THE STUDY OF DEFECTS IN LOW MISFIT GE-SI
STRAINED LAYER HETEROSTRUCTURES
Mark Benedetto Stirpe
A thesis submined in conformity with the requirements for the degree of Master of Applied Science
Graduate Department of Metallurgy and Materials Science University of Toronto
0 Copyright by Mark Benedetto Stirpe 1998
National Library ofcamcia
Bibliothèque nationale du Canada
Acquisitions and Acquisitions et Bibliographie Services services bibliographiques
395 Wellington Street 395, rue Wellingtm OttawaON K 1 A W Ottawa ON K1A ON4 Canada canada
The author has granted a non- L'auteur a accordé une licence non exclusive licence allowing the exclusive permettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, Ioan, distxibute or seii reproduire, prêter, distribuer ou copies of tbis thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfiche/nIm. de
reproduction sur papier ou sur format électronique.
The author retains ownership of the L'auteur conserve la propriété du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation,
The Study of Defmts in Low Misfit Ge-Si Sfrained Layer Heterostructures
Master of Applied Science Mark Benedetto Stirpe
Graduate Department of Metaiiurgy and Materials Science University of Toronto
In order to study the defect structures in strained layer superlattices, a series of
Ge,SiI.,/Si (100) (0.09 c x < 0.13) heterostructures were grown by u l a g h vacuum
chernical vapour deposition (UHV-CVD) and molecular beam epitaxy (MBE) methods.
Activation energies were determined to be consistent with the widely accepted values (Q,,
= 2.5 + 0.5 eV and Q, = 4.2 + 0.5 eV for nucleation of misfit dislocations and overd1
suain relaxation, respectively). The low-temperature MBE growths yielded Q,, = 0.5 k
0.05 eV and Q, = 2.0 f 0.5 eV for nucleation and overall strain relaxation, respectively.
Samples implanted with Si ions displayed significant decreases in nucleation rate and
misfit dislocation densities. Based on the quantitative study conducted on GexSiiJSi
heterostructures, it has been demonstrated that the onset of strain reIaxation via misfit
dislocations can be controlled by point defcct injection via ion implantation.
I would like to thank my supervisor, Professor D. D. Perovic for his advice and
encouragement throughout the course of this thesis. 1 wodd also like to thank Dr. H.
Lafontaine, Dr. J.-M. Baribeau, and Dr. D. C. Houghton at the National Research Council
of Canada (NRCC) for participating in useful discussions and providing excellent
samples for study. Speciai thanks also go out to Dr. R Goldberg, Dr. P. Simpson, and
Dr. 1. Mitchell at the University of Western Ontario for the ion implantations and positron
annihilation spectroscopy.
1 wodd also like to extend my appreciation to Mr. Sd Boccia, Mr. Fred Neub, Ms.
Azita Ariapour, and Mr. Bahi Bahierathan at the University of Toronto for their help with
sample preparation and TEM instruction.
Finaily, 1 am grateful to the N a W Sciences and Engineering Research Councii of
Canada and the University of Toronto Fellowship Award for financial support and the
National Research Council of Canada for travel suppoa through the Visiting Research
Graduate Program.
TABLE OF CONTENTS
Page #
Abstract
Acknowledgements
Table of Contents
List of Figures
List of Tables
Acronyms
Chap ter 1- Introduction
1 - 1 Ge-S i Heterostructures
1.2 Molecular Beam Epitaxy
1 -3 Chernical Vapour Deposition
1.4 Ion Implantation
1.5 Positron b h i l a t i o n Spectroscopy
1 -6 Microscopy Techniques
1.6.1 Nomarski Interference Microscope (NIM)
1 A.2 Field Ernission Scanning Electron Microscope (FE-SEM)
1 -6.3 Transmission Electron Microscope (TEM)
Chap ter 2 - Experimental
2.1 Heterostmcture Growth Conditions
2.2 Ion Implantation Conditions
2.3 Rapid Thermal Annealhg (RTA)
2.4 Nomarski Interference Microscopy (NIM)
2.5 Field Emission Scanning Electron Microscopy (FE-SEM)
2.6 Transmission EIectron Microscopy (TEM)
iii
2.7 Positron Annihilation Conditions
Chapter 3 - Results and Discussion
3.1 Strained Layer Geometry
3 -2 Buik Quantitative Measurements
3.3 Sources of Nucleation
3.4 Ion implantation
3.5 Low-Temperature MBE Heterostnictures
Chapter 4 - Conclusions and Recommendations
Appendix
Effective Stress and Kinetic Mode1
Energy Balance Approach
Bulk Quantitative Raw Data
TRlM Simulation
References
LIST OF FIGURES
Figure 1.1.1 - Heteroj unction Bipolar Transistor
Figure 1.1 -2 - Diagram of alternative modes of epitaxial growth
Figure 1.1.3 - Kinetically Iimited 'critical' thickness curves
Figure 1.1.4 - Schematic diagram of prismatic loops at Ge-rich platelets, half-loops and expansion of lwps
Figure 1 -2.1 - Schematic diagram of MBE process
Figure 1.2.2 - Plot of epitaxiai film morphology as a function of growth temperature and Ge composition for t 000 A Ge&,
Figure 1 -3.1 - Schematic diagram of UHV-CVD process
Figure 1 -4.1 - Photo of typical Ion bearn implantation device
Figure 1.5.1 - Schernatic diagram of positron's reaction in solids
Figure 1 S.2 - (a) Doppler broadening (b) positron lifetimes, and (c) S-parameter
Figure 1.5.3 - Schematic diagram of back reflection geometry in positron beam production
Figure 1.6.1 - Nomarski interference microscope
Figure 1.6.2 - FE-SEM gun configuration
Figure 1.6.3 - Relationship between total emitted e- coefficient and beam energy
Figure 2.1.1 - Czochralski growth method
Figure 2.4.1 - Typicai Nomarsici micrograph
Figure 2.6.1 - Schematic process of TEM cross-section preparation
Figure 2.6.2 - (a) Schematic of difhction pattern of diamond cubic lattice with beam direction z = [O0 11 (b) Actuai TEM image showing two-barn condition g = [400] (c) Schematic of cross-section conditions with z = [O 1 11
Figure 3.1.1 - Geornetncal distribution in strained layer superlattice
Figure 36.1 - Nomarski images for temperatures ranging h m 700 - 1 0OO0C
Figure 3.2.2 - Nucleation rate vs. F' of UHV-CVD and MBE matenal
Figure 3 -2.3 - Misfit dislocation density vs. K" for UHV-CVD and MBE Matenal
Figure 3.2.4 - Nucleation rate vs. K" for UHV-CM as-gmwn and implanted material
Figure 3 .ZS - Nomarski micrograph showing as-grown and ïmplanted sides after RTA at 850°C for 30 s
Figure 3.2.6 - Misfit dislocation density vs. EC1 for CVD-61 as-grown and irnplanted matenal
Figure 3.2.7 - Nucleation rate vs. Kf for UHV-CVD and low-T MBE Material
Figure 3.2.8 - Misfit dislocation density vs. K" for UHV-CVD and low-T MBE matenal
Figure 3.2.9 - Positron annihilation spectroscopy for Iow-T MBE material And CVD-9
Figure 3.3.1 - TEM DF image Ge-rich pïatelets, g = [220]
Figure 3.3.2 - TEM BF image of C M - 9 after RTA for 5 s at 1000°C
- l q - l Figure J .J -3 - TEM DF image of CVD-9 d e r RTA for 5s at 1 OOO°C
Figure 3.4- 1 - TEM BF image of implantation profile in cross-section with z = [O1 11 for RTA CVD-9 at 850°C for 30 s
Figure 3.4.2 - TEM plan-view image of implanted CVD-9 RTA at 850°C for 30 s
Figure 3.5.1 - Morphological instability phase diagram for Ge&30.74Si
Figure 3-52 - FE-SEM SE-detector image showing presence of voids
Figure 3.5.3 - TEM cross-section image revealing cusps wiîh ( 1 1 1 1- oriented facets
Figure 3 -5.4 - Doppler-broadened annihilation line-shape parameter S as a function of positron energy
LIST OF TABLES
Table 2.1.1 (a) Growth parameters of UHV-CVD G&Si 1JSi materiai
Table 2.1.1 (b) Growth parameten of MBE Ge&.,JSi matenai
Table 2.3.1 (b) Rapid Thermal Anneai TemperatundTime parameters
HBT - heterojunction bipolar transistor
IC - integrated circuit
LED - light - emitting diode
MBE - molecular b a r n epitaxy
UHV-CVD - ultrahigh vacuum chemical vapour deposition
TEM - transmission electron microscope
No - inifial misfii source density present in as-grown material
Q. - activation energy for nucleation of misfit dislocations
Q, - activation energy for overall h n relaxation
RTA - rapid thermal anneai(s)
LPCVD - low pressure chemical vapour deposition
sccm - standard cubic centimeten per minute
AES - Auger electron spectroscopy
XRD - X-ray diffraction
PL - photoluminescence
SIMS - secondary ion rnass spectroscopy
LIBI - linear ion beam implantation
y - ray - gamma rays produced in positron annihilation spectroscopy
NIM - Nomarski interference microscopy
FE-SEM - field emission scanning electron microscope
BS - backscattered electrons
SE - secondary electrons
DP - diffraction pattern
z - beam direction
BF - bright field
DF - dark field
SLS - strained layer superlattice
RTCVD - rapid thermal chemical vapour deposition
Er - energy of vacancy formation
viii
E, - energy of vacancy migration
Low-T - Iow-temperature
TRIM - transport of ions in matter
BCA - binary collision approximation
MD - molecular dynarnics
CHAPTER 1 - INTRODUCTION
1.1 Ge-Si Heterostnictures
Within the past decade, techniques for growing crystalline materiais have advanced to
the point whexe germanium can be selectively introduced into silicon. Germanium is
grouped with silicon in the periodic table and possesses the same diamond cubic crystal
structure and the same bonding orbitais (Le. four tetragonal sp3 hybnds). When
germanium is doped to the base region of a silicon bipolar transistor, the result is a Si-Ge
heterojunction bipolar transistor (HBTs)(Figure 1.1 -1). The device properties can be
tailoreci tightly to the needs of the desired application, because the profile and
concentration of the germanium can be wntrolled with great accuracy. Hence the IC
designer's new-found ability to devise hi&-speed, lowast chips using heterostruchired
layers (Cressler, 1 995).
Figure 1.1-1:
Al-CU
n' polysilicon Silicon-Germanium
P+
Schematic d i a m of a heterojunction bipolar transistor (HBT) in an advanced SiGe IC
The HBT is present in a self-aligned SiGe IC. Each HBT is isolated by I .O pm-wide
trenches etched 4.0 - 5.0 pn into the silicon subsîrate and overgrown with a composite of
polysilicon and oxide. Heavily doped d and P* polysilicon layers form the contacts to
the emitter and base legions, respectively. The AI-Cu is used for electrical contacts.
Due to the tremendous advancement in Si-based technology, considerable attention
has been given to the development of strained layer heterostruchues consisting of the
group-IV alloys Si-Ge. The retention of structurai perfection during epitaxial growdi and
thermal processing is crucial for the development of GqSil,/Si seained layer devices
such as HBTs, resonant tunuehg diodes, and Light-emitting diodes (LED). Such
heterostructures are, in generd, metastable and can relax through the injection of misfit
dislocations at the Ge,&,/Si interfaces upon elevated temperature exposure (Houghton,
1991). Misfit dislocation propagation must be suppressed to avoid the shultaneous
formation of threading dislocations, which may penetnite heterojunctions and enhance
current leakage. Successful use of semiconductor heterostructures in electronic and
optoelectronic devices depends strongly on the reduction of dislocation density below 10'
cm-2 (Rajan er al., 1991). The 4.2% lattice mismatch that exists between Si and Ge is
enough to cause this disruption in the crystalline order at the GeSVSi interface.
The concept of strained layer epitaxy was considered fim by Frank and van der
Mewe (1 949) and then M e r developed by Matthews and CO-workers (1972). Epitaxy
can be defined as an oriented overgrowth in which each successive layer tends to assume
the lattice orientation of the layer beneath it (Perovic, 1988). in the case of G ~ s i ~ . ~ / S i
heterosuucnires, the Larger lattice parameter of the epilayer, GeSi, must elastically
cornpress biaxially in the interfacial plane to match the lattice parameter of the substrate
(Le. Si). The misfit strain is then accommodated by a tetragonal distortion of the growllig
epilayer in the growth direction normal to the interface. Figure 1.1.2 shows a diagram of
the pmcess. The formation of misfit dislocations at the interface relieve this strah. The
misfit dislocations, typically 60' al2 4 10> in character, are parallel to the interfaces and
the thrmding dislocations gliding on the { 1 1 1 } -variants are the geometries expected in
partially relaxed heterostructures (Houghton, 199 1).
EpitaxiaI layer
Substrate crystal
(a) unsmined
(b) strained
Figure 1.1.2: Epitaxial growth that gives (a) an unstrained layer with misfit dislocations at the interface and (b) a straïned layer which deforms to match the lanice spacing
There exists a 'critical thickness' for strained layes where coherent dislocation-fke
growth takes place. The primary components for such growth are the control of the Ge
composition and thicknesses of layers. Figure 1.13 shows the kineticaily limited cntical
thickness of previous studies in Gedi i,/Si research at various temperatures. This criticai
thickness detemiines the metastability limiîs for strained layers. Layers grown above the
cntical thickness will nucleate interfacial misfit dislocations. Layea grown below the
critical thic kness will not. However, metastable strained layers can be produced which
are thicker than their equilibrïurn critical thickness. The metastability l imits are
determined kinetically by the availability of thermally activateci nucleation sites for misfit
dislocations and their rate of extension by glide. and by growth parameters such as
temperature and deposition rate (Houghtoh 1995). Growth techniques such as low
temperature rnolecular beam epitavy (MBE) and reduced temperature chemical vapour
deposition (CVD) are utilized to develop these structures. Under proper groowrh
conditions, strained layer epitavy allows one to repeatedly grow Ge,Sii, strained layers
separated by Si layen so that the whole layer sequence maintains the same lanice
parameter as the Si substrate.
10 20 30 40 50 60 Ge concentration (% 1
Figure 1 -3.1 : Kineticall y limited "cri tical" thickness for Ge,Si &i (1 00) growth
In studying strained layer heterostructures, it is also of importance to consider the
structurai limitations of growth techniques; in particular, low-temperature Si-MBE
growth. Growth at lower temperatures is desired for increasing the thicknesses of layers
while still maintahhg coherency among the layers. However, void formation cm be
associated wiîh low temperature MBE (<400°C). A study of void formation has been
conducted and will be discussed.
There have been rnany different treatments of the critical thickness model in the past,
but the treatment of Matthews and CO-workers has proved to be the most popular
(Perovic, 1995). The theory of Matthews et d. (1970) takes into consideration pre-
existing sources of nucleation for misfit dislocations such as substrate threading
dislocations. By takllig this into account, one can relate the equilibrium critical thickness
to the first misfit dislocation that is inîmduced at a misfihg interface (Perovic et al.,
1995).
In order to study the s& relaxation effects in metastable strained layer
heterostnictures, non-equi librium treatments of the pro blem would have to be
investigated. People and Bean (1985) aîtempted to describe the critical thiclmcss in
GeSi/Si using an energy balance model that considered spontaneous homogeneous
interfacial nucleation of misfit dislocations in the absence of pre-existing dislocations.
There are kinetic (i.e. Peieris) barriers, which iimit the extent of s& relaxation in
metastable structures for given temperature and tirne. Therefore, although exceeding the
critical thickness will result in misfit dislocations, it is not sunicient in determinhg the
onset of strain relaxation. Therefore, it is not possible to compare metastable structures
with equilibriurn theory (Perovic et al., 1995).
To smdy the effects of seain relaxation on sttained layer heterostructures, kinetic
barriers would have to be eliminated via post-growth annealing, and methods of detecting
thefint stages of misfit dislocation would have to be implemented. An excellent method
for determining the onset of strain relaxation is through the use of chernical
etching/lrlomarski interference microscopy, which will be discussed in later sections.
This technique is able to detect c 104 dislocations/cm2 unlike X-ray difiction, TEM, ion
channeling, Raman spectmscopy, photoliIminescence spectroscopy, and laser light
scattering. Houghton et ai. (1990) and Zhao (1993) were able to use Nomarski
microscopy to obtain results in excellent agreement with equilibriurn predictions.
The sources of nucleation of misfit dislocations were extensively exarnined by the
group of Perovic et ai. (1990). It was show using transmission electron microscopy
(TEM) that interfacial nucleation of 60' misfit dislocations was associateci with sub-
nanometer sized Ge-rich platelets that evolve fiom strain-induced growth (Perovic and
Houghton 1992, 1993). Rowell et al. (1993) found that the pIatelet formation has been
attributed to the onset of 2-D islanding under step growth conditions. The Ge
preferentially accumulates at the periphery of atornic terrace kink sites and is then
covered by the next atomic Iayer during the growth procedure (Perovic et ai. (1995).
This study M e r considered the nucleation processes in low misfit heterostnicnires of
Ge,SilJSi, where xc 0.13.
Low misfit heterostnicnires were chosen for this study to maintain consistency with
the nucleating mechanisms taking place Ï n the "Stage I" regime (dislocation densities
1 O' cd2) . Perovic and Houghton (1 995) defined this regime to be nucleation-limited,
where misfit dislocation nucleation was observed to increase Linearly with tirne, beyond
some background density, No (see Appendix). The misfit dislocation velocity remains
constant through this regime resulting in a hear increase in the oved l strain relaxation
rate with time. Activation energies were observed to be Q, = 2.5 + 0.5 eV and Qr = 2.25
f 0.05 eV for nucleation and glide of misfit dislocations, respectively (Houghton, 1991).
The activation energy for strain relaxation in the Stage 1 regime was fomd to be Q, =
4.75 f 0.5 eV. This value is consistent with combining the nucleation and glide energies
(Le. Qr = Qn + Qv) in series.
Interfacial perturbations
( 100) Substrate
Figure 1.1.4: Schematic diagram illustrating the nucleation of prismatic dislocation loops (vacancy-type) at the Ge-ptatelets localized at the strained layer interface. The prismatic loops generate the haif-loops on { 1 1 1 }variants which then expand to forrn misfit dislocations with an extemal driving force (RTA).
As in the work done by Perovic and Houghton (1992, 1993), the nucleation of misfit
dislocations of the strained layer heterostructures of this study will be considered as a
two-step process. The fmt step is the nucleation of the initial pnsmatic dislocation loop
(Le. with the loss of coherency at the Ge interstitial perturbation) and the second step is
the expansion of the loop to fonn a stable misfit dislocation. Figure 1.1.4 shows a
schematic diagram of the pmcess. The prismatic loops generate 60' shear loops, which
form half-loops in the epilayer andor multiple loops in the bufEer layer. With the aid of
some extemal driving force (Le. rapid themial annealhg (RTA)), the half-loops expand
leaving behind 60' misfit dislocations in two orthogonal 4 10> directions.
1.2 Molecular Beam Epitaxv
Ge,Sii., heterostnicnires can be grown at low temperatures (350-700°C) using the
MBE process. The advantages of p w i n g heterostructures at low temperatures is to
reduce the vibrational energies of thé growing film to decrease the disorder and
interdiffusion of Ge in Si (Hull et al., 1985). Figure 1.2.1 shows a schematic illustration
of the process. MBE is a technique for growing overlayers on crystals by the
condensation of low-vapour-pressure molecules or atoms from a molecular beam in a
high vacuum. The low vapour pressure of the materiai is required to provide nearly
instantaneous control of the deposition fluxes. Decomposition of the molecules at the
surface is avoided by using simple molecules or atoms.
O Heated rotating substrate crystal
Doping Gun
Figure 1.2.1 : Schematic diagram of MBE process
8
Si-based MBE consists of the coevaporation of sficon and dopants and acceleration
onto a slightly heated substrate. Silicon and germanium condense once they reach the
surface due to their negligible room temperature vapour pressures. This method of
elemental evaporation p e d t s the growth of crystalline layers at temperatures where
solid-state diffusion is negligible. Suice chernical decomposition is not required,
deposition species need acquire only enough energy to migrate dong the substrate
surface to a cry stalline bonding site (Bean, 1 98 1). This requires only minimal substrate
heating.
A Czochralski-grown silicon d e r of suitable orientation (i.e. <100>) is placed
polished side down in an ultra-high vacuum chamber. An in-situ cleaning technique is
used to produce an atornically ciean surface pnor to growth. Several methods can be
used such as ~ r ' sputtering/annealing and thermal desorption (1200°C) (Perovic, 1988).
Afier the substrate wafer has been cleaned, it is heated to its desired growth temperature.
The heated substrate ensures that single crystal growth will take place because the
thermal energy produced increases the sdace migration energy of atoms so they reach
proper bonding sites at the fke sdace. Deposition rates and concentrations are
controlled by rehctory metal shutters which are moved into and out of the evaporatiun
path. This process is cornputer controlled.
If proper conditions are met, smooth 2-D growth will take place. Lower growth
temperatures and lower Ge concentrations will ensure 2-D growth. However, if the
temperature and Ge concentration are raised accordingiy, 3-D growth will take place.
The kinetic pathways of the 2-D-to-343 transition are detailed by the group of Jesson et
al. (1996). Figure 1.2.2 displays a plot of epitaxial film morphology as a bction of
growth temperature and Ge concentration for a 1000 A thick GexSil, füm. Several in-
situ techniques are used to monitor the epitaxial growth parameters. A quartz crystd
deposition meter has a fmluency which changes during excitation as the crystal thickens.
Growth rates as low as - 1 A/sec can be realized with an accuracy of approximately 10%.
As well, there is a quadraple mass spectrometer which can be used to obtain molecuiar
beam flux and residual gas analyses. The surface crystallinity of the growïng crystal can
also be monitored. A surface electron difiction technique such as high energy electron
diffraction or low energy electron di f ic t ion can be utilized (Perovic, 1 98 8).
Figure 1.2.2: Plot of epitaxial film morphology vs. growth temperature and film composition for 1600 A thick GexSiI, films. (Bean et al., 1984)
1.3 Chemical Vapour Deposition
Chemical vapour deposition is among the most widely used film growth methods,
utilized in the deposition of critical device layers to the formation of lightweight carbon
fiber composite disk brake rotors for aircraft use (Meyerson, 1992). Characteristics such
as excellent film unifomiity, c o n f o d t y , mmpatibility with large area processing, and
relatively low apparatus costs compared with physical deposition techniques makes CVD
very attractive in the commercial market.
A low temperature process developed at IBM cded ultrahïgh vacuum chemical
vapour deposition (UHVCVD) will be discussed next. A schematic diagram of the
process is shown in figure 1.3.1. The UHV-CVD system is purnped at al1 times using a
turbomolecular pump, both when i d h g at base pressure. in the range of 1-5 x 1 oe9 torr,
and during film growth at 1 mtorr. The vacuum operation and high-density coaxial M e r
packing as in conventionai low pressure chemical vapour deposition (LPCVD), were
driven by the need of high-precision film growth. If films are to be accurate to several
Angstmms in thickness, the film growth rates need to be reduced to 1-10 Almin, and the
multi-wafer geometry of LPCVD is optimum to compensate for this (Meyerson, 1992).
Treatment of the silicon substrate before deposition is perfomed, is a requirement for
forming heterostructures of hi& quaiity. Rather than utilizing high-temperature oxide
desorption for surface preparation, UHV-CVD uses more conventional means. UHV-
CVD relies upon hydrogen passivation. An adlayer of hydrogen is formed in the process
of etching Si in hydrofluonc acid (HF), and was s h o w to reduce by 13 orders of
magnitude the reactivity of the silicon surface with respect to oxidants such as water and
oxygen (Meyerson et al., 1990). Therefore, in-situ cleaning methods as in MBE were
eliminated for UHV-CVD. The overall initial suface preparation for UHV-CVD is
straightforward involvïng a peroxide based chemical oxidation, followed by a 10:1 de-
ionized water/HF dip for 10 seconds. This pre-clean strategy produces a robust initiai
growth interface. As long as the hydrogen passivation is maintainecl, the silicon surface
remains free of contaminants.
Figure 1.3.1 : Schernatic of UHV-CVD process (Me yerson, 1992)
Afier the surface of the silicon substrate has been prepared for 10-35 wafers, held in a
quartz wafer boat, they are inserted into a load charnber which is pumped below 1 x 10"
torr. The wafer boat is then transferred under a hydrogen flow into the reactor, nominally
set to 550°C (- 500°C wafer temperature) at a pressure in excess of 200 rntorr. Al1
pumps remain in operation, resulting in a division of flow both down the reactor tube and
into the load charnber. This prevents the cross-contamination of the UHV side of the
apparatus by residuals in the load chamber. The entire cleaning and loading cycle for a
typical batch takes less than 15 min (Lafontaine et al., 19%). Upon isolating the growth
side of the apparatus, hydrogen flow is terminated, and silane flow is begun. At the flows
employed, typically 30 sccm (standard cubic centimeters per minute), system pressun
remains at - 1 .O mtorr.
The gaseous sources employed for the purpose of alloying or doping are germane,
diborane, acd phosphine, where these sources are diluted to produce films of deskd
chemical content. Helium is used as the carrier gas to facilitate diagnostics of the tlow
control systems, where it is straightfoxward to detect the He diluent gas compared wiîh
the highly diluent species. Film dopant content is linear in the atomic fiaction present in
the growth source, the same behg true for germanium incorporation (Meyerson et al.,
1987). Extremely precise dopant and d o y control is the result.
Auger electron spectroscopy (AES) and x-ray dihction (XRD) are used to measure
germanium concentrations. Photoluminescence (PL) is also used for M e r confirmation
of accurate Ge fraction. Thicknesses of deposited layers are determined by transmission
electron microscopy (TEM) and XRD. Background contamination effects and amount of
dopants incorporated are typically measured by secondary ion m a s spectroscopy (SIMS)
(Lafontaine et al., 1996).
1.4 Ion Implantation
Ion implantation is a process in which an ion of choice is embedded in the structure of
a selected material. ln a typical linear ion beam implantation (LIBI) (Figure 1.4.1). a
large voltage is used to bring the ion source niaterial up to a specified energy. For any
implantation to occur, the ions must be able to travel a distance of several meters without
collisions. This requires a vacuum of at Ieast 10-~ torr (Giedd et al., 1996). Severai
different kinds of purnps are utilized to achieve this pressure. Once the vacuum bas been
obtained, the ion source is filled with a gas. This gas is the ion source material, and can
be of a variety of compositions. The gas and ion source chamber are placed at a high
potential relative to ground. It is important that the entire source is well iosulated,
because at high potentials, dangerously potent arcs are possible (Hadey, 1976). A
current is run through the arc gap, aiming gas into a plasma.
The ions can be extracted h m the plasma using a magnet or strong electric field.
These extracted ions move as a beam towards the sample (Grovenor, 1992). The ions are
next passed through an analyzing magnet to reject those ions that are not of the desired
element. The operator can tune the magnetic field to allow onIy the ions of a certain
mass to p a s This beam then passes through an aperture that ennires the homogeneity of
the beam. At these apertures, a beam stop is present. The beam stop usually consists of a
large graphite block that on be dropped in fiont of the beam to block its passage. It is
here that an arnrneter is used to monitor the bearn current. The beam, screened of al1
undesired materials, is finally allowed to impact on the sample. The ions penetrate into
the material to some depth, altering the surface and bulk structure and the electncal
propenies of the matenal.
This method of doping materials allows for more exact control of where impurities
are placed in a material as well as more precise control of the impurity concentration in
the target material. Masks can be used to isolate areas of the target for doping. Lateral
motion of the dopants is much smaller in ion implantation since the diffraction of ions is
negligible (Giedd et al., 1996). The careful control of the beam current allows one to
monitor the dopant lrvels. Measurements of how many ions are impacting per unit time
c m be made. By multiplying the desired fluence by the entire scan area of the beam and
dividing by the bearn current, it is possible to calculate the implantation time necessary to
achieve the desired fluence.
The changes caused by ion implantation are ofien difficult if not impossible to
produce by other methods. The ability to control dopant parameters of the process aliow
the user to tailor the expairnent to produce correct concentrations and location of the
dopant materials.
Figure 1.4.1 : Photo of typical Ion beam implantation device (courtesy of University of Western Ontario)
1.5 Positron Annihilation S~ectrosco~v
The positron (e") which is an antiparticle of the electron (e'), has the same mass
within current experimental limits (51 1.003410.0014 kev/c2) and the same spin (%), but
with opposite charge and magnetic moment (Schultz et ai-, 198 8). It is stable in vacuum,
but rapidly themalizes in metais, and annihilates with an electron predominantly via 2-7-
ray decay (- 5 1 1 keV) with a mean lifetime that is typically only a few hundred
picoseconds (psec).
The positron's reaction in solids make it a very usehl method of studying defects on
the surface and in the bulk of materials When a positron from some radioactive source
enters a solid, it rapidly loses its kinetic energy (- 10 psec) until it is near thermal energy,
scattering between Bloch States to d i f i e through the solid. Figure 1.5.1 shows the
schematic of this process. The thermalbation tirne is short relative to the average
lifetime of the positron in the solid. Mer themalization, the positron rernains essentially
as a "free" or delocalized particle, although strongly correlated with conduction electrons
in its environment, until it annihilates in the bulk solid (Schultz et al., 1988).
DIFFUSION
POTENTIAL
Figure 1 .S. 1 : Schematic of positron's reaction in solids
The positron annihilation procedure is used to study lattice defects because a k l y
d i fh ing positron can localize in regions of minimum potential in a periodic lattice such
as vacancies as seen in Figure 1.5.1. This process is referred to as the Doppler
broadening of the energy of the annihilation y rays and relates to the positron trapping
into a vacancy. Figure 1.5.2 shows the effects of Doppler broadening (a) when a positron
is trapped in a vacancy, which also leads to longer positron lifetimes (b). Figure 1.5.2 (c)
shows how trapped positrons at defects, or at the surface can be parameterized by the line
shape parameter "Y, shown in the figure.
Energy
Figure 1.5.2: (a) Doppler Broadening (b) Positron lifetimes and (c) Formation of the S-parameter
The S-parameter is sensitive to vacancy concentrations of - 10-'crn-~. For each
implantation energy, the resulting S-parameter is
S = FbSb + FsSs + FdSd
where Fb Fs, and Fd conespond to the fiaction annihilating in the bufk, d a c e and at
defects and Sc, S,, and Sd are the S-parameters of the bulk, d a c e , and defect
respectively.
If a positron is trapped at a vacancy, it wiii be more likely to annihilate with the iow
momentum valence e1ectmn.s than the higher momenhim core electrons. Therefore the y
ray from the annihilation of trapped positrons will have a smaller Doppler shift and a
higher S-parameter.
The mean depth (D) at which positrons annihilate depends on their implantation
energy (E in keV). The group of Simpson et al. (1991) have simplified this relationçhip
as follows
D(in A) = 172 (E)'"
Positron beams al1 start with primary sources that have continuous energy
distributions arranged near a positron moderator or converter (Schultz et al., 1988). The
most cornmonly used primary sources are radioactive isotopes. A back-reflection
geometry (Figure 1.5.3) is typical with a '*CO source in front of a single-crystai
moderator to produce the bearn. Once the positron beam is extracted fiom the moderator,
it is magnetically-guided through a vacuum system to the target For the purposes of this
study, the positron beam utilized has a large Ge detector present for spectroscopy of
annihilation y rays for defect studies.
e+ Figure 1.5.3: Schematic of back reflection
geometry in e' beam production
1.6 Microsco~v Techniaues
Three different microscopy techniques have been utilized throughout this study of
defects in low misfit strained layer heterostructures. Each microscope possesses unique
characteristics which make them valuable tools for research These characteristics wiil
be discussed in detail in the following sections for (a) Nomarski Interference
Microscopy (NIM), (b) Field Ernission Scanning Electron Microscopy (FE-SEM), and (c)
Transmission Electron Microscopy (TEM). These microscopes are al1 located at the
Department of Metallurgy and Materials Science, University of Toronto.
1.6.1 Nomarski Interference Microscope (MM)
With slight modifications, a common optical microscope can be used for Nomarski
interference microscopy (NIM)(Figure 1 -6.1). A Wollaston prism and polarking lenses
are used to perform this procedure. Variations in surface level c m be seen by the
interference fringe contours or by changes in contrast performed by interference
microscopical techniques (Zhao, 1993). Coherent beams of light waves that are out of
phase by half a wavelength interfere destnictively which produce slightly displaced
images. These images cancel each other out at al1 points except those where the
displaced images are out of step, thus revealing the structure.
A parallel-sided double quartz pnsm .(Wollaston prism) produces the double image.
The angle of the wedges in the prism detemines the separation of the double image. An .
analyser which rotates in the column of the microscope produces coloured images
revealing certain wavelengths. This procedure illuminates and reveals desired structures
on the surface of the sample and can also provide optimum conditions for
photomicrographs by enhancing contrast.
I eyepiece
I I analyser
Inclincd giass slip illuminaior
objective lem sample
Wollaston prism
Figure 1.6.1 : Schematic diagram of Nomarski Interference Microscope (Haynes, 1 984)
1.6.2 Field Emission Scanninpr Electron Microscope (FE-SEM)
The Full capabilities of the FE-SEM will not be discussed here, however, certain
aspects of this powemil tool will be mentioned and their importance in this snidy. The
electron source in the FE-SEM is different fkom other conventional SEMs. Mead of
relying on high temperature to enable a fiaction of the free electrons in the cathode
material to overcome the barrier of the work function and leave (thennionic emission),
the FE-SEM utilizes an electric field at the tip to overcome the barrier (Goldstein et al.,
1981). The emitter for FE-SEM most commonly used is the <310> single crystal of
Tungsten (W). The 0 10> orientation has been chosen because the electmns were fomd
to travel in this direction most efficiently. Due to the Iarge mechanical stresses brought
on by the large electnc field at the tip (> 107 Vlcm), ody very strong materials can
withstand this stress without failing. Figure 1.62 displays the gun configuration of the
FE-SEM.
field emission tip 4 fmt anode a //\
Figure 1.6.2: FE-SEM gun configuration (Goldstein et al., L 98 1)
The three factors which govem the quality of a SEM image are: (a) electron probe
size, (b) probe brightness, and (c) interaction volume. A high intensity signai and small
probe size combine to produce the best imaging results.
Of the primary electrons that hit the sarnple fiom the incident beam, the two main
types of electrons which are most commoniy used for imaging are backscattered electrons
(BSE) and secondary electrons (SE). niirty percent of the pnmary electrons generate
BSE on average. BSE resdt fiom a shgie scattering event with the electron path greater
than 90' kom the incident beam direction. However, for the purposes of this study, only
SE will be outlined in detail. SE are the electrons most ofien used for sdace imaging in
the SEM. SE are defined as those electmns emitted h m the sample with an energy less
than 50eV (Goldstein et al., 198 1 ). interactions between the energetic beam and weakly
bound conduction electmns produce SE.
Due to the low kinetic energy of emission in which SE are generated, SE have a
shailow sarnpling depth ('; lm). SE are strongiy attenuated during motion in a solid by
energy loss due to inelastic scattering, which has a high probability in low-energy
electrons (Goldstein el al., 1981). The probability of SE escaping decreases
exponentiall y with depth in solid according to the following relationship,
P a exp (-dl)
Where z is the depth below the surface and A. is the SE mean fiee path. h is
approximately O S - 1.5 nm for metds and 10-20 for insulators.
Low accelerating voltages (< 5 kV) produce small interaction volumes in the SEM.
At accelerating voltages near 1 kV, spatial resolution can get as good as 5-7 A. Higher
SE yields are the result of lower accelerating voltages. Figure 1.6.3 shows the
relationship between the total emitted electron coefficient (backscanered and secondary)
and the beam energy. The secondary electron coefficient (6= nse/ne), where n s ~ is the
number of SE emitted fiom the sarnple and n~ is the number of beam electrons, rises to
reach unity at 1 kV. 6 increases to a value of 5 for nonmetais at 1-2 kV and then
continues to decrease to a beam energy of 0.1 at 20 kV for meîals.
The influence of specimen tilt on the SE cwfficient follows a secant law, 6(8) = 6.
sece, where 6, is S(B = 04 (Kanter, 1961). The SE CO-efficient increases with sec6. BSE
have a strong dependence of tilt angle. BSE increase with increasing tilt. Since SE are
also generated fiom the sample which suGers a backscattering event, the number of
secondary electrons will increase (Goldstein et al., 198 1).
LOG E,,
Figure 1 -6.3 : Relationship between total emitted e- coefficient and beam energy. (Goldstein et al., 198 1)
1.63 Transmission Electron Microscope CïEM)
Transmission electron microscopy (TEM) has proven to be a very powemil tool in
the study of defects in strained layer heterostructures. Individual layers can be isolateci
and observed with great accuracy in cross-section. Plan-view TEM allows one to scan
over large areas representative of the bulk matenal. Individual dislocations and
dislocation sources can dso be examined. There are situations where TEM results are
not representative of the bulk material due to structurai distortions induced by the shear
stresses which accompany any lanice parameter modulation. These distortions result
fiom TEM thinning during preparation in a direction perpendicular to the modulation
direction such that elastic relaxation of the internai stress-field responsible for the
tetragonal distortion occurs near the surface of the thin foi1 (Perovic, 1988).
For the study of dislocations and small defects such as precipitates or perturbations,
TEM is an indispensable tool. The combination of bright and dark field imaging
techniques and dif ict ion pattern adysis is essentiai in the chanicterization procedure.
Coherent elastic scattering produces difhction conhast DZhction contrast is simply a
special form of amplitude conh=ast beuiuse the scattering occurs at special (Bragg) angles
(Williams et al., 1 996). From the ciiffiaction pattern, which is usually known for single
crystal orientations, the 'NO-bearn' condition shouid be implemented. Tilting the foi1
and obtaining a recognizable symmetry is a convenient starting point (Thomas et al.,
1979). Because of the 180' ambiguity in spot patterns, the spot pattern by itself does not
yield a unique foi1 orientation. However, a Kikuchi pattern can be indexed uniquely and
is facilitated with a Kikuchi map. For vacancy or interstitiai loops, defects or faults, this
information is needed to d e t e d e the sense of the strain fields.
Defects with zero or small strain fields (e.g. voids, inclusions of second phase with
zero misfit) are visible only through "scattering factor" contrast-columns through such
defects appear to be of different thickness from those through adjacent regions of matrix
crystal. They are thus only visible in situations where strong thickness fringes are seen,
that is, in thin areas when the crystal is at the reflecting position (Thomas et al., 1979).
Defects with greater strain fields produce small "black-white" contrast (small dark
areas on a light background) when imaged at deviated positions. This black-white
contrast is oscillatory with the depth of the defect (with a penod equal to that of thickness
fnnges) (Thomas et al., 1979). The sense of the g-vector is useful in determining the
nature of the defect (interstitial or vacancy-type). Smdl Ge perturbations found in Ge-Si
strained layer heterostructures were determineci to be interstitial in nature due to the
contrast on the perturbation which went fiom light to dark following the g-vector in the
[220] direction (Perovic et al., 1992).
CHAPTER 2 - EXPERIMENTAL
2.1 Heterostmcture Growth Conditions
The two growth techniques used in this study were MBE and UHV-CVD. Both of
these facilities are located at the National Research Council of Canada ( M E C ) in
Ottawa. The heterostruchires were grown on 4 inch diameter (100) Czochralski Si
substrate wafers (6 inch for CVD). The Czochralski method is show in Figure 2.1.1.
Figure 2.1.1 : Czochralski growth method
UHV-CVD conditions for growth were taken fiom Lafontaine et al. (1996). The
UHV-CVD samples were grown using a "Sirius" hot wall UHV-CVD reactor. The
growth chamber consists of a quartz tube heated by a hirnace and evacuated by a
turbmolecular/roots blowedro tary pump system. The base pressure was 1.5 x 1 0" mbar
at T = 52S°C. Silane (100%) and germane (10% in helium), of Matheson ULSI grade,
were used as precursors. Customdesigned mass flow controlles (MKS, type 1449A)
were used to control gas flows.
M e r a standard "RCA" clean and an HF:H20 (1 : 10) dip for 10 seconds, d e r s were
introduced in the chamber through a loadiock which was pumped d o m to a base pressure
of 1 O" mbar in 5 minutes. A hydrogen flow of 300 (sccm) was used during d e r
transfer into the main chamber in order to prevent contamination from the loadlock
Growth was initiated immediately after the t r d e r is completed, Si& injection. The
entire cleaning and loading cycle for a typical batch takes less than 15 minutes. Table
2.1.1 shows the growth parameters of the UHV-CVD samples used.
(see Appendix for definition of reK, N, h, and H)
The MBE procedures foliowed those of Baribeau et al. (1994). GeXSii.JSi
superlattices were produced in a VG Semicon V80 MBE system. Unlike UHV-CVD,
only one wafer was grown at a tirne in this system. A 4 inch wafer was placed face down
inside the ultra-high vacuum chamber. An in-situ pre-clean was perfomed on the wafer
above 850°C under 0.1 A/s Si flux for 15 minutes. The wafer was then heated to the
desired growth temperature (360-500°C) (see Table 2.1 . 1 (b) for details). Each wafer
received a 15 nrn buffer layer of Si prior to growth of the superlattice. Samples 1688-
1703 were grown at rates of 5 &sec. Samples 1706 and 1707 were grown at slightly
lower growth rates. The growth rates were controlled by a Sentinel III in situ monitor.
The vacuum pressure during growth was maintained at 5 x IO-'* torr. Accurate
TC f i
( M W 1 03 57
composition readings were made by doublecrystal x-ray rocking curves.
N Periods
10 10
Sample #
CVD-9 CVD-61
Si)
21 29
X
0.10 0.13
h(siGe)
(nm) 15 7
Table 2.1 .llbl- Structurai Panuneters for MBE Ge,Sil.&
2.2 Ion Implantation Conditions
Sample#
MBE1688 MBE 1689 MBE1690 MBE1691 MBE1702 MBE 1703 MBE 1706 MBE 1707
Poçt-growth ion implantation was performed on four of the ten wafes produced.
CVD-9 and CVD-61 were taken h m the UHV-CVD sarnples and MBE1702 and
( s e Appendix for definition of T = ~ , N, h, and Fi)
Growth Temperature
CC) 405 405 360
MBE 1703 h m the low temperature MBE samples. Ion implantation procedures follow
those of Goldberg et ai. (1993) and Labrie et al. (1996). The GexSii,/Si superlanices
x
0.9 0.9 0.9
were self-implanted with 540 keV Si ions to a fluence of 2 x 1014 ions/cm2 ushg a 1.7
MV Tandetron accelerator located in the Department of Physics, at the University of
h(='
(m)
15 10 15
360 450 500 400
Western Ontario. Wafen were secured to a nickel block with thermally conducting paint
0.9 0-85 0.84 O. 12
to ensure good thermal contact. The temperature of the block was maintained at 25 k 1°C
IO 15 17 11
during irradiation. An implant flux of - 1 15 &cm2 was kept constant throughout the
Gtr
@Pa)
103
SI)
(m)
21
IO 10 10 I O
30 23 24 15
bombardment. Channeling effects were avoided by rotating the sample - '7' about axes
N Periods
10
47 89 97 137
10 14 425
both perpendicdar and parallel to the incoming beam.
30 21
147 0.13 1 10
10 I O
47 1 03
23 Rapid Thermal AnneaIin~ (RTA)
To eliminate kinetic barriers, pst-growth annealing via RTA was performed on
CVD-9, CVD-6 1, MBE-1702, MBE-1703, MBE-1706, and MBE-I 707. Table 2.3.1
shows the RTA temperatures and h e s for these samples. These temperature/time
combinations were chosen according to the equivalent strain relaxation contours of
Houghton et d (1993). The RTA were performed using a Heatpulse 410, in a dynamic
nitrogen atmosphere. The Heatpulse 410 is located at NRCC, Ottawa, To avoid
contamination of the samples fiom dust particles, the RTA experiments were conducted
in controlled airflow cleanrooms and full body suits had to be wom, as they were for
CVD and MBE growths. Small 1 cm x 1 cm samples were placed on a 4 inch diameter
Si wafer substrate which was used as a stage in the Heatpulse 410. RTA temperature
ramping was controlled via cornputer software.
Table 2.3-1 - Temperature and T h e Parameters for RTA
2.4 Nomarski Interference Microscopy (NIM)
Misfit dislocation nucleation rates and densities were calcuiated Eom images
produced on an optical microscope, modified for Nomarski imaging. Specimen
preparation for NIM was performed on the samples used for RTA. The 1 cm x 1 cm
wafer pieces were immersed in a dilute Schimmel etch (4 parts 48% HF : 5 parts 0.3-M
CQ) for desired etching times. The etching times were calculated using a Dektak 3-30
Temperature CC) 700 800 850 900 950 1 O00
RTA time (s) 1 O0 100 30 10 5 5
Version 2.13 depth profiler. Small drops of a black wax, immune to the etching solution,
were placed in the corner of each wafer. After submersion in the Schimmel etch for
approximately 30 seconds, the samples were profiled in the Dektak across the
rtchedhon-crched boundary line. Etching rates were determined for each sample. The
misfit dislocations nucleated at the interface of the substrate and the first Ge-Si layer,
therefore the proper rtching depth to reveal the rnisfits was calculated by using the
geometry of the layes (e-g. for CVD-9: 10 x (21 nm Si + 15 nm GeSi) = 360 nrn total
height to be etched). Sarnples needed for TEM observation were not etched to preserve
the lattice planes.
Figure 2.4.1 shows a typical Nomarski micrograph. Care was taken to observe
representative areas (several cm') away From scratches or edges to determine the nurnber
of nzw misfit segments formed per unit area per unit time. Dislocations can nucleate
from scratches. clsaved edges and other stress concentrations and cause inaccurats
readings of the tme strain relaxation data of the heterostructure.
Figure 2.4.1 : Micrograph of a typical Nomarski image showing UHV-CVD sample RTA for 100 s at 700°C. Individual misfit segments can be seen.
2.5 Field Emission Scanning Electron Mieroscom WE-SEMI
FE-SEM exgnhtion was carried out on the S4500 FE-SEM. Specimen preparation
for the FE-SEM was the easiest to carry out compareci to al1 other techniques. MBE-
1688, MBE-1689, MBE-1690, and MBE-1691 were the samples used for FE-SEM
observation Small(2 mm x 3 mm) pieces were cleaved using a diamond tip cutter. No
other preparation procedures were required. The fkeshly cleaved edge was placed in the
rotating sample holder and loaded into the vacuum chamber of the FE-SEM.
Low accelerating voltage (1-2 kV) experîments were conducted on the samples with
4-6 mm working distances. The secondary electron (SE) detector was used to examine
the surfaces of the MBE growths. The usual stigrnator corrections were performed to
pmduce the high-resolution images. Slow scanning speeds were used to obtain the
highest level of detail and contrast. Micrographs were acquired using the Quartz-PCI
software present in the FE-SEM laboratory.
2.6 Transmission Electron Microscopv (TEM)
A Hitachi H-800 TEM operating at 200 kV was used for specirnen characterization.
Sarnples were prepared for both cross-section and plan-view anaiysis. The cross-section
procedures are as follows. Several strips (1 mm x 5 mm) were cut fiom the 400>
oriented wafen using an automated cross-cut saw, located at NRCC. The strips were h t
immersed in acetic acid and then ethanol to clean the samples of any contaminants. Next,
two strips were turned 90' so the fih sides faced each other. A silver epoxy was used to
glue them together. This produced <O 1 1> surface normals. See figure 2.6.1 for
schematic of procedure. The epoxy was then cured at 100°C for approximately 3 hours.
The next step was to core the sample using a Precision Instruments drill. The sample
was placed on a g l a s slide and kept there using a fast cooling wax A 3 mm diameter
disc was cored using a 1 pm particle diamond paste slurry. The disc was then mounted
ont0 a polishing tool using the same wax. The disc needed to be mechanically dimpled to
a thickness of - 50 pm i m g the 1 pm diamoed paste. A D500 Dimpler was used for this
procedure.
The final step for the cross-section specirnen preparation was ion-milling. A Gatan
Ion Mill was used for this procedure. Argon ion sputtering was performed on both sides
of the disc at the same time until the sarnple was electron-transparent in the region of
interest. A sputtering angle of 15' and a voltage of 4.5 keV were used for the first 5
hours of operation to mil1 through the buik of the matenal. A shailower angle of 1 1' was
then used to min& the depth of the amorphous surface layers which result nom
mechanical damage in the sputtering process (Perovic, 1988). If done correctly, the
cross-section sample had four areas of observation. The entire process can take up to 25
hours per sarnple Cor TEM preparation.
/ ' / / A
slice sections
1 mm size strips
SOO A thick Figure 2.6.1 : TEM cross-section
preparation
The plan-view sampies were easier to prepare because there was no epoxy present,
therefore no weak adhesive bonds to break. A 4 mm x 4 mm sample is placed face down
on a glass slide and glued with the same wax as above. The coring, dimpling. and ion-
milling procedures were al1 the same as above. The only clifference is that fih side of
the sample never cornes in contact with the dimpling wheel or the argon sputtering stream
of ions. Thus, oniy the backside of the of the plan-view sample was ion-rnilled at a tirne.
prolonging this procedure.
In order to produce quality TEM micrographs for observation, detailed procedures
had to be followed. After producing suitable sarnples for the TEM, the next step was to
perform TEM imaging. For the purposes of this study, acquiring a working difhction
pattern (DP) and obtaining the mo-beam condirion were essential for TEM observations
to be made. Figure 2.6.2 shows (a), a schematic of the DP of a diamond cubic lattice
with [O011 beam direction and (b), an actual image of the pattern obtained in the TEM
with g = [400]. The beam direction or zone-axis z = [OOI] was used for plan-view
images and z = [O 1 1 ] for cross-section images (Figure 2.6.2 (c)).
To obtain good strong diffraction contrast in both bright field (BF) and dark field
(DF), the specimen was tilted to two beam conditions. in which only one difhcted beam
was strong as seen in figure 2.6.2(b). After obtaining the two-bearn condition, the
objective aperture was centered on the incident b a r n to produce BF images and centered
on one d ihc ted beam to produce DF images of the desired g-vector orientations.
Strain field contrasting was also performed on the TEM. In order to locate small
perturbations in the stained layer superlattices (SLS), first discovered by Perovic et al.
(1990). lattice stra in effects around the perturbations had to be examuied These lobes of
low intensity are symbolic of strain fields (Williams et al., 1996). Strain field contrast
images were obtained in DF mode.
Figure 2 - 6 2 (a) schematic of DP of diamond cubic lanice (beam direction, z = [O0 11) (b) actual TEM micrograph with g = [400] displaying 2-beam condition (c) schematic of DP showing cross-section arrangement (z = [O 1 11)
2.7 Positron Annihilation Conditions
The positron laboratory is located at the University of Western Ontario. It was here
that al1 the S-parameters were modelled and statistically calculated. One of the two
variable-energy positron beams was used to characterize the defect structures in the
samples (i.e. voids, vacancies, etc.). The positron beam started at the primary source,
which had a continuous energy distribution arranged at the positron moderator. The
radioactive isotope used was the S 8 ~ o source in front of the single-crystal moderator.
Magnets existing in the apparatus guided the beam through the vacuum chamber to the
target A large Ge detector was used for spectroscopy of the annihilation y rays for defect
c haracterization.
The S-parameter, discussed in Section 1.5, was determined for each sample in which
positron annihilation was performed. The correspondhg equations were also determined
(i.e. the mean depth at which positrons annihilate). By using these equations, different
positron annihilation models were calculated and defect concentrations were determined.
CHAPTER 3 - RESULTS AND DISCUSSION
The following section considers the observations and results obtained h m the
experiments and characterization techniques. In this examination of seain relaxation
effects of low misfit heterost~ictures, evidence of dislocation multiplication has not been
obtained. The experimental procedures outlined thus far have centered around seain
relaxation effects in the fim stages of relaxation (i.e. Stage 1 regime, dislocation densities
between O - 10' cnf2). It is in this area that there is little known about the precise
injection mechanisms of misfit dislocations. Multiplication processes and dislocation
interaction effîcts can be ignored in this regime. It is not until dislocation densities > 10'
that these issues are of concem. Dislocation interactions introduce a host of new
problems (Le. dislocation blocking, pile-ups, etc.) in trying to detemiine s t n h relaxation
mechanisms (Schwarz, 1997, Schwarz and Teaoff, 1996).
A novel method for controlling the generation of misfit dislocations by point defect
injection via ion-implantation in m e d Iayen has been examined. Bulk measurements
were made to detennine the misfit disiocation nucieation rates and densities for various
Ge,Si&i (x 5 0.13) heterostructures grown by MBE and W - C V D methods.
Nomarski interference microscopy was used for these bulk measurements. TEM has
been used to locate sources of nucleation and to examine the dislocation structures both
in plan-view and in cross-section.
Positron annihilation was perfomed on samples which were thought to have
substantid vacancy concentrations. On some of the low-temperature MBE growths,
where incoherent growth resulted, positron annihilation was used to mode1 the void
distribution in the niperlattices with the S-parameter. Along with FE-SEM. the void
study of strained layer superlattices (SLS) yielded some interesting resuits which were
consistent with a "morphological instability phase diagram" proposed by Perovic et al.
(1 993).
3.1 Strained Laver Geometrv
The strained layer geomehies play a major role in detennining how the
heterostructure will relax and through what mechanisrns. Figure 3.1.1 illustrates the
typical geometrical distribution in the SLSs studied. Each individual strained layer (i.e. h
or H for GeSi and Si thicknesses, respectively) was grown below their criticai thickness.
N represents the nurnber of periods of repeating (h + H) layers. According to this
configuration, compressive strain is generated in the layers as growth proceeded and
relaxation was attributed to the extension of misfit dislocations at the first GexSil,,/Si
interface. The unbalanced force responsible for the misfit dislocation nucleation and
propagation in strained epitaxial layers exceeding equilibnum critical thickness is the
effective stress, r,n (see Appendix), which was k t defmed by Matthews, Mader, and
Light (Houghton, 1991). This effective stress represents the driving force for misfit
nucleation and propagation.
A range of reff was examined in this study to obtain varyhg nucleation rates and
misfit densities. The effective stress is detemiined by the imbalance between the
resolved shear stress acting on the threading dislocation slip system and the iine tension
in the extending a/2 < 1 1 O> misfit dislocation segment (Houghton, 1 99 1). Typically,
SLSs grown with higher r,a values displayed greater nucleation rate and misfit
dislocation density values. However other factors were aiso important in detamining the
nucieation rates and dislocation densities.
Figure 3.1.1 : Geometrical distribution in the superlattice
3.2 Bulk Ouantitative Measurements
Large area diagnostic techniques were used to determine the misfit dislocation
nucleation rates and mis fit dislocation densities of the heterostructures. A comprehensive
strain relaxation rnodel (Houghton (1 99 1 ), Perovic and Houghton (1 993)) has been
developed based on bulk measurements of rnisfit dislocation nucleation and glide for a
range of metastable (1 00)aiented Ge,&,/Si (0.03-0.25) heterostructures grown by
MBE, RTCVD, and UHV-CVD. The same tests were conducted in this study following
their mode1 using GexSii../Si (0 .09~0 .13) heterostructures grown by MBE and UHV-
CVD.
Post-growth annealing (5-100 s, 700-1000°C) was carried out on the metastable,
coherently suained superlatticw (i.e. initially misfit dislocation-free) according to the
procedures outlined in Section 2.3. Afier a Schimrnel etch, quantitative measurements
were conducted on the samples (severai cm2) away fiom edges or other stress
concentrations. Therefore, the number of new rnisfit segments generated per unit area
Figure 3 2 . 1 : Nomarski images of initial stages of relaxation for CVD-6 1 after RTA for (a) 100 s at 700°C (b) 100 s at 800°C ( c ) 10 s at 900°C and (d) 5 s at 1 OOO°C
per unit t h e were determinecf for the nucleation rates (dN(t)/)/dt) (sez AppAppe in (cm-2
s-') for a given driving force (snr). In order to calculate the misfit dislocation densities
(N(f)) in (cm-'), the inverse of the distances between the misfit dislocation lines for a
given rnagnification were measured.
These values were measured directly off the Nomarski micmgraphs such as in Figure
3 -2.1. This figure illustrates the evolution of orthogonal a/2 cl 1 O> misfit dislocations for
the CVD-6 1 multilayer of Table 2.1.l(a). At 700°C, little misfit injection has o c c d
whereas at 800°C a low density of misfit segments is apparent The two higher anneals
(Le. 900°C and 1000°C) indicate a rapid increase in both average misfit segment length
and an increase in the density of activated nuclei at elevated annealing temperatures.
Figure 3.2.2 displays the nucleation rate vs. inverse temperature (KI) for various
sarnples grown by UHV-CM and MBE processes. An important fact to consider here is
that although the nucleation rates Vary between samples, due to layer geometry and
effective stresses, the activation energy for nucleation is consistent at a value of Q. = 2.5
+ 0.5 eV. This value is found by taking the dope of the line on the nucleation rate curve.
It represents the exponentiai evolution of misfit dislocation density (N(t)) as a h c t i o n of
temperature (7). This value is consistent with the work done by Houghton (1991) and
Zhao (1993) and recently by the group of Wickenhêuser et al. (1997). n i e group of
Wickenhauser et al. (1 997) were able to isolate the three different mechanisms involved
in stmùi relaxation (Le. nucleation, propagation, and multiplication) and study them
independently by selective epitaxy. The group of Tanaka et al. (1996) found the
activation energy to be 2.5 eV by studying photoluminescence (PL) spectra of deformeci
bulk Si-Ge alloys. By studying the change in alloy compositions around dislocations,
caused by elastic interactions between constituent atoms (i.e. Si and Ge), pecuhar peak
shifts of D 1 and D2 lines were observeci.
Ternperature (OC)
7.6 8.1 8.6 9.1 9.6
Ternperature (1 /K) (1 E-4)
Figure 5 -2.2: Nucleation rates of misfit dislocations vs inverse temperature
Other attempts have also been made in trying to determine the strain relaxation
mechanisrns. The group of Hull et al. (1989) used in situ TEM measurements to produce
activation energies for strain relaxation. Their low values for activation (Le. Qv = 1.1 eV
- 2.2 eV and Q,, = 0.3 eV for glide and nucleation, respectively) indicate that other
mec hanisms are taking place (i .e. dislocation interactions). Due to the srnall dimensions
of the thin foils used in the TEM, oxidation-induced vacancy injection at the surface cm
occur and vacancy migration in the Si substrate wodd be the rate-limiting process and
would account for the Qn = 0.3 eV measured by Hull et al. (Perovic et al., 1995).
It is also important to note the nucleation rate incrrases with t h u g h a power-law
dependence, n = 2.5. The nucleation rate varies hearly with the initial misfit dislocation
source density (No) present in the as-grown material at t = O (Le. subsîrate threading
dislocations, residual substrate-baer layer precipitates, etc.). Different substrate
cleaning procedures pnor to growth result in varying values of the initial source density,
No, for MBE (No > 10) cnf2), and UHV-CVD (No S 102 cm"). The linear increase with
time proceeds to densities well beyond No indicating the themial activation of new
sources in the early stages.
Figure 3.2.3 displays a plot of misfit dislocation density vs. inverse temperature for
UHV-CVD and MBE samples. In these samples the activation energies were found to be
Qr = 4.2 + 0.5 eV. This value is double that of misfit glide and close to the value of 4.75
eV reporteci by Houghton (199 1). The reason for the higher activation energy is that this
value takes into account the overdl strain relaxation in this stage of relaxation. Therefore
the nucleation and glide energies are in series to produce this value (i.e. Q, = Q, + Q,).
Figures 3 2.2 and 3.2.3 represent measwements taken from the initial stages of strain
relaxation (i-e. Stage 1 regime). This regime is said to be nucleation-limjted, b u s e the
misfit dislocation nucleation increases linearly with tirne, beyond No. Houghton et al.
( 1 990) found that the misfit dislocation velocity was effectively constant which resdted
in a linear increase in the ovemll seain relaxation rate with time. The rate-limiting step
in rnisfit dislocation nucleation has k e n attributed to vacancy formation andor migration
during the generation of the incipient dislocation loop at growth-induced interstitial
perturbations (i.e. Ge-rich platelets) which will be discussed in detail in Section 3.3.
Temperature ("C)
Figure 3 -2.3: Misfit dislocation density vs. inverse temperature for UHV-CVD and MBE matenal
Controlling misfit dislocation generation in strained layer superlattices has been a
major concern for researchers and manufacturers of electronic devices. To achieve
optimum performance fiom high-speed GeXSiiJSi strain layer devices, the retention of
structural perfection is a necessity. A novet procedure for controlling and reducing misfit
dislocations in these structures has been examined here. A detailed shidy on G+Sii,/Si
as-grown heterostructures verses 540 keV self-irradiated Si samples has been conducted.
Figure 3.2.4 shows a plot of misfit dislocation nucleation rate data acquired for the
UHV-CVD matenal shown in figure 3.2.2 (Le. CVD-61) and a plot of the same material
implanted with 2 x 1 014 Si ions/cm2. Figure 3 . î .S shows a Nomarski micrograph of the
CVD-6 I heterostnicture. An intentionally introduced scribed line, to assist nucleation of
new misfits, separates the as-grown and implanted sides. After a RTA of 850°C for 30
seconds, the following features were observed. Note how the misfit dislocations extend
Temperature (OC)
7.6 8.1 8.6 9.1 9.6
Temperature (1 /K) (1 E-4)
Figure 32.4: Nucleation rate data for UHV-CVD as-grown and implanted material
Figure 3 -2.5: Nomarski micrograph showing as-grown and implanted sides of UHV-CVD matend with rnisfits nucleating from scribed line. The scribed line separates the implanted and as-grown sides (see text)
M e r into sample on the a s - p k side. Wiîh time and increased temperatUres, the
misfit lines will extend across the entire sample. The scribeci line was introduced after
implantation to separate the two sides and RTA was then subsequently performed.
Figure 3.2.6 shows a plot of the misnt dislocation density vs. inverse temperature.
The dislocation densities dif5er by as much as a factor of eight for CVD-61 and a factor
of five for CVD-9 (not shown here)- A dramatic decrease has been observed in the
nucleation and densities of misfit segments on the implanted sample. However, the
activation energies for nucleation and overd strain relaxation remain constant at Qn = 2.5
+ 0.5 eV and Q, = 4.2 t 0.5 eV respectively, uidicating that the same mechanisms are
present even afier implantation.
Temperature (OC) 1000 950 900 850 800
7.6 8.1 8.6 9.1 9.6
Temperature (1 /K) (1 E-4)
Figure 3.2.6: Misfit dislocation density data for CVD-6 1 (as-grown and implanted)
The results shown here are consistent with a mechanism where Ge-rich platelets,
which possess an interstitiai eIastic field, can be relieved by the injection of point defects
and hence reduce the overall stress concentrations surrounding the perturbation which
assist the rnisfit dislocation generation Therefore, an overd decrease in sûaîn relaxation
via misfit dislocation is expected, which is in agreement with the resuits.
There are many theoretical and experimental atomic diaision studies about Si and Ge
and well-defined values for energies of vacancy formation (Er) and migration (E,). Van
Vecheten (1980) observed values for E~'' = 0.33 eV, E:' = 2.3-2.5 eV, E? = 1.0 eV,
and E? = 2.0 eV. Therefore, with the Q, = 2.5 f 0.5 eV found in this work, the energy
term can be attributed to: (a) vacancy formation in Si or (b) vacancy formation in Ge
followed by migration in Si. The latter is consistent with these resdts. This information
lead to the next set of experiments.
It is now accepted that the e l d c stress-field smunding the interstitial platelets (i.e.
Ge perturbations) can be relaxed by condensation of vacancies, which would Ieave
behind a vacancy-type pnsmatic loop. Section 3.3 will show evidence of the Ge-platelets
and the resultant behavîour after RTA-
With a large enough vacancy supersaturation during growth of the heterostructure,
dislocation loop nucleation will be controlled solely by vacancy migration. Vacancy
formation (E:' = 2.0 eV) wouid not be required. Therefore, the activation energy for
nucleation (QJ wouid be sufncientiy reduced near the value of E~'' = 0.33 documented
by Van Vecheten (1980). Low temperature MBE growth procedures would have to be
utilized in order for enough vacancies to be grown into the materid. However, if the
growth temperature is to low and the growth rates are not modifieci, nano-scale void
formation will r e d t (Section 3.5) due to v m c y condensation. In order to produce
these structures with a sufnciently large vacancy concentration, the growth rates of Si and
Ge would have to be decreased to d o w enough tirne for the surface migration of atoms
on the substrate at these reduced growth temperatures (Le. 400-450°C). UHV-CVD
couid not be used for these low temperature experiments because the temperature mut
remain constant at approximately 52S°C to maintain a base pressure of 1.5 x 10" mbar
necessary in the growth chamber.
Figure 32.7 shows the misfit dislocation nucleation rates for the CVD-61 material
and a low-temperature MBE material (i:e. MBE-1707). It is obvious here that a different
stmin relaxation mechanism is fiinctioning in the low-temperature MBE matenal. The
slope of the plot of this material was reduced giving an activation energy of 0.5 f 0.05
eV. Therefore, a suficiently large vacancy supersaturation h a . dominated here with
respect to the other structures grown at higher temperatures and growth rates. Vacancy
migration is the operative mechanism for low-T MBE material. The densities of misfit
dislocations were also affected as seen in Figure 3.2.8. The resuiting activation energy
dropped to 2.0 f 0.5 eV, exactly 2.3 eV lower than the overall strain relaxation values for
UHV-CVD matenal. Thus, for the low-T MBE growths, 0.5 t 0.05 eV c m be attributed
to the vacancy migration. These values are consistent with the predicted strain relaxation
mode1 proposed above.
To get a more accurate assessrnent of the functioning mechanism, positron
annihilation spectroscopy has ken conducted on the MBE material and the as-grown
CVD-9 matend as a control specimen, due to the low density of initial grown-in defects
Temperature (OC)
7.6 8.1 8.6 9.1 9.6
Temperature (1 /K) (1 €4)
Figure 3 -2.7: Nucleation rate data for UHV-CVD and low-T MBE material
Temperature (OC)
7.6 8.1 8.6 9.1 9.6
Temperature (1 /K) (1 €4)
Figure 3.2.8: Misfit dislocation deasity data for UHV-CVD and low-T MBE material
Depth (angstroms)
Positron energy (keV)
Figure 3.2.9: Positron Annihilation Spectmscopy data for low-T and UHV-CVD material
(No 1o2 cm'2). According to Figure 3.2.9, the S-parameter has been modeled for the
various MBE structures. The S-parameter in the graph corresponds to the ratio of the
number o f y-rays in the central region of the peak (5 10.27 to 5 1 1.73 keV) to the total
number in the peak (504.51 to 517.49 keV). The lower the Doppler bmadening, the
higher the S-parameter-
Ail the samples have the same S-panuneter for high positron implantation energies
indicating that there are no signifi~ant ciifferences deep in the samples. For some of the
samples (1702, 1703, and 1706), the graph is less steep in the beginning for positron
energies less than 2 keV. This means that the hction annihilating at the sdace does not
change much for increased implantation energy. This codd be caused by an electric field
driving the positrons back to the surface. Another possible expianation is that the= codd
be a layer at the surface which is different in some way, like a layer of oxide. This codd
be around 400 A for sample 1702. CVD-9 could be fit well with a model having no
defects. However, 1707 could be fitted using a model with a different sudace S-
parameter and assiiming a fiactional vacancy concentration of about 5 x 1017
(Simpson et a[., 1 994). Therefore, MBE- 1707 may have displayed a behaviour caused by
a high vacancy concentration after growth, which is consistent with the bulk
rneasurements conducted (Figures 3.2.7 and 3.2.8).
3.3 Sources of Nucleation
The sources of nucleation and the subsequent expansion of misfit dislocations have
been studied by many researchea. Perovic et al. (1989) found that coherent /?-Sic
precipitate plates, localized at the Ge&,/Si interface acted as efficient sources of 60'
misfit dislocations. Hurnphreys et al. (1991) observed the "diamond defea", which was
a faulted disiocation loop which acted as a heterogeneous source of 60' misfits.
However, heterogeneous sources like these can be controlled and are a result of specific
materials systems or growth conditions. Therefore, a generaüzed nucleation process for
low misfit heterostructures was required that was consistent with measined activation
energies regardless of growth conditions and procedm (Le. MBE CVD, RTCVD etc.)
for low misfit heterostructura.
The formation of dislocation loops in the interface has been recognked as the widely
accepted source for nucleation of the misnt segments (Jhmat et al., 1990, Jain et al.,
1995, Zou et al., 1996). Hirsch (1997) examined the conditions under which the Ge-
platelets act as sites for nucleation of "double half-loops" which expand in the epitaxial
layer and act as sources for the 60' misfit dislocations. Hirsch's theory of nucleation is
consistent with the mode1 developed by Perovic and Houghton (1992) and has dso been
examined here.
The sources of nucleation for the dislocation loops and the expansion of the incipient
loop to fonn stable misfits are attributed to the loss of coherency at interstitiai
perturbations (Le. Ge-rich platelets). Figure 3.3.1 shows a TEM image obtained in strain
field contrast of the Ge-nch perturbations. The g = 12201 was used to find the small
platelets (diameters - 1.5 nm, - 2-3 monolayers thick). DF imaging was used for the
strain field images because of the anomalous absorption effects. The high intensity
region corresponds to a region of good transmission where Bloch wave II predominates
since its intensity is concentraied between the difbcting planes. The adjacent region
trmsmits electrons poorly since Bloch wave 1, with its intensi~ peak at the dificting
planes, dominates and thus appears dark relative to the background (Perovic et al., 199 1 ) .
Images in BF are symmeîrical and therefore, it is more difficult to show this wntrast
behaviour. The contrast over the platelet goes h m Light to dark in the direction of the g-
vector, which means that the platelet is interstitially strained. Plateiet formation was
assumed to be an inherent feature of MBE growth (Rowell et ai., 1993). It was assumed
that the hydrogen-passivated alloy surface typical in CVD prohibited these features From
forming. However, platelets were indeed found in the CVD material as well as the MBE
material. The suain relaxation mode1 held for both growth techniques as seen in Section
3.2.
The group of Rowell et al. ( 1993) found that upon tilting the sample in the TEM to
othsr zone avrs other than the [001], the associated image conuast relationship indicated
that the perturbations are atornic-scale regions possessing plate-shaped symmetry and a
Figure 3.3.1: TEM dark field image of Ge-rich platelets in CVD-9 sample (g = [220])
larger larrice parameter than the adjacent matrix region. They found perturbations in a
wide range of single and multiple layer heterostructures. The defects were never
observed in the thin foi1 regions away h m the heterostnicture thus d i n g out specimen
damage-related artifacts as possible defect sources. A broad photoluminescence (PL)
peak, shifted lower in energy than the phonon-resolved PL, has been assigned to the
formation of these perturbations indicating an increase in the strain energy density. The
shift occurs since excitons in the Ge&, are localized in the lower band-gap Ge-rich
platelets (Rowell et al., 1993).
With the use of an energy balance approach (see Appendix), Perovic and Houghton
(1992) determineci that a criticai radius condition can be detennined where the strained
perturbations can iower its energy upon creating a prismatic loop at its interface. It was
also found that for large misfit strains (Le. x 2 0.85, for GexSii,/Si), Ge platelets
embedded in Si will homogeneously generate an interfacial dislocation loop (vacancy-
type) without having to overcome an activation energy through pst-growth ameaiing.
Hence, the term "banierless" misfit dislocation nucleation was coined- At Lower rnisfit
strains, the critical radius is important, and requires pst-growth annealing to generate
misfit dislocations at the interface. Hirsch (1997) has also argued that loop nucleation is
more likely in larger perturbations and that the rate controlling process is the diffusion of
vacancies to the perturbations.
After the generation of the prismatic dislocation loop, the nucleation of the half-loops
and their expansion in the superlattice is the next step (see Figure 1.1 4. In accordance
with Hirth and Lothe (1 982), it is possible that prismatic vacancy loops on ( 100) planes
can act as sources of dislocation shear loops on (1 11) planes in the presence of a
resolved stress on the glide plane (i.e. retr). Therefore, segments of the { 100) prismatic
loops can bow out ont0 the { 1 1 1 } planes, and form haif-loops.
Perovic and Houghton (1 992) used the total Helmoltz energy equation: E, = E, - Ed t
Es - TS, where Ed is the dislocation half-loop self-energy, E, is the misfit strain energy, Es
is the surface step energy, and TS is the entropy of the loop (see Appendix for the
complete goveming equations). The activation energy for nucleation of misfit
dislocations is obtained from this expression by substituting in the cntical radius (R') of
the Ge perturbation. Therefore there will exist a range of effective stresses, increasing
with perturbation size that facilitate a driving force for reduction of the activation barrier
for rnistit dislocation injection. [t is these sources that f o m misfit dislocations after
RTA. This mode1 is consistent with the results found in this study.
Figure 3 - 3 2 and figure 3 - 3 3 are TEM images of CVD-9 after RTA for 5 seconds at
1 OOOUC. Figure 3.3.2 is a BF image of orthogonal misfit expanding out from an original
nucleating source. Figure 3.3.3 shows a DF image of orthogonal segments nucleating
fiom the Ge-platelet source. Note that the DF image dislocation lines display higher
resolution due to the isolation of the diffracted beam conditions achieved with the
objective aperture.
Figure 3 -3 -2: TEM BF image of CVD-9 after RTA for 5 s at 1 OOO°C
Figure 3.3.3: TEM DF image of CVD-9 after RTA for 5 s at 1000°C
3.4 Ion Imptantation
bLuch work has been published on ion implantation of heterostructures to snidy the
intermixing of quantum weki (Labrie er al., 1996 and Charbonneau et al., 1995) and the
formation of secondary and extended defect structures (Goldberg et al., 1995). Jaraiz er
ul. ( 1995) produced atomistic caiculations of ion implantation in Si. Their study of point
defsct and transient enhanced diffusion (TED) of dopants in Si is consistent with the now
îàmous -'+I" mode[, which States that the Si interstitial excess is assurned to equal the
implanted dose. It was aiso observed that d e r subsequent anneaiing, the fiee vacancies,
with their higher diffusivities, diffuse and annihilate most of the interstitials produced
during ion implantation. The dominant process being recombination. Some of the
vacancies reach the surface like a perfect sink and annihilate there, Ieaving behind an
excess of interstitials in the bulk materiai (Jaraiz et al., 1995).
In this study, the area of concem has been focused on the direct primary eff- of ion
implantation in the superiattice. As stated earlier, the Ge-rich platelets, which possess an
interstitial elastic saain field, are relieved by the injection of point defects via ion
implantation. These small Ge perturbations require the diBision of vacancies to produce
the incipient dislocation loop. The excess interstitiah, as stated above, make this pro-
more dificuit by "rnopping up" the vacancies left behind. The relationship between the
spatial distribution of the secondary defects and the original primary effeçts is unclear
(Goldberg et al., 1995).
Ion implantation range profiles are very easily calcdated using wmputer simulations
(Le. TRLM and binary collision approximation @CA)) due to the relative simplicity of
the collision processes. However, ion implantation damage after subsequent annealhg is
far more complex. Molecuiar Dynamics (MD) provides detailed predictions of the
damage but only for the first nanoseconds because of its heavy compuiational burden
(Jaraiz et al., 1995). It has been demonstrated in Section 3 2 that the primary effets of
ion implantation have assisted in the reduction of misfit dislocations by a vacancy
migration mechanism after RTA.
Secondary defect formation has also been observed in the cross-section TEM studies
(Figure 3.4.1). TRIM (Ziegler et al., 1985) calculations have been performed and the
depth profiles of the implanted ions have k e n produced for a 10-period Ge,&, (x =
0.09) (hsioe = 150 A, Hsi = 210 A) heterostnicture implanted with 540 keV Si ions,
sirnilar to the CVD-9 materia! of this shidy (see Appendk). The results were consistent
with those found in this TEM obsewation and aui be seen in the Appendix. The Si
implanted ions end-of-range lie between 6000 - 10000 A. This is the region where al1 the
secondary darnage can be observed. The top four layen of this CVD-9 cross-section
sample in figure 3.4.1 have been previously etched away in attempt to reveal the misfit
dislocations at the interface for Nomarski study.
After an 8j0°C RTA for 30 seconds, dislocation loops, dislocation dipoles and <3 1 1>
md dekcts were the structures making up the secondary damage. consistent with the
observations of Goldberg et al. (1 993). Figure 3.1.2 shows a plan-view image of the
same sample. In plan-virw, the depth of the implantation darnage c m not be determined
because from a top view, dl structures seem to be lying on the sarne plane. However, the
long orthogonal misfit lines running dong <110> directions can cleariy be distinguished
Figure 3 - 4 1 : TEM BF image of implantation profile in cross-section (z = [O 1 11) of CVD-9 RTA for 30 s at 850°C
from the rod-like defects and dislocation loops formed from the implantation and RTA.
Therefore. what has been introduced here is a very feasible solution for decreasing the
drnsities of misfit dislocations. The thin film of Ge,Sii., is what we are concerned with
in designing new electronic devices. The Si substrate wafer is of no use for the ultimate
device fabrication and will be discarded. Thus, the secondary defect stnictures present in
the substrate will not affect the outcome of the devices.
Figure 3.42: TEM plan-view image of implanted CVD-9 RTA for 30 s at 8jO0C
3.5 Low-Temperature MBE Heterostructures
The use of low temperature MBE has resulted in the irnprovement of coevaporative
doping control and the production of highly metastable, clastically strained
heterostructures. However, there exists a limit for the growth temperanire whereby
planar 2-D growth stops and the breakdown of the (100)-oriented growth occurs. Low
temperature growth of heterostructures have been examined in the past by Jorke et ai.
( 1989). They studied the breakdown of the epitaxiai growth at low temperatures as a
hnction of Si deposition rate for both constant and variable temperature growths. The
transition From single crystalline to polycrysralline (twinned) and finally arnorphous
growth was observed. Perovic et al. (1 991) studied the microvoid formation in low-
temperature MBE structures. They discovered the presence of small spheaicd structures
which accompanied the breakdown of the (100)sriented growth of thae superlattices via
the formation of { 1 1 1 aiented facets.
An interesting phenornenon has taken place in the low-temperature (- 400°C) MBE-
grown heterostnictures with the breakdown of (100)-oriented growth. TEM imaging has
been perfonned in cross-section to reveal the (100) surface of the superlattices. FE-SEM
secondas electron (SE) detector imaging has been performed to show the surface of
these structures and the orientation of the pits. Positron annihilation spectroswpy was
conducted as an additional study to observe the nahue of the S-parameter as a h c t i o n of
implant energy. An idea of the defect concentrations (i.e. voids) was observed fiom this
data.
Samples MBE-1688 and MBE-1690 were chosen for this study. These samples were
initially grown to closely resemble the structured geometry of CVD-9. However, with
the reduced growth temperatures, it was reveaied by a Shïmmel etch and Nomarski
interference microscopy that misfit disiocations were not being nucleated at the interface.
Therefore, other methods of examination had to be employed to detemiine what
mechankm was taking place. M e r reviewing some literature (Perovic et al., 1993),
Figure 3-51 wvas found. According to the growth parameters used to fabricate MBE-
1688 and MBE-1690, the formation of voids was eminentiy probable. Although Figure
3.5.1 shows the morphoiogicai instability phase diagram for Geo-xSio.7s/Si, an
approximate estimation for the behaviour of G e ~ & i ~ - ~ ~ / S i (Le. MBE-1688, -1690) was
made. For a total thickness of 360 nm for the entire superlattice, a growth temperature of
- 400°C, and a growth rate of 0.5 d s , void formation would be the result.
200 300 400 500 600 700 800 . 900 1 0 0
Temperature ('C)
Figure 3 S. 1 : Morphological instability phase diagram describing cntical thickness instabilitia as a function of growth temperature
FE-SEM SE-detector imaging was used to examine the surface of the samples. The
low-kV (i.e. 1 kV) experiment was used to obtain images with s m d interaction volumes.
Figure 3.5.2 shows a SE-detector image with the presence of pits on the surface of the
MBE- 1688 sample. It was found later, through TEM observation that these features were
actually surface cusps with { l l l}onented facets. Figure 3.5.3 shows the TEM cross-
section image. The faceted growth suiface can be cleady resolved here. The (100)
surface breaks down into a series of cusps and peaks, with each cusp having a pyramidal
structure bounded by ( 1 1 1 } facets. It has been demonstrated by Perovic et al. (199 1) that
if the Si epitaxial thickness is increased, cylindrical voids are le& in the wake of the
migrating surface at each cusp. These cylindrical channels become morphologically
unstable and break up to fom a senes of aiigned voids (i.e. microvoids). Due to the
small thicknesses of the Si spacer layers of MBE-1688 and MBE-1690 (Le. 2 1 nm of Si),
the spherical microvoids were never observed in these samples.
Figure 3-52: FE-SEM SE-detector image showing presence of pits on surface
Figure 3 3.3 : TEM cross-section image revealing cusps with ( 1 1 I }-oriented facets
The low growth temperatures encountered in the MBE faci!lty accounted for the
formation of the surface cusps with { I 1 1 } -0riented facets. If allowed to proceed, with
larger Si spacer thicknesses, microvoids would form in the wake of the migrating cusp.
At these Iow growth temperatUres, the thermal energy produceci is low enough to
decrease the surface migration energy of atoms and the proper bonding sites at the fke
surface are not reached. Vacancies agglomerate and form these cusp structures.
Figure 3.5.4 shows the positron annihilation spectroscopy data for MBE-1688 and
MBE- 1690 performed at the University of Western Ontario. The beam of monoenergetic
positrons is implanted into the sarnple where the positron loses its energy (thennalization)
and then diffiises through the soiid until annihilating either from this fieely diffushg state
or &om a trapped state (bulk defect or a surface state) as described in Section 1.5. Two
specimens were taken from each sample (i.e. an as-grown sample and an annealed
sample). It was observed here that the S-parameten of the as-grown samples both
indicate the presence of cusps due to the steep incline leading to the sharp peaks at
approxirnately 5 keV. Interestingly, the S-parameter is higher in the material grown at
higher temperature (Le. MBE- 1688). Alîhough the higher S-parameter indicates that the
vacancy concentration is higher, it may mean that the defect structure is slightly different
in the two samples. The fiaction annihilating at the surface (i.e. Fs see section 1.5) is the
dominant factor in the S-panuneter expression and yields the steep peaks in the S-
parameter of the as-grown materiai. The presence of the large cusps bounded by the
{ I 1 1 1-oriented facets resdted in a high Doppler shift giving a high S-panuneter. M e r
annealhg the samples at 950°C for 3 0 seconds, the S- parameters greati y decreased,
indicating that the vacancy structures diffirsed during the annealing pmcess leaving a
lower vacancy concentration behind as shown in positron annihilation data of Figure
3 -2.9.
a. 1 = 16a8 M8E as grown
a P 1690MBE A r r A A CIZA
I
Figure 3.5.4: Doppler-broadened annihilation line-shape parameter S as a function of positron energy
The S-parameter represented above displays a value of Sd = 0.56 for the defected
region at approximately 5 keV for MBE 1688. At higher energies (> 10 keV), al1
annihilations occur in the fieely diffusing state (Le. void-net, defect-fiee) and is
represented by an S-parameter value of Sr = 0.5 1. The ratio (Sr / Sd) is 0.91, which is
consistent with other reported values for silicon for monovacancies or divacancies,
impurity-vacanc y or inters titial complexes (Perovic et al., 1 99 1).
CHAPTER 4 - CONCLUSIONS AND RECOMMENDATIONS
The study of defect formation in low misfit Ge,Sii.JSi heterostruchtres grown by
UHV-CVD and MBE has been examineci and based on a quantitative @y, it has
been dernonstrateci that the onset of strain relaxation via misfit disiocations can be
controlled by point defect injection. Point defects, which are generated by ion
implantation, will diffuse during post-growth rapid thermal annealhg (RTA) and
subsequentiy relieve the stress concentrations surroundhg growth-induced Ge
perturbations.
These Ge-cich platelets, fouad only by straïn field contrasting in the TEM, were
previously assumed to be inherent features of MBE. This study has lead to their
discovery in CVD material as weli.
The activation energy for nucleation of misfit dislocations, which was determined by
large ara bulk measurements (Le. Nomarski Interference Microscopy), was found to
be Q. = 2.5 f 0.5 eV. This universally accepted value for nucleation of rnisfit
dislocations remained constant for as-grown and implanteci samples regardless of
growth method (i.e. MBE or CVD), indicating that misfit dislocation nucleation is the
rate-limiting step in the strain relaxation process.
The overall strain relaxation activation energy was found to be Q, = 4 2 f 0.5 eV for
the samples, again regardes of gtowth methud, which is consistent with combining
the nucleation and glide energy terms in series (i.e. Q, = Q. + Qv).
0 The low-temperature MBE growth experiments yielded a value of 0.5 f 0.05 eV,
which represented the energy associated with vacancy migration. With the large
supersaturation of vacancies encountered with the low-temperature MBE growths, the
nucleation of the dislocation loops would be solely controlled by vacancy migration if
vacancy aggregation to form srnail cluster nuclei had already occumd.
a The vacancy formation energy term was dropped. Accordingly, the overall strain
relaxation energy value also decreased to a value of 2.0 f 0.5 eV, indicating once
more that vacancy migration is the rate-lirniting step in the Iow-temperature MBE
growths.
r Without monitoring the growth rates in low-temperature MBE material, the growth
surface of these structures was no longer found to remain planar. Low-kV FE-SEM
SE-detector imaging revealed surface pits on the samples. A senes of cusps with
(1 l 1)-oriented facets were observed with cross-sectional TEM. If the Si spacer
layen were increased, these facets were assumeci to generate linear arrays of spherical
microvoids in the wake of their growth.
By combining rnacroscopic and microscopie analyses dong with positron
annihilation spectroscopy and ion implantation, the semi-empincal predictive model
for strain relaxation developed by Perovic and Houghton (1992, 1993) has been
thoroughly examined to describe the early stages of misfit stralli relief for Ge,SiiJSi
(100). The behaviour of misfit dislocations in the Stage I regime (misfit densities <
10' cm-2) was consistent wîth the predicted model. Even in the presence of a large
vacancy supersaturation, the results still remain consistent with the model.
Perhaps fiiture studies can be conducted to fully understand the entire relaxation
processes across al1 the stages (Le. Stage I - III). Multiplication and interaction
effects of dislocations would need to be investigated to explain these stages of seain
relaxation.
Finally, a novel approach to reducing misfit dislocation formation via ion
implantation has been investigated. This process can be incorporated into the
fabrication of these heterostnictures in the manufachuhg of electronic devices such
as heterojunction bipolar transistors (HBT) and light-emitting diodes (LED).
Effective Stress and Kinetic Mode1 (Houghton et ai., 1994)
The effective stress is determined by the imbalance between the resolved shear stress
acting on the t h d i n g dislocation slip system and the iine tension in the extending a/2
4 10> misfit dislocation segment The effective stress due to misfit strain, expenenced
by the threading a m of an a/2 4 IO>-type 60' dislocation of length h, assuming isotmpic
behaviour and equai elastic moduli, is given by:
where yl is the angle between the strained interface normal and the slip plane, B is the
angle between the dislocation iine and its Burgers vector, and R is the angle between the
Burgers vector and the direction in the interface, normal to the dislocation line. p is the
shear modulus, v Poisson's ratio, ~ ( h ) the in-plane strain at position h, H is the thickness
of the layer through wfuch the threading dislocation cut, Hi is the displacement of the ith
misfit dislocation fiom the fkee surface. The core parameter ,&? = 4 is assumed. The Si in
the second term represents the increase in misfit dislocation tension due to dislocation-
dislocation interaction for the case where Aa(h) AT is introduced to account for the extra
strain due to differentiai expansion on heating to the anneal temperature.
For a single GexSii,/Si mdtilayer structure with N periods of altemating GeSi and Si
of thicknesses h and H, and with strain relaxation at the first GqSii,/Si interface, the
effective stress r , ~ (in GPa) h m above may be expressed as:
0.55 ln 1 ON(h + X))
The misfit nucleation rate ( d 2 s - ' ) for any (100) oriented Ge&.JSi (x < 0.5)
heterostructure grown by MBE or CVD cm be summarized by the following semi-
empîrical relationship:
where No is the initial misfit dislocation source density present at t = O(i.e. substrate
threading dislocations, residual subshate-buffer layer precipitates etc.), k is the
Boltzmann constant equal to 8.62 x 10" eVK and T is the anneal temperature in Kelvin.
r , ~ is given in units of MPa respectively. Q, is the activation energy for nucleation, n and
B are material constants and ,K is the shear modulus.
Using the classic Orowan plasticity equation (dddt = N(t) V(I)b.d, where befl is the
effective Burgers vector magnitude of a 60' dislocation projected in the strained
interface. the overall straïn relaxation rate (s") is given by:
where al1 parameten are material constants is the shear modulus) or experirnentally
detennined for the given materials system. Qn + Qv represent the overoll strah relaxation
activation energy, Q,.
Enerw Balance A~proach (Perovic and Houghton, 1992)
To determine the nucleation of misfit dislocations, the first step in the process was to
consider the generation of the interfacial prismatic dislocation loops via the loss of
coherency at Ge-nch platelets. Weatherly (1 968), Ashby and Johnson (1 969) and Brown
and Woolhouse (1970) nrst developed the theory to explain coherency breakdown at
precipitates in metallic systems. (EJ, the elastic energy of a wherent platelet is baianced
against the self-energy of a dislocation loop (Ed) to get the foiiowing total energy Et = E,
- Ed:
where , p is the shear modulus, v is Poisson's ratio, e is the Naperian base, b is th
magnitude of the Burgers vector and &' is the unconseained lattice misfit strain between
the precipitate and rnatrix. R is the dislocation loop radius where the inclusion sûah
energy is a minimum. O is the correction factor for the approximate infuite body self-
energy. The dislocation core parameter, a I 1.
The next step is to consider whether these loops can lead to misfit dislocations at strained
layer interfaces. The total Helmholtz energy balance equation (2) was used. E = E, - Ed
f Es - TS, where Ed is the dislocation half-loop self-energy including the appropriate
correction factor O for a semi-infinite body and a = 1. Es represents the energy gained or
lost during the creation or removal of a surface step and TS is the entropy. The governing
equation for (100)-oriented interfaces in the diamond cubic slip system ((1 1 1 } 4 IO>)
can be written as:
The Ec term has been modifieci by replacing the average lattice misfit strain (G) by an
effective strain Gr which is a superposition of the strain field of the perturbation with that
of the strained layer. From equation (21, one cm calculate the critical radius for
nucleation (R') upon maximizing with respect to R. The activation energy for nucleation
(E') is obtained by substituthg R ' into (2).
Bulk Quantitative Raw Data
Misfit Dislocation Nudeation Rates (~rn-~sec-9
Trials Temperature (K-')
0.000786 O.OOO8 18 0.000853 0.00089 1 0.000932
MBE- 1 703 ( A s - ~ w I ~ ~
CVD-9 (As-grown) Triais
Temperature O<-') 0.000786 0.0008 18 0.000853 0.00089 1 0.000932
MBE- 1707 (As-grown) - Trials
Temperature (Ki) 0,000786 0.0008 18 0.000853 0.000891 0.000932
1
900 400 130 40 10
1
300 250 170 120 80
2
879 356 135 36 8
3
940 378 146 32 16
2
287 236 164 127 74
3
326 289 156 115 76
Misfit Dislocation Densities (cm-')
CVD-6 1 (Irnplanted) ,
t
Triais Temperature (K")
0.000786 0.0008 18 0.000853 O-O0089 1 0.000932
Trials Temperature (KL)
0.000786 0-0008 18 0.000853 0.00089 1 0.000932
Triais Temperature (RI)
0.000786 0.0008 18 0.000853 0.00089 1 0.000932
1
900 190 65 18 6
2
845 1 84 57 26 9
MBE- 1703 (AS-grown)
3
930 206 75 24 11
~tials Temperature (Ki)
0.000786 0.0008 18 0.000853 0.00089 1 0-000932
1
3 100 980 1 70 70 17
2
3087 954 156 73 14
3
3024 932 179 61 19
Low-T MBE MBE-1707 (As-gro~n)
Trials Temperature (TC')
0,000786 0.0008 18 0.000853 0.00089 1 0.000932
2
135 53 23 15 7
1
150 60 30 14 7
3
165 58 35 14 6
TRIM Simulation @y R Goldberg, UWO)
6000
Depth (A)
Ashby M. F. and Johnson L. 1969 Phil. Mug. 20 1009
Baribeau J-M, Headnck EL L., and Maigné 1994 ScaMug Mimoscopy 8 No. 4 813-826
Bean J. C., Feldman L. C., Fiory A. T., Nakahara S., and Robioson 1. K. 1984 3. Vac. Sci. Technol. A2 436
Bean J. C. 198 1 Impurity Doping Processes in Silicon, edited by Wang F. F. Y. (Amsterdam: Elsevier/North Holland) 183
Brown L. M. and Woolhouse G. R 1970 Phil. Mag. 21 329
Charbonneau S., Poole P. J., Piva P. G., Aers G. C., Koteies E. S., Faliahi M., He J-J., McCafEey J. P., Buchanan M., Dion M., Goldberg R. D., and Mitchell 1. V. 1995 J. Appl. Phys. 78 3697
Cressler J. D. 1995 IEEE Spec- 49
Frank F. C. and Van der Mewe J . 1949 Proc. Roy. Soc. Lond. A 198 216
Giedd R. E., Moss M. G., Ka* J., Wang Y. Q. 1996 Electrical Applications of ion-Imp/anted Po&mer Films
Goldberg R. D., Simpson T. W., Mitchell 1. V, Simpson P. J., Pduyl M., and Weatherly G. C . 1995 hrucl. Imtr and Meth. B 106 2 16-221
Goldstein J. t., Newbury D. E., Echlin P., Joy D. C., Fiori C., Lifshin E., 1981 Scanrzhg Electron Microscopy and X-Rqy Microanalysis Plenum Press, New York
Grovenor C. R. M. 1992 Microelectronic Maferials (London: IOP Publishing Ltd.)
Hartley N. E. W. 1976 Appkztionr oflon Beam tu Mnterials edited by Carter Ge, Colligon J. S., and Grant W. A. (London: IOP Publishing Ltd.) 210
Haynes R. 1984 Optical Microscopy ofMateriaZs, International Textbook Company
Hirsch P. B. 1 997 Imt. P hys. ConJ Ser. In press
Hirth J. P. and Lothe J. 1982 Theory of Dislocations, 2"d ed., (New York: Wiley) 23 1 757
Houghton D. C., Baribeau J.-M., and Rowell N. L. 1995 J. Mater. Sci. in Elec. 6 280-291
Houghton D. C., Davies M., Sudersena Rao T., and Dion M. 1994 1 Cyst. Gr. 136 56-
63
Houghton D. C., Noël LP., Rowell N. L., and Perovic D. D. 1993 Si-Ge Strained Luyer Heterostructures: Device Possibilities and Process Limitations
Houghton D. C. 199 1 J. Appl. Phys. 70 (4) 2 136
Houghton D. C. 1990 Appi. Phys. Let?. 57 (14) 1434
Houghton D. C. 1990 Appl. Phys. Lett. 57 (20) 2 I N
Hull R and Bean J. C. 1989 J. Vac. $ci- Techmi. A7 (4) 2580
Bull R, Gibson J. M., and Bean J. C. 1985 Appl. Phys. Lett. 46 179
HumphRys C. J.?Maher D. M., Eaglesham D. J., Kvam E. P., and Salisbury 1. G. 199 1 J. de Phys. III 1 119
Jain U., Jain S. C., Harker A. H., and Bdough R 1995 J. Appl. Phys. 77 103
Jaraiz M., Gilmer G. H., Poate J. M., and de la Rubia T. D. 1996 Appi. Phys. Let?. 68 (3) 409
Jesson D. E., Chen K. M., and Pemycook S. J. 1996 MRS Bulletin 21 (4) 3 1
Jorke H., Herzog H.-J., and Kibbel H. 1989 Phys. Rev. B40 2005
Kamat S. V. and Hiah J. P. 1990 J. Appl. Phys. 67 (1 1) 6844
Kanter H. 1961 Phys. Rev. 121 677
Labne D., Aers C., Lafontaine H., Williams R L., and Charbonneau S. 1996 Appl. Phys. Left. 69 3866
Lafontaine H., Houghton D. C., Elliot D., Rowell N. L., Baribeau J.-M., Laframboise G., Sproule G. I., and Rolfe S. J. 1996 J. Vac. Sci. Technol. BI4 (3 ) 1675
Matthews J. W. 1972 Surf Sci. 32 1
Meyerson B. S. 1992 Proc. IEEE 80 (1 0) 1592
Meyerson B. S., Legoues F. K., Nguyen, and Harame D. L. 1987 Appl. Phys. Lett. 61 785-787
People R. and Bean L C. 1985 Appl. Phys. Lert. 47 322
Perovic D. D. and Houghton D. C. 1995 Inst. Phys. Conf: Ser. (146) 1 17
Perovic D. D. and Houghton D. C. 1993 Phys. Stat. Sol. (a) 138 425
Perovic D. D., Bahierathan B., Houghton D. C., Lafontaine H., and Baribeau J.-M. 1993 Mat. Res. Soc. Symp. Proc. 399 325
Perovic D. D. and Houghton D. C. 1992 Mut. Res. Soc. Symp. Proc. 263 391
Perovic D. D., Weatherly O. C., Simpson P. J., Schultz P. I., Jackman T. E., Aers G. C., Noël J.-P., and Houghton D. C. 1991 Phys. Rev. B43 (17) 14257
Perovic D. D. and Weatherly G. C. 199 1 (nimmicroscopy 35 27 1
Perovic D. D., Weatherly G. C., and Houghton D. C. 1990 Mut. Res. Soc. S ' p . Proc, 160 65
Perovic D. D. and Weatherly G. C. 1989 Thin Solid Films 183 14 1 - 156
Perovic D. D. 1988 M. A. Sc. Thesis, Department of Metallurgy and Materials Science, Univ, of Toronto
Rajan K, Fitzgerald E., Jagannadham K., and Jesser W. A. 199 1 TMS 1401 861
Rowell N. L., Noël J.-P., Houghton D. C., Wang A, Lenchyshyn L. C., Thewalt M. L. W., and Perovic D. D. 1993 J. Appl. Phys. 74 2790
Schultz P. J. and Lynn K. G. 1988 Rev. Mod Phys. 60 701
Schwarz K. W. 1997 Rev Phys. Len. 78 (25) 4785
Schwarz K. W. and Tersoff J. 1996 Appl. Phys. Lett. 69 (9) 1220
Simpson P. J., Vos M., Mitchell 1. V., Wu C., and Schultz P. J. 1994 Mater. Res. Soc. Symp Proc. B44 (22) 121 80
Stirpe M. B., Perovic D. D., Lafontaine H., Goldberg R. D. 1997 IW Phys. Conf: Ser. (1 57) 127
Tanaka K., Suezawa M., and Yonenaga 1.1996 J. Appl. Phys. 80 (1 2) 699 1
Thomas G. and Goringe M. J. 1979 Trrrnrnission Electron Microscopy of Materiais, John Wiley and Sons, Inc.
Van Vecheten J. A. 1980 Hmidbook on Semiconductors, Vol. 3 edited by T. S. Moss (Amsterdam, North Hoiland) 1
Weatherly G. C. 1968 Phil. Mag. 17 791
Wickenhauser S., Vescan L., Schmidt K., and Liith H. 1997 Appl. Phys. Lett. 70 324
Williams D. B. and Carter B. C. 1996 Tranrmission Elechon Microscopy Vol. I I I Plenum Press, New York
Zhao A. 1993 B. A. Sc. Thesis, Dept. of MetaIlurgy and Materials Science, Univ. of Toronto
Ziegler J. F. and Biersack J. P. 1985 The Stopping Range of Ions in Malfer, Pergamon Press, New York
IMAGE EVALUATION TEST TARGET (QA-3)
APPLIED - IMAGE. lnc = 1653 East Main Street - -. , . Rochester. NY 14ôû9 USA -- -- - - m e : 71 W482-O3OO -- -- - - Fax: 716i288-5989