Citation
Electron scattering studies of surface phonon-plasmon modes of semiconductors

Material Information

Title:
Electron scattering studies of surface phonon-plasmon modes of semiconductors
Creator:
Seo, Jae Myung, 1955-
Publication Date:
Language:
English
Physical Description:
xiv, 269 leaves : ill. ; 28 cm.

Subjects

Subjects / Keywords:
Alloys ( jstor )
Annealing ( jstor )
Atoms ( jstor )
Electrons ( jstor )
Energy ( jstor )
Hydrogen ( jstor )
Oxides ( jstor )
Oxygen ( jstor )
Phonons ( jstor )
Plasmons ( jstor )
Dissertations, Academic -- Physics -- UF
Electrons -- Scattering ( lcsh )
High resolution spectroscopy ( lcsh )
Phonons -- Spectra ( lcsh )
Physics thesis Ph. D
Plasmons (Physics) ( lcsh )
Surfaces (Physics) -- Optical properties ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1989.
Bibliography:
Includes bibliographical references (leaves 262-268)
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Jae Myung Seo.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
021090295 ( ALEPH )
22150156 ( OCLC )

Downloads

This item has the following downloads:


Full Text










ELECTRON SCATTERING STUDIES OF SURFACE PHONON-PLASMON MODES OF SEMICONDUCTORS

















By

JAE MYUNG SEO


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1989













ACKNOWLEDGEMENTS


I wish to express my sincere appreciation to all those who have provided support during this endeavor and who helped see it through to completion. Without these people I would not have a successful graduate career.

An unforgettable thank you is extended to the chairman of the supervisory committee, Dr. John E. Rowe, for his unending support and motivation. The research ethic he showed to me, his first Ph.D. candidate, will be greatly appreciated throughout my career.

I am deeply indebted to Dr. Paul Holloway for sharing his precious time and allowing me to use the state-of-the-art surface science equipment in his group. His optimism during my discouraging times will be a good lesson for my life.

I would like to thank members of my supervisory committee--Drs. Elizabeth Seiberling, Stephen Nagler, David Micha and David Tanner--for their guidance and support during my whole research career at the University of Florida. The input of Dr. Dale Doering, as a creative master of vacuum science is also deeply appreciated. Additionally, I would like to thank the following persons for their technical support: Scott Black for computer interfacing, Eric Lambers








for scanning Auger experiments, Paul Lyman for Shiraki oxide samples, William Wresh for Rutherford Back Scattering experiments, Chris Dykstal for Ni experiments, Larry Phelps and his colleagues for electrical equipment, and Harvey Nachtrieb and his colleagues for expert machining.

A special thank you is directed to Mrs. Glenda Smith for her help in the Physics Department. I appreciate the many discussions and advice on printing problems that Eric Lambers and Kelly Truman contributed to this dissertation.

My final thank you goes to my family members, my parents who allowed, encouraged and financially supported me in order to achieve this goal and my wife and our two daughters, Eunkyung, Hyosuk and Yesuk, who sacrificed much of their time and energy in completion of this goal.


iii














TABLE OF CONTENTS





ACKNOWLEDGEMENTS........................................ii

LIST OF FIGURES........................................vii

GLOSSARY OF SURFACE PHYSICS TERMINOLOGY................. xii

ABSTRACT ............................................... xiii

CHAPTERS

1 INTRODUCTION. ........................................ 1

1.1. Overview of Surface Physics Techniques........1 1.2. Overview of Specific Experimental Studies.....6

2 EXPERIMENTAL PROCEDURE .............................13

2.1. Overview.....................................13
2.2. Ultra High Vacuum............................16
2.3. Auger Electron Spectroscopy.................. 21
2.4. Sample Preparation ................... ......28

2.4.1. Overview.............................28
2.4.2. Specific Examples...................31

2.5. Low Energy Electron Diffraction ..............34
2.6. High Resolution Electron Energy Loss
Spectroscopy................................. 35

2.6.1. Basic Theory of 1270 Capacitor
Electron Optics ...................... 36
2.6.2. Resolution and Sweeping Mode..........40
2.6.3. Intensity Angular Profile.............43

2.7. Computer Interface ...........................45
2.8. Oxidation and Hydrogen Titration.............48
2.9. Film Growth under UHV ........................ 52








3 THEORETICAL BACKGROUND.............................56

3.1. Overview............. ... ............. ...... 56
3.2. Elastic Electron Scattering..................60

3.2.1. Diffraction from a Bulk Crystal
: 3-dimensional Diffraction..........60
3.2.2. Low Energy Electron Diffraction
: Surface Diffraction.................64

3.3. Inelastic Electron Scattering: HREELS........68

3.3.1. Semi-Classical Approach............... 70
3.3.2. Dielectric Function Theory............ 78
3.3.3. Impact Scattering: Off-specular
Scattering....................... ....83

3.4. Examples of Inelastic Electron Scattering....87

3.4.1. Surface Optical Phonon Excitation.....87
3.4.2. Surface Plasmon Loss and Dispersion:
Relation of Homopolar Semiconductors, and Plasmon-Phonon Coupling of Polar
Semiconductors......................... 89
3.4.3. Two-Layer Dielectric-Function Model
......................................97
3.4.4. Surface Phonon Dispersion for Semiinfinite Metallic Surface............ 102

4 ANGLE-RESOLVED SURFACE PHONON AND PLASMON MODES
AT Si(lll) AND GaAs(100) SURFACES.................107

4.1. Overview and Motivation.....................107
4.2. Experimental Results........................110
4.3. Discussion ..................................123
4.4. Summary.....................................130

5 VIBRATIONAL MODES FROM OXIDE LAYERS ON Ni(lll)
AND Ni(110).......................................133

5.1. Overview and Motivation......................133
5.2. Experimental Results........................ 134
5.3. Discussion................................... 145
5.4. Summary.............................. .. ...... 150

6 VIBRATIONAL MODES FROM OXIDE LAYERS ON Si(lll)....151

6.1. Overview and Motivation...................151
6.2. Experimental Results........................ 157
6.3. Discussion............ ....................... 168
6.4. Summary.............. ...... ...... ... ......182








7 SURFACE EXCITATION AND RELAXATION OF Ge/Si
ALLOY FILMS .......................................184

7.1. Overview and Motivation.....................184
7.2. Sample Preparation ..........................190
7.3. Surface Excitation from Ge/Si(111)-5x5......191

7.3.1. Adatom Vibration ...................191
7.3.2. Hydrogen Titration...................195

7.4. Surface Relaxation Measurements
Using Digital LEED ..........................203
7.5. Summary.....................................208

8 CONCLUSIONS....................................... 209

8.1. Conclusions from Present Experimental
Results.....................................209
8.2. Recommendations for Future HREELS Studies...212

APPENDICES

A FURTHER EXPERIMENTAL DETAILS......................214

A.1. Gas Handling System for UF HREELS System....214 A.2. Design of Evaporation System................214
A.3. Quantitative AES Analysis Using Standards...219 A.4. Digital LEED................................220
A.5. Tuning of HREELS Spectrometer............... 222
A.6. Computer Programs and Data Handling.........231

B FURTHER THEORETICAL DETAILS........................235

B.1. Dipole Scattering Cross Section............. 235
B.2. Transformation of Coordinates...............237
B.3. Fuchs-Kliewer Modes
: Lattice Dynamical Framework............... 239

C GROWTH AND CHARACTERIZATION OF Ge/Si ALLOY FILM...244

C.l. Alloy Film (1000A) Grown on Hot
Substrate (5600C) ..........................244
C.2. Depth Profile of Thin Alloy Film
without Ge-rich Islands.....................254
C.3. Depth Profile of Thick Alloy Film
with Ge-rich Islands........................ 256

REFERENCES. ............................................262

BIOGRAPHICAL SKETCH........................... .... ......269














LIST OF FIGURES


FIGURE TITLE PAGE

1 HREELS chamber. .......................................18

2 Electron escape depth...............................22

3 HREELS spectrometer ..................................37

4 Angular profile of the reflected electron beam of
the HREELS system.
.. ..................................................44

5 Block diagram of data acquisition system.............47

6 Si and Ge evaporation parameters..................... 54

7 Electron energy distribution .........................57

8 The reciprocal lattice and Ewald construction.
(a) Three dimensions; (b) Two dimensions.
.............. .................... ...... ........ ..63

9 Schematic diagrams of semi-classical dipole
scattering.(a) Electron trajectory and molecular dipole moment; (b) Transferred momentum; (c) Polar
plot of scattering intensity.
.. .................................................71

10 Plasmon loss calculation and plasmon-phonon coupling
calculation. (a) Plasmon loss calculation for Si(lll) at the specular geometry; (b) Plasmon dispersion curve for Si(lll); (c) Plasmon-phonon
coupling for GaAs(100).
... ................................................92

11 Two layer mode dispersion and polarization.
(a) Surface eigenmode dispersion for a slab of material with dielectric function c,(w) on a substrate with a dielectric constant, eb(>0); (b)
The electric fields for w, mode.
-o o...o....... o o o.....ooo................101


vii








12 Polarization of surface modes. The displacements in
the surface layer for the three surface modes at the x point of the two-dimensional Brillouin zone for an fcc crystal with nearest-neighbor central force interactions and a (100) surface. Modes S, and S, involve displacements mainly parallel to the surface while mode S, involves perpendicular
displacements.
.. .. e. ... ... .. ............ 0. .. .......0.. .......104

13 ARHREELS spectra obtained from partially ordered
Si(ll1)lxl for non-specular (AO 0) scattering.
...................................................111

14 ARHREELS spectra obtained from well ordered
Si(111) 7x7.
..................................................113

15 Dispersion curves of Si(111) surfaces............... 115

16 HREELS spectra obtained from a native oxide layer
on GaAs(100). (a) As introduced; (b) Annealed at
580C; (c) Annealed at 6650C.
...................................................116

17 ARHREELS spectra obtained from semi-insulating
GaAs(100) for specular (AO=0) and non-specular
scattering.
..e.......o............... e oo.........................120

18 ARHREELS spectra obtained from highly doped
GaAs(100) for specular (AO=0) and non-specular
scattering.
...................................................121

19 Experimental data and calculated dispersion curves
for highly doped GaAs(100).
.............................................. .....122

20 HREELS spectra for specular scattering obtained from
a Ni(lll)-p(2x2)-O surface with 0.25 monolayer of
chemisorbed oxygen.
.... ......... ......................................136

21 ARHREELS spectra obtained from Ni(ll1)-p(2x2)-O.
(a) 1.50off-specular; (b) 30off-specular; (c) 4.5O
off-specular.
0. ...-- ................00... .. .. ..... 137


viii








22 HREELS spectra obtained from nominally clean Ni(110)
showing a strong elastic peak and a weak loss peak
due to residual impurities.
.............................................. ......140

23 HREELS spectra obtained from thermally grown oxide
on Ni(110) under UHV.
..ooe.......oe.........................................141

24 HREELS spectra obtained from oxides on Ni(110)grown
under UHV at room temperature.
(a) 300L, 0,; (b) 1.8x10'L, 2O,; (c) 300L, air.
... ......................... .....................142

25 HREELS spectra obtained from Shiraki oxide on
Si(111). (a) As introduced; (b) After annealing at
500C;(c) After annealing at 900C.
.............oo.............o....oooo....................158

26 HREELS spectra obtained from water-preserved oxide
on Si(lll). (a) As introduced after preserving in deionized water for three weeks; (b) After annealing
at 500C;(c) After annealing at 9000C.
...................................................162
27 HREELS spectra obtained from native oxide on
Si(lll). (a) As introduced; (b) After annealing at
520C;(c) After annealing at 10100C.
................................................165
28 HREELS spectra obtained from thermal oxide grown on
a 700 K Si(lll) substrate. (a) 10L exposure; (b) 100L exposure; (c) lOOOL exposure; (d) 10 kL
exposure.
o..oo.........o...............o........................169
29 HREELS spectra obtained from thermally grown oxide
after annealing at 700 K.
.................. ....... .......................... 171

30 HREELS spectra obtained from thermally grown oxide
after annealing at 1100 K.
.......- .............................................172

31 Asymmetric stretching mode variations of thermal
oxides grown at 700 K and 900 K.
..oo**** **o** **o....** ** *... ... e.............181
32 HREELS spectra obtained from Ge0,Si0,.(111)-5x5
showing a strong elastic peak (E=0) and a weak loss peak (Ezl05meV) due to residual carbon impurities.
** .** *.** * *- -*-. *- ...* *.* *.... ...... ...... 192








33 ARHREELS spectra obtained from GeoSi05(ll111)-5x5
for specular (AO=0) and non-specular scattering
geometries.
....................................................193

34 Hydrogen titration on Ge,.Si0,(lll)-5x5.
(a) 2.5 L, H; (b) 5 L, H ............................199


35 Deuterium titration on Ge0.,S i0,(11l)-5x5.
(a) 2.5 L, D; (b) 5 L,D .............................201

36 LEED intensity profile of Ge05Si05(lll)-5x5
measured by digital LEED.
.. .......... ... ... ...... ........ ................205

37 Surface lattice relaxation.........................206

38 Schematic diagram of gas handling system............215

39 Design of evaporation system.
(a) Collimators; (b) Cross sectional view;
(c) Sideview. ........................................216

40 Schematic diagram of digital LEED....................221

41 Zoom trials of GaAs at different energies.
(a) Zoom at the elastic peak; (b) Zoom at 270 meV:
(c) Zoom at 360 meV.
..................................................227

42 Plate voltage versus capacitor voltage to correct
tuning.
........ ............................. ..............230

43 Screens of "LHMAIN" program.
(a) Filenames; (b) Selection of commands;
(c) Setting of parameters.
..................................................232

44 Ge evaporation control using AES.(a) Normalized
Si(92eV) intensity versus Ge coverage by Gossmann et at.[102]; (b) Normalized Si(92eV) intensity
versus Ge evaporation time to calibrate
evaporation rate.
*...... *..........*.. *..... .... ..................245
45 Thermal evolution of evaporated Ge film on Si(11l).
..*.... .............................. .............247

46 RBS data from 1000A of Ge/Si alloy film.............248








47 SEM photograph of 1000A of Ge/Si alloy film. Dark
plateau is area 1 and bright islands are area 2.
....................................................250

48 Sputter-AES of 1000A of Ge/Si alloy film.
(a) AES spectra obtained from area 1 as introduced; (b) AES spectra obtained from area 2 as introduced; (c) AES spectra obtained from area 1 after 30 sec.
sputtering; (d) AES spectra obtained from area 2
after 30sec. sputtering.
..................................................251

49 Sputter-AES of thin (~10A) Ge/Si alloy film.
(a) Sputtered edge profile of thin pure Ge film; (b) Sputtered edge profile of thin Ge/Si alloy film.
.................................... # ........... .....255

50 SEM and sputter-AES of thick (~200A)
GeoSi0,./Si(111)5x5 film.(a) SEM photograph with white islands;(b) AES line scan across one of white islands; (c) SEM photograph after point edge sputtering. (d) AES line scan across the sputtered
edge. (e) AES depth profile from another area.
........... ........................................257














GLOSSARY OF SURFACE PHYSICS TERMINOLOGY


AES

ARHREELS


CMA DAS

ELS FK FWHM HREELS

IR LEED MBE

RBS RHEED SAM SEM

SK STM

TEM UHV

UPS XPS


Auger electron spectroscopy Angle resolved high resolution electron energy loss spectroscopy Cylindrical mirror analyzer Dimer adatom stacking fault Energy loss spectroscopy Fuchs-Kliewer

Full width at half maximum High resolution electron energy loss spectroscopy Infrared

Low energy electron diffraction Molecular beam epitaxy Rutherford back scattering Reflection high energy electron diffraction Scanning Auger microprobe Scanning electron microscopy Stranski-Krastanov Scanning tunneling microscopy Transmission electron microscopy Ultrahigh vacuum Ultraviolet photoelectron spectroscopy X-ray photoelectron spectroscopy


xii













Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy


ELECTRON SCATTERING STUDIES OF
SURFACE PHONON-PLASMON MODES OF SEMICONDUCTORS


By

Jae Myung Seo

August 1989

Chairman: Professor John E. Rowe Major Department: Physics

Surface excitations such as surface phonons and surface plasmons of Si(111), GaAs(100), silicon oxide, nickel oxide and Ge.05Si,05(111) alloys have been studied under ultrahigh vacuum using high resolution electron energy loss spectroscopy (HREELS), low electron energy diffraction and Auger electron spectroscopy. It has been found that the surface free-carrier plasmons from sputtered and insufficiently annealed p-type Si(111) surfaces are losses due to a collective motion of acceptor-type carriers localized in the space charge region. The measured dispersion curve of these plasmons achieved by angle resolved HREELS agrees with that predicted by dielectric function theory using the Lindhard dielectric function and parameters appropriate to highly doped semiconductors.


xiii








Subsequent, high temperature annealing above 9500C induces a 7x7 reconstructed surface from which a surface phonon due to adatom vibrations has been detected. Surface plasmons due to sputter damage and Fuchs-Kliewer phonons have been detected from sputter-annealed GaAs(100) surfaces. The phonon-plasmon coupling effect has been observed experimentally and a dispersion curve has been measured by angle resolved HREELS (ARHREELS). It has been found from oxidation of Ni(110) and Si(lll) surfaces under ultrahigh vacuum that energies of oxygen-related surface vibrational modes increase as the thickness of the oxide layer increase in agreement with previous empirical correlations. This indicates that the ionic character of the oxide layer increases with increase of the thickness. Protective oxide layers on Si(lll) surfaces have thinner oxide layers on starting surfaces that are atomically smooth as compared to typical native oxide layers and do not have many active sites to react with residual gas. Also, this thin oxide layer requires a much lower temperature annealing without sputtering to be removed. In the study of Ge,Si05(lll)5x5 surfaces, it has been found that this alloy surface has similar adatom vibrational modes to the Si(lll)7x7 surface using ARHREELS and these adatoms are mostly (-90%) Ge atoms using hydrogen titration. For Ge0,Si0,(lll) a StranskiKrastanov growth mode has been detected and continuous surface relaxation of the strained overlayer has been found as the film thickness increases in the range 0-1000A.


xiv













CHAPTER 1
INTRODUCTION


1.1. Overview of Surface Physics Techniques


The potential energy function experienced by atoms at a surface is different from that of the bulk due to the absence of atoms above the surface. The atoms at the surface which are affected by this potential will arrange themselves in order to determine the lowest energy state of the combined system: surface and bulk. This results in the surface having structural, electrical and chemical properties which may differ greatly from corresponding bulk properties. This dissertation describes a sequence of surface physics studies using high resolution electron energy loss spectroscopy (HREELS) to measure surface vibrational modes. A variety of supporting experiments are also performed and discussed, since it is necessary to prepare, characterize and study a single atomically clean surface in the same ultrahigh vacuum (UHV) chamber.

It was not possible to monitor an atomically clean surface before UHV techniques were developed. Due to the short time during which a surface stays clean, experiments must be performed before the clean surface is contaminated by residual








2

gas impurities. If the pressure of the chamber is less than 1x10~9 torr, it takes about 103 seconds for the residual gas to cover one monolayer (1 ML) assuming the sticking probabilities are near one. At the present time, typical ion-pumped and well-baked systems usually have a pressure below 1xl10-'O torr, which allows about three hours of experimental measurements after surface cleaning. After the fundamental condition of modern surface physics experiments (low UHV pressures) had been achieved, various routine characterization methods of surface studies soon followed. For example, low energy electron diffraction (LEED), first demonstrated in 1927 by Davisson and Germer, has become a routine characterization measurement during the late 1960s and early 1970s, about ten years after the first commercial UHV ion pumps in 1961. Another characterization tool, Auger electron spectroscopy (AES) became routine in the mid 1970s although its feasibility was first demonstrated by J. J. Lander in 1953.

Surface studies can be classified by the probe particles such as electrons, ions, atoms and photons. The electron is the most common and convenient probe particle. The electron is a light, charged particle; it is easily focussed to become a non-destructive probe at low incident energies. Since the electron has a larger scattering or interaction cross-section than the photon, the finite electron escape depth from the surface enables one to collect the electrons originating mainly from the outermost 1-3 atomic layers, i.e. within the










escape depth of -5 A. The detected electrons can be the incident electrons which are scattered from the surface within the escape depth as in a LEED or HREELS experiment, or they can be secondary electrons ejected from surface atoms due to some other excitation mechanism such as Auger relaxation of excited core-hole states. The electron escape depth depends upon electron energy and to a lesser extent on the atomic number of the surface species. Usually electrons with energies from a few eV to several thousand eV escape from 520 A under the surface. In HREELS, the incident electrons with an energy of 1-20 eV penetrate 3-10 A under the surface, but the electric field which determines the so-called dipole scattering cross section extends in semiconductor and insulator materials down to 200-500 A below the surface. Photons, ions or neutral atoms can be also used as probes in a surface scattering experiment. Typically, several surface techniques are combined in order to achieve the overall purpose of experiments including surface characterization, surface preparation and final measurements.

The structure and elementary vibrational excitations of bulk condensed matter can be studied by spectroscopies such as inelastic neutron scattering, infrared (IR) spectroscopy, Raman spectroscopy and inelastic electron tunneling. High Resolution Electron Energy Loss Spectroscopy (HREELS), Infrared absorption-reflection spectroscopy and neutral atom inelastic scattering are the main probes of the vibrational








4

properties for the surfaces of solids. Inelastic neutron scattering is used for determination of the energy-wavevector dispersion relation of bulk single crystals, but the neutron phonon scattering cross section is very small, so this technique is not sensitive enough for surface studies of small area single crystal samples (however, powder samples can be studied). Surface IR spectroscopy is a technique for studying the molecular structure of adsorbate atom complexes and can be used on metal, semiconductor and insulator surfaces. The electric field of the incident IR beam can couple to the phonon modes of a solid and the absorption is usually shown as a function of the incident photon energy. The high resolution (0.1 meV) of IR spectroscopy is an advantage of this technique but the small absorption cross section for infrared photons limits the sensitivity of this method. Unless special reflection geometries are used, most of the detected IR signal is due to the bulk optical properties and provides an unwanted background signal that degrades the sensitivity. Raman spectroscopy and inelastic electron tunneling spectroscopy also have a substrate background signal problem which limits their sensitivity for surface studies. This problem is partially solved by the improvement of surface enhancement of these last two spectroscopies. Neutral atom scattering uses inelastic scattering of He or Ne atom beams of low-energy typically 20-60 meV. The resolution of atom scattering is very good (-0.5 meV), but it is experimentally








5
difficult to produce highly monoenergetic beams of neutral atoms. Since the maximum incident energy is usually limited to -60meV, this technique is useful only for obtaining surface vibrational modes below this incident energy.

The technique we have used, HREELS, covers a wide range of vibrational energies (5-500 meV) which in principle is also covered by IR spectroscopy. HREELS is useful for probing surface phonon dispersion, surface plasmon excitation and intra- and interband transitions. The high surface sensitivity of HREELS enables one to detect 0.001 ML coverage at the surface if the species has a typical dipole scattering cross section (-10"6cm2). Therefore many surface impurities can be readily detected. Due to the low incident electron energy, HREELS is a non-destructive technique which can be used to probe weakly bound species such as atomic hydrogen on graphite. In HREELS the scattering occurs due to the coupling between the incident electron and the electric field of charge fluctuations under the surface. This long-ranged interaction of the Coulomb field makes it possible to detect interfacial modes under the surface as long as the interface contains strong dipole species and is within the detection limit (-200500 A). Therefore HREELS is a very surface and interface sensitive technique with a wide energy range. However, it also has several limitations at the present time. Even with the most advanced electron optics, the resolution in HREELS is about 3 meV. This limits the analysis of closely spaced








6

vibrational modes such as H-Si modes on Si(100) which are split by only 1.4 meV. Since the performance of the electron optics depends sensitively upon the work function uniformity and stability of the energy analyzer capacitors, HREELS experiments should be done under UHV conditions. For highest signal levels, the specimen should have a dynamic dipole moment which has a large component along the surface normal. At the present time HREELS is under development to improve the experimental resolution of spectrometers and to improve scattering theory which does not yet treat long-range dipole scattering and short-range impact scattering on an equal basis.


1.2. Overview of Specific Experimental Studies

The motivation of this dissertation is to systematically study semiconductor surfaces using HREELS and in one selected example (oxidation of silicon), to compare a similar study on a metal surface (oxidation of nickel). For practical use of semiconductors, these materials have excess electrons or holes; i.e., they are n-type or p-type. These electron or hole carriers due to impurity doping play an important role in forming potential barriers at the interfaces of semiconductors with metals (e.g. Schottky barriers or ohmic contacts). Plasmons due to excess carriers can be distinguished from phonons by energy-wavevector dispersion which can be measured by angular studies of HREELS. Even if








7
a semiconductor does not contact a metal, dangling-bond surface states pin the Fermi-level at a position in the gap usually different from the bulk. Due to this band bending potential at the interface or bare surface, the local surface concentration of carriers can be dramatically different from that of the bulk. Using HREELS we observed such differences in carrier concentration by monitoring the plasmon energy for Si(111) and GaAs(100) surfaces.

In the present studies HREELS is applied to well ordered nickel surfaces since the formation of NiO at nickel and nickel-chromium alloy surfaces is not yet well understood. The simple selection rules make it easy to understand the polarization of specific vibrational modes. The coalesced nickel oxide layer is not a single continuous layer like a silicon oxide layer. Recently, a new interpretation of multiple oxide phases at Ni surfaces was suggested by a high binding energy oxygen species of nickel oxide layers detected with x-ray photoelectron spectroscopy (XPS). HREELS measurements of O-Ni vibrational modes were correlated with these XPS results.

In additional experiments, silicon oxidation in conjunction with surface cleaning techniques is studied. The formation mechanism of oxides at Si-SiO, interfaces; i.e., the initial steps of Si oxidation are not well understood. The knowledge of very thin oxides is closely related to cleaning techniques of an Si substrate prior to UHV experiments. Thin








8

protective oxide layers can be used to prepare clean surfaces under UHV without sputtering and high-temperature annealing. For two different reasons HREELS is a useful technique to study silicon oxides. First, oxygen atoms form partially ionic bonds with Si atoms, and their strong dipole moment enables one to easily detect the local vibrational modes of each type of oxide. Second, the high surface sensitivity of HREELS can easily detect surface impurities present on oxides and on clean surfaces down to 0.001 ML. These may include hydrocarbons, hydrides or hydroxyl groups as well as silicon carbide. Similar oxidation mechanisms with different oxide local bonding sites are expected for many metal oxide interfaces.

Alloys of Si and Ge are another interesting area for surface studies since they have reconstructed surfaces that are similar to the reconstruction found on clean elemental Ge or Si surfaces. A Ge.05Si,05 alloy grows epitaxially on Si(lll)7x7 substrates and has a 5x5 LEED pattern. From surface reconstruction studies by LEED and scanning tunneling microscopy (STM) both 7x7 and 5x5 surfaces are reconstructed in a quite similar way with about h monolayer of adatoms on a dimer-stacking fault unit cell. The vibrational mode due to adatoms has been reported for Si(11l)7x7 using HREELS. The cross section for Ge-Si alloys may increase due to a non-zero dipole moment in the Ge-Si system compared to the pure elements in the bulk. We can compare HREELS results for 5x5








9
and 7x7 surfaces and thus attempt to determine if the bulk composition is present at the surface. Another advantage of HREELS is the fact that the vibrational energy depends upon the reduced mass of the local oscillator. By titrating H atoms on the 5x5 surface the relative numbers of Si and Ge atoms in this surface can be determined since surface Si-H or Ge-H species have different reduced masses and the corresponding stretching modes have an energy difference of ~10 meV. We have found that a large fraction of the 5x5 surface (-90%) covered with H contains Ge atoms rather than a bulk 0.5-0.5 mixture. This is likely due to Ge having a lower surface energy than Si on this alloy surface.

Summarizing, High Resolution Electron Energy Loss Spectroscopy is a highly surface sensitive and quantitative vibrational spectroscopic technique. Through novel applications of this technique (i.e. Angle Resolved HREELS), surface phonon and surface plasmon modes were investigated from the semiconductor surfaces such as (1) the clean Si(lll) surface (2) the oxide surface of Ni and Si and (3) the Ge/Si alloy surface.

Chapter 2 describes broadly the experimental procedures following the typical sequential order of experimental measurements. Therefore the same experimental procedures are not repeated at the chapters (4-7) unless some new specific procedure was used. In addition to the procedures used to collect the data, a brief review of equipment and its








10

operation are also included. Especially, the electron optics of HREELS is described in some detail in order to allow one to understand the complex tuning problem of this type of instrument. In addition to the main experimental chapter, further experimental details are presented in Appendix A.

Chapter 3 describes the theoretical background of electron scattering as applied to HREELS. Electron scattering theory is divided into an elastic scattering part, which consists of bulk diffraction and surface diffraction, and an inelastic electron scattering part which includes dipole and impact scattering HREELS theory. For a clear understanding of dipole scattering theory, a semi-classical approach for adsorbed molecules as an example is comprehensively reviewed. For more general applications of HREELS on semiconductors a two-layer dielectric function theory is reviewed. For the application of angle-resolved HREELS, impact scattering theory is also reviewed. In addition to a review of fundamental HREELS scattering theories, applications to specific systems are briefly mentioned. Detailed theoretical background information is attached in Appendix B.

Chapter 4 is one of the four experimental measurement chapters. In this chapter, investigations of pure Si(111) and GaAs(100) surfaces are discussed. In the first section, sputtered annealed Si(111) surfaces are discussed. The quasielastic peak broadening effect and p-type surface accumulation layer formation due to a sputter-cleaning procedure is








11

considered. In the second section, sputtered and annealed GaAs(100) surface studies are considered. Also sputtercleaning effects are discussed in general. In this chapter, angle-resolved HREELS results show dispersion relations of phonon modes, plasmon modes and coupled phonon-plasmon modes.

Chapter 5 is the second experimental chapter. In this chapter studies of the room temperature oxidation of Ni(111) and Ni(110) are presented. The origin of a high binding energy oxygen species detected in x-ray photoelectron spectroscopy (XPS) data from the coalesced oxide layer is considered. A new interpretation of thin NiO layers is postulated based upon HREELS data.

Chapter 6 is the third experimental measurement chapter. In this chapter studies of oxide layers on Si(11l) surfaces are presented. In the first part, studies of a Shiraki oxide, a deionized-water preserved oxide and a native oxide formed in air on Si(lll) surfaces are discussed. The effect on the clean surface of removing oxide layers by annealing is considered. In the second part, thermal UHV oxidation studies of Si(lll) surfaces are presented. Differences in thin oxides are clearly evident in our HREELS results which indicate that the best thin oxide layers are produced by the Shiraki chemical etching method.

Chapter 7 is the fourth experimental measurement chapter. This chapter includes the film growth, characterization, relaxation measurement and excitation measurement through a








12
combination of various surface science techniques. In this chapter Ge05.,Si.,5 alloy films grown on Si(lll) are investigated. In the first part, surface vibrational excitations of the alloy film and its atomic hydrogen titrated surface were investigated by HREELS. Surface phonon and surface elemental species of the alloy film were considered. The origin of the 5x5 reconstruction is discussed. In the second part pseudomorphic growth studies and surface strain relaxation studied by digital LEED are presented. The critical thickness for pseudomorphic growth is discussed in the context of our measurements on (111) pseudomorphic films. In addition to the previous parts studies of the growth mechanism and morphology of the films studied by LEED, AES and Auger microscopy are presented in the Appendix C.

In Chapter 8 the experimental conclusions are summarized. In addition to experimental conclusions, recommendations for future studies are suggested.













CHAPTER 2
EXPERIMENTAL PROCEDURES


2.1. Overview


Surface studies have been made possible with the development of new pumping techniques during the 1960s. The competition between pumping speed and gas desorbing from chamber walls or leaking of the chamber determines the ultimate pressure. The basic condition for surface studies is that the pressure in the chamber must be kept as low as possible in order to prevent contamination of the surface by residual gas during the experimental procedures. To achieve the ultimate pressure for ultrahigh vacuum (UHV) (510-9 torr), combined pumping techniques such as turbo molecular pumping, bakeout at 200-2500C and sputter-ion pumping are generally used.

Another factor determining the quality of data is the preparation of a clean surface. A clean surface can be achieved by chemical pre-treatment before introducing it into the chamber with subsequent ion-bombardment sputtering followed by thermal heat treatment under UHV. The common goal of cleaning is reducing the level of contamination and defects








14

on the sample surface to well below 10' cm-', i.e. about 0.001 monolayer (ML). If pre-treatment such as oxidation, hydrogen adsorption or film growth under UHV is required, it is usually done in the same UHV chamber before data acquisition. It is important in surface physics experiments to prepare the required surface for analysis under UHV, since the transfer procedure from one UHV chamber to another may possibly bring contamination from the air. In some surface physics experiments several UHV chambers are interconnected with UHV transfer between each. However, in the University of Florida high resolution electron energy loss spectroscopy (HREELS) chamber several different sections of a single UHV chamber are used for various surface preparation, characterization and HREELS measurements. The advantage of pretreatment under UHV is in the control of gas adsorption and/or deposition of thin (0-10 ML) overlayers. Compared to an atmospheric pressure environment or low vacuum condition, the number of residual gas molecules in UHV is so small that the effective operation time can be long enough not to contaminate the surface with the residual gas during the measurement procedure (e.g. HREELS or low energy electron diffraction (LEED)). Basic surface analysis techniques using Auger electron spectroscopy (AES) for detecting surface chemical composition and LEED for determining the long range periodicity of surfaces follow cleaning procedures.








15

Once a clean and smooth surface is prepared under UHV, in our experiments the main analytical technique, HREELS, can be performed. It must be considered that the total measurement time includes tuning of the spectrometer as well as data acquisition. The critical point in tuning a HREELS system is to achieve the best resolution of the electron optics. The procedure of tuning can be divided into two parts, namely, detection of the optimum geometrical positions of monochromator and sample manipulator, and adjustment of the potentials of the electrodes and capacitors so as not to get the largest intensity of the elastic peak but to get the smallest width of the elastic peak. In angle resolved HREELS, the incident angle of monochromator is changed to vary the momentum transferred to the sample surface. It is necessary to understand the kinematical geometry of the spectrometer for determining the transferred momentum. Sometimes a diffracted order of the electron beam is needed instead of the specular (0,0) beam to suppress the high intensity of the elastic peak in order not to saturate the detector which would result in a nonlinear response. In this case understanding the orientation of the sample surface is essential in order to find the diffracted beam. The next step for tuning is to optimize the various potential settings of the spectrometer. This will be discussed in more detail later in this chapter. Data acquisition using a computer is useful for analysis of data by magnification, comparison of two or more spectra,








16

differentiation, smoothing and replotting. But a more important advantage of interfacing a computer with this system is in the reproducibility of small changes which can not be done manually. For example, an angle changed by a stepping motor controlled by a computer is essential to see the peak evolution at off-specular geometry since less than 0.10 steps are necessary to measure the angular width of the scattered beam.

In this chapter the entire range of experimental procedures will be covered. The major analytical technique, HREELS, will be comprehensively discussed while AES, back-view LEED, film deposition, UHV pumping techniques, computerinterfacing and sample preparation will be reviewed in the typical order of experimental use. Finally it should be emphasized that all procedures are essential to achieve high quality data, but well prepared sample surfaces and a thorough understanding of the electron optics of HREELS are key points for the surface physics experiments described in this dissertation.


2.2. Ultra High Vacuum


A high resolution electron energy loss spectrometer should be operated at a pressure below 1xl0-9torr. The work function of capacitors, monochromators and analyzers can be changed by contamination from residual gas, which results in










de-focussing of the HREELS electron optics. At the same time the sample surface is also quickly contaminated by the residual gas. If the sticking coefficient of the residual gas is assumed as one, then it takes 1000s (~17min) at 1x109 torr to cover the sample surface with one monolayer of residual gas. Since HREELS is a very sensitive technique, in order to eliminate impurity contributions to the data it is a prerequisite condition to achieve pressures below 1x10-"torr (i.e. 170 min for covering 1 monolayer, which consists of the time necessary for tuning HREELS and time for data taking by computer). The vacuum chamber overall diagram is shown in Fig. 1. Due to the chamber geometry and position of the pumping connections, the pressure at the entrance of ion-pump is almost half of the pressure near the hemispherical analyzer as measured by conventional UHV ion gauges. An HREELS spectrometer is just above the ion-pump since that position is near the lowest pressure in the chamber. A sample manipulator horizontally transports a specimen from the HREELS spectrometer to the electron analyzer. An electron gun and 5-crucible e-gun source are in one UHV section aimed at a common point which is just below the electron analyzer used for AES. Next to this electron analyzer are gas leak valves and an ion-gauge. Next to gas leak valves is the LEED optics section (see Fig. 1). In this section an 8"OD viewport allows one to observe the LEED screen as well as the motion of the sample manipulator. In order to allow a more complete view








LEYBOLD-HERAEUS SYSTEM


UPS xPs / XPS AES












5 CRUCIBLE E-BEAM EVAPORATOR


SAMPLE MANIPULATOR
X,Y,Z, 0


GAS INLET MANIFOLD


MBE TITANIUM KNUDSON SUBL I MAT I ON CELLS and
SON PUMPS
Figure 1. HREELS chamber.


TURBO PUMP


PRESSURE: lxlO-10


torr


(ELS-22)








19
of the diffraction pattern (Laue geometry), a rear view LEED system was constructed from a commercial LEED optics using the procedure developed by E.E.Chaban and co-workers[(l]. Photographs of our LEED patterns are usually taken through this rear view LEED port.

Pumping procedures to achieve UHV from the opening to the air of the University of Florida HREELS chamber are as follows.

(1) As soon as the sample is mounted, the electrical connection of sample relative to ground is checked to ensure the sample has not accidentally been grounded. Then all the ports opened during atmosphere operation are closed.

(2) The turbo-molecular pump line is opened and the mechanical pump is turned on to achieve low pressure (-1 torr) at the foreline of the turbo-molecular pump.

(3) If the pressure is below 1 torr between the turbo molecular pump and the mechanical pump, the turbo molecular pump and water chiller for cooling the turbo molecular pump are turned on.

(4) The turbo molecular pump is used during the backing procedure. The time required to reach the ultimate pressure in one hypothetical example of an unbaked metal-gasket system is 104'-105 hours, even though there is no leak in the chamber[2].








20

(5) Baking takes -20 hours at 2000C. During this bakeout time period the pressure increases to a maximum of p-10-6 torr and then decreases slowly over 10-15 hours to p~-5x10 torr.

(6) After cooling for several hours, the ion-pump is switched on and most of the filaments in the chamber including the titanium sublimation pump are outgassed. Briefly heating the titanium sublimation pump -2 times will reduce the pressure to ~Ix10-'o torr after the chamber reaches room temperature.

(7) If the liquid nitrogen cold trap is filled, the pressure can be lowered by a factor of two or three into the 10-" torr range.

The most important procedure discussed above is the bakeout procedure since the proper length of the bakeout time and temperature will determine the final pressure. Also baking reduces the total time of desorption from the surface of chamber walls and diffusion from inside surfaces of chamber walls. Design of vacuum systems including chamber and pumps depends upon the kind of experiments to be performed. The total pumping capacity is the main factor to determining the efficiency of the system. The order of design of UHV system is the determination of usage of chamber, determination of the available accessories, estimation of gas flow rate and determination of pumps. In our system a Leybold-Heraeus (model TRIVAC-A dual stage D16A) mechanical pump, with capacity of 400 1/s and a Leybold-Heraeus (model Turbovac 150) turbo molecular pump with capacity of 150 1/s are used. It








21

takes 5 min from air pressure, to reach one torr which is the proper pressure to start the turbo molecular pump. The turbo molecular pump is used during baking procedure and outgassing of the filaments. Finally an ion-pump (Leybold-Heraeus model IZ-270 with capacity:270-300 1/s) and a titanium sublimation pump are used in conjunction with a liquid nitrogen cold trap. In addition to pumping systems, a gas handling system is described in Appendix A.1.


2.3. Auger Electron Spectroscopy


Auger electron spectroscopy is a surface analytical technique for determining the elemental composition of the outer atomic layers of a solid. It provides high sensitivity, elemental selectivity for all atoms except hydrogen and helium and good lateral resolution (10"6-10-7m) of the analyzer which enables the detection of specific atomic species situated in the first few atom layers (5-10A). In AES, surface sensitivity is mainly determined by the escape depth of the observed electron, since the incident electron (E, = 3-5keV) penetrates much more deeply into the crystal. The inelastic mean free path (uan) of electrons as a function of energy (E,) shown in Fig. 2. is defined as the distance that an electron of energy E travels with a probability e-' of not being scattered by inelastic events that degrade its energy by more than several electron volts [3]. Therefore, in one type of




























0
rx~


10 20


50 100 200
Electron Energy


500
(eV)


1000 1500


Figure 2. Electron escape depth.


I w 11111I 'I 1 ,1111111'i ..
Ag
Au o"



Mo
C

5 lNi 1iIl I I Il ll I It Ji titil I








23

atomic Auger transitions, higher energy electrons usually escape from deeper layers. Besides the elemental information from the surface, the fine structure of Auger peaks is sensitive to the chemical environment. This change arises from transitions involving a surface molecular orbital formed by electrons in chemical bonds between the adsorbed species and the outermost surface atoms. The localized density of states at adsorption sites will be different from the clean surface. It is difficult in AES to relate peak shifts to changes of certain levels (a few electron volts) like XPS since three electrons are involved. But qualitative detection of chemical shifts in AES are at least indicative of differences in chemical bonds between elements. This can be found in silicon oxidation which produces a native silicon oxide layer.

Another aspect of AES is quantitative analysis. Two possible problems in using AES as a quantitative and elemental analytical technique are the determination of the total number of detected Auger electron separated from the background and the understanding of relation between the intensity of the spectrum and the atomic density in the selvedge of the specimen. At this point, these procedures are not fully developed. The typical AES experimental procedures are as follows [4].







24

(1) The derivative spectrum dN(E)/dE of the output of the spectrometer is integrated with respect to energy E to get

N(E).

(2) The background under peaks is determined by means of spline-approximation.

(3) Substraction of background from N(E) results in the distribution N,(E) of true Auger electrons.

(4) The area under each peak consists of multiplets and loss feature is Auger current which results from the corresponding transitions.

(5) This Auger current depends upon the angle of acceptance of the spectrometer, the primary current, backscattering factor, the attenuation by scattering processes after the transition, ionization cross-section for electron induced ionization of the specific level, the transition probability for the whole series related to a specific level, and the desired atomic density. By choosing appropriate values for each case the desired atomic density can be calculated.

It is generally assumed that the signal from a fraction of 102 of a monolayer can be detected by AES. Without going through details shown above, relative measurements (using the standard AES sensitivity factors) can give quantitative information on atomic density. The detailed procedures will be discussed in Appendix A, section A.3., Quantitative AES analysis using standards. Summarizing the introduction of AES, AES is useful for detecting elements at the surface since








25

it is very sensitive to most of elements and has good lateral resolution. AES also gives qualitative information on the chemical reaction between two elements if they are chemically reactive. Finally AES has the big potential for giving quantitative information of the surface composition.

An Auger transition converts an atom with a hole in one core level (i.e. i-th level) to a final state that consists of the atom with holes in other orbitals (i.e. 'j' and 'k') and the fast electron whose kinetic energy is Ek. In the initial process before Auger emission, a core hole is produced by electron impact ionization, i.e. element A,


A- A-(i) + e" (2-1)


where A'(i) means the hole in 'A' is in level 'i'. The Auger process is


A'(i)- A"(j,k) + eauge, (2-2)


where the emitted electron e-auge, has kinetic energy Ek from the vacuum level and A"(j,k) has left the atom doubly ionized with holes in 'j' and 'k' levels.

To understand Auger electron energy which can be measured by the analyzer, energies of the initial and final states of the Auger process should be known. The energy of the initial state of the Auger process (2-2) can be measured by x-ray








26
photoelectron spectroscopy, in which the i-th shell is photoionized in the process (2-1) [5]. The binding energy relative to the Fermi level of the electron in level 'i' is


Eb(i) = he (Ek + s ) (2-3)


where Ek is the kinetic energy of photoelectron emitted from the sample with work function 4,. But the final state energy is not the sum Eb(j) + Eb(k) of separate binding energies which can be measured like the initial state, since interactions between the two holes can occur and relaxation events to fill core holes are nonlinear. Considering ccc Auger process (i.e., the initial state has a core hole and final states have two core holes), the derivation of the final states A"(j,k) can be decomposed into two processes.


A- A* (j) + e- (2-4)


A'(j)-. A"(j,k) + e- (2-5)


In the process (2-4), Eb(j) is required, where EM(j) is the energy required for the threshold process which places the ionized electron at the Fermi level. In the process (2-5), Eb(k) is also required for the threshold process. Using the equations (2-2), (2-4) and (2-5),








27

EbM(i)= Eb(j) + Eb'(k) + (Ek + 0, ) (2-6)


where Ek is the kinetic energy of the ejected Auger electron. Energy required to produce the two holes in the levels 'j',

'k', namely Eb(j,k), is suggested as


Eb(j,k) = Eb(j) + Eb'(k)
= Eb(j) + Eb(k) + F(j,k;x) (2-8)


where F(j,k;x) describes the interaction (namely repulsive between holes) energy of the two holes. Shirley has pointed out that including the relaxation of other electrons in and around the atom (i.e. more screening) that takes place in the process (2-4), Eb(j,k) can be expressed as follows,


Eb(j,k) = Eb(j) + Eb(k) + F(j,k;x) R(x) (2-9)


which gives excellent agreement over the entire range of atomic number. In CVV and CCV case, the final state includes valence orbitals. According to the localization of the final state valence bond, the interaction term should be considered. It is not well known yet about the line shape estimation of CVV or CCV. Both energy and line shape change with chemical environment for Auger transitions having final states made up of valence orbitals involved in bonding, namely, CVV and CCV processes. Both chemical shifts and line shape variation have








28

been observed by AES. Energy shifts of Auger peaks have also been attributed to changes in relaxation effects associated with changes in chemical environment [6,7].

Sputtering-depth profiling with AES is a powerful analytical technique for detecting the depth distribution of elements in the film. In combination with rare gas sputtering, depth profile analysis uses the surface sensitivity of AES electron. But it is hard to get the quantitative information of the depth of each layer. Because of differential sputter rates caused by crystallite orientation and surface contamination, the depth resolution generally decreases as the film thickness increases [8,9]. Also combined with scanning electron microscopy (SEM), AES can probe at the special point due to its high lateral resolution while looking at the surface morphology. As another application of AES, quantitative estimation using standards will be introduced in Appendix A.3.


2.4. Sample Preparation


2.4.1. Overview

As mentioned in the experimental overview, one of important factors which determines the quality of data is the preparation of a clean and smooth sample surface. Recently this has become more important since thin film techniques such as molecular beam epitaxy and chemical vapor deposition








29

need atomically smooth surfaces. Conventional cleaning techniques, namely rare gas sputtering and high temperature annealing, work well for most substrates, but these procedures induce defects on the surface which cause unexpected impurities to migrate to the surface and vary the desired doping profile. To minimize the problem related to cleaning under UHV, it is recommended that the specimen be pretreated before introducing into UHV. Another reason for preparation of clean surfaces without sputtering and annealing is that the band bending at the semi-conductor surface, due to introducing surface states, results in depletion of the carrier at the surface. Ohmic contacts with metal require a thin barrier or a large number of carries at the interface.

In this section the normal cleaning procedures, namely sputtering and annealing, cleaving and chemical pretreatment on the semiconductor surfaces will be discussed. Especially for Si(111) substrates, a thin chemical oxide treatment the so called Shiraki method[10] and for the MBE grown GaAs(100), the As capping technique will be examined[11]. Besides the cleaning technique, the temperature measurement and heating will be discussed.

Some materials (e.g. Si, GaAs etc.) cleave easily to a direction of a favorable surface. In highly oriented pyrolitic graphite (HOPG), adhesive tape can be used to cleave the surface in the air since the graphite surfaces are nonreactive with air contamination. In III-V compound material,








30

a strong vacuum compatible epoxy can be used for cleaving the surface. These procedures, so called cleaving, can be used to prepare clean surfaces under UHV. Alkali halide (NaCl, LiF, NaF, KCl etc.) (100)faces, (lll)faces of materials like CaF2, Si, Fe, Zn and Be will cleave at liquid nitrogen temperatures[12], but this technique is limited to the small surfaces and a single orientation.

Bombardment of rare gas ions such as Ar* and Ne" (~10A/cm2 at 500eV-1000eV) can be used to clean a surface. Impurity atoms receive enough energy from the incident gas ion to be ejected from the surface. However ion bombardment can cause damage leaving the surface in an amorphous-like phase [13]. Some problems may arise from occlusion of the inert gas atoms on preferential sputtering of one of the components in a binary alloy or compound [14]. On an Si(lll) substrate after Ar'sputtering, annealing at 10000C for 2 min will produce the reconstructed surface with a 7x7 LEED pattern. But it is hard to estimate how many defect sites are left on the surface. On a GaAs(100) substrate after light Ne* sputtering, annealing around 500-5500C for 2min will produce a reconstructed surface with LEED pattern that will vary according to the stoichiometry of two components at surface [15]. It is also known that the proper choice of reactive gas will help the cleaning of the surface, such as oxygen to remove the carbon atoms in a suitable pretreatment and temperature.











2.4.2. Specific Examples

In cleaning an Si(111) substrate, along with the techniques listed in the previous section, laser annealing [16] and galliation [17] in which the substrate is exposed to a gallium vapor beam at about 8000C substrate temperature in UHV can be used. Galliation is known to be effective to remove silicon oxide, but inward Ga-diffusion or removal of Ga atoms by evaporation has not been confirmed.

The chemical treatment of an Si surface consists of two steps. The first is stripping the native oxide with HF acid and the second is applying a wet oxidizing chemical treatment like the RCA cleaning method [18]. But the oxide produced by this technique is relatively thick, such that thermal annealing at relatively low temperature (1000-1100 K) can not desorb oxygen species from the surface. To avoid high temperature annealing, it is necessary to prepare a very thin passive oxide layer which can be desorbed at a relatively low temperature (-1000 K) at which defects do not generate or the impurity profile does not change. This is the basic idea of the Shiraki cleaning technique[10] of which the procedure is listed in section 2.8.. The experimental results of this sample using HREELS are reported in chapter 5. The temperature of oxide of surface starts to desorb was 850C, but at 950C the oxide layer totally desorbs. The LEED pattern is very clean (i.e. most of the fractional order beams








32
can be clearly identified at 35eV) and HREELS shows a very broad elastic peak, which is a typical result for the impurity free surface. The preparation procedure takes 3-4 hours. Preserving the sample, prepared by Shiraki method, in D-I water for 20 days caused further oxidation of the Si(111) surface (refer chapter 5).

In GaAs grown by MBE, elemental As covers the final layer to protect it from contamination while transporting the sample from one chamber to another chamber. Since elemental As covered the GaAs MBE layer, the oxide AsO6 was formed in the air environment. After introduction into the UHV, annealing at ~4000C removed the volatile top oxide layer resulting in surface where the carrier is not depleted. The critical point in preparation of As capped surfaces is complete coverage with elemental As. If some portion is exposed, then air contamination will induce strong oxidation of the GaAs surface. Subsequently a low temperature (~4000C) anneal is not enough to get rid of the surface oxide on the GaAs layer.

Temperature measurement is critical in observing desorption temperature of adsorbed gases during the sample preparation. The temperature is measured by a Chromel-Alumel thermocouple which covers from -100 K to 1700K, and infraredoptical pyrometer which covers 3500C to 1500C. The operation of thermocouple is the current due to work function difference of two different metals. When the contact becomes hot, electrons are always transferred from the lower work function








33

metal to higher work function metal. The optical pyrometer is based on the principle that the color of the light emitted by an object is a function of the temperature of the object. The hot object is imaged by a lens in a plane where a filament of light bulb is situated. The current of this filament is regulated by a variable resistor until the brightness of the object and that of the filament is equal. This current reading using an ammeter can be calibrated directly in centigrade.

Annealing samples by resistive heating, namely flowing currents through the sample, works well with the sample which has a low resistance. In case of annealing a sample which has a high resistivity (>10 ohm-cm), a power supply with a high voltage (70V-350V) and a regulated current can be used. If a voltage source such as variac is used, the sample may be quickly destroyed. For the semi-insulating sample the high voltage (up to ~350V) is not enough to give initial heating. In a semiconductor, if the initial heating does not produce thermally excited carriers, it is impossible to heat resistively. Using another low resistive sample at the back of the sample, the initial temperature of the sample can be increased by heat conduction from the back substrate. Another method of heating samples with high resistivity is depositing a conducting film such as Ni at the back side of the film, then the current flows through the sample at relatively low voltage and starts to heat the sample.








34

2.5. Low Energy Electron Diffraction


In this section, the experimental set-up for LEED measurements will be introduced and digital LEED used for measurement of lateral surface lattice constants will be briefly described. A LEED optics consist of an electron gun and a hemispherical grid system. The LEED optics used is a modified Perkin-Elmer 15-120 LEED optics. Originally it was a front-view LEED system, but after changing the screen into a transparent glass screen, images can be seen from the other side of the sample (so called Back-view LEED). Electrons from filaments with primary beam voltage V, pass through lens electrodes with potentials V,, V, and V3 to focus into a parallel beam at the sample surface. The voltages V,, V2 and V3 are used to preserve the focus for a fixed ratio (V, V,)/(V, V,). Hemispherical grids allow one to achieve a field-free region between the sample and the first grid, G,, as well as to remove inelastic electrons with a potential barrier, G, and G,. After the elastically scattered electrons pass through the hemispherical grids, a concentric collector electrode (coated with phosphor) accelerates them to produce a bright visible image in a Laue diffraction pattern arrangement.

Electrons interacting with the sample will have various energies and directions. A hemispherical grid system is used to image the diffraction pattern by allowing only diffracted








35

electrons to reach the display screen. Usually the first and the fourth grid are kept at the same potential as the sample (i.e. grounded) to help shield the electron trajectories. Two inner grids are operated near filament potential to select almost elastically scattered electrons. The screen voltage (i.e. collector bias) is operated at 2-5kV to provide sufficient energy to excite the phosphor to reduce stray magnetic and electrostatic fields which can cause loss or distortion of LEED pattern. Insulators near the sample are shielded with copper plate. After obtaining a LEED pattern, digital LEED measurement using a vidicon camera (details in next section) or photo graphs are taken from the back-side view port. Photograph was taken using a commercial camera attached with zoom lens to focus at the screen. Normal 100ASA black and white film was used with manual exposure of 10-12 seconds.


2.6. High Resolution Electron Energy Loss Spectroscopy


The electron optics of the UF HREELS spectrometer (model ELS-22 Leybold-Heraeus System) will be discussed in this section from the point of view of describing the tuning procedure. It has been emphasized that understanding the geometry of the spectrometer and the relationship of each potential adjustment knob determines the resolution. Detailed tuning procedure is discussed in Appendix A.5.








36

The electron spectrometer of HREELS uses the deflection of the particle beam (i.e. electron beam) in a electrostatic field which disperses electrons in energy, so that a narrow energy band can be filtered by a slit. A schematic diagram of a spectrometer is shown in Fig. 3. Electrons emitted from the cathode pass through a repeller to favor forward emission. A three-element electrostatic lens is used to achieve focussing into the monochromator entrance. Passing through the electrostatic field of tandem-1270-cylindrical-deflectortype monochromators, a highly monochromatic (AE=5meV) beam is selected. Before interacting with the sample, monochromatized electrons pass through accelerating optics. After interacting with the sample, scattered electrons pass an almost symmetric path; namely decelerating optics and two 1270 cylindrical deflector type analyzers. Finally electrons are collected by a channeltron which multiplies the number of electrons into a short current pulse to enhance the intensity of the signal. To ensure monoenergetic electrons pass through monochromators or analyzers three connected slit plates are used at each entrance and exit of capacitors. To prevent electrons reflected by walls of capacitors, saw tooth type surfaces are engraved on walls of capacitors.


2.6.1. Basic Theory of 1270 Capacitor Electron Optics

The differential equation of motion of an electron in a cylindrical electric field is







High R Energy


esolut ion


Loss


Spectromneter


Sideview Toward


Detector 127 Analyzer



Target

1270
Monochr oinat


Sahnple


Manipulator


Electron


Figure 3. HREELS spectrometer.


Electron










d2y/d02 + y = Eo/(y*E*cos2a) (2-10)


where y is r/ro, r. is radius of parallel path to cylinder, E. is an incident energy of electrons, and (r,O) and E are a cylindrical coordinate and an energy of electrons inside the cylinder. The value a is the angle between the trajectory and parallel path at the same point, namely angular aberration. A first order solution of equation (2-10) is [19],


y = Eo(l-cos/2-4)/(E cos20) + cosfi 0 -{tanasinF2 #)//2.

(2-11)

Since the last term contains the largest angular aberration term, to achieve the best focussing the last term should vanish, i.e. sin!2 0 = 0. /2 = 7r gives 0 = 1270.

The potential inside a infinite cylinder is


V(r,) = a In r, + b (2-12)


where r, is a radius of trajectory. The potential of the outer plate is V(R), that of inner plate is V(r), the difference between them is A = V(R) V(r) and V(r0){=O) is the potential of main path which is parallel to the capacitor plate. Using above conditions, constants a and b in equation (2-12) can be expressed in terms of R,r, A and r, as follows,









V(r,) = [A/ln(R/r)]-ln(r,/r.) (2-13)


Electrons with their kinetic energy equal to central path energy, and entering the sector field in the direction of the tangent of the corresponding radius will travel in a circle when the centrifugal force and the electric force are equal. The electric force on an electron using equation (2-13) is


eE = e6V(r,)/&r, = (eA/[ln(R/r)]){(1/r,) (2-14)


The centrifugal force is mv/r,. Equating these two forces, the kinetic energy of an electron passing through the slit is


(1/2)mV2 = Ek,. = eA/[21n(R/r)] (2-15)


which is same as e(C + U,) where C is contact potential between the cathode and the slit and U, is the slit potential. From this relation, the capacitor potential A is linearly proportional to the slit potential U,. In the Leybold-Heraeus ELS-22 spectrometer, R = 41mm and r = 31mm for the main capacitors, so the theoretical slope will be 0.56 (i.e., 21n (41/31)) when plotting of A (of main elements) versus U,.

If the potential of the main path (i.e. V(r.)) is different from that of the slit potential, then electrons that enter the sector field will have to cross a potential step which causes angular distortion and decreasing output








40

intensity. In tandem capacitor systems, slits of both capacitors were designed to locate at the corresponding main radii. Electrons entering the sector field at a distance y0 from the main path with a velocity deviation P will leave the sector field at y, (=-y0+2r~) [20]. Since P has to be the same value for both capacitors for different r,, all slits should be located on the corresponding main radii. So the expected trajectory of tandem capacitors is the main path which is parallel to capacitor plate.


2.6.2. Resolution and Sweeping Mode

For the deflection analyzers, the base resolution may be expressed as a function of geometrical parameters in the general form as follows [21]:


AE'E. = A AS + B a" + C 82 (2-16)


where A, B and C are constants, a(maximum angular divergence) and P( the mean slit height) are the semiangular apertures and AS is the aperture or slit width at the entrance and exit.

The second term (4/3)a" causes poor resolution when targets enlarge the angular distribution of the reflected beam. For further development of the resolution, defining y,, as the radial deviation of the electron trajectory measured by dropping a perpendicular down to the central path,








41

the equation of the electron trajectory up to second order in a and using 1270 condition is [22]


y,= -y. + (AE/E.)r. -(4/3)roa' (2-17)



where y, is the deviation from the central path. With S being the slit width, the fastest electron that can pass through is that which enters the capacitor at S/2 with a,, and leaves at

S/2. Its relative energetic deviation is


AE'/E = S/ro + (4/3)a.x' (2-18)



The slowest electron enters at -S/ro with a=0 and leaves at

-S/rO with a=0. Its relative energetic deviation is AE-/E =

-S/ro. The total energy width within the plane of deflection is


AE,,/E = AE'/E AE-/E = 2S/r + (4/3) a.' (2-19)




For an electron which has a velocity component perpendicular to the plane of deflection (z-direction),


V = Vz + VO (2-20)


is perpendicular component of velocity v. Since


where v,







42

Vov, v,- (time of flight)=h, 0(1270) ro=v4 (time of flight), therefore v,=h/t=(h-vO)/(# r). A kinetic energy of electrons is


(1/2) mv2=m (v+v,2) /2

=(1/2) m2[ l+(h'/ (0'2 ro2) } ]. (2-21)


So the total energy width for 127 cylindrical deflection analyzer is


AEf/E = (2/ro)S + (4/3)a.,' + h2/(2r2r.') (2-22)


From above results, two terms of geometric deviation of beam profile add intensity to high energy side. Especially due to space charge, a monochromator system will contain too many electrons with higher energy than the pass energy. But the analyzer acts in an opposite way to the monochromator, namely a folding process with the analyzer transmission window, the resulting profile at the detector will be symmetric unless the beam is not disturbed by targets. Most of the energy broadening can be understood as an effective target disturbance which tends to increase the angular divergence.

Electrons which lose energies from the target surface and enter the analyzer with a kinetic energy of e(U,+C)-hw will pass the analyzer when the following condition is satisfied (see equation 2-15),











e(U,+C)-hw = 2A/[21n(R/r)] (2-23)


Since R, r and C are constant, to compensate hw, A or Usa must be changed. For HREELS, the AE constant mode which changes U. is preferred. So one moves the whole spectrum over a fixed transmission window and the electron optical system between the analyzer and the target acts as a zoom lens device.


2.6.3. Intensity Anqular Profile

The acceptance angle of an analyzer can be detected by measuring the intensity variation of elastic peak through the variation of incident angle of the monochromator. For an Si(lll)-7x7 surface, the reflected elastic peak intensity variation is shown in Fig. 4. Full width at half maximum (FWHM) is 1.950. But the FWHM of the elastic peak intensity variation of straight through mode (i.e. without sample) is

-1.30. This indicates that a portion of the reflected beam width is due to residual disorder of the sample surface. This may also due to a de-focussing of the reflected electrons by non-uniform fields such as contact potential fields of the sample holder components. Another variation for the elastic peak can be derived from the primary energy. Since the reflectivity from the sample can be varied by the primary incident electron energy, the best resolution can be achieved for different incident electron energies when the sample










300


Elastic Peak From Si(111)








FWHM
= 1.950










0 ......... 1 1 ... ...... ......
-3 -2 -1 0 1 2 3
Angle (degree)



Figure 4. Angular profile of the reflected electron beam
of the HREELS system .







45

surface is changed (e.g. desorption of oxygen or vice versa). From Si(lll) with a native oxide layer after annealing at 9200C for 3min under UHV, the measured resolution (i.e. FWHM of elastic peak) and intensity of elastic peaks for primary energies such as 5eV, 10eV, 15eV and 20 eV are as follows: (1) resolution is 12.1meV, 13meV, 7.5meV and 9.2meV respectively and relative intensity is 310, 360, 240 and 180 respectively.


2.7. Computer Interface


The high resolution electron energy loss spectrometer has been interfaced to an IBM pc/XT using IEEE-488 and Camac crate buses. The main purpose of computer controlled data acquisition is increased reproducibility of HREELS energy settings and reduction of noise by signal averaging. One of the limitations of HREELS which can be overcome by interfacing is in the scanning of pass energy of the analyzer via a motor driven potentiometer (ramp pot). Because of degradation of contacts, this potentiometer should be replaced regularly (

-6 months) however by using a programmable voltage source the pot lifetime is extended. In this section, the interface of the energy loss spectrometer to a personal computer to control the analyzer potential and incident angle of monochromator will be discussed. Also the structure of computer programs and data handling will be discussed in APPendix A,6.








46

A schematic diagram of interfacing is shown in Fig. 5. A Kepco 488-122 programmable power supply, with a resolution of 2.442x10' volts per step, has been placed in series with the ramp potentiometer in the HREELS power supply. The Kepco power supply which is used to set the pass energy on the analyzer has a resolution of 1/4096 with a maximum output voltage scale of one or ten volts. Then the ramp pot which was originally used for scanning can be used for detecting the elastic peak and monitoring the FWHM of elastic peak during tuning. In the same manner another Kepco power supply is connected to the HV amplifier sweep for ramping of hemispherical electron analyzer voltage for XPS, UPS and AES.

The output of channeltron detector is fed through an Ortec 109PC preamplifier to an Ortec 572 amplifier and finally to an Ortec 550 single channel analyzer set to transmit pulses only if their amplitude is above a preset threshold. This signal is fed either to a ratemeter for tuning or to CAMAC counting electronics. The CAMAC counting electronics consist of 3610 Hex scaler and 3655 timing generator. A second channel on the hex scaler is used to collect data from the TTL output of hemispherical electron analyzer electronics (Leybold Heraeus model LH-100) to collect a pulse counting signal from electron analyzer due to AES, UPS and XPS.

One of the important achievements of interfacing with the computer is in the precise reproducibility of the incident angle of the monochromator. A Klinger stepping motor










HREELS internal
supply


- to ana


IBM PC/XT


r aop


from detector


CAMAC ORTEC


Figuer 5. Block diagram of data acquisition system.








48

controller is placed on the IEEE bus and is used for obtaining angle resolved HREELS. Two stepping motors, one connected to monochromator and the other connected to sample manipulator, allow one to change angle of monochromator with a resolution of 0.05 degree of rotation, and to change the angle of sample with a resolution of 0.1 degree. A Keithley model 197 multimeter is connected to the IEEE bus to monitor the baseline (DC off-set) of HREELS analyzer. The Kepco supply, Klinger controller, Camac crate and Keithley multimeter are interfaced by an IEEE-488 bus to an IBM pc/XT running under DOS 2.10. Control of the IEEE bus is accomplished with a Tecmar IEEE interface card taking up one slot in the personal computer.


2.8. Oxidation and Hydrogen Titration


Using a gas handling system, reactive gases such as oxygen, hydrogen and deuterium can be chemisorbed or adsorbed on sample surfaces. In this section, two main procedures, namely oxidation and hydrogen-adsorption, will be introduced. Several different oxidation procedures, chemical wet oxidation, thermal oxidation and room temperature oxidation will be discussed. Hydrogen and deuterium adsorption on semiconductors will be also discussed.

Since -20% of air is oxygen and oxygen is a very reactive gas, silicon substrates can grow a relatively thin silicon







49

oxide layer in air. It is called a 'native oxide' and can be desorbed by thermal annealing at the temperature above 10000C under UHV. To prepare the native oxide on an Si substrate with a small amount of carbon impurity, degreasing the substrate is sufficient (i.e. acetone and methanol then blow drying). Removal of the native oxide only by high temperature annealing can not remove carbon impurities on Si surfaces in the temperature-limit of generation of no other defects. Wet chemical oxidation procedures have been developed. One of the methods which was used in these studies is the so called Shiraki technique which consist of a series of chemical treatments as follows [10]. An Si(lll) wafer 0.5mm thick, 1/4" x 3/4" rectangular, and p-type Boron doped with resistivity of 2.4 f-cm has been used. First the substrate must be degreased in hot (~800C) bath (whole procedure was done in hot bath except rinsing) in the sequence of methanol, acetone, trichloromethane, acetone and methanol. At least 10min should be spent in each bath. Then the substrate should be rinsed by deionized water (it will be called rinsing). Second, using HNO,, etching the surface region and forming an oxide layer for 10 min should be followed by an HF oxide etch for 10-15 secs and rinse. Repeating the HNO3 and HF sequential etch procedures must be continued until the surface dries uniformly. Third, in the combined chemicals, H20:NHOH:H20, = 4: 1: 1, the substrate must be oxidized for 5 min. Then rinsing, etching by HF : HO0 = 1:1 for 30 seconds







50

must follow to get rid of oxide. This is the so called alkali treatment. This alkali treatment should be repeated at least once more. Fourth, after a rinse in the combined chemicals, HO : HC1 : H20, = 5:1:1, the thin surface oxide must be made for 10 min using the alkali solution. Then a rinse, etching by HF : H,O = 1:1 for 30 secs and another rinse must follow to remove the thin oxide. Repeating with a solution of H,O: HCl: HO, = 5: 1: 1 the oxidation and rinse must be followed by etching using a new batch of HF: HO = 1: 1. Then without exposing the Si substrate to air, the HF: HO0 = 1: 1 solution must be diluted by adding deionized water. Finally the boiling, combined chemicals of H20: H202: HCI = 1: 1: 3 must be used to grow a thin oxide for 2 min. After waiting until it stops bubbling, the Si wafer must be rinsed for 10 min, spin dried and quickly transferred into the vacuum.

Besides chemical oxidation a hot Si substrate can be exposed to oxygen under UHV. The temperature of the substrate was increased up to 900 K during exposure to oxygen. If the substrate temperature is above 700 K, there are two advantages. One is oxygen molecule can dissociate on the substrate and become reactive. The other is adsorbed impurities related to hydrogen species desorb quickly from the substrate, leaving reactive sites for atomic oxygen. The oxidation procedure is as follows. Once the amount of exposure is determined, the total time and pressure should be calculated. Annealing the sample at a given temperature is







51

done after turning off the ion-pump or isolating the chamber by the butterfly valve. Oxygen molecules are inserted through a leak valve up to a higher than calculated pressure. Quickly opening the turbo-molecular pump line, the exact exposing pressure should be adjusted by the leak valve. After exposure, closing the oxygen line should be followed by turning off heating power. After the turbo-molecular pump reached its saturation level (in 2-3 min), ion-pump should be used for normal operation. The reason for using a turbo molecular pump during oxidation is to prevent hydrocarbon species from the ion-pump from adsorbing on the substrate and to preserve the pumping efficiency of the ion-pump by avoiding any high pressure exposure.

Oxidation of the substrate held at room temperature in vacuum can not be done by oxygen exposure alone. Positioning the substrate in front of a ion-gauge during exposure of oxygen molecules, helps to dissociate oxygen molecule into oxygen atoms which are reactive on the substrate. Even if all other procedures are quite similar to thermal oxidation, oxidation at room temperature is not as successful as thermal oxidation if surface impurities such as hydrocarbon species, hydrogen or hydroxyl group occupy reactive sites on the surface. Increasing the oxide thickness is limited by the cleanness of the starting surface.

Instead of an ion-gun, a tungsten filament can be used to dissociate hydrogen molecules into hydrogen atoms. On








52

semiconductors such as Si, GeSi,-x alloy and Ge, hydrogen was adsorbed by placing the sample in front of a hot tungsten filament. The effective exposure of atomic hydrogen can be calculated by assuming that only 10-' of the H2 would be dissociated into H-atoms. Thus the effective exposure of hydrogen atoms is 13.5 L for 15 min. at 1.5 x10" torr of H,. The exposure procedure is the same as the oxygen exposure case and the deuterium exposure is exactly same as hydrogen exposure. For hydrogen exposure residual gas contamination is expected and is often found due to reaction with the chamber walls.


2.9. Film Growth Under UHV


Since HREELS is a surface sensitive technique especially sensitive to impurities such as oxygen and carbon on an Si surface, another preparation method for clean surfaces is to grow a clean film in UHV. One additional reason for in-situ growth is that alloy sub-monolayer films are very fragile and cannot be prepared by the other methods described previously.

Monitoring the thickness of evaporated films is a critical point in growing thin films for HREELS studies. Especially in the two layer model the thickness of a film is one of the parameters which determines the loss energy. Thickness measurement by the variation of AES intensity and quartz crystal thickness monitor is discussed in Appendix C.







53

In this section, the growth rate monitored using a profilometer (Sloan, Dek-Tek II) will be discussed. Before the actual evaporation is conducted on the sample surface, the quartz glass crossed by tantalum wire is exposed to the evaporation source at the same position that the real substrate would be. After evaporation, the wire trace left among the evaporated spot will be detected as a groove by the profilometer. Since the thickness of this wire is 0.001", the sharp groove made it easy to level between each ends of groove. The evaporated film thickness is just the height difference between end of groove and center of groove. Even though the emission current knob (0-750 mA) in the power supply is the only variable to control the e-beam evaporation, the actual evaporated film thickness is not controlled precisely by the position of the knob, because the power is not stable. Further because the source is always changing, the position of the current knob does not give consistent results. Instead of the knob position the emission current reading provided a more consistent way to reproduce the beam evaporation conditions after several trial evaporations. In Fiq. 6, Ge source and Si source evaporation rates versus emission current are shown respectively. The evaporation rate as a function of emission current is different for each material. The initial vapor pressure curve for each material gives information on melting point and vapor pressure. A limitation in the use of the profilometer is that some




















0
C
-r-1


L
0 0l MD
UJ


35 40 45 Evaporation


Current


Figure 6. Si and Ge evaporation parameters.


60
(mA)







55

materials (e.g. Ni) will not make a uniform film on a quartz glass substrate and uneven island formations make it difficult to determine the average film thickness. In this case a glass substrate is replaced by another material such as Si which allows the material to grow as a uniform film.

Silicon and germanium are materials which make thin relatively smooth films on quartz glass to determine the evaporation rate. Once the emission current is determined the total evaporation time determined the actual evaporated film thickness after enough warming up the source. Finally details of design of evaporation system will be introduced in Appendix A.2.













Chapter 3
THEORETICAL BACKGROUND


3.1. Overview


The interaction of monoenergetic electrons with a sample surface will give rise to a typical energy distribution of scattered electrons due to incident primary electrons as well as secondary electrons from the sample surface. A typical energy spectrum, N(E), which is the emitted electron intensity per unit energy, is shown in Fig. 7 and is usually a strong function of emitted angle as far as relative intensities are concerned. The electron energy distribution, N(E), shows three features. A large maximum occurs at low energy. This first peak is due to electrons which suffer multiple inelastic collisions induced by a collision cascade process in the solid [23]. Near the incident electron energy, the sharp "elastic" peak which is comprised of the elastically scattered electrons plus the 'quasi-elastic' electrons that have lost a small amount of energy (0-200meV) occurs. Finally the intermediate region of the emission spectrum exhibits a series of smaller maxima with a small background. These losses result from












(1) Secondary Electrons

(2) Quasielastic Electrons

Elastic Electrons


(3)


Auger Electrons


Interband Transition Plasinon Excitation


(1)


(3)


Figure 7. Electron energy distribution.


(2)


Scattered Electron Energy Eo







58

plasmon excitation (due to valence band electrons), interband excitation and Auger electron excitation [24].

In this chapter we will briefly examine the theory of the second feature which includes elastic scattering as well as quasi-elastic scattering. Usually the total emitted electron yield increases as the primary energy increases up to several hundred electron volts. But the elastic yield alone of the second feature shown in Fig. 7, shows a different energy dependence. At normal incidence the elastic yield will be largest at primary energies less than 10eV, where it amounts to about 50% of the incident electron intensity and decreases with increasing primary energy.

The penetration of the primary electrons into the solid is limited by inelastic events and it is estimated that at typical energies (10-100 eV) the penetration depth is 3-10 A [25]. It follows that the elastic component of the emitted electrons can originate from a few atomic layers parallel to the surface. The wavelength of the electrons (1) is h/p by the de Broglie relation; in practical units


A= [150.4/E(eV)]" A (3-1)


where E is the kinetic energy in electron volts. At 100 eV the electron wave length, A, is of the order of 1.3 A so that a diffraction pattern of scattered electrons by the atomic lattice will occur.







59

As well as losing energy to other electrons in the crystal, the incident electron can also lose energy to the crystal lattice by scattering off a phonon, giving up some energy and momentum. Energy losses to phonons are very small of the order of 50-100meV. Electrons with such a small loss can not be filtered out by the energy selecting grids which are used in low energy electron diffraction (LEED) optics because their resolving power has a width of 0.5-1.0 eV; thus, these electrons are called quasi-elastic electrons. Besides phonon losses, the quasi-elastic peak can contain features due to all possible small energy loss mechanisms, e.g. surface plasmon losses due to doped carriers in semiconductor.

Elastic scattering is used to detect the long-range ordering of the surface using LEED optics and finite penetration depth of the incident electrons. Inelastic scattering with low energy loss (~100meV) is used to detect surface excitations related to phonons, plasmons and resonant electron-hole pair scattering. Two techniques, LEED and high resolution electron energy loss spectroscopy (HREELS), analyze the same elastic electrons with scattered electron background due to small inelastic energy losses but each of the emphasizes different features according to the purpose of the measurement. For structural analysis the angular intensity pattern of LEED is used and for determination of small energy loss features HREELS is used since this technique resolves the intensity features in a LEED angular pattern.







60

3.2. Elastic Electron Scattering

Before going into a description of elastic surface scattering theory, it is better to start from the kinematic description of scattering from X-ray diffraction from small crystals [26]. The first Born approximation is used for calculating kinematic scattering factors [27]. Also multiple scattering within the atom can be included in the scattering factor. Initially the ideal intensity of diffraction from a crystal will be derived. Then low-energy electrons will be considered an incident particles instead of X rays. Intensity attenuation relative to the penetration depth will be considered. Inner potential effects of the crystal as well as thermal vibration effects on the diffraction intensity will be discussed.


3.2.1. Diffraction From A Bulk Crystal
: 3-dimensional Diffraction

Assuming a sample is uniformly illuminated by the incident beam and ignoring multiple scattering, the scattering amplitude of an incident plane wave from N atoms, whose scattering factor is f,(8,E), is [26],


N
A = Ao Z f,(8,E) exp (iS'r,), (3-2)
1-1


where r, is the position of i-th atom and S = k k. is the transferred momentum. The factors k and k are the







61

propagation vectors of scattered and incident plane wave. Assuming each scatterer is at the lattice point

r,(=--m,a+m)b+mo) and atoms are located at x, in each unit cell n, the normalized amplitude is


A(S) = E f.(0,E) exp[i8- (r,+.)], (3-3)
i'n


where the sum is over the lattice sites i and the atoms within unit cell n. Separating the sums, then



A(8) = [E f (0, E) exp(iS x)j] E exp(iS r,)
n i

= F(0,E). Z exp(iS-r,), (3-4)
i


where F(8,E) is the crystal structure factor. The scattered intensity is then written as, I(s) = IA(s)12,


I(S) = IF(0,E) 1- EZ exp[iS. (r,-r)]
i,j

IF(e,E) 12. J(8). (3-5)


The interference function J(S) depends on the diffraction geometry through transferred wave vector S. Both F(O,E) and J(S) differ for different choices of non-primitive unit cells. For a parallelepiped with N,, N, and N3 lattice points, the interference function J(S) is







62

sin2 (N, 8 a/2) sin2 (N2S b/2) sin2 (N3S c/2) J(S) =
sin2(8S a/2) sin'(8-b/2) sin' (S c/2) .

(3-6)

The intensity becomes a maximum when S satisfies the Laue conditions, i.e.,


S*a = 2rl, S*b = 2,rm, and 8*c = 2rn. (3-7)


In terms of reciprocal lattice vectors A, B, and C, which are A=2v(bxc)/[a.bxc] and so on, the position of the

reciprocal lattice point (lmn) is G,., = 1A + mB + nC, which is a vector normal to the family of planes whose Miller indices are (lmn) with a magnitude IGsI = 2r/d., where d,.n is the interplanar distance. Then the Laue conditions are


S a = 2w1 = 27rG,. (a/2r) = G,,,*a and so on for b and c; thus, 8 = G,. = k k (3-8)


The diffraction geometry is displayed by the Ewald construction shown in Fig. 8(a). The incident wave vector has a fixed magnitude and direction and is terminated at the origin of the reciprocal lattice. The origin of the Ewald sphere whose radius is the magnitude of incident wave vector is at the same position as the origin of the incident wave vector in momentum space. Whenever this Ewald sphere passes through a reciprocal lattice point, the diffraction condition







x x Reciprocal Lattice


nI(S)
ko

xN x
8Gtmn" k ko



C ko K X B


(000) (10) (00) (01)

I(S): Interference function along a reciprocal-lattice rod.

(a) (b) Figure 8. The reciprocal lattice and Ewald construction.
(a) Three dimensions; (b) Two dimensions.







64

S = G,,, will be satisfied for the ray terminating at that point. Diffraction maxima gives the coordinate S and thus G,, and unit vectors a, b, c. From equation (3-5), the intensity of a diffraction maximum is I ,.1'2. But the modulus of F,. does not uniquely determine the atomic arrangement within a unit cell for non-Bravis lattices with more than one atom per cell.


3.2.2. Low Enerqy Electron Diffraction: Surface Diffraction

A beam of low-energy electrons passing through a material is attenuated by both elastic and inelastic processes. Inelastic processes are usually treated as simple attenuation by modifying the kinematic description derived above. Assuming the ratio of the amplitude contributed to the scattered beam by atoms in successively deeper planes is a = A,.,/ where n is the plane number. From equation (3-4), the total amplitude for a semi-infinite crystal is


A(S) = f (0, E) Z E exp[iS -(m,a+mb) ] *E a83 exp(im,S c)
m, m, m,

= f(8,E) Z Z exp[iS.(m,a+mb)]*[l-a exp(iS c)] .
m,m2
(3-9)
The corresponding intensity is



I() =f( sin' (N,8 a/2) sin2 (N, b/2) 1
sin' (S a/2) sin2 (8 *b/2) 1+a2 -2acos (S. c), (3-10)







65

the diffracted intensity now satisfies only two of the three Laue conditions, 8-a=2rl and Sb=2rm, and is confined along

lines in reciprocal space normal to the crystal surface, specified by S = G,. The reciprocal lattice and Ewald construction for a two dimensional lattice is shown in Fig 8bl. The lines are referred to as reciprocal-lattice rods and indexed by two integers (lm). These results which depend upon two dimensional periodicity of the surface are the basis for using LEED to determine surface structure. The modulation of the interference intensity function along the reciprocal lattice rod is determined by the factor [1+a2-2a cosS.c]-' The broad maxima exist at positions where the third Laue condition, 8-c=2rn, is satisfied. The breadth of these peaks is a consequence of attenuation. In the case of zero penetration, intensity depends only on the scattering factor jf(0,E) .

In addition to the attenuation factor, incident electrons experience a different potential while passing through the crystal. This periodic potential in the crystal half-space can be expressed as a Fourier expansion


V(r)=E V, exp(iG.r), (3-11)


and its spatial average V. is usually called the inner potential. Due to this inner potential V. the magnitude of wavevector in the vacuum is still k= 27r(E/150.4)" A-' but in







66

the crystal it is ks,=2v[(E+V0)/150.4]" A'. Conservation of parallel crystal momentum causes a refraction of the electron beam inside of crystal with refraction index


sin0 k,1[l+(V/E)]"
n = = (3-12)
sin0,, k .

The main contributions to the inner potential come from correlation and exchange interactions as well as the surface dipole layer potential and an imaginary part of the potential due to inelastic interactions.

Another factor determining the intensity is the thermal vibration of atoms in the crystal. The energy resolution in typical low-energy electron diffraction is insufficient to observe the loss or gain of phonon energies, so the intensity measured corresponds to both true elastic and integration over inelastic events due to phonon scattering. A diffraction experiment is equivalent to scattering from instantaneous and stationary configurations of scatterers. From equation (35), the instantaneous scattered intensity is


N
I(S) = If(0, E) 12 Z exp[iS (r,+u,-r,-u)], (3-13)
i,j

where u, is the instantaneous displacement of the i-th atom from its equilibrium position r,. The thermal average of intensity is










N
= if(0,E) I2 exp[-<(8-u) 2> E exp[iS.(r,-r)] i,j

{1+<(8-us) (S-u,)>+[exp<(s-u,) (8-u,)>-1-<(8-u,) (8-uj)>]).

(3-14)

The factor exp[-<(Su)2>] is Debye-Waller factor. The first term in the curly bracket is the zero-phonon scattering which is just the rigid-crystal scattering reduced by the DebyeWaller factor as discussed previously. The second term is the one phonon contribution to the thermal diffuse scattering. The third term is multiphonon scattering. So the effect of thermal vibration is to remove the intensity from the Bragg peaks and to redistribute the intensity throughout the Brillouin zone. Multiple scattering and inelastic process can be treated in self-consistent way to understand more detailed elements in diffraction problems. This is the goal of dynamical theory which will not covered here.

The theoretical treatment of elastic scattering of lowenergy electrons within a kinematic framework can be summarized as follows. Low energy electrons are scattered within a few atomic planes of the surface. Two dimensional diffraction pattern intensities with information on long-range order and the structure of the surface layer are modified due to both elastic and inelastic process. A full treatment involves the inner potential of the crystal, attenuation due







68

to electron-electron scattering, and thermal vibrational motion of the atoms (i.e. the Debye-Waller factor).


3.3. Inelastic Electron Scattering


Two conservation laws are the basic starting point for inelastic electron scattering theory and are summarized as AE = E, E, = hw and AkI= k,11- k11= q, + G11. (3-15)


Here AE is the energy transferred to the crystal and AkI is the momentum transferred parallel to the surface and G, is a reciprocal lattice vector of the crystal surface. Even if the vertical component of transferred momentum is large, a simple relation for it can not be established due to the broken symmetry normal to the surface. Approximate methods can be modelled to give periodicity of the z direction using periodic layers which consist of a large gap of vacuum to avoid the interaction between the surface and the substrate layers (-15 layers). Such a model will allow estimation of scattering near the surface which is quite similar to the system of vacuum and semi-infinite substrate.

From the law of conservation energy, an inelastically scattered electron can lose or gain energy depending on whether a phonon is created or annihilated. The relative probability of a phonon creation and annihilation is governed







69

by Bose factor, n=[exp(hw/kT)-1]'. The ratio of intensity of gain to loss peak (mn/(n+l)) is exp[-(hw/kT)]. From the law of conservation of momentum, the loss energy evolution relative to transferred momentum, either by changing geometry of scattering from specular to non-specular or by increasing primary energy at specular geometry, will give a dispersion relation of each mode. Based upon these two conservation laws, there are various applications of HREELS for the detection of surface excitations. The primary object of HREELS theory is then to describe a precise scattering differential cross-section which indicates the loss energy, its intensity and angular distribution.

In this section we will follow a semi-classical derivation of the differential cross section applied to a simple system which consists of an electron approaching an adsorbate on a metal surface [28]. The reasons it has been chosen are as follows: first it has several simple steps to derive the end result and is matched well with experiment, second it is easily connected to the actual geometry of the experimental set-up, and third it can be extended to further complicated, more advanced quantum mechanical theories. In addition to this semi-classical theory, the result of a quantum mechanical theory applied to a similar system by Persson will be compared with the semi-classical theory [29]. Also the result of a full quantum mechanical description applied to a more generalized system (i.e. dielectric function







70

theory for semi-infinite medium) will be introduced [30]. Even if impact scattering theory is not fully developed yet, the important characteristics of impact scattering have been reported. This impact scattering will be briefly discussed and finally examples and applications of inelastic scattering theory to specific systems will be presented.

3.3.1. Semi-Classical Approach

The electric field due to the specimen extends over the vacuum above the specimen. The longest range electric field is due to the surface dipole of the specimen. Considering the total range of this dipole field above the specimen, the time which the electron stays within this range is longer than the time it stays inside the specimen. This long range interaction is the so called 'dipole scattering' which is applied to specular geometry since the scattering intensity is sharply peaked around specular geometry. The scattered intensity of this dipole scattering applied to the system of an adsorbate on a metal surface will be derived following the description by Newns [28]. This theoretical approach to dipole scattering used here is the semi-classical one, since it treats an incident electron as a point particle traveling along a single trajectory remaining in the vacuum above the specimen at all times and exciting surface vibrations by means of its long-range coulomb field. The schematic diagram of the scattering is shown in Fia. 9(a). The coordinate of the electron is































Ro


r,: electron-dipole distance

(VlV): electron velocity

p: molecular instantaneous dipole
and its image in the surface

A,: resultant normal dipole





(a)


Figure 9. Schematic diagrams of semi-classical dipole
scattering.
(a) Electron trajectory and molecular dipole moment; (b) Transferred momentum; (c) Polar plot of
scattering intensity.









z





qX Sk,

ck0

x
0

w/k: parallel component of momentum change to reflected specular
beam

k$: perpendicular component of momentum change to reflected specular
beam

(b)

















a: 450



(c)


Figure 9.--continued.











r. = ( x+vt, v ItI ), (3-16)


where the origin is at the impact point and x. is the impact parameter in the surface plane. The term v, is the parallel velocity and v, is the normal component of velocity.

Since the dominant electron-molecule interaction comes from the part of electron trajectory where re is large, interaction between the incident electron and adsorbed molecule with dipole g can be written


V = Z V (1/r,) (a/aq,) q, (3-17)


The dipole is varying slowly at frequency wo compared to typical electron motion in a metal and the fast metal electrons will follow the instantaneous dipole motion adiabatically. Thus, the parallel component will have an opposite image dipole, while the vertical component will be approximately double the strength as long as the dipole is not imbedded into the metal. Only the normal component of the dipole moment has a non-zero perturbation. This is the consequence of the so-called normal dipole selection rule. The detailed derivation of this differential cross section is in Appendix B.1. For normal incidence vN=0 which gives ,Q=W0, and the differential cross section da is,









da = ( r2 4Q2V12 / [W0.2 +Qv1,2]2 } d2Q (3-18)


This function is strongly peaked near the characteristic wave vector Q-o~J/v. In typical experiments, wo=0.02 a.u.=540 meV and a primary beam energy E=0.20 a.u.=5.4 eV, then Q=0.01 a.u.. Since only small values of wave vector Q contribute to the scattering intensity, it can be deduced that the effective range of electron-molecule scattering is of order Qo' or at least 60 a.u. in the example. When 1/r, is expressed as a Fourier transform,


1 1 d2 Q[ dq exp(-iQ-x) exp(-iqz) re 27r2'j J Q2 + q2


= Q e-zQ
27j (3-19)


thus, the potential V has e-Qz factor, which also indicates that perturbation extends above the specimen up to Z~Qo"'. This is the underlying justification for using the long range dipole scattering interaction. One of the important reasons this semi-classical differential cross-section is chosen, is that the differential cross-section da is easily transformed to measurable quantities, e.g. k (incident wave vector) and measurable angles.







75

In Fig. 9(b), the outgoing elastic wave vector k, the outgoing wave vector k, after excitations of phonons and full scattering wave vector q are shown. Then


k = (v*.v,) = k(sina,0,cosa), (3-20)


where a is polar angle of k and q = k -k, = (Q,qz). Defining 8 and 0 as the polar angle and the azimuthal angle

respectively of k, relative to k (i.e. #=0 when k, is under k on the oxz plane), fully determines the scattered beam direction of k,. Energy conservation then gives



wo = k (k2-k,2 ) = k (k-k,), (3-21)


where h=e=m=l, and the second equation comes from relations q/k<


qA0=k-k,=k(0,0,1)-k,(sin0,0,cos0),

zk(-8,0,[k-k,]/k) a k(-0,0,0o/k2),

sk(-8,0,8), (3-22)


where the characteristic angle 0 is given by wo/(2E). The differential cross section derived by Newns is


4r2cosa f(0,O,a) Od0do
do = (3-23)
kc2 (92 +Ooz )








76

where f(0,0,a) is [(Ocos0-0otana)2+62sin2#sec'a]. The detailed angular transformation procedures are in B. 2. The denominator term shows that the differential cross section peaks strongly in the region OS8 where 0.=(j./ (2E). This strong forward scattering is a consequence of the small momentum transfer parallel to the surface. A polar plot of the scattering intensity of equation (3-23) for the case of a=450, 8o=0.1 rad., and 0=0 is shown in Fig. 9(c). The nodal value comes from f(0,0,450)=0, i.e. 9 = dotan450 in this case and forward scattering lobe is very distinct.

Considering the aperture of the analyzer (cf.refer the measurement of intensity profile Fig. 4 in chapter 2), the analyzer accepts scattered electrons lying in a cone of apex, 8,. The total cross-section, a, due to this finite acceptance of analyzer can be integrated up to 8, which is the polar angle of scattered electron with respect to specularly reflected direction.



[80 [2v 4r2cosa f(a, o, a)0
a =dd
J0 J0 k2 (02+0' )2

Frr2
= cosa [(t2-2)Y+(t2+2)lnX] (3-24)
E


where Y=0,2/( ,' +0o2), X=l+(0,2 /.2 ) and t=tana.








77

Assuming the surface coverage of adsorbates is n molecules per unit area, the current of one phonon loss (I,) versus that of unscattered specular reflection (I) will be


I,/Io = (wr n/E) cosa [(t2-2)Y+(t'+2)lnX]. (3-25)


The main feature of dipole scattering of the adsorbate on metal surface using a semi-classical approach is that the scattered electrons form a strong forward scattering lobe with vibrational modes polarized vertical to the surface.

Such dipole scattering was also described by Persson using a quantum mechanical approach [29]. In this treatment the electron wave function near scattering region is




where k'=k-2n(n-k) and 6 is the phase factor due to scattering. Using the Fermi-Golden Rule and a perturbing potential H'= -y*E due to the elastic field of external electrons and the image charge, the probability per unit time for vibrational excitation of the molecule was calculated. The differential cross section is


da/dG = (mpe/rEnh) [pp,/ (pocosa) ] [a,,/a' + (b1cos6+bisin6)/b2]2, (3-27)








78
where a=k-k,, b=k'-k,, p,=hk, and po=h11k. For the specular geometry condition, (a1/a' ) > (bi/b' ) or (bJb2 ) ,


da/dl = (mpe/ oh)2 [p,/(pocosa) ] [a,,/a2 2 (3-28)


The total cross section, a, for the electrons collected by a cone of half-angle 0, around the specularly reflected beam is



a = (Me/heov)2 -(cosa/2r) [(t2 -2)Y+(t2+2)X], (3-29)



where X,Y,t and a are defined the same as in equation (3-24). Comparing these results of da/dO and a with the semi-classical approach, we see that a quantum mechanical approach agrees with the semi-classical approach.


3.3.2. Dielectric Function Theory


In the previous section, electron scattering was limited to the system of adsorbed molecules on metal surfaces, but dipole scattering can be produced by any excitation of the sample accompanied by a fluctuation in charge density [31]. In this section the general description, using a quantum mechanical approach, for the differential cross section related to the dielectric function of the layer will be outlined following the derivation of Mills. The potential







79

outside the semi-infinite (z<0") specimen in which a time dependent charge fluctuation, n,(x,t), at a point x is



F n,(x' ,t)
4(x,t) = dx' (3-30)
J;.O. I x'l ,



where the integration extends over the specimen. Retardation effects are ignored since the time scale of information transfer from specimen to incident electron is very short compared to the period of excitation, which is also assumed in the previous section. Using the Fourier transform of I x x' I-' from equation (3-19), the potential seen by the electron outside the specimen can be written as



d'Qg1 (
0(x,t) = 27r[ e'q4' e-Q1 I n,(Q1z';t) ; eIz dz'
J Q1 J z..P
(3-31)

where

n,(Qllz';t) = d2'Y' e"'q n,(x',t) (3-32)

From the first integral, the factor e-' z indicates that the potential decays exponentially. The potential extension above the specimen (vacuum) is z=Qll", which is the height the electron starts to experience the potential due to a component







80
with wavevector QI. The factor eqIz in the second integral (of equation (3-31)) which is the charge source integral indicates that the contribution to the potential with component exp(iQ,-zl) is produced by charge fluctuations that extend down to a distance Q below the surface of the specimen.

The differential cross section can be obtained by inserting equation (3-31) into the Schrodinger equation and using Born approximation,



da 21R2v4 k' S(Q,)
(3-33)
dnk dhw hkr (ea)2 cos0, [VI2Q2 +(w-QI .V)2 32



The term ao is the Bohr radius, 0, is the angle of incidence relative to the surface normal, k and k' are the magnitudes of the incident and scattered wave vectors, IR i is the reflectivity for specular scattering, and Qg= 1-kg'. The spectral function S(Q1,w) is




S (Q,) = d2xy1 rdt exp [ iQn.-1-iwt ]
on1 J.W

x dz' dz exp[Q,(z+z')] , J J(3-34)
(3-34)







81

where x = x9+zz, and the brackets with subscript T denotes the average of the quantity enclosed over the appropriate statistical ensemble at the temperature T. Defining the scattering probability P(k,k') by,



1 do
P(k,k') =
IRI 2 df.d(hw)

2 v,' k' S(QI, W)

h7r (eao) 2 cos0, k [v2 Q12 + (w-Ql VI) 2 ]2

(3-35)

Thus P(k,k')dn.dhw is the probability an electron scattered into the solid angle dfln, in the energy range between hw and h(w+dw), normalized to the elastic intensity (here, ReReR was already assumed). The scattering probability shown in equation (3-35) has a kinematic factor which is equation (335) itself. This kinematic factor is independent of the property of specimen and peaks for small momentum transfers at Q1i~k(hw/2E). Since Q,' is the range of the potential as well as the probing depth, shown in equation (3-31) and its following discussions, the kinematic factor derived in equation (3-35) has shown a sharp forward scattering lobe by the potential with angular width of 2E0/(hw) and the effective probing depth of HREELS also turned out 2E0/(khw) from equation (3-34).







82

The quantity S(Q1, ) in equation (3-35) contains information on the specimen surface. Since a charge fluctuation n,(x,t) generates an electric field E(x,t) by Maxwell's equations, the charge density correlation function can be replaced by the correlation function of the electric field fluctuation, T. This can be related to the dielectric response functions of the substrate through use of the fluctuation-dissipation theorem [32]. The spectral density, S(Q1,w) can be constructed by means of a Green's function method [31], and the result is


S(QlW) = (2Q1/7r) N(w) Im[-1/{l+e(w))], (3-36)


where N(w) = [exp(hw/kT)-l]-' and e(w) is the dielectric function of specimen. The scattering probability is





4 v,' k' QI N(W)
P(k.k') =
17r2 (eao)2 cosO, k [v12 Q11 + (-QII. *vI)2 ]2

-1
x Im[ ] (3-37)



Equation (3-37) suggests that the information of scattering geometry and dielectric function of specimen are required to estimate the scattering probability.







83

Summarizing the results of this section, the scattering probability near specular scattering geometry consist of two factors. One is kinematic factor fully determined by the scattering geometry and the other is the spectral function, S(Qg,u) which contains information of surface property, e.g. a dielectric property. Also this general description of the dipole scattering probability of a semi-infinite medium confirms that the scattering probability sharply peaks near the specular direction (i.e. forward scattering). The electric potential due to the specimen (i.e. 0Oz-Qi-') extends up to a range of Q,,-, where QU is the transferred momentum.


3.3.3. Impact Scattering: Off-Specular Scattering

In dipole scattering theory, to simplify the derivation of scattering probability, major assumptions are single scattering and very small amount of momentum transfer parallel to the surface. Another regime of scattering accompanied with relatively large momentum transfer parallel to the surface is called impact scattering. To induce large momentum transfer, the incident electron energy can be increased at a specular geometry. Since the aperture of the analyzer has a fixed solid angle (AO-1), electrons passing the edge of the aperture of the analyzer have different momentum for different electron energies; i.e. the larger the energy is, the larger the transferred momentum is. The other way to increase transferred momentum is changing the geometry of spectrometer









from specular (i.e. 0, ,n=escideeed) to off-specular geometry at a fixed incident energy. But a mixing of both ways (i.e. energy and geometry) is not desirable for experiments observing the evolution of peak intensities since the reflectivity from the sample depends upon the incident energy very much. The first method can be simply used for observation of evolution of peak position. To detect the dispersion relation of surface excitation, incident energy should be increased up to a few hundred electron volts in both methods to cover the entire Brillouin zone whose size is typically a few A'. Another reason for using impact scattering is to determine local site symmetry from surface vibrations. Since large transferred momentum induces increased surface sensitivity

(e4Iz and eOI factors in potential and source, see equation (331)), a microscopic treatment is needed to interpret the spectrum. It is not so simple as the dipole scattering case to derive the scattering probability since the high incident energy (-300 eV) to cover the Brillouin zone edge causes a multiple scattering as we have already discussed in the overview of this chapter. So it is necessary to approach the scattering problem from other direction to get the differential cross-section of this multiple scattering. In this section a condensed account of Mill's derivation and results will be presented [33]. Also the selection rule for impact scattering will be briefly introduced.







85

The basic idea in impact scattering theory is the calculation of intensity of electrons which contribute to the thermal diffuse background of a diffraction pattern. Since dipole scattering is rapidly reduced at off-specular geometry, the intensity of the background is mostly due to impact scattering. When an electron encounters a solid, the positions of nuclei are not fixed and are displaced by thermal vibrations. The position of nucleus 'i' is then R,=R,,+u, where R0, is the position of the equilibrium site and u, is the displacement from equilibrium position, R,,. For small displacements of u, the scattering amplitude f(k,,k,;R) can be expanded in powers of ua ,


8f
f(ks,k,;R) = f(ks,k;R.) + Z (- )o*ua + (3-38) a aRa



where ua is the oth Cartesian component. Expressing ua in terms of the normal mode eigenvectors a,



h
ua = E (- )" a (as + as') (3-39)
20,M,


where 's' refers to a particular normal mode and a,,a,* are the annihilation and creation operators of vibrational quanta, and Mi is the ionic mass. When a particular vibrational quantum is emitted, the matrix element







86

M(k,,k,;+s) =

= (n,+l)"(h/2Ne,)"(8f/aQ,) (3-40)


where (af/8Q,) = E(8f/8Ra)o /M," and n,=[exp(hes/kT)-l]-'. Then the probability that the vibrational quanta (Q11ae) scatters the electron into the solid angle dn from the surface area 'A' is


dSa(k, ,k,) mE, cos'0,
= A IM(k,,k;Qlja) 1 (3-41)
dn 2r2 h2 cos,


where a contains all indices other than wavevector Q1, and 0, and 0, are an incident angle and a scattered angle. A further analysis of multiple scattering will not be covered here.

Selection rules for impact scattering provide a basis for obtaining direct information about symmetry of an adsorbed atom based on the general feature of the inelastic single scattering cross section. Selection rules are based on the symmetry of the substrate, time-reversal symmetry, scattering geometry and the direction of the polarization of the vibrational mode. The results by Tong et ial. were as follows [34]. If a normal mode is polarized out of the scattering plane as well as parallel to the surface, and the substrate has a reflection symmetry relative to the plane perpendicular to the scattering plane, the cross-section in any position in the scattering plane is zero. If the substrate has a rotational symmetry about z axis in the above case, the differential cross-section is zero at only specular geometry.




Full Text
123
angle change correspond approximately to a parallel momentum
change of 0.02 '1.
4.3. Discussion
Angle resolved energy-loss specular from Si(111) showed
one loss near 10-15meV and another loss near 22-30meV which
are assigned to surface plasmon and surface phonon modes,
respectively, based on dispersion behavior predicted from
theoretical models. The quasielastic peak in the specular
direction, which results from elastic scattering broadened by
unresolved inelastic losses, decreases rapidly in off-specular
directions, while other loss features may increase in relative
intensity. Possible Si(111) losses with energies in the range
of 0 to 30meV include acoustic phonon, optical phonon and
plasmon [56-58]. Each excitation may be dipole-like or
nondipole-like [33]. Loss energy versus transferred momentum
parallel to the sample surface (q,| ) is shown in Fig. 15 with
q,l = 2w (Eo/150. l)H(sin0,-sin0,) (4-1)
where E0 is incident electron energy, 6S is the scattered
electron angle and 9, is the incident electron angle relative
to the sample surface normal. To compare the results to
theoretical calculations, it was assumed that the loss peaks
are described by the poles of loss intensity function
Im[-1/(e (w,q,|) }+i], which gives the condition of e (<>#?,,) =1.


141
_C3
ln
G_3
Energy (meV)
Figure 23. HREELS spectra obtained from thermally grown
oxide on Ni(110) under UHV.


ACKNOWLE DGEMENTS
I wish to express my sincere appreciation to all those
who have provided support during this endeavor and who helped
see it through to completion. Without these people I would
not have a successful graduate career.
An unforgettable thank you is extended to the chairman
of the supervisory committee, Dr. John E. Rowe, for his
unending support and motivation. The research ethic he showed
to me, his first Ph.D. candidate, will be greatly appreciated
throughout my career.
I am deeply indebted to Dr. Paul Holloway for sharing
his precious time and allowing me to use the state-of-the-art
surface science equipment in his group. His optimism during
my discouraging times will be a good lesson for my life.
I would like to thank members of my supervisory
committeeDrs. Elizabeth Seiberling, Stephen Nagler, David
Micha and David Tannerfor their guidance and support during
my whole research career at the University of Florida. The
input of Dr. Dale Doering, as a creative master of vacuum
science is also deeply appreciated. Additionally, I would
like to thank the following persons for their technical
support: Scott Black for computer interfacing, Eric Lambers
ii


78
where a=k0-k1, b=k0'-k1, p,=hk, and p0=hk. For the specular
geometry condition, (an/a2 ) (bN/b2 ) or (byb2 ) ,
dcr/dn = (m/ie/7T60h)2 [p,/(p0cosa) ] [a,|/a2 ]2 (3-28)
The total cross section, a, for the electrons collected by a
cone of half-angle 8, around the specularly reflected beam is
o = (Me/he0v)2 (cosa/27T) [(t2-2)Y+(t2+2)X], (3-29)
where X,Y,t and a are defined the same as in equation (3-24).
Comparing these results of da/dn and a with the semi-classical
approach, we see that a quantum mechanical approach agrees
with the semi-classical approach.
3.3.2. Dielectric Function Theory
In the previous section, electron scattering was limited
to the system of adsorbed molecules on metal surfaces, but
dipole scattering can be produced by any excitation of the
sample accompanied by a fluctuation in charge density [31],
In this section the general description, using a quantum
mechanical approach, for the differential cross section
related to the dielectric function of the layer will be
outlined following the derivation of Mills. The potential


Figure 38. Schematic diagram of gas handling system.
215


82
The quantity S(Q(|,w) in equation (3-35) contains
information on the specimen surface. Since a charge
fluctuation n, (x,t) generates an electric field E(x,t) by
Maxwell's equations, the charge density correlation function
can be replaced by the correlation function of the electric
field fluctuation, T. This can be related
to the dielectric response functions of the substrate through
use of the fluctuation-dissipation theorem [32]. The spectral
density, S(Q)(,to) can be constructed by means of a Green's
function method [31], and the result is
S(Q,W) = (2Qn/tr) N(to) lm[-l/{l+e(w) }], (3-36)
where N(to) = [exp(hto/kT) -1 ]'' and e(to) is the dielectric
function of specimen. The scattering probability is
4 vA4 k' Qii N(to)
P(k.k') =
Iitt2 (ea)2 cos, k [vA2 Q2 + (w-Q,, v|()2 ]2
-1
x Im[ ] (3-37)
1+e(to)
Equation (3-37) suggests that the information of scattering
geometry and dielectric function of specimen are required to
estimate the scattering probability.


76
where f (0,0,a) is [ (0cos0-0otana)2+02 sin2 0sec2 a] The
detailed angular transformation procedures are in B. 2. The
denominator term shows that the differential cross section
peaks strongly in the region 8<60 where 0o=wo/(2E). This
strong forward scattering is a consequence of the small
momentum transfer parallel to the surface. A polar plot of
the scattering intensity of equation (3-23) for the case of
a=45, 0o=O.l rad., and 0=0 is shown in Fig. 9(c). The nodal
value comes from f {6,0,45) =0, i.e. 8 = 0otan45 in this case
and forward scattering lobe is very distinct.
Considering the aperture of the analyzer (cf.refer the
measurement of intensity profile Fig. 4 in chapter 2) the
analyzer accepts scattered electrons lying in a cone of apex,
8,. The total cross-section, a, due to this finite acceptance
of analyzer can be integrated up to 8, which is the polar
angle of scattered electron with respect to specularly
reflected direction.
a
'6, ?2n 4T2cosa f(0,0,a)0
d8 d0
0 J0 k2 (02+0o2)2
7rr2
cosa ( (t2-2)Y+(t2+2)lnX] (3-24)
E
where Y=0,2 / (0t2 +60> ) X=1 +(0,2/0o2) and t=tana.


Intensity
138
x40 x200
-50 0 50 100 150 200
Energy (meV)
Figure 21.continued


186
then more evaporated Ge starts to form a three dimensional
island. This typical growth mode was confirmed by techniques
such as reflection high energy electron diffraction (RHEED) ,
Auger electron diffraction (AES), transmission electron
microscopy (TEM), scanning electron microscopy (SEM), He-ion
channeling, and x-ray diffraction measurements. Chen, Belmont
and Sebenne reported sub-monolayer adsorption of Ge on the
cleaned Si(111) surface at room temperature using ultraviolet
photoelectron spectroscopy (UPS) [106]. At coverage up to
1/3 of a monolayer the Ge atom binds to three Si surface atoms
replacing Si dangling bonds by Si-Ge bonds plus one Ge
dangling bond. With further Ge evaporation the Ge dangling
bond states continue to develop and Ge-Ge interaction sets
in. Kasper and Herzog reported 8% Ge alloy films of
thickness, 0.1 Jim, were grown on Si (100) at 750C without
misfit dislocations (i.e. pseudomorphic growth) as determined
by x-ray diffraction and TEM [94]. Bean et al. using similar
techniques reported that the map of pseudomorphic growth of
0.1 jim alloy film indicates the substrate temperature for
pseudomorphic growth is related to Ge fraction in the alloy
[101]. Using RHEED Sakamoto et al. reported at a 450C
substrate temperature the critical thickness of pseudomorphic
growth rapidly decreases with increasing Ge [104]. Gossmann,
Feldman, and Gibson reported by AES that Ge starts to grow as
an island from 2-3 monolayers thickness at elevated
temperatures, (i.e. 300C, 520C) [102]. Toropovetal. showed


36
The electron spectrometer of HREELS uses the deflection
of the particle beam (i.e. electron beam) in a electrostatic
field which disperses electrons in energy, so that a narrow
energy band can be filtered by a slit. A schematic diagram
of a spectrometer is shown in Fig. 3. Electrons emitted from
the cathode pass through a repeller to favor forward emission.
A three-element electrostatic lens is used to achieve
focussing into the monochromator entrance. Passing through
the electrostatic field of tandem-127-cylindrical-deflector-
type monochromators, a highly monochromatic (AE5meV) beam is
selected. Before interacting with the sample,
monochromatized electrons pass through accelerating optics.
After interacting with the sample, scattered electrons pass
an almost symmetric path; namely decelerating optics and two
127 cylindrical deflector type analyzers. Finally electrons
are collected by a channeltron which multiplies the number of
electrons into a short current pulse to enhance the intensity
of the signal. To ensure monoenergetic electrons pass through
monochromators or analyzers three connected slit plates are
used at each entrance and exit of capacitors. To prevent
electrons reflected by walls of capacitors, saw tooth type
surfaces are engraved on walls of capacitors.
2.6.1. Basic Theory of 127 Capacitor Electron Optics
The differential equation of motion of an electron in a
cylindrical electric field is


43
e(Usa+C)-hw = 2A/[2ln(R/r)] (2-23)
Since R, r and C are constant, to compensate hu, A or Usa must
be changed. For HREELS, the AE constant mode which changes
Usa is preferred. So one moves the whole spectrum over a fixed
transmission window and the electron optical system between
the analyzer and the target acts as a zoom lens device.
2.6.3. Intensity Angular Profile
The acceptance angle of an analyzer can be detected by
measuring the intensity variation of elastic peak through the
variation of incident angle of the monochromator. For an
Si(111)-7x7 surface, the reflected elastic peak intensity
variation is shown in Fig. 4. Full width at half maximum
(FWHM) is 1.95. But the FWHM of the elastic peak intensity
variation of straight through mode (i.e. without sample) is
-1.3. This indicates that a portion of the reflected beam
width is due to residual disorder of the sample surface.
This may also due to a de-focussing of the reflected electrons
by non-uniform fields such as contact potential fields of the
sample holder components. Another variation for the elastic
peak can be derived from the primary energy. Since the
reflectivity from the sample can be varied by the primary
incident electron energy, the best resolution can be achieved
for different incident electron energies when the sample


Int ens ity
248
Figure 46. RBS data from 1000 of Ge/Si alloy film.


251
Area 1 As Introduced
(a)
Area 2 As Introduced
9
9
0)
V
u5
*
*4
u
v t
2
1
t
Figure 48. Sputter-AES of 1000 of Ge/Si alloy film.
(a) AES spectra obtained from area 1 as introduced;
(b) AES spectra obtained from area 2 as introduced;
(c) AES spectra obtained from area 1 after 30 sec.
sputtering; (d) AES spectra obtained from area 2
after 30 sec. sputtering.


42
v<£=v0, vz- (time of flight)=h,

therefore v2=h/t=(h- v^,)/ (<*> r0) A kinetic energy of electrons
is
(1/ 2) mv2=m (VQ+vz2) / 2
= (l/2)mv02[l+{h2/ (0J r0J) } ] (2-21)
So the total energy width for 127 cylindrical deflection
analyzer is
AEfw/E = (2/r0)S + (4/3) a^2 + h2/(27r2r02) (2-22)
From above results, two terms of geometric deviation of beam
profile add intensity to high energy side. Especially due to
space charge, a monochromator system will contain too many
electrons with higher energy than the pass energy. But the
analyzer acts in an opposite way to the monochromator, namely
a folding process with the analyzer transmission window, the
resulting profile at the detector will be symmetric unless the
beam is not disturbed by targets. Most of the energy
broadening can be understood as an effective target
disturbance which tends to increase the angular divergence.
Electrons which lose energies from the target surface
and enter the analyzer with a kinetic energy of e(Usa+C)-hu)
will pass the analyzer when the following condition is
satisfied (see eguation 2-15),


31
2.4.2. Specific Examples
In cleaning an Si(111) substrate, along with the
techniques listed in the previous section, laser annealing
[16] and galliation [17] in which the substrate is exposed to
a gallium vapor beam at about 800C substrate temperature in
UHV can be used. Galliation is known to be effective to
remove silicon oxide, but inward Ga-diffusion or removal of
Ga atoms by evaporation has not been confirmed.
The chemical treatment of an Si surface consists of two
steps. The first is stripping the native oxide with HF acid
and the second is applying a wet oxidizing chemical treatment
like the RCA cleaning method [18]. But the oxide produced by
this technique is relatively thick, such that thermal
annealing at relatively low temperature (1000-1100 K) can not
desorb oxygen species from the surface. To avoid high
temperature annealing, it is necessary to prepare a very thin
passive oxide layer which can be desorbed at a relatively low
temperature (-1000 K) at which defects do not generate or the
impurity profile does not change. This is the basic idea of
the Shiraki cleaning technique[10] of which the procedure is
listed in section 2.8.. The experimental results of this
sample using HREELS are reported in chapter 5. The
temperature of oxide of surface starts to desorb was 850C,
but at 950C the oxide layer totally desorbs. The LEED
pattern is very clean (i.e. most of the fractional order beams


120
Semi-insulating GaAs(lOO)
Energy (meV)
Figure 17. ARHREELS spectra obtained from semi-insulating
GaAs(lOO) for specular (A0=O) and non-specular
scattering.


208
7.5. Summary
Ge0 5Si0 5 alloy film growth on the hot (580C) substrate
follows the Stranski-Krastanov growth mode. The 5x5 LEED
pattern originated from Stranski layer (the layer between Ge-
rich islands) shows a very similar reconstruction to that of
an Si(111)7x7 surface. In HREELS experiments,
Ge0 5Si0 5(111) 5x5 surface (-200) also shows an adatom
vibrational mode and a quasi-elastic peak broadening. This
indicates a Ge/Si alloy surface is also metallic mainly due
to dangling bonds of adatom. Hydrogen titration to quench
this dangling bonds on a Ge0 5Si0 5(111)5x5 surface reveals that
the alloy surface is Ge-rich. Even if it is hard to identify
the origin of Ge-H mode, considering the total area of the
island is much smaller than that of the alloy layer at this
thickness (-200), Ge0 5Si0 5 (111) 5x5 surface has much portion of
Ge adatom whose dangling bond reacts with hydrogen atoms
during the titration procedures.
Surface relaxation of Ge0 5Si0 5(111) 5x5 alloy film grown
by the thermal evaporation using Knudsen-cell type evaporator
shows continuous relaxation of alloy film beyond -150 of film
thickness. Beyond 600 of film thickness, 5x5 and 7x7
patterns start to overlap and 7x7 LEED pattern becomes
dominant as the film becomes thick. Since there is an
apparent difference between the lattice constant of Si(111)7x7
surface and that of alloy 7x7 surface, the relaxed 7x7 pattern
from alloy layer is thought to be due to Ge-rich islands.


185
film deposited on the Si(111) surface, the lattice mismatch
is accommodated by a large strain of the Si substrate. The
growth of SiGe,_x alloy on Si substrate has been reported by
many authors using different techniques [94-105]. Initially
the SixGe,_x alloy film grows with the same lattice constant as
the Si substrate (i.e. pseudomorphic growth) and recovers its
own lattice constant after it becomes thick enough to allow
misfit dislocations to be introduced. With the right
combination of GexSi,.x composition and in the thickness limit
of pseudomorphic growth, Si layers and SixGe,.x alloy layers can
be used as a superlattice [101]. It has been reported that
the indirect bandgap of GexSi,.x strained layers on Si (001) have
confirmed the anticipated lowering of the indirect bandgap of
these alloys into 1.3-1.55 /im range (i.e. long wavelength
photodetector)[93]. Research related to the surfaces of
SixGe,.x alloy film from growth mechanism to film analysis has
already begun and the major results can be summarized as
follows:
Evaporated Ge on Si(111) surface can be intermixed to
make alloy films through post-thermal annealing and on a hot
Si(111) substrate, evaporated Ge can make alloy films through
intermixing at the interface. Both of results were reported
by many authors using various surface techniques[95-97,99-
105]. The growth mode of this alloy film is of Stranski-
Krastanov (SK) type which indicates that initially, to 2-3
monolayers, epitaxial and pseudomorphic films can be formed,


195
(0 = 2ir ( k / /i )* (7-1)
where k is the force constant between two atoms and m is the
reduced mass. The reduced mass n is 14 for Si-Si and n is 20
for Si-Ge. Assuming the force constant between Si and Si is
same as that between Si and Ge, the vibrational frequency
ratio of usl_sl to ui^c, is 1.2 (i.e., (20/14)'). From Raman
scattering data the frequency ratio is also 1.2 (i.e.,
62meV/52meV). The assumption that force constants are same
is correct in the bulk. For the surface, adatom vibrational
frequency is 30meV for Si(111)7x7 [refer to section 4.2. and
reference 31] and 23meV for Ge 5Si0 5(111) 5x5. The ratio of
adatom vibrational frequency is 1.3 (i.e., 30/23). The result
indicates that the combination of Ge and Si is dominant at the
surface assuming the force constants are similar. Also the
higher frequency ratio indicates a surface force constant
between atoms changes slightly from the bulk value or there
is more screening on 5x5 alloy surface than Si(111)7x7
surface.
7.3.2.Hvdroaen Titration
Since the mass of hydrogen-related vibrational mode has
a relatively large energy loss. Fundamental frequencies of


191
of hydrogen atoms is estimated (refer to section 2.8.) The
LEED pattern is monitored from the hydrogen titrated sample
after 2.5L effective H-atom exposure and shows still 5x5
periodicity.
For surface relaxation experiments Ge source is heated
inside alumina crucible surrounded by tungsten coil and
evaporates slowly with a rate of 1ML/5 min.. At this time
Si(111) substrate is held at 560C to induce intermixing of
Ge molecules on Si(111) surface. Calculation of an effective
thickness of Ge0 5Si0 5(111) film is shown in Appendix C. At
each step of 50 evaporation, lateral lattice constant is
monitored through digital LEED measurement.
7.3. Surface Excitations of SinGen. Alloy Films
bv HREELS.
7.3.1. Adatom Vibration
The SixGe,.x alloy film which was prepared on Si (111)
surface by e-beam evaporation and post-annealing at 600C, and
the HREELS spectrum obtained from the surface with 5x5 LEED
pattern is shown in Fig. 32. The similar LEED pattern was
shown from another Si05Ge05 alloy film which was grown on hot
(~580C) substrate. The angle resolved HREELS spectra, from
specular to off-specular up to 6, are shown in Fig. 33. At
specular geometry, the full width at half maximum (FWHM) is
15meV which means a broadened elastic peak. The cause of the


20
(5) Baking takes ~20 hours at 200C. During this bakeout time
period the pressure increases to a maximum of p~10'6 torr and
then decreases slowly over 10-15 hours to p-5xl0* torr.
(6) After cooling for several hours, the ion-pump is switched
on and most of the filaments in the chamber including the
titanium sublimation pump are outgassed. Briefly heating the
titanium sublimation pump ~2 times will reduce the pressure
to -1x10'' torr after the chamber reaches room temperature.
(7) If the liquid nitrogen cold trap is filled, the pressure
can be lowered by a factor of two or three into the 10'" torr
range.
The most important procedure discussed above is the
bakeout procedure since the proper length of the bakeout time
and temperature will determine the final pressure. Also
baking reduces the total time of desorption from the surface
of chamber walls and diffusion from inside surfaces of chamber
walls. Design of vacuum systems including chamber and pumps
depends upon the kind of experiments to be performed. The
total pumping capacity is the main factor to determining the
efficiency of the system. The order of design of UHV system
is the determination of usage of chamber, determination of the
available accessories, estimation of gas flow rate and
determination of pumps. In our system a Leybold-Heraeus
(model TRIVAC-A dual stage D16A) mechanical pump, with
capacity of 400 1/s and a Leybold-Heraeus (model Turbovac 150)
turbo molecular pump with capacity of 150 1/s are used. It


Subsequent, high temperature annealing above 950C induces a
7x7 reconstructed surface from which a surface phonon due to
adatom vibrations has been detected. Surface plasmons due to
sputter damage and Fuchs-Kliewer phonons have been detected
from sputter-annealed GaAs(lOO) surfaces. The phonon-plasmon
coupling effect has been observed experimentally and a
dispersion curve has been measured by angle resolved HREELS
(ARHREELS) It has been found from oxidation of Ni(110) and
Si(111) surfaces under ultrahigh vacuum that energies of
oxygen-related surface vibrational modes increase as the
thickness of the oxide layer increase in agreement with
previous empirical correlations. This indicates that the ionic
character of the oxide layer increases with increase of the
thickness. Protective oxide layers on Si(111) surfaces have
thinner oxide layers on starting surfaces that are atomically
smooth as compared to typical native oxide layers and do not
have many active sites to react with residual gas. Also, this
thin oxide layer requires a much lower temperature annealing
without sputtering to be removed. In the study of
Ge0 5Si0 ,(111) 5x5 surfaces, it has been found that this alloy
surface has similar adatom vibrational modes to the Si(111)7x7
surface using ARHREELS and these adatoms are mostly (-90%) Ge
atoms using hydrogen titration. For Ge0 ,Si0 ,(111) a Stranski-
Krastanov growth mode has been detected and continuous surface
relaxation of the strained overlayer has been found as the
film thickness increases in the range 0-1000.
xiv


157
6.2. Experimental Results
Using the Shiraki method (refer to section 2.8). Si(lll)
(Boron Doped, p-type and p= 2.4ii-cm) with a thin oxide layer
is introduced into the chamber and pumped to a pressure below
3xlO,0torr. The HREELS spectrum obtained from this sample is
shown in Fig. 25(a) The LEED pattern shows a lxl pattern
with three fold-symmetry only for high incident electron
energy (>180eV). No AES spectrum has not been obtained for
this sample in order not to induce any hydrocarbon species on
the sample surface. Then the sample is subsequently annealed
at 500C for 10 minutes without sputtering. Still the same
lxl LEED pattern is shown. The HREELS spectrum for this
annealed sample is shown in Fig. 25(b). The similar HREELS
spectrum and LEED pattern are shown after annealing at 800C
for 10 minutes. After annealing at 900C for 15 minutes the
clean 7x7 LEED pattern with most of the 49 spots in a unit
mesh are clearly shown. The HREELS spectrum of this sample
is shown in Fig. 25(c) and is essentially the same as that of
a clean sputtered and annealed Si(111)7x7 surface (described
in chapter 4).
The Shiraki-cleaned sample after preserving in deionized
water for three weeks was introduced into the chamber. The
HREELS spectra from this sample is shown in Fig. 26 fa). The
LEED pattern for this sample has no pattern even at high
incident electron energies indicating a thicker oxide layer


CHAPTER 7
SURFACE RELAXATION AND SURFACE EXCITATION OF Si/Ge ALLOY
FILMS
7.1. Overview and Motivation
Since the development new growth techniques of thin film
growth, thin film novel materials have opened new eras in
device fabrication as well as physics. Techniques such as
Molecular Beam Epitaxy (MBE) enables one to grow films layer
by layer. Optical properties due to differences of band gaps
of heterostructure of two different semiconductors grown by
MBE can be applied to devices such as photodetectors and field
effect transistors[92,93]. But the combination of
semiconductors was usually limited to the direct bandgap
materials, mainly III-V compound, lattice matched
semiconductors. Since the lattice matching reduces the defect
sites at the interface due to dangling bonds of unmatched
atoms, the carriers can be confined within one layer with high
mobility without being trapped or scattered by these defect
sites. But silicon, an indirect bandgap semiconductor in
Group IV, which the present industry based upon, has 2%
lattice mismatch relative to an Si05Ge0S alloy and 4% lattice
mismatch relative to Ge. In the case of thin SixGe,.x alloys
184


46
A schematic diagram of interfacing is shown in Fig. 5.
A Kepco 488-122 programmable power supply, with a resolution
of 2.442xl0'4 volts per step, has been placed in series with
the ramp potentiometer in the HREELS power supply. The Kepco
power supply which is used to set the pass energy on the
analyzer has a resolution of 1/4096 with a maximum output
voltage scale of one or ten volts. Then the ramp pot which
was originally used for scanning can be used for detecting the
elastic peak and monitoring the FWHM of elastic peak during
tuning. In the same manner another Kepco power supply is
connected to the HV amplifier sweep for ramping of
hemispherical electron analyzer voltage for XPS, UPS and AES.
The output of channeltron detector is fed through an
Ortec 109PC preamplifier to an Ortec 572 amplifier and finally
to an Ortec 550 single channel analyzer set to transmit pulses
only if their amplitude is above a preset threshold. This
signal is fed either to a ratemeter for tuning or to CAMAC
counting electronics. The CAMAC counting electronics consist
of 3610 Hex scaler and 3655 timing generator. A second
channel on the hex scaler is used to collect data from the TTL
output of hemispherical electron analyzer electronics (Leybold
Heraeus model LH-100) to collect a pulse counting signal from
electron analyzer due to AES, UPS and XPS.
One of the important achievements of interfacing with
the computer is in the precise reproducibility of the incident
angle of the monochromator. A Klinger stepping motor


70
theory for semi-infinite medium) will be introduced [30].
Even if impact scattering theory is not fully developed yet,
the important characteristics of impact scattering have been
reported. This impact scattering will be briefly discussed
and finally examples and applications of inelastic scattering
theory to specific systems will be presented.
3.3.1. Semi-Classical Approach
The electric field due to the specimen extends over the
vacuum above the specimen. The longest range electric field
is due to the surface dipole of the specimen. Considering
the total range of this dipole field above the specimen, the
time which the electron stays within this range is longer than
the time it stays inside the specimen. This long range
interaction is the so called 'dipole scattering' which is
applied to specular geometry since the scattering intensity
is sharply peaked around specular geometry. The scattered
intensity of this dipole scattering applied to the system of
an adsorbate on a metal surface will be derived following the
description by Newns [28]. This theoretical approach to
dipole scattering used here is the semi-classical one, since
it treats an incident electron as a point particle traveling
along a single trajectory remaining in the vacuum above the
specimen at all times and exciting surface vibrations by means
of its long-range coulomb field. The schematic diagram of the
scattering is shown in Fig. 9(a). The coordinate of the
electron is


5
difficult to produce highly monoenergetic beams of neutral
atoms. Since the maximum incident energy is usually limited
to ~60meV, this technigue is useful only for obtaining surface
vibrational modes below this incident energy.
The technique we have used, HREELS, covers a wide range
of vibrational energies (5-500 meV) which in principle is also
covered by IR spectroscopy. HREELS is useful for probing
surface phonon dispersion, surface plasmon excitation and
intra- and interband transitions. The high surface
sensitivity of HREELS enables one to detect 0.001 ML coverage
at the surface if the species has a typical dipole scattering
cross section (~10',6cm2) Therefore many surface impurities
can be readily detected. Due to the low incident electron
energy, HREELS is a non-destructive technique which can be
used to probe weakly bound species such as atomic hydrogen on
graphite. In HREELS the scattering occurs due to the coupling
between the incident electron and the electric field of charge
fluctuations under the surface. This long-ranged interaction
of the Coulomb field makes it possible to detect interfacial
modes under the surface as long as the interface contains
strong dipole species and is within the detection limit (-200-
500 ) Therefore HREELS is a very surface and interface
sensitive technique with a wide energy range. However, it
also has several limitations at the present time. Even with
the most advanced electron optics, the resolution in HREELS
is about 3 meV. This limits the analysis of closely spaced


260
One is that since the Ge island has high concentration of Ge,
removal of Ge island on the surface reduces the Ge level
rapidly. The other is that since the plateau area is wide
compared to Ge island, sputtering of this plateau area of
relatively thin Si.Ge,., alloy film decreases the Ge level
rapidly. It is not easy to determine at this point. Another
question is why is there a slow decay of Ge in band 3 ? One
possibility is that due to the thick Ge island, a small trace
of Ge island remains even after alloy film was removed.
Another possibility is thermal diffusion of Ge species into
the Si bulk. It is also not easy to determine at this point
whether there is diffused species as long as the Ge island
exists. Also it is impossible to estimate the thickness of
each band using a line scan of edge of sputter crater, since
the sputtering angle is quite glancing and the band width is
different position by position. Time-sputtered depth profile
shown in Fig. 50(e) was taken from the different area of the
same sample. Calibrating sputtering efficiency, sputtering
rate was 100 A/min. Since the lateral area of the island is
much smaller than the plateau area at the surface, the end of
the rapid decrease of Ge (~40 A) indicate the thickness of the
alloy. Also the long tail of Ge extends to 250 A of
thickness. The plateau area (area between islands) shown in
Fig. 48(a) had a rapid reduction of intensity of Ge(1148)
after 30 sec sputtering. Although, from the Ge island, Ge has
still the larger intensity after 5min more sputtering. so


174
Additional annealing at 800C for -lOrnin caused almost
all hydrogen species to desorb. The FWHM of the elastic peak
is reduced from 16.3meV to 15.3meV. This is due to hydrogen
desorption since adsorbed hydrogen species induces so called
a spurious ohmic conductance effect [62]. But the oxide layer
does not change; the LEED pattern is still lxl periodicity at
high incident energy and the 141meV Si-O-Si asymmetric stretch
mode does not change.
Final annealing at 900C for 15 minutes induces a clean
7x7 LEED pattern such that most of the 49 beams in a unit mesh
appear clearly at 36eV. The HREELS spectrum is shown in Fig.
25(c) with all the oxygen related peaks decreased to the noise
level. Contamination after annealing is not detected except
for very small traces with loss energy near HOmeV, which
indicates that this surface does not have many defect sites.
The tiny hump near HOmeV maybe due to bulk carbon impurity
which segregated to the surface during the annealing
procedure. The FWHM is still 15.8meV even though the oxide
layer has disappeared. Instead an of oxide layer effect this
broadening is due to dangling bonds of a clean 7x7 surface
[90].
The HREELS spectrum obtained from the sample preserved
in deionized water is shown in Fig. 26(a). The FWHM of the
quasielastic peak is 22.5meV and no LEED pattern is found from
this surface. These results indicate that the surface is very
disordered. The asymmetric stretch mode appears at 145meV.


196
the Si-H diatomic molecule and the Ge-H diatomic molecule are
244meV and 227meV respectively [33]. In case of hydrogen
adsorbed on an Si surface, the stretching mode appears at 258-
260meV depending upon the adsorbing configuration and the
direction of the substrate [113,114], The stretching mode of
adsorbed deuterium on Si is also reported at 189-190 meV which
is 7 meV larger the 2" factor, the mass factor between
deuterium and hydrogen. For the Ge(100) substrate, Papagno
et al. reported Ge-H stretch mode at 178meV which is also 6meV
larger than the 2* mass factor [115]. In addition to the
stretch mode, the bending mode and the dihydride scissor mode
of these two systems were listed in Table 1. The Si(111)
sample (p-type, Boron doped, p =10-20 D-cm) was annealed at
1010C for three minutes which induced clear a 7x7 surface.
About 2 00 A of Ge was evaporated on the room temperature
sample. After annealing at 600C for four minutes, the 5x5
LEED pattern appeared. The HREELS spectrum obtained from this
surface is shown in Fig. 32. The peak at 107 meV is due to
carbon contamination of the surface which makes a stable
carbide at the surface after annealing. First, adsorbing 2.5L
hydrogen still shows a slightly degraded 5x5 LEED pattern.
The HREELS spectrum of this surface is shown in Fig. 34 fal.
Two peaks at 3 62 meV and -170 meV are due to hydrocarbon
species adsorbed during the adsorption procedure.
Unfortunately, the broad peak near 104 meV made it impossible
to detect any SiH2 loss feature. Since 5x5 LEED pattern


212
similar to the Si(111)7x7 surface. This adatom species was
shown to be mostly germanium through titration of hydrogen and
HREELS.
8.2. Recommendations for Future HREELS Studies
Future research using HREELS can be divided into two
parts. First, improvement of the spectrometer will solve the
current resolution limit to a certain degree. Still the
electron spectrometer is lower in resolution compared to Raman
or IR photon spectroscopy. Second design of a novel
spectrometer which can change the scattering geometry to cover
larger momentum transfer will open new opportunities for
HREELS in angle resolved spectroscopy. This will give a
chance to complete the surface phonon dispersion of single
crystal systems to cover the whole first and second Brillouin
zones. In short, improvement of spectrometers can open new
research areas. The subjects of HREELS studies will increase
from simple geometry and simple adsorbates to many materials
such as non-crystalline surfaces and magnetic surfaces with
spin ordering. Since HREELS can see to a depth of -200 on
non-metal surfaces, one can study interface modes for rather
standard overlayer systems such as doped, quantum well
structures.
Finally the future research areas derived from this
thesis may be as follows: (1) Semiconductor surface relaxation


101
E> Q g
GJ_
(b)
Figure 11. Two layer mode dispersion and polarization.
(a) Surface eigenmode dispersion for a slab of
material with dielectric function es(w) on a
substrate with a dielectric constant, cb(>0);
(b) The electric fields for ut mode.


233
are selected. Either defining a starting and ending analyzer
potential or defining a starting potential and sweep energy
range will set the sweep energy range with respect to a
present setting of pass energy at the HREELS power supply.
Defining the number of points with maximum of 1024 points will
determine the energy step with the automatic calculation
referring to a previous energy range. With an option of angle
resolved HREELS, initial angle, final angle and step of angle
are determined in the same way as above. The total number of
scans to be averaged together and the time per point are
selected. With automatic calculation, estimations of each
run-time and total run-time are displayed. During data
acguisition, an x-y plot of data is shown on the monitor
screen with a numerical output of both analyzer potential and
counts. In case the count exceeds the expected count which
was chosen in the run parameter setting, y-axis is rescaled
automatically. Individual data scans can be viewed and the
average of all scans is recorded in the disk file at the end
of the scan. With the option of 'Test Run', the data can be
manually saved with the 'save' command.
Another important purposes of computer controlled data
acguisition is the handling of data. Data stored in disk file
can be retrieved by the program '2D plot' written in PASCAL,
which has various options to plot on the monitor screen or on
a piece of paper using a plotter. Since the analyzer
potential starts from zero in data acquisition, it is


80
with wavevector Q|(. The factor e0*1 z in the second integral (of
equation (3-31)) which is the charge source integral indicates
that the contribution to the potential with component
exp (iQ|| X||) is produced by charge fluctuations that extend down
to a distance Qn'1 below the surface of the specimen.
The differential cross section can be obtained by
inserting equation (3-31) into the Schrdinger equation and
using Born approximation,
da 2 | R |2 v/ k' S (Q|(, id)
dn, dhu hk7r(ea0)2 cos, [vx2 Q,,2 + (w-Q(| vH)2 ]2
(3-33)
The term a0 is the Bohr radius, is the angle of incidence
relative to the surface normal, k and k' are the magnitudes
of the incident and scattered wave vectors, | R|2 is the
reflectivity for specular scattering, and Q,|= kn-JCji'. The
spectral function S(Q||,w) is
S(Q||,U) =
00
d2X|,
dt exp [ iQ||-X||-iwt ]
CO
dz' | dz exp[Q,|(z+z') ] T ,
CO
J
DO
(3-34)


45
surface is changed (e.g. desorption of oxygen or vice versa).
From Si(111) with a native oxide layer after annealing at
920C for 3min under UHV, the measured resolution (i.e. FWHM
of elastic peak) and intensity of elastic peaks for primary
energies such as 5eV, lOeV, 15eV and 20 eV are as follows: (1)
resolution is 12.1meV, 13meV, 7.5meV and 9.2meV respectively
and relative intensity is 310, 360, 240 and 180 respectively.
2.7. Computer Interface
The high resolution electron energy loss spectrometer
has been interfaced to an IBM pc/XT using IEEE-488 and Camac
crate buses. The main purpose of computer controlled data
acguisition is increased reproducibility of HREELS energy
settings and reduction of noise by signal averaging. One of
the limitations of HREELS which can be overcome by interfacing
is in the scanning of pass energy of the analyzer via a motor
driven potentiometer (ramp pot). Because of degradation of
contacts, this potentiometer should be replaced regularly (
-6 months) however by using a programmable voltage source the
pot lifetime is extended. In this section, the interface of
the energy loss spectrometer to a personal computer to control
the analyzer potential and incident angle of monochromator
will be discussed. Also the structure of computer programs
and data handling will be discussed in Appendix A.6.


71
re: electron-dipole distance
(V|,,vx): electron velocity
H: molecular instantaneous dipole
and its image in the surface
Mi: resultant normal dipole
(a)
Figure 9. Schematic diagrams of semi-classical dipole
scattering.
(a) Electron trajectory and molecular dipole moment;
(b) Transferred momentum; (c) Polar plot of
scattering intensity.


>
-M
CD
C
CD
4->
C
hH
Energy (meV)
Figure 27.continued.
166


98
dielectric constant of the bulk which lies under the surface
layer, eb(w). Since the scattering geometry is not changed,
this modification only affects spectral density S(Q||,w). In
here, the result of the modification by Mills and Maradudin
will be introduced [38]. Two layer spectral density S2(Q(|,0))
can be derived by replacing c(w) in the spectral density
S(Q||,w), i.e. equation (3-36), by e,(w), where
1+A (w) exp(-2Q||d)
et(w) = es(w)[ ] (3-51)
1-A (w) exp (-2Qd)
where
[eb(w)-es(w) ]
A(w) = and N(u) = [exp(hu/kT)-l]'. (3-52)
[eb(w)+es(w) ]
If d-0,
et(w) = ,(w) [{1+A(w) )/{l-A(u) )] = eb(w). (3-53)
If Q,|dl,
e'2q,l<, l-2Qd, (3-54)
and the loss function Im[-l/(e, (w)+l) ] can be divided into two
terms,


Energy (meV)
(a)
Figure 27. HREELS spectra obtained from native oxide on Si(lll).
(a) As introduced; (b) After annealing at 520C; (c) After annealing
at 1010C.
165


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ELECTRON SCATTERING STUDIES OF
SURFACE PHONON-PLASMON MODES OF SEMICONDUCTORS
By
Jae Myung Seo
August 1989
Chairman: Professor John E. Rowe
Major Department: Physics
Surface excitations such as surface phonons and surface
plasmons of Si(lll), GaAs(lOO), silicon oxide, nickel oxide
and Ge0 5Si0 5(111) alloys have been studied under ultrahigh
vacuum using high resolution electron energy loss spectroscopy
(HREELS), low electron energy diffraction and Auger electron
spectroscopy. It has been found that the surface free-carrier
plasmons from sputtered and insufficiently annealed p-type
Si(111) surfaces are losses due to a collective motion of
acceptor-type carriers localized in the space charge region.
The measured dispersion curve of these plasmons achieved by
angle resolved HREELS agrees with that predicted by dielectric
function theory using the Lindhard dielectric function and
parameters appropriate to highly doped semiconductors.
xiii


220
be easily checked the concentration formula is just if all
the experimental setting are same as standard (i.e.Dx=l and
K=l) .
A.4. Digital LEED
Useful information can be obtained from the intensity of
diffracted beams as a function of incident electron beam
energy[1,118]. The purpose of digital LEED is to obtain the
quantitative intensity information with better resolution.
To achieve this goal, the LEED beam should not be blocked by
the sample manipulator. The front-view LEED system was
converted to the back-view LEED system.
A schematic diagram of the computer control system which
was used in the experiment for acquisition of two dimensional
intensity data is shown in Fig. 40. The retarding grids and
the beam voltage input of a constant-current LEED electron-
gun controller are connected to the outputs of two separate
computer-controlled digital-to-analog converters (D/A 1 and
D/A 2). The retarding field oscillates between a high value
(a few volts above the beam voltage) and low value (a few
volts below the beam voltage) with the same time interval.
The total intensities for high values of the retarding field
are subtracted from the total intensities for low values and
the difference, which is proportional to the intensity in an
energy window between the high and low values, is accumulated
and stored in the computer memory.


21
takes 5 min from air pressure, to reach one torr which is the
proper pressure to start the turbo molecular pump. The turbo
molecular pump is used during baking procedure and outgassing
of the filaments. Finally an ion-pump (Leybold-Heraeus model
IZ-270 with capacity:270-300 1/s) and a titanium sublimation
pump are used in conjunction with a liquid nitrogen cold trap.
In addition to pumping systems, a gas handling system is
described in Appendix A.l.
2.3. Auger Electron Spectroscopy
Auger electron spectroscopy is a surface analytical
technique for determining the elemental composition of the
outer atomic layers of a solid. It provides high sensitivity,
elemental selectivity for all atoms except hydrogen and helium
and good lateral resolution (10'6-10'7m) of the analyzer which
enables the detection of specific atomic species situated in
the first few atom layers (5-10). In AES, surface
sensitivity is mainly determined by the escape depth of the
observed electron, since the incident electron (Ep = 3-5keV)
penetrates much more deeply into the crystal. The inelastic
mean free path (X^) of electrons as a function of energy (Ep)
shown in Fig. 2. is defined as the distance that an electron
of energy E travels with a probability e'' of not being
scattered by inelastic events that degrade its energy by more
than several electron volts [3]. Therefore, in one type of


237
Substituting equation (B6) into equation (B1), then
integrating over x0 will give the total cross section a for
one-phonon excitation,
a
4 Q2 v2
[ nQ2 + Q2vA2 ]2
(B-8)
The fact that a parallel component of the wave vector Q is
the conserved quantity enables us to infer equation (B-8) can
be also applied to each d2Q Therefore
4 Q2 vA2
da = T2 d2Q (B-9)
[Q2 +Q2vx2]
B.2. Transformation of Coordinates
For the case of normal incidence (i.e. a=0) and 0 (the
rotational angle relative to the specular beam) is not zero,
, 60) (B-10)
cos0 sin0 0
-sin0 cos0 0
0 0 1
where M0=
is the rotation matrix about oz.


68
to electron-electron scattering, and thermal vibrational
motion of the atoms (i.e. the Debye-Waller factor).
3.3. Inelastic Electron Scattering
Two conservation laws are the basic starting point for
inelastic electron scattering theory and are summarized as
AE = E, Es = hw and Ak||= klj(- ksN= q,( + G(,. (3-15)
Here AE is the energy transferred to the crystal and Ak,, is
the momentum transferred parallel to the surface and GN is a
reciprocal lattice vector of the crystal surface. Even if
the vertical component of transferred momentum is large, a
simple relation for it can not be established due to the
broken symmetry normal to the surface. Approximate methods
can be modelled to give periodicity of the z direction using
periodic layers which consist of a large gap of vacuum to
avoid the interaction between the surface and the substrate
layers (-15 layers). Such a model will allow estimation of
scattering near the surface which is quite similar to the
system of vacuum and semi-infinite substrate.
From the law of conservation energy, an inelastically
scattered electron can lose or gain energy depending on
whether a phonon is created or annihilated. The relative
probability of a phonon creation and annihilation is governed


39
V(r,) = [A/ln(R/r) ] -ln(r,/r0) .
(2-13)
Electrons with their kinetic energy equal to central path
energy, and entering the sector field in the direction of the
tangent of the corresponding radius will travel in a circle
when the centrifugal force and the electric force are equal.
The electric force on an electron using equation (2-13) is
eE = eSV(rt)/Srt = {eA/[In (R/r) ] ) {l/rt)
(2-14)
The centrifugal force is mv02/rt. Equating these two forces,
the kinetic energy of an electron passing through the slit is
(1/2)mv02 = Ek,n = eA/[2ln(R/r) ] ,
(2-15)
which is same as e(C + Us) where C is contact potential
between the cathode and the slit and Us is the slit potential.
From this relation, the capacitor potential A is linearly
proportional to the slit potential Us. In the Leybold-Heraeus
ELS-22 spectrometer, R = 41mm and r = 31mm for the main
capacitors, so the theoretical slope will be 0.56 (i.e., 21n
(41/31)) when plotting of A (of main elements) versus Us.
If the potential of the main path (i.e. V(r0)) is
different from that of the slit potential, then electrons that
enter the sector field will have to cross a potential step
which causes angular distortion and decreasing output


110
4.2. Experimental Results
Angle resolved high-resolution energy-loss spectra from
Si (111) surface with the bulk doping density of Nb=lxlO'5 cm'3
are shown in Fig. 13. The sample is sputtered (Ar =
5. OxlO's torr, Vb(!a =1.0KV, emission current=25mA, sample
current=2.5/iA for 40min) and annealed at 800C for 15min. A
shoulder near lOmeV from the quasielastic peak can be seen in
the A6 =6 spectrum, and its energy increased to ~15mV when
A0 =18. The energy of this peak remains at 15mV until the
intensity becomes very small at 25.5 off-specular. For this
surface visual LEED pattern observations showed an apparent
lxl pattern, but the intensity of the elastic peak and the
15meV loss peak in our HREELS data increased when the momentum
transfer matched each of the seventh-order beams (1/7, 2/7,
3/7) which is consistent with a 7x7 periodicity. This
indicates that HREELS intensities are more sensitive than
visual LEED in detecting reconstruction of a long-range
ordered surface. When samples with a higher bulk doping
density (Nb =5.8xl0,s cm"3) were used, the peak in Fig. 13
appeared as a weak shoulder on the quasielastic peak. When
the highest doping (Nb =5.5x10" cm"3) sample was used, this peak
was broadened into the elastic peak, even at off-specular
directions and a detailed dispersion measurement could not be
obtained.


249
shows Ge is not evenly dispersed through the Ge05Si05 alloy
film. At the top a higher Ge fraction is shown compared to
the interface. Since the higher mass Ge scattered He back at
a higher energy, the independent peak at the right hand side
indicates the Ge is spread through the alloy film. The kink
in the left peak is the position of interface. This may
indicate that the thermal mixing technique (i.e. evaporation
of Ge on the hot substrate) is not able to produce a sharp
interface, only a smoothly varying Ge composition. The
thickness, estimated at the center of the Ge peak, is ~850.
At the bottom of Ge peak, the thickness is -1400.
Composition of Ge at the top is -60% and is smoothly reduces
as the thickness increases. To detect the surface morphology
and the depth profile of the film, the sample is transferred
to the Scanning Auger Microprobe Analysis chamber. The
morphology of the surface without any treatment is shown in
Fig. 47. This is not the area which was not radiated by 2 MeV
helium beam. Two distinct areas are shown. From the area 1
(grey color), AES shows Si(LMM: 91eV, KLL: 1620eV) peaks and
Ge(LMM: 51eV, KLL: 1150eV) peaks in Fig. 48(a). The peak to
peak ratio of Si(1620)/Ge(1150) is almost 1, which means Ge
is well intermixed with Si and they make an almost 50%-50%
alloy. Peaks such as C(275) and 0(511) are due to air
contamination. From the area 2 (white color) the peak to
peak AES intensity ratio of Si(1615)/Ge(1150) from Fig. 48 fb^.
is 0.65, which means Ge is richer than the area 1. After


350
Energy (meV)
(b)
Figure 35.continued.
202


Energy (meV)
(a)
Figure 26. HREELS spectra obtained from water-preserved oxide on Si(lll).
(a) As introduced after preserving in deionized water for three
weeks; (b) After annealing at 500C; (c) After annealing at 900C.
162


8
protective oxide layers can be used to prepare clean surfaces
under UHV without sputtering and high-temperature annealing.
For two different reasons HREELS is a useful technique to
study silicon oxides. First, oxygen atoms form partially
ionic bonds with Si atoms, and their strong dipole moment
enables one to easily detect the local vibrational modes of
each type of oxide. Second, the high surface sensitivity of
HREELS can easily detect surface impurities present on oxides
and on clean surfaces down to 0.001 ML. These may include
hydrocarbons, hydrides or hydroxyl groups as well as silicon
carbide. Similar oxidation mechanisms with different oxide
local bonding sites are expected for many metal oxide
interfaces.
Alloys of Si and Ge are another interesting area for
surface studies since they have reconstructed surfaces that
are similar to the reconstruction found on clean elemental Ge
or Si surfaces. A Ge0 5Si0 5 alloy grows epitaxially on
Si (111) 7x7 substrates and has a 5x5 LEED pattern. From
surface reconstruction studies by LEED and scanning tunneling
microscopy (STM) both 7x7 and 5x5 surfaces are reconstructed
in a quite similar way with about h monolayer of adatoms on
a dimer-stacking fault unit cell. The vibrational mode due
to adatoms has been reported for Si(111)7x7 using HREELS. The
cross section for Ge-Si alloys may increase due to a non-zero
dipole moment in the Ge-Si system compared to the pure
elements in the bulk. We can compare HREELS results for 5x5


127
and a peak near 180meV is due to deformation mode of CHX.
Near 70meV, a broad feature is due to second harmonic of 36meV
(i.e. 72meV), GaO(~70meV) Ga20(~90meV), AsO(~80meV) and
As304(110meV) From AES, major impurities are due to oxides
and carbon. After annealing at 580C, hydrocarbon species has
desorbed and resulted in better resolution of quasielastic
peak [62]. Oxygen peak does not reduce from AES data. The
strong Ga20 peak at 90meV and the second harmonic (72meV) of
FK mode (36meV) are shown from HREELS data, Fig. 16(b). The
remaining native oxide is mainly Ga20, and other oxide
desorbed at lower temperature (<580C) These results are
matched with other reported results [63,64]. After annealing
at higher temperature (665C) it is shown from HREELS data
that Ga20 has desorbed. A small carbon impurity is detected
from AES data, which does not appear as a peak in HREELS.
This indicate that the remaining carbon does not form a strong
dipole moment with Ga or As. A diffuse LEED pattern was shown
from this surface. This is due to the fact that thermal
annealing temperature is too high to keep the stoichiometry
detected. It is deduced from AES intensity ratio of Ga to As
that approximately 0.1 ML of the surface is As component.
Also carbon impurity can not be totally removed by thermal
annealing at 665C.
For highly-doped GaAs surfaces, a theoretical calculation
of the energy loss of coupled phonon-plasmon modes is shown
in Fig. 10 of chapter 3. An approach similar to that used in


242
Inserting equation (B-22) into equation (B-24), then the
solution is
47rine2 u2(0) [x +iz]
u0(x) exptiQuX-QuZ-iut],
M,(1+6) (aT02 a2)
(B-25)
when z = 0, the self-consistent relation
47rne'2 uoz(0)
uOz(0)=
Mr (I+600) (a2 -aT02 ) (B-26)
gives the Fuchs-Kliewer mode
47rne'2
a2 = as2 = aT02 +
M, (I+600) (B-27)
This can be expressed in terms of e(0) (refer equation (3-
42)), and 6 as follow
e(0) + 1
ws = wT0 ( )* ,
6 +1 (B-28)
which is the same result as equation (3-44).
The two approaches are identical in physical content.
'Fuchs-Kliewer mode' due to lattice dynamical theory and


229
Table 2, HREELS Tuning Examples.
Sample
Semi-insulating
GaAs(100)-new
Semi-insulating
GaAs(100)-old
Primary Energy (eV)
5.000
5.054
External Voltage (V)
4,890
4.981
Electron Gun
Filament (10"'A)
2.00
2.00
(V)
Repeller
-0.116
-1.298
Anode 1
-4.246
-5.940
Asymmetry 1
-0.205
0.313
Anode 2
22.50
19.00
Asymmetry 2
3.431
-0.735
Anode 3
-4.361
-4.103
Asymmetry 3
0.102
-0.043
Acceleration
Electrode 1
-0.570
-1.275
Optics
Asymmetry 1
0.165
-0.010
(V)
Electrode 2
1.174
-0.604
Asymmetry 2
0.040
0.008
Deceleration
Electrode 3
0.480
0.057
Optics
Electrode 4
-0.193
-1.361
(V)
Asymmetry 4
0.156
-0.011
Zoom Knob
Electrode 3
7.20
7.50
Position
Electrode 4
7.14
7.50
Monochro-
Contact potential -0.004
-0.005
mator
Pass energy Slit
-0.270
0.042
(V)
Precapacitor rpm
0.120
0.271
Rp*
-0.118
-0.257
rPB-RpB
0.237
0.527
Maincapacitor r^
0.115
0.267
Rim
-0.182
-0.418
r-R
0.297
0.683
Ramp position
7.28
4.13
Analyzer
Contact potential -0.046
-0.329
(V)
Pass energy Slit
-0.305
-0.488
Maincapacitor rm
0.118
0.158
R*
-0.181
-0.246
r-R*
0.298
0.402
2ndcapacitor rsa
0.100
-0.213
RSa
-0.180
-0.438
rsa-Rsa
0.279
0.225
Count rate (/sec) 1.2xl03 7.4xl03
Resolution (meV) 15.0 14.5


37
High Resolution Electron
Energy Loss Spectrometer
Sideview Toward Sample Manipulator
Figure 3. HREELS spectrometer.


CHAPTER 5
VIBRATIONAL MODES FROM OXIDE LAYERS ON Ni(lll) AND Ni(110)
5.1. Overview and Motivation
Oxidation of a nickel surface has been one of the famous
oxidation systems using various surface physics experimental
techniques. It is well known that the oxidation reaction at
room temperature can be divided into the following three
regimes: which are dissociative chemisorption, oxide
nucleation and growth to coalescence, and thickening of this
coalesced oxide layer up to saturation level thickness [8,67-
69]. The first two regimes are well known, but the third
regime is still controversial since there are relatively many
experimental factors to control. Quantitative evaluation
results from many different experimental techniques is very
difficult since the preparation of this third species is not
well defined. One of these controversial points is in
identification of high-binding energy oxygen species from x-
ray photoelectron spectroscopy (XPS) data (e.g. whether it is
an OH species which exists as an adsorbed species of a hydro
oxide layer or it is not OH-related at all but a more complex
Ni-0 species) [70-72]. If this species is identified, it will
133


9
and 7x7 surfaces and thus attempt to determine if the bulk
composition is present at the surface. Another advantage of
HREELS is the fact that the vibrational energy depends upon
the reduced mass of the local oscillator. By titrating H
atoms on the 5x5 surface the relative numbers of Si and Ge
atoms in this surface can be determined since surface Si-H or
Ge-H species have different reduced masses and the
corresponding stretching modes have an energy difference of
~10 meV. We have found that a large fraction of the 5x5
surface (-90%) covered with H contains Ge atoms rather than
a bulk 0.5-0.5 mixture. This is likely due to Ge having a
lower surface energy than Si on this alloy surface.
Summarizing, High Resolution Electron Energy Loss
Spectroscopy is a highly surface sensitive and quantitative
vibrational spectroscopic technique. Through novel
applications of this technique (i.e. Angle Resolved HREELS),
surface phonon and surface plasmon modes were investigated
from the semiconductor surfaces such as (1) the clean Si(111)
surface (2) the oxide surface of Ni and Si and (3) the Ge/Si
alloy surface.
Chapter 2 describes broadly the experimental procedures
following the typical sequential order of experimental
measurements. Therefore the same experimental procedures are
not repeated at the chapters (4-7) unless some new specific
procedure was used. In addition to the procedures used to
collect the data, a brief review of equipment and its


205
1000A
400A
Si(lll)7x7^J \J\sJ\jJ\J\)
Figure 36. LEED intensity profile of Ge0 5Si0 5(111)-5x5
measured by digital LEED.


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
4.
Pauf H. Holloway
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
David B. Tanner
Professor of Physics
This dissertation was submitted to the Graduate Faculty
of the Department of Physics in the College of Liberal Arts
and Sciences and to the Graduate School and was accepted as
partial fulfillment of the requirements for the degree of
Doctor of Philosophy.
August 1989
Dean, Graduate School


12
13
14
15
16
17
18
19
20
21
Polarization of surface inodes. The displacements in
the surface layer for the three surface modes at
the x point of the two-dimensional Brillouin zone
for an fee crystal with nearest-neighbor central
force interactions and a (100) surface. Modes S6
and S, involve displacements mainly parallel to the
surface while mode S4 involves perpendicular
displacements.
104
ARHREELS spectra obtained from partially ordered
Si(111)lxl for non-specular (L8?0) scattering.
Ill
ARHREELS spectra obtained from well ordered
Si(111)7x7.
113
Dispersion curves of Si(lll) surfaces 115
HREELS spectra obtained from a native oxide layer
on GaAs(lOO). (a) As introduced; (b) Annealed at
580C; (c) Annealed at 665C.
116
ARHREELS spectra obtained from semi-insulating
GaAs(lOO) for specular (A0=O) and non-specular
scattering.
120
ARHREELS spectra obtained from highly doped
GaAs(100) for specular (A0=O) and non-specular
scattering.
121
Experimental data and calculated dispersion curves
for highly doped GaAs(100).
122
HREELS spectra for specular scattering obtained from
a Ni(111)-p(2x2)-O surface with 0.25 monolayer of
chemisorbed oxygen.
136
ARHREELS spectra obtained from Ni(111)-p(2x2)-O.
(a) 1.5 off-specular; (b) 3 off-specular; (c) 4.5
off-specular.
137
viii


155
oxidation can be controlled under UHV conditions. Observation
of the evolution of three oxygen-related peaks in HREELS may
give information on the structural change of the intermediate
oxide as the thickness of the oxide changes. In our
experiments substrate temperature and exposure of oxygen have
been varied independently. As the oxide becomes thicker,
oxygen atoms on the surface migrate into the surface and
oxygen atoms start to bond with silicon atoms buried under the
surface layer. The structure related to this initial step
will be quiet different for Si02 layers due to the different
oxygen concentration in the different oxide layers. In
oxidation under UHV, an OH species is always created due to
background H20 gas produced mainly by wall reactions. One
possible mechanism is that oxygen molecules react with
hydrogen atoms adsorbed on the chamber walls or with
hydrocarbon species buried in the walls of the ion pump.
Another possible source is that water molecules in the
residual gas are broken into H and OH species on the surface.
Since an OH species is very reactive its possible effect
should be considered in the case of relatively long exposures
of oxygen. This effect is a distinct possibility in case of
exposing oxygen to a silicon substrate which has many active
sites since these active sites are easily occupied by OH
species from the gas phase. The passivation effects under UHV
by OH can be eliminated in thermal oxidation since most of
hydrogen species on the surface desorbs below 350C.


22
Electron Energy (eV)
Figure 2. Electron escape depth.


(B)
10 20 30 40
r
A
Plasmon Energy Loss (meV)


97
where A*[uro*/(1+A) ] + [p*/(1+ebo) ], B=wT02 (1+A) ,C=wp2 wT0J/(&ofl),
and 1+A=(e0+1)/(&<+l) Dispersion curves (i.e. energy loss
vs. momentum transfer) with neff=6xl0" /cm3 of equation (3-50)
are shown in Fig. 10 f c) High coupled mode behaves like a
optical phonon initially at small momentum transfer, while
low coupled mode behaves like a plasmon. After cg,| pass cross
over point, these two behaviors are exchanged. Coupling has
the effect of repelling these two mode further apart, i.e.
lower mode becomes lower and higher mode becomes higher energy
loss.
3.4.3. Two-laver dielectric-function model
Ideally bulk terminated surface can not be found since
the source of the potential above the surface is zero. So
the surface layer always exists when the surface is produced.
In this chapter at the section on dielectric function theory,
the probing depth is estimated as -O/' where QN is the
transferred momentum. If the surface layer is thicker than
Qil'', then dielectric function theory described by equation (3-
37) can be used as e(w)=es(w) where es(w) is the dielectric
function of the surface layer. If the thickness of surface
layer (sd) is smaller than QN"' then equation (3-37) can not
be used since different dielectric functions contribute to the
loss. Equation (3-37) can be modified through including
parameters such as thickness of the surface layer, d, the
dielectric constant of the surface layer, es(w), and the


>
4J
iH
cn
c
QJ
4->
C
II
Energy (meV)
(b)
Figure 34.continued.
200


173
related peaks are shown: 141meV (Si-O-Si asymmetric stretch
mode), 105.5meV (Si-O-Si symmetric stretch mode and Si-OH
stretch mode) and 51meV (Si-O-Si bending mode). It has been
reported by many authors that the asymmetric stretch mode
energy varies with the effective thickness of oxide layers
grown on silicon substrate [82,83,88]. The thickness
corresponding to 141 meV is -1.1 monolayer (ML) or 2.9 of the
oxide layer referring to results of the previous authors who
combined HREELS with x-ray photoelectron spectroscopy in order
to calculate the effective thickness. Since the ionic
character of a thin oxide layer is much weaker than a thick
oxide layer, an energy shift of this asymmetric stretch mode
to a lower value can be understood. The 0-H stretch mode was
shown at 452meV and C-H stretch mode was shown at 359.5meV,
which indicates adsorbed hydroxyl groups and adsorbed
hydrocarbon species on the oxide layer. A broad hump near
285meV is the second harmonic of the 141meV asymmetric stretch
mode.
The HREELS spectrum obtained from the annealed at 500C
is shown in Fig. 25(b) The energy and intensity of an
asymmetric stretch mode 141meV does not change significantly
which indicates that at this annealing temperature oxide
species does not change structure or desorb. The hydrogen
species (357meV: C-Hx, 457meV: 0-H) are largely suppressed
since the annealing temperature of 500C is high enough to
desorb most hydrogen species [89],


87
If the normal mode is in the scattering plane as well as
parallel to the surface, and the substrate has a reflection
symmetry plane, the cross-section is zero at the specular
geometry. If the substrate has a rotational symmetry about
the z axis in the above case, the cross-section is zero at the
specular geometry. If the normal mode is vertical to the
surface, the cross-section is almost zero at off-specular
geometry which was already stated in dipole scattering theory.
But these selection rules assume the condition of single
scattering and they can not be applied in case of rapid
variation of the reflectance or high energy multiple
scattering [35].
3.4. Examples of Inelastic Scattering
3.4.1. Surface Optical Phonon Excitation
Surface optical phonons which have finite frequency in
the limit Q(|-* 0, are due to specimen with more than one atom
in one unit cell. When a crystal lattice has two or more
atoms in one unit cell and each atom is electrically neutral,
like Si, Ge and some metals, only forces of short range become
important. Then, in the limit Q(|-> 0, the atomic displacement
related to the surface optical phonon are non-zero only in the
outermost few layers. On the other hand, in ionic crystals
where the Coulomb interactions between the ions produce
couplings extended over large distances, the optical phonon


16
differentiation, smoothing and replotting. But a more
important advantage of interfacing a computer with this system
is in the reproducibility of small changes which can not be
done manually. For example, an angle changed by a stepping
motor controlled by a computer is essential to see the peak
evolution at off-specular geometry since less than 0.1 steps
are necessary to measure the angular width of the scattered
beam.
In this chapter the entire range of experimental
procedures will be covered. The major analytical technigue,
HREELS, will be comprehensively discussed while AES, back-view
LEED, film deposition, UHV pumping techniques, computer
interfacing and sample preparation will be reviewed in the
typical order of experimental use. Finally it should be
emphasized that all procedures are essential to achieve high
quality data, but well prepared sample surfaces and a thorough
understanding of the electron optics of HREELS are key points
for the surface physics experiments described in this
dissertation.
2.2. Ultra High Vacuum
A high resolution electron energy loss spectrometer
should be operated at a pressure below lxlCTtorr. The work
function of capacitors, monochromators and analyzers can be
changed by contamination from residual gas, which results in


239
4r2 cosa f (0,0,a) 6d6d da
(B-16)
k2 (O'+O02 )
where f(0,0,a) is defined in equation (B-13) .
B.3. Fuchs-Kliewer Modes: Lattice Dynamical Framework
If u(l) denote the normal coordinate that describes the
relative motion of ions in the unit cell located at 1, the
equation of motion is
ii0(l) + w02u0(l) = (e'/Mr) E (1) ,
(B-17)
where e' is the transverse effective charge associated with
the unit cell, Mf is the reduced mass of the ions in unit cell
and w0 is the oscillation frequency due to a mechanical
restoring force between the two sublattices. The long-ranged
interaction between the ions is included in the electric field
E(l) generated by the ion motion. Due to dipole-dipole
interaction between each unit cell, E(l) is
e
31^(1,l')i^(l,l') 6qg
}u^(l') ,(B-18)


140
_CD
CU
-100 0 100 200 300 400 500
Energy (meV)
Figure 22. HREELS spectra obtained from nominally clean
Ni(110) showing a strong elastic peak and a weak
loss peak due to residual impurities.


113
-50 0 50 100 150 POO
Energy (meV)
Figure 14. ARHREELS spectra obtained from well ordered
Si(111)-7x7


Intensity
171
FWHM=12.8 meV
Energy (meV)
Figure 29. HREELS spectra obtained from thermally grown
oxide after annealing at 700 K.


252
Area 1 After 30 sec. Sputtering
(c)
Area 2 After 30 sec. Sputtering
Figure 48.continued.


161
(>5) than the Shiraki oxide sample above. After annealing
at 500C for 10 minutes the HREELS spectrum obtained from this
sample is shown in Fig. 26(b) No LEED was obtained from this
annealed surface. After annealing at 900C for 10 minutes
HREELS spectrum obtained from this sample is shown in Fig.
26(c). The LEED pattern from this surface shows a 7x7 pattern
where 3/7 and 4/7 fractional order beams have strong
intensities, but the HREELS data indicate residual surface
impurities of C, H and OH probably at a level of -0.1 ML.
After degreasing an Si(111) surface which had been
preserved in the air longer than two years, the HREELS
spectrum obtained from the sample without sputtering is shown
in Fig. 27(a) Upon annealing from 320C to 1010C with 100C
intervals, the intensity of the 350meV peak starts to reduce
at 325C and at 520C it has almost disappeared. The HREELS
data from this sample after annealing at 520C is shown in
Fig. 27(b). No LEED pattern is obtained for this native oxide
sample even after annealing at 920C and the HREELS spectrum
obtained is similar to Fig. 27(b). After annealing at 1010C
for three minutes the HREELS spectrum is shown in Fig. 27(c)-
-bottom curve. The HREELS spectrum obtained after 20 minutes
without anneal is shown in Fig. 27(c)top curve and the LEED
pattern of this sample shows a relatively clean 7x7 structure.
From AES data, no specific impurity except carbon
(C(236)/Si(92)=l/50) is detected; however, the HREELS data
show that the surface is still not clean.


149
detected by XPS [85], a transverse optical phonon mode
(w50meV) can not be expected from this coalesced oxide layer.
So it is suggested that an OH species due to OH of residual
gas covers up the coalesced oxide layer in the form of
precipitates of an Ni(OH)2 layer. Also 50meV can be the
bending mode of Ni-OH [79]. If 0-H species has the component
parallel to the surface like Ni(OH)2, 50meV peak will be shown
in specular geometry. This loss feature is clearly shown on
gain side since higher energy intensity at gain side is
suppressed more by the Boltzmann factor. Distinct hump at
24meV indicate the Ni(110) surface phonon induced by the next
oxide layer, which was also shown in thermal oxide.
Higher exposure (1.8xl09L 02, 1 torr 30min) spectrum in
Fig. 24(b) shows two overlapped peak near 60meV. Due to the
limit of resolution, second harmonic was chosen for clear
separation of two peaks. Second harmonic of NiO peak appeared
at 124meV and that of Ni(OH)2 peak appeared at 106meV. Ni-0
mode is at 62meV and Ni-OH mode in Ni(OH)2 is at 53meV.
Increase of Ni(OH)2 peak from 49meV (300L exposure) to 53meV
at 1.8x109L exposure and Ni-0 mode decrease from 63.3meV to
62.0meV indicates that thickness of Ni(OH)2 layer increase
while the thickness of NiO layer slightly reduced, since the
ionic character of Ni(OH)2 layer also become strong as the
thickness of the layer increase. Asymmetric tail near 20meV
of elastic peak indicates that the low intensity of the
surface phonon of Ni(110) induced by NiO layer.


231
group (~450meV) are not easily monitored. Besides the typical
example of 'poor' tuning, capacitor charging, wrong geometry,
and wrong bias on the sample can cause a 'poor' tuning.
A.6. Computer Program and Data Handling
All programs are written in PASCAL including the software
needed to drive the IEEE board which is originally written in
BASIC (i.e. converted to PASCAL). The program LHMAIN for data
acquisition and displaying is designed to select parameters.
The data collection program has the capability of being
actuated from the command line of DOS with one parameter being
the file containing pre-established run-parameters and the
other parameter being the filename where the data is stored.
This allows the user to set parameters and achieve a series
of runs without operator intervention. This setting of
command lines has the function of a human operator, namely
setting monochromator angle, sample angle, pass energies and
monitoring counts.
The detailed procedure to use the actual program is as
follows. In Fig. 43. three screens for filenames, selection
of commands and setting parameters are shown. Initially the
filename under which each run is stored is coded to indicate
user initials, date and run index. This runindex is
automatically incremented within the program each time a data
file is saved with the first value being entered by the user
initially at runtime. From the second screen run parameters


GLOSSARY OF SURFACE PHYSICS TERMINOLOGY
AES
Auger electron spectroscopy
ARHREELS
Angle resolved high resolution electron energy
loss spectroscopy
CMA
Cylindrical mirror analyzer
DAS
Dimer adatom stacking fault
ELS
Energy loss spectroscopy
FK
Fuchs-Kliewer
FWHM
Full width at half maximum
HREELS
High resolution electron energy loss spectroscopy
IR
Infrared
LEED
Low energy electron diffraction
MBE
Molecular beam epitaxy
RBS
Rutherford back scattering
RHEED
Reflection high energy electron diffraction
SAM
Scanning Auger microprobe
SEM
Scanning electron microscopy
SK
Stranski-Krastanov
STM
Scanning tunneling microscopy
TEM
Transmission electron microscopy
UHV
Ultrahigh vacuum
UPS
Ultraviolet photoelectron spectroscopy
XPS
X-ray photoelectron spectroscopy
xii


Intensity
Energy (meV)
(b)
Figure 16.continued.
117


ELECTRON SCATTERING STUDIES OF SURFACE PHONON-PLASMON MODES
OF SEMICONDUCTORS
By
JAE MYUNG SEO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1989

ACKNOWLE DGEMENTS
I wish to express my sincere appreciation to all those
who have provided support during this endeavor and who helped
see it through to completion. Without these people I would
not have a successful graduate career.
An unforgettable thank you is extended to the chairman
of the supervisory committee, Dr. John E. Rowe, for his
unending support and motivation. The research ethic he showed
to me, his first Ph.D. candidate, will be greatly appreciated
throughout my career.
I am deeply indebted to Dr. Paul Holloway for sharing
his precious time and allowing me to use the state-of-the-art
surface science equipment in his group. His optimism during
my discouraging times will be a good lesson for my life.
I would like to thank members of my supervisory
committeeDrs. Elizabeth Seiberling, Stephen Nagler, David
Micha and David Tannerfor their guidance and support during
my whole research career at the University of Florida. The
input of Dr. Dale Doering, as a creative master of vacuum
science is also deeply appreciated. Additionally, I would
like to thank the following persons for their technical
support: Scott Black for computer interfacing, Eric Lambers
ii

for scanning Auger experiments, Paul Lyman for Shiraki oxide
samples, William Wresh for Rutherford Back Scattering
experiments, Chris Dykstal for Ni experiments, Larry Phelps
and his colleagues for electrical equipment, and Harvey
Nachtrieb and his colleagues for expert machining.
A special thank you is directed to Mrs. Glenda Smith for
her help in the Physics Department. I appreciate the many
discussions and advice on printing problems that Eric Lambers
and Kelly Truman contributed to this dissertation.
My final thank you goes to my family members, my parents
who allowed, encouraged and financially supported me in order
to achieve this goal and my wife and our two daughters,
Eunkyung, Hyosuk and Yesuk, who sacrificed much of their time
and energy in completion of this goal.
iii

TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS
LIST OF FIGURES V
GLOSSARY OF SURFACE PHYSICS TERMINOLOGY xii
ABSTRACT xiii
CHAPTERS
1 INTRODUCTION 1
1.1. Overview of Surface Physics Techniques 1
1.2. Overview of Specific Experimental Studies 6
2 EXPERIMENTAL PROCEDURE 13
2.1. Overview 13
2.2. Ultra High Vacuum 16
2.3. Auger Electron Spectroscopy 21
2.4. Sample Preparation 28
2.4.1. Overview 28
2.4.2. Specific Examples 31
2.5. Low Energy Electron Diffraction 34
2.6. High Resolution Electron Energy Loss
Spectroscopy 35
2.6.1. Basic Theory of 127 Capacitor
Electron Optics 36
2.6.2. Resolution and Sweeping Mode 4 0
2.6.3. Intensity Angular Profile 43
2.7. Computer Interface 45
2.8. Oxidation and Hydrogen Titration 48
2.9. Film Growth under UHV 52
iv

3 THEORETICAL BACKGROUND
56
3.1. Overview 56
3.2. Elastic Electron Scattering 60
3.2.1. Diffraction from a Bulk Crystal
: 3-dimensional Diffraction 60
3.2.2. Low Energy Electron Diffraction
: Surface Diffraction 64
3.3. Inelastic Electron Scattering: HREELS 68
3.3.1. Semi-Classical Approach 70
3.3.2. Dielectric Function Theory 78
3.3.3. Impact Scattering: Off-specular
Scattering 83
3.4. Examples of Inelastic Electron Scattering....87
3.4.1. Surface Optical Phonon Excitation 87
3.4.2. Surface Plasmon Loss and Dispersion:
Relation of Homopolar Semiconductors,
and Plasmon-Phonon Coupling of Polar
Semiconductors 89
3.4.3. Two-Layer Dielectric-Function Model
97
3.4.4. Surface Phonon Dispersion for Semi
infinite Metallic Surface 102
4 ANGLE-RESOLVED SURFACE PHONON AND PLASMON MODES
AT Si (111) AND GaAs(lOO) SURFACES 107
4.1. Overview and Motivation 107
4.2. Experimental Results 110
4.3. Discussion 123
4.4. Summary 130
5 VIBRATIONAL MODES FROM OXIDE LAYERS ON Ni(lll)
AND Ni (110) 133
5.1. Overview and Motivation 133
5.2. Experimental Results 134
5.3. Discussion 145
5.4. Summary 150
6 VIBRATIONAL MODES FROM OXIDE LAYERS ON Si(111)....151
6.1. Overview and Motivation 151
6.2. Experimental Results 157
6.3. Discussion 168
6.4. Summary 182
v

7 SURFACE EXCITATION AND RELAXATION OF Ge/Si
ALLOY FILMS 184
7.1. Overview and Motivation 184
7.2. Sample Preparation 190
7.3. Surface Excitation from Ge/Si(111)-5x5 191
7.3.1. Adatom Vibration 191
7.3.2. Hydrogen Titration 195
7.4. Surface Relaxation Measurements
Using Digital LEED 203
7.5. Summary 208
8 CONCLUSIONS 209
8.1. Conclusions from Present Experimental
Results 209
8.2. Recommendations for Future HREELS Studies... 212
APPENDICES
A FURTHER EXPERIMENTAL DETAILS 214
A.1. Gas Handling System for UF HREELS System.... 214
A.2. Design of Evaporation System 214
A.3. Quantitative AES Analysis Using Standards... 219
A.4. Digital LEED 220
A.5. Tuning of HREELS Spectrometer 222
A. 6. Computer Programs and Data Handling 231
B FURTHER THEORETICAL DETAILS 235
B.l. Dipole Scattering Cross Section 235
B.2. Transformation of Coordinates 237
B.3. Fuchs-Kliewer Modes
: Lattice Dynamical Framework 239
C GROWTH AND CHARACTERIZATION OF Ge/Si ALLOY FILM...244
C.l. Alloy Film (1000) Grown on Hot
Substrate (560C) 244
C.2. Depth Profile of Thin Alloy Film
without Ge-rich Islands 254
C.3. Depth Profile of Thick Alloy Film
with Ge-rich Islands 256
REFERENCES 262
BIOGRAPHICAL SKETCH 269
vi

LIST OF FIGURES
FIGURE TITLE PAGE
1 HREELS chamber 18
2 Electron escape depth 22
3 HREELS spectrometer 37
4 Angular profile of the reflected electron beam of
the HREELS system.
44
5 Block diagram of data acquisition system 47
6 Si and Ge evaporation parameters 54
7 Electron energy distribution 57
8 The reciprocal lattice and Ewald construction.
(a) Three dimensions; (b) Two dimensions.
63
9 Schematic diagrams of semi-classical dipole
scattering.(a) Electron trajectory and molecular
dipole moment; (b) Transferred momentum; (c) Polar
plot of scattering intensity.
71
10 Plasmon loss calculation and plasmon-phonon coupling
calculation. (a) Plasmon loss calculation for
Si(111) at the specular geometry; (b) Plasmon
dispersion curve for Si(lll); (c) Plasmon-phonon
coupling for GaAs(lOO).
92
11 Two layer mode dispersion and polarization.
(a) Surface eigenmode dispersion for a slab of
material with dielectric function es(w) on a
substrate with a dielectric constant, e(>0); (b)
The electric fields for ut mode.
101
vii

12
13
14
15
16
17
18
19
20
21
Polarization of surface inodes. The displacements in
the surface layer for the three surface modes at
the x point of the two-dimensional Brillouin zone
for an fee crystal with nearest-neighbor central
force interactions and a (100) surface. Modes S6
and S, involve displacements mainly parallel to the
surface while mode S4 involves perpendicular
displacements.
104
ARHREELS spectra obtained from partially ordered
Si(111)lxl for non-specular (L8?0) scattering.
Ill
ARHREELS spectra obtained from well ordered
Si(111)7x7.
113
Dispersion curves of Si(lll) surfaces 115
HREELS spectra obtained from a native oxide layer
on GaAs(lOO). (a) As introduced; (b) Annealed at
580C; (c) Annealed at 665C.
116
ARHREELS spectra obtained from semi-insulating
GaAs(lOO) for specular (A0=O) and non-specular
scattering.
120
ARHREELS spectra obtained from highly doped
GaAs(100) for specular (A0=O) and non-specular
scattering.
121
Experimental data and calculated dispersion curves
for highly doped GaAs(100).
122
HREELS spectra for specular scattering obtained from
a Ni(111)-p(2x2)-O surface with 0.25 monolayer of
chemisorbed oxygen.
136
ARHREELS spectra obtained from Ni(111)-p(2x2)-O.
(a) 1.5 off-specular; (b) 3 off-specular; (c) 4.5
off-specular.
137
viii

22
23
24
25
26
27
28
29
30
31
32
140
141
142
158
162
165
169
171
172
181
192
HREELS spectra obtained from nominally clean Ni(110)
showing a strong elastic peak and a weak loss peak
due to residual impurities.
HREELS spectra obtained from thermally grown oxide
on Ni(110) under UHV.
HREELS spectra obtained from oxides on Ni(110)grown
under UHV at room temperature.
(a) 300L, 02; (b) 1.8xlOL, 02; (c) 300L, air.
HREELS spectra obtained from Shiraki oxide on
Si(111) (a) As introduced; (b) After annealing at
500C;(c) After annealing at 900C.
HREELS spectra obtained from water-preserved oxide
on Si(lll). (a) As introduced after preserving in
deionized water for three weeks; (b) After annealing
at 500C; (c) After annealing at 900C.
HREELS spectra obtained from native oxide on
Si(lll). (a) As introduced; (b) After annealing at
520C;(c) After annealing at 1010C.
HREELS spectra obtained from thermal oxide grown on
a 700 K Si(111) substrate, (a) 10L exposure; (b)
100L exposure; (c) 1000L exposure; (d) 10 kL
exposure.
HREELS spectra obtained from thermally grown oxide
after annealing at 700 K.
HREELS spectra obtained from thermally grown oxide
after annealing at 1100 K.
Asymmetric stretching mode variations of thermal
oxides grown at 700 K and 900 K.
HREELS spectra obtained from Ge0 5Si0 5(111) -5x5
showing a strong elastic peak (E=0) and a weak loss
peak (E105meV) due to residual carbon impurities.
ix

33
34
35
36
37
38
39
40
41
42
43
44
45
46
ARHREELS spectra obtained from Ge0 ,Si0, (111)-5x5
for specular (A0=O) and non-specular scattering
geometries.
193
Hydrogen titration on Ge0 5Si0 5(111)-5x5.
(a) 2.5 L, H; (b) 5 L, H 199
Deuterium titration on Ge0 ,Si0 ,(111)-5x5.
(a) 2.5 L, D; (b) 5 L, D 201
LEED intensity profile of Ge0 ,Si0,(111)-5x5
measured by digital LEED.
205
Surface lattice relaxation 206
Schematic diagram of gas handling system 215
Design of evaporation system.
(a) Collimators; (b) Cross sectional view;
(c) Sideview 216
Schematic diagram of digital LEED 221
Zoom trials of GaAs at different energies.
(a) Zoom at the elastic peak; (b) Zoom at 270 meV:
(c) Zoom at 360 meV.
227
Plate voltage versus capacitor voltage to correct
tuning.
230
Screens of "LHMAIN" program.
(a) Filenames; (b) Selection of commands;
(c) Setting of parameters.
232
Ge evaporation control using AES.(a) Normalized
Si(92eV) intensity versus Ge coverage by Gossmann
et at.[102]; (b) Normalized Si(92eV) intensity
versus Ge evaporation time to calibrate
evaporation rate.
245
Thermal evolution of evaporated Ge film on Si(lll).
247
RBS data from 1000 of Ge/Si alloy film 248
x

47 SEM photograph of 1000 of Ge/Si alloy film. Dark
plateau is area 1 and bright islands are area 2.
250
48 Sputter-AES of 1000 of Ge/Si alloy film.
(a) AES spectra obtained from area 1 as introduced;
(b) AES spectra obtained from area 2 as introduced;
(c) AES spectra obtained from area 1 after 30 sec.
sputtering; (d) AES spectra obtained from area 2
after 30sec. sputtering.
251
49 Sputter-AES of thin (~10A) Ge/Si alloy film.
(a) Sputtered edge profile of thin pure Ge film;
(b) Sputtered edge profile of thin Ge/Si alloy film.
255
50 SEM and sputter-AES of thick (-200)
Ge0 5Si0 5/Si (111) 5x5 film, (a) SEM photograph with
white islands;(b) AES line scan across one of white
islands; (c) SEM photograph after point edge
sputtering, (d) AES line scan across the sputtered
edge, (e) AES depth profile from another area.
257
xi

GLOSSARY OF SURFACE PHYSICS TERMINOLOGY
AES
Auger electron spectroscopy
ARHREELS
Angle resolved high resolution electron energy
loss spectroscopy
CMA
Cylindrical mirror analyzer
DAS
Dimer adatom stacking fault
ELS
Energy loss spectroscopy
FK
Fuchs-Kliewer
FWHM
Full width at half maximum
HREELS
High resolution electron energy loss spectroscopy
IR
Infrared
LEED
Low energy electron diffraction
MBE
Molecular beam epitaxy
RBS
Rutherford back scattering
RHEED
Reflection high energy electron diffraction
SAM
Scanning Auger microprobe
SEM
Scanning electron microscopy
SK
Stranski-Krastanov
STM
Scanning tunneling microscopy
TEM
Transmission electron microscopy
UHV
Ultrahigh vacuum
UPS
Ultraviolet photoelectron spectroscopy
XPS
X-ray photoelectron spectroscopy
xii

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ELECTRON SCATTERING STUDIES OF
SURFACE PHONON-PLASMON MODES OF SEMICONDUCTORS
By
Jae Myung Seo
August 1989
Chairman: Professor John E. Rowe
Major Department: Physics
Surface excitations such as surface phonons and surface
plasmons of Si(lll), GaAs(lOO), silicon oxide, nickel oxide
and Ge0 5Si0 5(111) alloys have been studied under ultrahigh
vacuum using high resolution electron energy loss spectroscopy
(HREELS), low electron energy diffraction and Auger electron
spectroscopy. It has been found that the surface free-carrier
plasmons from sputtered and insufficiently annealed p-type
Si(111) surfaces are losses due to a collective motion of
acceptor-type carriers localized in the space charge region.
The measured dispersion curve of these plasmons achieved by
angle resolved HREELS agrees with that predicted by dielectric
function theory using the Lindhard dielectric function and
parameters appropriate to highly doped semiconductors.
xiii

Subsequent, high temperature annealing above 950C induces a
7x7 reconstructed surface from which a surface phonon due to
adatom vibrations has been detected. Surface plasmons due to
sputter damage and Fuchs-Kliewer phonons have been detected
from sputter-annealed GaAs(lOO) surfaces. The phonon-plasmon
coupling effect has been observed experimentally and a
dispersion curve has been measured by angle resolved HREELS
(ARHREELS) It has been found from oxidation of Ni(110) and
Si(111) surfaces under ultrahigh vacuum that energies of
oxygen-related surface vibrational modes increase as the
thickness of the oxide layer increase in agreement with
previous empirical correlations. This indicates that the ionic
character of the oxide layer increases with increase of the
thickness. Protective oxide layers on Si(111) surfaces have
thinner oxide layers on starting surfaces that are atomically
smooth as compared to typical native oxide layers and do not
have many active sites to react with residual gas. Also, this
thin oxide layer requires a much lower temperature annealing
without sputtering to be removed. In the study of
Ge0 5Si0 ,(111) 5x5 surfaces, it has been found that this alloy
surface has similar adatom vibrational modes to the Si(111)7x7
surface using ARHREELS and these adatoms are mostly (-90%) Ge
atoms using hydrogen titration. For Ge0 ,Si0 ,(111) a Stranski-
Krastanov growth mode has been detected and continuous surface
relaxation of the strained overlayer has been found as the
film thickness increases in the range 0-1000.
xiv

CHAPTER 1
INTRODUCTION
1.1. Overview of Surface Physics Techniques
The potential energy function experienced by atoms at a
surface is different from that of the bulk due to the absence
of atoms above the surface. The atoms at the surface which
are affected by this potential will arrange themselves in
order to determine the lowest energy state of the combined
system: surface and bulk. This results in the surface having
structural, electrical and chemical properties which may
differ greatly from corresponding bulk properties. This
dissertation describes a sequence of surface physics studies
using high resolution electron energy loss spectroscopy
(HREELS) to measure surface vibrational modes. A variety of
supporting experiments are also performed and discussed, since
it is necessary to prepare, characterize and study a single
atomically clean surface in the same ultrahigh vacuum (UHV)
chamber.
It was not possible to monitor an atomically clean surface
before UHV techniques were developed. Due to the short time
during which a surface stays clean, experiments must be
performed before the clean surface is contaminated by residual
1

2
gas impurities. If the pressure of the chamber is less than
lxlO'9 torr, it takes about 103 seconds for the residual gas to
cover one monolayer (1 ML) assuming the sticking probabilities
are near one. At the present time, typical ion-pumped and
well-baked systems usually have a pressure below 1x10"' torr,
which allows about three hours of experimental measurements
after surface cleaning. After the fundamental condition of
modern surface physics experiments (low UHV pressures) had
been achieved, various routine characterization methods of
surface studies soon followed. For example, low energy
electron diffraction (LEED), first demonstrated in 1927 by
Davisson and Germer, has become a routine characterization
measurement during the late 1960s and early 1970s, about ten
years after the first commercial UHV ion pumps in 1961.
Another characterization tool, Auger electron spectroscopy
(AES) became routine in the mid 1970s although its feasibility
was first demonstrated by J. J. Lander in 1953.
Surface studies can be classified by the probe particles
such as electrons, ions, atoms and photons. The electron is
the most common and convenient probe particle. The electron
is a light, charged particle; it is easily focussed to become
a non-destructive probe at low incident energies. Since the
electron has a larger scattering or interaction cross-section
than the photon, the finite electron escape depth from the
surface enables one to collect the electrons originating
mainly from the outermost 1-3 atomic layers, i.e. within the

3
escape depth of -5 A. The detected electrons can be the
incident electrons which are scattered from the surface within
the escape depth as in a LEED or HREELS experiment, or they
can be secondary electrons ejected from surface atoms due to
some other excitation mechanism such as Auger relaxation of
excited core-hole states. The electron escape depth depends
upon electron energy and to a lesser extent on the atomic
number of the surface species. Usually electrons with
energies from a few eV to several thousand eV escape from 5-
20 A under the surface. In HREELS, the incident electrons
with an energy of 1-20 eV penetrate 3-10 A under the surface,
but the electric field which determines the so-called dipole
scattering cross section extends in semiconductor and
insulator materials down to 200-500 A below the surface.
Photons, ions or neutral atoms can be also used as probes in
a surface scattering experiment. Typically, several surface
techniques are combined in order to achieve the overall
purpose of experiments including surface characterization,
surface preparation and final measurements.
The structure and elementary vibrational excitations of
bulk condensed matter can be studied by spectroscopies such
as inelastic neutron scattering, infrared (IR) spectroscopy,
Raman spectroscopy and inelastic electron tunneling. High
Resolution Electron Energy Loss Spectroscopy (HREELS),
Infrared absorption-reflection spectroscopy and neutral atom
inelastic scattering are the main probes of the vibrational

4
properties for the surfaces of solids. Inelastic neutron
scattering is used for determination of the energy-wavevector
dispersion relation of bulk single crystals, but the neutron
phonon scattering cross section is very small, so this
technique is not sensitive enough for surface studies of small
area single crystal samples (however, powder samples can be
studied). Surface IR spectroscopy is a technique for studying
the molecular structure of adsorbate atom complexes and can
be used on metal, semiconductor and insulator surfaces. The
electric field of the incident IR beam can couple to the
phonon modes of a solid and the absorption is usually shown
as a function of the incident photon energy. The high
resolution (0.1 meV) of IR spectroscopy is an advantage of
this technique but the small absorption cross section for
infrared photons limits the sensitivity of this method.
Unless special reflection geometries are used, most of the
detected IR signal is due to the bulk optical properties and
provides an unwanted background signal that degrades the
sensitivity. Raman spectroscopy and inelastic electron
tunneling spectroscopy also have a substrate background signal
problem which limits their sensitivity for surface studies.
This problem is partially solved by the improvement of surface
enhancement of these last two spectroscopies. Neutral atom
scattering uses inelastic scattering of He or Ne atom beams
of low-energy typically 20-60 meV. The resolution of atom
scattering is very good (-0.5 meV), but it is experimentally

5
difficult to produce highly monoenergetic beams of neutral
atoms. Since the maximum incident energy is usually limited
to ~60meV, this technigue is useful only for obtaining surface
vibrational modes below this incident energy.
The technique we have used, HREELS, covers a wide range
of vibrational energies (5-500 meV) which in principle is also
covered by IR spectroscopy. HREELS is useful for probing
surface phonon dispersion, surface plasmon excitation and
intra- and interband transitions. The high surface
sensitivity of HREELS enables one to detect 0.001 ML coverage
at the surface if the species has a typical dipole scattering
cross section (~10',6cm2) Therefore many surface impurities
can be readily detected. Due to the low incident electron
energy, HREELS is a non-destructive technique which can be
used to probe weakly bound species such as atomic hydrogen on
graphite. In HREELS the scattering occurs due to the coupling
between the incident electron and the electric field of charge
fluctuations under the surface. This long-ranged interaction
of the Coulomb field makes it possible to detect interfacial
modes under the surface as long as the interface contains
strong dipole species and is within the detection limit (-200-
500 ) Therefore HREELS is a very surface and interface
sensitive technique with a wide energy range. However, it
also has several limitations at the present time. Even with
the most advanced electron optics, the resolution in HREELS
is about 3 meV. This limits the analysis of closely spaced

6
vibrational modes such as H-Si modes on Si(100) which are
split by only 1.4 meV. Since the performance of the electron
optics depends sensitively upon the work function uniformity
and stability of the energy analyzer capacitors, HREELS
experiments should be done under UHV conditions. For highest
signal levels, the specimen should have a dynamic dipole
moment which has a large component along the surface normal.
At the present time HREELS is under development to improve the
experimental resolution of spectrometers and to improve
scattering theory which does not yet treat long-range dipole
scattering and short-range impact scattering on an equal
basis.
1.2. Overview of Specific Experimental Studies
The motivation of this dissertation is to systematically
study semiconductor surfaces using HREELS and in one selected
example (oxidation of silicon), to compare a similar study on
a metal surface (oxidation of nickel). For practical use of
semiconductors, these materials have excess electrons or
holes; i.e., they are n-type or p-type. These electron or
hole carriers due to impurity doping play an important role
in forming potential barriers at the interfaces of
semiconductors with metals (e.g. Schottky barriers or ohmic
contacts). Plasmons due to excess carriers can be
distinguished from phonons by energy-wavevector dispersion
which can be measured by angular studies of HREELS. Even if

7
a semiconductor does not contact a metal, dangling-bond
surface states pin the Fermi-level at a position in the gap
usually different from the bulk. Due to this band bending
potential at the interface or bare surface, the local surface
concentration of carriers can be dramatically different from
that of the bulk. Using HREELS we observed such differences
in carrier concentration by monitoring the plasmon energy for
Si(lll) and GaAs(lOO) surfaces.
In the present studies HREELS is applied to well ordered
nickel surfaces since the formation of NiO at nickel and
nickel-chromium alloy surfaces is not yet well understood.
The simple selection rules make it easy to understand the
polarization of specific vibrational modes. The coalesced
nickel oxide layer is not a single continuous layer like a
silicon oxide layer. Recently, a new interpretation of
multiple oxide phases at Ni surfaces was suggested by a high
binding energy oxygen species of nickel oxide layers detected
with x-ray photoelectron spectroscopy (XPS). HREELS
measurements of O-Ni vibrational modes were correlated with
these XPS results.
In additional experiments, silicon oxidation in
conjunction with surface cleaning techniques is studied. The
formation mechanism of oxides at Si-SiOx interfaces; i.e., the
initial steps of Si oxidation are not well understood. The
knowledge of very thin oxides is closely related to cleaning
techniques of an Si substrate prior to UHV experiments. Thin

8
protective oxide layers can be used to prepare clean surfaces
under UHV without sputtering and high-temperature annealing.
For two different reasons HREELS is a useful technique to
study silicon oxides. First, oxygen atoms form partially
ionic bonds with Si atoms, and their strong dipole moment
enables one to easily detect the local vibrational modes of
each type of oxide. Second, the high surface sensitivity of
HREELS can easily detect surface impurities present on oxides
and on clean surfaces down to 0.001 ML. These may include
hydrocarbons, hydrides or hydroxyl groups as well as silicon
carbide. Similar oxidation mechanisms with different oxide
local bonding sites are expected for many metal oxide
interfaces.
Alloys of Si and Ge are another interesting area for
surface studies since they have reconstructed surfaces that
are similar to the reconstruction found on clean elemental Ge
or Si surfaces. A Ge0 5Si0 5 alloy grows epitaxially on
Si (111) 7x7 substrates and has a 5x5 LEED pattern. From
surface reconstruction studies by LEED and scanning tunneling
microscopy (STM) both 7x7 and 5x5 surfaces are reconstructed
in a quite similar way with about h monolayer of adatoms on
a dimer-stacking fault unit cell. The vibrational mode due
to adatoms has been reported for Si(111)7x7 using HREELS. The
cross section for Ge-Si alloys may increase due to a non-zero
dipole moment in the Ge-Si system compared to the pure
elements in the bulk. We can compare HREELS results for 5x5

9
and 7x7 surfaces and thus attempt to determine if the bulk
composition is present at the surface. Another advantage of
HREELS is the fact that the vibrational energy depends upon
the reduced mass of the local oscillator. By titrating H
atoms on the 5x5 surface the relative numbers of Si and Ge
atoms in this surface can be determined since surface Si-H or
Ge-H species have different reduced masses and the
corresponding stretching modes have an energy difference of
~10 meV. We have found that a large fraction of the 5x5
surface (-90%) covered with H contains Ge atoms rather than
a bulk 0.5-0.5 mixture. This is likely due to Ge having a
lower surface energy than Si on this alloy surface.
Summarizing, High Resolution Electron Energy Loss
Spectroscopy is a highly surface sensitive and quantitative
vibrational spectroscopic technique. Through novel
applications of this technique (i.e. Angle Resolved HREELS),
surface phonon and surface plasmon modes were investigated
from the semiconductor surfaces such as (1) the clean Si(111)
surface (2) the oxide surface of Ni and Si and (3) the Ge/Si
alloy surface.
Chapter 2 describes broadly the experimental procedures
following the typical sequential order of experimental
measurements. Therefore the same experimental procedures are
not repeated at the chapters (4-7) unless some new specific
procedure was used. In addition to the procedures used to
collect the data, a brief review of equipment and its

10
operation are also included. Especially, the electron optics
of HREELS is described in some detail in order to allow one
to understand the complex tuning problem of this type of
instrument. In addition to the main experimental chapter,
further experimental details are presented in Appendix A.
Chapter 3 describes the theoretical background of
electron scattering as applied to HREELS. Electron scattering
theory is divided into an elastic scattering part, which
consists of bulk diffraction and surface diffraction, and an
inelastic electron scattering part which includes dipole and
impact scattering HREELS theory. For a clear understanding
of dipole scattering theory, a semi-classical approach for
adsorbed molecules as an example is comprehensively reviewed.
For more general applications of HREELS on semiconductors a
two-layer dielectric function theory is reviewed. For the
application of angle-resolved HREELS, impact scattering theory
is also reviewed. In addition to a review of fundamental
HREELS scattering theories, applications to specific systems
are briefly mentioned. Detailed theoretical background
information is attached in Appendix B.
Chapter 4 is one of the four experimental measurement
chapters. In this chapter, investigations of pure Si(111)
and GaAs(lOO) surfaces are discussed. In the first section,
sputtered annealed Si(111) surfaces are discussed. The quasi
elastic peak broadening effect and p-type surface accumulation
layer formation due to a sputter-cleaning procedure is

11
considered. In the second section, sputtered and annealed
GaAs(lOO) surface studies are considered. Also sputter
cleaning effects are discussed in general. In this chapter,
angle-resolved HREELS results show dispersion relations of
phonon inodes, plasmon inodes and coupled phonon-plasmon inodes.
Chapter 5 is the second experimental chapter. In this
chapter studies of the room temperature oxidation of Ni(lll)
and Ni(110) are presented. The origin of a high binding
energy oxygen species detected in x-ray photoelectron
spectroscopy (XPS) data from the coalesced oxide layer is
considered. A new interpretation of thin NiO layers is
postulated based upon HREELS data.
Chapter 6 is the third experimental measurement chapter.
In this chapter studies of oxide layers on Si(111) surfaces
are presented. In the first part, studies of a Shiraki oxide,
a deionized-water preserved oxide and a native oxide formed
in air on Si(111) surfaces are discussed. The effect on the
clean surface of removing oxide layers by annealing is
considered. In the second part, thermal UHV oxidation studies
of Si(111) surfaces are presented. Differences in thin oxides
are clearly evident in our HREELS results which indicate that
the best thin oxide layers are produced by the Shiraki
chemical etching method.
Chapter 7 is the fourth experimental measurement chapter.
This chapter includes the film growth, characterization,
relaxation measurement and excitation measurement through a

12
combination of various surface science techniques. In this
chapter Ge0 5Si0 5 alloy films grown on Si(111) are investigated.
In the first part, surface vibrational excitations of the
alloy film and its atomic hydrogen titrated surface were
investigated by HREELS. Surface phonon and surface elemental
species of the alloy film were considered. The origin of the
5x5 reconstruction is discussed. In the second part
pseudomorphic growth studies and surface strain relaxation
studied by digital LEED are presented. The critical thickness
for pseudomorphic growth is discussed in the context of our
measurements on (111) pseudomorphic films. In addition to
the previous parts studies of the growth mechanism and
morphology of the films studied by LEED, AES and Auger
microscopy are presented in the Appendix C.
In Chapter 8 the experimental conclusions are summarized.
In addition to experimental conclusions, recommendations for
future studies are suggested.

CHAPTER 2
EXPERIMENTAL PROCEDURES
2.1. Overview
Surface studies have been made possible with the
development of new pumping techniques during the 1960s. The
competition between pumping speed and gas desorbing from
chamber walls or leaking of the chamber determines the
ultimate pressure. The basic condition for surface studies
is that the pressure in the chamber must be kept as low as
possible in order to prevent contamination of the surface by
residual gas during the experimental procedures. To achieve
the ultimate pressure for ultrahigh vacuum (UHV) (<10'9 torr) ,
combined pumping techniques such as turbo molecular pumping,
bakeout at 200-250C and sputter-ion pumping are generally
used.
Another factor determining the quality of data is the
preparation of a clean surface. A clean surface can be
achieved by chemical pre-treatment before introducing it into
the chamber with subsequent ion-bombardment sputtering
followed by thermal heat treatment under UHV. The common goal
of cleaning is reducing the level of contamination and defects
13

14
on the sample surface to well below 1012 cm'2, i.e. about 0.001
monolayer (ML). If pre-treatment such as oxidation, hydrogen
adsorption or film growth under UHV is required, it is usually
done in the same UHV chamber before data acquisition. It is
important in surface physics experiments to prepare the
required surface for analysis under UHV, since the transfer
procedure from one UHV chamber to another may possibly bring
contamination from the air. In some surface physics
experiments several UHV chambers are interconnected with UHV
transfer between each. However, in the University of Florida
high resolution electron energy loss spectroscopy (HREELS)
chamber several different sections of a single UHV chamber are
used for various surface preparation, characterization and
HREELS measurements. The advantage of pretreatment under UHV
is in the control of gas adsorption and/or deposition of thin
(0-10 ML) overlayers. Compared to an atmospheric pressure
environment or low vacuum condition, the number of residual
gas molecules in UHV is so small that the effective operation
time can be long enough not to contaminate the surface with
the residual gas during the measurement procedure (e.g. HREELS
or low energy electron diffraction (LEED)). Basic surface
analysis techniques using Auger electron spectroscopy (AES)
for detecting surface chemical composition and LEED for
determining the long range periodicity of surfaces follow
cleaning procedures.

15
Once a clean and smooth surface is prepared under UHV,
in our experiments the main analytical technique, HREELS, can
be performed. It must be considered that the total
measurement time includes tuning of the spectrometer as well
as data acquisition. The critical point in tuning a HREELS
system is to achieve the best resolution of the electron
optics. The procedure of tuning can be divided into two
parts, namely, detection of the optimum geometrical positions
of monochromator and sample manipulator, and adjustment of the
potentials of the electrodes and capacitors so as not to get
the largest intensity of the elastic peak but to get the
smallest width of the elastic peak. In angle resolved HREELS,
the incident angle of monochromator is changed to vary the
momentum transferred to the sample surface. It is necessary
to understand the kinematical geometry of the spectrometer for
determining the transferred momentum. Sometimes a diffracted
order of the electron beam is needed instead of the specular
(0,0) beam to suppress the high intensity of the elastic peak
in order not to saturate the detector which would result in
a nonlinear response. In this case understanding the
orientation of the sample surface is essential in order to
find the diffracted beam. The next step for tuning is to
optimize the various potential settings of the spectrometer.
This will be discussed in more detail later in this chapter.
Data acquisition using a computer is useful for analysis of
data by magnification, comparison of two or more spectra,

16
differentiation, smoothing and replotting. But a more
important advantage of interfacing a computer with this system
is in the reproducibility of small changes which can not be
done manually. For example, an angle changed by a stepping
motor controlled by a computer is essential to see the peak
evolution at off-specular geometry since less than 0.1 steps
are necessary to measure the angular width of the scattered
beam.
In this chapter the entire range of experimental
procedures will be covered. The major analytical technigue,
HREELS, will be comprehensively discussed while AES, back-view
LEED, film deposition, UHV pumping techniques, computer
interfacing and sample preparation will be reviewed in the
typical order of experimental use. Finally it should be
emphasized that all procedures are essential to achieve high
quality data, but well prepared sample surfaces and a thorough
understanding of the electron optics of HREELS are key points
for the surface physics experiments described in this
dissertation.
2.2. Ultra High Vacuum
A high resolution electron energy loss spectrometer
should be operated at a pressure below lxlCTtorr. The work
function of capacitors, monochromators and analyzers can be
changed by contamination from residual gas, which results in

17
de-focussing of the HREELS electron optics. At the same time
the sample surface is also quickly contaminated by the
residual gas. If the sticking coefficient of the residual
gas is assumed as one, then it takes 1000s (~17min) at lxlO'9
torr to cover the sample surface with one monolayer of
residual gas. Since HREELS is a very sensitive technique, in
order to eliminate impurity contributions to the data it is
a prerequisite condition to achieve pressures below lxlO',0torr
(i.e. 170 min for covering 1 monolayer, which consists of the
time necessary for tuning HREELS and time for data taking by
computer). The vacuum chamber overall diagram is shown in
Fia. 1. Due to the chamber geometry and position of the
pumping connections, the pressure at the entrance of ion-pump
is almost half of the pressure near the hemispherical analyzer
as measured by conventional UHV ion gauges. An HREELS
spectrometer is just above the ion-pump since that position
is near the lowest pressure in the chamber. A sample
manipulator horizontally transports a specimen from the HREELS
spectrometer to the electron analyzer. An electron gun and
5-crucible e-gun source are in one UHV section aimed at a
common point which is just below the electron analyzer used
for AES. Next to this electron analyzer are gas leak valves
and an ion-gauge. Next to gas leak valves is the LEED optics
section (see Fig. 1) In this section an 8"OD viewport allows
one to observe the LEED screen as well as the motion of the
sample manipulator. In order to allow a more complete view

LEYBOLD-HERAEUS SYSTEM (ELS-22)
7 HREELS
/
5 CRUCIBLE
E-BEAM
EVAPORATOR
SAMPLE MANIPULATOR
x.Y.z.e
GAS INLET
MANIFOLD
MBE
KNUDSON
CELLS
TITANIUM
SUBLIMATION
and
ION PUMPS
TURBO PUMP
-10
PRESSURE: 1x10 torr
Figure 1. HREELS chamber.

19
of the diffraction pattern (Laue geometry), a rear view LEED
system was constructed from a commercial LEED optics using the
procedure developed by E.E.Chaban and co-workers[1].
Photographs of our LEED patterns are usually taken through
this rear view LEED port.
Pumping procedures to achieve UHV from the opening to
the air of the University of Florida HREELS chamber are as
follows.
(1) As soon as the sample is mounted, the electrical
connection of sample relative to ground is checked to ensure
the sample has not accidentally been grounded. Then all the
ports opened during atmosphere operation are closed.
(2) The turbo-molecular pump line is opened and the mechanical
pump is turned on to achieve low pressure (-1 torr) at the
foreline of the turbo-molecular pump.
(3) If the pressure is below 1 torr between the turbo
molecular pump and the mechanical pump, the turbo molecular
pump and water chiller for cooling the turbo molecular pump
are turned on.
(4) The turbo molecular pump is used during the backing
procedure. The time required to reach the ultimate pressure
in one hypothetical example of an unbaked metal-gasket system
is lO'-lO5 hours, even though there is no leak in the
chamber[2].

20
(5) Baking takes ~20 hours at 200C. During this bakeout time
period the pressure increases to a maximum of p~10'6 torr and
then decreases slowly over 10-15 hours to p-5xl0* torr.
(6) After cooling for several hours, the ion-pump is switched
on and most of the filaments in the chamber including the
titanium sublimation pump are outgassed. Briefly heating the
titanium sublimation pump ~2 times will reduce the pressure
to -1x10'' torr after the chamber reaches room temperature.
(7) If the liquid nitrogen cold trap is filled, the pressure
can be lowered by a factor of two or three into the 10'" torr
range.
The most important procedure discussed above is the
bakeout procedure since the proper length of the bakeout time
and temperature will determine the final pressure. Also
baking reduces the total time of desorption from the surface
of chamber walls and diffusion from inside surfaces of chamber
walls. Design of vacuum systems including chamber and pumps
depends upon the kind of experiments to be performed. The
total pumping capacity is the main factor to determining the
efficiency of the system. The order of design of UHV system
is the determination of usage of chamber, determination of the
available accessories, estimation of gas flow rate and
determination of pumps. In our system a Leybold-Heraeus
(model TRIVAC-A dual stage D16A) mechanical pump, with
capacity of 400 1/s and a Leybold-Heraeus (model Turbovac 150)
turbo molecular pump with capacity of 150 1/s are used. It

21
takes 5 min from air pressure, to reach one torr which is the
proper pressure to start the turbo molecular pump. The turbo
molecular pump is used during baking procedure and outgassing
of the filaments. Finally an ion-pump (Leybold-Heraeus model
IZ-270 with capacity:270-300 1/s) and a titanium sublimation
pump are used in conjunction with a liquid nitrogen cold trap.
In addition to pumping systems, a gas handling system is
described in Appendix A.l.
2.3. Auger Electron Spectroscopy
Auger electron spectroscopy is a surface analytical
technique for determining the elemental composition of the
outer atomic layers of a solid. It provides high sensitivity,
elemental selectivity for all atoms except hydrogen and helium
and good lateral resolution (10'6-10'7m) of the analyzer which
enables the detection of specific atomic species situated in
the first few atom layers (5-10). In AES, surface
sensitivity is mainly determined by the escape depth of the
observed electron, since the incident electron (Ep = 3-5keV)
penetrates much more deeply into the crystal. The inelastic
mean free path (X^) of electrons as a function of energy (Ep)
shown in Fig. 2. is defined as the distance that an electron
of energy E travels with a probability e'' of not being
scattered by inelastic events that degrade its energy by more
than several electron volts [3]. Therefore, in one type of

22
Electron Energy (eV)
Figure 2. Electron escape depth.

23
atomic Auger transitions, higher energy electrons usually
escape from deeper layers. Besides the elemental information
from the surface, the fine structure of Auger peaks is
sensitive to the chemical environment. This change arises
from transitions involving a surface molecular orbital formed
by electrons in chemical bonds between the adsorbed species
and the outermost surface atoms. The localized density of
states at adsorption sites will be different from the clean
surface. It is difficult in AES to relate peak shifts to
changes of certain levels (a few electron volts) like XPS
since three electrons are involved. But gualitative detection
of chemical shifts in AES are at least indicative of
differences in chemical bonds between elements. This can be
found in silicon oxidation which produces a native silicon
oxide layer.
Another aspect of AES is guantitative analysis. Two
possible problems in using AES as a quantitative and elemental
analytical technique are the determination of the total number
of detected Auger electron separated from the background and
the understanding of relation between the intensity of the
spectrum and the atomic density in the selvedge of the
specimen. At this point, these procedures are not fully
developed. The typical AES experimental procedures are as
follows [4].

24
(1) The derivative spectrum dN(E)/dE of the output of the
spectrometer is integrated with respect to energy E to get
N(E) .
(2) The background under peaks is determined by means of
spline-approximation.
(3) Substraction of background from N(E) results in the
distribution NA(E) of true Auger electrons.
(4) The area under each peak consists of multiplets and loss
feature is Auger current which results from the corresponding
transitions.
(5) This Auger current depends upon the angle of acceptance
of the spectrometer, the primary current, backscattering
factor, the attenuation by scattering processes after the
transition, ionization cross-section for electron induced
ionization of the specific level, the transition probability
for the whole series related to a specific level, and the
desired atomic density. By choosing appropriate values for
each case the desired atomic density can be calculated.
It is generally assumed that the signal from a fraction
of 10'2 of a monolayer can be detected by AES. Without going
through details shown above, relative measurements (using the
standard AES sensitivity factors) can give quantitative
information on atomic density. The detailed procedures will
be discussed in Appendix A. section A.3., Quantitative AES
analysis using standards. Summarizing the introduction of
AES, AES is useful for detecting elements at the surface since

25
it is very sensitive to most of elements and has good lateral
resolution. AES also gives qualitative information on the
chemical reaction between two elements if they are chemically
reactive. Finally AES has the big potential for giving
quantitative information of the surface composition.
An Auger transition converts an atom with a hole in one
core level (i.e. i-th level) to a final state that consists
of the atom with holes in other orbitals (i.e. *j1 and 'k')
and the fast electron whose kinetic energy is Ek. In the
initial process before Auger emission, a core hole is produced
by electron impact ionization, i.e. element A,
A- A4(i) + e" (2-1)
where A4(i) means the hole in 'A' is in level 'i'. The Auger
process is
A* (i)* A44 (j k) + e'auger (2-2)
where the emitted electron e'auger has kinetic energy Ek from
the vacuum level and A44(j,k) has left the atom doubly ionized
with holes in 'j' and k' levels.
To understand Auger electron energy which can be measured
by the analyzer, energies of the initial and final states of
the Auger process should be known. The energy of the initial
state of the Auger process (2-2) can be measured by x-ray

26
photoelectron spectroscopy, in which the i-th shell is
photoionized in the process (2-1) [5]. The binding energy
relative to the Fermi level of the electron in level 'i' is
Eb(i) = hu> (Ek + 0S ) (2-3)
where Ek is the kinetic energy of photoelectron emitted from
the sample with work function is not the sum Eb(j) + Eb(k) of separate binding energies which
can be measured like the initial state, since interactions
between the two holes can occur and relaxation events to fill
core holes are nonlinear. Considering ccc Auger process
(i.e., the initial state has a core hole and final states have
two core holes), the derivation of the final states A**(j,k)
can be decomposed into two processes.
A- A*
(j) + e* ,
(2-4)
A*(j)-
A**(j,k) + e- .
(2-5)
In the process (2-4), Eb(j) is required, where Eb(j) is the
energy required for the threshold process which places the
ionized electron at the Fermi level. In the process (2-5),
Eb'(k) is also required for the threshold process. Using the
equations (2-2), (2-4) and (2-5),

27
Eb(i)= Eb(j) + Eb (k) + (Ek + <*>s ) ,
(2-6)
where Ek is the kinetic energy of the ejected Auger electron.
Energy required to produce the two holes in the levels j ',
1 k', namely Eb(j,k), is suggested as
E( j ,k) = Eb(j) + Eb (k)
= Eb(j) + Eb(k) + F(j ,k;x)
(2-8)
where F(j,k;x) describes the interaction (namely repulsive
between holes) energy of the two holes. Shirley has pointed
out that including the relaxation of other electrons in and
around the atom (i.e. more screening) that takes place in the
process (2-4), Eb(j,k) can be expressed as follows,
Eb(j,k) = Eb (j ) + Eb(k) + F (j k;x) R(x)
(2-9)
which gives excellent agreement over the entire range of
atomic number. In CW and CCV case, the final state includes
valence orbitals. According to the localization of the final
state valence bond, the interaction term should be considered.
It is not well known yet about the line shape estimation of
CW or CCV. Both energy and line shape change with chemical
environment for Auger transitions having final states made up
of valence orbitals involved in bonding, namely, CW and CCV
processes. Both chemical shifts and line shape variation have

28
been observed by AES. Energy shifts of Auger peaks have also
been attributed to changes in relaxation effects associated
with changes in chemical environment [6,7].
Sputtering-depth profiling with AES is a powerful
analytical technique for detecting the depth distribution of
elements in the film. In combination with rare gas
sputtering, depth profile analysis uses the surface
sensitivity of AES electron. But it is hard to get the
quantitative information of the depth of each layer. Because
of differential sputter rates caused by crystallite
orientation and surface contamination, the depth resolution
generally decreases as the film thickness increases [8,9].
Also combined with scanning electron microscopy (SEM), AES
can probe at the special point due to its high lateral
resolution while looking at the surface morphology. As
another application of AES, quantitative estimation using
standards will be introduced in Appendix A.3.
2.4. Sample Preparation
2.4.1. Overview
As mentioned in the experimental overview, one of
important factors which determines the quality of data is the
preparation of a clean and smooth sample surface. Recently
this has become more important since thin film techniques
such as molecular beam epitaxy and chemical vapor deposition

29
need atomically smooth surfaces. Conventional cleaning
techniques, namely rare gas sputtering and high temperature
annealing, work well for most substrates, but these procedures
induce defects on the surface which cause unexpected
impurities to migrate to the surface and vary the desired
doping profile. To minimize the problem related to cleaning
under UHV, it is recommended that the specimen be pretreated
before introducing into UHV. Another reason for preparation
of clean surfaces without sputtering and annealing is that the
band bending at the semi-conductor surface, due to introducing
surface states, results in depletion of the carrier at the
surface. Ohmic contacts with metal require a thin barrier or
a large number of carries at the interface.
In this section the normal cleaning procedures, namely
sputtering and annealing, cleaving and chemical pretreatment
on the semiconductor surfaces will be discussed. Especially
for Si(111) substrates, a thin chemical oxide treatment the
so called Shiraki method[10] and for the MBE grown GaAs(lOO),
the As capping technique will be examined[ll]. Besides the
cleaning technique, the temperature measurement and heating
will be discussed.
Some materials (e.g. Si, GaAs etc.) cleave easily to a
direction of a favorable surface. In highly oriented
pyrolitic graphite (HOPG), adhesive tape can be used to cleave
the surface in the air since the graphite surfaces are non
reactive with air contamination. In III-V compound material,

30
a strong vacuum compatible epoxy can be used for cleaving the
surface. These procedures, so called cleaving, can be used
to prepare clean surfaces under UHV. Alkali halide (NaCl,
LiF, NaF, KC1 etc.) (100)faces, (111)faces of materials like
CaF2, Si, Fe, Zn and Be will cleave at liquid nitrogen
temperatures[12], but this technique is limited to the small
surfaces and a single orientation.
Bombardment of rare gas ions such as Ar* and Ne*
(~10/LiA/cm2 at 500eV-1000eV) can be used to clean a surface.
Impurity atoms receive enough energy from the incident gas
ion to be ejected from the surface. However ion bombardment
can cause damage leaving the surface in an amorphous-like
phase [13]. Some problems may arise from occlusion of the
inert gas atoms on preferential sputtering of one of the
components in a binary alloy or compound [14]. On an Si(lll)
substrate after Ar* sputtering, annealing at 1000C for 2 min
will produce the reconstructed surface with a 7x7 LEED
pattern. But it is hard to estimate how many defect sites are
left on the surface. On a GaAs(lOO) substrate after light Ne4
sputtering, annealing around 500-550C for 2min will produce
a reconstructed surface with LEED pattern that will vary
according to the stoichiometry of two components at surface
[15], It is also known that the proper choice of reactive gas
will help the cleaning of the surface, such as oxygen to
remove the carbon atoms in a suitable pretreatment and
temperature.

31
2.4.2. Specific Examples
In cleaning an Si(111) substrate, along with the
techniques listed in the previous section, laser annealing
[16] and galliation [17] in which the substrate is exposed to
a gallium vapor beam at about 800C substrate temperature in
UHV can be used. Galliation is known to be effective to
remove silicon oxide, but inward Ga-diffusion or removal of
Ga atoms by evaporation has not been confirmed.
The chemical treatment of an Si surface consists of two
steps. The first is stripping the native oxide with HF acid
and the second is applying a wet oxidizing chemical treatment
like the RCA cleaning method [18]. But the oxide produced by
this technique is relatively thick, such that thermal
annealing at relatively low temperature (1000-1100 K) can not
desorb oxygen species from the surface. To avoid high
temperature annealing, it is necessary to prepare a very thin
passive oxide layer which can be desorbed at a relatively low
temperature (-1000 K) at which defects do not generate or the
impurity profile does not change. This is the basic idea of
the Shiraki cleaning technique[10] of which the procedure is
listed in section 2.8.. The experimental results of this
sample using HREELS are reported in chapter 5. The
temperature of oxide of surface starts to desorb was 850C,
but at 950C the oxide layer totally desorbs. The LEED
pattern is very clean (i.e. most of the fractional order beams

32
can be clearly identified at 35eV) and HREELS shows a very
broad elastic peak, which is a typical result for the impurity
free surface. The preparation procedure takes 3-4 hours.
Preserving the sample, prepared by Shiraki method, in D-I
water for 20 days caused further oxidation of the Si (111)
surface (refer chapter 5).
In GaAs grown by MBE, elemental As covers the final layer
to protect it from contamination while transporting the sample
from one chamber to another chamber. Since elemental As
covered the GaAs MBE layer, the oxide As406 was formed in the
air environment. After introduction into the UHV, annealing
at -400C removed the volatile top oxide layer resulting in
surface where the carrier is not depleted. The critical point
in preparation of As capped surfaces is complete coverage with
elemental As. If some portion is exposed, then air
contamination will induce strong oxidation of the GaAs
surface. Subsequently a low temperature (~400C) anneal is
not enough to get rid of the surface oxide on the GaAs layer.
Temperature measurement is critical in observing
desorption temperature of adsorbed gases during the sample
preparation. The temperature is measured by a Chrome1-Alumel
thermocouple which covers from -100 K to 1700 K, and infrared-
optical pyrometer which covers 350C to 1500C. The operation
of thermocouple is the current due to work function difference
of two different metals. When the contact becomes hot,
electrons are always transferred from the lower work function

33
metal to higher work function metal. The optical pyrometer
is based on the principle that the color of the light emitted
by an object is a function of the temperature of the object.
The hot object is imaged by a lens in a plane where a filament
of light bulb is situated. The current of this filament is
regulated by a variable resistor until the brightness of the
object and that of the filament is equal. This current
reading using an ammeter can be calibrated directly in
centigrade.
Annealing samples by resistive heating, namely flowing
currents through the sample, works well with the sample which
has a low resistance. In case of annealing a sample which has
a high resistivity (>10 ohm-cm), a power supply with a high
voltage (70V-350V) and a regulated current can be used. If
a voltage source such as variac is used, the sample may be
quickly destroyed. For the semi-insulating sample the high
voltage (up to -350V) is not enough to give initial heating.
In a semiconductor, if the initial heating does not produce
thermally excited carriers, it is impossible to heat
resistively. Using another low resistive sample at the back
of the sample, the initial temperature of the sample can be
increased by heat conduction from the back substrate. Another
method of heating samples with high resistivity is depositing
a conducting film such as Ni at the back side of the film,
then the current flows through the sample at relatively low
voltage and starts to heat the sample.

34
2.5. Low Energy Electron Diffraction
In this section, the experimental set-up for LEED
measurements will be introduced and digital LEED used for
measurement of lateral surface lattice constants will be
briefly described. A LEED optics consist of an electron gun
and a hemispherical grid system. The LEED optics used is a
modified Perkin-Elmer 15-120 LEED optics. Originally it was
a front-view LEED system, but after changing the screen into
a transparent glass screen, images can be seen from the other
side of the sample (so called Back-view LEED). Electrons from
filaments with primary beam voltage V, pass through lens
electrodes with potentials V,, V2 and V3 to focus into a
parallel beam at the sample surface. The voltages V,, V2 and
V3 are used to preserve the focus for a fixed ratio (V2 -
V,)/(V3 V,) Hemispherical grids allow one to achieve a
field-free region between the sample and the first grid, G,,
as well as to remove inelastic electrons with a potential
barrier, G2 and G3. After the elastically scattered electrons
pass through the hemispherical grids, a concentric collector
electrode (coated with phosphor) accelerates them to produce
a bright visible image in a Laue diffraction pattern
arrangement.
Electrons interacting with the sample will have various
energies and directions. A hemispherical grid system is used
to image the diffraction pattern by allowing only diffracted

35
electrons to reach the display screen. Usually the first and
the fourth grid are kept at the same potential as the sample
(i.e. grounded) to help shield the electron trajectories. Two
inner grids are operated near filament potential to select
almost elastically scattered electrons. The screen voltage
(i.e. collector bias) is operated at 2-5kV to provide
sufficient energy to excite the phosphor to reduce stray
magnetic and electrostatic fields which can cause loss or
distortion of LEED pattern. Insulators near the sample are
shielded with copper plate. After obtaining a LEED pattern,
digital LEED measurement using a vidicon camera (details in
next section) or photo graphs are taken from the back-side
view port. Photograph was taken using a commercial camera
attached with zoom lens to focus at the screen. Normal 100ASA
black and white film was used with manual exposure of 10-12
seconds.
2.6. High Resolution Electron Energy Loss Spectroscopy
The electron optics of the UF HREELS spectrometer (model
ELS-22 Leybold-Heraeus System) will be discussed in this
section from the point of view of describing the tuning
procedure. It has been emphasized that understanding the
geometry of the spectrometer and the relationship of each
potential adjustment knob determines the resolution. Detailed
tuning procedure is discussed in Appendix A.5.

36
The electron spectrometer of HREELS uses the deflection
of the particle beam (i.e. electron beam) in a electrostatic
field which disperses electrons in energy, so that a narrow
energy band can be filtered by a slit. A schematic diagram
of a spectrometer is shown in Fig. 3. Electrons emitted from
the cathode pass through a repeller to favor forward emission.
A three-element electrostatic lens is used to achieve
focussing into the monochromator entrance. Passing through
the electrostatic field of tandem-127-cylindrical-deflector-
type monochromators, a highly monochromatic (AE5meV) beam is
selected. Before interacting with the sample,
monochromatized electrons pass through accelerating optics.
After interacting with the sample, scattered electrons pass
an almost symmetric path; namely decelerating optics and two
127 cylindrical deflector type analyzers. Finally electrons
are collected by a channeltron which multiplies the number of
electrons into a short current pulse to enhance the intensity
of the signal. To ensure monoenergetic electrons pass through
monochromators or analyzers three connected slit plates are
used at each entrance and exit of capacitors. To prevent
electrons reflected by walls of capacitors, saw tooth type
surfaces are engraved on walls of capacitors.
2.6.1. Basic Theory of 127 Capacitor Electron Optics
The differential equation of motion of an electron in a
cylindrical electric field is

37
High Resolution Electron
Energy Loss Spectrometer
Sideview Toward Sample Manipulator
Figure 3. HREELS spectrometer.

38
d2y/d02 + y = E0/ (y E cos2a) (2-10)
where y is r/r0, r0 is radius of parallel path to cylinder,
E0 is an incident energy of electrons, and (r,0) and E are a
cylindrical coordinate and an energy of electrons inside the
cylinder. The value a is the angle between the trajectory
and parallel path at the same point, namely angular
aberration. A first order solution of equation (2-10) is
[19],
y E0(l-cos/2 0) / (E cos20) + cos/2 0 -{tana sin/2^ 0}//2.
(2-11)
Since the last term contains the largest angular aberration
term, to achieve the best focussing the last term should
vanish, i.e. sin/2 0 = 0. J2

The potential inside a infinite cylinder is
V(rt) = a In rt + b (2-12)
where rt is a radius of trajectory. The potential of the
outer plate is V(R), that of inner plate is V(r) the
difference between them is A = V(R) V(r) and V(ro){=0) is
the potential of main path which is parallel to the capacitor
plate. Using above conditions, constants a and b in equation
(2-12) can be expressed in terms of R,r, A and r0 as follows,

39
V(r,) = [A/ln(R/r) ] -ln(r,/r0) .
(2-13)
Electrons with their kinetic energy equal to central path
energy, and entering the sector field in the direction of the
tangent of the corresponding radius will travel in a circle
when the centrifugal force and the electric force are equal.
The electric force on an electron using equation (2-13) is
eE = eSV(rt)/Srt = {eA/[In (R/r) ] ) {l/rt)
(2-14)
The centrifugal force is mv02/rt. Equating these two forces,
the kinetic energy of an electron passing through the slit is
(1/2)mv02 = Ek,n = eA/[2ln(R/r) ] ,
(2-15)
which is same as e(C + Us) where C is contact potential
between the cathode and the slit and Us is the slit potential.
From this relation, the capacitor potential A is linearly
proportional to the slit potential Us. In the Leybold-Heraeus
ELS-22 spectrometer, R = 41mm and r = 31mm for the main
capacitors, so the theoretical slope will be 0.56 (i.e., 21n
(41/31)) when plotting of A (of main elements) versus Us.
If the potential of the main path (i.e. V(r0)) is
different from that of the slit potential, then electrons that
enter the sector field will have to cross a potential step
which causes angular distortion and decreasing output

40
intensity. In tandem capacitor systems, slits of both
capacitors were designed to locate at the corresponding main
radii. Electrons entering the sector field at a distance y0
from the main path with a velocity deviation f3 will leave the
sector field at y, (=-yo+2ro0) [20]. Since /3 has to be the same
value for both capacitors for different r, all slits should
be located on the corresponding main radii. So the expected
trajectory of tandem capacitors is the main path which is
parallel to capacitor plate.
2.6.2. Resolution and Sweeping Mode
For the deflection analyzers, the base resolution may be
expressed as a function of geometrical parameters in the
general form as follows [21]:
AEb/E0 = A AS + B a" + C /32 (2-16)
where A, B and C are constants, a(maximum angular divergence)
and p( the mean slit height) are the semiangular apertures
and AS is the aperture or slit width at the entrance and exit.
The second term (4/3)a" causes poor resolution when
targets enlarge the angular distribution of the reflected
beam. For further development of the resolution, defining
y,, as the radial deviation of the electron trajectory
measured by dropping a perpendicular down to the central path,

41
the equation of the electron trajectory up to second order in
a and using 127 condition is [22]
y, = -y0 + (AE/E0) r0 -(4/3)r0a2 (2-17)
where y0 is the deviation from the central path. With S being
the slit width, the fastest electron that can pass through is
that which enters the capacitor at S/2 with aIHX and leaves at
S/2. Its relative energetic deviation is
AE7E = S/r0 + (4/3) a^2 (2-18)
The slowest electron enters at -S/r0 with a=0 and leaves at
-S/r0 with q=0. Its relative energetic deviation is AE'/E =
-S/r0. The total energy width within the plane of deflection
is
AEf/E = AE*/E AE'/E = 2S/r0 + (4/3) 0tmx3 (2-19)
For an electron which has a velocity component
perpendicular to the plane of deflection (z-direction),
v = vz + v0 (2-20)
where v,
is perpendicular component of velocity v. Since

42
v<£=v0, vz- (time of flight)=h,

therefore v2=h/t=(h- v^,)/ (<*> r0) A kinetic energy of electrons
is
(1/ 2) mv2=m (VQ+vz2) / 2
= (l/2)mv02[l+{h2/ (0J r0J) } ] (2-21)
So the total energy width for 127 cylindrical deflection
analyzer is
AEfw/E = (2/r0)S + (4/3) a^2 + h2/(27r2r02) (2-22)
From above results, two terms of geometric deviation of beam
profile add intensity to high energy side. Especially due to
space charge, a monochromator system will contain too many
electrons with higher energy than the pass energy. But the
analyzer acts in an opposite way to the monochromator, namely
a folding process with the analyzer transmission window, the
resulting profile at the detector will be symmetric unless the
beam is not disturbed by targets. Most of the energy
broadening can be understood as an effective target
disturbance which tends to increase the angular divergence.
Electrons which lose energies from the target surface
and enter the analyzer with a kinetic energy of e(Usa+C)-hu)
will pass the analyzer when the following condition is
satisfied (see eguation 2-15),

43
e(Usa+C)-hw = 2A/[2ln(R/r)] (2-23)
Since R, r and C are constant, to compensate hu, A or Usa must
be changed. For HREELS, the AE constant mode which changes
Usa is preferred. So one moves the whole spectrum over a fixed
transmission window and the electron optical system between
the analyzer and the target acts as a zoom lens device.
2.6.3. Intensity Angular Profile
The acceptance angle of an analyzer can be detected by
measuring the intensity variation of elastic peak through the
variation of incident angle of the monochromator. For an
Si(111)-7x7 surface, the reflected elastic peak intensity
variation is shown in Fig. 4. Full width at half maximum
(FWHM) is 1.95. But the FWHM of the elastic peak intensity
variation of straight through mode (i.e. without sample) is
-1.3. This indicates that a portion of the reflected beam
width is due to residual disorder of the sample surface.
This may also due to a de-focussing of the reflected electrons
by non-uniform fields such as contact potential fields of the
sample holder components. Another variation for the elastic
peak can be derived from the primary energy. Since the
reflectivity from the sample can be varied by the primary
incident electron energy, the best resolution can be achieved
for different incident electron energies when the sample

Intensity
44
Angle (degree)
Figure 4. Angular profile of the reflected electron beam
of the HREELS system .

45
surface is changed (e.g. desorption of oxygen or vice versa).
From Si(111) with a native oxide layer after annealing at
920C for 3min under UHV, the measured resolution (i.e. FWHM
of elastic peak) and intensity of elastic peaks for primary
energies such as 5eV, lOeV, 15eV and 20 eV are as follows: (1)
resolution is 12.1meV, 13meV, 7.5meV and 9.2meV respectively
and relative intensity is 310, 360, 240 and 180 respectively.
2.7. Computer Interface
The high resolution electron energy loss spectrometer
has been interfaced to an IBM pc/XT using IEEE-488 and Camac
crate buses. The main purpose of computer controlled data
acguisition is increased reproducibility of HREELS energy
settings and reduction of noise by signal averaging. One of
the limitations of HREELS which can be overcome by interfacing
is in the scanning of pass energy of the analyzer via a motor
driven potentiometer (ramp pot). Because of degradation of
contacts, this potentiometer should be replaced regularly (
-6 months) however by using a programmable voltage source the
pot lifetime is extended. In this section, the interface of
the energy loss spectrometer to a personal computer to control
the analyzer potential and incident angle of monochromator
will be discussed. Also the structure of computer programs
and data handling will be discussed in Appendix A.6.

46
A schematic diagram of interfacing is shown in Fig. 5.
A Kepco 488-122 programmable power supply, with a resolution
of 2.442xl0'4 volts per step, has been placed in series with
the ramp potentiometer in the HREELS power supply. The Kepco
power supply which is used to set the pass energy on the
analyzer has a resolution of 1/4096 with a maximum output
voltage scale of one or ten volts. Then the ramp pot which
was originally used for scanning can be used for detecting the
elastic peak and monitoring the FWHM of elastic peak during
tuning. In the same manner another Kepco power supply is
connected to the HV amplifier sweep for ramping of
hemispherical electron analyzer voltage for XPS, UPS and AES.
The output of channeltron detector is fed through an
Ortec 109PC preamplifier to an Ortec 572 amplifier and finally
to an Ortec 550 single channel analyzer set to transmit pulses
only if their amplitude is above a preset threshold. This
signal is fed either to a ratemeter for tuning or to CAMAC
counting electronics. The CAMAC counting electronics consist
of 3610 Hex scaler and 3655 timing generator. A second
channel on the hex scaler is used to collect data from the TTL
output of hemispherical electron analyzer electronics (Leybold
Heraeus model LH-100) to collect a pulse counting signal from
electron analyzer due to AES, UPS and XPS.
One of the important achievements of interfacing with
the computer is in the precise reproducibility of the incident
angle of the monochromator. A Klinger stepping motor

IEEE488 bus
L_
1
tmrzi
/ M% m rn ginnTt>
IBM PC/XT
D/A
KEPCO
controller
t iining
generat or
count er
CAMAC
HREELS
int erna 1
s upp1y
V.
1 K
(0)
ramp
to analyzer
from detector
ORTEC
Figuer 5. Block diagram of data acquisition system.

48
controller is placed on the IEEE bus and is used for obtaining
angle resolved HREELS. Two stepping motors, one connected to
monochromator and the other connected to sample manipulator,
allow one to change angle of monochromator with a resolution
of 0.05 degree of rotation, and to change the angle of sample
with a resolution of 0.1 degree. A Keithley model 197
multimeter is connected to the IEEE bus to monitor the
baseline (DC off-set) of HREELS analyzer. The Kepco supply,
Klinger controller, Camac crate and Keithley multimeter are
interfaced by an IEEE-488 bus to an IBM pc/XT running under
DOS 2.10. Control of the IEEE bus is accomplished with a
Tecmar IEEE interface card taking up one slot in the personal
computer.
2.8. Oxidation and Hydrogen Titration
Using a gas handling system, reactive gases such as
oxygen, hydrogen and deuterium can be chemisorbed or adsorbed
on sample surfaces. In this section, two main procedures,
namely oxidation and hydrogen-adsorption, will be introduced.
Several different oxidation procedures, chemical wet
oxidation, thermal oxidation and room temperature oxidation
will be discussed. Hydrogen and deuterium adsorption on
semiconductors will be also discussed.
Since -20% of air is oxygen and oxygen is a very reactive
gas, silicon substrates can grow a relatively thin silicon

49
oxide layer in air. It is called a 'native oxide' and can be
desorbed by thermal annealing at the temperature above 1000C
under UHV. To prepare the native oxide on an Si substrate
with a small amount of carbon impurity, degreasing the
substrate is sufficient (i.e. acetone and methanol then blow
drying). Removal of the native oxide only by high temperature
annealing can not remove carbon impurities on Si surfaces in
the temperature-limit of generation of no other defects. Wet
chemical oxidation procedures have been developed. One of the
methods which was used in these studies is the so called
Shiraki technique which consist of a series of chemical
treatments as follows [10]. An Si(lll) wafer 0.5mm thick,
1/4" x 3/4" rectangular, and p-type Boron doped with
resistivity of 2.4 n-cm has been used. First the substrate
must be degreased in hot (~80C) bath (whole procedure was
done in hot bath except rinsing) in the sequence of methanol,
acetone, trichloromethane, acetone and methanol. At least
lOmin should be spent in each bath. Then the substrate should
be rinsed by deionized water (it will be called rinsing).
Second, using HN03, etching the surface region and forming an
oxide layer for 10 min should be followed by an HF oxide etch
for 10-15 secs and rinse. Repeating the HN03 and HF
sequential etch procedures must be continued until the surface
dries uniformly. Third, in the combined chemicals,
HjO:NH40H:HjOj = 4: 1: l, the substrate must be oxidized for 5
min. Then rinsing, etching by HF : H20 = 1:1 for 30 seconds

50
must follow to get rid of oxide. This is the so called alkali
treatment. This alkali treatment should be repeated at least
once more. Fourth, after a rinse in the combined chemicals,
H20 : HC1 : H202 = 5: 1: 1, the thin surface oxide must be made
for 10 min using the alkali solution. Then a rinse, etching
by HF : H20 = 1:1 for 30 secs and another rinse must follow to
remove the thin oxide. Repeating with a solution of H20: HC1:
H202 = 5: 1: 1 the oxidation and rinse must be followed by
etching using a new batch of HF: H20 = 1: 1. Then without
exposing the Si substrate to air, the HF: H20 = 1: 1 solution
must be diluted by adding deionized water. Finally the
boiling, combined chemicals of H20: H202: HC1 = 1: 1: 3 must be
used to grow a thin oxide for 2 min. After waiting until it
stops bubbling, the Si wafer must be rinsed for 10 min, spin
dried and quickly transferred into the vacuum.
Besides chemical oxidation a hot Si substrate can be
exposed to oxygen under UHV. The temperature of the substrate
was increased up to 900 K during exposure to oxygen. If the
substrate temperature is above 700 K, there are two
advantages. One is oxygen molecule can dissociate on the
substrate and become reactive. The other is adsorbed
impurities related to hydrogen species desorb quickly from
the substrate, leaving reactive sites for atomic oxygen. The
oxidation procedure is as follows. Once the amount of
exposure is determined, the total time and pressure should be
calculated. Annealing the sample at a given temperature is

51
done after turning off the ion-pump or isolating the chamber
by the butterfly valve. Oxygen molecules are inserted through
a leak valve up to a higher than calculated pressure. Quickly
opening the turbo-molecular pump line, the exact exposing
pressure should be adjusted by the leak valve. After
exposure, closing the oxygen line should be followed by
turning off heating power. After the turbo-molecular pump
reached its saturation level (in 2-3 min), ion-pump should be
used for normal operation. The reason for using a turbo
molecular pump during oxidation is to prevent hydrocarbon
species from the ion-pump from adsorbing on the substrate and
to preserve the pumping efficiency of the ion-pump by avoiding
any high pressure exposure.
Oxidation of the substrate held at room temperature in
vacuum can not be done by oxygen exposure alone. Positioning
the substrate in front of a ion-gauge during exposure of
oxygen molecules, helps to dissociate oxygen molecule into
oxygen atoms which are reactive on the substrate. Even if all
other procedures are quite similar to thermal oxidation,
oxidation at room temperature is not as successful as thermal
oxidation if surface impurities such as hydrocarbon species,
hydrogen or hydroxyl group occupy reactive sites on the
surface. Increasing the oxide thickness is limited by the
cleanness of the starting surface.
Instead of an ion-gun, a tungsten filament can be used
to dissociate hydrogen molecules into hydrogen atoms. On

52
semiconductors such as Si, GexSi,.x alloy and Ge, hydrogen was
adsorbed by placing the sample in front of a hot tungsten
filament. The effective exposure of atomic hydrogen can be
calculated by assuming that only 10'J of the H2 would be
dissociated into H-atoms. Thus the effective exposure of
hydrogen atoms is 13.5 L for 15 min. at 1.5 xlO'5 torr of H2.
The exposure procedure is the same as the oxygen exposure case
and the deuterium exposure is exactly same as hydrogen
exposure. For hydrogen exposure residual gas contamination
is expected and is often found due to reaction with the
chamber walls.
2.9. Film Growth Under UHV
Since HREELS is a surface sensitive technique especially
sensitive to impurities such as oxygen and carbon on an Si
surface, another preparation method for clean surfaces is to
grow a clean film in UHV. One additional reason for in-situ
growth is that alloy sub-monolayer films are very fragile and
cannot be prepared by the other methods described previously.
Monitoring the thickness of evaporated films is a
critical point in growing thin films for HREELS studies.
Especially in the two layer model the thickness of a film is
one of the parameters which determines the loss energy.
Thickness measurement by the variation of AES intensity and
quartz crystal thickness monitor is discussed in Appendix C.

53
In this section, the growth rate monitored using a
profilometer (Sloan, Dek-Tek II) will be discussed. Before
the actual evaporation is conducted on the sample surface, the
quartz glass crossed by tantalum wire is exposed to the
evaporation source at the same position that the real
substrate would be. After evaporation, the wire trace left
among the evaporated spot will be detected as a groove by the
profilometer. Since the thickness of this wire is 0.001, the
sharp groove made it easy to level between each ends of
groove. The evaporated film thickness is just the height
difference between end of groove and center of groove. Even
though the emission current knob (0-750 mA) in the power
supply is the only variable to control the e-beam evaporation,
the actual evaporated film thickness is not controlled
precisely by the position of the knob, because the power is
not stable. Further because the source is always changing,
the position of the current knob does not give consistent
results. Instead of the knob position the emission current
reading provided a more consistent way to reproduce the beam
evaporation conditions after several trial evaporations. In
Fig. 6. Ge source and Si source evaporation rates versus
emission current are shown respectively. The evaporation rate
as a function of emission current is different for each
material. The initial vapor pressure curve for each material
gives information on melting point and vapor pressure. A
limitation in the use of the profilometer is that some

2
Q I.III..I.I,., I I I I L
30 35 40 45 50 55 60 65
Evaporation Current (mA)
Figure 6. Si and Ge evaporation parameters.

55
materials (e.g. Ni) will not make a uniform film on a quartz
glass substrate and uneven island formations make it difficult
to determine the average film thickness. In this case a glass
substrate is replaced by another material such as Si which
allows the material to grow as a uniform film.
Silicon and germanium are materials which make thin
relatively smooth films on quartz glass to determine the
evaporation rate. Once the emission current is determined
the total evaporation time determined the actual evaporated
film thickness after enough warming up the source. Finally
details of design of evaporation system will be introduced in
Appendix A.2.

Chapter 3
THEORETICAL BACKGROUND
3.1. Overview
The interaction of monoenergetic electrons with a sample
surface will give rise to a typical energy distribution of
scattered electrons due to incident primary electrons as well
as secondary electrons from the sample surface. A typical
energy spectrum, N(E), which is the emitted electron intensity
per unit energy, is shown in Fig. 7 and is usually a strong
function of emitted angle as far as relative intensities are
concerned. The electron energy distribution, N(E), shows
three features. A large maximum occurs at low energy. This
first peak is due to electrons which suffer multiple inelastic
collisions induced by a collision cascade process in the solid
[23]. Near the incident electron energy, the sharp "elastic"
peak which is comprised of the elastically scattered electrons
plus the 'quasi-elastic' electrons that have lost a small
amount of energy (0-200meV) occurs. Finally the intermediate
region of the emission spectrum exhibits a series of smaller
maxima with a small background. These losses result from
56

dl/dE
57
(1) Secondary Electrons
(2) Quasielastic Electrons
Elastic Electrons
(3) Auger Electrons
Interband Transition
Plasinon Excitation
Figure 7. Electron energy distribution.

58
plasmon excitation (due to valence band electrons), interband
excitation and Auger electron excitation (24].
In this chapter we will briefly examine the theory of
the second feature which includes elastic scattering as well
as quasi-elastic scattering. Usually the total emitted
electron yield increases as the primary energy increases up
to several hundred electron volts. But the elastic yield
alone of the second feature shown in Fig. 7. shows a different
energy dependence. At normal incidence the elastic yield will
be largest at primary energies less than lOeV, where it
amounts to about 50% of the incident electron intensity and
decreases with increasing primary energy.
The penetration of the primary electrons into the solid
is limited by inelastic events and it is estimated that at
typical energies (10-100 eV) the penetration depth is 3-10 A
[25], It follows that the elastic component of the emitted
electrons can originate from a few atomic layers parallel to
the surface. The wavelength of the electrons (1) is h/p by
the de Broglie relation; in practical units
A= [150.4/E(eV)]H A (3-1)
where E is the kinetic energy in electron volts. At 100 eV
the electron wave length, X, is of the order of 1.3 A so that
a diffraction pattern of scattered electrons by the atomic
lattice will occur.

59
As well as losing energy to other electrons in the
crystal, the incident electron can also lose energy to the
crystal lattice by scattering off a phonon, giving up some
energy and momentum. Energy losses to phonons are very small
of the order of 50-100meV. Electrons with such a small loss
can not be filtered out by the energy selecting grids which
are used in low energy electron diffraction (LEED) optics
because their resolving power has a width of 0.5-1.0 eV; thus,
these electrons are called quasi-elastic electrons. Besides
phonon losses, the quasi-elastic peak can contain features due
to all possible small energy loss mechanisms, e.g. surface
plasmon losses due to doped carriers in semiconductor.
Elastic scattering is used to detect the long-range
ordering of the surface using LEED optics and finite
penetration depth of the incident electrons. Inelastic
scattering with low energy loss (-lOOmeV) is used to detect
surface excitations related to phonons, plasmons and resonant
electron-hole pair scattering. Two techniques, LEED and high
resolution electron energy loss spectroscopy (HREELS) analyze
the same elastic electrons with scattered electron background
due to small inelastic energy losses but each of the
emphasizes different features according to the purpose of the
measurement. For structural analysis the angular intensity
pattern of LEED is used and for determination of small energy
loss features HREELS is used since this technique resolves the
intensity features in a LEED angular pattern.

60
3.2. Elastic Electron Scattering
Before going into a description of elastic surface
scattering theory, it is better to start from the kinematic
description of scattering from X-ray diffraction from small
crystals [26]. The first Born approximation is used for
calculating kinematic scattering factors [27]. Also multiple
scattering within the atom can be included in the scattering
factor. Initially the ideal intensity of diffraction from a
crystal will be derived. Then low-energy electrons will be
considered an incident particles instead of X rays. Intensity
attenuation relative to the penetration depth will be
considered. Inner potential effects of the crystal as well
as thermal vibration effects on the diffraction intensity will
be discussed.
3.2.1. Diffraction From A Bulk Crystal
: 3-dimensional Diffraction
Assuming a sample is uniformly illuminated by the
incident beam and ignoring multiple scattering, the scattering
amplitude of an incident plane wave from N atoms, whose
scattering factor is ft(^,E), is [26],
A = A* 2 f i (0 E) exp (iSr,), (3-2)
where r, is the position of i-th atom and S = k k is the
transferred momentum. The factors k and k<, are the

61
propagation vectors of scattered and incident plane wave.
Assuming each scatterer is at the lattice point
r¡ (=m,a+m2b+m3c) and atoms are located at xn in each unit cell
n, the normalized amplitude is
A(S) = S f(0,E) exp[iS- (r,+xn)], (3-3)
i .n
where the sum is over the lattice sites i and the atoms within
unit cell n. Separating the sums, then
A(S) = [E fn(0 ,E) exp(iSxn) ] E exp(iS r,)
n i
= F(0,E)- E exp (is-r,), (3-4)
i
where F(0,E) is the crystal structure factor. The scattered
intensity is then written as, I(s) = |A(s)|2,
l(S) = | F(0 ,E) |2 E exp[is-(r.-r,)]
i, j
= |F(0,E)|2 J(S). (3-5)
The interference function J(S) depends on the diffraction
geometry through transferred wave vector S. Both F(0,E) and
J(S) differ for different choices of non-primitive unit cells.
For a parallelepiped with N,, N2 and N, lattice points, the
interference function J(S) is

62
sin2(N,S a/2) sin2(N2S b/2) sin2(N3Sc/2)
j(S) :
sin2(S-a/2) sin2(Sb/2) sin(S-c/2)
(3-6)
The intensity becomes a maximum when S satisfies the Laue
conditions, i.e.,
S a = 27rl, S b = 27rm, and S c = 2nn. (3-7)
In terms of reciprocal lattice vectors A, B, and C, which
are A=2ir (bxc) / [a*bxc] and so on, the position of the
reciprocal lattice point (lmn) is GlBn = 1A + mB + nC, which
is a vector normal to the family of planes whose Miller
indices are (lmn) with a magnitude 16lmn | = 2n/dlnn where d,n
is the interplanar distance. Then the Laue conditions are
S a = 27rl = 27rGllnn-(a/27r) = GlBn a and so on for b and c; thus,
S = GlBn = k K (3-8)
The diffraction geometry is displayed by the Ewald
construction shown in Fig. 8(a). The incident wave vector
has a fixed magnitude and direction and is terminated at the
origin of the reciprocal lattice. The origin of the Ewald
sphere whose radius is the magnitude of incident wave vector
is at the same position as the origin of the incident wave
vector in momentum space. Whenever this Ewald sphere passes
through a reciprocal lattice point, the diffraction condition

X Reciprocal Lattice
I(S): Interference function
along a reciprocal-lattice rod.
(b)
ON
w
Figure 8. The reciprocal lattice and Ewald construction,
(a) Three dimensions; (b) Two dimensions.

64
S = Ginn will be satisfied for the ray terminating at that
point. Diffraction maxima gives the coordinate 6 and thus G,n
and unit vectors a, b, c. From equation (3-5), the intensity
of a diffraction maximum is |F|J2. But the modulus of F|W) does
not uniquely determine the atomic arrangement within a unit
cell for non-Bravis lattices with more than one atom per cell.
3.2.2. Low Energy Electron Diffraction: Surface Diffraction
A beam of low-energy electrons passing through a material
is attenuated by both elastic and inelastic processes.
Inelastic processes are usually treated as simple attenuation
by modifying the kinematic description derived above.
Assuming the ratio of the amplitude contributed to the
scattered beam by atoms in successively deeper planes is a =
An.t/An where n is the plane number. From equation (3-4), the
total amplitude for a semi-infinite crystal is
A (S) = f(0,E) E E exp [is (m,a+m2b) ] E a"3 exp(im3Sc)
1^1 1^2 HI 3
= f (0 E) E E exp [is (m,a+m2b) ] [1-a exp(is-c)]'1
m, m2
The corresponding intensity is
(3-9)
sin2(N,S a/2) sin2(N2S b/2) 1
I(S) =|f(*,E)|2
sin2(S a/2) sin2 (S-b/2) 1+a2 -2acos (S c) ,
(3-10)

65
the diffracted intensity now satisfies only two of the three
Laue conditions, Sa=27rl and Sb=27rm, and is confined along
lines in reciprocal space normal to the crystal surface,
specified by S = G,. The reciprocal lattice and Ewald
construction for a two dimensional lattice is shown in Fia.
8(b). The lines are referred to as reciprocal-lattice rods
and indexed by two integers (lm). These results which depend
upon two dimensional periodicity of the surface are the basis
for using LEED to determine surface structure. The modulation
of the interference intensity function along the reciprocal
lattice rod is determined by the factor [l+a2-2a cosS-c]'1 .
The broad maxima exist at positions where the third Laue
condition, 8c=27rn, is satisfied. The breadth of these peaks
is a consequence of attenuation. In the case of zero
penetration, intensity depends only on the scattering factor
|f(*,E) |2 .
In addition to the attenuation factor, incident electrons
experience a different potential while passing through the
crystal. This periodic potential in the crystal half-space
can be expressed as a Fourier expansion
V(r)=E Vc exp(iG-r), (3-11)
and its spatial average V0 is usually called the inner
potential. Due to this inner potential V0 the magnitude of
wavevector in the vacuum is still k= 27r(E/150.4)>' A"' but in

66
the crystal it is kln=27r[ (E+V0)/150.4]* A'1. Conservation of
parallel crystal momentum causes a refraction of the electron
beam inside of crystal with refraction index
sin* kln[l+(V0/E) ]"
n = = (3-12)
sinfl(n k
The main contributions to the inner potential come from
correlation and exchange interactions as well as the surface
dipole layer potential and an imaginary part of the potential
due to inelastic interactions.
Another factor determining the intensity is the thermal
vibration of atoms in the crystal. The energy resolution in
typical low-energy electron diffraction is insufficient to
observe the loss or gain of phonon energies, so the intensity
measured corresponds to both true elastic and integration over
inelastic events due to phonon scattering. A diffraction
experiment is equivalent to scattering from instantaneous and
stationary configurations of scatterers. From equation (3-
5), the instantaneous scattered intensity is
N
I(S) = | f (0 ,E) |2 2 exp[is-(r.+u.-r.-u,)], (3-13)
i j
where u, is the instantaneous displacement of the i-th atom
from its equilibrium position r,. The thermal average of
intensity is

67
N
= |f(0 E) |2 exp[-<(Su)2> E exp[iS-(r.-rj ]
i j
(1+<(SU,) (S U,) >+ [exp< (8 u,) (8 Uj) >-l-< (8 U|) (8 !!,)>] }.
(3-14)
The factor exp[-<(Su)2>] is Debye-Waller factor. The first
term in the curly bracket is the zero-phonon scattering which
is just the rigid-crystal scattering reduced by the Debye-
Waller factor as discussed previously. The second term is the
one phonon contribution to the thermal diffuse scattering.
The third term is multiphonon scattering. So the effect of
thermal vibration is to remove the intensity from the Bragg
peaks and to redistribute the intensity throughout the
Brillouin zone. Multiple scattering and inelastic process can
be treated in self-consistent way to understand more detailed
elements in diffraction problems. This is the goal of
dynamical theory which will not covered here.
The theoretical treatment of elastic scattering of low-
energy electrons within a kinematic framework can be
summarized as follows. Low energy electrons are scattered
within a few atomic planes of the surface. Two dimensional
diffraction pattern intensities with information on long-range
order and the structure of the surface layer are modified due
to both elastic and inelastic process. A full treatment
involves the inner potential of the crystal, attenuation due

68
to electron-electron scattering, and thermal vibrational
motion of the atoms (i.e. the Debye-Waller factor).
3.3. Inelastic Electron Scattering
Two conservation laws are the basic starting point for
inelastic electron scattering theory and are summarized as
AE = E, Es = hw and Ak||= klj(- ksN= q,( + G(,. (3-15)
Here AE is the energy transferred to the crystal and Ak,, is
the momentum transferred parallel to the surface and GN is a
reciprocal lattice vector of the crystal surface. Even if
the vertical component of transferred momentum is large, a
simple relation for it can not be established due to the
broken symmetry normal to the surface. Approximate methods
can be modelled to give periodicity of the z direction using
periodic layers which consist of a large gap of vacuum to
avoid the interaction between the surface and the substrate
layers (-15 layers). Such a model will allow estimation of
scattering near the surface which is quite similar to the
system of vacuum and semi-infinite substrate.
From the law of conservation energy, an inelastically
scattered electron can lose or gain energy depending on
whether a phonon is created or annihilated. The relative
probability of a phonon creation and annihilation is governed

69
by Bose factor, n=[exp(hw/kT)-l]'1. The ratio of intensity of
gain to loss peak (sn/(n+l)) is exp[-(hw/kT)].
From the law of conservation of momentum, the loss energy
evolution relative to transferred momentum, either by changing
geometry of scattering from specular to non-specular or by
increasing primary energy at specular geometry, will give a
dispersion relation of each mode. Based upon these two
conservation laws, there are various applications of HREELS
for the detection of surface excitations. The primary object
of HREELS theory is then to describe a precise scattering
differential cross-section which indicates the loss energy,
its intensity and angular distribution.
In this section we will follow a semi-classical
derivation of the differential cross section applied to a
simple system which consists of an electron approaching an
adsorbate on a metal surface [28]. The reasons it has been
chosen are as follows: first it has several simple steps to
derive the end result and is matched well with experiment,
second it is easily connected to the actual geometry of the
experimental set-up, and third it can be extended to further
complicated, more advanced quantum mechanical theories. In
addition to this semi-classical theory, the result of a
quantum mechanical theory applied to a similar system by
Persson will be compared with the semi-classical theory [29].
Also the result of a full quantum mechanical description
applied to a more generalized system (i.e. dielectric function

70
theory for semi-infinite medium) will be introduced [30].
Even if impact scattering theory is not fully developed yet,
the important characteristics of impact scattering have been
reported. This impact scattering will be briefly discussed
and finally examples and applications of inelastic scattering
theory to specific systems will be presented.
3.3.1. Semi-Classical Approach
The electric field due to the specimen extends over the
vacuum above the specimen. The longest range electric field
is due to the surface dipole of the specimen. Considering
the total range of this dipole field above the specimen, the
time which the electron stays within this range is longer than
the time it stays inside the specimen. This long range
interaction is the so called 'dipole scattering' which is
applied to specular geometry since the scattering intensity
is sharply peaked around specular geometry. The scattered
intensity of this dipole scattering applied to the system of
an adsorbate on a metal surface will be derived following the
description by Newns [28]. This theoretical approach to
dipole scattering used here is the semi-classical one, since
it treats an incident electron as a point particle traveling
along a single trajectory remaining in the vacuum above the
specimen at all times and exciting surface vibrations by means
of its long-range coulomb field. The schematic diagram of the
scattering is shown in Fig. 9(a). The coordinate of the
electron is

71
re: electron-dipole distance
(V|,,vx): electron velocity
H: molecular instantaneous dipole
and its image in the surface
Mi: resultant normal dipole
(a)
Figure 9. Schematic diagrams of semi-classical dipole
scattering.
(a) Electron trajectory and molecular dipole moment;
(b) Transferred momentum; (c) Polar plot of
scattering intensity.

72
w0/k: parallel component of momentum
change to reflected specular
beam
k0: perpendicular component of momentum
change to reflected specular
beam
(b)
a: 45
Figure 9.continued.

73
re = ( x0+vt, vjt| ),
(3-16)
where the origin is at the impact point and xQ is the impact
parameter in the surface plane. The term is the parallel
velocity and vx is the normal component of velocity.
Since the dominant electron-molecule interaction comes
from the part of electron trajectory where re is large,
interaction between the incident electron and adsorbed
molecule with dipole / can be written
V = S V (l/re) (dn/dq,) q, .
(3-17)
The dipole is varying slowly at frequency w0 compared to
typical electron motion in a metal and the fast metal
electrons will follow the instantaneous dipole motion
adiabatically. Thus, the parallel component will have an
opposite image dipole, while the vertical component will be
approximately double the strength as long as the dipole is
not imbedded into the metal. Only the normal component of
the dipole moment has a non-zero perturbation. This is the
consequence of the so-called normal dipole selection rule.
The detailed derivation of this differential cross section is
in Appendix B.l. For normal incidence vM=0 which gives nQ=w0
, and the differential cross section da is,

74
da = { rJ 4QV / [W02 +Q2vx2]2 } d2Q (3-18)
This function is strongly peaked near the characteristic wave
vector Qo~Uo/vx. In typical experiments, uo=0.02 a.u.=54 0 meV
and a primary beam energy E=0.20 a.u.=5.4 eV, then Qo0.01
a.u.. Since only small values of wave vector Q contribute to
the scattering intensity, it can be deduced that the effective
range of electron-molecule scattering is of order Q0"1 or at
least 60 a.u. in the example. When l/re is expressed as a
Fourier transform,
1
re
1
2n2
d2 Q
dq
exp(-iQx) exp(-iqz)
Q2 + q2
1
2n
d2 Q e'iQ
(3-19)
thus, the potential V has e"Q^ factor, which also indicates
that perturbation extends above the specimen up to z~Q0'.
This is the underlying justification for using the long range
dipole scattering interaction. One of the important reasons
this semi-classical differential cross-section is chosen, is
that the differential cross-section da is easily transformed
to measurable quantities, e.g. k (incident wave vector) and
measurable angles.

75
In Fig. 9 (b) the outgoing elastic wave vector k, the
outgoing wave vector k, after excitations of phonons and full
scattering wave vector q are shown. Then
k = (Vy.vJ = k(sina, 0, cosa) (3-20)
where a is polar angle of k and q = k -k, = (Q,q2). Defining
6 and 0 as the polar angle and the azimuthal angle
respectively of kt relative to k (i.e. 0=0 when k, is under
k on the oxz plane), fully determines the scattered beam
direction of k,. Energy conservation then gives
w0 = h (k2 -k,2 ) k (k-k,) (3-21)
where h=e=m=l, and the second equation comes from relations
q/kl and 81. For the normal incidence case (i.e. a=0) and
0=0 ;
q^0=k-k,=k(0,0,1) -k, (sin0,0, cos) ,
~k(-0,0, [k-k,]/k) k(-0,O,wo/k2 ) ,
ak(-^,0,^o), (3-22)
where the characteristic angle 60 is given by u0/(2E). The
differential cross section derived by Newns is
4T2 cosa f(0,0,a) 86d(p
da
k2 (02+8o> )
(3-23)

76
where f (0,0,a) is [ (0cos0-0otana)2+02 sin2 0sec2 a] The
detailed angular transformation procedures are in B. 2. The
denominator term shows that the differential cross section
peaks strongly in the region 8<60 where 0o=wo/(2E). This
strong forward scattering is a consequence of the small
momentum transfer parallel to the surface. A polar plot of
the scattering intensity of equation (3-23) for the case of
a=45, 0o=O.l rad., and 0=0 is shown in Fig. 9(c). The nodal
value comes from f {6,0,45) =0, i.e. 8 = 0otan45 in this case
and forward scattering lobe is very distinct.
Considering the aperture of the analyzer (cf.refer the
measurement of intensity profile Fig. 4 in chapter 2) the
analyzer accepts scattered electrons lying in a cone of apex,
8,. The total cross-section, a, due to this finite acceptance
of analyzer can be integrated up to 8, which is the polar
angle of scattered electron with respect to specularly
reflected direction.
a
'6, ?2n 4T2cosa f(0,0,a)0
d8 d0
0 J0 k2 (02+0o2)2
7rr2
cosa ( (t2-2)Y+(t2+2)lnX] (3-24)
E
where Y=0,2 / (0t2 +60> ) X=1 +(0,2/0o2) and t=tana.

77
Assuming the surface coverage of adsorbates is n molecules
per unit area, the current of one phonon loss (I,) versus that
of unscattered specular reflection (I0) will be
I,/I0 = (77TJn/E) cosa [(t2-2)Y+(t2+2)lnX]
(3-25)
The main feature of dipole scattering of the adsorbate on
metal surface using a semi-classical approach is that the
scattered electrons form a strong forward scattering lobe with
vibrational modes polarized vertical to the surface.
Such dipole scattering was also described by Persson
using a quantum mechanical approach [29], In this treatment
the electron wave function near scattering region is
k (27T) [ e + e ] e
(3-26)
where k'=k-2n(nk) and 5 is the phase factor due to
scattering. Using the Fermi-Golden Rule and a perturbing
potential H'= -|x-E due to the elastic field of external
electrons and the image charge, the probability per unit time
for vibrational excitation of the molecule was calculated.
The differential cross section is
da/dn = (m/xe/7re0h)2 [p,/(p0cosa) ] [a/a2 + (bnCosS+^sinS) /b2 ]2 ,
(3-27)

78
where a=k0-k1, b=k0'-k1, p,=hk, and p0=hk. For the specular
geometry condition, (an/a2 ) (bN/b2 ) or (byb2 ) ,
dcr/dn = (m/ie/7T60h)2 [p,/(p0cosa) ] [a,|/a2 ]2 (3-28)
The total cross section, a, for the electrons collected by a
cone of half-angle 8, around the specularly reflected beam is
o = (Me/he0v)2 (cosa/27T) [(t2-2)Y+(t2+2)X], (3-29)
where X,Y,t and a are defined the same as in equation (3-24).
Comparing these results of da/dn and a with the semi-classical
approach, we see that a quantum mechanical approach agrees
with the semi-classical approach.
3.3.2. Dielectric Function Theory
In the previous section, electron scattering was limited
to the system of adsorbed molecules on metal surfaces, but
dipole scattering can be produced by any excitation of the
sample accompanied by a fluctuation in charge density [31],
In this section the general description, using a quantum
mechanical approach, for the differential cross section
related to the dielectric function of the layer will be
outlined following the derivation of Mills. The potential

79
outside the semi-infinite (z<0) specimen in which a time
dependent charge fluctuation, n,(x,t), at a point x is
<*>(x,t) =
n,(x' ,t)
dx'
| x x'| ,
(3-30)
where the integration extends over the specimen. Retardation
effects are ignored since the time scale of information
transfer from specimen to incident electron is very short
compared to the period of excitation, which is also assumed
in the previous section. Using the Fourier transform of
x x' |_1 from equation (3-19), the potential seen by the
electron outside the specimen can be written as
0(X,t)
e'inr e-fl"z
n, (Q||Z ;t) e5"7 dz '
(3-31)
where
n,(Q||Z';t) = Jd2x' e*1*" "N n,(x',t) (3-32)
From the first integral, the factor e'<5)l 7 indicates that the
potential decays exponentially. The potential extension above
the specimen (vacuum) is z=Q||*, which is the height the
electron starts to experience the potential due to a component

80
with wavevector Q|(. The factor e0*1 z in the second integral (of
equation (3-31)) which is the charge source integral indicates
that the contribution to the potential with component
exp (iQ|| X||) is produced by charge fluctuations that extend down
to a distance Qn'1 below the surface of the specimen.
The differential cross section can be obtained by
inserting equation (3-31) into the Schrdinger equation and
using Born approximation,
da 2 | R |2 v/ k' S (Q|(, id)
dn, dhu hk7r(ea0)2 cos, [vx2 Q,,2 + (w-Q(| vH)2 ]2
(3-33)
The term a0 is the Bohr radius, is the angle of incidence
relative to the surface normal, k and k' are the magnitudes
of the incident and scattered wave vectors, | R|2 is the
reflectivity for specular scattering, and Q,|= kn-JCji'. The
spectral function S(Q||,w) is
S(Q||,U) =
00
d2X|,
dt exp [ iQ||-X||-iwt ]
CO
dz' | dz exp[Q,|(z+z') ] T ,
CO
J
DO
(3-34)

81
where x = x,,+zz, and the brackets with subscript T denotes the
average of the quantity enclosed over the appropriate
statistical ensemble at the temperature T. Defining the
scattering probability P(k,k') by,
1 da
S(Q|,,w)
[VxJ Q,i2 +(w-Qli'V|l)2 ]2
(3-35)
Thus P(k,k' )dflk dhu is the probability an electron scattered
into the solid angle di\ in the energy range between hw and
h(u+diJ), normalized to the elastic intensity (here, R^R^-R was
already assumed). The scattering probability shown in
equation (3-35) has a kinematic factor which is equation (3-
35) itself. This kinematic factor is independent of the
property of specimen and peaks for small momentum transfers
at Qn~k(hw/2E) Since QN* is the range of the potential as
well as the probing depth, shown in equation (3-31) and its
following discussions, the kinematic factor derived in
equation (3-35) has shown a sharp forward scattering lobe by
the potential with angular width of 2E0/(hw) and the effective
probing depth of HREELS also turned out 2E0/(khw) from
equation (3-34).
|R|2 dnk d(hu)
2 Vj4 k'
h7T (ea0)2 cos, k

82
The quantity S(Q(|,w) in equation (3-35) contains
information on the specimen surface. Since a charge
fluctuation n, (x,t) generates an electric field E(x,t) by
Maxwell's equations, the charge density correlation function
can be replaced by the correlation function of the electric
field fluctuation, T. This can be related
to the dielectric response functions of the substrate through
use of the fluctuation-dissipation theorem [32]. The spectral
density, S(Q)(,to) can be constructed by means of a Green's
function method [31], and the result is
S(Q,W) = (2Qn/tr) N(to) lm[-l/{l+e(w) }], (3-36)
where N(to) = [exp(hto/kT) -1 ]'' and e(to) is the dielectric
function of specimen. The scattering probability is
4 vA4 k' Qii N(to)
P(k.k') =
Iitt2 (ea)2 cos, k [vA2 Q2 + (w-Q,, v|()2 ]2
-1
x Im[ ] (3-37)
1+e(to)
Equation (3-37) suggests that the information of scattering
geometry and dielectric function of specimen are required to
estimate the scattering probability.

83
Summarizing the results of this section, the scattering
probability near specular scattering geometry consist of two
factors. One is kinematic factor fully determined by the
scattering geometry and the other is the spectral function,
S(Qh,w) which contains information of surface property, e.g.
a dielectric property. Also this general description of the
dipole scattering probability of a semi-infinite medium
confirms that the scattering probability sharply peaks near
the specular direction (i.e. forward scattering). The
electric potential due to the specimen (i.e. 0>z>-Q||'') extends
up to a range of Q||"', where Qn is the transferred momentum.
3.3.3. Impact Scattering: Off-Specular Scattering
In dipole scattering theory, to simplify the derivation
of scattering probability, major assumptions are single
scattering and very small amount of momentum transfer parallel
to the surface. Another regime of scattering accompanied with
relatively large momentum transfer parallel to the surface is
called impact scattering. To induce large momentum transfer,
the incident electron energy can be increased at a specular
geometry. Since the aperture of the analyzer has a fixed
solid angle (A0~1) electrons passing the edge of the
aperture of the analyzer have different momentum for different
electron energies; i.e. the larger the energy is, the larger
the transferred momentum is. The other way to increase
transferred momentum is changing the geometry of spectrometer

84
from specular (i.e. 0,nci a fixed incident energy. But a mixing of both ways (i.e.
energy and geometry) is not desirable for experiments
observing the evolution of peak intensities since the
reflectivity from the sample depends upon the incident energy
very much. The first method can be simply used for
observation of evolution of peak position. To detect the
dispersion relation of surface excitation, incident energy
should be increased up to a few hundred electron volts in both
methods to cover the entire Brillouin zone whose size is
typically a few A'1. Another reason for using impact
scattering is to determine local site symmetry from surface
vibrations. Since large transferred momentum induces
increased surface sensitivity
(e'9"2 and e^z factors in potential and source, see equation (3-
31)), a microscopic treatment is needed to interpret the
spectrum. It is not so simple as the dipole scattering case
to derive the scattering probability since the high incident
energy (-300 eV) to cover the Brillouin zone edge causes a
multiple scattering as we have already discussed in the
overview of this chapter. So it is necessary to approach the
scattering problem from other direction to get the
differential cross-section of this multiple scattering. In
this section a condensed account of Mill's derivation and
results will be presented [33]. Also the selection rule for
impact scattering will be briefly introduced.

85
The basic idea in impact scattering theory is the
calculation of intensity of electrons which contribute to the
thermal diffuse background of a diffraction pattern. Since
dipole scattering is rapidly reduced at off-specular geometry,
the intensity of the background is mostly due to impact
scattering. When an electron encounters a solid, the
positions of nuclei are not fixed and are displaced by thermal
vibrations. The position of nucleus i1 is then R^R^+Ui where
Ro, is the position of the equilibrium site and u¡ is the
displacement from equilibrium position, R,,,. For small
displacements of u, the scattering amplitude f(ks,k, ;R) can be
expanded in powers of ua ,
df
f(ks,k,;R) = f (ks,k, ,Ro) + E ( )0ua + (3-38)
a 3R a
where ua is the ath Cartesian component. Expressing u0 in
terms of the normal mode eigenvectors §a!,
h
ua = 2 ( )* §as (as + as*) (3-39)
5 2SM,
where 's' refers to a particular normal mode and as,as* are the
annihilation and creation operators of vibrational quanta, and
Mi is the ionic mass. When a particular vibrational quantum
is emitted, the matrix element

86
M(k,, Jcs;+s) =
= (ns+l)*(h/2Nws)>'(af/aQs) (3-40)
where (3f/3Qs) = S(3f/aRa)0 §a/MlH and ns=[exp(hws/kT) 1). Then
the probability that the vibrational quanta (Q(|Qt) scatters the
electron into the solid angle dil from the surface area 'A1 is
d£a(k,,ks) mE, cos20s
= A |M(k(,k,;Qlla)|* (3-41)
dil 2?r2 h2 cos0,
where a contains all indices other than wavevector QN, and 6,
and 8S are an incident angle and a scattered angle. A further
analysis of multiple scattering will not be covered here.
Selection rules for impact scattering provide a basis
for obtaining direct information about symmetry of an adsorbed
atom based on the general feature of the inelastic single
scattering cross section. Selection rules are based on the
symmetry of the substrate, time-reversal symmetry, scattering
geometry and the direction of the polarization of the
vibrational mode. The results by Tong et al. were as follows
[34], If a normal mode is polarized out of the scattering
plane as well as parallel to the surface, and the substrate
has a reflection symmetry relative to the plane perpendicular
to the scattering plane, the cross-section in any position in
the scattering plane is zero. If the substrate has a
rotational symmetry about z axis in the above case, the
differential cross-section is zero at only specular geometry.

87
If the normal mode is in the scattering plane as well as
parallel to the surface, and the substrate has a reflection
symmetry plane, the cross-section is zero at the specular
geometry. If the substrate has a rotational symmetry about
the z axis in the above case, the cross-section is zero at the
specular geometry. If the normal mode is vertical to the
surface, the cross-section is almost zero at off-specular
geometry which was already stated in dipole scattering theory.
But these selection rules assume the condition of single
scattering and they can not be applied in case of rapid
variation of the reflectance or high energy multiple
scattering [35].
3.4. Examples of Inelastic Scattering
3.4.1. Surface Optical Phonon Excitation
Surface optical phonons which have finite frequency in
the limit Q(|-* 0, are due to specimen with more than one atom
in one unit cell. When a crystal lattice has two or more
atoms in one unit cell and each atom is electrically neutral,
like Si, Ge and some metals, only forces of short range become
important. Then, in the limit Q(|-> 0, the atomic displacement
related to the surface optical phonon are non-zero only in the
outermost few layers. On the other hand, in ionic crystals
where the Coulomb interactions between the ions produce
couplings extended over large distances, the optical phonon

88
amplitude penetrates deeply into the crystal as Q,|-> 0. This
surface optical phonon mode from ionic crystals, e.g. ZnO,
GaAs etc., is called 'Fuchs-Kliewer' mode [36]. There are two
ways to derive this 'Fuchs-Kliewer' mode. In this section the
method using the result of dielectric function theory for
infinite crystals derived in the previous section. The other
is the method using lattice dynamical frame work which is
derived in Appendix B.3.
If the crystal has an ideally terminated-surface like
bulk, the dielectric function of a cubic material with one
infrared active bulk transverse optical phonon is
47rne'2 1
eb(w) = cbo + (3-42)
Mr wT02 u2 iwr(w)
where &o is the dielectric constant at high frequency limit,
n is the number of unit cells per unit volume, e' is
transverse effective charge, Mr is the reduced mass of the
unit cell, and r(w) is the damping function of the oscillator
[37]. The static dielectric constant e(0) is
e(0) = feo + (47rne'2/Mr) (wT02) (3-43)
From equation (3-37) the loss function has a peak at the pole
of the loss function, Im[-l/{l+eb(w) ) ]. The loss peak, which
is the Fuchs-Kliewer mode, is

89
e(0)+l
Us = Mto [ ]* (3-44)
600 +1
where the damping function r(u) is assumed very small. The
Fuchs-Kliewer mode us is larger than the bulk transverse
optical phonon wT0 and smaller than the bulk longitudinal
optical phonon ul0 (=wT0[ &x/e (0) ]*) .
3.4.2. Surface Plasmon Loss and Dispersion: Relation of
Homopolar Semiconductors, and Plasmon-Phonon Coupling
of Polar Semiconductors
In the infrared regime, there are three elementary
excitations such as infrared active optical phonon in ionic
crystal lattice, free carrier plasmon in semiconductor, and
direct interband transition in small gap material. Among
them, a plasmon is due to a collective motion of charge
carriers. The frequency, wp depends upon the number of
effective carrier. If carriers are loosely bound valence band
electrons, the surface plasmon energy is several electron
volts, which can be detected in low resolution electron energy
loss spectroscopy. But, if the carrier concentration is due
to free carriers doped in a semiconductor whose number density
ne(( 10,5-10/cm3, the plasmon energy varies from a few meV to
-lOOmeV which can be detected by HREELS.
Polar-semiconductors (e.q. GaAs, NiO, ZnO, ...) have a
strong surface optical phonon mode due to the ionic character
of each unit cell coupled to plasmon mode while homopolar

90
semiconductors (e.q. Si, Ge, SixGet.) do not have a strong
optical phonon mode. In this section, for both semiconductors
the plasmon loss and their dispersion relation will be
examined.
For a homopolar semiconductor the loss function,
Im[-l/{ e(u)+l} ], from the semi-infinite crystal with
dielectric constant e(w) is shown in equation (3-37). For
the homopolar semiconductor (from now on Si will be used as
an example), dielectric function is
Up2
eb(u) = to (3-45)
u[u+{i/T(w))]
where 6 is the background term, wp2 =47rnefle2 /m, and t (u) is a
frequency dependent relaxation time [37]. The second term is
the Drude term contributed by the free carriers and a surface
optical phonon term is ignored since an Si is a non-polar
semiconductor. The loss function combined by this dielectric
function is
-1 W(dsp2
I*>{ } = [(wsp2-w2)2 + w2/t2(u)]" (3-46)
l+eb(w) (l+&o)r(w)
where usp2 =up2 / (l+e) and w2/r2 determines the breadth of the
loss peak in spectrum. For different doping densities (neft=

91
10,5-5xl0" /cm3) the calculated intensity of loss function,
(hw)''lm[-l/{eb(w)+l) } ], is plotted in Fig. 10 fa) The
parameters 6*5, m'(=0.232 mj for p-type semiconductors and r(u)
have been chosen to represent silicon at room temperature.
The loss function has a maximum at the surface plasmon energy
>Sb2 (=4jrne2/(m[ l+&o] }) This loss peak is due to specular
scattering geometry. If the dielectric function is replaced
by the Lindhard dielectric function which describe the highly
doped semiconductor and contains momentum transferred parallel
to the surface, q,,, then the loss function, Im[-l/{ e (o, q^j) } + l] ,
will give the dispersion relation, i.e. ws vs. q|(. The surface
plasmon loss, ws, comes from the condition
op2 u2
-1 = 6- (3-47)
U2 O2 -0.6 q||2vf2
where vf (=4.20x10'* a0/rs cm/sec) is the average Fermi velocity
of free carrier, a0 is Bohr radius and rs=[3/(47rnefl) ],/3.
Transferred momentum, qj(, comes from
q,! = 2n (Eo/150.1) (sin0,-sin0,) (3-48)
where E0=E,ES and 0S,0, are scattered and incident angle with
respect to surface normal. For different doping densities
(nef,=10'5-10'* /cm3), the dispersion curves (ws vs. g,,) were

Figure 10. Plasmon loss calculation and plasmon-phonon
coupling calculation.
(a) Plasmon loss calculation for Si(111) at the
specular geometry;
(b) Plasmon dispersion curve for Si(lll);
(c) Plasmon-phonon coupling for GaAs(lOO).

(B)
10 20 30 40
r
A
Plasmon Energy Loss (meV)

94
Momentum Transfer ('1)
(b)
Figure 10.continued.

95
(c)
Figure 10.continued.

96
plotted in Fia. 10 (b) When q,|=0, i.e. at specular geometry,
the losses matched with Fia. 10 fa).
The cut-off wave vector (qc) above which there is
continuum limit, is wp/v,. Therefore qc also depends upon the
doping density by net,,/6. So fitting the curve to the actual
data can be checked by this cut-off wave vector.
On the other hand for polar semiconductors the surface
plasmon frequency is very close to that of the surface optical
phonon, these two modes can not be an independent mode, but
they couple to form two new modes which are admixtures of a
surface plasmon and a surface phonon. Then the dielectric
function for the ionic crystal with carriers can be expressed
in terms of dielectric function of longitudinal optical
phonon, i.e. (equation (3-42)), and Lindhard dielectric
function for highly doped semiconductors as follows:
(c0&o) wT02 up2
(W, qj,) = &o + (3-49)
wT02 w2 iwr u2-qs2 + (iw/T)
where r and 1/r are damping terms, qs2=0.6 q^ vF2 and all
others defined same as the previous section. From the pole
of the loss function, £(w,q|()=-l, the coupled phonon-plasmon
losses are
W2 (qs2 ) = h [A +qs { (A+qs2 )2 -4Bqs2 -4C)*] ,
(3-50)

97
where A*[uro*/(1+A) ] + [p*/(1+ebo) ], B=wT02 (1+A) ,C=wp2 wT0J/(&ofl),
and 1+A=(e0+1)/(&<+l) Dispersion curves (i.e. energy loss
vs. momentum transfer) with neff=6xl0" /cm3 of equation (3-50)
are shown in Fig. 10 f c) High coupled mode behaves like a
optical phonon initially at small momentum transfer, while
low coupled mode behaves like a plasmon. After cg,| pass cross
over point, these two behaviors are exchanged. Coupling has
the effect of repelling these two mode further apart, i.e.
lower mode becomes lower and higher mode becomes higher energy
loss.
3.4.3. Two-laver dielectric-function model
Ideally bulk terminated surface can not be found since
the source of the potential above the surface is zero. So
the surface layer always exists when the surface is produced.
In this chapter at the section on dielectric function theory,
the probing depth is estimated as -O/' where QN is the
transferred momentum. If the surface layer is thicker than
Qil'', then dielectric function theory described by equation (3-
37) can be used as e(w)=es(w) where es(w) is the dielectric
function of the surface layer. If the thickness of surface
layer (sd) is smaller than QN"' then equation (3-37) can not
be used since different dielectric functions contribute to the
loss. Equation (3-37) can be modified through including
parameters such as thickness of the surface layer, d, the
dielectric constant of the surface layer, es(w), and the

98
dielectric constant of the bulk which lies under the surface
layer, eb(w). Since the scattering geometry is not changed,
this modification only affects spectral density S(Q||,w). In
here, the result of the modification by Mills and Maradudin
will be introduced [38]. Two layer spectral density S2(Q(|,0))
can be derived by replacing c(w) in the spectral density
S(Q||,w), i.e. equation (3-36), by e,(w), where
1+A (w) exp(-2Q||d)
et(w) = es(w)[ ] (3-51)
1-A (w) exp (-2Qd)
where
[eb(w)-es(w) ]
A(w) = and N(u) = [exp(hu/kT)-l]'. (3-52)
[eb(w)+es(w) ]
If d-0,
et(w) = ,(w) [{1+A(w) )/{l-A(u) )] = eb(w). (3-53)
If Q,|dl,
e'2q,l<, l-2Qd, (3-54)
and the loss function Im[-l/(e, (w)+l) ] can be divided into two
terms,

99
-1
Im
[e,()+l]
-1
Im
[b(M)+l]
+ Im[ (2Q,.d) {
3[-l/(e,(to)+l}]
d (-2Q|,d)
}]
-2Qj|d = 0
-1 esJ (w) ebJ (to)
= Im +(Qnd)-Im[ ] (3-55)
[eb(w)+l] et(u) {6b(w)+l}J
The first term describes loss function due to electron
scattering by electric field fluctuations in the vacuum above
the crystal produced by excitations in the bulk. The second
term, which is proportional to d, describes loss function due
to scattering by electric fields produced by fluctuations
within the surface layer. This second term is divided into
two terms as follows:
eb2 -1 Qd
(Q,|d) Im[ ] + Imes(to) (3-56)
(eb+l)J e,(u) (b+l)*
where eb is assumed to be real and not a function of
frequency. The pole of the first term of equation (3-56) is
the form of the longitudinal bulk phonon mode of the surface
layer (cf. es(to)-+0 at wL) and the pole of the second term of
equation (3-56) is the form of the transverse bulk phonon mode
of the surface layer (cf. es(u)- at uT) It is helpful to

100
plot the pole of two layer spectral density S2 (QM,w), which is
et(w)=-l from equation (3-51), in order to justify these modes
as real surface loss. Dispersion curves combined with
dielectric function, es(w) of the surface layer were shown in
Fig. 11 [33], Since the dielectric function is negative
between w=wT and w=wL, the surface wave will not propagate into
the medium, but will be reflected at the boundary. As Q()->0,
the modes become longitudinal bulk mode and transverse bulk
mode of the surface layer. For Q,(d 1, these modes become two
surface modes at each end of surface layer. In the finite
Qnd, the electric fields of the two branch were shown in Fig.
11 [33]. The branch related to wT represents a polarization
parallel to surface, which is forbidden in dipole selection
rule. The branch related to wL is polarized perpendicular to
the surface. This indicates that only the excitation related
to wL will be detected.
In the two layer model, in the limit of Qndl, two modes,
namely one is due to the bulk-layer interface under the
surface layer and the other is due to the longitudinal surface
layer mode, can be detected. In the limit of Q,|dl, only the
surface mode of the surface layer can be detected by the limit
of detection depth. There are a few limits in this two layer
model. Since the surface is not ideally terminated bulk, it
always has sources of imperfection which results in a
different dielectric function from the bulk, which is smoothly
varying. Strictly speaking, it is very difficult to find

101
E> Q g
GJ_
(b)
Figure 11. Two layer mode dispersion and polarization.
(a) Surface eigenmode dispersion for a slab of
material with dielectric function es(w) on a
substrate with a dielectric constant, cb(>0);
(b) The electric fields for ut mode.

102
ideal two layers from the actual surface in most cases. For
example, sub-monolayer coverage of adsorbate can not be a
smooth layer which can not be treated as one layer. If the
evaporated film is thin enough for the detection depth to
exceed the thickness of film and the evaporated film also has
its own surface, this system can be considered as three-layer
system not two layers. The dielectric function theory should
be generalized in order to apply for n-layer systems, which
will derive correct losses due to multilayer thin film
systems. For the sub-monolayer adsorbate system, microscopic
treatment is reasonable since the HREELS spectrometer can
detect down to 0.001 monolayer.
3.4.4. Surface Phonon Dispersion for Semi-infinite Metallic
Surface
Lattice dynamics of crystal surface can be calculated by
the finite slab method for semi-infinite metallic crystals
[39-41]. Information on the polarization and amplitude of
vibrational modes, two dimensional dispersion relation of
these modes and the spectral densities which describe the
number of vibrational states per unit energy can be estimated
by this method. Summarizing the basic ideas of the finite
slab calculation; the thickness of the slab is chosen as
about 15 layers of substrate layers to minimize the
interactions between the surfaces. Type parameters, which are
constrained by the bulk crystal structure, surface unit cell

103
characteristics and the substrate bulk phonon structure, are
used to distinguish the positions and the elements in an unit
cell, which result in different types of central-potential for
different types of pair. Instead of first-nearest neighbor
interaction, many neighbors interactions, if necessary, can
be easily included in the finite slab model. The dynamical
matrix is generated from above information and diagonalized
to give eigenvectors of normal modes of the slab and
eigenvalues for vibrational frequencies.
A few systems have been tested on the surface phonon
dispersion, since the relatively high incident electron energy
(200-300 eV) with the similar resolution as a low energy
incident electron beam has recently made it possible to
produce momentum transfer parallel to the surface up to edge
of the two dimensional Brillouin zone. Ni(001) and Cu(001)
surfaces (both of them are fcc-structure metals) had been
experimentally tested by a few authors [42]. Examining
results of both systems and comparing those with the
theoretical calculations, we can draw a few conclusions of
the intrinsic surface phonon. On Ni(001) surface, along the
T-X direction, two surface modes, the S4 mode and the S6 mode
can be detected, but the S, mode can not be detected due to
violation of an impact selection rule [42]. The schematic of
vibrational modes S,,S4, and S6 are shown in Fig. 12. The S4
mode is a surface acoustical phonon mode, so called Rayleigh
mode, and the S6 mode is a higher frequency mode. If the bulk

Figure 12. Polarization of surface inodes. The displacements
in the surface layer for the three surface modes
at the x point of the two-dimensional Brillouin
zone for an fee crystal with nearest-neighbor
central force interactions and a (100) surface.
Modes S6 and S, involve displacements mainly
parallel to the surface while mode S4 involves
perpendicular displacements.

105
Q Q,,
*t>
^ V ^
x x
**> O *0 Q,
V Q^ *D
t\, t 0**0
o-f 0- 04- O-
04- O- 04-
O- 04 O- 04-
O- 0-4 O-
04- O 04- O-
04- O- 04
0- 04- O- 04
/>
j> f o*
J> f />
* j> J f
S1mode
+ : up
S mode
4
S6 mode
down

106
force constant is used for calculation of S4, the loss energy
of the S4 mode is always lower than the experimentally
measured energy near the zone edge. If a 20% higher force
constant is used, the calculated value is matched well with
experimental value at the zone boundary. This indicates that
by the termination of the bulk, the spacing between the first
and second layer contracts compared to the bulk layer spacing
(3.2% contraction of spacing between the first and the second
layer was estimated [42]). The S6 mode is only detected near
the zone edge. In Cu(001) case, the surface force constant
increased by 16-20% and showed a similar contraction as
Ni(001). Surface phonon dispersion curves are mostly
influenced by the surface structure than the specify metallic
element.
It was pointed out in the dielectric function theory that
if Qj|<0 (in this case the Rayleigh mode near zone center) the
mode penetrate deeply into the bulk. In this regime, the role
of the contracted surface is very small and the dispersion
curve is matched well with the dispersion curve which used the
bulk value of a force constant. As Q|( approaches zone
boundary, the mode begins to localized at the surface, the
discrepancy become distinct near the zone boundary [42]. This
reguires a new theoretical model which could produce the
effect of surface contraction [43].

CHAPTER 4
SURFACE PHONON AND PLASMON MODES ON
Si(111) AND GaAs(lOO) SURFACES
4.1. Overview and Motivation
The energy-loss spectrum of electrons reflected
specularly from a clean Si(111)-7x7 surface is a broadened
elastic peak [33,44-46]. It is widely accepted that this
broadening does not result from phonon losses since Si is a
nonpolar semiconductor with no phonon excitations in the
dipole limit [33]. Instead this broadened elastic peak has
been attributed to losses due to a two-dimensional metallic
state created by dangling bonds from the odd number of atoms
in the surface 7x7 unit mesh or cell [44,45]. It has also
been interpreted as a conduction-band surface-plasmon
excitation [46], which is an oscillation of n-type bulk
carriers localized below the surface in a space-charge layer.
Because of typical low carrier concentration and limited
resolution of typical high resolution electron energy loss
spectrometers, these plasmon energies have been too small to
be separated from the large elastic peak in a specular
scattering geometry (A0=0). In this chapter new experiments
where the elastic peak intensity is suppressed by changing
107

108
from specular to nonspecular incidence angles are presented.
Outside of the specular scattering limit ( A >1.3 ) the
intensity of the elastic scattering peak in high resolution
electron energy loss spectroscopy (HREELS) is drastically
reduced and the loss peak becomes distinguishable from the
elastic peak. As the parallel momentum transfer (i.e., angle
from specular scattering A6 ) is changed further, the loss
peak shifts in a characteristic way which allows one to
identify peaks as plasmon-like or phonon-like. Such HREELS
dispersion effects have not been previously reported for
Si(lll) or GaAs(lOO) surfaces.
With good resolution (AE<5 meV), it is possible to
separate the plasmon peak from the elastic peak in specular
scattering. The effective near-surface space charge carrier
concentration of Si(111) can be estimated using the energy of
the surface-plasmon loss peak. If the effective near-surface
space charge carrier concentration is larger than 10'7 cm3,
the plasmon peak should be separated from the elastic peak
with the resolution of our spectrometer (AE5 meV). In this
chapter the bulk doping density and the annealing temperature
of several samples are varied independently. For 10,s to 1016
cm'3 boron doped silicon, high temperature annealing under
ultrahigh vacuum caused the surface carrier concentration to
increase at the surface [47-49] to ~6xl06 cm'3 and resulted in
a surface plasmon loss peak distinguishable from the elastic
peak (hwD6meV) A similar dependence of elastic and loss

109
peaks upon the strength of elastic low energy electron
diffraction (LEED) intensities for Si(111)7x7 surfaces has
been recently reported by Daum, Ibach and Muller for the
sample with n-type bulk doping and resistivity of 400-800
n-cm [50]. However these authors apparently studied only
specular scattering and high resistivity sample.
Previous studies of polar-semiconductors have tended to
emphasize polar surface phonons, i.e. the Fuchs-Kliewer (FK)
modes, observed in a specular geometry as well as surface
plasmon modes from cleaved GaAs(llO) and InSb(llO) surfaces
[51,52]. Besides cleaved surfaces, As-capped MBE grown
GaAs(100) surfaces also exhibit surface plasmon at specular
geometry [53]. Also for n-type GaAs(100) surfaces after
sputtering and annealing the free-carrier concentration was
found to be compensated by acceptor defects in the near
surface region introduced by the sputtering procedure [54].
In addition temperature-dependent broadening of the
guasielastic scattered electron peak was interpreted as due
to excitation of an unresolved surface plasmon mode with an
effective carrier concentration that differs significantly
from that of the bulk. The surface plasmon mode will be
resolved at off-specular geometry in the same way as Si(111)
surfaces. The effective carrier concentration deduced from
this plasmon loss peak will show the sputtering effect on
semiinsulating and highly doped, p-type semiconductor
surfaces.

110
4.2. Experimental Results
Angle resolved high-resolution energy-loss spectra from
Si (111) surface with the bulk doping density of Nb=lxlO'5 cm'3
are shown in Fig. 13. The sample is sputtered (Ar =
5. OxlO's torr, Vb(!a =1.0KV, emission current=25mA, sample
current=2.5/iA for 40min) and annealed at 800C for 15min. A
shoulder near lOmeV from the quasielastic peak can be seen in
the A6 =6 spectrum, and its energy increased to ~15mV when
A0 =18. The energy of this peak remains at 15mV until the
intensity becomes very small at 25.5 off-specular. For this
surface visual LEED pattern observations showed an apparent
lxl pattern, but the intensity of the elastic peak and the
15meV loss peak in our HREELS data increased when the momentum
transfer matched each of the seventh-order beams (1/7, 2/7,
3/7) which is consistent with a 7x7 periodicity. This
indicates that HREELS intensities are more sensitive than
visual LEED in detecting reconstruction of a long-range
ordered surface. When samples with a higher bulk doping
density (Nb =5.8xl0,s cm"3) were used, the peak in Fig. 13
appeared as a weak shoulder on the quasielastic peak. When
the highest doping (Nb =5.5x10" cm"3) sample was used, this peak
was broadened into the elastic peak, even at off-specular
directions and a detailed dispersion measurement could not be
obtained.

Ill
Si ( 111) nb = 1.0xl015/cm 3
Energy (meV)
Figure 13. ARHREELS spectra obtained from partially ordered
Si(111)lxl for non-specular (L8?0) scattering.

112
After this experiment the sample was sputter cleaned and
annealed at a higher temperature of 950C for 5min. A
stronger 7x7 pattern appeared during LEED observations and
AES intensity ratio from this surface was Si(92)/C(273)~100.
The specular beam had a full width at half maximum (FWHM) of
lOmeV and a loss peak can be seen at 26meV to 30meV in Fia.
14. As the angle was changed from specular to A0=4.4 off
specular the apparent loss energy increased from 26 meV to
30meV then decreased, but this is probably due to the overlap
with the elastic peak that changes intensity in this range.
The absolute intensity of the loss was maximum when the
geometry was just off specular ( A0 2 ) The peak intensity
decreases as in previous measurements assigned to dipole
scattering [55]. The intensity ratio of elastic peak to loss
peak of Fig. 14 was much larger than Fig. 13. However, it is
likely that the increased ordering produced surface states in
the band gap which depleted the surface space-charge region
of the carriers implied by the data in Fig. 13.
Further sputter cleaning and higher annealing temperature
(up to 1250 C) gave results similar to Fig. 14. A sharp,
intense 7x7 pattern was found by LEED and AES showed no
detectable carbon or oxygen peaks [intensity relative to
Si(92)<2x103 ]. However, the annealing at higher temperature
shifted the apparent maximum of the energy-loss peak from
30meV to 24meV. For higher doping (Nb =5.8xl0,s cm'3 ) density
samples, annealed up to 1000 C, the loss peak at maximum

113
-50 0 50 100 150 POO
Energy (meV)
Figure 14. ARHREELS spectra obtained from well ordered
Si(111)-7x7

114
intensity was measured to be AE=33meV. The highest density
(N =5.5x10" cm'3) sample had a maximum loss energy of 26meV at
1000 C annealing. Dispersion curves for these three
different losses are shown in Fia. 15.
Semi-insulating GaAs(lOO) samples were prepared by
degreasing and introduced into the HREELS chamber similar to
the Si(111) samples described above. These samples had a
disordered surface oxide layer since we observed no LEED
pattern and Auger electron spectroscopy (AES) shows
O(510)/C(273) = 4/1, Ga(1070)/As(1128) = 4/3 and Ga/O = 1:1.
Spectra taken from this degreased sample using HREELS are
shown in Fig. 16(a). After subsequent annealing at 580C for
2.5 min without sputtering we obtained HREELS data shown in
Fig. 16(b). and AES shows similar impurity levels to a native
oxide unannealed sample except for a decrease in carbon as
measured by AES to a ratio, C/0 = 1/6. After annealing at
665C for three minutes we obtained another HREELS data shown
in Fig. 16(c) and the LEED pattern from this surface was
C(4x4). From AES data, no oxygen is detected but carbon
impurity was detected (C(273)/Ga(1070) = 1:4 and
Ga(1070)/As(1128) = 4:3).
Undoped, semi-insulating GaAs(100) was sputtered and
annealed at 500 C for five minutes. No LEED pattern was
shown and no impurity was detected from the AES data. For
the clean and annealed surfaces of GaAs(100) the principal
HREELS loss feature at an energy of 36meV is a Fuchs-Kliewer

115
o
I._J
Ta=950C
CO
a

9 J
i
, Phonon (adatom)
>
0)
B
W
S
o--
(C)
CV2
>>
tac
u
0)
C
w
| TA1200 C
Phonon
:14eV
:1OeV
( "o
(screened adatom]
/
/
/
/
/
1 /
\/ *
_ V
A
o -
(A) .
/
/
/ J-

y
o
T =800 C/"
* /
*
qc=0.2ir
fcu,(0) =6.3 meV
0.0 0.1 0.2 0.3
Momentum Transfer (X1)
Figure 15. Dispersion curves of Si(111) surfaces.

Intensity
Energy (meV)
Figure 16. HREELS spectra obtained from a native oxide layer
on GaAs(lOO). (a) As introduced; (b) Annealed at
580C; (c) Annealed at 665C.
116

Intensity
Energy (meV)
(b)
Figure 16.continued.
117

Intensity
-100 0 100 200 300 400
Energy (meV)
(c)
Figure 16.continued.
118

119
(FK) loss peaks of phonon mode. The full width at half
maximum (FWHM) of quasi-elastic peak is lOmeV and the FK peak
is well resolved. Weaker peaks due to surface plasmons are
detected at slightly different energies due to different
densities of free carriers near the surface space charge layer
introduced by the different surface cleaning procedures. Such
plasmon peaks ultimately broadened the quasielastic loss peak
and as reported previously by others [45,46,54]. Angle-
resolved HREELS data obtained from specular geometry to 6.0
off-specular geometry are shown in Fig. 17 from another semi-
insulating GaAs(lOO) surface.
A degreased, Zn-doped, and p-type GaAs(lOO) with bulk
carrier density Nb =1.4x10" cm"3 was cleaned by sputtering (Ne
ions at 500eV and 12nA for 60min) and annealing at 490 C for
11 minutes. The LEED pattern from this sample showed a lxl
periodicity and AES showed no carbon contamination except a
small oxygen peak (-0.05 monolayer). After several sputtering
and annealing cycles a LEED pattern showing 6x1 periodicity
was obtained. For these heavily doped samples the plasmon
peak appears near 25meV and is resolved more clearly than that
of semi-insulating sample at off-angles from the specular
scattering geometry since the intensity of the elastic peak
is greatly reduced. Data shown in Fig. 18 are the HREELS data
obtained at incident angles from specular to 3 off-specular
scattering geometry. In Fig. 19. the dispersion for this loss
is shown. An incident electron energy, 14 eV, and the chosen

120
Semi-insulating GaAs(lOO)
Energy (meV)
Figure 17. ARHREELS spectra obtained from semi-insulating
GaAs(lOO) for specular (A0=O) and non-specular
scattering.

121
. ¡. 1 ,
-100 -50 O 50 100
Energy (meV)
Figure 18. ARHREELS spectra obtained from highly doped
GaAs(lOO) for specular (A0=O) and non-specular
scattering.

122
Combination of
Figure 19. Experimental data and calculated dispersion
curves for highly doped GaAs(lOO).

123
angle change correspond approximately to a parallel momentum
change of 0.02 '1.
4.3. Discussion
Angle resolved energy-loss specular from Si(111) showed
one loss near 10-15meV and another loss near 22-30meV which
are assigned to surface plasmon and surface phonon modes,
respectively, based on dispersion behavior predicted from
theoretical models. The quasielastic peak in the specular
direction, which results from elastic scattering broadened by
unresolved inelastic losses, decreases rapidly in off-specular
directions, while other loss features may increase in relative
intensity. Possible Si(111) losses with energies in the range
of 0 to 30meV include acoustic phonon, optical phonon and
plasmon [56-58]. Each excitation may be dipole-like or
nondipole-like [33]. Loss energy versus transferred momentum
parallel to the sample surface (q,| ) is shown in Fig. 15 with
q,l = 2w (Eo/150. l)H(sin0,-sin0,) (4-1)
where E0 is incident electron energy, 6S is the scattered
electron angle and 9, is the incident electron angle relative
to the sample surface normal. To compare the results to
theoretical calculations, it was assumed that the loss peaks
are described by the poles of loss intensity function
Im[-1/(e (w,q,|) }+i], which gives the condition of e (<>#?,,) =1.

124
Using e(u,q) as a Lindhard dielectric function for highly
doped semiconductors, then
w2 id2
e ( w2 u2 -0.6 q,|2 vf2
where e is the high-frequency limit of the dielectric
function, q,| is momentum transfer parallel to the surface, vF
is average Fermi velocity of free carriers with plasmon
frequency, and wp=(47rneffe2 /m)\ Here net, is the effective
carrier density and m is the hole effective mass, which is
0.23 times of the free electron mass. When an nel( of 6.0xl0'6
cm'3 is assumed, the predicted loss energy versus q^ [dashed
line in data set (A) of Fig. 151 matches reasonably well with
our experimental data. Another prediction of this model is
that the plasmon loss energy saturates at a cutoff wave vector
qc where electron-hole transitions broaden the elastic peak
and saturate the dispersion [59]. The cutoff wave vector
calculated for nef, =6.0xl0'6 cm'3 is qc=0.21'' which is shown as
a dotted line in Fig. 15. Finally, for q||=0, AE=hw(0)=5.3 meV
for the above parameters which is much smaller than the FWHM
of the elastic peak. Therefore some broadening of the elastic
peak would be expected. Assuming Gaussian line shapes for
this loss and for the elastic peak, an upper limit of ~8meV
for the unresolved peak due to the broadening from 9 to 12meV

125
in the straight-through and the reflected geometries was
estimated.
Higher energy-loss data [data set (B) and (C)] are also
plotted in Fig. 15. Although an effective carrier density
can be calculated for q,|, the corresponding calculated cutoff
wave vector is not consistent with this density. Also the
more rapid decrease in intensity suggests a second loss
mechanism must be operative on surface annealed at higher
temperatures. Annealing at higher temperatures reduced the
density of sputtering defects near the surface and improved
the ordering of the stable 7x7 reconstructed surface. The
change of intensities of the loss features at nonspecular
angles, shown in Fig. 14. indicates that the losses were
probably dipole-like [56]. Our 26-30meV loss peak has a
similar angular dependence to the 130meV mode from an ordered
monolayer of hydrogen on W(100) (i.e., dipole scattering)
[55]. At an annealing temperature of 1050C, the plasmon loss
peaks appear to remain but cannot be as well resolved;
therefore a distribution of carrier density probably occurs.
At the highest annealing temperature of 1250C, the loss
energy [data points (c) in Fig. 151 was reduced by ~8meV.
This behavior is again attributed to ordering of the 7x7
surface periodicity. For the clean well-ordered surface one
expects the adatoms to be the last features to order [60],
If the adatoms of the reconstructed surface have a net
repulsive interaction, then it is possible that both their

126
equilibrium positions and vibrational frequencies are
different for the partially ordered and well-ordered surfaces.
Daum, Ibach and Muller also reported [50] that the vibrational
electron energy loss spectrum of the Si(111)-7x7 surface
exhibited a well defined loss at 71meV and a structure at 25-
34meV. These features were detected from a 400-800 n-cm, n-
doped Si(111)-7x7 wafer after annealing at 800C without
sputtering. They also assigned these two modes to localized
vibrations involving Si-adatoms and the surface atoms lying
underneath using Ab initio total-energy calculations of the
dynamical matrix. The higher energy mode is due to out-of
phase vibration and the lower energy mode is due to in-phase
vibration. When these results are compared to our case, the
higher mode is not shown in our case since it is a more
localized mode and also screened by doping carriers. This is
easily simulated in the case of adsorbed oxygen on Ni(lll)
[61] where 75% of the spectral density function is due to the
frequency regime below the maximum phonon frequency (i.e.,
0-37meV) and 25% of that is associated with higher-frequency
72meV mode.
A majority of native oxide and hydrogen species from
HREELS data of the introduced semi-insulating GaAs(100) sample
are detected in Fig. 16(a). The FK optical phonon mode at
36meV is shown since this mode is due to a strong dipole
moment at the GaAs(100) surface covered with oxide and
impurities. A peak near 360meV is due to C-Hx stretch mode

127
and a peak near 180meV is due to deformation mode of CHX.
Near 70meV, a broad feature is due to second harmonic of 36meV
(i.e. 72meV), GaO(~70meV) Ga20(~90meV), AsO(~80meV) and
As304(110meV) From AES, major impurities are due to oxides
and carbon. After annealing at 580C, hydrocarbon species has
desorbed and resulted in better resolution of quasielastic
peak [62]. Oxygen peak does not reduce from AES data. The
strong Ga20 peak at 90meV and the second harmonic (72meV) of
FK mode (36meV) are shown from HREELS data, Fig. 16(b). The
remaining native oxide is mainly Ga20, and other oxide
desorbed at lower temperature (<580C) These results are
matched with other reported results [63,64]. After annealing
at higher temperature (665C) it is shown from HREELS data
that Ga20 has desorbed. A small carbon impurity is detected
from AES data, which does not appear as a peak in HREELS.
This indicate that the remaining carbon does not form a strong
dipole moment with Ga or As. A diffuse LEED pattern was shown
from this surface. This is due to the fact that thermal
annealing temperature is too high to keep the stoichiometry
detected. It is deduced from AES intensity ratio of Ga to As
that approximately 0.1 ML of the surface is As component.
Also carbon impurity can not be totally removed by thermal
annealing at 665C.
For highly-doped GaAs surfaces, a theoretical calculation
of the energy loss of coupled phonon-plasmon modes is shown
in Fig. 10 of chapter 3. An approach similar to that used in

128
the calculation of losses from Si(111) surface, which has used
the Lindhard dielectric function and dielectric function
theory for loss function intensity. Therefore in the
GaAs(lOO) case, effective carrier density is also a parameter
that can be used to fit the data to the theoretical curve.
The only difference is that a phonon term is added in the
dielectric function since GaAs is a polar-semiconductor and
has a FK optical phonon mode. Shown in Fia. 19 is the result
of plotting energy loss versus momentum transfer Also
shown are the calculated dispersion curves for GaAs with
effective carrier density of 6x10'* cm'3 (for solid line
[eo-&o]/[&ofl]=0.168) Dash lines show the dispersion curves
for the uncoupled system. Note that the data points in the
lower branch seem to fit the solid theoretical curve while
data points for the ho. branch seem to fit the dashed curve.
This indicates that our simple theory based on dielectric
response is inadequate. Filled squares correspond to observed
loss features at energies larger than the phonon energy with
the dot-dashed curve being the plasmon branch curve shifted
by 0.77 of the FK mode energy. Thus these latter points
correspond to an interesting combination of plasmon plus
phonon modes rather than sequential losses. The experimental
data are in good agreement with the calculated dispersion
curves with some disagreement at the lower q^values. It is
thought that this discrepancy is due to a surface migration
effect due to repeated sputter and anneal cycles. Angle

129
resolution is directly related to the detection depth (-q^1)
of incident electron with 14eV. At the near specular geometry
which has the deepest detection depth, the average of the
effective carrier density can be lower than that of off-
specular geometry in case of the uneven distribution of
carriers. This indicates that the surface has a highly
enhanced acceptor density due to sputtering. Compared to the
bulk doping density (nb=1.4xl0 cm'3 ), the effective carrier
density (nefl =6.0x10'* cm"3) is due to band bending reduced after
Fermi-level is pinned near the center of the bandgap. But the
depletion of carrier is not perfect, and the surface has an
increased carrier density due to sputtering acceptor density
due to sputtering since it is p-type semiconductor.
For the sputter-cleaned semi-insulating GaAs(lOO), the
plasmon loss peaks at specular geometry is reduced as the
annealing time is increased. Thus this plasmon is due to
surface defect related carriers created during sputter anneal
cleaning cycles. This plasmon energy shifts to lower energy
as defect density reduces with increased annealing time, which
means that defects act as acceptors near the surface for p-
type sample.
Angle resolved spectra of semi-insulating GaAs, shown in
Fig. 18. exhibits the dipole characteristic of FK mode which
is the rapid decrease of intensity for A6 >0. Also rapid
decrease of elastic intensity helps to identify the plasmon
peak whose intensity does not decrease rapidly at off-specular

130
geometry. This plasmon is also due to acceptor carrier
density due to sputtering and annealing cycle. This result,
which shows the increased acceptor density near the surface
due to sputter cleaning procedures, matches well with the
result of Dubois et.al. on n-type surface which shows
compensation of carriers at the surface due to acceptor-like
sputtering defects [54]. Previously our result of p-type
Si(111) surface also shows increased carrier density after
sputter-cleaning.
4.4. Summary
The dispersion of plasmon-loss peaks measured with angle
resolved electron energy-loss spectroscopy gives an evidence
of increased surface carrier concentration from B-doped (10'5 -
10'6 cm'3) Si (111) surfaces which have been sputtered and
annealed surfaces at temperatures between 800 C and 1250 C.
The plasmon loss dispersion curve (AE vs.Ak,|) agreed with the
calculated plasmon energy and cutoff wave vector for an
effective carrier concentration about 10-50 times greater
than the bulk doping density. Annealing at temperatures,
T>950 C reduced the surface plasmon peak presumably by
annealing defects, and allowed detection of a surface phonon
loss peak at 22meV to 30meV probably associated with the
adatoms of the 7x7 reconstructed surfaces. These losses

131
(22meV-30meV) which shift according to the degree of ordering
display dipole-type selection rules.
It has been shown that the HREELS data for GaAs(lOO) at
T=300 K can be quantitatively described by a coupled plasmon-
phonon model using an effective carrier density that is in the
range of 10 to 10 cm'3 for both heavily doped p-type and
undoped material. For p-type (n =1.4x10 cm'3) GaAs(lOO), a
dispersion in the range q||=0.012'' was found which is close to
the cross-over point for uncoupled phonons and plasmons at the
bulk carrier density [59]. The effective carrier
concentration was determined to be 6.0x10 cm'3 to a precision
of 50%. This effective density will depend on the bulk
doping as well as the sputtering and annealing procedures
which produce acceptor-like defects in the near-surface bulk.
These acceptor like defects produced by sputtering and
annealing were also detected in the undoped sample.
From two different kinds of semiconductor surface studies
angle resolved HREELS can yield detailed information about the
surface carrier concentration for semiconductors which can not
be obtained from other surface methods. Since these HREELS
results are shown to be strongly influenced by surface defects
one can obviously use the angular dependence of HREELS to test
defect models of interface formation including Schottky
barriers [65]. Some of these possible experiments might
include in situ growth of semiconductor overlayers including

132
superlattices as well as the formation of Schottky barriers
by metal overlayer depositions [65,66].

CHAPTER 5
VIBRATIONAL MODES FROM OXIDE LAYERS ON Ni(lll) AND Ni(110)
5.1. Overview and Motivation
Oxidation of a nickel surface has been one of the famous
oxidation systems using various surface physics experimental
techniques. It is well known that the oxidation reaction at
room temperature can be divided into the following three
regimes: which are dissociative chemisorption, oxide
nucleation and growth to coalescence, and thickening of this
coalesced oxide layer up to saturation level thickness [8,67-
69]. The first two regimes are well known, but the third
regime is still controversial since there are relatively many
experimental factors to control. Quantitative evaluation
results from many different experimental techniques is very
difficult since the preparation of this third species is not
well defined. One of these controversial points is in
identification of high-binding energy oxygen species from x-
ray photoelectron spectroscopy (XPS) data (e.g. whether it is
an OH species which exists as an adsorbed species of a hydro
oxide layer or it is not OH-related at all but a more complex
Ni-0 species) [70-72]. If this species is identified, it will
133

134
help to understand the kinetics of Ni-oxidation at room
temperature.
In this chapter the following points are considered:
(i) Surface phonon modes from single crystal nickel oxide
films grown thermally (300C) on Ni(110) surface. Are there
any other modes in addition to surface optical phonon modes
(Fuchs-Kliewer (FK) modes)?
(ii) Surface phonon modes from coalesced nickel oxide and
nickel hydroxide layers grown at room temperature for
different exposures of oxygen (300L and l.SxlOL). Are these
the same or different and can one correlate high resolution
electron energy loss spectroscopy (HREELS) data and XPS data?
(iii) Surface phonon modes from the clean Ni(110) sample
exposed to 300L of air at room temperature. Is this oxide
layer the same as that produced by ultrahigh vacuum (UHV)
exposure of pure oxygen?
5.2. Experimental Results
A degreased Ni(lll) single crystal sample was introduced
to the chamber. The AES spectra before cleaning indicated
impurities such as 0, S, Cl, K and C. Three cycles of
cleaning by sputtering on hot (~400C) sample and post
annealing at 750C for 3 min result in a clean surface with
lxl LEED pattern. No specific impurity is detected from AES

135
spectrum. Oxygen exposure ( 3L ;5xl0'* torr, 60 sec) induced
p(2x2) LEED pattern. The HREELS spectrum for this sample is
shown in Fia. 20. Angle resolved HREELS spectra were taken
from specular (Fig. 20) to 4.5off specular direction. Off
specular HREELS spectra are shown in Fig. 21. Oxygen (25L)
was exposed to the fresh cleaned sample. The LEED pattern
changed from lxl to p(2x2), and finally to lxl pattern. The
HREELS spectra from this sample are similar to Fig. 20.
The HREELS spectrum from clean Ni(110) surface is shown
in Fig. 22. Oxygen (18kL) was exposed to the sample held at
300C. The LEED pattern was changed from lxl rectangular
pattern to Ni0(100)-lxl pattern, i.e. a square pattern. The
HREELS spectrum is shown in Fia. 23. Oxygen of 300 L (2.5x
10'7torr, 1200 sec) was exposed to clean sample held at room
temperature. NO LEED pattern was shown. The HREELS spectrum
is shown in Fig. 24(a) Oxygen of 1.8x10 L (ltorr, 1800 s)
was exposed to clean sample held at room temperature. The
HREELS spectrum is shown in Fig. 24(b). Air (300 L; 2.5x10'
7 torr, 1200 C) was exposed to clean sample held at room
temperature. A diffused lxl LEED pattern was shown. The
HREELS spectrum is shown in Fig. 24 fc). All room temperature
oxidation samples did not show any LEED pattern indicating a
disordered oxide layer of thickness greater than 3-4.

Intensity
136
-50
0 50 100 150 200
Energy (meV)
Figure 20. HREELS spectra for specular scattering obtained
from a Ni(111)-p(2x2)-0 surface with 0.25 monolayer
of chemisorbed oxygen.

Intensity
137
x40 x200
Energy (meV)
(a)
Figure 21. ARHREELS spectra obtained from Ni(111)-p(2x2)-0.
(a) 1.5off-specular; (b) 3off-specular;
(c) 4.5off-specular.

Intensity
138
x40 x200
-50 0 50 100 150 200
Energy (meV)
Figure 21.continued

Intensity
139
x40 x200
-50 0 50 100 150 200
Energy (meV)
(c)
Figure 21.continued.

140
_CD
CU
-100 0 100 200 300 400 500
Energy (meV)
Figure 22. HREELS spectra obtained from nominally clean
Ni(110) showing a strong elastic peak and a weak
loss peak due to residual impurities.

141
_C3
ln
G_3
Energy (meV)
Figure 23. HREELS spectra obtained from thermally grown
oxide on Ni(110) under UHV.

142
_CD
CD
G_3
Energy (meV)
Figure 24. HREELS spectra obtained from oxides on Ni(110)
grown under UHV at room temperature.
(a) 300L, 02; (b) 1.8xlOL, 02; (c) 300L, air.

143
Energy (meV)
(b)
Figure 24.continued.

144
Energy (meV)
( c)
Figure 24.continued.

145
5.3. Discussion
For Ni(lll), as the oxygen induced a p(2x2) LEED pattern,
it is known as 1/4 monolayer coverage of the surface. Upton
and Goddard had carried out calculation on the system where
oxygen atom adsorbed on the Ni(lll) surface [40]. As the
oxygen atom approaches the surface along the line
perpendicular to it, the minimum potential energy site is
found at the hollow site of three fold symmetry and at a
distance Rx =1.20 above the plane which contain the nickel
nuclei. If Rx is determined, then using the potential of
Upton and Gaddard and the frequency calculation of the hollow
site at the three fold in the reference [39], Calculation
results are 70.4meV for vertical mode and 62.6meV for parallel
mode. In our case 72meV (Fia. 20) from p(2x2) pattern is
matched with vertical mode. Ibach et al. also has done the
same experiment which gave 71.6meV in their case [73]. The
p(2x2) pattern indicate surface Brillouin zone is not same as
the bulk Brillouin zone since the oxygen species induced twice
larger periodicity on the surface. For the specular direction
we only see the r point in the surface Brillouin zone of (111)
surface which is folded by 1/2 due to the change of the
surface periodicity. The zone edge loss can be the loss in
this case. From Ni bulk normal modes of vibration, the 26meV
mode can be found at K point (right now it is r point because
of zone folding) [74]. This mode is a longitudinal surface
phonon mode. From angle resolved HREELS spectra in Fig. 20.

146
and Fia. 21. both modes are showing dipole-like intensity
variation. Energy changes are not found. The primary
energy,7.0 eV, and the total angle change are not enough to
see the energy variation for the longitudinal phonon to the
end of Brillouin zone edge.
Oxygen exposure (25L) induced lxl LEED pattern (Fig. 20).
Even though surface periodicity went back to lxl, the position
of major oxygen atom doesn't change so much from the hollow
site of three fold symmetry. It also found that oxygen still
induce zone folding.
For Ni(110) surface, from Fig. 22 a very small trace of
residual gas can be detected around 60meV, which is considered
as dissociated water species (namely, O-H; 450meV, Ni-OH;
50meV, Ni-0; 60meV). The HREELS spectrum of the sample after
18kL of oxygen exposure on hot substrate (3 00C) is shown in
Fig. 23. The LEED pattern from single crystal oxide film on
Ni(110) surface showed Ni(100)-lxl pattern, since the relative
size of oxygen is larger by transferring electrons from Ni,
which cause a little rearrangement of original lattice
position while making NiO. Strong intensity at 67.2meV is
due to the FK mode of NiO single crystal films. This peak
from NiO on Ni(100) was reported earlier [75-77]. Also this
peak of easily identified using NiO bulk band structure that
indicates longitudinal optical phonon at 72meV at r point
since it is diatomic single crystal [78]. A big shoulder near
50meV can be clearly identified from the gain side. Either

147
adsorbed hydroxyl group mode (50meV) or transverse optical
mode quite similar to adsorbed oxygen vibration on Ni(100)
surface(49meV) can be the loss at 49.5meV [42,79]. A little
hump near 20meV is interfacial phonon mode between Ni(110) and
NiO(lOO). More precisely speaking, oxygen at the NiO side
induce the surface mode at the Ni(110) surface through
Brillouin zone folding similar to Ni(100) case [74,80].
Comparing with pure surface (Fig. 22] and thick oxide (Fig.
22) surface, such a huge Ni-OH intensity from NiO surface can
not be expected from dissociation of H20 in residual gas.
Benndorf et al. reported H20 adsorption at 300 K is possible
only if there are unoccupied Ni(110) sites in the neighborhood
of adsorbed oxygen [78]. In our case thick NiO covered the
Ni(110) surface and furthermore NiO is thermally grown single
crystal. O-H stretch mode (450meV) couldn't be detected
either. So three fundamental modes from NiO on Ni(110)
surface are NiO F-K mode (67.2meV), NiO transverse optical
mode (49.5meV) and Ni(110) surface mode (20.8meV) induced by
top oxide layer. Long tail up to 140meV is due to second
harmonics of these two phonon modes. High intensity and
energy of 67.2meV indicate strong ionic character of this
film. Dalmai-Imelik et al. [75], Cox and Williams [77], and
Andersson and Davenport [76] who reported values for this mode
as 67.5, 69.5, and 65meV respectively. The main reason why
the loss(69.5meV) of Cox and Williams (single crystal) is
different from the other authors including us (single crystal

148
film on Ni substrate), is because the ionic character of the
NiO single crystal is different from the NiO single crystal
film on Ni substrate. In the case of Cox and Williams,
69.5meV is much less than 71.5meV of bulk loss of longitudinal
optical phonon mode. Ionic character which determines the
strength of bonding changes the loss energy. In the case of
oxidation of Si, asymmetric stretch mode of Si02 varies from
12.5 meV to 150meV due to changing of ionic character of oxide
layer as the thickness becomes larger [81,82,83]. The
thickness of these oxide films are thicker than 10 due to
LEED pattern of the crystal film.
The spectrum after 300L of oxygen exposure is shown in
Fig. 24(a) and from AES intensity variation, the 0(510)
intensity reached saturation level at 50L exposure at room
temperature [84], The amount of exposure 300L is enough to
saturate the oxide formation on the Ni(110) surface at room
temperature. The loss at 63.3meV is the FK mode of NiO layer
on Ni(110) surface. Energy change from 67.2meV to 63.3meV is
due to thinner thickness of NiO compared to thermal oxide as
well as the poor ordering of NiO at room temperature (poor
ordering indicates loss of bonding between the coalesced oxide
islands). This poor ordering reduced the intensity of this
peak. A small hump near 49meV is also shown. Scanning up to
higher energy, 450meV O-H stretch mode does not appear, so OH
is not an adsorbed species. As the oxide ordered poorly and
covered by a layer which has high binding oxygen species

149
detected by XPS [85], a transverse optical phonon mode
(w50meV) can not be expected from this coalesced oxide layer.
So it is suggested that an OH species due to OH of residual
gas covers up the coalesced oxide layer in the form of
precipitates of an Ni(OH)2 layer. Also 50meV can be the
bending mode of Ni-OH [79]. If 0-H species has the component
parallel to the surface like Ni(OH)2, 50meV peak will be shown
in specular geometry. This loss feature is clearly shown on
gain side since higher energy intensity at gain side is
suppressed more by the Boltzmann factor. Distinct hump at
24meV indicate the Ni(110) surface phonon induced by the next
oxide layer, which was also shown in thermal oxide.
Higher exposure (1.8xl09L 02, 1 torr 30min) spectrum in
Fig. 24(b) shows two overlapped peak near 60meV. Due to the
limit of resolution, second harmonic was chosen for clear
separation of two peaks. Second harmonic of NiO peak appeared
at 124meV and that of Ni(OH)2 peak appeared at 106meV. Ni-0
mode is at 62meV and Ni-OH mode in Ni(OH)2 is at 53meV.
Increase of Ni(OH)2 peak from 49meV (300L exposure) to 53meV
at 1.8x109L exposure and Ni-0 mode decrease from 63.3meV to
62.0meV indicates that thickness of Ni(OH)2 layer increase
while the thickness of NiO layer slightly reduced, since the
ionic character of Ni(OH)2 layer also become strong as the
thickness of the layer increase. Asymmetric tail near 20meV
of elastic peak indicates that the low intensity of the
surface phonon of Ni(110) induced by NiO layer.

150
For 300L exposure of air, the peak intensity of 48meV
shown in Fig. 24(c) becomes much larger than that of 60meV.
When the air is exposed, much higher portion of water mixed
with oxygen compared to pure oxygen adsorption (but still
number of oxygen molecules is much larger than that of water)
simultaneously adsorbed on the surface. Effective exposure
of oxygen is about 100L. Before oxygen makes NiO layer up to
saturation level, water start to passivate the oxide layer.
60meV Ni-0 loss with low intensity indicate very thin and not
smooth oxide layer. Ni-OH 48meV in Ni(OH)2 indicate very thin
Ni(0H)2 layer covered the thin NiO oxide in Ni(110).
5.4. Summary
Thick thermally (300C) grown NiO layer on Ni(110) has
shown Fuchs-Kliewer (FK) mode of NiO layer, surface transverse
optical phonon mode of NiO layer and surface phonon mode of
Ni(110) induced by NiO layer. At the room temperature (300
K) oxidation through high exposure (>100L) initially grows
coalesced NiO layer up to saturation level at which OH in
residual gas starts to cover NiO layer with Ni(OH)2 layer.
Further high exposure induce thickening of Ni(0H)2 layer. The
loss energy of both NiO layer and Ni(OH)2 layer are shifted by
the thickness of their layer, which indicates that the thicker
the layer is the stronger the ionic character of the layer is.
Air exposure induces early passivation of NiO by Ni(OH)2,
before NiO reaches at the saturation level.

CHAPTER 6
VIBRATIONAL MODES FROM OXIDE LAYERS ON Si(111) SURFACE
6.1. Overview and Motivation
Molecular oxygen is a very reactive gas phase species
and forms strongly bound chemisorbed layers with most metal
and semiconductor materials. With different experimental
conditions different oxide phases can be formed and it is also
possible to form different oxide layers by using H20 instead
of Oj or by using mixtures of H20 and 02. When oxygen is
chemisorbed the surface atoms transfer outer shell electrons
from other species and oxide bonding shows a strong ionic
character. For different bonding phases the dipole moment of
this oxide bond will be changed and the magnitude of the
dipole moment is related to the ionicity per unit cell. From
the lattice dynamical point of view, the surface optical
phonons penetrate deeply into the crystal as the wave vector
Qi|-*-0 Since these modes have displacement fields that
penetrate deeply into the crystal as Q||->0, the electric field
in the vacuum above the ionic crystal becomes very strong.
Therefore an oxide is a promising candidate whose surface
structure can be detected by high resolution electron energy
151

152
loss spectroscopy (HREELS). For example, ZnO was the first
successful example of an oxide which was studied using HREELS
[86]. It has been reported that 60% of the electrons which
emerges from the crystal are contained in the one-phonon loss
peak at 69meV. Thick Si02 (>500) thermally grown on Si
substrate in the air is a non-crystalline oxide layer. Thiry
et al. reported HREELS data obtained from thick Si02 (950)
grown on Si(100). Since this thickness is larger than the
detection limit of HREELS (-200), they must have detected a
silicon dioxide layer near the vacuum-solid interface not the
solid-solid interface [87]. They shows that dielectric
function theory can be applied for this thick and homogeneous
silicon dioxide layer; but, as the thickness of the oxide
layer is reduced, the dielectric function theory can not
explain the evolution of three peaks related to the oxygen
species. This might be due to the fact that the dielectric
function theory is too simple to explain a complicated
interfacial layer or an inhomogeneous layer. In this chapter
we will investigate the SiOx/Si(lll) interface by preparation
of thin layers of oxygen-related species (X^intermediate
oxide) on Si(111) surfaces. We have investigated four
differently prepared intermediate oxide layers such as Shiraki
wet-chemical intermediate oxide (Shiraki oxide), deionized
water-preserved intermediate oxide (water oxide), air exposed
native intermediate oxide (native oxide) and thermally grown

153
chemisorbed intermediate oxide under ultrahigh vacuum (UHV)
(thermal oxide).
Many chemical etching technigues for cleaning an Si
surface have been reported. The common and basic idea of
chemical etching is removal of a native oxide by the chemical
etchant thus reducing surface impurities and keeping the
surface as clean as possible before introducing the sample
into UHV. The Shiraki technique consists of a wet chemical
treatment to eliminate contaminants on the Si substrate and
thin oxide formation to protect the clean Si surface from
contamination (refer to chapter 2.8.(101). An oxide film
grown by chemical treatment on the Si substrate has a function
as a passivation film. Since the oxide surface is much less
active than the bare Si surface there are few carbon
contaminants on the oxidized Si surface, and they can be
removed more easily than those on the Si surface. Also
forming a smooth surface through the repetition of the
chemical oxidation and etching process help to reduce active
sites on the oxide surface where contaminants are easily
adsorbed. The remaining active sites are filled with Cl atoms
which act as unstable adsorbents blocking active surface sites
before carbon atoms are adsorbed.
An Si(111) wafer was preserved in deionized water for
three weeks in order to isolate the chemically cleaned wafer
from oxygen molecule as well as other contaminants in the air.
Even though the water isolated the wafer from the air, water

154
by itself is an oxygen species. Therefore the wafer will be
oxidized by OH in the water. Since the oxidation by OH always
accompanies hydrogen species, the mechanism and the
intermediate oxide layer produced by the hydride and hydroxyl
group environment may be different from other intermediate
oxide layers.
When a clean Si(111) surface is exposed to air, reactive
species in the air will adsorb on the bare silicon substrate.
Even though oxygen molecules are the dominant reactive species
additional impurities will passivate the surface and oxidation
will stop. Therefore a native oxide is a very thin
intermediate oxide layer formed under a mixture of gases in
the air such as nitrogen, oxygen, water, hydrocarbon as well
as rare gases. Investigating this native oxide will also give
information on air contamination effects. Detailed
information on surface impurities supplied by HREELS with a
0.001ML detection limit might help to improve the vacuum
cleaning techniques. In detecting impurities from silicon
surfaces UHV analytical techniques such as low energy electron
diffraction (LEED) and Auger electron spectroscopy (AES) are
limited since LEED patterns for different amounts of carbon
are quite similar and AES can not detect a hydrogen species.
The combination of HREELS with these techniques is helpful to
a comprehensive understanding of surface impurities.
Thermal oxidation has an advantage over the protective
oxide layers discussed above since the initial stages of

155
oxidation can be controlled under UHV conditions. Observation
of the evolution of three oxygen-related peaks in HREELS may
give information on the structural change of the intermediate
oxide as the thickness of the oxide changes. In our
experiments substrate temperature and exposure of oxygen have
been varied independently. As the oxide becomes thicker,
oxygen atoms on the surface migrate into the surface and
oxygen atoms start to bond with silicon atoms buried under the
surface layer. The structure related to this initial step
will be quiet different for Si02 layers due to the different
oxygen concentration in the different oxide layers. In
oxidation under UHV, an OH species is always created due to
background H20 gas produced mainly by wall reactions. One
possible mechanism is that oxygen molecules react with
hydrogen atoms adsorbed on the chamber walls or with
hydrocarbon species buried in the walls of the ion pump.
Another possible source is that water molecules in the
residual gas are broken into H and OH species on the surface.
Since an OH species is very reactive its possible effect
should be considered in the case of relatively long exposures
of oxygen. This effect is a distinct possibility in case of
exposing oxygen to a silicon substrate which has many active
sites since these active sites are easily occupied by OH
species from the gas phase. The passivation effects under UHV
by OH can be eliminated in thermal oxidation since most of
hydrogen species on the surface desorbs below 350C.

156
Therefore we can expect continuous and controlled oxidation
by the amount of oxygen exposure on samples heated above
~350C.
Finally the main experimental motivation of this chapter
can be summarized as follows. Thermal silicon oxide layers
are noncrystalline oxide layers which can be grown at high
pressure on silicon substrate and have been previously studied
by HREELS. As the oxide layer becomes thinner the
stoichiometry of the oxide layer will be different from Si02
and the simple dielectric function theory may not explain the
evolution of the HREELS loss peaks related to oxygen on such
inhomogeneous surfaces. Through an investigation of the
initial stages of oxidation using controlled oxidation under
UHV we may obtain the information on evolution of the
structure of this intermediate oxide layer. In addition
several different preparation conditions were used to produce
intermediate oxide layers which may have different structures.
Besides thin intermediate thermal oxides grown by pure oxygen
environment under UHV, we have also investigated wet-
chemically prepared Shiraki oxides, deionized water preserved
oxides and air exposed native oxides. All of these
intermediate oxide layers are very thin compared to silicon
dioxide and we can expect to observe the apparent differences
among them using HREELS.

157
6.2. Experimental Results
Using the Shiraki method (refer to section 2.8). Si(lll)
(Boron Doped, p-type and p= 2.4ii-cm) with a thin oxide layer
is introduced into the chamber and pumped to a pressure below
3xlO,0torr. The HREELS spectrum obtained from this sample is
shown in Fig. 25(a) The LEED pattern shows a lxl pattern
with three fold-symmetry only for high incident electron
energy (>180eV). No AES spectrum has not been obtained for
this sample in order not to induce any hydrocarbon species on
the sample surface. Then the sample is subsequently annealed
at 500C for 10 minutes without sputtering. Still the same
lxl LEED pattern is shown. The HREELS spectrum for this
annealed sample is shown in Fig. 25(b). The similar HREELS
spectrum and LEED pattern are shown after annealing at 800C
for 10 minutes. After annealing at 900C for 15 minutes the
clean 7x7 LEED pattern with most of the 49 spots in a unit
mesh are clearly shown. The HREELS spectrum of this sample
is shown in Fig. 25(c) and is essentially the same as that of
a clean sputtered and annealed Si(111)7x7 surface (described
in chapter 4).
The Shiraki-cleaned sample after preserving in deionized
water for three weeks was introduced into the chamber. The
HREELS spectra from this sample is shown in Fig. 26 fa). The
LEED pattern for this sample has no pattern even at high
incident electron energies indicating a thicker oxide layer

-100 0 100 200 300 400 500
Energy (meV)
(a)
Figure 25. HREELS spectra obtained from Shiraki oxide on Si(lll).
(a) As introduced; (b) After annealing at 500C; (c) After annealing at 900C.
158

100
Energy (meV)
(b)
Figure 25.continued.
159

600
Energy (meV)
(C)
Figure 25.continued.
160

161
(>5) than the Shiraki oxide sample above. After annealing
at 500C for 10 minutes the HREELS spectrum obtained from this
sample is shown in Fig. 26(b) No LEED was obtained from this
annealed surface. After annealing at 900C for 10 minutes
HREELS spectrum obtained from this sample is shown in Fig.
26(c). The LEED pattern from this surface shows a 7x7 pattern
where 3/7 and 4/7 fractional order beams have strong
intensities, but the HREELS data indicate residual surface
impurities of C, H and OH probably at a level of -0.1 ML.
After degreasing an Si(111) surface which had been
preserved in the air longer than two years, the HREELS
spectrum obtained from the sample without sputtering is shown
in Fig. 27(a) Upon annealing from 320C to 1010C with 100C
intervals, the intensity of the 350meV peak starts to reduce
at 325C and at 520C it has almost disappeared. The HREELS
data from this sample after annealing at 520C is shown in
Fig. 27(b). No LEED pattern is obtained for this native oxide
sample even after annealing at 920C and the HREELS spectrum
obtained is similar to Fig. 27(b). After annealing at 1010C
for three minutes the HREELS spectrum is shown in Fig. 27(c)-
-bottom curve. The HREELS spectrum obtained after 20 minutes
without anneal is shown in Fig. 27(c)top curve and the LEED
pattern of this sample shows a relatively clean 7x7 structure.
From AES data, no specific impurity except carbon
(C(236)/Si(92)=l/50) is detected; however, the HREELS data
show that the surface is still not clean.

Energy (meV)
(a)
Figure 26. HREELS spectra obtained from water-preserved oxide on Si(lll).
(a) As introduced after preserving in deionized water for three
weeks; (b) After annealing at 500C; (c) After annealing at 900C.
162

Energy (meV)
(b)
Figure 26.continued.
163

1500
4->
(D
C
CD
4-J
C
II
x300 FWHM13.4 meV
Ta900C
-100 0 100 200 300 400 500
Energy (meV)
(c)
Figure 26.continued.
164

Energy (meV)
(a)
Figure 27. HREELS spectra obtained from native oxide on Si(lll).
(a) As introduced; (b) After annealing at 520C; (c) After annealing
at 1010C.
165

>
-M
CD
C
CD
4->
C
hH
Energy (meV)
Figure 27.continued.
166

1000
100
Energy (meV)
(c)
Figure 27.continued.
167

168
After annealing the same Si(111) surface at 1000C
oxygen molecules are exposed to the surface held at 700 K.
The HREELS spectra obtained from 10L-10kL exposed samples are
shown in Fig. 28. For three oxygen-related modes (-55, -95
and ~140meV), vibrational mode energy shifts as a function of
oxygen exposure are observed. In order to determine the
relative stability of this intermediate oxide layer, the 10
kL-exposed sample is annealed at 700 K for 10 minutes, then
annealed at 1000 K for three minutes and finally annealed at
1100 K for three minutes. The HREELS obtained from 700 K and
1100 K annealed samples are shown in Fig. 29 and Fig. 30. The
LEED pattern of this final sample was /19x/l9 indicating the
presence of some impurities such as Ni in the near surface
region. The same experimental procedures on Si(111) held at
900 K result in the similar evolution of three oxygen-related
peaks except the oxidation rate.
6.3. Discussion
Since the thickness of a Shiraki oxide layer is very thin
the LEED pattern shows a lxl periodicity with three fold
symmetry at an incident energy higher than 180 eV. This long
range ordered structure is not due to the Shiraki oxide which
is amorphous but due to a smooth silicon subsurface. Only
electrons with high energies can penetrate the thin oxide and
detect the ordered subsurface structure. From the HREELS
spectrum of the sample shown in Fig. 25(a) three oxygen-

Energy (meV)
(a)
Figure 28. HREELS spectra obtained
substrate.(a) 10L ; (b)
Energy (meV)
(b)
from thermal oxide grown on a 700 K Si (111)
100L ; (c) 1000L ; (d) 10 kL exposure.
169

Energy (meV)
(c)
Figure 28
120
-50 0 50 100 150 200
Energy (meV)
(d)
continued.
170

Intensity
171
FWHM=12.8 meV
Energy (meV)
Figure 29. HREELS spectra obtained from thermally grown
oxide after annealing at 700 K.

Intensity
172
Energy (meV)
Figure 30. HREELS spectra obtained from thermally grown
oxide after annealing at 1100 K.

173
related peaks are shown: 141meV (Si-O-Si asymmetric stretch
mode), 105.5meV (Si-O-Si symmetric stretch mode and Si-OH
stretch mode) and 51meV (Si-O-Si bending mode). It has been
reported by many authors that the asymmetric stretch mode
energy varies with the effective thickness of oxide layers
grown on silicon substrate [82,83,88]. The thickness
corresponding to 141 meV is -1.1 monolayer (ML) or 2.9 of the
oxide layer referring to results of the previous authors who
combined HREELS with x-ray photoelectron spectroscopy in order
to calculate the effective thickness. Since the ionic
character of a thin oxide layer is much weaker than a thick
oxide layer, an energy shift of this asymmetric stretch mode
to a lower value can be understood. The 0-H stretch mode was
shown at 452meV and C-H stretch mode was shown at 359.5meV,
which indicates adsorbed hydroxyl groups and adsorbed
hydrocarbon species on the oxide layer. A broad hump near
285meV is the second harmonic of the 141meV asymmetric stretch
mode.
The HREELS spectrum obtained from the annealed at 500C
is shown in Fig. 25(b) The energy and intensity of an
asymmetric stretch mode 141meV does not change significantly
which indicates that at this annealing temperature oxide
species does not change structure or desorb. The hydrogen
species (357meV: C-Hx, 457meV: 0-H) are largely suppressed
since the annealing temperature of 500C is high enough to
desorb most hydrogen species [89],

174
Additional annealing at 800C for -lOrnin caused almost
all hydrogen species to desorb. The FWHM of the elastic peak
is reduced from 16.3meV to 15.3meV. This is due to hydrogen
desorption since adsorbed hydrogen species induces so called
a spurious ohmic conductance effect [62]. But the oxide layer
does not change; the LEED pattern is still lxl periodicity at
high incident energy and the 141meV Si-O-Si asymmetric stretch
mode does not change.
Final annealing at 900C for 15 minutes induces a clean
7x7 LEED pattern such that most of the 49 beams in a unit mesh
appear clearly at 36eV. The HREELS spectrum is shown in Fig.
25(c) with all the oxygen related peaks decreased to the noise
level. Contamination after annealing is not detected except
for very small traces with loss energy near HOmeV, which
indicates that this surface does not have many defect sites.
The tiny hump near HOmeV maybe due to bulk carbon impurity
which segregated to the surface during the annealing
procedure. The FWHM is still 15.8meV even though the oxide
layer has disappeared. Instead an of oxide layer effect this
broadening is due to dangling bonds of a clean 7x7 surface
[90].
The HREELS spectrum obtained from the sample preserved
in deionized water is shown in Fig. 26(a). The FWHM of the
quasielastic peak is 22.5meV and no LEED pattern is found from
this surface. These results indicate that the surface is very
disordered. The asymmetric stretch mode appears at 145meV.

175
This energy corresponds to -2ML of oxide layer which is twice
the thickness of the Shiraki oxide sample. We can deduce that
the intermediate oxide layer formed in deionized water reaches
to the saturation thickness -2ML and is passivated. In the
water the OH activity is high enough to attack the Si-Si bond
forming Si-OH and Si-H. Only the OH species due to water,
instead of 02 in the air, can contribute to the oxide species
in Si-O-Si structure. A similar result was reported by
Grundner and Schulz for the sample rinsed with water after HF
etching [91] indicating that a few minutes is sufficient time
in H20 to produce this surface. The C-Hx mode (359meV), 0-H
mode (455meV) C-H deformation mode (178meV) and Si-H
(254meV) mode are similar to the Shiraki oxide sample results
shown in Fig.25
After annealing at 500C the resolution decreased to
18meV due to desorption of hydrogen species as expected (Fig.
26(b)). The 145meV asymmetric stretch mode did not shift at
all, which means the thickness of oxide layer does not change
at this annealing temperature. Additional annealing at 900C
induced a 7x7 LEED pattern on the surface. The 3/7 and 4/7
fractional order beam intensities are particularly strong.
The FWHM of the quasielastic peak of the HREELS spectrum [Fig.
26(c)) is l3.5meV which is much smaller than that of the
Shiraki cleaned surface which was 15.8meV. These differences
are due to the fact that the dangling bonds which are
understood as the origin of the metallic characteristic of the

176
Si(111)7x7 surface are quenched by the residual gas impurities
such as OH and H species which are detected in HREELS spectrum
for this surface. Even if dangling bonds of adatoms are
quenched, the 7x7 periodicity does not change from LEED
pattern observations so it can be understood that these OH and
H adsorbates do not break the bonds which are responsible for
the 7x7 periodicity. The Si-Hx stretch (X=l or 2) mode at
246meV, Si-H bending mode around 80meV Si-H2 scissor mode
HOmeV 0-H stretch mode at 450meV and Si-OH mode near lOOmeV
indicate major impurities are present on this surface.
Hydrocarbon stretch modes at 360meV are not detected since
hydrocarbon species did not adsorb on the surface after
annealing.
The HREELS spectrum obtained from a degreased native
oxide sample is shown in Fig. 27 fa). The asymmetric stretch
mode at 145meV indicates the saturation thickness (-2ML or 5-
6 A) of native oxide exposed to air. This energy, 145meV, is
almost same as that of a wafer preserved in deionized water
yet the thickness of such a native oxide layer is expected to
be nearly 20-30 [87] or 4-6 times that of the water-dosed
Shiraki oxide sample. This implies that the thickness
calibration reported previously may have a limited range of
validity. The strong intensity of the hydrocarbon stretch
mode at 360meV (i.e. six times larger than on the Shiraki
oxide) indicates that a large portion of active sites are

177
covered with hydrocarbon species compared to the two previous
samples.
The HREELS spectrum obtained from the sample annealed at
520C is shown in Fia. 27(b). The hydrocarbon 360meV stretch
mode is below the detection limit and the FWHM of the
quasielastic peak is reduced to 12.6meV from 16meV. This
effect is also due to hydrogen desorption. Upon annealing up
to 920C, the three oxide related peaks did not change.
Annealing at 1010C induces oxygen species to desorb from
the surface leaving an Si(111)-7x7 surface with only carbon
impurities. In Fia. 27(c). the HREELS spectrum after
annealing at 1010C is shown with a resolution of the elastic
peak of 9.3meV which is less than that of the previous two
samples. The peak at 113meV is due to adsorbed carbon species
which have transformed to Sic during the annealing procedure
and the peak at 99meV is due to Si-OH. The top HREELS
spectrum shown in Fig. 27(c) was taken subsequently just after
the bottom spectrum ( i.e. 20min time interval between two
spectra). Comparison between top and bottom spectrum shows
rapid increase of intensity at 99meV while 113meV does not
change so much. We can deduce that an OH species of residual
gas (dissociated H20 on Si(111) surface) adsorbed on active
sites left after desorption of oxygen species. We can also
deduce that many of the hydrocarbon species shown in Fig.
27(a) are directly adsorbed on the Si substrate so the strong
intensity of the Sic species is found even after the oxide

178
layer is totally desorbed. The oxygen desorbing temperature
(1010C) of native oxide is higher than that of the water
preserved oxide (900C) .
In thermal oxidation at 700 K passivating species such
as OH and H can not restrain oxidation procedure since
hydrogen desorbs from the surface below 650 K [89]. Therefore
passivation by the water dissociated species is not a problem
in thermal oxidation. The substrate temperature 700 K is not
high enough to desorb the oxygen species. The HREELS spectra
obtained from thermal oxides are shown in Fig. 28. From the
HREELS spectrum obtained from a 10L oxygen exposed sample,
Fig. 28(a) we find the asymmetric stretch mode of Si-O-Si
(124meV), symmetric stretch mode of Si-O-Si (91meV), bending
mode of Si-O-Si (47meV), a small peak at 164meV and a broad
feature around lOOmeV. The small peak at 164meV may be due
a peroxy 0-0 mode (155meV) in a superoxide-like species. This
peak is always detected from the sample exposed to a small
amount of oxygen. If the first and second layer of Si is
replaced by Si-O-Si bridging bonds the electronegativity of
oxygen in Si-O-Si will deplete the electron in dangling bond
at the top as well as the bonding electron between second
layer of Si surface. The asymmetric stretch mode has a small
perpendicular component compared to the other two modes, i.e.
the symmetric stretch mode and the bending mode since Si-Si
direction of first second layer is already tilted from the
vertical to the surface. The HREELS spectrum obtained from

179
this sample after further exposure oxygen (100L) is shown in
Fig. 28(b). The superoxide-like mode at 160meV disappears
while Si-O-Si mode energies evolve. The asymmetric stretch
mode at 130meV correspond to thickness of 0.5ML. At this
exposure most of oxygen atoms are still in bridging sites
between the first layer and the second layer. A small number
of oxygen atoms start to break the bond between the second and
the third layer and the symmetric stretch mode and bending
mode have more than one value. After 1000L oxygen exposure
the HREELS spectrum obtained from the surface is shown in Fia.
28(c). Asymmetric stretch mode at 135meV corresponds to the
thickness of 0.9ML. Since the top two layers are tilted, the
thickness of 0.9ML indicates that many oxygen atoms occupy the
bridging bond between the second layer and the third layer
which is vertical to the surface. This the reason why the
intensity of asymmetric stretch mode increases rapidly due to
dipole selection rule for the vertical mode polarization.
After lOkL exposure the HREELS spectrum obtained from the
surface is shown in Fig. 28(d). The thickness of the oxide
increases to 1.5ML and the direction of the Si-O-Si structures
seems to be random now since the intensity ratio of bending
modes (50meV) to asymmetric stretch modes (138meV) starts to
increase. From the previously reported HREELS spectrum
obtained from the thick Si02-like oxide (950) the intensities
of these two mode are similar [87].

180
Instead of further exposure to oxygen post-annealing at
the exposure temperature (700 K) for 10 min increased the
ratio of the intensity of 140meV to 50meV. Post-thermal
annealing appears to cause the inhomogeneous oxide layer to
become more uniform with sharper HREELS peaks (Fig_i__29) .
Post-annealing at higher temperature (1000 K) increases this
tendency and decreases the loss energy at 140meV by ~2meV
which means a small amount of oxygen starts to desorb.
Annealing at 1100 K induces the oxygen species to desorb
completely from the surface. In Fig. 30 the 113meV mode is
distinct and is due to silicon carbide which is produced by
hydrocarbon contamination dissolving into the sub-surface
after thermal annealing.
Compared to the substrate annealed at 700 K, the
substrate held at 900 K has more rapid oxidation speed.
Annealing at 900 K has the same effect as larger amount of
exposure at lower temperature (e.g. 700 K) since more oxygen
molecules which arrive at the surface can be dissociated into
oxygen atoms compared to the lower temperature substrate.
Except for the oxidation rate the evolution of three oxygen-
related peaks are almost same at the two temperatures.
Asymmetric stretching mode energies of two thermally grown
oxides (700 K and 900 K) versus to oxygen exposure are shown
in Fig. 31. Initially the oxidation rate of the 900 K sample
is about ten times faster than that of the 700 K sample.

145
140
135
130
125
120
1 2 3 4 5 6
og Oxygen Exposure (in Langmuir)
31. Asymmetric stretching mode variations of thermal
oxides grown at 700 K and 900 K.
181

182
6.4. Summary
From the variation of the asymmetric stretching mode
intensity and energy of the oxide layer thermal oxidation
stages can be monitored to a degree. Initially oxygen
molecules adsorbed on dangling bond of silicon substrate.
After dissociated oxygen atoms chemisorbed between the first
layer and the second layer of Si substrate a superoxide-like
species disappeared. As the oxidation continues oxygen atoms
start to chemisorb between the second layer and the third
layer of Si substrate. Thereafter the oxide layer starts to
be similar to amorphous Si02, i.e. the oxide layer starts to
lose its directional orientation with respect to the diamond
structure. We have also found that differently prepared
oxides have different passivation thickness according to their
oxygen supply and passivating impurities. For Shiraki
oxidation, very smooth and thin oxide (1.52ML or 4) covers
the surface and surface is inert to impurities after removing
the oxide layer. While the water preserved oxide is
disordered and saturation thickness of the oxide (2ML or 5.4)
is not even. After removing this uneven oxide layer the
surface is easily contaminated by impurities such as OH and
H. Native oxide formed by air has a similar HREELS spectrum
but this surface is initially contaminated by carbon
impurities and has a stable Sic after removing oxide layer by
annealing. In this experiment we tested the thermal stability

183
of each oxide by increasing the annealing temperature. The
asymmetric stretch mode of the HREELS spectrum obtained from
these intermediate oxides indicates that the oxide thickness
and oxygen desorbing temperature seems dependent upon this
thickness.

CHAPTER 7
SURFACE RELAXATION AND SURFACE EXCITATION OF Si/Ge ALLOY
FILMS
7.1. Overview and Motivation
Since the development new growth techniques of thin film
growth, thin film novel materials have opened new eras in
device fabrication as well as physics. Techniques such as
Molecular Beam Epitaxy (MBE) enables one to grow films layer
by layer. Optical properties due to differences of band gaps
of heterostructure of two different semiconductors grown by
MBE can be applied to devices such as photodetectors and field
effect transistors[92,93]. But the combination of
semiconductors was usually limited to the direct bandgap
materials, mainly III-V compound, lattice matched
semiconductors. Since the lattice matching reduces the defect
sites at the interface due to dangling bonds of unmatched
atoms, the carriers can be confined within one layer with high
mobility without being trapped or scattered by these defect
sites. But silicon, an indirect bandgap semiconductor in
Group IV, which the present industry based upon, has 2%
lattice mismatch relative to an Si05Ge0S alloy and 4% lattice
mismatch relative to Ge. In the case of thin SixGe,.x alloys
184

185
film deposited on the Si(111) surface, the lattice mismatch
is accommodated by a large strain of the Si substrate. The
growth of SiGe,_x alloy on Si substrate has been reported by
many authors using different techniques [94-105]. Initially
the SixGe,_x alloy film grows with the same lattice constant as
the Si substrate (i.e. pseudomorphic growth) and recovers its
own lattice constant after it becomes thick enough to allow
misfit dislocations to be introduced. With the right
combination of GexSi,.x composition and in the thickness limit
of pseudomorphic growth, Si layers and SixGe,.x alloy layers can
be used as a superlattice [101]. It has been reported that
the indirect bandgap of GexSi,.x strained layers on Si (001) have
confirmed the anticipated lowering of the indirect bandgap of
these alloys into 1.3-1.55 /im range (i.e. long wavelength
photodetector)[93]. Research related to the surfaces of
SixGe,.x alloy film from growth mechanism to film analysis has
already begun and the major results can be summarized as
follows:
Evaporated Ge on Si(111) surface can be intermixed to
make alloy films through post-thermal annealing and on a hot
Si(111) substrate, evaporated Ge can make alloy films through
intermixing at the interface. Both of results were reported
by many authors using various surface techniques[95-97,99-
105]. The growth mode of this alloy film is of Stranski-
Krastanov (SK) type which indicates that initially, to 2-3
monolayers, epitaxial and pseudomorphic films can be formed,

186
then more evaporated Ge starts to form a three dimensional
island. This typical growth mode was confirmed by techniques
such as reflection high energy electron diffraction (RHEED) ,
Auger electron diffraction (AES), transmission electron
microscopy (TEM), scanning electron microscopy (SEM), He-ion
channeling, and x-ray diffraction measurements. Chen, Belmont
and Sebenne reported sub-monolayer adsorption of Ge on the
cleaned Si(111) surface at room temperature using ultraviolet
photoelectron spectroscopy (UPS) [106]. At coverage up to
1/3 of a monolayer the Ge atom binds to three Si surface atoms
replacing Si dangling bonds by Si-Ge bonds plus one Ge
dangling bond. With further Ge evaporation the Ge dangling
bond states continue to develop and Ge-Ge interaction sets
in. Kasper and Herzog reported 8% Ge alloy films of
thickness, 0.1 Jim, were grown on Si (100) at 750C without
misfit dislocations (i.e. pseudomorphic growth) as determined
by x-ray diffraction and TEM [94]. Bean et al. using similar
techniques reported that the map of pseudomorphic growth of
0.1 jim alloy film indicates the substrate temperature for
pseudomorphic growth is related to Ge fraction in the alloy
[101]. Using RHEED Sakamoto et al. reported at a 450C
substrate temperature the critical thickness of pseudomorphic
growth rapidly decreases with increasing Ge [104]. Gossmann,
Feldman, and Gibson reported by AES that Ge starts to grow as
an island from 2-3 monolayers thickness at elevated
temperatures, (i.e. 300C, 520C) [102]. Toropovetal. showed

187
SEM photographs with the Ge islands in diameter -0.2 nm with
an effective film thickness of 25 A grown at a substrate
temperature 580C [96], The low energy electron diffraction
(LEED) pattern of the Si0 5Ge0, alloy film grown on Si (111)
substrate is 5x5 [99,102,103,107]. After deposition of 1.9ML,
Si(111)-7x7 LEED pattern become diffused (i.e. a 7x7 pattern
mixed with a 5x5 pattern), then above 1.9ML Ge coverage the
LEED pattern changed to 5x5, and then further annealing at a
high temperature (>770C) without additional Ge induces a 7x7
LEED pattern. One point of interest is this 5x5 LEED pattern
which has a similar structural analogy to the 7x7 LEED pattern
of Si(111) [107]. First, intense fractional order spots along
the lines joining neighboring integer-order spots forming a
6-pointed star formation centered on each integer order spot.
This star formations has been attributed to the shape
transformations of triangular subunits of the unit mesh. Off-
star fractional order spots on the perimeters of a hexagon of
side 2/5, centered on some integer order spot, and its six
fold rotational symmetric spots has relatively high
intensities. The intense (3/7, 4/7) spots are due to dimers
lying along the perimeter of the triangular submits. Recently
Becker, Swartzentruber and Vickers monitored the
SiGe,_x(lll) 5x5 and showed scanning tunneling microscopy (STM)
images which have six adatoms in a unit mesh, which is
analogous to the 12 adatoms on an Si(lll)7x7 surface [108].

188
The results of LEED and STM confirms that dimer-adatom-
stacking fault (DAS) model can be also applied to the
SixGe,_(111) alloy film case [109]. The driving force in
triangular-dimer model of the Si(111)-7x7 reconstruction is
the lateral compression of the two outermost double-layers.
The 7x7 reconstruction of Si(111) surface, consisting of the
removal of 1/7 of the second- and third- layer atoms and
dissociation of the resulting dislocations, is the means to
relieve this compressive stress. In SixGe,_x(Ill) -5x5 case,
since Ge has 4% larger covalent radius than Si, this alloy
has the enhanced lateral compressive stress resulting in
removal of 1/5 of the second- and third layer atoms. Schaefer
reported hydrogen interaction with the GexSi,_x (100) alloy using
HREELS to detect the Ge surface concentration according to
annealing temperature [110]. Compared to the 20% Ge bulk
concentration, the Ge surface concentration was estimated at
up to 75% at high annealing temperature (temperature is not
reported.). This is quite different from GexSi,_ alloy films
since high temperature annealing causes the varying of the Ge
profile in the film. Note also that the (100) face has a
higher step density than the (111) face, which can be easily
ignored in an adsorption experiment. The step contains a
larger number of dangling bonds which act as active sites on
the surface. Farrel et al. reported water adsorption on the
MBE grown SixGe,.x(100) alloy films [111]. In this case, the
sample was sputtered and annealed to get rid of air

189
contamination since the sample was prepared outside the
chamber. Also they reported Ge enrichment at the surface due
to high temperature annealing and water adsorption is quite
similar to Si(100) (dissociative chemisorption of H and OH
species).
At this point we want to suggest the problems to
consider. First, since Si0 5Ge0 5(111) surface has also adatoms
like Si(111)-7x7 surface, small amounts of hydrogen titration
on the alloy film surface may enable identification of adatom
species on this surface through only reaction with adatoms.
Vibrational energy loss in HREELS is inversely proportional
to me,,*. The adatom and hydrogen atom will cause a strong
dipole moment vertical to the surface which can be easily
detected at specular geometry. Deuterium instead of hydrogen
can confirm the adatom species since deuterium has similar
chemical properties to hydrogen atom. Second, since the
x-ray diffraction data of Bean et al. is so small (namely 0%
lateral relaxation at 100 A and 1.2% lateral relaxation at
500 A) it is necessary to check where this relaxation has
started and whether it is smooth or abrupt through slow
deposition of a Ge film [101]. Third, since the
GexSi,_x(111) 5x5 pattern starts to show at the same point as
the beginning of the Ge-rich island growth (2-3ML) [94-105],
and Ge island also grows epitaxially, the origin of this 5x5
pattern is uncertain, i.e. whether 5x5 pattern comes from Ge-
islands or from other area between Ge-islands or both.

190
In this chapter these questions are considered and
experiments were conducted as follows: First, the surface
excitation from SixGe,_x alloy films are monitored by HREELS
and a hydrogen (and deuterium) titration technique. Second,
the fine relaxation of this SixGe,.x alloy is monitored by
digital-LEED intensity measurements to detect the critical
thickness. Finally, the growth mechanism of this SixGe,_x alloy
film will be examined using thermal evaporation, LEED and AES
techniques. The SixGe,_x alloy film has been characterized by
RBS and Scanning Auger Microprobe after removal from the
analysis chamber. The growth and characterization of GexSi,.x
alloy film will be shown in Appendix C.
7.2. Sample preparation
Si(111) surface has been Ar-ion sputtered and annealed
at 900C to prepare a clean 7x7 reconstructed surface. Ge
source is heated by e-beam and Ge is thermally evaporated on
the Si (111) substrate. Either on a hot substrate held at
580C or on a substrate at room temperature and post-annealed
up to 580C, Ge intermixes at the Si surface and produces 5x5
LEED pattern. Ge thickness calibration is shown in section
2.9.. For hydrogen titration experiments hydrogen molecules
are dosed on the Ge0 5Si0 ¡(111) 5x5 surface in front of the hot
tungsten filaments (~2000C) Using ammeter through measuring
the current passing through the sample, an effective exposure

191
of hydrogen atoms is estimated (refer to section 2.8.) The
LEED pattern is monitored from the hydrogen titrated sample
after 2.5L effective H-atom exposure and shows still 5x5
periodicity.
For surface relaxation experiments Ge source is heated
inside alumina crucible surrounded by tungsten coil and
evaporates slowly with a rate of 1ML/5 min.. At this time
Si(111) substrate is held at 560C to induce intermixing of
Ge molecules on Si(111) surface. Calculation of an effective
thickness of Ge0 5Si0 5(111) film is shown in Appendix C. At
each step of 50 evaporation, lateral lattice constant is
monitored through digital LEED measurement.
7.3. Surface Excitations of SinGen. Alloy Films
bv HREELS.
7.3.1. Adatom Vibration
The SixGe,.x alloy film which was prepared on Si (111)
surface by e-beam evaporation and post-annealing at 600C, and
the HREELS spectrum obtained from the surface with 5x5 LEED
pattern is shown in Fig. 32. The similar LEED pattern was
shown from another Si05Ge05 alloy film which was grown on hot
(~580C) substrate. The angle resolved HREELS spectra, from
specular to off-specular up to 6, are shown in Fig. 33. At
specular geometry, the full width at half maximum (FWHM) is
15meV which means a broadened elastic peak. The cause of the

360
Energy (meV)
Figure 32. HREELS spectra obtained f rom Ge0 ,Si0 5( 111)-5x5 showing a strong
elastic peak (E=0) and a weak loss peak (E105meV) due to
residual carbon impurities.
192

Energy (meV) Energy (meV)
2.5
3
4
Figure 33. ARHREELS spectra obtained from Ge0 ,Si0, (111)-5x5 for specular
(Aff=0) and non-specular scattering geometries.
193

194
broadening may be a disordered surface due to evaporation.
But the rapid decrease of the intensity of the elastic peak
suggests the surface is quite smooth. Since Si0 5Ge0 5(111)-5x5
surface has the similar surface reconstruction as Si(111)-7x7,
it is considered that the adatom on the surface induces
metallic characteristics at this alloy surface. Therefore the
broadening is due to small transitions in this metallic band.
When the angle difference exceeds the limit of the aperture
(-1.3), the broad shoulder around 23 meV starts to develop.
The distinct peak is shown at A0=2.5, and the peak becomes
indistinguishable below A0=4 This is the typical peak
evolution of dipole active loss near the elastic peak which
was shown in chapter 4. The adatom vibration (phonon) induces
this peak. The broad feature may be due to a disordered
surface. But, as the alloy consist of Si and Ge, there are
three kinds of bonding at the surface, namely Ge-Ge, Ge-Si and
Si-Si. These bondings have slightly different frequencies,
since they are covalently bonded. From the Raman scattering
measurement, longitudinal optical (LO) modes of these three
bondings appeared at -36 meV, -52 meV and -62 meV respectively
in the bulk alloy [112]. At least 10 different bonding sites
are involved in this adatom vibration considering only the
nearest nine local atoms and this broadening is expected from
various combinations of these three kinds of bondings.
Using diatomic simple oscillator model the vibrational
frequency of diatomic molecule is

195
(0 = 2ir ( k / /i )* (7-1)
where k is the force constant between two atoms and m is the
reduced mass. The reduced mass n is 14 for Si-Si and n is 20
for Si-Ge. Assuming the force constant between Si and Si is
same as that between Si and Ge, the vibrational frequency
ratio of usl_sl to ui^c, is 1.2 (i.e., (20/14)'). From Raman
scattering data the frequency ratio is also 1.2 (i.e.,
62meV/52meV). The assumption that force constants are same
is correct in the bulk. For the surface, adatom vibrational
frequency is 30meV for Si(111)7x7 [refer to section 4.2. and
reference 31] and 23meV for Ge 5Si0 5(111) 5x5. The ratio of
adatom vibrational frequency is 1.3 (i.e., 30/23). The result
indicates that the combination of Ge and Si is dominant at the
surface assuming the force constants are similar. Also the
higher frequency ratio indicates a surface force constant
between atoms changes slightly from the bulk value or there
is more screening on 5x5 alloy surface than Si(111)7x7
surface.
7.3.2.Hvdroaen Titration
Since the mass of hydrogen-related vibrational mode has
a relatively large energy loss. Fundamental frequencies of

196
the Si-H diatomic molecule and the Ge-H diatomic molecule are
244meV and 227meV respectively [33]. In case of hydrogen
adsorbed on an Si surface, the stretching mode appears at 258-
260meV depending upon the adsorbing configuration and the
direction of the substrate [113,114], The stretching mode of
adsorbed deuterium on Si is also reported at 189-190 meV which
is 7 meV larger the 2" factor, the mass factor between
deuterium and hydrogen. For the Ge(100) substrate, Papagno
et al. reported Ge-H stretch mode at 178meV which is also 6meV
larger than the 2* mass factor [115]. In addition to the
stretch mode, the bending mode and the dihydride scissor mode
of these two systems were listed in Table 1. The Si(111)
sample (p-type, Boron doped, p =10-20 D-cm) was annealed at
1010C for three minutes which induced clear a 7x7 surface.
About 2 00 A of Ge was evaporated on the room temperature
sample. After annealing at 600C for four minutes, the 5x5
LEED pattern appeared. The HREELS spectrum obtained from this
surface is shown in Fig. 32. The peak at 107 meV is due to
carbon contamination of the surface which makes a stable
carbide at the surface after annealing. First, adsorbing 2.5L
hydrogen still shows a slightly degraded 5x5 LEED pattern.
The HREELS spectrum of this surface is shown in Fig. 34 fal.
Two peaks at 3 62 meV and -170 meV are due to hydrocarbon
species adsorbed during the adsorption procedure.
Unfortunately, the broad peak near 104 meV made it impossible
to detect any SiH2 loss feature. Since 5x5 LEED pattern

197
still remains after exposure, the triangular dimer is not
broken. Intensity of 104 meV does not increase, compared to
Fia. 32. This suggests that only adatoms have reacted with
hydrogen atoms. The losses at 246meV (Ge-H stretch mode) and
71meV (Ge-H bending mode) indicate that Ge-Hx mode is dominant
(see Table 1) A 5L exposure of hydrogen induced a lxl
pattern. The HREELS spectrum (Fig. 34(b)) obtained from this
surface shows little change of two peak positions (i.e. 246
meV and 71 meV) The loss due to Ge-Hx is also dominant at
this surface. For a 2.5L exposure of deuterium on
Ge0 5Si0 5(111)-(5x5) surface, the LEED pattern is not changed.
The HREELS spectrum is shown in Fig. 35(a). Compared to
hydrogen adsorption the contamination level due to hydrocarbon
species is guite reduced. The loss at 106meV is due to the
carbide. Two peaks at 50.3meV and 177meV are due to bending
and stretching modes of Ge-Dx (see Table 1). Further exposure
of deuterium up to 5L induced lxl LEED pattern. The same
177meV peak and three small peaks between 52meV-75meV are
shown in Fig. 35(b) taken from this surface. Still Ge-Dx is
the dominant peak and small peaks between 50-80 meV indicate
mixture of Ge-Dx and Si-Dx. From these results, we can
conclude that Ge0 5Si0 5(111)-(5x5) surface has a dominant Ge
atom dangling bond in comparison to an Si atom dangling bond.
There might be a guestion on this conclusion. Since the
annealed surface has many Ge islands, the reacted dangling
bond might be at the Ge-island not at Ge0 5Si0 5(111) (5x5)

198
Table 1, Energy Loss for Hydrogen and Deuterium Titration
on Si
and Ge
surfaces
[110,113,
115] .
system
ref.
bending
mode
(meV)
scissor
mode
(meV)
stretching
mode comment
(meV)
H/Si(111)7x7
[113]
80(78)
110
260(258)
(after
Si-H2
desorbed)
D/Si(111)-7x7
[113]
54,64
(53,64)
81
190(189)
(after
Si-D2
desorbed)
H/Si(111)-2x1
[113]
80(79)
109
258(259)
(after
Si-H2
desorbed)
H/Ge(100)
[115]
66
105
245
ln2
temperature
D/Ge(100)
[115]
50.7
71.8
178
ln2
temperature
H/Ge 2Si ,
[110]
78 (Si-H)
258(Si-HJ
(100)2x1
70 (Ge-Hx)
247 (Ge-Hx)

>
-4-J
tn
c
QJ
C
Energy (meV)
(a)
Figure 34. Hydrogen titration on Ge0 ,Si0 5(111) -5x5.
(a) 2.5 L, H; (b) 5L, H.
199

>
4J
iH
cn
c
QJ
4->
C
II
Energy (meV)
(b)
Figure 34.continued.
200

450
>
C/l
C
CD
C
II
xlOO
GeD bending
FWHM=20 meV
2.5L Dat om
50
11
1 106
Ge D stretching
/
/

\ 177 ^
~
361
-100 0 100 200 300 400 500
Energy (meV)
(a)
Figure 35. Deuterium titration on Ge0 5Si0,(111)-5x5.
(a) 2.5 L, D; (b) 5L, D.
201

350
Energy (meV)
(b)
Figure 35.continued.
202

203
surface which has almost the same amount of evaporated Ge.
The total area of Ge-island is quite small compared to that
of alloy layer. Even if Ge island also has a dangling bond,
the total number of dangling bond in Ge0 5Si 5(111) alloy
portion is much larger than that of the island and the
dominant LEED pattern is 5x5 not 7x7 due to Ge island (see
next section 7.4.). Since the reactivity of Ge with a
hydrogen atom is similar to that of Si, the possibility of
impartial reaction with one species can be excluded.
Concluding this section, Ge0 5Si0 5 (111) (5x5) surface has
the similar adatom vibrational mode to Si(111)-7x7 surface.
Quasi-elastic broadening from the alloy surface is also
detected, which means that the alloy surface is metallic.
Even if it is hard to distinguish the origin of Ge-H, mode,
considering the total area of the island is much smaller than
that of the alloy layer, Ge0 5Si 5(lll)-(5x5) surface has much
portion of Ge dangling bond.
7.4. Surface Relaxation Measurements
Using Digital LEED
The experimental procedure of film growth and
experimental set up of digital LEED measurement is discussed
in Appendix C and Appendix A. The sample used here is a 1000
A alloy film grown on a hot Si(lll) substrate (560C) The
intensity profile (primary energy 35eV) of the line connecting

204
the (10) beam and the (01) beam from the LEED screen were
shown in Fia. 36. For three different film thickness (e.g.
d=0, 400 A, 1000 A) intensity profiles show different peaks.
For d=0, Si(111)-7x7 peaks are clearly shown and for d=400 A
5x5 peaks are shown. But for d=1000 A, 7x7 and 5x5 admixtures
are shown, since two center peaks are broad (i.e. 2/5+3/7 and
3/5+4/7). This peak intensity ratio resembles a 5x5 pattern
but the peak positions favor a 7x7 pattern. The lattice
relaxation as a function of alloy film thickness is shown in
Fia. 37. The relaxation calculation from the channel distance
measurement is { (d/dR)-1 }xl00%, where d is the channel
distance between clean Si(111)-7x7 integer order beams and dR
is the channel distance between Ge0 5Si0 5(111)-5x5 film's
integer order beams. Since the relaxation is measured for
five different energies (i.e. 35eV-55eV), the standard
deviation of relaxation ratio is also shown. The lattice
constant of film matches that of Si substrate for thickness
up to 150 A. At greater thicknesses, however, a gradual
increase in the lattice constant is observed. After a film
of 1000 A was grown, the lattice constant was about 1.5%
greater than that of Si. This is consistent with the 2%
difference expected between the bulk alloy and the bulk Si
dimensions and a 4% difference is expected between Si and Ge
bulk. Interestingly, it was observed that for a thickness of
600 A there was evidence of two overlapping LEED patterns of
5x5 pattern and 7x7 pattern as we have seen in Fig. 36. But

205
1000A
400A
Si(lll)7x7^J \J\sJ\jJ\J\)
Figure 36. LEED intensity profile of Ge0 5Si0 5(111)-5x5
measured by digital LEED.

206
c
o
4->
fD
X
ro
Q)
CT
0
200
Ge/Si
400 600 BOO
Film Thickness
1000
()
Figure 37. Surface lattice relaxation.

207
the lattice constant of the 7x7 pattern is larger than that
of Si(111)-7x7. This 7x7 pattern may result from domains of
pure Ge on the surface of the alloy. Such a 7x7
reconstruction of pure Ge implies that this part of the film
is not fully relaxed to the 4% larger lattice constant of bulk
Ge compared to bulk Si since the relaxed Ge surfaces has a 2x8
reconstruction. From the post-film analysis, SEM photographs
(Fig. 47 in Appendix CM showed Ge-rich islands of the size of
a few jum. This is not surprising since thin films of strained
pure Ge with 7x7 pattern have been formed on Si (111) by Ge
deposition at high growth rate [116]. This result may
indicate that the deposition rate exceeds that of dissolution
rate into bulk. In other words, if the deposition rate is
kept constant and the diffusion length becomes longer, then
the dissolution rate becomes smaller and finally no
dissolution will occur. Also, even if the estimated thickness
of alloy was 1000 A, the actual thickness is less than 1000
A due to island growth of the film. From AES analysis, there
are thick Ge rich islands whose typical size is a few /m on
the final surface, and the alloy film thickness is quite
shallow (<50 A) compared to the thickness of island. So the
measurement of relaxation at 1000 A is mostly due to Ge-island
films. This is the reason why the relaxation graph has a
large error bars, since the two 5x5 unrelaxed and 7x7 relaxed
patterns are overlapped.

208
7.5. Summary
Ge0 5Si0 5 alloy film growth on the hot (580C) substrate
follows the Stranski-Krastanov growth mode. The 5x5 LEED
pattern originated from Stranski layer (the layer between Ge-
rich islands) shows a very similar reconstruction to that of
an Si(111)7x7 surface. In HREELS experiments,
Ge0 5Si0 5(111) 5x5 surface (-200) also shows an adatom
vibrational mode and a quasi-elastic peak broadening. This
indicates a Ge/Si alloy surface is also metallic mainly due
to dangling bonds of adatom. Hydrogen titration to quench
this dangling bonds on a Ge0 5Si0 5(111)5x5 surface reveals that
the alloy surface is Ge-rich. Even if it is hard to identify
the origin of Ge-H mode, considering the total area of the
island is much smaller than that of the alloy layer at this
thickness (-200), Ge0 5Si0 5 (111) 5x5 surface has much portion of
Ge adatom whose dangling bond reacts with hydrogen atoms
during the titration procedures.
Surface relaxation of Ge0 5Si0 5(111) 5x5 alloy film grown
by the thermal evaporation using Knudsen-cell type evaporator
shows continuous relaxation of alloy film beyond -150 of film
thickness. Beyond 600 of film thickness, 5x5 and 7x7
patterns start to overlap and 7x7 LEED pattern becomes
dominant as the film becomes thick. Since there is an
apparent difference between the lattice constant of Si(111)7x7
surface and that of alloy 7x7 surface, the relaxed 7x7 pattern
from alloy layer is thought to be due to Ge-rich islands.

CHAPTER 8
CONCLUSIONS
8.1. Conclusions about Present Experimental Results
It has been observed that semiconductor surfaces have
various states corresponding to various cleaning procedures.
First, the normal UHV cleaning procedure such as sputtering
and post-annealing, induces surface defects which act as an
acceptor-type carrier on the p-type semiconductor surfaces
such as Si(111), GaAs(lOO) and semi-insulating GaAs(lOO).
This effective carrier density was determined by the angle-
resolved HREELS technique which measured the dispersion
relation of the surface free-carrier plasmon. The intrinsic
bulk dopant density is not the major contribution to this
plasmon since band bending by the Fermi-level pinning depletes
the bulk carriers. It can be concluded that sputter-annealing
induces a p-type surface layer in the space charge region on
the semiconductor surface. By annealing at higher
temperature, the defects diffuse into the bulk, the near
surface carrier density is decreased and the plasmon is no
longer detected. Instead, another vibrational mode which
209

210
follows the dipole selection rule is detected. This is the
adatom phonon mode from the reconstructed Si (111)-7x7 surface.
The 'protective-oxide' annealing technique under UHV
produces a clean surface initially; but, if the sub-surface
is not smooth, the surface is quickly passivated by residual
gas impurities. These impurities consist mainly of water-
related species such as OH and H dissociated by the remaining
active sites after removal of the oxide layer. So it is
emphasized that the protective oxide should be thin (<1 ML)
to desorb at low temperature (<900C) and the surface under
the oxide layer should be smooth enough not to have many
active sites in order that a clean surface can be obtained.
It has been shown that the initial thermal oxidation rate
in UHV depends upon the substrate temperature and the final
saturation thickness of oxide layer is nearly independent of
substrate temperature since the saturation thickness occurs
due to slow bulk diffusion through the oxide which is nearly
temperature independent in this temperature range. The oxide
has ionic bonding between an oxygen atom and the substrate
species (such as Si or Ni) and the vibrational frequency
varies according to the degree of the bonding strength. In
the oxidation of Ni(110) at room temperature, the variation
of a vibrational frequency due to oxide thickness was also
monitored. For room temperature oxidation of Ni(110) with
high exposure (>100 L) of oxygen, the coalesced NiO layer
grows to saturation level at which OH in the residual gas

211
starts to cover NiO layer with Ni(OH)2 layer. Further
exposure of oxygen induces thickening of this Ni(OH)2 layer.
The surface morphology, the growth mechanism and
vibrational modes of the Ge0 5Si0 5(111)-5x5 film were observed
by AES, LEED, SEM and HREELS. These alloy films (thickness>10
A) were grown on a hot (560-590C) substrate and they followed
the Stranski-Krastanov growth mode. It has been shown that
the origin of the 5x5 LEED pattern of the alloy film is not
Ge-rich islands, and these Ge-islands which are distributed
evenly at the surface relieve the strain of Ge0 5Si0 5 alloy
layer. This is the reason why the clean 5x5 LEED pattern and
Ge-rich islands appear at almost the same time. For further
Ge-evaporation, the thickness of the alloy layer does not
increase. For slow deposition of Ge on hot (-580 C) Si(111)
substrate the intermixed GexSi,.x alloy film (x~0.5) grows
pseudo-morphically up to 150 A. For more evaporation, a
gradual increase in the lattice constant was observed. At the
'effective' alloy film thickness of -1000 A, two overlapped
LEED patterns were observed. One is the 5x5 LEED pattern due
to a shallow alloy layer and the other is the unrelaxed
Ge(111)7x7 pattern due to thick Ge-rich islands.
The adatom vibration mode was also observed from the
Ge0 5Si0 5 (111)-5x5 surface at an energy similar to the Si (111)-
7x7 surface. Quasi-elastic peak broadening from the alloy
surface is observed, which means that the alloy surface also
has metallic surface states due to adatom dangling bonds

212
similar to the Si(111)7x7 surface. This adatom species was
shown to be mostly germanium through titration of hydrogen and
HREELS.
8.2. Recommendations for Future HREELS Studies
Future research using HREELS can be divided into two
parts. First, improvement of the spectrometer will solve the
current resolution limit to a certain degree. Still the
electron spectrometer is lower in resolution compared to Raman
or IR photon spectroscopy. Second design of a novel
spectrometer which can change the scattering geometry to cover
larger momentum transfer will open new opportunities for
HREELS in angle resolved spectroscopy. This will give a
chance to complete the surface phonon dispersion of single
crystal systems to cover the whole first and second Brillouin
zones. In short, improvement of spectrometers can open new
research areas. The subjects of HREELS studies will increase
from simple geometry and simple adsorbates to many materials
such as non-crystalline surfaces and magnetic surfaces with
spin ordering. Since HREELS can see to a depth of -200 on
non-metal surfaces, one can study interface modes for rather
standard overlayer systems such as doped, quantum well
structures.
Finally the future research areas derived from this
thesis may be as follows: (1) Semiconductor surface relaxation

213
can be studied by HREELS equipped with high incident energy
to measure dispersion throughout the entire first Brillouin
zone. (2) Interfaces between semiconductors, metals and
insulators can be probed by HREELS in combination with
photoemission. Especially the initial stage of the interface
formation which is still not well understood. (3) Oxidation
combined with hydrogen titration can detect oxidation steps
of metal, semiconductor and alloy surfaces. Hydrogen and
deuterium titration will solve the problem of insensitivity
of HREELS to non-polar surface to a certain degree. (4) New
materials containing oxygen such as surfaces of high Tc
superconductors are good candidates for future HREELS studies.

APPENDIX A
FURTHER EXPERIMENTAL DETAILS
A.1. Gas Handling System for UF HREELS System
Gas lines are connected to UHV chamber through the leak
valve to control the rate of leaking. Two separate branches
are used for sputtering gases such as Ne and Ar, and a
reactive line for oxygen, hydrogen, deuterium and ammonia.
The schematic diagram is shown in Fig. 38. For rare gas lines
like Ar and Ne, a LN2 cold trap can be used just before
opening the valve to condense the impurity gases which may
have mixed with rare gases due to a leak or desorption from
the line. But the gas in the line should be replaced by the
fresh gas just before using if the gas has been in the line
longer than 1 week. Two gas lines are used separately to
avoid mixing sputtering gas with reactive gas.
A.2. Design of Evaporation System
The evaporation system consists of a commercial electron
gun with five crucibles (Thermionics laboratory, model 100-
0050) collimators with shutters and a titanium sublimation
pump with a liquid nitrogen shroud. The design is shown in
Fig. 39. When the filament is heated by a power supply,
electrons emitted from the filament which is positioned under
214

Figure 38. Schematic diagram of gas handling system.
215

17 1/4
216
Figure 39. Design of evaporation system.(a) Collimators;
(b) Cross sectional view;(c) Side view.

217
cathode shield are repelled by cathode shield. As soon as
electrons come out of shield, the anode attracts electrons
toward the crucible. Since this motion occurs in a
perpendicular magnetic field, the trajectory of electron will
be circular and its destination will be inside of crucible.
The electron gun power supply has a high output voltage fixed
at 4 kV and the emission current can be varied from 0 to
750mA. The voltage and current of gun filament are variable
between 0-6 volts and 0-25 A each. A well focused electron
beam can heat the source up to 3500C. Deposition rates are
different for each material (e.g. Nickel 2250 A/ min at full
power). Crucible liners made of graphite or tantalum plate
have been used. But a tantalum plate can not be used for
nickel source due to alloy formation. Vitreous Graphite is
usually non-reactive with source but it can be broken for fast
temperature variation. Crucible liner is essential to keep
the crucible clean and a safe method for preventing damage to
the crucible with electron beam in case the source is totally
evaporated.
Once the evaporation source is heated by electron beam,
the vapor diffuses due to the pressure difference. The vapor
spreads radially and can cover a large area of the chamber
which can include the aperture of the electron analyzer and
viewports if collimators are not installed. Two collimators
installed between the sample and evaporation source (distance
17 1/4") reduced the evaporated film diameter to 3/4" at the

218
position of sample. The rest of the beam not captured by the
sample (l/4"x3/4") is blocked by the shutter positioned at the
back of the sample. Another shutter between two collimators
can control the beam before it arrives at the sample. The top
copper collimator is attached to the copper gasket to prevent
gas passing through along side collimator. The bottom
collimator which is attached to a liquid nitrogen cylinder
captures the evaporated material. This design is limited by
high pressures inside of evaporator section which is almost
two order of magnitude higher than the sample side. Two
viewports (2 3/4") aimed at the filament and crucible allow
one to observe the state of a source during evaporation. Two
flanges at the top of source chamber are used to refill the
source without detaching the whole evaporation system from the
chamber. Center section of -9" OD cylinder secures the space
for electrons to move without hitting the chamber walls.
As mentioned earlier, the small diameter of collimator
reduces conductance of gas and results in higher pressure (~2
orders) at the source side than the sample side during
evaporation. This can not be allowed if a thin film
(submonolayer-100 A) is desired, since slow deposition rates
for better control require relatively longer periods of
evaporation. Then impurities from the high background
pressure can quickly saturate the sample surface. Therefore
an independent pumping system with a similar capacity as the
sample side is necessary. A large surface of liquid nitrogen

219
jacket and titanium sublimation pump are appropriate pumps for
temporary use during evaporation. Along with pumps, an ion
gauge for independent measurement of pressure and mass
spectrometer for analysis of residual gas inside evaporation
section have been installed.
A.3. Quantitative AES Analysis Using Standards
This procedure mainly follow in the PHI AES handbook
[117]. Usually clean silver target is used as a standard.
First, the relative sensitivity, Sx, between the element x and
silver standard.
SX(EP) = [ (A+B) /A] IXM/ (Kx IAgH) (A-l)
where A, B are ideal composition of chemical formula, Kx is
the scale factor. IXH and IAgH are peak to peak amplitude of
each peak in standard for element x and silver. The atomic
concentration of element x is
C* Ix/ (I*g S, Dx) (A2)
where Ix, IAg are the peak to peak heights from the spectrum
of element x and silver target and Dx is relative scale factor
between the spectra for test specimen and silver. Dx is
multiplication of ratios of lock-in amplifier sensitivity,
modulation energy and primary beam current setting. It can

220
be easily checked the concentration formula is just if all
the experimental setting are same as standard (i.e.Dx=l and
K=l) .
A.4. Digital LEED
Useful information can be obtained from the intensity of
diffracted beams as a function of incident electron beam
energy[1,118]. The purpose of digital LEED is to obtain the
quantitative intensity information with better resolution.
To achieve this goal, the LEED beam should not be blocked by
the sample manipulator. The front-view LEED system was
converted to the back-view LEED system.
A schematic diagram of the computer control system which
was used in the experiment for acquisition of two dimensional
intensity data is shown in Fig. 40. The retarding grids and
the beam voltage input of a constant-current LEED electron-
gun controller are connected to the outputs of two separate
computer-controlled digital-to-analog converters (D/A 1 and
D/A 2). The retarding field oscillates between a high value
(a few volts above the beam voltage) and low value (a few
volts below the beam voltage) with the same time interval.
The total intensities for high values of the retarding field
are subtracted from the total intensities for low values and
the difference, which is proportional to the intensity in an
energy window between the high and low values, is accumulated
and stored in the computer memory.

221
LEED Opt i es
Samp 1e
e-gun
Vidicon
Camera
D/A
D/A
Int erf ace
i
_jj t'
Computer CPU
4t
Disk Stor
age Unit
OMA
Termina 1
Figure 40. Schematic diagram of digital LEED.

222
To measure the lateral lattice constant, it is necessary
to scan the line which contains two sequential integer order
beams since the distance between two integer order beams in
the reciprocal surface is inversely proportional to the real
lateral lattice constant. Usually a electron beam energy of
35eV gives a relatively large distance between two integer
beams in our experimental set-up. For example among an
available 500-points in a line, the distance between (01) beam
and (10) beam is 292 points for 35eV while 235 points for 55eV
in the Si(111)-7x7 case.
A.5. Tuning of HREELS Spectrometer
To achieve the best resolution from experimental set-up,
before adjusting knobs which change the potential inside of
spectrometer, the sample should be cleaned properly and the
vacuum pressure should be low enough so as not to contaminate
the sample until the end of experiment. A geometrical
understanding of sample position relative to monochromator
exit or analyzer entrance and ideal specular direction of
scattering should be first achieved. There are variable
factors to determine before data acquisition such as sweep
energy range, sample geometry, primary energy of incident
electron, and sample bias. Also there are many potential
energies to optimize, tuning must begin from a simple geometry
and ideal setting of capacitor potentials. Most of the
potential knobs of the ELS-22 power supply, except analyzer

223
part, change independently so that the potential on any given
knob can be changed in order to reach the optimum maximum
intensity in most cases. In here the full tuning procedure
from the very beginning will be presented. Tuning problems
such as asymmetry and proper zooming will be discussed.
Finally tuning examples will be compared using setting
records.
Initially to detect an electron beam inside the
monochromator, a current from the outer capacitor (R^J of main
monochromator should be maximized using a picoammeter.
Initially the filament current for the cathode is set to 2A,
the repeller to -1 V, anode 2 to 10 V, anode 3 to -2 V. The
contact potential of monochromator is varied from the most
negative value to higher values to detect the current pass
through the premonochromator until changing the difference of
premonochromator (Apm) reduces the maximum current of R*,, .
Reconnecting R, to the power supply, the current at the target
position is detected. At the straight through geometry, the
target as a beam probe is put into the center of the
scattering section to receive the current. The electrode 1
of acceleration optics is set 3 V, electrode 2 to 3 V, and
primary energy 5 V. The difference of main monochromator is
varied in order to detect the current from the target.
Achieving the current of target of -10,0A, all the knob in
cathode system and monochromator system can be adjusted.
Removing beam probe (i.e. target) outside of scattering

224
section, the slit potential of the analyzer (=A) should be set
to the same value as the slit potential of monochromator (=M) .
Connecting the outer plate of main analyzer to a picoammeter
to detect the current in the same way as a monochromator.
Setting electrode 4 of deceleration optics to the slit
potential (A) electrode 3 is varied to detect the current
from the picoammeter. The current can be maximized by
adjusting e,-e4. Reconnecting main analyzer to power supply
and setting difference of main analyzer, AM, to difference of
main monochromator, A^,, the difference of second analyzer, Asa,
is varied until the current is detected from the channeltron
(the high voltage source is set to 3kV) Once a straight-
through beam is detected, readjusting most of knobs to
optimize the signal is followed until the symmetric and sharp
(at least less than 20meV) intensity signal is achieved.
Since the optimum setting at straight through geometry is not
related to the sample, this setting can be recorded and used
for different samples (I, = l.5xlO',0A, rate = 104/sec and
resolution 5-7 meV) as long as the work function of the
capacitors are not changed.
Current reflected from the sample can be optimized.
Rotating the sample surface parallel to the direction of
straight-through electron beam, the sample is transferred
slowly to the scattering section. As the sample approaches
to the scattering point, the intensity of the straight-through
beam should decrease due to the bias on the sample. The lost

225
intensity can be recovered by adjusting the external bias
knob. When the sample is at the scattered position, the
surface of sample can be viewed through the port. At this
point the count rate is almost zero since the beam is almost
blocked by the sample. The sample is rotated 30 counter
clockwise if the ideal specular geometry is 60. The
ratemeter should be at the lowest count rate position and
filament current should be increased. After rotating sample
(e.g. 30 counter clockwise), the monochromator part should be
rotated through twice the angle that the sample is rotated
(e.g. 60 which correspond to 2 cm down from the top notch)
since the analyzer part is fixed. At near the specular
geometry the ratemeter will be guickly saturated. By
adjusting the position of sample, the maximum countrate should
be achieved. If the rate meter saturates at 104/ s, the
filament current is reduced. The ELS-22 power supply
potential can be adjusted to find the maximum countrate. The
countrate is guite different according to the state of sample
surface since resolution depends upon angular divergence from
the target. By checking the full width at half maximum (FWHM)
of the slit potential of analyzer using ramp knob, the
resolution of setting is measured. Sensitivity of the
potential is guite different for samples and geometries.
When the target causes the angular divergence the energy
broadening will give an asymmetric profile of the elastic
peak. In this case, by adjusting electrode '4' of

226
deceleration optics, the asymmetric tail can be suppressed.
First moving the position of recorder pen to the asymmetric
tail near elastic peak, electrode '4' of deceleration optics
is adjusted to reduce the intensity of the asymmetric tail.
If this procedure reduces the intensity of elastic peak more
than 25%, electrode '4' is changed in the opposite way to
reduce the asymmetric tail. After adjusting electrode '2' of
acceleration optics, electrode '1' and electrode 31 can be
sequentially used to increase intensity of the elastic peak.
Another important feature in tuning is zoom. If zoom is
at the elastic peak, most of the intensity is consumed by the
elastic peak so that it is very hard to detect losses beyond
200meV where important information of hydrogen related species
always exists. So as not to lose high loss energy
information, the position of zoom is around 450meV. It shows
in Fig. 41 that a different zoom position affects intensity
of the loss peak. The zooming procedure is as follows. As
mentioned earlier in tuning, most potential knobs are
independent. But electrode '3' and electrode '4' of the
deceleration optics have to be changed during the sweep (i.e.
slit potential changing) to fulfill their imaging condition.
To do that, the highest loss energy (e.g. 450meV, 0-H stretch
mode) should be the zoom position. Initially electrode '3'
and electrode '4' settings are recorded at the elastic
position (=lst setting). Sweeping to the chosen loss peak
(e.g. 450meV), the loss at the chosen peak is maximized using

227
-100 0 100 200 300 400
Energy (meV)
Figure 41. Zoom trials of GaAs at different energies.
(a) Zoom at the elastic peak; (b) Zoom at 270 meV:
(c) Zoom at 360 meV.

228
electrode '3' and electrode '4' whose settings are also
recorded (s2nd setting). Returning to the elastic peak,
electrode '3' and electrode '4' values are reset using the
first setting values while zoom knobs are set to zero.
Sweeping again to the chosen loss peak (e.g. 450meV), the
second setting values of electrode *3' and electrode '4' are
recovered using the corresponding zoom knobs. Then the energy
window can be swept from the elastic peak. Through zooming
operation the maximum intensity of the elastic peak reduces,
but the total intensity are distributed to the whole range of
the spectrum.
There is no absolute criteria for determining of good
tuning. But typical examples of 'poor' tuning and 'good'
tuning will be shown in Table 2. First the difference between
them is in the 2nd capacitor of analyzer. Since the
difference between rsa and Rsa is 0.225eV which is much smaller
than other capacitor differences such as 0.527eV, 0.683eV,
0.402eV, it is impossible to increase to those values with the
same polarity at both side of plates. The graph of plate
voltage (r,*,, R,., r_, R,*, r, R*,, rsa, Rsa) versus capacitor
voltages (A,,., A,., Au,Asa) is shown in Fig. 42. All capacitors
except second analyzer have the right sign of slope compared
to the slope of standard in the manual. Second the zoom is
at the elastic peak in 'poor' tuning case. It is hard to
detect high energy losses. Especially in detecting adsorbed
impurities such as hydrocarbon species (~360meV) or hydroxyl

229
Table 2, HREELS Tuning Examples.
Sample
Semi-insulating
GaAs(100)-new
Semi-insulating
GaAs(100)-old
Primary Energy (eV)
5.000
5.054
External Voltage (V)
4,890
4.981
Electron Gun
Filament (10"'A)
2.00
2.00
(V)
Repeller
-0.116
-1.298
Anode 1
-4.246
-5.940
Asymmetry 1
-0.205
0.313
Anode 2
22.50
19.00
Asymmetry 2
3.431
-0.735
Anode 3
-4.361
-4.103
Asymmetry 3
0.102
-0.043
Acceleration
Electrode 1
-0.570
-1.275
Optics
Asymmetry 1
0.165
-0.010
(V)
Electrode 2
1.174
-0.604
Asymmetry 2
0.040
0.008
Deceleration
Electrode 3
0.480
0.057
Optics
Electrode 4
-0.193
-1.361
(V)
Asymmetry 4
0.156
-0.011
Zoom Knob
Electrode 3
7.20
7.50
Position
Electrode 4
7.14
7.50
Monochro-
Contact potential -0.004
-0.005
mator
Pass energy Slit
-0.270
0.042
(V)
Precapacitor rpm
0.120
0.271
Rp*
-0.118
-0.257
rPB-RpB
0.237
0.527
Maincapacitor r^
0.115
0.267
Rim
-0.182
-0.418
r-R
0.297
0.683
Ramp position
7.28
4.13
Analyzer
Contact potential -0.046
-0.329
(V)
Pass energy Slit
-0.305
-0.488
Maincapacitor rm
0.118
0.158
R*
-0.181
-0.246
r-R*
0.298
0.402
2ndcapacitor rsa
0.100
-0.213
RSa
-0.180
-0.438
rsa-Rsa
0.279
0.225
Count rate (/sec) 1.2xl03 7.4xl03
Resolution (meV) 15.0 14.5

230
Figure 42. Plate voltage versus capacitor voltage to correct
tuning.

231
group (~450meV) are not easily monitored. Besides the typical
example of 'poor' tuning, capacitor charging, wrong geometry,
and wrong bias on the sample can cause a 'poor' tuning.
A.6. Computer Program and Data Handling
All programs are written in PASCAL including the software
needed to drive the IEEE board which is originally written in
BASIC (i.e. converted to PASCAL). The program LHMAIN for data
acquisition and displaying is designed to select parameters.
The data collection program has the capability of being
actuated from the command line of DOS with one parameter being
the file containing pre-established run-parameters and the
other parameter being the filename where the data is stored.
This allows the user to set parameters and achieve a series
of runs without operator intervention. This setting of
command lines has the function of a human operator, namely
setting monochromator angle, sample angle, pass energies and
monitoring counts.
The detailed procedure to use the actual program is as
follows. In Fig. 43. three screens for filenames, selection
of commands and setting parameters are shown. Initially the
filename under which each run is stored is coded to indicate
user initials, date and run index. This runindex is
automatically incremented within the program each time a data
file is saved with the first value being entered by the user
initially at runtime. From the second screen run parameters

232
Enter your initials please:ims:
Enter todays dataset :51:
Enter beginning run number:701:
(a)
Choose run parameters
Elastic peak tuner
Trial run (no save)
Run
View data runs
Save data
Quit
(b)
High Resolution EELS
Baseline is at -7038.2000 meV
Kepco voltage scale IV or 10V
1
Start Energy (meV)
0.00
Energy range (meV)
249.82
Final Energy (meV)
249.82
Maximum number of points
1023
Time per point
2.00
Number of averaged data scans
3
Count rate
3.000E+004
Enter number of angles to step through :
10
Enter starting angle for monochromator :
0.00
Enter final angle for monochromator
(referenced from current zero
position) :
9
Sweep will be from -7.04 to
242.78 by
0.24
Total time per run (min) :
35.8
Total time per set (min) ;
107.4
Total time for all angles (hrs]
Everything OK Y/N ?
:
17.90
(c)
Figure 43. Screens of "LHMAIN" program.
(a) Filenames;
(b) Selection of commands;
(c) Setting of parameters.

233
are selected. Either defining a starting and ending analyzer
potential or defining a starting potential and sweep energy
range will set the sweep energy range with respect to a
present setting of pass energy at the HREELS power supply.
Defining the number of points with maximum of 1024 points will
determine the energy step with the automatic calculation
referring to a previous energy range. With an option of angle
resolved HREELS, initial angle, final angle and step of angle
are determined in the same way as above. The total number of
scans to be averaged together and the time per point are
selected. With automatic calculation, estimations of each
run-time and total run-time are displayed. During data
acguisition, an x-y plot of data is shown on the monitor
screen with a numerical output of both analyzer potential and
counts. In case the count exceeds the expected count which
was chosen in the run parameter setting, y-axis is rescaled
automatically. Individual data scans can be viewed and the
average of all scans is recorded in the disk file at the end
of the scan. With the option of 'Test Run', the data can be
manually saved with the 'save' command.
Another important purposes of computer controlled data
acguisition is the handling of data. Data stored in disk file
can be retrieved by the program '2D plot' written in PASCAL,
which has various options to plot on the monitor screen or on
a piece of paper using a plotter. Since the analyzer
potential starts from zero in data acquisition, it is

234
necessary to shift elastic peak position to zero loss position
to find real loss energy. Magnification is an essential
procedure in recognizing loss features with reasonable
intensity. Precise measurement of loss energy, intensity and
FWHM is obtained by the cursor on the curve. Especially in
angle resolved HREELS data analysis, plotting all the curves
for different incident angles on one sheet makes comparison
easy. In case of plotting noisy data, smoothing by three
point average technique helps to display better-looking curve.
But smoothing technique causes to remove detailed information
from a spectrum, so both spectra with and without smoothing
which are overlapped in one sheet sometimes avoids both
extremes. Besides options listed above, many mathematical
options including derivative and integration are also included
in the plotting program '2D plot'.

APPENDIX B
FURTHER THEORETICAL DETAILS
B.1. Dipole Scattering Cross Section
Since the electron trajectory re depends upon time, from
equation (3-16), the potential V also depends on time. The
probability (=P,) of the dipole with normal mode of frequency
u0 being found in the first excited state at t=oo if initially
(at t=-oo) it was at the ground state, is from the second
order time dependent perturbation theory,
(B-l)
Calculating using equation (3-17),
= r (d/aze) |i/re(t) | (b-2)
where r = <0| q (d/dq)/x 11> which is dynamic dipole moment.
Using the relation q = (2u0)'"( b* + b ),
r = (2u0)-*dii/dq (B-3)
P, =
exp(iu0t) dt
. ¡o
235

236
Since re=(x,z), l/re can be expressed using Fourier
transform,
1
r.
2n2
d2 Q
dq
exp(-iQx) exp(-iqz)
Q2 + q2
1
2n
d2 Q e
¡Q.X-Q lU-1
(B-4)
From equation (B-2), then
a 1
3ze re
1
27T,
1
2n
d2 Q e
iQ.x-Q 2
d2Q exp[-iQ. (x0+vl|t) -QvJ 11 ]
(B 5)
Integrating over time and using equation (B-l) and equation
(B-2),
dt exp(iw0t) <01 V(t) | l>dt =
so
-1
7r
d2 Q
exp (-iQ x0) vAr
[Oq2+Q2v2] ,
(B 6)
where
n= Wo-Q-v|| .
(B-7)

237
Substituting equation (B6) into equation (B1), then
integrating over x0 will give the total cross section a for
one-phonon excitation,
a
4 Q2 v2
[ nQ2 + Q2vA2 ]2
(B-8)
The fact that a parallel component of the wave vector Q is
the conserved quantity enables us to infer equation (B-8) can
be also applied to each d2Q Therefore
4 Q2 vA2
da = T2 d2Q (B-9)
[Q2 +Q2vx2]
B.2. Transformation of Coordinates
For the case of normal incidence (i.e. a=0) and 0 (the
rotational angle relative to the specular beam) is not zero,
, 60) (B-10)
cos0 sin0 0
-sin0 cos0 0
0 0 1
where M0=
is the rotation matrix about oz.

238
For the general case where 07*0 and q = McFtyitkoteo
= k ( -0cosacos0+0osina, 6sin where Ife=
cosa 0 sina
0 10
-sina 0 cosa
is rotation matrix about OY.
Q the parallel component of q to the surface, is
Q = k(-0cosacos0 +0osina, 0sin0),
(B-12)
Q2 = k2 cos2 a [ (0cos0-0otana)2 +62 sin2 0sec2 a]
= k2 cos2 a f (6,0, a) ,
(B-13)
nQ2 + Q2v2 = (W0-Q-v||)2 +Q2va2
= k4cos2 a (602 +62 ) ,
where 0=0 is used.
d2 Q
Jacobian
Q, Qy
6 0
= k2 6cosa dfld0 .
(B-14)
(B-15)
Substituting equations (B-13,14,15) into equation (B-9), the
differential cross section is

239
4r2 cosa f (0,0,a) 6d6d da
(B-16)
k2 (O'+O02 )
where f(0,0,a) is defined in equation (B-13) .
B.3. Fuchs-Kliewer Modes: Lattice Dynamical Framework
If u(l) denote the normal coordinate that describes the
relative motion of ions in the unit cell located at 1, the
equation of motion is
ii0(l) + w02u0(l) = (e'/Mr) E (1) ,
(B-17)
where e' is the transverse effective charge associated with
the unit cell, Mf is the reduced mass of the ions in unit cell
and w0 is the oscillation frequency due to a mechanical
restoring force between the two sublattices. The long-ranged
interaction between the ions is included in the electric field
E(l) generated by the ion motion. Due to dipole-dipole
interaction between each unit cell, E(l) is
e
31^(1,l')i^(l,l') 6qg
}u^(l') ,(B-18)

240
where e is the background dielectric constant. Under the
assumption of slow variation of u0(l) and E(l), the equation
of motion for u^x), after some algebra combining equation
(B-17) and equation (B-18), is
4wne2
^(x) + [w02 ]u,a(x)
3 Mr 6oo
ne'2
M, £oo
d[ V Uoix')
dx'
3xcJ | x -x' |
(B-19)
This has the solution of the form u0(x,t)=u0(Q)exp(iQx-iwt)
in the bulk. If Uo(Q)iQ (transverse mode), then the right
hand side of equation (B-19) vanishes. Then bulk transverse
optical mode is
wto2= w02-47rne'2 / (3Mr&o) (B-20)
If u0(Q)I Q (i.e. longitudinal mode), then the right hand side
of equation (B-19) becomes (ne2/(Mr&o) ) (-47rufla(x) }. Then bulk
longitudinal optical mode is
= w02 + 87rne'2/(3Mrebo) (B-21)
Considering semi-infinite ionic crystal in the half space
z>0, and taking the divergence of both sides of equation (B-
19), then

241
d 2
[V u0 (x) ] + wL02 [V-u,(x) ] = 0
(B-22)
at2
inside of the crystal. All possible solutions of equation
(B-22) except those with frequency of wl0 must have the
relation of Vuo=0. Then the solution of equation (B-19) can
be expressed as an excitation that propaqate parallel to the
x direction with wave vector Q||.
u0(x,!, z) = u0(z) exp(iQnX-iwt) ; for z>0
; for z<0
(B-23)
0
where u0(z) must satisfy equation (B-22). For this case
V'u0(x') = array of sources located on the crystal surface. If the
charge q is placed on the surface of a dielectric, the
potential inside of medium is not due to q but due to
effective charge 2&cq/ (feo+-l) at the position of original
position of charge by the screening effect [15]. Then
equation (B-19) becomes
(B-24)

242
Inserting equation (B-22) into equation (B-24), then the
solution is
47rine2 u2(0) [x +iz]
u0(x) exptiQuX-QuZ-iut],
M,(1+6) (aT02 a2)
(B-25)
when z = 0, the self-consistent relation
47rne'2 uoz(0)
uOz(0)=
Mr (I+600) (a2 -aT02 ) (B-26)
gives the Fuchs-Kliewer mode
47rne'2
a2 = as2 = aT02 +
M, (I+600) (B-27)
This can be expressed in terms of e(0) (refer equation (3-
42)), and 6 as follow
e(0) + 1
ws = wT0 ( )* ,
6 +1 (B-28)
which is the same result as equation (3-44).
The two approaches are identical in physical content.
'Fuchs-Kliewer mode' due to lattice dynamical theory and

243
'Surface polariton' due to electromagnetic theory were named
according to the way of description of these surface mode.
Both of them ignored the retardation effect since Q,|ws/c.
The frequency-dependent dielectric function can contain more
terms to describe other excitations besides optical surface
phonon, e.q. surface plasmon due to free carriers.

APPENDIX C
GROWTH AND CHARACTERIZATION OF Ge/Si ALLOY FILMS
C.1. Alloy Film (lOOOA) Grown on Hot Substrate (560C)
Sputtering and annealing at 800C Si(111) surface induce
(7x7) reconstructed surface. The evaporation source (Ge) was
held at a fixed temperature of 910C during deposition.
Assuming the evaporation rate is constant at this temperature,
Ge is evaporated on the prepared Si substrate held at room
temperature with a period of three minutes. During each
interval, Si and Ge AES peaks were monitored. The intensity
of the Si 92-eV AES peak as a function of the evaporation time
was found to decrease exponentially. The normalized Si AES
intensities as a function of evaporation time were compared
with those of Gossmann et.al in Fig. 44 [102]. The
evaporation rate was calibrated to be about 0.2ML/min and was
constant throughout the experiments. To detect the 50%-50%
alloy mixing point, about 4 ML Ge was evaporated onto the
Si(111) substrate held at room temperature. Annealed slowly
from 100C for 2min, at each interval AES intensities and LEED
patterns were monitored. This indicated when the Ge started
to intermix at the interface. From 310C, AES intensity ratio
of Si(92) to Ge(52) starts to grow and LEED pattern is shown
as lxl. From 4 60C, LEED pattern is shown as 5x5 and AES
244

(a) (b)
Figure 44. Ge evaporation control using AES.
(a) Normalized Si (92eV) intensity versus Ge coverage
by Gossmann et at.[102]; (b) Normalized Si(92eV)
intensity versus Ge evaporation time to calibrate
evaporation rate.
245

246
ratio [Si(92)/Ge(52)] is 6, which means Ge mole fraction is
near 50%. Above 650C, annealing AES intensity ratio is above
10 and a 7x7 LEED pattern reappeared. At higher temperature
annealing (~800C) the Ge peak has totally disappeared since
Ge diffuses into the bulk and Ge is not expected to sublimate
due to the low annealing temperature. A graph of this
annealing procedure is shown in Fig. 45. We chose the
temperature 560C for subsequent Ge evaporation, since Ge may
start to diffuse into the surface a little faster and the
exposed Si atom can make an alloy with freshly arriving Ge
atoms. After cleaning once more, Ge was evaporated onto the
Si (111) substrate held at 560C at the rate of 0.2ML/min. At
the 50 A interval, the distance between two integer beams was
monitored by digital LEED measurement using a Vidicon Camera
interfaced with NOVA computer. The 50 A alloy thickness was
assumed to be twice the thickness of evaporated Ge. For a
given film thickness, intensity profiles are measured at
several electron beam energies ranging from 35eV to 55eV.
Each line scan consists of 512 channels and the integral order
spots were typically separated by about 300 to 350 channels
depending on the LEED beam energy. The uncertainty in the
spot separation was about one channel or 0.3% of the
separation. After a film of 1000 A was grown, the sample was
transferred to RBS chamber. Two different spots were
monitored using 2.0 MeV a particles. The data is in Fig. 46

m m.
10
AES Int ensit y
Ratio of 5
S i ( 92) /Ge( 52)
0








i
Annealing
Temperat ure
100 200 300 400 500 600 700 (C)
LEED Pattern
No LEED pattern
lxl
5x5
7x7
Surf ace
Compos ition
Ge Amorphous Film
Ge S i,
* l-x
x> 0.5
Ge Si
* l-x
x~0.5
Clean
Si( 111)
Figure 45. Thermal evolution of evaporated Ge film on
Si(111).
247

Int ens ity
248
Figure 46. RBS data from 1000 of Ge/Si alloy film.

249
shows Ge is not evenly dispersed through the Ge05Si05 alloy
film. At the top a higher Ge fraction is shown compared to
the interface. Since the higher mass Ge scattered He back at
a higher energy, the independent peak at the right hand side
indicates the Ge is spread through the alloy film. The kink
in the left peak is the position of interface. This may
indicate that the thermal mixing technique (i.e. evaporation
of Ge on the hot substrate) is not able to produce a sharp
interface, only a smoothly varying Ge composition. The
thickness, estimated at the center of the Ge peak, is ~850.
At the bottom of Ge peak, the thickness is -1400.
Composition of Ge at the top is -60% and is smoothly reduces
as the thickness increases. To detect the surface morphology
and the depth profile of the film, the sample is transferred
to the Scanning Auger Microprobe Analysis chamber. The
morphology of the surface without any treatment is shown in
Fig. 47. This is not the area which was not radiated by 2 MeV
helium beam. Two distinct areas are shown. From the area 1
(grey color), AES shows Si(LMM: 91eV, KLL: 1620eV) peaks and
Ge(LMM: 51eV, KLL: 1150eV) peaks in Fig. 48(a). The peak to
peak ratio of Si(1620)/Ge(1150) is almost 1, which means Ge
is well intermixed with Si and they make an almost 50%-50%
alloy. Peaks such as C(275) and 0(511) are due to air
contamination. From the area 2 (white color) the peak to
peak AES intensity ratio of Si(1615)/Ge(1150) from Fig. 48 fb^.
is 0.65, which means Ge is richer than the area 1. After

250
Figure 47. SEM photograph of 1000 of Ge/Si alloy film. Dark
plateau is area 1 and bright islands are area 2.

251
Area 1 As Introduced
(a)
Area 2 As Introduced
9
9
0)
V
u5
*
*4
u
v t
2
1
t
Figure 48. Sputter-AES of 1000 of Ge/Si alloy film.
(a) AES spectra obtained from area 1 as introduced;
(b) AES spectra obtained from area 2 as introduced;
(c) AES spectra obtained from area 1 after 30 sec.
sputtering; (d) AES spectra obtained from area 2
after 30 sec. sputtering.

252
Area 1 After 30 sec. Sputtering
(c)
Area 2 After 30 sec. Sputtering
Figure 48.continued.

253
3Osee sputtering with raster of 3mmx3mm, Ge reduced to noise
level from area 1 shown in Fig. 48 (c) : while Ge level increase
in area 2 (shown in Fig. 48(d): Si(1617)/Ge(1148) = 0.5). In
area 1, relatively thin Ge0 5Si0 5 alloy film is sputtered off
and Si-bulk is exposed. In area 2, as the air contamination
disappears the Ge level is enhanced. The fine film quality
(i.e. no C or O) is detected from the AES profile of the
sputtered area. For further sputtering (up to five minutes
more) two areas do not show any change. The helium damaged
area (due to RBS measurement) a very similar morphology. The
same analysis procedures as for the undamaged area were
followed. More impurities (Na(990eV), C and O) were detected
from the unsputtered area than the undamaged area. These are
surface impurities induced by the helium beam. For the
sputtered surface, no impurities were detected and the
intensity profile was the same as the undamaged area. From
SEM photo, three bands were identified, which are unsputtered
area, alloy area and bulk area.
The growth mechanism of thick (-1000 A) Si^Ge,.,, alloy film
can be summarized as follows: On a room temperature substrate,
Ge can grow epitaxially as an amorphous layer. Subsequent
annealing induces the intermixing at the interface and the
film is crystallized, producing 5x5 LEED pattern and Ge
diffuses totally into the bulk for higher temperature(>650C)
annealing if the evaporated Ge is less than 3 ML. In case of
deposition on a hot (~560C) substrate, the evaporated Ge

254
intermixes with Si initially as it arrives at the surface, but
the intermixing is not even. Some area remain as a thin
Ge0 5Si0, alloy films and other areas are occupied by 3-D Ge
rich island (2-10 /m) This result is matched with the known
growth mode so called "Stranski-Krastanov" growth.
C.2. Depth Profile of Thin Alloy Film without Ge-rich Islands
In the previous section, the depth profile of a thick
(-1000 A) Ge05Si05 alloy film was studied. In this section
the depth profile of thin alloy film (<10 A) before forming
Ge-island will be presented. In Fig. 49 fa). the line scan of
edge of sputter crater of the thin pure Ge film is shown.
With no thermal heating, Ge does not diffuse into the bulk.
The kink around 272.5 is the boundary between sputtered and
unsputtered film and the point near 726.7 where 0, C, and Ge
reduces to zero is the sharp boundary between Ge film and Si
substrate. From the SEM image, no special feature (i.e. no
Ge island) is shown from this surface. So Ge evaporation on
the room temperature substrate results in an epitaxial Ge
layer which does not intermix with bulk Si.
From another sample covered with a thin (<10 A)
Ge05Si05 alloy film which was grown on the hot(-580C)
substrate, a split 5x5 LEED pattern was shown just after
growth of the film. The SEM image does not show any island
formation at all. In Fig. 49(b). the sputtered edge profile
of this alloy film is shown. Due to the low intensity of Ge,

255
(a)
Figure 49. Sputter-AES of thin (-10) Ge/Si alloy film.
(a) Sputtered edge profile of thin pure Ge film;
(b) Sputtered edge profile of thin Ge/Si alloy film;

256
it is hard to identify whether Ge diffuse into the bulk. But
this surface contains the important information of the origin
of 5x5 LEED pattern because 5x5 LEED pattern started to appear
without large Ge islands.
C.3. Depth Profile of Thick Alloy Film with Ge-rich Islands
Germanium(~200 ) was evaporated on hot Si(111)-7x7
surface held at 590C using electron gun evaporator. LEED
pattern after cooling down to room temperature was a clean
5x5 pattern. Transferring it to SAM chamber, an SEM image
(Fig. 50 fa)) was taken. Island distribution is relatively
even. Total area survey of the sample using AES shows
Ge(1147)/Si(1615) 1. An AES line scan through one of the
islands is shown in Fig. 50(b). The island is mostly composed
of Ge. The dark area consist of Si and Ge, and Ge component
is smoothly distributed. The photo of the surface after point
sputtering is shown in Fig. 50 fc) The AES line scans
indicates 4 bands which consist of unsputtered area (band 1),
sputtered area with Ge island (band 2), sputtered area with
long tail of Ge mixed with Si (band 3), and Si bulk (band 4).
The AES line scans which extends from band 1 to band 3 is
shown in Fig. 50 (d) Since the dark plateau area in Fig.
50(a). is exposed to air during transportation, the point
where impurities reduce to zero is the surface boundary of
this plateau. The reason that the level of Ge reduces rapidly
passing through this point, can be interpreted in two ways.

257
i 1
5/03/88 10.0kV 3.0kX 10.0wm
§ 3 GeSi
(a)
(b)
Figure 50. SEM and sputter-AES of thick (~200)
Ge0 5Si0 5/Si (111) 5x5 film.
(a) SEM photograph with white islands;(b) AES line
scan across one of white islands; (c) SEM photograph
after point edge sputtering, (d) AES line scan
across the sputtered edge, (e) AES depth profile
from another area.

500.0Mm
5/05/88 3.0kV
i 35X 3GeSi
Figure 50.continued.

259
. I :>
0 1.5 min. 3
(e)
Figure 50.continued.

260
One is that since the Ge island has high concentration of Ge,
removal of Ge island on the surface reduces the Ge level
rapidly. The other is that since the plateau area is wide
compared to Ge island, sputtering of this plateau area of
relatively thin Si.Ge,., alloy film decreases the Ge level
rapidly. It is not easy to determine at this point. Another
question is why is there a slow decay of Ge in band 3 ? One
possibility is that due to the thick Ge island, a small trace
of Ge island remains even after alloy film was removed.
Another possibility is thermal diffusion of Ge species into
the Si bulk. It is also not easy to determine at this point
whether there is diffused species as long as the Ge island
exists. Also it is impossible to estimate the thickness of
each band using a line scan of edge of sputter crater, since
the sputtering angle is quite glancing and the band width is
different position by position. Time-sputtered depth profile
shown in Fig. 50(e) was taken from the different area of the
same sample. Calibrating sputtering efficiency, sputtering
rate was 100 A/min. Since the lateral area of the island is
much smaller than the plateau area at the surface, the end of
the rapid decrease of Ge (~40 A) indicate the thickness of the
alloy. Also the long tail of Ge extends to 250 A of
thickness. The plateau area (area between islands) shown in
Fig. 48(a) had a rapid reduction of intensity of Ge(1148)
after 30 sec sputtering. Although, from the Ge island, Ge has
still the larger intensity after 5min more sputtering. so

261
this long tail of Ge is due to the thick Ge island, not the
diffusion of Ge into the Si substrate.
In conclusion Ge0 5Si0, alloy film growth on the hot (560-
590C) substrate follows the Stranski-Krastanov growth mode.
Initially at the very thin (<10 A) alloy layer without Ge
island in SEM image, split 5x5 LEED pattern is shown, which
indicates that the origin of 5x5 layer is not Ge-rich island.
But the function of Ge-rich islands, which are distributed
evenly at the surface, is essential to relieve the strain of
Ge05Si05 alloy layer. That is the reason why a clean 5x5 LEED
pattern and the Ge-rich island appear almost at the same time.
This growth method does not increase the thickness of alloy
for further Ge-evaporation, but does increase the width and
the thickness of the island.

REFERENCES
1. S.P. Weeks, J.E. Rowe, S.B. Christman and E.E. Chaban,
Rev.Sci.Instrum. 50(10), 1249(1979).
2. J.F. O'Hanlon, "A User's Guide to Vacuum Technology, "John
Wiley & Sons, New York, 1980.
3. C.R. Brundle, J.Vac.Sci.Technol. 11, 212(1974).
4. P. Staib, J. Kirschner, Appl.Physics, 3, 421(1974).
5. K. Siegbahn, C. Nordling, A. Fahlman, R. Nordberg, K.
Hamrin, J. Hedman, G. Johannson, T. Belgmark, s.
Karlsson, I. Lindgren and B. Lindberg,"ESCA Applied to
Free Molecules," North-Holland Publ., Amsterdam, 1969.
6. D.A. Shirley, Chem.Phys.Lett. 17, 312(1972); F.J.
Szalkowski and G. A. Somorjai, J.Chem.Phys. 56,
6097(1972), and J.T. Grant and T.W. Hass, Surf.Sci. 24,
332(1971).
7. S.P. Kowalczyk, L. Ley, F.R. McFeely, R.A. Poliak and
D.A.Shirley, Phys.Rev.B. 9, 381(1974).
8. P.H. Holloway and J.B. Hudson, Surf.Sci. 43, 123,
141(1974) .
9. P.H. Holloway and G.E. McGuire, Thin Solid Films, 53,
3 (1978) .
10. A. Ishizaka and Y. Shiraki, J.Electronchem.Soc.,
133,666(1986).
11. Z.J. Gray-Grychowski, R.G. Egdell, B.A. Joyce, R.A.
Stradling and K. Woodbridge, Surf.Sci. 186, 482(1987).
12. M. Prutton, "Surface Physics," Clarendon Press, Oxford,
1975.
13. M.Tabe, Jpn.J.Appl.Phys. 21, 534(1982).
14. J.C. Bean, G.E. Becker, P.M. Petroff and T.E.
Siedel,J.Appl.Phys. 48, 907(1977), and J.C. Bean, G.A.
Razgonyl, Appl.Phys.Lett. 41, 752(1982).
262

263
15. R.Z. Bachrach, R.S. Bayer, R. Chiaradia and G.V.
Hansson, J.Vac.Sci.Technol. 18, 797(1981).
16. D.M. Zehner, C.W. White and G.W. Ownby, Appl.Phys.Lett.
36, 56(1980).
17. S. Wright and H. Kroemer, Appl.Phys.Lett. 36, 210(1980).
18. W. Kern, D.A. Puotinen, RCA Rev. 31, 187(1970).
19. M.E. Rudd, in "Low Energy Electron Spectrometry," K.D.
Sevier ed., John Wiley and Sons-Interscience, New York,
1972, p.17, and A. L. Hughes and V. Rojansky, Phys.Rev.
34, 291(1929).
20. R. Herzog, Z.Phys., 89, 447(1934) and E.G. Johnson and
A.O. Nier, Phys.Rev. 91, 10(1953).
21. J.A. Simpson, in "Methods of Experimental Physics,"
vol.4A, ed. by V.W. Hughes and H.L. Schultz, Academic,
New York, London, 1967, pp.124-135.
22. D. Roy and J-D. Carette, Can.J.Phys. 49, 2138(1971).
23. M.P. Seah, Surf.Sci., 17, 132 (1969).
24. C.C. Chang, Surf.Sci., 25, 53 (1971).
25. B. Bauer, J.Vac.Sci.Technol., 7, 3 (1970).
26. R.W. James, "The Optical Principles of the Diffraction
of X-Rays.," Bell, London, 1962; A. Guinier, "X-Ray
Diffraction,"Freeman, San Francisco, California, 1963.
27. R. Shankar, "Principles of Quantum Mechanics," Plenum
Press, New York, 1980.
28. D.M. Newns, Phys.Letts., 60A, 461 (1977), and D.M.Newns,
in "Vibrational Spectroscopy of Adsorbates," R.F. Willis
ed., Springer-Verlag, New York, 1980, p.7.
29. B.J.N. Persson, Solid State Commun., 24, 573 (1977).
30. E. Evans and D.L. Mills, Phys.Rev., B5, 4126 (1972).
31. D.L. Mills, Surf.Sci., 48, 59 (1975).
32. A.A. Abrikov, L.P. Gor'kov, and I. Dzyaloshinski, in
"Methods of Quantum Field Theory in Statistical
Physics," Chapter 6, Prentice-Hall, Englewood Cliffs,
New Jersey, 1963.

264
33. H. Ibach and D.L. Mills, "Electron Energy Loss
Spectroscopy and Surface Vibrations," Academic Press,
New York, 1982.
34. S.Y. Tong, C.H. Li, and D.L. Mills, Phys .Rev. Lett. 44,
407 (1980); C.H. Li, S.Y. Tong, and D.L.Mills,
Phys.Rev., B21, 3057 (1980), and S.Y. Tong, C.H. Li,
and D.L. Mills, Phys.Rev., 24, 806 (1981).
35. J.P. Woods and J.L. Erskine, Phys.Rev.Lett. 50, 1277
(1983) .
36. R. Fuchs and K.L. Kliewer, Phys.Rev., 140, A2076 (1965).
37. J.D. Jackson, "Classical Electrodynamics," 2nd ed., John
Wiley & Sons, Inc., New York, 1975.
38. D.L. Mills and A.A. Maradudin, Phys.Rev. B12, 2943
(1975).
39. J.E. Black, in "Vibrations at Surface," R. Caudano,
R.Giles, A.A.Lucas eds., Plenum, New York, 1981, p.38,
and J.E. Black, Surf.Sci. 100, 555 (1980).
40. J.E. Black, B. Laks, and D.L. Mills, Phys.Rev., B22,
1818 (1986), and T.H. Upton and W.A. Goddard III,
Phys.Rev.Lett., 46, 1635 (1981).
41. R.E. Allan, G.P. Aldredge, and F.W. de Witt,
Phys.Rev.,B4, 1648 (1971); B4, 1661 (1971); B4, 1682
(1971).
42. J. Szeftel and S. Lehwald, Surf.Sci., 143, 11 (1984); M.
Wuttig, R. Franchy, and H. Ibach, Solid State Commun.,
143, 11 (1984); R. Franchy, and H. Ibach, Solid State
Commun. 57, 445 (1986), and L.L. Kesmodel, M.L. Xu, and
S.Y. Tong, Phys.Rev., B34, 2010 (1986).
43. J.S. Nelson, E.C. Sowa, and M.S. Daw, to be published at
Phys.Rev.Lett.
44. U. Backes and H. Ibach, Solid State Commun. 40,
575(1981).
45. B.N.J. Perrson and J.E. Demuth, Phys.Rev. B 30,
5968(1984).
46. J.A. Stroscio and W.Ho, Phys.Rev.Lett. 54, 1573(1985).
47. B.I. Boltaks and S.I.Budarina, Soc.Phys.Solid State 11,
330(1969).

265
48. L.N. Safronov, L.N. Aleksandrov, and R.N. Lovyagin,
Phys.Status Solidi B107, 461(1981).
49. R.T. Tung, K.K. Ng, J.M. Gibson, and A.F.J. Levi,
Phys.Rev. B 33, 7077(1986).
50. W. Daum, H. Ibach and J.E. Muller, Phys. Rev. Lett. 60,
2527(1987).
51. R. Matz and H. Luth, Phys. Rev. Lett. 46, 500(1981).
52. A. Ritz and H. Luth, Phys. Rev. Lett. 52, 1242(1984),
and A. Ritz and H. Luth, Surf. Sci. 168, 773(1986).
53. E.J. Gray-Grychowski, R.G.Egdell, B.A. Joyce, R.A.
Stradling and K. Woodbridge, Surf. Sci. 186, 482(1987).
54. L.H. Dubois, B.R. Zegarski, and B.N.J. Persson, Phys.Rev.
B 35, 9128(1987).
55. W. Ho, R.F. Willis, and E.W. Plummer, Phys.Rev.Lett. 40,
1463(1978).
56. H. Ibach and T.S. Rahman, in Chemistry and Physics Solid
Surface V, edited by R. Vanselow and R. Howe, Springer,
NewYork, 1984, p.456.
57. D.H. Ehlers and D.L. Mills, Phys.Rev. B36, 1051(1987).
58. U. Harten, J.P. Toennies, and Ch. Woll, Phys.Rev.Lett.
57, 2947(1986).
59. H. Raether, in "Excitation of Plasmons and Interband
Transitions by Electrons,"Springer, New York, 1980,p.9.
60. R.S. Becker has performed scanning tunneling microscopy
on Si(111) surfaces with surface defects due to
incomplete 7x7 ordering which show missing adatoms
(unpublished results).
61. J.E. Black, Surf. Sci. 100, 555(1980).
62. M. Grundner and H. Jacob, Appl.Phys. A39, 73(1986) and
references there in (22).
63. J.R. Arthur, J.Appl.Phys. 38, 4023(1967).
64. A.D.Buonaquisti, Y.-X.Wang, and P.H. Holloway, J.
Vac.Sci. Technol. Al(2),776(1983).
65. W.E. Spicer, P.W. Chye, P.R. Skeath, C.Y. Su, and I.
Lindau, J. Vac. Sci. Technol. 16, 1422(1979).

266
66. L.H. Dubois, G.P. Schwartz, R.E. Camley, and D.L. Mills,
Phys. Rev. B 29, 3208(1984).
67. F.P. Fehlner and N.F. Mott, Oxidation of Metal 2,
52(1970).
68. R.N.Bloomer, Brit. J. Appl. Phys. 8, 321(1957).
69. W.H. Orr, Thesis, Cornell University, Ithaca, N.Y.(1962).
70. K.S. Kim and N. Winograd, Surface Sci. 43, 625(1974).
71. L.M. Moroney, R.St.C. Smart, and M.W. Roberts, J. Chem.
Soc. Faraday Trans. I, 79, 1769(1983).
72. C. Benndorf, C. Nobl, and T. Madey, Surface Sci. 138
292(1984).
73. H. Ibach and D. Bruchmann, Phys.Rev.Lett. 44, 36(1980).
74. R.J. Birgeneau, J. Cordes, G. Polling and A.D.B. Woods,
Phys. Rev. 136, A1359(1964).
75. G. Dalmei-Imelik, J.C. Bertolini, and J. Rousseau,
Surface Sci. 63, 67(1977).
76. S. Andersson and J.W. Davenport, Solid State Commun. 28,
677(1978).
77. P.A. Cox and A.A. Williams, Surface Sci. 152/153,
791(1985).
78. W. Reichardt, V. Wagner, W. Kress, J. Phys. C8,
3955(1975).
79. M. Hock, U. Seip, I. Bassingnana, K. Wagemann and J.
Kuppers, Surface Sci. 177, L978(1986).
80. It was reported that randomly adsorbed oxygen on Ni(100)
also induced the another loss. R. Franchy, M. Wuttig and
H. Ibach at the National Symposium of AVS in Atlanta,
Georgia, October 6, 1988.
81. M. Nishijima, K. Edamoto, Y. Kubota, H. Kobayashi and M.
Onchi, Surface Sci. 158, 422(1985).
82. J.A. Schaefer, F. Stucki, D.J. Frankel, W.Gpel and G.J.
Lapeyre,J.vac.Sci.Technol. B2(3), 359(1984).
83. M.P. Seah and W.A.Dench, Surf.Interface Anal.1,1(1979).

267

00
R.G. Smeenk, R.M. Tromp, J.F. Van Der Veen, and F.W.
Saris, Surface Sci. 95, 156(1980).

in
00
C.N. Dykstal, M.S. Thesis, University of Florida (1988).

CO
H. Ibach, Phys. Rev. Lett., 24, 1416(1970).

r~
00
P. A. Thiry, M. Liehr, J.J. Pireaux and R. Caudano,
J.Elec.Spec.Rela. Phen. 39, 69(1986).

00
00
H. Ibach, H.D. Bruchmann and H. Wagner, Appl.Phys. A29,
113(1982).
89.
H. Ibach, H. Wagner and D. Bruchmann, Solid State Commun.
42, 457(1982).
90.
U. Backes, H.Ibach, Solid State Commun. 40, 575(1981).
91.
D. Grf, M. Grundner and R. Schulz, J.Vac.Sci.Technol.
A7(3), 808(1989).
92.
J.C. Bean, Physics Today Oct., 36 (1986).
93.
R. People, IEEE J.Quan.Elec. QE-22, 1696 (1986).
94.
E. Kasper and H.-J. Herzog, Thin Solid Films 44,
357(1977).
95.
T. Narusawa and Gibson, Phys.Rev.Lett. 47, 1459 (1981).
96.
A.I. Toropov, L.V. Sokolov, O.P. Pchelyakov and S.I.
Stenin, Sov.Phys.Crystallogr. 27, 450 (1982).
97.
T. Narusawa and W.M. Gibson, J.Vac.Sci.Technol. 20,
709(1982).

CO
cr\
J.C. Bean, T.T. Sheng, L.C. Feldman, A.T. Fiory and R.T.
Lynch, Appl.Phys.Lett. 44, 102 (1984).
99.
K. Shoji, M. Hyodo, H. Ueba and C. Tatsuyama,
Jpn.J.Appl.Phys. 22, L200 (1983).
100.
T. Ichikawa and S. Ino, Surf.Sci. 136, 267 (1984).
101.
J.C. Bean, L.C. Feldman, A.T. Fiory, S. Nakahara and I.K.
Robinson, J.Vac.Sci.Technol. A2, 436 (1984).
102.
H.-J. Gossmann, L.C. Feldman, and W.M. Gibson, Surf.Sci.
155, 413 (1985).
103.
J.M. Seo, D.L. Doering, D.S. Black and J.E. Rowe,
J.Vac.Sci.Technol. A4, 894 (1986).

268
104.
K. Sakamoto, T. Sakamoto, S. Nagao, G. Hashiguchi,
K.Kuniyoshi and Y. Bando, Jpn.J.Appl.Phys. 26, 666
(1987).
105.
P.M.J. Maree, K. Nakagawa, F.M. Mulders, J.F. Van Der
Veen and K. L. Kavanagh, Surf.Sci. 191, 305 (1987).
106.
P. Chen, D. Bolmont and C. Sebenne, Solid State Commun.
44, 1191 (1982).
107.
E.G. McRae, H.-J. Gossmann and L.C. Feldman, Surf.Sci.
146, L540 (1984).
108.
R.S. Becker, B.S. Swartzentruber and J.S. Vickers,
J.Vac.Sci.Technol. A6, 472 (1988).
109.
K. Takayanagi, Y. Tanishiro, M. Takahashi,
J.Vac.Sci.Technol. A3, 1502 (1985).
110.
J.A. Schaefer, J.Q. Broughton, J.C. Bean and H.H.
Farrell, Phys.Rev. B33, 2999 (1986); J.A. Schaefer,
Surf.Sci. 189/190, 127(1987), and J.A. Schaefer,
Surf.Sci. 178, 90 (1986).
111.
H.H. Farrell, J.A. Schaefer, J.Q. Broughton and J.C.
Bean, in "The Structure of Surfaces," M.A. Van Hove and
S.Y. Tong eds., Springer Verlag, New York, 1985, p.l.
112.
G. Abstreiter, H. Brugger, T. Wolf, H. Jorke and H.J.
Herzog, Surf.Sci. 174, 640 (1986).
113.
H. Froitzheim, U. Kohler and H. Lammering, Surf.Sci. 149,
537 (1985).
114.
R. Butz, E.M. Oellig, H. Ibach and H. Wagner, Surf.Sci.
147, 537 (1984).
115.
L. Papagno, X.Y. Shen, J. Anderson, G. Spagnolo and G.J.
Lapeyre, Phys,Rev. B34, 7188 (1986), and L. Papagno, L.S.
Caputi, D. Frankel, Y. Chen and G.J. Lapeyre, Surf.Sci.
189/190, 199 (1987).
116.
H.J. Gossmann, J.C. Bean, L.C. Feldman, E.G. McRae and
I.K. Robinson, Phys.Rev.Lett. 55, 1106 (1985).
117.
L.E. Davis, N.C. MacDonald, "Handbook of Auger Electron
Spectroscopy," Physical Electronics Industries, Inc.,
Eden Prairie, Minnesota.
118.
K. Heinz and K. Muller, in Springer Tracts in Modern
Physics, 91, 1 (1982).

BIOGRAPHICAL SKETCH
Jae Myung Seo was born in Seoul, Republic of Korea, on
December 27, 1955. In February, 1978, he received his B.S.
in physics from Seoul National University in Seoul. After
three years of military service as an R.O.T.C. officer, he
came to the United States of America in July 1981 and entered
the University of Florida to pursue the degree of Doctor of
Philosophy in experimental surface physics.
His research interests include physical and chemical
properties of semiconductor surfaces, metal/semiconductor
interfaces and thin films using electron spectroscopies and
photoemission spectroscopies. He is a member of the American
Physical Society, the American Vacuum Society and Korean
Scientists and Engineers Association in America.
He married the former Eunkyung Lee on June 23, 1981, and
they have two daughters, Hyo-suk and Yesuk, six and one years
old each.
269

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
[o|in E. Rowe, Chairman
>fessor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
£. OA. Pj^Po
Lucy E. aeiberling /
Associate Professor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Assistant Professor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
David A. Micha
Professor of Physics

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
4.
Pauf H. Holloway
Professor of Materials Science
and Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
David B. Tanner
Professor of Physics
This dissertation was submitted to the Graduate Faculty
of the Department of Physics in the College of Liberal Arts
and Sciences and to the Graduate School and was accepted as
partial fulfillment of the requirements for the degree of
Doctor of Philosophy.
August 1989
Dean, Graduate School



Intensity
44
Angle (degree)
Figure 4. Angular profile of the reflected electron beam
of the HREELS system .


254
intermixes with Si initially as it arrives at the surface, but
the intermixing is not even. Some area remain as a thin
Ge0 5Si0, alloy films and other areas are occupied by 3-D Ge
rich island (2-10 /m) This result is matched with the known
growth mode so called "Stranski-Krastanov" growth.
C.2. Depth Profile of Thin Alloy Film without Ge-rich Islands
In the previous section, the depth profile of a thick
(-1000 A) Ge05Si05 alloy film was studied. In this section
the depth profile of thin alloy film (<10 A) before forming
Ge-island will be presented. In Fig. 49 fa). the line scan of
edge of sputter crater of the thin pure Ge film is shown.
With no thermal heating, Ge does not diffuse into the bulk.
The kink around 272.5 is the boundary between sputtered and
unsputtered film and the point near 726.7 where 0, C, and Ge
reduces to zero is the sharp boundary between Ge film and Si
substrate. From the SEM image, no special feature (i.e. no
Ge island) is shown from this surface. So Ge evaporation on
the room temperature substrate results in an epitaxial Ge
layer which does not intermix with bulk Si.
From another sample covered with a thin (<10 A)
Ge05Si05 alloy film which was grown on the hot(-580C)
substrate, a split 5x5 LEED pattern was shown just after
growth of the film. The SEM image does not show any island
formation at all. In Fig. 49(b). the sputtered edge profile
of this alloy film is shown. Due to the low intensity of Ge,


115
o
I._J
Ta=950C
CO
a

9 J
i
, Phonon (adatom)
>
0)
B
W
S
o--
(C)
CV2
>>
tac
u
0)
C
w
| TA1200 C
Phonon
:14eV
:1OeV
( "o
(screened adatom]
/
/
/
/
/
1 /
\/ *
_ V
A
o -
(A) .
/
/
/ J-

y
o
T =800 C/"
* /
*
qc=0.2ir
fcu,(0) =6.3 meV
0.0 0.1 0.2 0.3
Momentum Transfer (X1)
Figure 15. Dispersion curves of Si(111) surfaces.


dl/dE
57
(1) Secondary Electrons
(2) Quasielastic Electrons
Elastic Electrons
(3) Auger Electrons
Interband Transition
Plasinon Excitation
Figure 7. Electron energy distribution.


2
Q I.III..I.I,., I I I I L
30 35 40 45 50 55 60 65
Evaporation Current (mA)
Figure 6. Si and Ge evaporation parameters.


APPENDIX C
GROWTH AND CHARACTERIZATION OF Ge/Si ALLOY FILMS
C.1. Alloy Film (lOOOA) Grown on Hot Substrate (560C)
Sputtering and annealing at 800C Si(111) surface induce
(7x7) reconstructed surface. The evaporation source (Ge) was
held at a fixed temperature of 910C during deposition.
Assuming the evaporation rate is constant at this temperature,
Ge is evaporated on the prepared Si substrate held at room
temperature with a period of three minutes. During each
interval, Si and Ge AES peaks were monitored. The intensity
of the Si 92-eV AES peak as a function of the evaporation time
was found to decrease exponentially. The normalized Si AES
intensities as a function of evaporation time were compared
with those of Gossmann et.al in Fig. 44 [102]. The
evaporation rate was calibrated to be about 0.2ML/min and was
constant throughout the experiments. To detect the 50%-50%
alloy mixing point, about 4 ML Ge was evaporated onto the
Si(111) substrate held at room temperature. Annealed slowly
from 100C for 2min, at each interval AES intensities and LEED
patterns were monitored. This indicated when the Ge started
to intermix at the interface. From 310C, AES intensity ratio
of Si(92) to Ge(52) starts to grow and LEED pattern is shown
as lxl. From 4 60C, LEED pattern is shown as 5x5 and AES
244


79
outside the semi-infinite (z<0) specimen in which a time
dependent charge fluctuation, n,(x,t), at a point x is
<*>(x,t) =
n,(x' ,t)
dx'
| x x'| ,
(3-30)
where the integration extends over the specimen. Retardation
effects are ignored since the time scale of information
transfer from specimen to incident electron is very short
compared to the period of excitation, which is also assumed
in the previous section. Using the Fourier transform of
x x' |_1 from equation (3-19), the potential seen by the
electron outside the specimen can be written as
0(X,t)
e'inr e-fl"z
n, (Q||Z ;t) e5"7 dz '
(3-31)
where
n,(Q||Z';t) = Jd2x' e*1*" "N n,(x',t) (3-32)
From the first integral, the factor e'<5)l 7 indicates that the
potential decays exponentially. The potential extension above
the specimen (vacuum) is z=Q||*, which is the height the
electron starts to experience the potential due to a component


148
film on Ni substrate), is because the ionic character of the
NiO single crystal is different from the NiO single crystal
film on Ni substrate. In the case of Cox and Williams,
69.5meV is much less than 71.5meV of bulk loss of longitudinal
optical phonon mode. Ionic character which determines the
strength of bonding changes the loss energy. In the case of
oxidation of Si, asymmetric stretch mode of Si02 varies from
12.5 meV to 150meV due to changing of ionic character of oxide
layer as the thickness becomes larger [81,82,83]. The
thickness of these oxide films are thicker than 10 due to
LEED pattern of the crystal film.
The spectrum after 300L of oxygen exposure is shown in
Fig. 24(a) and from AES intensity variation, the 0(510)
intensity reached saturation level at 50L exposure at room
temperature [84], The amount of exposure 300L is enough to
saturate the oxide formation on the Ni(110) surface at room
temperature. The loss at 63.3meV is the FK mode of NiO layer
on Ni(110) surface. Energy change from 67.2meV to 63.3meV is
due to thinner thickness of NiO compared to thermal oxide as
well as the poor ordering of NiO at room temperature (poor
ordering indicates loss of bonding between the coalesced oxide
islands). This poor ordering reduced the intensity of this
peak. A small hump near 49meV is also shown. Scanning up to
higher energy, 450meV O-H stretch mode does not appear, so OH
is not an adsorbed species. As the oxide ordered poorly and
covered by a layer which has high binding oxygen species


132
superlattices as well as the formation of Schottky barriers
by metal overlayer depositions [65,66].


2
gas impurities. If the pressure of the chamber is less than
lxlO'9 torr, it takes about 103 seconds for the residual gas to
cover one monolayer (1 ML) assuming the sticking probabilities
are near one. At the present time, typical ion-pumped and
well-baked systems usually have a pressure below 1x10"' torr,
which allows about three hours of experimental measurements
after surface cleaning. After the fundamental condition of
modern surface physics experiments (low UHV pressures) had
been achieved, various routine characterization methods of
surface studies soon followed. For example, low energy
electron diffraction (LEED), first demonstrated in 1927 by
Davisson and Germer, has become a routine characterization
measurement during the late 1960s and early 1970s, about ten
years after the first commercial UHV ion pumps in 1961.
Another characterization tool, Auger electron spectroscopy
(AES) became routine in the mid 1970s although its feasibility
was first demonstrated by J. J. Lander in 1953.
Surface studies can be classified by the probe particles
such as electrons, ions, atoms and photons. The electron is
the most common and convenient probe particle. The electron
is a light, charged particle; it is easily focussed to become
a non-destructive probe at low incident energies. Since the
electron has a larger scattering or interaction cross-section
than the photon, the finite electron escape depth from the
surface enables one to collect the electrons originating
mainly from the outermost 1-3 atomic layers, i.e. within the


238
For the general case where 07*0 and q = McFtyitkoteo
= k ( -0cosacos0+0osina, 6sin where Ife=
cosa 0 sina
0 10
-sina 0 cosa
is rotation matrix about OY.
Q the parallel component of q to the surface, is
Q = k(-0cosacos0 +0osina, 0sin0),
(B-12)
Q2 = k2 cos2 a [ (0cos0-0otana)2 +62 sin2 0sec2 a]
= k2 cos2 a f (6,0, a) ,
(B-13)
nQ2 + Q2v2 = (W0-Q-v||)2 +Q2va2
= k4cos2 a (602 +62 ) ,
where 0=0 is used.
d2 Q
Jacobian
Q, Qy
6 0
= k2 6cosa dfld0 .
(B-14)
(B-15)
Substituting equations (B-13,14,15) into equation (B-9), the
differential cross section is


250
Figure 47. SEM photograph of 1000 of Ge/Si alloy film. Dark
plateau is area 1 and bright islands are area 2.


28
been observed by AES. Energy shifts of Auger peaks have also
been attributed to changes in relaxation effects associated
with changes in chemical environment [6,7].
Sputtering-depth profiling with AES is a powerful
analytical technique for detecting the depth distribution of
elements in the film. In combination with rare gas
sputtering, depth profile analysis uses the surface
sensitivity of AES electron. But it is hard to get the
quantitative information of the depth of each layer. Because
of differential sputter rates caused by crystallite
orientation and surface contamination, the depth resolution
generally decreases as the film thickness increases [8,9].
Also combined with scanning electron microscopy (SEM), AES
can probe at the special point due to its high lateral
resolution while looking at the surface morphology. As
another application of AES, quantitative estimation using
standards will be introduced in Appendix A.3.
2.4. Sample Preparation
2.4.1. Overview
As mentioned in the experimental overview, one of
important factors which determines the quality of data is the
preparation of a clean and smooth sample surface. Recently
this has become more important since thin film techniques
such as molecular beam epitaxy and chemical vapor deposition


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EACW64TJ3_YBGRJD INGEST_TIME 2015-04-20T21:09:26Z PACKAGE AA00029800_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


102
ideal two layers from the actual surface in most cases. For
example, sub-monolayer coverage of adsorbate can not be a
smooth layer which can not be treated as one layer. If the
evaporated film is thin enough for the detection depth to
exceed the thickness of film and the evaporated film also has
its own surface, this system can be considered as three-layer
system not two layers. The dielectric function theory should
be generalized in order to apply for n-layer systems, which
will derive correct losses due to multilayer thin film
systems. For the sub-monolayer adsorbate system, microscopic
treatment is reasonable since the HREELS spectrometer can
detect down to 0.001 monolayer.
3.4.4. Surface Phonon Dispersion for Semi-infinite Metallic
Surface
Lattice dynamics of crystal surface can be calculated by
the finite slab method for semi-infinite metallic crystals
[39-41]. Information on the polarization and amplitude of
vibrational modes, two dimensional dispersion relation of
these modes and the spectral densities which describe the
number of vibrational states per unit energy can be estimated
by this method. Summarizing the basic ideas of the finite
slab calculation; the thickness of the slab is chosen as
about 15 layers of substrate layers to minimize the
interactions between the surfaces. Type parameters, which are
constrained by the bulk crystal structure, surface unit cell


Figure 10. Plasmon loss calculation and plasmon-phonon
coupling calculation.
(a) Plasmon loss calculation for Si(111) at the
specular geometry;
(b) Plasmon dispersion curve for Si(lll);
(c) Plasmon-phonon coupling for GaAs(lOO).


210
follows the dipole selection rule is detected. This is the
adatom phonon mode from the reconstructed Si (111)-7x7 surface.
The 'protective-oxide' annealing technique under UHV
produces a clean surface initially; but, if the sub-surface
is not smooth, the surface is quickly passivated by residual
gas impurities. These impurities consist mainly of water-
related species such as OH and H dissociated by the remaining
active sites after removal of the oxide layer. So it is
emphasized that the protective oxide should be thin (<1 ML)
to desorb at low temperature (<900C) and the surface under
the oxide layer should be smooth enough not to have many
active sites in order that a clean surface can be obtained.
It has been shown that the initial thermal oxidation rate
in UHV depends upon the substrate temperature and the final
saturation thickness of oxide layer is nearly independent of
substrate temperature since the saturation thickness occurs
due to slow bulk diffusion through the oxide which is nearly
temperature independent in this temperature range. The oxide
has ionic bonding between an oxygen atom and the substrate
species (such as Si or Ni) and the vibrational frequency
varies according to the degree of the bonding strength. In
the oxidation of Ni(110) at room temperature, the variation
of a vibrational frequency due to oxide thickness was also
monitored. For room temperature oxidation of Ni(110) with
high exposure (>100 L) of oxygen, the coalesced NiO layer
grows to saturation level at which OH in the residual gas


150
For 300L exposure of air, the peak intensity of 48meV
shown in Fig. 24(c) becomes much larger than that of 60meV.
When the air is exposed, much higher portion of water mixed
with oxygen compared to pure oxygen adsorption (but still
number of oxygen molecules is much larger than that of water)
simultaneously adsorbed on the surface. Effective exposure
of oxygen is about 100L. Before oxygen makes NiO layer up to
saturation level, water start to passivate the oxide layer.
60meV Ni-0 loss with low intensity indicate very thin and not
smooth oxide layer. Ni-OH 48meV in Ni(OH)2 indicate very thin
Ni(0H)2 layer covered the thin NiO oxide in Ni(110).
5.4. Summary
Thick thermally (300C) grown NiO layer on Ni(110) has
shown Fuchs-Kliewer (FK) mode of NiO layer, surface transverse
optical phonon mode of NiO layer and surface phonon mode of
Ni(110) induced by NiO layer. At the room temperature (300
K) oxidation through high exposure (>100L) initially grows
coalesced NiO layer up to saturation level at which OH in
residual gas starts to cover NiO layer with Ni(OH)2 layer.
Further high exposure induce thickening of Ni(0H)2 layer. The
loss energy of both NiO layer and Ni(OH)2 layer are shifted by
the thickness of their layer, which indicates that the thicker
the layer is the stronger the ionic character of the layer is.
Air exposure induces early passivation of NiO by Ni(OH)2,
before NiO reaches at the saturation level.


24
(1) The derivative spectrum dN(E)/dE of the output of the
spectrometer is integrated with respect to energy E to get
N(E) .
(2) The background under peaks is determined by means of
spline-approximation.
(3) Substraction of background from N(E) results in the
distribution NA(E) of true Auger electrons.
(4) The area under each peak consists of multiplets and loss
feature is Auger current which results from the corresponding
transitions.
(5) This Auger current depends upon the angle of acceptance
of the spectrometer, the primary current, backscattering
factor, the attenuation by scattering processes after the
transition, ionization cross-section for electron induced
ionization of the specific level, the transition probability
for the whole series related to a specific level, and the
desired atomic density. By choosing appropriate values for
each case the desired atomic density can be calculated.
It is generally assumed that the signal from a fraction
of 10'2 of a monolayer can be detected by AES. Without going
through details shown above, relative measurements (using the
standard AES sensitivity factors) can give quantitative
information on atomic density. The detailed procedures will
be discussed in Appendix A. section A.3., Quantitative AES
analysis using standards. Summarizing the introduction of
AES, AES is useful for detecting elements at the surface since


33
metal to higher work function metal. The optical pyrometer
is based on the principle that the color of the light emitted
by an object is a function of the temperature of the object.
The hot object is imaged by a lens in a plane where a filament
of light bulb is situated. The current of this filament is
regulated by a variable resistor until the brightness of the
object and that of the filament is equal. This current
reading using an ammeter can be calibrated directly in
centigrade.
Annealing samples by resistive heating, namely flowing
currents through the sample, works well with the sample which
has a low resistance. In case of annealing a sample which has
a high resistivity (>10 ohm-cm), a power supply with a high
voltage (70V-350V) and a regulated current can be used. If
a voltage source such as variac is used, the sample may be
quickly destroyed. For the semi-insulating sample the high
voltage (up to -350V) is not enough to give initial heating.
In a semiconductor, if the initial heating does not produce
thermally excited carriers, it is impossible to heat
resistively. Using another low resistive sample at the back
of the sample, the initial temperature of the sample can be
increased by heat conduction from the back substrate. Another
method of heating samples with high resistivity is depositing
a conducting film such as Ni at the back side of the film,
then the current flows through the sample at relatively low
voltage and starts to heat the sample.


176
Si(111)7x7 surface are quenched by the residual gas impurities
such as OH and H species which are detected in HREELS spectrum
for this surface. Even if dangling bonds of adatoms are
quenched, the 7x7 periodicity does not change from LEED
pattern observations so it can be understood that these OH and
H adsorbates do not break the bonds which are responsible for
the 7x7 periodicity. The Si-Hx stretch (X=l or 2) mode at
246meV, Si-H bending mode around 80meV Si-H2 scissor mode
HOmeV 0-H stretch mode at 450meV and Si-OH mode near lOOmeV
indicate major impurities are present on this surface.
Hydrocarbon stretch modes at 360meV are not detected since
hydrocarbon species did not adsorb on the surface after
annealing.
The HREELS spectrum obtained from a degreased native
oxide sample is shown in Fig. 27 fa). The asymmetric stretch
mode at 145meV indicates the saturation thickness (-2ML or 5-
6 A) of native oxide exposed to air. This energy, 145meV, is
almost same as that of a wafer preserved in deionized water
yet the thickness of such a native oxide layer is expected to
be nearly 20-30 [87] or 4-6 times that of the water-dosed
Shiraki oxide sample. This implies that the thickness
calibration reported previously may have a limited range of
validity. The strong intensity of the hydrocarbon stretch
mode at 360meV (i.e. six times larger than on the Shiraki
oxide) indicates that a large portion of active sites are


3
escape depth of -5 A. The detected electrons can be the
incident electrons which are scattered from the surface within
the escape depth as in a LEED or HREELS experiment, or they
can be secondary electrons ejected from surface atoms due to
some other excitation mechanism such as Auger relaxation of
excited core-hole states. The electron escape depth depends
upon electron energy and to a lesser extent on the atomic
number of the surface species. Usually electrons with
energies from a few eV to several thousand eV escape from 5-
20 A under the surface. In HREELS, the incident electrons
with an energy of 1-20 eV penetrate 3-10 A under the surface,
but the electric field which determines the so-called dipole
scattering cross section extends in semiconductor and
insulator materials down to 200-500 A below the surface.
Photons, ions or neutral atoms can be also used as probes in
a surface scattering experiment. Typically, several surface
techniques are combined in order to achieve the overall
purpose of experiments including surface characterization,
surface preparation and final measurements.
The structure and elementary vibrational excitations of
bulk condensed matter can be studied by spectroscopies such
as inelastic neutron scattering, infrared (IR) spectroscopy,
Raman spectroscopy and inelastic electron tunneling. High
Resolution Electron Energy Loss Spectroscopy (HREELS),
Infrared absorption-reflection spectroscopy and neutral atom
inelastic scattering are the main probes of the vibrational


263
15. R.Z. Bachrach, R.S. Bayer, R. Chiaradia and G.V.
Hansson, J.Vac.Sci.Technol. 18, 797(1981).
16. D.M. Zehner, C.W. White and G.W. Ownby, Appl.Phys.Lett.
36, 56(1980).
17. S. Wright and H. Kroemer, Appl.Phys.Lett. 36, 210(1980).
18. W. Kern, D.A. Puotinen, RCA Rev. 31, 187(1970).
19. M.E. Rudd, in "Low Energy Electron Spectrometry," K.D.
Sevier ed., John Wiley and Sons-Interscience, New York,
1972, p.17, and A. L. Hughes and V. Rojansky, Phys.Rev.
34, 291(1929).
20. R. Herzog, Z.Phys., 89, 447(1934) and E.G. Johnson and
A.O. Nier, Phys.Rev. 91, 10(1953).
21. J.A. Simpson, in "Methods of Experimental Physics,"
vol.4A, ed. by V.W. Hughes and H.L. Schultz, Academic,
New York, London, 1967, pp.124-135.
22. D. Roy and J-D. Carette, Can.J.Phys. 49, 2138(1971).
23. M.P. Seah, Surf.Sci., 17, 132 (1969).
24. C.C. Chang, Surf.Sci., 25, 53 (1971).
25. B. Bauer, J.Vac.Sci.Technol., 7, 3 (1970).
26. R.W. James, "The Optical Principles of the Diffraction
of X-Rays.," Bell, London, 1962; A. Guinier, "X-Ray
Diffraction,"Freeman, San Francisco, California, 1963.
27. R. Shankar, "Principles of Quantum Mechanics," Plenum
Press, New York, 1980.
28. D.M. Newns, Phys.Letts., 60A, 461 (1977), and D.M.Newns,
in "Vibrational Spectroscopy of Adsorbates," R.F. Willis
ed., Springer-Verlag, New York, 1980, p.7.
29. B.J.N. Persson, Solid State Commun., 24, 573 (1977).
30. E. Evans and D.L. Mills, Phys.Rev., B5, 4126 (1972).
31. D.L. Mills, Surf.Sci., 48, 59 (1975).
32. A.A. Abrikov, L.P. Gor'kov, and I. Dzyaloshinski, in
"Methods of Quantum Field Theory in Statistical
Physics," Chapter 6, Prentice-Hall, Englewood Cliffs,
New Jersey, 1963.


38
d2y/d02 + y = E0/ (y E cos2a) (2-10)
where y is r/r0, r0 is radius of parallel path to cylinder,
E0 is an incident energy of electrons, and (r,0) and E are a
cylindrical coordinate and an energy of electrons inside the
cylinder. The value a is the angle between the trajectory
and parallel path at the same point, namely angular
aberration. A first order solution of equation (2-10) is
[19],
y E0(l-cos/2 0) / (E cos20) + cos/2 0 -{tana sin/2^ 0}//2.
(2-11)
Since the last term contains the largest angular aberration
term, to achieve the best focussing the last term should
vanish, i.e. sin/2 0 = 0. J2

The potential inside a infinite cylinder is
V(rt) = a In rt + b (2-12)
where rt is a radius of trajectory. The potential of the
outer plate is V(R), that of inner plate is V(r) the
difference between them is A = V(R) V(r) and V(ro){=0) is
the potential of main path which is parallel to the capacitor
plate. Using above conditions, constants a and b in equation
(2-12) can be expressed in terms of R,r, A and r0 as follows,


LIST OF FIGURES
FIGURE TITLE PAGE
1 HREELS chamber 18
2 Electron escape depth 22
3 HREELS spectrometer 37
4 Angular profile of the reflected electron beam of
the HREELS system.
44
5 Block diagram of data acquisition system 47
6 Si and Ge evaporation parameters 54
7 Electron energy distribution 57
8 The reciprocal lattice and Ewald construction.
(a) Three dimensions; (b) Two dimensions.
63
9 Schematic diagrams of semi-classical dipole
scattering.(a) Electron trajectory and molecular
dipole moment; (b) Transferred momentum; (c) Polar
plot of scattering intensity.
71
10 Plasmon loss calculation and plasmon-phonon coupling
calculation. (a) Plasmon loss calculation for
Si(111) at the specular geometry; (b) Plasmon
dispersion curve for Si(lll); (c) Plasmon-phonon
coupling for GaAs(lOO).
92
11 Two layer mode dispersion and polarization.
(a) Surface eigenmode dispersion for a slab of
material with dielectric function es(w) on a
substrate with a dielectric constant, e(>0); (b)
The electric fields for ut mode.
101
vii


190
In this chapter these questions are considered and
experiments were conducted as follows: First, the surface
excitation from SixGe,_x alloy films are monitored by HREELS
and a hydrogen (and deuterium) titration technique. Second,
the fine relaxation of this SixGe,.x alloy is monitored by
digital-LEED intensity measurements to detect the critical
thickness. Finally, the growth mechanism of this SixGe,_x alloy
film will be examined using thermal evaporation, LEED and AES
techniques. The SixGe,_x alloy film has been characterized by
RBS and Scanning Auger Microprobe after removal from the
analysis chamber. The growth and characterization of GexSi,.x
alloy film will be shown in Appendix C.
7.2. Sample preparation
Si(111) surface has been Ar-ion sputtered and annealed
at 900C to prepare a clean 7x7 reconstructed surface. Ge
source is heated by e-beam and Ge is thermally evaporated on
the Si (111) substrate. Either on a hot substrate held at
580C or on a substrate at room temperature and post-annealed
up to 580C, Ge intermixes at the Si surface and produces 5x5
LEED pattern. Ge thickness calibration is shown in section
2.9.. For hydrogen titration experiments hydrogen molecules
are dosed on the Ge0 5Si0 ¡(111) 5x5 surface in front of the hot
tungsten filaments (~2000C) Using ammeter through measuring
the current passing through the sample, an effective exposure


19
of the diffraction pattern (Laue geometry), a rear view LEED
system was constructed from a commercial LEED optics using the
procedure developed by E.E.Chaban and co-workers[1].
Photographs of our LEED patterns are usually taken through
this rear view LEED port.
Pumping procedures to achieve UHV from the opening to
the air of the University of Florida HREELS chamber are as
follows.
(1) As soon as the sample is mounted, the electrical
connection of sample relative to ground is checked to ensure
the sample has not accidentally been grounded. Then all the
ports opened during atmosphere operation are closed.
(2) The turbo-molecular pump line is opened and the mechanical
pump is turned on to achieve low pressure (-1 torr) at the
foreline of the turbo-molecular pump.
(3) If the pressure is below 1 torr between the turbo
molecular pump and the mechanical pump, the turbo molecular
pump and water chiller for cooling the turbo molecular pump
are turned on.
(4) The turbo molecular pump is used during the backing
procedure. The time required to reach the ultimate pressure
in one hypothetical example of an unbaked metal-gasket system
is lO'-lO5 hours, even though there is no leak in the
chamber[2].


256
it is hard to identify whether Ge diffuse into the bulk. But
this surface contains the important information of the origin
of 5x5 LEED pattern because 5x5 LEED pattern started to appear
without large Ge islands.
C.3. Depth Profile of Thick Alloy Film with Ge-rich Islands
Germanium(~200 ) was evaporated on hot Si(111)-7x7
surface held at 590C using electron gun evaporator. LEED
pattern after cooling down to room temperature was a clean
5x5 pattern. Transferring it to SAM chamber, an SEM image
(Fig. 50 fa)) was taken. Island distribution is relatively
even. Total area survey of the sample using AES shows
Ge(1147)/Si(1615) 1. An AES line scan through one of the
islands is shown in Fig. 50(b). The island is mostly composed
of Ge. The dark area consist of Si and Ge, and Ge component
is smoothly distributed. The photo of the surface after point
sputtering is shown in Fig. 50 fc) The AES line scans
indicates 4 bands which consist of unsputtered area (band 1),
sputtered area with Ge island (band 2), sputtered area with
long tail of Ge mixed with Si (band 3), and Si bulk (band 4).
The AES line scans which extends from band 1 to band 3 is
shown in Fig. 50 (d) Since the dark plateau area in Fig.
50(a). is exposed to air during transportation, the point
where impurities reduce to zero is the surface boundary of
this plateau. The reason that the level of Ge reduces rapidly
passing through this point, can be interpreted in two ways.


ELECTRON SCATTERING STUDIES OF SURFACE PHONON-PLASMON MODES
OF SEMICONDUCTORS
By
JAE MYUNG SEO
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1989


194
broadening may be a disordered surface due to evaporation.
But the rapid decrease of the intensity of the elastic peak
suggests the surface is quite smooth. Since Si0 5Ge0 5(111)-5x5
surface has the similar surface reconstruction as Si(111)-7x7,
it is considered that the adatom on the surface induces
metallic characteristics at this alloy surface. Therefore the
broadening is due to small transitions in this metallic band.
When the angle difference exceeds the limit of the aperture
(-1.3), the broad shoulder around 23 meV starts to develop.
The distinct peak is shown at A0=2.5, and the peak becomes
indistinguishable below A0=4 This is the typical peak
evolution of dipole active loss near the elastic peak which
was shown in chapter 4. The adatom vibration (phonon) induces
this peak. The broad feature may be due to a disordered
surface. But, as the alloy consist of Si and Ge, there are
three kinds of bonding at the surface, namely Ge-Ge, Ge-Si and
Si-Si. These bondings have slightly different frequencies,
since they are covalently bonded. From the Raman scattering
measurement, longitudinal optical (LO) modes of these three
bondings appeared at -36 meV, -52 meV and -62 meV respectively
in the bulk alloy [112]. At least 10 different bonding sites
are involved in this adatom vibration considering only the
nearest nine local atoms and this broadening is expected from
various combinations of these three kinds of bondings.
Using diatomic simple oscillator model the vibrational
frequency of diatomic molecule is


224
section, the slit potential of the analyzer (=A) should be set
to the same value as the slit potential of monochromator (=M) .
Connecting the outer plate of main analyzer to a picoammeter
to detect the current in the same way as a monochromator.
Setting electrode 4 of deceleration optics to the slit
potential (A) electrode 3 is varied to detect the current
from the picoammeter. The current can be maximized by
adjusting e,-e4. Reconnecting main analyzer to power supply
and setting difference of main analyzer, AM, to difference of
main monochromator, A^,, the difference of second analyzer, Asa,
is varied until the current is detected from the channeltron
(the high voltage source is set to 3kV) Once a straight-
through beam is detected, readjusting most of knobs to
optimize the signal is followed until the symmetric and sharp
(at least less than 20meV) intensity signal is achieved.
Since the optimum setting at straight through geometry is not
related to the sample, this setting can be recorded and used
for different samples (I, = l.5xlO',0A, rate = 104/sec and
resolution 5-7 meV) as long as the work function of the
capacitors are not changed.
Current reflected from the sample can be optimized.
Rotating the sample surface parallel to the direction of
straight-through electron beam, the sample is transferred
slowly to the scattering section. As the sample approaches
to the scattering point, the intensity of the straight-through
beam should decrease due to the bias on the sample. The lost


223
part, change independently so that the potential on any given
knob can be changed in order to reach the optimum maximum
intensity in most cases. In here the full tuning procedure
from the very beginning will be presented. Tuning problems
such as asymmetry and proper zooming will be discussed.
Finally tuning examples will be compared using setting
records.
Initially to detect an electron beam inside the
monochromator, a current from the outer capacitor (R^J of main
monochromator should be maximized using a picoammeter.
Initially the filament current for the cathode is set to 2A,
the repeller to -1 V, anode 2 to 10 V, anode 3 to -2 V. The
contact potential of monochromator is varied from the most
negative value to higher values to detect the current pass
through the premonochromator until changing the difference of
premonochromator (Apm) reduces the maximum current of R*,, .
Reconnecting R, to the power supply, the current at the target
position is detected. At the straight through geometry, the
target as a beam probe is put into the center of the
scattering section to receive the current. The electrode 1
of acceleration optics is set 3 V, electrode 2 to 3 V, and
primary energy 5 V. The difference of main monochromator is
varied in order to detect the current from the target.
Achieving the current of target of -10,0A, all the knob in
cathode system and monochromator system can be adjusted.
Removing beam probe (i.e. target) outside of scattering


128
the calculation of losses from Si(111) surface, which has used
the Lindhard dielectric function and dielectric function
theory for loss function intensity. Therefore in the
GaAs(lOO) case, effective carrier density is also a parameter
that can be used to fit the data to the theoretical curve.
The only difference is that a phonon term is added in the
dielectric function since GaAs is a polar-semiconductor and
has a FK optical phonon mode. Shown in Fia. 19 is the result
of plotting energy loss versus momentum transfer Also
shown are the calculated dispersion curves for GaAs with
effective carrier density of 6x10'* cm'3 (for solid line
[eo-&o]/[&ofl]=0.168) Dash lines show the dispersion curves
for the uncoupled system. Note that the data points in the
lower branch seem to fit the solid theoretical curve while
data points for the ho. branch seem to fit the dashed curve.
This indicates that our simple theory based on dielectric
response is inadequate. Filled squares correspond to observed
loss features at energies larger than the phonon energy with
the dot-dashed curve being the plasmon branch curve shifted
by 0.77 of the FK mode energy. Thus these latter points
correspond to an interesting combination of plasmon plus
phonon modes rather than sequential losses. The experimental
data are in good agreement with the calculated dispersion
curves with some disagreement at the lower q^values. It is
thought that this discrepancy is due to a surface migration
effect due to repeated sputter and anneal cycles. Angle


450
>
C/l
C
CD
C
II
xlOO
GeD bending
FWHM=20 meV
2.5L Dat om
50
11
1 106
Ge D stretching
/
/

\ 177 ^
~
361
-100 0 100 200 300 400 500
Energy (meV)
(a)
Figure 35. Deuterium titration on Ge0 5Si0,(111)-5x5.
(a) 2.5 L, D; (b) 5L, D.
201


257
i 1
5/03/88 10.0kV 3.0kX 10.0wm
§ 3 GeSi
(a)
(b)
Figure 50. SEM and sputter-AES of thick (~200)
Ge0 5Si0 5/Si (111) 5x5 film.
(a) SEM photograph with white islands;(b) AES line
scan across one of white islands; (c) SEM photograph
after point edge sputtering, (d) AES line scan
across the sputtered edge, (e) AES depth profile
from another area.


203
surface which has almost the same amount of evaporated Ge.
The total area of Ge-island is quite small compared to that
of alloy layer. Even if Ge island also has a dangling bond,
the total number of dangling bond in Ge0 5Si 5(111) alloy
portion is much larger than that of the island and the
dominant LEED pattern is 5x5 not 7x7 due to Ge island (see
next section 7.4.). Since the reactivity of Ge with a
hydrogen atom is similar to that of Si, the possibility of
impartial reaction with one species can be excluded.
Concluding this section, Ge0 5Si0 5 (111) (5x5) surface has
the similar adatom vibrational mode to Si(111)-7x7 surface.
Quasi-elastic broadening from the alloy surface is also
detected, which means that the alloy surface is metallic.
Even if it is hard to distinguish the origin of Ge-H, mode,
considering the total area of the island is much smaller than
that of the alloy layer, Ge0 5Si 5(lll)-(5x5) surface has much
portion of Ge dangling bond.
7.4. Surface Relaxation Measurements
Using Digital LEED
The experimental procedure of film growth and
experimental set up of digital LEED measurement is discussed
in Appendix C and Appendix A. The sample used here is a 1000
A alloy film grown on a hot Si(lll) substrate (560C) The
intensity profile (primary energy 35eV) of the line connecting


48
controller is placed on the IEEE bus and is used for obtaining
angle resolved HREELS. Two stepping motors, one connected to
monochromator and the other connected to sample manipulator,
allow one to change angle of monochromator with a resolution
of 0.05 degree of rotation, and to change the angle of sample
with a resolution of 0.1 degree. A Keithley model 197
multimeter is connected to the IEEE bus to monitor the
baseline (DC off-set) of HREELS analyzer. The Kepco supply,
Klinger controller, Camac crate and Keithley multimeter are
interfaced by an IEEE-488 bus to an IBM pc/XT running under
DOS 2.10. Control of the IEEE bus is accomplished with a
Tecmar IEEE interface card taking up one slot in the personal
computer.
2.8. Oxidation and Hydrogen Titration
Using a gas handling system, reactive gases such as
oxygen, hydrogen and deuterium can be chemisorbed or adsorbed
on sample surfaces. In this section, two main procedures,
namely oxidation and hydrogen-adsorption, will be introduced.
Several different oxidation procedures, chemical wet
oxidation, thermal oxidation and room temperature oxidation
will be discussed. Hydrogen and deuterium adsorption on
semiconductors will be also discussed.
Since -20% of air is oxygen and oxygen is a very reactive
gas, silicon substrates can grow a relatively thin silicon


360
Energy (meV)
Figure 32. HREELS spectra obtained f rom Ge0 ,Si0 5( 111)-5x5 showing a strong
elastic peak (E=0) and a weak loss peak (E105meV) due to
residual carbon impurities.
192


Chapter 3
THEORETICAL BACKGROUND
3.1. Overview
The interaction of monoenergetic electrons with a sample
surface will give rise to a typical energy distribution of
scattered electrons due to incident primary electrons as well
as secondary electrons from the sample surface. A typical
energy spectrum, N(E), which is the emitted electron intensity
per unit energy, is shown in Fig. 7 and is usually a strong
function of emitted angle as far as relative intensities are
concerned. The electron energy distribution, N(E), shows
three features. A large maximum occurs at low energy. This
first peak is due to electrons which suffer multiple inelastic
collisions induced by a collision cascade process in the solid
[23]. Near the incident electron energy, the sharp "elastic"
peak which is comprised of the elastically scattered electrons
plus the 'quasi-elastic' electrons that have lost a small
amount of energy (0-200meV) occurs. Finally the intermediate
region of the emission spectrum exhibits a series of smaller
maxima with a small background. These losses result from
56


BIOGRAPHICAL SKETCH
Jae Myung Seo was born in Seoul, Republic of Korea, on
December 27, 1955. In February, 1978, he received his B.S.
in physics from Seoul National University in Seoul. After
three years of military service as an R.O.T.C. officer, he
came to the United States of America in July 1981 and entered
the University of Florida to pursue the degree of Doctor of
Philosophy in experimental surface physics.
His research interests include physical and chemical
properties of semiconductor surfaces, metal/semiconductor
interfaces and thin films using electron spectroscopies and
photoemission spectroscopies. He is a member of the American
Physical Society, the American Vacuum Society and Korean
Scientists and Engineers Association in America.
He married the former Eunkyung Lee on June 23, 1981, and
they have two daughters, Hyo-suk and Yesuk, six and one years
old each.
269


106
force constant is used for calculation of S4, the loss energy
of the S4 mode is always lower than the experimentally
measured energy near the zone edge. If a 20% higher force
constant is used, the calculated value is matched well with
experimental value at the zone boundary. This indicates that
by the termination of the bulk, the spacing between the first
and second layer contracts compared to the bulk layer spacing
(3.2% contraction of spacing between the first and the second
layer was estimated [42]). The S6 mode is only detected near
the zone edge. In Cu(001) case, the surface force constant
increased by 16-20% and showed a similar contraction as
Ni(001). Surface phonon dispersion curves are mostly
influenced by the surface structure than the specify metallic
element.
It was pointed out in the dielectric function theory that
if Qj|<0 (in this case the Rayleigh mode near zone center) the
mode penetrate deeply into the bulk. In this regime, the role
of the contracted surface is very small and the dispersion
curve is matched well with the dispersion curve which used the
bulk value of a force constant. As Q|( approaches zone
boundary, the mode begins to localized at the surface, the
discrepancy become distinct near the zone boundary [42]. This
reguires a new theoretical model which could produce the
effect of surface contraction [43].


125
in the straight-through and the reflected geometries was
estimated.
Higher energy-loss data [data set (B) and (C)] are also
plotted in Fig. 15. Although an effective carrier density
can be calculated for q,|, the corresponding calculated cutoff
wave vector is not consistent with this density. Also the
more rapid decrease in intensity suggests a second loss
mechanism must be operative on surface annealed at higher
temperatures. Annealing at higher temperatures reduced the
density of sputtering defects near the surface and improved
the ordering of the stable 7x7 reconstructed surface. The
change of intensities of the loss features at nonspecular
angles, shown in Fig. 14. indicates that the losses were
probably dipole-like [56]. Our 26-30meV loss peak has a
similar angular dependence to the 130meV mode from an ordered
monolayer of hydrogen on W(100) (i.e., dipole scattering)
[55]. At an annealing temperature of 1050C, the plasmon loss
peaks appear to remain but cannot be as well resolved;
therefore a distribution of carrier density probably occurs.
At the highest annealing temperature of 1250C, the loss
energy [data points (c) in Fig. 151 was reduced by ~8meV.
This behavior is again attributed to ordering of the 7x7
surface periodicity. For the clean well-ordered surface one
expects the adatoms to be the last features to order [60],
If the adatoms of the reconstructed surface have a net
repulsive interaction, then it is possible that both their


69
by Bose factor, n=[exp(hw/kT)-l]'1. The ratio of intensity of
gain to loss peak (sn/(n+l)) is exp[-(hw/kT)].
From the law of conservation of momentum, the loss energy
evolution relative to transferred momentum, either by changing
geometry of scattering from specular to non-specular or by
increasing primary energy at specular geometry, will give a
dispersion relation of each mode. Based upon these two
conservation laws, there are various applications of HREELS
for the detection of surface excitations. The primary object
of HREELS theory is then to describe a precise scattering
differential cross-section which indicates the loss energy,
its intensity and angular distribution.
In this section we will follow a semi-classical
derivation of the differential cross section applied to a
simple system which consists of an electron approaching an
adsorbate on a metal surface [28]. The reasons it has been
chosen are as follows: first it has several simple steps to
derive the end result and is matched well with experiment,
second it is easily connected to the actual geometry of the
experimental set-up, and third it can be extended to further
complicated, more advanced quantum mechanical theories. In
addition to this semi-classical theory, the result of a
quantum mechanical theory applied to a similar system by
Persson will be compared with the semi-classical theory [29].
Also the result of a full quantum mechanical description
applied to a more generalized system (i.e. dielectric function


Energy (meV)
(c)
Figure 28
120
-50 0 50 100 150 200
Energy (meV)
(d)
continued.
170


96
plotted in Fia. 10 (b) When q,|=0, i.e. at specular geometry,
the losses matched with Fia. 10 fa).
The cut-off wave vector (qc) above which there is
continuum limit, is wp/v,. Therefore qc also depends upon the
doping density by net,,/6. So fitting the curve to the actual
data can be checked by this cut-off wave vector.
On the other hand for polar semiconductors the surface
plasmon frequency is very close to that of the surface optical
phonon, these two modes can not be an independent mode, but
they couple to form two new modes which are admixtures of a
surface plasmon and a surface phonon. Then the dielectric
function for the ionic crystal with carriers can be expressed
in terms of dielectric function of longitudinal optical
phonon, i.e. (equation (3-42)), and Lindhard dielectric
function for highly doped semiconductors as follows:
(c0&o) wT02 up2
(W, qj,) = &o + (3-49)
wT02 w2 iwr u2-qs2 + (iw/T)
where r and 1/r are damping terms, qs2=0.6 q^ vF2 and all
others defined same as the previous section. From the pole
of the loss function, £(w,q|()=-l, the coupled phonon-plasmon
losses are
W2 (qs2 ) = h [A +qs { (A+qs2 )2 -4Bqs2 -4C)*] ,
(3-50)


67
N
= |f(0 E) |2 exp[-<(Su)2> E exp[iS-(r.-rj ]
i j
(1+<(SU,) (S U,) >+ [exp< (8 u,) (8 Uj) >-l-< (8 U|) (8 !!,)>] }.
(3-14)
The factor exp[-<(Su)2>] is Debye-Waller factor. The first
term in the curly bracket is the zero-phonon scattering which
is just the rigid-crystal scattering reduced by the Debye-
Waller factor as discussed previously. The second term is the
one phonon contribution to the thermal diffuse scattering.
The third term is multiphonon scattering. So the effect of
thermal vibration is to remove the intensity from the Bragg
peaks and to redistribute the intensity throughout the
Brillouin zone. Multiple scattering and inelastic process can
be treated in self-consistent way to understand more detailed
elements in diffraction problems. This is the goal of
dynamical theory which will not covered here.
The theoretical treatment of elastic scattering of low-
energy electrons within a kinematic framework can be
summarized as follows. Low energy electrons are scattered
within a few atomic planes of the surface. Two dimensional
diffraction pattern intensities with information on long-range
order and the structure of the surface layer are modified due
to both elastic and inelastic process. A full treatment
involves the inner potential of the crystal, attenuation due


143
Energy (meV)
(b)
Figure 24.continued.


121
. ¡. 1 ,
-100 -50 O 50 100
Energy (meV)
Figure 18. ARHREELS spectra obtained from highly doped
GaAs(lOO) for specular (A0=O) and non-specular
scattering.


Figure 12. Polarization of surface inodes. The displacements
in the surface layer for the three surface modes
at the x point of the two-dimensional Brillouin
zone for an fee crystal with nearest-neighbor
central force interactions and a (100) surface.
Modes S6 and S, involve displacements mainly
parallel to the surface while mode S4 involves
perpendicular displacements.


182
6.4. Summary
From the variation of the asymmetric stretching mode
intensity and energy of the oxide layer thermal oxidation
stages can be monitored to a degree. Initially oxygen
molecules adsorbed on dangling bond of silicon substrate.
After dissociated oxygen atoms chemisorbed between the first
layer and the second layer of Si substrate a superoxide-like
species disappeared. As the oxidation continues oxygen atoms
start to chemisorb between the second layer and the third
layer of Si substrate. Thereafter the oxide layer starts to
be similar to amorphous Si02, i.e. the oxide layer starts to
lose its directional orientation with respect to the diamond
structure. We have also found that differently prepared
oxides have different passivation thickness according to their
oxygen supply and passivating impurities. For Shiraki
oxidation, very smooth and thin oxide (1.52ML or 4) covers
the surface and surface is inert to impurities after removing
the oxide layer. While the water preserved oxide is
disordered and saturation thickness of the oxide (2ML or 5.4)
is not even. After removing this uneven oxide layer the
surface is easily contaminated by impurities such as OH and
H. Native oxide formed by air has a similar HREELS spectrum
but this surface is initially contaminated by carbon
impurities and has a stable Sic after removing oxide layer by
annealing. In this experiment we tested the thermal stability


122
Combination of
Figure 19. Experimental data and calculated dispersion
curves for highly doped GaAs(lOO).


142
_CD
CD
G_3
Energy (meV)
Figure 24. HREELS spectra obtained from oxides on Ni(110)
grown under UHV at room temperature.
(a) 300L, 02; (b) 1.8xlOL, 02; (c) 300L, air.


62
sin2(N,S a/2) sin2(N2S b/2) sin2(N3Sc/2)
j(S) :
sin2(S-a/2) sin2(Sb/2) sin(S-c/2)
(3-6)
The intensity becomes a maximum when S satisfies the Laue
conditions, i.e.,
S a = 27rl, S b = 27rm, and S c = 2nn. (3-7)
In terms of reciprocal lattice vectors A, B, and C, which
are A=2ir (bxc) / [a*bxc] and so on, the position of the
reciprocal lattice point (lmn) is GlBn = 1A + mB + nC, which
is a vector normal to the family of planes whose Miller
indices are (lmn) with a magnitude 16lmn | = 2n/dlnn where d,n
is the interplanar distance. Then the Laue conditions are
S a = 27rl = 27rGllnn-(a/27r) = GlBn a and so on for b and c; thus,
S = GlBn = k K (3-8)
The diffraction geometry is displayed by the Ewald
construction shown in Fig. 8(a). The incident wave vector
has a fixed magnitude and direction and is terminated at the
origin of the reciprocal lattice. The origin of the Ewald
sphere whose radius is the magnitude of incident wave vector
is at the same position as the origin of the incident wave
vector in momentum space. Whenever this Ewald sphere passes
through a reciprocal lattice point, the diffraction condition


183
of each oxide by increasing the annealing temperature. The
asymmetric stretch mode of the HREELS spectrum obtained from
these intermediate oxides indicates that the oxide thickness
and oxygen desorbing temperature seems dependent upon this
thickness.


153
chemisorbed intermediate oxide under ultrahigh vacuum (UHV)
(thermal oxide).
Many chemical etching technigues for cleaning an Si
surface have been reported. The common and basic idea of
chemical etching is removal of a native oxide by the chemical
etchant thus reducing surface impurities and keeping the
surface as clean as possible before introducing the sample
into UHV. The Shiraki technique consists of a wet chemical
treatment to eliminate contaminants on the Si substrate and
thin oxide formation to protect the clean Si surface from
contamination (refer to chapter 2.8.(101). An oxide film
grown by chemical treatment on the Si substrate has a function
as a passivation film. Since the oxide surface is much less
active than the bare Si surface there are few carbon
contaminants on the oxidized Si surface, and they can be
removed more easily than those on the Si surface. Also
forming a smooth surface through the repetition of the
chemical oxidation and etching process help to reduce active
sites on the oxide surface where contaminants are easily
adsorbed. The remaining active sites are filled with Cl atoms
which act as unstable adsorbents blocking active surface sites
before carbon atoms are adsorbed.
An Si(111) wafer was preserved in deionized water for
three weeks in order to isolate the chemically cleaned wafer
from oxygen molecule as well as other contaminants in the air.
Even though the water isolated the wafer from the air, water


66
the crystal it is kln=27r[ (E+V0)/150.4]* A'1. Conservation of
parallel crystal momentum causes a refraction of the electron
beam inside of crystal with refraction index
sin* kln[l+(V0/E) ]"
n = = (3-12)
sinfl(n k
The main contributions to the inner potential come from
correlation and exchange interactions as well as the surface
dipole layer potential and an imaginary part of the potential
due to inelastic interactions.
Another factor determining the intensity is the thermal
vibration of atoms in the crystal. The energy resolution in
typical low-energy electron diffraction is insufficient to
observe the loss or gain of phonon energies, so the intensity
measured corresponds to both true elastic and integration over
inelastic events due to phonon scattering. A diffraction
experiment is equivalent to scattering from instantaneous and
stationary configurations of scatterers. From equation (3-
5), the instantaneous scattered intensity is
N
I(S) = | f (0 ,E) |2 2 exp[is-(r.+u.-r.-u,)], (3-13)
i j
where u, is the instantaneous displacement of the i-th atom
from its equilibrium position r,. The thermal average of
intensity is


12
combination of various surface science techniques. In this
chapter Ge0 5Si0 5 alloy films grown on Si(111) are investigated.
In the first part, surface vibrational excitations of the
alloy film and its atomic hydrogen titrated surface were
investigated by HREELS. Surface phonon and surface elemental
species of the alloy film were considered. The origin of the
5x5 reconstruction is discussed. In the second part
pseudomorphic growth studies and surface strain relaxation
studied by digital LEED are presented. The critical thickness
for pseudomorphic growth is discussed in the context of our
measurements on (111) pseudomorphic films. In addition to
the previous parts studies of the growth mechanism and
morphology of the films studied by LEED, AES and Auger
microscopy are presented in the Appendix C.
In Chapter 8 the experimental conclusions are summarized.
In addition to experimental conclusions, recommendations for
future studies are suggested.


264
33. H. Ibach and D.L. Mills, "Electron Energy Loss
Spectroscopy and Surface Vibrations," Academic Press,
New York, 1982.
34. S.Y. Tong, C.H. Li, and D.L. Mills, Phys .Rev. Lett. 44,
407 (1980); C.H. Li, S.Y. Tong, and D.L.Mills,
Phys.Rev., B21, 3057 (1980), and S.Y. Tong, C.H. Li,
and D.L. Mills, Phys.Rev., 24, 806 (1981).
35. J.P. Woods and J.L. Erskine, Phys.Rev.Lett. 50, 1277
(1983) .
36. R. Fuchs and K.L. Kliewer, Phys.Rev., 140, A2076 (1965).
37. J.D. Jackson, "Classical Electrodynamics," 2nd ed., John
Wiley & Sons, Inc., New York, 1975.
38. D.L. Mills and A.A. Maradudin, Phys.Rev. B12, 2943
(1975).
39. J.E. Black, in "Vibrations at Surface," R. Caudano,
R.Giles, A.A.Lucas eds., Plenum, New York, 1981, p.38,
and J.E. Black, Surf.Sci. 100, 555 (1980).
40. J.E. Black, B. Laks, and D.L. Mills, Phys.Rev., B22,
1818 (1986), and T.H. Upton and W.A. Goddard III,
Phys.Rev.Lett., 46, 1635 (1981).
41. R.E. Allan, G.P. Aldredge, and F.W. de Witt,
Phys.Rev.,B4, 1648 (1971); B4, 1661 (1971); B4, 1682
(1971).
42. J. Szeftel and S. Lehwald, Surf.Sci., 143, 11 (1984); M.
Wuttig, R. Franchy, and H. Ibach, Solid State Commun.,
143, 11 (1984); R. Franchy, and H. Ibach, Solid State
Commun. 57, 445 (1986), and L.L. Kesmodel, M.L. Xu, and
S.Y. Tong, Phys.Rev., B34, 2010 (1986).
43. J.S. Nelson, E.C. Sowa, and M.S. Daw, to be published at
Phys.Rev.Lett.
44. U. Backes and H. Ibach, Solid State Commun. 40,
575(1981).
45. B.N.J. Perrson and J.E. Demuth, Phys.Rev. B 30,
5968(1984).
46. J.A. Stroscio and W.Ho, Phys.Rev.Lett. 54, 1573(1985).
47. B.I. Boltaks and S.I.Budarina, Soc.Phys.Solid State 11,
330(1969).


124
Using e(u,q) as a Lindhard dielectric function for highly
doped semiconductors, then
w2 id2
e ( w2 u2 -0.6 q,|2 vf2
where e is the high-frequency limit of the dielectric
function, q,| is momentum transfer parallel to the surface, vF
is average Fermi velocity of free carriers with plasmon
frequency, and wp=(47rneffe2 /m)\ Here net, is the effective
carrier density and m is the hole effective mass, which is
0.23 times of the free electron mass. When an nel( of 6.0xl0'6
cm'3 is assumed, the predicted loss energy versus q^ [dashed
line in data set (A) of Fig. 151 matches reasonably well with
our experimental data. Another prediction of this model is
that the plasmon loss energy saturates at a cutoff wave vector
qc where electron-hole transitions broaden the elastic peak
and saturate the dispersion [59]. The cutoff wave vector
calculated for nef, =6.0xl0'6 cm'3 is qc=0.21'' which is shown as
a dotted line in Fig. 15. Finally, for q||=0, AE=hw(0)=5.3 meV
for the above parameters which is much smaller than the FWHM
of the elastic peak. Therefore some broadening of the elastic
peak would be expected. Assuming Gaussian line shapes for
this loss and for the elastic peak, an upper limit of ~8meV
for the unresolved peak due to the broadening from 9 to 12meV


206
c
o
4->
fD
X
ro
Q)
CT
0
200
Ge/Si
400 600 BOO
Film Thickness
1000
()
Figure 37. Surface lattice relaxation.


14
on the sample surface to well below 1012 cm'2, i.e. about 0.001
monolayer (ML). If pre-treatment such as oxidation, hydrogen
adsorption or film growth under UHV is required, it is usually
done in the same UHV chamber before data acquisition. It is
important in surface physics experiments to prepare the
required surface for analysis under UHV, since the transfer
procedure from one UHV chamber to another may possibly bring
contamination from the air. In some surface physics
experiments several UHV chambers are interconnected with UHV
transfer between each. However, in the University of Florida
high resolution electron energy loss spectroscopy (HREELS)
chamber several different sections of a single UHV chamber are
used for various surface preparation, characterization and
HREELS measurements. The advantage of pretreatment under UHV
is in the control of gas adsorption and/or deposition of thin
(0-10 ML) overlayers. Compared to an atmospheric pressure
environment or low vacuum condition, the number of residual
gas molecules in UHV is so small that the effective operation
time can be long enough not to contaminate the surface with
the residual gas during the measurement procedure (e.g. HREELS
or low energy electron diffraction (LEED)). Basic surface
analysis techniques using Auger electron spectroscopy (AES)
for detecting surface chemical composition and LEED for
determining the long range periodicity of surfaces follow
cleaning procedures.


30
a strong vacuum compatible epoxy can be used for cleaving the
surface. These procedures, so called cleaving, can be used
to prepare clean surfaces under UHV. Alkali halide (NaCl,
LiF, NaF, KC1 etc.) (100)faces, (111)faces of materials like
CaF2, Si, Fe, Zn and Be will cleave at liquid nitrogen
temperatures[12], but this technique is limited to the small
surfaces and a single orientation.
Bombardment of rare gas ions such as Ar* and Ne*
(~10/LiA/cm2 at 500eV-1000eV) can be used to clean a surface.
Impurity atoms receive enough energy from the incident gas
ion to be ejected from the surface. However ion bombardment
can cause damage leaving the surface in an amorphous-like
phase [13]. Some problems may arise from occlusion of the
inert gas atoms on preferential sputtering of one of the
components in a binary alloy or compound [14]. On an Si(lll)
substrate after Ar* sputtering, annealing at 1000C for 2 min
will produce the reconstructed surface with a 7x7 LEED
pattern. But it is hard to estimate how many defect sites are
left on the surface. On a GaAs(lOO) substrate after light Ne4
sputtering, annealing around 500-550C for 2min will produce
a reconstructed surface with LEED pattern that will vary
according to the stoichiometry of two components at surface
[15], It is also known that the proper choice of reactive gas
will help the cleaning of the surface, such as oxygen to
remove the carbon atoms in a suitable pretreatment and
temperature.


198
Table 1, Energy Loss for Hydrogen and Deuterium Titration
on Si
and Ge
surfaces
[110,113,
115] .
system
ref.
bending
mode
(meV)
scissor
mode
(meV)
stretching
mode comment
(meV)
H/Si(111)7x7
[113]
80(78)
110
260(258)
(after
Si-H2
desorbed)
D/Si(111)-7x7
[113]
54,64
(53,64)
81
190(189)
(after
Si-D2
desorbed)
H/Si(111)-2x1
[113]
80(79)
109
258(259)
(after
Si-H2
desorbed)
H/Ge(100)
[115]
66
105
245
ln2
temperature
D/Ge(100)
[115]
50.7
71.8
178
ln2
temperature
H/Ge 2Si ,
[110]
78 (Si-H)
258(Si-HJ
(100)2x1
70 (Ge-Hx)
247 (Ge-Hx)


22
23
24
25
26
27
28
29
30
31
32
140
141
142
158
162
165
169
171
172
181
192
HREELS spectra obtained from nominally clean Ni(110)
showing a strong elastic peak and a weak loss peak
due to residual impurities.
HREELS spectra obtained from thermally grown oxide
on Ni(110) under UHV.
HREELS spectra obtained from oxides on Ni(110)grown
under UHV at room temperature.
(a) 300L, 02; (b) 1.8xlOL, 02; (c) 300L, air.
HREELS spectra obtained from Shiraki oxide on
Si(111) (a) As introduced; (b) After annealing at
500C;(c) After annealing at 900C.
HREELS spectra obtained from water-preserved oxide
on Si(lll). (a) As introduced after preserving in
deionized water for three weeks; (b) After annealing
at 500C; (c) After annealing at 900C.
HREELS spectra obtained from native oxide on
Si(lll). (a) As introduced; (b) After annealing at
520C;(c) After annealing at 1010C.
HREELS spectra obtained from thermal oxide grown on
a 700 K Si(111) substrate, (a) 10L exposure; (b)
100L exposure; (c) 1000L exposure; (d) 10 kL
exposure.
HREELS spectra obtained from thermally grown oxide
after annealing at 700 K.
HREELS spectra obtained from thermally grown oxide
after annealing at 1100 K.
Asymmetric stretching mode variations of thermal
oxides grown at 700 K and 900 K.
HREELS spectra obtained from Ge0 5Si0 5(111) -5x5
showing a strong elastic peak (E=0) and a weak loss
peak (E105meV) due to residual carbon impurities.
ix


188
The results of LEED and STM confirms that dimer-adatom-
stacking fault (DAS) model can be also applied to the
SixGe,_(111) alloy film case [109]. The driving force in
triangular-dimer model of the Si(111)-7x7 reconstruction is
the lateral compression of the two outermost double-layers.
The 7x7 reconstruction of Si(111) surface, consisting of the
removal of 1/7 of the second- and third- layer atoms and
dissociation of the resulting dislocations, is the means to
relieve this compressive stress. In SixGe,_x(Ill) -5x5 case,
since Ge has 4% larger covalent radius than Si, this alloy
has the enhanced lateral compressive stress resulting in
removal of 1/5 of the second- and third layer atoms. Schaefer
reported hydrogen interaction with the GexSi,_x (100) alloy using
HREELS to detect the Ge surface concentration according to
annealing temperature [110]. Compared to the 20% Ge bulk
concentration, the Ge surface concentration was estimated at
up to 75% at high annealing temperature (temperature is not
reported.). This is quite different from GexSi,_ alloy films
since high temperature annealing causes the varying of the Ge
profile in the film. Note also that the (100) face has a
higher step density than the (111) face, which can be easily
ignored in an adsorption experiment. The step contains a
larger number of dangling bonds which act as active sites on
the surface. Farrel et al. reported water adsorption on the
MBE grown SixGe,.x(100) alloy films [111]. In this case, the
sample was sputtered and annealed to get rid of air


APPENDIX B
FURTHER THEORETICAL DETAILS
B.1. Dipole Scattering Cross Section
Since the electron trajectory re depends upon time, from
equation (3-16), the potential V also depends on time. The
probability (=P,) of the dipole with normal mode of frequency
u0 being found in the first excited state at t=oo if initially
(at t=-oo) it was at the ground state, is from the second
order time dependent perturbation theory,
(B-l)
Calculating using equation (3-17),
= r (d/aze) |i/re(t) | (b-2)
where r = <0| q (d/dq)/x 11> which is dynamic dipole moment.
Using the relation q = (2u0)'"( b* + b ),
r = (2u0)-*dii/dq (B-3)
P, =
exp(iu0t) dt
. ¡o
235


7
a semiconductor does not contact a metal, dangling-bond
surface states pin the Fermi-level at a position in the gap
usually different from the bulk. Due to this band bending
potential at the interface or bare surface, the local surface
concentration of carriers can be dramatically different from
that of the bulk. Using HREELS we observed such differences
in carrier concentration by monitoring the plasmon energy for
Si(lll) and GaAs(lOO) surfaces.
In the present studies HREELS is applied to well ordered
nickel surfaces since the formation of NiO at nickel and
nickel-chromium alloy surfaces is not yet well understood.
The simple selection rules make it easy to understand the
polarization of specific vibrational modes. The coalesced
nickel oxide layer is not a single continuous layer like a
silicon oxide layer. Recently, a new interpretation of
multiple oxide phases at Ni surfaces was suggested by a high
binding energy oxygen species of nickel oxide layers detected
with x-ray photoelectron spectroscopy (XPS). HREELS
measurements of O-Ni vibrational modes were correlated with
these XPS results.
In additional experiments, silicon oxidation in
conjunction with surface cleaning techniques is studied. The
formation mechanism of oxides at Si-SiOx interfaces; i.e., the
initial steps of Si oxidation are not well understood. The
knowledge of very thin oxides is closely related to cleaning
techniques of an Si substrate prior to UHV experiments. Thin


85
The basic idea in impact scattering theory is the
calculation of intensity of electrons which contribute to the
thermal diffuse background of a diffraction pattern. Since
dipole scattering is rapidly reduced at off-specular geometry,
the intensity of the background is mostly due to impact
scattering. When an electron encounters a solid, the
positions of nuclei are not fixed and are displaced by thermal
vibrations. The position of nucleus i1 is then R^R^+Ui where
Ro, is the position of the equilibrium site and u¡ is the
displacement from equilibrium position, R,,,. For small
displacements of u, the scattering amplitude f(ks,k, ;R) can be
expanded in powers of ua ,
df
f(ks,k,;R) = f (ks,k, ,Ro) + E ( )0ua + (3-38)
a 3R a
where ua is the ath Cartesian component. Expressing u0 in
terms of the normal mode eigenvectors §a!,
h
ua = 2 ( )* §as (as + as*) (3-39)
5 2SM,
where 's' refers to a particular normal mode and as,as* are the
annihilation and creation operators of vibrational quanta, and
Mi is the ionic mass. When a particular vibrational quantum
is emitted, the matrix element


222
To measure the lateral lattice constant, it is necessary
to scan the line which contains two sequential integer order
beams since the distance between two integer order beams in
the reciprocal surface is inversely proportional to the real
lateral lattice constant. Usually a electron beam energy of
35eV gives a relatively large distance between two integer
beams in our experimental set-up. For example among an
available 500-points in a line, the distance between (01) beam
and (10) beam is 292 points for 35eV while 235 points for 55eV
in the Si(111)-7x7 case.
A.5. Tuning of HREELS Spectrometer
To achieve the best resolution from experimental set-up,
before adjusting knobs which change the potential inside of
spectrometer, the sample should be cleaned properly and the
vacuum pressure should be low enough so as not to contaminate
the sample until the end of experiment. A geometrical
understanding of sample position relative to monochromator
exit or analyzer entrance and ideal specular direction of
scattering should be first achieved. There are variable
factors to determine before data acquisition such as sweep
energy range, sample geometry, primary energy of incident
electron, and sample bias. Also there are many potential
energies to optimize, tuning must begin from a simple geometry
and ideal setting of capacitor potentials. Most of the
potential knobs of the ELS-22 power supply, except analyzer


73
re = ( x0+vt, vjt| ),
(3-16)
where the origin is at the impact point and xQ is the impact
parameter in the surface plane. The term is the parallel
velocity and vx is the normal component of velocity.
Since the dominant electron-molecule interaction comes
from the part of electron trajectory where re is large,
interaction between the incident electron and adsorbed
molecule with dipole / can be written
V = S V (l/re) (dn/dq,) q, .
(3-17)
The dipole is varying slowly at frequency w0 compared to
typical electron motion in a metal and the fast metal
electrons will follow the instantaneous dipole motion
adiabatically. Thus, the parallel component will have an
opposite image dipole, while the vertical component will be
approximately double the strength as long as the dipole is
not imbedded into the metal. Only the normal component of
the dipole moment has a non-zero perturbation. This is the
consequence of the so-called normal dipole selection rule.
The detailed derivation of this differential cross section is
in Appendix B.l. For normal incidence vM=0 which gives nQ=w0
, and the differential cross section da is,


Energy (meV)
(b)
Figure 26.continued.
163


255
(a)
Figure 49. Sputter-AES of thin (-10) Ge/Si alloy film.
(a) Sputtered edge profile of thin pure Ge film;
(b) Sputtered edge profile of thin Ge/Si alloy film;


88
amplitude penetrates deeply into the crystal as Q,|-> 0. This
surface optical phonon mode from ionic crystals, e.g. ZnO,
GaAs etc., is called 'Fuchs-Kliewer' mode [36]. There are two
ways to derive this 'Fuchs-Kliewer' mode. In this section the
method using the result of dielectric function theory for
infinite crystals derived in the previous section. The other
is the method using lattice dynamical frame work which is
derived in Appendix B.3.
If the crystal has an ideally terminated-surface like
bulk, the dielectric function of a cubic material with one
infrared active bulk transverse optical phonon is
47rne'2 1
eb(w) = cbo + (3-42)
Mr wT02 u2 iwr(w)
where &o is the dielectric constant at high frequency limit,
n is the number of unit cells per unit volume, e' is
transverse effective charge, Mr is the reduced mass of the
unit cell, and r(w) is the damping function of the oscillator
[37]. The static dielectric constant e(0) is
e(0) = feo + (47rne'2/Mr) (wT02) (3-43)
From equation (3-37) the loss function has a peak at the pole
of the loss function, Im[-l/{l+eb(w) ) ]. The loss peak, which
is the Fuchs-Kliewer mode, is


213
can be studied by HREELS equipped with high incident energy
to measure dispersion throughout the entire first Brillouin
zone. (2) Interfaces between semiconductors, metals and
insulators can be probed by HREELS in combination with
photoemission. Especially the initial stage of the interface
formation which is still not well understood. (3) Oxidation
combined with hydrogen titration can detect oxidation steps
of metal, semiconductor and alloy surfaces. Hydrogen and
deuterium titration will solve the problem of insensitivity
of HREELS to non-polar surface to a certain degree. (4) New
materials containing oxygen such as surfaces of high Tc
superconductors are good candidates for future HREELS studies.


241
d 2
[V u0 (x) ] + wL02 [V-u,(x) ] = 0
(B-22)
at2
inside of the crystal. All possible solutions of equation
(B-22) except those with frequency of wl0 must have the
relation of Vuo=0. Then the solution of equation (B-19) can
be expressed as an excitation that propaqate parallel to the
x direction with wave vector Q||.
u0(x,!, z) = u0(z) exp(iQnX-iwt) ; for z>0
; for z<0
(B-23)
0
where u0(z) must satisfy equation (B-22). For this case
V'u0(x') = array of sources located on the crystal surface. If the
charge q is placed on the surface of a dielectric, the
potential inside of medium is not due to q but due to
effective charge 2&cq/ (feo+-l) at the position of original
position of charge by the screening effect [15]. Then
equation (B-19) becomes
(B-24)


TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS
LIST OF FIGURES V
GLOSSARY OF SURFACE PHYSICS TERMINOLOGY xii
ABSTRACT xiii
CHAPTERS
1 INTRODUCTION 1
1.1. Overview of Surface Physics Techniques 1
1.2. Overview of Specific Experimental Studies 6
2 EXPERIMENTAL PROCEDURE 13
2.1. Overview 13
2.2. Ultra High Vacuum 16
2.3. Auger Electron Spectroscopy 21
2.4. Sample Preparation 28
2.4.1. Overview 28
2.4.2. Specific Examples 31
2.5. Low Energy Electron Diffraction 34
2.6. High Resolution Electron Energy Loss
Spectroscopy 35
2.6.1. Basic Theory of 127 Capacitor
Electron Optics 36
2.6.2. Resolution and Sweeping Mode 4 0
2.6.3. Intensity Angular Profile 43
2.7. Computer Interface 45
2.8. Oxidation and Hydrogen Titration 48
2.9. Film Growth under UHV 52
iv


217
cathode shield are repelled by cathode shield. As soon as
electrons come out of shield, the anode attracts electrons
toward the crucible. Since this motion occurs in a
perpendicular magnetic field, the trajectory of electron will
be circular and its destination will be inside of crucible.
The electron gun power supply has a high output voltage fixed
at 4 kV and the emission current can be varied from 0 to
750mA. The voltage and current of gun filament are variable
between 0-6 volts and 0-25 A each. A well focused electron
beam can heat the source up to 3500C. Deposition rates are
different for each material (e.g. Nickel 2250 A/ min at full
power). Crucible liners made of graphite or tantalum plate
have been used. But a tantalum plate can not be used for
nickel source due to alloy formation. Vitreous Graphite is
usually non-reactive with source but it can be broken for fast
temperature variation. Crucible liner is essential to keep
the crucible clean and a safe method for preventing damage to
the crucible with electron beam in case the source is totally
evaporated.
Once the evaporation source is heated by electron beam,
the vapor diffuses due to the pressure difference. The vapor
spreads radially and can cover a large area of the chamber
which can include the aperture of the electron analyzer and
viewports if collimators are not installed. Two collimators
installed between the sample and evaporation source (distance
17 1/4") reduced the evaporated film diameter to 3/4" at the


15
Once a clean and smooth surface is prepared under UHV,
in our experiments the main analytical technique, HREELS, can
be performed. It must be considered that the total
measurement time includes tuning of the spectrometer as well
as data acquisition. The critical point in tuning a HREELS
system is to achieve the best resolution of the electron
optics. The procedure of tuning can be divided into two
parts, namely, detection of the optimum geometrical positions
of monochromator and sample manipulator, and adjustment of the
potentials of the electrodes and capacitors so as not to get
the largest intensity of the elastic peak but to get the
smallest width of the elastic peak. In angle resolved HREELS,
the incident angle of monochromator is changed to vary the
momentum transferred to the sample surface. It is necessary
to understand the kinematical geometry of the spectrometer for
determining the transferred momentum. Sometimes a diffracted
order of the electron beam is needed instead of the specular
(0,0) beam to suppress the high intensity of the elastic peak
in order not to saturate the detector which would result in
a nonlinear response. In this case understanding the
orientation of the sample surface is essential in order to
find the diffracted beam. The next step for tuning is to
optimize the various potential settings of the spectrometer.
This will be discussed in more detail later in this chapter.
Data acquisition using a computer is useful for analysis of
data by magnification, comparison of two or more spectra,


74
da = { rJ 4QV / [W02 +Q2vx2]2 } d2Q (3-18)
This function is strongly peaked near the characteristic wave
vector Qo~Uo/vx. In typical experiments, uo=0.02 a.u.=54 0 meV
and a primary beam energy E=0.20 a.u.=5.4 eV, then Qo0.01
a.u.. Since only small values of wave vector Q contribute to
the scattering intensity, it can be deduced that the effective
range of electron-molecule scattering is of order Q0"1 or at
least 60 a.u. in the example. When l/re is expressed as a
Fourier transform,
1
re
1
2n2
d2 Q
dq
exp(-iQx) exp(-iqz)
Q2 + q2
1
2n
d2 Q e'iQ
(3-19)
thus, the potential V has e"Q^ factor, which also indicates
that perturbation extends above the specimen up to z~Q0'.
This is the underlying justification for using the long range
dipole scattering interaction. One of the important reasons
this semi-classical differential cross-section is chosen, is
that the differential cross-section da is easily transformed
to measurable quantities, e.g. k (incident wave vector) and
measurable angles.


LEYBOLD-HERAEUS SYSTEM (ELS-22)
7 HREELS
/
5 CRUCIBLE
E-BEAM
EVAPORATOR
SAMPLE MANIPULATOR
x.Y.z.e
GAS INLET
MANIFOLD
MBE
KNUDSON
CELLS
TITANIUM
SUBLIMATION
and
ION PUMPS
TURBO PUMP
-10
PRESSURE: 1x10 torr
Figure 1. HREELS chamber.


267

00
R.G. Smeenk, R.M. Tromp, J.F. Van Der Veen, and F.W.
Saris, Surface Sci. 95, 156(1980).

in
00
C.N. Dykstal, M.S. Thesis, University of Florida (1988).

CO
H. Ibach, Phys. Rev. Lett., 24, 1416(1970).

r~
00
P. A. Thiry, M. Liehr, J.J. Pireaux and R. Caudano,
J.Elec.Spec.Rela. Phen. 39, 69(1986).

00
00
H. Ibach, H.D. Bruchmann and H. Wagner, Appl.Phys. A29,
113(1982).
89.
H. Ibach, H. Wagner and D. Bruchmann, Solid State Commun.
42, 457(1982).
90.
U. Backes, H.Ibach, Solid State Commun. 40, 575(1981).
91.
D. Grf, M. Grundner and R. Schulz, J.Vac.Sci.Technol.
A7(3), 808(1989).
92.
J.C. Bean, Physics Today Oct., 36 (1986).
93.
R. People, IEEE J.Quan.Elec. QE-22, 1696 (1986).
94.
E. Kasper and H.-J. Herzog, Thin Solid Films 44,
357(1977).
95.
T. Narusawa and Gibson, Phys.Rev.Lett. 47, 1459 (1981).
96.
A.I. Toropov, L.V. Sokolov, O.P. Pchelyakov and S.I.
Stenin, Sov.Phys.Crystallogr. 27, 450 (1982).
97.
T. Narusawa and W.M. Gibson, J.Vac.Sci.Technol. 20,
709(1982).

CO
cr\
J.C. Bean, T.T. Sheng, L.C. Feldman, A.T. Fiory and R.T.
Lynch, Appl.Phys.Lett. 44, 102 (1984).
99.
K. Shoji, M. Hyodo, H. Ueba and C. Tatsuyama,
Jpn.J.Appl.Phys. 22, L200 (1983).
100.
T. Ichikawa and S. Ino, Surf.Sci. 136, 267 (1984).
101.
J.C. Bean, L.C. Feldman, A.T. Fiory, S. Nakahara and I.K.
Robinson, J.Vac.Sci.Technol. A2, 436 (1984).
102.
H.-J. Gossmann, L.C. Feldman, and W.M. Gibson, Surf.Sci.
155, 413 (1985).
103.
J.M. Seo, D.L. Doering, D.S. Black and J.E. Rowe,
J.Vac.Sci.Technol. A4, 894 (1986).


221
LEED Opt i es
Samp 1e
e-gun
Vidicon
Camera
D/A
D/A
Int erf ace
i
_jj t'
Computer CPU
4t
Disk Stor
age Unit
OMA
Termina 1
Figure 40. Schematic diagram of digital LEED.


266
66. L.H. Dubois, G.P. Schwartz, R.E. Camley, and D.L. Mills,
Phys. Rev. B 29, 3208(1984).
67. F.P. Fehlner and N.F. Mott, Oxidation of Metal 2,
52(1970).
68. R.N.Bloomer, Brit. J. Appl. Phys. 8, 321(1957).
69. W.H. Orr, Thesis, Cornell University, Ithaca, N.Y.(1962).
70. K.S. Kim and N. Winograd, Surface Sci. 43, 625(1974).
71. L.M. Moroney, R.St.C. Smart, and M.W. Roberts, J. Chem.
Soc. Faraday Trans. I, 79, 1769(1983).
72. C. Benndorf, C. Nobl, and T. Madey, Surface Sci. 138
292(1984).
73. H. Ibach and D. Bruchmann, Phys.Rev.Lett. 44, 36(1980).
74. R.J. Birgeneau, J. Cordes, G. Polling and A.D.B. Woods,
Phys. Rev. 136, A1359(1964).
75. G. Dalmei-Imelik, J.C. Bertolini, and J. Rousseau,
Surface Sci. 63, 67(1977).
76. S. Andersson and J.W. Davenport, Solid State Commun. 28,
677(1978).
77. P.A. Cox and A.A. Williams, Surface Sci. 152/153,
791(1985).
78. W. Reichardt, V. Wagner, W. Kress, J. Phys. C8,
3955(1975).
79. M. Hock, U. Seip, I. Bassingnana, K. Wagemann and J.
Kuppers, Surface Sci. 177, L978(1986).
80. It was reported that randomly adsorbed oxygen on Ni(100)
also induced the another loss. R. Franchy, M. Wuttig and
H. Ibach at the National Symposium of AVS in Atlanta,
Georgia, October 6, 1988.
81. M. Nishijima, K. Edamoto, Y. Kubota, H. Kobayashi and M.
Onchi, Surface Sci. 158, 422(1985).
82. J.A. Schaefer, F. Stucki, D.J. Frankel, W.Gpel and G.J.
Lapeyre,J.vac.Sci.Technol. B2(3), 359(1984).
83. M.P. Seah and W.A.Dench, Surf.Interface Anal.1,1(1979).


179
this sample after further exposure oxygen (100L) is shown in
Fig. 28(b). The superoxide-like mode at 160meV disappears
while Si-O-Si mode energies evolve. The asymmetric stretch
mode at 130meV correspond to thickness of 0.5ML. At this
exposure most of oxygen atoms are still in bridging sites
between the first layer and the second layer. A small number
of oxygen atoms start to break the bond between the second and
the third layer and the symmetric stretch mode and bending
mode have more than one value. After 1000L oxygen exposure
the HREELS spectrum obtained from the surface is shown in Fia.
28(c). Asymmetric stretch mode at 135meV corresponds to the
thickness of 0.9ML. Since the top two layers are tilted, the
thickness of 0.9ML indicates that many oxygen atoms occupy the
bridging bond between the second layer and the third layer
which is vertical to the surface. This the reason why the
intensity of asymmetric stretch mode increases rapidly due to
dipole selection rule for the vertical mode polarization.
After lOkL exposure the HREELS spectrum obtained from the
surface is shown in Fig. 28(d). The thickness of the oxide
increases to 1.5ML and the direction of the Si-O-Si structures
seems to be random now since the intensity ratio of bending
modes (50meV) to asymmetric stretch modes (138meV) starts to
increase. From the previously reported HREELS spectrum
obtained from the thick Si02-like oxide (950) the intensities
of these two mode are similar [87].


26
photoelectron spectroscopy, in which the i-th shell is
photoionized in the process (2-1) [5]. The binding energy
relative to the Fermi level of the electron in level 'i' is
Eb(i) = hu> (Ek + 0S ) (2-3)
where Ek is the kinetic energy of photoelectron emitted from
the sample with work function is not the sum Eb(j) + Eb(k) of separate binding energies which
can be measured like the initial state, since interactions
between the two holes can occur and relaxation events to fill
core holes are nonlinear. Considering ccc Auger process
(i.e., the initial state has a core hole and final states have
two core holes), the derivation of the final states A**(j,k)
can be decomposed into two processes.
A- A*
(j) + e* ,
(2-4)
A*(j)-
A**(j,k) + e- .
(2-5)
In the process (2-4), Eb(j) is required, where Eb(j) is the
energy required for the threshold process which places the
ionized electron at the Fermi level. In the process (2-5),
Eb'(k) is also required for the threshold process. Using the
equations (2-2), (2-4) and (2-5),


4
properties for the surfaces of solids. Inelastic neutron
scattering is used for determination of the energy-wavevector
dispersion relation of bulk single crystals, but the neutron
phonon scattering cross section is very small, so this
technique is not sensitive enough for surface studies of small
area single crystal samples (however, powder samples can be
studied). Surface IR spectroscopy is a technique for studying
the molecular structure of adsorbate atom complexes and can
be used on metal, semiconductor and insulator surfaces. The
electric field of the incident IR beam can couple to the
phonon modes of a solid and the absorption is usually shown
as a function of the incident photon energy. The high
resolution (0.1 meV) of IR spectroscopy is an advantage of
this technique but the small absorption cross section for
infrared photons limits the sensitivity of this method.
Unless special reflection geometries are used, most of the
detected IR signal is due to the bulk optical properties and
provides an unwanted background signal that degrades the
sensitivity. Raman spectroscopy and inelastic electron
tunneling spectroscopy also have a substrate background signal
problem which limits their sensitivity for surface studies.
This problem is partially solved by the improvement of surface
enhancement of these last two spectroscopies. Neutral atom
scattering uses inelastic scattering of He or Ne atom beams
of low-energy typically 20-60 meV. The resolution of atom
scattering is very good (-0.5 meV), but it is experimentally


211
starts to cover NiO layer with Ni(OH)2 layer. Further
exposure of oxygen induces thickening of this Ni(OH)2 layer.
The surface morphology, the growth mechanism and
vibrational modes of the Ge0 5Si0 5(111)-5x5 film were observed
by AES, LEED, SEM and HREELS. These alloy films (thickness>10
A) were grown on a hot (560-590C) substrate and they followed
the Stranski-Krastanov growth mode. It has been shown that
the origin of the 5x5 LEED pattern of the alloy film is not
Ge-rich islands, and these Ge-islands which are distributed
evenly at the surface relieve the strain of Ge0 5Si0 5 alloy
layer. This is the reason why the clean 5x5 LEED pattern and
Ge-rich islands appear at almost the same time. For further
Ge-evaporation, the thickness of the alloy layer does not
increase. For slow deposition of Ge on hot (-580 C) Si(111)
substrate the intermixed GexSi,.x alloy film (x~0.5) grows
pseudo-morphically up to 150 A. For more evaporation, a
gradual increase in the lattice constant was observed. At the
'effective' alloy film thickness of -1000 A, two overlapped
LEED patterns were observed. One is the 5x5 LEED pattern due
to a shallow alloy layer and the other is the unrelaxed
Ge(111)7x7 pattern due to thick Ge-rich islands.
The adatom vibration mode was also observed from the
Ge0 5Si0 5 (111)-5x5 surface at an energy similar to the Si (111)-
7x7 surface. Quasi-elastic peak broadening from the alloy
surface is observed, which means that the alloy surface also
has metallic surface states due to adatom dangling bonds


253
3Osee sputtering with raster of 3mmx3mm, Ge reduced to noise
level from area 1 shown in Fig. 48 (c) : while Ge level increase
in area 2 (shown in Fig. 48(d): Si(1617)/Ge(1148) = 0.5). In
area 1, relatively thin Ge0 5Si0 5 alloy film is sputtered off
and Si-bulk is exposed. In area 2, as the air contamination
disappears the Ge level is enhanced. The fine film quality
(i.e. no C or O) is detected from the AES profile of the
sputtered area. For further sputtering (up to five minutes
more) two areas do not show any change. The helium damaged
area (due to RBS measurement) a very similar morphology. The
same analysis procedures as for the undamaged area were
followed. More impurities (Na(990eV), C and O) were detected
from the unsputtered area than the undamaged area. These are
surface impurities induced by the helium beam. For the
sputtered surface, no impurities were detected and the
intensity profile was the same as the undamaged area. From
SEM photo, three bands were identified, which are unsputtered
area, alloy area and bulk area.
The growth mechanism of thick (-1000 A) Si^Ge,.,, alloy film
can be summarized as follows: On a room temperature substrate,
Ge can grow epitaxially as an amorphous layer. Subsequent
annealing induces the intermixing at the interface and the
film is crystallized, producing 5x5 LEED pattern and Ge
diffuses totally into the bulk for higher temperature(>650C)
annealing if the evaporated Ge is less than 3 ML. In case of
deposition on a hot (~560C) substrate, the evaporated Ge


81
where x = x,,+zz, and the brackets with subscript T denotes the
average of the quantity enclosed over the appropriate
statistical ensemble at the temperature T. Defining the
scattering probability P(k,k') by,
1 da
S(Q|,,w)
[VxJ Q,i2 +(w-Qli'V|l)2 ]2
(3-35)
Thus P(k,k' )dflk dhu is the probability an electron scattered
into the solid angle di\ in the energy range between hw and
h(u+diJ), normalized to the elastic intensity (here, R^R^-R was
already assumed). The scattering probability shown in
equation (3-35) has a kinematic factor which is equation (3-
35) itself. This kinematic factor is independent of the
property of specimen and peaks for small momentum transfers
at Qn~k(hw/2E) Since QN* is the range of the potential as
well as the probing depth, shown in equation (3-31) and its
following discussions, the kinematic factor derived in
equation (3-35) has shown a sharp forward scattering lobe by
the potential with angular width of 2E0/(hw) and the effective
probing depth of HREELS also turned out 2E0/(khw) from
equation (3-34).
|R|2 dnk d(hu)
2 Vj4 k'
h7T (ea0)2 cos, k


10
operation are also included. Especially, the electron optics
of HREELS is described in some detail in order to allow one
to understand the complex tuning problem of this type of
instrument. In addition to the main experimental chapter,
further experimental details are presented in Appendix A.
Chapter 3 describes the theoretical background of
electron scattering as applied to HREELS. Electron scattering
theory is divided into an elastic scattering part, which
consists of bulk diffraction and surface diffraction, and an
inelastic electron scattering part which includes dipole and
impact scattering HREELS theory. For a clear understanding
of dipole scattering theory, a semi-classical approach for
adsorbed molecules as an example is comprehensively reviewed.
For more general applications of HREELS on semiconductors a
two-layer dielectric function theory is reviewed. For the
application of angle-resolved HREELS, impact scattering theory
is also reviewed. In addition to a review of fundamental
HREELS scattering theories, applications to specific systems
are briefly mentioned. Detailed theoretical background
information is attached in Appendix B.
Chapter 4 is one of the four experimental measurement
chapters. In this chapter, investigations of pure Si(111)
and GaAs(lOO) surfaces are discussed. In the first section,
sputtered annealed Si(111) surfaces are discussed. The quasi
elastic peak broadening effect and p-type surface accumulation
layer formation due to a sputter-cleaning procedure is


204
the (10) beam and the (01) beam from the LEED screen were
shown in Fia. 36. For three different film thickness (e.g.
d=0, 400 A, 1000 A) intensity profiles show different peaks.
For d=0, Si(111)-7x7 peaks are clearly shown and for d=400 A
5x5 peaks are shown. But for d=1000 A, 7x7 and 5x5 admixtures
are shown, since two center peaks are broad (i.e. 2/5+3/7 and
3/5+4/7). This peak intensity ratio resembles a 5x5 pattern
but the peak positions favor a 7x7 pattern. The lattice
relaxation as a function of alloy film thickness is shown in
Fia. 37. The relaxation calculation from the channel distance
measurement is { (d/dR)-1 }xl00%, where d is the channel
distance between clean Si(111)-7x7 integer order beams and dR
is the channel distance between Ge0 5Si0 5(111)-5x5 film's
integer order beams. Since the relaxation is measured for
five different energies (i.e. 35eV-55eV), the standard
deviation of relaxation ratio is also shown. The lattice
constant of film matches that of Si substrate for thickness
up to 150 A. At greater thicknesses, however, a gradual
increase in the lattice constant is observed. After a film
of 1000 A was grown, the lattice constant was about 1.5%
greater than that of Si. This is consistent with the 2%
difference expected between the bulk alloy and the bulk Si
dimensions and a 4% difference is expected between Si and Ge
bulk. Interestingly, it was observed that for a thickness of
600 A there was evidence of two overlapping LEED patterns of
5x5 pattern and 7x7 pattern as we have seen in Fig. 36. But


86
M(k,, Jcs;+s) =
= (ns+l)*(h/2Nws)>'(af/aQs) (3-40)
where (3f/3Qs) = S(3f/aRa)0 §a/MlH and ns=[exp(hws/kT) 1). Then
the probability that the vibrational quanta (Q(|Qt) scatters the
electron into the solid angle dil from the surface area 'A1 is
d£a(k,,ks) mE, cos20s
= A |M(k(,k,;Qlla)|* (3-41)
dil 2?r2 h2 cos0,
where a contains all indices other than wavevector QN, and 6,
and 8S are an incident angle and a scattered angle. A further
analysis of multiple scattering will not be covered here.
Selection rules for impact scattering provide a basis
for obtaining direct information about symmetry of an adsorbed
atom based on the general feature of the inelastic single
scattering cross section. Selection rules are based on the
symmetry of the substrate, time-reversal symmetry, scattering
geometry and the direction of the polarization of the
vibrational mode. The results by Tong et al. were as follows
[34], If a normal mode is polarized out of the scattering
plane as well as parallel to the surface, and the substrate
has a reflection symmetry relative to the plane perpendicular
to the scattering plane, the cross-section in any position in
the scattering plane is zero. If the substrate has a
rotational symmetry about z axis in the above case, the
differential cross-section is zero at only specular geometry.


180
Instead of further exposure to oxygen post-annealing at
the exposure temperature (700 K) for 10 min increased the
ratio of the intensity of 140meV to 50meV. Post-thermal
annealing appears to cause the inhomogeneous oxide layer to
become more uniform with sharper HREELS peaks (Fig_i__29) .
Post-annealing at higher temperature (1000 K) increases this
tendency and decreases the loss energy at 140meV by ~2meV
which means a small amount of oxygen starts to desorb.
Annealing at 1100 K induces the oxygen species to desorb
completely from the surface. In Fig. 30 the 113meV mode is
distinct and is due to silicon carbide which is produced by
hydrocarbon contamination dissolving into the sub-surface
after thermal annealing.
Compared to the substrate annealed at 700 K, the
substrate held at 900 K has more rapid oxidation speed.
Annealing at 900 K has the same effect as larger amount of
exposure at lower temperature (e.g. 700 K) since more oxygen
molecules which arrive at the surface can be dissociated into
oxygen atoms compared to the lower temperature substrate.
Except for the oxidation rate the evolution of three oxygen-
related peaks are almost same at the two temperatures.
Asymmetric stretching mode energies of two thermally grown
oxides (700 K and 900 K) versus to oxygen exposure are shown
in Fig. 31. Initially the oxidation rate of the 900 K sample
is about ten times faster than that of the 700 K sample.


232
Enter your initials please:ims:
Enter todays dataset :51:
Enter beginning run number:701:
(a)
Choose run parameters
Elastic peak tuner
Trial run (no save)
Run
View data runs
Save data
Quit
(b)
High Resolution EELS
Baseline is at -7038.2000 meV
Kepco voltage scale IV or 10V
1
Start Energy (meV)
0.00
Energy range (meV)
249.82
Final Energy (meV)
249.82
Maximum number of points
1023
Time per point
2.00
Number of averaged data scans
3
Count rate
3.000E+004
Enter number of angles to step through :
10
Enter starting angle for monochromator :
0.00
Enter final angle for monochromator
(referenced from current zero
position) :
9
Sweep will be from -7.04 to
242.78 by
0.24
Total time per run (min) :
35.8
Total time per set (min) ;
107.4
Total time for all angles (hrs]
Everything OK Y/N ?
:
17.90
(c)
Figure 43. Screens of "LHMAIN" program.
(a) Filenames;
(b) Selection of commands;
(c) Setting of parameters.


265
48. L.N. Safronov, L.N. Aleksandrov, and R.N. Lovyagin,
Phys.Status Solidi B107, 461(1981).
49. R.T. Tung, K.K. Ng, J.M. Gibson, and A.F.J. Levi,
Phys.Rev. B 33, 7077(1986).
50. W. Daum, H. Ibach and J.E. Muller, Phys. Rev. Lett. 60,
2527(1987).
51. R. Matz and H. Luth, Phys. Rev. Lett. 46, 500(1981).
52. A. Ritz and H. Luth, Phys. Rev. Lett. 52, 1242(1984),
and A. Ritz and H. Luth, Surf. Sci. 168, 773(1986).
53. E.J. Gray-Grychowski, R.G.Egdell, B.A. Joyce, R.A.
Stradling and K. Woodbridge, Surf. Sci. 186, 482(1987).
54. L.H. Dubois, B.R. Zegarski, and B.N.J. Persson, Phys.Rev.
B 35, 9128(1987).
55. W. Ho, R.F. Willis, and E.W. Plummer, Phys.Rev.Lett. 40,
1463(1978).
56. H. Ibach and T.S. Rahman, in Chemistry and Physics Solid
Surface V, edited by R. Vanselow and R. Howe, Springer,
NewYork, 1984, p.456.
57. D.H. Ehlers and D.L. Mills, Phys.Rev. B36, 1051(1987).
58. U. Harten, J.P. Toennies, and Ch. Woll, Phys.Rev.Lett.
57, 2947(1986).
59. H. Raether, in "Excitation of Plasmons and Interband
Transitions by Electrons,"Springer, New York, 1980,p.9.
60. R.S. Becker has performed scanning tunneling microscopy
on Si(111) surfaces with surface defects due to
incomplete 7x7 ordering which show missing adatoms
(unpublished results).
61. J.E. Black, Surf. Sci. 100, 555(1980).
62. M. Grundner and H. Jacob, Appl.Phys. A39, 73(1986) and
references there in (22).
63. J.R. Arthur, J.Appl.Phys. 38, 4023(1967).
64. A.D.Buonaquisti, Y.-X.Wang, and P.H. Holloway, J.
Vac.Sci. Technol. Al(2),776(1983).
65. W.E. Spicer, P.W. Chye, P.R. Skeath, C.Y. Su, and I.
Lindau, J. Vac. Sci. Technol. 16, 1422(1979).


268
104.
K. Sakamoto, T. Sakamoto, S. Nagao, G. Hashiguchi,
K.Kuniyoshi and Y. Bando, Jpn.J.Appl.Phys. 26, 666
(1987).
105.
P.M.J. Maree, K. Nakagawa, F.M. Mulders, J.F. Van Der
Veen and K. L. Kavanagh, Surf.Sci. 191, 305 (1987).
106.
P. Chen, D. Bolmont and C. Sebenne, Solid State Commun.
44, 1191 (1982).
107.
E.G. McRae, H.-J. Gossmann and L.C. Feldman, Surf.Sci.
146, L540 (1984).
108.
R.S. Becker, B.S. Swartzentruber and J.S. Vickers,
J.Vac.Sci.Technol. A6, 472 (1988).
109.
K. Takayanagi, Y. Tanishiro, M. Takahashi,
J.Vac.Sci.Technol. A3, 1502 (1985).
110.
J.A. Schaefer, J.Q. Broughton, J.C. Bean and H.H.
Farrell, Phys.Rev. B33, 2999 (1986); J.A. Schaefer,
Surf.Sci. 189/190, 127(1987), and J.A. Schaefer,
Surf.Sci. 178, 90 (1986).
111.
H.H. Farrell, J.A. Schaefer, J.Q. Broughton and J.C.
Bean, in "The Structure of Surfaces," M.A. Van Hove and
S.Y. Tong eds., Springer Verlag, New York, 1985, p.l.
112.
G. Abstreiter, H. Brugger, T. Wolf, H. Jorke and H.J.
Herzog, Surf.Sci. 174, 640 (1986).
113.
H. Froitzheim, U. Kohler and H. Lammering, Surf.Sci. 149,
537 (1985).
114.
R. Butz, E.M. Oellig, H. Ibach and H. Wagner, Surf.Sci.
147, 537 (1984).
115.
L. Papagno, X.Y. Shen, J. Anderson, G. Spagnolo and G.J.
Lapeyre, Phys,Rev. B34, 7188 (1986), and L. Papagno, L.S.
Caputi, D. Frankel, Y. Chen and G.J. Lapeyre, Surf.Sci.
189/190, 199 (1987).
116.
H.J. Gossmann, J.C. Bean, L.C. Feldman, E.G. McRae and
I.K. Robinson, Phys.Rev.Lett. 55, 1106 (1985).
117.
L.E. Davis, N.C. MacDonald, "Handbook of Auger Electron
Spectroscopy," Physical Electronics Industries, Inc.,
Eden Prairie, Minnesota.
118.
K. Heinz and K. Muller, in Springer Tracts in Modern
Physics, 91, 1 (1982).


178
layer is totally desorbed. The oxygen desorbing temperature
(1010C) of native oxide is higher than that of the water
preserved oxide (900C) .
In thermal oxidation at 700 K passivating species such
as OH and H can not restrain oxidation procedure since
hydrogen desorbs from the surface below 650 K [89]. Therefore
passivation by the water dissociated species is not a problem
in thermal oxidation. The substrate temperature 700 K is not
high enough to desorb the oxygen species. The HREELS spectra
obtained from thermal oxides are shown in Fig. 28. From the
HREELS spectrum obtained from a 10L oxygen exposed sample,
Fig. 28(a) we find the asymmetric stretch mode of Si-O-Si
(124meV), symmetric stretch mode of Si-O-Si (91meV), bending
mode of Si-O-Si (47meV), a small peak at 164meV and a broad
feature around lOOmeV. The small peak at 164meV may be due
a peroxy 0-0 mode (155meV) in a superoxide-like species. This
peak is always detected from the sample exposed to a small
amount of oxygen. If the first and second layer of Si is
replaced by Si-O-Si bridging bonds the electronegativity of
oxygen in Si-O-Si will deplete the electron in dangling bond
at the top as well as the bonding electron between second
layer of Si surface. The asymmetric stretch mode has a small
perpendicular component compared to the other two modes, i.e.
the symmetric stretch mode and the bending mode since Si-Si
direction of first second layer is already tilted from the
vertical to the surface. The HREELS spectrum obtained from


145
5.3. Discussion
For Ni(lll), as the oxygen induced a p(2x2) LEED pattern,
it is known as 1/4 monolayer coverage of the surface. Upton
and Goddard had carried out calculation on the system where
oxygen atom adsorbed on the Ni(lll) surface [40]. As the
oxygen atom approaches the surface along the line
perpendicular to it, the minimum potential energy site is
found at the hollow site of three fold symmetry and at a
distance Rx =1.20 above the plane which contain the nickel
nuclei. If Rx is determined, then using the potential of
Upton and Gaddard and the frequency calculation of the hollow
site at the three fold in the reference [39], Calculation
results are 70.4meV for vertical mode and 62.6meV for parallel
mode. In our case 72meV (Fia. 20) from p(2x2) pattern is
matched with vertical mode. Ibach et al. also has done the
same experiment which gave 71.6meV in their case [73]. The
p(2x2) pattern indicate surface Brillouin zone is not same as
the bulk Brillouin zone since the oxygen species induced twice
larger periodicity on the surface. For the specular direction
we only see the r point in the surface Brillouin zone of (111)
surface which is folded by 1/2 due to the change of the
surface periodicity. The zone edge loss can be the loss in
this case. From Ni bulk normal modes of vibration, the 26meV
mode can be found at K point (right now it is r point because
of zone folding) [74]. This mode is a longitudinal surface
phonon mode. From angle resolved HREELS spectra in Fig. 20.


145
140
135
130
125
120
1 2 3 4 5 6
og Oxygen Exposure (in Langmuir)
31. Asymmetric stretching mode variations of thermal
oxides grown at 700 K and 900 K.
181


>
-4-J
tn
c
QJ
C
Energy (meV)
(a)
Figure 34. Hydrogen titration on Ge0 ,Si0 5(111) -5x5.
(a) 2.5 L, H; (b) 5L, H.
199


Intensity
136
-50
0 50 100 150 200
Energy (meV)
Figure 20. HREELS spectra for specular scattering obtained
from a Ni(111)-p(2x2)-0 surface with 0.25 monolayer
of chemisorbed oxygen.


Ill
Si ( 111) nb = 1.0xl015/cm 3
Energy (meV)
Figure 13. ARHREELS spectra obtained from partially ordered
Si(111)lxl for non-specular (L8?0) scattering.


29
need atomically smooth surfaces. Conventional cleaning
techniques, namely rare gas sputtering and high temperature
annealing, work well for most substrates, but these procedures
induce defects on the surface which cause unexpected
impurities to migrate to the surface and vary the desired
doping profile. To minimize the problem related to cleaning
under UHV, it is recommended that the specimen be pretreated
before introducing into UHV. Another reason for preparation
of clean surfaces without sputtering and annealing is that the
band bending at the semi-conductor surface, due to introducing
surface states, results in depletion of the carrier at the
surface. Ohmic contacts with metal require a thin barrier or
a large number of carries at the interface.
In this section the normal cleaning procedures, namely
sputtering and annealing, cleaving and chemical pretreatment
on the semiconductor surfaces will be discussed. Especially
for Si(111) substrates, a thin chemical oxide treatment the
so called Shiraki method[10] and for the MBE grown GaAs(lOO),
the As capping technique will be examined[ll]. Besides the
cleaning technique, the temperature measurement and heating
will be discussed.
Some materials (e.g. Si, GaAs etc.) cleave easily to a
direction of a favorable surface. In highly oriented
pyrolitic graphite (HOPG), adhesive tape can be used to cleave
the surface in the air since the graphite surfaces are non
reactive with air contamination. In III-V compound material,


144
Energy (meV)
( c)
Figure 24.continued.


134
help to understand the kinetics of Ni-oxidation at room
temperature.
In this chapter the following points are considered:
(i) Surface phonon modes from single crystal nickel oxide
films grown thermally (300C) on Ni(110) surface. Are there
any other modes in addition to surface optical phonon modes
(Fuchs-Kliewer (FK) modes)?
(ii) Surface phonon modes from coalesced nickel oxide and
nickel hydroxide layers grown at room temperature for
different exposures of oxygen (300L and l.SxlOL). Are these
the same or different and can one correlate high resolution
electron energy loss spectroscopy (HREELS) data and XPS data?
(iii) Surface phonon modes from the clean Ni(110) sample
exposed to 300L of air at room temperature. Is this oxide
layer the same as that produced by ultrahigh vacuum (UHV)
exposure of pure oxygen?
5.2. Experimental Results
A degreased Ni(lll) single crystal sample was introduced
to the chamber. The AES spectra before cleaning indicated
impurities such as 0, S, Cl, K and C. Three cycles of
cleaning by sputtering on hot (~400C) sample and post
annealing at 750C for 3 min result in a clean surface with
lxl LEED pattern. No specific impurity is detected from AES


17 1/4
216
Figure 39. Design of evaporation system.(a) Collimators;
(b) Cross sectional view;(c) Side view.


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
[o|in E. Rowe, Chairman
>fessor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
£. OA. Pj^Po
Lucy E. aeiberling /
Associate Professor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Assistant Professor of Physics
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
David A. Micha
Professor of Physics


IEEE488 bus
L_
1
tmrzi
/ M% m rn ginnTt>
IBM PC/XT
D/A
KEPCO
controller
t iining
generat or
count er
CAMAC
HREELS
int erna 1
s upp1y
V.
1 K
(0)
ramp
to analyzer
from detector
ORTEC
Figuer 5. Block diagram of data acquisition system.


168
After annealing the same Si(111) surface at 1000C
oxygen molecules are exposed to the surface held at 700 K.
The HREELS spectra obtained from 10L-10kL exposed samples are
shown in Fig. 28. For three oxygen-related modes (-55, -95
and ~140meV), vibrational mode energy shifts as a function of
oxygen exposure are observed. In order to determine the
relative stability of this intermediate oxide layer, the 10
kL-exposed sample is annealed at 700 K for 10 minutes, then
annealed at 1000 K for three minutes and finally annealed at
1100 K for three minutes. The HREELS obtained from 700 K and
1100 K annealed samples are shown in Fig. 29 and Fig. 30. The
LEED pattern of this final sample was /19x/l9 indicating the
presence of some impurities such as Ni in the near surface
region. The same experimental procedures on Si(111) held at
900 K result in the similar evolution of three oxygen-related
peaks except the oxidation rate.
6.3. Discussion
Since the thickness of a Shiraki oxide layer is very thin
the LEED pattern shows a lxl periodicity with three fold
symmetry at an incident energy higher than 180 eV. This long
range ordered structure is not due to the Shiraki oxide which
is amorphous but due to a smooth silicon subsurface. Only
electrons with high energies can penetrate the thin oxide and
detect the ordered subsurface structure. From the HREELS
spectrum of the sample shown in Fig. 25(a) three oxygen-


25
it is very sensitive to most of elements and has good lateral
resolution. AES also gives qualitative information on the
chemical reaction between two elements if they are chemically
reactive. Finally AES has the big potential for giving
quantitative information of the surface composition.
An Auger transition converts an atom with a hole in one
core level (i.e. i-th level) to a final state that consists
of the atom with holes in other orbitals (i.e. *j1 and 'k')
and the fast electron whose kinetic energy is Ek. In the
initial process before Auger emission, a core hole is produced
by electron impact ionization, i.e. element A,
A- A4(i) + e" (2-1)
where A4(i) means the hole in 'A' is in level 'i'. The Auger
process is
A* (i)* A44 (j k) + e'auger (2-2)
where the emitted electron e'auger has kinetic energy Ek from
the vacuum level and A44(j,k) has left the atom doubly ionized
with holes in 'j' and k' levels.
To understand Auger electron energy which can be measured
by the analyzer, energies of the initial and final states of
the Auger process should be known. The energy of the initial
state of the Auger process (2-2) can be measured by x-ray


52
semiconductors such as Si, GexSi,.x alloy and Ge, hydrogen was
adsorbed by placing the sample in front of a hot tungsten
filament. The effective exposure of atomic hydrogen can be
calculated by assuming that only 10'J of the H2 would be
dissociated into H-atoms. Thus the effective exposure of
hydrogen atoms is 13.5 L for 15 min. at 1.5 xlO'5 torr of H2.
The exposure procedure is the same as the oxygen exposure case
and the deuterium exposure is exactly same as hydrogen
exposure. For hydrogen exposure residual gas contamination
is expected and is often found due to reaction with the
chamber walls.
2.9. Film Growth Under UHV
Since HREELS is a surface sensitive technique especially
sensitive to impurities such as oxygen and carbon on an Si
surface, another preparation method for clean surfaces is to
grow a clean film in UHV. One additional reason for in-situ
growth is that alloy sub-monolayer films are very fragile and
cannot be prepared by the other methods described previously.
Monitoring the thickness of evaporated films is a
critical point in growing thin films for HREELS studies.
Especially in the two layer model the thickness of a film is
one of the parameters which determines the loss energy.
Thickness measurement by the variation of AES intensity and
quartz crystal thickness monitor is discussed in Appendix C.


50
must follow to get rid of oxide. This is the so called alkali
treatment. This alkali treatment should be repeated at least
once more. Fourth, after a rinse in the combined chemicals,
H20 : HC1 : H202 = 5: 1: 1, the thin surface oxide must be made
for 10 min using the alkali solution. Then a rinse, etching
by HF : H20 = 1:1 for 30 secs and another rinse must follow to
remove the thin oxide. Repeating with a solution of H20: HC1:
H202 = 5: 1: 1 the oxidation and rinse must be followed by
etching using a new batch of HF: H20 = 1: 1. Then without
exposing the Si substrate to air, the HF: H20 = 1: 1 solution
must be diluted by adding deionized water. Finally the
boiling, combined chemicals of H20: H202: HC1 = 1: 1: 3 must be
used to grow a thin oxide for 2 min. After waiting until it
stops bubbling, the Si wafer must be rinsed for 10 min, spin
dried and quickly transferred into the vacuum.
Besides chemical oxidation a hot Si substrate can be
exposed to oxygen under UHV. The temperature of the substrate
was increased up to 900 K during exposure to oxygen. If the
substrate temperature is above 700 K, there are two
advantages. One is oxygen molecule can dissociate on the
substrate and become reactive. The other is adsorbed
impurities related to hydrogen species desorb quickly from
the substrate, leaving reactive sites for atomic oxygen. The
oxidation procedure is as follows. Once the amount of
exposure is determined, the total time and pressure should be
calculated. Annealing the sample at a given temperature is


X Reciprocal Lattice
I(S): Interference function
along a reciprocal-lattice rod.
(b)
ON
w
Figure 8. The reciprocal lattice and Ewald construction,
(a) Three dimensions; (b) Two dimensions.


34
2.5. Low Energy Electron Diffraction
In this section, the experimental set-up for LEED
measurements will be introduced and digital LEED used for
measurement of lateral surface lattice constants will be
briefly described. A LEED optics consist of an electron gun
and a hemispherical grid system. The LEED optics used is a
modified Perkin-Elmer 15-120 LEED optics. Originally it was
a front-view LEED system, but after changing the screen into
a transparent glass screen, images can be seen from the other
side of the sample (so called Back-view LEED). Electrons from
filaments with primary beam voltage V, pass through lens
electrodes with potentials V,, V2 and V3 to focus into a
parallel beam at the sample surface. The voltages V,, V2 and
V3 are used to preserve the focus for a fixed ratio (V2 -
V,)/(V3 V,) Hemispherical grids allow one to achieve a
field-free region between the sample and the first grid, G,,
as well as to remove inelastic electrons with a potential
barrier, G2 and G3. After the elastically scattered electrons
pass through the hemispherical grids, a concentric collector
electrode (coated with phosphor) accelerates them to produce
a bright visible image in a Laue diffraction pattern
arrangement.
Electrons interacting with the sample will have various
energies and directions. A hemispherical grid system is used
to image the diffraction pattern by allowing only diffracted


95
(c)
Figure 10.continued.


177
covered with hydrocarbon species compared to the two previous
samples.
The HREELS spectrum obtained from the sample annealed at
520C is shown in Fia. 27(b). The hydrocarbon 360meV stretch
mode is below the detection limit and the FWHM of the
quasielastic peak is reduced to 12.6meV from 16meV. This
effect is also due to hydrogen desorption. Upon annealing up
to 920C, the three oxide related peaks did not change.
Annealing at 1010C induces oxygen species to desorb from
the surface leaving an Si(111)-7x7 surface with only carbon
impurities. In Fia. 27(c). the HREELS spectrum after
annealing at 1010C is shown with a resolution of the elastic
peak of 9.3meV which is less than that of the previous two
samples. The peak at 113meV is due to adsorbed carbon species
which have transformed to Sic during the annealing procedure
and the peak at 99meV is due to Si-OH. The top HREELS
spectrum shown in Fig. 27(c) was taken subsequently just after
the bottom spectrum ( i.e. 20min time interval between two
spectra). Comparison between top and bottom spectrum shows
rapid increase of intensity at 99meV while 113meV does not
change so much. We can deduce that an OH species of residual
gas (dissociated H20 on Si(111) surface) adsorbed on active
sites left after desorption of oxygen species. We can also
deduce that many of the hydrocarbon species shown in Fig.
27(a) are directly adsorbed on the Si substrate so the strong
intensity of the Sic species is found even after the oxide


59
As well as losing energy to other electrons in the
crystal, the incident electron can also lose energy to the
crystal lattice by scattering off a phonon, giving up some
energy and momentum. Energy losses to phonons are very small
of the order of 50-100meV. Electrons with such a small loss
can not be filtered out by the energy selecting grids which
are used in low energy electron diffraction (LEED) optics
because their resolving power has a width of 0.5-1.0 eV; thus,
these electrons are called quasi-elastic electrons. Besides
phonon losses, the quasi-elastic peak can contain features due
to all possible small energy loss mechanisms, e.g. surface
plasmon losses due to doped carriers in semiconductor.
Elastic scattering is used to detect the long-range
ordering of the surface using LEED optics and finite
penetration depth of the incident electrons. Inelastic
scattering with low energy loss (-lOOmeV) is used to detect
surface excitations related to phonons, plasmons and resonant
electron-hole pair scattering. Two techniques, LEED and high
resolution electron energy loss spectroscopy (HREELS) analyze
the same elastic electrons with scattered electron background
due to small inelastic energy losses but each of the
emphasizes different features according to the purpose of the
measurement. For structural analysis the angular intensity
pattern of LEED is used and for determination of small energy
loss features HREELS is used since this technique resolves the
intensity features in a LEED angular pattern.


Intensity
-100 0 100 200 300 400
Energy (meV)
(c)
Figure 16.continued.
118


23
atomic Auger transitions, higher energy electrons usually
escape from deeper layers. Besides the elemental information
from the surface, the fine structure of Auger peaks is
sensitive to the chemical environment. This change arises
from transitions involving a surface molecular orbital formed
by electrons in chemical bonds between the adsorbed species
and the outermost surface atoms. The localized density of
states at adsorption sites will be different from the clean
surface. It is difficult in AES to relate peak shifts to
changes of certain levels (a few electron volts) like XPS
since three electrons are involved. But gualitative detection
of chemical shifts in AES are at least indicative of
differences in chemical bonds between elements. This can be
found in silicon oxidation which produces a native silicon
oxide layer.
Another aspect of AES is guantitative analysis. Two
possible problems in using AES as a quantitative and elemental
analytical technique are the determination of the total number
of detected Auger electron separated from the background and
the understanding of relation between the intensity of the
spectrum and the atomic density in the selvedge of the
specimen. At this point, these procedures are not fully
developed. The typical AES experimental procedures are as
follows [4].


146
and Fia. 21. both modes are showing dipole-like intensity
variation. Energy changes are not found. The primary
energy,7.0 eV, and the total angle change are not enough to
see the energy variation for the longitudinal phonon to the
end of Brillouin zone edge.
Oxygen exposure (25L) induced lxl LEED pattern (Fig. 20).
Even though surface periodicity went back to lxl, the position
of major oxygen atom doesn't change so much from the hollow
site of three fold symmetry. It also found that oxygen still
induce zone folding.
For Ni(110) surface, from Fig. 22 a very small trace of
residual gas can be detected around 60meV, which is considered
as dissociated water species (namely, O-H; 450meV, Ni-OH;
50meV, Ni-0; 60meV). The HREELS spectrum of the sample after
18kL of oxygen exposure on hot substrate (3 00C) is shown in
Fig. 23. The LEED pattern from single crystal oxide film on
Ni(110) surface showed Ni(100)-lxl pattern, since the relative
size of oxygen is larger by transferring electrons from Ni,
which cause a little rearrangement of original lattice
position while making NiO. Strong intensity at 67.2meV is
due to the FK mode of NiO single crystal films. This peak
from NiO on Ni(100) was reported earlier [75-77]. Also this
peak of easily identified using NiO bulk band structure that
indicates longitudinal optical phonon at 72meV at r point
since it is diatomic single crystal [78]. A big shoulder near
50meV can be clearly identified from the gain side. Either


175
This energy corresponds to -2ML of oxide layer which is twice
the thickness of the Shiraki oxide sample. We can deduce that
the intermediate oxide layer formed in deionized water reaches
to the saturation thickness -2ML and is passivated. In the
water the OH activity is high enough to attack the Si-Si bond
forming Si-OH and Si-H. Only the OH species due to water,
instead of 02 in the air, can contribute to the oxide species
in Si-O-Si structure. A similar result was reported by
Grundner and Schulz for the sample rinsed with water after HF
etching [91] indicating that a few minutes is sufficient time
in H20 to produce this surface. The C-Hx mode (359meV), 0-H
mode (455meV) C-H deformation mode (178meV) and Si-H
(254meV) mode are similar to the Shiraki oxide sample results
shown in Fig.25
After annealing at 500C the resolution decreased to
18meV due to desorption of hydrogen species as expected (Fig.
26(b)). The 145meV asymmetric stretch mode did not shift at
all, which means the thickness of oxide layer does not change
at this annealing temperature. Additional annealing at 900C
induced a 7x7 LEED pattern on the surface. The 3/7 and 4/7
fractional order beam intensities are particularly strong.
The FWHM of the quasielastic peak of the HREELS spectrum [Fig.
26(c)) is l3.5meV which is much smaller than that of the
Shiraki cleaned surface which was 15.8meV. These differences
are due to the fact that the dangling bonds which are
understood as the origin of the metallic characteristic of the


114
intensity was measured to be AE=33meV. The highest density
(N =5.5x10" cm'3) sample had a maximum loss energy of 26meV at
1000 C annealing. Dispersion curves for these three
different losses are shown in Fia. 15.
Semi-insulating GaAs(lOO) samples were prepared by
degreasing and introduced into the HREELS chamber similar to
the Si(111) samples described above. These samples had a
disordered surface oxide layer since we observed no LEED
pattern and Auger electron spectroscopy (AES) shows
O(510)/C(273) = 4/1, Ga(1070)/As(1128) = 4/3 and Ga/O = 1:1.
Spectra taken from this degreased sample using HREELS are
shown in Fig. 16(a). After subsequent annealing at 580C for
2.5 min without sputtering we obtained HREELS data shown in
Fig. 16(b). and AES shows similar impurity levels to a native
oxide unannealed sample except for a decrease in carbon as
measured by AES to a ratio, C/0 = 1/6. After annealing at
665C for three minutes we obtained another HREELS data shown
in Fig. 16(c) and the LEED pattern from this surface was
C(4x4). From AES data, no oxygen is detected but carbon
impurity was detected (C(273)/Ga(1070) = 1:4 and
Ga(1070)/As(1128) = 4:3).
Undoped, semi-insulating GaAs(100) was sputtered and
annealed at 500 C for five minutes. No LEED pattern was
shown and no impurity was detected from the AES data. For
the clean and annealed surfaces of GaAs(100) the principal
HREELS loss feature at an energy of 36meV is a Fuchs-Kliewer


83
Summarizing the results of this section, the scattering
probability near specular scattering geometry consist of two
factors. One is kinematic factor fully determined by the
scattering geometry and the other is the spectral function,
S(Qh,w) which contains information of surface property, e.g.
a dielectric property. Also this general description of the
dipole scattering probability of a semi-infinite medium
confirms that the scattering probability sharply peaks near
the specular direction (i.e. forward scattering). The
electric potential due to the specimen (i.e. 0>z>-Q||'') extends
up to a range of Q||"', where Qn is the transferred momentum.
3.3.3. Impact Scattering: Off-Specular Scattering
In dipole scattering theory, to simplify the derivation
of scattering probability, major assumptions are single
scattering and very small amount of momentum transfer parallel
to the surface. Another regime of scattering accompanied with
relatively large momentum transfer parallel to the surface is
called impact scattering. To induce large momentum transfer,
the incident electron energy can be increased at a specular
geometry. Since the aperture of the analyzer has a fixed
solid angle (A0~1) electrons passing the edge of the
aperture of the analyzer have different momentum for different
electron energies; i.e. the larger the energy is, the larger
the transferred momentum is. The other way to increase
transferred momentum is changing the geometry of spectrometer


129
resolution is directly related to the detection depth (-q^1)
of incident electron with 14eV. At the near specular geometry
which has the deepest detection depth, the average of the
effective carrier density can be lower than that of off-
specular geometry in case of the uneven distribution of
carriers. This indicates that the surface has a highly
enhanced acceptor density due to sputtering. Compared to the
bulk doping density (nb=1.4xl0 cm'3 ), the effective carrier
density (nefl =6.0x10'* cm"3) is due to band bending reduced after
Fermi-level is pinned near the center of the bandgap. But the
depletion of carrier is not perfect, and the surface has an
increased carrier density due to sputtering acceptor density
due to sputtering since it is p-type semiconductor.
For the sputter-cleaned semi-insulating GaAs(lOO), the
plasmon loss peaks at specular geometry is reduced as the
annealing time is increased. Thus this plasmon is due to
surface defect related carriers created during sputter anneal
cleaning cycles. This plasmon energy shifts to lower energy
as defect density reduces with increased annealing time, which
means that defects act as acceptors near the surface for p-
type sample.
Angle resolved spectra of semi-insulating GaAs, shown in
Fig. 18. exhibits the dipole characteristic of FK mode which
is the rapid decrease of intensity for A6 >0. Also rapid
decrease of elastic intensity helps to identify the plasmon
peak whose intensity does not decrease rapidly at off-specular


CHAPTER 2
EXPERIMENTAL PROCEDURES
2.1. Overview
Surface studies have been made possible with the
development of new pumping techniques during the 1960s. The
competition between pumping speed and gas desorbing from
chamber walls or leaking of the chamber determines the
ultimate pressure. The basic condition for surface studies
is that the pressure in the chamber must be kept as low as
possible in order to prevent contamination of the surface by
residual gas during the experimental procedures. To achieve
the ultimate pressure for ultrahigh vacuum (UHV) (<10'9 torr) ,
combined pumping techniques such as turbo molecular pumping,
bakeout at 200-250C and sputter-ion pumping are generally
used.
Another factor determining the quality of data is the
preparation of a clean surface. A clean surface can be
achieved by chemical pre-treatment before introducing it into
the chamber with subsequent ion-bombardment sputtering
followed by thermal heat treatment under UHV. The common goal
of cleaning is reducing the level of contamination and defects
13


135
spectrum. Oxygen exposure ( 3L ;5xl0'* torr, 60 sec) induced
p(2x2) LEED pattern. The HREELS spectrum for this sample is
shown in Fia. 20. Angle resolved HREELS spectra were taken
from specular (Fig. 20) to 4.5off specular direction. Off
specular HREELS spectra are shown in Fig. 21. Oxygen (25L)
was exposed to the fresh cleaned sample. The LEED pattern
changed from lxl to p(2x2), and finally to lxl pattern. The
HREELS spectra from this sample are similar to Fig. 20.
The HREELS spectrum from clean Ni(110) surface is shown
in Fig. 22. Oxygen (18kL) was exposed to the sample held at
300C. The LEED pattern was changed from lxl rectangular
pattern to Ni0(100)-lxl pattern, i.e. a square pattern. The
HREELS spectrum is shown in Fia. 23. Oxygen of 300 L (2.5x
10'7torr, 1200 sec) was exposed to clean sample held at room
temperature. NO LEED pattern was shown. The HREELS spectrum
is shown in Fig. 24(a) Oxygen of 1.8x10 L (ltorr, 1800 s)
was exposed to clean sample held at room temperature. The
HREELS spectrum is shown in Fig. 24(b). Air (300 L; 2.5x10'
7 torr, 1200 C) was exposed to clean sample held at room
temperature. A diffused lxl LEED pattern was shown. The
HREELS spectrum is shown in Fig. 24 fc). All room temperature
oxidation samples did not show any LEED pattern indicating a
disordered oxide layer of thickness greater than 3-4.


131
(22meV-30meV) which shift according to the degree of ordering
display dipole-type selection rules.
It has been shown that the HREELS data for GaAs(lOO) at
T=300 K can be quantitatively described by a coupled plasmon-
phonon model using an effective carrier density that is in the
range of 10 to 10 cm'3 for both heavily doped p-type and
undoped material. For p-type (n =1.4x10 cm'3) GaAs(lOO), a
dispersion in the range q||=0.012'' was found which is close to
the cross-over point for uncoupled phonons and plasmons at the
bulk carrier density [59]. The effective carrier
concentration was determined to be 6.0x10 cm'3 to a precision
of 50%. This effective density will depend on the bulk
doping as well as the sputtering and annealing procedures
which produce acceptor-like defects in the near-surface bulk.
These acceptor like defects produced by sputtering and
annealing were also detected in the undoped sample.
From two different kinds of semiconductor surface studies
angle resolved HREELS can yield detailed information about the
surface carrier concentration for semiconductors which can not
be obtained from other surface methods. Since these HREELS
results are shown to be strongly influenced by surface defects
one can obviously use the angular dependence of HREELS to test
defect models of interface formation including Schottky
barriers [65]. Some of these possible experiments might
include in situ growth of semiconductor overlayers including


53
In this section, the growth rate monitored using a
profilometer (Sloan, Dek-Tek II) will be discussed. Before
the actual evaporation is conducted on the sample surface, the
quartz glass crossed by tantalum wire is exposed to the
evaporation source at the same position that the real
substrate would be. After evaporation, the wire trace left
among the evaporated spot will be detected as a groove by the
profilometer. Since the thickness of this wire is 0.001, the
sharp groove made it easy to level between each ends of
groove. The evaporated film thickness is just the height
difference between end of groove and center of groove. Even
though the emission current knob (0-750 mA) in the power
supply is the only variable to control the e-beam evaporation,
the actual evaporated film thickness is not controlled
precisely by the position of the knob, because the power is
not stable. Further because the source is always changing,
the position of the current knob does not give consistent
results. Instead of the knob position the emission current
reading provided a more consistent way to reproduce the beam
evaporation conditions after several trial evaporations. In
Fig. 6. Ge source and Si source evaporation rates versus
emission current are shown respectively. The evaporation rate
as a function of emission current is different for each
material. The initial vapor pressure curve for each material
gives information on melting point and vapor pressure. A
limitation in the use of the profilometer is that some


156
Therefore we can expect continuous and controlled oxidation
by the amount of oxygen exposure on samples heated above
~350C.
Finally the main experimental motivation of this chapter
can be summarized as follows. Thermal silicon oxide layers
are noncrystalline oxide layers which can be grown at high
pressure on silicon substrate and have been previously studied
by HREELS. As the oxide layer becomes thinner the
stoichiometry of the oxide layer will be different from Si02
and the simple dielectric function theory may not explain the
evolution of the HREELS loss peaks related to oxygen on such
inhomogeneous surfaces. Through an investigation of the
initial stages of oxidation using controlled oxidation under
UHV we may obtain the information on evolution of the
structure of this intermediate oxide layer. In addition
several different preparation conditions were used to produce
intermediate oxide layers which may have different structures.
Besides thin intermediate thermal oxides grown by pure oxygen
environment under UHV, we have also investigated wet-
chemically prepared Shiraki oxides, deionized water preserved
oxides and air exposed native oxides. All of these
intermediate oxide layers are very thin compared to silicon
dioxide and we can expect to observe the apparent differences
among them using HREELS.


600
Energy (meV)
(C)
Figure 25.continued.
160


64
S = Ginn will be satisfied for the ray terminating at that
point. Diffraction maxima gives the coordinate 6 and thus G,n
and unit vectors a, b, c. From equation (3-5), the intensity
of a diffraction maximum is |F|J2. But the modulus of F|W) does
not uniquely determine the atomic arrangement within a unit
cell for non-Bravis lattices with more than one atom per cell.
3.2.2. Low Energy Electron Diffraction: Surface Diffraction
A beam of low-energy electrons passing through a material
is attenuated by both elastic and inelastic processes.
Inelastic processes are usually treated as simple attenuation
by modifying the kinematic description derived above.
Assuming the ratio of the amplitude contributed to the
scattered beam by atoms in successively deeper planes is a =
An.t/An where n is the plane number. From equation (3-4), the
total amplitude for a semi-infinite crystal is
A (S) = f(0,E) E E exp [is (m,a+m2b) ] E a"3 exp(im3Sc)
1^1 1^2 HI 3
= f (0 E) E E exp [is (m,a+m2b) ] [1-a exp(is-c)]'1
m, m2
The corresponding intensity is
(3-9)
sin2(N,S a/2) sin2(N2S b/2) 1
I(S) =|f(*,E)|2
sin2(S a/2) sin2 (S-b/2) 1+a2 -2acos (S c) ,
(3-10)


CHAPTER 1
INTRODUCTION
1.1. Overview of Surface Physics Techniques
The potential energy function experienced by atoms at a
surface is different from that of the bulk due to the absence
of atoms above the surface. The atoms at the surface which
are affected by this potential will arrange themselves in
order to determine the lowest energy state of the combined
system: surface and bulk. This results in the surface having
structural, electrical and chemical properties which may
differ greatly from corresponding bulk properties. This
dissertation describes a sequence of surface physics studies
using high resolution electron energy loss spectroscopy
(HREELS) to measure surface vibrational modes. A variety of
supporting experiments are also performed and discussed, since
it is necessary to prepare, characterize and study a single
atomically clean surface in the same ultrahigh vacuum (UHV)
chamber.
It was not possible to monitor an atomically clean surface
before UHV techniques were developed. Due to the short time
during which a surface stays clean, experiments must be
performed before the clean surface is contaminated by residual
1


33
34
35
36
37
38
39
40
41
42
43
44
45
46
ARHREELS spectra obtained from Ge0 ,Si0, (111)-5x5
for specular (A0=O) and non-specular scattering
geometries.
193
Hydrogen titration on Ge0 5Si0 5(111)-5x5.
(a) 2.5 L, H; (b) 5 L, H 199
Deuterium titration on Ge0 ,Si0 ,(111)-5x5.
(a) 2.5 L, D; (b) 5 L, D 201
LEED intensity profile of Ge0 ,Si0,(111)-5x5
measured by digital LEED.
205
Surface lattice relaxation 206
Schematic diagram of gas handling system 215
Design of evaporation system.
(a) Collimators; (b) Cross sectional view;
(c) Sideview 216
Schematic diagram of digital LEED 221
Zoom trials of GaAs at different energies.
(a) Zoom at the elastic peak; (b) Zoom at 270 meV:
(c) Zoom at 360 meV.
227
Plate voltage versus capacitor voltage to correct
tuning.
230
Screens of "LHMAIN" program.
(a) Filenames; (b) Selection of commands;
(c) Setting of parameters.
232
Ge evaporation control using AES.(a) Normalized
Si(92eV) intensity versus Ge coverage by Gossmann
et at.[102]; (b) Normalized Si(92eV) intensity
versus Ge evaporation time to calibrate
evaporation rate.
245
Thermal evolution of evaporated Ge film on Si(lll).
247
RBS data from 1000 of Ge/Si alloy film 248
x


94
Momentum Transfer ('1)
(b)
Figure 10.continued.


58
plasmon excitation (due to valence band electrons), interband
excitation and Auger electron excitation (24].
In this chapter we will briefly examine the theory of
the second feature which includes elastic scattering as well
as quasi-elastic scattering. Usually the total emitted
electron yield increases as the primary energy increases up
to several hundred electron volts. But the elastic yield
alone of the second feature shown in Fig. 7. shows a different
energy dependence. At normal incidence the elastic yield will
be largest at primary energies less than lOeV, where it
amounts to about 50% of the incident electron intensity and
decreases with increasing primary energy.
The penetration of the primary electrons into the solid
is limited by inelastic events and it is estimated that at
typical energies (10-100 eV) the penetration depth is 3-10 A
[25], It follows that the elastic component of the emitted
electrons can originate from a few atomic layers parallel to
the surface. The wavelength of the electrons (1) is h/p by
the de Broglie relation; in practical units
A= [150.4/E(eV)]H A (3-1)
where E is the kinetic energy in electron volts. At 100 eV
the electron wave length, X, is of the order of 1.3 A so that
a diffraction pattern of scattered electrons by the atomic
lattice will occur.


227
-100 0 100 200 300 400
Energy (meV)
Figure 41. Zoom trials of GaAs at different energies.
(a) Zoom at the elastic peak; (b) Zoom at 270 meV:
(c) Zoom at 360 meV.


60
3.2. Elastic Electron Scattering
Before going into a description of elastic surface
scattering theory, it is better to start from the kinematic
description of scattering from X-ray diffraction from small
crystals [26]. The first Born approximation is used for
calculating kinematic scattering factors [27]. Also multiple
scattering within the atom can be included in the scattering
factor. Initially the ideal intensity of diffraction from a
crystal will be derived. Then low-energy electrons will be
considered an incident particles instead of X rays. Intensity
attenuation relative to the penetration depth will be
considered. Inner potential effects of the crystal as well
as thermal vibration effects on the diffraction intensity will
be discussed.
3.2.1. Diffraction From A Bulk Crystal
: 3-dimensional Diffraction
Assuming a sample is uniformly illuminated by the
incident beam and ignoring multiple scattering, the scattering
amplitude of an incident plane wave from N atoms, whose
scattering factor is ft(^,E), is [26],
A = A* 2 f i (0 E) exp (iSr,), (3-2)
where r, is the position of i-th atom and S = k k is the
transferred momentum. The factors k and k<, are the


207
the lattice constant of the 7x7 pattern is larger than that
of Si(111)-7x7. This 7x7 pattern may result from domains of
pure Ge on the surface of the alloy. Such a 7x7
reconstruction of pure Ge implies that this part of the film
is not fully relaxed to the 4% larger lattice constant of bulk
Ge compared to bulk Si since the relaxed Ge surfaces has a 2x8
reconstruction. From the post-film analysis, SEM photographs
(Fig. 47 in Appendix CM showed Ge-rich islands of the size of
a few jum. This is not surprising since thin films of strained
pure Ge with 7x7 pattern have been formed on Si (111) by Ge
deposition at high growth rate [116]. This result may
indicate that the deposition rate exceeds that of dissolution
rate into bulk. In other words, if the deposition rate is
kept constant and the diffusion length becomes longer, then
the dissolution rate becomes smaller and finally no
dissolution will occur. Also, even if the estimated thickness
of alloy was 1000 A, the actual thickness is less than 1000
A due to island growth of the film. From AES analysis, there
are thick Ge rich islands whose typical size is a few /m on
the final surface, and the alloy film thickness is quite
shallow (<50 A) compared to the thickness of island. So the
measurement of relaxation at 1000 A is mostly due to Ge-island
films. This is the reason why the relaxation graph has a
large error bars, since the two 5x5 unrelaxed and 7x7 relaxed
patterns are overlapped.


1000
100
Energy (meV)
(c)
Figure 27.continued.
167


Intensity
139
x40 x200
-50 0 50 100 150 200
Energy (meV)
(c)
Figure 21.continued.


49
oxide layer in air. It is called a 'native oxide' and can be
desorbed by thermal annealing at the temperature above 1000C
under UHV. To prepare the native oxide on an Si substrate
with a small amount of carbon impurity, degreasing the
substrate is sufficient (i.e. acetone and methanol then blow
drying). Removal of the native oxide only by high temperature
annealing can not remove carbon impurities on Si surfaces in
the temperature-limit of generation of no other defects. Wet
chemical oxidation procedures have been developed. One of the
methods which was used in these studies is the so called
Shiraki technique which consist of a series of chemical
treatments as follows [10]. An Si(lll) wafer 0.5mm thick,
1/4" x 3/4" rectangular, and p-type Boron doped with
resistivity of 2.4 n-cm has been used. First the substrate
must be degreased in hot (~80C) bath (whole procedure was
done in hot bath except rinsing) in the sequence of methanol,
acetone, trichloromethane, acetone and methanol. At least
lOmin should be spent in each bath. Then the substrate should
be rinsed by deionized water (it will be called rinsing).
Second, using HN03, etching the surface region and forming an
oxide layer for 10 min should be followed by an HF oxide etch
for 10-15 secs and rinse. Repeating the HN03 and HF
sequential etch procedures must be continued until the surface
dries uniformly. Third, in the combined chemicals,
HjO:NH40H:HjOj = 4: 1: l, the substrate must be oxidized for 5
min. Then rinsing, etching by HF : H20 = 1:1 for 30 seconds


CHAPTER 4
SURFACE PHONON AND PLASMON MODES ON
Si(111) AND GaAs(lOO) SURFACES
4.1. Overview and Motivation
The energy-loss spectrum of electrons reflected
specularly from a clean Si(111)-7x7 surface is a broadened
elastic peak [33,44-46]. It is widely accepted that this
broadening does not result from phonon losses since Si is a
nonpolar semiconductor with no phonon excitations in the
dipole limit [33]. Instead this broadened elastic peak has
been attributed to losses due to a two-dimensional metallic
state created by dangling bonds from the odd number of atoms
in the surface 7x7 unit mesh or cell [44,45]. It has also
been interpreted as a conduction-band surface-plasmon
excitation [46], which is an oscillation of n-type bulk
carriers localized below the surface in a space-charge layer.
Because of typical low carrier concentration and limited
resolution of typical high resolution electron energy loss
spectrometers, these plasmon energies have been too small to
be separated from the large elastic peak in a specular
scattering geometry (A0=0). In this chapter new experiments
where the elastic peak intensity is suppressed by changing
107


246
ratio [Si(92)/Ge(52)] is 6, which means Ge mole fraction is
near 50%. Above 650C, annealing AES intensity ratio is above
10 and a 7x7 LEED pattern reappeared. At higher temperature
annealing (~800C) the Ge peak has totally disappeared since
Ge diffuses into the bulk and Ge is not expected to sublimate
due to the low annealing temperature. A graph of this
annealing procedure is shown in Fig. 45. We chose the
temperature 560C for subsequent Ge evaporation, since Ge may
start to diffuse into the surface a little faster and the
exposed Si atom can make an alloy with freshly arriving Ge
atoms. After cleaning once more, Ge was evaporated onto the
Si (111) substrate held at 560C at the rate of 0.2ML/min. At
the 50 A interval, the distance between two integer beams was
monitored by digital LEED measurement using a Vidicon Camera
interfaced with NOVA computer. The 50 A alloy thickness was
assumed to be twice the thickness of evaporated Ge. For a
given film thickness, intensity profiles are measured at
several electron beam energies ranging from 35eV to 55eV.
Each line scan consists of 512 channels and the integral order
spots were typically separated by about 300 to 350 channels
depending on the LEED beam energy. The uncertainty in the
spot separation was about one channel or 0.3% of the
separation. After a film of 1000 A was grown, the sample was
transferred to RBS chamber. Two different spots were
monitored using 2.0 MeV a particles. The data is in Fig. 46


259
. I :>
0 1.5 min. 3
(e)
Figure 50.continued.


261
this long tail of Ge is due to the thick Ge island, not the
diffusion of Ge into the Si substrate.
In conclusion Ge0 5Si0, alloy film growth on the hot (560-
590C) substrate follows the Stranski-Krastanov growth mode.
Initially at the very thin (<10 A) alloy layer without Ge
island in SEM image, split 5x5 LEED pattern is shown, which
indicates that the origin of 5x5 layer is not Ge-rich island.
But the function of Ge-rich islands, which are distributed
evenly at the surface, is essential to relieve the strain of
Ge05Si05 alloy layer. That is the reason why a clean 5x5 LEED
pattern and the Ge-rich island appear almost at the same time.
This growth method does not increase the thickness of alloy
for further Ge-evaporation, but does increase the width and
the thickness of the island.


17
de-focussing of the HREELS electron optics. At the same time
the sample surface is also quickly contaminated by the
residual gas. If the sticking coefficient of the residual
gas is assumed as one, then it takes 1000s (~17min) at lxlO'9
torr to cover the sample surface with one monolayer of
residual gas. Since HREELS is a very sensitive technique, in
order to eliminate impurity contributions to the data it is
a prerequisite condition to achieve pressures below lxlO',0torr
(i.e. 170 min for covering 1 monolayer, which consists of the
time necessary for tuning HREELS and time for data taking by
computer). The vacuum chamber overall diagram is shown in
Fia. 1. Due to the chamber geometry and position of the
pumping connections, the pressure at the entrance of ion-pump
is almost half of the pressure near the hemispherical analyzer
as measured by conventional UHV ion gauges. An HREELS
spectrometer is just above the ion-pump since that position
is near the lowest pressure in the chamber. A sample
manipulator horizontally transports a specimen from the HREELS
spectrometer to the electron analyzer. An electron gun and
5-crucible e-gun source are in one UHV section aimed at a
common point which is just below the electron analyzer used
for AES. Next to this electron analyzer are gas leak valves
and an ion-gauge. Next to gas leak valves is the LEED optics
section (see Fig. 1) In this section an 8"OD viewport allows
one to observe the LEED screen as well as the motion of the
sample manipulator. In order to allow a more complete view


6
vibrational modes such as H-Si modes on Si(100) which are
split by only 1.4 meV. Since the performance of the electron
optics depends sensitively upon the work function uniformity
and stability of the energy analyzer capacitors, HREELS
experiments should be done under UHV conditions. For highest
signal levels, the specimen should have a dynamic dipole
moment which has a large component along the surface normal.
At the present time HREELS is under development to improve the
experimental resolution of spectrometers and to improve
scattering theory which does not yet treat long-range dipole
scattering and short-range impact scattering on an equal
basis.
1.2. Overview of Specific Experimental Studies
The motivation of this dissertation is to systematically
study semiconductor surfaces using HREELS and in one selected
example (oxidation of silicon), to compare a similar study on
a metal surface (oxidation of nickel). For practical use of
semiconductors, these materials have excess electrons or
holes; i.e., they are n-type or p-type. These electron or
hole carriers due to impurity doping play an important role
in forming potential barriers at the interfaces of
semiconductors with metals (e.g. Schottky barriers or ohmic
contacts). Plasmons due to excess carriers can be
distinguished from phonons by energy-wavevector dispersion
which can be measured by angular studies of HREELS. Even if


225
intensity can be recovered by adjusting the external bias
knob. When the sample is at the scattered position, the
surface of sample can be viewed through the port. At this
point the count rate is almost zero since the beam is almost
blocked by the sample. The sample is rotated 30 counter
clockwise if the ideal specular geometry is 60. The
ratemeter should be at the lowest count rate position and
filament current should be increased. After rotating sample
(e.g. 30 counter clockwise), the monochromator part should be
rotated through twice the angle that the sample is rotated
(e.g. 60 which correspond to 2 cm down from the top notch)
since the analyzer part is fixed. At near the specular
geometry the ratemeter will be guickly saturated. By
adjusting the position of sample, the maximum countrate should
be achieved. If the rate meter saturates at 104/ s, the
filament current is reduced. The ELS-22 power supply
potential can be adjusted to find the maximum countrate. The
countrate is guite different according to the state of sample
surface since resolution depends upon angular divergence from
the target. By checking the full width at half maximum (FWHM)
of the slit potential of analyzer using ramp knob, the
resolution of setting is measured. Sensitivity of the
potential is guite different for samples and geometries.
When the target causes the angular divergence the energy
broadening will give an asymmetric profile of the elastic
peak. In this case, by adjusting electrode '4' of


218
position of sample. The rest of the beam not captured by the
sample (l/4"x3/4") is blocked by the shutter positioned at the
back of the sample. Another shutter between two collimators
can control the beam before it arrives at the sample. The top
copper collimator is attached to the copper gasket to prevent
gas passing through along side collimator. The bottom
collimator which is attached to a liquid nitrogen cylinder
captures the evaporated material. This design is limited by
high pressures inside of evaporator section which is almost
two order of magnitude higher than the sample side. Two
viewports (2 3/4") aimed at the filament and crucible allow
one to observe the state of a source during evaporation. Two
flanges at the top of source chamber are used to refill the
source without detaching the whole evaporation system from the
chamber. Center section of -9" OD cylinder secures the space
for electrons to move without hitting the chamber walls.
As mentioned earlier, the small diameter of collimator
reduces conductance of gas and results in higher pressure (~2
orders) at the source side than the sample side during
evaporation. This can not be allowed if a thin film
(submonolayer-100 A) is desired, since slow deposition rates
for better control require relatively longer periods of
evaporation. Then impurities from the high background
pressure can quickly saturate the sample surface. Therefore
an independent pumping system with a similar capacity as the
sample side is necessary. A large surface of liquid nitrogen


230
Figure 42. Plate voltage versus capacitor voltage to correct
tuning.


72
w0/k: parallel component of momentum
change to reflected specular
beam
k0: perpendicular component of momentum
change to reflected specular
beam
(b)
a: 45
Figure 9.continued.


234
necessary to shift elastic peak position to zero loss position
to find real loss energy. Magnification is an essential
procedure in recognizing loss features with reasonable
intensity. Precise measurement of loss energy, intensity and
FWHM is obtained by the cursor on the curve. Especially in
angle resolved HREELS data analysis, plotting all the curves
for different incident angles on one sheet makes comparison
easy. In case of plotting noisy data, smoothing by three
point average technique helps to display better-looking curve.
But smoothing technique causes to remove detailed information
from a spectrum, so both spectra with and without smoothing
which are overlapped in one sheet sometimes avoids both
extremes. Besides options listed above, many mathematical
options including derivative and integration are also included
in the plotting program '2D plot'.


147
adsorbed hydroxyl group mode (50meV) or transverse optical
mode quite similar to adsorbed oxygen vibration on Ni(100)
surface(49meV) can be the loss at 49.5meV [42,79]. A little
hump near 20meV is interfacial phonon mode between Ni(110) and
NiO(lOO). More precisely speaking, oxygen at the NiO side
induce the surface mode at the Ni(110) surface through
Brillouin zone folding similar to Ni(100) case [74,80].
Comparing with pure surface (Fig. 22] and thick oxide (Fig.
22) surface, such a huge Ni-OH intensity from NiO surface can
not be expected from dissociation of H20 in residual gas.
Benndorf et al. reported H20 adsorption at 300 K is possible
only if there are unoccupied Ni(110) sites in the neighborhood
of adsorbed oxygen [78]. In our case thick NiO covered the
Ni(110) surface and furthermore NiO is thermally grown single
crystal. O-H stretch mode (450meV) couldn't be detected
either. So three fundamental modes from NiO on Ni(110)
surface are NiO F-K mode (67.2meV), NiO transverse optical
mode (49.5meV) and Ni(110) surface mode (20.8meV) induced by
top oxide layer. Long tail up to 140meV is due to second
harmonics of these two phonon modes. High intensity and
energy of 67.2meV indicate strong ionic character of this
film. Dalmai-Imelik et al. [75], Cox and Williams [77], and
Andersson and Davenport [76] who reported values for this mode
as 67.5, 69.5, and 65meV respectively. The main reason why
the loss(69.5meV) of Cox and Williams (single crystal) is
different from the other authors including us (single crystal


75
In Fig. 9 (b) the outgoing elastic wave vector k, the
outgoing wave vector k, after excitations of phonons and full
scattering wave vector q are shown. Then
k = (Vy.vJ = k(sina, 0, cosa) (3-20)
where a is polar angle of k and q = k -k, = (Q,q2). Defining
6 and 0 as the polar angle and the azimuthal angle
respectively of kt relative to k (i.e. 0=0 when k, is under
k on the oxz plane), fully determines the scattered beam
direction of k,. Energy conservation then gives
w0 = h (k2 -k,2 ) k (k-k,) (3-21)
where h=e=m=l, and the second equation comes from relations
q/kl and 81. For the normal incidence case (i.e. a=0) and
0=0 ;
q^0=k-k,=k(0,0,1) -k, (sin0,0, cos) ,
~k(-0,0, [k-k,]/k) k(-0,O,wo/k2 ) ,
ak(-^,0,^o), (3-22)
where the characteristic angle 60 is given by u0/(2E). The
differential cross section derived by Newns is
4T2 cosa f(0,0,a) 86d(p
da
k2 (02+8o> )
(3-23)


-100 0 100 200 300 400 500
Energy (meV)
(a)
Figure 25. HREELS spectra obtained from Shiraki oxide on Si(lll).
(a) As introduced; (b) After annealing at 500C; (c) After annealing at 900C.
158


61
propagation vectors of scattered and incident plane wave.
Assuming each scatterer is at the lattice point
r¡ (=m,a+m2b+m3c) and atoms are located at xn in each unit cell
n, the normalized amplitude is
A(S) = S f(0,E) exp[iS- (r,+xn)], (3-3)
i .n
where the sum is over the lattice sites i and the atoms within
unit cell n. Separating the sums, then
A(S) = [E fn(0 ,E) exp(iSxn) ] E exp(iS r,)
n i
= F(0,E)- E exp (is-r,), (3-4)
i
where F(0,E) is the crystal structure factor. The scattered
intensity is then written as, I(s) = |A(s)|2,
l(S) = | F(0 ,E) |2 E exp[is-(r.-r,)]
i, j
= |F(0,E)|2 J(S). (3-5)
The interference function J(S) depends on the diffraction
geometry through transferred wave vector S. Both F(0,E) and
J(S) differ for different choices of non-primitive unit cells.
For a parallelepiped with N,, N2 and N, lattice points, the
interference function J(S) is


77
Assuming the surface coverage of adsorbates is n molecules
per unit area, the current of one phonon loss (I,) versus that
of unscattered specular reflection (I0) will be
I,/I0 = (77TJn/E) cosa [(t2-2)Y+(t2+2)lnX]
(3-25)
The main feature of dipole scattering of the adsorbate on
metal surface using a semi-classical approach is that the
scattered electrons form a strong forward scattering lobe with
vibrational modes polarized vertical to the surface.
Such dipole scattering was also described by Persson
using a quantum mechanical approach [29], In this treatment
the electron wave function near scattering region is
k (27T) [ e + e ] e
(3-26)
where k'=k-2n(nk) and 5 is the phase factor due to
scattering. Using the Fermi-Golden Rule and a perturbing
potential H'= -|x-E due to the elastic field of external
electrons and the image charge, the probability per unit time
for vibrational excitation of the molecule was calculated.
The differential cross section is
da/dn = (m/xe/7re0h)2 [p,/(p0cosa) ] [a/a2 + (bnCosS+^sinS) /b2 ]2 ,
(3-27)


91
10,5-5xl0" /cm3) the calculated intensity of loss function,
(hw)''lm[-l/{eb(w)+l) } ], is plotted in Fig. 10 fa) The
parameters 6*5, m'(=0.232 mj for p-type semiconductors and r(u)
have been chosen to represent silicon at room temperature.
The loss function has a maximum at the surface plasmon energy
>Sb2 (=4jrne2/(m[ l+&o] }) This loss peak is due to specular
scattering geometry. If the dielectric function is replaced
by the Lindhard dielectric function which describe the highly
doped semiconductor and contains momentum transferred parallel
to the surface, q,,, then the loss function, Im[-l/{ e (o, q^j) } + l] ,
will give the dispersion relation, i.e. ws vs. q|(. The surface
plasmon loss, ws, comes from the condition
op2 u2
-1 = 6- (3-47)
U2 O2 -0.6 q||2vf2
where vf (=4.20x10'* a0/rs cm/sec) is the average Fermi velocity
of free carrier, a0 is Bohr radius and rs=[3/(47rnefl) ],/3.
Transferred momentum, qj(, comes from
q,! = 2n (Eo/150.1) (sin0,-sin0,) (3-48)
where E0=E,ES and 0S,0, are scattered and incident angle with
respect to surface normal. For different doping densities
(nef,=10'5-10'* /cm3), the dispersion curves (ws vs. g,,) were


84
from specular (i.e. 0,nci a fixed incident energy. But a mixing of both ways (i.e.
energy and geometry) is not desirable for experiments
observing the evolution of peak intensities since the
reflectivity from the sample depends upon the incident energy
very much. The first method can be simply used for
observation of evolution of peak position. To detect the
dispersion relation of surface excitation, incident energy
should be increased up to a few hundred electron volts in both
methods to cover the entire Brillouin zone whose size is
typically a few A'1. Another reason for using impact
scattering is to determine local site symmetry from surface
vibrations. Since large transferred momentum induces
increased surface sensitivity
(e'9"2 and e^z factors in potential and source, see equation (3-
31)), a microscopic treatment is needed to interpret the
spectrum. It is not so simple as the dipole scattering case
to derive the scattering probability since the high incident
energy (-300 eV) to cover the Brillouin zone edge causes a
multiple scattering as we have already discussed in the
overview of this chapter. So it is necessary to approach the
scattering problem from other direction to get the
differential cross-section of this multiple scattering. In
this section a condensed account of Mill's derivation and
results will be presented [33]. Also the selection rule for
impact scattering will be briefly introduced.


103
characteristics and the substrate bulk phonon structure, are
used to distinguish the positions and the elements in an unit
cell, which result in different types of central-potential for
different types of pair. Instead of first-nearest neighbor
interaction, many neighbors interactions, if necessary, can
be easily included in the finite slab model. The dynamical
matrix is generated from above information and diagonalized
to give eigenvectors of normal modes of the slab and
eigenvalues for vibrational frequencies.
A few systems have been tested on the surface phonon
dispersion, since the relatively high incident electron energy
(200-300 eV) with the similar resolution as a low energy
incident electron beam has recently made it possible to
produce momentum transfer parallel to the surface up to edge
of the two dimensional Brillouin zone. Ni(001) and Cu(001)
surfaces (both of them are fcc-structure metals) had been
experimentally tested by a few authors [42]. Examining
results of both systems and comparing those with the
theoretical calculations, we can draw a few conclusions of
the intrinsic surface phonon. On Ni(001) surface, along the
T-X direction, two surface modes, the S4 mode and the S6 mode
can be detected, but the S, mode can not be detected due to
violation of an impact selection rule [42]. The schematic of
vibrational modes S,,S4, and S6 are shown in Fig. 12. The S4
mode is a surface acoustical phonon mode, so called Rayleigh
mode, and the S6 mode is a higher frequency mode. If the bulk


Energy (meV) Energy (meV)
2.5
3
4
Figure 33. ARHREELS spectra obtained from Ge0 ,Si0, (111)-5x5 for specular
(Aff=0) and non-specular scattering geometries.
193


100
Energy (meV)
(b)
Figure 25.continued.
159


90
semiconductors (e.q. Si, Ge, SixGet.) do not have a strong
optical phonon mode. In this section, for both semiconductors
the plasmon loss and their dispersion relation will be
examined.
For a homopolar semiconductor the loss function,
Im[-l/{ e(u)+l} ], from the semi-infinite crystal with
dielectric constant e(w) is shown in equation (3-37). For
the homopolar semiconductor (from now on Si will be used as
an example), dielectric function is
Up2
eb(u) = to (3-45)
u[u+{i/T(w))]
where 6 is the background term, wp2 =47rnefle2 /m, and t (u) is a
frequency dependent relaxation time [37]. The second term is
the Drude term contributed by the free carriers and a surface
optical phonon term is ignored since an Si is a non-polar
semiconductor. The loss function combined by this dielectric
function is
-1 W(dsp2
I*>{ } = [(wsp2-w2)2 + w2/t2(u)]" (3-46)
l+eb(w) (l+&o)r(w)
where usp2 =up2 / (l+e) and w2/r2 determines the breadth of the
loss peak in spectrum. For different doping densities (neft=


226
deceleration optics, the asymmetric tail can be suppressed.
First moving the position of recorder pen to the asymmetric
tail near elastic peak, electrode '4' of deceleration optics
is adjusted to reduce the intensity of the asymmetric tail.
If this procedure reduces the intensity of elastic peak more
than 25%, electrode '4' is changed in the opposite way to
reduce the asymmetric tail. After adjusting electrode '2' of
acceleration optics, electrode '1' and electrode 31 can be
sequentially used to increase intensity of the elastic peak.
Another important feature in tuning is zoom. If zoom is
at the elastic peak, most of the intensity is consumed by the
elastic peak so that it is very hard to detect losses beyond
200meV where important information of hydrogen related species
always exists. So as not to lose high loss energy
information, the position of zoom is around 450meV. It shows
in Fig. 41 that a different zoom position affects intensity
of the loss peak. The zooming procedure is as follows. As
mentioned earlier in tuning, most potential knobs are
independent. But electrode '3' and electrode '4' of the
deceleration optics have to be changed during the sweep (i.e.
slit potential changing) to fulfill their imaging condition.
To do that, the highest loss energy (e.g. 450meV, 0-H stretch
mode) should be the zoom position. Initially electrode '3'
and electrode '4' settings are recorded at the elastic
position (=lst setting). Sweeping to the chosen loss peak
(e.g. 450meV), the loss at the chosen peak is maximized using


3 THEORETICAL BACKGROUND
56
3.1. Overview 56
3.2. Elastic Electron Scattering 60
3.2.1. Diffraction from a Bulk Crystal
: 3-dimensional Diffraction 60
3.2.2. Low Energy Electron Diffraction
: Surface Diffraction 64
3.3. Inelastic Electron Scattering: HREELS 68
3.3.1. Semi-Classical Approach 70
3.3.2. Dielectric Function Theory 78
3.3.3. Impact Scattering: Off-specular
Scattering 83
3.4. Examples of Inelastic Electron Scattering....87
3.4.1. Surface Optical Phonon Excitation 87
3.4.2. Surface Plasmon Loss and Dispersion:
Relation of Homopolar Semiconductors,
and Plasmon-Phonon Coupling of Polar
Semiconductors 89
3.4.3. Two-Layer Dielectric-Function Model
97
3.4.4. Surface Phonon Dispersion for Semi
infinite Metallic Surface 102
4 ANGLE-RESOLVED SURFACE PHONON AND PLASMON MODES
AT Si (111) AND GaAs(lOO) SURFACES 107
4.1. Overview and Motivation 107
4.2. Experimental Results 110
4.3. Discussion 123
4.4. Summary 130
5 VIBRATIONAL MODES FROM OXIDE LAYERS ON Ni(lll)
AND Ni (110) 133
5.1. Overview and Motivation 133
5.2. Experimental Results 134
5.3. Discussion 145
5.4. Summary 150
6 VIBRATIONAL MODES FROM OXIDE LAYERS ON Si(111)....151
6.1. Overview and Motivation 151
6.2. Experimental Results 157
6.3. Discussion 168
6.4. Summary 182
v


236
Since re=(x,z), l/re can be expressed using Fourier
transform,
1
r.
2n2
d2 Q
dq
exp(-iQx) exp(-iqz)
Q2 + q2
1
2n
d2 Q e
¡Q.X-Q lU-1
(B-4)
From equation (B-2), then
a 1
3ze re
1
27T,
1
2n
d2 Q e
iQ.x-Q 2
d2Q exp[-iQ. (x0+vl|t) -QvJ 11 ]
(B 5)
Integrating over time and using equation (B-l) and equation
(B-2),
dt exp(iw0t) <01 V(t) | l>dt =
so
-1
7r
d2 Q
exp (-iQ x0) vAr
[Oq2+Q2v2] ,
(B 6)
where
n= Wo-Q-v|| .
(B-7)


7 SURFACE EXCITATION AND RELAXATION OF Ge/Si
ALLOY FILMS 184
7.1. Overview and Motivation 184
7.2. Sample Preparation 190
7.3. Surface Excitation from Ge/Si(111)-5x5 191
7.3.1. Adatom Vibration 191
7.3.2. Hydrogen Titration 195
7.4. Surface Relaxation Measurements
Using Digital LEED 203
7.5. Summary 208
8 CONCLUSIONS 209
8.1. Conclusions from Present Experimental
Results 209
8.2. Recommendations for Future HREELS Studies... 212
APPENDICES
A FURTHER EXPERIMENTAL DETAILS 214
A.1. Gas Handling System for UF HREELS System.... 214
A.2. Design of Evaporation System 214
A.3. Quantitative AES Analysis Using Standards... 219
A.4. Digital LEED 220
A.5. Tuning of HREELS Spectrometer 222
A. 6. Computer Programs and Data Handling 231
B FURTHER THEORETICAL DETAILS 235
B.l. Dipole Scattering Cross Section 235
B.2. Transformation of Coordinates 237
B.3. Fuchs-Kliewer Modes
: Lattice Dynamical Framework 239
C GROWTH AND CHARACTERIZATION OF Ge/Si ALLOY FILM...244
C.l. Alloy Film (1000) Grown on Hot
Substrate (560C) 244
C.2. Depth Profile of Thin Alloy Film
without Ge-rich Islands 254
C.3. Depth Profile of Thick Alloy Film
with Ge-rich Islands 256
REFERENCES 262
BIOGRAPHICAL SKETCH 269
vi


Intensity
137
x40 x200
Energy (meV)
(a)
Figure 21. ARHREELS spectra obtained from Ni(111)-p(2x2)-0.
(a) 1.5off-specular; (b) 3off-specular;
(c) 4.5off-specular.


Intensity
Energy (meV)
Figure 16. HREELS spectra obtained from a native oxide layer
on GaAs(lOO). (a) As introduced; (b) Annealed at
580C; (c) Annealed at 665C.
116


197
still remains after exposure, the triangular dimer is not
broken. Intensity of 104 meV does not increase, compared to
Fia. 32. This suggests that only adatoms have reacted with
hydrogen atoms. The losses at 246meV (Ge-H stretch mode) and
71meV (Ge-H bending mode) indicate that Ge-Hx mode is dominant
(see Table 1) A 5L exposure of hydrogen induced a lxl
pattern. The HREELS spectrum (Fig. 34(b)) obtained from this
surface shows little change of two peak positions (i.e. 246
meV and 71 meV) The loss due to Ge-Hx is also dominant at
this surface. For a 2.5L exposure of deuterium on
Ge0 5Si0 5(111)-(5x5) surface, the LEED pattern is not changed.
The HREELS spectrum is shown in Fig. 35(a). Compared to
hydrogen adsorption the contamination level due to hydrocarbon
species is guite reduced. The loss at 106meV is due to the
carbide. Two peaks at 50.3meV and 177meV are due to bending
and stretching modes of Ge-Dx (see Table 1). Further exposure
of deuterium up to 5L induced lxl LEED pattern. The same
177meV peak and three small peaks between 52meV-75meV are
shown in Fig. 35(b) taken from this surface. Still Ge-Dx is
the dominant peak and small peaks between 50-80 meV indicate
mixture of Ge-Dx and Si-Dx. From these results, we can
conclude that Ge0 5Si0 5(111)-(5x5) surface has a dominant Ge
atom dangling bond in comparison to an Si atom dangling bond.
There might be a guestion on this conclusion. Since the
annealed surface has many Ge islands, the reacted dangling
bond might be at the Ge-island not at Ge0 5Si0 5(111) (5x5)


32
can be clearly identified at 35eV) and HREELS shows a very
broad elastic peak, which is a typical result for the impurity
free surface. The preparation procedure takes 3-4 hours.
Preserving the sample, prepared by Shiraki method, in D-I
water for 20 days caused further oxidation of the Si (111)
surface (refer chapter 5).
In GaAs grown by MBE, elemental As covers the final layer
to protect it from contamination while transporting the sample
from one chamber to another chamber. Since elemental As
covered the GaAs MBE layer, the oxide As406 was formed in the
air environment. After introduction into the UHV, annealing
at -400C removed the volatile top oxide layer resulting in
surface where the carrier is not depleted. The critical point
in preparation of As capped surfaces is complete coverage with
elemental As. If some portion is exposed, then air
contamination will induce strong oxidation of the GaAs
surface. Subsequently a low temperature (~400C) anneal is
not enough to get rid of the surface oxide on the GaAs layer.
Temperature measurement is critical in observing
desorption temperature of adsorbed gases during the sample
preparation. The temperature is measured by a Chrome1-Alumel
thermocouple which covers from -100 K to 1700 K, and infrared-
optical pyrometer which covers 350C to 1500C. The operation
of thermocouple is the current due to work function difference
of two different metals. When the contact becomes hot,
electrons are always transferred from the lower work function


109
peaks upon the strength of elastic low energy electron
diffraction (LEED) intensities for Si(111)7x7 surfaces has
been recently reported by Daum, Ibach and Muller for the
sample with n-type bulk doping and resistivity of 400-800
n-cm [50]. However these authors apparently studied only
specular scattering and high resistivity sample.
Previous studies of polar-semiconductors have tended to
emphasize polar surface phonons, i.e. the Fuchs-Kliewer (FK)
modes, observed in a specular geometry as well as surface
plasmon modes from cleaved GaAs(llO) and InSb(llO) surfaces
[51,52]. Besides cleaved surfaces, As-capped MBE grown
GaAs(100) surfaces also exhibit surface plasmon at specular
geometry [53]. Also for n-type GaAs(100) surfaces after
sputtering and annealing the free-carrier concentration was
found to be compensated by acceptor defects in the near
surface region introduced by the sputtering procedure [54].
In addition temperature-dependent broadening of the
guasielastic scattered electron peak was interpreted as due
to excitation of an unresolved surface plasmon mode with an
effective carrier concentration that differs significantly
from that of the bulk. The surface plasmon mode will be
resolved at off-specular geometry in the same way as Si(111)
surfaces. The effective carrier concentration deduced from
this plasmon loss peak will show the sputtering effect on
semiinsulating and highly doped, p-type semiconductor
surfaces.


105
Q Q,,
*t>
^ V ^
x x
**> O *0 Q,
V Q^ *D
t\, t 0**0
o-f 0- 04- O-
04- O- 04-
O- 04 O- 04-
O- 0-4 O-
04- O 04- O-
04- O- 04
0- 04- O- 04
/>
j> f o*
J> f />
* j> J f
S1mode
+ : up
S mode
4
S6 mode
down


108
from specular to nonspecular incidence angles are presented.
Outside of the specular scattering limit ( A >1.3 ) the
intensity of the elastic scattering peak in high resolution
electron energy loss spectroscopy (HREELS) is drastically
reduced and the loss peak becomes distinguishable from the
elastic peak. As the parallel momentum transfer (i.e., angle
from specular scattering A6 ) is changed further, the loss
peak shifts in a characteristic way which allows one to
identify peaks as plasmon-like or phonon-like. Such HREELS
dispersion effects have not been previously reported for
Si(lll) or GaAs(lOO) surfaces.
With good resolution (AE<5 meV), it is possible to
separate the plasmon peak from the elastic peak in specular
scattering. The effective near-surface space charge carrier
concentration of Si(111) can be estimated using the energy of
the surface-plasmon loss peak. If the effective near-surface
space charge carrier concentration is larger than 10'7 cm3,
the plasmon peak should be separated from the elastic peak
with the resolution of our spectrometer (AE5 meV). In this
chapter the bulk doping density and the annealing temperature
of several samples are varied independently. For 10,s to 1016
cm'3 boron doped silicon, high temperature annealing under
ultrahigh vacuum caused the surface carrier concentration to
increase at the surface [47-49] to ~6xl06 cm'3 and resulted in
a surface plasmon loss peak distinguishable from the elastic
peak (hwD6meV) A similar dependence of elastic and loss


112
After this experiment the sample was sputter cleaned and
annealed at a higher temperature of 950C for 5min. A
stronger 7x7 pattern appeared during LEED observations and
AES intensity ratio from this surface was Si(92)/C(273)~100.
The specular beam had a full width at half maximum (FWHM) of
lOmeV and a loss peak can be seen at 26meV to 30meV in Fia.
14. As the angle was changed from specular to A0=4.4 off
specular the apparent loss energy increased from 26 meV to
30meV then decreased, but this is probably due to the overlap
with the elastic peak that changes intensity in this range.
The absolute intensity of the loss was maximum when the
geometry was just off specular ( A0 2 ) The peak intensity
decreases as in previous measurements assigned to dipole
scattering [55]. The intensity ratio of elastic peak to loss
peak of Fig. 14 was much larger than Fig. 13. However, it is
likely that the increased ordering produced surface states in
the band gap which depleted the surface space-charge region
of the carriers implied by the data in Fig. 13.
Further sputter cleaning and higher annealing temperature
(up to 1250 C) gave results similar to Fig. 14. A sharp,
intense 7x7 pattern was found by LEED and AES showed no
detectable carbon or oxygen peaks [intensity relative to
Si(92)<2x103 ]. However, the annealing at higher temperature
shifted the apparent maximum of the energy-loss peak from
30meV to 24meV. For higher doping (Nb =5.8xl0,s cm'3 ) density
samples, annealed up to 1000 C, the loss peak at maximum


119
(FK) loss peaks of phonon mode. The full width at half
maximum (FWHM) of quasi-elastic peak is lOmeV and the FK peak
is well resolved. Weaker peaks due to surface plasmons are
detected at slightly different energies due to different
densities of free carriers near the surface space charge layer
introduced by the different surface cleaning procedures. Such
plasmon peaks ultimately broadened the quasielastic loss peak
and as reported previously by others [45,46,54]. Angle-
resolved HREELS data obtained from specular geometry to 6.0
off-specular geometry are shown in Fig. 17 from another semi-
insulating GaAs(lOO) surface.
A degreased, Zn-doped, and p-type GaAs(lOO) with bulk
carrier density Nb =1.4x10" cm"3 was cleaned by sputtering (Ne
ions at 500eV and 12nA for 60min) and annealing at 490 C for
11 minutes. The LEED pattern from this sample showed a lxl
periodicity and AES showed no carbon contamination except a
small oxygen peak (-0.05 monolayer). After several sputtering
and annealing cycles a LEED pattern showing 6x1 periodicity
was obtained. For these heavily doped samples the plasmon
peak appears near 25meV and is resolved more clearly than that
of semi-insulating sample at off-angles from the specular
scattering geometry since the intensity of the elastic peak
is greatly reduced. Data shown in Fig. 18 are the HREELS data
obtained at incident angles from specular to 3 off-specular
scattering geometry. In Fig. 19. the dispersion for this loss
is shown. An incident electron energy, 14 eV, and the chosen


130
geometry. This plasmon is also due to acceptor carrier
density due to sputtering and annealing cycle. This result,
which shows the increased acceptor density near the surface
due to sputter cleaning procedures, matches well with the
result of Dubois et.al. on n-type surface which shows
compensation of carriers at the surface due to acceptor-like
sputtering defects [54]. Previously our result of p-type
Si(111) surface also shows increased carrier density after
sputter-cleaning.
4.4. Summary
The dispersion of plasmon-loss peaks measured with angle
resolved electron energy-loss spectroscopy gives an evidence
of increased surface carrier concentration from B-doped (10'5 -
10'6 cm'3) Si (111) surfaces which have been sputtered and
annealed surfaces at temperatures between 800 C and 1250 C.
The plasmon loss dispersion curve (AE vs.Ak,|) agreed with the
calculated plasmon energy and cutoff wave vector for an
effective carrier concentration about 10-50 times greater
than the bulk doping density. Annealing at temperatures,
T>950 C reduced the surface plasmon peak presumably by
annealing defects, and allowed detection of a surface phonon
loss peak at 22meV to 30meV probably associated with the
adatoms of the 7x7 reconstructed surfaces. These losses


40
intensity. In tandem capacitor systems, slits of both
capacitors were designed to locate at the corresponding main
radii. Electrons entering the sector field at a distance y0
from the main path with a velocity deviation f3 will leave the
sector field at y, (=-yo+2ro0) [20]. Since /3 has to be the same
value for both capacitors for different r, all slits should
be located on the corresponding main radii. So the expected
trajectory of tandem capacitors is the main path which is
parallel to capacitor plate.
2.6.2. Resolution and Sweeping Mode
For the deflection analyzers, the base resolution may be
expressed as a function of geometrical parameters in the
general form as follows [21]:
AEb/E0 = A AS + B a" + C /32 (2-16)
where A, B and C are constants, a(maximum angular divergence)
and p( the mean slit height) are the semiangular apertures
and AS is the aperture or slit width at the entrance and exit.
The second term (4/3)a" causes poor resolution when
targets enlarge the angular distribution of the reflected
beam. For further development of the resolution, defining
y,, as the radial deviation of the electron trajectory
measured by dropping a perpendicular down to the central path,


CHAPTER 6
VIBRATIONAL MODES FROM OXIDE LAYERS ON Si(111) SURFACE
6.1. Overview and Motivation
Molecular oxygen is a very reactive gas phase species
and forms strongly bound chemisorbed layers with most metal
and semiconductor materials. With different experimental
conditions different oxide phases can be formed and it is also
possible to form different oxide layers by using H20 instead
of Oj or by using mixtures of H20 and 02. When oxygen is
chemisorbed the surface atoms transfer outer shell electrons
from other species and oxide bonding shows a strong ionic
character. For different bonding phases the dipole moment of
this oxide bond will be changed and the magnitude of the
dipole moment is related to the ionicity per unit cell. From
the lattice dynamical point of view, the surface optical
phonons penetrate deeply into the crystal as the wave vector
Qi|-*-0 Since these modes have displacement fields that
penetrate deeply into the crystal as Q||->0, the electric field
in the vacuum above the ionic crystal becomes very strong.
Therefore an oxide is a promising candidate whose surface
structure can be detected by high resolution electron energy
151


CHAPTER 8
CONCLUSIONS
8.1. Conclusions about Present Experimental Results
It has been observed that semiconductor surfaces have
various states corresponding to various cleaning procedures.
First, the normal UHV cleaning procedure such as sputtering
and post-annealing, induces surface defects which act as an
acceptor-type carrier on the p-type semiconductor surfaces
such as Si(111), GaAs(lOO) and semi-insulating GaAs(lOO).
This effective carrier density was determined by the angle-
resolved HREELS technique which measured the dispersion
relation of the surface free-carrier plasmon. The intrinsic
bulk dopant density is not the major contribution to this
plasmon since band bending by the Fermi-level pinning depletes
the bulk carriers. It can be concluded that sputter-annealing
induces a p-type surface layer in the space charge region on
the semiconductor surface. By annealing at higher
temperature, the defects diffuse into the bulk, the near
surface carrier density is decreased and the plasmon is no
longer detected. Instead, another vibrational mode which
209


219
jacket and titanium sublimation pump are appropriate pumps for
temporary use during evaporation. Along with pumps, an ion
gauge for independent measurement of pressure and mass
spectrometer for analysis of residual gas inside evaporation
section have been installed.
A.3. Quantitative AES Analysis Using Standards
This procedure mainly follow in the PHI AES handbook
[117]. Usually clean silver target is used as a standard.
First, the relative sensitivity, Sx, between the element x and
silver standard.
SX(EP) = [ (A+B) /A] IXM/ (Kx IAgH) (A-l)
where A, B are ideal composition of chemical formula, Kx is
the scale factor. IXH and IAgH are peak to peak amplitude of
each peak in standard for element x and silver. The atomic
concentration of element x is
C* Ix/ (I*g S, Dx) (A2)
where Ix, IAg are the peak to peak heights from the spectrum
of element x and silver target and Dx is relative scale factor
between the spectra for test specimen and silver. Dx is
multiplication of ratios of lock-in amplifier sensitivity,
modulation energy and primary beam current setting. It can


51
done after turning off the ion-pump or isolating the chamber
by the butterfly valve. Oxygen molecules are inserted through
a leak valve up to a higher than calculated pressure. Quickly
opening the turbo-molecular pump line, the exact exposing
pressure should be adjusted by the leak valve. After
exposure, closing the oxygen line should be followed by
turning off heating power. After the turbo-molecular pump
reached its saturation level (in 2-3 min), ion-pump should be
used for normal operation. The reason for using a turbo
molecular pump during oxidation is to prevent hydrocarbon
species from the ion-pump from adsorbing on the substrate and
to preserve the pumping efficiency of the ion-pump by avoiding
any high pressure exposure.
Oxidation of the substrate held at room temperature in
vacuum can not be done by oxygen exposure alone. Positioning
the substrate in front of a ion-gauge during exposure of
oxygen molecules, helps to dissociate oxygen molecule into
oxygen atoms which are reactive on the substrate. Even if all
other procedures are quite similar to thermal oxidation,
oxidation at room temperature is not as successful as thermal
oxidation if surface impurities such as hydrocarbon species,
hydrogen or hydroxyl group occupy reactive sites on the
surface. Increasing the oxide thickness is limited by the
cleanness of the starting surface.
Instead of an ion-gun, a tungsten filament can be used
to dissociate hydrogen molecules into hydrogen atoms. On


Intensity
172
Energy (meV)
Figure 30. HREELS spectra obtained from thermally grown
oxide after annealing at 1100 K.


47 SEM photograph of 1000 of Ge/Si alloy film. Dark
plateau is area 1 and bright islands are area 2.
250
48 Sputter-AES of 1000 of Ge/Si alloy film.
(a) AES spectra obtained from area 1 as introduced;
(b) AES spectra obtained from area 2 as introduced;
(c) AES spectra obtained from area 1 after 30 sec.
sputtering; (d) AES spectra obtained from area 2
after 30sec. sputtering.
251
49 Sputter-AES of thin (~10A) Ge/Si alloy film.
(a) Sputtered edge profile of thin pure Ge film;
(b) Sputtered edge profile of thin Ge/Si alloy film.
255
50 SEM and sputter-AES of thick (-200)
Ge0 5Si0 5/Si (111) 5x5 film, (a) SEM photograph with
white islands;(b) AES line scan across one of white
islands; (c) SEM photograph after point edge
sputtering, (d) AES line scan across the sputtered
edge, (e) AES depth profile from another area.
257
xi


100
plot the pole of two layer spectral density S2 (QM,w), which is
et(w)=-l from equation (3-51), in order to justify these modes
as real surface loss. Dispersion curves combined with
dielectric function, es(w) of the surface layer were shown in
Fig. 11 [33], Since the dielectric function is negative
between w=wT and w=wL, the surface wave will not propagate into
the medium, but will be reflected at the boundary. As Q()->0,
the modes become longitudinal bulk mode and transverse bulk
mode of the surface layer. For Q,(d 1, these modes become two
surface modes at each end of surface layer. In the finite
Qnd, the electric fields of the two branch were shown in Fig.
11 [33]. The branch related to wT represents a polarization
parallel to surface, which is forbidden in dipole selection
rule. The branch related to wL is polarized perpendicular to
the surface. This indicates that only the excitation related
to wL will be detected.
In the two layer model, in the limit of Qndl, two modes,
namely one is due to the bulk-layer interface under the
surface layer and the other is due to the longitudinal surface
layer mode, can be detected. In the limit of Q,|dl, only the
surface mode of the surface layer can be detected by the limit
of detection depth. There are a few limits in this two layer
model. Since the surface is not ideally terminated bulk, it
always has sources of imperfection which results in a
different dielectric function from the bulk, which is smoothly
varying. Strictly speaking, it is very difficult to find


(a) (b)
Figure 44. Ge evaporation control using AES.
(a) Normalized Si (92eV) intensity versus Ge coverage
by Gossmann et at.[102]; (b) Normalized Si(92eV)
intensity versus Ge evaporation time to calibrate
evaporation rate.
245


89
e(0)+l
Us = Mto [ ]* (3-44)
600 +1
where the damping function r(u) is assumed very small. The
Fuchs-Kliewer mode us is larger than the bulk transverse
optical phonon wT0 and smaller than the bulk longitudinal
optical phonon ul0 (=wT0[ &x/e (0) ]*) .
3.4.2. Surface Plasmon Loss and Dispersion: Relation of
Homopolar Semiconductors, and Plasmon-Phonon Coupling
of Polar Semiconductors
In the infrared regime, there are three elementary
excitations such as infrared active optical phonon in ionic
crystal lattice, free carrier plasmon in semiconductor, and
direct interband transition in small gap material. Among
them, a plasmon is due to a collective motion of charge
carriers. The frequency, wp depends upon the number of
effective carrier. If carriers are loosely bound valence band
electrons, the surface plasmon energy is several electron
volts, which can be detected in low resolution electron energy
loss spectroscopy. But, if the carrier concentration is due
to free carriers doped in a semiconductor whose number density
ne(( 10,5-10/cm3, the plasmon energy varies from a few meV to
-lOOmeV which can be detected by HREELS.
Polar-semiconductors (e.q. GaAs, NiO, ZnO, ...) have a
strong surface optical phonon mode due to the ionic character
of each unit cell coupled to plasmon mode while homopolar


27
Eb(i)= Eb(j) + Eb (k) + (Ek + <*>s ) ,
(2-6)
where Ek is the kinetic energy of the ejected Auger electron.
Energy required to produce the two holes in the levels j ',
1 k', namely Eb(j,k), is suggested as
E( j ,k) = Eb(j) + Eb (k)
= Eb(j) + Eb(k) + F(j ,k;x)
(2-8)
where F(j,k;x) describes the interaction (namely repulsive
between holes) energy of the two holes. Shirley has pointed
out that including the relaxation of other electrons in and
around the atom (i.e. more screening) that takes place in the
process (2-4), Eb(j,k) can be expressed as follows,
Eb(j,k) = Eb (j ) + Eb(k) + F (j k;x) R(x)
(2-9)
which gives excellent agreement over the entire range of
atomic number. In CW and CCV case, the final state includes
valence orbitals. According to the localization of the final
state valence bond, the interaction term should be considered.
It is not well known yet about the line shape estimation of
CW or CCV. Both energy and line shape change with chemical
environment for Auger transitions having final states made up
of valence orbitals involved in bonding, namely, CW and CCV
processes. Both chemical shifts and line shape variation have


Energy (meV)
(a)
Figure 28. HREELS spectra obtained
substrate.(a) 10L ; (b)
Energy (meV)
(b)
from thermal oxide grown on a 700 K Si (111)
100L ; (c) 1000L ; (d) 10 kL exposure.
169


189
contamination since the sample was prepared outside the
chamber. Also they reported Ge enrichment at the surface due
to high temperature annealing and water adsorption is quite
similar to Si(100) (dissociative chemisorption of H and OH
species).
At this point we want to suggest the problems to
consider. First, since Si0 5Ge0 5(111) surface has also adatoms
like Si(111)-7x7 surface, small amounts of hydrogen titration
on the alloy film surface may enable identification of adatom
species on this surface through only reaction with adatoms.
Vibrational energy loss in HREELS is inversely proportional
to me,,*. The adatom and hydrogen atom will cause a strong
dipole moment vertical to the surface which can be easily
detected at specular geometry. Deuterium instead of hydrogen
can confirm the adatom species since deuterium has similar
chemical properties to hydrogen atom. Second, since the
x-ray diffraction data of Bean et al. is so small (namely 0%
lateral relaxation at 100 A and 1.2% lateral relaxation at
500 A) it is necessary to check where this relaxation has
started and whether it is smooth or abrupt through slow
deposition of a Ge film [101]. Third, since the
GexSi,_x(111) 5x5 pattern starts to show at the same point as
the beginning of the Ge-rich island growth (2-3ML) [94-105],
and Ge island also grows epitaxially, the origin of this 5x5
pattern is uncertain, i.e. whether 5x5 pattern comes from Ge-
islands or from other area between Ge-islands or both.


65
the diffracted intensity now satisfies only two of the three
Laue conditions, Sa=27rl and Sb=27rm, and is confined along
lines in reciprocal space normal to the crystal surface,
specified by S = G,. The reciprocal lattice and Ewald
construction for a two dimensional lattice is shown in Fia.
8(b). The lines are referred to as reciprocal-lattice rods
and indexed by two integers (lm). These results which depend
upon two dimensional periodicity of the surface are the basis
for using LEED to determine surface structure. The modulation
of the interference intensity function along the reciprocal
lattice rod is determined by the factor [l+a2-2a cosS-c]'1 .
The broad maxima exist at positions where the third Laue
condition, 8c=27rn, is satisfied. The breadth of these peaks
is a consequence of attenuation. In the case of zero
penetration, intensity depends only on the scattering factor
|f(*,E) |2 .
In addition to the attenuation factor, incident electrons
experience a different potential while passing through the
crystal. This periodic potential in the crystal half-space
can be expressed as a Fourier expansion
V(r)=E Vc exp(iG-r), (3-11)
and its spatial average V0 is usually called the inner
potential. Due to this inner potential V0 the magnitude of
wavevector in the vacuum is still k= 27r(E/150.4)>' A"' but in


1500
4->
(D
C
CD
4-J
C
II
x300 FWHM13.4 meV
Ta900C
-100 0 100 200 300 400 500
Energy (meV)
(c)
Figure 26.continued.
164


11
considered. In the second section, sputtered and annealed
GaAs(lOO) surface studies are considered. Also sputter
cleaning effects are discussed in general. In this chapter,
angle-resolved HREELS results show dispersion relations of
phonon inodes, plasmon inodes and coupled phonon-plasmon inodes.
Chapter 5 is the second experimental chapter. In this
chapter studies of the room temperature oxidation of Ni(lll)
and Ni(110) are presented. The origin of a high binding
energy oxygen species detected in x-ray photoelectron
spectroscopy (XPS) data from the coalesced oxide layer is
considered. A new interpretation of thin NiO layers is
postulated based upon HREELS data.
Chapter 6 is the third experimental measurement chapter.
In this chapter studies of oxide layers on Si(111) surfaces
are presented. In the first part, studies of a Shiraki oxide,
a deionized-water preserved oxide and a native oxide formed
in air on Si(111) surfaces are discussed. The effect on the
clean surface of removing oxide layers by annealing is
considered. In the second part, thermal UHV oxidation studies
of Si(111) surfaces are presented. Differences in thin oxides
are clearly evident in our HREELS results which indicate that
the best thin oxide layers are produced by the Shiraki
chemical etching method.
Chapter 7 is the fourth experimental measurement chapter.
This chapter includes the film growth, characterization,
relaxation measurement and excitation measurement through a


500.0Mm
5/05/88 3.0kV
i 35X 3GeSi
Figure 50.continued.


for scanning Auger experiments, Paul Lyman for Shiraki oxide
samples, William Wresh for Rutherford Back Scattering
experiments, Chris Dykstal for Ni experiments, Larry Phelps
and his colleagues for electrical equipment, and Harvey
Nachtrieb and his colleagues for expert machining.
A special thank you is directed to Mrs. Glenda Smith for
her help in the Physics Department. I appreciate the many
discussions and advice on printing problems that Eric Lambers
and Kelly Truman contributed to this dissertation.
My final thank you goes to my family members, my parents
who allowed, encouraged and financially supported me in order
to achieve this goal and my wife and our two daughters,
Eunkyung, Hyosuk and Yesuk, who sacrificed much of their time
and energy in completion of this goal.
iii


152
loss spectroscopy (HREELS). For example, ZnO was the first
successful example of an oxide which was studied using HREELS
[86]. It has been reported that 60% of the electrons which
emerges from the crystal are contained in the one-phonon loss
peak at 69meV. Thick Si02 (>500) thermally grown on Si
substrate in the air is a non-crystalline oxide layer. Thiry
et al. reported HREELS data obtained from thick Si02 (950)
grown on Si(100). Since this thickness is larger than the
detection limit of HREELS (-200), they must have detected a
silicon dioxide layer near the vacuum-solid interface not the
solid-solid interface [87]. They shows that dielectric
function theory can be applied for this thick and homogeneous
silicon dioxide layer; but, as the thickness of the oxide
layer is reduced, the dielectric function theory can not
explain the evolution of three peaks related to the oxygen
species. This might be due to the fact that the dielectric
function theory is too simple to explain a complicated
interfacial layer or an inhomogeneous layer. In this chapter
we will investigate the SiOx/Si(lll) interface by preparation
of thin layers of oxygen-related species (X^intermediate
oxide) on Si(111) surfaces. We have investigated four
differently prepared intermediate oxide layers such as Shiraki
wet-chemical intermediate oxide (Shiraki oxide), deionized
water-preserved intermediate oxide (water oxide), air exposed
native intermediate oxide (native oxide) and thermally grown


228
electrode '3' and electrode '4' whose settings are also
recorded (s2nd setting). Returning to the elastic peak,
electrode '3' and electrode '4' values are reset using the
first setting values while zoom knobs are set to zero.
Sweeping again to the chosen loss peak (e.g. 450meV), the
second setting values of electrode *3' and electrode '4' are
recovered using the corresponding zoom knobs. Then the energy
window can be swept from the elastic peak. Through zooming
operation the maximum intensity of the elastic peak reduces,
but the total intensity are distributed to the whole range of
the spectrum.
There is no absolute criteria for determining of good
tuning. But typical examples of 'poor' tuning and 'good'
tuning will be shown in Table 2. First the difference between
them is in the 2nd capacitor of analyzer. Since the
difference between rsa and Rsa is 0.225eV which is much smaller
than other capacitor differences such as 0.527eV, 0.683eV,
0.402eV, it is impossible to increase to those values with the
same polarity at both side of plates. The graph of plate
voltage (r,*,, R,., r_, R,*, r, R*,, rsa, Rsa) versus capacitor
voltages (A,,., A,., Au,Asa) is shown in Fig. 42. All capacitors
except second analyzer have the right sign of slope compared
to the slope of standard in the manual. Second the zoom is
at the elastic peak in 'poor' tuning case. It is hard to
detect high energy losses. Especially in detecting adsorbed
impurities such as hydrocarbon species (~360meV) or hydroxyl


154
by itself is an oxygen species. Therefore the wafer will be
oxidized by OH in the water. Since the oxidation by OH always
accompanies hydrogen species, the mechanism and the
intermediate oxide layer produced by the hydride and hydroxyl
group environment may be different from other intermediate
oxide layers.
When a clean Si(111) surface is exposed to air, reactive
species in the air will adsorb on the bare silicon substrate.
Even though oxygen molecules are the dominant reactive species
additional impurities will passivate the surface and oxidation
will stop. Therefore a native oxide is a very thin
intermediate oxide layer formed under a mixture of gases in
the air such as nitrogen, oxygen, water, hydrocarbon as well
as rare gases. Investigating this native oxide will also give
information on air contamination effects. Detailed
information on surface impurities supplied by HREELS with a
0.001ML detection limit might help to improve the vacuum
cleaning techniques. In detecting impurities from silicon
surfaces UHV analytical techniques such as low energy electron
diffraction (LEED) and Auger electron spectroscopy (AES) are
limited since LEED patterns for different amounts of carbon
are quite similar and AES can not detect a hydrogen species.
The combination of HREELS with these techniques is helpful to
a comprehensive understanding of surface impurities.
Thermal oxidation has an advantage over the protective
oxide layers discussed above since the initial stages of


55
materials (e.g. Ni) will not make a uniform film on a quartz
glass substrate and uneven island formations make it difficult
to determine the average film thickness. In this case a glass
substrate is replaced by another material such as Si which
allows the material to grow as a uniform film.
Silicon and germanium are materials which make thin
relatively smooth films on quartz glass to determine the
evaporation rate. Once the emission current is determined
the total evaporation time determined the actual evaporated
film thickness after enough warming up the source. Finally
details of design of evaporation system will be introduced in
Appendix A.2.


243
'Surface polariton' due to electromagnetic theory were named
according to the way of description of these surface mode.
Both of them ignored the retardation effect since Q,|ws/c.
The frequency-dependent dielectric function can contain more
terms to describe other excitations besides optical surface
phonon, e.q. surface plasmon due to free carriers.


m m.
10
AES Int ensit y
Ratio of 5
S i ( 92) /Ge( 52)
0








i
Annealing
Temperat ure
100 200 300 400 500 600 700 (C)
LEED Pattern
No LEED pattern
lxl
5x5
7x7
Surf ace
Compos ition
Ge Amorphous Film
Ge S i,
* l-x
x> 0.5
Ge Si
* l-x
x~0.5
Clean
Si( 111)
Figure 45. Thermal evolution of evaporated Ge film on
Si(111).
247


APPENDIX A
FURTHER EXPERIMENTAL DETAILS
A.1. Gas Handling System for UF HREELS System
Gas lines are connected to UHV chamber through the leak
valve to control the rate of leaking. Two separate branches
are used for sputtering gases such as Ne and Ar, and a
reactive line for oxygen, hydrogen, deuterium and ammonia.
The schematic diagram is shown in Fig. 38. For rare gas lines
like Ar and Ne, a LN2 cold trap can be used just before
opening the valve to condense the impurity gases which may
have mixed with rare gases due to a leak or desorption from
the line. But the gas in the line should be replaced by the
fresh gas just before using if the gas has been in the line
longer than 1 week. Two gas lines are used separately to
avoid mixing sputtering gas with reactive gas.
A.2. Design of Evaporation System
The evaporation system consists of a commercial electron
gun with five crucibles (Thermionics laboratory, model 100-
0050) collimators with shutters and a titanium sublimation
pump with a liquid nitrogen shroud. The design is shown in
Fig. 39. When the filament is heated by a power supply,
electrons emitted from the filament which is positioned under
214


187
SEM photographs with the Ge islands in diameter -0.2 nm with
an effective film thickness of 25 A grown at a substrate
temperature 580C [96], The low energy electron diffraction
(LEED) pattern of the Si0 5Ge0, alloy film grown on Si (111)
substrate is 5x5 [99,102,103,107]. After deposition of 1.9ML,
Si(111)-7x7 LEED pattern become diffused (i.e. a 7x7 pattern
mixed with a 5x5 pattern), then above 1.9ML Ge coverage the
LEED pattern changed to 5x5, and then further annealing at a
high temperature (>770C) without additional Ge induces a 7x7
LEED pattern. One point of interest is this 5x5 LEED pattern
which has a similar structural analogy to the 7x7 LEED pattern
of Si(111) [107]. First, intense fractional order spots along
the lines joining neighboring integer-order spots forming a
6-pointed star formation centered on each integer order spot.
This star formations has been attributed to the shape
transformations of triangular subunits of the unit mesh. Off-
star fractional order spots on the perimeters of a hexagon of
side 2/5, centered on some integer order spot, and its six
fold rotational symmetric spots has relatively high
intensities. The intense (3/7, 4/7) spots are due to dimers
lying along the perimeter of the triangular submits. Recently
Becker, Swartzentruber and Vickers monitored the
SiGe,_x(lll) 5x5 and showed scanning tunneling microscopy (STM)
images which have six adatoms in a unit mesh, which is
analogous to the 12 adatoms on an Si(lll)7x7 surface [108].


41
the equation of the electron trajectory up to second order in
a and using 127 condition is [22]
y, = -y0 + (AE/E0) r0 -(4/3)r0a2 (2-17)
where y0 is the deviation from the central path. With S being
the slit width, the fastest electron that can pass through is
that which enters the capacitor at S/2 with aIHX and leaves at
S/2. Its relative energetic deviation is
AE7E = S/r0 + (4/3) a^2 (2-18)
The slowest electron enters at -S/r0 with a=0 and leaves at
-S/r0 with q=0. Its relative energetic deviation is AE'/E =
-S/r0. The total energy width within the plane of deflection
is
AEf/E = AE*/E AE'/E = 2S/r0 + (4/3) 0tmx3 (2-19)
For an electron which has a velocity component
perpendicular to the plane of deflection (z-direction),
v = vz + v0 (2-20)
where v,
is perpendicular component of velocity v. Since


REFERENCES
1. S.P. Weeks, J.E. Rowe, S.B. Christman and E.E. Chaban,
Rev.Sci.Instrum. 50(10), 1249(1979).
2. J.F. O'Hanlon, "A User's Guide to Vacuum Technology, "John
Wiley & Sons, New York, 1980.
3. C.R. Brundle, J.Vac.Sci.Technol. 11, 212(1974).
4. P. Staib, J. Kirschner, Appl.Physics, 3, 421(1974).
5. K. Siegbahn, C. Nordling, A. Fahlman, R. Nordberg, K.
Hamrin, J. Hedman, G. Johannson, T. Belgmark, s.
Karlsson, I. Lindgren and B. Lindberg,"ESCA Applied to
Free Molecules," North-Holland Publ., Amsterdam, 1969.
6. D.A. Shirley, Chem.Phys.Lett. 17, 312(1972); F.J.
Szalkowski and G. A. Somorjai, J.Chem.Phys. 56,
6097(1972), and J.T. Grant and T.W. Hass, Surf.Sci. 24,
332(1971).
7. S.P. Kowalczyk, L. Ley, F.R. McFeely, R.A. Poliak and
D.A.Shirley, Phys.Rev.B. 9, 381(1974).
8. P.H. Holloway and J.B. Hudson, Surf.Sci. 43, 123,
141(1974) .
9. P.H. Holloway and G.E. McGuire, Thin Solid Films, 53,
3 (1978) .
10. A. Ishizaka and Y. Shiraki, J.Electronchem.Soc.,
133,666(1986).
11. Z.J. Gray-Grychowski, R.G. Egdell, B.A. Joyce, R.A.
Stradling and K. Woodbridge, Surf.Sci. 186, 482(1987).
12. M. Prutton, "Surface Physics," Clarendon Press, Oxford,
1975.
13. M.Tabe, Jpn.J.Appl.Phys. 21, 534(1982).
14. J.C. Bean, G.E. Becker, P.M. Petroff and T.E.
Siedel,J.Appl.Phys. 48, 907(1977), and J.C. Bean, G.A.
Razgonyl, Appl.Phys.Lett. 41, 752(1982).
262


35
electrons to reach the display screen. Usually the first and
the fourth grid are kept at the same potential as the sample
(i.e. grounded) to help shield the electron trajectories. Two
inner grids are operated near filament potential to select
almost elastically scattered electrons. The screen voltage
(i.e. collector bias) is operated at 2-5kV to provide
sufficient energy to excite the phosphor to reduce stray
magnetic and electrostatic fields which can cause loss or
distortion of LEED pattern. Insulators near the sample are
shielded with copper plate. After obtaining a LEED pattern,
digital LEED measurement using a vidicon camera (details in
next section) or photo graphs are taken from the back-side
view port. Photograph was taken using a commercial camera
attached with zoom lens to focus at the screen. Normal 100ASA
black and white film was used with manual exposure of 10-12
seconds.
2.6. High Resolution Electron Energy Loss Spectroscopy
The electron optics of the UF HREELS spectrometer (model
ELS-22 Leybold-Heraeus System) will be discussed in this
section from the point of view of describing the tuning
procedure. It has been emphasized that understanding the
geometry of the spectrometer and the relationship of each
potential adjustment knob determines the resolution. Detailed
tuning procedure is discussed in Appendix A.5.


99
-1
Im
[e,()+l]
-1
Im
[b(M)+l]
+ Im[ (2Q,.d) {
3[-l/(e,(to)+l}]
d (-2Q|,d)
}]
-2Qj|d = 0
-1 esJ (w) ebJ (to)
= Im +(Qnd)-Im[ ] (3-55)
[eb(w)+l] et(u) {6b(w)+l}J
The first term describes loss function due to electron
scattering by electric field fluctuations in the vacuum above
the crystal produced by excitations in the bulk. The second
term, which is proportional to d, describes loss function due
to scattering by electric fields produced by fluctuations
within the surface layer. This second term is divided into
two terms as follows:
eb2 -1 Qd
(Q,|d) Im[ ] + Imes(to) (3-56)
(eb+l)J e,(u) (b+l)*
where eb is assumed to be real and not a function of
frequency. The pole of the first term of equation (3-56) is
the form of the longitudinal bulk phonon mode of the surface
layer (cf. es(to)-+0 at wL) and the pole of the second term of
equation (3-56) is the form of the transverse bulk phonon mode
of the surface layer (cf. es(u)- at uT) It is helpful to


126
equilibrium positions and vibrational frequencies are
different for the partially ordered and well-ordered surfaces.
Daum, Ibach and Muller also reported [50] that the vibrational
electron energy loss spectrum of the Si(111)-7x7 surface
exhibited a well defined loss at 71meV and a structure at 25-
34meV. These features were detected from a 400-800 n-cm, n-
doped Si(111)-7x7 wafer after annealing at 800C without
sputtering. They also assigned these two modes to localized
vibrations involving Si-adatoms and the surface atoms lying
underneath using Ab initio total-energy calculations of the
dynamical matrix. The higher energy mode is due to out-of
phase vibration and the lower energy mode is due to in-phase
vibration. When these results are compared to our case, the
higher mode is not shown in our case since it is a more
localized mode and also screened by doping carriers. This is
easily simulated in the case of adsorbed oxygen on Ni(lll)
[61] where 75% of the spectral density function is due to the
frequency regime below the maximum phonon frequency (i.e.,
0-37meV) and 25% of that is associated with higher-frequency
72meV mode.
A majority of native oxide and hydrogen species from
HREELS data of the introduced semi-insulating GaAs(100) sample
are detected in Fig. 16(a). The FK optical phonon mode at
36meV is shown since this mode is due to a strong dipole
moment at the GaAs(100) surface covered with oxide and
impurities. A peak near 360meV is due to C-Hx stretch mode


240
where e is the background dielectric constant. Under the
assumption of slow variation of u0(l) and E(l), the equation
of motion for u^x), after some algebra combining equation
(B-17) and equation (B-18), is
4wne2
^(x) + [w02 ]u,a(x)
3 Mr 6oo
ne'2
M, £oo
d[ V Uoix')
dx'
3xcJ | x -x' |
(B-19)
This has the solution of the form u0(x,t)=u0(Q)exp(iQx-iwt)
in the bulk. If Uo(Q)iQ (transverse mode), then the right
hand side of equation (B-19) vanishes. Then bulk transverse
optical mode is
wto2= w02-47rne'2 / (3Mr&o) (B-20)
If u0(Q)I Q (i.e. longitudinal mode), then the right hand side
of equation (B-19) becomes (ne2/(Mr&o) ) (-47rufla(x) }. Then bulk
longitudinal optical mode is
= w02 + 87rne'2/(3Mrebo) (B-21)
Considering semi-infinite ionic crystal in the half space
z>0, and taking the divergence of both sides of equation (B-
19), then