:3; ,
3.. an: .,

 

. . 15-13:. ,H ‘ . 10.1.
.v .31 . ., a. 3.. w ,

 

 

 

 

IHFSlS

 

LIBRARY ‘
Michigan State
University

 

 

 

This is to certify that the

dissertation entitled

SCANNING TUNNELING MICROSCOPY
STUDY OF Dy,Gd DISILICIDE NANOSTRUCTURES ON Si (001)

presented by

Bangzhi Liu

has been accepted towards fulfillment
of the requirements for

Ph.D. Materials Science

degree in
and Engineering

 

 

 

9th /\——-

Major pro sor

Date '2/ I’L—l 07’

MSUi: an Affirmative Action/Equal Opportunity Institution 0- 12771

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/Date0ue.p65-pr15

-_————— _ _ __.

 

SCANNING TUNNELING MICROSCOPY
STUDY OF Dy,Gd DISILICIDE NANOSTRUCTURES ON Si (001)

By

Ban gzhi Liu

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Department of Chemical Engineering and Materials Science

2002

ABSTRACT

SCANNING TUNNELING MICROSCOPY
STUDY OF Dy, Gd DISILICIDE NANOSTRUCTURES ON Si (001)

By
Bangzhi Liu
The epitaxial growth of rare earth (RE) silicides on the silicon (001) surface has
received some recent interest after it was discovered that self-assembled silicide
nanowires form in the initial stages of growth. In this dissertation, Dy and Gd are chosen
as prototypical systems for the study of RE silicide growth on Si (001) using scanning
tunneling microscopy and low energy electron diffraction, with an aim of optimizing

nanowire growth.

When depositing Dy and Gd on the Si (001) surface, three types of nanostructures can
be formed: 2D reconstructions, nanowires, and 3D islands. The growth behavior of all of
these structures is a function of growth temperature, annealing duration, and metal
coverage. Two types of 2D reconstructions (2x4 and 2x7) are found, and their close
relationship is revealed. Dy and Gd disilicide nanowire growth is achieved on both
normal Si (001) and vicinal Si (001) substrates. 3D silicide islands which grow in
competition with the nanowires can be suppressed by manipulating growth conditions. A
relationship between the grown nanostructures and the crystallographic structures of bulk

RE disilicides is established.

ACKNOWLEDGMENTS

I would first like to express my sincere thanks to my advisor, Professor Jun Nogami. I
was so lucky to be his first student here at MSU, and have been fortunate to learn the
details of running STM from him hand to hand. I know how much time he has spent on
me in doing things like taking images, processing data, reparing problems in STM,
writing manuscripts and finally finishing this thesis. I have also been deeply impacted by
Jun’s character qualities and his pleasant personality. He always greets others with a
smile and I never saw him getting angry, even when I made mistakes! Without Jun’s
excellent supervision and huge contribution, it would not have been possible for me to

complete this endeavor so successfully nor to graduate so soon.

Thanks also to my labmates Michael Katkov, Chigusa Ohbuchi, and Gangfeng Ye for
the help you gave me and for the companionship and fun we had together these last few

years. Each of you have truly become a part of my life.

I also want to express my appreciation to Professor Norman Birge for his help with
transport measurements and to Professor Martin Crimp for his guidance on TEM
experiments. Thanks also to these two professors who, along with Professor Tim Hogan,

served on my graduate committee and for all the help you gave me on this thesis.

I would like to give my thanks to my lovely wife, Yi Zhang. She faithfully stood by
me during these last four years and especially during this arduous period of thesis
writing. I was always deeply moved by her when I still saw her awake wating for me
back home as late as 2:00 am in the morning. Without her love and support, none of this

would have been possible.

iii

Special thanks to my friends, Richard and Gretchen Roberts for making me and my
wife a part of their family so we could really know them and know what life is like in an
America family and for praying for us to successfully complete this difficult time with
peace and joy in our hearts.

Finally, I would like to give my thanks to my parents for allowing me to go so “far
away" to pursue my dreams. I understand that it is a big sacrifice to them for my not
being around them and for their not being able to see their son and his wife for such a
long time. I also want to extend my thanks to my sisters and their families, to my wife’s

family, and to friends here in the US. and back in China for caring about me.

iv

TABLE OF CONTENTS

 

 

 

 

 

 

LIST OF TABLES - ...... _ - - -- vii
LIST OF FIGURES - - -- - - _ - - viii
CHAPTER 1 INTRODUCTION - -- -- - 1
1.1 BACKGROUND: RE SILICIDE NANOWIRES ................................................................... l
1.2 SILICIDES .................................................................................................................... 4
1.2.1 Transition metal silicides ................................................................................... 4
1.2.2 RE silicides ......................................................................................................... 4
1.2.3 Epitaxial growth of RE disilicides ................................................................... 10
1.2.3.1 Epitaxial growth on Si (111) ..................................................................... 11

1.2.3.2 Epitaxial growth on Si (001) ..................................................................... 13

1.2.4 Resistivity of RE silicides ................................................................................ 16

1.3 THESIS ORGANIZATION ............................................................................................. 17
CHAPTER 2 EXPERIMENTAL BACKGROUND 19
2.1 EXPERIMENTAL METHODS ........................................................................................ 19

. 2.2 SI (001) SUBSTRATES ................................................................................................ 20
2.3 SCANNING TUNNELING MICROSCOPY AND IMAGE PROCESSING ................................. 25
2.4 INDEXING OF HEXAGONAL CLOSE-PACKED STRUCTURE ............................................ 27
CHAPTER 3 THE Si (001)(2x4)-Dy SURFACE - -- - ---28
3.1 EXPERIMENTAL OBSERVATIONS ............................................................................... 28
3.2 DISCUSSION .............................................................................................................. 35
3.3 CONCLUSIONS .......................................................................................................... 39
CHAPTER 4 Dy SILICIDE NANOWIRE GROWTH ON Si (001) --.---40
4.1 EXPERIMENTAL OBSERVATIONS ............................................................................... 40
4.1.1 Coverage dependence ....................................................................................... 40
4.1.2 Width and height distribution of nanowires ..................................................... 43
4.1.3 Nanowire top surface reconstruction ................................................................ 45
4.1.4 Nanowire nucleation ........................................................................................ 47
4.1.5 Effect of annealing duration ............................................................................. 48

4.2 DISCUSSION .............................................................................................................. 49
4.3 CONCLUSIONS .......................................................................................................... 55
CHAPTER 5 THE Si (001)(2x7)-Gd, Dy SURFACES ............................................. 56
5.1 EXPERIMENTAL OBSERVATIONS ............................................................................... 56
5.1.1 General observations ........................................................................................ 56
5.1.2 Bias dependence ............................................................................................... 61
5.1.3 Atomic Structure ............................................................................................... 63
5.1.4 Coverage dependence ....................................................................................... 65

5.2 DISCUSSION .............................................................................................................. 69

5.3 CONCLUSIONS .............................................................................................................. 71

CHAPTER 6 A COMPARISON OF Dy AND Gd SILICIDE NANOWIRE

 

 

 

GROWTH - 72

6.1 GROWTH ON NORMAL SI (001) ................................................................................. 72
6.1.1 General observations ........................................................................................ 72
6.1.2 Gd silicide nanowire growth ............................................................................ 73
6.1.3 Dy silicide nanowire growth ............................................................................ 76
6.1.4 3D islands ......................................................................................................... 78
6.1.5 Height quantization .......................................................................................... 81
6.1.6 Coverage dependence ....................................................................................... 82
6.1.7 Growth behaviour at 500°C ............................................................................. 86
6.1.8 Room temperature growth with PDA ............................................................... 87
6.1.9 Higher temperature growth .............................................................................. 89

6.2 GROWTH ON VICINAL SI (001) .................................................................................. 89
6.3 DISCUSSION .............................................................................................................. 94
6.3.1 Effects of annealing and growth temperature .................................................. 94
6.3.2 Multiphase coexistence and atomic model ....................................................... 94
6.3.3 Lattice mismatch and strain relaxation ............................................................. 98

6.4 CONCLUSIONS ........................................................................................................ 101
CHAPTER 7 FUTURE WORK -- - - -- - ....... 102
7.1 EPITAXY ................................................................................................................. 102
7.2 LATTICE MISMATCH ............................................................................................... 104
7.3 STRAIN RELAXATION .............................................................................................. 107
7.4 POSSIBLE SYSTEMS TO FORM NANOWIRES .............................................................. 108
7.5 FUTURE WORK ........................................................................................................ 110
7.5.1 Growth of nanowires with other systems ....................................................... 110
7.5.2 Strain relaxation in RE silicide/Si (001) systems ........................................... 112
7.5.3 TEM work ...................................................................................................... 113
APPENDIX _ - - - - _ _ ...... -115
REFERENCES - .......... 116

 

Vi

LIST OF TABLES

Table 1.1 Technology roadmap for semiconductors

Table 1.2 Known bulk metal silicides

Table 1.3 Comparison between three classes of transition metal silicides
Table 1.4 RE metals in periodic table

Table 1.5 Space group and stoichiometry of RE disilicides

Table 1.6 RESig lattice parameters and mismatches with Si substrate
Table 1.7 Relationship between rare earth metals and disilicide structures

Table 1.8 Growth conditions, and orientation of appeared phases in RESlz growth on
Si (001)

Table 1.9 Resistivity of RE silicides (RT)
Table 4.1 Measured mismatches between the C(2x2) structure and Si substrate
Table 6.1 Measured amplitude, wavelength and height of wavy islands.

Table 7.1 Lattice matching of some RE disilicides with Si (001)

vii

LIST OF FIGURES

Figure 1.1 (a) Hexagonal AlB2 Structure, and (b) tetragonal ThSlz or orthorhombic GdSlz
structure.

Figure 1.2 Relationship between (b) hexagonal Ale structure and (c) tetragonal
ThSiZ/orthorhombic GdSiz structures, which are built from the same unit block (a).

Figure 1.3 Lattice matching between (a) RESiz (0001) and Si (111), and (b) RESlg (1100)
and Si (001).

Figure 2.1 (a) Diamond structure of Silicon, and (b) ideal Si (001) terminated surface.

Figure 2.2 Features observed under (a) filled and (b) empty states STM imaging, and
their correspondence with the dimer model. The 2x1 unit cell is marked by the two
rectangular boxes in both images.

Figure 2.3 Schematic drawing of Si:2xl reconstruction, as well as SA and SB types of
steps.

Figure 2.4 Filled states image of the two types of steps on Si (001) surface.
Figure 2.5 Empty states image (19 nmx19 nm) of 4° tilted vicinal Si (001) surface.
Figure 2.6 Indices of hep structure.

Figure 3.1 STM topography of Si (001) surface with 0.5 ML Dy deposited. At a sample
bias of +1.39 V, the maximum apparent height of the nanowire is about 0.4 nm, the
width is about 5 nm.

Figure 3.2 2x4 LEED pattern of 0.53 ML Dy on Si (001), grown at 600°C. The beam
energy is 49.2 eV. The Si: 1x1 spots are indexed. Two 2x4 reciprocal unit cells are
marked by two white boxes which are 90° rotated, reflecting the two domain
structure of the Si (001) surface.

Figure 3.3 Large empty states image (50x50 nmz) of the reconstructed surface with 0.5
ML Dy deposited, VT=+1.05 V. The inset is an FFT of this image. The spots in the
right half of the FFT pattern are accentuated by the black dots. The Si: 2x1 reciprocal
unit cell is marked by the crosses, and the 2x4 by the open circles. A close-up of the
area enclosed in the square box is Shown in Figure 3.4, and that in the rectangular
box is Shown in Figure 3.5.

Figure 3.4 (a) and (b) are two empty states images of the area enclosed in the square box

(10x10 nmz) in Figure 3.3, taken at +1.05 V and +0.46 V, respectively. Clean Si
areas (A) and metal reacted areas (B) are marked. Image (a) Shows all the maxima in

viii

the metal covered areas have similar height. Image (b) Shows two types of maxima,
one higher than the other. The broken white boxes in both images Show the 2x4 unit
cell. The two small boxes in (b) mark the HLH (type I) and LHL (type II) features.

Figure 3.5 (a)-(c) are three empty states images of the area enclosed in the rectangular
box (15x12 nmz) in Figure 3.3, taken at +0.46, +1.05 and +1.85 V, respectively. The
heights marked by the black and white arrows along the profiles correspond to the
positions marked by the same arrows along the dotted traces.

Figure 3.6 Filled states image (15x15 nmz) of the surface with 0.17 ML deposited, VT:-
1.93 V. The surface is separated into areas of clean Si (A) and areas reacted with the
metal (B).

Figure 3.7 Registry of the maxima features in the 2x4 structure with respect to adjacent
clean Si dimer positions. A unit cell is marked by a rectangular box.

Figure 3.8 3D model of 2x4 reconstruction, in which the orientation relationship between
nanowires and Si (001) surface is marked. Bonds are added to the Dy atoms for
illustration purpose only.

Figure 3.9 Local position of Dy atom in the bulk silicide, which is similar to that on the
2D surface.

Figure 4.1 (a)-(d) are STM images (all 300x300 nm) which Show coverage dependent
topography of the Dy/Si (001) surface.

Figure 4.2 3D islands formed at the junction of the nanowires. The image size is
150x150nm, taken from the same sample surface as Figure 4.1(b). The inset is a
height profile of the trace marked by the broken line.

Figure 4.3 Filled states STM image (50x50nm, VT: - 1.2 V) Shows that Dy silicide
nanowires coexist with 2x4 and 2x7 reconstructed Si (001) surface. The nanowires
bundle together and form wider structures. A Single nanowire can be seen in the
lower right comer. The Si dimer row direction is marked in the image.

Figure 4.4 Examples of height and width measurement of the nanowires. The height
profiles of traces (1) and (2) in the left-hand Side panel are shown in the right-hand
Side panel, respectively. The image Size is 70x70 nm, VT=+1.04V.

Figure 4.5 (a) Width distribution of bundled nanowires, and (b) height distribution of
additional layers on nanowires.

Figure 4.6 Various reconstructions of the nanowire top surfaces. All the images have the
same Size, 10x10 nm. Roman numerals in images (b), (d)-(f) correspond to the
marked positions in Figure 4.4.

Figure 4.7 Nanowire nucleates (a) on a flat terrace, forming a new step around it by
consuming the Si atoms in the neighborhood, and (b) at a step edge, interacting with
Si steps on both ends during growth.

Figure 4.8 STM image showing that only a few nanowires survive when continuing to
anneal a sample at 600°C for 30 min after deposition of 0.68 ML Dy, mostly large
square and elongated islands are formed. The image size is 400x400 nm.

Figure 4.9 (a) Hexagonal Ale type Structure of Dy disilicide; (b) lattice matching
between the silicide (1 T00) plane and Si (001) surface; (c) Side View of Dy disilicide
growing on Si (001) substrate.

Figure 5.1 STM images (50 nm x 50 nm) Showing nanowires coexisting with 2x4 and
2x7 superstructures on a (a) 0.18 ML Dy deposited Si (001) surface, VT: -1.25 V,
and (b) 0.63 ML Gd deposited Si surface, VT: +0.6 V.

Figure 5.2 (a) STM image (60 nm x 60 nm) showing a 2x7 reconstructed surface with
0.46 ML Gd deposited at 500°C, VT: +1.17 V. (b) LEED pattern from the sample
shown in (a). (c) FFT pattern of image (a). The black box and white box in (b) and
(c) denote Si: 1x1 and Gd: 2x7 reciprocal unit cells, respectively.

Figure 5.3 Close ups (15 nm x 18 nm) of (a) Gd: 2x7 superstructure, VT: +1.17 V, and
(b) Dy: 2x7 and 2x4 mixtures, VT: +1.24 V. Image (a) is taken from the same
sample as Figure 5.2.

Figure 5.4 STM images (12 nm x 15 nm) Showing 2x7 bias dependence in the empty
states (a)-(c) and in the filled States ((1) and (e). Relative Shifts of the maxima in both
the subunit rows and the bright strips are marked near the straight lines in (b), (d) and

(c).

Figure 5.5 Bias voltage ranges for three different regimes in empty states imaging of the
subunits in the Gdz2x7, Dy:2x7 and Dy:2x4 structures. Bias voltages for the
individual measurements are shown.

Figure 5.6 Diagram showing the atomic registry of the maxima in the 2x7 superstructure
with respect to the Si (001) substrate. The 2x7 unit cell is marked by a black box, and
a broken line is drawn to Show the anti-phase Shift of both the subunit maxima and
bright strip maxima.

Figure 5.7 Coverage dependence of 2D and 3D phases appearing for (a) Gd grown on Si
at 600°C with post deposition annealing; and Dy on Si at 600°C without (b) and with
(c) post deposition annealing.

Figure 5.8 The calculated coverage of the three cases shown in Figure 5.7 is plotted
against the nominal coverage.

Figure 5.9 Sum of squared deviation between the actual and calculated coverage plotted
versus the assumed number of metal atoms in the 2x7 superstructure.

Figure 6.1 Topography of Si (001) surface with 0.4 ML Dy deposited, showing
coexistence of nanowires and 3D islands, and second layer growth.

Figure 6.2 (a)-(c) (350 nm x 350 nm) and (d) (300 nm x 300 nm) are STM images which
Show coverage dependent topography of the Gd/Si (001) surface. (6) (60 nm x 60
nm) is a closer View of the area enclosed in the white box in (b). (f) (75 nm x 75 nm)
is a closer View of the white box in (d).

Figure 6.3 Width distribution of Single Gd silicide nanowires in bundle.

Figure 6.4 (a)-(c) are STM images (all 350 nmx350 nm) Showing coverage dependent
topography of the Dy/Si (001) surface prepared by PDQ. (d) (75 nmx75 nm) is a
closer View of the area enclosed in the square box in (b).

Figure 6.5 Topography of (a) 0.6 ML Dy deposited sample prepared by 10 min PDA,
image size is 200 nmx200 nm, and (b) the same sample shown in (a) which was
annealed another 1 hour, image size is 350 nmx350 nm. The inset in (b) is the height
profile of island (1) along the broken line.

Figure 6.6 (a) 3D rendering of wavy island (2) shown in Figure 6.5(b). (b) The height
profile of the wavy island along the broken line in (a). (c) Wavy island formed by
deposition of 0.92 ML Gd.

Figure 6.7 Coverage dependence of 2D and 3D phases appearing for (a) Gd grown on Si
at 600°C with PDA; and Dy on Si at 600°C with (b) PDQ and (c) PDA.

Figure 6.8 The calculated coverage of the three cases shown in Figure 6.7 is plotted
versus the nominal coverage.

Figure 6.9 STM images (180 nmx180 nm) Showing coverage dependent topography of
the Gd/Si (001) surface grown at 500°C with (a) 0.46 ML Gd deposited, taken at VT:
+0.44 V; and (b) 0.54 ML Gd deposited, taken at VT: +1.12 V. (c) is a high
resolution image (40 nmx40 nm, VT: —0.67 V) of the 2D reconstruction appeared in
(b). (d) is an STM image (180 nmx180 nm) of the same sample shown in (b) which
was annealed at 600°C, taken at VT: +1.35 V.

Figure 6.10 (a) and (b) (Size: 60 nmx70 nm), and (c) and ((1) (size: 45 nmx45 nm) are two
pairs of dual bias images of a sample prepared by depositing 0.16 ML Dy onto Si at
room temperature then annealing at 600°C.

Figure 6.11 Scanning tunneling microscope (STM) images showing (a) a clean vicinal Si

(001) surface with 4° miscut (19 nmx19 nm), and (b) the vicinal surface after
deposition of Dy at 600 °C with 10 min postannealing (300 nmx300 nm).

xi

Figure 6.12 (a) STM image (800 nmx800 nm) Showing parallel nanowires formed on a
0.38 ML Dy deposited vicinal Si (001) surface without PDA. A small number of 3D
islands Still can be seen. After annealing this sample 30 min, the topography is Shown
in (b) of image size 420 nmx420 nm. 3D islands of various shapes appeared, and
only a small number of nanowires survived.

Figure 6.13 STM images Showing a parallel nanowire covered surface free of islands
formed by depositing 0.77 ML Gd on a vicinal Si (001) substrate (300 nmx300 nm).
The sample was prepared with 5 min PDA.

Figure 6.14 Side views of (b) hexagonal and (c) tetragonal RE silicides grown on Si (001)
substrate, both of which can be built from the unit block shown in (a).

Figure 7.1 Equilibrium phase diagram in function of the coverage H and misfit e. The
small panels on the top and the bottom illustrate the morphology of the surface.

Figure 7.2 Lattice matching in heteroepitaxial systems, where 0, denotes substrate lattice
parameter, and a0 denotes the overlayer lattice parameter. The overlayer lattice is the
Shaded area.

Figure 7.3 Registry between overlayer and substrate.

Figure 7.4 (a) Bright field image (lumx0.9um) of a Dy/Si (001) sample at a metal
coverage of 3.4 ML. (b) The diffraction pattern from the same sample.

xii

Chapter 1 Introduction

1.1 Background: RE silicide nanowires

Rare-earth (RE) silicides have been studied since 1979 as a natural extension of the
study of transition metal silicides [1], and attracted more attention Shortly after Tu et al.
found that they have the lowest Schottky barrier height on n-type Silicon [2]. In 1986, it
was found that the RE disilicides grow epitaxially on Si (111) [3]. Thereafter, most of the

work on RE silicide growth has been on Si (111) substrates.

In 1998, Preinesberger et al. found that nanowires can be formed by depositing a
small amount of Dy on the Si (001) surface with subsequent high temperature annealing
[4]. However, nobody paid attention to this phenomenon until the November 1, 1999, the
New York Times reported that Hewlett-Packard (HP) research labs in Palo Alto made a
breakthrough in making nanowires on Si surface [5]. Later, Chen et a1. [6] from the HP
group reported that the formation of the self-assembled erbium disilicide nanowires on Si
(001) is due to the anisotropic lattice mismatch between the hexagonal AlBg type
structure of the disilicide and Si (001) surface, which can work as a strategy for the
intentional epitaxial growth of nanowires for other RE disilicides. Following the NY
Times article, we arrived at this conclusion independently and started studying the
growth of RE silicides on Si (001) extensively, and found that Ho, Y and Gd disilicides
form nanowires, but that other silicides, such as NdSlg only form compact islands,
depending on the details of the lattice mismatch between the RE silicides and the Si (001)
surface [7-9]. Fairly recently. another RE silicide, scandium disilicide, was also found to

form nanowires [10].

As Shown in Table 1.1, the minimum feature Size achievable with current
photolithographic technology is 65 nm, and this minimum is projected to reach 9 nm in
15 years as the scaling down of integrated circuits continues. RE silicide nanowire widths
can be as small as 0.8 nm, which is probably beyond the limits of photolithography.
Therefore, it would be of technological interest to grow and of scientific interest to
understand the properties of these Wires at such a small scale. To date, very few data have
been published on the RE silicide nanowires [4, 6, 7, 10-12], and their growth behavior
and properties are still unclear.

Dysprosium iS a prototypical nanowire forming RE metal. Preinesberger et al.
published only the barest outlines of the growth behavior in this system in their first
paper [4]. In this thesis, the growth of Dy on Si (001) is studied in detail, with different
chapters focusing on various aspects such as Silicon surface reconstruction, detailed
nanowire morpholog , growth kinetics, and formation of non-nanowire silicide islands.
The primary experimental techniques were scanning tunneling microscopy (STM) and
low energy electron diffraction (LEED). In addition, some data on Gd growth is
presented and compared with Dy, and some insight into the behavior of both metals is
gained by this comparison. The contents of this thesis correspond to four or five papers
that have either been published, or are at some stage along the path to publication [13-
17]. In the context of the other work on RE silicide nanowires, almost all of which
appeared in print after this thesis research was well underway, these studies on Dy and
Gd present a significant addition to this literature.

In the remainder of this introductory chapter, the current understanding of RE

silicides and RE silicide thin film properties will be reviewed.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

££ £6
m. N: . . . Ewe: £sz Ema. £ch a E 3ng . 560
unmet.“ Ema.» EmEe :mem :55: Ema 2%: EEO £=m Mam £2 ££ £3
$3 in; EEG. 55 ion $5 55. £5 55m 385 A662 mmmm 530
£25 £3.» £25 £5 £8: £§d £5 £60 . . £ a £3
«6.5
£& N £§ ££L .
Ema :2 £3 £8 aux mama £5 £E .3: £3
£5 .9 £3 £5 £8m as? £39 Em swam £34
£5 . £3 £5 £5 a 3 £25 . £23
. £5 m m
m N Ewe—2 NEON N m a >
£8 £5. £§ s 8 £2 .ch £ LN £>
. S r .m 2 m w . D E
.m cm _ a . _m 32 .m N s. A
«mmwm imp—M .mm M MmmmO—Z mmvnz EWEN ~mth -mm >
£8; ££m £ 3. £82 . £3 £5 £3,
m—Wmum
2 o ££ £52 £6 a £F -.
3&2 532 M000 Eon— EEZ ECU 2%.» EC. Emmom
£26 £2 £sz £26 £5 £262 £5 £9 £5 mmwm
m—mmuz ~Wm—Z mmmOU mmmvnm MWMCE mmmuu . :mW—r—t MM Mom-w
£32: 252: 636: .Soz £52: reuuacom 5.. ~—
m_SoE 556:9;
«Eel
ENS: 826:6 EOE 0:3 SEEM N; 635.
- . a? com Seems. Saw
a m“ w _ mm mm mv mo Ow _ com - ASE—85.2286
om: emu omw com .
no ow 02 of
8 mm 3 a 89 32 3% E as;
o _ om m 5m 0 _ om 50m meow 80m 30m mm

 

 

 

 

 

 

 

 

 

 

 

 

S: 3963:8155 Sm 92:38 328:8... _.~ 653.

 

 

1.2 Silicides
1.2.1 Transition metal silicides

Many elements in the periodic table react with silicon to form different types of
compounds. The compounds formed between metals and silicon are called silicides. Of
all these, the transition metal silicides are the largest group, of which RE silicides are a
subclass, as shown in Table 1.2 [19-21]. The other two subclasses are refractory metal
silicides, and near noble metal silicides. The silicides occur in three major
stoichiometries: metal rich silicides, in the form of M5813, M5Si4; monosilicides, in the
form of MSi; and silicon rich Silicides, in the form of M812, M3Si5. In much of the
published work in these Silicides, the compounds are prepared by depositing metal or co-
depositing metal and silicon onto Si substrates with high temperature annealing. The final
phase formed is the most Si-rich silicide which is frequently the disilicide [19, 22-24]. In
addition, most of the silicides used in integrated circuit technology are disilicides, such as
T1813, C0812, W812, T1813 and F6813, etc [19, 22, 25]. Therefore. our interest is narrowed
down to the discussion of disilicides.

The RE silicides are classified as a separate class of Silicides due to their distinct and
interesting physical and chemical properties, amongst them a low temperature of
formation, the lowest known Schottky barrier heights on n type Si, good electrical and
thermal conductivity, and unusual kinetics of formation [23, 26]. A detailed comparison
of the three classes of Silicides is listed in Table 1.3.

1.2.2 RE silicides
Speaking of RE metals, traditionally this category refers to lanthanides inclusively,

but now it also tends to include Y and Sc since they have very similar properties [27].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 02 e: E E O em EU :2. 5 Hz D 8 fi 2
3 A; E am a: 5 Be .5 am Em Em uz 5 so 3
85526 536.23 5on SEED—O 536-28 Ems
cm 2 on. E E E. w: 3.. a : 5 am 3 me E 3 am SO
ox L we £ £ 5 8 w< E am am 8e oz 32 .N > 5 .E
Q um um m< so ”WO 5 5U _z 00 on :2 .O > F om aO M
E 6 m A. a Z w: «2
02 E o z O m am 5
a: m
use 28:8 E 23.2: mm 3 29a.
:2 02-8 2182.: a: 29% 83 83-3. £22m:

 

.NQ corona

UO__o::oO-:o_m:.L.EO

.mfl cocoa“:
8:3::oOéoEEtEBE

_mm ._m_ 226:8
co__o::oO-:o:3_O_E

55.258 me 8:25—

 

 

 

 

 

 

am 2:52: .3 .2 :m am .8. £ 868% magmas—s Eases.

Rm .9: 908-com EN .0: PS? .3 .E Oocovzmmm 252358 59258

EN .mmzmnz 5 .ms £2 :2 £2 BEE 8% E:

as >a 93-8.3 as >a 53.3 as RES 5 8%.: .5 Ewe: BEE 3:23
8235:.

.32: 22¢: :32

 

822:2“. .32: Defining

 

8222.. 5.8 2am

 

 

826:6 :32: 522%: CO Emma—O 66:: 5653 :omtmanU m; 2an

 

Chemically, the rare earths are classified as members of the IIIB subgroup of the periodic

table, as shown in Table 1.4. Based largely on trivalent ionic radii, the RE elements are

broadly classified into two groups:

0 Light rare earth elements (LREE), or the cerium subgroup, comprise the first seven
lanthanides (atomic number 57 through 63):

0 Heavy rare earth elements (HREE), or the yttrium subgroup, comprises the remaining
lanthanides and yttrium (atomic numbers 64 through 71 and 39).

Yttrium is categorized with the HREE because of its occurrence, ionic radius and other

similar properties. Scandium (atomic number 21) occurs widely distributed in nature at

low concentrations and is not typically classified as either a LREE or HREE [27].

Table 1.5 Space group and stoichiometry of RE disilicides [21, 26, 33, 34]

 

 

 

 

 

 

 

structure Space group stoichiometry
tetragonal ThSlz I4/amd RESiz
orthorhombic GdSlz Imma RESi1_75..1_9
hexagonal AlBg P6/mmm RESi 1,67(Re3Si5)

 

 

Accordingly. the RE disilicides are classified into two groups, light RE silicides and
heavy RE Silicides. They crystallize in three different structures: tetragonal ThSlz,
hexagonal A182, and orthorhombic GdSlz. Usually the stoichiometry of the disilicide is
written as RESig.x due to the existence of Si vacancies [33, 35]. The tetragonal phase has
a composition close to the stoichiometric disilicide (x=0) and the hexagonal phase forms
the RE3Si5 stoichiometry (x=0.25), whereas the orthorhombic phase exists for
intermediate compositions, as Shown in Table 1.5. It is believed that there is one silicon
atom out of six is missing in the hexagonal A1B2 type Structure, giving a stoichiometry of

3/5'~‘1.67 or 1.7. Therefore, the disilicides are usually written as RESi1_7, RESilm, or

RESig.x with x203 [1. 36—38]. The notation RE3815 is reserved for such compounds as
Yb3Si5 with their own structure, not for the defected AlBg structure [1].

The three types of structures are closely related. The hexagonal Ale type Structure is
of the P6/mmm space group; the unit cell is shown in Figure 1.1(a). The tetragonal ThSig
type structure is of the I4/amd space group; the unit cell is Shown in Figure 1.1(b). The
orthorhombic GdSiz type structure is of the Imma Space group, which is a deformation of
the tetragonal ThSiz type structure. As shown in Figure 1.2, the three structures can be
described as different arrangements of trilayers of the same building element: a
parallelepiped with eight Silicon atoms as apexes, and a RE atom at the center. These
trilayers are stacked along the [001] direction in the tetragonal and orthorhombic
structures, and along the [10T0] direction in the hexagonal one. The hexagonal structure

also differs from the others by the stacking sequence: successive plates are systematically
translated by a/2 in the hexagonal structure, while they are alternatively translated by a/2
or b/2 in the tetragonal and orthorhombic structures. In addition, the largest face of the
building element is square in the tetragonal structure, and rectangular in the orthorhombic

and hexagonal structures [33]. As a result. the lattice parameters of the tetragonal and

hexagonal structures are related, (1:1):((1,,,.,.+c,,,._,)/2 and ('22 am ,5 [39]. Most of the

known RE disilicides lattice parameters are listed in Table 1.6.

For bulk disilicides, the type of structure adopted has a dependence on the particular
RE metal, as shown in Table 1.7. Although there are discrepancies between different
references, a tendency can be seen that the tetragonal and/or orthorhombic Structures
appear mainly for light RE Silicides, the hexagonal structure appears mainly for heavy RE

Silicides, and a mixture of these structures appears for the RE silicides of medium weight.

C).-. .

«g. a. .

 

A471,.

0
\
US$910

(a) (b)

Figure 1.1 (a) Hexagonal Ale structure, and (b) tetragonal ThSiz or orthorhombic GdSiz
structure.

a. a
a. lav
v

4

O
I
1'

I C
I‘ Q a
a ' 6
a 4'.
I a

 

A
O
v

Figure 1.2 Relationship between (b) hexagonal Ale structure and (c) tetragonal
ThSi2/orthorhombic GdSi2 structures, which are built from the same unit block (a).

NW4 Vh.V—HMO.~N .< VWMHHGG Am

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NLmUO 63805055
cos cos ow.£ 009 :34 :34 “GE. BeowEB £ch
~m=< :Eowwxo:
3d 86 3.3 $09 god 036 £95 oEEofiocto
£fi 35mg“: £>
m2. mod- mw.2 ~96 omwd mm? Ecowmxuc
:mv- _ .6 oQO mwwd wood oEEoEocto
mmm- mmm- N3: 3:4 wwwfi wmna Ecoweba "mmom
cod wmé- 3.3 PM: $0M. Nm2 Ecowmxoc
Nico omnEoEocto
2 .m 2 .m 3.2 3.2 cafi cod swab 320mg“: «figm—
cvc mo. H - mmhfi waV wand mm? Ecowwxo:
KN 3.4. 3.2 3.2 349.4“ mod mmmco ORE—office
NEE. Ecowgou Ewe:
mod mod- no.2 5:4 GEM. mm? Ecowexo;
Sum Kw #52 2%.: mead X34 DmuO 65565558
mod moi VGE wee: 8% 3% «if. Ecowebe £sz
mm..\. mmd- 052 3:4 _mw.m ~m_< Ecowmxoc
91V GB 9}: 3;. L 5% mod emuo oEEoEocto
2.6 KG _w.£ 6.2 36 0:4 Emit Ecowwbe £va
mow cod 2.3 mid Ema mm? Ecowexo:
Q; O Q; a GE 5
saga—5:2 $3 3.5. 3.9 O Q: a JD a BBS—hm 639:5

 

 

 

 

 

 

 

8v .mm gm .2 23353 a 2:3 monenEmE GEN EBoESmm 83:: Emma m4 2an

 

Table 1.7 Relationship between rare earth metals and disilicide structures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Sc Y
T [41] O/H [33]
Light rare-earth disilicides Heavy rare-earth disilicides
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
T/O O/H H [33]
T/O [42] H/O/T H

 

 

 

 

 

H: hexagonal. T: Tetragonal, O: Orthorhombic

1.2.3 Epitaxial growth of RE disilicides

Silicide films on monocrystalline Si substrates, in general, exhibit a polycrystalline
microstructure [22]. But almost all transition-metal silicides can grow epitaxially to a
certain extent on Silicon under controlled growth conditions [22, 40, 43, 44]. All silicides
currently used in device structures are usually polycrystalline due to the difficulty in
making epitaxial Silicides [45, 46]. and that the crystallinity and interface perfection are
not of overriding importance in these applications [47]. However, there has been
tremendous attention on the growth of epitaxial Silicide films out of scientific interest in
studying perfect interfaces and technological interest in potential applications in novel
devices [22, 45, 48-50]. Also, epitaxial silicide layers have an increasing advantage over
polycrystalline Silicide layers when junction depths continue to shrink. The good layer
uniformity and high thermal stability of epitaxial Silicides are very attractive for Shallow
junction formation. Therefore, the development of reliable and convenient epitaxial
silicide technologies is likely to be important for sub 0.1 pm devices [45].

It is generally accepted that the reactions between deposited RE metal films and Si
substrates are nucleation controlled [24, 26, 51]. For nucleation controlled kinetics, the

film nucleates in isolated regions, and then grows rapidly both vertically and laterally,

10

resulting in a rough, pitted surface [23, 24, 38, 52, 53]. Additionally, nucleation usually
occurs over a very limited temperature range (325-400°C for RE silicides grown on Si
(111)), below which nothing happens and above which the reactions are too fast to
measure [25, 32]. It was also suggested that the contamination layer at the interface is the
primary cause of surface pitting [38, 40]. But this seems to be questionable Since pinholes
can be formed in the RE silicide films grown on a pre—cleaned substrate in UHV (ultra
high vacuum) [54-56].

The most common methods used to produce thin epitaxial Silicide films on top of
silicon are solid phase epitaxy (SPE, i.e., depositing a thin metal layer on an atomically
clean substrate at room temperature or elevated temperature under UHV conditions, with
or without subsequent annealing at higher temperature) and molecular beam epitaxy
(MBE, the metal and silicon are co-deposited onto the Silicon substrate under Similar
conditions as SPE) [36, 39, 40, 43, 57-59]. Other techniques have been developed to
improve the surface morphology, such as rapid annealing [3, 60], and intermediate
template layers [37, 61, 62]. Sometimes an additional thin layer of Si or metal, such as Pt,
W, Mo. etc., is deposited on top of the RE Silicide films to keep them from oxidizing [3,
23, 60].
1.2.3.1 Epitaxial growth on Si (111)

AS mentioned above, the bulk heavy RE disilicides crystallize in the hexagonal Ale
type structure. However, the real Situation is more complicated when growing RE
disilicide thin films on Silicon substrates.

Good-quality epitaxial growth of hexagonal RE silicides is expected on Si (111)

substrates since the atoms in the plane normal to c axis in the silicides are essentially

11

identical to those on the Si (111) surface, as shown in Figure 1.3(a), resulting in a nearly
ideal epitaxial arrangement. Typically, the mismatch is within 1.5%, as shown in Table
1.6. Therefore, major attention has been concentrated on epitaxial growth of RE silicide
films on Si (111) surface in the past. It has been found that all of the heavy RE metal
disilicides, including YSig-,,. can grow epitaxially on Si (111) substrates, and nearly all of

them crystallize in the hexagonal AlBg type structure with a unique epitaxial relationship:
RESig-,(0001)// Si (111), RESi2_x[l2lO]//Si[OT 1] [1, 3, 40, 58, 61-66]. A few references

showed that orthorhombic structure could appear for GdSiz.x after the hexagonal structure

is formed, with a dependence on thickness and annealing temperature [67, 68].

 

Figure 1.3 Lattice matching between (a) RESig (0001) and Si (1 l l), and (b) RESlz (1100)
and Si (001).

Many techniques, such as LEED, STM and transmission electron microscopy (TEM)
etc, have been used in studying the surface reconstruction and vacancy ordering of RE

Silicides. The hexagonal RE silicides on Si (1 11) generally Show a p(lx1) reconstruction

below 1ML and J3x x/3 reconstruction above 1 ML [54, 55, 63, 69, 70], and this
behavior may be temperature dependent [62, 71]. Experiments and theoretical models
Show that both of the 2D reconstructions result from a buckled Si top layer Without

vacancies, instead of the graphite-like Si termination expected for a simple truncation of

the bulk Silicide structure [69, 72-79,]. The J3 x\/3 periodicity is possibly a reflection of
the 3D vacancy ordering superstructure within the film, which has been seen by TEM
[63. 64].

1.2.3.2 Epitaxial growth on Si (001)

Compared to Si (1 l 1), less research activity has been dedicated to growth on Si (001)
Since the lattice match between disilicides and Si (001) is not as good as that on Si (111),
and the epitaxial relationship is less straightforward. Most of the data on epitaxial growth
of continuous RE silicide films on Si (001) are summarized'in Table 1.8, which Shows
that all of the three types of structures, hexagonal AlBg, orthorhombic GdSlg, and
tetragonal ThSlz, can grow epitaxially on Si (001), and that the hexagonal phase can
coexist with orthorhombic or tetragonal phases, depending on the detailed growth

conditions. It is generally observed that there are two types of orientations between the

hexagonal Structure and Si (001) surface, RESig.x (1100) //Si (100) with RESig.x [0001]
//Si[01T] and RESi2-x(1lOO) //Si (100) with RESi2-, [0001]//Si[0TT] [39, 64, 80, 81].

These two orientations are normal to each other, and equivalent in volume fraction and

13

GEEoEofieO ”O Jagged“; F 42035: n:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.m E _ 882.8: .5: § 88:22 .8: Se 8&8: n+9 08288880
5: 32 38 Ms .E 53:82 .socioce be: 888 .wofimmm. ..£.m
. p 535% Hmfi: H2 3502 .80: \\ 588:8 I mm oQoowémv HmBEEEB

:8 .3 28883 .2 8.582 8:2 snfouemeso: m; 3 mm 88588 03-8.. SE

8..:82 8 E .2 . ......S§S.m: .2 8.. .88 Ease: 88.8880 .

. - a
:2 e £83....8 : a: .2 85.822 .28: 8.5.x _ 09 E. 8 888 owwfimeMw +£E
.8. 0+: 82:8 88.88 2 £an
E :5 08 0888 2 3.888»

:a £8: 8 e ESE \\ a a: .2 8 e .2 58 meso: A 8-8 88$-8w 84-84 ..£.m
5. .SL Ooomm s 2.888... O 878 8?on E 2.80
RE .5289 :28: o 8mm 88 O08 £5

83 Ewe \\ x2.88. .838... 8.:
. ._ - : 0% e38 E £>

SSS .803 28:: .86

E: 2528......82 as? 53......82 $882.52: 2 8. 0888 E ..£.m
.28. m z. E

a; 53.....582 emsogéssc : 8 28w 0 4 .m

$2 80:88.86.CSE x2:82 : 8 ~58 Hm £8839
.88.... 82: a S SS .0582 .
8 85 8.5 .35.: CL
aim—5:29.. :eSSEEO -a .8535 any Esau. 25H. 3mm:—
nm £856.89 5.59.0

 

 

 

 

308 E :o £38m Emma E 822m @2383 e8 coaficoto was £52256 5380 w; 2an

l4

 

crystalline quality [81]. The orthorhombic and tetragonal structures match onto Si (001)
surface in the same way, RESig.x (001 )tetr // Si (001) with [100]]..[rl/ Si[Oll], which is

consistent with the close relationship between them. The occurrence of the orthorhombic
and tetragonal phases may have a dependence on the species of RE metal. For example,
the orthorhombic GdSig.x phase was usually reported [67, 82, 83], but for ErSi2_x,

tetragonal phase was reported [57, 85].
When we match the hexagonal RESlg (1100) to the Si (100) surface as shown in

Figure 1.3(b), anisotropic lattice mismatch results: quite small (0% for YSi2-x) along axis
a, but quite large (8.65% for GdSlg-x) along axis c, as listed in Table 1.6. This is the
proposed mechanism for the RE silicide nanowire formation: the silicides grow without
restriction along the small lattice mismatch direction, but with a restriction along the
large lattice mismatch direction, resulting in nanowire-like structure [4, 6, 7, 10-12].

Although the growth of RE silicide films on Si (001) has been studied since 1980 [1],
the discovery of the RE silicide nanowires had been delayed until fairly recently [4] for
two reasons. Firstly, people usually grew continuous thick films (e.g., 5-1000 nm, as
listed in Table 1.8), but the nanowires form only at submonolayer coverage (<1 nm);
secondly, it is the invention of the STM that made it possible to “see" nanowires at the
atomic scale.

Experiments have shown the presence of Si vacancies and 3D vacancy ordering in

tetragonal ErSi3-x [39, 84] and hexagonal R1335. (REsz, Er, and Y) (e.g., J3 x f3
superstructure, same as that within the silicides grown on Si (11 1) [64]). Several surface
reconstructions have been found, such as c(2x2) on Gd/Si (001) [86], and 3x1 and 3x2 on

Yb/Si (001) [87]. However, many important issues still remain open, such as what

15

species constitutes the topmost layer and what structure it has, due to the limited
publications.

By comparing the behaviors of the epitaxial growth of RE silicides on Si (001) and Si
(1 l 1), we can see that the substrate orientation affects the growth significantly through
the lattice mismatch, which has been realized by some other authors [36, 67, 83, 88]. The
epitaxial growth of different phases can also be controlled by using different growth
techniques (deposition, codeposition, multilayers, and use of templates), and by
controlling the growth conditions. Four factors: growth temperature, annealing
temperature, annealing duration, and thickness have been proved to be critical for the
epitaxial growth of RE silicides on Si (001), such as the resulting phases and the ratio
between coexisting phases [39, 68, 85]. In our experiments, only thermal evaporation of
pure metal with subsequent reaction with the substrate is used to form the silicide. The
growth temperature, annealing temperature, annealing duration, and metal coverage are
varied to study the growth behavior of RE silicide nanowires.
1.2.4 Resistivity of RE silicides

The majority of transition metal silicides are metallic [89], and some are
semiconducting (CfSig, B-F6812 ,etc) [90]. The resistivity of the refractory and near noble
metal silicides is in the range of 10l to 102 p!) cm at room temperature [89]. Metal
silicides have been widely used as ohmic contacts, Schottky barriers, gate electrodes, and
interconnects in microelectronic devices [19, 25, 44, 91].

The study of the electrical properties of RE silicides has been less extensive
compared to the other silicides. Most of the available data for heavy RE silicides are

listed in Table 1.9. As shown in the table, the resistivity of RE silicides is in a similar

16

range as the other transition metal silicides, but varies with formation conditions and
substrate orientations. The RE silicides have potential in applications such as ohmic
contacts and infrared detectors because of their low Schottky barrier heights [2, 92].
However, two major problems have delayed their practical application [26]. One is that
the RE silicides react with oxygen easily, forming ternary phases with different
physicochemical properties. The second problem is the rough morphology of the films

formed on Si substrates.

Table 1.9 Resistivity of RE silicides (RT)

 

 

 

 

 

 

 

 

 

 

 

 

. . . Resistivity (p9 cm)
81 d
l 'c‘ 8 Bulk Film/Si (001) Film/Si (111)
NdSiz 349 [93]
GdSiz 263 [93] 210 o [34]
110 [94]
DySi2 3020 [93] 76 T [61] 114 H [61]
264 H
ErSiz 59 [94] 217 H [85] 34 H [95]
136 H
129 H+T
SCSilj 38 H [41]

 

 

 

 

 

 

 

 

 

 

1.3 Thesis organization
This thesis is organized as follows.

0 Chapter 2 is a brief description of experimental methods, as well as a brief
introduction to the working mechanism of STM and the structure of the Si (001) clean

surface.

17

Chapters 3 and 5 describe two dimensional (2D) reconstructions of the Si (001)
surface induced by RE metal deposition. These 2D reconstructions co-exist with the
3D silicide structures that grow on the surface and are closely related.

Chapter 4 summarizes the growth behavior of Dy on Si (001), focusing in detail on
nanowire formation, and nanowire morphology.

Chapter 6 is a comparative study of Dy and Gd silicide growth. Here, Dy growth is
studied over a wider range of conditions than in Chapter 4, and the differences in
behavior of the two metals are examined. This chapter also includes data on a
particular variant of nanowire growth: growth of parallel arrays of Dy and Gd
nanowires on single-domain, vicinal Si (001) surfaces.

Chapter 7 is a summary of potential future research directions.

18

Chapter 2 Experimental Background

In this chapter, the experimental methods used, the basic properties of the Si (001)
substrate, and the basic mechanism of STM are introduced. The indexing of the Si (001)
surface and the hexagonal structure are also given since both are used repeatedly in

subsequent chapters.

2.1 Experimental methods

All experiments were performed in a UHV chamber with a base pressure of 2x10“)
Torr. The chamber was equipped with a commercial STM (Omicron, Gmbh,
Taunusstein, Germany), low energy electron diffraction (LEED) optics, and metal
deposition facilities. Clean Si (001) 2x1 surfaces were prepared by flashing above
1150°C for 30 s, quickly cooling to 980°C and holding for 10 min, then slowly turning
down the heating current. The clean surfaces were monitored by both LEED and STM.
Gadolinium and dysprosium were evaporated from tungsten filaments. Deposition rates
were measured with a quartz crystal monitor. Typical deposition rates were 0.1
monolayers (ML) per minute. The nominal coverage was determined by timed exposure
to the source. The samples were kept at elevated temperature during deposition, and then
annealed for several minutes after deposition before cooling down to room temperature.
Four different growth temperatures, 500°C, 600°C, 650°C, and 700°C, were used in
experiments. Three methods were primarily used to prepare RE disilicide films:

0 depositing RE metal onto a heated Si (001) substrate, mostly at 600°C with 2 to

10 min post deposition annealing

19

0 the same deposition conditions as above, but cutting off the heating current
immediately after deposition
0 annealing a sample at 600°C after it has been cooled down to room temperature
The metal coverage (0.1 — 3.5 ML) was varied under each condition. All STM and LEED
observations were at room temperature. The tunneling voltages are specified as sample

bias, with the STM tip held at virtual ground.

2.2 Si (001) substrates
Bulk silicon has a diamond structure of the space group Fd3m, as shown in Figure
2.1(a), with a lattice parameter cum/1. = 0.543 nm. An ideally terminated Si (001) surface is

shown in Figure 2.1(b), which is a square 2D lattice with a lattice parameter of as): (1,,qu
Ix/E =0.384 nm. A 1x1 unit cell is defined in the figure, with two basic vectors running in
[1T0] and [110], and one Si atom in the unit cell.

Generally the unit of a monolayer (ML) is used to measure the coverage of overlayer
atoms on a single crystal substrate. One monolayer is defined as an areal density equal to
the number of atoms per unit area in a single atomic layer of the substrate material
parallel to the surface. In the absence of reconstruction, this is the same as the number
density of atoms in the top atomic layer of the substrate [96]. According to this definition,
for Si (001) is 1 ML=1/a,‘,.2 = 1/ (0.3842 nmz) = 6.78x10l4 atoms/cmz.

As shown in Figure 2.1, each Si atom is bonded with four Si atoms in the bulk
structure. However, on the Si (001) surface, each of the first layer atoms is bonded with
two atoms in the second layer, leaving two dangling bonds on each surface atom. Since

this configuration is not energetically favorable, the surface reconstructs to lower the

20

surface energy. The most common reconstruction observed is 2x1, in which two
neighboring surface atoms are bonded together to form a Si dimer, which reduces the
number of dangling bonds from two per atom to one per atom. The reconstruction
establishes two characteristic directions on the surface: either along or perpendicular to
the surface dimer rows, in orthogonal <110> directions, which will be called the dimer

row direction and the dimerization direction, respectively. In this thesis, the Si dimer row

direction is defined as [1T0] and the Si dimerization direction as [110], as already

indicated in Figure 2.1(b).

 

 

o 0 O O

0.38 . .
o o

0.543 . .

0m - 100

0110 o 11(1- o

O

010

O O O

 

(b)

Figure 2.1 (a) Diamond structure of silicon, and (b) ideal Si (001) terminated surface.

Theories [97-99] Show that the electronic structure of these dimers consists of an
occupied 1: state below the Fermi energy E: and an unoccupied 1t* state above E13. During
STM imaging, the spatial distribution of these n-bonding and 7t*-antibonding states are
probed by electrons tunneling out of and into the surface, giving the filled and empty

states images, respectively. The filled states image emphasize the filled dimer bond lying

21

between the two atoms, and the empty states image shows the dangling bonds at the end
of each dimer pair. As a result, each Si dimer appears as a bean—shaped feature at
negative sample bias, as shown in Figure 2.2(a), but as two well-distinguished small

features at positive sample bias as shown in Figure 2.2(b), respectively.

 

(a) filled states (b) empty states

Figure 2.2 Features observed under (a) filled and (b) empty states STM imaging, and
their correspondence with the dimer model. The 2x1 unit cell is marked by the two
rectangular boxes in both images.

In reality, the Si (001): 2x1 surface is stressed since atoms in a dimer are closer than
they are in the bulk. There is a tensile stress in the dimer bond direction, while along the
dimer row direction, the stress is compressive (or at least less tensile) [99, 100]. The 2x1
structure is shown schematically in Figure 2.3, in which a 2x1 unit cell is marked.
Because of the stress anisotropy, the surface is expected to break up into two degenerate
stress domains with single—atomic-steps as their boundaries, with a step height of

abulk/4=0.136 nm. The dimer rows rotate 90° from one terrace domain to the next owing

22

to the tetrahedral coordination of diamond structure of silicon. Thus, two types of
monoatomic steps coexist. Following the notations of Chadi [101], the SA step is parallel
to the upper-terrace dimer rows and the 83 step is perpendicular to the upper-terrace
dimer rows. The energy required to form kinks in an SA step is higher than that in an S3
step, hence at experimentally accessible temperatures, S A steps typically appear smooth,

while SB steps appear rough, as shown in Figure 2.4.

 

Figure 2.3 Schematic drawing of Siz2xl reconstruction, as well as SA and 53 types of
steps.

When the vicinal angle is high, double-atomic-layer high steps are favored on a Si
(001) surface. One can again define purely from geometrical considerations two different
types of double steps, DA and D3, which are parallel and perpendicular to the upper—
terrace dimer rows, respectively. But in reality, because the DA steps are much higher in

energy, a Si (001) surface with a large vicinal angle contains only DB steps. Figure 2.5 is

23

an STM image of a 4° tilted vicinal surface with DB steps. Here, about 95% of the surface
is single domain, with the Si dimer rows running from lower left to upper right. A few
small areas of the alternative domain are seen where the double height DB step has split

into a pair of single height steps.

 

Figure 2.5 Empty states image (19 nmx19 nm) of 4° tilted vicinal Si (001) surface.

24

2.3 Scanning tunneling microscopy and image processing

Scanning tunneling microscopy (STM) has become a very powerful technique for the
determination of the structural and electronic properties of surfaces over the last twenty
years [96, 102-104]. The STM is primarily used as a high resolution imaging tool, by
which a lateral resolution of 0.1 A and a vertical resolution of 0.01 A can be achieved.
This is sufficient resolution so that the STM can actually “see” single atoms on a surface.

The mechanism of STM is based on the phenomenon of quantum mechanical
tunneling. When a sharp metallic tip is brought close enough (~ 5 to 10 A) to a
conducting surface, electron tunneling occurs and a tunneling current (nA range) can be
detected between the tip and sample when a bias voltage (mV to 5V) is applied. To
obtain an image, the tip is scanned in the two lateral dimensions over the surface driven
by piezoelectric elements. The STM can be operated in several imaging modes, such as
constant current and constant height imaging modes. All of the STM images included in
this dissertation are taken in the constant current mode, which is introduced briefly as
follows.

In the constant current imaging mode, a computer-controlled feedback loop adjusts
the tip height to maintain a pre-set constant current. Since the tunneling current is
exponentially dependent on the tunneling gap, constant current implies constant tip to
sample distance to a very high accuracy. The tip motion, as measured by the voltage
applied to the z-piezo, is then recorded versus the scan position. Essentially, the
corrugation in the STM image reflects the local density of states integrated over a range
of energies between the tip to sample voltage and the Fermi level, at the position of the

tip apex. For surfaces which have relatively uniform electronic properties, such as metals,

25

the STM image effectively represents the surface tOpography. For electronic
inhomogeneous surfaces, such as semiconductors, the electronic structure becomes
important in determining the nature of the image.

Generally, the STM image of semiconductors shows a very strong bias dependence,
reflecting the spatial distribution of the valence and conduction band structure. A typical
example is the pair of filled and empty states images of the Si (001): 2x1 reconstructed
surface shown in Figure 2.2.

The most common method of displaying STM data is to construct a grayscale image,
where the brightness is keyed to height. Bright features in the image represent higher
features on the surface. In virtually all of the images in this thesis, only minimal image
processing has been done, consisting of planar background subtraction and contrast
adjustment in order to make full use of the grayscale for any given image. In some
instances, a lateral linear mapping has been applied to the image to correct for the effects
of thermal drift during the acquisition time of the image (typically 1 - 20 minutes).

In a few instances, additional processing has been done to emphasize small
corrugations on atomic terraces at different heights. This is certainly true of all of the
pseudo-3D rendered images, for example Figure 2.4. When this type of contrast
enhancement is applied to 2D images, it is noted in the text.

Most of the image processing was done with Image SXM, which is a front end for
NIH Image written by Steve Barrett at the University of Liverpool specifically for

scanned probe images of all kinds [105]

26

2.4 Indexing of hexagonal close-packed structure

Due to the repeated usage of indices of hexagonal close-packed (hcp) structure, a
fully indexed bulk structure is shown in Figure 2.6 following the four-index notation, in
which the indices of any direction can be written as [u v t w], where t = —(u+v), and the

indices of any plane can be written as (h k i l), where I = —(h+k).

  
 

[0110]

    

[1010]»

Figure 2.6 Indices of hcp structure.

27

Chapter 3 The Si (001)(2x4)-Dy surface

In this chapter, the 2x4 reconstruction induced by submonolayer deposition of Dy on

Si (001) is introduced.

3.1 Experimental observations

Deposition of Dy on Si (001) at 600°C leads to the formation of nanowires on a
reconstructed surface, as shown in Figure 3.1. The 3D rendering shows a ridged nanowire
in the upper part and a single-atomic—step in the lower part of the image. On the upper

terrace in this image the Si dimer row direction ([1T0]) is vertical. As we can see, the

long dimension of the nanowire is perpendicular to this direction, which is the general
trend observed for nanowire growth. The nanowire is significantly higher than the single-
atomic-step height (1.36 A). Aside from the nanowire, the remainder of the surface is
covered by both bare 2x1 reconstructed silicon and metal reconstructed areas, with the

details depending on the metal coverage.

Generally, a 2x4 LEED pattern was observed over the coverage range 0.17-1.6 ML,
and it is sharpest around 0.5 ML, as shown in Figure 3.2. Figure 3.3 shows a 50 nm
square empty states image of a surface with 0.5 ML Dy deposited. A single-atomic-step
runs horizontally through the image. Each terrace has areas of bare Si as well as
reconstructed areas that we associate with the Dy. The reconstructed patches show a
stronger corrugation, and trenches separated by three rows of maxima. The inset is an
FFT of this image that shows a 2x4 superstructure that agrees with the LEED pattern of

this exact sample.

28

     
 

._'..‘:1

 
 

' i I

 
   

v11

“I “I”
'Nuu)!

   

Figure 3.1 STM topography of Si (001) surface with 0.5 ML Dy deposited. At a sample
bias of +1.39 V, the maximum apparent height of the nanowire is about 0.4 nm, the width
is about 5 nm.

 

Figure 3.2 2x4 LEED pattern of 0.53 ML Dy on Si (001), grown at 600°C. The beam
energy is 49.2 eV. The Si: 1x1 spots are indexed. Two 2x4 reciprocal unit cells are
marked by two white boxes which are 90° rotated, reflecting the two domain structure of
the Si (001) surface.

29

 

Figure 3.3 Large empty states image (50x50 nmz) of the reconstructed surface with 0.5
ML Dy deposited, VT=+1.05 V. The inset is an FFT of this image. The spots in the right
half of the FFT pattern are accentuated by the black dots. The Si: 2x1 reciprocal unit cell
is marked by the crosses, and the 2x4 by the open circles. A close-up of the area enclosed
in the square box is shown in Figure 3.4, and that in the rectangular box is shown in
Figure 3.5.

 

Figure 3.4 (a) and (b) are two empty states images of the area enclosed in the square box
(10x10 nmz) in Figure 3.3, taken at +1.05 V and +0.46 V, respectively. Clean Si areas
(A) and metal reacted areas (B) are marked. Image (a) shows all the maxima in the metal
covered areas have similar height. Image (b) shows two types of maxima, one higher than
the other. The broken white boxes in both images show the 2x4 unit cell. The two small
boxes in (b) mark the HLH (type I) and LHL (type II) features.

30

Figures 3.4(a) and (b) are two empty states images of the surface shown in Figure 3.3,
taken at two different biases. Both images are close-ups of the area marked by the square
box in Figure 3.3. The surface is separated into areas of clean Si (marked A) and areas
reacted with the metal (marked B). In Figure 3.4(a), taken at a sample bias of +1.05 V, all
the maxima in the metal covered areas have similar heights. In Figure 3.4(b), at +0.46 V,
however, two types of maxima can be seen, with one much higher than the other. These
maxima are in registry with the positions of Si dimers in both directions, along and
perpendicular to the Si dimer rows. If we use H to denote the high maxima and L to
denote the low maxima in (b), two arrangements along the Si dimer row direction can be
differentiated, HLH (type I feature) and LHL (type II feature). These two types of
features alternate along the Si dimer rows, and there is usually a Si dimer vacancy in
between. In a few cases, the two features are adjacent directly to each other. Along the
perpendicular direction, no obvious alternation between type I and type II was observed.
Statistics show that the probability of having a like neighbor in the Si dimerization

direction is much higher than that of having a different neighbor.

Figure 3.5(a)_(c) shows three empty states images over the area marked by the
rectangular box in Figure 3.3, taken at different biases. Once again, significant
differences in maxima height in the metal covered areas can be seen, implying that the
maxima heights are strongly bias-dependent in the empty states. The detailed appearance
of the metal covered areas varies as well. Observing the arrangement of three rows of
maxima in the white box in these three images, the middle row is higher in (a), all three
rows are at similar heights in (b), and the middle row is almost invisible in (c). Therefore,

we can draw a conclusion here that not all maxima are visible at +1.85 V (c), but these

31

missing maxima show up again with stronger intensity at +0.46 V (a), and with similar
intensity at +1.05 V (b). The bias-dependence of maxima height can also be observed
from the height profiles along the same trace marked by dotted lines in these three
images. The black and white arrows in the images show the correspondence between the

specific maxima and their heights.

 

Figure 3.5 (a)—(c) are three empty states images of the area enclosed in the rectangular
box (15x12 nmz) in Figure 3.3, taken at +0.46, +1.05 and +1.85 V, respectively. The
heights marked by the black and white arrows along the profiles correspond to the
positions marked by the same arrows along the dotted traces.

As mentioned above, type I and type II features alternate along the Si dimer rows. If

this alternation were strict, the long range periodicity would be 2x8. However, the real

32

order is 2x4, as seen by LEED and also the FFT of our STM data. We ascribe this

ordering to the basic 2x4 building block which can be either type I or type II.

Figure 3.6 is a filled states image of a surface with 0.17 ML Dy. Areas of clean Si and
reacted Si are marked by A and B, respectively. Overall, the reacted areas appear lower
than the Si dimers. The 2x4 ordering is not as clear as in the empty states images. This is
due to a variation in the heights and sizes of maxima in the reacted areas. A one-for-one
correspondence can be found between maxima in the filled and the empty states, but not
all maxima that appeared in the empty states are visible in the filled states. It is clear from

Figure 3.6 that the filled states maxima line up with Si dimers in the same row.

 

Figure 3.6 Filled states image (15x15 nmz) of the surface with 0.17 ML deposited, VT:—
1.93 V. The surface is separated into areas of clean Si (A) and areas reacted with the
metal (B).

In many metal induced reconstructions of Si surfaces, it is possible to evaluate the

number of metal atoms per unit area by measuring the fraction of reconstructed surface

33

and comparing it to the metal coverage. However, it is very difficult to do the same thing
in this particular case since a significant fraction of the metal can be drawn into the
silicide nanowires. Comparing images of surfaces at 0.17 ML (Figure 3.6) and 0.5 ML
(Figure 3.4(a)), it is true that more of the surface is reconstructed at higher metal
coverage, but this increase is not in direct proportion to the change in coverage. Since we
observe three maxima within each 2x4 unit cell, the simplest possible assumption is to
associate each maximum with one metal atom. It is certainly possible that the metal
atoms might be below the top layer and that the observed maxima are in fact Si atoms,
but in the absence of other information, we make this assumption for the purpose of
estimating metal coverage. Usually, images taken at around +1 V are chosen for the
coverage counting because all maxima are visible at this voltage. Under this assumption,
if the surface is covered by the perfect 2x4 structure, the saturation coverage would be
3/8=O.375 ML. However, a completely 2x4 reconstructed surface was never observed in
our experiments. The 2x4 always coexists with nanowires and with clean Si which do not
disappear even when the coverage is above 3 ML. If we only count metal atoms on the
2D reconstructed areas, 0.15 ML was obtained for nominal coverage 0.5 ML, and 0.2 ML
for 1 ML. This suggests that about 2/3 of the metal is in the nanowires at 0.5 ML and
about 4/5 in the nanowires at 1 ML. To check the accuracy of this statement, we
estimated the amount of metal corresponding to the measured volume of silicide
nanowires. There is an additional 0.35 ML at 0.5 ML nominal coverage, and 0.6 ML at 1
ML. Therefore, all of the metal can be accounted for within experimental error (about

20%) if we combine the 2D surface and the 3D nanowires together. It is also possible that

34

some of the metal atoms diffuse into the bulk which would slightly reduce the amount of
metal in the nanowires.

Note that one other possible assumption is that only the high or the low maxima in the
reconstructed areas are metal atoms. Under either assumption, the density of maxima on
the terraces combined with the metal density associated with the nanowires is not
sufficient to account for the deposited metal, and so both the high and low maxima in the

2x4 structure are likely to be associated with metal atoms.

3.2 Discussion

The relative heights of maxima and surrounding Si dimers shown in Figures 3.4-6
suggest that they are in the same layer, i.e., the Si dimers in the reacted areas were
removed and the maxima sit on the next lower layer. As discussed above, the maxima are
centered over the Si dimer rows and also in line with Si dimers, which implies that the
maxima are exactly in the place of the missing dimers. Figure 3.7 is a model which shows
the positions of the maxima in the 2x4 structure with respect to the surrounding Si
dimers. The alternation between the two types of features (HLH and LHL) along the
dimer rows, the Si dimer vacancies between different features, and the statistical
character along the Si dimerization direction are also shown in this figure. A 3D model is
shown in Figure 3.8 (note that the difference between H and L maxima are not shown

here).

The presence of the two types of maxima (H and L) could be explained by a buckling

of the Dy atoms based on the previous work done on Si (111) [3, 55, 63, 75, 106]. It is

35

C O . ll,
OQOEOOO
090 0°C

.0 o o 0230 o o 0110
°§
0

Ill

3

o 0 05-10 0 o 0‘30
9050

0 0 053.0 0 o 05.0

-I'

o o 05.0 o o 0:30

 

'o
(‘1
90°0qOQO
0000

.. -— , .\ = . =

000000

0
O

0 underlying Si atom 0 O Si dimer 5:1”: Si dimer vacancy

. . . typell'eature(HLH) . . . typellfeature(LHL)

Figure 3.7 Registry of the maxima features in the 2x4 structure with respect to adjacent
clean Si dimer positions. A unit cell is marked by a rectangular box.

     

. Si [110]
DySig [00011

   

DySi2 [1.1201

0 Si Q Dy “‘30 Si dimer

Figure 3.8 3D model of 2x4 reconstruction, in which the orientation relationship between
nanowires and Si (001) surface is marked. Bonds are added to the Dy atoms for
illustration purpose only.

36

known that the RE disilicides grown on Si (1 l 1) are a defected Ale structure, which
includes about 15—20% Si vacancies [3]. The vacancies are believed to play a role in
strain relaxation. Ordered arrangement of these vacancies has been observed in the bulk
silicide [3, 75] and also in silicide films on Si (1 1 l) [55, 75]. These vacancies induce
shifts in the Si and metal positions within the silicide [106] and in the surface [75]. The
buckling will mimic this behavior. where the metal atoms shifting up and down cause the
high and low maxima, as well as the alternation between the two features, HLH and LHL.
The bias-dependence of the maxima in empty states could come from the buckling as
well, which would lead to a change in the electronic charge distribution on the surface,

and in turn, might affect the appearance of the maxima at different biases.

(Sixk

J
\

0i 15‘“

)4”

 

Figure 3.9 Local position of Dy atom in the bulk silicide, which is similar to that on the
2D surface.

If the assumption that each maximum in the empty states corresponds to one Dy atom
is correct, then the local environment of the Dy atom on the surface, as shown in Figure
3.8, is similar to that in the bulk silicide, as shown in Figure 3.9. Comparing Figure 3.8

and Figure 3.9, the c axis of the silicide can be visualized as lying along the Si dimer row

37

direction. Therefore, the reconstructed 2D surface can be considered as the first step to

forming the silicide, with DySi2(01T0)//Si (001) and [0001] parallel to the dimer row
direction (Si [1 T0] ). Referring back to Figure 3.1, the long dimension of the nanowire

would be along [1120] (direction of axis a).

For DySig, the lattice parameter a is 0.3831 nm, which matches the Si surface
periodicity as. (=0.384 nm) very well, but 6 is 0.4121 nm, which has a significant
mismatch with as,- [3]. So strain will develop along what would be the c axis of the
silicide, which is parallel to the Si dimer row direction as observed from the nanowires.
Therefore, the Si dimer vacancies along the Si dimer rows in the 2x4 surface would
relieve the strain. Likewise, the growth of the silicide islands (nanowires) in the [0001]

direction (the Si dimer row direction) will be limited due to the strain developed, but they
will grow without limitation in the direction of [1 120] due to the small mismatch. This is

consistent with the observed orientation of the silicide nanowire, as shown in Figure 3.1.

It is entirely possible that the maxima seen in the STM images are due to rearranged
Si atoms, with Dy atoms having moved below the topmost layer of atoms. We cannot
rule this out on the basis of the STM data, and an appropriate model cannot be proposed
for such a structure at this moment. However, the analogies between the 2x4 surface and
the bulk silicide structure in regard to strain accommodation by buckling and Si dimer

vacancies hold true regardless of the details of the surface structure.

38

3.3 Conclusions

When Dy is deposited on Si (001) at 600°C, highly elongated islands of Dy silicide
are formed, and the surrounding substrate is reconstructed into a 2x4 periodicity. STM
measurements show that the appearance of the 2x4 surface is bias-dependent. An atomic
model is proposed where a Dy atom replaces a Si dimer on the top layer, sitting in an
atomic position that is almost the same as that in the bulk silicide. Groups of three Dy
atoms and a Si dimer vacancy form the 2x4 unit cell. The bias—dependence of the STM
images is explained in terms of a buckling of the Dy atoms. Both the buckling and the
presence of Si dimer vacancies contribute to strain relaxation along the Si dimer row

direction.

39

Chapter 4 Dy silicide nanowire growth on Si (001)

In this chapter, the growth behavior of Dy on Si (001) is described in detail, including
nanowire and non-nanowire island formation, surface reconstructions of the nanowires
and the substrates, nanowire nucleation, and interactions between nanowires and Si steps

during growth.

4.1 Experimental observations

4.1.1 Coverage dependence

 

Figure 4.1 (a)-(d) are STM images (all 300x300 nm) which show coverage dependent
topography of the Dy/Si (001) surface.

Figures 4.1(a-d) show a sequence of STM images at metal coverages up to 3.3 ML.

At this image magnification, only the overall surface morphology is visible. All samples

40

in Figure 4.1 were prepared at a growth temperature of 600°C with 2-10 minutes of post-
annealing. At the lowest coverage (0.17 ML) only very narrow nanowires are formed.
These nanowires run in two perpendicular <1 TO > directions and form a network
structure, reflecting the two domain structure of the substrate surface. Single height
atomic steps are visible on the uncovered areas of substrate. As coverage increases in (b)
and (c), the nanowires become denser and wider, and some rectangular islands are
formed, mostly at the junctions of the nanowires, as shown in Figure 4.2, which was
taken from the same sample surface as Figure 4.1(b) with a higher resolution. At 3.3 ML

in (d), a network structure is formed with significant formation of irregular 3D islands.

 

Figure 4.2 3D islands formed at the junction of the nanowires. The image size is
150x150nm, taken from the same sample surface as Figure 4.1(b). The inset is a height
profile of the trace marked by the broken line.

Figure 4.3 shows a close up of a surface with a narrow and wide nanowire at 0.53
ML. Here it is clear that what was described as a wider nanowire in Figure 4.1 is actually
a bundle of parallel nanowires separated by narrow trenches. Some second layer silicide

growth is seen as well, exhibiting a layer-by-layer growth mode, where each second layer

41

island does not cross the trenches in the first layer. We make a distinction between this
type of second layer growth which we interpret as a continuation of nanowire growth,

and the growth of larger and wider islands seen at nanowire junctions.

 

Figure 4.3 Filled states STM image (50x50nm, VT: - 1.2 V) shows that Dy silicide
nanowires coexist with 2x4 and 2x7 reconstructed Si (001) surface. The nanowires
bundle together and form wider structures. A single nanowire can be seen in the lower
right corner. The Si dimer row direction is marked in the image.

Two different surface reconstructions, 2x4 and 2x7, are visible in the areas of the
substrate adjacent to the nanowires in Figure 4.3. The Si dimer row direction of the
substrate surface is marked, which is perpendicular to the nanowires on the same terrace.
Furthermore, the 2x dimensions of the 2x4 and 2x7 superstructures run parallel to the
nanowires, or perpendicular to the Si dimer row direction. The details of the 2x4
reconstruction have been presented chapter 3, and the 2x7 reconstruction will be

described in chapter 5. Generally, both the 2x4 and 2x7 reconstructions could appear

42

separately or simultaneously with the nanowires, with a dependence on the detailed
growth conditions. Unreconstructed Si patches usually coexist with the 2x4 reconstructed

areas. as discussed in the previous chapter, but not with the 2x7 reconstructed areas.

 

 

 

 

 

1.2-

  

0,.) - 0.36

 

 

 

0.6Illllllllllll
93% 0 20 40 60nm

Figure 4.4 Examples of height and width measurement of the nanowires. The height
profiles of traces (l) and (2) in the left-hand side panel are shown in the right-hand side
panel, respectively. The image size is 70x70 nm, VT=+1.04V.

4.1.2 Width and height distribution of nanowires

By comparing the width of the nanowires to the periodicity measured from Si dimer
rows or 2x4/2x7 superstructures in the same image, it was found that the width is
quantized in multiples of the Si surface lattice parameter as; (=0.384 nm). One example of
the width measurement is shown in the right-hand side panel of Figure 4.4, where the
curve 1 is the height profile of trace (1) in the left-hand side panel. Along this trace, the
last two nanowires are 4as,- and 503,’ wide, respectively, and the trench between them is
lag, wide. Note that the width measurement was calibrated by the periodicity measured

within the same image. Figure 4.5(a) shows the width distribution of the bundled

43

nanowires measured from many images at different metal coverages. The minimum
width is 2a5,- (0.77nm), and the maximum 16a5,- (6.14nm). Mostly, the nanowire widths
are between 2-7a5,-. All of the trenches between adjacent nanowires are measured to be
lag,- wide. For isolated nanowires which are not in bundles, the widths range between 2-

405i-

 

(a)

15-

 

number of nanowires

  
  

 

 

 

 

 

 

 

 

 

q.
i."

123 4 6 7 8910111213141516

Width (aSi)
_ (b)

 

 

w :5 Ur O\
l

N

number of nanowires

‘
A
I

 

 

0 . . , ; a... ‘
0.28 0.3 0.32 0.34 0.36 0.38 0.4 0.42
height (nm)

Figure 4.5 (a) Width distribution of bundled nanowires, and (b) height distribution of
additional layers on nanowires.

Without considering any additional layers, the minimum nanowire heights measured
with respect to the surrounding Si surface range between 0.48 nm and 0.56 nm, which
varying slightly with bias in both filled and empty states. One example of this

measurement is shown in Figure 4.4. The nanowire height along trace (1) in the left panel

44

is 0.53 nm, as indicated in profile (1) in the right-hand side panel. These measured
values could be different from the geometrical heights since the reconstructed Si surface
and nanowire surface have different electronic states, resulting in different apparent
heights. To avoid this difficulty, heights of the islands growing atop nanowires were
measured with respect to the surface of the nanowires underneath since we believe that
they have the same surface. The height difference distribution of nanowires is shown in
Figure 4.5(b), in which all the measured heights range from 0.31 nm to 0.39 nm with an
average of 0.34:0.02 nm. One example of the height measurement is shown by profile
(2) in the right-hand side panel of Figure 4.4, showing that the height of the islands along
trace (2) in the left-hand side panel is 0.36 nm.
4.1.3 Nanowire top surface reconstruction

The nanowire top surfaces exhibit various superstructures, which may depend on the
nanowire width. Figure 4.6 shows a collection of high resolution images of nanowire top
surfaces, most of which were taken from Figure 4.4 (a) except nanowire I and IV. In
Figure 4.6(a) the reconstruction of nanowire I turns from p(1x1) in the narrow part into a
different superstructure when the width changes from 2 as,- to 4 as). In (b), nanowire II of
width 2 as,- has maxima on the top which are spaced 2 as, apart along the nanowire long
dimension. Maxima of 1 as,- apart are also possible, as seen on nanowire III of the same
width, but less common. Nanowires IV and V with width of 5 as; in (c) have other
complicated superstructures with holes in the center, with a periodicity of 2 as,- along the
nanowire for V but larger for IV. We did not observe the well-ordered 2x1 nanowire

reconstruction recently reported by Preinesberger et al.[l2].

45

 

Figure 4.6 Various reconstructions of the nanowire top surfaces. All the images have the
same size, 10x10 nm. Roman numerals in images (b), (d)-(f) correspond to the marked
positions in Figure 4.4.

Table 4.1 Measured mismatches between the c(2x2) structure and Si substrate

 

 

 

 

 

island # Longitudinal mismatch (%) Lateral mismatch (%)
VI 0.8 7.6
VII 2.1 10.3
VIII 2.7 10.4

 

 

 

 

46

The most common superstructure is c(2x2) when the nanowire width is larger than 4
(15,, as shown in (d). The reconstructed surface is atomically flat, and one unit cell is
marked in the image. Figures 4.6(e) and (i) show that islands VII and VIII have the same
surface reconstruction as nanowire VI underneath, although both surfaces are curved
across their short dimensions. The c(2x2) unit cell in Figures 4.6(d) and (e) is not square
if observed closely. The maxima spacings (x) along the longitudinal and lateral
directions of the nanowires are compared with the measured Si periodicities along both

directions to correct for any possible image distortion. The mismatches, (x — a5,)/ 615,-

expressed as percentages are listed in Table 4.1. Referring to the image shown in Figure
4.4, island V1 is one height quantum, and islands VII and VIII are two height quanta,
respectively, above the surface of the minimum height nanowires in this image, with each
quantum being three layers of atoms. In all cases, the longitudinal mismatch along the
long axis of the nanowire or island is small, and the lateral mismatch is large. The
apparent mismatch changes as a function of height. As a further point of reference, there
is no measurable mismatch along the long axis of the minimum height nanowires (i.e. the
level below VI), and the lateral mismatch is not measurable since the reconstructions of
these wires are complex and varied.
4.1.4 Nanowire nucleation

Short nanowires are often observed at a step edge, or on a flat terrace, as marked by
the arrows in Figure 4.1(b) and Figure 4.4, respectively. This suggests that the nanowire
could nucleate at a step edge or on a flat terrace. In the first case, as the nanowire grows,
the Si atoms around it will be consumed to construct the nanowire, causing new steps to

form, as shown in Figure 4.7(a). In the second case, the nanowire takes Si atoms directly

47

from Si step edges, causing Si steps to bunch up at the end of the nanowire and forming a
“snow-plow” effect, as noted by Chen et al.[6, 11]. A typical example is shown in Figure
4.7(b). Due to the consumption of Si atoms from the step edges, the steps retract and
form a channel in the upper terraces at the lower right end, and produce a quantity of run-
on steps and form a significant “snow-plow” effect in the lower terraces during growth.
The nanowire can affect the terraces significantly below and above the terrace on which it

grows, but always grows on a single flat terrace, never crosses a step.

 

Figure 4.7 Nanowire nucleates (a) on a flat terrace, forming a new step around it by
consuming the Si atoms in the neighborhood, and (b) at a step edge, interacting with Si
steps on both ends during growth.

4.1.5 Effect of annealing duration

All the samples discussed above were prepared by depositing Dy onto Si (001) at
600°C with post-annealing less than 10 min. When Dy deposition is followed by a long
period of post-annealing, larger, more three dimensional (3D) silicide islands are formed.
Figure 4.8 shows a surface with 0.68 ML Dy annealed for 30 minutes at the deposition

temperature of 600°C. The metal coverage is between those shown in Figures 4.2(b) and

48

(c), and yet the topography is very different with almost no nanowires present. This
makes it clear that for Dy in this general coverage range, the morphology of the silicide
formed is a function of both metal coverage and annealing time, and that the nanowires
are a metastable phase and consumed by the post-annealing process. Islands 1 and 2 are
about 8 nm wide and 300 nm long, and islands 3 and 4 are about 25 nm wide. It is hard to
measure the heights of these islands since they are situated in trenches. But our
measurements still showed that all of the islands 1-4 have a minimum height of 0.97 nm

with respect to the bottom of the surrounding trenches.

 

Figure 4.8 STM image showing that only a few nanowires survive when continuing to
anneal a sample at 600°C for 30 min after deposition of 0.68 ML Dy, mostly large square
and elongated islands are formed. The image size is 400x400 nm.

4.2 Discussion
The nanowire formation has been explained by anisotropic lattice mismatch between
hexagonal RE disilicides and Si (001) surface [6, 7], which can be described with an

atomic model shown in Figure 4.9. The hexagonal Ale type structure is shown in (a). If

49

 

1
%:© | T
00.
llqggnn} ail-lg) [ITUI
.

 

(1 mo, 53?."
a
o—n°°°°°°°°°m°0‘
111000000000 i[1120]
0 0 0 O O[0001]

0330 O 0 m0 (3 0

nm

O 0

{Op :3 o 0, cliff/n a [001]
S' . _
l g o o 6 11101

Figure 4.9 (a) Hexagonal Ale type structure of Dy disilicide; (b) lattice matching
between the silicide (l T00) plane and Si (001) surface; (c) side view of Dy disilicide
growing on Si (001) substrate.

we match the (l T00) onto the Si (100) surface lattice as shown in (b), for certain RE

silicides, a is well matched to the silicon substrate, but c is not. For example, for DySiz,

a=0.383nm and c=0.421nm which correspond to mismatches of -0.36% and 7.3%,

respectively[3]. Figure 4.9(c) shows the side view of the hexagonal DySiz on Si (001)

surface lattice seen from the [0001] direction. Here we make an assumption that the first

layer of Dy atoms are located in the center of the Si (1x1) unit cells, which reserve the

same positions as that of the 2x4 reconstructed Dy atoms, as suggested in the previous

50

chapter. If we continue to assume that the Dy atoms are 0.11 nm above the Si (001) plane
underneath so that these Dy atoms in the interface have almost the same environment as
they are in the bulk, providing a perfect epitaxial condition for the silicide to grow.
Essentially, this model is the same as the one proposed by Chen et a1. [6, 11]. The wire
grows to an arbitrary length in the direction of small mismatch (horizontally in the plane
of the page) and is limited in its ability to grow coherently with the substrate in the lateral
direction (perpendicular to the page).

The quantized height of 0.34 nm can be explained by the model shown in Figure
4.9(c). As seen from the side view, the silicide is comprised of one layer of metal atoms
and one bilayer of Si atoms alternatively along the vertical direction. The spacing
between each layer is 0.11 nm, and the height of one unit cell is 0.33 nm. The average
height of islands on nanowires (0.34 nm) is very close to this number, which suggests
that the layer height on top of the nanowires is just one silicide unit cell high.

The orientation relationship between hexagonal structure and Si (001 )substrate

shown in Figure 4.9 is consistent with that observed during RE silicide film growth, i.e.,
the hexagonal phase grows on Si (001) surface in two orientations, RESI2-x (1100) //Si
(001) with RESiz.x [0001] //Si [1 T0]. and RESiz.x (1100) //Si (100) with RESi2-x [0001] //
Si [T TO] [39, 64, 80, 81], which are normal to each other, and equivalent in volume

fraction and crystalline quality due to the four-fold symmetry of the Si (001) substrate.

Our STM data add more details to this statement. On a single 2x1 reconstructed Si (001)

terrace, the nanowires run only in one specific direction, the long dimension [1120]

perpendicular to the Si dimer row direction[l T0], as seen in Figures 4.4 & 4.5. The

presence of the two perpendicular orientations of the hexagonal phases is caused by the

51

fact that the Si dimer rows run in perpendicular directions on adjacent terraces. Although
the Si (001) substrate has four-fold symmetry, a single terrace only has two-fold
symmetry regulated by the 2x1 reconstruction. The preferred orientation could result
from the chemical bonding between silicide and Si substrate, as discussed in references
[6, 11].

It has been determined that the hexagonal RE disilicides grown on Si (1 l l) terminate
with a buckled Si bilayer instead of the graphite-like Si termination expected for a simple
truncation of the bulk silicide structure [69, 72]. The RE silicides on Si (111) generally
show p(lx1) reconstruction below 1 ML and J3 x J3 reconstruction above 1 ML [54, 55,
63, 69, 70]. Both of the 2D reconstructions result from a buckled Si top layer without
vacancies, but the J3 x J3 periodicity is a modulation of the p(lxl) periodicity caused
by the 3D vacancy ordering superstructure within the film[69, 72-78]. The same issue,
however, still remains open for the RE silicides on Si (001) since no experimental data
and theoretical simulation have been done on this system. As shown in Figure 4.9(c), the
nanowires could be terminated by either by a Si atomic layer or a metal atomic layer.
Compared to Si (111), it would be natural to believe that the top layer of the RE silicides
on Si (001) has a Si bilayer as a termination since it is exactly like the Si (001)
termination of bulk silicon, as in reference [12] which suggests that a 2x1 surface seen
on nanowires is terminated with Si dimers. The p(lxl) reconstruction shown in Figure
4.6(a) would also be a proof for this statement. It would not be hard to construct a model
for the c(2x2) superstructure by removing some Si atoms from the top layer , similar to
that proposed for the c(2x2) superstructure seen on larger Er silicide islands grown on Si

(001)[107]. It is also possible that these superstructures result from the 3D vacancy

52

””511

12:-

ordering within the nanowire, as discussed above for Si (1 l l). The existence of the 3D
vacancy ordering has been observed for RE silicide films on Si (001) by TEM[64].
However, the holes appearing in the top layer of nanowire V, as shown in Figure 4.6(e),
could result from the Si vacancies in the surface, different from the vacancy free top
surface of RE silicides on Si (1 l 1).

It is well established that strain relaxation plays a very important role in
heteroepitaxial growth due to the mismatch between dissimilar materials[108-110]. For
large lattice mismatch systems (>2%), three dimensional island growth is favored over
the layer-by-layer growth mode[l l l, 112]. Theoretical simulation shows that the early
stages of binary epilayer growth on a solid surface may self-assemble into ordered
parallel stripes with identical width in nanoscale due to anisotropic surface stress [113,
114]. The bundled nanowires may resemble this behavior, but the width distributes in a
wide range. No ordering can be derived from the distribution of single nanowires,
probably resulting from the random nucleation of the nanowires.

As discussed above, the Si atoms move from Si step edges or adjacent areas on the
same terrace to form the nanowires, which is consistent with the fact that the Si atoms are
the dominant diffusive species during the RE silicide growth on Si (111) [23, 115]. In the
second case, new steps will be formed and make the surface rough. But the nanowires
always grow on a flat terrace and never cross a step, whose growth could be stopped by
another perpendicular nanowire, or at step edges, as seen in Figures 4.2, 3, and 8. Figure
4.7(b) is a typical example of a nanowire that is stopped by one nanowire on one end and
at Si step edges on the other end. The nanowire could also stop in the middle of a terrace

due to shortage of metal atoms, as seen in Figures 4.2(b) and (c), where the only short

53

nanowires that do not terminate at either a step or an intersecting nanowire are those
(circled) that are entirely framed by other nanowires. In this situation, the nanowire must
compete with the surroundings for metal atoms.

As seen in Figure 4.8, some 3D islands can be fairly elongated. However, we still
call them ‘islands’ since their widths lie well outside the range of width distribution
defined for ‘nanowires’ in Figure 4.4(a), and also, they have similar minimum height as
that of compact islands shown in Figures 4.3 and 4.9, much higher than that of the
nanowires. One example of the height measurement is shown in Figure 4.2 that the island
marked by the broken line is 0.96 nm, as indicated by the height profile in the inset.

The formation of the nanowires has been explained by the hexagonal Ale type
structure in this chapter and other references [6, 7, 11], and the restricted width is related
to the poor mismatch along 0 direction [7]. Evidently, we are going to encounter
difficulties to explain these compact and elongated islands with the same structure.
Considering the fact that the hexagonal phase usually coexists with orthorhombic or
tetragonal phase in the RE silicide films on Si (001), and their mismatches with the
silicon substrate are not nearly as anisotropic as for the hexagonal phase, it is natural to
propose that these rectangular and square shaped islands have orthorhombic or tetragonal
structure. This is consistent with the TEM observation that only the tetragonal phase is
seen in thicker films that were annealed for 30 minutes [61]. Similar 3D Er silicide
islands on Si (001) have also been assigned to the tetragonal phase [107], which are
definitely different from the Er disilicide nanowires discussed by Chen et a1. [6, 11].

How do these 3D islands compare with the rectangular islands at the nanowire

junctions seen in Figure 4.4? Close examination shows that the islands in Figure 4.4 have

54

geometries that are still defined or restricted by the underlying or adjacent nanowire
structure. In this respect, these islands are considered as upper layers of nanowire layer
by layer growth. It is possible that the highest of these small islands are right at the
transition between nanowire and 3D island growth. Close examination of the nanowires,
islands at nanowire junctions, and the large 3D islands with high resolution diffraction
techniques will be necessary to understand the different crystal structures. However, at
this point in time, all that we can conclude is that true nanowires (corresponding to the
hexagonal silicide) have widths less than 6 nm and minimum height 0.5 nm, whereas 3D

islands have a minimum width of 8 nm, and a larger minimum height.

4.3 Conclusions

When Dy is deposited onto Si (001) surface at 600°C, a network of nanowires is
formed, and the width and density of the nanowires strongly depend on the deposited
coverage. These nanowires grow only on a flat terrace, running perpendicular to the Si
dimer rows. The nanowires coexist with a 2x4 and/or 2x7 reconstructed Si surface. The
top surface of the nanowires are reconstructed into p(1x1), c(2x2), and other
superstructures. 3D islands appear with increasing coverages and longer annealing
duration, which could have orthorhombic or tetragonal structure, different from the
hexagonal structure of the nanowires. The nanowire width is quantized in terms of as),
with a range from 2a5,- to 16615;. The nanowires grow layer-by-layer with a quantized step

height of one unit cell of the hexagonal Ale type structure.

55

Chapter 5 The Si (001)(2x7)-Gd, Dy surfaces

As discussed in Chapter 3 and 4, deposition of Dy onto Si (001) leads to the
formation of a 2D reconstructed Si surface, nanowires and 3D islands. Growth of Gd on
Si (001) shows very similar behavior. For example, both metals can induce 2x4 and 2x7
reconstructions of the Si (001) substrate. In this chapter, a detailed study of these 2D
phases is presented. The structure of the 2x7 phase will be discussed in terms of what is
known about the 2x4. The similarities and differences between the reconstructions

induced by the two different metals will also be discussed.

5.1 Experimental observations
5.1.1 General observations

Deposition of RE metals (Dy, Gd in this study) leads to the formation of RE disilicide
nanowires on a reconstructed surface, which can be a 2x4 or 2x7 superstructure, or a
mixture of both. This latter case is shown in Figure 5.1(a) for Dy and (b) for Gd,
respectively. Morphological similarities can be seen for Dy and Gd with nanowires or
nanowire bundles coexisting with the reconstructed substrate.

Figure 5 2(a) is an empty states image of an area that is covered entirely with the 2x7
phase. Figure 5.2(b) is a low energy electron diffraction (LEED) pattern of the sample
shown in (a). The 7x periodicity is clearly discemable from the pattern. Streaks appear at
the half order positions, caused by anti-phase disorder[116]. Figure 5.2(c) is the fast
Fourier transform (FFI‘) of image (a), which matches with the LEED pattern very well.
The 2x7 reciprocal unit cell is marked by a white box, and the Si: 1x1 reciprocal unit cell

is marked by a black square box in both (b) and (c).

56

 

A

Figure 5.1 STM images (50 nm x 50 nm) showing nanowires coexisting with 2x4 and
2x7 superstructures on a (a) 0.18 ML Dy deposited Si (001) surface, VT: -1.25 V, and
(b) 0.63 ML Gd deposited Si surface, VT: +0.6 V.

 

Figure 5 .2 (a) STM image (60 nm x 60 nm) showing a 2x7 reconstructed surface with
0.46 ML Gd deposited at 500°C, VT = +1.17 V. (b) LEED pattern from the sample shown
in (a). (c) FFT pattern of image (a). The black box and white box in (b) and (c) denote

Si: 1x1 and Gd: 2x7 reciprocal unit cells, respectively.

57

 

Figure 5.3 Close ups (15 nm x 18 nm) of (a) Gd: 2x7 superstructure, VT: +1.17 V, and
(b) Dy: 2x7 and 2x4 mixtures, VT: +1.24 V. Image (a) is taken from the same sample as
Figure 5.2.

Figure 5.3(a) is a close up of the Gd22x7 superstructure, with a single-atomic-step
running across the image from the upper left comer to the lower right comer. Although
all the structure rotates by 90° at the step edge, we will denote the Si dimer row direction
on any given terrace as [1 I0], and the Si dimerization direction as [110]. At this

particular sample bias of +1.17 V(empty states), the 2x7 superstructure appears as rows

of three maxima divided by a bright strip. A 2x7 unit cell is marked by a black box in the

58

image, and the 2x7 periodicity can be easily measured by comparison to the bare Si
dimers marked along the step edge. If we define groups of three maxima as a subunit,
then each 2x7 unit cell has one subunit at each corner. Each unit cell also crosses one of
the bright strips. The subunits could be in line or shifted half a unit cell with respect to
the adjacent subunits along [110], as marked by the straight lines in the lower and upper
terraces in (a). This half unit cell shift appears very often, resulting in antiphase disorder
along the 2x direction which produces the streaks in the 2D diffraction pattern seen in
Figure 5.2 (b) and (c).

The 2x4 unit cell is also composed of a three maxima subunit plus a “missing Si
dimer” at similar bias, as discussed in Chapter 3. One can ask what the relationship is
between the three maxima subunits appearing in both 2x4 and 2x7 superstructures.
Figure 5.3(b) is an STM image for Dy taken at a bias of +1.24 V, showing that the same
three maxima subunits can form both 2x4 and 2x7 local superstructures, whose unit cells
are marked by black boxes in the image. The bright strip between two rows of subunits
combines them to form a row of 2x7 unit cells. Comparing the marked bright strips with
the maxima along the same row and with adjacent maxima rows in Figure 5.3(a) and (b),
it can be seen that the bright strip is two maxima wide, and has a trench one Si dimer
wide separating it from the adjacent three maxima subunits on both sides. Therefore, the
three maxima subunits have the same geometry for both 2x4 and 2x7 superstructures.
Also, it can be seen that all of the maxima within each subunit in (a) and (b) share the

same position as a Si dimer by comparing them to the bare Si dimers as marked.

59

 

Figure 5.4 STM images (12 mm x 15 nm) showing 2x7 bias dependence in the empty
states (a)—(c) and in the filled states ((1) and (e). Relative shifts of the maxima in both the
subunit rows and the bright strips are marked near the straight lines in (b), (d) and (e).

60

5.1.2 Bias dependence

Significant differences can be observed in the appearance of the 2x7 superstructure
shown in Figures 5.1-3 due to a strong bias dependence in the empty states. Generally,
three distinct types of appearance for the groups of three maxima can be differentiated.
Since the bias dependences of Gd: 2x7 and Dy: 2x7 are very similar, as evidenced by the
fact that the appearance of the 2x7 superstructure taken at similar bias shown in Figure
5.3(a) for Gd and (b) for Dy is actually identical, the best images are selected from both
metal reconstructions to show the bias dependence features. Figure 5.4(a) is an STM
image at +0.8 V taken from a Gd deposited sample, where the three maxima in the
subunit appear with similar intensity (regime II). In image (b) taken from a 0.2 ML Dy
deposited sample at +0.28 V, the center row of maxima are very strong, and the other two
rows are much less bright (regime 1). Image (c) is of the same area as image (a) taken at a
bias of +2.06 V, showing that the center row is totally invisible, but the other two rows
appear with significant intensity (regime III). The same regimes have been observed in
the Dy: 2x4 superstructure[l3], which is not surprising to us since both 2x4 and 2x7
superstructures are composed of the same groups of three maxima as discussed above.
Another feature we can see from the three images is the intensity change in the bright
strips, especially in (b) where the strip breaks into a row of maxima of the same 2a5,-
(lagi=0.384 nm) periodicity along Si [110] as the subunits.

Further observations show that each regime does not appear only at a specific bias,
but in a certain range. Additionally, the same regime in 2x4 and 2x7 superstructures
appears in similar ranges, as illustrated in Figure 5.5, in which the measurement biases

are separated into the three regimes. The measurements for Dy: 2x4 are numerous

61

enough to assign a specific range for each regime, i.e., 0.08 ~ 1.0 V for regime I, 1.04 ~
1.39 V for regime II, and 1.4 ~ 3.01 V for regime 111 (an extension from 2.01 V to 3.01 V
is made to match with Dy: 2x7).The boundaries of the corresponding regime for Dy: 2x7
are shown as matching with Dy: 2x4 where data is absent. For Gd: 2x7, a certain range
can also be assigned to each regime, but the available data are not enough to determine
the exact boundary between regimes. However, one thing that is clear is the bias range of
regime II of Gd: 2x7 overlaps with that of regime I of Dy: 2x7 or Dy: 2x4. This

difference is a reflection of the difference in the electronic structure between Gd and Dy.

 

 

 

IE II- III M Gd: 2x7

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l | HI Dy: 2x7
H: c t
I ll fllll Dy: 2x4

 

 

 

 

I I I

0 0.4 0.8 1.2 1'6 2 2.4 2.8
Bias(V)

Figure 5.5 Bias voltage ranges for three different regimes in empty states imaging of the
subunits in the Gd:2x7, Dy:2x7 and Dy:2x4 structures. Bias voltages for the individual
measurements are shown.

The situation for the filled states is much simpler compared to the empty states.
Generally, the appearance of the 2x7 superstructure is the same in a bias range of -0.45 to
-3.26 V for both Dy and Gd, as shown in Figure 5.4(d). Only the center row of the three

subunit maxima is visible (marked (1)), and the bright strip breaks into a row of maxima

62

(marked (2)) with a 2aSi periodicity along Si [110]. In some circumstances, the bright

strip separates into two maxima along Si[l T0], as seen in (6). Note that Figure 5.4(d)

and (e), and Figure 5.4(a) and (c) are STM images of the same area.
5.1.3 Atomic structure

As discussed for Figure 5.3, the three maxima subunits are the same for both 2x4 and
2x7 superstructures, and these maxima have the same registry with respect to the
underlying Si, that is, in the place of a removed Si dimer. The next question is to
determine the atomic configuration of the bright strip.

The structure of the bright strip can be extracted from the STM images shown in
Figure 5.4(b), (d) and (e), since the bright strips in these images break into maxima. First

thing we can see is that these maxima have the same periodicity as the three maxima

subunits, which is 2(15) apart along Si [110]. If we draw straight lines parallel to Si [1T0]
in these images, antiphase shift (note that 1/2 unit cell shift along 2x direction in the 2x7
superstructure is 1615,’ different) of the three maxima subunits can be observed in (d) and
(e). Moreover, we can see that the bright strip maxima are not in line with the subunit
maxima; they are about M: unit cell off both sides of the straight line. Since we know the
spacing between the straight lines is 2a3,-, hence, the shift between the maxima of strips
and subunits is 1/2 (15,. Therefore, the strip maxima share the same atomic position in both
filled and empty states, which is V2 Si unit cell shift from the subunits along Si [110]. The
shift between maxima along the bright strip in different rows shown in (b) implies an

antiphase shift, similar to that of the three maxima subunits.

63

A

011:.- O" )0 :4 0C? 0

\
J

103000000

9»

O {50 0;". 0:, 0;: 0‘

O

000
O

0
oo

o'o'o (00: Geo o o o

O
o .

0

00-0

 
 
   

i. O 02:

O
O
O

D

00000
060

"i
I
a.

too:

O
0
‘000

I

0.9
U

00
0000000

C1

00
0
oo

'00 C:

A

06 o( _lo\

( .-.

o o 0
COO“) 0000-30000 0

00
000
000

G

o o 0 0.09000 0 o
O‘anooooo

O

000
9.0
oooo

0 COO) 0.0

00000000
0

000-0
co
oooor.
oo
0.

0000

,0

030530

0 O O 0 O
C.)
O

O O

OCOO)
oo o!

.0
6

O O
Q
0
(OO)
O t- O
:00»
O

'u'
r t
.4

'.
3-4
f .
v

‘00) 0°00

wovoOolo

Co 000 o
O Co;;000;‘1.0.

". O 03.“)

A
v

O
(-
O

t) (‘00) 0000000 .

0
oo
o

00

O

o

00
o
o
0

000000

underlying Si atom C subunit maximum
Si O a Si dimer 0 bright strip maximum

Figure 5 .6 Diagram showing the atomic registry of the maxima in the 2x7 superstructure
with respect to the Si (001) substrate. The 2x7 unit cell is marked by a black box, and a
broken line is drawn to show the anti-phase shift of both the subunit maxima and bright
strip maxima.

Based on this information, a diagram of the atomic scale features in the 2x7
superstructure is shown in Figure 5 .6. The subunit maxima sit at the center of the 1x1
unit cells in the underlying layer, and the strip maxima sit in between two Si atoms,
which is shifted exactly 1/2 Si unit cell away from the atomic position of the subunit
maxima. The diagram also shows that the bright strips and the subunit rows actually have .
a gap between them. A 2x7 unit cell is marked, and the antiphase shift of both the
subunits and the strip maxima is shown in the figure. One row and column of Si dimers
are also shown to illustrate the registry of the 2x7 structure with the surrounding Si

surface.

64

5.1.4 Coverage dependence

The occurrence of 2x4 and 2x7 phases are strongly coverage dependent in both Gd/Si
(001) and Dy/Si (001) systems, as shown in Figure 5.7. The percentage of the surface
covered by different phases is plotted against the nominal metal coverage. As seen in (a)
for Gd, the Si surface is firstly 2x4 reconstructed up to 0.2 ML. Subsequently, the 2x7
reconstruction starts to appear at the cost of decreasing the 2x4 coverage up to 0.33ML,
at which the surface is almost entirely 2x7 reconstructed. Starting from 0.36 ML, 3D
nanowires appear and occupy more area with increasing coverage up to 0.92 ML,
decreasing the area covered by both 2D reconstructions. The 2x7 reconstruction dies off
gradually, but the total percentage of surface covered by 2x4 reconstruction and 2x1
clean Si remains approximately constant after about 0.8 ML. This diagram implies that
the whole surface could be totally 2x7 reconstructed at a certain coverage between 0.33 ~
0.36 ML, which is also the coverage at which the nanowires start to nucleate.

Two methods have been used to grow Dy disilicide nanowires: depositing Dy onto a
Si surface at 600°C, with and without post deposition annealing. As we can see from
Figure 5.7(b) and (c), the nanowires start to appear at a much earlier coverage compared
to Gd for either preparation. Details of nanowire and 3D island growth are discussed in
the next chapter. What is clear is that the appearance of the Dy: 2x7 phase strongly
depends on the post—annealing. It is promoted by quenching the sample immediately after
the end of metal deposition at 600°C. There is no critical coverage at which the 2x7
covers the whole surface; nor is there a clear relationship between when the 2x7 phase
decreases, and the nanowires appear. Nevertheless, the Dy 2x7 phase seems to be

associated with coverages between 0.3 and 0.7 ML. A similar behavior is seen for the

65

l 00%

 

800/0 ~-— — I
i \‘ x ‘l
\ \
~. t
K‘ Cl 2X] I

 

 

 
 
 
 

  
 

 

 

 
 

   

 

 

 

I
L»
U

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

   

 

 

 

 

   
    

  

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

00/0 J
O l 0.’7 0.49 0 55 0.63 086 0.92
(a)
1009/0 ' .; .:
I 2x7 gflgi
800/0 :1 I:
60% 2‘4 I
!
Cl 2 1l S -’:
4090 2: I
\ 1
4 5.
I 3D sl~
3090
09/6 I 7 I I I I I I I I I I I I I I Y I I I I I I
0.26 0.45 0.5 0.53 0.72 0.83
1009/0 ’ ’
I 2X7
80% ~- '—
- ‘4‘ 2x4
600/0 — 1— § \
S s S
”Q t s D 221
40% -- — 3 ‘4 3:
x it s
\ \ \
§I§ s I so
20% 4 e b \
\
00/0“ IIYTI fi—r IIIII Tr!
0.1 0.17 0.3 0.3_6 0.4 0.5 0.53 0.6 0.8 I

nominal coverage

(C)

Figure 5.7 Coverage dependence of 2D and 3D phases appearing for (a) Gd grown on Si
at 600°C with post deposition annealing; and Dy on Si at 600°C without (b) and with (c)
post deposition annealing.

66

Ho/Si (001) system where the 2x7 is seen to coexist with the 2x4 in a range of coverage
around 0.5 ML[1 17].

It is easiest to deduce a relationship between coverage dependence and the metal
content of the 2x7 phase for the case of Gd. If we make the same assumption that was
made for the Dy: 2x4 structure in Chapter 3, i.e., the number of metal atoms and empty
states maxima in the STM images are the same, then the metal density in the 2x4 phase is
3/8 20.375 ML, and the metal density in the 2x7 phase is 5/ 14 = 0.357 ML. Both of these
numbers are consistent with the behavior shown in Figure 5.7(a), at least up to 0.36 ML.
The agreement beyond this point depends on our assumption about the metal content of
the 3D nanowires. If we assume that there is only one layer of RE metal atoms inside the
nanowires, then they have an areal density of 1 ML, as implied by the RE disilicide bulk
structure[12]. Hence, a total metal coverage can be calculated based on the data given in
Figure 5.7, which is plotted versus the nominal coverage as shown in Figure 5.8 for all
three cases. The straight line in the figure is a reference line assuming an ideal match
between calculated coverage and nominal coverage. The three curves match with the
reference line very well in the beginning then start to deviate from it at a certain
coverage, about 0.5 ML for both Gd and Dy-cutoff curves, and 0.36 ML for Dy-annealed
curve. This deviation is caused by both second layer nanowire growth and 3D island
formation, which is discussed in the next chapter. Nevertheless, the agreement between
each curve and the reference line gives us a clue that the atom assignment to different
phases is reasonable. We can further test this conclusion by assuming that there are other
possibilities for the RE atom density assigned to the 2x7 superstructure, e.g., 2, 3, 4, 6, 7,

and 8 atoms per unit cell. The total coverage is recalculated according to each assignment

67

 

nominal
—o— Gd

- A - Dy-cutoff

 

 

 

 

 

 

 

 

3 X Dy-anncaled
E 0.8
_, X
E) 0.6 ‘
C
L)
7:3
.75 0.4 X
'_3 X
75 I
" 0.2‘ X
0B . . v . . .

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 l [.1
nominal coverage (ML)

Figure 5.8 The calculated coverage of the three cases shown in Figure 5.7 is plotted
against the nominal coverage.

 

 

0.16
+ Gd

—fl— Dy-cutoff [

\ —A— Dy-annealedxlO /

C
R)
%

 

 

   
 

 

0.08 4.

 

i

Q
Q
.L'.
1

 

sum of squared deviation

 

l

 

 

 

 

2 3 4 5 6 7 8
# ofatoms/Zx7 unit cell

Figure 5.9 Sum of squared deviation between the actual and calculated coverage plotted
versus the assumed number of metal atoms in the 2x7 superstructure.

68

for the data in Figure 5.8 up to the coverage that starts to deviate from the reference line
significantly, resulting in a deviation from the nominal coverage at each data point. These
deviations are squared and summed up for each case and plotted in Figure 5.9. It can be
seen that the minimum error is 5/14 for Gd, 6/ 14 for Dy-cutoff, and 4/ 14 for Dy-
annealed. Allowing for experimental error, and the fact that the 2x7 superstructure should
be the same for all cases, 5/ 14 ML RE atoms is the best assignment for the 2x7

SUPEI‘SII‘UCIUI’C.

5.2 Discussion

The similarities in the appearance and bias dependence of the 2x7 reconstructions
formed by Gd and Dy suggest that they are actually the same superstructure. However,
the chemical identities of the metals still play a role, as suggested by the difference in the
details of bias dependence shown in Figure 5 .5, and the difference in the coverage
dependence of the 2x7 superstructure shown in Figure 5.7.

Similarities also exist in the 2x4 and 2x7 superstructures since they share the same
three maxima subunits and these maxima have exactly the same atomic position and bias
dependence. Considering the close relationship between 2x4 and 2x7 superstructures,
the evolution from 2x4 superstructure to 2x7 superstructure could be a result of
increasing strain in the 2x4 reconstructed areas, which means the appearance of bright

strips between the three maxima subunit rows can ease the strain along Si[1T0]. The

intensity of the three maxima within each subunit in the 2x7 superstructure are more

regular compared to the 2x4 superstructure, in which an alternation of atomic buckling

(LHL, HLH) along [1T0] is observed and explained by atomic buckling caused by strain

69

[13]. The absence of this alternation in the 2x7 also suggests that the 2x7 superstructure
is less strained due to the formation of the bright strips.

The 2x7 reconstruction appears only after the 2x4 reconstruction covers a certain
percentage, about 50%. The 2x4 phase almost always coexists with areas of clean
surface. The 2x7 phase appears only when the 2x4 phase is also present. In cases where
areas of 2x7, 2x4 and clean Si are present, the areas of clean surface are bordered by the
2x4 but not the 2x7. Also the 2x7 reconstruction dies off at higher coverage because of
the decrease in the 2D reconstructed areas due to the consumption of metal atoms by
nanowires or 3D islands. The behavior of the 2x7 phase is consistent with it being
associated with more metal on the surface than the 2x4 phase. This seems to be
inconsistent with the fact that the areal density of metal in the 2x7 phase is slightly less
than that for the 2x4 phase. However, the total amount of metal accommodated in the
2x7 phase is generally higher than that in the 2x4 phase since the percentage of the
surface covered in the 2x7 phase, when it is present, is usually higher than the percentage
covered by the 2x4.

A specific atomic structure for the 2x7 phase is not being proposed here. However,
any model must be consistent with the registry of atomic scale features shown in Figure
5.6. While it is highly suggestive that the number of empty states maxima and metal
atoms is the same, it is not possible to state that the maxima correspond to metal atoms.
In fact it is possible that the metal atoms occupy sub surface positions and that the surface
is terminated by Si atoms, as has been suggested for the surfaces of RE silicides grown

on both Si (111) and Si (001)[107, 118].

70

Furthermore, the metal content of the 2x4 and the 2x7 phases may not be exactly the
same for different RE metals. The 2x4 phase appears to be the same for Gd and Dy. In
addition, the Yb22x4 structure also has three atoms per unit cell[119]. On the other hand,
the Hoz2x4 structure appears to have only 1.5 metal atoms per unit cell[l 17]. More
analysis by other analytical techniques will be necessary to solve this structure in these

RE metal / Si (001) systems.

5.3 Conclusions

Detailed examination of the 2x7 reconstruction of the Si (001) surface induced by Gd
and Dy shows that this phase appears to be the same for both metals, and is similar to the
2x7 phase induced by Ho[117]. Furthermore, the 2x7 structure contains subunits whose
appearance and bias dependence in STM images is the same as the 2x4 superstructure
induced by Gd, Dy, and several other RE metals. The registration of the maxima seen in
the STM images with respect to the surrounding Si substrate has been determined for
both the 2x4 and the 2x7 phases. Coverage dependent measurements show that the metal
content of the 2x4 and the 2x7 phases is 3/8 and 5/ 14 ML, respectively. Any model for
the atomic structure of the 2x7 phase will be constrained by the STM images and the
measured metal content. Aside from the structural similarity of the Dy and the Gd
induced phases, there are some differences between the two metals, particularly with

respect to the coverage dependent growth behavior.

71

Chapter 6 A comparison of Dy and Gd silicide nanowire growth

In this chapter, more data on Gd silicide growth is presented, and the behavior of the
two metals is compared. In particular, some insight into the differences between the two
metals is obtained by comparing the Gd data with Dy growth for a wider range of
conditions than what has been previously discussed in Chapter 4. This chapter also
includes data on a particular variant of nanowire growth: growth of parallel arrays of Dy
and Gd nanowires on single-domain, vicinal Si (001) surfaces. Experimental results show
that the growth temperature, annealing duration, metal coverage and metal varieties are

important factors for RE silicide growth.

6.1 Growth on normal Si (001)

6.1.1 General observations

Generally, three types of structures can be formed when depositing RE metals onto Si
(001) substrates at elevated temperature -- nanowires, 3D islands, and 2D reconstructions.
All of these structures can coexist together on the same surface, as shown in Figure 6.1.
This sample was prepared by depositing 0.4 ML Dy on Si (001) substrate at 600°C with 5
min post deposition annealing (PDA). As seen in the figure, long islands (nanowires or
nanowire bundles) run in two perpendicular directions, with compact rectangular 3D
islands lying at the nanowire junctions. Additionally, second layer growth can be
observed clearly on top of both the nanowires and the 3D islands. There are two types of
3D islands can be seen, one with a flat surface and one with a wavy surface. Although it
is not apparent in this rendering, the substrate is not bare Si, but is reconstructed by the

RE metal. The 2D reconstructions, usually 2x4 and 2x7 for Dy and Gd, have been

72

discussed in Chapter 3 and 5. The nanowire and the 3D island growth of both Dy and Gd
disilicides show strong coverage dependence. The Dy silicide 3D island is very sensitive

to the annealing duration, as will see from the comparative experiments with and without

PDA.

 

Figure 6.1 Topography of Si (001) surface with 0.4 ML Dy deposited, showing
coexistence of nanowires and 3D islands, and second layer growth.

6.1.2 Gd silicide nanowire growth

When depositing Gd onto Si (001) surface at 600°C with 5-10 min PDA, only 2D
reconstructions (2x4 and/or 2x7) are formed up to about 0.33 ML. The Si surface is
almost entirely 2x7 reconstructed at 0.3 ML mixed with small areas of 2x4
reconstruction and 2x1 clean Si. When an additional 0.06 ML Gd is deposited onto this
sample, nanowires appear, as shown in Figure 6.2(a), implying the existence of a critical
coverage for nanowire nucleation between 0.3 ML and 0.36 ML. The surface in (a) is

fairly rough due to the existence of a large number of Si steps. It is clear that each

73

 

Figure 6.2 (a)-(c) (350 nm x 350 nm) and (d) (300 nm x 300 nm) are STM images which
show coverage dependent topography of the Gd/Si (001) surface. (e) (60 nm x 60 nm) is
a closer View of the area enclosed in the white box in (b). (f) (75 nm x 75 nm) is a closer
view of the white box in (d).

74

nanowire grows on a flat terrace, 90° rotated from those on adjacent terraces. It is worthy
to note that the nanowires can cross a step, an example of which is marked by an arrow in
the image. This is not observed when depositing metal in one evaporation onto a clean
sample. We believe this occurs due to extended annealing that occurs during evaporation
calibration when adding additional metal onto an old sample. Similar phenomenon has
been observed when annealing a room temperature (RT) Dy deposited sample up to
600°C, as will be discussed later on.

Surfaces with more Gd deposition are shown in Figure 6.2(b) at 0.63 ML, (c) at 0.86
ML, and (d) at 1.9 ML. At this image magnification, only the overall surface morphology
is visible. As coverage increases in (b) and (c), the nanowires become denser and wider,
with additional layer growth on top. The wide nanowires are actually nanowire bundles,
consisting of parallel nanowires separated by narrow trenches, as seen in (e), which is a
closer view of the area enclosed in the white box shown in (b). Gd silicide nanowires are
more often observed as nanowire bundles compared to Dy silicide nanowires. At the
highest coverage shown in (d), 3D islands are formed. The wide wires are still nanowire
bundles, as shown in (f) which is a closer view of the area enclosed in the white box in
(d), in which 3D islands with flat top and wavy surface can also be observed.

The nanowire widths are measured as multiples of the Si periodicity as,- = 0.384 nm,
as we have done for Dy nanowires in Chapter 3. A minimum nanowire width of 2 a5;
(0.77nm) is found, which is the same as that for Dy nanowire. The width distribution of
Gd silicide nanowires, ranging between 2a5,- and 505,‘ as shown in Figure 6.3, is much

narrower than that of Dy nanowires. which ranges between 2615) andl6agi. Nanowire

75

bundles can consist of a large number of single nanowires, and the total width of the

bundle can be as much as 100 nm.

 

 

 

number of nanowire

 

 

 

 

 

 

 

Id

3 4 5
width ((1 g, )

Figure 6.3 Width distribution of single Gd silicide nanowires in bundle.

6.1.3 Dy silicide nanowire growth

When using the same method to prepare Dy silicide nanowires, i.e., deposition with
PDA, nanowires start to appear at a coverage as early as 0.1 ML, with significant 3D
island formation above 0.3 ML. Figure 6.1 is a good example showing coexistence of
nanowires and 3D islands at 0.4 ML. More details of growth with PDA have been
discussed in Chapter 4. The growth of 3D islands is not desirable from the viewpoint of
optimizing nanowire growth. We noticed in our experiments that the number of islands
decrease with shorter annealing duration, implying the growth of 3D islands might be
suppressed by shortening the annealing duration. Therefore, we tried another method to
grow Dy silicide nanowires, shortening the annealing duration to zero by cutting off the
heating current immediately after deposition, which we call post deposition quenching
(PDQ). Figure 6.4(a)-(c) show the nanowire growth with increasing coverage with PDQ.

No 3D islands are formed at a coverage as high as 0.53 ML in (b), and only a small

76

number of islands can be found at the junctions of nanowires at a coverage 0.72 ML in
(c). One other significant difference from FDA growth is that most of the nanowires
formed under 0.5 ML are single nanowires, as seen in (a). A small number of nanowire
bundles with slight additional layer growth can be seen in (d) at 0.53 ML, which is a

closer view of the area inside the square box in (b).

 

Figure 6.4 (a)-(c) are STM images (all 350 nmx350 nm) showing coverage dependent
topography of the Dy/Si (001) surface prepared by PDQ. (d) (75 nmx75 nm) is a closer
view of the area enclosed in the square box in (b).

77

6.1.4 30 islands

As discussed above, the growth of 3D islands can be suppressed by PDQ, at least at
lower coverage. On the other hand, what happens if we increase the post deposition
annealing duration? Figure 6.5 (a) shows a sample prepared by depositing 0.6 ML Dy
onto Si surface with 10 min post annealing. Both nanowires and 3D islands can be
observed as expected. If we continue to anneal this sample at 600°C for 1 hour, most of
the nanowires disappear and large 3D islands appear as shown in (b), implying that for
Dy, long annealing duration promotes 3D island growth. The 3D islands can grow
downward below the Si surface, e.g., island (1), or above the Si surface, e.g., island (2).
The inset is the height profile of island (1) along the broken line, showing the island is
situated completely in the trench, which is caused by consumption of Si atoms during

formation of the island.

 

Figure 6.5 Topography of (a) 0.6 ML Dy deposited sample prepared by 10 min PDA,
image size is 200 nmx200 nm, and (b) the same sample shown in (a) which was annealed
another 1 hour, image size is 350 nmx350 nm. The inset in (b) is the height profile of
island (1) along the broken line.

78

 

amplitude

wave length

height

 

 

 

 

Figure 6.6 (a) 3D rendering of wavy island (2) shown in Figure 6.5(b). (b) The height
profile of the wavy island along the broken line in (a). (c) Wavy island formed by
deposition of 0.92 ML Gd.

79

Again, there are two types of islands can be seen, one with a flat top, e.g., island (1).
and one with a wavy surface, e.g., island (2), which is consistent with the islands shown
in Figure 6.1. A 3D rendering of the wavy island (2) is shown in Figure 6.6(a), and the
surface undulates sinusoidally, as shown by the height profile in Figure 6.6(b). Wavy
surfaces have also been observed in Gd silicide islands at high coverage. One example is
shown in Figure 6.6(c). The wave length, the amplitude of the undulation, and the island
height measured from Si surface for a number of islands are listed in Table 6.1. It shows
that the undulation amplitude is very close for all of the wavy islands, on average 0.13
nm. The wavelength varies, but no obvious relationship can be established between the
wave length and the island height. Note that the amplitude and wavelength are not bias

dependent

Table 6.1 Measured amplitude, wavelength and height of wavy islands.

 

 

 

 

 

 

 

 

 

 

 

amplitude wavelength island height
(nm) (nm) (nm)
0.10 5.93 1.23
. 0.14 6.24 1.05
Dys‘z 0.13 6.92 1.42
0.15 7.16 1.50
GdSI-z 0.13 5.58 1.07

 

Both of the flat top islands and wavy islands can grow into elongated shapes that look
like wires. The 3D rendering in Figure 6.6(a) shows that the members (ii) and (iii) are
wavy, and elongated in a direction perpendicular to the undulation direction. This is
different from member (i) and the Gd island in Figure 6.6(c), which are both elongated
parallel to the undulation direction but have much smaller lengths than (ii) and (iii). A
general conclusion about a correlation between island elongation and undulation direction

cannot be given due to the limited number of observations of these wavy islands.

80

We draw a distinction between these “elongated islands” and nanowires based on the
fact that they are much wider; the width of these islands is in the range of 8 ~ 17 nm for
Dy silicide, much larger than the width defined for nanowires, 0.8 ~ 6 nm for single Dy
silicide nanowires, and 0.8 ~ 2 nm for single Gd silicide nanowires. These elongated
islands can grow fairly long. The longest island we observed is 280 nm long, comparable
to nanowires. These elongated islands coexist with even larger square or rectangular 3D
islands whose widths are in the range of 20-30 nm.

6.1.5 Height quantization

Both the nanowires and the islands exhibit layer-by-layer growth, which is evident
from Figure 6.1, 2 and 5. As we have done for Dy nanowires discussed in section 4.1.2,
the height of Gd nanowires were measured with respect to the substrate surface, giving us
a minimum height ranging between 0.48 nm and 0.50 nm, with an average of 0.49 nm.
This height range is smaller than that of Dy nanowires, which range between 0.48 nm and
0.56 nm in both filled and empty states, with an average of 0.52 nm, as mentioned in
Chapter 4. However, both averages are comparable. The height of 3D islands is hard to
measure since they can grow on top of the nanowires (Figure 6.2(f)), or in a trench
(Figure 6.5(b)). Generally, 3D islands have heights above 1 nm. A consistent minimum
height cannot be found for the 3D islands, as shown in Table 6.1, which could be a clue
that these 3D islands are different phase from nanowires.

Once again, it is useful to measure the height of additional layers with respect to the
top surface of adjacent or underlying nanowires or islands in order to minimize the
effects of different surface structure. It was mentioned in Chapter 4 that the height of

additional layers on Dy silicide nanowires is actually quantized in increments of 0.35 nm.

81

The additional layers on top of wavy islands are also quantized in the same thickness, but
that of flat islands have a quantum height of 0.66 nm, and another possible value larger
than 0.84 nm. Similarly, the additional layers grown atop Gd silicide nanowires also have
an average height of 0.35 nm. We do not have enough data for Gd silicide islands since
they grow so infrequently.
6.1.6 Coverage dependence

The appearance of each phase is strongly related to the metal coverage under different
growth conditions, as shown in Figure 6.7, which has more data above 1 ML compared to
Figure 5.7. The diagrams in the figure are plotted with the percentage of the surface
covered by different phases versus the nominal coverage, where 3D phases include both
nanowires and 3D islands. All the samples in (a) were prepared by depositing Gd onto
600°C heated Si substrates with 5 to 10 min PDA. And all the samples shown in (b) and
(c) were prepared by depositing Dy onto Si surface of 600°C with PDA and PDQ,
respectively. These diagrams show that the area covered by 3D phases increases with
increasing coverage before 1 ML, but stop increasing after 1 ML and even drops,
although it increases at the highest coverage measured in (c). A specific minimum
coverage for Gd silicide nanowire growth can be seen in (a), between 0.33 ML and 0.36
ML, which might correspond to the saturation coverage of 2x7 superstructure mentioned
in the previous chapter. But such a critical coverage can not be seen for Dy from the
available data as illustrated in (b) and (c), and the 3D phases start to appear at a much

earlier coverage compared to Gd in (a).

82

 

 

 

 

 

 

.
.
u
n
u
a
n
u
o
u
.
u
a
o

 

 

 

Gd
100% :3 3: '3 -1 :~ .g .3
:3 £2 23 2: g: 3 “,7
80%“ 3: :3 1: I: ;: :
f: 3' \ s 5
60%- I S S .2X4
, I; :2 I § 3
40%‘ 3 33 If :i s b,
:2 21 3: i: S i
“' § \' : :f 3: q 3D
20%-\ t S ' r: :z p. '
\ \ \ ._ .
s t ‘~: S 7.- 3
0%4 ‘ 3‘ -\ \ \ H H IIIIII
0. l 02 0.27 0.3 0.33036 0.49 0.55 0.63 0.86 0.92 l.9

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10096 , ,
_ 3§ 3; I 2x7
80% - g; 33 3
35 E5 5: 2x4
600/0 a ”i I
T: I. 3.
:g s 3. 1:1 2x1
40% « i; 3 ' ~
-I \
E 3 ~ I 3D
20% - ~‘ ~
00/0 IIT—rr rrrrrrrrrr Irr1 rrr TIIIIrII 11"
0.26 0.45 0.50.53 0.72 0.83 2.4
Dy-PDA
100% ’
80%1
S \
60%1 § ~: S
t t i
\ \‘ x x
k ‘x‘ s s
40%“ S S SS 4‘ x
S S S S
20% :3 s 3‘ S
s
0%1

 

 

 

 

0.10.17' ' ' 103' ' 036 0.4 0.50.5306' ' ' 1 0.8 1 1 1.6 3.3
nominal coverage

Figure 6.7 Coverage dependence of 2D and 3D phases appearing for (a) Gd grown on Si
at 600°C with PDA; and Dy on Si at 600°C with (b) PDQ and (0) FDA.

83

Again, the total coverage can be calculated based on the same assumption made in the
previous chapter, i.e., the metal density in Dy: 2x4 structure is 0.375 ML, and that in the
2x7 phase is 0.357 ML, and 1 ML in the nanowires of minimum height. The calculated
results are plotted versus the nominal coverage as shown in Figure 6.8 for all three cases.
The straight line in the figure is a reference line assuming an ideal match between
calculated coverage and nominal coverage. The three curves match with the reference
line very well in the beginning then start to deviate from it at a certain coverage, between
0.36 and 0.5 ML for all three curves. This deviation is caused by both second layer

nanowire growth and 3D island formation, as shown in Figure 6.2 and 4.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2
1.8 / .
1 6 / —-— nommal
E?) 14 / - 0- - Gd
g 1 2 / —A——Dy-PDQ
g 1 / ——><——Dy-PDA
3:: /
3 0.8
733 0.6 / - /<
0.4
0.2 -
O a ' I l I l l
0 0.5 1 1.5 2 2.5 3 3.5

nominal coverage

Figure 6.8 The calculated coverage of the three cases shown in Figure 6.7 is plotted
versus the nominal coverage.

For Dy coverage less than 0.5 ML (Figure 6.4(a)), only single nanowires are formed.

At 0.53 ML (Figure 6.4(b)). second layer growth and nanowire bundles start to appear. At

84

0.72 ML (Figure 6.4(c)), the nanowires are highly bundled with some population of
second layer nanowire growth and 3D islands formed at the nanowire junctions. Since the
second layer nanowires and 3D islands contain more metal atoms than one layer of
nanowires, certain amount of metal atoms will be miscounted during calculation based on
the assumption of 1 ML metal atoms in the nanowires.

The same conclusion can be obtained by observing the coverage dependence of Gd
silicide growth shown in Figure 6.2. Generally, the Gd disilicide nanowires are more
likely to appear as nanowire bundles, as shown in Figure 6.2(b)-(d). At 0.63 ML, second
layer growth can be seen. At 0.86 ML in (c), the nanowires are highly bundled and
second layer growth getting more and more populated, and 3D island growth can be seen
in (d).

The coverage dependence of Dy silicide growth prepared with FDA has been
discussed in Chapter 4. As shown in Figure 6.8, the Dy-PDA curve deviates from the
reference line at about 0.36 ML, much earlier than that of Gd, and Dy-PDQ, which is
caused by the early formation of 3D islands and second layer growth, as shown in Figure
6.1 at 0.4 ML with PDA. When the coverage is well above 1 ML, generally large 3D
islands are formed no matter what condition is used. The irregular shape of the Dy-anneal
curve might be a result of annealing duration since it is different from sample to sample
in the range of2 to 10 min.

Nevertheless, the degree of agreement between nominal and calculated coverages up
to 0.5 ML shows that the assignment of metal density to the different phases is

reasonable, as was discussed in Chapter 5.

85

 

Figure 6.9 STM images (180 nmx180 nm) showing coverage dependent topography of
the Gd/Si (001) surface grown at 500°C with (a) 0.46 ML Gd deposited, taken at VT:
+0.44 V; and (b) 0.54 ML Gd deposited, taken at VT: +1.12 V. (c) is a high resolution
image (40 nmx40 nm, VT: —0.67 V) of the 2D reconstruction appeared in (b). (d) is an
STM image (180 nmx180 nm) of the same sample shown in (b) which was annealed at
600°C, taken at VT: +1.35 V.

6.1. 7 Growth behaviour at 500 ‘C
Figure 6.9 shows the growth behavior of Gd deposited onto Si (001) surface at
500°C. It can be seen that only 2x7 reconstruction is formed without nanowires at a

coverage of 0.46 ML in (a), and nanowires start to appear at a coverage of 0.54 ML in

86

(b). Obviously, the 2D reconstruction in (b) is different from the 2x7 reconstruction in
(a). As shown in (c), which is a high resolution image of the 2D reconstruction in (b), part
of the area is still 2x7 reconstructed as marked, but the other areas are covered by an
unknown structure. When the sample shown in (b) is annealed at 600°C, nanowires
appear and the 2x7 reconstruction is recovered, as shown in (d). It is possible that the
unknown structure is 2x7 with an excess of metal, and that this excess is able to form
nanowires during the 600°C annealing.

The deposition of Dy at 500°C shows similar behavior. No nanowires are formed at
0.38 ML but 2x7 reconstruction, and large number of short nanowires are observed at
0.66 ML. Both samples were prepared without PDA. When annealed the sample in (a)
was annealed at 600°C for 20 min, nanowire-like structures appear on a completely 2x1
reconstructed Si surface. Therefore, a conclusion here is that the metal atoms are more
likely to be accommodated in a 2D reconstruction rather than in nanowires at lower
growth temperature.

Preinesberger et a1. [4] show one image of 1 A Dy (= 0.47 ML) on Si (001) grown at
500°C. They described the growth mode as "close-packed Dy silicide wires", but the
image looks very similar to Figure 6.9(a) which is a 2x7 reconstructed surface, not wires.
6.1.8 Room temperature growth with PDA

When depositing 0.16 ML Dy onto Si surface at room temperature, then annealing it
at 600°C for 10 min, a black line is observed, as shown in Figure 6.10(a) in the filled
states, and this line becomes bright in the empty states in (b). Figure 6.10(d) shows a
“wire” across a step in the empty states, but it disappear in the filled states in (c).

Apparently, the long nanowire-like features shown in Figure 6.10 are totally different

87

from the nanowires discussed elsewhere. Similar “black lines” have been observed after
annealing a sample with nanowires, which could be traces of nanowires removed by
annealing. Some other features could be described as nanowires that are buried so that
their top surface is roughly level with the substrate. In any case, the growth behavior for
RT deposited and post annealed samples is different than that seen for samples growth at

elevated temperature, particularly in the initial stages of nanowire nucleation.

+1.08 V

 

Figure 6.10 (a) and (b) (size: 60 nmx70 nm), and (c) and (d) (size: 45 nmx45 nm) are two
pairs of dual bias images of a sample prepared by depositing 0.16 ML Dy onto Si at room
temperature then annealing at 600°C.

88

6.1.9 Higher temperature growth

Preliminary experiments have been done on growth of Gd silicide at higher
temperature than 600°C. As we discussed, nanowires can be formed at 600°C when the
coverage is above 0.357 ML. When we increase the growth temperature to 650°C, only
2x7 reconstruction without nanowires is seen at 0.45 ML for Gd. When depositing 0.4
ML Gd at 700°C, the Si surface is slightly 2x4 reconstructed, leaving most of the surface
2x1 reconstructed. Similarly, when depositing Dy at 700°C, no nanowires are formed at
0.48 ML. All of the nanowires are removed when the 0.78 ML sample with nanowires
grown at 600°C is annealed at 700°C. In both cases for Dy, the 2x1 reconstructed clean
Si surface is recovered with small area 2x4 reconstructed. Therefore, nanowires and
islands cannot survive at 700°C for both Gd and Dy silicides. Furthermore, the apparent
metal coverage on the surface is much less at either 650°C or 700°C than at 600°C,
implying that deposited metal is lost from the surface, either by re-evaporation or

diffusion into the substrate.

6.2 Growth on vicinal Si (001)

For practical applications of these nanowires, it is of paramount importance to control
where and how they grow. In particular, one proposed molecular electronics architecture
has two perpendicular, planar nanowire arrays at two vertical levels and have switching
molecules at nanowire junctions[120]. In order for RE silicide nanowires to be used as a
component in this structure, growth of an array of parallel nanowires has to be realized.
In the context of previous observations of nanowire growth on normal Si (001) substrates,

the network structure composed of nanowires running in two perpendicular directions has

89

to be eliminated and the growth of secondary phases such as the 3D islands has to be

suppressed.

 

Figure 6.11 Scanning tunneling microscope (STM) images showing (a) a clean vicinal Si
(001) surface with 4° miscut (l9 nmxl9 nm), and (b) the vicinal surface after deposition
of Dy at 600 °C with 10 min postannealing (300 nmx300 nm).

The most straightforward approach to grow a parallel array of nanowires is to use a
vicinal Si (001) substrate, as shown in Figure 6.11(a). This particular substrate is tilted 4°
off towards Si [1 T0] , resulting in a double height step spacing of 39A: 10as,-. All the
steps are perpendicular to the Si dimer row direction, and all Si dimers are parallel to the
step edge. This image shows a typical area of the surface with less than 5% minority
dimer orientation, due to local splitting of the double height steps. Since nanowires grow
perpendicular to the Si dimer row direction on the substrate, they would grow parallel to
the step edges on the vicinal surface. Figure 6.1 1(b) shows the result of Dy deposition on
the single domain Si (001) surface under typical growth conditions used for growth on
flat substrates, i.e. deposition at a substrate temperature of 600°C, and several minutes of
PDA at the same temperature. Dy silicide growth has disrupted the step distribution on

the surface, and as a result, there are short nanowires running in both directions as well as

90

a significant population of rectangular silicide islands. This is clearly not the way to
prepare a single domain nanowire array.

We have tried two different approaches to form parallel nanowire arrays on single
domain vicinal Si (001). The first is to vary the growth kinetics, and the second is to
optimize the choice of RE metal. In either case, the central issue is to minimize the
growth of the large 3D islands, whose larger volume in comparison to nanowires implies

a greater local consumption of silicon on the substrate, which in turn greatly affects the

i
1

step distribution on the surface. This is clearly happening in the case of the 4° miscut
vicinal surface in Figure 6.11(b).

For Dy, it is possible to greatly reduce the relative proportion of 3D islands to
nanowires by cutting off the substrate heating current immediately at the point at which
the metal deposition ends (i.e., PDQ, or without PDA), as discussed sections 6.1.3 and
6.1.4. It is clear that without PDA, the density of nanowires is higher and there are
virtually no rectangular islands on the surface, whereas annealing time increases the
relative density of 3D islands versus nanowires. This behavior suggests that the
nanowires are in fact a metastable phase.

With this in mind, Dy growth experiments were repeated on vicinal surfaces, without

PDA. A typical result is shown in Figure 6.12(a). The number of rectangular islands is

 

greatly reduced, and furthermore most of the deposited Dy has formed into parallel

nanowires whose length is limited only by step wandering, or misorientation between the
step edge directions and the [1 T0] direction. When annealing this sample for another 30

min, nanowires disappear, and 3D islands appear. the same as what we discussed for

91

growth on normal Si substrates. Besides rectangular islands, a fairly long elongated island

forms running from the upper right to lower left as seen in (b).

 

Figure 6.12 (a) STM image (800 nmx800 nm) showing parallel nanowires formed on a
0.38 ML Dy deposited vicinal Si (001) surface without PDA. A small number of 3D
islands still can be seen. After annealing this sample 30 min, the topography is shown in
(b) of image size 420 nmx420 nm. 3D islands of various shapes appeared, and only a
small number of nanowires survived.

 

Figure 6.13 STM images showing a parallel nanowire covered surface free of islands
formed by depositing 0.77 ML Gd on a vicinal Si (001) substrate (300 nmx300 nm). The
sample was prepared with 5 min PDA.

92

Another approach to improving nanowire growth is the judicious choice of metal.
Since these nanowires arise due to an anisotropy in lattice mismatch between the silicide
and the substrate, and in particular since there is a good match between the silicide and
the silicon along the long dimension of the nanowire, one possible criterion for choice of

metal would be the best possible match between the a dimension of the hexagonal

silicide. and the lattice constant (as,- = 3.84A) of the Si (001) surface along < 1T0 >
directions. For example, Dy has a mismatch of -0.23% in this direction, and in general,
nanowire formation is seen for metals with less than 1.2% mismatch [7]. Y would appear
to be an even better choice with 0.0% mismatch. However, we have found that Y
nanowire growth can be problematic due to growth of 3D islands that are not suppressed
by quenching.

A more objective figure of merit for growth is a high ratio of nanowires to 3D islands.
We have obtained the best results for Gd which has +0.96% mismatch along a. Figure
6.2(b) shows 0.63 ML of Gd grown on flat Si (001), with PDA. These are similar growth
conditions as the Dy surface shown in Figure 6.5(a). There are only well-ordered bundles
of nanowires on the surface, with no islands seen even at the nanowire junctions. Gd
nanowire growth on the vicinal surface is shown in Figure 6.13. In this case, the surface
was grown under very similar conditions as on the flat surface, i.e. with 5 minutes PDA.
The surface is covered in parallel nanowires, and 3D island growth is absent, just as on

the flat surface.

93

6.3 Discussion
6.3.1 Effects of annealing and growth temperature

Annealing duration plays a very important role in the growth of Dy silicide growth.
Longer annealing duration promotes 3D island growth. With PDA, the 3D islands can
appear as early as 0.3 ML, but appear after 0.5 ML with PDQ. Further annealing drives
metal atoms to leave nanowires and form large 3D islands. Annealing can also affect the
nanowire bundling. Similar to island growth, the nanowires start to bundle at about 0.3
ML with PDA, but are still in single at 0.5 ML with PDQ. Additionally, annealing can
affect the growth of 2D phases, for example, the 2x7 reconstruction appears in a very
narrow range with PDA.

The effect of annealing on Gd silicide growth does not appear to be as strong as on
Dy silicide growth. The Gd nanowires are very stable upon annealing without
transforming into islands, and the island growth depends mainly on coverage.
Unfortunately, we have not done comparable experiments on Gd silicide growth with
PDQ and with longer time annealing.

The growth temperature is also an important factor during RE silicide growth. At low
temperature like 500°C, the nanowires start to appear at higher coverage compared to
600°C, more metal atoms enter the 2D phases. But at higher temperature than 600°C, the
surface starts to lose metal atoms due to re-evaporation or diffusion, and it seems that
silicides do not survive extended annealing at 700°C.

6.3.2 Multiphase coexistence and atomic model
The effect of annealing duration on the growth of nanowires and islands could result

from the competition between tetragonal/orthorhombic and hexagonal phases, which has

94

been observed for thick films (>20 nm) of Gd silicide on Si substrates by Molnar et al.
[67, 68, 121]. The hexagonal phase has lower activation energy of formation than the
orthorhombic phase, but the difference is very small, and the ratio of two phases strongly
depends on the annealing temperature, annealing duration and thickness. This is relevant
to submonolayer growth of RE silicides on Si (001) if we assign hexagonal structure to
nanowires and orthorhombic/tetragonal structure to 3D islands as has been done for the
case of Dy silicide in Chapter 4.

The three types of structures are closely related. When growing on Si (001), the
hexagonal structure has two orientations, RESI2-x ( l l00)//Si (100) with RESiz.x [0001]
//Si[01l] and RESiz.X (1100) //Si (100) with RESI2-x [0001] // Si[OTT] [39, 64, 80, 81],

and the orthorhombic and tetragonal structures have the same orientation (note that the

orthorhombic Gd812 type structure is only a deformation of the tetragonal ThSiz type
structure), RESi2_x (001) //Si (001) with [100]//Si[ll0] [57, 61, 82]. Based on this

information, atomic models can be built for the growth of hexagonal and tetragonal
phases, as shown in Figure 6.14, in which (b) and (c) are the side views of the first and
second layer nanowire growth and 3D island growth, respectively. 14(b) is a revised
version of the model shown in Figure 4.9. It can be seen that at the interface between
silicide and Si substrate in both models, the metal atoms occupy a position essentially the
same as that in the bulk silicide of both structures.

The two silicide structures in (b) and (c) can be described as different arrangements of
the same building unit block: a parallelepiped with eight silicon atoms as apexes, and a
rare earth atom at the center [33], which is shown in (a) and marked by a rectangular box

in both models. These unit blocks are stacked along the [001] direction in the tetragonal

95

 

[001] [1100]
11120]

  

[170]

 
 

[000 1 ]

  
  
  
     

[0011 [0011
$1100] 7

   

Figure 6.14 Side views of (b) hexagonal and (c) tetragonal RE silicides grown on Si (001)
substrate, both of which can be built from the unit block shown in (a).

and orthorhombic structures, and along the [1T00] direction in the hexagonal structure.

All of the blocks in (b) are in the same vertical plane, translated half unit block between

each other in the hexagonal structure. While in (c), the layers (2) and (3) are in the same

96

vertical plane and reserve the same relationship between blocks as that in the hexagonal
structure, but layers (1) and (4) are in another vertical plane, translated half unit block
into the paper with respect to layers (2) and (3), as marked by different shades. In
addition, the largest face of the building element is square in the tetragonal structure, and

rectangular in the orthorhombic and hexagonal ones. The lattice parameters of the

tetragonal and hexagonal structures are related, azbz (a;,,.,,.+c;,,._,—)/2 and (‘22 fig,“ [39].

The first rationale for the assignment of 3D islands to the tetragonal or orthorhombic
phase is their compact rectangular shapes. Experiments have also shown that for the
similar Er/Si (001) system where both nanowires and 3D islands are seen, the crystal
structure of the islands is tetragonal [107]. Even though it has not been proved by any
experimental techniques that Gd or Dy nanowires are of hexagonal A1132 type structure,
the anisotropic lattice mismatch to Si (001) surface explains well the nanowire formation
mechanism [6, 7, 14]. Furthermore, the hexagonal structure can also explain the 0.35 nm

height quantum of Dy silicide nanowires, which is the unit cell height (0.332 nm) along

[1T00] [14], as shown in (b). Similarly, the quantum height (0.35nm) for Gd silicide
nanowires also agrees with this model, in which the unit cell height of Gd disilicide is
0.336 nm. Following this line of reasoning, the quantum height 0.66 nm of flat 3D islands
can be explained by the tetragonal ThSiz model, which is the total height (0.62 nm) of
layers 3 and 4 in (c), including two layers of metal atoms inside.

Notice that in both models the top surface is shown as terminated with a Si bilayer,
similar to the Si (001) termination of bulk silicon. This is analogy to the fact that RE
silicides grown on Si (111) terminate with a buckled Si bilayer [69, 72]. If this is the case,

the first layer of nanowires should consist of one layer of metal atoms (as discussed in

97

Section 3.4), and one Si bi-layer as a termination, as shown in (b). The apparent height of
the nanowires of minimum height, 0.52 nm for Dy, and 0.49 nm for Gd, is larger than the
measured quantum height of 0.35 nm. This is possibly caused by the different electronic
states of the top surface of the nanowire, and the substrate which can be either clean Si
(001 )2x1, or either the 2x4 or the 2x7 reconstructions.

Note that the first layer of tetragonal phase in the 3D islands is drawn including two
layers of metal atoms in (c) for convenience. The thickness may vary in the real situation
as suggested by the measurements that no unique number can be found for the minimum
height for the 3D islands. However, it is worthy to note that the distinction between the
hexagonal and the other two silicide phases only arises when there is more than one layer
of metal atoms present.

6.3.3 Lattice mismatch and strain relaxation

Apparently, the competition between nanowires and islands, or between the
hexagonal phase and the tetragonal/orthorhombic phases, is a function of annealing
duration and metal coverage, as well as the metal species since the coexistence of
nanowires and islands is easier to realize for Dy silicides compared to Gd silicides. The

hexagonal Gd silicide nanowires are more stable than Dy silicide nanowires, but islands

 

do appear for Gd silicides at higher coverage and with longer annealing duration.

Essentially, the competition is controlled by the ease of epitaxial growth of each
phase. As we know, lattice mismatch and chemical bonding at the interface are two
factors to determine the realization of a heteroepitaxial system [108, 109, 122]. Due to
the existence of lattice mismatch, a pair of superlattices with similar translational

symmetry across the interface can be defined instead of a common superlattice (or

98

coincidence lattice) [108, 122]. The mismatch between the two superlattices and the unit
cell area of the superlattices are two parameters that characterize the lattice match [88].
For RE silicide/Si (001) systems, the disilicides of different structures can form almost
the same interface, implying similar chemical bonding configuration, as shown in Figure
6.11. Therefore, the lattice mismatch would be crucial for the epitaxial growth. As shown
in Table 1.6, comparing the hexagonal to either the orthorhombic or the tetragonal
silicide phases, a large uniaxial strain in the former can be redistributed over two axes in
the latter. Unit cell area is a measure by which one can assess the effect of biaxial strain
on epitaxial growth [88, 123]. We can use the ratio of the unit cell area of the tetragonal
phase to the hexagonal phase as a measure of strain, and the percentage deviation from
1.0 as an areal mismatch (AM), where AM(tet/hex)=(areatct/areahcx-l). Then AM(tet/hex)
is 40% higher for Gd than Dy. Similarly, AM(orth/hex) is 50% higher for Gd than Dy. In
both cases, this would support the tendency for Gd to be less likely to form non-
hexagonal phases than Dy. If we consider the case of Er, AM(tet/hex) is very small which
is consistent with the fact that small coverages of Er form at least as many 3D islands
(tet) as nanowires (hex). In any case, the mismatch in unit cell area correctly predicts
trends in the different behavior of different RE metals, even if it cannot be used to predict
the details of the growth behavior for a particular system.

The presence of elongated 3D islands could also be related to strain relaxation.
Firstly, we believe that these elongated islands are different from the nanowires in that
they usually appear after annealing and are much wider and higher than nanowires. These
islands could result from a shape transformation proposed by Tersoff and Tromp [124],

where compact, symmetric islands are favored below a critical island size, while

99

elongated islands with high aspect ratios are the preferred shape for large islands. Thus
the shape transformations taking place as a result of the system relaxing its strain may
provide a convenient way of producing quasi—one-dimensional quantum-wire structures,
without the necessity of an anisotropic lattice mismatch.

Sinusoidal undulated 3D islands could be of orthorhombic or hexagonal structure
since a one dimensional undulation is more like to be related to anisotropic strain
relaxation, and the measured quantized height 0.35 nm is the same as that of the
nanowires, but different from that of the flat 3D islands. Similar phenomenon has been
found in film growth and explained by variations in elastic strain energy and surface
energy of the film related to lattice mismatch [110, 125-127].

Therefore, many phenomena apparent in the growth of RE silicides on Si (001) can
be related to effect of strain relaxation. Based on the STM results, we can do little more
than enumerate the different growth morphologies that can occur. The detailed
connection between strain and the growth behavior, as well as the assignment of
nanowires and the 3D islands to different silicide phases, awaits confirmation from other
analytical techniques.

It is illuminating to draw an analogy to the extensive literature on the growth of Ge on
Si (001). This is a system that is in principle easier to understand than the RE silicide /Si
(001) system since the growth is pseudomorphic, with a mismatch (4%) that is similar to
that between tetragonal RE silicides and silicon [128, 129]. Nevertheless, the behavior of
this system is very complex, with three different forms of islands formed (hut clusters,
pyramids, and domes), and it is only after many years of effort that a basic understanding

of the growth in terms of strain relaxation has been achieved.

100

6.4 Conclusions

The growth of Dy, Gd silicide nanostructures is a function of growth temperature,
annealing duration, and metal coverage. At temperatures lower than 600°C, metal
accommodation in 2D reconstructions is preferable, and at higher temperatures, metal is
lost from the surface. Longer annealing duration promotes the formation of 3D islands.
The proportions of 2D reconstructions, nanowires and 3D islands in the surface are also
related to metal coverage. The Gd is accommodated in 2D phases up to a critical
coverage, 0.357 ML, and then nanowires start to nucleate. 3D islands form for both Gd
and Dy at higher coverage. Gd grows nanowires without 3D islands over a wider range of
conditions than Dy. The differences in behavior between the metals can be correlated
with the areal mismatch of the surface lattice unit cells which is a measure of biaxial
strain. Both the nanowires and islands grow layer-by-layer with quantized step heights
which are consistent with the bulk RE silicide crystallographic structures by assigning
nanowires to the hexagonal phase, and 3D islands to the orthorhombic or tetragonal
phases. The RE silicide growth on the vicinal Si substrate shows similar kinetics as on the
normal Si substrate. In order to grow a single domain, parallel arrays of nanowires on the
vicinal surface, it is crucial to suppress the growth of 3D islands which are accompanied

by significant Si redistribution and step redistribution.

101

Chapter 7 Future work

As discussed in the introduction, the idea of using anisotropic lattice mismatch
systems to produce nanowires was just realized two years ago, and it seems that RE
silicides/Si (001) are the only systems under study. Generally, people are more familiar
with biaxially lattice mismatched systems, and most of the knowledge comes from the
extensive and intensive studies of heteroepitaxial growth of semiconductor thin films and
quantum dots. Essentially, most of the phenomena during heteroepitaxial growth can be
related to strain relaxation in these systems, which originates from the biaxial lattice
mismatch. In this chapter, a brief review is given of the effect of lattice mismatch and
strain relaxation on epitaxy, and some possible nanowire forming systems are pointed
out. More research work is needed to better understand the strain related effects in the
growth of RE silicides, and to find other systems in which uniaxial strain results in the

formation of nanowires.

7.1 Epitaxy

Epitaxial growth is a key process in modern microelectronics technology and has
been studied extensively and intensely to pursue new classes of artificially structured
materials to produce novel devices [109, 130, 131].

The term epitaxy is derived from the Greek words epi (meaning "on") and taxis
(meaning "arrangement") [132]. The epitaxial growth refers to the deposition of a thin
layer of material onto the surface of a single crystal substrate in such a manner that the
layer is also a single crystal and has a fixed and predetermined crystal orientation with

respect to the substrate. If the material grown is the same as the substrate such as silicon

102

on silicon, then it has the same crystal orientation as that of substrate. This type of
epitaxy growth is called homoepitaxy. If the materials are different, such as Ge on Si, this
is called heteroepitaxy. Obviously, the growth of silicides on silicon substrates belongs to
the second case.

It is generally accepted that the epitaxial growth proceeds by three growth modes at
thermodynamic equilibrium: layer-by-layer [Frank-van der Merwe (FM)], island
[Volmer-Weber (VW)], and layer-by-layer growth followed by 3D islands [Stranski-
Krastanov (8K)] growth modes [129, 130, 133, 134].

In the lattice matched systems, the growth mode is governed by the interface and
surface energies only. If the sum of the surface free energy of the epitaxial film and the
free energy of the film/substrate interface is smaller than the surface free energy of the
substrate, layer-by-layer mode occurs. The opposite case results in island growth mode.
In the intermediate case, Stranski—Krastanov mode occurs. The situation is further
complicated if lattice mismatch is present between film and substrate. Figure 7.5 is an
equilibrium phase diagram of a lattice-mismatched heteroepitaxial system [135], showing
that the FM mode only occurs at small lattice mismatch (up to 10%), but at large lattice
mismatch, SK mode and VW mode are favored. Transition from FM mode to SK mode
could happen at a certain thickness for a given mismatch due to the increasing strain
energy with thickness during growth.

In practice, epitaxial growth (or deposition) is a non-equilibrium process determined
by kinetic phenomena, growth morphologies will deviate more or less from the
thermodynamic picture [130, 134]. The growth mode strongly depends on the detailed

growth conditions, and also on the combination of film and substrate. For semiconductor

103

heteroepitaxy, for example, GeSi/Si [136] and InGaAS/GaAs [137, 138], three
dimensional island growth is favored over the layer-by-layer growth mode when the

lattice mismatch is larger than 2% [111].

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5 .

4 >-

3 >—

I

2 i.

1 a

O I A

0.0 0.1 0.2

8
FM SK VW
”fl (IRE [31 /\

 

 

 

 

 

 

 

Figure 7.1 Equilibrium phase diagram in function of the coverage H and misfit a. The
small panels on the top and the bottom illustrate the morphology of the surface.

7.2 Lattice Mismatch
As we can see, lattice mismatch plays a central role in determining the final overlayer
morphology. A brief discussion on lattice mismatch would be helpful in finding new

systems to form nanowires.

104

Zur and McGill [88]suggested a criterion for lattice match in the heteroepitaxial
system that the interface translational symmetry could be compatible with the symmetry
on both sides of the interface within a given precision. A pair of superlattices with similar
translational symmetry across the interface can be defined instead of a common
superlattice (or coincidence lattice) [108, 122] due to the existence of mismatch. The
mismatch between the two superlattices and the unit cell area of these superlattices are
two parameters that characterize the lattice match. Following this criterion, a number of
ways can be defined to match the two lattices, but actual orientation will be determined
by the chemical bonding at the interface [109]. The simplest case for heteroepitaxy is that
the overlayer and substrate have similar matching lattices and similar lattice parameters
(mismatch <10%), then the matching between the two cubic lattices can be described in

Figure 7.2(a). In this case, the mismatch can be defined as f = (a0 — a, )/ as (note that this
is only one type of definition commonly in use). For high mismatch systems (>10%), a
coincidence lattice can be defined with lattice parameter na0 or mas , as shown in Figure

7.2(b), where m/ n is the closest integer ratio of a0 / ax. In this case, the mismatch is

defined as f = (nan - "160/ max [108, 122]. Sometimes, a rotation between the two
lattices is involved to match them together, as shown in Figure 7.2(c). Obviously, the
lattice matching between hexagonal RE disilicide (0110) and Si (001) substrate belong to
the first case. And that between cubic C3F2 (110) with Si (001) belong to the second case
since it has a 45% mismatch along one direction, which amounts to -4.64% if we use the

mismatch definition for coincidence lattice.

105

 

 

 

 

do

 

 

 

(a)

 

 

 

 

_ (b)

 

200 2 3a, \

 

(L
M
\Q
Q
Q
/\
\

 

 

 

 

 

 

(C) (d)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 7.2 Lattice matching in heteroepitaxial systems, where as denotes substrate lattice
parameter, and (:0 denotes the overlayer lattice parameter. The overlayer lattice is the
shaded area.

   

(a) Pseudomorphic growth (b) Semicoherent film

   

(c) Dislocated island ((1) Coherent island

Figure 7.3 Registry between overlayer and substrate.

106

7.3 Strain relaxation

In the early stages of epitaxial growth, the overlayer will strain elastically to match
the substrate lattice coherently at the interface as a wetting layer, which is usually called
“pseudomorphic” growth [139], as shown in Figure 7.3(a). Up to a critical thickness,
strain built up during growth will be relieved by forming misfit dislocations in the
interface, resulting in a semicoherent film [108, 140, 141], as shown in (b), or by a
transition from the layer-by-layer growth mode to SK mode by forming islands on top of
the wetting layer [142]. The islands could be unstrained with dislocation at the interface
[143, 144], which is called ‘dislocated islands’, as shown in (c), or could be strained
elastically to keep in registry with wetting layer lattice, which is called ‘coherent islands’
[129, 134], as shown in (d). It is generally stated that the islands formed in the VW
system are incoherent or dislocated [143]. Only recently, it was reported that coherent
islands can be formed in Gd/Si (l l 1) system which proceed by VW growth mode [145].

Misfit dislocation is just one way for strain relaxation. The strain caused by lattice
mismatch can also be relieved by surface roughening and other defects resulting from the
interaction between misfit dislocation and surface roughening during heteroepitaxial
growth [110, 131, 146]. All of these mechanisms are closely related to the growth
modes, determining the final morphologies [111, 146]. For systems with small mismatch
(<2%), growth proceeds by layer-by-layer mode and ‘cross-hatch’ morphology usually
forms [111, 147]. For systems with larger mismatch, growth proceeds by coherent or
incoherent islanding (SK or VW mode), resulting in a series of strain-relief
morphologies[110]. Theory and experiments show that systems with tensile strain could

be more resistant to roughening [133, 148, 149].

107

Strain-induced roughening can be problematic in the fabrication of coherently
strained device structures, but on the other hand, it is beneficial for the fabrication of
quantum dots [150], for example, GeSi quantum dots on Si (001) [128, 129, 151, 152]
and InGaAs quantum dots on GaAs [153, 154]. The quantum dots could form ordering
arrays spontaneously, called self-organization, driven by long-range elastic interaction
[129,134,155].

The formation of quantum dots generally involves two square lattices which are
highly mismatched biaxially, thus the growth is limited along the two matching
crystallographic axes, forming square-shaped islands. There are circumstances that the
symmetry is broken then islands grow into elongated shapes, forming wire-like structure.
For example, the formation of Ge ‘hut’ shaped islands on Si (001) is due to surface
energy anisotropy [128, 156], and the formation of InAs/InP (001) quantum wires results
from anisotropic stress relaxation [157]. In these two cases, the geometric symmetry is
still reserved. If the geometric symmetry is destroyed, anisotropic strain will result,
leading to the formation of nanowires. There are two ways to break the geometric
symmetry, using low symmetry substrate, such as Ge nanowires on Si (113) [158], and
using low symmetry overlayer, such as Can nanowires on Si (001) [159, 160].
Obviously, the formation of RE disilicide nanowires on Si (001) under investigation

belongs to the latter case.

7.4 Possible systems to form nanowires

Out of scientific curiosity and technological interest, it is desirable to seek if there are

other substrate/overlayer combinations that have anisotropic mismatch and might form

108

nanowires of metals, insulators, and semiconductors. For nanowire forming RE
silicide/Si (001) systems, the mismatches range between —l.09% and 0.96% along the
good match direction, and between 6.5% and 8.6% along the poor match direction (-
4.58% for ScSi2/Si (001)). We can conclude here that a system might be used to form
nanowires if the mismatch between the two lattices fall in a similar range as the RE
silicide/Si (001) systems. One interesting system that would be a straightforward
extension of the present work is RE silicide/SiGe (001). As mentioned above, we can
manipulate the lattice parameters of SiGe alloy by changing the ratio between Si and Ge,
hence it is possible to make zero mismatch along the good match direction, but keep the
mismatch along the poor match direction still within the required range, for example,
GdSIz/Slojjceogs (0%, 7.61%), in which it is possible to manipulate the width and length
of the nanowires which are regulated by the according mismatch. This system can also be
used to manipulate the mismatch anisotropy to test the lattice mismatch model.

Our understanding on nanowire formation with systems of anisotropic superlattice
mismatch is very limited since we only have one example of Can /Si (001). The Can is
of face-centered cubic structure, having a lattice parameter 5.463A. Can matches onto Si
(001) surface with (3an (110) // Si (001) and (3an [110] parallel to one of the Si<110>
directions on Si surface, as shown in Figure 7.3(d). The mismatch is 0.6% along the
longitudinal direction, and -4.64% along the lateral direction of the nanowire. The Can
nanowire is about 5-10 nm wide, and typically 1pm long [159]. There are a number of
systems which have similar anisotropy as Can (110) // Si (001), as shown in Table 7.1,
which are possible to form nanowires. All of the metal silicides listed in the table belong

to the category of refractory metal (RM) silicides, as shown in Table 1.2.

109

— 41:1

Ire-f-

It is also possible to use lower symmetry substrates, such as Si (1 10) to form
nanowires. For example, if tetragonal GdSlg (001) can match onto the Si (110) surface
with GdSlg [100] // Si [001], then we have anisotropic mismatches of 0.68% along one
direction and 6.77% along the other direction. Therefore, nanowires could be formed

with this system.

7 .5 Future work

7.5.1 Growth of nanowires with other systems

As we can see, following the anisotropic lattice mismatch model, it is possible to find
new systems to form nanowires, such as the refractory metal (Nb, Ta, Mo and W)
silicides mentioned above, and other systems on lower symmetry substrates, such as
GdSiZ/Si (110). Although it is easy to form nanowires with RE/Si (001) systems, it is not
the best system due to their low thermal stability and high reactivity with oxygen. In
contrary, RM silicides are very stable against heating and oxygen [19]. Therefore, the
nanowires formed by RM silicides would be a good alternative to RE disilicide
nanowires. Although the transition metal silicides have been studied extensively, the
formation of nanowires is not observed possibly because attention has been concentrated
on continuous films, and the early stages of growth are not very fully studied. Similar to
RE/Si systems, RM/Si systems have rich phase diagrams with a variety of stoichiometries
(see Table 1.2). Growth conditions need to be manipulated to produce the correct phase

that might potentially form nanowires.

110

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m ~ .o voé 3.8 we.» v0.8 3.» 8w 2: 8w HE ESOOVEEBA _ 88

.2 .o 36 328 was Samm 38 SH 2: 8w HQ EQOGSNEOS: 508
z: E Hmmvmfi a

god 5.5 NO: was 00.: mm.w 1 m\\ I m H GSOQVEENCNGE.
SE a: 85:2me
_ 8a: 3 Reg

m: 2.0 G.: we.» #09: Ex m 1\\ mnz 1 _m:o€\\~_mn28o_ C
asttmapzzmg 1

age was Ga 2 Ga a Q; a Q; a

853.89% 9233.595 Eva—3:38 REESE mesa—a uni—832
35:25::
Scum—mm «Beam

 

 

:i A as a 5:, 82256 mm 2:8 to wages: 02:3 E 038

111

7.5.2 Strain relaxation in RE silicide/Si ( 001 ) systems

As discussed above, strain relaxation plays an important role in heteroepitaxial
growth. Generally, the heteroepitaxy of GeSi alloy on Si has been studied as a model
system for strain relaxation due to its technological importance, simplicity, and the fact
that the lattice mismatch can be tuned by changing the ratio between Ge and Si in the
film [128, 129]. The growth of RE silicides on Si (001) would also be a good system for
studying strain effect on heteroepitaxial growth since they form the same crystal
structures and the lattice mismatch varies from metal to metal. Most interestingly, the
mismatch could be biaxial or uniaxial, depending on the detailed silicide structure, which
provides new insight into the heteroepitaxy. Unfortunately, this issue is left unstudied for
RESlg/Sl (001) system for several reasons. Firstly, additional complications have to be
considered due to the difference in chemical properties between different RE metals, such
as valency and chemical bonding, compared to Ge/Si system. Secondly, attention is
focused on improvement of surface morphology and properties of RESIz/SI interfaces.
And most probably, the RE silicides have not been widely used in the semiconductor
industry.

In our experiments, many interesting phenomena have been found in the early stages
of RE (Dy, Gd, and Er) silicide growth on Si (001), e.g., transformation between 2x4
and 2x7 superstructures, nanowire bundling, wavy islands, elongated islands, all of which
can be related to strain relaxation effect. More experimental and theoretical work need to
be done to understand these phenomena. A good understanding of these strain related
issues in the RE silicide/Si (001) systems will not only contribute to the growth of low

dimensional materials, but also contribute to the theory of heteroepitaxy.

112

Another important issue is the difference in growth behavior of different RE metals,
which could be controlled by strain relaxation due to different lattice mismatch in
different systems. As an example, the SCSIz/SI (001) system is worthy to try since Sc
silicide has negative mismatch (tensile strained) along the poor mismatch direction
([10l0] ). opposite to that of Dy and Gd silicides which are compressive strained on Si
(001). As we know, the relaxation of compressive and tensile strained systems is
different. Therefore, it would be interesting to see what differences we can observe in the
growth of Sc silicide on Si (001).

7.5.3 TEM work

In this thesis, we have made different plausibility arguments for the assignment of the
hexagonal silicide phase to nanowires, and tetragonal or orthorhombic phase to large 3D
islands. Transmission electron microscopy (TEM) is needed to examine these structures
individually to test these assignments. Convergent beam diffraction has to be used since

most of the structures are very small.

   

.
.
“4'

a)

Figure 7.4 (a) Bright field image (1pmx0.9um) of a Dy/Si (001) sample at a metal
coverage of 3.4 ML. (b) The diffraction pattern from the same sample.

We have done a preliminary TEM study of the Dy/Si (001) surface. The sample was

prepared by gluing a Mo ring on the film side of 3.4 ML Dy deposited sample, then

113

gluing the Mo ring to a sapphire and thinning the sample from back side to 10 pm by
mechanical polishing and ion milling. H-800 TEM was used for observation. Plan view
bright field TEM images of DySim islands show diffraction/scattering contrast. The
morphology is the same as that seen by STM on the same sample, indicating that the
wires are stable after exposure to air. The lack of extra spots in the diffraction pattern
from the overlayer, along with the absence of Moire fringes in the images, indicate that
the islands are coherent with the substrate, at least in the direction along the long axis of
the elongated islands.

STM measurements shown in Chapter 4, Table 4.1 show that the top surface of
islands is relaxed to a certain degree with a dependence on thickness. The best
explanation would be that the islands grow coherently with the substrate at the interface,
but relaxed at the top, matching the model shown in Figure 7.3(c). Further experiments
with cross sectional TEM observation is also needed to characterize the interface

structure between islands (or nanowires) and Si substrate.

114

APPENDIX

List of Publications

1.

k)

Liu, 8.2. and J. Nogami, An STM study of Cu on Si(001) in the c(8x8) structure.
Surf. Sci., 2000. 453: p. 173.

Liu, B2. and J. Nogami, An STM study ofthe Si( 001 )( 2x4 )-Dy surface. Surf.
Sci., 2001. 488: p. 399-405.

Nogami, J., 32. Liu, M.V. Katkov, C. Ohbuchi, and NO. Birge, Self-assembled
rare-earth silicide nanowires on Si (001). Phys. Rev. B, 2001. 63: p. 233305.

Liu, B2. and J. Nogami, An STM study of Dysprosium silicide nanowire growth
on Si (001). J. Appl. Phys., 2003. 93(1): p. 593.

Liu, 8.2. and J. Nogami, Growth of parallel rare earth silicide nanowire arrays
on vicinal Si(001). in preparation.

Liu, 8.2. and J. Nogami, An STM study ofthe Si(001)(2X7)-Gd, Dy surface. in
preparation.

Liu, B2. and J. Nogami, A comparison of Dy and Gd silicide nanowire growth. in
preparation.

115

 

10.

11.

REFERENCES

Baglin, J .E., FM. d'Heurle, and CS. Petersson, The formation of silicides from
thin films of some rare-earth metals. Appl. Phys. Lett., 1980. 36(7): p. 594.

Tu, K.N., R.D. Thompson, and B.Y. Tsaur, low Schottky barrier of rare-earth
silicide on n-Si. Appl. Phys. Lett., 1981. 38(8): p. 626.

Knapp, J .A. and ST. Picraux, Epitaxial growth of rare-earth silicides on (111) Si.
Appl. Phys. Lett., 1986. 48(7): p. 466.

Preinesberger, C., S. Vandre’, T. Kalka, and M. Dahne-Prietsch, Formation of
dysprosium silicide wires on Si (001). J. Phys. D: Appl. Phys, 1998. 31: p. L43.

Markoff, J ., Computer scientists are poised for revolution on a tiny scale, in The
New York Times. November 1, 1999.

Chen, Y., D.A.A. Ohlberg, G. Medeiros-Ribeiro, Y.A. Chang, and RS. Williams,
Self-assembled growth of epitaxial erbium disilicide nanowires on silicon (001).
Appl. Phys. Lett., 2000. 76(26): p. 4004.

Nogami, J ., B.Z. Liu, M.V. Katkov, C. Ohbuchi, and NO. Birge, Self-assembled
rare-earth silicide nanowires on Si (001). Phys. Rev. B, 2001. 63: p. 233305.

Katkov, M. and J. Nogami, Y growth on Si(001). Bull. Am. Phys. Soc., March
meeting, 2002, 2002. 47(1): p. 283.

Liu, B2. and J. Nogami, Optimizing the growth of rare-earth silicide nanowires
on Si(001). Bull. Am. Phys. Soc., March meeting, 2002, 2002. 47(1): p. 169.

Chen, Y., D.A.A. Ohlberg, and RS. Williams, Nanowires of four epitaxial
hexagonal silicides grown on Si(001). J. Appl. Phys., 2002. 91(5): p. 3213.

Chen, Y., D.A.A. Ohlberg, and RS. Williams, Epitaxial growth of erbium silicide
nanowires on silicon( 001 ). Mater. Sci. Eng. B, 2001. 87: p. 222—226.

116

12.

l3.

14.

15.

16.

17.

18.

19.

20.

21.

22.

Preinesberger, C., S.K. Becker, S. Vandré, T. Kalka, and M. Dahne-Prietsch,
Structure of DySiz nanowires on Si(001). J. Appl. Phys., 2002. 91(3): p. 1695.

Liu, B.Z. and J. Nogami, An STM study ofthe Si(001)(2x4)-Dy surface. Surf.
Sci., 2001. 488: p. 399-405.

Liu, B.Z. and J. Nogami, An STM study of Dysprosium silicide nanowire growth
on Si (001). J. Appl. Phys., 2003. 93(1): p. 593.

Liu, B.Z. and J. Nogami, Growth ofparallel rare earth silicide nanowire arrays
on vicinal Si(001). in preparation.

Liu, B.Z. and J. Nogami, An STM study of the Si(001)(2X7)-Gd, Dy surface. in
preparation.

Liu, B.Z. and J. Nogami, A comparison of Dy and Gd silicide nanowire growth. J.
Appl. Phys., in preparation.

 

http://public.itrs.net/, International Technology Roadmap for Semiconductors
(2001).

Murarka, S.P., Silicides for VLSI Applications. 1983, New York: Academic Press,
Inc.

Maex, K., M. Van.Rossum, and A. Reader, Crystal structure of TM silicides, in
Properties of Metal Silicides, K. Maex and M. Van.Rossum, Editors. 1995,
INSPEC: London. United Kingdom. p. 3.

Xu, 2., Basic physical properties of RE silicides, in Properties of Metal Silicides,
K. Maex and M.V. Rossum, Editors. 1995, INSPEC: London. United Kingdom. p.
45.

Reader, A.H., A.H. van.Ommen, P.J.W. Weijs, R.A.M. Wolters, and DJ. Oostra,
Transition metal silicides in silicon technology. Rep. Prog. Phys., 1992. 56: p.
1397-1467.

117

 

23.

25.

27.

28.

29.

31.

32.

33.

Thompson, RD. and K.N. Tu, Comparison of the three classes( rare earth,
refractory and near-noble) of silicide contacts. Thin Solid Films, 1982. 93: p.
265-274.

Thompson, R.D., B.Y. Tsaur, and K.N. Tu, Contact reaction between Si and rare
earth metals. Appl. Phys. Lett., 1981. 38(7): p. 535.

Gambino, J .P. and EC. Colgan, Silicides and ohmic contacts. Materials
Chemistry and Physics, 1998. 52: p. 99-146.

Netzer. F.P., Rare earth over-layers on silicon. J. Phys.: Condens. Matter, 1995. 7:
p. 991.

Hedrick, J .B., Rare Earths. The American Ceramic Society Bulletin, 2000. 79(8):
p. 74.

Xu, 2., Schottky barrier heights of RE silicides on Si, in Properties of Metal
Silicides, K. Maex and M.V. Rossum, Editors. 1995, INSPEC: London, United
Kingdom. p. 169.

Derrien, J ., Schottky barrier heights of TM silicides on Si and GaAs, in Properties
of Metal Silicides, K. Maex and M.V. Rossum, Editors. 1995, INSPEC: London.
United Kingdom. p. 164.

Baglin, J .E.E., FM. d'Heurle, and CS. Petersson, Diffusion marker experiments
with rare-earth silicides and germanides: Relative mobilities of the two atom
species. J. Appl. Phys., 1981. 52(4): p. 2841-2846.

Anderson, R., J. Baglin, J. Dempsey, W. Hammer, F. d'Heurle, and CS.
Petersson, Nucleation-controlled thin-film interactions-some silicides. Appl. Phys.
Lett., 1979. 35(3): p. 285-287.

d'Heurle, FM. and P. Gas, Kinetics of formation of silicides: A review. J. Mater.
Res., 1986. 1(1): p. 205.

Houssay, E., A. Rouault, O. Thomas, R. Madar, and J .P. Senateur, Metallurgical
reinvestigation of rare earth silicides. Appl. Surf. Sci., 1989. 38: p. 156-161.

118

34.

35.

36.

37.

39.

40.

41.

42.

43.

Travlos, A., P. Aloupogiannis, E. Rokofyllou, and C. Papasta'ikoudis, Growth,
characterization and electrical properties of gadolinium silicide thin layers.
Philos. Mag. B, 1993. 67(4): p. 485-495.

Perri, J ., I. Binder, and B. Post, Rare earth metal "disilicides". J. Phys. Chem.,
l959.63:p.616.

Frangis, N., G. Van.Tendeloo, J .V. Landuyt, P. Muret, and T.T.A. Nguyen,
Structural characterization of erbium silicide thin films on an Si(I 11) substrate. J.
Alloys and compounds, 1996. 234: p. 244-250.

Siegal, M.P., F.H. Kaatz, W.R. Graham, J .J . Santiago, and J. VanderSpiegel,
Formation 0f Epitaxial Yttrium and Erbium Silicide 0n Si(I 1 I) in Ultrahigh-
Vacuum. Appl. Surf. Sci., 1989. 38: p. 162.

Knapp, J .A., S.T. Picraux, C.S. Wu, and SS. Lau, Kinetics and morphology of
erbium silicide formation. J. Appl. Phys., 1985. 58(10): p. 3747.

Frangis, N., J. Van.Landuyt, G. Kaltsas, A. Travlos, and AG. Nassiopoulos,
Growth of erbium-silicide films on (100) silicon as characterised by electron
microscopy and diffusion. J. Cryst. Growth, 1997. 172: p. 175-182.

Xu, 2., Epitaxial growth of RE silicides, in Properties of Metal Silicides, K. Maex
and M.V. Rossum, Editors. 1995, INSPEC: London, United Kingdom. p. 78.

Thomas, 0., CS. Petersson, and FM. d'Heurle, The Reaction of Scandium Thin-
Films with Silicon-diffusion, Nucleation, Resistivities. Appl. Surf. Sci., 1991. 53:
p. 138-146.

Vantomme, A., M.F. Wu, S. Hogg, U. Wahl, W. Deweerd, H. Pattyn, G.
Langouche, S. J in, and H. Bender, Stabilisation and phase transformation of
hexagonal rare-earth silicides on Si (111). Nucl. Instrum. Methods in Phys. Res.
B, 1999. 147: p. 261-266.

Clevenger, L.A. and R.W. Mann, Formation of epitaxial TM silicides, in
Properties of Metal Silicides, K. Maex and M.V. Rossum, Editors. 1995,
INSPEC: London, United Kingdom. p. 61.

119

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

Chen, L]. and K.N. Tu, Epitaxial-growth of Transition-metal Silicides on Silicon.
Mater. Sci. Rep., 1991. 6(2-3): p. 53-140.

Tung, RT. and S. Ohmi, Epitaxial silicide interfaces in microelectronics. Thin
Solid Films, 2000. 369: p. 233-239.

Mantl, S. and BL. Bay, New method for epitaxial heterostructure layer growth.
Appl. Phys. Lett., 1992. 61(3): p. 267.

Hensel, J .C., A.F.J. Levi, R.T. Tung, and J .M. Gibson, Transistor Action In
Si/CoSiz/Si Heterostructures. Appl. Phys. Lett., 1985. 47: p. 151.

Tung, R.T., Epitaxial CoSiz And NiSiz Thin-Films. Mater. Chem. Phys., 1992.
32((2)): p. 107-133.

Tung, R.T., J .M. Poate, J .C. Bean, J .M. Gibson, and DC. Jacobson, Epitaxial
Silicides. Thin Solid Films, 1982. 93((1-2)): p. 77-90.

Yu, B.D., Y. Miyamoto, O. Sugino, A. Sakai, T. Sasaki, and T. Ohno, Structural
and Electronic Properties of Metal-Silicide/Silicon Interfaces: A F irst-Principles
Study. J. Vac. Sci. Technol. B, 2001. 19(4): p. 1180-1185.

Xu, Z., Thin film formation of RE silicides, in Properties of Metal Silicides, K.
Maex and M.V. Rossum, Editors. 1995, INSPEC: London, United Kingdom. p.
73.

Lau, 3.8., OS. Pai, C.S. Wu, T.F. Kuech, and B.X. Liu, Surface-Morphology 0f
Erbium Silicide. Appl. Phys. Lett., 1982. 41: p. 77.

Wu, CS, 8.8. Lau, T.F. Kuech, and B.X. Liu, Surface-Morphology and
Electronic-Properties of ErSiz. Thin Solid Films, 1983. 104: p. 175.

Polop, C., C. Rogero, J .L. Sacedén, and J .A. Martin-Gago, Surface morphology of
yttrium silicides epitaxially grown on Si(I I I) by S TM. Surf. Sci., 2001. 482-485:
p. 1337-1342.

120

 

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

Vandré, S., T. Kalka, C. Preinesberger, and M. D'ahne-Prietsch, Epitaxial growth

and electronic structure of lanthanide silicides an n-type Si(I I I). J. Vac. Sci.
Technol. B, 1999. 17(4): p. 1682.

Shen, G.H., J .C. Chen, C.H. Lou, S.L. Cheng, and L.J. Chen, The growth of
pinhole-free epitaxial DySi2-x films on atomically clean Si(I I I). J. Appl. Phys.,
1998. 84(7): p. 3630.

Kaltsas, G., A. Travlos, A.G. Nassiopoulos, N. Frangis, and J. Van.Landuyt, High
crystalline quality erbium silicidefilms on (100) silicon, grown in high vacuum.
Appl. Surf. Sci., 1996. 102: p. 151-155.

Siegal, M.P., W.R. Graham, and J .J . Santiago-Aviles, Growth of pinhole-free
epitaxial yttrium silicide on Si( 11 I ). J. Appl. Phys., 1990. 68(2): p. 574.

Bennett, PA. and H. vonK'anel, Scanning tunneling microscopy studies of
silicides. J. Phys. D: Appl. Phys., 1999. 32: p. R71.

Hsu, C.C., Y.X. Wang, J. Hu, J. Ho, and J .J . Qian, The effect of thermal treatment
on the thin-film reaction of La, Ce, and Nd on silicon surfaces. J. Vac. Sci.
Technol. A, 1989. 7(5): p. 3016.

Travlos, A., N. Salamouras, and N. Boukos, Epitaxial dysprosium silicide films on
silicon: growth, structure and electrical properties. Thin Solid Films, 2001. 397:
p. 138-142.

Kaatz, F.H., W.R. Graham, and J. VanderSpiegel, Modification of the
microstructure in epitaxial erbium silicide. Appl. Phys. Lett., 1993. 62(15): p.
1748.

Shen, G.H., J .C. Chen, and L.J. Chen, The structures and variation of vacancy
ordering in epitaxial DySi2-,, thin films on (I I I)Si. Appl. Surf. Sci., 1999. 142: p.
300-304.

Luo, C.H., G.H. Shen, and L.J. Chen, Vacancy ordering structures in epitaxial
RESi2-x thinfilms on (III)Si and (001)Si. Appl. Surf. Sci., 1997. 113/114: p. 457.

Gurvitch, M., A.F.J. Levi, R.T. Tung, and N. S., Epitaxial yttrium silicide on
(I II) silicon by vacuum annealing. Appl. Phys. Lett., 1987. 51(5): p. 311.

121

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

Travlos, A., N. Salamouras, and N. Boukos, Growth, structure and electrical
properties of epitaxial thulium silicide thin films on silicon. J. Appl. Phys., 1997.
81(3): p. 1217.

Molnar, G.L., G. Peto, E. Zsoldos, N.Q. Khanh, and Z.E. Horvath, Amorphous
alloy formation and thickness dependent growth of Gd-silicide in solid phase thin
film reaction. Thin Solid Films, 1998. 317: p. 417-420.

Molnar, G., I. Gerocs, G. Peto, E. Zsoldos, E. Jaroli, and J. Gyulai, Thickness-
dependent formation of Gd-silicide compounds. J. Appl. Phys., 1988. 64(12): p.
6746.

Wetzel, P., S. Saintenoy, C. Pirri, D. Bolmont, G. Gewinner, T.P. Roge, F.
Palmino, C. Savall, and J .C. Labrune, STM investigation of 2- and 3-dimensional
Er disilicide grown epitaxially on Si(III). Surf. Sci., 1996. 355: p. 13-20.

Roge, T.P., F. Palmino, C. Savall, J .C. Labrune, S. Saintenoy, P. Wetzel, C. Pirri,
D. Bolmont, and G. Gewinner, Initial growth mode of Er silicide on Si(I 11) by
solid phase epitaxy. Surf. Sci., 1996. 352-354: p. 622-627.

d’Avitaya, F.A., A. Perio, J .-C. Oberlin, Y. Campidelli, and J .A. Chroboczek,
Fabrication and structure of epitaxial Er silicide films on (I 1 I ) Si. Appl. Phys.
Lett., 1989. 54(22): p. 2198.

Wetzel, P., S. Saintenoy, C. Pirri, D. Bolmont, and G. Gewinner, Surface states
and reconstruction of epitaxial root 3 x root 3R30 Er silicide on Si( I I I ). Phys.
Rev. B, 1994. 50(15): p. 10886.

Lohmeier, M., W.J. Huisman, G.t. Horst, P.M. Zagwijn, E. Vlieg, C.L. Nicklin,
and TS. Turner, Atomic structure and thermal stability of two-dimensional Er
silicide on Si(III). Phys. Rev. B, 1996. 54: p. 2004.

Spence, D.J., S.P. Tear, T.C.Q. Noakes, and P. Bailey, Medium-energy ion
scattering studies of two-dimensional rare-earth silicides. Phys. Rev. B, 2000. 61:
p.5707.

Baptist, R., S. Ferrer, G. Grenet, and HG Poon, Surface Crystallography of YSi2-
x Films Epitaxially Grown on Si(I I I): An X—Ray Photoelectron Diffraction Study.
Phys. Rev. Lett., 1990. 64(3): p. 311.

122

 

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

Martin-Gago, J .A., J.M. Gomez-Rodriguez, and J .Y. Veuillen, Growth and
morphology of epitaxial ErSi 1.7 films on Si(I I I)7x 7 studied by scanning
tunneling microscopy. Surf. Sci., 1996. 366: p. 491-500.

Roge, T.P., F. Palmino, C. Savall, J .C. Labrune, P. Wetzel, C. Pirri, and G.
Gewinner, Surface reconstruction of ErSi I . 7( 0001 ) investigated by scanning
tunneling microscopy. Phys. Rev. B, 1995. 51(16): p. 10998.

Tuilier, M.-H., P. Wetzel, C. Pirri, D. Bolmont, and G. Gewinner, Interfacial
structure of two-dimensional epitaxial Er silicide on Si(I I 1). Phys. Rev. B, 1994.
50:p.2333.

Martin-Gago, J.A. and J .M. Gomez-Rodriguez, Surface atomic structure of
epitaxially grown erbium silicidefilms on Si(III)7x 7. Phys. Rev. B, 1997. 55(8):
p.5136.

Lee, Y.K., N. Fujimura, and T. Ito, Epitaxial-Growth of Yttrium Silicide YSi2.x on
(100) Si. J. Alloys and Compounds, 1993. 193(1-2): p. 289-291.

Lee, Y.K., N. Fujimura, T. Ito, and N. Itoh, Epitaxial-Growth and Structural
Characterization of Erbium Silicide Formed on (100)Si Through a Solid-Phase
Reaction. J. Cryst. Growth, 1993. 134(3-4): p. 247-254.

Gerocs, 1., G. Molnar, E. Jaroli, E. Zsoldos, G. Peto, and J. Gyulai, Epitaxy of
orthorhombic gadolinium disilicide on <100> silicon. Appl. Phys. Lett., 1987.
51(25): p. 2144.

Molnar, G., G. Peto, E. Zsoldos, Z.E. Horvath, and NO. Khanh, The efi‘ect of
silicon substrate orientation on the formation of Gd-silicide phases. Appl. Surf.
Sci., 1996. 102: p. 159-162.

Frangis, N., G. Van.Tendeloo, J. Van.Landuyt, G. Kaltsas, A. Travlos, and A.G.
Nassiopoulos, New Erbium Silicide Superstructures: A Study by High Resolution
Electron Microscopy. phys. stat. sol. (a), 1996. 158: p. 107.

Travlos, A., N. Salamouras, and E. Flouda, Epitaxial erbium silicide films on
(100) silicon: growth, structure and electrical properties. Appl. Surf. Sci., 1997.
120: p. 355-364.

123

86.

87.

88.

89.

90.

91.

92.

93.

94.

95.

96.

Henle, W.A., M.G. Ramsey, and RP. Netzer. The gadolinium-Si(100)2x1
interface. Surf. Sci., 1991. 243: p. 141-150.

Pasquali, L., S. D’Addato, and S. Nannarone, Growth of epitaxial Yb silicide on
Si( 100) studied by metastable atom deexcitation spectroscopy and photoemission.
Phys. Rev. B, 2002. 65: p. 115417.

Zur, A. and TC. McGill, Lattice match: An application to heteroepitaxy. J. Appl.
Phys., 1984. 55(2): p. 378.

Gottlieb, U., F. Nava, M. Affronte, O. Laborde, and R. Madar, Electrical
transport in metallic TM silicides, in Properties of Metal Silicides, K. Maex and
M.V. Rossum, Editors. 1995, INSPEC: London. United Kingdom. p. 189.

Gottlieb, U., F. Nava, M. Affronte, O. Laborde, and R. Madar, Electrical
transport in semiconducting TM silicides, in Properties of Metal Silicides, K.
Maex and M.V. Rossum, Editors. 1995, INSPEC: London. United Kingdom. p.
205.

d'Heurle, F.M., Silicide Interfaces in Silicon Technology. Journal of Electronic
Materials, 1998. 27(11): p. 1138.

Unewisse, M.H. and J .W.V. Storey, Electrical and infrared investigation of
erbium silicide. J. Appl. Phys., 1992. 72(6): p. 2367.

Xu, 2., Electrical transport in RE silicides, in Properties of Metal Silicides, K.
Maex and M.V. Rossum, Editors. 1995, INSPEC: London, United Kingdom. p.
217.

Guizzetti, G., E. Mazzega, M. Michelini, F. Nava, A. Borghesi, and A. Piaggi,
Electrical and optical characterization of GdSiz and ErSiz alloy thin films. J.
Appl. Phys., 1990. 67(7): p. 3393.

Duboz, J .Y., P.A. Badoz, F.A. d'Avitaya, and J .A. Chroboczek, Electronic
transport properties of epitaxial erbium silicide/silicon heterostructures. Appl.
Phys. Lett., 1989. 55(1): p. 84.

Woodruff, DP. and TA. Delchar, Modern Techniques of Surface Science. 2nd ed.
1994: Cambridge University Press.

124

97.

98.

99.

100.

101.

102.

104.

105.

106.

107.

Kubby, J .A. and J .J . Boland, Scanning tunneling microscopy of semiconductor
surfaces. Surf. Sci. Rep., 1996.26: p. 61-204.

Appelbaum, J .A., G.A. Baraff, and DR. Hamann, Si( 100) Surface
Reconstruction: Spectroscopic Selection of a Structural Model. Phys. Rev. Lett.,
1975. 35(729).

Becker, R. and R. Wolkow, Semiconductor surfaces, in Scanning tunneling
microscopy, J .A. Stroscio and W.J. Kaiser, Editors. 1993, Academic Press: San
Diego. p. 149.

Zhang, 2., F. Wu, and M.G. Legally, An atomistic view ofSi(001) homoepitaxy.
Ann. Rev. Mater. Sci., 1997.27: p. 525.

Chadi, D.J., Stabilities of single-layer and bilayer steps on Si(001) surfaces. Phys.
Rev. Lett., 1987. 59: p. 1691.

Stroscio, J .A. and W.J. Kaiser, eds. Scanning Tunneling Microscopy. Methods of
Experimental Physics. Vol. 27. 1993, Academic Press, Inc.: San Diego.

Vickerman, J .C., ed. Surface Analysis - The Principal Techniques. 1997, John
Wiley & Sons Ltd.

Baski, A.A., Scanning Tunneling Microscopy of Metal Growth and
Reconstruction on Si( 100) and Si(] 1 I ), in Edward L. Ginzton Laboratory. 1991,
Stanford University: Stanford, California.

http://reg.ssci.liv.ac.uk/.

 

Magaud, J ., A. Pasture], G. Kresse, and J. Hafner, Role of silicon vacancies in
yttrium-disilicide compounds from ab-initio calculations. Phys. Rev. B, 1997.
55(20): p. 13479. '

Chen, G., J. Wan, J. Yang, X. Ding, L. Ye, and X. Wang, Surface structures of
erbium silicide ultra-thin films formed by solid phase epitaxy on Si(100). Surf.
Sci., 2002. 513(1): p. 203.

125

108.

109.

110.

111.

113.

114.

115.

116.

117.

118.

119.

Matthews, J .W., Misfit Dislocations, in Dislocations in Solids, F .R.N . Nabarro,
Editor. 1979, North-Holland Publishing Company. p. 463.

Palmstrom, C.J., Epitaxy ofDissimilar Materials. Ann. Rev. Mater. Sci., 1995.
25: p. 389-415.

Gao, H. and W.D. Nix, Surface roughening of heteroepitaxial thin films. Ann.
Rev. Mater. Sci., 1999.29: p. 173-209.

Andrews, A.M. and J .S. Speck, Modeling cross-hatch surface morphology in
growing mismatched layers. J. Appl. Phys., 2002. 91(4): p. 1933.

Chang, K.H., R. Gibala, D.J. Srolovitz, P.K. Bhattacharya, and J .F. Mansfield,
C rosshatched surface morphology in strained III-V semiconductor films. J. Appl.
Phys., 1990. 67(9): p. 4093.

Lu, W. and Z. Suo, Symmetry breaking in self-assembled monolayers on solid
surfaces: Anisotropic surface stress. Phys. Rev. B, 2002. 65: p. 85401.

Suo, Z. and W. Lu, Forces that drive nanoscale self-assembly on solid surfaces.
Journal of Nanoparticle Research, 2000. 2: p. 333-344.

Baglin, J .E.E., FM. d'Heurle, and CS. Petersson, Diffusion marker experiments
with rare-earth silicides and germanides: Relative mobilities of the two atom
species. J. Appl. Phys., 1981. 52(4): p. 2841.

Lipson, H. and KB. Singer, Disorder in a film of gold deposited on silicon:
investigation by low-energy electron diffraction. J. Phys. C: Solid State Phys.,
1974.7:p.l2.

Ohbuchi, C. and J. Nogami, Holmium growth on Si( 001 ): surface reconstructions
and nanowire formation. Phy. Rev. B, 2002. 66: p. 233305.

Wetzel, P., C. Pirri, P. Paki, D. Bolmont, and G. Gewinner, Structure of a two-

dimensional epitaxial Er silicide on Si(] 1 1) investigated by Auger-electron
diflraction. Phys. Rev. B, 1993. 47: p. 3677.

Katkov, M.V. and J. Nogami, Yb and Nd growth on Si(001). Surf. Sci., 2002.

126

121.

122.

123.

124.

125.

126.

127.

128.

129.

130.

Heath, J .R., P]. Kuekes, G.S. Snider, and RS. Williams, A defect-tolerant
computer architecture: opportunities for nanotechnology. Science, 1998. 280: p.
1716.

Molnar, G., G. Peto, and E. Zsoldos, Solid phase reaction and epitaxy ode-
silicide films on Si substrates. Appl. Surf. Sci., 1993. 70/71: p. 466.

Trampert, A. and K.H. Ploog, Heteroepitaxy of large-misfit systems: role of
coincidence lattice. Cryst. Res. Technol., 2000. 35(6-7): p. 793-806.

Zur, A., T.C. McGill, and M.-A. Nicolet, Transition-metal silicides lattice-
matched to silicon. J. Appl. Phys., 1985. 57(2): p. 600.

Tersoff, J. and RM. Tromp, Shape transition in growth of strained islands -
spontaneous formation of quantum wires. Phys. Rev. Lett., 1993. 70: p. 2782.

Grinfeld, M.A., The stress driven "rearrangement" instability in crystalline films.
J. Intell. Mater. Syst. Struct., 1993. 4(76).

Grilhe, J ., Study of roughness formation induced by homogeneous stress at the
free surfaces of solids. Acta metall. mater., 1993. 41(3): p. 909.

Cullis, A.G., Strain-induced modulations in the surface morphology of
heteroepitaxial layers. MRS Bulletin, 1996. April: p. 21.

Voigtl'ander, B., Fundamental process in Si/Si and Ge/Si epitaxy studied by
scanning tunneling microscopy during growth. Surf. Sci. Rep., 2001. 43: p. 127.

Teichert, C., Self-organization of nanostructures in semiconductor heteroepitaxy.
Physics Reports, 2002. 365: p. 335-432.

Bauer, E.G., et al., Fundamental issues in heteroepitaxy-A Department of Energy,
Council on Materials Science Panel Report. J. Mater. Res., 1990. 5(4): p. 852.

Schowalter, L.J., Heteroepitaxy and Strain: Applications to Electronic and
Optoelectronic Materials. MRS Bulletin, 1996(April): p. 18.

127

132.

133.

134.

135.

136.

137.

138.

139.

140.

141.

142.

143.

$26, S.M., Physics of semiconductor devices. 1981, New York: John Wiley &
Sons.

Markov, I., Recent theoretical developments in epitaxy. Mater. Chem. Phys.,
1993.36: p. 1-30.

Shchukin, V.A. and D. Bimberg, Spontaneous ordering of nanostructures on
crystal surfaces. Rev. Mod. Phys., 1999. 71(4): p. 1125.

Daruka, I. and A.-L. Barabasi, Dislocation-free island formation in
heteroepitaxial growth: a study at equilibrium. Phys. Rev. Lett., 1997. 79(19): p.
3708.

Bean, J .C., L.C. Feldman, A.T. Fiory, S. Nakahara, and I.K. Robinson, Ge,,.Si 1-,/Si
strained-layer superlattice grown by molecular beam epitaxy. J. Vac. Sci.
Technol. A, 1984. 2: p. 436.

Snyder, C.W., B.G. Orr, D. Kessler, and L.M. Sander, Effect of strain on surface
morphology in highly strained InGaAs films. Phys. Rev. Lett., 1991. 66: p. 3032.

Leonard, D., K. Pond, and PM. Petroff, Critical layer thickness for self-
assembled InAs islands on GaAs. Phys. Rev. B, 1994. 50: p. 11687.

Pashley, D.W., The nucleation, growth, structure and epitaxy of thin surface
films. Advances in physics, 1965. 14: p. 327.

vanderMerwe, J .H. and CA. Ball, in Epitaxial growth, J .W. Matthews, Editor.
1975, Academic: New York. p. 493.

Fitzgerald, E.A., Dislocations in strained-layer epitaxy - theory, experiment, and
applications. Mater. Sci. Rep., 1991. 7(3): p. 91.

Politi, P., G. Grenet, A. Marty, A. Ponchet, and J. Villain, Instabilities in crystal
growth by atomic or molecular beams. Phys. Rep., 2000. 324: p. 271-404.

Kern, R., G. LeLay, and J .J . Metois, in Current topics in materials science, 1201.3,
E. Kaldis, Editor. 1979, North-Holland: Amsterdam. p. 128-419.

128

144.

145.

146.

147.

148.

149.

150.

151.

152.

153.

Markov, I. and S. Stoyanov, Mechanisms ofepitaxial-growth. Contemp. Phys.,
l987.28:p.267.

Raviswaran, A., C.P. Liu, J. Kim, D.G. Cahill, and J.M. Gibson, Evolution of
coherent islands during strained-layer Volmer- Weber growth of Si on Ge( 1 I 1 ).
Phys. Rev. B, 2001. 63(12): p. 125314.

LeGoues, F .K., The eflect of strain on the formation of dislocations at the SiGe/Si
interface. MRS Bulletin, 1996. April: p. 38.

Lutz, M.A., R.M. Feenstra, F.K. LeGoues, P.M. Mooney, and J .O. Chu, Influence
of misfit dislocations on the surface morphology of Si ”Ge, films. Appl. Phys.
Lett., 1995. 66(6): p. 724.

Guyer, J.E. and PW. Voorhees, Morphological stability of alloy thin films. Phys.
Rev. Lett., 1995. 74: p. 4031.

Xie, Y.H., et al., Semiconductor surface roughness: dependence on sign and
magnitude of bulk strain. Phys. Rev. Lett., 1994. 73: p. 3006.

Jesson, D.E., K.M. Chen, and SJ. Pennycook, Kinetic pathways to strain
relaxation in the Si-Ge system. MRS Bulletin, 1996. April: p. 31.

Eaglesham, DJ. and M. Cerullo, Dislocation-free Stranski-Krastanow growth of
Ge on Si(100). Phys. Rev. Lett., 1990. 64(16): p. 1943.

Koch, R., G. Welder, and B. Wassermann, Evolution of stress and strain
relaxation of Ge and SiGe alloy films on Si(001). Appl. Surf. Sci., 2002. 190(1—4):
p. 422.

Tillmann, K. and A. Forster, Critical dimensions for the formation of interfacial
misfit dislocations of Ino_6Ga0,4As islands on GaAs(001). Thin Solid Films, 2000.
368(1): p.93.

Ramachandran, TR, A. Madhukar, I. Mukhametzhanov, R. Heitz, A. Kalburge,
Q. Xie, and P. Chen, Nature of Stranski-Krastanow growth of InAs on GaAs(001).
J. Vac. Sci. Technol. B, 1998. 16(3): p. 1330.

129

155.

156.

157.

158.

159.

160.

N6tzel, R., Self-organized growth of quantum-dot structures. Semicond. Sci.
Technol., 1996. 11: p. 1365-1379.

Zhang, Y.W., Self-organization, shape transition, and stability of epitaxially
strained islands. Phys. Rev. B, 2000. 61(3): p. 10388.

Garcia, J .M., L. Gonzalez, M.U. Gonzalez, J .P. Silveira, Y. Gonzalez, and F.
Briones, InAs/InP(001) quantum wire formation due to anisotropic stress
relaxation: in situ stress measurements. J. Cryst. Growth, 2001. 227-228: p. 975.

Omi, H. and T. Ogino, Self-assembled Ge nanowires grown on Si(113). Appl.
Phys. Lett., 1997. 71(15): p. 2163.

Loretto, D., F.M. Ross, and CA. Lucas, Quasi-one-dimensional Can islands
formed on Si(001) by molecular beam epitaxy. Appl. Phys. Lett., 1996. 68(17): p.
2363.

Pasquali, L., S. D’Addato, G. Selvaggi, S. Nannarone, N.S. Sokolov, S.M.
Suturin, and H. Zogg, Formation of Can nanostructures on Si(001).
Nanotechnology, 2001. 12: p. 403.

130