new all rights reservedepubs.surrey.ac.uk/844322/1/10148804.pdf · 2018. 6. 20. · simox...
TRANSCRIPT
All rights reserved
INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a com p le te manuscript and there are missing pages, these will be noted. Also, if materia! had to be removed,
a note will indicate the deletion.
Published by ProQuest LLC (2017). Copyright of the Dissertation is held by the Author.
All rights reserved.This work is protected against unauthorized copying under Title 17, United States C ode
Microform Edition © ProQuest LLC.
ProQuest LLC.789 East Eisenhower Parkway
P.O. Box 1346 Ann Arbor, Ml 48106- 1346
C h a ra c te r iz a t io n a j t f C o n t r o l o f C ry s ta llo g ra p h ic D e fe c ts T h in f i l m S IM O X M a te r ia ls I H I w
V
Submitted for the degree of
Doctor of Philosophy
to
Department of Electronic and Electrical Engineering
University of Surrey
Luis Felipe Giles
December 1994.
0 1994, Luis Felipe Giles
A b s t r a c t
Crystallographic defects present in the silicon overlayer of thin (< 1000 A) SIMOX material have been characterized using a newly developed etchant and by transmission electron microscopy. Rutherford backscattering spectroscopy has also been used to determine the chemical composition and thickness of the synthesised layers. The thin SIMOX layers were produced by two different methods, namely (i) sacrificial oxidation of the silicon overlayer of thick (2000 A) SIMOX films and (ii) low energy 0 + implantation.
The main crystallographic defects present in materials prepared by sacrificial oxidation are threading dislocations and oxidation induced stacking faults (OISF) whose density and size depends mainly upon the oxidation conditions (temperature, time) and also on the number of nucleation sites present before oxidation. The nucleation of these OISF has been investigated and it has been observed that stacking fault (SF) complexes are the main nucleation sites. The density of threading dislocations within the Si overlayer does not increase during oxidation. The lowest density of OISF (5,0xl02 cm"2) was observed in layers thinned by dry oxidation although some layers contained up to 5,0x105 cm'2. This difference in OISF densities was attributed to differences in the densities of the SF complexes before oxidation.
Thin film SIMOX structures formed by low energy implantations also contain two principal defect types namely, threading dislocations and stacking fault complexes, where the densities depend upon the implantation and annealing conditions. A low density of threading dislocations is only obtained when parameters, such as dose and implantation temperature are optimised. Furthermore, it is also observed that thermal and mechanical stresses produced in the silicon overlayer during implantation or annealing, need to be minimised in order to obtain low defect density material. It has been shown by whole wafer defect mapping of six inch SIMOX wafers implanted under optimised conditions, that a uniform distribution of defects can be achieved, having an average defect density as low as 1.0xl04cm'2.
Additionally, the effects of implantation damage on the formation of secondary defects have also been investigated. The results have shown that the coalescence of point defects generated during implantation produce a high density of dislocation loops that, depending upon the annealing treatment, develop into threading dislocations or OISF.
These experiments confirm that careful optimisation of the processing conditions, such as implantation temperature and dose uniformity, can significantly reduce defect densities thus enhancing the prospect of thin film SIMOX as a suitable substrate for folly depleted MOS devices.
ii Abstract
cDedicated to my parents
Barbara and Antonio,
to my w ife Danielle,
to my brothers & sisters,
Antonio, Stylanolo, Carla, JQudana
and io Quillermo Sfiniunez de Cdayo/o.
Acknowledgem ents
I would like to express my gratitude to my supervisor Dr. P.L.F. Hemment for his
support and encouragement throughout the course of this research project. My thanks
also to Dr. A. Nejim for his guidance and support during the preparation of the SIMOX
wafers at the University of Surrey and for organizing SIMS analysis of selected samples. I
would like also to acknowledge Eaton Corporation (Mr. N. Meyyappan) for the thin film
SIMOX substrates prepared using the NV-10-80 implanter and LETI (Dr. J. Margail),
SOITEC (Dr. A. Auberton-Herve) and IBIS Corporation (Dr. A. Witthower) for the
thick film SIMOX substrates prepared in commercial NV-200 machines. Dr. M. Lourenco
is also acknowledged for the X-ray diffraction analyses.
I am also grateful to Dr. A. Claverie (CNRS, France) and Dr. C. Marsh (University of
Oxford) for very helpful discussions.Thanks are also extended to my friends and
colleagues Mr. F. Cristiano and Mr. G. Curello for preparing some of the TEM samples
and also for many constructive discussions. I would like also to express my thanks to the
staff of the D.R. Chick Laboratory and the staff of the Microstructure Studies Unit
(MSSU) of the Department of Materials Science who helped me to carry out many of
the experiments described in this work.
I would like to acknowledge the financial support of CAPES (Brazil) without which this
project would not have been possible. Finally, my warmest thanks must go to my wife
Danielle for her help, dedication and support throughout these years.
iv AclmowMedgements
A b b r e v ia tio n s
AES- Auger Electron Spectroscopyy BOX - Buried Oxide BF- Bright FieldCMOS - Complementary Metal Oxide SemiconductorCVD - Chemical Vapour DepositionDF- Dark FieldDL - Dislocation loopELO - Epitaxial Lateral OvergrowthFIPOS - Full Insulation by Porous Oxidised SiliconFWHM - Full Width Half MaximumHTA - High Thermal AnnealingIIB - Ion Implanted BoundaryIR - Infra RedLSI - Lar ge Scale IntegrationMBE -Molecular Beam EpitaxyMOS- Metal Oxide SemiconductorNSF - Narrow Stacking FaultOISF - Oxidation Induced Stacking FaultOM - Optical MicroscopyPVTEM - Plan View Transmission Electron MicroscopyRBS - Rutherford Backscattering SpectroscopyRED - Radiation Enhanced DiffusionSDL - Semi Dislocation LoopSEM - Scanning Electron MicroscopySF - Stacking FaultSFP - Stacking Fault PyramidalSFT - Stacking Fault TetrahedronSI A - Sequential Implants and AnnealsSIMOX - Separation by Implanted OxygenSIMS - Secondary Ion Mass SpectroscopySOI - Silicon On InsulatorSOS - Silicon On SaphireSPE - Solid Phase EpitaxyTD - Threading DislocationTEM - Transmission Electron MicroscopyVLSI - Very Large Scale IntegrationWB-Weak BeamXRD - X-Ray DiffractionXTEM - Cross Section Transmission Electron Microscopy ZMR - Zone Melt and Regrowth
v Abbreviations
Contents
Acknowledgem ents iv
A bbreviations v
Chapter One: In tro d u c tio n
1. Introduction 1
Chapter Two: L ite ra tu re Survey
2.1 Introduction 7
2.2 Evolution of SIMOX technology 8
C hapterThree: B ackground
3.1 Introduction 223.2 Ion beam synthesis of Si02 23
3.2.1 Formation of a buried oxide layer 243.3 Crystallographic defects in oxygen implanted silicon 34
3.3.1 Radiation damage as a source of secondary defects 343.3.2 Internal oxidation of silicon as a source of defects 37
C hapter F o u r : E xperim enta l details
4.1 Sample description 394.1.1 Oxygen implantations at the University of surrey 404.1.2 Oxygen implantations at Eaton Corporation 464.1.3 Standard SIMOX substrates 484.1.4 BF2+5 Ge+ and Si+ implantations 494.1.5 Sacrificial oxidation 52
4.2 Characterization methods 544.2.1 Chemical defect etching 54
4.2.1.1 Etching experiment details 554.2.2 TEM 574.2.3 RBS 574.2.4 Other methods 58
A b s t r a c t ii
vi C o n te n ts
C h a p te r F iv e : N e w e tc h a n t
5.1 Introduction 595.2 Development of an optimum defect etchant for thin Si layer 60
5.2.1 Dilute silicon etchants 605.2.2 New etchant 72
5.3 Distribution of defects in thick SIMOX 815.3.1 Lateral distribution of defect 815.3.2 Defect distribution-depth 84
Chapter S ix : Results I - S a c rific ia lly oxid ised S IM O X
6.1 Introduction 926.2 Defects in sacrificiallyoxidised SIMOX 93
6.2.1 Crystallographic defects in standard SIMOX 936.2.2 Crystallographic defects in sacrificially oxidised SIMOX 93
6.3 Sacrificial oxidation of ion implanted SIMOX materials 110
Chapter Seven: Results I I - L o w energy oxygen im plants
7.1 Introduction 1267.2 Wafers prepared by Eaton Corporation 127
7.2.1 General description 1277.2.2 Microstructure 1287.2.3 Defect distribution-depth 1327.2.4 Defect distribution-lateral 1347.2.5 Crystallographic defects in the silicon substrate 1447.2.6 Defects associated with edge effects 145
7.3 Wafers prepared at the University of Surrey 1477.3.1 General description 1477.3.2 SIMOX structures 1487.3.3 Defect distribution-lateral 1507.3.4 Evolution of threading dislocations during annealing 1577.3.5 Defect density-dependence upon implant temperaturre 1587.3.6 Edge effects-stress related 1607.3.7 Defect densities in the silicon overlayer 1657.3.8 Crystallographic defects in the Si substrate 167
vii C o n te n ts
C h a p te r E ig h t: D is c u s s io n
8.1 Introduction 1708.2 Crystallographic defects in sacrificially oxidised SIMOX 171
8.2.1 Nucleation process of OISF in SIMOX 1718.2.2 Evolution of OISF during oxidation 179
8.3 Crystallographic defects in low energy and low dose SIMOX 1878.3.1 Effects of thermal stresses on the nucleation of
defects in SIMOX 1898.3.2 Effects of contamination on the formation of
crystallographic defects 191
Chapter N ine : Conclusions and Suggestions fo r fu tu re w ork
9.1 Crystallographic defects in SIMOX thinned by sacrificial oxidation194
9.2 Crystallographic defect in low energy, low dose SIMOX195
9.2.1 SIMOX wafers produced at the University of Surrey 1959.2.2 SIMOX wafers produced at Eaton Corporation 196
9.3 Suggestions for future work 196
References 198
A ppendix A - S O I Technologies 208A ppendix B - Chem ical defect etch ing 211A ppendix C - D islocations and p la s tic flo w in the diam ond structu re 221A ppendix D - P repara tion o f T E M samples 232A ppendix E - L is t o f s a c rific ia l oxid ised samples 23 4P ublica tions 243
viii C o n te n ts
C h a p t e r O n e : I n t r o d u c t i o n
Complementary metal oxide semiconductor (CMOS) technology is the most common
device architecture used for very large scale integration (VLSI) microelectronics and
although bulk silicon is still used as its main substrate, it is expected that for the next
generation devices ( say < 0.25 pm), silicon on insulator (SOI) substrates could replace
bulk silicon in order to reduce parasitic effects between neighbouring devices and to
increase speed [1]. Indeed, the improved performance of devices formed on SOI
substrates results from reduced parasitic capacitance, latch-up suppression due to the
dielectric insulation and inherent radiation hardness[2]. In addition it is possible that SOI
substrates will be used, in the near future, for low power circuits for "hand held"
equipments [3].
During the last thirty years, a great deal of work has been dedicated to improve the
quality of SOI substrates. Indeed, during this time several SOI technologies were
developed and among the most important are Silicon on Sapphire[4](SOS), Fully
1 Introduction
Insulated Porous Oxidized Silicon[5] (FIPOS), Wafer Bonding,[6], Zone Melt
Regrowth[7] (ZMR) and Separation by Implanted Oxygen (SIMOX). However, in spite
of all the technological advances made during the 1980s these materials are only now
just beginning to reach a desirable quality ( defect density, lateral uniformity) to compete
with bulk Si technology (appendix A). Control of crystal quality, reduction of wafer cost
and designing of new compatible architectures are some of the main problems currently
being addressed ( see appendix A).
Among all of the SOI technologies, SIMOX is arguably the most promising material for
high complexity CMOS circuits and even bipolar applications[8]. It has been successfully
used as the substrate for 1 Mbit SRAM[9] memories and many others[10]. Furthermore,
fully depleted MOS devices have been successfully fabricated in thin film SIMOX ( <
1000 A) and it has been shown that higher drive currents and frequencies of operation
can be achieved and that deleterious effects, arising from parasitic capacitances and short
channel effects, can be reduced leading to further improvement to circuit
performance[l 1].
SIMOX substrates are fabricated by a two step process which involves the implantation
of high doses of energetic oxygen ions into (100) silicon wafers followed by a
post-implantation thermal annealing used to annihilate defects and to redistribute the
implanted oxygen. Standard SIMOX substrates having a silicon film thickness of about
3000 A and an oxide thickness of about 4000 A are normally produced by implanting a
dose of 1.8x1018 0 +cm'2 at energies of 180 keV to 200 keV and annealing at 1300°C for
typically six hours in an argon plus oxygen ambient[12]. Thin film SIMOX can be
obtained either by sacrificial thermal oxidation of the overlayer of thick SIMOX or by
implanting with lower energies and doses.
2 Introduction
As with other SOI technologies, SIMOX has still to overcome some obstacles before
becoming the preferred SOI substrate for mainstream CMOS technology. The main
problems facing SIMOX technology are the presence of residual defects (usually >
104cm'2) in the silicon overlayer, the quality of the buried oxide layer and the wafer cost
[13] . Although the quality of the silicon overlayer has improved dramatically from the
early days ( from 109 cm'2 to 104 cm'2 and lower) it still needs to be reduced to values
typical of bulk silicon before it will be acceptable for bipolar applications. In order to
achieve this goal ( defect free SIMOX) it is important to identify the mechanisms of
formation of crystallographic defects in SIMOX.
Paradoxically, the success in producing relatively low defect density materials ( 104 cm'2)
created new problems which impeded studies of the mechanisms of formation of
crystallographic defects due to the lack or inadequacies of available characterization
methods to delineate such low defect densities.
Among the direct characterization methods available to delineate crystallographic defects
in silicon, the most versatile are plan view transmission electron microscopy (PVTEM),
cross-section transmission electron microscopy (XTEM) and X-ray topography.
Arguably, XTEM is the best technique, however it is unfortunate that the high
magnification which is necessary to resolve the defects gives poor statistics for low defect
densities ( < 107cm*2) and, due to the complexity of sample preparation, the throughput of
samples is generally low. The spatial resolution of plan view TEM means that it is
impractical to use the technique to determine the areal density and the defect population
in samples containing defect densities lower than 104cm'2.
3 Introduction
X-ray topography[14] is another direct method extensively used to characterize defects in
bulk silicon. Like TEM, the contrast observed originates from regions that have a non
periodical arrangement of atoms and therefore, do not obey the Bragg law[15]. The main
difference between X-ray and electron diffraction is the radiation penetration depth,
which in the case of X-rays is almost two orders of magnitude greater (100 pm) than
electron penetration [16].This feature makes the use of X-ray diffraction topography
extremely difficult for the analysis of thin films since the defect contrast from the thin
films (<1.0 pm) is extremely weak compared with the contrast of defects present in
the bulk of the material. Therefore, although very attractive, since large areas can be
analysed at the same time, X-ray topography is generally not practical for the analysis of
defects in the silicon overlayer of SIMOX substrates.
Indirect methods such as RB S/channeling [17], photo luminescence[18], device leakage
currents[19] and chemical defect etching[20] could be used as alternative methods to
TEM analyses since these methods are sensitive to the presence of crystallographic
defects. However, only chemical defect etching can be used to accurately estimate the
density of crystallographic defects. The method is based upon the formation of etch pits
at the intersection of defects with the surface. Its advantage is that it is a simple technique
and only requires SEM or even optical microscopy to visualise the etch pits and,
therefore, can be used to analyse samples with a wide range of defect densities. Indeed,
chemical etching may be used to analyse samples containing a very low density (say 10
cm'2) up to a density of 109 cm'2 ( see Fig. 1.1).
4 In tw duction
XTEMC hem ica l e tch in g
105 - 11
PVTEM 1 01 0 5
Fig. 1.1 - Schematic showing the range o f defect densities for the different characterization methods.
However, standard bulk silicon defect etchants can not be applied directly to the study of
thin film SOI substrates because they require at least 1 pm of Si to be removed in order
to form resolvable etch pits (by optical microscopy). If less silicon is removed the
technique is found to be unreliable. Indeed, the reliability of the etchant will depend
upon its effectiveness to nucleate an etch pit at each crystallographic defect and this will
depend upon the actual thickness of the silicon film since the probability to nucleate all
defects will increase with etching time, that is to say with thickness of silicon removed.
In this project a systematic analysis of different silicon etchant systems has been carried
out in order to identify and optimise an etchant suitable for thin silicon films ( < 1000
A). This etchant has subsequently been used to identify crystallographic defects and to
facilitate a study of the evolution and annihilation of extended defects.
The overall aim of this project was to characterize the extended defects in thin film
SIMOX material and determine methods to control the nucleation and growth of these
defects.For this purpose a new defect etchant suitable for ultra thin silicon films has been
developed. The development of a novel characterization technique which is at the same
5 Introduction
time simple, quick and reliable opens the way not only to assess the quality of materials
produced in different ways but also enables the processing conditions to be optimised in
order to control crystallographic defect formation and, therefore, to obtain materials with
lower defect densities.
Furthermore, the availability of this new characterization technique can be used, in the
future, not only in SOI technology but also in other group IV materials such as Si/SiGe
hetero-epitaxy structures or ion beam synthesised Si/silicide structures which also call for
analysis of thin films of silicon.
This thesis contains nine chapters including this introductory one. In chapter 2, we
present a literature survey of publications relevant to this work. A background to the
basic physical aspects of ion beam synthesis of Si02 and to the main sources of defects in
SIMOX are presented in chapter 3 whilst in chapter 4, we describe the experimental
conditions and procedures used to prepare and analyse the thin films. In chapters 5,6 and
7 we present the experimental results. In chapter 5, we investigate and optimise a new
etchant suitable for thin silicon overlayers and in chapter 6, we apply it to the analysis
of thin film SIMOX prepared by sacrificial oxidation. In chapter 7 we report further use
of the the new etchant together with TEM analysis of thin film SIMOX produced by low
energy and low dose oxygen implantation. In chapter 8, we discuss the results presented
in chapters 6 and 7 . Finally, in chapter 9 we present our conclusion and suggestions for
future work.
6 Inti'oduction
C h a p te r T w o : L ite ra tu re S u rv e y
2 .1 I n tr o d u c t io n .
In this chapter we present a literature survey of the processing and materials
characterization of SIMOX structures. We first review the advances made in SIMOX
technology from the early idea of using ion implantation as a method to produce a buried
layer of Si02 up to todays ultra thin SIMOX structures. We also report on the
characterization methods that have been used to delineate crystallographic defects in the
SIMOX materials.
7 Literature Survey
The first report on ion implantation as a method to introduce impurities in a controllable
way into a substrate was made by Smith et al. [21] in 1956. However, it was eleven years
later (1967) that Watanabe and Tooi [22] reported for the first time, the successful
synthesis of Si02 by 0 + implantation into bulk silicon. They used a relatively low energy
(40 keV) and high dose (2.0x1018 0 +cm'2) and successfully formed a surface oxide which
was identified by infra-red (IR) spectroscopy and ellipsometry measurements. They
speculated on the use of ion beam synthesis as an alternative to thermal oxidation for the
formation of thin oxide layers. In the same year, Pavlov and Shitova[23] analyzed the
properties of the synthesised oxide using infra red spectroscopy and electron diffraction
and concluded that the oxide layer consisted of a mixture of Si02 and sub oxides (SiOx,
where x <2).
At this time, the implantation energies were low, typically 40 keV to 60 keV, because the
main objective was to produce a surface oxide. During the early 1970s, the quality of the
synthesised oxides was further assessed and the consensus was that they were inferior
to thermal oxides[24]. However, IR spectroscopy and I-V measurements [25] showed
that at least some of their properties (IR and passivation properties) were comparable to
a thermal oxide.
In 1977, the use of a synthesised buried oxide layer to isolate semiconductor devices was
proposed by Badawi and Anand[26]. They suggested that high energy and a medium
dose would enable a device worthy buried oxide layer to be formed. In the following
year, Izumi et al.[27] reported the successful synthesis of a buried Si02 layer formed by
2 .2 E v o lu t io n o f S I M O X te c h n o lo e v .
8 Literature Survey
implanting a high dose (<j)= 2.0xl018 0 + cm'2 ) of energetic oxygen ions (E= 150 keV) into
(100) silicon followed by an anneal at 1150°C for 2 hours. Afterwards, they grew an
epitaxial layer on top of the silicon overlayer in which CMOS devices, having operation
speeds twice as faster as devices on bulk silicon, were successfully fabricated. They
proposed the acronym SIMOX (Separation by IMplanted OXygen) for this new SOI
technology.
Subsequently, several groups[28,29] were involved in the formation and analysis of
SIMOX structures. The University of Surrey in collaboration with Middlesex Polytechnic
were the first groups to use Rutherford Backscattering (RBS) of 1.5MeV He* to
characterize both the silicon overlayer and buried oxide layer of SIMOX structures. This
analysis technique enabled the oxygen depth profiles to be determined and Gill [30]
investigated the evolution of the buried oxide layer with increasing oxygen dose using
RBS. He observed that by increasing the oxygen dose a Gaussian-like depth distribution
of oxygen evolved into a flat topped distribution. Gill and Wilson [31] suggested that
after the oxygen reached the stoichimetric value of oxygen in Si02 (4.48x1022 0 + cm'3)
the excess oxygen (at the peak of concentration depth-profile) diffused to the wings of the
distribution enabling the growth of the buried oxide layer.
Meanwhile in Japan, Hayashi et al.[32] were the first to use transmission electron
microscopy (TEM) to determine the microstructure of SIMOX materials. They analyzed
samples implanted at 150 keV with a wide range of doses (0.3xl018 to 2.4xl018 0 +
cm)'2 and annealed at 1150°C for 2 hours. They observed that the quality of the silicon
overlayer in samples implanted with the highest doses was very poor, with the presence of
microtwins and polycrystalline silicon. Hayashi et al. [33] used Auger Electron
Spectroscopy (AES) to determine the concentration-depth profiles of silicon and oxygen
9 Literature Survey
in the buried oxide layers in samples implanted with 1.8x10'* O' cm2 and found that the
concentrations saturated at the values of stoichiometric SiOz, namely, 33% and 67%
respectively. In addition, they found that the interface between the implanted buried oxide
layer and the silicon overlayer was abrupt, suggesting that no intermediate regions were
present at these interfaces.
The application of cross-sectional TEM (XTEM) by Booker and co-authors[34] in 1980
gave more information about the microstructure of the buried oxide layers and the quality
of the silicon overlayers Subsequently, the application of Secondary Ion Mass
Spectroscopy (SIMS) [35,36,37] enabled the evolution of the oxygen distributions to be
followed with increasing dose and anneal conditions. Studies carried out by Hemment et
al [38] showed that oxygen diffusion was preferentially driven to the upper Si SiO:
interface They attributed this behaviour to a radiation enhanced diffusion (RED)
process[39] due to a higher population of point defects in the vicinity of the upper
Si/Si02 interface (see Fig. 2 1).
0c =14* 10lfc*0 cmJal 200 KeV
Fig. 2.1 - Schematic o f the evolution o f the SiOj layer with increasing oxygen dose. [38]
10 Literature Survey
In spite of this worldwide research effort, the relatively poor quality of the silicon
overlayer impeded the further development of SIMOX technology. Analysis by XTEM
[40] showed that the main defects present in the silicon overlayer were Si02 precipitates
and dislocations which in the early 1980s had a density, typically of 10n cm'2 . It was
observed that two different types of dislocations were present namely, type I, which were
dislocations that thread from the buried oxide layer to the surface. These dislocations
were named "threading dislocations" and were generally observed in pairs separated by a
distance of less than 0.1 pm. Type II dislocations, started from a Si02 precipitate and
ended at a neighbouring precipitate remaining parallel to the surface. The threading
dislocations were later characterized by Krause et al [41] as being edge dislocations
having 1/2 <110> Burgers vectors. Although the mechanisms of formation of the
threading dislocations were not understood, it was clear at that time that the use of higher
implantation temperatures would lead to lower residual defect densities due to the
enhanced self annealing during implantation. In 1984, Holland et al [42] were the first
to introduce background heating to increase and control the implantation temperature. By
so doing, they successfully reduced the level of dislocations in the silicon overlayer by
three orders of magnitude (from 1011 to 10s cm'2). However, the introduction of higher
implantation temperatures did not produce a reduction in the number of Si02 precipitates
in the silicon overlayer[43].
The big breakthrough in SIMOX technology occurred in 1986, when the mechanisms of
coalescence and dissolution of Si02 precipitates became better understood. Stoemenos
et al. [44] proposed that by using higher temperature annealings (1300°C) the Si02
precipitates present in the silicon overlayer could be completely eliminated. He
11 Literature Survey
demonstrated that at these high temperatures only the dissolution of the Si02 precipitates
occurred whilst the buried oxide layer remained thermally stable (see chapter 3)
The improvements in the quality of the silicon overlayer, both by the elimination of Si02
precipitates and reduction of dislocations, suddenly, placed SIMOX in the position of the
most promising SOI technology. Moreover, in the spring of 1986, the commissioning of
the first 100 mA high-current oxygen implanter (NV-200) , where the engineering
development was funded by Nippon Telephone and Telegraphs (NTT) [45] and the
machine was built by the Eaton Corporation, put SIMOX a step closer to implementation
as the preferred SOI substrate for radiation hard and small geometry CMOS circuits. This
change in the status of SIMOX technology was reflected by an increase in the number of
papers presented at the IEEE conferences during the late 1980s[46] ( see Fig 2 2)
100
£ 80<DCX03cx 60
f t - .oV 40
E2 20
083 84 85 86 87 88 89 90
Y e a r
Fig. 2.2 - Number o f SIMOX papers presented at the IEEE SOS/SOI technology workshops over the period 1983 to 1993 [ 46].
Device Physics
■ 3D□ Other SOI
E SIMOX□ SOS
12 Literature Survey
During 1987 to 1992, several groups [47,48,49] were involved in the materials
characterization of SIMOX with the primary objective of determining the mechanisms
responsible for the formation of threading dislocations. Although several mechanisms
were proposed none of them was experimentally verified, however, all of the researchers
agreed that the origin of threading dislocations was either due (i) to the evolution of
defects present in the as-implanted material ( before annealing) or (ii) to the creation of
defects during annealing.
Jaussaud et al [47] proposed that dislocations could be created by precipitate-induced
stresses during thermal ramping and annealing. Nevertheless, they concluded that under
optimum conditions (to give a stress free silicon overlayer) defect creation during
annealing was not the dominant mechanism but that the principal mechanism was due to
the evolution of defects present in the as-implanted state. This idea was shared by other
researchers and indeed, Stoemenos et al [48] proposed that the formation of paired
threading dislocations was due to the growth of surface defects (semi dislocation loops)
during annealing. They attributed the formation of these defects to the inability of the
surface to accommodate the flux of self interstitials, arriving at the wafer surface and
leading to a non ideal internal epitaxial growth. The defects were thus viewed as "growth
defects". On the other hand, Visitsemgtrakul et al. [49] proposed another mechanism in
which intrinsic stacking faults (SF) located near the surface grow towards the buried
oxide layer during annealing. They suggested that threading dislocations were formed by
the unfaulting of the SF during the expansion. De Veirman et al [50] also found near
surface defects on {111} habit planes and suggested that these defects were the
percursors of threading dislocations. Finally, Venables et al [51] suggested that
implantation-induced stresses in the silicon overlayer played a role in dislocation
13 Literature Survey
formation. Based on X-ray rocking curve determinations of the strain within the
overlayer, they showed that the upper portion of the silicon overlayer was in a stress state
of biaxial tension, probably due to the formation of vacancy-oxygen clusters. The stress
was believed to be of sufficient magnitude to provide a climb force for the expansion of
prismatic extrinsic half loops. These half loops could, then, expand over the whole silicon
overlayer and form paired dislocations.
As a result of these detailed studies a better understanding of the mechanism of defect
formation was achieved leading to the successful optimisation of the implantation
conditions and consequently to lower dislocation densities (< 105 cm'2). The introduction
of multiple implants and anneals by Hill et al [52] enabled the formation of SIMOX
structures with dislocation densities as low as 104 cm'2 to be achieved. The reason for
using multiple 0 + implants instead of the more conventional single 0 + implants, was to
avoid exceeding an experimentally established critical dose for crystallographic defect
formation[53].
Other methods to produce low defect density SIMOX appeared later, for example, El
Ghor et al.[54] demonstrated that implanting at high current densities together with high
implantation temperatures ( 600-675 °C) enabled the formation of cavities along the path
of the incident ions. These cavities act as sinks for point defects and reduce the stress
state of the silicon overlayer resulting in a low defect density.
Another method proposed by Van Ommen[55] is the implantation of oxygen ions in
channelled directions using low current beam densities . In this way, a cubic superlattice
of Si02 precipitates was formed . The ordering of the precipitates resulted in a decrease in
14 Literature Survey
the stress state of the silicon overlayer due to easier accommodation of the strain and, as a
consequence, the defect density was lower, being typically less than 105 cm'2.
At the end of the 1980s, when the first commercial SIMOX wafers appeared in the
market a recipe for good quality device worthy substrates was already well known ( see
table 2.1).[56]
Implantation Dose (0 + cm'2)
Energy(keV)
Implantationtemperature(°C)
Annealingtemperature(°C)
Annealing time (hrs)
Single 1.8xl018 190 700 1300 6
Multiple 3x (6x1017) 190 700 1300 3x2
Table 2.1 - Implantation and annealing conditions to obtain good quality SIMOX
Thin film SIMOX
At the beginning of the 1990s SIMOX was already well established as the most
promising SOI materials technology for high performance CMOS devices. However,
the advent of fully depleted CMOS devices[l 1] necessitated the fabrication of thin film (<
1000 A) SIMOX substrates. Additionally, the desirability of forming thin buried oxide
(BOX) layers to lower the dose and, hence, cost of the substrates shifted the attention of
researchers and substrate vendors to low dose and low energy implantations.
In 1990, Nakashima and Izumi [57] reported the formation of a continuous thin buried
oxide layer in samples implanted with low 0 + doses. They implanted 160 + into 100 mm
(100) silicon wafers at an energy of 180 keV over a large range of doses (0.1 to 2.4xl018
0 + cm'2) and observed extremely low defect densities (103cm'2 ) in substrates implanted
with doses (j) < 0.7xl018 0 + cm'2 (see Fig. 2.3).
15 Literature Survey
Fig. 2.3 - Dependence of defect density upon oxygen dose. [ref. 57 ]
Subsequently, several groups[58,59] in different continents ( Europe and USA) used low
dose 0 + implantations to form thin buried oxide layers. The SPIRE Corporation^8] was
a pioneer in the use of low dose and low energy 0 + implantations to produce thin
SIMOX structures. The low doses enabled thin buried oxide layers to be formed whilst
the use of low energies ensured that the silicon overlayer was thin also. Although they
succeeded in producing continuous buried oxide layers, no reports on the quality of the
silicon overlayer appeared in the open literature.
During the period of 1990 to 1993 the University of Surrey in collaboration with three
others laboratories (Imperial College, Oxford University and Liverpool University)
characterized thin SIMOX structures [59] prepared by low dose and low energy 0 +
implantations.In 1991, Li et al. [59] reported, for the first time, the production of good
16 Literature Survey
quality thin silicon overlayers (1000 A) on top of continuous thin buried oxide layers.
They produced these device worthy structures by implanting low oxygen doses ( (j) <
l.OxlO18 0 + cm'2 ) using a wide range of energies ( 40 to 90 keV) and annealed at
1300°C for 6 hours. Furthermore, they proposed a semi-empirical model to determine
the minimum oxygen dose ((j>cA) needed to form a continuous buried oxide as a function
of the implantation energy. They suggested that the necessary condition to obtain a single
oxide layer was that the separation between the peak of maximum damage (Rd) and the
peak of concentration (Rp) had to be smaller or equal to the estimated thickness (Tsi02) of
the synthesised layer of oxide, given by:
Tsioj = - f - (2.1)N?i02
Where d> is the oxygen dose and N°si02 is the oxygen concentration in stoichiometric
SiO, (N°Si0 - 4.48x1022 cm'3).
Therefore, according to Li et al. <I>AC is:
O ?=aR pN°si02 (2.2)
where a is:
01 = — (23)
From eq. (2.2) it is possible to obtain the critical dose for the formation of a continuous
buried oxide layer. Li et al found a good agreement between the experimental values and
those obtained from eq. (2.2) (see Table. 2.2).
17 Literature Survey
Energy ('*0+ keV)
( X 10l7/craJ) <*>1 ( X 10'7/cur)Expt. Calc. Expt. Calc.
15 2.34 2.3 0.5750 = 5 5.4 =2 1.370 = 7 6.7 = 3 2.2
140 = 10 9.9 3 < <t>1 < 7 4.1150 10.2 4 < <t>J < 4.5 4.4200 12-13 11.6 5, 6 5.6
Table 2.2 - Estimated and experimental values for Oac according to Li et al [59].
An alternative method to produce ultra thin silicon overlayers (tSi02< 1000 A) is to employ
sacrificial oxidation of the overlayer (tsi02.= 3000 A) of conventional (200 keV) SIMOX
material. The method consists of thermal oxidation of a thick film (3000 A) structure and
subsequent removal of the oxide by a HF etch. First reports [60, 61] suggested that the
quality of the silicon overlayer was not degraded during the extra oxidation step.
Low defect density studies
Until 1988, the main method used to characterise crystallographic defects in SIMOX
materials was TEM in planar and cross section views . However, with the improvements
in the quality of the silicon overlayer and much reduced defect densities, TEM rapidly
became inefficient for the analysis of samples containing low defect densities (< 105cm'2)
because direct observation methods which iavolve very high magnifications were not
compatible with the low areal density of defects[62].
18 Literature Survey
Conventional characterization methods used in bulk silicon technology such as X-ray
topography[14], electrochemical etching[63] and chemical defect etching[20] were
investigated as alternatives to TEM analyses since the spatial resolution limits of these
methods facilitated the analysis of samples containing extremely low defect densities.
A recent report on the use of X-ray topography to the analysis of crytallographic defects
in SOI materials (prepared by Wafer Bonding) was given by Theunissen et al [64], They
succeeded in separating the images of the silicon overlayer from the substrate due to a
misorientation between the layers. They intentionally introduced this misorientation by
bonding two wafers of different orientations, namely (001) and (111). However, they did
not succeed in analysing SIMOX since for these materials, the silicon overlayer and the
substrate have the same orientation and, therefore, a distinction between the image of
defects from the silicon overlayer and from the substrates was not possible
The application of chemical defect etching to the analysis of SIMOX substrates is also
very difficult and there have been only relatively few reports on the successful use of
that method to delineate defects in SIMOX. Probably, the first report was by Homma
et al.[53] in 1987. They grew a thick epitaxial silicon layer (0.6 p- 1.0 pm) on top of the
original silicon overlayer using molecular beam epitaxy (MBE) prior to the etching
process. The etch pits formed in these relatively thick layers ( «1 pm) were large enough
to be observed by scanning electron microscopy (SEM). However, it was not possible, at
that time, to determine a one to one correlation between the etch pits observed in the
epitaxial layer and the "threading dislocations" in the silicon overlayer since no
cross-correlations with TEM analysis were carried out. In 1990, Liaw et al.[65] reported
extensive analyses by TEM of epitaxial SIMOX samples and determined that not all the
"threading dislocations" (TD) propagated into the epitaxial layer. Indeed, they found that
19 Literature Survey
the dislocation density measured at the top surface decreased as a function of the epitaxial
silicon thickness due to the "looping" of threading dislocations during the epitaxial
growth.
In 1990 Fechner et al. [66] introduced a second step in the etching process which
enabled the etch pits to be observed by optical microscopy without growing an epitaxial
layer. The second etching step consisted of a dip in a HF(40%) solution which dissolved
the Si02 layer below the etch pit. Using this second etch step, large cavities (1-2 jim)
were formed underneath the etch pits ( see Fig. 2.4). The "decorated" etch pits could
then be easily observed by optical microscopy. Margail et al[62] used a similar method to
delineate defects in commercial SIMOX and reported that the double etch method was
only reliable for samples ha\ing a silicon overlayer thickness greater than 1000 A.
In 1993, Giles et al.[67] reported a new etchant capable of reliably delineating
crystallograpic defects (dislocations and SF) in silicon overlayers as thin as 500 A. (see
chapter 5).
20 Literature Survey
! t
V)cJZ ^ A r w v v >
i tuQ
o
21 Literature Survey
Fig.
2.4
- CVy/
v.v s
ectio
n sc
hem
atic
of
the
"dec
orat
ed"
etch
pits.
C h a p t e r T h r e e : B a c k g r o u n d
3 .1 I n tr o d u c t io n
An understanding of some physical aspects, such as diffiisivity and solid solubility of
oxygen in silicon is fundamental to a better appreciation of the mechanisms of the
formation of the buried oxide layer in SIMOX materials. In this chapter, we present
basics aspects related to the synthesis of a continuous buried oxide layer by 0 + ion
implantation. In addition, we also present some basic aspects related to the formation of
crystallographic defects in oxygen implanted silicon.
22 Background
3 .2 I o n b e a m s y n th e s is o f S i 0 2
Ion implantation is a non-equilibrium process in which highly energetic ions are injected
into a target in order to change the mechanical, optical or electrical properties of the
target material [68,69]. In some cases (for extremely high doses), the volume
concentration of the ions (impurities) introduced by the implantation process can exceed
the solid solubility limits and the formation of new chemical phases is likely to occur. The
use of ion implantation as a method to synthesise new phases has been widely used in the
past, to form compounds like silicon dioxide, silicon nitrides[70] and more recently, to
form silicides such as iron silicides [71], cobalt silicides[72] and many others [73],
However, the complex nature of the structures created by ion beam synthesis makes
theoretical descriptions of the evolution of the new phase and identification of the
dominant processes and prediction of the final structure rather difficult.
The fundamental processes that occur during the nucleation and segregation of a second
phase depend on the processing conditions but the general scheme of the evolution of an
incipient buried layer can be considered as consisting of four overlapping stages[74j.
These stages are namely, supersaturation of implanted atoms (ions), nucleation of
precipitates, precipitate growth and, finally, the coalescence of precipitates into a planar
continuous layer with abrupt interfaces. In the next sub-section we briefly describe the
physical processes which control the four stages of the growth of the BOX layer.
23 Background
3 .2 .1 F o r m a tio n o f a b u r ie d o x id e layer.
A. Diffusivity o f oxygen in silicon.
Below the solubility limits, oxygen in single crystal silicon is known[75] to be a
covalently bonded interstitial impurity, incorporated in slightly off-centred interstitial
positions between two neighbouring silicon atoms ( see Fig. 3.1).
n \ \ <m> 1 1 \\
5 1 U
i i 3 f lP3 i I 2 / /
'
A Si K
Fig. 3.1- Configuration o f interstitial oxygen in silicon [75J.
The diffusion of interstitial oxygen is an activated process with a diffusion coefficient D
given by [76]:
D(T) = Do exp(-£7*67) (3.1)
Where T is the absolute temperature, is the Boltzmann constant, D0= 0.13 crn2/s and
E*=2.53 eV. Fig 3.2 shows the diffusion coefficient of interstitial oxygen in silicon as a
function of temperature.
24 Background
TEMPERATURE(x lO C)
Fig. 3.2 - Diffusion coefficient o f oxygen in silicon as a function o f temperature 176].________________________________________________________________
In some special cases, such as for example in ion implanted silicon, anomalous fast
diffusing oxygen has been reported[ 77,78], It has been suggested that this anomalous
diffusion is due to the diffusion of molecular oxygen (0 2) formed during heat treatments
at moderate temperatures.
B. Solid solubility o f oxygen in silicon
The solid solubility of oxygen in silicon is given by[79]:
C” '(7) = q r'exp (-A m k 7) (3.2)
Where C0,ol= 9xl022 cm*3 and AE = 1.52 cV.
Fig. 3.3 shows the dependence of the solubility upon temperature [79],
25 Background
t e m p e ra tu re [°CJ
104/T [K.-1]
Fig. 3.3- C°l dependence upon temperature [ 79].
In oxygen implanted silicon, the oxygen concentration, CQ (z), varies significantly with
depth, having its maximum value near the projected range (Rp) and decreasing
progressively towards the surface and at greater depths into the substrate. To a first
approximation, neglecting effects such as crystal orientation, C0(z) is given approximately
by a Gaussian distribution:
Co = Cpeak e x p - ~ L £ ) - . (3.3)2o
where 5 is the standard deviation (straggle) and q _ ♦.5
For CG(z) > C0so1 , segregation of the implanted oxygen into a Si02 phase will occur
through the nucleation and growth of Si02 precipitates (see 3.2.1 C). Mantl[80] has
shown that this phase transformation is driven by the decrease in the free energy of the
system. The change of Gibbs free energy, AG, when a new phase of volume, V, and
interface area, A, is formed is given by[80]:
AG = V(AGv + AGe) + Act (3.4)
26 Background
where AGV is the change in free energy per unit volume associated with the formation of
the Si02 phase, AGE is the increase in elastic energy per unit volume and a is the specific
interface energy. Since the molar volume of the Si02 is approximately twice as large as
the molar volume of Si[ 81], the formation of the Si02 phase will produce internal
strains, and therefore will tend to increase the elastic energy of the crystal. The typical
dependence of the change of Gibbs free energy AG upon the radius of a precipitate is
presented in Fig, 3.4 [82], Precipitates with radii smaller than a critical radius rc ( shown
in the figure as rcl at a temperature T,) are unstable and tend to dissolve in the matrix.
27 Background
In oxygen implanted silicon, several mechanisms of Si02 precipitation can occur
simultaneously depending upon the processing conditions ( oxygen dose, implanted
temperature, and current density). We can basically separate these mechanism into three
major processes, that are namely, homogeneous[83], heterogeneous[84] and
'etherogeneous' precipitation.
Homogeneous precipitation [86,87] is the dominant process when the oxygen
concentration exceeds a minimum value for which spontaneous precipitation occurs
independently of the presence of nucleation sites. Experimental evidence of homogeneous
precipitation in high dose oxygen implanted silicon ( <|>= 1.8xl0,s 0 + cm'2) has been
extensively reported by Hemment et al[86] and Marsh et al[87] .
On the other hand, Cerofolini et al [88] have suggested that heterogeneous precipitation
is the dominant process for lower oxygen concentrations (low dose regimes) and high
implantation temperatures. For these conditions the oxygen supersaturation inside the
matrix is not high enough to enable homogeneous precipitation to occur, and therefore
the formation of the Si02 precipitates will depend upon the existence of preferencial
nucleation sites. Several authors [89,90] have identified vacancy clusters as being the
preferential nucleation sites for Si02 precipitates. Preferential nucleation at vacancy
clusters occurs because the elastic energy associated with the formation of the Si02
phase is considerably reduced in these localized areas [91]. In oxygen implanted silicon,
vacancy clusters are predominantly located near the peak of damage (Rp) and, therefore,
heterogeneous precipitation will occur preferentially near that region. Experimental
C. P r e c ip ita tio n p r o c e s s
28 Background
evidence that heterogeneous precipitation occurs near the peak of damage has been
recently shown by, Li et al [ 92] and Aspar et al [ 93], for low dose oxygen implanted
silicon submitted to high temperature annealings. These authors showed that in these
cases two bands of Si02 precipitates are formed . One located near the peak of the
damage and the other close to the peak of the oxygen concentration-depth profile. The
formation of the first band of precipitates is explained by the radiation enhanced
diffusion[39] of the implanted oxygen to the region of higher concentration of vacancy
clusters where the heterogeneous precipitation process takes place.
More recently, Cerofolini et al [94] have proposed a new precipitation mechanism
named 'etherogeneous' precipitation. This new mechanism differs from the conventional
heterogeneous precipitation because it takes into account the effects of the implantation
process. In their model, they propose that for certain conditions (low implantation
temperature and high beam currents) Si02 precipitates can be heterogeneously nucleated
inside highly disordered regions formed by the implantation process.The driving force for
this precipitation process is the local heating created by the collisional events and
according to their model only projectiles that release an energy higher than a minimum
threshold energy are able to induce the formation of Si02 precipitates [95 ].
D. G rowth and d issolution process o f S iO 2 precipitates.
The evolution of the Si02 precipitates during thermal processing is initiated by second
phase nucleation, and then controlled by a process in which only precipitates having a
radius larger than rc (see Fig. 3.4) will remain stable while smaller ones will
dissolve[96]. Vanhellemont and Claeys[ 97] have derived an expression for the critical
29 Background
radius rc which takes into account the influence of oxygen, vacancy and self-interstitial
concentrations in the Si matrix. This expression is given by [ 98]:
with cr is the interface energy per unit area, Estr the strain energy and K a constant
associated with the silicon oxide polymorph (SiOx). The parameters C0 , Cv and CT are
respectively the oxygen, vacancy and self-interstitial concentrations in the silicon matrix.
Their thermal equilibrium values are given by C*0 , 0% and C \ . In addition a is the
number of vacancies absorbed from the matrix and p the number of self-interstitials
injected into the matrix. For high temperature anneals ( say > 1250°C) the recombination
rate of vacancies and interstitials is very high [ 99] and therefore, Cv and Cr rapidly
approach the thermal equilibrium values C*v and C*r Also the interstitial oxygen
concentration rapidly approaches its equilibrium value due to the combined effect of
oxygen out-diffusion (at these high temperatures the diffusion length of oxygen in silicon
is larger than the silicon overlayer thickness) and precipitate growth. This makes the
denominator of eq. 3.5 very small provided that Estr is also small. A quantitative value
for Estr is not well known, however, at these high temperatures several strain relief
mechanisms are active [100] and therefore Estr is expected to be small. The decrease of
the denominator of eq. 3.5, means that the critical radius increases at these high
temperatures (1250°C). Therefore, as we have already suggested, precipitates having a
radius smaller than rc will become unstable and will dissolve. The oxygen atoms released
by the dissolving precipitates can out-diffuse or diffuse towards the larger precipitates.
30 Backgi'ound
The oxygen diffusion from the dissolving precipitates towards the large precipitates can
be described in terms of the Gibbs-Thomson equation, given by [101]:
Cs(f) = CoV rkB T }
(3.6)
where Cs is the solute concentration at the surface precipitate of radius r and
denotes the solute concentration at a planar interface (r = oo ), Va the atomic volume, a
is the specific interface energy and kD the Boltzman constant. From equation 3.6 it is
observed that Cs(r2) > Cs(r,) for r2> rc> rl ( see Fig. 3.5).
growth dissolution
|
[80].
31 Background
Therefore, the oxygen released by the dissolving precitate ( r, < r c ) will diffuse towards
the growing precipitate ( *2 > rc) .The competitive growth of precipitates, in which
larger ones grow at the expense of the smaller ones, is known as Ostwald ripening [102].
The growth of precipitates larger than the critical radius is, then, a diffusion limited
process in which the limiting factor is the diffusion of oxygen to the periphery of the
precipitate. Wada et al.[103] have expressed the volume increase of a precipitate with
time above the critical radius as:
I Elr ~ (2+ ~ Dt)m (3.7)
Go ~ 0 o
where r is the radius of the precipitate, is the initial oxygen concentration expressed in
atoms per cubic centimeter, C0E is the equilibrium oxygen concentration at the annealing
temperature and Cp0 is the oxygen concentration in the precipitate.
E. Coalescence o f S iO 2 precipitates.
A further "coarsening" mode observed in oxygen implanted silicon is the coalescence of
Si02 precipitates. If the density of these large precipitates is high enough, coalescence of
adjacent precipitates may occur as the two phase system will change in order to reduce
the surface energy[104]. The coalescence process is mainly controlled by solvent
diffusion from the matrix to the 'neck' region at the point of contact of the precipitates(
see Fig. 3.6). Brown [105] calculated the solute concentration at the neck Cs(p) as
being:
a s )
where p is the radius of the neck, as depicted in Fig. 3.6.
32 Background
F l u x o f O x y g e n
Fig. 3.6 - Schematic of two coalescent precipitates.
The equation (3 8) resembles the Gibbs-Thomson equation (eq 3.6), but with the
difference that the solute concentration at the neck is smaller than the concentration in the
matrix, in contrast to the concentration at the surface of a small spherical particle This is
a consequence of the 'negative curvature' at the neck of the two touching precipitates
(see Fig. 3.6)
Mantl [80] showed that the solute flux towards the neck is approximately:
a C .D V , p2 K*7-
A consequence of this solute flux towards the neck is that the radius of the neck will
increase leading to a localized planarization of the resulting precipitate Thus, precipitate
coalescence can promote planar layer formation provided the degree of overlapping of
the large precipitates is high enough to enable their coalescence into a single continuous
layer.
33 Background
3.3 C rvsta llosravh ic defects in oxvzen im planted silicon.
In a comprehensive review of defects in SIMOX materials, Jaussaud et al [46] proposed
that crystallographic defects in oxygen implanted silicon are mainly formed by two
mechanisms, namely (i) condensation of point defects produced both by the implantation
and by internal oxidation and (ii) plastic deformation of the silicon structure caused by
the large volume mismatch between atomic Si and the Si02 molecule. Therefore, we can
see that the two main sources of defects are the radiation damage itself and the internal
oxidation of silicon and in the following sub-sections we describe the consequential
formation of crystallographic defects in the implanted layer.
3.3.1 Radiation damage as a source o f secondary defects
The loss of kinetic energy by the implanted ions occurs through electronic ( inelastic) and
nuclear (elastic) interactions with the atoms of the target[69]. In silicon, elastic collisions
are mainly responsible for the primary implantation damage. Collisions with energy
transfers above a threshold, of about 15 eV, will displace Si atoms from their lattice sites
[106]. The displaced atoms may in turn displace further atoms resulting in highly
disordered regions. The displacement of an atom from its regular lattice position to an
interstitial one forms an interstitial/vacancy pair known as a Frenkel pair[107].
Therefore, the highly disordered regions produced by the implantation process consist
predominantly of a non equilibrium concentration of vacancies and interstitials. Since
the internal energy of isolated point defects is high, interstitials and vacancies will tend to
recombine with each other or condensate into aggregates in order to decrease the total
energy.
34 Background
In diamond structures, the condensation of vacancies (interstitials) occurs preferentially
along {111} planes, since they are the lowest surface energy planes in these
structures[108] (see appendix C). The growth of aggregates occurs because there is a net
reduction in energy when an additional vacancy (interstitial) is captured by the aggregate.
However, Thomas et al [109] have proposed that the continuous growth of the vacancy
(interstitial) aggregate is limited by the possible collapse into a dislocation loop (DL)
with a lower energy configuration. Since the stacking fault energy in silicon is relatively
low (y « 50 erg cm'2) [110] both faulted ( Frank sessile DL) and perfect ( prismatic DL)
dislocation loops can be formed from the collapse of those aggregates. Moreover, it has
also been proposed [109] that a vacancy aggregate can collapse into a stacking fault
tetrahedron (SFT)[111] when this configuration is a lower energy state.
Further, Thomas et al [109] have shown that the relative energies of intrinsic ( vacancy
type) faulted DL, SFT and perfect DL are given approximately by:
where Ep, ET and Ep are the energies of a triangular faulted DL, a SF tetrahedron and a
perfect loop, respectively, each having side /. Additionally,G is the shear modulus, a is the
unit cell dimension, r0 is the dislocation core radius, v is Poisson's ratio and y is the
stacking fault energy.
It is observed from eq. (3.10 a-c) that the relative energies of these three different types
of defects depend on the size / of the defect. For illustrative purposes we show in Fig.
3.7 the dependence on the energy of the defect upon the size / [109].
(3.10 a)
(3.10 b)
(3.10 c)
35 Background
100 200 300 400 500Size I (A)
Fig. 3.7 - Energy dependence with size I o f faulted DL, SFT and perfect loop in Ag [109].
It is observed that SFT are more stable ( have lowest energy) below a critical size, / ,
while perfect loops are always favoured above an upper critical size (330 A). Table 3.1
lists the critical value, lo , for which the SF tetrahedron (SFT) is the most stable defect
in different metals [109],
Material y (erg cm'2) /0( A)Cu-Zn 20 1.6xl04
Au 33 430Cu 40 515Ag 40 330
Table 3.1 - Calculated values o f lo below which SFT are the most stable defect (after Thomas et al [109])
Transformation of one type of defect to another can occur if the energy associated with
the defect (see eq. 3.10) is decreased by the transformation and also if the activation
energy for this process is low [112]. The transformation of a faulted DL ( Frank type, b
36 Background
=1/3<111> ) to a perfect DL ( b =1/2<110>) was first considered by Saada[113] who
showed that it basically involves the nucleation of a Shockley partial ( b = 1/6 <211>) in
the plane of the fault. The unfaulting mechanism can be written, then, as:
± < 1 1 1 > + | < 2 1 1 > —> i < 1 10 > (3.11)
Frank Shockley perfect
Similarly, SFT can arise from a multiple dissociation mechanism of an intrinsic Frank DL.
The dissociation mechanism that enables the formation of the SFT from an intrinsic
Frank DL is described i n appendix C.
3 3 .2 In te rn a l oxidation o f s ilicon as a source o f defects.
The molar volume of the Si02 molecule is approximately 2.25 larger than that of Si (Vsi-
2.23x1022 cm'3)[81] and therefore, the formation of the Si02 phase will produce large
internal stresses unless an efficient stress relief mechanism occurs, The extra volume can
be accommodated either by (a) emission of self-interstitials away from the Si-Si02
interface or by (b) plastic deformation of the silicon matrix near the Si02 precipitate.
A. Emission o f Si interstitials
The large volume change associated with Si02 formation requires that silicon atoms will
be displaced from the vicinity of the Si02/Si interface. This may be accomplished by
generation of silicon interstitials, Sii5 at the Si02/Si interface. Self interstitials are created
in this region because the crystal elastic energy associated with the stress generated by
the Si02 precipitate generally exceeds the formation energy of the [114]. Therefore,
37 Background
the internal oxidation of silicon involves the formation of Sij according to the
following reaction[l 15] :
(r + 1) Si + 0 2 -> Si02 + r Si* (3.12)
where r = 1.25.
From eq. 3.12, it is observed that a supersaturation of Sij will be created near the Si02/Si
interface and, consequently, condensation of interstitials into aggregates is likely to
occur giving rise to the formation of crystallographic defects through the mechanisms
described in section 3.2.1.
B. Plastic deformation
Plastic deformation of the silicon matrix can arise if the compressive stresses around the
Si02 precipitate are not completely relieved. The silicon matrix will deform plastically
through the nucleation of prismatic dislocation loops. This process was first described by
Seitz et al [116] as "prismatic punching". They pointed out that a pair of prismatic loops
of opposite nature, one interstitial and one vacancy type, can be generated under the
influence of a misfit stress field. This is accomplished by expansion of "generator" loops
around a glide prism.
When the precipitate is compressed by the matrix, the interstitial loop is punched out
along the glide directions, <110>, while the vacancy loop moves onto the precipitate
surface where it is annihilated, thereby reducing the misfit strain. This mechanism can be
repeated several times and can produce a family of loops "punched" out along
crystallographic directions which defines the axis of the glide surface. In some cases,
pinning of the prismatic dislocation loop to the Si02 precipitate can occur and the
expansion of the dislocation is only possible by a climb process [117].
38 Background
C h a p te r F o u r : E x p e r im e n ta l D e ta ils
In this chapter we describe the experimental conditions in which samples were prepared
and we give details on the different methods used to characterize the SIMOX materials.
4.1 Sample description.
Two different sets of samples have been used in this project. The first group of samples
consisted of, (100) Cz, Si wafers implanted with oxygen using a wide range of energies
(30 keV-200 keV) and a wide range of doses ( 0.2-1.0 xlO18 0 + cm'2).The samples were
implanted either at the University of Surrey or at Eaton Corporation. The wafers
prepared at the University of Surrey were produced as part of a SERC funded
programme (IED 1777) to produce and characterize low energy and low dose SEMOX
materials whilst the wafers prepared at Eaton Corporation were obtained to support a
collaborative activity involving that company and the University of Surrey. Annealing of
the Surrey implanted wafers were carried out at Southampton University. Further details
of the implants and anneals are included in section 4.1.1 and 4.1.2.
A second group of specimens consisted of SIMOX samples thinned by sacrificial
oxidation. The original thick SIMOX substrates were supplied by LETI [118] and IBIS
39 Experimental details
Corporation [119]. Details of the implantation and annealing conditions are given in
section 4.1.3 whilst the details of the sacrificial oxidation are given in section 4.1.5, In
addition, some SIMOX samples were implanted with Si+, Ge+ or BF+2, to amorphise the
silicon overlayer in the vicinity of the Si/BOX interface prior to oxidation (see section
4.1.4). Details of all the samples used in this project are listed in tables 4.1 to 4.8 and in
appendix E.
The SIMOX wafers were then cleaved into small samples ( typically 5x5 mm2) and
cleaned using acetone, isopropyl alcohol and trichloroethylene prior to a dip in an
ultrasonic bath in order to remove particles from the surface, The samples were then
ready to be analyzed using the different analytical methods described in section 4.2.
Furthermore, selected SIMOX samples were etched using a KOH solution (1 molar) at
70 °C in order to selectively remove the silicon overlayer. Then, the buried oxide layer
was removed using a HF(40%) solution. This two-fold etch treatment enabled both the
silicon overlayer and BOX layer to be removed, thus exposing the silicon substrate for
further analysis.
4.1 J Oxvzen im planta tions a t the U niversity o f Surrey.
Ion implantation was carried out on the 500 kV implanter at the University of Surrey.
This machine has an open air terminal from which the ions are accelerated to their frill
velocity, (maximum voltage of 500 kV) prior to analysis at the earth potential. The ion
beam is electrostatically scanned in the x and y directions over an area of maximum size
2"x2". A schematic of the implanter showing the main components is presented in Fig.
4.1. From Fig. 4.1 we can see that the implanter has six major components; namely: (i)
40 Experimental details
ion source, (ii) focussing element, (iii) acceleration tube, (iv) analysing magnet (v)
electrostatic scan and (vi) the target chamber.
High voltage terminalr <kK)
Analysingmagnet
Graded accelerator tube
<DTogoS/ u h . V N V N V V hIU
Resistor chain
Aperture discAnode supply
Filament (heater)
XXX
}x X X \X X XT I j
Heat<SU|
>ateePW
magnet
Bun earth Basel lens
Extraction supply
TU
Plates
| _o=-X plates
Electrostatic deflection and scanning system HV
Supply 0-500 kV
Gate valve (pneumatic)
T
Sample holder |Target chamber
Fig. 4.1 - Schematic o f the 500k V implanter.
The sample holder was specially designed in order to provide background heating of the
wafer during implantation using a bank of quartz halogen lamps mounted behind the
wafer. A thermocouple was inserted between the wafer and the lamps to control the
41 Experimental details
implantation temperature. Furthermore, the sample holder was water cooled in order to
dissipate the excess heat generated by the lamps. A schematic of the sample holder is
shown in Fig. 4.2.
coolingwater
micrometfk
movement
up •downmovement
St suppression plates
ha log- lamps/ dummy
waterSilicon Si iwafaraperture
Liquid Nitrogen reservoir
Fig. 4.2 - Schematic o f the halogen lamp heater sample holder.
The 3 inch (100) Si wafer was placed inside the sample holder facing the beam but with
the wafer normal at an angle of 7° in order to reduce the occurrence of channelling.
Additionally, another silicon wafer was also incorporated into the sample holder to be
used as a 'dummy'. This enabled the beam parameters to be set up before the beam was
directed onto the wafer. A silicon aperture was used to define the implant area which is
smaller than the wafer's diameter. In these experiments the implanted area was
approximately 6.5 cm 2. Fig.4.3 shows a schematic of the 3 inch wafer and the implanted
area.
42 Experimental details
Fig. 4.3 - Schematic showing the implanted region o f a 3" wafer
Both the beam defining aperture and suppressor electrode, biased to - 300 V to suppress
secondary electrons, were faced with silicon to minimise metallic contamination due to
sputtering. The target chamber was pumped by a conventional diffusion and rotary pump
system to a pressure of less than 10'6 Torr before commencing the implantation. The
chamber included a liquid N2 cold finger to condense hydrocarbons.
The ion source used was a r-f source which was capable of producing total beam currents
of up to 100 to 150 pA . After scanning the total current incident upon the sample was 50
pA to 70 pA with a current density of typically 3-10 pA/cm2. Molecular singly charged
oxygen ( 0 +2) were implanted at energies of 140, 180 and 400 keV . Doses in the range
of 3xl017 0 + cm'2 to 7xl017O+cm'2 were selected to be close to the critical dose for the
formation of a continuous buried oxide layer [59], although some wafers were also
implanted with higher doses (l.OxlO18 0 + cm'2)to form a thicker oxide layer.
43 Experimental details
After implantation, all wafers were annealed at 1300 °C or 1320° C for 2 or 6 hours
under argon (Ar + 1/2% 0 2 ) or nitrogen (N2 + 1/2% 0 2) ambients. The anneals were
carried out either at Southampton University or Imperial College. Prior to annealing, the
samples were capped with a low temperature CVD oxide of thickness 5000 A in order to
protect the surface from contamination, unintentional oxidation and thermal etching.
Implantation and annealing details of the samples prepared at the University of Surrey are
listed in table 4.1.
44 Experimental details
Anneal
ingtim
e(hr
s.)
C+ *0 <N
Anneal
ingtem
peratu
re(°
Q 1300
= = = = 1320
= 1300
= - = = =
Anneal
ingamb
ient
csON?oxin+
< !
= - -
C"O0syn
OveoxLO+
•*
= = = -
Area
Implan
ted(cm
'2) 6.45
Implan
tation
temper
ature
(°Q
500/55
0/600/
650/70
0700
500/55
0/600
650/70
0700
Curren
t den
sity
, ( pA
/cm r—HCO = = = 7.7
56.2 7.7
57.7
5 i— iCO CO 4.65
4.3
<0 & °
£ o i—i o O x )
3.3 3.3 in 3.3 or—1 in in VO vo c-- in
Energy
(keV)
70 70 06 06 90 200 =
SIMO
Xwaf
er
G1/G
2/G4
G5/G
6G3 G7/G
8/G9
/G1
0/G1
1G1
2G1
3G1
4G1
5G1
6G1
7G1
8G1
9G2
0G2
1
45 Experimental details
Table
4.1
- S
IMOX
wa
fers
pre
pare
d at
the
Unive
rsity
of
Surr
ey.
4.1.2 Oxygen implantation at Eaton Corporation.
Six inch S I M O X wafers were prepared at Eaton Corporation using a prototype commercial implanter, namely a modified Eaton N V - 10-80 high current batch implanter [120]. This machine was used to implant batches of 6" (100) C z wafers using mechanical scanning to achieve lateral dose uniformity. The silicon wafers were implanted using either atomic or molecular singly charged (0+2 or 0 +) ions at energies of 60 keV(
0*, implantations), 30 k e V or 40 k e V with a range of doses which varied from 2.2x1017 O V c m ' 2 to 3.0xl017 O fcm'2. This range of doses was selected in order to form a continuous buried oxide layer[59]. Background heating of the wafers was obtained using quartz halogen lamps which enabled the wafers to be pre-heated at 500°C before commencing the implantations. The average wafer temperature was measured using an optical fibre thermometer placed 3 m m in front of the wafers. The complete 6 inch wafer was exposed to the ion b ea m except for small areas at the periphery beneath the wafer clips ( see fig. 4.4 ).
High temperature annealing ( H T A ) was performed subsequently to implantation at a temperature of 1300°C for six hours in a nitrogen ambient mixed with 0.25% oxygen. Table 4.2 shows the implantation and annealing details of the complete batch of wafers prepared at Eaton Corporation.
46 Experimental details
Fig. 4.4 - Schematic o f the implanted areas o f the 6 " wafers supplied by Eaton Corporation.
S I M O X oxygenmassimplanted
Dose xlO17 0 cm'2Energyk e V Implantationtemperature(°C)
Annealingtemperature(°C)
AnnealingambientEl 32 2.2 60 530 1300 N 2 +0.25% 0 2E 2 32 2.2 60 530 1300 tlE3 16 2.2 30 530 1300 II
E4 16 2.2 40 530 1300 II
E5 16 2.4 40 550 1300 II
E6 16 2.4 40 600 1300 If
E7 16 3.0 40 510 1300 It
Table 4.2- Implantation details o f wafers produced by Eaton Corporation using the modified NV-10-80 implanter. All wafers were annealed for six hours.
47 Experimental details
4.1.3 Standard SIMOX Substrates.
Device grade 4" S I M O X substrates were supplied by two different vendors (IBIS [ 118], L ET I[ 1 19 ] ) w h o prepared the substrates using commercial Eaton N V - 2 0 0 implanters either by single or multiple step oxygen implantations (SIA SIMOX). In these samples the complete wafer was implanted. The implantations were carried out at high temperatures, typically 700°C, by pre-heating the wafers using halogen lamps. The implanted wafers were subsequently annealed for six hours at 1300°C in an Ar + 1/2% 0 2 atmosphere. Details of the implantation conditions are listed in table 4.3.
S I M O X Batch Source Implantmethod Nominal dose (cm2) Energy(kevy Siliconoverlayerthickness(A)Al IBIS single 1.8xl018 190 2000A 2 It single II 11 2000A3 11 single 11 II 2000A 4 If multiple ft It 1800B1 L E T I multiple II II 2100B 2 fl single II 200 2500B 3 11 single II 11 2500B 4 II single 1.78xl018 190 2100B 5 II single 11 It 2100B 6 II single If II 500
Table 4.3- Implantation details o f standard SIMOX substrates supplied by IBIS and LETI.
48 Experimental details
It is observed from Table 4.3, that all samples had a relatively thick silicon overlayer (typically > 2000 A) except wafer B6, which was sacrificially oxidised by the vendor (LETI) in order to reduce the silicon overlayer thickness. W e n o w designate these thick S I M O X substrates as "standard" S I M O X .
4.1.4 B F \ . Ge+ and S t implantations.
Selected samples cut from standard S I M O X wafers (namely wafers Bl, B3(2) and
B4(2)) were subsequently implanted with B F 2+, G e + and Si+ ions at the University of
Surrey. The purpose of the implantations was to amorphise the silicon overlayer in the vicinity of the B O X layer in these standard S I M O X wafers. The implantations were carried out at r o o m temperature using a carousel sample holder, in which a m a x i m u m of 16 samples ( approximately 1 c m 2) could be loaded. The samples were lightly clamped
with aluminium clips and the total area implanted was 1 cm' . The B F 2 implantations were carried out at energies of 270 k e V with doses which varied from 8.0xl013 cm'2 to
2.0xl014B F +2cm'2 (see table 4.4). For the Si+ implantations the energy used was 190 k e V
and the doses varied from l.OxlO13 cm"2 to 5.0xl015 cm"2 (see table 4.5). Finally, the
G e + implantations were carried out at energies of 330 k e V with doses which varied from lxl013 cm'2 to 3.0xl014 cm"2 ( see table 4.6). The energies and doses were selected
using the L U P I N [121] simulation programme in order to place the peak of the damage profile near the Si/Si02 interface, and therefore selectively amorphise that region.
49 Experimental details
sampleID S I M O Xwafer Ionimplanted Energy(keV)
Dos e ( xl014cm'2)Bl/BF,7l B 1 B F +, 270 0.8B1/BF/V2 II II it 1B1/BF//3 II It it 2B1/BF//4 II II it 5B1/BF,75 It II n 8B1/BF,76 II II H 10B3/BF,7l B3(2) II it 0.8B3/BF,72 II II ii 1B3/BF,73 It 11 ii 2B3/BF,74 II II ii 5B3/BF,75 II II ii 8B3/BF,76 II II ii 10B4/BF7/1 B4(2) II ii 0.8B4/BF,72 ■I II n 1B4/BF,73 ii II it 2B4/BF,74 n II ii 5B4/BF,75 n II ii 8B4/BF,76 ii II ii 10B4/BF,77 it II ii 20
Table 4.4- Details o f samples implanted with BE*2 ions.
SampleID S I M O Xwafer Ionimplanted Energy(kef) D os e (xlO13 c m ’2)B4/Si7l B4(2) Si+ 190 1B4/Si72 ii II ii 2B4/Si73 ii II ii 5B4/Si74 n II ii 7.5B4/Si75 n II M 10B4/Si76 ii II ii 20B4/Si77 ti II it 30B4/Si78 ii II n 50B4/Si79 ii 11 n 60B4/Si7l0 ii II n 80B4/Si7l 1 ii II n 100B4/Si7l2 n II ii 250B4/Si7l3 ii II ii 300B4/Si7l4 H II ii 500Table 4.5- Details o f samples implanted with Si+ ions.
50 Experimental details
SampleID S I M O Xwafer Ionimplanted Energy(keV) Dose (xlO13 cm'2)B3/Ge7l B3(2) Ge* 330 1B 3/ Ge 7 2 II fi II 2.5B3/Ge73 li ii II 5B3/Ge7l it if II 7.5B3/Ge+/2 it ii II 10B3/Ge73 if it It 12.5B3/Ge7l n m tl 15B 3/ Ge 7 2 li « l l 17B3/Ge73 li if II 20B3/Ge7l if i i 11 25B 3/ Ge 7 2 i i ii 11 30B4/Ge7l B4(2) ii 11 1B 4/Ge72 ll M II 2.5B4/Ge73 li
" 11 5B4/ Ge 7 4 i i ii l l 7.5B4/Ge75 ll ii ll 10B4/ Ge 7 6 n If 11 12.5B4/Ge77 i i 11 11 15B4/Ge78 it 11 II 17B4/Ge79 ti 11 11 20B4/Ge7lO ti M 11 25B4/Ge7l 1 11 II 11 30
Tabled. 6- Details o f samples implanted with Ge ions
51 Experimental details
Standard S I M O X as well as BF^, G e + and Si* implanted S I M O X samples were thermally oxidised at the University of Surrey using a dedicated dry oxidation furnace. The furnace consisted of a quartz semi-open tube ( 3 inch in diameter) which enabled that a flux of oxygen to be flown continuously during the oxidation treatment. The temperature inside the furnace was monitored using a Pt-Pt (13%)/Rh thermocouple and the m a x i m u m temperature variation along the central zone of the furnace ( along a zone of length 10 cm) was estimated to be ± 2 °C. A flow of 1.5 Iitre/min of pure oxygen gas was bled into the tube, approximately 10 min. before loading the samples and then reduced to 1.0 litre/min. a couple of minutes before the oxidation in order to perform the thermal treatment under standard dry oxidation conditions [122], The samples were, then, horizontally loaded into a quartz sample holder (5 c m long) which was placed inside the tube near its extremity. The quartz sample holder was maintained at that position until the air inside the tube was blown away ( typically for 3 min.). Then, the sample holder was pushed into the centre of the furnace using a quartz rod. The loading time was kept short, typically one minute, in order to minimise the thermal oxidation during the loading procedure. The range of oxidation temperatures used varied from 800°C to 1100°C and the oxidation times from 30 min. to 74 hours. After the thermal oxidation, the samples were immersed in a H F( 40 % ) solution for 2 min. to strip off the thermal oxide.The thickness of the thermal oxide was estimated measuring the height of an oxide window using a Rank Taylor Tallystep stylus instrument. The oxide window was formed by protecting a selected area of the sample with a H F ( 4 0 % ) resistant mask. The final thickness of the silicon overlayer was determined by 1.5 M e V R B S analyses. Table 4.7 shows a summarised list of the samples oxidised in this project. A complete list of all
4.1 .5 S acrific ia l oxidation
52 Experimental details
sacrificially oxidised samples is presented in Appendix E.
Sample S I M O Xwafer Oxidation temperature (°C)
Oxidation time (hours)
Bl/1 v Bl/20— > B1 850/900/9501000/1050/1100 0.5 60B2/1 B2/22 B 2 900/1000/10501100 0.6 74
B3/1 B3/23 — > B3 900/950/10001050/1100 0.5 74B4/1 B4/23-» B 4 850/900/9501000/1100 0.5 60
B5/1 B5/22 B5 900/950/10501100 0.5 — > 60
Bl/BFyi/1 -> Bl/BF+2/6/l B1 1100 4
B3/BF'2/1/1 -> B3/BFV6/1 B3(2) 1100 4B4/BFyi/l -> B4/BF+2/7/l B4(2) 1100 4
B4/Si71/1 -> B4/Si7l4/1 B4(2) 1100 4
B4/Ge71/1 -> B4/Ge7l 1/1 B4(2) 1100 4
B3/Ge71/1 -» B3/Ge7l 1/1 B3(2) 1100 4
Table 4.7 - Sacrificially oxidised S I M O X samples.
53 Experimental details
4.2 Characterization methods.
Crystallographic defects in the S I M O X samples were characterized using chemical defect etching, plan view transmission electron microscopy (PVTEM), cross-section transmission electron microscopy ( X T E M ) . Rutherford backscattering (RBS), double crystal X-ray diffraction ( X R D ) analysis and Secondary Ion Mass Spectroscopy (SIMS) were also used to analyze a limited number of samples.
4.2.1 Chemical defect etching.
Chemical defect etching was used to delineate crystallographic defects in the siliconoverlayer and in the silicon substrate just beneath the B O X . A number of standard bulkSi etchants were used in this project (see Table 4.8). However, for the analysis of thin silicon overlayers ( <1000 A) it w as found necessary to develop a n e w defect etchant capable of reliably deliniating crystallographic defects in these thin layers. Experimentalresults regarding the chemical composition and the main characteristics of this n ewetchant are presented in chapter 5.
SiliconEtchant Formula Etch pit morphologySecco[123] 2 HF(49%): 1 K 7Cr?0 7 (0.15 M ) isotropicSchimmel[124] 2 HF(49%): 1 CrO, (1 M ) isotropicYang[125] 1 HF(49%): 1 CrO, (1.5 M ) AnisotropicWright [126] 60 ml HF(49%): 30 ml H N O v 30 ml C r0 3 ( 1 M): 60 HAc: 60 H 20 ( 1 gr. C u N 0 3 )
Anisotropic.
Table 4.8 - Bulk silicon etchants used in this project.
54 Experimental details
4.2 .1 .1 E tc h in s ex p erim en t details.
All the etching experiments were carried out in a climate controlled r o o m (20°C ±1 °C) at the University of Surrey. Additionally, the etchant temperature was further controlled by placing the beaker containing the etchant in a temperature controlled bath, for a couple of hours, before the commencement of the etching experiment. In this way, the etchant temperature was controlled over the range from 21 ± 1 °C to 35 ± 1°C.
II Etch rates
The etch rates of the chemical solutions were estimated by etching a partially masked silicon sample for a predetermined time. After removing the mask ( H F resistant tape) , the step height between the etched and non etched region was determined using a R an k Taylor Tallystep stylus instrument. Furthermore, the etch rates were double checked by estimating the etching time needed to completely remove the silicon overlayer of a standard S I M O X sample of k n o w n thickness ( Tsi= 2000 A). B y so doing, the uncertainty was estimated to be less than ± 5 0 A/min.
C Ageing o f the chemical solution.
In order to avoid problems with the deterioration of the etchant during storage, fresh solutions were mixed for each set of analyses. Additionally, the etch rates of the fresh solutions were always measured before beginning a n e w set of experiments.
A. Etcliine temperature
55 Experimental details
All the S I M O X samples submitted to chemical etching were, additionally, immersed in a H F ( 4 0 % ) solution for a m i n i m u m of 30 sec. and a m a x i m u m of 5 min. in order to form cavities underneath the etch pits ( see Fig. 2.6). The double etched samples were, then, observed either by optical microscopy ( O M ) using an Olympus microscope or by scanning electron microscopy ( S E M ) using a Cambridge S250 Stereoscan electron microscope. The defect densities were estimated by counting the etch pits in a m in im u m of ten micrographs ( optical microscopy or S E M ) taken from the area of interest. The m ea n etch pit density and the standard deviation could then be estimated for every region of interest. The lateral distribution of defects in the S I M O X wafers was determined by measuring the etch pit densities in different areas of samples cut along the [110] diameter of the S I M O X wafers. In some cases, defect etch pit maps of whole S I M O X wafers were obtained. Furthermore, the depth distribution of defects w as also obtained by determining the etch pit density as a function of the thickness of silicon removed (e.g. etching time). For this purpose, small samples cut from adjacent areas were etched for different times and the thickness removed was estimated measuring the step height at the edge of a non etched region. Whenever possible, the etch pit results were correlated with plan view T E M or even X T E M analysis in order to determine the nature and geometry of the crystallographic defects.
D. Determination o f the defect etch pit densities.
56 Experimental details
4.2.2 Transmission electron microscopy
Characterization of crystallographic defects was carried out by transmission electron microscopy. Samples were observed under plan view and cross section views using the Jeol 200 C X and 2000 F X high resolution microscopes in the Microstructural Studies Unit ( M S S U ) at the University of Surrey. The preparation of the samples for T E M analyses is described in appendix D. After preparation the samples were loaded in a double tilt sample holder in order to accurately obtain the two beam condition[127] and also to scan through different zone axes. Determination of the Burgers vectors of the crystallographic defects was not carried out in this work. The areas analyzed by plan view T E M were larger than 1000 p m 2 enabling the determination of
5 2defect densities as low as 1.0x10 cm" .
4.2.3 Rutherford Backscatteriii2 (RBS)
Rutherford backscattering (RBS) and channelling ([100] direction) analyses were used to determine the composition and crystalline quality of selected S I M O X samples. In this project, w e have used the 2.0 M V V a n de: Graaff accelerator at the University of Surrey, to produce a 1.5 M e V or 1.0 M e V H e + beam. The backscattered He* ions were collected using a silicon surface barrier detector, having an energy resolution of 16 keV. This detector w as located inside the target chamber at an approximate angle of 32° to the incident b e a m axis . The energy spectrum of the backscattered He* ions w as used to determine the mass and depth distribution of the component elements of the target. Furthermore, simulations of the R B S spectra were carried out, using the X R B S S I M [ 1 2 8 ]
57 Experimental details
simulation programme, in order to accurately determine both the thickness of the silicon overlayer and the buried oxide layer.
4.2.4 Other methods.
Double crystal X-ray diffraction ( X R D ) analysis was also used to determine the stress near the implanted/ non implanted area of a typical 'as-implanted' S I M O X wafer implanted with a dose of 3.0xl017 0 + c m ' 2 and energies of 70 keV.The X R D experiments were performed by Dr. M . Lourenco using the 50 k V X-ray generator at the University of Surrey. The rocking curves were obtained by scanning through the (004) diffracted
b e a m ( 0B= 34.58°) using the characteristic C u K a line (1.5432 A ) .
Furthermore, Secondary Ion M as s Spectroscopy (SIMS) was also used to quantify the C u concentration present in the (100) Cz- Si wafers implanted at the University of Surrey and at Eaton Corporation. Dr. A. Nejim is acknowledged for arranging for the S I M S analyses to be earned out.
58 Experimental details
C h a p t e r F i v e : N e w e t c h a n t f o r t h i n S O I .
5J Introduction
In this chapter w e present the experimental results associated with the development and optimisation of the n e w defect etchant, which was subsequently used in the study of crystallographic defects in thin silicon films (Chapters 6 and 7). T E M analyses have been used to validate the interpretation of the data obtained by chemical defect etching. Furthermore, the lateral and depth distributions of crystallographic defects in thick film S I M O X have been determined to investigate the uniformity of device grade wafers and also to s ho w the improved selectivity of the n e w etchant.
This chapter contains two sections. In the first one (5.2), w e present the results of the n e w defect etchant developed to delineate defects in ultra thin film silicon layers, while in the second section (5.3), w e apply this n e w etchant together with T E M analysis to study the lateral and depth distribution of crystallographic defects present in thick S I M O X .
59 New etchant
5.2 Development o f an ovtimum defect etchant for thin silicon layers
5.2.1 Dilute silicon etchants
In chapter two, w e have seen that although different etching systems have been developed for the study of bulk silicon materials, so far, no defect etchant has been specially developed for ultra thin silicon layers (< 1000 A ) . Bulk silicon etchants, such as the Wright[123], Secco[124], Y a n g [125 ] and Schimmel [126], can be used to analyse thin silicon layers, however these solutions have etch rates of about 1.0 gm/min. or faster (see Fig. 5.1-a) and therefore, the removal of thin silicon overlayers is difficult to control, specially if the layer is as thin as 500 A .
A simple w a y of decreasing the etching rate is to dilute the chemical solution by increasing the H ^ O content. In figure 5.1 b) w e plot the etch rates of different silicon etchants upon dilution. F r o m Fig. 5.1 b) it is observed that the etch rates of these etchants decrease considerably as a function of the dilution. Indeed, the etch rate of the ten fold dilute Secco solution is decreased to approximately 200 A/min, which is slow enough to allow good control of the removal of the silicon overlayer. However, a defect etchant can only be considered as reliable if all the crystallographic defects intersecting the surface are delineated.Therefore, it is also important to determine the effectiveness of the dilute solutions to delineate defects.
60 New> etchant
~oCD>oECDv _
i f ) CI) 0 co
J ZI—
2 3E t c h i n g t i m e ( m i n . )
c'E
0•A—*Ctf
LU
0 6Dilution
8 10
Fig. 5.1 - a) Thickness removed upon etching time and b) etch rate upon the number o f times the solution has been diluted for the Wright, Yang,Schimmel and Secco etchants.
61 New etchant
S I M O X wafers produced commercially by the IBIS Corporation, namely wafers Al andA 2 (silicon overlayer thickness of 2000 A , see table 4.3), were etched, following theprocedure described in section 4.2.1, and observed by optical and scanning electronmicroscopy (SEM). Selected samples cut from each of the wafers were analysed by planview T E M and the threading dislocation density in these samples were estimated to be (1.0 ±0.5 )xl06/cm2for wafer Al and (9.0 ± 1 ) xl06/cm2 for wafer A2.Figure 5 .2 shows S E M micrographs of samples prepared from adjacent regions of waferAl, etched with a two-fold dilute Wright solution for different etching times (15 sec, 30sec and 35 sec). It is seen (see fig. 5.2 a to c) that the etch pit density increases withetching time and only reaches a value which corresponds to the crystallographic defectdensity determined from the T E M analyses w he n 1900 A of silicon has been removed (e.g for 35 sec. etching time ).
In order to assess the effects of the dilution on the formation of etch pits, two standard
Fig. 5.2 a-( continued)
62 New e/chant
Figure 5.2- SEM micrographs o f samples from wafer A l etched with the 2 fold Wright etch fo ra ) 15 sec. ( 850 A ) b) 30 sec (1650 A) c) 35 sec. ( 1900 A)
Similar experiments have been carried out for the other etchants and in Figure 5.3 w e plot the etch pit density upon the thickness of silicon removed for the Schimmcl.Ynng, Wright and Secco dilute etchants using as standard material, samples prepared from adjacent
regions of wafer A l .
63 New etchant
Etch
pit
de
nsity
(■
cm
-2)
500 1000 1500T h ic k n e s s of S i re m o ve d (A)
2000
Fig. 5.3 «- b (continued)
64 New etchant
Etch
pit
de
nsit
y ( c
m'2
.
0 500 1000 1500T h ic k n e ss of S i re m o ved <A)
2000
10'
10*
i o :
10'
1 0 :
1 0 2
1 0 1
1 0 c
400 800 1200 1600Thickness of Si removed (A)
2000
Figure 5.3- Etch pit density against thickness o f silicon removed for a) Wright, b) Yang , c) Schimme! and d) Secco dilute etchants (data collected from wafer A l: 1.0x1 & cm 1 ; etchant temperature: 20±I°Cy
65 New etchant
= 10c*E
0CO
JZoLU
From Fig. 5.3 (a-d) it is observed that for each of the etchants the etch pit density only reached a saturation value that corresponds to the defect density determined by plan view T E M (e.g. 1.0x106/cm) when a certain minimum thickness of silicon had been removed. The minimum thickness (Tm) of silicon that needs to be removed to obtain reliable results depends upon the etchant and also upon the dilution of the solution. For example, for the 5-fold diluted Secco etch (Fig. 5.3 d), it is necessary to remove about 1100 A in order to delineate all of the defects (density : 1 .OxlO6 cm'2).
5 — -------- ---------- ---------------------------------------------------------------------------------------------------------------------------10
10'
10‘
"Sec C O
:2,5x~dt nte:
TJX
Sohimmel
dilute
2x
10x dilute
FrSx-dtlttte -:
■ ...dilute
ang Vy right= * r .
2.0>: dilute ,
500 900 1300T m ( A )
1700
Figure 5.4- Etch rate against Tm fo r the Wright, Yang, Schimmel and Secco based solutions.
Fig. 5.4 summarises the results of the different etchants in terms of etch rate against T m. It is evident from Fig. 5.4 that silicon etchants such as the Wright and Yang etchants, give poor results since most of the silicon overlayer has to be removed ( T > 1500 A) to obtain a one to one correlation between etch pits and defects imaged in the T E M . This
66 New etchant
result is not unexpected since both the Ya n g and Wright etchants form well defined etch pits due to preferential etching of silicon in the {111} planes near the dislocation core and consequently, the formation of etch pits using these systems is very slow. U p o n dilution the capability to form etch pits is further decreased and fewer than 1 % of the crystallographic defects (see Fig. 5.3 a) are delineated, even after removing almost all the silicon overlayer.
Preferential etchants that form isotropic etch pits ( Schimmel and Secco) give better results. However, it is observed that Schimmel based solutions are inferior to the Secco based ones because more material has to be removed to obtain reliable results (see Fig. 5.4). T ra for the standard Schimmel etchant is 1200 A and increases rapidly for the dilute solutions. The best results were obtained for solutions having potassium dichromate (K^C^Oy) as the oxidising agent (e.g. Secco solutions). Indeed, the min im u m thickness for the standard Secco is 700 A and increases only to 1250 A for the 10 fold dilute solution.
Considering that the initial thickness of the silicon overlayer was 2000 A , these results mean, in practical terms, that all the dilute CrO, based solutions analysed in this work (e.g Yang, Wright and Schimmel etchants) are not suitable for the study of ultra thin film S I M O X ( < 1000 A ) since the formation of etch pits is not fast enough to allow the delineation of all crystallographic defects in the thin films. It is important to point out that the main difference between Secco and Schimmel based solutions is the presence of the alkali ions (IC) in the former. W e believe that the presence of IC ions in the solution increases the selectivity of the etchant. This assumption is backed by the early w or k from Gilman et al[ 129] which showed that the selectivity of chemical defect etchants was improved by the introduction of metal ions. They showed that the metals ions are
67 New etchant
Etch
ra
te (A
/min
.)
generally adsorbed on the active sites of the surface and inhibit the retreat of monomolecular steps across the surface, reducing their velocity (V,) without modifying the etch rate (Vn) at the crystallographic defect (see appendix B).
The etching temperature is another parameter that influences the kinetics of the reaction. For example, Figure 5.5 shows the dependence of the etch rate upon temperature for the dilute Secco solutions. It is seen, that the etch rate increases linearly with temperature for the three Secco solutions investigated.
Figure 5.5 - Etch rate dependence o f dilute Secco solutions upon temperature.
Figure 5.6 (a to d) show the temperature dependence of the etch pit density upon the thickness of silicon removed for samples prepared from adjacent areas of wafer A 2 ( defect density: 9.0xl06cm'2). All four sets of data, which were recorded using samples similarly etched at different temperatures, namely 21°C, 26°C, 31°C and 35°C, respectively show the same trend. The etch pit density reaches a saturation value between 5.0x106 cm'2 and 8.0x106 cm'2 after a particular thickness of silicon ( T J has been removed.
68 New etchant
'Eo
COczCDQ
oLLJ
-Eo
COc=CD
■q _j z :oLLJ
Figure 5.6 - Etch p it density versus thickness o f silicon removed for samples prepared from wafer A2 at a) 21° C , b) 26 °C , b) 31° C and d) 35° C
70 New etchant
It is also observed that the value of the minimum thickness T m is insensitive to temperature and depends only on the dilution of the solution, as previously observed (see fig. 5 .4). This result suggests that the temperature of the solution does not accelerate the nucleation of etch pits.
Figure 5.7 is a typical plan view T E M micrograph from a sample prepared from wafer A 2 etched for 2 min. (1000 A removed) using the 5 fold dilute Secco solution. Three well delineated etch pits are observed while no defect contrast was seen in other regions indicating that all of the crystallographic defects were delineated.
Figure 5.7 - Typical plan vieyv TEM micrograph o f sample A2 etched for 2 min. using the 5 fo ld dilute Secco
It is concluded that crystallographic defects can be reliably delineated using the 5 fold dilute Secco solution providing the silicon overlayer is greater than 1000 A. However, for silicon layers thinner than 1000 A, a n e w etchant is required.
71 New etchant
5.2.2 New etchant
In section 5.2.1 w e have shown that the effectiveness of the defect etchants was controllably reduced upon dilution of the solution (see Fig. 5.4). W e believe that this decrease in the effectiveness of the dilute solutions is due to a decrease in the concentration of the oxidising agent reducing its etching capabilities ( see appendix B), O n the other hand, w e have also showed, that the preferential etching characteristic of the solution will be enhanced by the presence of alkali ions (K+) in the solution. F r o m these findings, w e concluded that to obtain a highly preferential etchant w e need to develop a n e w system with a strong oxidising power (i.e with a high concentration of the oxidising agent )[125] and at the same time having a high concentration of alkali (K+) and metal (Cu"'") ions. The introduction of the metal ions being aimed to further inhibit the retreat of monomolecular steps and by so doing reduce the planar etch rate of the solution [129],
The n e w etchant that w e developed in this project is based on the H F : H N 0 3:H20 : K 2Cr20 7 : C u ( N 0 2 )3 system . The oxidising power of this system , compared to the H F : H N 0 3:H20 , is increased by the presence of Cr++ ions (see appendix B) and at the same time the introduction of K + and C u ++ ions in the solution will inhibit the retreat of monomolecular steps. A n optimum formula was obtained by changing the relative quantities of the etchant components in order to obtain the highest etching anisotropy with the slowest planar etch rate. The optimum formula for this n ew system is 50 ml HF(40%):70 ml H N 0 3(70%):170 ml FL,0 (1.0 gr. K ^ C r ^ + 4,0gr. C u ( N 0 2):H20, which was established by carrying out a large number of etching experiments.
72 New etchant
In Figs, 5.8 and 5.9 w e present calibration data for the n ew chemical solution. Figure 5.8 (a) and (b) show the etch rate and dependence of the etch rate with temperature, respectively.
The n e w silicon etchant is found to be well behaved, as the removal of silicon varieslinearly with increasing time and also the etch rate has a linear dependence upontemperature. The etch rate of the n e w solution at r oo m temperature (21°C) is 450 A/min, which ensures good control of the dissolution of silicon from the thinnest films ( < 500A ) .
Fig 5.9 (a) and (b) show the dependence of the etch pit density in samples cut from wafers A l and A 2 upon the thickness of silicon removed. For comparison w e have also plotted data obtained using dilute Secco solutions. It is clear from Fig. 5.9 (a) and (b) that the m in im u m thickness T m for the n e w solution is about 500 A for samples from both wafers. The increase in the effectiveness of the n e w etchant confirms our supposition that the introduction of Cu** and K + ions inhibits the retreat of monomolecular steps across the surface, reducing the dissolution rate of the surface without modifying the etch rate at the crystallographic defect and, therefore, enabling the formation of large etch pits even w h e n the thickness of silicon removed is small.
The evolution of etch pits using this n e w etchant is shown in Figure 5,10 (a-d) which includes plan-view T E M micrographs of samples prepared from wafer A 2 after different etching times. For an etching time of 7 sec. (Fig. 5.10 a) w e observe that no preferential etching of dislocations has occured. The small amount of material removed ( about 50 A )
is insufficient to show nucleation of etch pits at the site of the dislocations. The onset of nucleation (Fig. 5.10 b) appears to occur after the removal of 100 A (14 sec.) of material.
73 New etchant
It is noted that the growth of the etch pits appears not to be homogeneous during the initial stage and that initially not all the dislocations have resolvable etch pits.
1 5 0 0
"O0>oE0
(7)ocoto0co Ic I—
1000
5 0 0
00 1 2
E t c h i n g t i m e ( m i n . )7 0 0 1 i ■ i i . - r
6 0 0■ b
-
c 5 0 0 -
Et i 4 0 0 .
00 i_ 3 0 0 • r
_ Co •+— '
L U 2 0 0 - -
1 0 0 - -
0 - - - - - - - - - - - - - -- - - - - - - - - - - - - 1_ _ _ _ _ _ _ _ i_ _ _ _ _ _ _ _ 1_ _ _ _ 1 - - - - - - - - - - - - - - 1- - - - - - - - - - - - - 1- - - - - - - - - - - - - 1
10 1 4 1 8 2 2E t c h i n g t e m p e r a t u r e f C )
2 6 3 0
Fig. 5.8 a) Thickness removed upon etching time and b) dependence o f the etch rate upon temperature fo r the new etchant.
74 New etchant
0 400 800 1200 1600 2000T h i c k n e s s o f S i r e m o v e d ( A )
1 0
1 0
CNi<—< 1 0
so 1 0
> -*—* 1 0'(/)c=0TD 1 0 :
'o_ 1 0 :J ZoL U 1 0
1 0 (
0 400 800 1200 1600 2000T h i c k n e s s o f S i r e m o v e d ( A )
Fig. 5.9 - Etch pit density upon thickness o f silicon removed for samples prepared from wafers a) A l and b) A2.
75 New etchant
Fig. 5.10 - Plan view TEM micrographs o f samples prepared from adjacent samples from wafer A2 etched fo r a) 7 sec., h) 14 sec., c) 24 sec., d)60 sec., and e) 120 sec. using the neyv etchant
77 New etchant
After the removal of 150 A of material (Fig 5.10 c) etch pits are visible at the sites of all resolvable dislocations, however the diffraction contrast indicates that the etch pits have not penetrated completely through the silicon overlayer. W h e n 450 A of the silicon overlayer has been removed (Fig. 5.10 d), it is clear that well defined etch pits have formed, with all the silicon around all of the dislocations being etched away. These observations agree closely with the data shown in Fig. 5.9. U p o n a further increase in the amount of material removed (900 A , Fig. 5. 10 e) the etch pit size increases considerably but the etch pit density is unchanged, since there are no further dislocations to nucleate additional etch pits in this material. Analysis by plan view T E M of samples etched using the 5 fold dilute Secco solution showed similar trends, although the amount of material needed to be removed for the onset of nucleation was considerably greater (250 A ) and the etch pits only penetrated through the top silicon layer w h e n about 900 A of silicon had been removed. These results give confirmation that the formation of etch pits is indeed a nucleation-dependent process. F r o m the nucleation theory of etch pits (Appendix B) it is k n o w n that the probability that an etch pit will be nucleated at a dislocation is directly proportional to the undersaturation of the m e d i u m ( Si surface/etchant) and to the square of the Burgers vector of the defect. In terms of this theory the different etching times for the etchants (new etchant and dilute Secco etchant) to nucleate etch pits can be explained by the presence of inhibitors that decrease the free enthalpy required for nucleation[20]
Fig. 5.11 shows the diameter of the etch pits as a function of the thickness of silicon removed, where the data has been collected from plan view T E M micrographs. It is seen that the n e w chemical solution forms etch pits that are larger by a factor of two than those formed by the five-fold dilute Secco. For 900 A of silicon removed the diameter of
78 New etchant
the etch pit produced by the n e w chemical etch is 0.2 pm, whilst it is less than 0.1 p m for the 5 fold dilute Secco. This feature illustrates, once more, the increased effectiveness of the n e w solution over the 5 fold dilute Secco.
E
7CD
O4 — >
CDCD
CD4 — '
CDEojb
T h i c k n e s s o f S i r e m o v e d ( A )
Fig. 5.11 - Etch pit diameter against thickness of silicon removed.
The use of this new etchant is further reported in section 5.3 of this chapter and in chapters 6 and 7. It is shown that threading dislocations are successfully delineated in thin silicon layers by using the n e w etchant. Stacking faults (SF) in thin layers (Sample B6, silicon overlayer thickness: 500 A ) are also delineated using this etchant (see fig. 5,12). Identification of stacking faults has been made possible by the faceted nature of the etch pits and their orientation along <110> directions. Fig. 5.13 shows SF present in another sample which has been thermally oxidised. It is obseived that SF can be distinguished from a pair of dislocations if the projected length of the fault is larger than 0.5pm.
79 New etchant
Fig 5.12- SEM micrograph o f SF observed in thin SIMOX (Ti = 500 A).
Fig 5.13- SEM micrograph o f threading dislocations and SF observed in sacrificially oxidised SIMOX. It is observed that SF can be distinguished from a pair o f dislocations if the projected length o f the SF is greater than 0.5 pm.
80 New etchant
5.3 Distribution of defects in thick SIMOX.
In this section w e use the n e w chemical defect etchant to determine the distribution of defects present in the silicon overlayer of standard S I M O X wafers produced commercially by the IBIS and L E T I Corporations.
5.3.1 Lateral distribution o f defects.
Figure 5.14 a) shows the lateral distribution of defects from the centre to the edge of wafer A 2 . It is observed that although small variations are evident (from 8.0xl06cm'2 to 6.0x106 cm'2), the overall defect distribution is relatively constant suggesting that the vendors have good control of the processing parameters during implantation and annealing. Lateral uniformity could also be evaluated from the uniform interference colour (green) observed over the whole wafer and was found qualitatively to be good.
This result is contrasted with the one presented in Fig. 5.14 b) obtained from another S I M O X wafer (A3) which had a large colour difference between the centre (green) and the edge of the wafer (pink). A difference in the defect density of one order of magnitude is observed from the centre (6.0x10° cm'2) to the edge (6.0x10' cm'2) of this wafer (Fig. 5.14 b).
81 New etchant
y r u u f i
C MIEo
C OLUfe•4— ’
'coc0T3
*Q-4— ■LU
Distance from centre to the [110] edge of wafer (mm)
b
icr: 10 10 20 30 40 50
Distance from centre to the [110] edge of wafer (mm)
g re e n m ix e d p in k
Fig. 5.14 - Etch p it density distribution from the centre to the edge o f a) wafer A2 and b) wafer A3. The colour (by eye) o f the layers is indicated at the top o f the figure.
82 New etchant
C hannel Num ber
Fig. 5.15 - RBS spectra from two different areas o f wafer A3.
R B S analysis of different areas of the wafer A 3 ( see Fig. 5 15) confirmed that the silicon overlayer and buried oxide layer thicknesses were significantly different in the centre and the edge of the wafer. With the aid of an R B S computer simulation programme it was possible to determine that the thickness of the silicon overlayer near the edge of the wafer was 2100 ±60 A whilst it was only 1900 ± 60 A close to the centre O n the other hand, the B O X thickess near the edge was thicker ( 4100 ± 50 A ) than in the centre (3900 ± 50 A ). These results, associated with the different interference colours ( green to pink), indicate that lateral dose inhomogeneities occured during implantation. The increased defect density in the centre of the wafer is attributed to the higher implanted dose which will increase the number of point defects created during implantation and lead to greater lattice disorder.
83 New etchant
Etch
pit
den
sity
(c
m
The depth distribution of defects in the silicon overlayer was also determined Fig 5 16 shows the etch pit density upon the thickness of silicon removed for two multiple implanted S I M O X substrates, namely wafers A4 and B1 (see table 3.2) It is observed that the etch pit density ( threading dislocations ) for both materials is around 1 OxlO3 cm'2. However, an increase in the etch pit density for thickness higher than 1000 A is
observed for both wafers which saturates at a value of about 3 0x1 0 6 c m 2 W e speculate that this increase in the etch pit density is caused either by the presence of localized defects or by irregularities found near the Si/Si02 interface
5.3.2 Defect distribution- depth
0 250 500 750 1000 1250 1500 1750T h i c k n e s s o f S i r e m o v e d ( A )
Fig. 5.16 - Etch pit density upon thickness o f silicon removed for a) A4 and b)Bl multiple implanted SIMOX wafers.
Plan view TEM analysis of a sample prepared from wafer A4, confirmed that the
84 New etchant
threading dislocation density is below the resolution limit for this technique ( e g < 1.0x1 O' cm'2) and also revealed the existence of small extended defects in the silicon overlayer. Fig. 5.17 shows a plan view T E M micrograph of a sample prepared from wafer A 4 observed under weak beam conditions. It is possible to observe alpha fringes inside two opposite {111} faulted planes. This contrast is characteristic of a stacking fault pyramid: (SFP)[130] The SF P density measured was l .OxlO7 cm'2, explaining theincrease in the etch pit density observed in Fig. 5.16 From Fig. 5.16 and 5.17, w e conclude that the increase in the etch pit density is due to the presence of SFP localized near the back of the Si overlayer. Although the etch pit density is approximately 3 times lower than the actual SFP density estimated by plan view T E M analyses, w e speculate that this difference is due to the different magnifications used to estimate the defect densities
500 A
5.17 Plan view TEM micrograph o f sample A4 under weak beam conditions. An extended defect is observed. Fringes inside two faulted planes are clearly visible.
85 New etchant
The presence of non threading defects was also observed in other S I M O X wafers. Fig. 5.18 (a-b) include plan view T E M micrographs of a sample prepared from wafer B4. In Fig. 5.18 a) two paired threading dislocations can be observed. The threading dislocation density estimated by plan view T E M analysis of this particular sample was 5.0x105 cm'2
In Fig. 5.18 b) an extended defect is observed which w e conclude is an elongated SFP since the edges of the projected defect are orientated along <110> directions, and contrast from the (111) and (111) planes are visible under (g= 220 ) and invisible under(g=220).
j 0.20 p m
Fig. 5.18- Plan view TEM o f a sample prepared from wafer B4. a) A pair o f threading dislocations and b) an elongated SFP are observed.
86 New etchant
Fig. 5.19 (a-b) are X T E M micrographs of sample prepared from wafer B4. Fig. 5.19 a) and b) are respectively a bright field and dark field micrographs of the silicon/BOX interface of a sample prepared from wafer B4. Fro m Fig. 5.19, it is observed that a small stacking fault complex is localized in a region close to the Si/Si02 interface. D u e to poor
statistics a meaningful estimation of the small S F complex density was not possible, however the height of this particular defect was estimated to be of 100 A. Furthermore, X T E M analyses have also shown the presence of these defects in the silicon substrate below the B O X layer ( see Fig. 5.20). The presence of small SF complexes far from the substrate/BOX interface suggests that some of them are stacking fault tetrahedra (SFT).
P V T E M analyses of a similar sample prepared from wafer B4, enabled us to determine the density of these defects ( SFP/SFT) as 2.0xl07 cm'2. A distribution of sizes is apparent with a mean length of 0.1 p m ( see Fig. 5.21).
87 New etchant
88 New etchant
Fig.
5. 19
- XT
EM
mic
rogr
aphs
from
a
sam
ple
prep
ared
from
wa
fer
B2
unde
r a)
brig
ht fi
eld
and
b) da
rk f
ield
Figure- 5.20- Bright field XTEM micrograph from a sample prepared from wafer B4 showing the presence o f a SF complex below the BOX layer
89 New etchant
Table 5.1 summarises the crystallographic defect densities observed in standard S I M O X wafers produced by IBIS and L E T I during the period 1989 to 1992. The dislocation densities were estimated using the n e w chemical etchant whilst the SFP/SFT densities were determined by P V T E M analyses. In general w e find that in single implanted materials the threading dislocation density is lower for L E T I than for IBIS samples being ~ 1 OxlO5 cm'2 and 1 OxlO6 cm'2 respectively. O n the other hand, IBIS material contains the lower SFP/SFT densities.
90 New etchant
S I M O Xwafer
Supplier Implantationmethod
Dislocationdensity(cm'2)
SFP/SFTdensity(cm'2)
A l IBIS single l.OxlO6 1.0x10s
A 2 ti it 8.0xl06 5.0xl05
A 3 ii ii 50.0 /6.0 xlO5 ❖ ❖
A 4 M multiple 1.0x10s t'.oV,©t—4
B 1 L E T I multiple l.OxlO3 < l.OxlO5
B 2 II single 5.0xl04 l.OxlO6
B 3 II n 5.0xl04 l.OxlO7
B3(2) II ii 5.0xl04 l.OxlO7
B 4 II ii 5.0xl05 2.0xl07
B4(2) II n 5.0x10s 2.0X107
B 5 II ii 8.0xl05 3.0xl07
B6* II n 1.0x10s **
Table 5.1 - Threading dislocation and SFP/SFT densities in standard SIMOX
* : The silicon overlayer thickness of wafer B 6 was 500 ±50 A. **: SFP/SFT density not measured.
91 New etchant
C h a p t e r S i x : R e s u l t s I
Sacrificially oxidised SIMOX
6.1 Introduction
In this chapter w e report the use of the n e w etchant to study the nature and density of defects present in thin S I M O X materials produced by sacrificial oxidation. Furthermore, data from plan view T E M and R B S analyses are also reported to provide complementary characterizations.
This chapter is divided in two sections. In the first one (6.2) w e characterize the thin silicon overlayer in terms of the nature and density of crystallographic defects produced during sacrificial oxidation of thick S I M O X materials whilst in section 6.3 w e describe an experiment to control the density of O IS F by the implantation of B F +2, Ge* or Si+ ions prior to the oxidation. The aim of this latter experiment is to annihilate the nucleation sites for O I S F by selectively amorphising the region near the Si/Si02 interface, leaving the near surface region relatively free of damage in order for it to be used as a seed for a solid phase epitaxial (SPE) regrowth during a subsequent thermal treatment.
92 Results I
6.2 Defects in sacrificiallv oxidised SIMOX.
6.2.1 Crvstalloeravhic defects in standard SIMOX.
In section 5.3 w e have shown that the main defects present in the silicon overlayer of thick S I M O X substrates are threading dislocations and small SF complexes (SFP/SFT). W e have also shown that S I M O X wafers showing a uniform light interference colour also showed a uniform distribution of defects. Furthermore, w e observed that although the nominal implantation and annealing conditions were the same, samples implanted in different batches contained different defect densities. Indeed, the range of defect densities observed in wafers from different batches varied from 103 cm'2 to 8x106 cm'2 for threading dislocations and from 105 cm'2 to 3.0xl07cm'2 for SF complexes (see table 5.1).
6.2.2 Crvstalloeravhic defects in sacrificiallv oxidised SIMOX
Selected samples ( 1 c m 2 ) cut from different S I M O X wafers ( namely B2,B3, B 4 and B5, see table 5.1) were sacrificially oxidised following the procedure described in section 4.1.5. The thicknesses of the silicon overlayer of the sacrificially oxidised S I M O X samples were measured by R B S analyses. Fig. 6.1 shows 1.5 M e V He* R B S spectra from samples (cut from wafer B2) submitted to dry oxidations at 1100°C for different times. The R B S analyses were carried out after removing the sacrificial oxide using a H F ( 4 0 % ) solution. F r o m Fig. 6.1, it is observed that the thickness of the silicon overlayer decreases according to the oxidation time. The estimation of the thicknesses of the silicon overlayers was carried out by simulation using the X R B S S I M programme[128]. The simulated data
93 Results I
Counts
showed that the thicknesses of the silicon overlayer (T=2500 A ) were reduced to 1000 A
and 500 A after oxidation times of 4 hrs. and 8 hrs., respectively. The thicknesses of the sacrificial oxide layers were previously estimated by measuring the step height of an oxide win do w using a Rank Taylor Tallystep stylus instrument ( see section 4.1.5 ).The R B S data together with the Tallystep trace analysis showed that the ratio between the thickness of the oxide and the consumed silicon was approximately 2.3 which is consistent with published values[60].
non oxidiseda a a a 1100°C, 4 hrs o o o o 1100°C, 8 hrs
C hannel N um ber
Fig. 6.1- 1.5 MeV He+ RBS spectra from wafer B2 and samples sacrificially oxidised at 1100 °C for 4 hrs, and 8 hrs.
94 Results 1
Oxid
e th
ickn
ess
(A)
Fig. 6.2 shows the dependence of the oxide thickness upon oxidation time for S I M O X samples oxidised in dry oxygen at different temperatures. For comparison w e have also plotted the theoretical values obtained from the Deal-Grove model [131]. It is observed that the oxidation rate of the silicon overlayer in our S I M O X materials is similar to that predicted for bulk silicon, and thus, suggesting that the existence of the buried oxide layer ( B O X ) does not affect the oxidation rate.
Oxidation time (hrs.)
Fig. 6.2- Dependence o f the oxide thickness upon time for SIMOX samples oxidised in dry 0 2 at different temperatures.
Fig. 6.3 shows the dependence of the etch pit density upon thickness of silicon removed for a sample cut from wafer B 2 and oxidised at 900°C under dry oxidation conditions for 2 hours. The thickness of silicon consumed was 200 A. As a reference, data from a non oxidised sample cut from wafer B 2 is also included.The data presented in Fig. 6.3
95 Results I
was collected by estimating the etch pit density in a m inimum of ten ( S E M ) micrographs for each point ( see section 4.2.1) and shows the depth distribution of the crystallographic defects in the oxidised and non oxidised material. F ro m Fig. 6.3 two different regions are observed. U p to 1500 A of silicon removed, the etch pit density in both oxidised and non oxidised samples is found to be the same (5.0 ±1x104 cm'2). However, the etch pit density for the oxidised sample increases by a factor of six close to the back of the silicon overlayer (at an estimated depth of 1700 A ), suggesting an increase in the defect density at approximately 800 A from the top Si/Si02 interface.
800 1200 1600 2000 Thickness of Si removed (A)
Fig. 6.3- Etch pit density upon thickness o f silicon removed from a SIMOX substrate (B2) oxidised at 900°C fo r 2 hrs. Data from wafer B2 before oxidation is included as a reference.
Fig. 6.4 a-b-c) show plan view T E M micrographs of samples cut from wafer B 2 oxidised at 900°C for 2, 10 and 24 hours respectively. In Fig. 6.4 a) w e can observe a well defined oxidation induced stacking fault (OISF) bounded by a dislocation loop (arrowed). The
96 Results I
density and mean size of the O IS F have been determined from a set of micrographs recorded using the two b e a m - condition (g= 220) at a magnification of 30 thousand times.The O IS F density estimated in this sample is 5.0xl0'cm‘2 and the mean size of the OISF is 0.2 pm. Furthermore, detailed analysis by plan view T E M has shown that the density of threading dislocations is not modified during a 2-hour dry oxidation at 900°C. This result together with data presented in Fig 6 3 suggests that for short oxidation times ( < 2 hours), the O IS F are formed in the silicon overlayer close to the Si overlayer/BOX interface. The micrographs shown in Fig. 6.4 b and c) show that the O IS F grow larger during longer oxidation times at this temperature. In Fig. 6.4 b, (10 hours), it is observed that the O IS F grow and eventually reach the silicon surface (arrowed) which w e propose is by capturing interstitials emitted during oxidation [132], U p o n a further increase in the oxidation time ( 24 hours) the O IS F grows laterally, being bounded by two partial dislocations to a length of approximately 2 pm.(see fig 6.4 c).
0.2 ( m
Fig. 6.4 (a) continued
97 Results I
b)
Fig. 6.4 - Plan view TEM o f samples prepared from wafer B2 after dry oxidation at 900°C for a) 2 hrs. b) 10 hrs and c) 24 hrs
Fig. 6.5 (a-b) are S E M micrographs of OISF in samples prepared from wafer B 2 and B 3 submitted to 900°C (10 hrs) and 1100°C (4 hrs.) respectively. The OISF have been delineated by etching the samples for 1 min. using the new etchant.
98 Results I
4HM 20KV 01 025 S
Fig. 6.5- SEM micrographs o f OISF which are delineated by the new etchant a) wafer B2 after dry oxidations at 900 °C fo r 10 hours and b) wafer B3 after oxidation at 1100°C for 4 hours.
It is observed that the mean O I S F length is strongly dependent upon the oxidation temperature having an average length of 1.0 p m ( 900 °C ,10 hrs) and 35 p m (1100°C, 4
99 Results I
hrs) in these two representative samples. Fig. 6.6 shows the dependence of the mean length of the O IS F upon the oxide thickness (i.e.oxidation time, see Fig 6.2) for samples prepared from wafer B2. It is observed that the OISF length increases linearly with increasing oxide thickness for different temperatures. Similar results were obtained from samples prepared from other wafers . A complete list showing the length and densities of O I S F of all the sacrificial oxidised S I M O X samples is presented in Appendix E.
Oxide thickness (A)
Fig. 6.6 Dependence o f the mean length o f OISF upon oxide thickness( wafer B2).
Fig. 6.7 shows the dependence of the OISF density upon the thermal oxide thickness for wafers Bl, B 2 and B3. From Fig. 6.7, it is observed that before the oxidation treatment,
100 Results I
the OIS F density measured was below the sensitivity limit of the optical microscope ( typically less than 102 cm"2). However, after growing a thin oxide ( say 500 A) all samples cut from the same wafer, presented a constant OISF density. Furthermore, it is also observed that the OISF density is insensitive to temperature and to oxide thickness (i.e. oxidation time).
The areal density of OISF along the [110] diameter of wafer B 3 was also quantified. Fig. 6.8 shows the O IS F distribution in a sample cut along the diameter of wafer B3 and oxidised at 1000°C for 16 hours. It is observed that the O IS F density is constant within the mean standard deviation which is typically ± 50%, except at a distance of less than 5 m m from the [110] flat edge. The data points do appear to show a periodic variation which m a y be reflecting a lateral non uniformity. However, as the variation is within the estimated uncertainty further measurements would be required to establish if this variation is real or an experimental artifact. For comparison w e have also plotted the distribution of O I S F along the diameter of a wafer sacrificially oxidised in the clean room facility at the L E T I laboratory ( wafer B6). It is observed that the OISF density is very similar to the values (1.5 ±0.5 x 104 c m ’2) determined from the oxidised sample cut from wafer B3 (see Fig. 6.8), which w e present as evidence for the growth of the O IS F not being a feature of the particular equipment or laboratory.
101 Results I
- tCH
a
S iOQ
h*H
l i ml1*1
K > »Hi
K
o yo u o o o o o oO LO O t-cn cj) t- i—
m
CQ
' M i : i i
ooo
OOOC O
feCTO(f)C D
O CIoo O C M _ c
CD■gxO
ooo
o IT )o o o CMo o
Q uid ) A)|SU8p J S IO
1?=*§
t:es<Nas
O1Nr*-
<3§
1>•*
1fefe3*c<s:ij
3I§§■
l yk, <5>«; *-ife C
53CUOIs8-aQ %7 S * o
o
«, Cn
102 Results I
(sji) 91. ‘o oooOea
•
L - U I O ) A t j s u e p J S I O
103 Results /
Fig.
6.8-
Dens
ity
of OI
SF
along
a
diam
eter
of
wafer
B
3. Da
ta fro
m wa
fer
B6 is
also
pre
sent
ed f
or
com
pari
son.
W e conclude from the results presented in Fig. 6.7 and 6.8 that the nucleation of OISF in our sacrificially oxidised S I M O X materials is controlled by the presence of nucleation sites prior to oxidation. It is important to emphasise that all of the S I M O X substrates chosen for the oxidation experiments were produced in the same laboratory ( LETI) and that the main difference between the substrates was the density of crystallographic defects.These results, then, suggest that crystallographic defects ( SF complexes, threading dislocations) present in the silicon overlayer before oxidation are the main nucleation sites for the OISF. However, w e conclude that the nucleation sites are not threading dislocations since specimens B3/22 and B5/23 (see table 6.1) had different threading dislocation densities, yet, the density of the OISF is very similar. In table 6.1 w e also list the SFP/SFT densities observed in different samples before oxidation. Although the experimental data is limited, there is a trend for the density of the O IS F to be lower in samples with a higher density of SFT/SFP before oxidation. This behaviour seems to indicate that the small SF complexes (SFP/SFT) act as effective sinks for silicon interstitials that are emitted during the thermal oxidation[133]. However, it is important to point out that this trend was not observed for a multiple implanted S I M O X sample ( Bl/19, see table 6.1) in which no SFP/SFT were resolved under plan view T E M (e.g < l.OxlO5 cm'2) before oxidation and the O I S F density was very low (5.0xl02 cm'2). Therefore, it is not clear if the SF complexes act as sinks for self interstitials or are the effective nucleation sites for OISF. In order to test whether the SFP ( or SFT) act as effective sinks for point defects, samples cut from wafer B 2 were annealed at high temperatures (1150°C) for a prolonged time (24 hours) in a N 2 ambient. B y so doing, vacancies generated during the nitridation process[133] could promote the growth or shrinkage of the SF complexes.
104 Results I
wfl<L>rtIrtCOl-Ho
S' aS3 OJrt <x3g j l^ ^ ♦r*H <£j j Irt o Cfl J \fl §•S'S<3 rtJ ‘g" sg>€jS x>1•* 1 'SH £ art v-/
Q
rtI<Dsfl0• r>»j131 B
XO £§ i*co §
©CO
X©
X©»ri
<DI 1
CT<NCT(2Q
X©<N
X©
X©
<NCN|CO(rt
£
Xo<N
X©
rcfC4(rt
x©oo
©<—< X oCO
Moo
COCTi©ea
&in
XoV
X©
< I—Iw'a. * *—<itiI
ON
PQ
I.1Ii4
05
.11
2l"Si
I9k<fl>*-<fl*S*$I5p
«05iH
1
.5s>S3*8■C
I
i
*«5"flflK
105 Results I
Fig. 6.9 shows a plan view T E M micrograph of an extended defect observed in a sample cut from wafer B 2 and submitted to a thermal annealing at 1150°C for 24 hours in a N 2 ambient. The contrast of this defect is very similar to that observed in Fig. 5.18 b, suggesting that it is an elongated S FP (1 =0.5 pm). Fig. 6.10 (a-b) are plan view T E M micrographs in a) dark and b) bright field contrast of another region of the same sample In Fig. 6 a) and b) an extended defect ( 1 =1.0 pm) is clearly observed. The extended defect was again identified as being an S FP since the two elongated faulted planes are orientated along the [110] direction and the contrast from the two small planes are clearly observed under dark field contrast (see Fig 6.10 b).
Fig. 6.9- Plan view TEM micrograph o f a SIMOX sample submitted to a thermal annealing in a N 2 ambient (1150°C)
106 Results I
107 Results I
Fig.
6.10-
Plan
vie
w TEM
m
icro
grap
hs i
n a)
brig
ht a
nd
b) da
rk fie
ld co
ntra
st of
a lon
g SF
P ob
sen'e
d in
a sa
mple
(B3)
after
a
prol
onge
d an
neal
ing
at 11
50°C
fo
r 24
hour
s in
/V, a
mbi
ent
The increase in size of the SFP from 0.1 pm ( before the N2annealing, see Fig. 5.18 b) to
1.0 pm ( after the N2 annealing) suggests that SFP can grow by capturing vacancies and
therefore, can act as effective sinks for point defects. However, it is important to point
out that not all the SFP have grown to sizes as large as 1.0 pm. In fact, the density of
long SFP (1 = 1.0 pm) was estimated to be 105 cm'2.This result could suggest that not all
the SFP in SIMOX are intrinsic and therefore, it is probable that some of them can also
be eliminated during a thermal anneal in a nitrogen ambient.
To further investigate if single extrinsic SF can also be eliminated by the capture of
vacancies, we annealed ( 1150°C, 24 hours in N2 ) a sample which was previously
oxidised at 900°C for 24 hours ( cut from wafer B2) , resulting in an OISF density of
3.0xl05 cm’2. Fig. 6.11 shows the lateral variation of the OISF density in the silicon
overlayer along a 30 mm wide sample (cut from wafer B2 and oxidised at 900°C for 24
hours), before and after the annealing in the N2 ambient. The data was obtained by etching
(SEM) adjacent areas of the same sample before and after the high temperature annealing
in the N2 ambient. From Fig. 6.11, it is observed that after the annealing, the OISF
density is reduced from 3.0 ±1x105 cm'2 to 1.0±0.3xl04 cm'2.
108 Results I
Etch
pit
de
nsity
( c
m'
10
1 0 '
1 0 ‘
1 0 k
Before N annealing
* # « I * i • • • • • ♦ . i
I tI \ T / I O
After N annealing
A
- r T ' S - JT i
0 1 0 2 0
D i s t a n c e f r o m t h e [ 1 1 0 ] e d g e o f w a f e r ( m m )3 0
Fig. 6.11 - OISF densities in a sample ( cut from wafer B2) which was thermally oxidised at 900°C for 24 hours and then further annealed in a nitrogen ambient at 1150°C for 24 hours. The OISF density before annealing is also plotted for comparison.
109 Results I
6.3 S a c r ific ia l o x id a tio n o f io n im p la n te d S I M O X m ateria ls.
In section 6.2, we have shown that the OISF density in sacrificially oxidised SIMOX
materials depends upon the presence of SFP/SFT at the back of the silicon overlayer. A
method to determine the influence of these defects on the formation of OISF is to
annihilate the SFP/SFT by selectively amorphising the silicon in the vicinity of the Si/Si02
interface by ion implantation. Subsequently, the amorphous region can be (thermally)
epitaxially regrown using the top of the silicon overlayer as a seed.
Fig. 6.12 (a - c) show 1.5 MeV He * RB S/channelling spectra from SIMOX samples
implanted with different doses of a) BF2+ , b) Ge* and Si*. The samples implanted with
BF*2 were cut from wafer B1 while samples implanted with Ge* and Sf were cut from
wafer B3(2) ( see table 4.3). From Fig. 6.12 (a-c) it is observed that for doses of
8.0xl014 BFycm'2, 1.7xl014Ge*/cm'2 and l.OxlO15 Si+/cm'2, the channelled yield between
ch. 220-260 is the same as the non channelled (random) spectra. Therefore, we conclude
that the threshold dose for the amorphisation is approximately 8.0xl014 BF'/cm'2,
1.7xl014 Ge*/cm'2 and l.OxlO13 Si*/cm‘2 for the BF*2, Ge* and Si* implantations
respectively. We also conclude that the amorphous region is confined to the back of the
silicon overlayer and that the top silicon layer remains relatively free of damage.
110 Results I
80 135 190
Channel Number
245 300
1768
1326
8 884
442
112
Random (non-implanted)
Channelled (non-implanted) Channelled: l x l 0 15Si+ cm "2
o - o - o o Channelled: 2.5x 1 0 15 S iA nf2
S i surface
174
Channel Number236 298
Fig. 6.12 (a-b) continued.
I l l Results I
Random (non-implanted)
Channel Number
Fig. 6.12- 1.5 MeV He+ RBS/channelling spectra from SIMOX samples implanted with different doses o f a) B F f b) Si+ and c) Ge+ ions
Fig. 6.13 (a-b) are XTEM micrographs of samples ( cut from wafer B3(2)) implanted
with cj)= l.OxlO15 Si+ cm'2 under a) bright and b) dark field contrast. From the dark field
micrograph (Fig. 6.13 b) we can see that an amorphous region has been formed near the
back of the silicon overlayer confirming the results obtained by RBS analyses ( see Fig.
6.12 b ). It is also observed that the silicon surface, although defective, remains
crystalline.
112 Results I
Fig.
6.13-
XTE
M mi
crog
raph
s of
a sa
mple
( cut
from
wafer
B3(
2))
impl
anted
wit
h (f>=
1.0x
1015
St cm
2 (19
0 kev
) un
der
a)
brig
ht a
nd
b) da
rk fie
ld co
ntra
st.
The silicon overlayer was regrown during thermal treatments in Ar ( 700°C, 30 min.) and
in dry oxygen (1100°C, 10 min.) ambients. Fig. 6.14 is a XTEM micrograph of a sample
(wafer B3(2)) implanted with a dose of <}>=5.0xl014 BF+2cm'2 and thermally treated at
1100°C for 10 min. It is observed that the silicon overlayer is completely regrown after
the thermal treatment. Similar results were obtained for the thermal treatments in Ar
ambients (700° C, 30 min ).
ermal oxide
Si overlayer
BOX layer
Si substrate
Fig. 6.14- XTEM micrograph o f SIMOX B3(2) implanted with 5.0x1014 BF2cm 2 and submitted to a thermal treatment ( 1100°C, 10 min., dry O )
114 Results I
Analysis by plan view TEM of samples implanted with BFfe Si+ and Ge+ ions and
thermally treated at 1100 °C for 10 min in dry 0 2 were carried out in order to determine
the quality of the regrown silicon overlayer. Fig. 6 15 is a plan view TEM micrograph of a
sample (wafer B3(2)) implanted with <]>= 1.0xl01? BF*2 cm'2 and thermally treated at
1100°C for 10 min. in dry 0 2. It is observed that a high density of dislocation loops is
formed ( 2.0x10* cm'2) during the regrowth process. Similar results were obtained for
samples implanted with Si* with the threshold dose for amorphisation ( <J> = l.OxlO15 Sf
cm'2) and thermal treated under the same conditions (see Fig. 6.16).
0 6 um
Fig. 6.15- Plan view TEM micrograph from a sample (wafer B3(2)) implanted with 1.0x1015 B E cm'2 and thermally treated at 1100°C for 10 min. in dry 0 T The DL density estimated from PV TEM observations was 2.0x1 (fcm'2.
115 Results I
Fig. 6.16- Plan view' TEM micrograph from a sample (wafer B3(2)) implanted with l.OxlO15 S t cm 2 and thermally treated at 1100°C for 10 min. in dry 0 2 . The DL density estimated from PV TEM obser\>ations was 1. Oxl 0scm 2.
On the other hand, no dislocation loops nor SF complexes were observed by plan view
TEM ( < 10? cm'2) of samples implanted with Ge+ ions with the threshold dose for
amorphisation (<}) = 1.7xl014Ge* cm'2) and submitted to a thermal treatment at 1100°C for
10 min in dry 0 2. However, upon further increasing the Ge+ dose (<j) = 2.5xl014 Ge+
cm'2), the dislocation loop density increases to a value of 7.0xl06 cm'2 (see Fig. 6 17).
116 Results 1
Fig. 6.17 - a) bright and b) dark field plan view TEM micrographs o f a sample (cut from wafer B3(2)) implanted with 2.5x1014 GeVcm'2 (1100°C, 10 min. dry O f A DL density o f 7. Oxl O6 cm 2 was observed
117 Results I
Fig. 6.18 is a SEM micrograph of a sample (cut from wafer B3(2)) implanted with Ge+ (<{>
= 1.7xl014 Ge+ cm'2) and oxidised at 1100°C for 4 hours in dry 0 2 after defect etching.
The mean OISF density estimated from a set of SEM micrographs was 5 OxlO2 cm'2
Fig. 6.18 - SEM micrograph o f sample (wafer B3(2)) implanted with 1.7x1 O'4 Ge+ cm 2 and thermally treated at 1100°C for 4 hours in dry oxygen. An OISF density o f 5. OxlO2 cm 2 was measured after extensive SEM analysis.
On the other hand, samples implanted with the threshold dose for amorphisation with
BF2 ions showed a high density of OISF after an oxidation at 1100°C for 4 hours in dry
0 2 (see Fig. 6.19) The OISF density estimated by plan view TEM analysis was 1 OxlO7
cm'2. Chemical etching plus SEM analysis of an adjacent area of the same sample showed
that a network of SF existed, which we suggest are formed due to the interaction of
OISF lying in different {111} planes (see Fig. 6.20).
118 Results 1
Fig. 6.19 - Plan view TEM micrograph o f a sample ( prepared from wafer B3(2)) implanted with 0= l.OxlO15 B E 2 cm 2 and oxidised at 1100°C, 4 hours in a dry oxygen ambient. OISF are clearly identified by the characteristic fringe contrast in the plane o f the fault and also by the partial dislocations that bound the fault.
Fig. 6.20- SEM micrograph o f a sample (prepared from wafer B3(2)) implanted with 0=1. Oxl O'5 B E 2cm 2 and oxidised at 1100°C for 4 hours in dry oxygen. The sample was etched using the new etchant for 1 min.
119 Results I
Fig. 6.21 (a-c) show the dependence of the OISF density upon the ion dose after oxidising
the samples at 1100°C for 4 hours in dry oxygen.From Fig. 6.21 a) it is observed that for
doses close to the threshold dose for Ge* the OISF density decreases to approximately
S.OxlO2 cm'2 indicating that a very low density of nucleation sites are present in the
silicon overlayer before the oxidation. However, for higher doses ( <j)= 2.5xl014 Ge* cm'2)
the OISF density increases significantly due to the increase in dislocation loop density
during the regrowth process (see Fig. 6.17). We conclude that the effective annihilation of
SF complexes gives rise to a low OISF density,thus, suggesting that they act as effective
nucleation sites for OISF.
In Fig. 6.21 b) we observe that although a decrease in the OISF density is not obtained,
an interesting trend is apparent. In the high dose regimes ( O > 5.0x1014 BF*2 cm'2) the
OISF density is strongly dependent upon the BF*2 dose. This is due to a high density of
dislocation loops (2.0xl08 cm'2) that are formed during the regrowth process (see Fig.
6.14). We believe that the dislocation loops act as the main nucleation sites for OISF .
For intermediate doses (l.OxlO14 BF*2 cm'2< ft) < 5.0xl014 BF*2 cm'2) , the nucleation
process of OISF is more complex since the SFT/SFP are not completely removed and
therefore, they coexist with the dislocation loops produced during the regrowth process.
We, therefore, conclude that in this case, the OISF density is primarily dependent on the
SFP/SFT density found in the original SIMOX (see table 5.1) and only partially dependent
upon the implanted dose (see Fig. 6.21b ). For lower doses (<j) < 1014 BF*2 cm'2) no
significant increase in the OISF density has been observed indicating that crystallographic
defects present before the amorphisation process are the main nucleation sites for OISF.
In Fig. 6.21 c) it is observed that for sub-threshold doses (<D < l.OxlO15 Si* cm'2) the OISF
density is independent of the Si* dose.
120 Results I
eo
LU
XCDC/3O
*0■00
CLECD
(3
o cvO
( j-1 0 0 ) A i j s u e p - j s i O
121 Results I
Fig.
6.21
a) OI
SF
dens
ity
agai
nst
Ge+
dose
after
oxida
tion
at 11
00°C
for
24
hour
s in
dry
oxyg
en.
( z _uio ) A ; ] S U 0 P - i s i o
122 Results J
Fig.
6.21
b) OI
SF
dens
ity
agai
nst
RE
2 do
se aft
er o
xidati
on
at 11
00°C
for
4 ho
urs
in dry
ox
ygen
.
( j - i u o . ) A i j s u e p J S I O
123 Results 1
Fig.
6.21
c) - O
ISF
dens
ity
agai
nst
St do
se aft
er o
xidat
ion
at 11
00°C
fo
r 4
hour
s in
dry
oxyg
en.
However, for doses equal or higher than l.OxlO15 Si+ cm'2 , an increase in the OISF
density is observed. This increase, as in the case of Ge+ and BF+2 implantations, is due
to the nucleation of the dislocation loops during the solid phase epitaxial regrowth
process ( see Fig. 6.15).
In order to determine the effects of the regrowth ambient on the formation of secondary
defects, short thermal annealings ( 1100°C , 10 min.) under N2 ambients were used to
regrow the amorphous regions of the BF+2, Si+ and Ge+ implanted SIMOX samples. Once
more, low DL densities were determined for samples implanted with <j) = 1.7xl014 Ge+
cm'2(p < 10s cm'2) while a considerably higher density of dislocation loops ( ~ 10s cm'2)
were observed for the BF+2 implanted samples, indicating that, for our processing
conditions, the formation of DL seems not to depend upon the regrowth ambient. The
reason why the amorphisation with Ge+ ions gives better results is not well known,
however recent work by Laanab et al [134] have shown that the density of secondary
defects ( known as 'end of range' defects [135]) formed during the solid phase epitaxial
(SPE) regrowth of amorphous silicon depends upon the number of excess interstitials at
the amorphous/crystalline interface. Moreover, Laanab[136] have shown that Ge+
implants ( compared to Si+ implants) give rise to a lower number of excess interstials, and
perhaps, explains the low DL densities observed in our Ge+ Implanted samples.
Finally, an interesting result observed for samples implanted with Si+ ions and annealed at
1100°C for 10 min. in a N2 ambient is the evolution of the dislocation loops into paired
threading dislocations (arrowed in Fig.6.22 a-b). We presume that this process occurs by
the climb of vacancy type perfect dislocation loops due to the capture of vacancies that
are found in super saturation under those conditions[133]. This result is interesting since
it confirms the assumption, made separately by Stoemenos [48] and Venables [51], that
124 Results I
the formation of threading dislocations in SIMOX occurs through a climb process of
perfect dislocation loops.
Fig. 6.22 - Plan view TEM micrographs o f samples implanted with Si a) (ft - l.OxlO15 S t cm'2 and b) <f> = 2.5x 1015 S t cm 2 annealed at 1100°C for 10 min. in aN2 ambient.
125 Results 1
C h a p te r S e ven : R e su lts I I
Low energy oxygen implants
7.1 Introduction.
In this chapter, we report experimental studies of crystallographic defects in thin film
SIMOX substrates produced by low energy and low dose O' implantation, The new
defect etchant (chapter 5) was used to delineate defects and complementary data is
obtained by RBS, XTEM and PV TEM methods, Additionally, SIMS and double crystal
X-ray diffraction analyses were also used to further characterize a selected number of
samples.
This chapter contains two sections. In Section 7.2 we report experiments to characterize
thin film SIMOX wafers produced by Eaton Corporation while in Section 7.3 we
characterize crystallographic defects in thin film SIMOX substrates produced at the
University of Surrey.
126 Results II
7.2 Wafers prepared by Eaton Corporation.
7.2.1 General description.
The wafers prepared by Eaton Corporation were implanted with low energies (30 keV
or 40 keV) and low doses ( typically 2.2x1017 0 + cm'2) with the specific purpose of
producing a thin overlayer above a thin continuous buried oxide (BOX). The doses were
selected to be close to the critical dose Oac for the formation of a continuous BOX layer
according to the semi-empirical model proposed by Li et al[59] ( see section 2.1). Table
7.1 shows the main analytical methods used in this project to determine the
micro structure and the crystalline quality of the wafers prepared by Eaton Corporation.
Wafer Ionimplanted
Energy(keV)
Dose xlO17 (0+ cm'2)
Analytical Methods
El 320 +2 60 2.2 RB S/XTEM/P VTEMChemical etching + SEM/ Optical microscopy
E2 32o+2 60 2.2 RB S/XTEM/P V TEMChemical etching + SEM/Optical microscopy
E3 16Q+ 30 2.2 RBS/PV TEMChemical etching + SEM/Optical microscopy
E4 160+ 40 2.2 RBS/PV TEMChemical etching + SEM/Optical microscopy
E5 16q+ 40 2.4 RBS/PV TEMChemical etching + SEM/Optical microscopy
E6 ,6o + 40 2.4 RBS/ PV TEMChemical etching + SEM/Optical microscopy
E7 K>0 * 40 3 RBS/ PV TEMChemical etching + SEM/Optical microscopy
Table 7.1 - Description o f the analytical methods to determine the microstmcture and crystalline quality o f the wafers prepared by Eaton Corporation.
127 Results II
The microstructure of the SIMOX wafers was determined by RBS and XTEM analyses
while the nature and density of crystallographic defects in the silicon overlayer and in the
substrate below the BOX were determined by chemical defect etching (SEM/optical
microscopy) and PVTEM analyses.
7,2.2 Microstmcture
Fig. 7.1 shows a typical 1.0 MeV He+ RBS spectrum from wafer El. The two main
features of this spectrum are marked by the letters A and B. In region A, the reduced
backscattered yield of the He+ ions is due to the reduced density of silicon atoms in the
buried oxide layer whilst the increase in the backscattered yield, in region B, is due to the
overlapping of the oxygen and silicon signals. From Fig. 7.1, due to the abruptness of the
features A and B, we conclude that wafer El has a well defined buried oxide layer. In
order to determine if the signal from region A corresponds to stoichiometric Si02, we
have carried out simulations using the XRBSSIM programme[128]. The accurate fitting
to the experimental data of the simulated spectrum for a test structure containing a BOX
layer of 500 A confirms that a stoichiometric Si02 layer exists in the Eaton wafer.
Furthermore, the thicknesses of the silicon overlayer and buried oxide layer were also
determined by matching the experimental spectrum with the simulated one. RBS
analyses of other Eaton wafers were carried out and table 7.2 summarises the determined
thicknesses of the silicon overlayers and buried oxide layers.
128 Results II
B
C hann el N u m b e r
Fig 7.1 - Typical RBS spectrum from wafer El. A simulated spectrum is also shown.
SIMOXwafer
Ionimplanted
Energy(keV)
Dose (0+ cm'2)
Implantationtemperature(°C)
Sioverlayerthickness(A)
BOXthickness(A)
El 320 +, 60 2.2xlOn 530 500 ± 50 500± 30E2 H 60 2.2xlOn 530 530 ± 50 500 ± 30
E3 .6q + 30 2.2xl0n 530 550 ± 50 600 ± 30
E4 II 40 2.2xlOn 530 630 ± 50 620 ± 30
E5 11 40 2.4xl017 550 650 ± 50 600 ± 30E6 11 40 2.4xI017 600 600 ± 50 600 ± 30
E7 ti 40 3.0xl017 510 700 ± 50 680 ± 30
Table 7.2 - Si overlayer and BOX layer thicknesses estimated by using computer simulation o f the RBS spectra.
129 Results I 1
Fig. 7.2 (a-b) arc typical XTEM micrographs taken from a representative area of wafer
El showing respectively, bright and dark field images. From these and other
micrographs, it was found that the structure of wafer El consists of a single crystal silicon
overlayer of thickness 500 ±20 A, beneath which a thin continuous buried oxide layer
exists of thickness 500 ± 20 A. All of these thicknesses determinations agreed consistently
with the RBS data (table 7.2). No silicon islands [137] were found inside the BOX layer
after extensive observations by XTEM suggesting that their density is below the
practical sensitivity limit for XTEM analysis, which is typically 107cm'2. Additionally,
crystallographic defects were not observed either in the silicon overlayer or in the
silicon substrate confirming that the crystallographic defect density in these layers also is
below the sensitivity limit for XTEM analysis.
130 Results II
Fig.
7.2 - X
TEM
mic
rogr
aphs
of
a typ
ical
regio
n of
wafer
El
in a)
brig
ht a
nd
b) da
rk fi
eld
Results II
Chemical defect etching (optical microscopy and SEM) and PVTEM analyses have been
used to quantify the density and depth distribution of crystallographic defects in the thin
silicon overlayers of wafers prepared by Eaton.
Fig. 7.3 shows the dependence of the etch pit density against thickness of silicon
removed from a representative sample ( N° 6, Fig. 7.6) cleaved from wafer El. It is
observed that for half of the thickness removed the etch pit density reaches a saturation
value of 4.0 ±2 xlO4 cm'2. However, upon etching to greater depths (> 350 A) the etch
pit density increases to approximately l.OxlO7 cm'2. From Fig. 7.3 and results presented
in section 5.3 (see for example Fig. 5.16), we conclude that although the density of
threading dislocations is low (4.0x104 cm'2), a high density of extended but non threading
defects ( ~ l.OxlO7 cm'2) is present at the back of the silicon overlayer. Similar results
7.2.3 Defect distribution - depth
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0T h i c k n e s s o f S i r e m o v e d ( A )
Fig. 7.3 - Dependence o f the etch pit density upon thickness o f silicon removed from a representative area o f wafer El.
132 Results II
s m a l l S F P ( a r r o w e d ) a t t h e b a c k o f t h e s i l i c o n o v e r l a y e r ( s e e F i g . 7 . 4 ) . T h e m a x i m u m s i z e
o f t h e s e d e f e c t s w a s m e a s u r e d t o b e 3 0 0 A a n d t h e S F P d e n s i t y w a s « 8 . 0 x 1 0 6 c m ' 2 .
F u r t h e r m o r e , n o t h r e a d i n g d i s l o c a t i o n s w e r e o b s e r v e d b y p l a n v i e w T E M a n a l y s i s
confirming that the threading dislocation density in wafer El is lower than the
s e n s i t i v i t y l i m i t o f p l a n v i e w T E M ( t y p i c a l l y l e s s t h a n 1 0 ' c m ' 2) . T h e p l a n v i e w T E M
a n a l y s e s t o g e t h e r w i t h t h e r e s u l t s p r e s e n t e d i n F i g . 7 . 3 s u g g e s t t h a t a h i g h d e n s i t y o f S F P
a r e l o c a t e d c l o s e t o t h e s i l i c o n / B O X i n t e r f a c e . I t i s i m p o r t a n t t o p o i n t o u t t h a t t h e
p r e s e n c e o f S F c o m p l e x e s a t t h e b a c k o f t h e s i l i c o n o v e r l a y e r i s n o t s i m p l y r e l a t e d t o t h e
e n e r g y a n d d o s e o f t h e o x y g e n i m p l a n t a t i o n s i n c e t h e s e d e f e c t s w e r e a l s o o b s e r v e d i n
s a m p l e s i m p l a n t e d w i t h h i g h e n e r g y a n d h i g h d o s e s ( s e e s e c t i o n 5 . 3 ) . T h e f o r m a t i o n o f S F
c o m p l e x e s i n S I M O X i s d i s c u s s e d i n c h a p t e r 8 .
A nalyses by plan v iew TEM o f sam ples prepared from wafer E 1 indicates the presence o f
________iooo A
Fig. 7.4- Plan view TEM micrographs o f a sample prepared from wafer El showing two small SFP.
133 Results II
The lateral distribution of threading dislocations across the diameter of wafer El is
shown in Fig. 7.5. while in Fig. 7.6 we show a map of the density of etch pits for the
same wafer (the schematic showing the 48 areas analyzed by chemical etching is also
shown).
From Fig. 7.5 and 7.6, it is observed that the average defect density near the centre of
the wafer is 1.0xl04cm'2. However, small variations from 3.0xl03 cm"2 to 2.0xl04cm'2,
which are greater than the experimental uncertainty, were observed. It is speculated that
this non uniformity is associated with temperature inhomogeneities during implantation or
during the subsequent thermal annealing.
7.2.4 D efect distribution - lateral
134 Results II
EE .
X
o CO
( z.u i o ) A | | S U 0 p ? jd i p i g
135 Results II
Fig.
7.5
- Lat
eral
var
iatio
n of
the
etch
pit
dens
ity
along
the
[1
10]
diam
eter
of wa
fer
El.
136 Results II
Fig.
7.6
- W
hole
wafer
defe
ct etc
h pit
den
sity
map
(wafe
r El
). A
sche
matic
of
the
area
s an
alyz
ed
is als
o sh
mvn
.
Defect densities higher than 107 cm'2 were observed near the borders of the areas
masked from the oxygen ion beam ( see fig. 7.6). These highly defective areas extended
several millimetres inside the implanted area and are symmetrically distributed around the
wafer. Fig. 7.7 (a-c) show plan view TEM micrographs of samples prepared from the
highly defective areas. From Fig. 7.7 b) it is observed that a high density of SF complexes
are present in these regions. The density of these defects is very high and has been
estimated to be 5.OxlO9 cm'2. From Fig. 7.7 c) we can clearly see that the size of these
defects is not uniform, and they consist mainly of SFP (labelled X) and SFT ( labelled
Y). The distinction between the SFP and SFT was carried out by determing the direction
of projection of the faulted planes. The SF complexes having the projected edges
orientated along <110> directions were identified as being SFP whilst defects having the
projected edges orientated along <100> directions were identified as being SFT[130].
The origin and growth of the SFP/SFT in SIMOX are discussed in chapter 8.
137 Results II
/ e %c)
Fig. 7. 7- (a-c) Plan view TEM micrographs o f highly defective areas o f wafer El
The existence of high defect densities in areas near the non-implanted regions were
observed in other Eaton wafers. Fig 7.8 (a-b) show etch pit density maps close to the
non-implanted area for wafers E2 and E3. It is observ ed that a high defect density ( > 107
cm'2) is always observed close to the edge of the non-implanted area The data shown in
Fig. 7 8 suggests that the increased defect density is related to processing problems
during implantation or annealing of the 6 inch wafers A further discussion of the reasons
for this anomalous behaviour is given in chapter 8.
139 Results II
Non implanted area
/• # • • •3E4 1.5E4 2E4 1.5E4 2E4 1E4
• • • • •
T
6E3 1E4 7.5E3 ••
9E3•
1E4•
1.3E4 1E4 8E3 1E4 • •
1E4 1E4•5E4 2E4 1.5E4 1E4 1E4
• • • • •[> 1E7 8E6 1E5 6E4 1E4 1E4
. • • . •|> 1E7 > 1E7 3E5 8E4 2E4 2E4
1 cm
Non implanted area
/. . . . • •
7E5 3.5E5 4.4E5 2E5 1.7E5 1.5E52E6 . . . • •
1.5E6 5E5 4E5 2E5 2.5E51E7 *. 5E6
1.3E7 * • •. 8E6 1.5E6 5E5 4E5 2E5 ✓
>1E7 # ✓/. >1E7 2E5 /Vr
WE7 ./>1E7# 3E5 2E5 2.5E5 3E5 ✓✓/
>#1E7 >1E7 ✓sss*
1 cmFig. 7.8 Defect maps o f regions close to the non-implanted border fo r wafers a)E2
and b)E3.
140 Results II
At a distance of approximately > 1cm from the implanted/non implanted boundary, the
defect density is considerably lower for all of the Eaton wafers. Fig 7.9 (a- b) show
typical SEM micrographs of defect pits in wafer E7 etched using the new etchant for 35
sec. It is observed that although a relatively small amount of silicon has been removed
(250 A), well defined paired etch pits can be clearly seen, identified by the arrows in Fig
7.9 b.
a)
4 H M 2 0 K V 0 5 0 9 3 S
K T
<Fu 0O
v>
2HM 20KV 06 066 S
Fig. 7.9 a) and b) SEM micrographs o f an area close to the centre o f wafer E7 etched using the new etchant for 35 sec.
141 Results II
The etch pit density distribution along a 35 mm long sample cut from a region close to
the centre of wafer E7 is shown in Fig 7.10. It is observed that the defect density has a
mean value of 2.0x106 cm'2.
sQ
’(/)C0~o-t—;‘CL_Co•*—I1
LU
X ( m m )
Fig. 7.10- Lateral etch pit density distribution along a 35 mm sample cut from a region near the centre o f wafer E7.
The mean threading dislocation densities measured for all Eaton wafers are listed in
table 7.3 where it be can seen that the lowest defect densities were observed in wafers
implanted with the lowest doses and energies. Another interesting result is that the
defect density of wafers implanted with molecular oxygen (160 +2) (El and E2) were lower
(by more than one order of magnitude) than wafers implanted with atomic oxygen (160 +).
142 Results II
This result, although not conclusive (since further analysis of other wafers are needed to
confirm the trend), suggests at least that the defect density is not increased due to the use
of molecular oxygen and, therefore, opens a way to achieve low cost thin film SIMOX
structures. Moreover, the results presented in table 7.3 suggest that good quality thin film
SIMOX structures can be obtained using existing equipment (NV-10-80 implanter) and
hence development costs for a new generation of low voltage machines will be avoided,
providing a route to achieve thin film SIMOX substrates without incurring a cost penalty.
SIMOXwafer
Ionimplanted
Energy(keV)
Dose(CTcm-2)
Implantationtemperature(°C)
Si overlayer thickness(A)
Dislocationdensity
El 320 +, 60 2.2x1017 530 500 ± 50 l.OxlO4E2 II 11 fl It 530 ±50 l.OxlO4E3 I60 + 30 2.2xl017 ft 550 ±50 5.OxlO5E4 If 40 It If 630 ±50 5.0x105E5 It II 2.4xl017 550 650 ± 50 2.OxlO5E6 ft 11 tl 600 600 ± 50 5.OxlO6|E7 It 11 3.OxlO17 510 700 ± 50 2.OxlO6
Table 7.3 - Mean threading dislocation densities found in Eaton thin film SIMOX wafers.
143 Results II
T h e c r y s t a l l i n e q u a l i t y o f t h e s i l i c o n s u b s t r a t e b e l o w t h e B O X l a y e r w a s a l s o a s s e s s e d
u s i n g t h e n e w d e f e c t e t c h a n t . B e f o r e t h e d e f e c t e t c h a n t w a s a p p l i e d t h e s i l i c o n o v e r l a y e r
o f t h e S I M O X s a m p l e w a s s t r i p p e d o f f u s i n g a K O H e t c h a n d t h e b u r i e d o x i d e l a y e r w a s
r e m o v e d b y H F ( 4 0 % ) e t c h i n g ( s e e s e c t i o n 4 . 1 ) . I t i s i m p o r t a n t t o p o i n t o u t t h a t t h e
e x p o s e d s u r f a c e w a s f e a t u r e l e s s a n d w a s f r e e o f e t c h p i t s p r i o r t o t h e c h e m i c a l d e f e c t
e t c h i n g .
F i g . 7 . 1 1 s h o w s a t y p i c a l S E M m i c r o g r a p h o f t h e s i l i c o n s u b s t r a t e b e l o w t h e B O X l a y e r
o f w a f e r E l a f t e r d e f e c t e t c h i n g . T h e m a i n d e f e c t s o b s e r v e d i n t h i s a n d o t h e r
m i c r o g r a p h s w e r e s t a c k i n g f a u l t s ( S F ) h a v i n g a m e a n d e n s i t y o f 5 . 0 x 1 0 ' c m 2 ( l a b e l l e d
X ) . A n a l y s i s o f t h e s i l i c o n s u b s t r a t e b e l o w t h e B O X o f o t h e r E a t o n w a f e r s h a v e a l s o
s h o w n t h e p r e s e n c e o f S F w i t h a s i m i l a r d e n s i t y ( 5 . 0 x 1 0 ' c m ' 2) .
7.2.5 Crvstalloeravhic defects in the silicon substrate.
x
20KV 01 027Fig. 7.11 - SEM micrograph o f the substrate below the BOX layer o f wafer E l after chemical etching.
x
1 0 H M
144 Results 11
The crystalline quality of the non-implanted area of the Eaton wafers was also assessed
Analysis by SEM and optical microscopy (after chemical defect etching) of the
non-implanted areas of wafers El, E2 and E3 show the presence of stacking faults with a
mean density of 5.0x10' cm'2 The overall distribution of SF in the non implanted area is
uniform, although at a distance of approximately 20 pm away from the visible ion
implanted boundary ( 1IB), SF are distributed in bands separated by a denuded zone 20
pm wide (see Fig. 7 12 a-b) Analysis of other areas of wafers E1,E2 and E3 have shown
that the SF are essentially parallel to the visible IIB. We believe that these bands appear
due to a stress related process caused by the lateral pinning of the silicon overlayer It is
important to point out that no SF were observed inside the implanted area A discussion
about the stress related edge effects is given in section 8.3 1.
7.2.6 Defects associated with edge effects.
IIB
Implanted area
\y hands of SP
Fig. 7.12a- (continued).
145 Results II
I IB SB
Fig. 7.12 a) and b) SEM micrographs showing the SF band close to the ion implanted boundary o f wafer El observed after defect etching
band o f
146 Results II
7.3 W a fers p re p a re d a t th e U n iversity o f Surrey .
7.3.1 General description.
The wafers prepared at the University of Surrey were implanted with doses equal or
greater than the critical dose for the formation of a continuous buried oxide layer at
energies of 70 keV, 90 keV and 200 keV [59], Table 7.4 shows the main analytical
methods used to determine the microstructure and characterize the crystallographic
defects.
Wafer Energy(keV)
Dose(0+cm‘2xlO17
Implantationtemperature(°C)
Analytical methods
G1/G2/G3G5/G6
70 3.3 650/550/700700/600/500
RB S/channelingChemical etching (SEM/Optical microscopy) PVTEM
G4 70 3.3 700 RB S/channelingChemical etching (SEM/Optical microscopy) PVTEMDouble X ray diffraction
G7/G8/G9G10/G11G12/G13
90 5/5/55/5
3/10
700/650/600550/500700/550
RB S/channelingChemical etching (SEM/Optical microscopy) PVTEM
G14/G15G16/G17G18/G19G20/G21
200 4/55/66/77/5
700 RBS/channellingChemical etching (SEM/Optical microscopy) PVTEM
(100) Cz SiWacker
** ** ** SIMSChemical etching
* * : s a m p le n o n i m p la n t e d .
Table 7.4 - Description o f the analytic methods used to determine the crystalline quality o f the SIMOX wafers prepared at the University o f Surrey.
147 Results II
The structure of the wafers prepared at the University of Surrey were determined using
RB S/channeling analyses. Fig, 7.13 shows typical 1,5 MeV He4 RBS spectra from wafers
G12, G10 , G13 implanted at energies of 90 keV and doses of 3.0x1 O’7 O’ cm'2 , 5x1 O'7
0 + cm'2 and l.OxlO18 0* cm'2, respectively. From Fig. 7.13, it is observed that the
thickness of the BOX layer increases with increasing ion dose, The thicknesses of the Si
overlayer and BOX layer have been determined by curve fitting of computed spectra,
generated using XRBSSIM [128], to the experimental data.The thicknesses estimated by
RBS for the wafers prepared at the University of Surrey are summarised in table 7.5.
G10
++++++ G12
7.3.2 S IM O X Structures
Channel Number
Fig, 7.13- 1.5 MeV RBS spectra o f samples G12, G10 and G13 implanted at 90 keV
with doses o f 3.0x1017cm'7, 5,0xl0t7cm’2 and 1.0x10l* cm’2, respectively.
148 Results II
BOX
laye
r th
ickn
ess
(A)
± 60
A
oooooo00 * oo00
ooCO
oooo
00910091 15
00
1500
1500 oooo 2500 * 950
1000
1400
1500 * -X- *
Silic
onth
ickn
ess
(A)
± 60
A
ooCNooCN W oo<N
oocnoom
oomoocn
oomoom
oomoooo oovo •X-
oonrnroonrnr
oo<Nroocnnr * * *
Ann
eal
tim
e (h
rs.)
VO VO CN vo = CN
Ann
eal
ambi
ent
<5*£«qo+
CNOO'in+CNZ
CNov°O 'in §o'in+R
= =
Impl
ant
tem
pera
ture
(°C
)
oinvoO<n>n
OR
OOr-OOvo
OOinOo O<nVO
OOvoO<n<n
oo>nOR
oininoot>
OOf -
ooi>OOl>
oot->oot-
ool-~Oot-
Cur
rent
den
sity
(g
A/c
m'2) cri
inr-r-‘
CNvoinr--A
inr~-Kr—<cri cri
invqnfcnnj-
ot—H.
& a& B o + Q O
cncn cncri COcri cncn cncn cncn in >n >n «n in cn or—< nr in in VO vo r- r- n
Ene
rgy
(keV
)
Ot-- or-~ ot" ot~- or-- o oOv oOv
oOV oov oOv
oOv
oovooCN
ooCN
ooCN
ooCN
OoCN
ooCN
ooCN
ooCN
Sam
ple
r—tO CNo cnO nrO
«nO cS t -O OOo ovo
ot—Ho
T—1i-HO
CNt—1o
cnO
T f
oinO
VOr - lO
t''r—.o
OO !---1o
ovO
oCNO
*—HCNO
149 Results II
Tabl
e- 7.5
Thic
knes
s of
the S
i ov
erla
yer
and
SiO
2 lay
er.
I t i s f o u n d t h a t t h e s i l i c o n o v e r l a y e r a n d B O X t h i c k n e s s e s o f t h e s a m p l e s i m p l a n t e d a t 7 0
k e V w i t h t h e c r i t i c a l d o s e f o r t h e f o r m a t i o n o f a c o n t i n u o u s b u r i e d o x i d e l a y e r ( ( j ) =
3 . 3 x 1 O ' 7 O ' c m ' 2) a r e 1 2 0 0 ± 6 0 A a n d 8 0 0 ± 6 0 A , r e s p e c t i v e l y . S i m i l a r s i l i c o n o v e r l a y e r
a n d B O X t h i c k n e s s e s w e r e o b t a i n e d f o r s a m p l e s i m p l a n t e d u n d e r s i m i l a r c o n d i t i o n s b u t
w i t h d i f f e r e n t i m p l a n t a t i o n t e m p e r a t u r e s s h o w i n g a g o o d r e p r o d u c i b i l i t y i n t h e s a m p l e s
p r o d u c e d a t t h e U n i v e r s i t y o f S u r r e y . I t i s a l s o o b s e r v e d t h a t t h e t h i n n e s t f i l m ( 6 0 0 A )
w e r e f o r m e d i n s a m p l e s i m p l a n t e d w i t h a d o s e o f < })= 1 O x 1 0 1 8 O c m ' 2 a n d e n e r g y o f 9 0
k e V ( w a f e r G 1 3 , s e e F i g . 7 . 1 3 ) . O n t h e o t h e r h a n d , t h e t h i c k e s t s i l i c o n o v e r l a y e r s ( 4 3 0 0
A ) w e r e o b s e r v e d i n l o w d o s e s a m p l e s i m p l a n t e d w i t h 2 0 0 k e V .
7.3.3 Defect distribution - lateral.
C h e m i c a l d e f e c t e t c h i n g a s s o c i a t e d w i t h S E M o b s e r v a t i o n s w e r e u s e d t o d e t e r m i n e t h e
l a t e r a l d i s t r i b u t i o n o f t h e a r e a l d e n s i t y o f c r y s t a l l o g r a p h i c d e f e c t s w i t h i n a 3 0 m m c e n t r a l
i m p l a n t e d r e g i o n . F i g . 7 . 1 4 s h o w s a s c h e m a t i c o f t h e c e n t r a l i m p l a n t e d a r e a o f t h e 3 "
w a f e r s u s e d t o d e t e r m i n e t h e l a t e r a l v a r i a t i o n o f t h e c r y s t a l l o g r a p h i c d e f e c t s .
7.14- Schematic showing the central area o f the 3 inch SIMOX used to determine the lateral distribution o f the areal density.
150 Results II
Fig. 7.15 (a-b) show SEM micrographs from two different areas of wafer G1 ( E= 70
keV , 3.3xl017 cm'2,Ti = 650 °C), namely areas A and B (see Fig. 7.14), etched using the
new etchant for 1 min It is observed that the etch pit density is considerably different
in distinct areas of wafer Gl. Indeed, a difference of more than one order of magnitude
was observed from one side ( region A: 5. OxlO6 cm'2) to the other of the implanted area
( region B: 1.0x10* cm'2) of wafer Gl.
1 H M 2 Q K V
Oo 0
i ow v f.c
I ur c & v ..0 OD 0 A O C>C<
b0■o
©
c: <^ Cc a0
c<r
?To o
0 *S e ■ **>0a v c auO r U <£
U
O 0 v Q Cu
« Pow
u 8
. c ~ o c C t c7 $ / : o •
<5: o CM-
cO'
o
1HN 2 u h V io:Fi#. 7.75- SEM micrographs o f a) region A and b) region B o f wafer Gl etched using the new etchant for 1 min.
151 Results 11
G1 and G2. It is seen that for specimen G1 ( Ti =650 °C) a low etch pit density is only
observed on one side of the implanted area where the density is « 5.0x106 cm'2. Furthermore, it is observed that for wafer G2, the etch pit density is relatively uniform
across the implanted area. It is important to point out that both wafers G1 and G2 were
implanted under the same conditions except that the implantation temperature for wafer
G1 was 650°C whilst it was only 550 °C for wafer G2.
Fig. 7 .16 show s the distribution o f the etch pit density across the implanted area o f wafer
1 0 '
£ 8 ° 1 0
C/3C0
"O
f e o 7o
LU
1 0
, S f f * # * # • * *
f .......
0
■©
• ........... • T i = 5 5 0 ° CO ........... O T i = 6 5 0 ° C
G 1
’-o . QT-
8 1 2 1 6 2 0 2 4 2 8 3 2X ( m m )
Fig. 7,16 - Lateral distribution o f the etch pit density in wafers G1 and G2.
Plan view TEM analyses were carried out in order to determine the nature of the defects
present in wafer Gl. Fig. 7.17 is a typical plan view TEM micrograph of the silicon
overlayer of a sample taken from the highly defective area of wafer Gl. It is observed that
152 Results II
the most numerous crystallographic defects present in the silicon overlayer are threading
dislocations ( labelled X), with a density of 2.0x10s cm'2. Identification of the defects as
threading dislocations was possible due to the classical stress contrast associated with
inclined dislocations [127]
Fig. 7.17 - Plan view TEM micrograph from a sample cut from area A o f wafer Gl.
Fig. 7.18 shows the lateral distribution of the areal density of defects obtained by
chemical defect etching (SEM observations) across the implanted area of wafer Gl 8 (<h
= 6.0x1 O'7 CT cm'2, E= 200 keV, T; = 700°C) confirming that a high lateral non uniformity
is typical of the SIMOX substrates prepared at Surrey at high implantation temperatures.
153 Results II
Etch
pit
de
nsity
(
cm"2
)
1 06
f»
0 1 0 2 0 3 0X ( m m )
Fig. 7.IS - Lateral etch pit distribution across wafer G18 (<j>-6xl0'7 O' cm’1 , 200 keV).
From the results presented in Fig. 7.16 and 7.18, it is evident that the non uniformity of
the defect densities is more accentuated for the wafers implanted at temperatures greater
or equal to 650°C . From Figure 7.18, it is also apparent that the presence of areas, in
the same wafer, containing different defect densities is independent of the oxygen dose
and energy. We, therefore, speculate that this behaviour is due to thermal gradients across
the wafer during implantation which cause large thermal stresses. These thermal
inhomogeneities are accentuated because only a 6.45 cm'2 area is implanted causing a
large gradient of temperature across the wafer. Indeed, an approximate temperature
difference of 70° C between the centre and edge of the wafer was determined using a
pyrometer for samples implanted at 700°C[138,139]
In order to better understand the reasons for the non uniform nucleation of
crystallographic defects inside the implanted area, we annealed for shorter times ( 2
154 Results II
Etch
pi
t de
nsit
y (
hours, 1300 °C) a wafer implanted under the same conditions as wafer Gl, namely
wafer G3. Fig. 7.19 shows the lateral variation of the defects inside the implanted central
area of wafer G3 delineated by chemical etching plus SEM observations. Again a similar
pattern is evident, namely non uniformity of the areal defect density confirming that the
effect is independent of the post implantation annealing process.
10®
1 0 6
1 0 50 1 0 2 0 3 0
X ( m m )
Fig. 7.19 - Lateral etch pit density distribution across the implanted area o f wafer G3 , The data was obtained after etching the sample fo r 1 min. using the neiv etchant.
Fig. 7.20 a) and b) are SEM micrographs from two different regions of wafer G3. It is
observed that in the region of high defect density , etch pits are distributed in larger
clusters (7.20 a), whilst in the region of lower defect density the etch pits are single or
paired (7.20 b)
155 Results II
Fig. 7.20 - SEM micrographs o f samples from the a) high defect density region and b) low defect density region o f wafer G3 ( after I min. etching).
F r o m F i g 7 . 2 0 , i t i s c o n c l u d e d t h a t t h e p r e s e n c e o f r e g i o n s c o n t a i n i n g t h e h i g h d e f e c t
d e n s i t i e s o c c u r s d u e t o a h i g h e r d e g r e e o f o v e r l a p p i n g o f l a r g e d e f e c t c l u s t e r s . W e
p r o p o s e t h a t t h e s e c l u s t e r s o f d i s l o c a t i o n s a r e n u c l e a t e d i n t h e r e g i o n s o f h i g h e s t s t r e s s
( d i s c u s s e d f u r t h e r i n s e c t i o n 8 . 3 . 1 ) .
156 Results II
7.3.4 Evolution o f threadine dislocations during annealing.
The evolution of threading dislocations during the high temperature thermal annealing
process was assessed by plan view TEM analysis of a sample implanted with <j)=
6.0xl017 CT cm'2 (200 keV) and annealed for 2 hours at 1300°C , namely wafer G21.
Fig. 7.21 is a plan view TEM micrograph of the silicon overlayer of a sample taken from
the implanted area of wafer G21. It is observed that a high density of dislocations (
labelled D) are pinned to large SiOz precipitates ( labelled P) . Some of the dislocations,
however, are only partially pined (labelled T) to the SiO, precipitates and thread through
the silicon overlayer. The observation of dislocations totally and partially pinned to large
Si02 precipitates suggests that the former are the precursors of the threading
dislocations in low dose SIMOX samples. We propose that during the annealing these
dislocations evolve into threading dislocations by a non conservative climb process
during the anneal[117]. This assumption is backed by recent proposals made by Venables
et al [ 51] that the evolution of threading dislocations in high dose SIMOX materials
occur through a stress assisted climb process ( see section 2. 2.).
157 Results II
T P
2000 A
Fig. 7.21 - Plan view TEM micrograph from a sample cut from wafer G21. Dislocations pined ( labelled D) and partially pinned (labelled T) to SiO7 precipitates (labelled P) are observed
7.3.5 Defect density - dependence upon implant temperature (T)
Figure 7.22 a) shows the etch pit density dependence upon temperature for samples
implanted at 70 keV with a dose of 3.3x10' 7 O cm2 and Fig. 7.22 b) for samples
implanted at 90 keV with doses of 5.0x1017O*cm’2 From the figure it is observed that the
defect density decreases considerably for temperatures equal or higher than 650°C. This
data agrees closely with recent results presented by Nakashima and Izumi[140] which
showed a large decrease of the defect density with temperature. However, it is important
to point out that, in the samples implanted at the University of Surrey ,this decrease does
not occur uniformly across the implanted area, which we propose is due to machine
specific temperature gradients.
158 Results II
Etch
pit
de
nsit
y ( c
m
4 5 0 5 0 0 5 5 0 6 0 0 6 5 0 7 0 0 7 5 0I m p l a n t a t i o n t e m p e r a t u r e ( ° C )
I m p l a n t a t i o n t e m p e r a t u r e ( ° c )
Fig. 7.22- Dependence o f the defect density upon temperature for a) samples implantedat70keV, 3.3x1017O' cm 2and b) at 90keV, 5.0x1017O' cm 2.
T h e d e f e c t d e n s i t ie s p l o t t e d i n F i g . 7 . 2 2 a r e f r o n t t h e c e n t r e o f a p a r t i c u l a r w a f e r .
1 5 9 Results II
7.3.6 Edge effects - stress related
Fig. 7.23 is a typical SEM micrograph from the non implanted area of wafer G1 (
previously etched for 1 min) at a distance of about 100 pm from the ion implanted
boundary (IIB, see schematic). It is noted that bands of etch pits (A) exist separated by
zones (B) which are relatively free of defects.
Fig. 7.23 - SEM micrograph o f etch pit bands observed in the non implanted area from wa fer Gl. A schematic showing the region o f analyses is included.
Fig. 7.24 (a-b) shows low magnification ( 100 x and 1000 x) optical micrographs of the
non implanted area of the same sample ( after defect etching) Bands of etch pits are
found to follow the implanted/non-implanted boundary as can be seen in the lower
magnification optical micrograph (fig. 7.24 a). Furthermore, it is observed that the
separation between the etch pit bands increases as a function of the distance from the
IIB until they completely disappear at a distance of about 200 pm from the IIB. The
defects responsible for the bands of etch pits were identified as faulted dislocation
loops[109], since it was possible to grow them into large OISF ( 1= 20 pm) during a
160 Results II
subsequent dry oxidation treatment[122] ( 2 hours at 1100 °C). The OISF were
identified by SEM observations after being defect etched
a )
20 um
Fig. 7.24 (a-b) Optical micrographs o f the etch pits bands following the ionimplanted boundary (IIB).
161 Residts II
The existence of defect bands parallel to the IIB suggests that large stresses are present
in this region due to the pinning of the silicon overlayer ( see section 7.2.6).
In Fig. 7.25 (a-b) we show SEM micrographs ( after defect etching) from the implanted
area of wafer G14 which was implanted at a higher current density than normal, namely at
7.75 pA/cm'2. From these micrographs, it is observed that bands of defects, which are
orientated along <110> directions, exist inside the implanted area. We postulate that this
effect is due to slip within the silicon overlayer. Similar results were observed in other
wafers implanted with current densities of 6.0 pA/cm'2 and 7.75 pA/cm'2, namely wafers
G15, G16 and G17. The occurance of slip is assumed to be related to the higher thermal
gradients[141] in wafers implanted at the higher current densities ( greater power
loading).
In order to confirm that large stress fields can exist in these wafers, double crystal X-ray
diffraction analyses have been carried out. Fig. 7 26 a) shows a X-ray diffraction rocking
curve (Cu Ka ) from the non implanted area at a distance of 0.5 cm from the visible IIB
border of an as-implanted wafer (<j)= 3.OxlO1' 0 “ cm'2, E= 70 kV). Fig. 7.26 b) shows
the rocking curve of the non implanted area of the same wafer, far removed (3cm) from
the IIB. We observe that a broadening of the peak occurs with the FWHM increasing
from 14.15' to 25.27' close to the IIB.This broadening of the FWHM confirms an
increase in the stress state of that region [142], A discussion on the effects of large
stress on the nucleation of crystallographic defects is given in section 8.3.1.
162 Results I I
Fig. 7.25 (a-b) SEM micrograph o f defect bands in the implanted area o f wafer G14. The defect bands are orientated along <110> directions.
163 Results II
Countrate
C o u n tn a te
Fig. 7.26 - X-rays diffraction rocking cun’es from the non implanted region a) 0.5 cm and b) 3 cm cnvay from the visible IIB. 3.3xl017 0 + cm'2 at E =70 keV (as-implanted).
164 Results II
7.3.7 - D efect densities in the silicon overlaver
The defect densities in all wafers implanted at the University of Surrey are summarised in
table 7.6. It is observed that optimisation of the implantation parameters such as dose,
current beam density and temperature can lead to the formation of thin SIMOX
structures with a relative low defect density (= 105 cm'2). Indeed, for the low energy
SIMOX samples the lowest defect density (3.0xl05 cm'2) was observed in wafer G4
which was implanted at an energy of 70 keV and a dose of ch = 3.3x1017 0 + cm'2 (T{
=700°C). On the other hand, the highest defect density was observed in wafer G13 which
was implanted at an energy of 90 keV and dose of ^= l.OxlO18 O' cm'2 (T;=700 °C). This
high defect density is attributed to the non optimised dose , which leads to an extremely
high defect density due to the high concentration of point defects produced by the
implantation process and by the internal oxidation of silicon ( see section 3.3). We
conclude that, the best results are obtained for samples implanted with doses close to the
threshold dose [59] for the formation of the buried oxide layer, which are, for the 70 keV
and 90 keV implantations, 3.3xl017 0 + cm'2 and 5.0xl017 O' cm'2 respectively.
For the 200 keV SIMOX samples , the lowest defect density (l.OxlO4 cm'2) was
observed in wafer G14 which was implanted with a dose of <J)= 4.0xl017 O' cm'2 at T; =
700°C. However, it is important to point out that this wafer was implanted with a high
current beam density (7.75 pA/cm'2) and that slip was observed in part of the silicon
overlayer (see section 7.3.6).
165 Results I I
dislo
catio
n de
nsity
(c
m'2)
© 1—MX©VO©r—4XCN
00ofHXoco’
©fHXo
NOot-4XoVO
©1—4XoCO<ooT—1Xovo
©fHXCOCV
000 1 1XpCV
voOrHXpcoi>OpHX
ot—H
00ot—4XpT—4
ooo1—4XoON
COoi—4XpCO
00ot—4Xp00
©rHXoOn
Q\©t—4Xvocv
lbfHXpt—4
©t—4X001
©t—4XpA
©1—1Xpcv
©fHXovo©iHXpcv
r~OfH
S
©pH
S.cv
jjjj
° f t
i t j
l i t
i—4CO
r—t CO
1—4co
»—4co
<—4CO
1—4CO
r—4co’
=-4CO
■—(CO
1—4CO
t—4CO
i—4CO
t—4CO
voC-fi*
CVNO
RfN R T—1
CO1—4CO
vortA
G0
1 1 1 I I
ovoNOovovo
oR
oR
oONOoovo
oR
OvoNOoo<o
Ovovooovo
oot"oR
oR
ooC"
ool>
oR
Oo oR
oR
cn<D E: Rg ° -ua ■ 3 °
COCO
coco CO
COco’
COCO
COCO vo VO vo vo vo CO
CO*ofH vo vo NO NO
Ene
rgy
(keV
)
R R R R R R oON oON oON oON oON oON oONooCV
oocvooCV
ooCVooCV
oocvooCV
lCO
rH0
CN0
cn0 s 0
r t0
K0
000 s
o4--10
1—41—40
CN1—40
CO1—40 0
woH0
r t1—40
r~-rH0
oo1—40
ONfH0
oCN0
166 Results II
Table
7.6
- Di
sloca
tion
dens
ities
in
wafer
s im
plan
ted
at the
Un
iversi
ty of
Surr
ey
The crystalline quality of the silicon substrate below the BOX was also assessed using
chemical defect etching. Before applying the defect etchant the silicon overlayer was
removed using a KOH etch and the BOX layer was removed using a HF(40%) solution
(section 4.1). Fig. 7.27 (a-b) are typical SEM micrographs of the silicon substrate of
wafer G7 after being defect etched for 1 min. The etch pits characteristic of stacking
faults are clearly observed, and are labelled SF in the figures.
7.3.8 C rvstallosravhic defects in the S i substrate
Fig. 7.27 (a-b)- Typical SEM micrographs o f the silicon substrate o f wafer G7 after being etched fo r 1 min.
167 Results II
The SF density measured in the silicon substrate was (5.0 ± 2.0)xl05 cm'2. Furthermore,
from Fig. 7.27 b) it is observed that the SF are decorated since the etch pits spread also
along <100> directions. Plan view TEM analyses of the silicon substrate of a similar
sample, carried out at the University of Oxford [143], have confirmed that the SF are
decorated by Cu colonies orientated along <100> directions. SF were also observed in
the non implanted area of the SIMOX wafers as described for the Eaton substrates in
Section 7.2.6. It is important to point out that the Si wafers used both at the University
of Surrey and at Eaton Corporation were supplied by the same vendor (Wacker) and
are expected to have similar concentrations of contaminants and growth defects. In order
to determine if the nucleation of SF in these wafers was related to the implantation
process or the high thermal treatment, a non-implanted bulk silicon wafer was annealed
at 1300°C for 6 hours and examined in the SEM after chemical defect etching. The
annealed bulk silicon wafer showed that SF ( density: l.OxlO6 cm'2) were formed during
the annealing process. This result suggests that the nucleation sites for the SF are not
introduced during the 0 + but pre-exist in the silicon wafer prior to the 0 + implantation.
SIMS analysis (see fig. 7.28 a-b) of Wacker silicon wafers have shown an increase in the
Cu concentration in the near surface region from 3x1015 /cm3 to more than 2x1018/ cm'3
during high temperature annealing (1300°C, 6 hours). In the absence of other
experimental data we assume that this increase is due to diffusion and segregation of Cu
from the bulk of the Si wafer. This increase in the metal contamination near the surface
could explain the formation of the decorated SF observed in the non implanted area of
the SIMOX wafers.
168 Results I I
DEPTH (aicrons)
DEPTH (oicrons)
Fig. 7.28 - SIM S copper concentration-depth profiles o f (100) Si wafer (supplied by Wacker) a) non annealed and b) annealed at 1300°C fo r 6 hours.
169 Results II
C h a p te r E i g h t : D i s c u s s i o n .
8.1 Introduction.
In this chapter, we discuss the main results obtained in chapter 6 and 7 and give some
interpretations on the possible mechanisms of formation of crystallographic defects in
thin film SIMOX.
This chapter has two sections. In the first (8.2), we discuss the formation of
crystallographic defects in sacrificially oxidised SIMOX and in the second (8.3) we
discuss the origin of defects in low energy, low dose SIMOX.
170 Discussion
8 .2 C rv s ta llo e ra v h ic d e fec ts in sa crific ia llv o x id ise d S IM O X
In chapter 6, we have shown that during the thermal oxidation of SIMOX materials,
oxidation induced stacking faults (OISF) are formed. We have also shown that the OISF
density, which varies from 5.0xl02 cm'2 and 5.0xl05 cm'2 in the particular samples under
investigaton is strongly dependent upon the density of SF complexes present in the
silicon overlayer before oxidation. In the next sub-section, the nucleation and evolution of
OISF are discussed in more detail.
8.2.1 Nucleation process o f OISF in SIMOX
The conversion of Si to S i02 involves a 2.2 fold increase in molar volume and therefore,
unless there is an efficient mechanism to account for this extra volume, high stress fields
will be formed at the Si/Si02 interface. Although several stress relief mechanisms may be
active during the thermal oxidation of silicon[144,145,146] the most widely accepted one,
involves the emission of silicon interstitials or absorption of vacancies[147]. This model
was first proposed by Dunham et al [147] and they suggested that most o f the
interstitials ejected are immediately captured by the superficial oxide with only a few,
perhaps less than one in 104, diffusing into the bulk silicon. Therefore, we can see that
the thermal oxidation of bulk silicon creates a perturbation in the point defect equilibrium
concentration.
171 Discussion
In the case of SIMOX, a super-saturation o f silicon interstitials in the silicon overlayer is
expected since the diffusivity o f Si through the buried oxide (BOX) layer is extremely
low ( D= l.OxlO'20 cm'2 s'1 at 1100°C) [ 148] and, therefore, the BOX layer acts as an
effective barrier for the diffusion o f the interstitials . Indeed, several authors [149, 150]
have suggested that the formation of {113} defects below the buried oxide layer o f
SIMOX is due to a super-saturation o f Si interstitials that are unable to diffuse in the
opposite direction to the surface through the buried oxide layer. Recently, Guillemot et al
[151] have proposed that Sij could diffuse through the buried oxide layer as SiO since the
diffusivity of SiO in S i02 (at 1000 °C) is four orders of magnitude ( 1.6xl0'16 cm 'V1)
higher than that of Si in S i02 and therefore, the supersaturation of silicon interstitials
could be reduced through the formation o f SiO molecules. However, this mechanism is
unlikely to occur at the temperatures used in our oxidation studies ( 800°C- 1100°C)
since the activation energy for the dissociation o f SiO, into SiO is very high (4.4eV)
[151]. We conclude that the BOX layer is an effective.barrier for the downwards diffusion
(towards the bulk) of Si interstitials, giving rise to an interstitial super-saturation in the
silicon overlayer.
In crystalline materials the free energy associated with a point defect is high [152] and
therefore, recombination with other point defects or annihilation at sinks is generally
energetically favourable. In SIMOX the surface and the Si/Si02 interface are the natural
sinks for interstitials, since point defects can be easily incorporated in surface 'kinks'
[153],
Other sinks for interstitials are the crystallographic defects present in the silicon overlayer.
It is nowadays widely accepted[154] that crystallographic defects can act as effective
sinks for point defects. Indeed, proposed annealing models for point defect in metals
172 Discussion
[155] generally take into account the presence of defects such as SFT, dislocation loops
and others.
In section 5.3, we have shown that threading dislocations and SF complexes are the
main crystallographic defects present in the silicon overlayer of SIMOX materials.
However, we have also shown (see chapter 6) that threading dislocations do not
influence the formation of OISF. In order to understand why threading dislocations do not
act as effective sinks for interstitials it is important to know the origin and nature o f the
threading dislocations. Stoemenos et al [48] have proposed that threading dislocations
originate from the evolution of perfect semi dislocation loops (SDL) (b =1/2<110>) that
expand by a climb mechanism through the capture of point defects. Once the SDL
reaches the surface it opens up into a paired dislocation by the escape of the segment
parallel to the surface. However, it has been recently suggested by Lee et al [156], that
after reaching the surface the perfect dislocation can dissociate into two Shockley partials
through a mechanism given by[ 157]:
}[! 10] —» j[2 U ] + j [121 ]
perfect Shockley Shockley
This dissociation reaction can occur when a segment of the originally perfect dislocation
lies along the intersection of the {110} habit plane of the half loop [ 157] and a {111}
slip plane. This process forms a narrow stacking fault (NSF) in a {111} plane bounded by
the two partials. However, it is unlikely that the NSF can act as an efficient sink for
interstitials since the Shockley partials that bound, the fault can not efficiently capture
point defects[158]. Therefore, we conclude that the threading dislocations will not
influence directly the formation o f OISF.
173 Discussion
On the other hand, in chapter 6, we have also presented extensive evidence to show that
SF complexes act as efficient sinks for interstitials and influence the nucleation process o f
OISF in SIMOX. Although the nature and origin o f these defects in SIMOX is still
uncertain some models for their formation have been presented recently by
Stoemenos[159j and Venables [160] and these are discussed below.
A. Mechanisms for the formation of SF complexes (SFT/SFP).
1. Stacking fa u lt tetrahedra (SFT)
Stacking fault tetrahedra (SFT) were first reported in quenched gold (polycrystalline Au,
99.998%) by Silcox and Hirsch in 1959[111], They attributed the formation to the
dissociation of intrinsic Frank dislocation loops through a mechanism described in
Appendix C. Another mechanism for the formation o f SFT in metals (Al, Cu) was later
proposed by Jong and Koehler[161] who suggested that SFT could be formed by the
aggregation of tetravacancies into a SFT by the capture o f a divacancy ( where the
six-vacancy cluster is the smallest tetrahedron). Although these two models differ
substantially in nature, they agree in the fact that SFT arise directly or indirectly due to
the condensation of vacancies.
The nature and origin o f SFT in SIMOX have still not been determined. However, several
authors [159,162] have suggested that they are formed due to the condensation o f
vacancies produced during the dissolution o f the S i02 precipitates during the latest stages
of the annealing process. This proposal is backed by the fact that intrinsic SFT have been
observed in P+ implanted Si [163],
174 Discussion
As in the case of SFT, the nature of SFP in SIMOX has still not been determined,
however, several mechanisms for the formation of SFP have been proposed recently.
A recent proposal made by Stoemenos et al [159], is that SFP are "as-grown" defects
formed during the high temperature anneal as a result o f the considerable silicon
redistribution near the Si/Si02 interface that occurs during the dissolution o f large S i02
precipitates in the vicinity o f the interface. Stoemenos also suggested that the sequence of
faults inside the SFP has to be intrinsic/extrinsic/intrinsic/extrinsic in order to satisfy
conditions imposed by the stair rod partial dislocations that bounds the fault[ 163],
A second mechanism recently proposed by Venables et al [160] is that the formation of
SFP occurs due to the interaction between two or more narrow stacking faults (NSF)
that are pinned by a S i02 precipitate. In this way the NSF can interact with each other
forming a SF complex. Upon the dissolution of the S i02 precipitate the interacting NSF
can further interact and develop into a complete SFP.
Here, it is our supposition, that the formation of SFP occurs due to the interaction of
two adjacent Frank dislocation loops pinned by a Si02 precipitate through a mechanism
similar to that proposed by Venables (see Fig. 8.1 ).We believe that the Frank dislocation
loops can be nucleated either (i) by the condensation of vacancies produced by the
dissolution of the S i02 precipitates or (ii) by a prismatic punching process [116] (see
section 3.3.2) in which an intrinsic and an extrinsic dislocation loop are formed near the
S i02 precipitate. This supposition is substantiated by experimental evidence shown in
chapter 6, that suggests that the SFP are effective sinks for vacancies (see Fig.6.11 ).
Although the model proposed by Venables can account for the formation o f square like
2. Stacking Fault Pyramidals (SFP)
175 Discussion
SFP, it cannot explain the presence of elongated SFP (see Fig 6.9), since the Shockley
partials are not efficient sinks for point defects [ 158],
SiO 2 precipitate
Frank loop
D issociated F. loop
/
8.1- Schematic showing a proposed mechanism fo r the formation o f SFP during the dissolution o f a SiO , precipitate.
176 Discussion
B. SF complexes as nucleation sites for OISF
1. Stacking fau lt tetrahedra (SFT).
Koehler and Lund [165] have reported that the activation energy for the annealing rate
o f SFT in gold is 2.2 eV. Furthermore, Cotterill and Jones [ 166] have observed that
SFT present in quenched gold can be completely eliminated when submitted to a
particle-irradiation. They suggested that the SFT are removed by the capture o f the
excess interstitials generated by the irradiation. Moreover, Jong and Koehler [161] have
proposed that the incorporation o f point defects in the SFT occurs through the capture of
point defects by the stair-rod dislocations that bound the fault. During the incorporation
o f point defects the SFT can grow or shrink depending upon the nature of the point
defects. Indeed , Silcox and Hirsch [111] have suggested that under certain conditions,
the SFT can collapse into a Frank sessile dislocation loop on any of the four {111} planes
associated with the SFT.
Although the nature of SFT in SIMOX is unknown it is our supposition that they also
arise from the condensation of vacancies and therefore, they will also act as effective sinks
for point defects. However, according to the annealing model proposed by Jong and
Koehler [161] the capture o f interstitials by the SFT will cause the defect to shrink until
it is completely eliminated. We conclude that, if the nature of SFT in SIMOX is intrinsic,
they will not act as nucleation sites for OISF.
On the other hand, if the SFT in SEMOX are extrinsic, the capture of interstitials by the
stair rod dislocations will lead to the growth o f the defect until its size reaches a critical
value (10) when it will become unstable and will collapse into an extrinsic Frank loop,
according to the model by Silcox and Hirsch[l 11]. In this case the SFT will act as direct
177 Discussion
nucleation sites for OISF, since the extrinsic Frank loop can further capture interstitials
and expand into a large OISF.
2. Stacking fau lt pyramidals
Although the nature of the SFP have not yet been determined, the different mechanisms
of formation of these defects suggest that they can be either intrinsic, extrinsic or
mixed.If the nature of a SFP is extrinsic then, the incorporation of interstitials will lead
to the growth of the defect until it collapses into extrinsic Frank loops in a similar
manner to the SFT. The Frank loops can then grow by capturing further interstitials and
evolve into OISF. In the case o f an intrinsic/extrinsic SFP, the two opposite planes will
grow through the capture o f interstitials while the other two intrinsic planes will shrink
and probably disappear. Once again, upon further growth, the defect can collapse and
give rise to the formation o f OISF.
In the case of a pure intrinsic SFP, the capture of interstitials will lead to a reduction in
the size of the defect and, therefore, although they will act as efficient sinks for
interstitials they will not nucleate OISF.
178 Discussion
8 ,2 .2 E vo lu tion o f O IS F d u rin g oxidation .
Once an embyonic OISF is nucleated, it will evolve through the capture of self
interstitials by the Frank partial that bounds the fault. It is important to remember that the
Frank dislocation ( b=l/3 <111 >) loop that bounds the fault is not glissible and therefore
it can only evolve through a non-conservative motion. This can only be a climb process
involving the capture of interstitials. Fig. 8.2 is a schematic showing different stages in
the evolution of the OISF in SIMOX. In the first stage o f evolution an embryonic OISF is
formed near the Si/Si02 interface probably through the collapse of a SF complex which
will form one Frank dislocation loop (See Fig. 8.2 a). Upon further oxidation, the Frank
partial that bounds the fault will grow by a climb process through the capture of
interstitials (see Fig. 8.2 b ). Once the partial loop reaches the surface it opens up into
two Frank partial dislocations in a similar way to that observed for threading dislocations.
After splitting of the partial, the fault is only able to grow laterally by the incorporation
o f interstitials by the two partials (See Fig. 8.2 )
179 Discussion
A. Activation energy for the growth o f OISF in SIMOX
In chapter six we have shown ( see for example Fig. 6.6) that the OISF length was
strongly dependent upon temperature. Using data presented in Fig. 6.6 and in appendix E,
we have estimated the activation energy for the growth of the OISF in SIMOX. The
results are displayed in Fig. 8.3, showing In R (rate of growth) as a function of T"1.
Performing a linear regression on the data, it was possible to estimate an activation energy
Ea = 1.5 ± 0.5 eV. This value is comparable to the values for bulk silicon (2.3 eV) [167]
when the rather large experimental uncertainty is taken into account.
3
2
QCJ=
1
°7.0 7.5 8.0 8.5 9.01/T ( x1 E4 1C1)
Fig. 8.3 - In R(rate o f growth o f OISF) as a function o f T 1.
181 Discussion
B. Estimation o f the number o f self interstitials emitted during dry oxidation.
Assuming that the number o f interstitials captured by the OISF is proportional to the
total number of interstitials emitted during oxidation, we can make an estimation o f the
interstitial super-saturation at different oxidation temperatures under dry oxidation
conditions. This can be done by estimating the total number o f interstitials inside the
OISF ( Ns,oisf) and correlating this with the total number of silicon atoms consumed by
the thermal oxidation (Nox). The total surface area of the OISF (SOISF) is expressed by
[151]:
S o i s f = D o i s f I o i s f Q i ! sin 5 5 °) (8.1)
where D01SF is the OISF density, LOISF is the OISF mean length and h is the thickness of
the silicon overlayer and 55° is the angle that {111} planes form with the (001) surface
(see Fig. 8.4 ). Since the number of atoms per cm'2 in the {111} extra plane is N0 =
1.6xl015 atoms/cm'2, NSlOISF is given by: Ns’OISF= N0 S0ISF. The total number of silicon
atoms consumed (Nox) can be estimated by knowing the thickness of the silicon
consumed and the molar density o f silicon (p = 12.1 cmVrnol.).
Frank partial
Fig. 8. 4 - Schematic showing an OISF inside the silicon overlayer.
182 Discussion
6-™°
/!S 01.91.x)
Fig. 8,5 (a-d) shows the dependence of the Ns'ols,.. as a function o fN 0X, From Fig. 8.5 it is
observed that the NS'01SF increases according to the temperature for a given SIMOX
substrate. This result is not unexpected since for high temperatures the oxidation rate
increases considerably and therefore the incorporation of the ejected interstitials by the
oxide interface is more difficult and as a consequence more interstitials will be able to
diffuse into the silicon overlayer[165]. The other feature observed from Fig. 8.5 is that the
number of silicon interstitials captured by the OISF depends upon the particular choice of
the SIMOX wafer. This result indicates that other sinks for interstitials exist, and as we
have already suggested we believe that this difference is due to the presence of SFT in the
silicon overlayer.
1 0 0 0 0
1 0 0 0
1 1 0 000z
100 2 4 6 8 10
Nox ( x 1E17 Si/cm'2)
Fig, 8.5 - a (continued) [ the lines are drawn only to guide the eyeJ.
183 Discussion
nS'o
i.st ( X
1 e1
0 Si
/ cm
-7 )
Nsi
( X1 e
10 Si
/ cm
'2)
Nox ( x 1E17 Si/cm'2 )
1 0 0
2 4 6 8Nox ( x 1E17 Si/ cm'2 )
8.5 b-c (continued)
184 Discussion
<Nso
o5o0 1-j”"X
u.00O
S=
10
0 2 4 6 8 10Nox ( x 1E17 Si/ cm'2 )
Fig. 8.5-N5' ^ versus Nox fo r substrates a) B2 b) B3 c) B5 and d)Bl
Using the data presented in Fig. 8.5, an estimation o f the number of interstitials emitted
per silicon atom consumed during oxidation (NS,OISF/Nox) can be obtained. In Fig. 8.6, we
plot the Ns’olSF/N0X ( for Nox=4 .OxlO17 cm'2) as a function of temperature for wafer B2. It
is observed that in the particular case o f wafer B2 the value for Nsi0iSF /Nox at 1100°C is
5.0x1 O'5, which is in close agreement to the value theoretically obtained by Dunham et al.
[147],
1100°C
d
J i_____i— ■ I ■ I-
185 Discussion
850 900 950 1000 1050 1100 1150Temperature ( ° c )
Fig. 8,6 - as a function o f temperature fo r wafer B2,
186 Discussion
8 .3 C rvs ta llo e ra vh ic d e fec ts in lo w en ergy a n d lo w d o se S IM O X .
The results presented in chapter 7 have shown that the main defects observed in the
silicon overlayer of low energy and low dose SIMOX samples were "threading"
dislocations and in some cases, SF complexes (see Fig. 7.7). We have also shown that
implantation parameters such as dose, energy and implantation temperature, have a great
effect on the final defect density. In fact, a wide range of defect densities ( from 104 cm'2
to 109 cm'2) was observed depending upon the implantation conditions.
In order to understand the reasons why the implantation parameters affect so dramatically
the formation of crystallographic defects it is important to refer back to chapter 3. In
chapter 3 we have shown that the main sources of defects in oxygen implanted silicon
were basically the radiation damage and the internal oxidation of silicon. We have also
indicated that the last factor contributed twice as a source of defects since it was
responsible for the generation o f self interstitials and also for the plastic deformation of
the silicon matrix around the S i02 precipitate. In this way , we would expect that low
dose O* implantations (close to <f>cA) at high temperatures (700°C) would minimise the
effects o f both the radiation damage and the internal oxidation of silicon, and therefore,
would produce low defect densities. Results by other groups[57] (see Fig. 2.3) have
suggested that extremely low defect densities (10J cm'2) are formed for low dose and high
implantation temperatures. However, by scaling down the implantation energy in order to
form thin silicon overlays, a new problem is introduced since, in doing so, we bring the
peak o f damage (R J closer to the surface and therefore, dislocations that are pinned to
187 Discussion
Def
ect
dens
ity
( c
m-2
)large S i02 precipitates (see Fig. 7.21) can escape more easily to the surface during the
annealing process.
In spite of that problem, our results have shown that low defect densities can be obtained
in thin film SIMOX produced by low energy and low dose 0 + implantations using high
implantation temperatures (Ti > 500°C) (see Fig. 8.7).
10 20 30 40 50 60 70 80 90 100Energy (keV)
Fig. 8.7 - Defect density upon implantation energy for doses close to (Pic [59J.The line is to guide the eye.
From Fig. 8.7 we can see that, interestingly, the best results (l.OxlO4 cm’2) were obtained
for the lowest energy (30 keV). However, it is important to point out that in Fig. 8.7. we
have plotted the results obtained from samples implanted at Eaton and Surrey without
making any distinction between them. Obviously, clear differences in the way that these
188 Discussion
| - r | I | i | i | i | i | r
: I • Eaton samples_ ♦ Surrey samples
i I i l ■
samples have been prepared (see chapter 4) may have influenced the final
crystallographic defect density but what emerges from Fig. 8.7 is that the defect density
does not necessarily increase with decreasing ion energy, indeed in these samples the
density decreases. These results indicate that the positioning of the peak of damage Rd
closer to the surface ( by using lower energies) does not increase significantly the defect
density as might be initially expected[62], Meda et al [169] have proposed that the
gradual merging of the damage and atomic concentration-depth profiles with decreasing
ion energy will facilitate the formation of a continuous BOX layer. The results presented
in Fig. 8.7 may be a consequence of the improved planarity of the final structures with the
proximity to the free surface having a significantly smaller effect.
Comparing the best results obtained for samples implanted with low dose and low energy
(104 cm'2) and those obtained from other groups[57], for low dose and high energy
implantations (103 cm'2) there is a difference of one order of magnitude between them.
We believe that this difference is, at least partially, due to. undesirable effects such as
thermal stresses produced during the local area implantation and the presence of
contaminants in the silicon substrates which are due to sputtering of metal components
within the accelerator vaccum system.
8.3.1 Effects o f thermal stresses on the nucleation of defects in SIMOX.
In both the wafers implanted at the University of Surrey and at Eaton, we have observed
that anomalous defect densities (see Fig. 7.11) were present in areas near the ion
implanted boundary (IIB). We believe that these highly defective areas appear as a
189 Discussion
consequence o f large stress fields caused both by the pinning o f the silicon overlayer and
by the formation o f large thermal gradients (see section 7.2). The effects of large stresses
on the generation o f crystallographic defects in SIMOX have been recently discussed by
Nejim et al [170] and by Venables et al [51]. Venables and co-authors have argued that
large stress fields can assist in the nucleation and climb of dislocation loops. These stress
fields can cause plastic deformation o f the silicon matrix by generation of dislocations,
either heterogeneously by the capture and multiplication o f dislocations from other
sources or homogeneously by the condensation of silicon interstitials[171].
Homogeneous nucleation occurs following a clustering of silicon interstitials which align
in the stress field and collapse into a dislocation loop [172]. The observation o f defect
bands entering the implanted area along <110> directions (see Fig. 7.25 ) is strong
evidence of the deformation process caused by these large stress fields.
We believe that, for the wafers prepared at Eaton Corporation (see table 7.3), the
nucleation o f defects caused by this process was localized in the regions close to the
implanted/non-implanted boundary because almost the whole 6 inch wafer was implanted.
However, for the wafers prepared at the University o f Surrey, the effect o f large stress
fields on the nucleation of defects affects the whole o f the implanted area because only a
6.45 cm'2 area of the 3 inch diameter wafer was implanted. Most probably, this is one o f
the reasons why the lowest defect densities observed in wafers prepared at the University
o f Surrey was 30 times higher than those obtained from wafers prepared at Eaton
Corporation. Therefore, we think that to further reduce the defect density it will be
necessary to implant the whole wafer. In doing so, thermal and mechanical stresses will
be reduced leading to lower defect densities.
190 Discussion
8.3.2 Effects o f contamination on the formation of crvstallozraphicdefects.
Carbon and metals contaminants (such as Fe, Ni, etc) can be accidentally introduced
during the implantation process. Carbon is likely to be incorporated into the wafer surface
by a "knock on" process of hydrocarbons deposited on the surface of the wafer, which
originate from the vacuum pump oils while metals can be sputtered from the walls of the
implanter or mounting clips during the implantation process[ 173],
Recently, the role of carbon in the nucleation o f defects has been studied [174], It has
been suggested that C absorbs silicon interstitials by the formation of beta SiC
precipitates. This process leads to volume contraction preventing the development of
stress caused by self-interstitial supersaturation[175]. If a supersaturation of vacancies is
present, the formation of beta SiC precipitates will further increase the vacancy
supersaturation, enhancing the condensation of vacancies and the formation of dislocation
loops through the collapse o f vacancy aggregates. Therefore, the detrimental or beneficial
role o f carbon in the formation o f defects will depend upon the concentration and type of
point defects in the material.
On the other hand, the detrimental effects o f metal contamination are widely known[176].
The formation of crystallographic defects due to the presence of metal contamination may
occur through the nucleation o f dislocation loops induced by the metal silicide
precipitate. These dislocation loops are nucleated due to the plastic deformation of the
silicon matrix caused by the extra volume o f the new phase (see section 3.3).
191 Discussion
In chapter 7, we have shown that a high concentration of Cu (see Fig. 7.26) was observed
in silicon wafers supplied by Wacker after high temperature treatments. We have shown
that decorated SF (5.0x105 cm '2 ) were formed below the buried oxide layer and in the
non-implanted area of the silicon wafer. In addition, we have also shown that Cu was
not introduced during the implantation process since it was already observed in non
implanted wafers after high temperature processing (1300°C, 6 hours in Ar + 1/2% 0 2).
These results suggest at first sight that the high Cu concentration originated from
contamination of the furnace tube. However, if this was the case, we would expect that
the incorporation of Cu would occur homogeneously along the silicon surface and
therefore, we would expect to find similar SF densities inside and outside the implanted
area. Since no SF were observed in the implanted area ( < 102 cm'2) this assumption seems
unlikely. Another possibility is that Cu was already present in the bulk of the silicon wafer
before annealing and during the thermal treatment diffuses towards the silicon surface.
This assumption could explain why SF were not observed in the silicon overlayer if we
consider that the BOX layer acts as a barrier for the diffusion of Cu. This assumption is
further supported by the results obtained from Wacker wafers annealed in a different
furnace (Eaton laboratories) in which a similar SF density of 5.0xl05 cm'2 was observed.
If the formation of SF was furnace dependent we would expect to find a completely
different SF density. Since the wafers were supplied by the same vendor it is probable that
the Cu contamination in these wafers was the same, giving rise to similar SF densities.
These findings suggest that further reduction o f defect densities could be achieved by the
reduction of contaminant levels using a cleaner implanter environment and also by using
silicon substrates more amenable to extremely high temperature.
192 Discussion
C o n c l u s i o n s a n d S u g g e s t i o n s f o r f u t u r e w o r k
C h a p t e r N i n e :
In this project we have extensively characterized thin SIMOX materials prepared by two
different methods namely, sacrificial oxidation o f thick SIMOX materials and low energy
and low dose oxygen implantations. For this purpose we have specially developed a new
defect etchant capable o f delineating stacking faults and threading dislocations in thin
silicon films (>500 A). Detailed analysis o f large areas of thin film SIMOX was possible
due to the speed with which the samples could be prepared and etch pit densities
obtained. Determination o f the lateral distributions of crystallographic defects in the
implanted and non-implanted areas o f the wafers was also carried out. Furthermore, in
conjuntion with XTEM and plan view TEM analysis, the depth distribution o f the
crystallographic defects was also determined. From these observations it was possible to
determine the nature and origin o f the different crystallographic defects, and by so
doing, control their final density. The main conclusions drawn from these extensive
analyses are presented below.
193 Conclusions
The main defects observed in oxidised SIMOX samples were OISF and threading
dislocations. The OISF densities found in the samples analysed in this project varied from
5.0xl02 to 5.0xl05 cm'2 and were found to depend upon the density o f nucleation sites
prior to the oxidation . The nucleation and growth o f the OISF have also been studied
and it has been determined that the main nucleation sites are SFP/SFT that are present
before oxidation and lie close to the Si/Si02 interface. The formation o f SFP in SIMOX
occurs during the high temperature annealing ( typically 1300°C) and although few
mechanisms of formation have been proposed in the literature we have suggested that
they arise from the interaction o f Frank dislocation loops pinned by large S i02
precipitates. Annihilation o f the SFP/SFT from the Si/Si02 interface by prior
amorphisation has shown that the effective annihilation of these defects enables the
substrates to be thinned whilst maintaining an extremely low OISF density (S.OxlO2 cm'2).
It has also been observed that the growth o f OISF in SIMOX is an activated process with
an activation energy o f 1.5 eV ± 0.5 eV. Furthermore, an estimate o f the number of
interstitials captured by the OISF has been made in order to determine the minimum
number o f interstitials ejected from the growing oxide during thermal oxidation. It has
been shown that the number o f ejected interstitials increases with oxidation temperature in
close agreement with the model proposed by Dunham et al [147], Therefore, in order to
avoid the formation of long OISF ( < 5 pm) , we suggest that dry oxidation treatments at
moderate temperatures ( 900°C or 950°C) be used to thin the silicon overlayers o f thick
SIMOX materials.
9.1 C ry s ta llo sra v h ic d e fe c ts in S I M O X th in n e d b y sa c r if ic ia l o x id a tio n
194 Conclusions
9.2. C rv s ta llo sra v h ic d e fe c ts in lo w en ergy, lo w d o se S IM O X .
9.2.1. SIM O X wafers produced at the University o f Surrey.
The main crystallographic defects present in the silicon overlayer of thin film structures
prepared at the University of Surrey are threading dislocations. The threading dislocation
density in samples implanted with a dose of 3.3xl017 0 + cm"2 at an energy of 70 keV
varied from 2.0x10* cm"2 to 3.OxlO5 cm"2 depending upon the implantation temperature.
Indeed, it was observed that the threading dislocation density decreases by 660 times
when the implantation temperature is increased from 550°C to 700°C. The same trend
was observed in samples implanted with 5.OxlO17 0 + cm'2 at an energy of 90 keV when
the dislocation density decreases from 1.0x10* cm"2 to 3.OxlO6 cm"2 when the implantation
temperature is increased from 550°C to 700°C. However, it was observed that for the
wafers implanted at the highest temperatures (650°C-700°C) a non uniform distribution of
threading dislocations across the implanted area was present. We have suggested that the
presence of different defect densities inside the same wafer, was either due to the pinning
of the silicon overlayer at its periphery by the surrounding bulk silicon or due to the
presence of thermal gradients in the wafer during the implantation. These assumptions
were validated by preliminary X-ray diffraction analyses which showed the existence of
large stress fields near the ion implanted boundaiy (IIB) in the as-implanted samples.
The relatively high threading dislocation densities can also be attributed to the presence of
contaminants or to the non adequate control of the implant conditions ( such as
particulates, current stability) at the University of Surrey.
195 Conclusions
9.2.2 SIMOX wafers produced at Eaton Corporation.
The main crystallographic defects present in the silicon overlayer of the specific wafers
examined were threading dislocations and SF complexes. The threading dislocation
density varied from l.OxlO4 cm'2 to 5.0xl06 cm'2, while the density o f the SF complexes
varied from 8.0xl06 cm'2 to 5.0xl09 cm'2. The lowest threading dislocation densities (
l.OxlO4 cm"2) were found in samples implanted using singly charged molecular oxygen
(0 +2) with a dose o f 2 .2xl017 0 + cm'2 and energy o f 30 keV (Ti= 530 °C).These results
are very interesting since they show that it is possible to obtain low defect density SIMOX
materials using low energy and low dose oxygen implants. Furthermore, complementary
analyses by XTEM showed that no silicon islands ( e.g < 107 cm'2) were present in the
buried oxide layer suggesting that these materials could be used, in the near future, as the
substrate for low power CMOS integrated circuits. Indeed, 0 + implants at 30 keV with
doses o f 2 .2xl017 0 +cm"2 produce, after annealing ( 1300°C for 6 hours), an ultra thin
silicon overlayer (500 A) containing a low density of threading dislocations ( l.OxlO4
cm'2).
9.3 Suggestions for future work.
In this project we have successfully characterized SIMOX materials and optimised the
processing conditions such as implantation temperature (700°C) and oxidation
temperature ( 900°C) to achieve reduced densities o f crystallographic defects in thin
SIMOX structures. Annihilation o f the SF complexes located near the Si/Si02 interface
has been achieved using a post amorphisation process. However, further improvements in
the quality o f the silicon overlayer are required in order to achieve dislocation free
196 Conclusions
SIMOX structures. For this purpose, we suggest that the nature and origin of SFP/SFT
have to be further studied. The determination of the nature (intrinsic/extrinsic) of
SFP/SFT in SEMOX by high resolution TEM analysis could give us more indications
about the actual mechanism o f formation o f these defects. Another experiment that could
be carried out in order to determine the mechanism o f formation o f SFP/SFT is to study
the effects o f quenching on the SFP/SFT density . This could be easily done by changing
the cooling down rates following the high temperature treatments. This experiment could
be used to further improve the thermal annealing process in order to produce SEMOX
structures with lower SF complex densities.
We also suggest that in a future project, the effects o f thermal and mechanical stresses on
the nucleation o f defects in low energy and low dose SIMOX has to be further studied.
Implantations of the whole surface o f the wafer would presumably reduce these effects
and therefore, comparison with partially implanted wafers or patterned wafers would
give us an indication about the influence o f the thermal/mechanical stresses on the
formation o f defects. Furthermore, the effects o f contaminants on the nucleation o f SF in
the silicon substrate need also to be studied. This could be done by implanting ultra pure
FZ-silicon wafers and comparing the results with those obtained with the Wacker Cz-
silicon wafers.
Finally, the quality o f the SIMOX wafers examined in this project has been found to be
very sensitive to the details o f the implantation and thermal processing. For this reason it
is suggested that any future experimental work should be a joint project involving a
SIMOX wafer, vendor so that tight control o f the processing conditions can be maintained
from batch to batch and wafer to wafer when meaningful defect densities could be
determined.
197 Conclusions
R e fe re n c e s
[1] - C.W. Mueller and P.H. Robinson, Proc. Inst. Electr. Eng. 52, 1487 (1964).
[2] - H.W. Lam, IEEE Circuit and Devices Magazine, page 6, July (1987).
[3] - S. Kawamura in " Ultra Thin Film SOI Technology And Its Application to NextGeneration CMOS Devices", Proc. IEEE SOI Conf., page 6 (1993).
[4] - S.S. Lau, S. Matterson, J.W. Meyer, P. Revesz, J. Byulai, J. Roth, Appl. Phys. Lett.34, page 76 (1979).
[5] - K. Imai, H. Unno and H. Takaoki, J. Crystal Growth 63. 547 (1983).
[6] - W. Maszara, Proc. SOI Technology and Devices, pag. 199, Edited by D. Schmidt,The Electrochemical Society, (1990).
[7] - K. Izumi, Proc. IV SOI Electrochem. Soc. Conf., page 31 (1990).
[8] - S. A. Park, C. Hu and P.K. Ko, IEEE Electron. Dev. Lett., 14, page234, (May 1993)
[9] - L. Cohn, H.H. Hosack, R. Cherne, P. Fechner, 158, Proc. IEEE SOI Conf., 158,(1993)
[10] - A.J. Auberton, Proc. 1993 IEEE SOI Conf., page 8, (1993).
[11]- J.P. Colinge, in " Advanced CMOS Devices Made in Thin SOI Films".Fifth international Workshop on Future Electron Devices, May 1988, Japan.
[12] - M. Guerra in "The status of SIMOX technology" Proc. SOI Technology andDevices, page 21, Edited by D. Schmidt, The Electrochemical Society, (1990).
[13] - B. Davari, H.J. Novel, G.G. Shahibi, Proc. IEEE SOI Conf., page 4, (1993).
[14] - B.K. Tanner, in " X ray Diffraction Topography", Edited by PergamonInternational Library (1976).
[15] - L. Azaroffin " Element of X-ray Crystallography" Me Graw-Hilll (1968).
[16] - M. Von Heimendahl, "Einfuhrung in die Elektronenmikroskopie" Friedrich Vieweg,(1970).
[17] - W.K. Chu, J.W. Mayer and M.A. Nicolet in " Backscattering Spectrometry",Edited by Academic Press, (1978)
198 References
[18] - J. Harrold, J. Hovel, Proe. IEEE SOI Conf. ,page 26 (October 1993).
[19] - MJhara, T. Kimura, H. Yamawaki, O. Ueda, Proc. IV SOI Electrochem. Soc.Conf., page 399 (1990).
[20] - R.B. Heimann in " Crystals: Silicon Chemical etching", page 175, Edited by J.Grabmaier, Springer-Verlag (1982).
[21] - M.L. Smith in "Electromagnetically Enriched Isotopes and Mass Spectrometry",Publishers: Butterworth Science, London, page 100, (1956)
[22] - M. Watanabe and A. Tooi, Jap. J. Appl. Phys _5, page 737, (1966).
[23] - P.V. Pavlov and E.V. Shitova, Sov. Phys. 12, page 11, (1967).
[24] -J.H. Freeman, Proc. Int. Conf. Applications of ion beams to semiconductor technology, Grenoble, page 75 (1967).
[25] - J. Dyleswsi and M.C. Yoshi, Thin Solid Films 3513). page 327, (1976).
[26] - M.H. Badawi and K.V. Anand, J. Phys. D 10. page 1931 (1977).
[27]- K . Izumi, M. Doken and M. Ariyoshi, Electron. Lett.JA, page 593 (1978).
[28] - H.W. Lam, A.F. Tarsh and R. F. Pinizzotto, in "Silicon-on-insulator for VLSI andVHSIC", VLSI Electronics Microstrusture Science 4, Ed. N.G. Einspruch, Academic Press, New York (1982).
[29]- D. Fathy, O.L. Krivanah, R.W. Carpenter and S.R. Wilson, Inst. Phys. Conf. Ser. 67, page 479 (1983).
[30] - S. S. Gill, PhD.Thesis, University of Surrey, Guildford, UK (1980).
[31] - S. S. Gill and I.H. Wilson, Thin Solid Films 55, 435, (1978).
[32] - T. Hashayi, S. Maeyama and S. Yoshi, Jap. J.Appl. Phys. 1916) , page 1111 (1980).
[33] - T. Hashayi, H. Okamoto and Y. Homma, Jap. J. Appl Phys. 1915) , page 1005(1980). *
[34] - K. Das, J. Butcher, M.C. Wilson, G.R. Booker, D.W. Wellby, P.L.F. Hemment and K.V. Annand, Inst. Phys. Conf. Ser. 60, page 307 (1981).
[35] - H. ICoyama, J. Appl. Phys. 51, page 3202 (1982)
[36] ]- Y. Homma, M. Oshima and T. Hayashi, Jap. J. Appl. Phys. 21, page 890 (1982).
199 References
[37] - J.A. Kilner, D.S. Littlewood, P.L.F. Hemment, E.A. Maydell-Ondruszand ICG. Stephens, Nucl. Instr. & Meths. 218. page 573 (1983).
[38] - P.L.F. Hemment, IC.J. Reeson, J.A. Kilner, R.J. Chater,C.D.Marsh,G.R.Booker,J.R. Davis, G. K. Celler, Nucl. Instr. & Methds. B21. page 129 (1987).
[39] - G. Carter and W. A. Grant in "Ion Implantation of Semiconductors"( Edited by Edward Arnold ), page 174, (1976).
[40] - J. Margail, PhD Thesis, Grenoble University (1987).
[41] - S. J. Krause, C. O. Jung, T.S. Ravi and D.E. Burke, Mat. Res. Soc. Soc.,Sym. Proc. 107, page 93, (1988).
[42] - O.W. Holland, T.P. Sjoren, D. Fathy and J. Narayan, Appl. Phys. Lett. 45.page 1081, (1984).
[43] - C. Jaussaud, J. Stoemenos, J. Margail, M. Dupuy, B. Blanchard and M. Bruel,Appl. Phys. Lett. 46, page 1064 (1985).
[44] - J. Stoemenos and J. Margail, Thin Solid films 135. page 119 (1986).
[45] - J.P. Ruffell, D.H. Douglas-Hamilton, R.E. Kaim and IC Izumi, Proc. 6th Int. Conf.Ion Impl. Techn., page 229, (1986).
[46] - P.L.F. Hemment, K.J. Reeson, J.A. Kilner and S.J. Krause, Vaccum 42(5/6) ,page 293, (1991).
[47] - C. Jaussaud, J. Stoemenos, J. Margail, A.M. Papon and M. Bruel, Vaccum42(5/6) page 341, (1991).
[48] - J. Stoemenos, IC J. Reeson, A.IC Robinson and P.L.F. Hemment, J. Appl. Phys.69, page 793, (1991).
[49] - S. Visitserngtrakul, C.O. Jung, T.S. Ravi, B. Cordts, D. Burke and S.J. KrauseInst. Pys. Conf. Ser. 100, page 557, (1989).
[50]- A. de Veirman, J. Van Landuyt, J. Vanhellemont, H. Maes and IC Yallup, Vaccum, 42, page 367, (1991).
[51] - D. Venables, K.S. Jones, F. Namavar and J. Manke, Mat. Res. Soc. Symp. Proc.235, page 103 (1992).
[52] - D. Hill, P. Fraundorf and G. Fraundorf, J. Appl. Phys. 63, page 4933 (1988).
[53] - Y. Homma, M. Yoshima and T. Hayashi, Jap. J. Appl. Phys. 2 \ , page 890(1982).
[54] - M.K. El Ghor and S. Pennycook, F. Navamar and N.H. Karam, Appl. Phys. Lett.57(2). page 156 (1990)
200 References
[55] - A. H. Van Ommen, Nucl .Instr. Methds. B39. page 194 (1989).
[56] - J. Margail et al. " A New Method for Maldng Improved SIMOX Structures",Proc. 1st European SOI W orkshor, Meylan, March 1988.
[57] - S. Nakashima, S. ICanamori and K. Izumi, Electron. Lett._2i, page 1229, (1989).
[58] - F. Navamar,E. Cortesi, B. Buchanan,P.Sioshansi, Proc. IEEE SOS/SOITechnology workshop (1989).
[59] - Y. Li, J.A. Kilner et al. , J. Appl. Phys. 70(7), page 3605 (1991).
[60] - C. Tsamis, D. Tsoukalas, N. Guillemot, J. Stoemenos, J. Margail, J. Electrochem.Soc. 138(9), page 2752 (1991).
[61] - D. Tsoukalas, C. Tamis, N. Guillemot, J. Stoemenos and J. Margail, MaterialsScience and Engineering B12. page 209 (1992).
[62] - J. Margail, J. M. Lamure and A.M, Papon, Materials Science and EngineeringB12, page 27 (1992).
[63] - M. J. Kelly, T.R. Guilinger, J.W. Medernach, S.S. Tsao, H.D. T. Jones and J. O.Stevenson Proc. SOI Technology and Devices, pag. 121, Edited by D. Schmidt,The Electrochemical Society, (1990).
[64] - M.J.J. Theunissen, A.H. Goemans, A.J.R. Kock and J. Haisma, Proc. SOITechnology and Devices, pag. 163, Edited by D. Schmidt, The Electrochemical Society, (1990).
[65] - H. Ming Liaw , M. Tomozane, P. L. Fejes, Proc. SOI Technology and Devices,page 129, Edited by D. Schmidt, The Electrochemical Society , (1990).
[66]- P.S. Fechner, G. R. Gardner and S.T. Liu, Materials letters, 9112) page 508, (1990).
[67] - L.F. Giles, A. Nejim and P.L.F. Hemment, Mater. Chem. Phys. 35. page 129(1993).
[68] - J.F. Ziegler in "Ion Implantation Science and Technology", page 51, edited by J.F.Ziegler, Academic Press (1984).
[69] - G. Dearnley, J.H. Freeman, R.S. Nelson and J. Stephen in " Defects in crystallinesolids: Ion Implantation" edited by S. Amelinckx, R. Gevers, J. Nihoul,North Holland Publishing Company (1973).
[70] - K.G. Stephens, K.J. Reeson, B,J. Sealy, R.M. Gwilliam and P.L.F. Hemment,Nucl. Instr. Meth. B50. page 350 (1990).
[71] - IC. Radermacher, S. Mantl, Ch. Dieker and H. Luth, Appl. Phys. Lett., 59, (1991).201 References
[72] - A.E. White, K.T. Short, R.C. Dynes, J.P. Garno and J.M. Gibson,Appl. Phys. Lett. 50, 95, (1987).
[73]- A. White, K.T. Short, K. Maex, R. Hull, Y.F. Hsieh, S.A. Audet, IC.W. Goosen, D.C. Jacobson and J.M. Poate , Nucl. Instr. & Methds. B59/60. 665, (1990).
[74] - R. Weber, R. Yankov, R. Muller, W. Skorupa, S. Reiss and IC.H. Heining, Mat.Res. Soc. Symp. 316. page 729, (1994).
[75] - U. Goselle, T. Y. Tan, Appl. Phys. A2§, page 79 (1982).
[76] - T. Lee and D. Nichols Appl. Phys. Lett. 47. page 1001 (1985)
[77] - T. Lee and D. Felluzer, Appl. Phys. Lett. 49, page 1793 (1985)
[78] - K. Graff, A. Gallrath, S. Ades, G. Goldbach, G. Tolg, Solid State E lect. 16,page 887 (1973).
[79] - F. Shimura in " Semiconductor Silicon Crystal Technology", pag. Academic PressHartcourt Brace Jonavich, Publishers (1989).
[80] - S. Mantl, Materials Science Report 8(1/2), page 1-95, (1992).
[81] - 1. H. Wilson; Ion Beam Modification of Insulators, Ed. P. Mazzoldi andG. Arnold, page 248, Elsevier, Amsterdam (1987).
[82] - R.D. Doherty in "Physical Metallurgy", Editors R.W. Cahn and P. Haasen(North Holland Physics Publishers), Page 933 (1983).
[83] - J. Osaka, N. Inuoue, K. W ada, Appl. Phys. Lett. 36(4), page 289 (1989).
[84] - Shimura, Tsuya, Kawamura, Appl. Phys. Lett 37(5) , page 485 (1980).
[85] - L. Meda, S. Bertoni, G.F. Cerofolini, C. Spaggiari Proc. SOI Conf, Edited byW. E. Bailey ( The Electrochem. Soc. Pennington, NJ) page 237 (1992 ).
[86] - P.L.F. Hemment, K.J. Reeson, J.A. Kilner, R.J. Chater, C.D. Marsh, G.R. Booker,G.K. Celler and J. Stoemenos, Vaccum 36 (11/12). page 877 (1986).
[87] - C.D. Marsh, G.R. Booker, K.J. reeson, P.L.F. Hemment, R.J. Chater, J.A. Kilnerand G.K. Celler, European SOI 88 Workshop, Meylan, France (March 1988).
[88] - G.F. Cerofolini, S. Bertoni, L. Meda and C. Spaggiari, Nucl. Instr. & Methds. B8Apage 234 (1994)
[89]- A. Bourret, J. Thibault Desseaux, D.N Seidman, J. Appl. Phys. 55(4), page 825 (1984).
202 References
[90] - F. Shimura, H. Tsuya, T. Kawamura, Appl. Phys. Lett. 37, page 483 (1980).
[91] - G.F. Cerofolini and L. Meda in "Physical Chemistry of, In and On Silicon, Chp. 3,Springer Verlag, Berlin (1989).
[92] - Y. Li, J.A. Kilner, Mater. Sci. Eng. B I2, page 77 (1992).
[93] - B. Aspar, C. Pudda, A.M. Papon, A.J. Auberton Herve, J.M. Lamure,Proc. Electrochem, Soc. SOI conf., page 62, (1994).
[94] - G.F. Cerofolini, S.Bertoni., P. Fumagalli., L. Meda, C. Spaggiari, Phys. Rev. B47,page 10 175, (1993).
[95] - S.Spaggiari, S. Bertoni., G.F. Cerofolini, P. Fumagalli and L. Meda,Ceramics International 19, page 399 (1993).
[96] - I.M. Lifshtz and V. V. Slyozof, J. Phys. Chem. Solids, 19(112). page 33 (1961).
[97] - J. Vanhellemont and C. Claeys, J. Appl. Phys. 62(9). page 3960 (1987).
[98] - A. De Veirman, J. Van Landuyt, J. Vanhellemont, H.E. Maes, IC Yallup,Vaccum 42(5/6) , page 367, (1991)
[99] - T.Y. Tan and U. Gosele, Appl. Phys. A37. page 1 (1985).
[100 ] - M.P. Maszara, J. Appl. Phys. 64, 123 (1988).
[101] - D. A. Porter and ICE. Easterling, in " Phase transformations in Metals and alloys,Van Nostrand Reinhold Co. , page 115, (1981).
[102] - W. Ostwald, Z. Phys. Chem.. 34. page 495 (1900).
[103] - K. Wada, N. Inoue, K. Kohra, J. Cryt. Growth 49 , page 749 (1980).
[104] - A. Yoshimo, Nucl. Inst. Methds. B39. page 204 (1989).
[105] - L.C. Brown, Scripta Metall. 21. page 693 (1987).
[106] - F. Gibbons, Proc. IEEE 60, page 1062 (1972).
[107] - M.W. Thompson in " Defects and Radiation Damage in Metals", page 45,Cambridge University Press (1969).
[108] - C. J. Gallagher, Phys. Rev. 88, page 721 (1952).
[109]- G. Thomas and J. Washburn, Rev. Mod. Phys., 35(4). page 992, (1963).
[110] - A. Gross, H. Teichlker, Phil. Mag. B64(4). page 413 (1991).203 References
[111] - J. Silcox, P. Hirsch, Phil. Mag. 4, page 72 (1959).
[112] - R. M. Cotterill, P h il. Mag. 6, page 1351 (1961).
[113] - G. Saada, Acta Met. JO, 551, (1962).
[114] - G.F. Cerofolini, S. Bertoni, S. Fumagaili, L. Meda, C. Spaggiari, Phys. Rev. B47(1993).
[115]- S.M. Hu in "Oxygen, Carbon, Hydrogen and Nitrogen in Crystalline Silicon", (edited by S.J. Pearton, J.W., corbett and S.J. Pennycook) Mat. Res. Soc. Proc. 59,249(1985).
[116] - F. Seitz, Phys. Rev. 79, page 723-724 (1950).
[117] - J. Silcox, M. Whelan, Phil. Mag._5, page 1, (1960).
[118] - IBIS Technology Corporation, 32A Cherry Hill Drive, Danvers,MA 01923-2583, USA
[119] - LETI- Site technologique ASTEC, Avenue des Martyrs, 38041 Grenoble, France
[120] - N. Meyyappan, T. Nakato, Proc. IEEE SOI Conf., Late News paper, (1993).
[121] - C. Vieu, A. Claverie, J. Faure and Beauvillain, Nucl. Instr. & Methds. B28.page 229 (1989).
[122] - S.M. Sze in " VLSI Technology", Me Graw Hill International Editions,page 121, (1988).
[123] - F. Secco d'Aragona, J. Electrochem. Soc., 119(7), page 947, (1972).
[124] - D.G. Schimmel, J. Electrochem. Soc. 126(3). page 479, (1979).
[125] - K.H. Yang, J. Electr. Soc., 131(5). page 1140, ( 1984)
[126] - Wright Jenkins, J. Electrochem. Soc., 124(51 757, (1977).
[127] - J.W. Edington in " Electron Diffraction in the Electron Microscope"( Macmillan Philips Technical Library) (1974).
[128] - A. Robinson, R. Webb, Internal Communication, University of Surrey (1993).
[129] - J.J. Gilman, W. G. Johnston, W. Sears, J. Appl. Phys. 29(5). page 774 (1958).
[130] - C. Marsh, private communication.
[131] - B. Deal and A.Grove, J. Appl. Phys. 36, page 3770 (1965).
204 References
1331 - P. Fahey, P. Griffin and J.Plummer, Rev. Mod. Phvs. 61(2). page 289-348,(1989).
134] - L. Laanab, C. Bergaud, M.M. Faye, J. Faure, A. Martinez and A. Claverie,Proc. Mat. Res. Soc. ( Boston ,1992),
135] - B.de Mauduit, L. Laanab, C. Bergaud. M.M. Faye, A. Martinez and A. Claverie, Nucl. Instr. 8c Methds., B84. page 190 (1994).
136] - L. Laanab, PhD thesis, Laboratoire d'Optique Electronique- CNRS , Toulouse(1993)
1371 - J. Vanhellemont and H, Maes, Vaccum 42(5/6), page 359, (1991).
1381 - A. Nejim, private communication.
1391 - P.L.F. Hemment, private communication.
1401 - S.Nakashima and K. Izumi, Nucl. Inst 8c Methds. B55. page 847 (1991).
1411 - J. Miltat, Phil. Mag. A55(5). page 543 (1987).
1421 - J. Miltat, Phil. Mag. A57(4). page 685 (1988).
1431 - C. Marsh, PhD Thesis, University of Oxford (1993).
1441 - S.MHu, J. Appl. Phys._45, page 1567, (1974).
1451- Y. Tan and U. Gosselle, Appl. Phys. Lett. 39, page 86 (1981).
1461 - A. M. L in , R.W. Dutton, D. A. Antoniadis and W. Tiller., J. Electrochem. Soc. 128, page 1121(1981).
147] - S.T. Dunham and J.D. Plummer., J. Appl. Phvs. 59 (7) , page 2541 (1986).
1481 - G. Brebec, R. Seguin, C. sella, J. Bevenot, C. Martin, Acta Met. 28, page 327,(1980).
1491 - S. Visitserngrtrakul, C.O. Jung and ICS. Krause, B.F. Cadts,Proc. Electrochem. Soc. SOI Conference, page 106, (1990).
150] - T. Y. Tan, H. Foil, S.M. Hu, Phil. Mag. 44A, page 127 (1981)
1511 - N. Guillemot, O.Tsoukalas, C. Tsamis, J. Margail, A.M.Papon and J. Stoemenos, J. Appl. Phys., 71(4). page 1713, (1992).
1521 - T. Y. Tan, Phil. Mas. 44A. 101 (1981).205 References
132] - B. Leroy, J. Appl. Phys. 50(12). page 7999 (1979).
[153] - W.K. Burton, N. Cabrera and F.C. Frank, Phil. Trans. Roy. Soc., A243. page 299(1950).
[154] - D.N. Seiman, R.W. Balluffi, Phys. Rev. J3 9 , A1824 (1965).
[155] - Kimura, Maddin, Wilsdorf, Acta Met._7, page 145, (1959).
[156] - J.D. Lee, J.C. Park., D. Venables, S.IC. ICrauseamd P. Roitman, Mat. Res.Soc.Symp. Proc. 316 , page 758 edited by the Materials Research Society (1994)
[157] - J.P. Hirth and J. Lothe in " Theory of Dislocations", page 315,Jonh Wiley & Sons publication (1982).
[158] - D. Venables, private communication.
[159] - J. Stoemenos, B. Aspar and J. Margail, Proc. Electrochem . Soc. SOI Conf.(1994).
[160] - D. Venables, S.K. Krause, S.J. Park, J.D. Lee, P. Roitman, Proc. IEEE SOIConf., 49, (1993)
[161] - M. Jong and J.S. Koehler, Phys. Rev. 129(1), page 49, (1963).
[162] - A. De Veirman, K. Yallup , J. Van Landuyt, H.E. Maes, S. Amelinckx, Inst. Phys.Conf. Ser. N° 87, page 403 (1987).
[163] - W. Coene, H. Bender and S. Amelinckx, Phil. Mag. A52(3). page 369, (1985).
[164] - G. Booker and R. Stickler, J. Appl. Phys. . 33(11), page 3281, (1962).
[165] - J.S. Koehler and C. lund - in "Lattice defects in quenched metals, pag. 1, editedby R.M.J. Cotterill, M.Doyama, J.J. Jackson., M. Meshi, Academic Press (1965).
[166] - R. M Coterill and M. W. Jones, J. Phys. Soc. Japan 18 , page 158 (1963).
[167] - I.R. Sanders, P.S. Dobson, Phil .M ag., page 881, (1969).
[168] - S. T. Dunham and J. Plummer, J. App. Phys. 59(71 page 2551 (1986).
[169] - L. Meda, S. Bertoni, G.F. Cerofolini and C. Spaggiari, Nucl. Instr. & Methds.B80/81. page 813 (1993).
[170] - A. Nejim, C.D. Marsh, L.F. Giles, P.L.F. Hemment, Y. Li, R. J. Chater,J.A. Kilner and G.R. Booker, Nucl. Intrs. & Methds. B84 , page 248, (1994).
[171] - K. Maex, L. Van den Hove. Mat. Sci. Eng., B4, page 321 (1989).
[172] - J. Vanhellemont, C. Claeys, J. Van Landuyt, Mat Sci. Forum 10/12. page 757(1986).
206 References
[173] - P.L.F. Hemment, Vaccum 29111/12L page 439 (1979).
[174] - J. Stoemenos, in " Crystalline Defects and Contamination: The impact and controldevice Manufacturing", Proc. Satellite Symp., page 130 Electrochemical Society,(1990).
[175] - R.J. Schreutelkamp F. W. Saris, J. Custer, R. Liefting, R. wijburg, H. Wallinga,Microelec. Eng. 19, page 352 (1992).
[176] - R. Falster in "Crystalline Defects and Contamination: The impact and controldevice Manufacturing", Proc. Satellite Symp. Electrochemical Society, page 149,
Grenoble (1994).
207 References
Appendix A
SOI technologies
During the last thirty years several SOI technologies have been developed. Among the
most important ones we can named, for instance, Silicon on Sapphire[l](SOS), Full
Insulation o f Porous Oxidized Si[2] (FIPOS), Zone Melt Regrowth (ZM R)[3], Epitaxial
Lateral Oxidation[4] (ELO) , Wafer Bonding[5] and Separation by IMplanted OXygen
(SIMOX).
In Silicon on Sapphire (SOS) a thin epitaxial film o f silicon is grown on a single-crystal
alpha-aluminum oxide (sapphire) substrate by chemical vapour deposition (CVD). SOS
was the first SOI technique applied to VLSI, however, due to the low quality o f the
silicon overlayer particularly near the Si/Al20 3 interface,it is now being replaced by
alternative substrates.
In Full Insulation of Porous Oxidized Silicon (FEPOS), porous silicon is formed by
electrochemical etching silicon islands in a HF aqueous electrolyte. The oxidation rates of
porous and monocrystalline silicon differ by an order o f magnitude and therefore, during
oxidation, the oxide is formed mainly underneath the silicon islands. The practical
drawback of this approach is that the electrochemical etching necessary to form porous
silicon formation is not compatible with other IC fabrication steps. However, interest has
recently shifted to the study o f light emission from porous silicon. As a consequence its
importance as a SOI material has decreased considerably.
In Zone Melt Regrowth (ZMR), the process starts with deposition of polycrystalline
silicon over S i02. Then the film is implanted with silicon. A channelled silicon ion beam
208 Appendix A
amorphises most of the polysilicon layer leaving intact grains with major crystallographic
axes that are aligned along the channelling direction. These grains are used as seeds for a
solid-state regrowth, which take place during a subsequent thermal treatment. The main
drawback of the technique is the poor quality of the silicon overlayer and residual defects
at the grain boundary interfaces.
Epitaxial lateral overgrowth (ELO) , consists of the growth of silicon by chemical
vapour deposition (CVD) within windows in a SiOz mask. If no nucleation of silicon on
the Si02 layer takes place, the silicon grows vertically to the mask level and then laterally
over the Si02 mask. The lateral growth enables small SOI islands to be formed. The main
drawback of the technique, is indeed, the small size of the islands.
The formation of SOI substrate by W afer Bonding uses the principle that two surfaces
in close contact may be bonded (Van der Walls forces) during a high temperature anneal.
This basic idea has evolved towards silicon wafer bonding with a thermal Si02 layer on
top. The main drawback of the technique is that wafer thinning processes have to be
applied, and consequently, due to its limitations, lateral non-uniformities of the thickness
of the Si overlayer will occur. However, during the last few years a new thinning process
based on plasma etching[6] has been developed, and is now being used as the main
method to produce thick (>1.0 pm) SOI substrates.
In SIMOX the formation of a buried Si02 layer is obtained by high energy and high
dose 0 + implantations. To facilitate dynamic annealing the wafers are maintained at a high
temperature (> 600°C) during O* implantation. Due to the radiation damage caused by
the incident 0 + ions ions, a post-implantation thermal annealing is needed both to
annihilate the defects and to redistribute the implanted oxygen. The presence of residual
209 Appendix A
defects ( dislocations, stacldng faults, and SFT) in the silicon overlayer are the main
disadvantage o f this technique. In Table 1 we summarise the main advantages and
drawbacks o f the different SOI technologies[7].
Technique Advantages Drawbacks
Silicon on Sapphire
(SOS)
1. Conventional processing2. First SOI applied in production of LSI/VLSI circuits.3. Application for : CMOS
1. Cannot be used for 3D integration.2. Long lifetime- no bipolar applications.3. High substrate cost.
Zone melt and regrowth (ZMR)
1. Conventional processing2. Application: 3D
1. Limited applicability due to the smallness ! of the stripes
Full Insulation of Porous Oxidized Si
(FIPOS)1. Low defect density.2. Application for: CMOS, bipolar.
1. Process noncompatible with IC processing.2. Limited applicability due to smallness of the islands.3. Cannot be used for 3D integration.
Epitaxial Lateral Oxidation
(ELO)
1. Conventional processing.2. low defects/warpage3. Applications : bipolar-, 3D integration.
1. Thinning necessary for CMOS high speed circuit.2. Low overgrowth ratio ( small areas)3. Seeding necessary.
Wafer Bonding1. Low defect density2. Application: 3D integration. 1. Thick SOI layers (1-5 pm)
Separation by Implanted OXygen
(SIMOX)
1. Conventional processing.2. Good quality SOI: low defect density/warpage3. Stage of development mature (complex circuits demonstrated)4. Application for : CMOS, bipolar.
1 .Residual defects in the Si overlayer.
Table 1 - Main advantages and drmvbacks o f SO I technologies.
References
1. M. Aindow J.Am. Cerm. Soc., 73(5). 1136 (1990)
2. IC. Imai, H. Unno and H. Takaoki, J. Crystal Growth 63, 547 (1983).
3. D. Kwizera, P. Reiff and D. Gandomenico, 1982 Electronic Materials Conference, Ft. Collins, CO, (1982).
4. L. Jastrzebsld, J. Crystal Growth 63, 493 (1983).
5. J. Haisma, G.A.C. Spierings, U.IC.P. Biermann, Japan. Journal Appl. Phys. 28(8),1426 (1989).
6. P.J. Clapis, A M . Ledger, IC.E. Daniell, Proceedings 1993 IEEE International SOI conference, pag.66, (1993).
7. L. Jastrzebsld, J. Crystal Growth 70, 253 (1984).
210 Appendix A
Appendix B
Chemical defect etching
1. Basic principles
A) types o f reactions
Chemical etching can be described as the dissolution process o f a material into an
undersaturated medium. The term dissolution covers many types o f reactions. Focusing
our attention to the dissolution o f solids in liquids, we can name three different dissolution
processes. Firstly, we can have the dissolution of a molecular crystal into an organic
solvent. In this reaction intermolecular crystal forces are overcome by solute-solvent
interaction, and although the environment o f a molecule in solution is different from that
in the crystal the molecule itself is unchanged. A second type o f dissolution is encountered
with ionic crystals. The characteristic properties of the solid result from the electrostatic
interaction of the regular array o f ions. In solution, the ions are separated and properties
characteristic of the solid are lost. Thirdly, there is dissolution brought about by chemical
change. More frequently oxidation is required to convert the solid into a soluble
material. This is expressed in general terms as the extraction o f electrons from the
material. The resultant material has then to be removed from the surface, usually with the
accompaniment of dissolution. This land o f reaction is the most common in
semiconductors and metals and therefore, it is the most important for the work discussed
in this thesis.
The rate o f dissolution o f a solid (semiconductor) in a given etchant is primarly
controlled by the speed of transfer o f the active ions in the solution to the dissolving
211 Appendix B
namely an initial oxidation followed by dissolution o f the oxide, the effectiveness o f the
oxidation will depend on the ability o f the surface to adsorb the oxidising ions in the
solution. This will undoubtedly be influenced, among other things ,by the atomic nature of
the surface, such as crystallographic orientation.
B) Kinematic description.
The kinematic aspects of crystal etching is generally considered to proceed through the
retreat o f monomolecular steps across the surface of the crystal[ 1], The active sites are
taken to be the places along the steps where single molecular rows end. These places,
called "kinks", are where individual molecules may be removed. When a perfect crystal
face is exposed to an etchant, dissolution begins by the nucleation of unit pits of one
molecule depth. These units pits grow as steps retreat across the crystal through the
action of the kinks. On such a perfect crystal face the nucleation of unit pits is random.
On a real crystal, defect such as dislocations may be preferential sites for nucleation o f
unit pits and repeated nucleation at a dislocation leads to the formation o f an etch pit. The
formation of visible dislocation etch pits depends on the nucleation rate for unit pits at a
dislocation and the rate of motion o f the steps across the crystal surface. These two
quantities are reflected in the step velocity normal to the surface at a dislocation Vn and
parallel to the surface V, ( see Fig. 1).
212 Appendix B
Figure 1 - Schematic showing the step velocities near a dislocation.
C) Nucleation Process
The driving force for the nucleation o f an etch pit is the energy around the dislocation.
This 'localized' energy consists o f the core energy plus a fraction of the total elastic strain
energy. A model to determine the general features and important parameters of such a
nucleation process was proposed by Cabrera et al. [2], In their model, they assumed that
the nucleus lies along the core o f the dislocation and that a cross-section o f the nucleus
perpendicular to the dislocation is circular.
The change of free enthalpy due to the formation o f a two-dimensional nucleus is given
by the equation[3]:
AGr - AGb - AG/r - AGz (B l)
213 Appendix B
The term AGB describes the work required for the formation of the (circular) periphery
of the pit and is given by the expression:
A Gb = 27trp (B2)
where r is the radius of the pit and p the specific free boundary tension. AGF is the
work gained by the transition from the solid to the solute,
with Q = atomic volume h = height of step
Ap = chemical potential difference between crystal and surrounding crystal.
Finally, AGZ describes the gain in strain energy and depends largerly on the type of
dislocation, the distance from the dislocation line, r, and the radius of the dislocation core
AGz = A h In (^) + B (r> r0) B4)
For a pure edge dislocation, A, is given by:
(B5)
and for a pure screw dislocation is:
(B6)
The term B in eq. B4 is the core energy and it must be considered because the driving
force for the formation of an etch pit is actually the release of the core energy
214 Appendix B
immediately at the point of emergence of a dislocation. The relationship between AGR
and r is shown graphically in Fig. 2.
AG
Figure 2 - Free enthalpy o f formation o f a p it near a dislocation site (AGR ) and free enthalpy offormation o f a p it in a perfect surface (A G ^J .
Rewriting Eq. I, we obtain for a screw dislocation:
AG = 2nrp -A esita fc L )£2 4n V/V (B7)
The first d e r i v a t i v e , y i e l d s two solutions when equated to zero ( the two equilibrium
points):
p • £2 i------= 2 A ^ ( ‘ + m
and:
(B9)
215 Appendix B
K Gb2h 2 Ap where: * =
For £, = 1, we obtain:
p * H rr = r . — ----- = — (BIO)
max mm 2 A \ x h 2 v }
The critical radius of the nucleus, rc, is derived from the treatment o f a dislocation-free
surface as follows:
AG ideal ~ AG B ~ AG/r=27irp - — ( Bl l )
(d ^ G )n=rc ~ o provides: dr
p£2rc -
A\ih
Eq. ( Bl l ) yields with rc , the following expression:
AG, = ^ 2 (BI2)Ap/7
For £,< 1, the critical free enthalpy for nucleation at the site of a screw dislocation is:
AGJ = AGrmax - AGrmm (B13)
and with Eqs. (B8) and (B9) we obtain:
AG* = Kprc j T J + f \ ) (B14)471 l + j l +Z ,
for r > r0 and < 1.
216 Appendix B
Equation (B14) yields the decrease of the work of nucleation at a dislocation site
compared with a perfect surface when setting £<1. Then:
AG'R = AGueoi / l D (B15)
The numerical value of £ is a function of the undersaturation.Thus etch pits occur when
the undersaturation in the immediate vicinity of the surface is smaller than a critical
undersaturation which causes spontaneous dissolution at a dislocation site. The presence
of inhibitors decrease the numerical value of £ , thus, decreasing the free enthalpy
required for nucleation.
2. Silicon etchants.
Several preferential etchants can be used to delineate crystallographic defects in silicon (
see table B l). The dissolution process of some of them are briefly discussed below.
A) The CKX-ELO-HF and K X r,O z-ELO-HF systems.
The dissolution process in CrC^-E^O-HF and K^CqO^H jO-HF etchants is based on the
oxidation of the silicon by the oxidizing agent ( C r03 or K2Cr20 7) and the subsequent
dissolution o f the oxide by the HF acid. For the CrC^-HF-HjO system (
Sirlt,Shimmel,Yang etchs) the behaviour of the etchant in terms of preferential attack and
formation of well defined etch pits, will depend on the concentration ratio m (given in
term o f %/M )[3], If m is too high or even too low, the etchant will lose both its
capability o f preferential etching and to form well defined etch pits, respectively. This can
be explained by the fact that the formation of a dislocation etch pit arises mainly from the
217 Appendix B
strain field associated with a dislocation ( see heading 1C). The strain field produces a
surface energy difference, AGr ,between the site of the dislocation and the surrounding
perfect crystal.If the C r03 concentration is too high the strong oxidising power of the
solution tends to minimise the surface potential to a negligible extent. As a result
preferential etching does not occur. Instead, if the concentration o f C r0 3 is too low,
etching due to HF dissolution is the dominant effect and the etch pits become ill defined.
The same principle applies to the K2Cr20 7 system ( Secco etch). However, due to the
low solubility of K2Cr20 7 in water, the concentration of the oxidizing agent can not rise to
a value high enough to give well defined etch pits. This is one of the reasons why the
Secco etch gives rise only to isotropic (circular) etch pits.
B) The HNQ3-HF-H2Q system
The dissolution mechanism o f silicon in HF-HN03 -H20 occurs mainly through an
electrolytic mechanism that allows the corrosion of the silicon by the formation o f local
anodic and cathodic sites on the surface o f the semicoductor[4].In addition to the
electrolytic attack, a chemical reaction also takes place on the silicon surface . The
electrolitic reactions on the cathodic and anodic sites may be written as:
4 e+ + Si + 6F —» SiF6“
4 e + 8 FT + 4 NOa-> N20 + 4 Hfi
The local chemical attack is a reduction-oxidation reaction followed by the dissolution of
the silicon dioxide formed. These reactions can be written a s :
218 Appendix B
Si + 4 HN03 —> S i O ^ N O ^ H ^ O
Si02 + 6 HF f t HjSiFg + 2 H20
The formation o f well defined etch pits will depend on the nature and quantity of the
inhibitors in the solution. The role of impurities is to decrease the shift velocity vt in order
to obtain a very small ratio v /vn ( see Fig. 1) giving rise to well defined etch pits ( see
Fig. IB).
Silicon Etchant Formula Characateristics
Secco [5]2 H F (49% ): 1 K2Cr20 7 (0.15 M)
Preferential etching in all Si planes.Etch pits not well defined.
Schimmel( [6] 2 H F(49% ): 1 Cr0 3 (1 M) Preferential etching in all Si planes. Etch pits not well defined.
Sirtt [7]4 H F (49% ): 5 Cr03 solution ( 50 g Cr0 3 for 100 g D.l. H20 )
Referential etching in (111) and ( 110) planes.Etch pits well defined.
Yang [3]1 H F(49%):1 CrO a solution (1.5 M)
Preferential etching in all planes.Etch pits well defined.
Wright [8]60 ml H F (49%):30 ml H N O 3:30 ml CfO3:60 ml HAc: 60 ml H20
Preferential etching in (111) and ( 110).Etch pits well defined.
Dash [9]1 H F (49% ): 3 HNO3(70% ): 10 HAc (glacial)
Preferential etching in all planes.Etch pits not well defined.
M EM C [10] 36 H F (49%):25 H N O 3(70%):18 HAc(glacial):21 H20 plus 1 g of Cu (N03)2:3 H20 per 100 ml of mixed acid.
Preferential etching in all planes.Etch pits well defined.
Table B1 - Preferential etchants for silicon.
219 Appendix B
References
[1J - J J . Gilman, W.G. Johnston, W. Sears, J. Appl. Phys. 29(5), 774 (1978).
[2] - N. Cabrera, M. Levine, J.S. Plashett, Phys. Rev., 96,1153 (1954).[3]- K.H. Yang, J. Electrochem. Soc. 131,1140, (1984).
[ 4 ] - H. Robins and B. Schwartz, /. Electrochem. Soc. 106(6), 505 (1959).
[5]- F. Secco d'Aragona, J. Electrochem. Soc., 119(7) 458, (1973).
[6]- D.G. Schimmel, J. Electrochem. Soc. 126(3) 479, (1979).
[7] - E. Sirlt, A. Adler, Z. Metallk, 52 , 529 (1961).
[8] - Wright. Jenkins, J. Electrochem. Soc., 124(5). 757, (1977).
[9J- W.C. Dash, J. Appl. Phys., 27, 1193 (1956)[10]- T.C. Chandler, J. Electrochem. Soc., 137(3), 947 (1990).
220 Appendix B
A p p e n d ix C
Dislocations and plastic flow in the Diamond Cubic structure
1. Introduction.
The first observation of the plastic flow of silicon and germanium crystals was made by
Gallagher [1] in 1952, They observed that although these crystals were completely brittle
at room temperature, they become increasingly ductile above 60% of their absolute
melting temperatures, that is about 450°C for Ge and 750°C for Si. Since the plastic glide
o f crystals operates by the movement o f dislocations, they attributed the low-temperature
brittleness o f diamond structure crystals to the high frictional resistance in shearing the
directional covalent bond. This frictional force against the movement of dislocations is
generally known as the Pierls-Nabarro force [2].
2.Lattice defects in the Diamond structure.
A.Geometry
The diamond structure (D.S.) [3] can be derived from the fee structure by doubling every
atom of the fee and shifting the second lattice by a/4[l 11] (see fig. 1 a). Thus, the
close-packed {111} planes o f the fee lattice become pairs o f {111} planes in the diamond
structure as is shown in Fig. 1 b , or in a two-dimensional projection, in Fig. Ic.
221 Appendix C
[ t i l ]
Fig. 1 (a)- The cubic unit cell o f the D.S. (b) Stacking o f (111) planes in the D.S.
c) projection o f fig. lb on the (110) plane.
It is observed, that if ABCABC is the stacking order of the {111} planes in the first fee
sublattice of the D.S, and abcabc is the stacking order of {111} planes in the second
produced by the a/4[l 11] shift from the fee, then aB, bC, cA, are closely spaced, double
planes connected by three times as many bonds as are between the three times wider
spaced Aa,Bb, Cc pair o f planes. Since the Bravais lattice of the D.S. still is the fee, the
slip system of the D.S, is expected to be similar to that of the fee. Therefore, one expects
the D.S. to glide on {111} planes in <110> directions. Indeed, experimental results have
222 Appendix C
shown that the {111} planes are the slip planes in silicon[lJ and germanium[l] and that
the slip direction is <110>.
.B. Perfect dislocations in the D.S.
Perfect dislocations in D.S. have b= | - < 1 1 0 > Burgers vectors and {111} glide
planes[4]. The structure of perfect dislocations in the D.S. structure was first considered
by Honstra [5] in 1958. He described various types of dislocations such as the 60° and
the screw dislocation, among others. Fig. 2 a) shows the representation
made by Honstra of the 60° dislocation (60° is the angle between the dislocation line £
and the Burgers vector b). It is observed that a dense row of "dangling" bonds are
present along the edge o f the extra half plane. Fig. 2 b) shows the representation made by
Honstra of a 'screw dislocation. It is observed that the screw dislocation, contrary to the
60° dislocation, has no "dangling" bonds .
Fig. 2 a- continued
223 Appendix C
Fig. 2 a)-60° dislocation and b) Screw dislocation in D.S. Heavy lines indicate the extra half-plane. (According to Honstra[5])
More recently, Hirth and Lothe[4] proposed that, due to the double occupancy o f the fee
bravais lattice, there are two different sets of dislocations in the diamond cubic lattice,
with one member of each set having a Burgers vector identical to one dislocation in the
other set.
In order to see the distinction between these two sets of dislocations let us consider the
simple case of a 60° dislocation ( see Fig. 4 a). From' Fig. 4 a, it is observed that the extra
plane terminates between layers of the same letter index, i.e. B and b. Therefore, glide of
this dislocation will occur between layers B and b ( or equivalent pairs). However, it is
also possible that the extra plane, instead o f terminating between layers of the same letter
could terminate between layers of different index, for example between layers a and B
(see Fig. 3). In this case, glide of this dislocation will occur between layers a and B.
Hirth and Lothe, named these two different set of dislocations as being the "glide"
(dislocations which glide between layers a and B) and the "shuffle" ( dislocations which
glide between layers B and b) set. According to Hirth and Lothe, perfect dislocations of
224 Appendix C
the "glide" set can dissociate into glissible Shockley partials[6] whilst perfect dislocations
of the "shuffle"set can not dissociate directly into partials, because such dissociation
would produce a high-energy fault of the type:
CcAaB 1 ciBbCc
F i g . 3 - 6 0 ° d i s l o c a t i o n o f t h e " g l i d e " s e t ( A c c o r d i n g t o H i r t h a n d L o t h e [ 4 ] )
C. Partial dislocations
Frank[6] suggested that the stability of a dislocation depended on the strain energy
associated with it and that, in a first approximation, the energy associated with a
dislocation was proportional to the square o f its Burgers vector. He, therefore,
proposed that a dislocation with Burgers vector b2 could dissociate into two other
dislocations b2 and b3 if:
b\ > b \ + b \ (C l)
Appendix C225
The dissociation o f perfect dislocations into partial ( imperfect) dislocations occurs as a
method to reduce the energy of the dislocation according to the Frank criterion. The
dissociation process always creates a narrow stacking fault between the two partials. This
configuration is generally known as an extended dislocation. ( see Fig. 4)
Eq. (C l) is generally known as the Frank energy criterionf 6],
Fig. 4 - Dissociation o f a perfect dislocation into Shockley partials.
In the D.S. structure, Shockley partials[7] can be created by the dissociation of a perfect
(b < 110 >) dislocation o f the 'glide' set (see heading B).
The dissociation reaction can be written as:
i [ i i o ] ^ i [ 2 n ] + i [ i 2 J ] (C2)
Perfect Shockley Shockley
Another possible dissociation reaction, is the formation of a Frank sessile[8] and a
Shockley partial. This reaction can be written as:
^ < 110 >—> | < 111 > + | < 2 1 1 > (C3)
Perfect Frank Shockley
226 Appendix C
Experimental evidence o f the existence of extended dislocations in silicon [9] and
germanium [10] has been widely reported, confirming that dissociation of perfect
dislocations occurs in D.S. crystals.
Thompson tetrahedron
In fee and similar structures (such as the D.S)., dissociation reactions can be easily written
using the notation proposed by Thompson[ 11].
Thompson suggested that by joining the atoms in a one-eighth unit cell by straight lines, a
imaginary tetrahedron ( see Fig. 5) containing the four possible {111} glide planes could
be formed.
Fig. 5 - A one-eigth unit cell o f edge length l/2a0 o f an fee lattice, showing the Thompson tetrahedron ABCD.
From Fig. 5, it is easily observed that the edges of the tetrahedron correspond to the six
<110> glide directions of the fee structure. The atom at the origin is labelled D and the
others are labelled ABC in the clockwise order indicated in Fig. 5. The midpoints o f the
faces opposite to A,B, C and D are labelled a,p ,y227 Appendix C
and 8 respectively. If the tetrahedron is open up at D, it can be folded out into the planar
arrangement displayed in Fig. 6.[11]
As it is clear in Fig. 6, the letters a,b,c and d represent the {111} planes; the perfect
dislocations \ < 110 > are represented by pairs of Roman letters- for example y[011]
becomes BD; and Shockley partials are represented by Greek-Roman pairs, for example
|[211] becomes D(3. Finally, the Frank partials are represented also by Greek-Roman letters, for example j [ l 11] becomes 5D.
D
Fig 6- A Thompson tetrahedron opened up at a com er D.
228 Appendix C
Formation o f a SFT from a Frank loon.
Silcox and Hirsch [12] proposed that the stacking fault tetrahedron could be formed by
the expansion of a triangular Frank dislocation loop ( 5D) ( see Fig. 7). They proposed
that the <110> sides o f the triangular Frank loop could dissociate according to the
reaction:
6 D - + 5 P + PD (C4)
Frank stair-rod Shockley
where 5p (b= ^[110]) is a low energy stair-rod dislocation[12] and pD is a Shockley
partial. This dissociation reaction can take place at the three sides o f the triangular loop,
yielding the configuration of Fig. 7 b. According to Silcox and Hirsch the Shocldey
partials bows out in their slip plane as they are repelled by the stair rods. Taking into
account the fact that opposite ends o f the loops are o f opposite signs , the Shockley
partials attract each other in pairs to form stair-rods along DA, BA and CA according to
the reaction:
.pD + y D- >yp (C5)
Shockley Shocldey stair-rod
The attraction forces lead to the development o f configurations depicted in Fig. 7 c and
Fig. 7 d . Therefore, the final result is a tetrahedron bounded completely by stacldng
faults and stair-rod dislocations.
229 Appendix C
D
Fig. 7 - Various stages (a-d), in the formation o f a stacking-fault tetrahedron from a Frank dislocation loop (according to Silcox and Hirsch[12] )
230 Appendix C.
R eferences
[1] - C.G. Gallagher, Phys. Rev., 88, 721, (1952)
[2] - R .E. Peierls, Proc. Phys. Soc. 52, 23, (1940).
[3] - H. A lexander and P. H aasen, S o lid S ta te Physics Research, 28, page 2 2 (1968)•
[4] - J.P. H irth a n d J. L o th e in " Theory o f D islocations" second edition,
Ed. Wiley & So n s (1982 )•
[ 5 J -J . H onstra, J. Phys. Chem. Solids, 5 ,1 2 9 (1958).
[6 ]- F.C. F rank, Physica, 15 ,131 , (1949).
[7J- R.D. H eindenreich a n d W. Shockley in "report o f a Conference on Strength o f
So lids" Physical Society, L ondon , p a g e 57 (1948).
[8] - K. V Ravi in " Im perfections a n d Im purities in Sem iconductor S ilicon " ( Wiley,
N ew York, 1981). page 101.
[9] - E. Aerts, P. Delavignette, R. S iem s a n d S. A m elinckx, J. A p p l Phys. 33,
page 3078 (1962).
[10] - A . Art, E. Aerts, P. D elavignette and S. Am elinclix, Appl. Phys. Lett. 2, page 40,
(1963).
[11] - N . Thom pson - Proc. Phys. Soc. 66B, 481 (1953).
[12] - J. S ilcox a n d P.B. H irsch, Phil. M ag. 4, 72 (1959).
231 Appendix C
A p p e n d i x D
P r e p a r a tio n o f th in s a m p le s f o r T E M o b s e r v a tio n s .
1. Preparation o f silicon sam ples fo r plan vieiv observations.
The thin film silcon samples analysed by plan view TEM were prepared by the "float o ff1
methodfl]. Briefly the following sample preparation procedure was used:
1. Cleave SIM O X wafers into sm all samples ( typically 5x5 mm).
2. Immerse the S IM O X samples in HF(40% ) to dissolve the B O X layer (typically 8 hrs.)
3. R inse in D I water.
4. The silicon thin film flo a ts to the surface a n d is scooped up using a 3 mm Cu grid.
5. Remove the excess water using fil te r paper.
6. D ry in air and place a second g rid on top o f the support g rid and glue them with a
sm all spot o f silver dag.
232 Appendix D
The samples prepared for XTEM analyses were thinned to electron transparency using ion
beam milling[2]. Briefly, the procedure was as follows:
1. Cleave S IM O X wafers into sm all samples ( typically 3x10 mm2).
2. Glue two samples together ( top surfaces in contact) using an epoxy resin.
3. Cut the g lued sample into 500 p m thin sections using a diam ond saw.
4. G rind and polish the 500 pm sections down to a thickness o f about 50 pm.
5. M ount the 50 pm thin section onto a 3 m m Cu g rid using silver dag .
6. Load the sample into the ion beam miller ( M odel: Gatan 600).
7. B om bard the sample using a dua l A r+ beam accelerated to 5 k e V with an angle o f
incidence o f 15° until electron tt'ansparency o f the sample is obtained.
References
[1] -P.J. Goodhew in "Electron Microscopy and Analysis", page 66. Wykeham Publications (1975).
[2] - P.J. Goodhew, in " Practical Methods in Electron Microscopy: Thin Foil Preparation for Electron Microscopy" Editor : A.M. Glauert, Elsevier Science Publishers (1985).
2. Preparation o f sam vles fo r cross section observations .
233 Appendix D
A p p e n d ix E
D e n s i ty a n d le n g th o f th e O I S F in th e s a c r i f ic ia l o x id i s e d S I M O X s a m p le s .
A list of all the sacrificial oxidised SIMOX together with the length and density of OISF is
presented in tables 1 to 8. We also list the thermal oxide thicknesses, the estimated
number of silicon atoms consumed by the oxidation (Nox), and the estimated number of
interstitials captured by the OISF(NsiOISF, see section 8.2.2 for more details). In order to
facilitate the interpretation of the da ta , we have grouped the sacrificial oxidised samples
in different tables corresponding to different SIMOX wafers ( table 1 to 5). Moreover, in
table 6, 7 and 8, we group the sacrificial oxidised SIMOX samples (1100°C for 4 hours)
pre-amorphised with BF+2, Si+ and Ge+ ions, respectively. From table 1 to 5, it is
observed that the NSl0ISF values increase with temperature and vary significantly for
different wafers ( see section 8.2.2 for more details).
234 Appendix E
g£o°S o
% X0-3
1!
a25.
......
..10.9
40.5
50.4
50.4
10.2
4 |
0.41
:0.6
20.8
30.5 - 0.5
10.8
52.7
82.2 1.5
4
CN 2.65
(1.3
4
So* 0O __j
2 x
3.25
8.25
3.75 ft ft 00 8.5 9.2 CO rt- 4.5
|
OO CN 2.2 4.25
5.87
6.25
6.7 7.5 00
OIS
Fle
ngth
(urn
)
+!CN rttr cd 10±2 CNrt
ftCN-Hox 10
±3 CNrtO rtrt- rtXOrtvqxd
CNrt00VOvS
ftrtVO
CNrtvoCOrtXOCN
CNrtXDCNco-HoxCN
CO-HrX
ftrt»r.
OISF
de
nsity
xlO
2(c
m'2)
rtV-VCNrt CNrtOX 5
CNrtcd
rtrO CN 4±1
5±1 CNrtvx
CNrtCO 4±1 CN
COCNrtvo
rpft 4±
2 COrtCOCNCNrtrt
CNrtW~i 4±1
Si ov
erla
yer
afte
rox
idat
ion
(A)
1530
665
1440 o0000
OOO00 700 OCNXO 530
1580
1405
|13
20 710
1750
1730
J13
60
1 0801 10
10 930
800
490
Ther
mal
Oxi
deth
ickn
ess
(A)
; 13
0033
0015
0028
00 oo00CN 3200
3400
3600
1200
1600
I 0081 32
00 ooooOvooo 17
0023
5025
0027
0030
0037
00
Oxi
datio
ntim
e(h
rs.)
! 28
.5 09 xd 00 VDCO vxCNCN 24 27 CN CO xO XO 0.5 0.6 xq 2.5 2.9 3.5 rr vx
Oxi
datio
nte
mpe
ratu
re(°C
)
ooo
006 950
950
950
950
950
950
1000
IOOO
1000
1000
00110011 11
0011
0011
0011
00
0011 1100
Waf
er 5
Sam
ple
lBl/1
lBl/2
lBl/3
lBl/4
lBl/5
lBl/6
IB 1/
7B
l/8B
l/9lB
l/10
Ibi/
iilB
l/12
[Bl/1
3lB
l/14
lBl/1
5B
l/16
Bl/1
7B
l/18
B1/
19B
l/20
235 Appendix E
Tabl
e 1
-Lis
t of
sacr
ifici
ally
ox
idis
ed
SIM
OX
prep
ared
fro
m wa
fer
B1
(ini
tial
thic
knes
s:
Ti =2
100A
).
fc2°©
f e X
43 156
r 22
234
232
01
562
|44
4[
406
| 77
384
6703
|
540
| 226
0 j
O
00 1545
422 00
00 1020
118
2332
4728
4232
70
|
fe*u v1>CN
"eo
O 2Z X
1.25
2.5 3 .4 4.26
| 8.2
5
C n r - r - 00
1 8‘5
. 1
9.2 rO fe 4.55 00 UV
CSco 5.75
;
CSc s 4.4
0000uv 7.5 oo
OIS
Fle
ngth
(u
m)
0.2
(S©-H
uv©41uv 41CS nuv ±100
uvo
CS-H00
CS-HOn
UVo41f eoo
CS43O
c s
3 41oo i4100
uvo41fecO 12
±0.6
c s
to? 18±3
uy4?uyuvc s
c o4100c s
c s41UVCO
CO-Hv oCO
OIS
Fde
nsity
xlO
5(c
m'2)
-Hw 41fe- 41fe 41uv 41co 541uvCS 2±
1 h3fe 5±0.7 2±1 41fe
feO41uvfe
UV©41NOCS
uvo
4
CS©41feCO
UVo41uvCS
nuvCOCS41CO
uyo41uvfe 5 41CO41uvCO
Si ov
erla
yer
thic
knes
s(A
) 2200
, 20
00j
1900
| 17
50j
1280 900
1300
1300
1100
1020
190
019
70 oo00 1720
1100
I | 0C6I 150
0 J
2130
1174
0 I
1450
I13
0012
0011
00
Ther
mal
Oxi
deth
ickn
ess
(A)
500
1000 Ouvcn 1700
3300
3600
2800
2800
3200
3400
;36
0012
00160
0 i
oooo 3200
1300
2300 850
1750
2350
:27
00300
0 ;
3200
Oxi
datio
ntim
e(h
rs.)
cs o 24)
09
! 74 OO
uvoo 22.5 24 27 CS co V© VC CS 4.5 0.6
9'I 2.5 3.5 fe UV
Oxi
datio
nte
mpe
ratu
re(°
C) 90
090
0 ooCN
! 90
090
0 006 950
950
950
950
950
1000
1000
1000
1000
I10
20 oUVo 110
0 I
1100
1100
1100
1100
1100
Waf
er
CS
Sam
ple
B2/1
B2/2
B2/3
B2/4
IB2/
5B2
/6 cs
B2/8
B2/9
B2/
10B2
/11
B2/
12B2
/13
B2/1
4IB
2/15
B2/
16B
2/17
IB2/
18IB
2/19
B2/20
i
B2/2
1IB
2/22
B2/2
3
236 Appendix
Tabl
e 2-
List
of
sacr
ifici
ally
ox
idise
d sa
mpl
es p
repa
red
from
wafe
r B2
( i
nitia
l th
ickn
ess:
Ti
=250
0 A
).
eftSu. Owo
.- o o Z x
2.6 :
00
11.12
voz 25.4 44 11.9
30.15
[25
.2127
.23 20.3
j13
.5641.
32 j
CScswowo 13.82
42.16
)21 33
.6 O'00 72.06
]
O;b00 98.4
87.6
!61
.32
ot-* o
2 X
1.25
2.5 3.4 4.26
8.25 OX 3.75
7.25
9.2 CO 3.25 00 3.1
3 ;
5.75
2.2 ;
2-2
COcqO' 6.3 J
6.7 7.5 00 WXCSOx
pH ^£3 r | o 4a 3 |
0.3
cso-HWOb+1wq 41cs 5±
1 41wo 41coCO41xqv£>
CS41wxr-*CS41O 41wo
co
xob41vqCO41CS 2cs00
b41cqCO41CS 5x
1 ft3
cs41woCS 26±2
|
CS41XOCS
CS4100csCO41co
00CO41cqwx*CO
r A£3 « ® *E0 - 0 * 3
41CS acs ncs 3±1 41CS 41wx
VIb41
b41°ooob4)
wqb41wowqb41wo
cs41OXb
3csCOb41CO
cqb41wqb-HCS
wob41wqb41CS
cob$bCSb4100b
wqb41CS
o-b41CSCOo41CS
COb41o
Si ov
erla
yer
thic
knes
s(A
)
j 22
0020
0019
0017
5013
00 900
1850
1300
1230 900
1980
1930
1800 wo0\
1950
|150
0 j
2150
|21
5017
3014
20132
0 1
1200
1100
I89
0
Ther
mal
Oxi
deth
ickn
ess
lA)_
____
_50
0 0001 1350
1700
3300
3600
1500
2800
2900
3600
1200
1300
1600 oooco 1250
123
00 850
j85
017
5025
0027
00300
0 I
3200
3500
Oxi
datio
ntim
e(h
rs.) CS O 24 36 60 74 6.5 00 WOOO* 27 CS 2.2
5 xo
2.25
4.5 :
wqb 0.6
1
xq
2.5 |
3.5 |
WO
Oxi
datio
nte
mpe
ratu
re(°
C) 90
090
0 006 900
900
900
950
950
950
950
1000
:10
0010
00100
0 I
ocso 1050
1100
j110
0 j
1100
! OOII 11
0011
0011
0011
10
Waf
er B3
Sam
ple
|B3/
1B3
/2B3
/3B3
/4B3
/5B3
/6B
3/7
IB3/
8IB
3/9
B3/1
0IB
3/11
B3/
12|B
3/13
IB3/1
4 I
IB3/
15IB
3/16
[IB
3/17
|IB
3/18
|IB
3/19
1IB
3/20
IB3/
21IB
3/22
B3/2
3B3
/24
237 Appendix E
Tabl
e 3-
List
of sa
crifi
cial
ly
oxid
ised
sam
ples
pre
pare
d fro
m wa
fer
B3
J o fe X r
34 i
9.8 VO ____
__J
cncs 18
.6 CO 11.6
'1
23.8 30 25.8
j33
.222
.629
.628.
7 j
fe00*cs 32 46 cnfe 89 104.4
|52
[75
.493
.6 fec-
ClSoc-
£ X 1 3.2
5 UV<N00 o r- 00
1 .
8‘5 ..
.9.2 2.7 CO 3.1
2 fe 4.25
4.5 4.55
i
00 00 3.13 fe cs UVCSfe 5.9 6.25
6-7 1
7.5 _
]
00
OIS
Fle
ngth
(gm
)
vo©A
4©uv
cs3uv
3K
1 10
±2|
11±2 CS41CS
UV©41COcs
CS41cofe
uv41feuv
CS41COVO 7±
2 CS41COs
fe©41fevd
CO41csvo
vq©41CSoo’41uv 41VO
cs©4!S
UVCS410000cn4VDcnCS
CS4VOcsCS400cs 30±
3 ,
vqcs4VOrV
OIS
Fde
nsity
(cm
'2) 1
uvr--Huv
CSoto?
cso4100o
uv©41CS
CSo2
41uvcs
cso41vouvo4100
fe©4100r-o’41fe
UVO4100vqo41cs
4100COo4CO
feo41ooUV©41cs
4100cs
UN©41fe
r-©4oo
VDO4
CS
fe©41fe©41fe
feo4©4fe
Si ov
erla
yer
thic
knes
s(A
)
, 15
30 665
J 90
0 006 !
700
620
530
1620
1580
1550
1400
1350
1320
1300
700
|700
[
1560
1140
0 [
1750
j135
0 I
©00© 1020
|92
080
070
0
Ther
mal
Oxi
deth
ickn
ess
(A)_
____
__13
001
3300
2800
1 28
00 ©©CScn©ofem 36
0011
0012
0012
5016
0017
00 ©©00 18
25320
0 |
3200
1250
1
0091 o©00 1700
|235
0 I
2500
1270
0 1
3000
I32
00
Oxi
datio
ntim
e(h
rs.)
1 28
.5 60 00 18.5
L 22.
5 I
24
LZ 1.45 CS 2.2
5 j
fe UV 5.75 vo
UVfeuv vo cs m 0.5 vq 2.5 2.9 3.5 fe UV
Oxi
datio
nte
mpe
ratu
re(°
C) 85
0j
900
; 95
095
095
095
095
010
00100
0 1
1000
1000
0001 1000
I 0001 100
0 I
1000
I102
0 1
1020
1100
111
00110
0 1
1100
I O
OU 11
0011
00
Waf
er B4
Sam
ple
IB4/
1|B
4/2
[B4/
3(B
4/4
IB4/
5IB
4/6
IB4/
7IB
4/8
B4/9
IB4/
10|B
4/11
(B4/
121B
4/13
|IB
4/14
IB4/
15[B
4/16
IB4/
17 00r-HP2 B
4/19
|B4/
20B4
/21
B4/
221B
4/23
IB4
/24
|B4
/25
238 Appendix E
Tabl
e 4-
List
of sa
crifi
cial
ly
oxid
ised
sam
ples
pre
pare
d fro
m wa
fer
B4
(ini
tial
thic
knes
s,
Ti =2
100
A).
&. OVi oo O
% X5.2
0000
7.23
4.72
10.93
|10
.97.2
310
.311.
5 ;
XOCNft 11.5
\ 11
.37 15.6
j8.5 6.6
9!3
.5 CNONCN 33.2
|
ftCNm 42.16 37.6 52 43.7
"gor-g o
8.25
3.7 6.62 ft 00 00 vxod
[ 9.2 2.8
8 cn 3.25 rt- 00 5.75 CN 2.2 4.2 4.25 vx 5.88
6.25
6.7 00
OIS
Fle
ngth
(pm
)
VXo-Hpvx
00ortvxcdrtvxXO
CNrtvxvxrto
00o4ON
ftOrtcnodCN41CN
vxOrtcnCN 4±0.5
VXrtvxvx
3vxrt*rt00ft
rtCNvxCNrtO nrt*
r-.3 16
.2±2 VXCNrt00
CNrtvxONCN-HVXCN 27±
4 |
cn41VX00CNCN41VXm
OIS
Fde
nsity
xlO
3cm
'2
CNrt00COrtCN CNrtXO
(OrtvxCNrtOO
CNrtCNxortCNft
CNrtOX00 16±3 rtVO
vxrtft
CNrt00cn3xd
rtxoftCNrtrt-
cn41OCN4100xd
CNrtftCNrtft rt00
CN41ftcnrto rt00
Si ov
erla
yer
thic
knes
s(A
)
665
1 14
50 OvxON
088 700
700
620
530
1 16
0015
5015
3014
00 700
;11
00175
0 ;
ocnft 1360
1350
1230
|108
0 1
1020
j93
0 oo00
Ther
mal
Oxi
deth
ickn
ess
(A)_
____ ooCOCO
1 14
751
2650
| 28
001
3200
3200
jI
3400 ooxocn 1150
J12
0013
00160
0 I
3200
1230
0 ! '
008 ovxOO 1700
1700
2000
j23
50250
0 |
2700
3000
Oxi
datio
ntim
e(h
rs.)
____
____
_
09
] 6.5 17
.5 VXCO 22.5
| 22
.6 24 27 1.75 CN 2.25 rt-
vxftvx in■rt 0.5 0.6 vx xo CN 2.5 j
2.9 3;5 rt*
Oxi
datio
nte
mpe
ratu
re(°C
)
006 ovxo\ 950
950
i 95
095
0|
950
950
1000
1000
1000
1000
1000
|105
0 1
1100
1110
0 j
1100
|11
00110
0 I
1100
111
0011
0011
00
Waf
er
vxCQ
Sam
ple
B5/1
B5/
2B5
/3IB
5/4
IB5/
5B5
/6B
5/7
B5/
8B5
/9 !
B5/
10B5
/11
B5/
12B5
/13
B5/
14B5
/15
B5/
16B
5/17
B5/
18B
5/19
B5/
20B5
/21
iB5
/22
|B5
/23
239 Appendix E
Table
5-
List
of sa
crifi
cial
oxid
ised
sam
ples
pre
pare
d fro
m wa
fer B
5 (i
nitia
l th
ickn
ess:
Ti
-210
0 A)
.
csaft oMO
- S o" ■—t£ x
L'ZZ'Z
5.45
43.3
940
3045
81.5 OoorH 408
4650
+To 1— 1 X cn
vnO«—HXONONr-H
OvO 82
oONr—i 273
•rfol“H£ooi—i
N-or-HXf t
r f
**
IN
g
J ^
NOrf
OISF
de
nsity
xl
03cm
"2
i—io'-Hino’
C N
o'
o'o' + i—H
H'-Hoo
(Nor—4X ✓—\ VOo'-Hd
COoi—iX /—\ in-Hcoi—i
ONHHor -H
CO-HONONh*r-H-HOm
OXoXino'-Hr-H
O!— 1Xino'-Hvooo
toor-HX ✓—■n-Hinm
in41oONh-+1mr -H
CO-HONON
oON+1Oin
o(H_X_i—i-Hoo
nvO--\cn -H o ON v_'
**
Oxide
tim
e (h
rs)
f t -
Oxid
ation
tem
pera
ture
(°C)
oor—I r-H
csgs-
<D Ot/1O X Q w
00o' - CN <n 00 oi—i 00o - ON •n oo or H
00o - 04 in 00 oH o
O N
Ener
gy(k
eV)
oftON
Ion
impl
anted
pq
SIM
OXw
afer
r-HPQ - * = = = CLcoPQ= = = = B oi
PQ= = = = = =
Sam
ple
rbtinear-H
pq
1— 1§
G9t—(PQ
r—i COts§S
>—i4
ear—tPQ
r-H
n"'cs£-tG3i—iPQ
i—H
VO
eai—iPQ
1—I r-H
eacnCQ
r-H
tfteaCOPQ
critfteacnCQ
tTeacnPQ
nMtneacoPQ
3S ’69cnPQ
r-H
§PQ
r-H
ON
heaH*PQ
r-H
co
§PQ
hi§PQ
r-H
<ri
eaH*PQ
<-(
tT§PQ
r-H
f t
tfteaPQ
240 Appendix E
Tabl
e 6
- Li
st
of sa
crif
icia
l ox
idis
ed
sam
ples
im
plan
ted
with
B
F*2
ions
.
CN6
s o■US r <X >4
13.6
LZ
L
8.75
601
601
rt-o'■—i 11
.5
8.2
12.6
14.7
tsor—1XN©
CNot-HXO'Nf
*-* **
CN'Boft
X X
4.6
\n § V •o gfe 2 O x
NO0 -HI©01
c©o ’-Hc©c©
n©O-Hv°.
c©O-H001
•©o'5001
c©O-HON
i©c©0-H( I01
rto-Hi©
NOo ’-Hr©oi
c©0 -H o-01
tsO51o '+1oI©,
tsoX/•—"N01 1-Hopi,
** *
Oxi
datio
n tim
e (h
rs).
Oxi
datio
nte
mpe
ratu
re(°C
)11
00
cs"*"'So
to 1Q S
i—< CN '© 7.5 © I—H 20 30 50
09 ooo oo 250
300
500
Eenr
gy(k
eV) oON
Ion
impl
ante
d
+c7>
SIM
OX
waf
er B4(
2)
Sam
ple
B4/
Si7l
/1
B4/
Si+/2
/l r-HCOf
« B4/
Si73
/1 rHrt?fiHCO©-pq
r—l
00rt?PQ B4
/S17
6/1 o+c©
4PQ B4/
Si78
/1
B4/
Si79
/1
B4/
Si7l
0/1
B4/
Si7l
l/1 CN+Pc©7PQ B4
/ST/
13/1
241 Appendix E
Tabl
e 7-
List
of sa
crifi
cial
oxi
dise
d sa
mpl
es p
revi
ously
im
plan
ted
with
Si i
ons.
'aoo§ o
1<25 X
roVDt—tINoor-H
voDfr—<oCDl—l
ooCDi—<CDVOi-H
INn-i—H
VO >— 1oo
Clor-HXINVO
cVor—1XinCD
** vo4in
vo4>n
CN00 N- vq4in
OnOr—HOn O !—<
VDCNp-H
CDtNCN
(NO«—HXDfo’CD
-X-*
CSSo
x O o£ X
vo4 *
•I? — § s-a <->fe 2 o &
VOaCN
-HCDCN“HCO -Hvo
<kD-H«Nr—<
cn+oCNino'-Hoor—t
ino’-Hi—i
C©or—(X /—> in N* O4inNf»—i
toOr-'MX>n-Hor-H
** in-HoCN-Ho -Hin
CN+1OO
■—ia
in-HoCNCN41oCN
r-H-HCDCN
i—io’-Hino
or-HXCDo’-HCDr-H
**
Oxida
tion
time
(hrs.
) re-
Oxida
tion
tempe
ratu
re
(°C
)___ oo
"gO0 <D TU to v—'& s
>Dcn
uo VIN O >DCN CD i> oCN 'nCN oID •nCN ID <DN o IDni *D N oCN incN oCD
W w
ocncn
Ion
impl
anted
4<DO
SIMOX
wa
fer sCDPPCNNPQ
4k&reoo
r-H
+0>0CDPQ
r-H
CN+<DoCDPQ
r-HCO+<DoCDPQ
r*HN+<D0CDPQ
in+<POCDPQ
rHvo+<uOCDPQ
N+<DOCDPQ
00+<uOCDPQ
i—IOn+<uOCDPQ
r-Hor—H
UOCDPQ
r—H
+<D0CDPQ
+<DO4PQ
i-HCN\>o4PQ
i-HCD+04PQ
»—i4-V04PQ
>n+<u04PQ
©+<D04PQ
\>04PQ
r-H
oo\>0NPQ
I—1ON+<D04PQ
Or-HU04PQ
\>0
PQ
242 Appendix E
Tabl
e 8
- List
of
sacr
ifici
al o
xidi
sed
SIM
OX
sam
ples
pre
\nou
sly
impl
ante
d wi
th G
e+ io
ns.
P u b lica tio n s
Part of this thesis has been published in several conferences and journals:
1. L.F. Giles, T.W. Fan, A. Nejim and P.L.F. Hemment in " Delineation of defects in SIMOX structures using a chemical etching technique." Vaccum 43(4) 297 (1992).
2. L.F. Giles, A. Nejim and P.L.F. Hemment in " New etchant for crystallographic defect studies in thin SOI materials (< 1000 A)", Electronics Letters 29(9) 788 (1993).
3. L.F. Giles, A. Nejim and P.L.F. Hemment in " New etchant for very thin SIMOX structures", Materials Chemistry and Physics, 35 129 (1993).
4. L.F. Giles, C.D. Marsh, A. Nejim, P.L.F. Hemment and G.R. Booker in " Crystallographic defect studies in SIMOX material thinned by sacrificial oxidation", Nucl. Inst. Methods in Phys. Researchs B84 242 (1994).
5. L.F. Giles, A. Nejim, C.D. Marsh, P.L.F. Hemment and G.R. Booker in " Formation of oxidation induced stacking faults during sacrificial thinning of SIMOX materials", Proceedings of the 1993 IEEE International SOI Conference, page 54 (1993).
6. L.F. Giles, A. Nejim, F. Cristiano and P.L.F. Hemment in " Nucleation of OISF in SIMOX materials", Proceedings of the 1994 Eletrochemical Society Conference, San Francisco, USA.
7. L.F. Giles, N. Meyyappan, A. Nejim, F. Cristiano and P.L.F. Hemment in "Delineation of crystallographic defects in low energy low dose SIMOX materials", Proceedings of the 1994 Electrochemical Society Conference, San Francisco, USA.
8. L.F. Giles, F. Cristiano, A. Nejim, G. Curello and P.L.F. Hemment in "Annihilation of SFT by ion implantation, Proc. IIT 94 conference, May 1994 (to be published in Nucl. Instr. & Methods).
9 A. Nejim, C.D. Marsh, L.F. Giles, P.L.F. Hemment, Y. Li, R.J. Chater, J.A. Kilner and G.R. Boker in " An investigation of the role of the time averaged ion beam current density upon the defect densities in thin film SIMOX" in Nucl. Inst. Methods in Phys. Researchs B84 248 (1994).
10. A. Nejim, Y. Li, C.D. Marsh, P.L.F. Hemment, R.J. Kilner, L.F. Giles and G.R. Booker in " Process optimisation for direct formation of device worthy thin film SIMOX structures using 90 KeV oxygen implantation", Nuclear Inbstruments and Methods B74 170 (1994).
243 Publications