are grown on InP (111)A substrates.
The growth conditions of InP and InAs nanowire are the SGC of those materials.
InAs layers are grown on the sidewalls of InP nanowires. Non-uniform sidewall
136
growth of InAs layers on the sidewalls bends entire growth structure. A few
nanowires have vertically stacked InAs structures on the top of InP nanowires The
top InAs nanowires are grown in the preferred growth direction, the nearest <111>B
direction.

137
Chapter 4 Endnotes
[1] S. Choi, Y. Chung, D. Kim, H. Chu, and K. Seo, Jpn. J. Appl. Phys. 39 2508-
2511 (2000).
[2] R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H.
Yoon, R. A. Sherif, and N. H. Karam, Appl. Phys. Lett. 90 183516 (2007).
[3] K. Tomioka, P. Mohan, J. Noborisaka, S. Hara, J. Motohisa, and T. Fukui, J.
Cryst. Growth 298 644-647 (2007).
[4] K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Nano Lett. 8 3475-3480
(2008).
[5] M. Noguchi, K. Hirakawa, and T. Ikoma, Phys. Rev. Lett. 66 2243 (1991).
[6] L. Ö. Olsson, C. B. M. Andersson, M. C. Håkansson, J. Kanski, L. Ilver, and
U. O. Karlsson, Phys. Rev. Lett. 76 3626 (1996).
[7] J. W. G. Wildöer, C. J. P. M. Harmans, and H. van Kempen, Phys. Rev. B 55
R16013 (1997).
[8] H. Hasegawa and M. Akazawa, Appl. Surf. Sci. 255 628-632 (2008).
[9] Shadi A. Dayeh, D. P. R. Aplin, X. Zhou, P. K. L. Yu, Edward T. Yu, and D.
Wang, Small 3 326-332 (2007).
[10] S. Yeon and K. Seo, Jpn. J. Appl. Phys. 47 2868-2871 (2008).
[11] J.-W. Lee, S.-W. Kim, K.-S. Seol, Y. Kwon, and K.-S. Seo, Microwave Opt.
Tech. Lett. 49 609-612 (2007).
[12] S. J. Choi, Z. Peng, Q. Yang, E. H. Hwang, and P. D. Dapkus, IEEE
Photonics Technol. Lett. 17 2101-2103 (2005).
138
[13] F. E. Ejeckam, C. L. Chua, Z. H. Zhu, Y. H. Lo, M. Hong, and R. Bhat, Appl.
Phys. Lett. 67 3936-3938 (1995).
[14] A. Srinivasan, S. Murtaza, J. C. Campbell, and B. G. Streetman, Appl. Phys.
Lett. 66 535-537 (1995).
[15] B. Hua, J. Motohisa, Y. Kobayashi, S. Hara, and T. Fukui, Nano Lett. 9 112-
116 (2008).
[16] H. Goto, K. Nosaki, K. Tomioka, S. Hara, K. Hiruma, J. Motohisa, and T.
Fukui, Appl. Phys. Express 2 035004 (2009).
[17] A. S. Jordan and A. Robertson, J. Vac. Sci. Techol. A, 12 204-215 (1994).
[18] G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, 2nd ed. San Diego:
Academic Press, 1999.
[19] M. Inari, J. Takeda, J. Motohisa, and T. Fukui, Physica E 21 620-624 (2004).
[20] P. Mohan, J. Motohisa, and T. Fukui, Appl. Phys. Lett. 88 013110 (2006).
[21] J. Motohisa, J. Noborisaka, J. Takeda, M. Inari, and T. Fukui, J. Cryst.
Growth 272 180-185 (2004).
[22] W. G. Oldham and R. Holmstrom, J. Electrochem. Soc. 114 381-388 (1967).
[23] O. Kayser, J. Cryst. Growth 107 989-998 (1991).
[24] E. Colas, C. Caneau, M. Frei, J. E. M. Clausen, W. E. Quinn, and M. S. Kim,
Appl. Phys. Lett. 59 2019-2021 (1991).

139
[25] K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Jpn. J. Appl. Phys. 46
L1102 (2007).
[26] M. Akabori, J. Takeda, J. Motohisa, and T. Fukui, Nanotechnology 14 1071-
1074 (2003).

140
Chapter 5: Growth Models of NS-SAG and Electron Beam
Induced Bundling
5.1 Introduction
5.1.1 Models of Selective Area Growth
In SAG, the presence of a dielectric mask on a substrate changes modulates
the adsorption of precursors. The mask also changes the diffusion length of cracked
reaction species such as Ga or As. Because the dielectric mask is not a favorable
place for growth to occur, the species on the mask region can diffuse longer before
ultimately forming crystals. If only very small amount of crystal formation occurs,
the surface diffusion length of precursors can assume to be approximately infinite.
The changes in adsorption and diffusion modulate the spatial precursor concentration
profile and cause local growth enhancement, non-uniform lateral thickness variations
and spatial composition variations of ternary and quaternary compound
semiconductor materials. As a result, a SAG model is required to estimate the
growth rate and composition of ternary and quaternary materials.
There are two major and widely used theories for SAG models. In surface
diffusion (SD) theory[1], precursor molecules are adsorbed on the mask and diffuse
to the bare semiconductor region because no growth reactions occurs on the mask.
141
On the surface of the bare semiconductor region the diffusion length is finite because
the precursors are consumed by forming epi-layers. The other theory is the vapor
phase diffusion (VPD) theory[2-4]. On the mask region, the concentration of vapor
phase precursors which are diffusing through the stagnant boundary layer is higher
than that of the bare semiconductor region, since there are no precursor adsorption
events on the mask. Whereas on the bare semiconductor region, the precursor
concentration drops due to adsorption events on the surface caused by crystal growth.
This precursor concentration difference drives vapor phase precursors on the mask
regions to diffuse into the bare semiconductor region. Then, the edge region shows
growth enhancement and composition change due to excessive precursor supply In
the VPD model, the vapor phase precursors at the surface of the wafer accumulate on
the dielectric mask regions. Either of the theories can model growth enhancement at
the mask edge region, since two diffusion mechanisms are not perfectly
distinguishable. Therefore, a diffusion model based on the VPD theory internally
has SD effects and vice versa.
Choi developed a SAG model based on the VPD theory[5]. The vapor phase
diffusion model is based on the stagnant layer theory which states that the net
velocity of molecules is zero which can be expressed by the Laplace equation.
Solving the Laplace equation with boundary conditions determined by patterning can
give a spatial precursor profile. Figure 5-1 is the simulation window and boundary
condition definitions in a two-dimensional (2D) case. In the simulation, the
142
precursor concentration N(x,y) satisfies 2D Laplace equation as expressed in
Equation (5-1)
 0 ,
2
2
2
2














y x N
y x
D
 
kN
y
y x N
D  

 ,  
0
,



y
y x N
D
 
0
,



y
y x N
D

Figure 5-1 Schematic diagram of two-dimensional simulation domain and boundary
conditions[5].
 0 ,
2
2
2
2














y x N
y x
D (5-1)
Here, D is the VPD coefficient. The top boundary is the end of the stagnant layer, if
the precursor supply from the outside of the stagnant layer is constant. This can be
modeled as a Dirichlet boundary condition shown in (5-2).
 
s
N h x N , (5-2)
, where h and u
s
are the height of the simulation window and precursor concentration
at the top of the stagnant layer. On the sidewall of windows, there is no precursor
consumption across the boundaries. Transparent boundary conditions can be used
for these.

   
0
, ,
0






  w x x
x
y x N
D
x
y x N
D (5-3)
143
w is width of the simulation window. The boundary condition of mask region is
determined by Fick’s first law of diffusion, which relates the diffusion flux to the
concentration field, by postulating that the flux goes from regions of high
concentration to regions of low concentration, with a magnitude that is proportional
to the concentration gradient. This can be expressed as
 y x N
y x
D J ,













 (5-4)
, where J is flux of diffusion. Because there are no adsorption events on the mask,
there is no concentration difference or diffusion in the surface normal direction, y.
Therefore, the boundary condition of the mask region is

 
0
,
mask , 0



 y
y
y x N
D (5-5)
The boundary condition of the bare semiconductor region is determined by
Fick’s first law of diffusion and the Langmuir isotherm[6]. The Langmuir isotherm
is a semi-empirical adsorption model derived from a proposed kinetic mechanism. It
is based on four assumptions:
#1. The surface of the adsorbent is uniform, that is, all the adsorption
sites are equivalent.
#2. Adsorbed molecules do not interact.
#3. All adsorption occurs through the same mechanism.
144
#4. At the maximum adsorption, only a monolayer is formed:
molecules of adsorbate do not deposit on other, already adsorbed,
molecules of adsorbate, only on the free surface of the adsorbent
Assume an adsorption reaction, S
e
+P
g
SP, where S
e
, P
g
, SP is an empty adsorption
site, a vapor phase precursor, and captured precursor in solid form, respectively. In
the equilibrium, the reaction rate constant k can be defined by

 

g e
P S
SP
k  (5-6)
Define surface coverage, θ, as the fraction of the adsorption sites occupied, [SP] is
proportional to θ, [S
e
] is proportional to (1- θ), and P
g
is proportional to the
concentration, N. Therefore,

N
k




1
(5-7)
Reorganize Equation (5-7) for θ to find

kN
kN


1
 (5-8)
If the concentration is low enough, Equation (5-8) becomes
kN   (5-9)
In equilibrium, the flux J must be equal to surface coverage, , because the surface
coverage is equal to consumption of precursors by adsorption. Then, according to
Fick’s first law, the boundary condition of the bare semiconductor region is defined
as
145

 
 0 ,
,
bare , 0
x kN
y
y x N
D
y
 



(5-10)
In the VPD model, the parameter D/k has a similar meaning to the lateral diffusion
length of precursor molecules. Choi set binary compounds such as InP or GaAs as
single molecules and empirically measured the diffusion length of each binary
compound by curve fitting[5].
5.1.2 Surface Reconstructions
Crystal growth is strongly affected by surface kinetics in the crystal
formation reaction. Moreover, in NS-SAG, many different crystal planes are
involved in growth, so understanding crystal surfaces is very important to understand
the growth behavior and designing growth runs. If a crystal surface is formed by
truncating bulk material, the truncated surface has many dangling bonds. The
dangling bonds must be stabilized for the stability of the entire system. Pashley has
proposed an electron counting model (ECM)[7]. According to the model, in order to
minimize the formation energy of a III-V compound semiconductor crystal surface,
the dangling bonds of group III atoms should be empty and the dangling bonds of
group V atoms should be filled with two electrons. To satisfy this rule, surface
dangling bond electrons are redistributed and sometimes, the surface itself is
reconstructed. Because all elements in III-V semiconductors have sp
3
orbital
hybridization, group III atoms make 4 bond legs with 3 valence electrons, which
146
corresponds to 0.75 electrons per bond. On the other hand, group V atoms make 4
bond legs with 5 valence elections, or 1.25 electrons per bond. To follow the ECM
rule, group III atoms should lose 0.75 electrons to empty a dangling bond, while
group V atom should add 0.75 electrons to fill a dangling bond.
If a surface has equal number of group III and V atoms and the number of
dangling bond per atom is the same, situation is simple. If the 0.75 electrons in each
group III dangling bond is transferred to the group V dangling bond, the ECM rule is
satisfied. In this case, the surface does not need to be reconstructed. However, after
the electron transfer, and because of charge imbalance between the group III and V
atom, the position of each atom is changed. This is called surface relaxation. The
{110} surfaces of ZB crystal have surface relaxation, because they consist of equal
number of group III and V atoms and each surface atom has one dangling bond.
The {111} planes of the ZB crystal and {0001} planes of the WZ crystal are
composed of one kind of atoms, so these surfaces must be reconstructed to satisfy the
ECM rule. Because the ECM has no unique solution, many reconstructions are
possible. Actually, many different types of reconstructions are reported by electron
diffraction or scanning tunneling microscopy. The ECM rule can be used to
understand these reconstructions. The surface reconstructions are commonly
expressed by Wood’s notation. The format of Wood’s notation is X(hkl)-(m×n)R ,
which describes the reconstruction of the (hkl) plane of X material. m and n are the
multipliers of the surface unit cell. This indicates the translation vector of surface is
m and n times larger than the a and b vectors, respectively, which are the translation
147
vectors of the bulk lattice. R  indicates the super cell is rotated by the angle . This
notation is often used to describe reconstructions concisely, but does not directly
indicate changes in the layer symmetry (for example, square to hexagonal).
Because III-V semiconductors are not centrosymmetric, in other words,
(hkl) ( h ¯ k ¯ l ¯ ), the (111)A and (111)B, which are (111) and (1 ¯ 1 ¯ 1 ¯ ), have different
polarity. It has been reported that the possible surface reconstructions of GaAs (111)
surface are (2×2) and ( 19× 19) reconstructions, regardless of polarity[8-9]. (2×2)
reconstruction is the dominant reconstruction of GaAs (111) surface in MBE or
MOCVD chambers[8]. In MOCVD chambers, GaAs(111)A-(2×2) reconstruction
can be achieved by two different phase, group III vacancy and group V trimer. The
amount of group V precursor overpressure determines the phase of the (2×2)
reconstruction. In As rich conditions, the surface forms (2×2) reconstruction with As
trimers[10]. In Ga rich conditions, the surface is reconstructed by (2×2) with Ga
vacancies[11]. GaAs(111)B reconstruction also has two different phases. In As rich
condition, it forms (2×2) reconstruction with As trimers[8]. In Ga rich conditions,
the surface forms ( 19× 19) reconstructions[8]. Figure 5-2 illustrates these
reconstructions. The reconstructions can be checked by the ECM rule. In
GaAs(111)B-(2×2) with As trimers as shown in Figure 5-2 (b), one surface unit cell
has 4 surface As atoms. Three form an As trimer structure with 3 As adatoms.
There are 4 dangling bonds in the surface unit cell. One is from the surface atom
without forming the trimer. The other three are from each As atom in the trimer.
When As-As bonds are formed for a trimer formation, each bond has 0.5 of
148
excessive electron, because an As bond arm has 1.25 electrons. In totality 3 excess
electrons are generated when forming an As trimer. To stabilize 4 dangling bonds, 4
of the 0.75 electron contributions are needed, in other words, 3 electrons in total to
fill the dangling bonds. If 3 of excess electrons from the bond are redistributed into
the dangling bonds, all dangling bond are filled with 2 electrons and the ECM rule is
satisfied. Consequently, the GaAs(111)B-(2×2) reconstruction with As trimers
obeys the ECM rules. The other (2×2) reconstructions also obey the ECM rules.

(a) (b)

(c)
Figure 5-2 Schematic diagrams of surface reconstructions, (a) GaAs(111)A-(2×2), (b)
GaAs(111)B-(2×2), and (c) InP(111)A-( 3× 3)R30 . Red dotted lines are
surface unit cell boundaries.
149
The InP (111)A surface shows different surface reconstruction. In In rich
conditions, (2×2) reconstruction with In vacancy like GaAs(111)A-(2×2) with Ga
vacancy is formed[12]. In P rich conditions, ( 3× 3)R30  reconstruction with P
trimer is observed[12]. Figure 5-2 (c) shows the ( 3× 3)R30  reconstruction.
Interestingly the InP(111)A-( 3× 3)R30  reconstruction does not satisfy the ECM
rule. To explain this, it is assumed that hydrogen atoms are attached to stabilize
dangling bonds, but there has been no direct observation yet. There has been no
direct observation of the surface reconstruction of InP(111)B surface, but it is
believed that in P rich conditions, the surface is stabilized by P trimers.
5.1.3 Scanning Electron Microscopy
To characterize grown nanowire structures, electron microscopy is essential
due to the small feature size of nanowires in nanometer range. Among all electron
optic measurements, SEM plays the central role in surface morphology
characterization. Compared to TEM, SEM has the advantage for easy sample
preparation, beause samples can be measured as is. The resolution of SEM is limited
compared to TEM. The electron optic column of SEM is very similar to that of EBL
systems described in chapter 2. EBL systems are actually modified SEM to write
patterns. For that purpose, they have beam blankers, pattern generators and high
precision stage movement system, such as piezoelectric motors and laser
interferometers.
150
SEM utilizes the interaction between electrons and the material. If a focused
electron beam impinges on a sample surface, the penetration depth is determined by
acceleration voltage of the electron in the column. To measure the surface
morphology of a sample, the acceleration voltage should be as low as possible, but
the electron must still have enough energy to interact with the material. Because of
aberration issues, achieving a small beam spot is challenging at low acceleration
voltages, which degrades SEM resolution at low acceleration voltage. Figure 5-3
shows the interaction between high energy electrons and a specimen. Secondary
electrons (SE) are generated by inelastic collisions between atoms and electrons.
This interaction is dominant if the primary electrons, which are from the electron
beam directly hitting the specimen surface, have low kinetic energy. Back scattered
electrons (BSE) are generated by elastic collision. This interaction is dominant if the
energy of the primary electrons is high. In SEM measurement at a few kV
acceleration voltage, SE is dominant, and SE detectors deliver better contrast
compared to BSE detectors.
Because an SEM image is constructed by detecting electrons, the image may
have artifacts. If a specimen is a dielectric, charging effects can be seen. Because
the ionization energy of dielectric materials is too high, the electron beam cannot
generate secondary electrons and instead makes the material negatively charged.
Because the electric field of charged specimen distorts the electron beam and
emission from other parts, the image of such a specimen is very dark and shows
rapid drift or distortion. Edge effect is another artifact. Sharp corners have better
151
electron emission due to the high field emission efficiency at the corners, sharp
edges on a specimen look brighter than other parts. Therefore, consideration should
be taken of the SEM images, as they can have some non-physical artifacts.

Figure 5-3 Origin and information depth of secondary electrons (SE), backscattered
electrons (BSE), Auger electrons (AE) and X-ray quanta (X) in the
diffused cloud of electron range R for normal incidence of the primary
electrons (PE)[13].
5.1.4 Chapter Overview
In this chapter, three different growth models and observation will be
discussed. As shown in the previous section, there are many macroscopic SAG
models and they show great correspondence with measured results. However, the
growth model of NS-SAG has not been reported. Because of the complex nature of
crystal facets in nanowires and their different growth behaviors, a dedicated model
for NS-SAG is required to estimate growth rate and other growth behaviors. In this
152
chapter, a growth model for NS-SAG is proposed. The model considers VPD and
SD on the sidewalls at the same time to increase accuracy. The average modeling
error of the model is less than 9% by virtue of the combination of two diffusion
mechanisms.
In chapter 3 and 4, it is reported that the NS-SAG behavior is strongly
affected by the polarity of substrates. Here, a model for dependencies as preferred
growth direction or SF generation on the polarity of substrates is proposed. These
dependences are caused by difference in surface reconstruction and chemical bond
strength of the III-V semiconductors. Group V trimer structures determine the ease
of growth on a surface and location of the adatom adsorption. This steers growth
promotion or suppression and determines the stacking sequences.
Lastly, strong mechanical deformation of nanowires under focused electron
beam scanning is investigated. Under electron beam, nanowires in an array are
observed to bundle with each other. In this work, this is referred to as nanowire
bundling. This bundling occurs in InP and GaAs nanowire arrays, regardless of
doping. The bundling is affected by the length and diameter of the nanowires
because those factors determine the ease of mechanical deformation or bending.
When this bundling occurs, the SEM images are strongly distorted and show strong
artifacts. However, this bundling never occurs in InAs nanowire arrays. In this
chapter, these differences will be explain this with surface state changes induced by
an electron beam.

153
5.2 Diffusion Model for NS-SAG
5.2.1 Experiments

Before a growth model for NS-SAG is developed, the macroscopic SAG
behavior in the growth conditions of NS-SAG must be checked, because the growth
behavior of NS-SAG and macroscopic SAG are closely related. Figure 5 is the test
pattern for macroscopic SAG in NS-SAG conditions and their results. Simple trench
patterns are located in the [112] direction. The width of the trench for InP and InAs
is 15 m and the width for GaAs is 1 mm, because the lateral diffusion length of Ga
is far longer than that of In. Here, the VPD based model for Choi is used to
characterize the diffusion[5]. In the SGC, the growth rate and lateral diffusion length,
(D/k) of InP, and GaAs are 5 nm/min and 40 m for InP, 2.1 nm/min and 181.5 m
for GaAs, respectively. Because the surface of InAs growth on GaAs or InP is
uneven due to the lattice mismatch, extracting the diffusion length of InAs is
challenging.
In order to see the macroscopic diffusion effect in a nanowire array island, a
1 mm × 1 mm size opening array is defined on InP (111)A. The center-to-center
spacing and opening diameter of the array is fixed at 200 nm and 90 nm respectively.
Some of nano opening arrays have different skirt lengths, which is the unpatterned
dielectric mask region in the perimeter of a nano opening array. The chosen skirt
154
widths are 150 m, 250 m, and 350 m. Figure 5-4 shows grown InP nanowire
arrays on the large scale nano opening array with a 175 m wide skirt. It is clear that
the nanowires at the edge are longer than the center ones, and this represents the
precursor diffusion from the skirt region. The arrays with no skirt have shorter
wires at the edges as shown in Figure 5-5, because the precursor concentration at the
edge region is lowered by diffusion from the array into the bare semiconductor
region.
350 m
1 mm
0.5 mm
Skirt
Patterned
Region
0
x
x=0
x=0.15 mm
x=0.5 mm

Figure 5-4 SEM images of an InP nanowire array with 350 m wide skirt grown by
NS-SAG with respect to the location in the array. Scale bars are 200 nm.
155

Figure 5-5 Surface morphology of InP nanowires at the edge of array. The array has
no skirt. Because of diffusion toward bare semiconductor region, The
growth is suppressed.
5.2.2 Theoretical Model
It is challenging to estimate the NS-SAG behavior because of the
complicated diffusion processes induced by the nano-patterned dielectric mask.
There are several examples [14] for this estimation; however, most of these are
empirical models and cannot give any information about the growth behavior. The
VPD model has been widely used to describe macroscopic SAG as described in the
previous section. The VPD model can be used for NS-SAG of nanowire arrays. The
VPD model is based on the stagnant layer theory, which assumes that the net
molecule velocity is zero. The simulation of the VPD model is as described in the
previous section. Figure 5-6 (a) is a schematic of the VPD model for NS-SAG. If
this scheme is applied to modeling nanowire array growth by NS-SAG, in the entire
simulation, nanometer size grid resolution is required over millimeter wide area to
describe a nano opening array. Due to the large simulation size, the VPD model of
156
nanowire array is difficult to perform on a typical PC. On the contrary, the diffusion
process is insensitive to rapid fluctuations of the adsorption if the periodicity of the
fluctuations is much shorter than the lateral diffusion length, D/k, which is of
micrometer scale. This allows the complicated nano-patterned region to be viewed
as a large area of a new semiconductor region whose adsorption is the weighted
average rate of the bare semiconductor region and the dielectric mask as shown in
Figure 5-6 (b). This is the average adsorption approximation. In this approximation,
the new lateral diffusion length of the nano opening array, (D/k)
new
becomes the
averaged diffusion length defined in (5-11)

  
  
1 1
1
2
1
2
1
/ 1 /
/ 1 / /
 
  
  








  
mask semi
mask semi new
k D f k D f
k D
A
r
k D
A
r
k D
 
(5-11)
, where (D/k)
semi
, (D/k)
mask
, A, f and r are diffusion length of the bare semiconductor
region and mask region, the unit cell area of the nano opening array which contains 1
opening, fill factor of the array and the radius of the opening, respectively. If the
diffusion length of mask region is assumed to be infinite, (D/k)
new
can be expressed
in a simple way.
  
1 1
2
1
/ / /
  
 
semi semi new
k D f k D
A
r
k D

(5-12)

157

  0 , ,
2
  z y x N D
k
dz
dN
D   0 
dz
dN
D
k
dz
dN
D  
0 
dy
dN
D
0 
dx
dN
D

(a)
  0 , ,
2
  z y x N D
new
k
dz
dN
D  
0 
dz
dN
D
k
dz
dN
D  
0 
dy
dN
D
0 
dx
dN
D

(b)
Figure 5-6 Schematic diagrams of VPD simulation, (a) full VPD model without any
approximation, and (b) VPD model with average adsorption
approximation.
158
For convenience, normalized precursor concentration (NPC), u which is
normalized by the precursor concentration of the bare semiconductor region far away
from the dielectric mask is defined. Choi numerically calculated the precursor
concentration of the bare semiconductor region for normalization[5]. However, the
precursor concentration of the bare semiconductor region, N
0
can be analytically
calculated. If there is no mask in the simulation window, the VPD model becomes a
one-dimensional (1D) differential equation. The 1D Laplace equation is

 
0
2
2



y
y N
D (5-13)
The solution of Equation (5-13) is
b ay y N   ) ( (5-14)
If we apply the boundary conditions, (5-2) and (5-10) on
h N b a
s
/ ) (   (5-15)
  b a k D
semi
 / (5-16)
From (5-15) and (5-16), we get

 h k D
h N
b N
semi
S
1 0
/ 1


  (5-17)
Then u is defined by

0
/ ) 0 , , ( ) , ( N y x N y x u  (5-18)
If the array is large enough, the center region of the array is not affected by diffusion
from the outside of the array. Then, u is the only function of fill factor. If the bare
semiconductor region and mask region cannot affect each other,
159

 
 f h k D f
h k D
y x u
semi
semi
1
/ 1
/ 1
,
1
1






(5-19)
Figure 5-7 illustrates a comparison between the finite element analysis of a
full VPD model and finite element calculation with average adsorption
approximation. With this approximation, though fine details of precursor
concentration in each nano opening are missing, the macroscopic profile of the
precursor concentration is exactly reproduced, which is important in the growth rate
estimation.
Figure 5-8 is the simulation result of the InP nanowire array. The size of
array is 1 mm × 1 mm, with a 350 m wide skirt. Figure 5-9 is length estimation
with the calculation result shown in Figure 5-8. The length is underestimated,
because it ignores heterogeneous pyrolysis of precursors and diffusion of cracked
precursor molecules on the sidewalls of the nanowires. In NS-SAG, growth on the
sidewalls is suppressed by choosing suitable growth conditions. Therefore, cracked
precursor species cannot settle at the sidewalls and instead diffuse along the
nanowire sidewalls. Some of the species reaching the top surface of the nanowire
can participate in the growth and increase the growth rate. As a result, surface
diffusion of cracked reaction species on the sidewalls must be considered to estimate
the growth rate of nanowires grown by NS-SAG.
160
1000 1020 1040 1060 1080 1100 1120 1140 1160 1180 1200
1.026
1.028
1.030
1.032
1.034
1.036
1.038
w/o Average adsorption approx.
w/ Average adsorption approx.

Normalized precursor concentration
Position (nm)

Figure 5-7 Simulation result of VPD model with and without average adsorption
approximation.
The way to add the effect of surface diffusion to the VPD model is to convert
the effect to a localized increase of the precursor concentration. If u
s
, the increase of
NPC is defined, the total NPC is the sum of the concentration calculated by VPD and
u
s
. To get a simple form of u
s
, two approximations are made. The precursor
concentration along the sidewall is assumed to be uniform. Pyrolysis of the
precursors on the sidewalls does not significantly change the precursor concentration.
Under these approximations, u
s
becomes
) , (
) , , ( 2
) , , (
2
y x u
r
t y x rL
t y x u
s


  (5-20)
L is the length of nanowires. If  is 1, u
s
is simply the total amount of precursor
molecules falling on the sidewalls divided by the area of the top surface. This means
all precursor molecules falling on the sidewalls diffuse onto the top surface.  is the
incorporation ratio, which describes how many precursor molecules are cracked and
161
are incorporated in the growth on the top surface. If the growth rate at the bare
semiconductor region, g
0
, is known, the growth rate can be expressed as


) , (
) , , ( 2
1
) , , ( ) , ( ) , , (
0
0
y x u
r
t y x L
g
t y x u y x u g t y x L
dt
d
s






 
 

(5-21)
From equation (5-21), the equation of nanowire length can be derived.














 1 ) , (
2
exp
2
) , , (
0
t y x u
r
g
r
t y x L


(5-22)

0 200 400 600 800 1000 1200 1400 1600
3
4
5
6
7
8
Skirt

Normalized precursor concentration
Location( m)
Skirt Nano-patterned region

(a) (b)
Figure 5-8 Simulation result of approximated VPD model. The size of array is 1 mm
× 1 mm, with 350 m wide skirt. The diameter and center-to-center
spacing is 90 nm and 200 nm, respectively. (a) 2D concentration profile is
expressed in color. (b) 1D concentration profile of AB section clearly
shows VPD from the skirt regions.
162
0.0 0.1 0.2 0.3 0.4 0.5
500
1000
1500
2000
2500
3000

Nanowire length (nm)
Position from the end of the pattern (mm)
Measured: 350 m skirt
Model, VPD Only: 350 m skirt

Figure 5-9 Estimation of nanowire length by the VPD based model. The model only
considers VPD mechanism. The size of the tested array is 1 mm × 1 mm,
with 350 m wide skirt. The diameter and center-to-center spacing is 90
nm and 200 nm, respectively.
5.2.3 Modeling Results
In the VPD model, all of the patterning parameters such as opening diameter
and center-to-center spacing and the growth rate on the bare semiconductor region
are extracted by HR-SEM observation. (D/k)
semi
is extracted from macroscopic SAG
experiments. Extracting  in the surface diffusion part is challenging, because it is
difficult to measure the efficiency of heterogeneous cracking and incorporation, so 
is treated as a fitting parameter. For InP nanowire growth modeling, the  value is
extracted from arrays whose size is 1 mm × 1mm with a 350 m skirt. The opening
diameter and center-to-center spacing are 90 nm and 200 nm respectively. The
163
extracted  value of InP SGC is 0.16. Figure 5-10 is the modeling results of InP
nanowire arrays. The size of arrays, center-to-center spacing and opening diameter
are the same as the array for  extraction, but the skirt widths are different. As seen
in figure 10, the extracted  from an array is valid for others if the arrays are grown
at same growth condition.
This model can be used for GaAs nanowire NS-SAG. Because the later
diffusion length,(D/k)
semi
of GaAs NS-SAG is 4.5 times larger than that of InP NS-
SAG, growth enhancement effect by VPD is not significant. On the other hand, the
extracted  value of GaAs SGC is 0.64, which indicates the SD component is 4 times
larger than that of InP nanowire NS-SAG. As described in chapter 4, GaAs has
strong SD component due to the heterogeneous cracking of AsH
3
. This effect is
reflected in  value of GaAs SGC. As a consequence, GaAs nanowire NS-SAG is
mainly driven by SD effect, except for very early stage of growth.
InAs nanowire NS-SAG modeling is still challenging. Because of the large
lattice mismatch, determining macroscopic SAG characteristics is hard. Moreover, if
the opening diameter is small, the growth rate fluctuates, so it is hard to apply
average adsorption approximation. Further model development is required for InAs
nanowire NS-SAG

164
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
500
1000
1500
2000
2500
3000

Nanowire length (nm)
Position from the end of the pattern (mm)
Measured: 350 m skirt
Measured: 250 m skirt
Measured: 150 m skirt
Measured: no skirt
The Model

Figure 5-10 Estimation of nanowire length by the VPD based model. The model only
considers VPD mechanism. The size of the tested array is 1 mm × 1 mm,
with a 350 m wide skirt. The diameter and center-to-center spacing is 90
nm and 200 nm, respectively.

5.3 Dependence of Nanowire Growth on the Polarity of Substrates
As shown in chapter 3 and 4, growth behaviors of III-V semiconductor
nanowire NS-SAG are affected by the polarity of substrates. Only InP nanowires
show a preference for growth in the [111]A direction. InP nanowire growth on
(111)B substrates shows strong growth suppression, with an occupation ratio of less
than 10%. There is no growth suppression for GaAs and InAs NS-SAG on (111)A
substrates; however, GaAs and InAs nanowires show a preference for growth in the
<111>B directions. The polarity of substrate influences the preferred growth
165
direction and also the crystal structure and SF density. As shown in chapter 3, InP
nanowires grown in the [111]A direction are in the wurtzite (WZ) structure with SFs.
In InP NS-SAG, the critical diameter of the structural transition to zincblende (ZB) is
larger than 125 nm. This result is much larger than the theoretical thermodynamic
stability criterion estimate of 32 nm[15], which indicates there is another mechanism
of WZ structure formation in InP nanowire NS-SAG other than thermodynamic
stability of structures.
Figure 5-11 represents the relation between the TMI partial pressure and SF
density in InP nanowires grown in [111]A. The graph illustrates that lower TMI
partial pressure and smaller diameter decreases SF density. PH
3
partial pressure
cannot affect the SF density in used partial pressure range. Interestingly, InP
nanowires grown in the [111]B direction are in the ZB structure with dense SFs,
which is similar to GaAs or InAs nanowires grown in the [111]B direction. As
described in chapter 4, the average repetition period of SF in the [111]B direction
GaAs nanowires grown at 750 C and 3.79×10
-7
atm of TMG partial pressure is 12.7
monolayers (MLs). In same TMG partial pressure, if GaAs nanowires are grown at
650 C, the average SF repetition period becomes 4.2 MLs. The repetition period is
also increased by lowering the TMG partial pressure. At 750 C, increasing TMG
partial pressure from 3.79×10
-7
atm to 7.58×10
-7
atm decreases the repetition period
from 12.7 MLs to 5.4 MLs. InAs nanowires in the [111]B direction show faster SF
generation (3 MLs), regardless of growth temperature and AsH
3
partial pressure.
This is consistent with previous InAs nanowire NS-SAG work[16]. According to
166
these observations, it is clear that preferred growth direction and crystal structure
formation are closely related to the polarity of the substrates. The only two
differences between A and B polar planes are the surface reconstruction and atomic
components composing the plane. Those two factors govern all growth behavior and
stacking sequence of the crystal.
1.2x10
-6
1.8x10
-6
2.4x10
-6
0
5
10
15
20
25
30
35
40
45
50

Stacking fault density ( m
-1
)
TMI partial pressure (atm)

Figure 5-11 SF density of InP nanowires with respect to TMI partial pressure. The
nanowires are grown in the [111]A direction. The average diameter of
nanowires is 50 nm.
The ideal (111)A and B surfaces consist of only group III or V atoms,
respectively. These ideal (111) surfaces undergo surface reconstruction to stabilize
surface dangling bonds by obeying the ECM[7]. As seen in section 5.1, under PH
3

overpressure, the InP (111)A surface is reconstructed as ( 3× 3)R30  with P
trimers[12]. Under AsH
3
overpressure, it is known that the GaAs (111)A and B
167
surfaces show (2×2) reconstruction with As trimers[10, 17]. The reconstruction of
InP (111)B in MOCVD is not yet clear, but it may also be stabilized by P trimers.
To start growth on the (111)B substrate, group V trimers should be
eliminated for the following group III layer. To dissociate group V trimers, As-As
bonds or P-P bonds must be broken. The bond dissociation energy (BDE) of In-P,
Ga-As, and In-As bonds are 197.9 kJ/mol, 202 kJ/mol, and 201 kJ/mol,
respectively[18]. The BDE of the P-P bond (201 kJ/mol) is almost equal to the BDE
of an In-P bond[19]. It is challenging to dissociate P trimers without also breaking
In-P bonds. Due to strong P-P bonds, P trimers disturb stacking of the following
group III layers and cause growth suppression on InP(111)B substrates. On the
contrary, by virtue of weak As-As bond (146 kJ/mol)[19], As trimers can be
selectively eliminated while allowing for GaAs or InAs growth. This explains why
GaAs and InAs nanowires are able to grow, but InP nanowire growth is suppressed
in the [111]B direction.
After the removal of surface trimers, the crystal structure is determined by the
thermodynamic stability of the structure[15]. The grown GaAs and InAs nanowires
in the [111]B direction are in the ZB structure, since the resolution limit of EBL
limits their minimum diameters at 25 nm, which is almost twice as large as the
estimated critical diameters by the thermodynamic stability criterion. Grown InP
nanowires in the [111]B direction are also in the ZB structure, because they cannot
reach their critical diameter (32nm) due to enhanced growth on the sidewalls. Rapid
SF generation of nanowires grown in the [111]B direction also can be explained by
168
the group V trimer structures. At low temperature or high group V partial pressure,
adsorption of group III atoms and layer stacking starts with high surface coverage of
group V trimers due to their slow desorption. For instance, in GaAs nanowire NS-
SAG in the [111]B direction, since the As anions in the trimers and As adatoms
attached Ga atoms repel each other, As adatoms are forced to the WZ lattice site as
shown in figure 13, which promotes SF generation. Therefore, high growth
temperature and low group AsH
3
partial pressure can reduce SF generation in the
[111]B direction GaAs nanowires by reducing the surface coverage of As trimers. It
is challenging to suppress dense SFs in the [111]B direction InAs nanowires due to
its small growth temperature range, which causes high surface coverage of As
trimers.

Figure 5-12 Schematic diagram of Adatom adsorption on GaAs (111)B surface with
high As trimer coverage. Red arrows represent strong repulsion and
orange arrows represent weak repulsion.
Compared to trimers on the (111)B surface, trimers on the (111)A surface can
be reused to form the following group V layer. Growth on the (111)A plane always
starts with high coverage of trimers, and subsequent adatom adsorption is affected by
169
their presence. This is very important to InP nanowire growth on InP (111)A
substrates. ( 3× 3)R30  reconstruction has 100% of P coverage, which means there
is no exposed In atom on the surface. If ( 3× 3)R30  reconstruction can be
maintained at the initial stage of growth, all In adatoms must make bonds with P
trimers. On the top of trimers, there is no stacking restriction. This is quite different
from (2×2) reconstruction. If the GaAs (111)A surface is terminated by (2×2)
reconstruction shown at Figure 5-13 (a), first As adatom adsorption occurs at the
surface Ga atom places due to its imperfect P coverage, 75%. In a (2×2)
reconstructed surface, following Ga adatoms sit at ZB crystal sites, where the
interaction between As trimers and Ga adatoms is maximized [10]. This is called
3Ga-As structure. The 3Ga-As structures are further stabilized by forming As
trimers on the top[10].
The presence of the mask affects adsorption on ( 3× 3)R30  reconstructed
surfaces as well as (2×2) reconstructed surfaces. Because the mask blocks part of
surface atoms, the location of group III adatom adsorption is determined by
minimizing dangling bond generation and maximizing the interaction to group V
trimers. Figure 5-13 illustrates that, in both GaAs(111)A-(2×2) and InP(111)A-
( 3× 3)R30  reconstructions, the optimal site of the first group III adatom is a ZB
lattice site; however, to maximize the interaction, the second adatom adsorption
rotates the stacking and forms WZ stacking. Therefore, in either case, ZB lattice
sites are not optimal sites for adsorption in terms of these criterions. Instead, the WZ
lattice sites are the optimal sites that can increase the interaction and decrease the
170
number of dangling bonds. WZ formations are stabilized further due to the diffusion
of precursors from the mask region and sidewalls. In Figure 5-14, all of the pits are
located in the center region and there are no pits in the edge regions. This indicates
the growth rate and stability of edge region is better than the center of the wire. In
NS-SAG, the contribution from the growth of the edge region in the entire growth
behavior is larger than in macroscopic SAG. As the wire diameter becomes larger,
the contribution from WZ stacking at the edge region cannot dominate the entire
growth behavior. If the TMI partial pressure is high, the stability of ZB structure
formation at the center of nanowires is improved due to increase of direct supply of
vapor phase precursors and can compete with WZ structure formation at the edge.
This competition increases the probability to form ZB stacking, which is why SF
density in [111]A direction InP nanowires is proportional to the TMI partial pressure.
(2×2) reconstruction of GaAs (111)A with As trimers is similar to
( 3× 3)R30  reconstruction of InP (111)A with P trimers. GaAs and InAs are more
likely to form ZB structure rather than WZ structure due to their lower ionicity[15].
This strong competition between WZ stacking at the edge and ZB stacking at the
center makes the (111)A plane a less favorable growth plane compared to other
planes, such as (111)B, which have no stacking competition. Though there is no
growth suppression as in InP NS-SAG on InP (111)B, GaAs and InAs form clusters
or nanowires in <111>B or <112>B, preferable growth directions where there is no
competition.

171
As trimer
Surface Ga atom
Mask
3Ga-As
adsorption
3Ga-As
adsorption
w/ As trimer
ZB stacking
at the edge
WZ stacking
at the edge
As adatom
Ga adatom
As trimer
2
nd
layer

(a)

(b)
Figure 5-13 Schematic diagram of Adatom adsorption on (a) GaAs (111)A surface
and (b) InP (111)A surface. The GaAs surface forms (2×2) reconstructed
surface and the InP surface forms ( 3× 3)R30  reconstruction.
172

Figure 5-14 Top view of 200 nm thick InP nanowire array. Black stains on the top are
pits on the top surface.
5.4 Electron Beam Induced Bundling of Nanowires
As mentioned in section 5.1, SEM plays the major role in surface
morphology characterization of grown nanowire arrays. Because the SEM utilizes
interaction between high energy electrons and specimens, unwanted artifact can
occur and deformation of specimens by high energy electron bombardment is also
possible. Figure 5-15 is an example of SEM artifacts which can be observed in
measuring nanowires grown by NS-SAG using a Hitachi S4800 cold field emission
SEM. This artifact looks wavy in the fast scan mode of SEM, and in slow scan mode,
images of nanowire are blurred and form ghost images.

173

(a) (b)
Figure 5-15 Typical artifact in SEM measurement of thin nanowires whose diameter
is smaller than 60 nm. The artifact looks like surface oscillation in (a) fast
scan view and smeared ghost image in (b) slow scan view. All
micrographs are taken by Hitachi S4800.
If the electron beam scanning is not stopped after this artifact appears, the
electron beam causes bundling of nanowires. Here, this is described as electron
beam induced bundling (EBIB). EBIB is spontaneous. It is very challenging to
capture the moment of the middle of EBIB process even with TV scan. This
indicates that it is not a slow mechanical bend by physical damage of wire. This
process needs setup time to appear and cannot be seen immediately after stating to
scan the electron beam. The setup time is proportional to nanowire diameter, and
center-to-center spacing, and inversely proportional to the nanowire length. The
setup time is also inversely proportional acceleration voltage, beam current and
magnification. Figure 5-16 is the evolution of EBIB in an electron beam scanning
window. The acceleration voltage of electron beam and extractor current value is 1
kV and 10 A, respectively. The actual beam current cannot be measured because
the Hitachi SEM has no Faraday cup.
174

(a) (b)

(c) (d)
Figure 5-16 Time evolution of EBIB. The snap shots are taken at (a) the beginning
of EBIB event and (b) 2 scans, (c) 4 scans and (d) 6 scans after the event.
To use a controlled beam, Raith e-Line Schottky field emission EBL systems
was used for EBIB study. The EBIB experiments are carried out at 5 kV and 10 kV
of acceleration voltage settings and corresponding beam current is 180pA, and 224
pA, respectively. The tested array is InP nanowire array on a n-type InP (111)A
substrate. The diameter and length of nanowires is 40 nm and 2.2 m, respectively.
The center-to-center spacing of the openings is 200 nm. As shown in Figure 5-17 (a),
the ghost images of nanowires can still be seen, and EBIB occurs. Figure 5-17 (b)
and (c) are the results of single spot electron beam projection. After 2 minute of
175
electron beam projection on a nanowire, The EBIB propagates along the direction of
electron beam. The nanowires in front of the target nanowire are not bundled.
Figure 5-17 (b) shows the damages on the target nanowire and neighboring
nanowires, which depicts that the primary electron can pass through the nanowire
and propagate backward. The EBIB propagation length was 3.6 m in the electron
beam incident direction and 3.7 m in the lateral direction at 5 kV acceleration
voltage. 10 kV acceleration voltage gives similar result, 4.1 m in the electron beam
incident direction and 4.4 m in the lateral direction. Tiling of the sample is the
limiting factor of the EBIB propagation. Because the electron beam is incident at
45  from the surface normal vector and the length of target nanowire is 2.2 m, the
electron beam hits the substrate at a point which is 2.2 m away from the target
nanowire, if there is no scattering.
From these observations, the EBIB is caused by strong Coulomb attraction
between nanowires. The ghost image can be viewed as a field emission driven by
strong electric field between nanowires. Normally, electron beam projection charges
nanowire negatively, but, as in Figure 5-18, if the kinetic energy of the primary
electrons is large enough, they generate secondary electrons and pass through the
nanowire. This secondary electron emission makes the nanowire positively charged.
Because of high surface density of nanowire, the nanowires are totally depleted.
Therefore, dissipating built-up charges by electron beam is difficult and charges
accumulate in the nanowires. If there are neighboring nanowires which are charged
in opposite polarity, they attract each other and become bundled.
176

100 nm
Beam
target

(a) (b) (c)
Figure 5-17 EBIB experiment in Raith e-Line system. Acceleration voltage and
beam current are 5 kV and 180 pA, respectively. As in Hitachi S-4800, (a)
ghost images appear before the EBIB event. After 2 minutes of single
spot electron beam projection, (b) the target and neighboring nanowires
are damaged by electron beam. (c) EBIB propagates along the beam
direction.
h
+
h
+
h
+
h
+
e
-
e
-
e
-
e
-
e
- e
-
e
-
e
-
Secondary
electron
emission
Charging
by e
-
beam

Figure 5-18 Schematic of opposite charge building in nanowires. If electron beam
passes through a nanowire, positive charges are built by secondary
electron emission. On the contrary, if the electron beam stops in the
middle of a nanowire, the nanowires are negatively charged. Due to
strong depletion, these charges cannot be dissipated effectively. Then,
strong Coulomb attraction moves the nanowires.
177

Figure 5-19 Calculation setup of nanowire bending force. Center-to-center spacing is
2d. The displacement of nanowire, x, by bending is equal on both sides. It
is assumed that the nanowires have prefect hexagonal cross section.
The amount of charge for EBIB can be calculated by a simple equation. Here,
a bending model for ZnO nanowires is modified[20]. For simplicity, assume that the
amount of charge is equal on both the positive and negative nanowire; the charges
are located at the center of the top surface, and they bend equally. Figure 5-19 is the
setup of this model. If n holes and electrons are at the top, the electrostatic force, F
E
,
between nanowires is


2
2 2
0
2 2 4
1
x d
e n
F
E



(5-23)
Here, 
0
and e are respectively permittivity of vacuum and electron charge. d and x
are defined in Figure 5-19. If a hexagonal nanowire is bent by x, the applied force,
F
B
is
Ix
L
E
F
B 3
3
 (5-24)
178
, where E, L, I are Young’s modulus, length of nanowire and moment of inertia.
Assume that the nanowires have hexagonal cross section with hexagon edge length, r.
Then, I becomes (5 3/16)r
4
. Plug this in (5-24),
x r
L
E
F
B
4
3
16
3 15
 (5-25)
To start to bend nanowires, the increase of the electrostatic force by bending should
be larger than increase of bending force. Therefore,

4
3 3
2 2
0
16
3 15
8
1
r
L
E
d
e n


(5-26)
Reorganize Equation (5-26), the minimum charge Q for bending is

4
3
0 2 2 2
2
3 15
r
L
d E
e n Q






 

(5-27)

2
2 / 3
2 / 1
0
2
3 15
r
L
d E
Q
















(5-28)
Assume an InP nanowire pair whose spacing, radius and length are respectively 200
nm and 25 nm, and 1200 nm, the amount of charge needed for EBIB is 2.0 fC. This
result is consistent to the observations. But, if the diameter of nanowire exceeds 70
nm, EBIB does not appear regardless of nanowire length and spacing. From
Equation (5-28), the required charge for EBIB is a few fC. This indicates thick
nanowires have a conducting channel at the center of the nanowire and the surface
depletion width is approximately 30 nm. Also, InAs nanowires show no EBIB
regardless length, center-to-center spacing, diameter, and electron beam parameters.
179
This is because the InAs nanowires are highly conductive due to electron
accumulation at the sidewalls.
The direction of EBIB can be steered by adjusting the acceleration voltage
and direction of the projection. Figure 5-19 is the schematic of controlled EBIB. In
the controlled EBIB, the projection angle of the electron beam is adjusted to allow
only two nanowires to have primary electron injection. Sufficiently low acceleration
voltage must be used to stop the electron beam in the second nanowire. In this
scheme one nanowire pair can be selectively charged. The circled region in Figure
5-20 (b) is the result of the controlled EBIB. The nanowires can be bundled pair-by-
pair in a controlled direction.

(a) (b)
Figure 5-20 The concept and experiment result of of controlled EBIB. (a) Incident
angle must be adjusted to project electron beam only on the first nanowire.
Electron beam pass should not touch the third nanowire. By controlled
EBIB, pair-wise bundling is possible in wanted direction. (b) The result of
coltrolled EBIB is circles in the micrograph.
180
5.5 Summary
For modeling, SAG parameters such as the growth rate on a bare
semiconductor region and the lateral diffusion length of the vapor phase precursors
are extracted by macroscopic SAG experiments. The measured lateral diffusion
length of InP and GaAs are 40 m and 181.5 m, respectively. VPD of SAG can be
simulated by solving Laplace equation with proper boundary conditions which are
determined by patterning parameters; however, simulation of VPD in NS-SAG is a
very computensively expensive calculation. It requires a nanometer size simulation
grid over a millimeter wide range for modeling the nano opening arrays. In this
chapter, the average adsorption approximation is proposed to reduce size of the VPD
simulation. In this approximation, the nano opening array region is viewed as a new
bare semiconductor region whose lateral diffusion length is the weighted average of
the lateral diffusion lengths of a bare substrate and a dielectric mask region. This
approximation can reduce the computation cost dramatically.
The VPD model cannot accurately estimate the growth rate of NS-SAG due
to SD on the sidewalls. SD is proportional to total precursor flux on the sidewalls
and area of the sidewalls and inversely proportional to area of the growth surface.
The SD term is added in the growth rate equation of VPD model. This hybrid model
has no underestimation and can simulate the effect of skirt regions. The average
modeling error is 9%. In the model of GaAs NS-SAG, the extracted incorporation
181
parameter of SD is 4 times larger than that of InP NS-SAG. This illustrates strong
SD effect in GaAs NS-SAG.
As shown in chapter 3 and 4, the preferred growth direction, structural crystal
translation, and SF density of III-V semiconductor nanowire NS-SAG are affected
by substrate polarity. These can be explained by surface reconstruction of the (111)
surface and bond strength difference of each semiconductors and V-V bonds. GaAs
and InP (111)B surfaces form (2×2) surface reconstruction cells with the group V
trimers. To grow nanowires, group V trimer must be eliminated for the following
group III plane stacking. Removal of As trimers is relatively easy because As-As
bond is weaker than the semiconductor bonds, whereas it is very hard to selectively
decompose P trimers owing to the strong P-P bond. This is why InP NS-SAG on InP
(111)A is suppressed. Because there is no restriction on the growth, the (111)B
surface is growth favorable for GaAs and InAs NS-SAG. The relation between SF
generation and TMG partial pressure of GaAs is also explained by As trimers. In
high growth rate condition, the growth starts with a high coverage of As trimers,
which forces As adatoms to stay at WZ lattice sites by repulsion.
Growth on the (111)A plane always starts with high coverage of trimers, and
subsequent adatom adsorption is affected by their presence. InP (111)A forms
( 3× 3)R30  reconstruction. At the mask edge region, In adatoms are forced to stay
at WZ lattice sites to maximize interaction with P trimers. Because of diffusion from
the mask region, these WZ structures at the edge of mask drive entire growth on the
surface. This is why [111]A direction InP nanowire have enhanced structural crystal
182
transition. In this case, excessive precursor concentration increases the stability of
the ZB structure at the center of nanowire and increases the possibility of ZB
stacking formation. A similar situation occurs on the GaAs (111)A surface; however
the competition between WZ stacking and ZB stacking is stronger owing to low
ionicity of Ga-As and In-As bonds, so the [111]B direction is preferred because it
has no such competition.
In electron microscopy of nanowire arrays, ghost image artifacts and
spontaneous bundling of nanowires is observed. EBIB experiments are carried out in
the Raith e-Line electron beam writer. The tested array is InP nanowire array on a n-
type InP (111)A substrate. The diameter and length of nanowires are 40 nm and 2.2
m, respectively. The center-to-center spacing of opening is 200 nm. The average
setup time, which is the time from electron beam scanning to the first bundling event,
is 43 s at 5 kV acceleration and 38 s at 10 kV acceleration. The setup time is quite
random and is affected by the scanning speed and magnification. By a single spot
projection experiment, propagation of EBIB is observed. This indicates the electron
beam can pass through multiple nanowires. According to these observations, it is
clear that the EBIB is caused by the strong Coulomb attraction between nanowires.
Because the electron beam can pass through nanowires, nanowires can be positively
charged by secondary electron emission. The calculated amount of charge for EBIB
is 2.0 fC. If the diameter of a nanowire exceeds 70 nm, EBIB does not appear
regardless of nanowire length and spacing. Also, InAs nanowires show no EBIB
regardless length, center-to-center spacing, diameter, and electron beam parameters.
183
These suggest that EBIB does not occur if the nanowires are conductive. The
direction of EBIB can be controlled by adjusting the acceleration voltage and
direction of projection. With this controlled EBIB, nanowires can be selectively
bundled pair-by-pair in a controlled direction.

184
Chapter 5 Endnotes
[1] V. J. Silvestri, R. Ghez, and T. O. Sedgwick, J. Electrochem. Soc. 119 245-
250 (1972).
[2] W. G. Oldham and R. Holmstrom, J. Electrochem. Soc. 114 381-388 (1967).
[3] E. Colas, C. Caneau, M. Frei, J. E. M. Clausen, W. E. Quinn, and M. S. Kim,
Appl. Phys. Lett. 59 2019-2021 (1991).
[4] O. Kayser, J. Cryst. Growth 107 989-998 (1991).
[5] S.-J. Choi, "Modeling of Selective Area Growth of InP-Related Alloys by
Metal Organic Chemical Vapor Deposition," Ph.D. Ph.D. Dissertation, Dept.
Material Science, University of Southern California, Los Angeles, CA, 2004.
[6] I. Langmuir, J. Am. Chem. Soc. 40 1361-1403 (1918).
[7] M. D. Pashley, Phys. Rev. B 40 10481 (1989).
[8] T. Nishida, K. Uwai, Y. Kobayashi, and N. Kobayashi, Jpn. J. Appl. Phys. 34
6326-6330 (1995).
[9] E. Kaxiras, Y. Bar-Yam, J. D. Joannopoulos, and K. C. Pandey, Phys. Rev. B
35 9625 (1987).
[10] A. Taguchi, K. Shiraishi, T. Ito, and Y. Kangawa, Surf. Sci. 493 173-177
(2001).
[11] A. Taguchi, K. Shiraishi, and T. Ito, Phys. Rev. B 61 12670 (2000).
[12] C. H. Li, Y. Sun, D. C. Law, S. B. Visbeck, and R. F. Hicks, Phys. Rev. B 68
085320 (2003).
185
[13] L. Reimer, Scanning Electron Microscopy - Physics of Image Formation and
Microanalysis. Berlin: Springer-Verlag, 1998.
[14] K. Tomioka, P. Mohan, J. Noborisaka, S. Hara, J. Motohisa, and T. Fukui, J.
Cryst. Growth 298 644-647 (2007).
[15] T. Akiyama, K. Sano, K. Nakamura, and T. Ito, Jpn. J. Appl. Phys. 45 L275
(2006).
[16] K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Jpn. J. Appl. Phys. 46
L1102 (2007).
[17] T. Nishida, K. Uwai, Y. Kobayashi, and N. Kobayashi, Jpn. J. Appl. Phys. 34
6326 (1995).
[18] Y.-R. Luo, Comprehensive handbook of chemical bond energies. Boca Raton:
CRC Press, 2007.
[19] T. L. Cottrell, The Strength of Chemical Bonds, 2nd ed. London:
Butterworths, 1958.
[20] J. Liu, S. Lee, K. Lee, Y. H. Ahn, J.-Y. Park, and K. H. Koh,
Nanotechnology 19 185607 (2008).

186
Chapter 6: Conclusions and Future Work
6.1 Nanoscale Selective Area MOCVD
As the importance of nanowires increases in the semiconductor industry, the
need for high quality, single crystal nanowire array growth increases. Nanoscale
selective area MOCVD growth (NS-SAG) is a promising candidate to fulfill these
needs. NS-SAG utilizes dielectric masks with nano opening arrays to form growth
clusters only in the nano openings. By selecting the proper growth conditions,
growth on the side facets can be minimized and epitaxial nanowires can grow
vertically. Since this technique only uses a dielectric growth mask, there is no
unwanted metal incorporation, which is a problem in vapor-liquid solid assisted
MOCVD techniques. NS-SAG is also suitable for uniformly distributed, dense
nanowire arrays. Because the size and location of nanowires are governed by
patterning, ideally, control over nanowire diameter and location may be equal to the
critical dimension of the nanolithography technique being used, which may extend
down to the nanometer level.
6.2 Electron Beam Lithography
EBL technique has been widely used for nano pattern generation. It is
capable of sub-100 nm resolution and is flexible in pattern modification. In EBL, the
187
pattern data is stored in an electronic file form. The system translates the pattern file
into deflection information for the electron beam to reproduce the pattern with the
trace of an electron beam. The pattern is written on an electron beam sensitive resist.
After the development process the pattern is reproduced pattern on the resist.
In this work, EBL parameters are optimized and to achieve the minimum
opening diameter and center-to-center spacing possible. Raith e-Line system at USC
is used for the lithography. Among all of the EBL parameters, acceleration voltage,
beam current, electron dose, working distance, and write field size are the most
important parameters for optimization. The acceleration voltage is set to 10 kV,
because at higher acceleration voltage, backward scattering may become a problem,
and up to 20 kV, there is no significant resolution improvement. The electron dose is
optimized by dose tests. Owing to proximity effect variation, the clearing dose,
which can clear positive resist (in this work, PMMA 950k C2), changes with respect
to the center-to-center spacing. At 200 nm center to center spacing, 1 fC dot dose is
the slight overexposure condition for 100 nm thick 950k PMMA. Though shorter
working distance delivers best result in terms of resolution, 10 mm working distance
is chosen to increase the depth of focus in the writing. (The minimum working
distance is 7 mm for the e-Line system). The write field is the maximum field size
that can be written without moving the stage. Smaller write fields have less
distortion at the edge of the field but have increased writing times because of stage
movement and stitching calibration. Here, 100 m × 100 m write field was chosen,
which is the standard field size of the system, and the typical pattern area used here
188
is smaller than 100 m × 100 m. From this optimization 25 nm diameter and 80
nm center-to-center spacing patterns were achieved.
Considering the poor dry etch selectivity of PMMA, a 100 nm thick 950k
PMMA process is developed to form nano opening array on a 30 nm thick SiN
X

layer. Dry etch and other process are designed for SAG template preparation. There
are two different approaches developed to generate templates. The two step
deposition process utilizes mix-and-match lithography between photolithography and
EBL. This process can reduce EBL time by forming the macroscopic shape of the
mask with photolithography. A one-step deposition process is developed to
minimize damage on the growth surface. Though the EBL time is far longer than
that of two-step approach, the growth surface is less exposed to plasma and chemical
processing. A bilayer lift off process is also developed for metal contact patterns.
6.3 Growth of Phosphide Nanowire Arrays
InP nanowire NS-SAG has been studied with TBP for the group V precursor.
The high pyrolysis efficiency of TBP at low temperature loosens the tight growth
window of InP NS-SAG and makes more room for growth temperature tuning. Only
a few InP NS-SAG studies have been performed with PH
3
as the P precursor. The
optimized growth condition for InP nanowire is 1.8×10
-6
atm of TMI partial pressure,
1.43×10
-4
atm of PH
3
partial pressure at 650 C. The V/III ratio for this condition is
almost 7 times higher than previous works due to the low cracking efficiency of PH
3
.
189
The preferred growth direction of InP nanowires is the [111]A direction. The
effective precursor concentration of nano opening arrays determines the surface
morphology of nanowires, because high precursor concentrations increase the
reaction on the sidewalls and other non-polar and semi-polar planes. The degree of
tapering and lateral growth increases and is reflected in the relation in the tapering
parameter and lateral growth parameter with respect to the precursor concentration.
Bumps also appear on sidewalls in high precursor concentration, especially of the
NPC is greater than 5.5. If the NPC is greater than 8.5, cluster growth starts to
appear. The appearance of cluster increases as the NPC increases, and if the NPC
exceeds 14, the occupation ratio of nanowire form growth drops to less than 10%
The crystal structure of [111]A direction InP nanowires is WZ structure. The
critical diameter of this structural transition is larger than 125 nm. This is far larger
than thermodynamic estimation, 32 nm. Average SF density is 20 m
-1
and is
proportional to the TMI partial pressure. The sidewalls of WZ InP nanowires are
{1100}. {1100} planes are known to be the dominant sidewall type of WZ InP
nanowires. HR-TEM observations illustrate the shape and origin of sidewall bumps.
A SF exposes other {111}A planes on the sidewall and in high precursor
concentration, this small section of {111}A can be the seed for sidewall bump
generation.
Dopants are able to affect the behavior of InP NS-SAG. Si
2
H
6
and DEZ are
used to test surface morphology changes by dopant incorporation. Both affect the
growth, especially the sidewall growth, but the impact of DEZ is stronger than that
190
of Si
2
H
6
. High DEZ flow degrades the uniformity of nanowire arrays by promoting
lateral growth and generating thick nanowires. InP NS-SAG on (111)A is very
sensitive to the incorporation of other elements such as Ga or As. Even with a very
small amount of incorporation of other elements, the growth cannot form nanowire
but instead mainly forms clusters. Growths in pillar form are toward the <111>B
directions.
InP NS-SAG on InP (111)B is not uniform. Most of the openings have no
growth and only a few openings have vertical nanowires in the [111]B direction.
This growth is sensitive to the pre-treatment of template. If the template is treated by
diluted sulfuric acid, vertical nanowires disappear and instead, tripod structures
appear. Through HR-TEM observations, it is found that the [111]B direction InP
nanowires are in the ZB structure with dense SFs. The average SF density is from
100 m
-1
to 200 m
-1
. InP NS-SAG on (111)B substrates is insensitive to the
incorporation of other elements. The surface morphology of InP nanowire arrays
grown on GaAs (111)B is very similar to those grown on InP (111)A.
GaP NS-SAG is tried on InP (111) substrates. The preferred growth direction
of GaP nanowires is the [111]B direction. The growth parameter window of GaP
NS-SAG is very narrow. It is challenging to achieve a reasonable nanowire growth
rate and also maintain sidewall growth suppression. Noticeable nanowire formation
occurs at 2.36×10
-6
atm of TMG partial pressure and 2.14×10
-4
of PH
3
partial
pressure at 700 C. At 750 C, the growth is totally suppressed. At 650 C, cluster
generation is reduced but the diameter of grown nanowires is 2.8 times larger than
191
that of openings. The cluster generation is stronger in larger openings, which
indicates that the growth behavior is affected by the large lattice mismatch between
InP substrate and the grown GaP clusters.
6.4 Growth of Arsenide and Heterostructure Nanowire Arrays
The growth temperature of GaAs NS-SAG is higher than that of GaAs film
growth on GaAs (100) substrates. The precursor partial pressure is far lower than
that of GaAs film growth in order to suppress the growth on non-polar sidewalls.
The growth temperature of GaAs nanowire NS-SAG ranges from 650 C to 750 C.
Trimethylgallium (TMG) and AsH
3
are used for group III and V precursors. The
SGC of GaAs NS-SAG is 3.79 ×10
-7
atm of TMG partial pressure and 2.42×10
-4
atm
of AsH
3
partial pressure.
GaAs NS-SAG on GaAs (111)A is not uniform and develops clusters and
many tilted nanowires toward the nearest <111>B directions. More than 75% of the
openings have grown clusters. GaAs NS-SAG on GaAs (111)B is uniform and
vertical. This represents that the [111]B direction is the preferred growth direction
of GaAs NS-SAG. There is no strong tapering and lateral growth of grown GaAs
nanowires. In high growth temperature and low AsH
3
partial pressure, localized
growth suppression appears. This is caused by a localized As deficiency at the early
stage of growth. Because openings are the only place for heterogeneous AsH
3

192
pyrolysis, localized As deficiency retards the growth where it occurss and causes
non-uniformity of the nanowire array.
There is another growth disorder in the opposite direction. At low growth
temperature and high AsH
3
partial pressure strong lateral growth occurs over
multiple opening and the epitaxial burial of nanowires happens. Epitaxial burial
easily occurs in densely packed GaAs nanowire arrays. As nanowires grow, As
supply increases, because more AsH
3
molecules are cracked on the sidewall due to
the expansion of the sidewall area. This excessive As supply triggers strong lateral
growth. To eliminate both disorders, AsH
3
flow modulation is introduced in the
growth. This technique can suppress both disorders effectively.
From HR-TEM observation results, GaAs nanowires in the [111]B direction
are in ZB structure with rapid SF generation. The average SF density ranges from
200 m
-1
to 400 m
-1
. The average SF repetition period is a function of the growth
temperature and of the TMG partial pressure. 3.79×10
-7
atm of TMG partial pressure
conditions can increase the SF repetition period up to 12.7 monolayers (MLs) at 750
C. If the growth temperature decreases down to 650 C or the TMG partial pressure
is doubled, the average SF repetition period becomes 4 to 5 MLs.
InAs nanowire NS-SAG shows similar growth kinetics to GaAs nanowire
NS-SAG. The diffusion length of In is much shorter than that of Ga. The growth
temperature ranges from 550 C to 600 C. The SGC of InAs NS-SAG is 3.00 ×10
-7

atm of TMI partial pressure and 1.00×10
-4
atm of AsH
3
partial pressure. The
preferred growth direction is the [111]B direction. The growth rate is much faster
193
than GaAs or InP nanowires, which shows the high incorporation of In atoms in the
openings. InAs nanowire arrays with smaller opening have strong growth rate
fluctuations, and this depicts the effect of SD on the sidewalls. Because the SD
effect is inversely proportional to the radius of nano openings, larger opening arrays
show better growth rate uniformity. HR-TEM observations indicate that the grown
InAs nanowires in the [111]B direction are of the ZB crystal structure with dense
SFs. The SF density is 400 m
-1
or higher, which indicates the SF repetition period
is approximately 3 ML per SF. The SF density of InAs nanowires cannot be
controlled by the growth conditions due to the low growth temperature.
InAsP NS-SAG is attemped to test the conflict of the preferred growth
direction in a ternary alloy. For InAsP nanowire growth, the growth temperature and
TMI partial pressure is fixed at 650 C, and 4.20×10
-7
atm, respectively. Two
different PH3/AsH
3
ratios, 1.17 and 9 are used to change the composition of InAsP
in a ternary compound. On InP (111)B substrates, as the composition of InP
increases, growth suppression appears. This is exactly same as InP growth behavior
on InP (111)B. This growth suppression is also related to the opening diameter,
which seems to be caused by AsH
3
deficiency in the early stages of growth.
InP/InAs heterostructure nanowire arrays are grown on InP (111)A substrates.
The growth conditions of InP and InAs nanowire are the SGC of those materials.
InAs layers are grown on the sidewalls of InP nanowires. Non-uniform sidewall
growth of InAs layers on the sidewalls bends the entire growth structure, and this
shows strong strain in the structure. A few nanowires have vertically stacked InAs
194
structures on the top of the InP nanowires The top InAs nanowires are grown in the
preferred growth direction, the nearest <111>B direction. This illustrates that
mismatch in the preferred growth directions can kink the heterostructure nanowires.
6.5 Growth Models of NS-SAG and Electron Beam Induced Bundling
To estimate the growth rate of nanowires and control the effective precursor
concentration precisely, a growth model which is capable of describing complex
diffusion processes is required. In this work, a growth model based Choi’s VPD
model[1] is proposed.
For modeling, SAG parameters such as the growth rate on a bare
semiconductor region and the lateral diffusion length of vapor phase precursors are
extracted by macroscopic SAG experiments. The measured lateral diffusion length
of InP and GaAs are 40 m and 181.5 m, respectively. A 1 mm × 1 mm size InP
nanowire array is grown for a macroscopic SAG effect test. The array clearly shows
dependence of nanowire length on the position in the array and presence of a skirt
region.
VPD of SAG can be simulated by solving the Laplace equation with proper
boundary conditions which are determined by the patterning parameters; however,
simulation of VPD in NS-SAG is a very computationally expensive calculation. It
requires a nanometer size simulation grid over a millimeter wide range for modeling
the nano opening array. The lateral diffusion lengths of precursors are far larger than
195
the nano opening diameter and center-to-center spacing of the array. The average
adsorption approximation is proposed to reduce the size of the VPD simulation. In
this approximation, the nano opening array region is viewed as a new bare
semiconductor region whose lateral diffusion length is the weighted average of the
lateral diffusion lengths of a bare substrate and dielectric mask region. This
approximation can reduce the computation cost dramatically and the size of
simulation becomes small enough for calculation on a desktop PC.
The VPD model cannot accurately estimate the growth rate of NS-SAG.
Since nanowires have a large surface area on their sidewalls, SD on the sidewalls
should be considered. The SD term is proportional to the total precursor flux on the
sidewalls and the area of the sidewalls and inversely proportional to area of the
growth surface, which is the top surface of the nanowire. This SD term added in the
growth rate equation of the VPD model. This hybrid model has no underestimation
and can simulate the effect of skirt regions. The average modeling error is 9%. Due
to the long diffusion length of GaAs precursors, the VPD effect in GaAs NS-SAG is
not as significant as in InP NS-SAG. The extracted incorporation parameter,  is 4
times larger than that of InP NS-SAG. This illustrates the strong SD effect in GaAs
NS-SAG, except during the very early stage of growth.
As shown in chapter 3 and 4, the preferred growth direction, structural
translation of crystal, and SF density of III-V semiconductor nanowire NS-SAG are
affected by the polarity of substrates. These can be explained by surface
reconstruction of the (111) surface and by the bond strength difference of each
196
semiconductors and V-V bonds. GaAs and InP (111)B surfaces form (2×2) surface
reconstruction cells with group V trimers. To grow nanowires, the group V trimer
must be eliminated for the following group III plane stacking. Removal of As
trimers is relatively easy because the bond dissociation energy of As trimers is
sufficiently lower than that of the semiconductor bonds, whereas the bond
dissociation energy of the P-P bond is comparable to the In-P bond. Therefore, it is
very hard to selectively decompose P trimers. This is why InP NS-SAG on InP
(111)A is suppressed. Because there are no restrictions on the growth, the (111)B
surface is growth favorable for GaAs and InAs NS-SAG. Rapid SF generation of
GaAs and InAs is also explained by As trimers. In high growth rate condition, the
growth starts with high coverage of As trimers. Because As trimers repel As adatoms
from WZ lattice sites, twining easily occurs.
Group V trimers on (111)A can be reused to form the following group V
layer. Growth on the (111)A plane always starts with high coverage of trimers, and
subsequent adatom adsorption is affected by their presence. InP (111)A forms
( 3× 3)R30  reconstruction which has 100% P coverage. The presence of the mask
affects adsorption on ( 3× 3)R30 reconstructed surfaces. At the mask edge region,
In adatoms are forced to stay at WZ lattice sites to maximize the interaction with P
trimers. Because of diffusion from the mask region, these WZ structures at the edge
of mask drive the entire growth on the surface. This is why the [111]A direction InP
nanowire have enhanced structural crystal transition. In this case, excessive
precursor concentration strengthens the stability of the ZB structure at the nanowire
197
center and increases the possibility of ZB stacking formation. A similar situation
occurs on the GaAs (111)A surface; however the competition between WZ stacking
and ZB stacking is stronger due to the low ionicity of Ga-As and In-As bonds. So,
they take the [111]B direction which has no such stacking competition.
In electron microscopy of nanowire arrays, ghost image artifacts and
spontaneous bundling of nanowires are observed. Electron beam induced bundling
(EBIB) experiments are carried out in the Raith e-Line electron beam writer. The
tested array is an InP nanowire array on a n-type InP (111)A substrate. The diameter
and length of nanowires are 40 nm and 2.2 m, respectively. The center-to-center
spacing of the openings is 200 nm. The average setup time, which is the time from
beginning electron beam scanning to the first bundling event, is 43 s at 5 kV
acceleration and 38 s at 10 kV acceleration. The setup time is random and is affected
by the scanning speed, and magnification. By a single spot projection experiment,
propagation of EBIB is observed. After 2 minutes of single spot projection, the
target nanowire and neighboring nanowires are damaged due to the high energy
focused electron beam. This indicates the electron beam can pass through multiple
nanowires.
According to these observations, it is clear that the EBIB is caused by strong
a Coulomb attraction between nanowires. Because the electron beam can pass
through nanowires, nanowires can become positively charged by secondary electron
emission, and this positive charge can attract negatively charged neighboring
nanowires. The equation of the Coulomb attraction force and mechanical bending
198
force predicts that the needed charge is EBIB for a few fC. If the diameter of the
nanowire exceeds 70 nm, EBIB does not appear regardless of nanowire length and
spacing. Also, InAs nanowires show no EBIB regardless length, center-to-center
spacing, diameter, and electron beam parameters, which suggests that EBIB does not
occur if the nanowires are conductive. The direction of EBIB can be controlled by
adjusting the acceleration voltage and projection direction. With controlled EBIB,
nanowires can be selectively bundled pair-by-pair in a selected direction.
6.6 Future Work
This study provides NS-SAG growth conditions for InP, GaAs, InAs, and
GaP. GaP NS-SAG is not as completely studied as the other materials; however,
GaP is important as a starting point for InGaP, which is a useful material for
photovoltaic applications and LEDs. Therefore, further study of growth optimization
of GaP and InGaP growth experiments would be a good research topic. Also, to
prove crystal quality of grown nanowires, test device fabrication is strongly
recommended. For device applications, especially photovoltaic or LED devices,
larger nano opening array patterns are needed. EBL is not a good lithography
technique for these large nano opening array matrixes due to its long writing time.
Larger scale nano lithography is required before the device fabrication is easily
possible. In substrate selection model, computational verification of the adsorption
model would provide better support for the model. Making direct observations of
199
the surface reconstruction changes would be an interesting research problem. In
EBIB, an experiment must be performed to make a quantitative analysis of the
relation between nanowire conductance and EBIB.

200
Chapter 6 Endnotes
[1] S.-J. Choi, "Modeling of Selective Area Growth of InP-Related Alloys by
Metal Organic Chemical Vapor Deposition," Ph.D. Ph.D. Dissertation, Dept.
Material Science, University of Southern California, Los Angeles, CA, 2004.

201
Comprehensive Bibliography
S. Adachi, Properties of group-IV, III-V and II-VI semiconductors. West Sussex:
John Wiley & Sons Ltd., 2005.
M. Akabori, J. Takeda, J. Motohisa, and T. Fukui, Nanotechnology 14 1071-1074
(2003).
T. Akiyama, K. Sano, K. Nakamura, and T. Ito, Jpn. J. Appl. Phys. 45 L275 (2006).
S. Choi, Y. Chung, D. Kim, H. Chu, and K. Seo, Jpn. J. Appl. Phys. 39 2508-2511
(2000).
S.-J. Choi, "Modeling of Selective Area Growth of InP-Related Alloys by Metal
Organic Chemical Vapor Deposition," Ph.D. Ph.D. Dissertation, Dept. Material
Science, University of Southern California, Los Angeles, CA, 2004.
S. J. Choi, K. Djordjev, S.J. Choi, P. D. Dapkus, W. Lin, G. Griffel, R. Menna, and J.
Connolly, IEEE Photonics Technol. Lett. 16 828-830 (2004).
S. J. Choi, Z. Peng, Q. Yang, E. H. Hwang, and P. D. Dapkus, IEEE Photonics
Technol. Lett. 17 2101-2103 (2005).
L. C. Chuang, M. Moewe, C. Chase, N. P. Kobayashi, and C. Chang-Hasnain, Appl.
Phys. Lett. 90 043115 (2007).
E. Colas, C. Caneau, M. Frei, J. E. M. Clausen, W. E. Quinn, and M. S. Kim, Appl.
Phys. Lett. 59 2019-2021 (1991).
D. G. Coronell and K. F. Jensen, J. Cryst. Growth 114 581-592 (1991).
T. L. Cottrell, The Strength of Chemical Bonds, 2nd ed. London: Butterworths, 1958.
Y. Cui, Z. Zhong, D. Wang, W. U. Wang, and C. M. Lieber, Nano Lett. 3 149-152
(2003).
202
P. D. Dapkus and J. J. Coleman, "Metalorganic Chemical Vapor Deposition," in III-
V Semiconductor Materials and Devices, R. J. Malik, Ed., 1st ed Amsterdam: North-
Holland, 1989.
Shadi A. Dayeh, D. P. R. Aplin, X. Zhou, P. K. L. Yu, Edward T. Yu, and D. Wang,
Small 3 326-332 (2007).
K. A. Dick, S. Kodambaka, M. C. Reuter, K. Deppert, L. Samuelson, W. Seifert, L.
R. Wallenberg, and F. M. Ross, Nano Lett. 7 1817-1822 (2007).
K. Djordjev, S.-J. Choi, W.-J. Choi, S.-J. Choi, I. Kim, and P. D. Dapkus, IEEE
Photonics Technol. Lett. 14 603-605 (2002).
J. P. Duchemin, M. Bonnet, F. Koelsch, and D. Huyghe, J. Cryst. Growth 45 181-
186 (1978).
F. E. Ejeckam, C. L. Chua, Z. H. Zhu, Y. H. Lo, M. Hong, and R. Bhat, Appl. Phys.
Lett. 67 3936-3938 (1995).
E. Ertekin, P. A. Greaney, D. C. Chrzan, and T. D. Sands, J. Appl. Phys. 97 114325
(2005).
N. C. Frateschi, P. D. Dapkus, S. S. Ou, J. J. Yang, and M. Jansen, IEEE Photonics
Technol. Lett. 5 741-743 (1993).
H. Goto, K. Nosaki, K. Tomioka, S. Hara, K. Hiruma, J. Motohisa, and T. Fukui,
Appl. Phys. Express 2 035004 (2009).
M. Fujii, H. Iwanaga, and N. Shibata, J. Cryst. Growth 99 179-182 (1990).
H. Hasegawa and M. Akazawa, Appl. Surf. Sci. 255 628-632 (2008).
K. Hiramatsu, K. Nishiyama, A. Motogaito, H. Miyake, Y. Iyechika, and T. Maeda,
Phys. Stat. Sol. (a) 176 535-543 (1999).
203
E. Hilner, U. Håkanson, L. E. Fröberg, M. Karlsson, P. Kratzer, E. Lundgren, L.
Samuelson, and A. Mikkelsen, Nano Lett. 8 3978-3982 (2008).
B. Hua, J. Motohisa, Y. Kobayashi, S. Hara, and T. Fukui, Nano Lett. 9 112-116
(2008).
M. Inari, J. Takeda, J. Motohisa, and T. Fukui, Physica E 21 620-624 (2004).
T. Ito, K. E. Khor, and S. Das Sarma, Phys. Rev. B 40 9715 (1989).
T. Ito, J. Appl. Phys. 77 4845-4886 (1995).
T. Ito, Jpn. J. Appl. Phys. 37 L1217 (1998).
A. S. Jordan and A. Robertson, J. Vac. Sci. Technol. A 12 204-215 (1994).
O. Kayser, J. Cryst. Growth 107 989-998 (1991).
E. Kaxiras, Y. Bar-Yam, J. D. Joannopoulos, and K. C. Pandey, Phys. Rev. B 35
9625 (1987).
K. E. Khor and S. Das Sarma, Phys. Rev. B 38 3318 (1988).
Y. Kitauchi, Y. Kobayashi, K. Tomioka, S. Hara, K. Hiruma, T. Fukui, and J.
Motohisa, Nano Lett. 10 1699-1703 (2010).
T. W. Kim, Y. J. Hong, G.-C. Yi, J.-H. Kwon, M. Kim, H. N. Han, D. H. Kim, K. H.
Oh, K.-J. Kong, and Y.-K. Kwon, J. Phys. D: Appl. Phys. 41 015406 (2008).
R. R. King, D. C. Law, K. M. Edmondson, C. M. Fetzer, G. S. Kinsey, H. Yoon, R.
A. Sherif, and N. H. Karam, Appl. Phys. Lett. 90 183516 (2007).
D. F. Kyser and N. S. Viswanathan, J. Vac. Sci. Technol. 12 1305-1308 (1975).
204
I. Langmuir, J. Am. Chem. Soc. 40 1361-1403 (1918).
J.-W. Lee, S.-W. Kim, K.-S. Seol, Y. Kwon, and K.-S. Seo, Microwave Opt.
Technol. Lett. 49 609-612 (2007).
T.-W. Lee, Y.-B. Moon, J.-H. Choi, D.-Y. Kim, and E. Yoon, J. Cryst. Growth 182
299-308 (1997).
C. H. Li, Y. Sun, D. C. Law, S. B. Visbeck, and R. F. Hicks, Phys. Rev. B 68 085320
(2003).
R. Li, "III-V compound semiconductor nanostructures by selective area growth using
block copolymer lithography," Ph.D., Material Sciences, University of Southern
California, Los Angeles, 2003.
J. Liang, H. Luo, R. Beresford, and J. Xu, Appl. Phys. Lett. 85 5974-5976 (2004).
J. Liu, S. Lee, K. Lee, Y. H. Ahn, J.-Y. Park, and K. H. Koh, Nanotechnology 19
185607 (2008).
Y.-R. Luo, Comprehensive handbook of chemical bond energies. Boca Raton: CRC
Press, 2007.
S. Luryi and E. Suhir, Appl. Phys. Lett. 49 140-142 (1986).
M. A. McCord and M. J. Rooks, "Electron beam lithography," in Handbook of
Microlithography, Micromachining, and Microfabrication. vol. 1, P. Rai-Choudhury,
Ed., ed Bellingham, WA: SPIE Optical Engineering Press, 1997.
H. M. Manasevit, Appl. Phys. Lett. 12 156-159 (1968).
P. Mohan, J. Motohisa, and T. Fukui, Nanotechnology 16 2903-2907 (2005).
P. Mohan, J. Motohisa, and T. Fukui, Appl. Phys. Lett. 88 013110 (2006).
205
J. Motohisa, J. Noborisaka, J. Takeda, M. Inari, and T. Fukui, J. Cryst. Growth 272
180-185 (2004).
"Nano PMMA and Copolymer Datasheet," Microchem Co. 2001.
T. Nishida, K. Uwai, Y. Kobayashi, and N. Kobayashi, Jpn. J. Appl. Phys. 34 6326-
6330 (1995).
J. Noborisaka, J. Motohisa, J. Takeda, M. Inari, Y. Miyoshi, N. Ooike, and T. Fukui,
"Growth of GaAs and InGaAs Nanowires by Utilizing Selective Area MOVPE," in
International Conference on Indium Phosphide and Related Materials, Kagoshima,
Japan, 2004, pp. 647-650.
M. Noguchi, K. Hirakawa, and T. Ikoma, Phys. Rev. Lett. 66 2243 (1991).
W. G. Oldham and R. Holmstrom, J. Electrochem. Soc. 114 381-388 (1967).
L. Ö. Olsson, C. B. M. Andersson, M. C. Håkansson, J. Kanski, L. Ilver, and U. O.
Karlsson, Phys .Rev. Lett. 76 3626 (1996).
J. S. Osinski, Y. Zou, P. Grodzinski, A. Mathur, and P. D. Dapkus, IEEE Photonics
Technol. Lett. 4 10-13 (1992).
M. D. Pashley, Phys. Rev. B 40 10481 (1989).
P. A. Peterson, Z. J. Radzimski, S. A. Schwalm, and P. E. Russell, J. Vac. Sci.
Technol. B 10 3088-3093 (1992).
J. C. Phillips, Rev. Mod. Phys. 42 317 (1970).
D. H. Reep and S. K. Ghandhi, J. Electrochem. Soc. 130 675-680 (1983).
L. Reimer, Scanning Electron Microscopy - Physics of Image Formation and
Microanalysis. Berlin: Springer-Verlag, 1998.
206
D. Ren, W. Zhou, and P. D. Dapkus, Appl. Phys. Lett. 86 111901 (2005).
T. Sadagopan, S. J. Choi, C. Sang Jun, K. Djordjev, and P. D. Dapkus, IEEE
Photonics Technol. Lett. 17 414-416 (2005).
V. Schmidt, H. Riel, S. Senz, S. Karg, W. Riess, and U. Gösele, Small 2 85-88
(2006).
L. Silvestre, A. Ougazzaden, D. Delprat, A. Ramdane, C. Daguet, and G. Patriarche,
J. Cryst. Growth 170 639-644 (1997).
V. J. Silvestri, R. Ghez, and T. O. Sedgwick, J. Electrochem. Soc. 119 245-250
(1972).
A. Srinivasan, S. Murtaza, J. C. Campbell, and B. G. Streetman, Appl. Phys. Lett. 66
535-537 (1995).
G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy, 2nd ed. San Diego:
Academic Press, 1999.
A. Taguchi, K. Shiraishi, T. Ito, and Y. Kangawa, Surf. Sci. 493 173-177 (2001).
A. Taguchi, K. Shiraishi, and T. Ito, Phys. Rev. B 61 12670 (2000).
K. Takemasa, M. Kubota, and H. Wada, IEEE Photonics Technol. Lett. 12 471-473
(2000).
T. Tanaka, K. Tomioka, S. Hara, J. Motohisa, E. Sano, and T. Fukui, Appl. Phys.
Express 3 025003 (2010).
"Technical Report: ZEP520A," ZEONREX Electronic Chemicals 2003.
B. Tomas, L.-E. Wernersson, T. Lowgren, and L. Samuelson, Nanotechnology 17
S227 (2006).
207
K. Tomioka, P. Mohan, J. Noborisaka, S. Hara, J. Motohisa, and T. Fukui, J. Cryst.
Growth 298 644-647 (2007).
K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Jpn. J. Appl. Phys. 46 L1102 (2007).
K. Tomioka, J. Motohisa, S. Hara, and T. Fukui, Nano Lett. 8 3475-3480 (2008).
A. Tribu, G. Sallen, T. Aichele, R. g. André, J.-P. Poizat, C. Bougerol, S. Tatarenko,
and K. Kheng, Nano Lett. 8 4326-4329 (2008).
J. Vuckovic, D. Fattal, C. Santori, G. S. Solomon, and Y. Yamamoto, Appl. Phys.
Lett. 82 3596-3598 (2003).
M. A. Wahab, Solid state physics: structure and properties of materials, 2nd ed.
Middlesex: Alpha Science International Ltd., 2005.
Y. Wang, V. Schmidt, S. Senz, and U. Gösele, Nature Nanotech. 1 186-189 (2006).
J. W. G. Wildöer, C. J. P. M. Harmans, and H. van Kempen, Phys. Rev. B 55
R16013 (1997).
Z. H. Wu, X. Mei, D. Kim, M. Blumin, H. E. Ruda, J. Q. Liu, and K. L. Kavanagh,
Appl. Phys. Lett. 83 3368-3370 (2003).
S. Yeon and K. Seo, Jpn. J. Appl. Phys. 47 2868-2871 (2008).
M. A. Zimmler, T. Voss, C. Ronning, and F. Capasso, Appl. Phys. Lett. 94 241120
(2009).
J. Zou, M. Paladugu, H. Wang, G. J. Auchterlonie, Y.-N. Guo, Y. Kim, Q. Gao, H. J.
Joyce, H. Tan, and C. Jagadish, Small 3 389-393 (2007).

Abstract (if available)

Abstract The interesting properties and potential applications of semiconductor nanowires have received significant attention. Nanoscale selective area growth using MOCVD (NS-SAG) has been demonstrated as an attractive growth technique for compound semiconductor nanowires. With this technique, the diameter and location of wires can be controlled, and no unwanted metal incorporation occurs. This technique is also suitable for large scale uniform nanowire arrays.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Gallium arsenide based semiconductor laser design and growth by metal-organic chemical vapor deposition

PDF

The selective area growth and coalescence of indium phosphide nanostripe arrays on silicon through MOCVD for NIR monolithic integration

PDF

Growth, characterization of gallium arsenide based nanowires and application in photovoltaic cells

PDF

Experimental studies and computer simulation of the preparation of nanoporous silicon-carbide membranes by chemical-vapor infiltration/chemical-vapor deposition techniques

PDF

GaAs nanowire optoelectronic and carbon nanotube electronic device applications

PDF

Inorganic/organic hybrid materials and grafted coatings via vapor phase deposition processes

PDF

Novel processing of liquid substrates via initiated chemical vapor depostion

PDF

Building blocks for 3D integrated circuits: single crystal compound semiconductor growth and device fabrication on amorphous substrates

PDF

Fabrication, deposition, and characterization of size-selected metal nanoclusters with a magnetron sputtering gas aggregation source

Asset Metadata

Creator Chu, Hyung-Joon (author)

Core Title The growth and characterization of III-V semiconductor nanowire arrays by nanoscale selective area metalorganic chemical vapor deposition

Contributor Electronically uploaded by the author (provenance)

School Andrew and Erna Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Electrical Engineering

Publication Date 07/30/2010

Defense Date 06/17/2010

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag MOCVD,nanoscale selective area growth,nanowire,OAI-PMH Harvest

Language English

Advisor Dapkus, P. Daniel (committee chair), Nakano, Aiichiro (committee member), Steier, William H. (committee member)

Creator Email hyungjoc@usc.edu,lssah1@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3231

Unique identifier UC154240

Identifier etd-Chu-3850 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-362863 (legacy record id),usctheses-m3231 (legacy record id)

Legacy Identifier etd-Chu-3850.pdf

Dmrecord 362863

Document Type Dissertation

Rights Chu, Hyung-Joon

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu