ABSTRACT
Title of Dissertation: IN SITU ENRICHMENT AND
EPITAXIAL GROWTH OF 28Si FILMS
VIA ION BEAM DEPOSITION
Kevin Joseph Dwyer, Doctor of Philosophy, 2017
Directed By: Professor John Cumings, Department of
Materials Science and Engineering
Isotopically enriched 28Si is an ideal material for solid state quantum comput-
ing because it interacts weakly with the spin states of embedded qubits (quantum
bits) resulting in long coherence times. This is the result of eliminating the roughly
4.7 % 29Si isotopes present in natural abundance Si, which possesses nuclear spin
I = 1/2 that is disruptive to qubit operation. However, high-quality 28Si is scarce
and the degree to which it improves the performance of a qubit is not well under-
stood. This leads to an important question in the Si-based quantum information
field, which can be stated as “how good is good enough?” regarding the perfection
of 28Si as a host medium for qubits. The focus of this thesis is to engineer a material
that can address this question, specifically in terms of the enrichment. Secondary
requirements for ideal 28Si films that are also pursued are crystalline perfection and
high chemical purity.
I report on the production and characterization of 28Si thin films that are the
most highly enriched of any known 28Si material ever produced with a maximum
28Si enrichment of 99.9999819(35) % and a residual 29Si isotopic concentration of
1.27(29)× 10−7. A hyperthermal energy ion beamline is used to produce this ex-
treme level of enrichment starting from a natural abundance silane gas (SiH4) source.
The Si is enriched in situ by mass separating the ions in a magnetic field just before
deposition onto Si(100) substrates. Initial proof of principle experiments enriching
22Ne and 12C were also conducted. In the course of achieving this 28Si enrichment, I
also pursue the epitaxial deposition of 28Si thin films. Characterizations of the film
morphology and crystallinity are presented showing that smooth, epitaxial 28Si films
are achieved using deposition temperatures between 349 ◦C and 460 ◦C. Crystalline
defects present in these films include {111} stacking faults. When using higher
deposition temperatures, I find that trace impurity compounds such as SiC cause
step pinning and faceting of the growth surface leading to severely rough films. As-
sessments of the chemical purity of 28Si films are also presented, which show major
impurities N, C, and O are present in the purest film at an atomic concentration of
approximately 1× 1019 cm−3, resulting in a Si purity of 99.96(2) %.
Additionally, I introduce a model that describes the residual 29Si and 30Si in
28Si films, i.e. the enrichment, as the result of adsorption of diffusive natural abun-
dance SiH4 gas from the ion source into the
28Si films during deposition. This model
correlates the measured enrichments of 28Si films with the SiH4 partial pressures dur-
ing deposition. An incorporation fraction for SiH4 adsorption at room temperature
of s = 6.8(3)× 10−4 is extracted. Finally, the temperature dependence of the sample
enrichment is analyzed using a thermally activated incorporation model that gives
an activation energy of Ec = 1.1(1) eV for the reactive sticking coefficient of SiH4.
IN SITU ENRICHMENT AND EPITAXIAL GROWTH OF 28Si
FILMS VIA ION BEAM DEPOSITION
by
Kevin Joseph Dwyer
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2017
Advisory Committee:
Professor John Cumings, Chair/Advisor
Dr. Joshua Pomeroy, Co-Advisor
Professor Lourdes Salamanca-Riba
Professor Ichiro Takeuchi
Professor Neil Goldsman
© Copyright by
Kevin Joseph Dwyer
2017
Acknowledgments
The path that led me to become a Ph.D. candidate studying materials science
and engineering (MSE) at the University of Maryland, College Park began as I
was an undergraduate physics major at Maryland and sought to join a lab to gain
research experience. I contacted Professor Gottlieb Oehrlein, who invited me to work
in his lab, for which I am grateful. His lab was a materials science lab researching
plasma-surface interactions. Later, when applying to graduate programs, Professor
Oehrlein’s graduate student, Robert Bruce, suggested that I apply to the Materials
Science and Engineering Department at Maryland. I did apply, and, after being
accepted, I decided to become a materials scientist.
While taking first year classes in graduate school, my professor, John Cum-
ings, put me in touch with a scientist at the National Institute of Standards and
Technology (NIST) looking for a graduate student. That scientist was Dr. Joshua
Pomeroy, and I soon joined his group at NIST in Gaithersburg after an impressive
tour. Josh became my co-advisor and John my official MSE advisor for the duration
of a fruitful six year collaboration that produced the work in this thesis. I thank
John for not only introducing me to Josh and being the “guy that signs my forms”,
but also for his practical advice for achieving academic milestones as well as the
growth of my scientific career. Working with Josh at NIST has been pivotal in my
development as a scientist, and I thank him for his role in guiding that development.
The combination of practical engineering and analytical science knowledge imparted
by Josh has made a strong impression on me and will no doubt serve me well moving
ii
forward. Additionally, Josh’s do-it-yourself attitude and adage that “data talks and
everything else walks” will always stick with me. I am very proud of what we have
accomplished and look forward to collaborating with him again.
I want to thank Dr. Russell Lake, Josh’s first graduate student, for making
me feel welcome after transitioning from Maryland to Josh’s group at NIST and for
many insights and scientific discussions. I thank Josh’s current group members, Dr.
Aruna Ramanayaka and Ke Tang for helpful discussions, and especially Hyun soo
Kim, with whom I spent countless hours in the lab (breaking and fixing things) in
the pursuit of data. I have had the fortune and pleasure of working with numerous
other smart people at NIST. Thank you to former and current members of the
Quantum Processes and Metrology group at NIST including Dr. Neil Zimmerman,
Dr. Ted Thorbeck, Dr. Panu Koppinen, Dr. Justin Perron, Dr. Michael (Stew)
Stewart, Dr. Roy Murray, Dr. Hamza Shakeel, and Zac Barcikowski for numerous
helpful discussions and brainstorming sessions over the years. Interactions with
these colleague have led to valuable scientific relationships and even more valuable
friendships. Additionally, thank you to Dr. Garnett Bryant for paying my stipend
during my tenure.
I gratefully acknowledge the Laboratory for Physical Sciences for partially
funding this work and look forward to joining them as a postdoctoral researcher.
Other collaborators at NIST I want to thank include Terry Moore, Dr. Rick
Silver, Dr. Kai Li, Dr. Pradeep Namboodiri, Xiqiao Wang, Dr. Kristen Steffens,
Dr. June Lau, Dr. Vald Oleshko, Dr. Joshua Schumacher, and Dr. Alline Myers.
I extend a special thank you to the most important collaborator I had the pleasure
iii
of working with, Dr. Dave Simons, whose expertise in isotope measurements and
willingness to push the boundaries of his work played a critical role in this thesis.
Thank you to the fellow students of my MSE Ph.D candidate class including
Alex, Jen, Amy, Colin, Elliot, Ben, and Jeff for helping me survive my first year
of graduate school. Thank you to the GRA NIST solfball team led by our fearless
leader Jack for helping me blow off steam at the plate and on the pitching mound
in between experiments. We may not have won any championships, but it was fun
nonetheless! Thank you also to my roommates Dana, Somak, and Evan for putting
up with my crazy hours while writing this thesis.
Most crucially, I want to acknowledge and extend a very warm thank you to my
parents Eddie and Linda and brother John as well as the rest of my family for their
endless love and support over the years! Thank you to my parents for cultivating,
encouraging, and enabling my scientific curiosity as a child. Lego building, physics
and chemistry kits, model rockets, a telescope, trips to the Franklin Institute, and
numerous other activities and discussions about science topics provided by my par-
ents formed the foundation for the scientist that I have become. Further financial
and emotional support from them throughout the years as I progressed through high
school, college, and finally, graduate school is greatly appreciated. Ultimately, the
work presented in this thesis would not have been possible without them.
Finally, thank you to my committee members Josh, John, Professor Lourdes
Salamanca-Riba, Professor Ichiro Takeuchi, and Professor Neil Goldsman for their
careful consideration of this thesis.
iv
Table of Contents
List of Tables ix
List of Figures x
1 Introduction 1
1.1 28Si for Quantum Information . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Si-Based Solid State Quantum Information . . . . . . . . . . . 1
1.1.2 Sources of 28Si . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.1.3 Single Spin Measurements in 28Si . . . . . . . . . . . . . . . . 15
1.2 Ion Beam Enrichment and Deposition . . . . . . . . . . . . . . . . . . 19
1.3 Objectives and Outline . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3.1 Project Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.3.2 Strategy and Impact of Results . . . . . . . . . . . . . . . . . 25
1.3.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2 Experimental Apparatus and Methods 30
2.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.1 Ultra-High Vacuum Deposition . . . . . . . . . . . . . . . . . 30
2.1.2 Previous Operation . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Hyperthermal Energy Ion Beamline . . . . . . . . . . . . . . . . . . . 33
2.2.1 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.2 Theory of Magnetic Mass Separation . . . . . . . . . . . . . . 40
2.2.3 Operating Parameters . . . . . . . . . . . . . . . . . . . . . . 50
2.2.4 Ion Beam Characterization . . . . . . . . . . . . . . . . . . . . 54
2.2.4.1 Ion Beam Mass Spectra . . . . . . . . . . . . . . . . 54
2.2.4.2 Ion Beam Energy, Ei . . . . . . . . . . . . . . . . . . 67
2.2.4.3 Ion Beam Spot Size . . . . . . . . . . . . . . . . . . 73
2.3 UHV Deposition and Analysis Chamber . . . . . . . . . . . . . . . . 75
2.3.1 Apparatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.3.2 Vacuum Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.3.3 Sample Manipulation . . . . . . . . . . . . . . . . . . . . . . . 79
2.3.4 In situ Sample Analysis . . . . . . . . . . . . . . . . . . . . . 87
v
3 Initial Experiments Enriching 22Ne and 12C 91
3.1 Context and Experimental Setup . . . . . . . . . . . . . . . . . . . . 91
3.2 22Ne Implantation and Characterization: Proof of Principle . . . . . . 94
3.2.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . 94
3.2.2 22Ne Implantation . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.2.3 Enrichment Measurements via SIMS . . . . . . . . . . . . . . 99
3.3 12C Deposition and Characterization: First Enriched Thin Films . . . 102
3.3.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
3.3.2 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . 103
3.3.3 Deposition of 12C . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.3.4 Enrichment Measurements via SIMS . . . . . . . . . . . . . . 109
3.4 Chapter 3 Summary: Outlook for 28Si . . . . . . . . . . . . . . . . . . 114
4 28Si Thin Film Deposition and Characterization Phase I: In Situ Enrichment116
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.1.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.1.2 Experimental Configurations for 28Si Deposition . . . . . . . . 119
4.2 Si Deposition Proof of Principle: Ion Beam Chamber Samples . . . . 125
4.2.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . 125
4.2.2 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . 128
4.2.3 Deposition of 28Si . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.2.4 Enrichment Measurements via SIMS for IC–1 Samples . . . . 134
4.2.5 Summary of Results for IC–1 Samples . . . . . . . . . . . . . 141
4.3 Achieving Highly Enriched 28Si: Lens Chamber Samples . . . . . . . 142
4.3.1 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . 142
4.3.2 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3.3 Deposition of 28Si . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.3.4 Enrichment Measurements via SIMS for LC–2 Samples . . . . 150
4.3.5 Crystallinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.3.6 Chemical Purity . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.3.6.1 SIMS . . . . . . . . . . . . . . . . . . . . . . . . . . 159
4.3.6.2 XPS . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
4.4 Chapter 4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
5 28Si Thin Film Deposition and Characterization Phase II: Crystallinity and
Chemical Purity 169
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.2 Experimental Setup for Improving Crystallinity and Chemical Purity:
Deposition Chamber Samples . . . . . . . . . . . . . . . . . . . . . . 175
5.3 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.3.1 Ex Situ Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.3.2 In Situ Preparation . . . . . . . . . . . . . . . . . . . . . . . . 183
5.4 Deposition of 28Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.5 Enrichment Measurements via SIMS for DC–3 Samples . . . . . . . . 201
5.5.1 Initial Tests at DC–3 . . . . . . . . . . . . . . . . . . . . . . . 201
vi
5.5.2 Enrichment Progression Timeline Samples . . . . . . . . . . . 205
5.5.3 Samples with Deposition T > 600 ◦C . . . . . . . . . . . . . . 214
5.5.4 High Pressure Mode Sample . . . . . . . . . . . . . . . . . . . 225
5.6 Epitaxial Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
5.6.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
5.6.2 Morphology of Films with Deposition T > 600 ◦C . . . . . . . 238
5.6.2.1 RHEED . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.6.2.2 STM . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
5.6.2.3 SEM . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
5.6.2.4 Step Pinning Induced Roughness . . . . . . . . . . . 250
5.6.3 Elimination Strategies for Step Pinning Sites . . . . . . . . . . 280
5.6.4 Morphology of Films with Deposition T < 600 ◦C . . . . . . . 292
5.6.4.1 RHEED . . . . . . . . . . . . . . . . . . . . . . . . . 292
5.6.4.2 STM . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
5.7 Chemical Purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
5.7.1 Context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
5.7.2 XPS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
5.7.3 SIMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
5.8 Crystallinity: Film Inspection via TEM . . . . . . . . . . . . . . . . . 334
5.9 Chapter 5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
6 Pressure and Temperature Dependent Adsorption of 29Si and 30Si During
28Si Deposition 352
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
6.2 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 357
6.2.1 28Si Samples and Enrichment Values . . . . . . . . . . . . . . 357
6.2.2 SiH4 Mass Spectrum and Mass Selectivity . . . . . . . . . . . 360
6.2.3 Determination of SiH4 Partial Pressures . . . . . . . . . . . . 365
6.2.4 Substrate Temperature Calibration . . . . . . . . . . . . . . . 369
6.3 Temperature Dependent Gas Incorporation Model . . . . . . . . . . . 370
6.4 Correlating Enrichment to SiH4 Partial Pressure . . . . . . . . . . . . 373
6.5 Temperature Dependence of 29Si and 30Si Adsorption . . . . . . . . . 379
6.6 Temperature Dependence of the Incorporation Fraction, s . . . . . . . 383
6.7 Determination of the Reactive Sticking Activation Energy, Ec . . . . 387
6.8 Chapter 6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
7 Summary of Results and Future Experiments 392
7.1 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 392
7.2 Proposals for Future Experiments . . . . . . . . . . . . . . . . . . . . 400
7.2.1 Targeted Levels of Enrichment . . . . . . . . . . . . . . . . . . 400
7.2.2 Enriched Si/Ge Deposition . . . . . . . . . . . . . . . . . . . . 403
7.2.3 28Si sublimation . . . . . . . . . . . . . . . . . . . . . . . . . . 404
7.2.4 Al Dopant Devices with Hydrogen Lithography . . . . . . . . 405
7.2.5 Electrical Measurements and T2 in
28Si . . . . . . . . . . . . . 407
vii
A Ion Source and Beamline: Additional Operating Parameters 410
B Experimental Apparatus Photographs 415
C Substrate Catalog 425
D Sample Catalogs 427
E SIMS Measurement Settings 441
F 28Si Deposition Fun Facts 448
Bibliography 449
viii
List of Tables
A.1 Ion source and beamline operating parameters for implanting 22Ne
on 5/4/11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
A.2 Ion source and beamline operating parameters for depositing 12C on
2/7/12 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
A.3 Ion source and beamline operating parameters for depositing 28Si on
12/18/15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414
C.1 Substrate Catalog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
D.1 22Ne Sample Catalog (3/1/11–5/4/11) . . . . . . . . . . . . . . . . . 428
D.2 12C Sample Catalog (1/25/12–2/7/12) . . . . . . . . . . . . . . . . . 429
D.3 28Si Sample Catalog: IC–1 (6/4/12–6/28/12) . . . . . . . . . . . . . . 430
D.4 28Si Sample Catalog: LC–2: I (2/4/13–3/4/13) . . . . . . . . . . . . . 431
D.5 28Si Sample Catalog: LC–2: II (3/7/13–9/24/13) . . . . . . . . . . . 432
D.6 28Si Sample Catalog: DC–3: I (2/7/14–6/26/14) . . . . . . . . . . . . 433
D.7 28Si Sample Catalog: DC–3: II (8/28/14–7/27/15) . . . . . . . . . . . 434
D.8 28Si Sample Catalog: DC–3: III (9/20/15–11/9/15) . . . . . . . . . . 435
D.9 28Si Sample Catalog: DC–3: IV (12/18/15–6/24/16) . . . . . . . . . . 436
D.10 Deposition parameters for the analysis of 28Si samples deposited at
room temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
D.11 Enrichment measurements and analysis results for 28Si samples de-
posited at room temperature . . . . . . . . . . . . . . . . . . . . . . . 438
D.12 Deposition parameters for the analysis of 28Si samples deposited at
elevated temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
D.13 Enrichment measurements and analysis results for 28Si samples de-
posited at elevated temperature . . . . . . . . . . . . . . . . . . . . . 440
ix
List of Figures
1.1 Si-based quantum dot and P-donor devices . . . . . . . . . . . . . . . 4
1.2 Schematic diagrams of the Bloch sphere for a qubit spin and the T2
coherence time due to spin dephasing . . . . . . . . . . . . . . . . . . 6
1.3 Cartoon depictions of a natural abundance Si crystal containing 29Si
nuclear spins and a nuclear spin-free 28Si crystal . . . . . . . . . . . . 8
1.4 T2 coherence time measurements of the electron and nuclear spins of
31P atoms in 28Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 31P in Si energy levels and optical transitions, photoluminescence
spectra of 31P in natural abundance Si and 28Si showing hyperfine
splitting, and ESR frequency tuning of a 31P electron spin in 28Si . . 10
1.6 Process flow chart for production of a 28Si crystal from centrifugation
of natural abundance SiF4 . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7 Thermal conductivity measurements of natural abundance Si and 28Si 13
1.8 Schematic and photographs of the 28Si single-crystal boule and final
sphere produced in the International Avogadro Coordination . . . . . 14
1.9 Predicted electron T∗2 coherence time dependance on 29Si concentra-
tion for qubits in Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.10 ESR measurement phase space diagram of 29Si concentration vs.
number of measured 31P spins . . . . . . . . . . . . . . . . . . . . . . 18
1.11 Dual source ion beam deposition system schematic and an enrichment
measurement of a deposited 28Si film from the Tsubouchi group . . . 22
1.12 Enrichment progression timeline of the best overall sample enrichments 26
2.1 Schematic of the ion beam deposition system . . . . . . . . . . . . . . 31
2.2 Ion beamline schematic with ion source inset and electrostatic lens
potential landscape . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3 RGA mass spectra of the ion beam chamber at two achieved base
pressures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.4 Ion beamline schematic showing SiH4 diffusion and mass separation . 38
2.5 Calculated sector mass analyzer magnetic field relation to selected mass 43
2.6 Calculated mass dependance of the spatial separation, λ1, of ions with
adjacent mass number in the ion beam . . . . . . . . . . . . . . . . . 45
x
2.7 Calculated accelerating voltage dependance of the spatial separation,
λE, of ions accelerated with voltages differing by ∆V in the ion beam 48
2.8 Calculated mass dependance of the mass resolving power of the ion
beamline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.9 Schematic diagrams of the gas-mode and solids-mode Penning ion
sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.10 Ar, N2, and CH4 ion beam mass spectra . . . . . . . . . . . . . . . . 56
2.11 SiH4 ion beam mass spectrum . . . . . . . . . . . . . . . . . . . . . . 58
2.12 SiH4 ion beam mass spectrum showing shoulder peaks . . . . . . . . . 61
2.13 Ion beam mass spectra of chemical contaminants acquired when using
SiH4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.14 SiH4 ion beam mass spectra showing the transition from the low pres-
sure mode to the high pressure mode . . . . . . . . . . . . . . . . . . 65
2.15 Ion beam mass spectrum of Si ions generated from Si cathodes while
using an Ar working gas . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.16 Average ion energy measurement for Ar ions using a roll-off curve . . 68
2.17 Ion energy dependance on the anode and arc voltages . . . . . . . . . 70
2.18 Ion energy dependance of the relative energy spread, ∆E/Ei . . . . . 71
2.19 Calculated sputter yields for 28Si and 12C ions on a 28Si and a 12C
surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.20 Ion beam spot 2D current map . . . . . . . . . . . . . . . . . . . . . 74
2.21 Schematic cross section through the deposition chamber showing the
sample location and relevant instruments . . . . . . . . . . . . . . . . 76
2.22 RGA mass spectrum of the deposition chamber at its base pressure . 78
2.23 Photograph and wiring diagram for heating a Si(100) substrate on
the manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.24 Au-Si eutectic phase diagram . . . . . . . . . . . . . . . . . . . . . . 85
2.25 DH power supply sample heating control curve . . . . . . . . . . . . . 86
2.26 RHEED diffraction pattern of a clean, flash annealed Si(100) substrate 89
3.1 Ne ion beam mass spectrum . . . . . . . . . . . . . . . . . . . . . . . 95
3.2 22Ne implantation depth for different ion energies based on TRIM . . 97
3.3 Optical micrograph of a 22Ne-implanted sample produced at room
temperature at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.4 SIMS “depth” profile of a 22Ne-implanted sample produced at room
temperature at LC–2 with 0.545 % 20Ne . . . . . . . . . . . . . . . . 100
3.5 CO2 ion beam mass spectrum . . . . . . . . . . . . . . . . . . . . . . 105
3.6 Optical micrograph of a 12C sample deposited at room temperature
at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.7 SEM cross-sectional micrograph of a 12C sample deposited at room
temperature at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.8 SIMS “depth” profile of a 12C sample deposited at room temperature
at LC–2 with 39.2 ppm 13C . . . . . . . . . . . . . . . . . . . . . . . 110
3.9 SIMS “depth” profile of a 12C sample deposited at room temperature
at LC–2 with 39.2 ppm 13C on a linear scale . . . . . . . . . . . . . . 112
xi
4.1 Schematic drawings of the ion beam chamber, lens chamber, and
deposition chamber experimental setups used to deposit 28Si samples 120
4.2 Enrichment progression timeline of the best 28Si sample enrichments . 124
4.3 SiH4 ion beam mass spectrum for samples deposited at room temper-
ature at IC–1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.4 SEM cross-sectional micrograph of a 28Si film deposited at room tem-
perature at IC–1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
4.5 SIMS mass spectrum of 29Si and 28SiH showing well resolved peaks . 135
4.6 SIMS depth profile for a 28Si sample deposited at room temperature
at IC–1 with 1130 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . 137
4.7 SIMS depth profile for the most highly enriched 28Si sample deposited
at room temperature at IC–1 with 9.5 ppm 29Si . . . . . . . . . . . . 140
4.8 SiH4 ion beam mass spectrum for samples deposited at LC–2 . . . . . 148
4.9 Optical micrograph of a 28Si sample deposited at room temperature
at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.10 SIMS depth profile of a 28Si sample deposited at room temperature
at LC–2 with 2.02 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . 151
4.11 SIMS depth profile of a 28Si sample deposited at room temperature
at LC–2 with 0.993 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . 154
4.12 29Si isotope fractions vs. deposition rate for multiple SIMS measure-
ments of a sample deposited at room temperature at LC–2 . . . . . . 156
4.13 SIMS depth profile of the most highly enriched 28Si sample deposited
at room temperature at LC–2 with 0.691 ppm 29Si . . . . . . . . . . . 157
4.14 HR-TEM cross-sectional micrograph of an amorphous 28Si film de-
posited at room temperature at LC–2 . . . . . . . . . . . . . . . . . . 158
4.15 SIMS depth profile measuring C in a 28Si sample deposited at room
temperature at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.16 SIMS depth profile measuring C in a 28Si sample deposited at room
temperature at LC–2 with varying deposition pressure . . . . . . . . 163
4.17 XPS spectra of a 28Si sample deposited at room temperature at LC–2
and a control Si sample . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.1 RHEED patterns of a Si(100) substrate before and after thermal des-
orption of the native SiO2 . . . . . . . . . . . . . . . . . . . . . . . . 187
5.2 STM topography images of clean Si(100) (2×1) surfaces prepared in
situ by flash annealing . . . . . . . . . . . . . . . . . . . . . . . . . . 189
5.3 RGA mass spectrum of the deposition chamber during deposition of
a 28Si sample at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . . . 193
5.4 SiH4 ion beam mass spectrum for samples deposited at DC–3 using
the low pressure mode . . . . . . . . . . . . . . . . . . . . . . . . . . 194
5.5 SiH4 ion beam mass spectrum for samples deposited at DC–3 using
the high pressure mode . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.6 Optical micrographs of three 28Si samples deposited at DC–3 . . . . . 198
5.7 SIMS depth profile of a 28Si sample deposited at room temperature
at DC–3 with 0.58 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . 203
xii
5.8 SIMS depth profile of a 28Si sample deposited at 249 ◦C at DC–3 with
0.79 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.9 SIMS depth profile of a 28Si sample deposited at 610 ◦C at DC–3 with
300 ppb 29Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.10 SIMS depth profile of a 28Si sample deposited at 712 ◦C at DC–3 with
132 ppb 29Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
5.11 SIMS depth profile of the most highly enriched 28Si sample deposited
at 502 ◦C at DC–3 with 127 ppb 29Si . . . . . . . . . . . . . . . . . . 212
5.12 SIMS depth profile of a 28Si sample deposited at 812 ◦C at DC–3 with
4.32 ppm 29Si . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
5.13 29Si concentration calculations in 28Si from isotope diffusion during
deposition at 700 ◦C to 900 ◦C . . . . . . . . . . . . . . . . . . . . . . 219
5.14 SEM cross-sectional micrograph of the surface roughness of a 28Si film
deposited at 708 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . 222
5.15 SIMS depth profile of a 28Si sample deposited at 808 ◦C at DC–3 with
3.97 ppm 29Si compared to a sample deposited at 812 ◦C . . . . . . . 224
5.16 SIMS depth profile of 28Si sample deposited at 421 ◦C at DC–3 using
the high pressure mode with 0.303 ppm 29Si . . . . . . . . . . . . . . 227
5.17 TEM cross-sectional micrograph of a Si film showing the epitaxial-
to-amorphous transition of low temperature epitaxy . . . . . . . . . . 231
5.18 Si ion beam deposition epitaxy phase diagrams of T vs. Ei . . . . . . 236
5.19 RHEED pattern of a rough crystalline surface of a 28Si sample de-
posited at 708 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . 239
5.20 RHEED patterns showing diffraction due to microfacets on rough
surfaces of 28Si samples deposited at 610 ◦C and 705 ◦C at DC–3 . . . 241
5.21 STM topography image of the rough surface of a 28Si sample de-
posited at 708 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . 244
5.22 SEM tilted micrograph of the rough surface morphology of a 28Si film
deposited at 708 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . 245
5.23 SEM top-down micrographs of the surface morphology variation of
28Si films deposited between 610 ◦C and 1041 ◦C at DC–3 . . . . . . 247
5.24 STM topography images of pits formed at contaminants on a 28Si film
deposited at 709 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . 256
5.25 STM topography images of pits formed from contaminants in a nat-
ural abundance Si film deposited at 713 ◦C at DC–3 . . . . . . . . . . 258
5.26 RHEED pattern of SiC contamination on a flash annealed Si(100)
substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
5.27 STM topography image of SiC clusters on a Si(100) substrate after
an HF etch and 900 ◦C anneal . . . . . . . . . . . . . . . . . . . . . . 265
5.28 STM topography images of contaminant clusters on Si(100) sub-
strates after an HF etch and flash annealing . . . . . . . . . . . . . . 268
5.29 STM topography images showing metal contamination on Si(100)
substrates after flash annealing . . . . . . . . . . . . . . . . . . . . . 271
5.30 Mound size vs. deposition temperature of rough 28Si films deposited
above 600 ◦C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
xiii
5.31 Arrhenius plot of ln(A), the natural log of the mound area, vs. inverse
deposition temperature for rough 28Si samples deposited above 600 ◦C 279
5.32 STM topography images of pits formed from contaminants on 28Si
and natural abundance Si films deposited at 712 ◦C at DC–3 after
CMOS cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
5.33 STM topography images of contaminant clusters on Si(100) sub-
strates after CMOS cleaning and flash annealing . . . . . . . . . . . . 287
5.34 SEM top-down micrograph of the rough surface morphology of a 28Si
film deposited after CMOS cleaning at 705 ◦C at DC–3 . . . . . . . . 289
5.35 RHEED pattern of a 28Si sample with a smooth crystalline surface
with islands deposited at 357 ◦C at DC–3 . . . . . . . . . . . . . . . . 293
5.36 RHEED pattern variation of 28Si samples deposited between 249 ◦C
and 920 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
5.37 STM topography images of islands on the smooth surface of a 28Si
sample deposited at 357 ◦C at DC–3 . . . . . . . . . . . . . . . . . . 298
5.38 STM topography images of islands on the smooth surfaces of 28Si
samples deposited at DC–3 at low temperature . . . . . . . . . . . . 301
5.39 STM topography images of the surface morphology variation of 28Si
samples deposited between 249 ◦C and 804 ◦C at DC–3 . . . . . . . . 304
5.40 XPS spectra of a 28Si sample deposited at 812 ◦C at DC–3 . . . . . . 310
5.41 SIMS depth profile measuring 22 potential contaminants in a 28Si film
deposited at 460 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . 312
5.42 Phase diagrams for the N-Si, C-Si, and O-Si systems near the solid
solubility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
5.43 SIMS depth profile measuring N, C, and Cl in a natSi film deposited
at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
5.44 RGA mass spectrum of the deposition chamber during substrate de-
gassing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
5.45 SiH4 ion beam mass spectrum for N-contaminated
28Si samples de-
posited at DC–3 using the low pressure mode . . . . . . . . . . . . . 323
5.46 SIMS depth profile measuring N, C, and O contaminants in a 28Si
sample deposited at 712 ◦C at DC–3 . . . . . . . . . . . . . . . . . . 325
5.47 SIMS depth profile measuring contaminants in a 28Si film deposited
at 460 ◦C at DC–3 after experimental improvements . . . . . . . . . . 329
5.48 N, C, and O atomic concentrations of 28Si samples vs. deposition ion
current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
5.49 TEM cross-sectional micrographs showing facets and stacking faults
of a rough 28Si sample deposited at 708 ◦C at DC–3 . . . . . . . . . . 337
5.50 HR-TEM cross-sectional micrograph of a rough 28Si sample deposited
at 708 ◦C at DC–3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
5.51 TEM cross-sectional micrographs at two magnifications of a smooth
28Si sample deposited at 460 ◦C at DC–3 using the low pressure mode 341
5.52 HR-TEM cross-sectional micrograph of a smooth 28Si sample de-
posited at 460 ◦C at DC–3 using the low pressure mode . . . . . . . . 343
xiv
5.53 HR-TEM cross-sectional micrograph of a smooth 28Si sample de-
posited at 421 ◦C at DC–3 using the high pressure mode . . . . . . . 345
5.54 Isotope reduction timeline of the isotope reduction factors of the best
28Si sample enrichments . . . . . . . . . . . . . . . . . . . . . . . . . 348
6.1 SiH4 CVD surface reaction . . . . . . . . . . . . . . . . . . . . . . . . 355
6.2 SiH4 ion beam mass spectrum measuring the geometric selectivity . . 362
6.3 Ion beam mass spectrum of ThO+ showing a gas scattering peak tail 364
6.4 RGA mass spectra of the deposition chamber prior to and during
operation of the ion beam with SiH4 . . . . . . . . . . . . . . . . . . 367
6.5 Gas sticking deposition model cartoon for 28Si deposition . . . . . . . 371
6.6 Room temperature correlation plot of isotope fraction vs. SiH4 flux
ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
6.7 Room temperature correlation plot of isotope fraction vs. SiH4 flux
ratio including high pressure 28Si samples deposited at IC–1 . . . . . 377
6.8 29Si/30Si isotope ratios . . . . . . . . . . . . . . . . . . . . . . . . . . 378
6.9 Adjusted isotope fraction, cz(sT , k502), vs. temperature . . . . . . . . 381
6.10 High temperature correlation plot of converted isotope fraction vs.
SiH4 flux ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
6.11 s vs. deposition temperature . . . . . . . . . . . . . . . . . . . . . . . 386
6.12 Arrhenius plot of ln(s) vs. inverse deposition temperature . . . . . . 389
7.1 Isotope reduction timeline of the isotope reduction factors of the best
22Ne, 12C, and 28Si sample enrichments . . . . . . . . . . . . . . . . . 394
7.2 Enrichment progression timeline of the best overall sample enrich-
ments with references . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
7.3 SIMS depth profile of a Si isotope heterostructure deposited at LC–2 401
7.4 Si sublimation rates for a 28Si chip and calculated values . . . . . . . 405
7.5 SIMS depth profile of an Al delta layer in Si . . . . . . . . . . . . . . 406
7.6 Cartoon schematic of a 28Si capacitor . . . . . . . . . . . . . . . . . . 408
A.1 Ion beamline lens element circuit diagram . . . . . . . . . . . . . . . 411
A.2 Ion source operating parameter scans . . . . . . . . . . . . . . . . . . 412
B.1 Photographs of the ion beam deposition system . . . . . . . . . . . . 416
B.2 Photograph of the gas manifold . . . . . . . . . . . . . . . . . . . . . 417
B.3 Photographs of the ion source elements . . . . . . . . . . . . . . . . . 417
B.4 Ion beamline electrostatic elements and magnetic sector mass analyzer418
B.5 Photograph of the mass-selecting aperture used for 22Ne and 12C sam-
ples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
B.6 Photographs of the mass-selecting aperture used for 28Si samples at
IC–1 and LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
B.7 Photograph of the mass-selecting aperture used for 28Si samples at
DC–3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
B.8 Photograph of a gas aperture on the inlet to the deceleration lenes in
the ion beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
xv
B.9 Photographs of the experimental setup and sample stage for 28Si sam-
ples deposited at IC–1 . . . . . . . . . . . . . . . . . . . . . . . . . . 421
B.10 Photograph of the experimental setup of the lens chamber for samples
produced at LC–2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
B.11 Photographs of the sample stage for producing sample at LC–2 . . . 422
B.12 Photographs of the manipulator at DC–3 . . . . . . . . . . . . . . . . 423
B.13 Photographs of interchangeable sample apertures at DC–3 . . . . . . 424
xvi
Chapter 1
Introduction
1.1 28Si for Quantum Information
1.1.1 Si-Based Solid State Quantum Information
Major technological advances are often driven by or require development of
new or improved materials [1]. Development of visible light LEDs and lasers in the
1960s was possible due to improvements in GaAs crystals with engineered compo-
sitions. Carbon-based materials have played an important role in both manufac-
turing advances and basic physics research. Carbon fibers and carbon reinforced
plastics were first developed in the 1960s and have found a wide variety of man-
ufacturing applications as strong, lightweight materials. At the microscopic level,
carbon nanotubes have spurred a tremendous amount of research into their unique
and impressive physical and electrical properties with numerous potential applica-
tions. Although they were initially discovered in 1952, interest in carbon nanotubes
increased after further observations in the 1990s. The widespread use of small, high-
energy density, rechargeable batteries for portable electronics was made possible by
1
the development and engineering of lithium ion-based electrodes such as LiCoO2 or
later LiFeO2 starting in the 1980s. Another major materials-based technology that
has made portable electronics and, more broadly, all modern computers possible
is the transistor. Developed initially in the 1950s, the ubiquitous use of semicon-
ductor transistors as the base computing component in integrated circuits in the
microelectronics industry first required the engineering of extremely high purity,
single-crystalline silicon material. Further advances in silicon processing for metal-
on-semiconductor transistor technology would come to be guided on a large scale
by the International Technology Roadmap for Semiconductors. The engineering of
the properties of silicon has also been integral to the development of photovoltaic
solar cells since the 1970s and silicon cells account for the majority of solar panels in
use. High-quality, i.e. nearly perfect, silicon has perhaps been the most important
material to the modern and increasingly computer oriented world.
In solid state quantum information (QI), isotopically enriched 28Si is a critical
material for the further development of Si-based quantum computing architectures.
The abundance of high-quality (unenriched) Si and the established microelectronics
infrastructure make Si an attractive medium for quantum computing, which holds
the promise of significant increases in computing speed for certain tasks over clas-
sical computers. For quantum coherent devices, both the classical aspects of device
operation and the states of the qubits (quantum bits) utilize the electronic band
structure of crystalline semiconductors like Si. A great deal of experimental QI re-
search has leveraged these advantages with a high degree of success, as described by
Zwanenburg et al. in a review article [2]. Si-based quantum computing architectures
2
include electron spin qubits in quantum dots defined electrostatically by gate elec-
trodes. Quantum dots can be formed in a variety of structures including undoped
Si surfaces, Si nanowires, and Si quantum wells in Si/SiGe heterostructures.
Another popular quantum computing architecture is based on a proposal by
Kane to use the nuclear spins of an array of single donor 31P atoms in Si as qubits [3].
In this design, metal “A-gates” on the surface manipulate the individual donor spins
and they interact via electron-mediated coupling, which is controlled by “J-gates”.
The proposal makes clear that for this scheme to be successfully implemented in a
quantum computer, a host material for the donors needs to be free of nuclear spins (I
= 0). This basic design principle of gate-controlled donors and dots has been adapted
to produce architectures including implanted 31P atoms in Si transistor devices or
in combination with quantum dots and 31P single atom transistors or quantum dots
in Si produced by scanning tunneling microscopy H-lithography. A schematic of
Kane’s architecture as well as examples of two physical implementations of these
types of Si quantum computing devices are shown in Fig. 1.1. Panel (a) is the Kane
schematic from Ref. [3] showing the relation between the 31P donors and the surface
control gates. Panel (b) is a false colored top-down scanning electron microscope
(SEM) micrograph of a Si quantum dot device from an experiment in Ref. [4] defined
by electrostatic metal gates on a 28Si epilayer. The location of the quantum dot is
shown by the representation of the cartoon electron spin (circle with arrow). A single
electron transistor (SET) is formed by the gates at the bottom of the micrograph.
A transmission line for sending microwave pulses to the dot to manipulate the qubit
spin states is seen at the left. A schematic cartoon of the quantum dot system is
3
Figure 1.1: Si-based quantum dot and P-donor devices. (a) Schematic for a pro-
posed quantum computing architecture for 31P donors in Si. 31P nuclear spin qubits
are manipulated by metal “A-gates” and interact via their electrons, controlled by
“J-gates” (from Ref. [3]). (b) False colored SEM micrograph of a Si quantum dot
(cartoon arrow) defined electrostatically by metal gates on 28Si epilayer. Gates defin-
ing a single electron transistor are seen below the quantum dot and a transmission
line for ESR pulses is at the left. A cartoon schematic of the quantum dot is at the
right (from Ref. [4]). (c) STM image of a H-terminated Si(100) surface with areas
selectively depassivated for 31P dosing that will define electric gates and a quantum
dot with single or multiple 31P donors (from Ref. [5]).
shown at the right. Panel (c) is a scanning tunneling microscope (STM) image of a
H-terminated Si(100) surface from Ref. [5] with electrostatic gates and a dot defined
by selectively removing H in the bright and outlined areas of the image. These areas
of bare Si substrate will be dosed with 31P atoms, producing conductive wires and
a dot formed of one or several 31P atoms.
Interest in 31P nuclear and electron spins as qubits (or memory) in Si has also
4
spurred research in electron spin resonance (ESR) and nuclear magnetic resonance
(NMR) of 31P spin ensembles in Si crystals. One of the key performance metrics of
spin qubits, which is measured in bulk ESR experiments, is their dephasing or coher-
ence time, T2. This is the time that coherent spins comprising the qubit maintain
their quantum phase before the information they encode is lost to the environment.
For a continuous wave ESR measurement, T2 is inversely related to the ESR signal
linewidth. For a viable quantum computer, the T2 time must be sufficiently long
as to be approximately 1× 106 times longer than the average single gate operation
time to account for dephasing errors. This is a general design rule which depends
on a number of factors including use of error reduction codes. Two qubit gate op-
erations with gate times of the order of 100 ns have been demonstrated [6]. If the
spin interacts with local inhomogeneities in the magnetic field, enhanced dephasing
will occur and the resulting coherence time is denoted as T∗2 . Certain manipula-
tions of the spins using specific pulse sequences such as a Hahn echo or dynamical
decoupling can reduce the effects of dephasing due to environmental noise and will
result in a T2 echo signal measurement. The dephasing of spins in an ensemble is
measured, for example, by projecting the spin states as spin-up (aligned to an ex-
ternal magnetic field) or spin-down (anti-aligned to an external magnetic field) and
the probability of the spin being in those projected states over time gives the decay
characterised by T2. Other more complicated projections to states not aligned to
the magnetic field are also possible. Figure 1.2 shows a cartoon schematic of the
Bloch sphere construction for a spin qubit as well as a schematic of a T2 determi-
nation from measurements of such a spin. Panel (a) shows the Bloch sphere for a
5
Figure 1.2: Schematic diagrams of the Bloch sphere for a qubit spin and the T2
coherence time due to spin dephasing (a) The Bloch sphere for two level quantum
system. The state of an electron spin qubit in a semiconductor quantum dot exists
as a vector on the sphere and is projected as spin-up or spin-down. A “σx” optical
control pulse rotates the qubit spin around the x-axis to perform gate control opera-
tions (from Ref. [7]). (b) Diagram of a T2 coherence time determination showing the
decay of measurements of the spin-up probability after gate control operations. The
T∗2 decay is caused by the dephasing of the spin ensemble (adapted from Ref. [8]).
two level quantum system from Ref. [7]. The state of an electron spin qubit in a
semiconductor quantum dot exists as a vector on the sphere and is projected by a
measurement onto a basis state such as spin-up or spin-down. ESR or NMR control
pulses rotate the qubit spin around an axis, e.g. the “σx” rotation shown in panel
(a), to perform gate control operations. When the spin vector with a basis state
aligned with the magnetic field in the positive z direction lies in the x-y plane, the
6
strongest dephasing can occur. Panel (b) shows how T∗2 is determined from many
measurements of a spin where the spin-up population oscillates with varying spin
evolution time, adapted from Ref. [8]. The spin evolution when the spin is dephas-
ing is shown for three representations of the Bloch sphere at the top of the figure.
The spin-up oscillations exhibit a certain decay related to T2 and environmental
inhomogeneities.
In Si, a significant source of inhomogeneous magnetic field noise is the Over-
hauser field generated by nuclear and impurity spins in the crystal. Natural abun-
dance Si is comprised of three stable isotopes, 28Si, 29Si, and 30Si, which have abun-
dances of approximately 92.2 %, 4.7 %, and 3.1 %, respectively. The 29Si isotope
has a nuclear spin I = 1/2, while 28Si and 30Si isotopes have no net nuclear spin. By
eliminating 29Si nuclei, pure, isotopically enriched 28Si becomes an ideal spin-free
environment in which to place the electron and nuclear spins of qubits. Without
a randomly fluctuating global Overhauser field present, spins in 28Si interact with
their environment far less than in unenriched material leading to a greatly enhanced
T∗2 coherence time. Consequently, 28Si has been dubbed a “semiconductor vacuum”
and is analogous to the isolation of trapped atoms in a vacuum chamber [9]. A
cartoon model depicting the composition of 28Si and natural abundance Si (natSi) is
shown in Fig. 1.3, adapted from Ref. [10].
Theoretical modeling and bulk ESR experiments predicted the enhancement
in T∗2 to be proportional to the reduction in 29Si concentration [11,12], which further
spurred interest in exploiting 28Si experimentally. Numerous research groups have
shown through bulk ESR and NMR experiments of 31P spins in 28Si that nuclear
7
Figure 1.3: Cartoon depictions of natural abundance Si and enriched 28Si crystals.
(a) Natural abundance Si contains 29Si atoms that possess nuclear spins as well as
nuclear spin-free 28Si and 30Si atoms. (b) The 28Si is free of nuclear spins. (adapted
from Ref. [10])
and electron spin coherence T2 times can exceed seconds [9, 10, 13, 14]. A recent
measurement of the T2 time of a single
31P electron spin resulted in a value of
559 ms [10]. Two examples of T2 measurements of spins in
28Si for both bulk
NMR and in the single spin regime are presented in Fig. 1.4. Panel (a) shows the
measurement results from Ref. [10] mentioned above for a single 31P atom implanted
into a 28Si quantum dot device. The dynamic decoupling pulse sequence used for
this measurement is referred to as CPMG [15]. The 31P nuclear spin measured in
this experiment was determined to have a T2 = 35.6 s. Those measurements were
done at cryogenic temperatures. Panel (b) shows a T2 determination measurement
from Ref. [14] of the nuclear spins of an ensemble of 31P atoms in 28Si in a bulk
NMR experiment. When using a so-called XY-16 decoupling pulse sequence [16],
the measurement resulted in a coherence time of T2 = 39 min, measured at room
8
Figure 1.4: T2 coherence time measurements of spins in
28Si. (a) An electron spin
of a single 31P atom in 28Si has a T2 = 559 ms after dynamical decoupling (from
Ref. [10]). (b) An ensemble of 31P atoms in 28Si have a nuclear spin T2 = 39 min
at room temperature (from Ref. [14]).
temperature.
Quantum computing architectures that stand to benefit from or already have
demonstrated benefits from using enriched 28Si include STM hydrogen lithography
Si:P devices [17–20], single dopant qubits implanted near SETs [2,21–23], 28Si quan-
tum wells in Si/Ge heterostructures [24,25], and fabrication of transistors (FinFETs)
for QI [26]. A few of these groups have shown both long T2 times and coherent
manipulation in 28Si for bulk donor spins [27] as well as single spins in quantum
wells [24] and quantum dots [10,28].
In addition to long coherence times, using 28Si as a medium for 31P opens
up the possibility of optical manipulation of the qubit system through the use of
hyperfine transitions, which are unresolvable in natural Si. Typically in solid state
QI systems, in the absence of optical addressability, electrostatic control gates are
needed in close proximity to the dot to manipulate the qubit states and have the
9
Figure 1.5: Examples of experimental capabilities for qubit control in 28Si (a) 31P
in Si energy level diagram for 12 optical transitions between the neutral donor (D0)
and the donor bound exciton (D0X) (from Ref. [14]). (b) Photoluminescence spectra
(bound exciton no-phonon) of 31P atoms in natural abundance Si and 28Si showing
hyperfine splitting only resolvable in 28Si, making optical addressing possible (from
Ref. [29]). (c) ESR frequency tuning of a 31P electron spin in 28Si using a control
gate voltage to induce a Stark shift. This is due to the very narrow ESR linewidth
of spins in 28Si (from Ref. [4]).
possibility of introducing charge noise into the system [30]. Additionally, qubit ma-
nipulation schemes, which have been proposed for arrays of quantum dot qubits,
and which involve tuning the qubit ESR frequency through Stark or Zeeman shifts,
have been demonstrated in single quantum dots in 28Si [4]. The ability to tune the
qubit ESR frequency relies on qubit spins that have very narrow inhomogeneous
ESR linewidths of a few kHz, which are only achievable in a material with homoge-
neous mass such as highly enriched 28Si with exceptionally small strain fields [9,31].
10
Examples of the 31P hyperfine splitting and the tuning of a 31P ESR frequency are
shown in Fig. 1.5. Panel (a) shows an energy level diagram for 12 hyperfine-split
optical transitions of a 31P in Si from Ref. [14]. The transitions are between the
neutral donor (D0) levels and the donor bound exciton (D0X) levels and can be
excited with optical control pulses. Panel (b) shows the photoluminescence spectra
(bound exciton no-phonon) of 31P atoms in natural abundance Si and 28Si from
Ref. [29]. The 12 transitions of the hyperfine splitting are clearly resolvable only in
the 28Si sample, which makes it possible to address the spins with the optical pulse
in panel (a). Panel (c) shows experimental data from Ref. [4] displaying the change
in the ESR frequency of a 31P electron spin in 28Si due to a Stark shift induced by
a gate voltage. Multiple qubits can be addressed by one pulse signal using such a
technique by bringing only one at a time into resonance with the control pulse.
1.1.2 Sources of 28Si
Despite the demonstrated advantages of 28Si for solid state quantum informa-
tion, only a very limited amount of highly enriched 28Si is available within the solid
state quantum computing community for use in QI experiments. 28Si is primar-
ily produced at great cost and effort through international collaborations requiring
large centrifuge facilities. This lack of readily available 28Si is one of the prime mo-
tivations for using ion beam enrichment and deposition to produce 28Si films in this
work.
The majority of the 28Si bulk crystals and epitaxial films that have been pro-
duced are grown from chemical vapor deposition (CVD) of enriched 28SiH4. The
11
Figure 1.6: Process flow chart for the production of a 28Si bulk crystal at IKZ by
28SiH4 CVD, generated from centrifugation of natural abundance SiF4 at Centrotech.
(from Ref. [32])
production of this 28SiH4 has predominately started at the Centrotech facility in St.
Petersburg, Russia. There, natural abundance SiF4 is enriched using industrial gas
centrifuges. The enriched 28SiF4 is then converted chemically into the silane (
28SiH4)
used to grow 28Si crystals at the Leibniz Institute (IKZ) in Berlin, Germany [32]. A
process flow chart for this production of 28Si is shown in Fig. 1.6.
Production of 28Si early on in the 1990s in these efforts was spurred on by
interest in the thermal conductivity of 28Si. Evidence existed that 28Si had a higher
thermal conductivity than natural abundance Si. A higher thermal conductivity
was advantageous because heat dissipation was a major problem in the microelec-
tronics industry at the time. This early 28Si had a 28Si enrichment of about 99.9 %.
Measurements of the thermal conductivity of 28Si revealed that it significantly ex-
ceeded that of natural abundance Si, but only at cryogenic temperatures [33]. This
meant that 28Si was not a viable option to solve the heat dissipation crisis and thus
12
Figure 1.7: Thermal conductivity measurements of natural abundance Si and 28Si
as a function of temperature showing a large enhancement in thermal conductivity
for 28Si compared to natural abundance Si at low temperatures. (adapted from
Ref. [33])
demand for 28Si plummeted. The thermal conductivity measurements of 28Si and
natural abundance Si are shown in Fig. 1.7, adapted from Ref. [33].
Multiple 28Si bulk crystals were grown by CVD and zone refinement from
enriched 28SiH4 at IKZ. The largest of these crystals were grown as part of the
International Avogadro Coordination (IAC), which seeks to use single crystal 28Si
spheres as a standard to measure the Avogadro constant, NA, by counting the num-
ber of atoms in a kg of 28Si [32,34,35]. This effort is also related to the redefinition of
the kg unit using the kg 28Si sphere. The accuracy of the measurements of NA using
these spheres relies on them being very nearly perfect 1 kg 28Si spheres. Measure-
ments of their properties show that they are almost perfect single crystals with no
detectable dislocations. They have chemical impurity concentrations, including C
and O, of approximately 5 ×1014 cm−3, and they have a 28Si enrichment of approx-
13
Figure 1.8: 28Si bulk crystals produced by the International Avogadro Coordination.
(a) Schematic of the cutting plan to produce two 28Si spheres from a single-crystal
boule and (b) a photograph of the 28Si boule (from Ref. [32]). (c) Photograph of
one the final 28Si spheres on a weighing apparatus (from Ref. [34]).
imately 99.995 % with a residual 29Si isotopic concentration, i.e. the concentration
among the three Si isotopes, of approximately 50 ppm (parts per million, equal to
the isotopic concentration times 106) [32]. A schematic cutting plan for producing
two 28Si spheres from the 28Si boule produced for the IAC is shown in Fig. 1.8 (a)
from Ref. [32]. Panel (b) shows a photograph from Ref. [32] of the IAC single-crystal
28Si boule and panel (c) shows a photograph from Ref. [34] of one of the final spheres
sitting on a weighing apparatus.
An effect of producing this much 28Si is that the pieces leftover after forming
the spheres from the boule as well as from other bulk crystals produced at IKZ
were then able to be used for research in the QI community. This of course is a
14
limited supply of 28Si and producing more in this manner using centrifuges is both
extremely expensive and time consuming.
Some enriched 28SiH4 was also acquired by the Isonics Corporation, USA as
well as Dr. Kohei Itoh in the early 2000s who collaborated to grow CVD epilayer
28Si films on natural abundance Si substrates. Itoh has also grown bulk 28Si crys-
tals from this source. The 28Si enrichment of these materials was measured to be
approximately 99.927 % with a residual 29Si isotope concentration of approximately
730 ppm [36, 37]. The epilayers and other material from Itoh have been used by
research groups for QI experiments as well [10], although this material is also lim-
ited in supply. Some other less abundant sources of 28Si have also been used in QI
research although the exact details of those sources are difficult to verify.
1.1.3 Single Spin Measurements in 28Si
In addition to there being a general need for 28Si in semiconductor quantum
computing research, a specific need exists for material with targeted levels of enrich-
ment to map the dependence of T2 on
29Si concentration in the few-spin or single-spin
regime. Recent ESR measurements of T∗2 for single 31P spins in 28Si [10, 28] have
disagreed with both the theoretical predictions for the same systems made by Witzel
et al. [12], as well as each other. These theoretical predictions are shown in Fig. 1.9.
The T2 and T
∗
2 coherence times for quantum dots and
31P electron spin qubits in
Si are shown vs. the concentration of 29Si in the system. Solid and dashed lines
represent T2 times for
31P-donor and quantum dots, respectfully, i.e. the two archi-
tectures shown in Fig. 1.1. The theory predicts that for every order of magnitude
15
Figure 1.9: Predicted electron T2 (solid and dashed lines) and T
∗
2 (solid and dashed
line and symbols) coherence time dependance on 29Si concentration for qubits in Si
including quantum dots and 31P-donor atoms. Experiments measuring T∗2 for single
31P electron spins in 28Si represented by the triangle (from Ref. [10]) and diamond
(from Ref. [28]) do not agree with the prediction value (open circle). (figure adapted
from Ref. [12])
decrease in 29Si concentration, there is roughly an order of magnitude increase in the
T2 time. Bulk electron paramagnetic resonance (EPR) and ESR experiments of
31P
spins in Si with various enrichments including Si enriched in 28Si to Si enriched in
29Si [11] have agreed with this theoretical work over a large portion of the predicted
curve, as discussed by Witzel et al. in Ref. [12]. The dashed line with symbols
shows the T∗2 times for a quantum dot, and the solid line with symbols shows the
T∗2 times for 31P electron spins in Si. Also shown are the two results from the previ-
ously mentioned ESR experiments (triangle and diamond) measuring a 31P electron
spin T∗2 in 28Si with residual 29Si isotopic concentrations of approximately 800 ppm.
16
Both of these experiments involve a single 31P atom implanted in a 28Si SET device,
which is used for readout, and manipulated with a nearby ESR line. One result by
Muhonen et al. [10] gave a value of T∗2 = 268 µs, and the result by Tracy et al. [28]
gave a value of T∗2 = 18 µs, which was believed to be limited by experimental non-
idealities, e.g. magnetic field noise. The value predicted by the theory of T∗2 ≈
2 µs is highlighted by the open circle, which lies below both experimental results.
Although the experimental results give longer and thus more desirable coherence
times than the theory, this outcome shows the fundamental mechanisms limiting
coherence at the single-spin level require further study.
In order for the field of solid state quantum computing, especially utilizing
31P or other donor spin qubits, to continue to progress, the effects of nuclear spins
such as 29Si isotopes near a single qubit spin needs to be better understood. This
requires further measurements of coherence times in the few or single-spin regime
with varying concentrations of 29Si in the host material around the qubit atoms. The
goal of such a measurement would be to recreate the 31P T∗2 curve from Fig. 1.9
with additional experimental data and compare it to the existing theoretical curve.
28Si material with 29Si concentrations as low as 1 ppm would make the measure-
ment more robust and complete. The ability to measure single spins and predict
coherence times may ultimately be required for a viable quantum computer. This
concept is highlighted by a schematic ESR measurement phase space diagram of 29Si
concentration vs. number of measured spins, Ns, in Fig. 1.10. Ovals represent ESR
measurements that have been demonstrated already in QI research. These include
bulk EPR experiments on a large number of donor spins (e.g. > 1010) for a range
17
Figure 1.10: ESR measurement phase space diagram of 29Si concentration vs. num-
ber of measured 31P spins, Ns, in Si. Ovals represent ESR measurements that have
been made in the QI field including bulk ESR (i.e. > 1010 spins) for 31P in Si with
a range of enrichments (29Si concentrations), a large number of ESR measurements
of hundreds or thousands of 31P in natural abundance Si, and single spin ESR mea-
surements in natural abundance Si. Only two single spin measurements are known
to have been made of 31P in 28Si, represented by the star. These two measurements
are the two shown in Fig. 1.9. For a viable quantum computer, it is probable that
single spin measurements in highly enriched 28Si is necessary, represented by the
shaded region of the phase space.
of 29Si concentrations down to approximately 800 ppm [11, 27, 38], represented by
the large oval on the right. Bulk ESR experiments have also been done on 28Si with
an approximately 50 ppm 29Si concentration [39]. Experiments on smaller numbers
of donor spins ranging from 100s to 1000s have also been done but only in natural
abundance Si [40,41], which is represented by the oval at the top. In the single spin
regime, a large number of QI experiments have measured 31P spin coherence times
in natural abundance Si, but only the two previously mentioned experiments have
used enriched 28Si with a residual 29Si concentrations of approximately 800 ppm,
18
represented by the star. These two measurements are the two shown in Fig. 1.9
from Ref. [28] and [12]. The far bottom left corner of the phase space is where it
is suspected that measurements required for a viable quantum computer will reside
when considering scale-up, represented by the shaded region. It is this region of
single 31P spins in 28Si with 1 ppm 29Si concentrations where further research is
required. Enabling such measurements through production of highly enriched 28Si
epitaxial films with targeted enrichments is another aspect of the goals of this work.
1.2 Ion Beam Enrichment and Deposition
Si thin film epitaxial deposition can proceed by several techniques including
CVD using SiH4, molecular beam epitaxy (MBE) by thermal evaporation, ion as-
sisted deposition (IAD) which uses a separate source of ions to enhance deposition,
and direct ion beam deposition or ion beam epitaxy (IBE). IBE has two advantages
over the other methods. First, the energy of the ions can be used to enhance the
deposition process leading to higher quality epitaxy [42, 43]. Typically, hyperther-
mal energies below 200 eV are used, and this will be discussed further in Chapter
5. Second, ions can be mass filtered in a magnetic field to select a single isotope
of an atom for deposition. This means that ion beam deposition can be used to
isotopically enrich a material during the deposition process itself starting from a
natural abundance source. This is referred to here as in situ enrichment because
the enrichment occurs along the flight path of the ions from the source before being
deposited on a substrate to grow a film of enriched material, e.g. 28Si. Mass sepa-
19
rated ion beam deposition and epitaxy is the technique used in this work to produce
highly enriched 28Si films.
One of the earliest and most well known uses of ion beam enrichment was in
the calutron mass spectrometer developed by Ernest Lawrence for the the Manhat-
tan Project in the United States in the 1940s [44]. The calutron was used to generate
enriched quantities of the isotope 235U for use as fissile material in the development
of nuclear weapons. This was accomplished by ionizing natural abundance U con-
taining over 99 % 238U, accelerating the ions using electric fields, and then deflecting
the ion trajectory using a magnetic field as in a mass spectrometer. The magnetic
field separates the ions by mass, which generates an enriched ion beam of 235U that
is collected at a target. In order for this process to produce significant quantities of
enriched 235U, large, industrial scale apparatus were required, which were produced
at great cost and effort during the Manhattan Project.
Laboratory scale ion beam deposition systems have been studied and developed
since the early 1970s. Fair developed an ion beam deposition system and demon-
strated deposition of thin In films with energies between 100 eV and 500 eV [45].
Around the same time, Aisenberg and Chabot demonstrated deposition of diamond-
like thin films at room temperature using a beam of C atoms with 40 eV of energy,
which was transferred into the film to enhance the deposition [46]. Neither of these
early experiments involved mass selecting the ions. Later, in the 1980s and 1990s, a
number of other groups began experimenting with mass separated ion beam deposi-
tion and IBE. Shimizu et al. developed an ion beam deposition capable of producing
mass separated ion beams with mA level current and ion energies down to 10 eV.
20
Ar ion beams and Ca deposition was demonstrated [47].
Various groups have also demonstrated enrichment and ion beam deposition
of materials significant to quantum computing research including semiconductors
such as Si and Ge, for production of enriched Si/SiGe heterostructures and quan-
tum wells, and C, for nitrogen-vacancy centers in enriched diamond. Herbots et
al. demonstrated deposition of both 30Si and 74Ge ions with energies of 40 eV at
deposition temperatures including 400 ◦C [48]. Zalm and Beckers deposited mass
separated 28Si ions on both Si and Ge substrates [49], and Yagi et al. likewise de-
posited 28Si as well as 74Ge films with energies of 100 eV at deposition temperatures
of 300 ◦C [50].
28Si IBE was extensively studied by several groups including by Tsubouchi et
al. [43] and Rabalais et al. [51] who both developed dual ion beam deposition systems
for single and compound enriched materials [52, 53]. 28Si epitaxial deposition was
achieved by both groups using ions with energies of typically 20 eV at very low
deposition temperatures of 100 ◦C to 400 ◦C. These parameters produced epitaxial
thin films of 28Si with low defect densities. The highest 28Si enrichment reported
by Rabalais et al. was a film enriched to approximately 99.99 %. Tsubouchi et
al. reported a 28Si sample with a higher enrichment of approximately 99.9982 %
with a residual 29Si isotopic concentration of approximately 16 ppm, which is more
highly enriched than the material produced for the IAC. An isotope measurement
of this enriched 28Si sample as well as a schematic of the ion beam system used for
deposition are shown in Fig. 1.11. Panel (a) shows a dual source ion beam deposition
system schematic drawing from Ref. [53]. Two ion sources are seen connected to
21
Figure 1.11: Ion beam enrichment by Tsubouchi et al. used for depositing 28Si.
Panel (a) Schematic drawing of a dual source ion beam deposition system for en-
riching and depositing single or multi-component materials (from Ref. [53]). Panel
(b) Isotope depth profile of a 28Si film deposited from the ion beam. The isotope
concentrations of 28Si, 29Si, and 30Si are shown vs. the depth from the surface into
the film and Si substrate. The 28Si film is clearly enriched with the 29Si and 30Si
concentrations reduced to below 1018 cm−3, or approximately 16 ppm for 29Si (from
Ref. [43]).
two 90◦ magnetic isotope separators that both feed into a deposition chamber. This
system can be used to enrich and deposit single or multi-component materials. Panel
(b) shows a depth profile measurement of the enrichment of a 28Si film deposited
with the system in (a) from Ref. [43]. The concentrations of isotopes 28Si, 29Si, and
30Si are shown vs. the depth from the surface into the film and Si substrate. 29Si
and 30Si concentrations in the 28Si film are reduced to below 1018 cm−3, resulting in
the aforementioned enrichment values.
Rabalais et al. have also used their ion beam system to implant 74Ge ions
in SiO2 to create enriched Ge quantum dots [54], and deposited
28Si16O2 using the
dual beam setup with 28Si and 16O [55]. Tsubouchi et al. have used the dual beam
22
system to deposit enriched compounds including 28Si12C and 12C14N [56,57].
Finally, the work presented here relies heavily on the previous work done
using this same ion beam deposition system by the lead researcher of this effort,
Dr. Joshua Pomeroy. This system was used to deposit thin films of Cu using mass
separated ions and observe the effects of the ion energy on the epitaxial quality of
the film using STM [58,59].
1.3 Objectives and Outline
1.3.1 Project Goals
Enrichment and thin film deposition of 28Si is pursued here with the objective
of producing high-quality enriched material for solid state quantum computing. 28Si
of sufficiently high quality (i.e. high enrichment, crystallinity, and purity) provides
an ideal solid state environment to host qubit spins, as discussed previously. Un-
wanted deviations from ideal 28Si material can be classified as three types of defects:
isotopic defects, structural defects, and chemical defects. Controlling and limiting
these defects is critical for successful integration of 28Si into quantum computing
architectures. The 28Si materials goals of this work are stated as follows:
(1) high enrichment in 28Si with a residual 29Si isotopic concentrations less than
50 ppm,
(2) single-crystalline and smooth epitaxial structure with a low dislocation density
below 1× 106 cm−3, and
(3) high chemical purity including C and O with atomic concentrations below
2× 1015 cm−3.
23
These are believed (but not known) to be the criteria needed for the 28Si to be
comparable to single-crystalline electronic grade (EGS) natural abundance Si as well
as the enriched Si currently available in the QI research community from the IAC.
Electronic grade Si that is purified and crystallized into single-crystalline boules
of Si can have dislocation densities below roughly 1× 106 cm−3 [60]. Float zone
refinement also produces Si with atomic concentrations of most residual impurities
below 5× 1013 cm−3 and atomic concentrations of some impurities such as O below
1× 1018 cm−3 [61]. As mentioned previously, Si produced for the IAC has C and
O concentrations below roughly 5× 1014 cm−3. The crystallinity of this material is
nearly perfect with no detectable dislocations and a vacancy related defect density
of roughly 3× 1014 cm−3 [35]. Additionally, the IAC 28Si has a residual 29Si isotopic
concentration as low as 50 ppm. Producing 28Si with 29Si isotopic concentrations as
low as 1 ppm is necessary to enable a robust and systematic study measuring electron
coherence times vs. 29Si concentration in the single spin regime and compare it to
the theoretical predictions discussed previously.
Ultimately, the goal of this work is to produce 28Si material that can answer
the question “how good is good enough?” for quantum information. This means
determining the levels of enrichment, purity, and crystallinity that are necessary
to satisfy the materials needs of QI and solid state quantum computing devices.
These goals will be achieved using processing methods that are both common (e.g.
vacuum deposition, sample heating) and fairly unique (e.g. mass selected ion beam
deposition) to engineer the properties, such as enrichment in 28Si and chemical
purity, and structure (crystallinity) of Si thin films.
24
1.3.2 Strategy and Impact of Results
The experimental strategy used to achieve the materials goals described above
for this work is to use an ultra-high vacuum (UHV) ion beam deposition system to
prepare and deposit 28Si thin films several hundred nm in thickness. A hyperther-
mal energy ion beam is used to achieve in situ enrichment to very high levels from
a natural abundance silane gas (SiH4) source and deposit it on Si(100) substrates.
Clean substrates are prepared in situ in a UHV environment for minimal incorpo-
ration of chemical impurities and heated during deposition to facilitate epitaxial
deposition. Proof of principle experiments enriching Ne and C are used to establish
experimental techniques.
Characterization methods used to support this effort include in situ analy-
sis of the surface and crystallinity of 28Si films by reflection high energy electron
diffraction (RHEED) and STM. These are used for quick feedback to fine tune the
deposition parameters. Ex situ characterization used to assess the quality of the
films includes secondary ion mass spectrometry (SIMS) to analyze the enrichment
as well as the chemical purity, x-ray photoelectron spectroscopy (XPS) to determine
chemical purity, SEM to inspect the surface morphology, and transmission electron
microscopy (TEM) to inspect the crystallinity of the films. These characterization
methods are used for feedback on the deposition process to make informed decisions
on experimental changes leading to higher quality 28Si films.
The main impact on the QI field of this work is that 28Si was produced with
extremely high levels of enrichment. The most highly enriched sample produced
25
Figure 1.12: Enrichment progression timeline. A timeline of the progression of
samples with the lowest residual isotope fractions of 20Ne (diamonds), 13C (circle),
29Si (squares), and 30Si (triangles), as measured by SIMS. These were achieved for
22Ne, 12C, and 28Si samples produced over approximately five years.
by this work had an overall enrichment in 28Si of 99.9999819(35) % with a residual
29Si isotopic concentration of 127(29) ppb (parts per billion, equal to the isotopic
concentration times 106). This level of enrichment exceeds that of all other known
sources of 28Si by a factor of approximately 125. The enrichment of this sample
and other highly enriched samples is seen in the enrichment progression timeline
in Fig. 1.12 showing the best isotope fractions, which are a measure of the sample
enrichment, measured by SIMS for the minor isotopes 29Si (squares) and 30Si (tri-
angles) vs. the deposition date for selected samples produced throughout this work.
Isotope fractions of a particular isotope are defined in a SIMS measurement as the
26
detected counts of that isotope divided by the total counts of the measurement. Also
shown in this figure are measurements of the isotope fractions of 20Ne (diamonds)
and 13C (circles). These measurements were part of initial proof of principle exper-
iments producing enriched 22Ne and 12C samples done in preparation for enriching
and depositing 28Si. The isotope fractions here are written generally as zX/Xtot.,
where z refers to the mass number of a particular isotope, e.g. 28 for 28Si, X refers
to a particular element, and therefore zX is the counts of a particular isotope in
the measurement. Xtot. is the sum of the counts of all isotopes being measured.
Uncertainties in the isotope fractions are shown for all samples and are derived from
isotope measurements described in later chapters. The samples depicted here in this
enrichment timeline are those that had the best enrichments of any sample produced
up to that point. In other words, Fig. 1.12 is a timeline of the record enrichments
achieved in this work.
1.3.3 Outline
ˆ Chapter 2: The experimental apparatus and methods used to produce 28Si
thin films are presented. Descriptions of the hyperthermal energy ion beam-
line used for in situ isotopic enrichment, as well as descriptions of the UHV
deposition and analysis chamber used for substrate preparation and in situ
sample analysis are included.
ˆ Chapter 3: Initial proof of principle experiments enriching 22Ne and 12C
are discussed. 22Ne is implanted into Si while 12C thin films are deposited
27
on Si substrates. Three 22Ne and three 12C samples were produced. SIMS
enrichment measurements are presented showing a maximum achieved 22Ne
isotope fraction of 99.455(36) % and a maximum achieved 12C isotope fraction
of 99.9961(4) %.
ˆ Chapter 4: Phase I of 28Si deposition is discussed involving two experimen-
tal configurations. First, 28Si thin films are deposited in a proof of principle
experiment for Si enrichment producing five 28Si samples out of 61 total 28Si
samples produced in this work. Adjustments of the deposition parameters are
discussed for improved depositions producing 16 28Si films. SIMS measure-
ments of the enrichments and chemical purity are presented. A maximum 28Si
isotope fraction of 99.999888(10) % was achieved with a residual 29Si isotope
fraction of 0.691(74) ppm.
ˆ Chapter 5: Phase II of 28Si deposition is discussed involving an experimental
configuration that leverages the capabilities of the full system while depositing
40 28Si samples. SIMS enrichment measurements are presented showing a
maximum achieved 28Si isotope fraction of 99.9999819(35) % with a residual
29Si isotope fraction of 127(29) ppb for the most highly enriched sample in
this work. New substrate preparation procedures are discussed as is substrate
heating to enable epitaxial deposition. RHEED, STM, and SEM analysis
of 28Si film morphology and TEM analysis of film crystallinity is presented.
Finally, chemical purity analysis of 28Si films by SIMS is presented. Smooth,
epitaxial 28Si films were achieved for deposition temperatures between about
28
349 ◦C and 460 ◦C. These samples contain atomic concentrations of N, C, and
O slightly below 1 ×1019 cm−3.
ˆ Chapter 6: A model describing the adsorption of natural abundance SiH4
into 28Si films during deposition is presented and discussed. SIMS enrichment
values and SiH4 partial pressures are correlated using the model to extract a
room temperature incorporation fraction, s = 6.8(3)×10−4. The temperature
dependance of the sample enrichment is explored and an activation energy for
reactive SiH4 adsorption is determined to be Ec = 1.1(1) eV.
ˆ Chapter 7: A summary of the main scientific results is presented. Then,
experimental proposals enabled by this work are discussed including 28Si sam-
ples with targeted levels of enrichment for measuring T2, deposition of enriched
28Si/28Si74Ge quantum well heterostructures, 28Si re-deposition, Al dopant de-
vices, and several electrical measurements.
29
Chapter 2
Experimental Apparatus and Methods
2.1 Context
2.1.1 Ultra-High Vacuum Deposition
These experiments involving the deposition of 28Si from a mass selected ion
beam are conducted primarily in an ultra-high vacuum chamber. This provides the
cleanest possible environment to prepare clean, flat surfaces on substrates before
deposition. While the ion beam itself is not UHV, a gate valve separates it from the
UHV deposition chamber and the bulk of the higher pressures gases generated by
the ion source are differentially pumped before reaching the sample position. The
surfaces of 28Si samples are also inspected in UHV after deposition, which is critical
to prevent contaminants from adsorbing that would obscure the measurements. A
top-down schematic of the UHV and ion beamline deposition system is shown in
Fig. 2.1. The deposition system consists of four connected but isolated vacuum
chambers. The hyperthermal energy ion beamline is pictured at the left. The
UHV deposition and analysis chamber is at the right with a load lock branching
30
Figure 2.1: Top-down schematic of the ion beam deposition system including the
hyperthermal energy ion beamline pictured at the left, the UHV deposition and
analysis chamber at the upper right, the load lock used for sample loading into the
deposition chamber at the right, and the scanning tunneling microscope pictured
at the bottom. These four sections roughly separate four vacuum environments. A
smaller scale schematic showing the full length of the magnetic transfer arm used
to move samples from the deposition chamber into the STM is in the lower left.
31
off further to the right for sample loading. Separated from this chamber is the
scanning tunneling microscope pictured at the bottom, which is separated from the
deposition chamber by a gate valve. Photographs of the system are shown in Fig. B.1
in Appendix B.
2.1.2 Previous Operation
This work relies heavily on the previous work done using this same ion beam
deposition system by the lead researcher of the broader enriched Si project, of which
this work is a part, Dr. Joshua Pomeroy. This system was used by Pomeroy at
Cornell University to deposit thin films of Cu using mass separated ions and observe
the effects of the ion energy on the epitaxial quality of the film using STM [58,59], as
previously mentioned. Additionally, Baumann and Bethge have extensively tested
and utilized the same type of Penning ion source used in these experiments [62–64],
and one study in particular served as a reference for the operation of the source in
this work under various settings [65]. These tests showed the dependence of the ion
current on the gas flow, ion source magnetic field, and anode and cathode voltages.
Versions of similar ion source parameter tests generated using the ion source in this
work are shown in Fig. A.2 in Appendix A. Other references of previous operation
of the ion source used here include studies by Nouri et al. [66] and the Handbook of
Ion Sources by Wolf [67].
32
2.2 Hyperthermal Energy Ion Beamline
2.2.1 Apparatus
Much of the hyperthermal energy ion beamline used in this work was obtained
from Physicon Corporation (MA, USA) including two ion sources, the first section of
the beamline and housing chamber, and the magnetic sector mass analyzer. The first
element of the ion beamline is the Penning-type ion source, also called a Penning
Ionization Gauge (PIG) ion source. The principle of gas discharge, which is the
basis of this type of ion source, was first investigate by Phillips [68]. The Penning
source generates a plasma from a working gas in an electric field generated by a high
voltage between the anode and the cathode that is contained radially by a magnetic
field. This plasma ionizes and cracks gas molecules that are injected into the ion
source. Mostly singly charged ions are then extracted by a high voltage applied
to the extractor electrode, VExt, into the beamline. An einzel lens with the focus
electrode, VF , focuses and transports the ions at the transport voltage, VT , which
is typically about -4 kV. The transport voltage is applied to the entire ion beam
vacuum chamber, which is isolated from the chamber frame and ground. This high
transport voltage is needed in order to minimize the effect of space charge repulsion
of the positively charged beam, which would cause the beam to expand and become
unfocused. A schematic of the ion beamline is shown in Fig. 2.2 (from Ref. [59]).
This diagram shows the electrostatic elements of the beamline. Inset is a schematic
of the gas-mode Penning ion source showing the anode, cathodes, electromagnet,
33
Figure 2.2: Ion beamline schematic showing the electrostatic lens elements used to
focus and control an ion beam. Inset is a schematic of the gas-mode Penning ion
source showing the anode, cathodes, electromagnet, and extraction cusp (extractor).
The gas plasma forms in the anode between the cathodes. The potential energy
landscape seen by ions in the beamline is represented at the bottom of the figure
with labels A through K referencing the electrostatic elements in the beamline. Ions
are created in the source at close to the anode potential (A) and extracted with a
large negative potential by the extractor (B). Ions are then focused by an einzel lens
(D) and transported at a large negative potential (C). Deceleration lenses (C to K)
lower the ion kinetic energy to their starting energy for deposition at a grounded
target sample. (from Ref. [59])
and extraction cusp (extractor). At the bottom of the figure is a representation of
the potential energy landscape seen by ions in the beamline with labels A through
K corresponding to lens elements. A is the anode potential and the potential at
which the ions are created, B is the extractor, C is the transport potential, D is the
34
focus, E is an electron suppression (rejection) electrode that has a potential fixed at
approximately 100 V more negative than VT . Before the electron suppressor, there
are X-Y deflectors for steering the ion beam and a skimmer element with monitoring
current, Isk. F through K are the individual elements of the deceleration lenses that
focus and slow the ions to deliver them onto the sample, which is at ground potential.
Six lens elements are independently tunable and are referred to as lenses A2, A3, B2,
B3, B4, and X. A typical voltage applied to some of these lenses is roughly -1 kV.
Voltages for the deceleration lenes that were used when depositing a 28Si sample
are given in Table A.3 in Appendix A. Additionally, a schematic circuit diagram of
the power supplies and wiring of the various lens elements within the beamline is
presented there in Fig. A.1.
In order to isolate the transport voltage of the ion beam chamber and the
cathode voltage applied to the housing of the ion source, an insulating flange is used
to connect the two. While most components and flanges comprising the ion beam
chamber use UHV seals, the insulating flange uses o-rings to form the vacuum seals.
The components of the ion source itself are also sealed using o-rings. This means
that these components of the vacuum chamber containing the beamline are only
rated to high vacuum, i.e. a minimum achievable pressure of roughly 1.3× 10−7 Pa
(1.0× 10−9 Torr). The ion beam chamber is pumped using a 350 L/s turbo pump
(Pfeiffer Vacuum). The base pressure of the ion beam chamber ranged from ap-
proximately 2.7× 10−6 Pa to 1.3× 10−5 Pa (2.0× 10−8 Torr to 1.0× 10−7 Torr)
throughout the work presented here. While initially at the higher end of this range,
the base pressure was reduced at one point by removing a gate valve that was also
35
only rated to high vacuum. The components of the vacuum for the first and sec-
ond achieved base pressures can be observed using the residual gas analyzer (RGA)
in the ion beam chamber. Residual gas mass spectra representative of these base
pressures are shown in Fig. 2.3. Base pressure 1 (line) corresponds to the earlier
base pressure before removing the gate valve. Base pressure 2 (diagonal line fill)
corresponds to the lower achieved base pressure after removing the gate valve. Com-
mon vacuum components are observed in both spectra as peaks vs. their mass un
u (unified atomic mass unit) including H2, C, N, H2O, CO and N2 at 28 u, O2, and
CO2. Ar is also observed due to it being used to vent the chamber previously. When
a lower base pressure was achieved, the partial pressures of N at 14 u, N2 and CO
at 28 u, and O2 at 32 u were all reduced within the chamber.
After the ions are extracted and focused into a beam, they are transported
into a re-focusing magnetic sector mass analyzer, which bends the ion trajectories in
a 90◦ arc. The sector mass analyzer is a large electromagnet that is used to separate
the ions according to their mass-to-charge ratio, m/q. Ions of different mass-to-
charge ratios have different resulting trajectories. When the sector mass analyzer
is tuned to transmit ions with a particular value of m/q, that value is referred to
in discussions and in graphical representations of data only by the mass number,
e.g. the mass analyzer is said to be tuned to a mass of 28 u for 28Si ions with m/q
= 28 u/e. This is because ions generated from the ion beam and discussed in this
work are assumed to be singly charged, unless otherwise noted. This mass analyzer
is used to select a particular mass, u, and propagate those ions down the beamline.
At the exit of the mass analyzer is the mass-selecting aperture. Nominally,
36
Figure 2.3: Residual gas mass spectra collected from the RGA in the ion beam
chamber for two achieved base pressures. Partial pressure peaks of typical compo-
nents of the vacuum are seen including H2, C, N, H2O, CO and N2 at 28 u, O2,
and CO2. Ar is also observed due to it being used to vent the chamber previously.
The initial base pressure 1 (line) corresponds to the use of a gate value only rated
to high vacuum, while a later base pressure 2 (diagonal line fill) was achieved after
removing that gate valve. The partial pressures of N, N2 and CO, and O2 were
reduced with base pressure 2.
ions of a single mass pass through the aperture, while those with other masses are
rejected. Three mass-selecting apertures were used in the work presented here. Ini-
tially, an aperture that consisted of a circular hole approximately 5 mm in diameter
in a stainless steel spacer that was 16 mm thick was used to produce 22Ne and 12C
samples. Then, a much thinner Cu gasket aperture in the shape of a slit approx-
imately 1 mm in width, i.e. the same direction that different mass ion beams are
spatially separated, was used for initial 28Si depositions. The aperture was also ap-
proximately 15.25 mm tall and 2 mm thick. Finally, a second Cu slit approximately
2 mm in width and 12 mm tall with a beveled slit opening to reduce ion scattering
37
Figure 2.4: Ion beamline schematic showing the diffusion path of natural abundance
SiH4 (“clouds”), including
29SiH4, from the ion source inlet (lower left), through the
magnetic sector mass analyzer (top left), past the mass-selecting aperture and into
the deposition chamber where the sample is located (upper right). The inset cartoon
shows a magnification of the aperture area with mass separated Si ions. 29Si ions
are blocked by the aperture while the 28Si ions as well as 29SiH4 (and other) gas
molecules pass into the depositions chamber.
off of the aperture was used to deposit the remaining 28Si samples. Photos of these
three mass-selecting apertures are shown in Fig. B.5 to B.7 in Appendix B.
Simultaneous with ions passing through the aperture, gas from the ion source
diffuses along the beam path and through the aperture as well. This source gas
is natural abundance and results in a partial pressure of unwanted isotopes at the
sample, which may be incorporated into the sample during deposition. This phe-
nomenon is discussed in great detail in Chapter 6. Figure 2.4 shows a schematic
38
of the ion beamline with a cartoon representation of the mass selection and gas
diffusion for the case of 28Si deposition. Natural abundance silane gas (SiH4) is used
as the source for generating Si ions in this work. This gas has a purity of 99.999 %
according to the gas vendor (Matheson Tri-Gas). Gas is injected into the ion source
using a UHV leak valve from a gas manifold used to regulate different gases used for
operation of the ion source. A photograph of the gas manifold is shown in Fig. B.2
in Appendix B. When depositing 28Si, the natural abundance SiH4 diffuses from the
ion source down the beamline and to the sample location in the deposition chamber,
as represented by the “clouds” in the figure. The inset shows a magnification of the
mass selection process occurring at the mass-selecting aperture. 29Si (and 30Si) ions
are blocked by the aperture while the 28Si ions and SiH4 gas molecules pass into the
deposition chamber.
Beyond the mass-selecting aperture are the deceleration lenses. As mentioned,
the purpose of these einzel lenses is to maintain the focus or de-focus of the ions while
decelerating them from the transport voltage mentioned above to ground potential
at the sample. The final kinetic energy of the ions depends on the voltage at which
they were created in the ion source (plasma potential), which is typically similar
to the positive voltage applied to the anode. This is discussed in more detail later
in this chapter. Photographs of the electrostatic and magnetic elements comprising
the ion beamline are shown in Fig. B.4 in Appendix B.
39
2.2.2 Theory of Magnetic Mass Separation
When charged ions with different masses but otherwise equal properties tra-
verse a region of magnetic field, they will follow circular motion. Each ion of a
particular mass will follow different circular trajectories, resulting in a physical sep-
aration of the ions by mass. This principle of magnetic mass separation of ions used
in mass spectrometers relies primarily on the magnetic Lorentz force, FB, which is
the force exerted on a charged particle in a magnetic field. The magnetic Lorentz
force is given by
FB = qv ×B, (2.1)
where q is the charge state of the particle, here an ion, v is the ion velocity, and B
is the magnetic field experienced by the ion. The cross product results in a force,
FB, acting on the ion in a direction perpendicular to the direction of motion. This
force thus results in circular motion of the ion.
In order to determine how ions of different masses are separated as a function
of the magnetic field in terms of the radii of curvature of their trajectories, a general
equation for circular motion is used. Circular motion of any object can be described
as being the result of a centripetal force, Fc, which acts on the object in the direction
of the center of the circular path. Fc is given by
Fc =
mv2
r
, (2.2)
where m is the mass of the object, v is the velocity, and r is the radius of curvature
of the circle defining the object’s trajectory. The centripetal force and the magnetic
40
Lorentz force can then be equated, FB = Fc, because they both describe circular
motion. This expression is solved to get the radius of curvature of the ion trajectory,
r, in terms of the mass, velocity, charge state, and magnetic field, yielding
r =
mv
qB
. (2.3)
This means that for ions with different masses but equal charge states and velocities,
the radius of curvature of the ion’s motion is proportional to its mass. However, for
ions with different masses generated in a beamline and accelerated by electric fields,
their velocities will not be equal.
An expression for the ion velocity in a beamline can be determined from con-
sidering the ion kinetic energy, EK , due to acceleration by an electric field generated
from an applied voltage. For this scenario, the potential energy, EP , of the ion in
the electric field is transformed into its kinetic energy. Thus, setting EK = EP gives
the expression
mv2
2
= qV, (2.4)
with the kinetic energy term on the left side of the equation and the potential energy
on the right. V is the accelerating voltage used to generate the ion beam. Then,
solving for v yields
v =
√
2qV
m
. (2.5)
This expression for v is substituted into Eq. (2.3) yielding a general expression for
the mass dependance of the radius of curvature of an ion trajectory given by
ri =
1
B
√
2miV
q
. (2.6)
41
Here, ri is the radius for an ion with index i corresponding to a mass of mi. In other
words, Eq. (2.6) shows that for ions with equal charge states of q and accelerated
with voltage V that pass through a magnetic field of B, different masses, mi, will
follow trajectories with different radii of curvature, ri. It is from this equation
that the mass-to-charge ratio, m/q, is seen to be the important parameter for mass
separation in an ion mass spectrum.
In a mass spectrometer and in the ion beamline used in this work, the radius
of curvature of the sector mass analyzer is fixed, giving a single value of the radius
for selected (transmitting) ions, r0. Rearranging Eq. (2.6) to get an expression for
B and substituting q = 1 e and r0 = 97.9 mm gives
B = (1.47× 10−3)
√
mV , (2.7)
which is the magnetic field in units of T required for selecting ions with a given
mass, m, in units of u using a given acceleration voltage, V , in units of V for the
beamline used in this work. The quoted value of r0 is an approximation determined
from analysis similar to Eq. (2.6) using experimental values. A plot of B vs. m
calculated from Eq. (2.7) is shown in Fig. 2.5 (a). This calculation uses a value for
the accelerating voltage into the sector mass analyzer of V = 4040 V, which is a
typical value for the operation of the beamline in this work. B ranges from 0.09 T
at 1 u to 0.84 T at 80 u. For 28Si at 28 u, B = 0.49 T in this calculation.
Panel (b) of Fig. 2.5 shows the relation between the current applied to the
sector mass analyzer magnet and the resulting mass of the ions being selected for
accelerating voltage values of 4040 V (line) and 4740 V (dashed line). An accel-
42
Figure 2.5: Calculated relation of the sector mass analyzer magnetic field and applied
current to the selected mass in the ion beam. These calculations are for singly
charged ions and use a value for the sector mass analyzer radius of curvature of r0 =
97.9 mm. (a) The magnetic field required to select a certain mass (line) is calculated
from Eq. (2.7) as a function of mass. This calculation uses an accelerating voltage V
= 4040 V, typical of the operation of the beamline. (b) The selected mass is shown
as a function of the current applied to the mass analyzer magnet for accelerating
voltages of V = 4040 V (line) and V = 4740 V (dashed line). These curves are
calculated from Eq. (2.9), which is an experimentally derived conversion from the
calculated field to the current.
erating voltage of V ≈ 4740 V is also used for operation of the ion beamline in
addition to 4040 V. These curves are determined from the curves in panel (a), by
converting the calculated values of B into the applied current based on experimen-
tal measurements. B and the applied magnet current, Imag, are linearly related, as
given by
B = (9.69× 10−3)Imag + (3.43× 10−3). (2.8)
The slope and intercept of this linear equation were experimentally determined.
Here, Imag is in units of A and B is in units of T. Substituting Eq. (2.8) for B in
Eq. (2.7) and solving for m yields an expression relating the selected mass and the
43
applied current given by
m =
((6.59)Imag + 2.33)
2
V
, (2.9)
where, Imag is in units of A, V is in units of V, and m is in units of u. In Fig. 2.5
(b), values of Imag between approximately 10 A and 90 A are needed to select ions
with masses from 1 u up to about 80 u. Inverting Eq. (2.9) yields an expression for
Imag as a function of m and V given by
Imag = (0.15)
√
mV − 0.35. (2.10)
Again, m is in units of u, V is in units of V, and Imag is in units of A. This equation
gives the applied current needed for the mass analyzer in this work to select ions
with a particular mass for a given accelerating voltage. The mass analyzer current
needed for 28Si at 28 u for V = 4040 V is Imag ≈ 50.1 A.
For a mass analyzer with a 90◦ bend, the physical separation, λ1, at a mass-
selecting aperture of the trajectories of ions with different masses can be calculated
using geometry and trigonometric relations. The two parameters that are required
for the calculation are the radius of curvature of the analyzer, r0, and the distance,
Y , from the exit of the analyzer to the aperture where mass selection occurs. An
expression for λ1 as a function of r0, Y , and two masses is given by
λ1 = r0
({
m1
m0
} 1
2
cos
{
sin−1
[
1−
(
m0
m1
) 1
2
]}
− 1
)
+ Y tan
{
sin−1
[
1−
(
m0
m1
) 1
2
]}
, (2.11)
where m0 is the mass of the ions being selected and passing through the aperture
and m1 is the mass of another ion not being selected. For this equation, m1 >
44
Figure 2.6: Calculated mass dependance of the spatial separation, λ1, of ions with
adjacent mass number in a mass separated ion beam at the mass-selecting aperture.
These calculations (line) use values for the physical parameters of the ion beamline
including the sector mass analyzer radius of curvature of r0 = 97.9 mm and the
distance from the analyzer outlet to the mass-selecting aperture of Y = 215 mm.
The spatial separation between ions of a given mass and ions of one mass unit, u,
higher decreases with increasing mass number from about 100 mm between masses
1 u and 2 u to about 2 mm between masses 80 u and 81 u. This calculated curve
is derived from Eq. (2.11). The datum (triangle) is the result of measuring the
distance between the aperture center and a deposited mark on the aperture due to
the 29 u ion beam.
m0 in general, and in this case, m1 is adjacent to m0, i.e. m1 = m0 + 1 u. λ1,
the distance between m0 and m1 ion trajectories, depends only on the physical
parameters (geometry) of the system and the two masses, and it is independent of
the energy of the ions or the magnetic field used.
The dependance calculated from Eq. (2.11) of λ1 as a function of mass, m0,
is shown in Fig. 2.6. A measured value for the distance to the aperture of Y ≈
215 mm was used for this calculation along with a value for the analyzer radius of
45
curvature of r0 = 97.9 mm. The calculated λ1 values (line) are highest at lower
masses, decreasing with increasing mass. This shows that the physical distance
between trajectories at the aperture of ions with adjacent mass numbers ranges from
approximately 100 mm between 1 u and 2 u to approximately 2 mm between 80 u
and 81 u. For 28Si at 28 u, the calculated distance to the 29Si beam at 29 u is λ1 ≈
5.5 mm. A 2 mm wide mass-selecting aperture then allows a mass-to-charge range
of approximately ± 0.18 u/e when selecting 28 u/e ions. This distance is compared
to a measurement of the distance between these two beams (triangle), which has
a similar value of approximately 6.2 mm. This measurement was made from the
center of the aperture to the center of a visible deposition spot on the aperture
created from the 29 u ion beam after depositing 28Si samples. A photograph of this
measured aperture is shown in Fig. B.6 in Appendix B. The significance of λ1 being
larger at lower masses is that ions at those masses have a better isolation with ions
at adjacent masses leading to larger geometric selectivites for a selected mass.
Ions with the same mass but different kinetic energies due to different accel-
erating voltages will also have different radii of curvature in a mass analyzer. The
physical separation of these ion trajectories at the aperture can be calculated in a
similar manner as Eq. (2.11). For the case of a small kinetic energy spread amongst
otherwise identical ions due to slight variations in the accelerating voltage, ∆V , a
slight spreading of the ion trajectories would result. The physical size of this spread-
ing of trajectories at the mass-selecting aperture, λE, from a selected ion to an ion
46
of the same mass but with a slightly higher kinetic energy is given by
λE = r0
({
1 +
∆V
V
} 1
2
cos
{
sin−1
[
1−
(
1 +
∆V
V
)− 1
2
]}
− 1
)
+ Y tan
{
sin−1
[
1−
(
1 +
∆V
V
)− 1
2
]}
. (2.12)
This expression shows that the spreading of ion trajectories due to an energy spread,
λE, for ions of a given mass depends only on the physical parameters of the system
and the accelerating voltage and is independent of mass or magnetic field. A plot
of λE calculated from Eq. (2.12) as a function of the accelerating voltage, V , for
different values of the variation in accelerating voltage, ∆V , is shown in Fig. 2.7.
Values in this calculation for the distance to the aperture (Y = 215 mm) and the
analyzer radius of curvature (r0 = 97.9 mm) were the same as previously used. λE is
calculated for ∆V = 5 V (line), ∆V = 7.5 V (dotted line), and ∆V = 7.5 V (dashed
line). For ions accelerated with a nominal voltage of 4040 V with a typical ∆V =
6 V, λE ≈ 0.23 mm. It is observed from these calculations that λE decreases with
increasing V , and although the effect is relatively small, it is therefore advantageous
to transport ions in the beamline at higher voltages. For V increased to 4740 V
in the previous calculation, λE is reduced to approximately 0.20 mm. For λE =
0.23 mm, the total width of the trajectories due to this spread is 0.46 mm, which is
equivalent to a mass range of 0.08 u at a mass of 28 u.
From the calculated values of λ1, the mass resolving power,
m
∆m
, of the ion
beam system and sector mass analyzer can be calculated as
m
∆m
=
m0λ1
(m1 −m0)∆λ , (2.13)
47
Figure 2.7: Calculated accelerating voltage dependance of the spatial separation,
λE, of ions accelerated with voltages differing by ∆V in the ion beam at the mass-
selecting aperture. These calculations use values for the physical parameters of
the ion beamline including the sector mass analyzer radius of curvature of r0 =
97.9 mm and the distance from the analyzer outlet to the mass-selecting aperture
of Y = 215 mm. λE is calculated for ∆V values of 5 V (line), 7.5 V (dotted line),
and 10 V (dashed line). λE decreases with increasing accelerating voltage by about
35 % from 3500 V to 5000 V. These calculated curves are derived from Eq. (2.12).
where ∆λ is the width of the selected ion beam at mass m0. m1 is the larger mass
used in the calculation of λ1 in Eq. (2.11) and here, m1 −m0 = 1 u. For the mass
resolving power of a mass spectrometer, the numerator represents the selected mass,
here m0, and the denominator, ∆m, is a variant of the width of the selected ion
beam. Calculated mass resolving power values from Eq. (2.11) and (2.13) are shown
as a function of mass in Fig. 2.8. Values in this calculation for the distance to the
aperture (Y = 215 mm) and the analyzer radius of curvature (r0 = 97.9 mm) were
the same as previously used. Additionally, a value of the beam width of ∆λ =
2 mm was used, which is equal to the aperture width used for much of this work.
48
Figure 2.8: Calculated mass dependance of the mass resolving power, m
∆m
, of the
ion beamline. These calculations (line) use values for the physical parameters of the
ion beamline including the sector mass analyzer radius of curvature of r0 = 97 mm,
and the distance from the analyzer outlet to the mass-selecting aperture of Y =
215 mm. Additionally, they use an ion beam width, ∆λ = 2 mm. This calculated
curve is derived from Eq. (2.13). The datum (triangle) is the highest measured mass
resolution for a 28Si ion current peak.
The calculated values of m
∆m
(line) increase with increasing mass, which is counter
intuitive because the ion beams of lower masses are better separated than those at
higher masses. m
∆m
increases sharply from a value of approximately 50.2 at a mass of
1 u to approximately 73.4 at 10 u and then increases much slower at higher masses
up to approximately 77.6 at 80 u. A single experimental value of m
∆m
(triangle)
for the best mass resolving power measured in this work is shown for comparison.
Although other experimental data exists, it can be problematic comparing them if
the total ion beam current at various masses is very different because the beam
widths can depend on the total current, which is often the case.
49
2.2.3 Operating Parameters
Two versions of the Penning ion source were used in this work. One with
a thin disc-like anode that is efficient at producing ions by sputtering the cathode
material, and another with a longer cylindrical anode that is more efficient at ionizing
atoms of the working gas. Schematic diagrams of the two ion sources are shown in
Fig. 2.9. These schematics are viewed as a cross-section through the middle of the
ion sources in a plane parallel to the axis of the ion beamline. Panel (a) shows
a schematic obtained from Physicon Corp. of the gas-mode ion source with the
gas inlet at the right and the ion beam exit facing the left with an outlet cone.
This source uses an anode that is long and cylindrical that is surrounded on either
side by the inlet and outlet cathodes. These cathodes have holes through their
middles to allow gas to enter the source at the inlet side and ions to exit at the
outlet. The anode and cathodes are isolated from each other by a ring shaped
insulator seen inside the back plate. As mentioned, the voltage between the anode
and cathodes, Varc, was typically about 3 kV, and it was used to generate the gas
plasma between the cathodes and inside the anode (shaded oval). Surrounding the
anode and cathode region is the source electromagnet solenoid. The anode is in the
middle of the solenoid in this source. The source magnet provides a magnetic field
(dashed curves) of typically 60 mT that radially confines the ions in the plasma
within the anode. Ions are extracted from the plasma by an electric field generated
by the high voltage applied to the extractor electrode, Vext. Photographs of the
anodes and cathodes of the two ion sources are shown in Fig. B.3 in Appendix B.
50
Figure 2.9: Schematic diagrams of the gas-mode and solids-mode Penning ion
sources. These sources are distinguished primarily by the size and configuration
of the anode and cathodes. (a) The gas-mode ion source is viewed in cross-section
through the middle with the gas inlet at the right and the ion beam exit at the left
(from Physicon Corp.). This source uses a long cylindrical anode surrounded by
inlet and outlet cathodes with holes through each. The voltage between the anode
and cathodes, Varc, was typically 3 kV and was used to generate the plasma between
the cathodes (shaded oval). The source electromagnet solenoid surrounds the anode
and cathode section, providing a magnetic field (dashed curves) of typically 60 mT
to radially confine the ions in the plasma. (b) The solids-mode ion source is viewed
in cross-section in a similar orientation to the schematic in (a) but with only the
middle area visible (from Ref. [62]). This source uses a short anode in the shape
of a disk with a hole in the center surrounded by an outlet cathode with a hole on
the left and a solid inlet cathode at the right. The plasma (shaded oval) is confined
between the cathodes in a smaller volume than in the source in (a). Although the
magnetic field is not shown, the electromagnet is used in a similar manner as in (a).
Panel (b) of Fig. 2.9 shows a schematic cross-section from Ref. [62] of a portion
of the solids-mode ion source in the same orientation as the source in (a). This source
uses an anode that is in the shape of a thin disk with a hole in the center. As with
the gas-mode source, the anode is surrounded by the inlet and outlet cathodes. The
51
outlet cathode on the right of the diagram is the same used for the outlet of the
gas-mode source with a hole through it for ions to escape. The inlet cathode is
larger and is solid without a hole in order to provide a surface for sputtering. Gas
is injected into this source through holes at the base of the inset cathode support.
Although not shown, the magnetic field used is similar to that of panel (b) but
with anode and cathodes offset from the center of the solenoid. The plasma region
is again between the cathodes and within the anode (shaded oval), however, the
volume is much smaller in this source than for the gas-mode source.
The solids-mode ion source was used to produce both 22Ne and 12C samples in
this work. Working pressures of the source gas injected into the ion source ranged
from 1.1× 10−3 Pa to ≈ 1.0× 10−2 Pa (8.0× 10−6 Torr to 7.8× 10−5 Torr). The
anode voltage, VA, was typically about 500 V or more and the cathode was typically
about -500 V. The source electromagnet typically current ranged between 1.0 A and
1.5 A. Arc currents, Iarc, between 15 mA and 30 mA were observed.
The gas-mode ion source was used to produce 28Si in this work. There are two
operating regimes of the gas-mode ion source. These are the typical “low pressure”
plasma mode operating condition used to deposit the majority of 28Si samples in this
work, as well as a “high pressure” mode that was used to deposit several samples.
The typical working pressure used when injecting gas into the source for the low
pressure mode was between 1.0× 10−4 Pa and 3.3× 10−4 Pa (7.5× 10−7 Torr to
2.5× 10−6 Torr). The operating pressure of the high pressure plasma mode was
typically 1.3× 10−3 Pa (1.0× 10−5 Torr). Anode voltages, VA, between +30 V and
+700 V were used for various depositions throughout this work to generate ions with
52
a similar range of final ion kinetic energies, Ei. The voltage of to the cathode, VC ,
was between -1 kV and -4 kV. This voltage was floating on top of the anode voltage,
and the applied difference between the two was referred to as the “arc” voltage, Varc.
The source electromagnet typically requires 1.5 A to 2 A to enable ignition of the
plasma, which corresponds to magnetic fields of approximately 40 mT to 80 mT.
A typical arc current, Iarc, of 0.5 mA was observed for this source. Additional
operating parameters for both ion sources including those used to deposit the most
highly enriched 28Si sample produced in this work are shown in Tables A.1 to A.3
in Appendix A.
The operation and tuning procedures used with the ion beamline to produce
a stable and focused ion beam are described in the following steps.
1. Inject the source gas into the ion source using the leak valve at an appropriate
pressure, depending on the source and plasma mode being used,
2. set the anode voltage, VA, to a small positive voltage and the arc voltage, Varc,
negative with reference to VA to define the cathode voltage, VC ,
3. set the transport voltage, VT , to typically -4 kV,
4. set the extractor voltage, VExt, which floats on and is more negative than VT ,
and
5. set the focus voltage, VF , which also floats on and is more negative than VT
and VExt.
6. With these voltages set, turn on the source electromagnet and as the current,
ISM , is increased, watch the current reading on the transport power supply,
IT . This will ignite a plasma in the source and then set ISM for when IT is
observed to increase and stabilize.
7. Then, set the magnetic sector mass analyzer to a current corresponding to
the desired ion mass and power on the deceleration lenses, monitoring the ion
current at the target.
53
8. Tune the ion beam by adjusting each control element starting from the ion
source and moving down the beamline to the deceleration lenses, including
the source pressure, ISM , Varc, VExt, VF , and the X-Y deflectors, all while
maximising the total detected ion current.
9. Finally, tune the deceleration lenses by adjusting the voltages of lenses A2, A3,
B2, B3, B4, and X in order while maximising the ion beam current through
the sample aperture.
2.2.4 Ion Beam Characterization
2.2.4.1 Ion Beam Mass Spectra
Several techniques were used to characterize each ion beam in preparation for
deposition of enriched materials. The components of the ion beam can be observed
by sweeping the current, and thus the magnetic field, of the sector mass analyzer
magnet, which corresponds to sweeping the selected mass, and recording the mea-
sured ion current at the sample location. This type of measurement generates a
mass spectrum of the components of the ion beam. Ion currents are measured using
a picoammeter with a typical noise floor of 10 pA for the range setting typically
used. Several working gases are used in the ion source throughout this work includ-
ing Ne, carbon dioxide (CO2), and SiH4 for
28Si deposition. Mass spectra for Ne and
CO2 are discussed in Chapter 3, and mass spectra for SiH4 are discussed later in
this chapter and throughout this work. Collecting and analyzing a mass spectrum
is an important aspect to preparing a mass separated ion beam for deposition. The
mass spectrum is used for tuning the mass analyzer to the desired ion, assessing the
isolation of that ion beam from adjacent beams, and detecting contaminants within
54
the vacuum system, among other things.
The mass spectra collected while using several other working gases of Ar, N2,
and methane (CH4) in the ion source are shown in Fig. 2.10. The measured ion
currents (circles and line) are shown as a function of the ion mass-to-charge ratio,
m/q. As discussed previously, in this work, mostly singly charged ions are produced,
so the mass-to-charge ratio is simply referred to and represented as the ion mass
when discussing mass spectra. The current applied to the sector mass analyzer used
for the magnetic field sweep is shown on the top axes. The spectra in panels (a) and
(b) were acquired using the gas-mode ion source, while the spectrum in panel (c)
was acquired using solids-mode source. Panel (a) is a portion of a mass spectrum
for Ar showing current peaks corresponding to 36Ar (36 u), 38Ar (38 u), and 40Ar
(40 u). The relative sizes of these peaks are similar to the natural abundance of Ar,
which is comprised of approximately 0.33 % 36Ar, 0.07 % 38Ar, and 99.6 % 40Ar.
Panel (b) of Fig. 2.10 is a portion of a mass spectrum for N2 showing current
peaks corresponding to 13C (13 u), 14N (14 u), 15N (15 u), and 16O (16 u). The
presence of C and O in the beam is due to partial pressures of molecular species
containing those elements within the vacuum chamber or possibly contamination in
the N2 gas itself. The peak at 14.5 u likely corresponds to a doubly-charged ion of
mass 29 u. A large shoulder peak is seen on the lower mass side of the 14N peak
and is likely a result of poor ion beam focusing causing ion scattering from a lens
element. The relative sizes of the 14 u and 15 u peaks are similar to the natural
abundance of N, which is comprised of approximately 99.6 % 14N and 0.4 % 15N.
Panel (c) of Fig. 2.10 is a portion of a mass spectrum for CH4 showing current
55
Figure 2.10: Ion beam mass spectra for different working gases used in the ion source.
Ion currents (circles and line) are recorded while sweeping the mass analyzer current,
and thus the magnetic field (top axes) (a) Ar mass spectrum showing current peaks
corresponding to 36Ar (36 u), 38Ar (38 u), and 40Ar (40 u). The relative sizes of
these peaks are similar to the natural abundance of Ar. (b) N2 mass spectrum
showing current peaks corresponding to 13C (13 u), 14N (14 u), 15N (15 u), and 16O
(16 u). The peak at 14.5 u likely corresponds to a doubly-charged ion of mass 29 u.
A large shoulder peak is seen on the lower mass side of the 14N peak due to poor ion
beam focusing causing scattering from a lens element. The relative sizes of the 14 u
and 15 u peaks are similar to the natural abundance of N. (c) CH4 mass spectrum
showing current peaks corresponding to 12C (12 u) and various C hydrides that are
cracked in the ion source plasma. The peak at 13 u is mostly 12CH with a small
amount of 13C. Likewise, the peaks at 14 u, 15 u, 16 u are mostly 12CH2,
12CH3,
12CH4, respectively.
56
peaks corresponding to 12C (12 u) as well as various C hydrides that are the result of
different numbers of H atoms being removed when CH4 is cracked in the ion source
plasma. The ion current peak at 13 u is mostly 12CH with a small amount of 13C.
Similarly, the peak at 14 u is mostly 12CH2 and possibly
14N, the peak at 15 u is
mostly 12CH3, and the peak at 16 u is mostly
12CH4 and possibly
16O.
As mentioned, natural abundance SiH4 gas is used for
28Si deposition and
several spectra representing various deposition conditions for different samples will
be presented in chapters 4, 5, and 6. Figure 2.11 shows a portion of a mass spectrum
of the ion beam collected while using SiH4 as the working gas. The corresponding
magnetic sector mass analyzer current used for the field sweep is shown on the top
axis. This spectrum was generated while using the gas-mode ion source. Ion current
peaks (circles) corresponding to 28Si, and various Si hydrides are observed from 28 u
to 33 u. The ion current peak at 28 u is 28Si, and the ion current peak at 29 u is
mostly 28SiH containing ≈ 5 % 29Si. This estimated relative 29Si is based on the peak
heights of 28 u and 29 u being similar and the expected natural abundance of 29Si
relative to the 28Si at 28 u. Similar to the peak at 29 u, the peak at 30 u is mostly
28SiH2, the peak at 31 u is mostly
28SiH3, and the peak at 32 u is mostly
28SiH4. The
peak at 33 u is likely a combination of 30SiH3 and
29SiH4. These molecular species
are formed by SiH4 being cracked in the ion source and losing different numbers of H
atoms. The overall efficiency of generating 28Si is fairly low as it makes up roughly
10 % of the total SiH4 ion current. The
28Si ion current observed in this spectrum
is Ii ≈ = 620 nA, which is typical for 28Si depositions.
The 28 u peak and the 29 u peaks in Fig. 2.11 show a high degree of separa-
57
Figure 2.11: SiH4 mass spectrum representative of the ion beam settings for deposi-
tion of many 28Si samples. The ion current (circles) is recorded while sweeping the
mass analyzer current, and thus the magnetic field (top axis). Gaussian fits (line,
Eq. (2.14)) to the 28 u, 29 u, 30 u, and 31 u peaks are superimposed on the data.
The peak at 28 u is 28Si and the peak at 29 u peak is 28SiH and ≈ 5 % 29Si. Sev-
eral higher order hydrides are also observed corresponding to mostly 28SiH2 (30 u),
28SiH3 (31 u), and
28SiH4 (32 u). These molecular species are cracked in the ion
source plasma. The centers of the 28 u and 29 u fits are separated by ≈ 10 σ.
tion with no detectable ion current signal occurring between the peaks. Secondary
electrons generated by the ion beam cause the current between the peaks to be
≈ -0.5 nA. Also shown is a sum of Gaussian fits to the data (line) of the form
Ii = I
0
i +
B
σ
√
2pi
exp
(
−1
2
(
m−mc
σ
)2)
, (2.14)
where Ii is the measured ion current, I
0
i is an offset to the current due to the
noise floor of the measurement, B is the area of the Gaussian, σ is the standard
deviation, m is the mass, and mc is the center mass of the peak. The value of
58
I0i used for fits to ion current peaks is typically 10 pA, which is the noise of the
picoammeter mentioned previously. One can see that the data matches the form of
a Gaussian very well. This indicates a symmetric and optimally tuned beam shape
with minimal perturbations such as scattering. The ion current peaks generated in
this system are expected to be approximately Gaussian due to several factors. A
Gaussian shaped current peak can result only if the width of the ion beam is greater
than the aperture width, which is the case for this system. For a beam with a width
smaller than the aperture width, the current peak would have a flat top as the full
beam traversed the aperture. The approximately Gaussian shaped profile represents
a spacial distribution of beam fluxes that likely originates at the ion source where
a higher flux of ions is extracted from the center of the source exit and a smaller
distribution coming from the edges.
For this mass spectrum in Fig. 2.11, the centers of the 28 u and 29 u fits are
separated by ≈ 10 σ (standard deviation). The mass resolving power of the system
in this configuration can also be derived from this mass spectrum as m
∆m
≈ 58 at
28 u. ∆m in this calculation is determined by taking the full width of the fit to
the 28 u peak at 10 % of the peak height, which is approximately 0.48 u. This
width is equivalent to a physical width at the mass-selecting aperture of 2.6 mm.
As mentioned previously, for a typical ion energy spread of ± 6 eV due to a spread of
accelerating voltages in the beam of ∆V = 6 V, the width in mass due to the energy
spread at mass 28 u is approximately 0.08 u. This width due to the energy spread is
therefore approximately 17 % of the total width of the 28Si beam as measured in this
mass spectrum. The more significant contribution to the full width is likely related
59
to the width of the ion beam exiting the ion source and may be an intrinsic width
for the system, although there may be different focusing conditions which could
reduce this width. The hole in the exit cathode of the ion source is approximately
2.54 mm, which is perhaps not coincidentally very similar to the measured 28Si beam
width. Ions with a small spread of angles entering the sector mass analyzer could
contribute to this width as well. If the total width was only due to the mentioned
energy spread, then the resulting mass resolving power would be m
∆m
≈ 350.
An example of a SiH4 mass spectrum where the ion current peaks are not
symmetric and cannot be described by single Gaussian fits is presented in Fig. 2.12.
This spectrum shows four ion current peaks (circles) including 28Si at 28 u and a
combination of 29Si and 28SiH at 29 u. The corresponding magnetic sector mass
analyzer current used for the field sweep is shown on the top axis. Unlike the peaks
in the mass spectrum in Fig. 2.11, the peaks here are not symmetric. Shoulders
appear on the current peaks to the lower mass side. This is likely due to imperfect
tuning and focusing of the ion beam causing a small amount of ion scattering from
a lens element. This type of asymmetry decreases the isolation of adjacent peaks,
which could possibly affect the realized selectivity and enrichment of a sample.
Gaussian fits (line, Eq. (2.14)) to the 28 u and 29 u peaks are superimposed on the
data. An additional Gaussian fit of a shoulder peak (dotted line) of the 29 u main
peak is also shown, which is needed to accurately determine the peak separation.
In this case, the separation from the 29 u main peak to 28 u is ≈ 13 σ, however,
when considering the shoulder peak, that separation drops to ≈ 10 σ. This value is
still very large, but it illustrates the effects of non-ideal beam tuning. Much more
60
Figure 2.12: SiH4 ion beam mass spectrum showing shoulder peaks on the main
ion current peaks to lower mass side. The ion current (circles) is recorded while
sweeping the mass analyzer current, and thus the magnetic field (top axis). The
peak at 28 u is 28Si and the peak at 29 u peak is 28SiH and ≈ 5 % 29Si. Gaussian
fits (line, Eq. (2.14)) to the 28 u and 29 u peaks are superimposed on the data. A
Gaussian fit of the shoulder peak (dotted line) of the 29 u main peak is also shown.
significant shoulder peaks are sometimes observed as well.
Other elements and contaminants in the ion beam generated from the chamber
or the gas being used can be observed in the mass spectra as well. Typically, when
analyzing the mass spectrum of SiH4, several other mass regions of the full spectrum
that contain common contaminants are inspected and recorded. Four of these ion
beam contaminant mass spectra acquired when using SiH4 as the working gas are
shown in Fig. 2.13. In each of these spectra, the ion current (circles and lines) is
recorded while sweeping the mass analyzer current, and thus the magnetic field,
61
Figure 2.13: Ion beam mass spectra of several chemical contaminants in the ion
beam. These spectra were acquired while using SiH4 as the working gas. In each of
these spectra, the ion current (circles and line) is recorded while sweeping the mass
analyzer current, and thus the magnetic field (top axes). (a) H (1 u) and H2 (2 u)
current peaks are typically observed in the ion beam. (b) Most of the contaminants
have masses between 11 u and 20 u, as seen here. These current peaks correspond
to elements including 12C (12 u), 14N (14 u), 16O (16 u), H2O at mass 18 u, and F
(19 u). Doubly charged Si-hydride peaks 28SiH2+ and 28SiH2+2 are also present at
masses 14.5 u and 15 u, respectively. Other unknown peaks appear as well. (c) A
CO2 current peak is also typically observed at 44 u, although usually at much lower
current levels than other peaks in (a) and (b). The species corresponding to the
other three peaks observed within this range are not known. (d) 63Cu (63 u) and
65Cu (65 u) ion current peaks are observed due to ions generated through sputtering
of the Cu cathodes in the ion source. The relative peak heights are similar to the
natural abundance of Cu.
62
which is shown on the top axes. The spectra in panels (a), (b), and (c) were
acquired while using the gas-mode ion source and the spectrum in panel (c) was
acquired while using the solids-mode source. Panel (a) shows H (1 u) and H2 (2 u)
current peaks, which are typically observed in the ion beam. These can originate
from vacuum chamber, which always contains H2, or from the ion source which is
cracking SiH4 and releasing H2. Panel (b) shows a spectrum with a mass range
between 11 u and 20 u containing the most common and abundant contaminants
observed. These current peaks correspond to elements including 12C (12 u), 14N
(14 u), 16O (16 u), H2O at mass 18 u, and F (19 u). Doubly charged Si-hydride
peaks 28SiH2+ and 28SiH2+2 are also present at masses 14.5 u and 15 u, respectively.
If a large 14N or 12C is present, it could indicate the presence of N2 or CO, which
both would appear at approximately 28 u along with 28Si.
Panel (c) of Fig. 2.13 shows a mass range beyond the SiH4 peaks where several
peaks are present. The current peak at 44 u is likely CO2, which is also typically
observed, although with lower currents than the peaks in panels (a) and (b). It is
not obvious what ionic species correspond to the other three peaks in this spectrum.
Panel (d) shows 63Cu (63 u) and 65Cu (65 u) ion current peaks, which are due to
ions generated through sputtering of the Cu cathodes in the ion source. The relative
peak heights are similar to the natural abundance of Cu, which is comprised of
approximately 69.2 % 63Cu and 30.8 % 65Cu. Note the linear scale of panel (d).
As previously mentioned, a low pressure plasma mode and an high pressure
plasma mode of the ion source were used to deposit 28Si samples in this work. These
working modes result in qualitatively and quantitatively different SiH4 ion beam
63
mass spectra. To demonstrate the transition between these two pressure modes,
SiH4 mass spectra were acquired for several different ion source pressures. These
spectra are presented in Fig. 2.14. Note that these spectra are shown on a linear
scale to emphasize the differences. Four current peaks are shown comprised of 28Si at
28 u, 28SiH and 29Si at 29 u, primarily 28SiH2 at 30 u, and primarily
28SiH3 at 31 u.
Initially, a typical working pressure for the low pressure mode of 1.2× 10−4 Pa
(9.0× 10−7 Torr) is used. This pressure results in a standard low pressure mode
spectrum (open circles and line), with the 28 u peak appearing smaller than the 30 u
and 31 u peaks. This is qualitatively similar to the spectrum shown in Fig. 2.11 on
a semi-log scale. When the pressure of the ion source is increased to 6.7× 10−4 Pa
(5.0× 10−6 Torr), the relative peak heights of the ion peaks in the spectrum shift
(open squares and lines). The 28 u peak increases as the 29 u, 30 u, and 31 u
peaks decrease. This is because the cracking efficiency of the plasma mode increases
with higher pressures and additional H atoms are being cracked from the hydrides,
producing more 28Si. This trend continues for the spectrum corresponding to a
pressure of 1.1× 10−3 Pa (8.0× 10−6 Torr), which shows the 28 u peak even larger
(dotted line) but still lower than the 30 u and 31 u peaks. Finally, when the working
pressure of the source is increased to 1.3× 10−3 Pa (9.5× 10−6 Torr), the mass
spectrum (line) appears as a nominal spectrum for the high pressure mode. Here
the 28 u peak is the largest of the four peaks resulting from a further increase in
the cracking efficiency of hydrides into 28Si. In these spectra, the total efficiency for
producing 28Si increased from approximately 8 % to 33 %. The highest efficiency
for producing 28Si observed for a high pressure mode spectrum was approximately
64
Figure 2.14: SiH4 ion beam mass spectra for several working pressures showing the
transition of the relative peak heights from the low pressure mode of the ion source to
the high pressure mode. Note the linear scale. The four current peaks are comprised
of 28Si at 28 u, mostly 28SiH and ≈ 5 % 29Si at 29 u, primarily 28SiH2 at 30 u, and
primarily 28SiH3 at 31 u. For a typical pressure used for the low pressure mode of
1.2× 10−4 Pa, a standard low pressure mode spectrum is observed (open circles and
line) with the 28 u peak appearing smaller than the 30 u and 31 u peaks. As the
pressure of SiH4 in the source is increased, leading to a higher cracking efficiency,
the 29 u, 30 u, and 31 u peaks decrease in height while the 28 u peak increases. For
a pressure of 1.3× 10−3 Pa, typical of the high pressure mode, the spectrum has
fully transitioned to a high pressure mode spectrum (line) with the 28 u peak larger
than the other peaks.
47 %. Note also that the total 28Si ion current increased roughly a factor of 2.7
up to approximately 1.5 µA. The high pressure mode is thus able to produce the
highest 28Si ion currents, which leads to the highest growth rates and thickest films.
While all 28Si samples produced in this work were deposited using SiH4 as the
Si source, the viability of using solid Si cathodes in the ion source that are sputtered
to generate ions was explored. Tests with Si cathodes were conducted using the gas-
65
Figure 2.15: Ion beam mass spectrum of Si ions generated by sputtering natural
abundance Si cathodes while using Ar as the working gas in the gas-mode ion source.
The ion current (circles and line) is recorded while sweeping the mass analyzer
current, and thus the magnetic field (top axis). Ion current peaks corresponding to
28Si (28 u), 29Si (29 u), and 30Si (30 u) are observed. The relative heights of the
three peaks are similar to the natural abundance of Si.
mode ion source instead of the solids-mode source because the gas-mode source was
deemed more reliable at the time. To sputter the Si cathodes, Ar was used as the
working gas. A Si mass spectrum acquired when using Si cathodes is shown in
Fig. 2.15. The ion current (circles and line) is recorded while sweeping the mass
analyzer current, and thus the magnetic field, which is shown on the top axes. Ion
current peaks corresponding to 28Si (28 u), 29Si (29 u), and 30Si (30 u) are seen.
There also appears to be a small peak at 29.5 u, but the origin of this peak is not
known. The relative heights of the three peaks approximately match the natural
abundance of Si. With only approximately 35 nA of 28Si ion beam current, which
66
is at least a factor of ten lower than when using SiH4 as a source, this approach was
abandoned in favor of continuing to use SiH4.
2.2.4.2 Ion Beam Energy, Ei
The average ion energy, Ei, of ions in the beam at the target can be determined
in a measurement referred to as a “roll-off” curve. In this measurement, a positive
bias voltage, Vbias, is applied to the target to repel the incoming ions and is increased
in steps while the ion current, Ii, is recorded. The ion current will drop significantly
at a voltage equal to the average ion energy. An example of a roll-off curve produced
in this work is shown in Fig. 2.16 (a). This roll-off curve is for Ar ions while using
an anode voltage VA = 30 V. The Ar ion current (circles and line) initially has a
value over 1.5 µA and initially decreases slowly as Vbias is increased. The current
then drops much more sharply from approximately 1.25 µA to near 0 µA between
20 V and 30 V. Finally, beyond 30 V, the ion current levels off and again degreases
slowly with further increasing Vbias. Note that the ion current becomes negative for
voltages above 30 V. This is because electrons are attracted to the positively biased
target. Panel (b) is the numerical derivative of the current signal in (a) with respect
to the bias voltage, |dIi/dVbias|. The absolute value of the derivative is taken for
clarity in the figure. The derivative (circles) shows the peak change in ion current
at the average ion energy. |dIi/dVbias| is then fit to a Gaussian (line) of the same
form as Eq. (2.14) but with different variables given by
∣∣∣∣ dIidVbias
∣∣∣∣ = C0 + Bσ√2pi exp
(
−1
2
(
Vbias − Ei/q
σ
)2)
, (2.15)
67
Figure 2.16: Average ion energy measurement for Ar ions using a roll-off curve.
(a) Roll-off curve for an anode voltage, VA = 30 V. As the positive bias voltage
on the sample, Vbias, is increased, positively charged Ar ions are repelled and the
ion current, Ii, decreases (circles and line). At a bias voltage corresponding to the
average ion energy, Ei, Ii decreases significantly. The measured current becomes
negative as electrons are attracted into the target. (b) The absolute value of the
derivative, |dIi/dVbias|, of the data in (a) (circles) shows a peak in the derivative at
the average ion energy. The data is fit to a Gaussian (line, Eq. (2.15)) giving a peak
center Ei ≈ 26 eV.
where C0 is the vertical offset from zero of the derivative, B is again the area of the
Gaussian, and the ion charge state, q = 1, is needed as a conversion of the voltage
value of the peak center to ion energy. From this fit, the average ion energy, i.e. the
energy corresponding to the peak center, is determined to be Ei ≈ 26 eV, which is
close to the 30 V applied to the anode, as expected for ions created at the plasma
potential defined by the anode in the source.
Ei can be mapped as a function of anode voltage, VA, to determine their re-
lationship. The plasma potential is also affected by the cathode voltage, however,
and so the relation between Ei and VA will change depending on the cathode volt-
age, or rather the arc voltage between the anode and cathode, Varc. Experimental
68
measurements of the dependance of Ei on VA and Varc is shown in Fig. 2.17. The
values of Ei are determined in roll-off curve measurements such as the one shown
in Fig. 2.16. These measurements are repeated for several values of VA to get their
relationship while keeping the value of Varc constant. This measurement was re-
peated for different values of Varc including 1.74 kV (circles), 2 kV (squares), 3 kV
(triangles), and 4 kV (diamonds). The resulting values of Ei are always smaller
than VA. Additionally, for a given value of VA, Ei increases with increasing values
of Varc. The data for each Varc value is fit to a line (solid, dashed, and dotted lines)
given by
Ei = αVA + β, (2.16)
with slope α and offset β. These fits give values of α ≈ 1 e, and in general α should
ideally be the ion charge state, q. This means that the offset fit parameter, β,
approximately describes the difference between the applied VA and the resulting Ei
of ions, written as β ≈ Ei − qVA. β is plotted as a function of Varc in the inset of
Fig. 2.17, which shows that the magnitude of β becomes smaller, i.e. less negative,
with increasing Varc.
The energy spread of the ions can also be determined from the fits to the
derivatives of the roll-off curves. The energy spread, ∆E, is determined as half
the peak width at half the height, with the full range of expected energies being
twice this value. For the measurements of Ei in Fig. 2.17, ∆E varies between
approximately 17 eV for the highest values of VA and 4 eV for the lowest ones. The
relative energy spread of the ion is then calculated as ∆E/Ei, which is plotted vs.
69
Figure 2.17: Dependance of the average ion energy, Ei, on the anode voltage, VA,
and arc voltage, Varc. Measured values of Ei are plotted vs. VA for a four values
of Varc: 1.74 kV (circles), 2 kV (squares), 3 kV (triangles), and 4 kV (diamonds).
The data for each arc voltage value are fit to a line (solid, dashed, and dotted lines,
Eq. (2.16)). The inset shows the dependance of the intercept fit parameter of the
linear fits, β, on Varc. With slopes of α ≈ 1 e for the linear fits, β ≈ Ei − qVA, the
magnitude of which becomes smaller, i.e. less negative, with increasing Varc.
Ei in Fig. 2.18. The measured values of ∆E/Ei (circles) increase with decreasing
ion energy from roughly 0.1 at 160 eV to 0.35 at 15 eV. This means that nominally
15 eV ions will have an approximate range of energies between 4.5 eV and 25.5 eV.
The uncertainty on the values of ∆E/Ei come from the standard error of the fit
values. This data is fit using a power low plus a constant given by
∆E
Ei
= (4.5)E−1i + 0.04, (2.17)
where the prefactor and the constant are the fit values, Ei is in units of eV, and
70
Figure 2.18: Dependance of the relative energy spread ∆E/Ei of ions on the average
ion energy, Ei. ∆E/Ei (circles) increases from below 0.1 at about 150 eV to above
0.3 at lower ion energies below about 20 eV. A fit to the data of the form of a power
law plus a constant (line, Eq. (2.17)) is also shown.
∆E/Ei is unitless. This fit equation is derived by first fitting ∆E vs. Ei to a
line. These quantities are expected to be proportional with a constant offset from
an intrinsic energy spread possibly due to other potentials near the source. The
resulting linear equation is then divided by Ei to get a fit for ∆E/Ei. For deposition
of 28Si, ions with average energies of Ei ≈ 35 eV were typically used, and these ions
would have a ∆E of approximately 5.9 eV and range in energy from approximately
29 eV to 41 eV. Additionally, at 28 u, a ∆E of 5.9 eV would represent a physical
width of of the beam at the mass-selecting aperture of approximately 0.23 mm and
a width in mass of approximately 0.08 u, as mentioned previously.
The ion energy is an important parameter in hyperthermal energy ion beam de-
position because ions with these energies will sputter the film as they are deposited.
71
Figure 2.19: Calculated sputter yields for 28Si ions bombarding a crystalline 28Si
surface (circles and line) and an amorphous 28Si surface (triangles and line) are
shown vs. the ion energy. Sputter yields for 12C ions bombarding an amorphous
12C surface (squares and line) are also shown. Sputter yield values are based on
TRIM calculations [69]. The 28Si sputter yield decreases from about 0.6 sputtered
atoms per ion down to about 0.05 when decreasing the ion energy from 550 eV to
50 eV. 12C sputter yields are lower and only reach about 0.25 at an ion energy of
550 eV.
If the sputter rate is too large, then the efficiency of the deposition decreases. Sput-
ter yields for 28Si and 12C deposition have been calculated using TRIM [69] and
are shown in Fig. 2.19. Sputter yields for the average number of atoms sputtered
for every incident ion were calculated for 28Si ions bombarding an amorphous 28Si
film (triangles and line), 28Si ions bombarding a crystalline 28Si film (circles and
line), and 12C ions bombarding an amorphous 12C film (squares and line). These
calculated sputter yields are shown as a function of the ion energy, Ei. As one would
expect, the sputter yields decrease with decreasing ion energy. The 28Si sputter yield
decreases from approximately 0.6 sputtered atoms per ion down to approximately
0.05 when decreasing the ion energy from 550 eV to 50 eV. 12C sputter yields are
72
lower and only reach approximately 0.25 at an ion energy of 550 eV, showing that
the issue of sputtering is more significant for Si deposition. For 28Si ions with Ei ≈
35 eV, the sputter yield is approximately 0.03, which means that for every 100 28Si
ions bombarding the surface, roughly 3 28Si atoms on the surface are sputtered off.
This results in a very high deposition efficiency.
2.2.4.3 Ion Beam Spot Size
The beam spot shape and ion distribution can be mapped using a small detec-
tor at the sample position that is scanned in two dimensions. The detector consists
of a collector plate behind an aperture referred to as the sample aperture. At the
detector, the ion beam passes through the roughly 3 mm diameter aperture and is
collected. The measured ion current through aperture is typically used for assessing
the beam flux to estimate the expected deposition rate when producing a sample.
A 2D current current map of an Ar ion beam spot is shown in Fig. 2.20. The max-
imum current measured in the map is approximately 550 nA. The size of the spot
is approximately 5 mm by 7 mm with a resulting area of approximately 35 mm2.
This spot size is convolved with the size of an aperture at the detector used for this
measurement. The aperture is a roughly 3 mm diameter circular hole, which is large
compared to the beam spot size. A simple estimate of the true beam spot size can
be made by subtracting half of the aperture diameter from each size of the spot.
This give dimensions for the ion beam spot size of roughly 2 mm by 4 mm with a
resulting area of roughly 8 mm2. 2D maps such as these can help to tune the beam
73
Figure 2.20: Ar ion beam spot 2D current map. The ion current is recorded while
moving the detector in two directions, y and z, perpendicular to the axis of the ion
beam to produce a 2D grid of data. The maximum current measured in the map is
about 550 nA. The area of the spot in this measurement is about 35 mm2, although
the spot size is convolved with the 3 mm diameter of the detector.
better and produce a more circular shape for deposition so the spot location on the
sample is better known.
While mass analyzed ion beam currents are typically of the order of 1 µA,
the ion source actually generate as much as 1.5 mA of ion current. This total
current can be measured on the power supply for the transport voltage, because
all ions exiting the ion source must be neutralized in the ion beam chamber except
those ions that pass through the mass-selecting aperture. This loss of ion current
(fluence) from the total output to the analyzed current at the target was investigated
by measuring ion currents on various electrostatic lens elements along the beamline.
These investigation found that for a total of 1.3 mA, 0.56 mA was captured on the
74
extractor, 0.11 mA was captured on the focus, 0.2 mA was captured on the skimmer,
and 0.02 mA was captured on the electron suppressor. A total of 0.28 mA was
transmitted into the sector mass analyzer. As previously mentioned, approximately
10 % of the total current of SiH4 ions is
28Si when using the low pressure mode, so
one would expect about 28 µA of 28Si current to be transmitted, which is much larger
than typical. This may be because the re-focusing properties of the analyzer only
affect ion trajectories in the plane of the bend of the analyzer and so ions are perhaps
spreading out in a vertical direction perpendicular to the bend of the analyzer. Some
modeling using SIMION ion trajectory software was used to model the electrostatic
lenes of the ion beamline to try to find better settings for transmitting more ions,
but a suitable solution was not found.
2.3 UHV Deposition and Analysis Chamber
2.3.1 Apparatus
The deposition and analysis chamber pictured in the right part of Fig. 2.1
consists of a UHV system connected to the ion beamline on one side and the STM
chamber on another. Samples can be loaded into the deposition chamber through
the load lock, which can accommodate up to four samples. The load lock is pumped
by a 67 L/s turbo pump (Agilent Technologies), which results in a base pressure
of approximately 1.3× 10−7 Pa (1.0× 10−9 Torr). Typically, the load lock must
be pumped out for one day before samples can be transferred into the deposition
chamber. At that time, the pressure is about 1.3× 10−6 Pa (1.0× 10−8 Torr). The
75
Figure 2.21: Schematic cross section through the deposition chamber in the plane
of the sample normal to the ion beam. The sample location on the manipulator and
instruments are shown including the ion beam lenses, the RHEED gun and screen,
and the natural abundance Si evaporator. Samples facing the ion beam during
deposition can simultaneously be monitored with RHEED. (from Ref. [59])
deposition chamber also contains a 5-axis sample manipulator, where the samples
sit during deposition. Several other analytic instruments are also present in the
chamber including UHV ion gauges, a residual gas analyzer (RGA), reflection high
energy electron diffraction (RHEED), a natural abundance Si electron beam thermal
evaporator, and an Auger electron spectrometer (Physical Electronics), which was
not used significantly in this work. Figure 2.21 shows a schematic cross section
through the deposition chamber at the sample location normal to the ion beam,
i.e. from the perspective of the ion beam (from Ref. [59]). The sample manipulator
is pictured at the right of the schematic with the sample at the center. Normal
to the sample (on the manipulator) are the deceleration lenses of ion beam optics,
and aligned in the plane of the sample are the RHEED gun and screen and the Si
76
evaporator. Samples facing the ion beam during deposition can simultaneously be
monitored with RHEED. The Si thermal evaporator (Omicron) is a model EFM
3 and is used for natural abundance Si (natSi) deposition. The EFM is capable of
producing deposition rates of typically 0.1 nm/min. The Si source used was either
standard Si wafer pieces in a crucible or a high purity Si rod.
Throughout the experiments discussed here, the deposition chamber has also
contained other instruments including a sputter gun for Ar sputter cleaning, an Al-
deposition source, a STM tip preparation tool, a H-passivation cracking filament,
and a quartz crystal microbalance (QCM). These components were also not signifi-
cantly used in the work reported here. Two external infrared pyrometers, one from
Omega and the other from Process Sensors, were used to measure the temperature
of samples through a window. The Process Sensors pyrometer has a measurement
range between 300 ◦C and 1300 ◦C, while the Omega pyrometer can only measure
temperatures above 600 ◦C. When comparing the reading from the two pyrometers,
a correction is applied to the Omega pyrometer, which can differ from the Process
Sensors readings by ≈ 25 ◦C. A long magnetic transfer rod is used to transport sam-
ples between the manipulator, the load lock, and the STM chamber. Additionally,
wobble sticks are used to manipulate samples in vacuum at sample stations.
2.3.2 Vacuum Analysis
The deposition chamber is pumped using two turbo pumps and two ion pumps.
The turbo pumps are a 300 L/s pump (Edwards Vacuum) located at the decel-
eration lens section, and a 685 L/s pump (Pfeiffer Vacuum) located at the main
77
Figure 2.22: Residual gas mass spectrum collected from the RGA in the deposition
chamber for a base pressure of 8.3× 10−9 Pa obtained after baking out the chamber
to UHV conditions. Partial pressure peaks of typical components of the vacuum are
seen including H2, C, a very small H2O peak, CO and N2, and CO2. F is also seen
in this system.
deposition chamber below the manipulator. Typically, the base pressure of the
deposition chamber for the experiments discussed in this work has been between
6.7× 10−9 Pa (5.0× 10−11 Torr) and about 1.3× 10−8 Pa (1.0× 10−10 Torr). This
UHV environment is achieved by heating the chamber to around 150 ◦C for sev-
eral days, commonly referred to as baking out. This process removes water and
other impurities from inside the chamber, which otherwise get pumped very slowly.
The RGA (SRS Vacuum Instruments) can detect the components of the vacuum
as partial pressures, giving insight into possible contaminants within the chamber.
A residual gas mass spectrum collected from the RGA of the deposition chamber
78
at a base pressure of approximately 8.3× 10−9 Pa (6.2× 10−11 Torr) is shown in
Fig. 2.22. Typical components of the vacuum are seen including H2, C, a very small
H2O peak, CO and N2, and CO2. F is also seen in this system. This indicates that
some component in the chamber is outgassing F, which may act as a contaminant in
the deposition of thin films. The impurities and vacuum pressures do increase from
these base values when the deposition chamber is exposed to the ion beam cham-
ber and ion beam. Another RGA residual gas mass spectrum from the deposition
chamber collected while the ion beam was being operated with SiH4 will be shown
and discussed in Fig. 6.4 in Chapter 6.
2.3.3 Sample Manipulation
As mentioned previously, a 5-axis manipulator is used to position samples
within the deposition chamber for various purposes. The manipulator itself was
produced by VG Scienta, but a sample stage from Omicron was attached to the
end. It has motion in x, y, and z directions as well as rotation about two axes.
One rotation axis is aligned with the insertion axis of the manipulator and is used
to face a sample towards or away from the ion beam as well as other instruments.
The other rotation axis is an azimuthal rotation about the normal to the sample
surface. This rotation is used to position the sample so that particular crystallo-
graphic directions of the substrate can be aligned with the RHEED electron beam.
The sample manipulator also provides heating capabilities to samples, which are
held by sample holders that are loaded onto the manipulator. Sample holders are
made of Mo and clamp the Si chips between clips, one of which is electrically iso-
79
lated from the base. Two methods of sample heating include radiative heating from
a tungsten wire back heater, which is referred to as “RH”, and direct current re-
sistive heating where current is passed through the Si substrate resistively heating
it, which is referred to as “DH”. Heating of the substrates and samples are used
for several purposes. Deposition at elevated temperatures is critical for facilitating
epitaxial deposition of the films. Additionally, substrates can be prepared in situ
to have clean surfaces by flash annealing them using the DH method for heating.
Flash annealing involves rapidly increasing the temperature of the substrate typi-
cally from approximately 600 ◦C to as high as 1250 ◦C in a few seconds. This high
temperature anneal step removes oxide from the substrate surface and reconstructs
the surface to form the well known Si(100) (2×1) dimer rows. The oxide typically
present on substrates introduced into the vacuum chamber in this work was either a
native SiO2 or a deliberately grown oxide. After remaining at a high temperature for
typically 10 s, the substrate is cooled quickly back to 600 ◦C. This flash annealing
method produces flat, nominally clean surfaces on the Si(100) substrates.
A wiring diagram on a photograph of the sample stage on the manipulator
is shown in Fig. 2.23 indicating the current path for the two heating methods. A
Si(100) substrate with dimensions of approximately 10 mm wide by 4 mm long is
seen mounted in a sample holder that is being held on the manipulator. The sample
is mounted in the holder ex situ using two Mo foil clips that clamp either end of
the substrate chip, as mentioned. The Si(100) substrate is glowing due to it being
at a temperature of roughly 1000 ◦C. This demonstrates the DH heating method
where current is being driven through the sample to resistively heat it. The DH
80
Figure 2.23: Photograph and wiring diagram for heating a Si(100) substrate on
the sample manipulator. The Si(100) substrate shown is heated to about 1000 ◦C
using the DH power supply to drive current through the substrate. The DH supply
connects to “finger” contact on the left of the sample holder and the grounded base
of the sample stage. The current and voltage of the DH supply used for sample
flashing to 1200 ◦C are IDH ≈ 9 A, and VDH ≈ 5 V, respectively. The RH power
supply connects to terminals of a tungsten wire heater beneath the sample, one of
which is grounded to the sample stage base. The RH supply uses a current and
voltage of IRH ≈ 8 A, and VRH ≈ 12 V, respectively, for sample degassing at about
600 C.
power supply is represented on the left side of the diagram with the positive lead
being connected to the “finger” contact on the left side of the sample stage. The
finger is isolated from the rest of the sample stage and contacts the left sample
holder mounting clip, which is also isolated from the rest of the sample holder.
Therefore, when a voltage, VDH is applied to the finger, a current, IDH flows through
the substrate. The other (right) side of the substrate is grounded through the sample
holder to the sample stage so that the current flows to the negative terminal pictured
on the right. The current and voltage of the DH supply used for sample flashing to
1200 ◦C are IDH ≈ 9 A, and VDH ≈ 5 V, respectively.
The RH tungsten back heater shares the negative terminal on the right side of
the sample stage in Fig. 2.23. The positive lead of the RH power supply represented
81
on the right connects to the positive terminal above the negative terminal at the
right. The positive terminal of the RH heater is isolated from the sample stage,
unlike the negative one. The tungsten wire is connected to these two terminals and
runs beneath the sample location in a snaking pattern. The RH supply uses a current
and voltage of IRH ≈ 8 A, and VRH ≈ 12 V, respectively, for degassing samples at
approximately 600 C. The RH heater is also typically required to pre-heat the Si
substrate before the DH method can be used. This is because the resistance of the
substrate is often too large at room temperature to begin flowing current. As the
sample is heated by the RH back heater, the resistance of the substrate drops and
the Si can then be heated using DH. Photographs of the manipulator and sample
stage are shown in Fig. B.12 in Appendix B.
When depositing a 28Si film while simultaneously heating the substrate, a
different wiring setup is used in order to be able to monitor the ion beam current on
the sample. In this setup, the RH leads are disconnected and the DH power supply
is isolated from ground with its leads remaining connected to the sample stage as
shown in Fig. 2.23. The input of a picoammeter, which is itself grounded, is then
connected to the negative terminal of the sample stage. This setup allows for the
ion beam current to flow from the sample through the picoammeter to be measured.
There is some leakage current of typically about -10 µA registered using this setup,
which produces an offset in the ion current measurement.
In order to achieve the sample heating procedures discussed in this section
including rapidly changing the sample temperature as well as using precise depo-
sition temperatures, accurate temperature measurements are important. As was
82
mentioned previously, an infrared pyrometer is used in this work to measure the
temperature of samples. The pyrometers are external to the deposition chamber
and view the sample surface through a window. The temperature reading of the
pyrometer depends on the value of the emissivity, e, to which it is referenced. There-
fore, the emissivity of the Si substrates, which is a temperature dependant quantity,
needs to be determined for the temperature range being measured. An indication
of the emissivity values for Si can be taken from literature, however, differences
in experimental setups such as the type of window that the pyrometer views the
sample through require calibration to the system being used. The pyrometer from
Process Sensors, which was used for the most samples, was calibrated for the work
reported here, while the Omega pyrometer was not. Experimental data and calcu-
lations show that for Si at temperatures above approximately 800 ◦C, e ≈ 0.68 and
below 100 ◦C, e can be as low as 0.1 [70,71].
To calibrate the pyrometer and determine e, two Si eutectics were used. These
Si compounds have melting temperatures that are much lower than the Si melting
temperature of approximately 1414 ◦C. By forming a Si eutectic in a small region
on a Si substrate, the substrate temperature can be brought to the melting point of
the eutectic while the Si substrate remains solid. To calibrate the pyrometer, it is
used to monitor the temperature of the substrate while the eutectic is heated until
it melts. With the temperature adjusted so that the eutectic is held at its melting
point, the emissivity of the pyrometer is adjusted until the temperature that it reads
matches the known melting temperature of the eutectic. This gives a calibration
for the pyrometer at a specific temperature for the experimental setup being used.
83
The two Si eutectic systems that were used in this work to calibrate the Process
Sensors pyrometer are Au-Si, which has a melting temperature of approximately
363 ◦C and Al-Si, which has a melting temperature of 577 ◦C. The Au-Si eutectic
phase diagram is shown in Fig. 2.24 taken from Ref. [72]. The phase boundary for
the liquid phase is shown as a function of Si atomic concentration as a percentage.
The eutectic forms at a Si concentration of approximately 18.6 % where the melting
temperature is at a minimum of approximately 363 ◦C.
The Au-Si eutectic was formed here by pressing a small Au wire onto the
edge of a Si substrate. The substrate was then heated in the vacuum chamber
using the DH method. The substrate temperature was increased until part of the
Au wire was seen to melt. This signified a Au-Si eutectic being formed due to a
small amount of Si diffusing into the Au where it contacting the substrate. Upon
repeated cycling of the temperature above and below the the point that the (mostly)
Au wire melts, the temperature at the melting point is observed to decrease as the
Si concentration equilibrates to the optimal eutectic Si concentration of 18.6 %.
When this occurs, the temperature reading at the melting and freezing of the wire
remains constant. This temperature is then known to be approximately the 363 ◦C
melting temperature of Au-Si and the pyrometer emissivity was adjusted to match
this temperature, yielding a value of e = 0.25. A similar procedure was followed
for the Al-Si eutectic although instead of a Al wire, an approximately 2 mm by
2 mm square of Al approximately 500 nm thick was deposited on the Si substrate.
This method was found to be less reliable than using a larger wire. For the Al-
Si calibration at 577 ◦C, an emissivity value of e = 0.42 was determined. These
84
Figure 2.24: Au-Si eutectic phase diagram showing the phase boundaries as a func-
tion of Si concentration. The melting temperature of a Au-Si system is lowest for
a Si atomic concentration of 18.6 %. This corresponds to the eutectic melting tem-
perature of 363 ◦C. (from Ref. [72])
measured values were then combined with the literate value for temperatures above
approximately 800 ◦C of e ≈ 0.68 to get an approximate function of e vs. T used for
experiments here. In this function, e = 0.25 is used for Si substrates temperatures
between room temperature and 470 ◦C, e = 0.42 is used between temperatures of
470 ◦C and 625 ◦C, and e = 0.68 is used for temperatures above 625 ◦C. These
crossover points were chosen based on both the measured e values and literature
values of e for a range of temperatures [70, 71].
The calibrated pyrometer was then used to determine a calibration curve for
the current applied for DH sample heating and flashing. This control curve of
substrate temperature as a function of the substrate current using the DH, IDH, is
shown in Fig. 2.25. To generate this control curve, the temperature of the substrate
was recorded for different values of IDH using the three previously discussed values
of e in their appropriate temperature ranges. These data are seen appearing in three
distinct curves (circles) marked with each value of e. Note that the current is shown
85
Figure 2.25: Experimental control curve for the DH power supply used for sam-
ple heating. Temperature measurements (circles) are shown vs. DH current, IDH,
applied through the substrate. Note that the current axis is shown as a log scale.
Temperatures are measured using an infrared pyrometer for three emissivity val-
ues, 0.25, 0.42, and 0.68. These three values correspond to measurements in three
temperature ranges with crossover points between the ranges at 470 ◦C and 625 ◦C.
To transition smoothly between the three regions, the data is fit to a sum of three
exponentials (line, Eq. (2.18)), which is then used for sample heating. This data is
for a Si(100), phosphorous-doped substrate 300 µm thick with a resistivity between
7 Ω · cm and 20 Ω · cm.
on a log axis. This data is for a Si(100), phosphorous-doped substrate 300 µm
thick with a resistivity between 7 Ω · cm and 20 Ω · cm acquired from Virginia
Semiconductor. Substrates with different specifications, especially the thickness,
will have different control curves. In order to smoothly transition between the three
ranges, the data is fit to a sum of three exponentials (line) given by
T = 1360− (290) exp
(−IDH
0.026
)
− (240) exp
(−IDH
0.024
)
− (810) exp
(−IDH
4.7
)
, (2.18)
which shows the fit parameters. IDH is in units of A and T is in units of
◦C.
86
This fit allows for the substrate temperature to be controlled continuously using
the applied DH current for both sample degassing and flash annealing. Including
the uncertainties from the pyrometer calibration, the temperature readings of the
substrate are estimated to have a 5 % relative uncertainty due to fluctuations in the
current used for sample heating as well as temperature gradients across the sample.
2.3.4 In situ Sample Analysis
After flash annealing a Si(100) substrate to clean and prepare it in situ, the
surface can be inspected by RHEED. This and other capabilities including STM
are used to produce 28Si thin films and analyze them in situ. The RHEED system
(STAIB Instruments) used here typically uses an electron energy of approximately
15 keV, and an electron gun filament current of approximately 1.55 mA. As men-
tioned previously, Si(100) substrates are positioned using the manipulator to be
able to do RHEED during deposition. The sample can be rotated about its azimuth
normal to the surface to align the RHEED electron beam along different crystal-
lographic directions, but typically the beam was aligned with the 〈110〉 direction.
The other axis of rotation of the manipulator can be used to adjust the angle of
incidence of the RHEED electron beam, and typically, an angle of approximately
2◦ was used for capturing diffraction patterns although angles as high as 5◦ were
sometimes used. The RHEED diffraction patterns are captured using a camera in
a dark enclosure that images the patterns produced by diffracted electrons on a
phosphorous screen.
An example RHEED diffraction pattern of a flashed Si(100) surface showing
87
diffraction corresponding to the (2×1) surface reconstruction is shown in Fig. 2.26,
which was acquired with the surface at approximately 600 ◦C. The observed semi-
circle of spots is the zeroth Laue zone, which is typically imaged. The outer diffrac-
tion spots are due to the bulk Si diffraction and the inner two spots are due to the
reconstruction. The reflected specular spot appears in the middle. As this pattern
is a representation of reciprocal space, the distance from the central (00) spot to the
bulk Si (11) spots are inversely related to the atomic spacing of atoms on the Si(100)
surface. Knowing the physical spacing of these spots on the screen and the working
distance to the sample, the Si lattice constant can be extracted. The spacing of
Si(100) (2×1) dimer rows on the surface is double the atomic spacing, and so the
(2×1) spots appear at a distance from the central spot that is half of that to the
bulk spots.
These diffraction spots and a lack of a diffuse background indicate a crystalline
surface that is fairly clean, i.e. free of oxide, although RHEED is not sensitive to
trace impurities. The observed diffraction patterns in RHEED are the result of
the combined diffraction from a macroscopic area on the sample (≈ 0.25 mm2) due
to an elongated electron beam spot on the surface of several mm resulting from
the low incidence angle. Typically, diffraction from a flat surface produces long
vertical streaks in the pattern, which can be seen in Fig. 2.26. Stronger intensity
spots in the middle of the streaks are the result of the finite and relatively large
size of atomic terraces (e.g. 100 nm) on the Si(100) surface with the streak length
being inversely proportional to the terrace size. Kikuchi lines are also seen running
diagonally through the pattern, which are due to surface resonances.
88
Figure 2.26: RHEED diffraction pattern of a clean, flash annealed Si(100) substrate
showing bulk Si diffraction spots and (2×1) reconstruction spots. A lack of a bright
diffuse background and the presence of the (2×1) spots indicates that the native
oxide has been removed. This pattern was acquired with the substrate at about
600 ◦C and the electron beam in the 〈110〉 direction.
RHEED is used during deposition of 28Si films to monitor the growth structure,
i.e. smooth crystalline vs. rough crystalline vs. amorphous, as a function of film
thickness. Additionally, by inspecting different areas of the substrate, the location
of the deposition spot, which is typically smaller than the substrate chip, can be
roughly located. This is because the diffraction pattern of a deposited film is never
exactly the same as the diffraction pattern of the bare Si(100) substrate. Thus the
spot can be located based on the changing pattern and by comparing those patterns
to the patterns of the substrate before deposition.
After depositing 28Si films, samples are typically moved into the STM chamber
for inspection of the film surface. As previously mentioned, the STM resides in a
UHV chamber separated from the deposition chamber by a gate valve. A magnetic
transfer arm was used to transfer samples between the two. The STM chamber
has a typical base pressure of approximately 6.7× 10−9 Pa (5.0× 10−11 Torr) and
89
was pumped by a 500 L/s ion pump (Varian) and a Ti sublimation pump (TSP). A
sample holding area in this chamber can store up to 12 samples at a time. The STM
used here is an Omicron variable temperature model. STM images were typically
acquired using a tip bias of -2 V for Si substrates and -1.8 V for 28Si samples.
Tunneling currents used were typically 150 pA for Si substrates and 100 pA for 28Si
samples. STM was used to assess the cleanliness and surface quality of Si(100)
substrates prepared in situ by flashing before deposition. Scanning was typical
done in an area on the surface of 100 nm by 100 nm where contaminants, residual
surface oxide, adsorbates, or other surface defects are clearly visible. The quality
of the visible Si(100) (2×1) dimer rows can also give an indication of the presence
of contaminants. After depositing 28Si films, STM was also used to image the film
surface morphology to gain qualitative information about island formation or other
features as well as determine some qualitative parameters such as the total local
height variation of a film surface. STM images presented in this work were acquired
in collaboration with Hyun soo Kim.
90
Chapter 3
Initial Experiments Enriching 22Ne and
12C
3.1 Context and Experimental Setup
Before depositing 28Si films, proof of principle experiments were conducted
involving the enrichment of 22Ne, which was implanted into Si substrates, and 12C,
which was deposited as thin films onto Si. These experiments helped establish the
experimental practices needed for later adapting the system to 28Si deposition. Ad-
ditionally, they showed that a high enough level of enrichment was possible to war-
rant investment in 28Si enrichment and deposition, as can be seen on the enrichment
progression timeline in Fig. 1.12 in Chapter 1.
The experiments discussed in this chapter were conducted with an experimen-
tal setup where the ion beamline was disconnected from the deposition chamber. A
schematic of the ion beam and lens chambers in this setup is shown in the upper right
section of Fig. 4.1 and is discussed further in Chapter 4. The mass-selecting aper-
ture used in these experiments was a circular hole that was approximately 5 mm
91
in diameter and 16 mm thick. In this configuration, samples were located after
the deceleration lenses, and they were mounted on the end of an electrical vacuum
feedthrough which acted as a sample stage. A sketch of the sample stage feedthrough
and sample location is shown in the blowup of the schematic of the “LC–2” setup
in Fig. 4.1. Samples were mounted using a strip of conductive carbon tape on the
back side of the chip. The purpose of the feedthrough was to isolate the sample elec-
trically from the chamber so that the ion beam current could be monitored during
beam tuning and deposition. Additional electrical feedthroughs on the sample stage
were used to mount a masking element above the sample location consisting of a
metal shim for collecting current. This was positioned between the sample and the
path of the ion beam and had a small circular aperture directly above the sample.
This mask and aperture was fixed in place for the duration of a deposition. The
fixed sample aperture was approximately 3 mm in diameter and provided a mech-
anism to monitor the focusing of the ion beam on the sample by maximizing the
ion current detected on the sample while minimizing the current detected on the
sample mask.
Additionally, the mask and sample aperture allowed for precise location of the
ion beam spot on the sample substrate. The feedthrough (and sample) were posi-
tioned on axis with the ion beamline optics, and the sample aperture allowed for the
ion beam to be tuned and steered onto the exact sample location under the aperture.
A photograph of a sample mounted on the vacuum feedthrough on this intermediate
sample stage with the sample mask is shown in Fig. B.11 in Appendix B. A lack
of sample motion in this setup means that the ion beam was constrained to be on
92
axis, which may not have been the optimal tuning position. There was no method
for sample heating in this setup.
One difference between the 12C and 22Ne experiments is that, as a gas at room
temperature, Ne needs to be implanted into a substrate as a means of capturing it so
that the enrichment can be measured later. Ne implantation required much higher
ion kinetic energies than the carbon deposition. To accomplish this, a bias voltage
as high as approximately -4 kV was applied to the substrates during deposition to
attract the positively charged 22Ne+ ions into the substrate. The final energy of
the 22Ne ions would then be the energy due to the bias potential plus the starting
energy of the ions, which was approximately 500 eV for Ne ions, for a total energy
of roughly 4.5 keV. As will be discussed below, this high energy was needed to
achieve an implantation depth of at least 10 nm for a reliable SIMS measurement
of the enrichment. A sample bias of -4 kV was the highest bias achievable for the
experimental setup used in Ne implantation experiments.
Section 3.2 of this chapter will discuss the enrichment and implantation of
22Ne. Section 3.3 will discuss the enrichment and deposition of 12C. Finally, Section
3.4 will summarize those results.
93
3.2 22Ne Implantation and Characterization:
Proof of Principle
3.2.1 Sample Preparation
For these 22Ne implantation tests, there was very little ex situ sample prepara-
tion. Substrates consisted of “lightly doped” natural abundance commercial Si(100)
wafers that were cleaved by hand into approximately 1.5 cm by 1.5 cm chips. The
chips were handled with clean teflon tweezers and mounted on the end of the
feedthrough using a strip of carbon tape on the back of the wafer, and they were
loaded into the vacuum chamber with a native oxide. No further sample prepara-
tion occurred in situ. The role of these substrates is to simply be a “catcher foil” to
collect the 22Ne ions. Typically, after being loaded, samples sat several day in the
vacuum chamber before deposition.
3.2.2 22Ne Implantation
For Ne implantation experiments, the solids-mode Penning ion source was
used (see discussion and Fig. 2.9 in Chapter 2). While the gas-mode ion source
should be more efficient for generating Ne ions, it was not yet purchased at the
time of this experiment. Natural abundance Ne gas was used to generate mostly
singly charged Ne ions with a working pressure in the ion source of ≈ 1.0× 10−2 Pa
(7.8× 10−5 Torr). Sweeping the magnetic field of the mass analyzer and monitoring
the ion current onto the vacuum feedthrough and sample through the fixed sample
94
Figure 3.1: A mass spectrum obtained using Ne as a working gas shows ion current
peaks (circles) corresponding to atomic species of 20Ne (20 u), 21Ne (21 u), and 22Ne
(22 u). The ion current is recorded while sweeping the mass analyzer current, and
thus the magnetic field (top axis). Gaussian fits (line, Eq. (2.14)) to the Ne peaks
are superimposed on the data. The centers of the 20 u and 22 u fits are separated
by ≈ 14 σ.
aperture at the end of the ion beamline, a mass spectrum of the individual molecular
and atomic species is generated. Figure 3.1 shows a portion of the mass spectrum
obtained using Ne as a source gas with ion current peaks (circles) at masses of 20 u,
21 u, and 22 u corresponding to atomic species of 20Ne, 21Ne, and 22Ne, respectively.
The top axis shows the applied current used to sweep the magnetic field of the
analyzer. The three stable Ne isotopes have a natural abundance of approximately
90.5 % 20Ne, 0.3 % 21Ne, and 9.2 % 22Ne, and the relative peak heights of the
three peaks in Fig. 3.1 fairly accurately reflect this abundance. A sum of Gaussian
fits (line, Eq. (2.14)) to the Ne peaks are superimposed on the data and show
95
fairly good agreement with some divergence to the low mass side in the tails of the
peaks, indicating possible ballistic scattering of a small portion of the ion beam.
As discussed in Chapter 2, the spacial distribution of ions in the beam is expected
to be roughly Gaussian for this system. The minor isotope, 22Ne, was chosen for
enrichment in these experiments because enriching in the minor isotope and rejecting
the major isotope is a stronger proof of principle demonstration of the enriching
power of the ion beam system than enriching the more abundant major isotope.
The Gaussian fits to the mass spectrum peaks can be used to indicate the separation
of the 22 u peak from the 20 u peak. From the fits, the 20 u peak has a standard
deviation of σ ≈ 0.14, resulting in a separation to the major isotope 20Ne at 20 u
of approximately 14 σ. The mass resolving power at the 22Ne peak of the ion beam
in this configuration is derived from the mass spectrum to be m
∆m
≈ 26 (measured
at 10 % of the peak height). This fairly low mass resolution and the presence of
an ion current of roughly 1 nA between the peaks indicates that the beam may not
be optimally focused at the mass-selecting aperture. Tuning the mass analyzer to
a mass of 22 u (unified atomic mass units) results in the 22Ne ion beam passing
through the mass-selecting aperture, through the focusing lenses, and to the sample
for implantation.
The 22Ne sample discussed below in the analysis of the sample enrichment
here was implanted with the substrate at room temperature with an unknown back-
ground pressure. This pressure was possibly around 6.5× 10−5 Pa (5.0× 10−7 Torr)
based on later experiments. This partial pressure during growth was mostly Ne leak-
age from the ion source as the base pressure of the deceleration lens chamber before
96
Figure 3.2: 22Ne stopping range (implantation depth) vs. ion energy (circles and
line) based on TRIM calculations [69]. In the calculation, the ions hit a crystalline
Si target with a 2 nm SiO2 surface layer.
implanting this sample was ≈ 3.1× 10−6 Pa (2.8 × 10−8 Torr). The anode voltage
used for generating Ne ion beams in these experiments was approximately 500 V
resulting in an initial energy of the ions in the ion source of 500 eV. As mentioned
previously, the sample was biased with a negative high voltage to achieve implanta-
tion. The starting energy of the ions is added to the energy gained from the sample
bias voltage of approximately -4 kV. The total average ion energy at the target was
then approximately 4.5 keV, as discussed previously. This high energy was required
to achieve an implant depth of at least 10 nm, which was needed for the SIMS mea-
surements of the enrichment to separate the Ne signals of the implant from that of
surface contaminants.
Calculations of the stopping range of energetic ions based on TRIM [69] predict
that 4.5 keV 22Ne ions will implant into Si with a native oxide at a peak implantation
97
Figure 3.3: Optical micrograph of an implanted 22Ne sample produced at room
temperature at LC–2. The Si(100) substrate is seen with scribe marks used to align
the chip to the expected ion beam spot location. The 22Ne implanted area is not
clear but is likely to the right of the crossing of the scribe marks in an expected
area of about 2 mm2. Other marks visible on the chip are due to sputtering from
the SIMS measurement.
depth of approximately 12.5 nm. The stopping range, i.e. the peak implantation
depth, of 22Ne ions hitting a Si target (circles and line) is shown for a range of ion
energies in Fig. 3.2. The Si target used in these calculations is crystalline and has
a 2 nm SiO2 surface layer representing the native oxide on the substrates used in
these experiments. The implantation depth varies between approximately 4 nm for
an ion energy of 1 keV to approximately 14.5 nm for an ion energy of 5.5 keV.
The ion current of 22Ne achieved for implantation was approximately 140 nA
over an implantation area, estimated from other samples, of approximately 2 mm2.
The associated ion dose for these parameters is approximately 7.7× 1016 cm2, re-
sulting in an atomic concentration of 22Ne in the Si substrate of approximately
4.9× 1022 cm−3, based on TRIM calculations. This means the implantation volume
in the Si substrate was roughly 50 % 22Ne.
An optical micrograph of the 22Ne sample analyzed for enrichment here is
98
shown in Fig. 3.3. The Si(100) substrate is seen in the micrograph with scribe
marks used to align the chip to the expected ion beam spot location just under the
fixed sample aperture. It is not obvious where the 22Ne-implanted area is, although
it is believed to be just to the right of the crossing of the scribe marks. Other
marks visible on the chip are due to sputtering from the SIMS measurement. A
total of three 22Ne samples were produced in this work, and two were analyzed for
enrichment, the second of which is discussed below.
3.2.3 Enrichment Measurements via SIMS
SIMS was used to determine the enrichment of two 22Ne-implanted samples,
both of which are represented on the enrichment progression timeline in Fig. 1.12.
For the SIMS measurement of these samples, the primary ion sputter beam was
composed of Cs+ ions with a current of 40 nA and an energy of 10 keV. The
sputter beam was raster-scanned over a nominal 280 µm by 430 µm area on the
sample, although the actual area sputtered was considerably larger because of the
finite size of the ion beam. Measurements of the counts of the major isotope, 20Ne,
and the minor isotope chosen for enrichment, 22Ne, were made after each sputter
cycle, producing a “depth” profile of the isotope ratios (assuming a constant sputter
rate). Isotope ratios are then used to determine the isotope fractions, which indicate
the enrichment level. As mentioned in Chapter 1, isotope fractions of a particular
isotope are defined as the counts of that isotope divided by the total counts of the
measurement. The isotope fractions of Ne can be written as zNe/Netot., where z is
the mass number denoted as 20 for 20Ne counts and 22 for 22Ne counts and Netot. is
99
Figure 3.4: SIMS “depth” profile of a 22Ne-implanted sample produced at room
temperature at LC–2. 22Ne was implanted into Si(100) with an energy of about
4.5 kV. Isotope ratios of 22Ne/20Ne (open diamonds and line) are shown vs. the
sputter time. The second point at 63 s is the largest ratio value (183), representing
the maximum enrichment in 22Ne. The inset shows the raw count rates vs. sputter
time for 22Ne (triangles and line) and 20Ne (diamonds and line).
the sum of 20Ne and 22Ne counts.
The first 22Ne sample produced, which was also the first enriched sample of
any kind produced in this work, had a lower level of enrichment than the second one.
This first sample had a 22Ne isotope fraction of 84.4(10) %, seen on the enrichment
progression timeline as the highest residual isotope fraction of those samples. The
second 22Ne sample was also analyzed by SIMS, and the resulting “depth” profile
is shown in Fig. 3.4. The isotope ratios of 22Ne/20Ne (open diamonds and line)
are shown vs. the sputter depth into the sample. At a sputter time of 63 s, which
likely corresponds to the peak implantation depth of the 22Ne (expected to be ap-
100
proximately 12.5 nm), the isotope ratio has a maximum value of 183. This value
represents a 22Ne isotope fraction (assumes no 21Ne is present) of 99.455(36) % and
a residual 20Ne isotope fraction of 0.545(36) %, as seen on the enrichment progres-
sion timeline. The inset of Fig. 3.4 shows the raw count rates vs. sputter time
of 22Ne (triangles and line) and 20Ne (diamonds and line) for reference. Another
useful parameter that describes the measured enrichment of a sample is the isotope
reduction factor, which gives the amount by which an excluded isotope is reduced
from the natural abundance. This reduction factor is determined by dividing the
natural abundance of an isotope of an element by the measured isotope fraction
described previously. The natural abundance of an isotope is represented by az,
where z again is the mass number of the isotope. For the reduction factor of the
excluded isotope 20Ne, az is a20, i.e. the natural abundance of
20Ne. The reduction
factor for 20Ne is thus written as a20/(
20Ne/Netot.). The measured isotope fractions
discussed here give an isotope reduction factor for 20Ne of 166(1), i.e. the 20Ne in
the sample is approximately 166 times lower than in natural abundance Ne. These
measurements indicate a realized mass selectivity of 1785:1 for the ion beam system
in this configuration. This value gives a sense of the performance of the system
independent of the natural abundance of Ne.
101
3.3 12C Deposition and Characterization: First
Enriched Thin Films
3.3.1 Context
Solid state QI using spin states of atoms or quantum dots as qubits is lim-
ited by the need for host materials that are as minimally-interacting as possible,
e.g., 28Si. Another potential host material for QI devices is diamond. Natural
abundance diamond has isotopes possessing a non-zero nuclear spin, and so isotopic
enrichment can increase the coherence time of qubits in diamond by reducing the
density of randomly fluctuating spins in their local environment. Qubits in 12C
enriched diamond have received a lot of attention in this context [73]. Isotopic en-
richment eliminates the 13C nuclear spins I = 1/2 present in natural abundance C
(natural stable isotope abundances: 98.9 % 12C, 1.1 % 13C). Nitrogen-vacancy (NV)
centers in enriched, highly pure diamond can have electron spin T2 coherence times
of milliseconds [74,75]. Additionally, qubits with long nuclear spin coherence times
approaching a second have been demonstrated for isolated 13C atoms coupled to NV
centers in 12C diamond [73]. This chapter describes the enrichment and deposition
of 12C films as a potential host for qubits, but it also serves as a demonstration for
the production of 28Si films. Significant portions of the data and analysis presented
in this section was previously published in Ref. [76].
The 12C reported here has an enrichment in the solid state comparable to
that of commercially available methane (up to 99.999 % 12CH4) used as a source
102
gas to grow enriched diamond by CVD [73, 77, 78], where enrichment and purity
may be diminished in processing. CVD diamonds grown from enriched 12CH4 can
suffer from incorporation of excess N (14N has nuclear spin I = 1) that can also
limit coherence times. Several groups growing enriched 12C diamond by CVD have
made efforts to reduce the atomic concentration of paramagnetic impurities like N
below 1× 10−9 cm−3 [74, 75, 78, 79]. The mass selecting method used here isolates
the 12C to avoid contaminants such as N and O. N may still be physisorbed from the
background vacuum but vacuum improvements and heated substrates could mitigate
this liability. As previously mentioned, some work has been done depositing enriched
C materials including 12C, 28Si12C and 12C14N [56, 57, 80]. The numerous research
efforts utilizing enriched diamond and other C allotropes [74,81,82] can also benefit
from the availability of 12C films.
3.3.2 Sample Preparation
Similar to the Ne samples, for these 12C deposition tests, there was very little ex
situ sample preparation. Substrates consisted of “lightly doped” natural abundance
commercial Si(100) wafers that were cleaved by hand into approximately 1.5 cm by
1.5 cm chips. Clean teflon tweezers were used to mount these substrates onto the
end of the same electrical vacuum feedthrough used in Ne implantation experiments
using a strip of carbon tape on the back of the wafer. Substrates were loaded into
the vacuum chamber with a native oxide. No further sample preparation occurred
in situ before the deposition of 12C films. Similar to the substrates used in the Ne
experiments, these substrates serve as a simple “catcher foil” to collect the 12C ions
103
as a film in these experiments. Typically, after being loaded, substrates sat several
days in the vacuum chamber before deposition occurred.
3.3.3 Deposition of 12C
In this experiment, natural abundance CO2 is used as the carbon source to gen-
erate 12C ions and grow 12C films. CH4 gas was also tested but was found to produce
a lower ion beam flux than CO2 (see Fig. 2.10 in Chapter 2). The solids-mode Pen-
ning ion source was used in these experiments (see Fig. 2.9 and discussion in Chapter
2). The working pressure of the ion source used in these experiments was between
approximately 1.1× 10−3 Pa (8.0× 10−6 Torr) and 1.3× 10−3 Pa (1.0× 10−5 Torr).
CO2 molecules are cracked and ionized by the plasma in the ion source to gener-
ate the ion beam. By scanning the magnetic sector mass analyzer and monitoring
the ion current through the mass-selecting aperture onto the sample stage, a mass
spectrum of the individual molecular and atomic species is generated. Figure 3.5
shows a portion of the mass spectrum acquired using CO2 as a source gas that is
representative of the ion beam conditions used to deposit samples discussed in this
section. Ion current peaks (circles) are visible at masses of 12 u, 13 u, 14 u, and
16 u corresponding to atomic species of 12C, 13C, 14N, and 16O, respectively. The
14N peak is due to a small amount of residual nitrogen background pressure present
in the ion source chamber. The top axis shows the applied current used to sweep the
magnetic field of the analyzer. The relative peak heights of the 12C and 13C peaks
fairly accurately reflect the natural abundance of stable C isotopes. The 12 u and
13 u mass peaks are fitted to Gaussian functions (line, Eq. (2.14)), which are super-
104
Figure 3.5: A mass spectrum obtained using CO2 as a working gas shows current
peaks (circles) corresponding to atomic species of 12C (12 u), 13C (13 u), 14N (14 u),
and 16O (16 u). The ion current is recorded while sweeping the mass analyzer
current, and thus the magnetic field (top axis). Gaussian fits (line, Eq. (2.14)) to
the 12C and 13C peaks with 95 % confidence bands are superimposed on the data.
The centers of the fits are separated by ≈ 4 σ.
imposed on the data along with 95 % confidence bands (shaded area). The peaks
are fairly symmetric and the data can be seen to match the form of a Gaussian,
indicating minimal scattering of ions by lens elements.
The mass resolving power of the ion beam at the 12C peak is derived from
the fits to the mass spectrum to be m
∆m
≈ 13.5 (measured at 10 % of the peak
height). This mass resolution is fairly low and the presence of a measurable ion
current of roughly 0.4 nA between the 12 u and 13 u peaks indicates that the beam
may not be optimally focused at the mass-selecting aperture. The separation of the
105
peaks can also be estimated from the Gaussian fits, which show that the center of
the 13C peak is ≈ 4 σ away from the center of the 12C peak. The overlap of the
13 u peak on the 12 u peak due to this high level of separation can be determined
using the parameters G12(m) and G13(m). These are the values (calculated at mass
m in units of u) of the Gaussian fits to the current peaks at 12 u and 13 u. The
geometric selectivity of the system for 12C relative to 13C for the conditions used
when producing a 12C beam is then determined from the ratio of the Gaussian fit
functions at 12 u, given by
G12(12)
G12(12) +G13(12)
= 99.9999(3) % 12C, (3.1)
where m = 12 for the above parameters signifying that the values of the Gaussian
fits are calculated at a mass of 12 u. The uncertainty is found in a similar manner
using the 95 % confidence bands of the fits. By further tuning the ion beam and
adjusting the parameters of the electrostatic lenses, better isolation of the masses
can be achieved in this setup at the expense of beam flux. This analysis sets an
upper bound for a sample enrichment grown under these conditions. By tuning the
magnetic field to the center of the peak at 12 u, 12C ions are transmitted with a high
degree of selectivity to the target substrate for deposition. Thus, the enrichment
should not be limited by the geometric mass selectivity of the beamline but by other
sources of contamination, e.g. neutral 13C from un-ionized source gas diffusing to
the sample during deposition or gas scattering of ions.
12C samples were grown with typical beam currents ranging from 600 nA to
900 nA over a ≈ 1 mm2 beam spot, resulting in deposition rates between approxi-
106
Figure 3.6: Optical micrograph of a12C sample deposited at room temperature at
LC–2. The Si(100) substrate is seen with the 12C deposition spot near the upper
center appearing darker. The spot is roughly 1.3 mm in diameter. Lighter dust is
also visible on the substrate surface. Inset is a lower magnification image with the
edges of the chip visible.
mately 0.5 nm/min and 0.75 nm/min. An optical micrograph of a 12C sample after
deposition is shown in Fig. 3.6. The Si(100) substrate is seen with the 12C deposition
spot visible near the upper center appearing darker than the substrate. The spot
is roughly 1.3 mm in diameter. Lighter dust is also visible on the substrate surface
from handling the chip. Inset is a lower magnification image of the sample with the
edges of the chip visible giving an indication of the chip size. The sample discussed
in the analysis of this section was deposited with an average ion energy of ≈ 560 eV
in a background pressure of ≈ 2.1× 10−5 Pa (1.6× 10−7 Torr). This partial pres-
sure during growth was mostly CO2 leakage from the ion source and the typical base
pressure for the deposition chamber is ≈ 1.3× 10−7 Pa (1.0× 10−9 Torr). A total
of three 12C samples were produced in this work, and two samples were analyzed
for enrichment, one of which is the focus of the analysis below.
SEM was used to inspect one of the deposited 12C films in cross-section and
107
Figure 3.7: SEM cross-sectional micrograph of a 12C sample deposited at room
temperature at LC–2. The top layer is ≈ 27 nm of Au-Pt deposited for imaging.
The middle layer is the deposited 12C film which is ≈ 66 nm thick. Below that is
the crystalline Si(100) substrate.
measure the thickness. Figure 3.7 is a cross-sectional SEM micrograph of a cleaved
edge of the analyzed 12C sample. The micrograph shows the crystalline silicon
substrate on the bottom that has diagonal lines going from upper left to lower right,
which are a consequence of the cleaving. On top of the silicon is the deposited 12C
film, which also shows lines running vertically through the entire thickness of the
film. These lines suggest continuity and uniformity throughout the material that
would not be expected if the sample is more polycrystalline consisting of multiple
grains from the bottom to the top of the film. In the area of the sample shown in
the micrograph, the film is ≈ 66 nm thick. Above the 12C film is ≈ 27 nm of Au-Pt
deposited for imaging. The upper and lower edges of the carbon film are nearly
parallel, indicating smooth growth. While there are no obvious signs that the film
108
is amorphous, it is not expected that a crystalline film can be deposited at room
temperature on a substrate with a surface oxide layer.
3.3.4 Enrichment Measurements via SIMS
The enrichment in 12C and the residual 13C in the carbon films were deter-
mined using SIMS. One of these analyzed samples is represented on the enrichment
progression timeline in Fig. 1.12 in Chapter 1. For the SIMS measurements, a Cs+
primary ion beam with an impact energy of 20 keV was raster-scanned to sputter a
150 µm by 150 µm area of the sample. Secondary negative ions of 12C and 13C sput-
tered from the surface were alternately directed by magnetic peak-switching into a
secondary electron multiplier where the counts were recorded. Under the measure-
ment conditions, the instrument has a mass resolving power m
∆m
≈ 5000, allowing it
to distinguish between 13C and 12CH peaks, which are separated by approximately
0.0045 u. A set of measurements were made by pre-sputtering the target area with
an 8 nA primary ion current and then recording counts of 40 isotopic ratio pairs
with a sputter beam current of 23 pA.
Isotope fractions of 12C and 13C were determined from the averages of the 40
ratios by calculating zC/Ctot., where z is the mass number denoted as 12 for the
12C average counts and 13 for the 13C average counts. Ctot. is the sum of the
12C
and 13C average counts. The isotope fractions of 12C and 13C from eight sets of
SIMS isotopic ratio measurements after successive pre-sputter cycles are presented
in Fig. 3.8. The 12C (squares and line) and 13C (circles and line) isotope fractions
are shown in a semi-log plot of the “depth” profile vs. the cumulative pre-sputter
109
Figure 3.8: SIMS “depth” profile of a 12C sample deposited at room temperature
at LC–2. Isotope fractions of 12C (squares and line) and 13C (circles and line)
representing eight successive data runs as the primary SIMS beam sputtered through
the film are shown vs. the pre-sputter time. The inset shows 40 measurements of
13C/12C ratios (open circle and line) vs. sputter time, which are averaged (line)
to determine the values of the isotope fractions of the sixth points at a pre-sputter
time of 540 s. The middle six points between 180 s and 900 s were averaged for
the final enrichment and 13C isotope fraction of 39.2(13) ppm. Also shown is the
natural abundance value for 13C (dashed line) as a reference.
time, i.e. different depths into the film. The inset shows one set of 40 isotopic ratio
measurements of 13C/12C (open circles and line) vs. sputter time into the film for
the sixth data points of the isotope fractions at a pre-sputter time of 540 s. The
average (line) of the 40 measurements is used to determine the 13C and 13C isotope
fractions. The natural abundance of 13C (dashed line) is also shown for reference
highlighting the reduction of 13C. To determine the average isotope fractions of the
film, only the middle six data between pre-sputter times of 180 s and 900 s were
110
averaged, excluding the first and last data points. The first datum at 90 s shows a
lower enrichment that is likely due to natural abundance carbon contamination on
the film surface. The last datum at 900 s has lower enrichment but also corresponds
to a dramatic drop in absolute 12C counts indicating that the interface between
the carbon film and the Si(100) substrate had likely been reached. This depth was
later estimated to be roughly 100 nm. For the remaining six points, the average
measured 13C/12C ratio is 3.79(5)× 10−5. After accounting for an instrumental
mass fractionation for carbon, the average measured 12C isotope fraction in the
film was determined to be 99.99608(13) %. The average residual isotope fraction
of 13C in the film was measured to be 39.2(13)× 10−6 or 39.2(13) ppm, as seen in
the enrichment progression timeline. The isotope reduction factor for 13C based
on the isotope fraction is written as a13/(
13C/Ctot.). The isotope fraction of the
measurement here represents an isotope reduction factor for 13C of 270(20), i.e. the
13C in the film is approximately 270 times lower than in natural abundance C.
The data from Fig. 3.8 is also shown on a linear scale and over a smaller isotope
fraction range in Fig. 3.9 to highlight the structure of the “depth” profile. Here, the
12C isotope fractions (squares and line) corresponding to the left axis and 13C isotope
fractions (circles and line) in ppm corresponding to the right axis are shown again
vs. the pre-sputter time. Small changes in the isotope fractions throughout the film
are clearly visible with the sixth points at a pre-sputter time of 540 s representing
the highest local enrichment within the film. As mentioned, the first datum at a pre-
sputter time of 90 s and the last datum at 1260 s (crosses) were discarded when the
averages were determined. The middle six points between 180 s and 900 s represent
111
Figure 3.9: SIMS “depth” profile of a 12C sample deposited at room temperature
at LC–2. The data from Fig. 3.8 is shown on a linear scale. Isotope fractions of
12C (squares and line, left axis) and 13C (circles and line, right axis) representing
eight successive data runs as the primary SIMS beam sputtered through the film
are shown vs. the pre-sputter time. The middle six points between 180 s and
900 s were averaged for the final enrichment. The outside points of the 12C isotope
fractions (crosses) at 90 s and 1260 s were discarded due to surface contamination
and a loss of signal moving into the Si(100) substrate, respectively, represented by
vertical dashed lines. The average 13C isotope fraction of the middle six points is
39.2(13) ppm.
the bulk of the deposited film, represented by the vertical dashed lines, with surface
contamination and the Si(100) substrate on either side. Each error bar is derived
from the standard deviation of the mean of the 40 individual measurements at each
data point. The enrichment of this sample is comparable to that of CVD grown
12C diamonds used for NV center QI experiments [74,78]. A second 12C sample was
also analyzed by SIMS as a check on the reproducibility, and it was found to have
a nearly identical 12C enrichment. These 12C enrichment measurements represent a
112
realized mass selectivity of 276:1 for 12C in this system. This value gives a sense of
the performance of the system independent of the natural abundance of C. The 12C
realized mass selectivity is less than that achieved for 22Ne, but that may be due
to the estimated spatial peak separation between 12C and 13C at the mass-selecting
aperture (12.4 mm) being smaller than that between 20Ne and 22Ne (14.7 mm).
As mentioned above, it is believed that the limiting factor for the enrichment
of these samples is not the ultimate mass selectivity of the ion beam system, but
rather other sources of unwanted isotopes such as background CO2 gas diffusing
from the ion source. Therefore, the contribution to the 13C concentration from
this CO2 partial pressure is considered. Using the deposition conditions of a CO2
partial pressure of 2.1× 10−5 Pa (1.6× 10−7 Torr) and a natural 13C abundance in
the gas of 1.1 %, an incorporation fraction, s, of about 0.037 is needed to account
for the measured isotope fraction of 13C in the film of ≈ 39 ppm. Representative
sticking coefficients found in the literature range from values that are similar in
magnitude to several times lower [83]. The former case provides a reasonable ex-
planation for the 13C observed while the latter may indicate an additional source of
contamination such as scattered ions in the beam path. These sources of 13C can
be mitigated by reducing the background pressure during deposition. For 13C due
to adsorption from the vacuum, every order of magnitude decrease in background
pressure would correspond to an order of magnitude increase in the enrichment
of the deposited material. Reducing the background gas in the above estimate to
2× 10−6 Pa (1.5× 10−8 Torr) would lower the isotope fraction of 13C to ≈ 4× 10−6
(4 ppm) under the same deposition conditions. An improved vacuum would also
113
reduce chemical impurities incorporated during deposition, which could be param-
agnetic. Increasing the beam flux and therefore deposition rate will further reduce
the effect of unwanted background gases adsorbing into the films by reducing the
relative 13C gas flux compared to the deposition rate.
Time-of-Flight SIMS (TOF-SIMS) was also used to look at chemical contami-
nants present in a sample similar to the one analyzed for enrichment. All signals of
organic and inorganic species being monitored for analysis were near the noise floor
of the measurement and at least 100 times lower than the 12C signal. The derived
concentrations of chemical impurities in the film are unreliable because the sample
was not being efficiently sputtered and ionized for detection. Any signal from O
adsorbed during deposition from a CO2 background partial pressure is below the
detection of this measurement. Further assessment needs to be done to determine
the concentrations of spin and chemical impurities present in these films.
3.4 Chapter 3 Summary: Outlook for 28Si
This chapter demonstrated successful proof of principle enrichment by im-
planting the minor isotope 22Ne enriched with an isotope fraction of 99.455(36) %
into Si as well as depositing thin films of 12C enriched to an isotope fraction of
99.9961(4) % onto Si using the mass selected ion beam system. A total of three
22Ne and three 12C were produced. The isotope fractions of 20Ne and 13C from these
measurements can be seen in the enrichment progression timeline in Fig. 1.12. Re-
alized mass selectivities of 1785:1 for 22Ne (selection between masses separated by
114
2 u) and 276:1 for 12C (selection between masses separated by 1 u) were achieved.
While these selectivities are not directly translatable to 28Si because the peaks of Si
are spatially closer than for Ne or C, these experiments do show that very high levels
of in situ enrichment are possible in thin film deposition using this system. Similar
levels of 28Si enrichment are likely possible, especially if background gas adsorption
has a more significant effect on the realized sample enrichment than does the mass
selectivity.
115
Chapter 4
28Si Thin Film Deposition and
Characterization Phase I: In Situ
Enrichment
4.1 Introduction
4.1.1 Context
In this chapter as well as Chapter 5, the ion beam deposition of 28Si films
enriched in situ is discussed as well as the characterization of their properties, par-
ticularly their enrichment level. Chapter 3 demonstrated successful proof of principle
enrichment experimenubots by implanting 22Ne enriched to 99.455(36) % into Si as
well as depositing thin films of 12C enriched to 99.9961(4) % using the mass selected
ion beam system. Realized mass selectivities of 1785:1 for 22Ne (selection between
masses separated by 2 u) and 276:1 for 12C (selection between masses separated
by 1 u) were achieved. Because the 28 u and 29 u mass peaks are spatially closer
together at the mass-selecting aperture than the Ne or C peaks, the selectivity for
116
Si deposition is lower than these values. Nevertheless, these experiments showed
that high levels of isotopic enrichment can be achieved with this method, laying the
groundwork for adapting the system for 28Si deposition.
As stated in Chapter 1, enrichment and thin film deposition of 28Si is pursued
here with the objective of producing high-quality enriched material for Si based solid
state quantum computing. 28Si of sufficiently high quality (i.e. high enrichment,
crystallinity, and purity) provides an ideal solid state environment to host qubit
spins. The minimal interactions between 28Si and nuclear and electron spins of
qubits result in a level of isolation akin to a trapped atom in a vacuum chamber.
This leads to extremely long coherence times which have earned 28Si the moniker of
“semiconductor vacuum” [9]. Unwanted deviations from ideal 28Si material can be
classified as three types of defects: isotopic defects, structural defects, and chemical
defects. Controlling and limiting these defects is critical for successful integration
of 28Si into quantum computing architectures. The 28Si materials goals of this work
are discussed in Chapter 1 and can be restated as follows:
(1) high enrichment in 28Si with a residual 29Si isotopic concentration less than
50 ppm,
(2) single-crystalline and smooth epitaxial structure with a low dislocation density
below 1× 106 cm−3, and
(3) high chemical purity including C and O with atomic concentrations below
2× 1015 cm−3.
These are believed to be the criteria needed for the 28Si to be comparable to elec-
tronics grade natural abundance Si as well as the enriched Si currently available
117
in the QI research community. The bulk of that 28Si is produced by the Interna-
tional Avogadro Coordination [32], which has a residual 29Si isotopic concentration
as low as 50 ppm. Producing 28Si with 29Si isotopic concentrations as low as 1 ppm
is necessary to enable a robust and systematic study measuring electron coherence
times vs. 29Si concentration in the single spin regime and compare it to theoretical
predictions (see Fig. 1.9), as discussed in Chapter 1 [12].
The experiments producing 28Si discussed in this chapter and Chapter 5 rely
not only on prior work using this ion beam deposition system, such as the Ne
and C experiments of Chapter 3 and previous thin metal film deposition experi-
ments [59], but also on previous work by other groups that have deposited Si via
an ion beam [43, 48–51]. The results from these groups were reviewed in Chap-
ter 1 showing that they demonstrated both enrichment in 28Si to approximately
99.9982 % in one experiment [43] as well as epitaxial deposition using hyperthermal
energy ions and a range of substrate temperatures [51].
The experiments described in this chapter and Chapter 5 seek to use process-
ing methods that are both common (e.g. vacuum deposition, sample heating) and
fairly unique (e.g. mass selected ion beam deposition) to engineer the properties,
such as enrichment in 28Si and chemical purity, and structure (crystallinity) of Si
thin films. Characterization methods including SIMS for assessing sample enrich-
ment and chemical purity, STM, RHEED, and SEM for inspecting the film surface
and crystallinity, TEM to inspect the bulk crystallinity of films, and XPS for de-
tecting chemical impurities are used to assess the 28Si films in terms of the materials
goals and guide the experimental adjustments needed to improve their quality. The
118
experiments discussed in this chapter focus on achieving very high levels of 28Si
enrichment. The experiments of Chapter 5 will focus on maintaining a high en-
richment while assessing and improving the chemical purity and crystallinity of 28Si
samples.
4.1.2 Experimental Configurations for 28Si Deposition
Three distinct experimental configurations were used for deposition of 28Si
films in this work. Experiments involving the first two will be discussed in this
chapter, and experiments involving the third one will be discussed in Chapter 5.
These experimental configurations are defined partly by the location of samples
during deposition at three positions in the vacuum system. Additionally, they are
defined as a chronology of initial, intermediate, and final experimental configura-
tions with materials characterizations and subsequent experimental improvements
occurring between segments of 28Si deposition at each one. The final configuration
is the final (last) setup used in this work, although not necessarily the final setup
used in the larger enriched Si project of which this work is a part. These three
setups are illustrated in Fig. 4.1, which shows top down schematics of the ion beam
chamber, deceleration lens chamber, and deposition and analysis chamber in the
three experimental configurations used to deposit 28Si samples. Schematic drawings
of these chambers were previously shown in Fig. 2.1 in Chapter 2 and described in
detail there. Highlighted here are the three sample locations used in each of the
setups, which are discussed further below.
The initial experimental configuration was one in which samples were located
119
Figure 4.1: Schematic top down drawings of the ion beam chamber, lens chamber,
and deposition chamber experimental configurations used to deposit 28Si samples.
The top left shows the setup for Sample Location 1 in the ion beam chamber (IC–1).
This setup consists of the ion beam with the sample located on a feedthrough at
the end of the beamline, shown in the blowup. The top right shows the setup for
Sample Location 2 in the lens chamber (LC–2). This setup consists of the ion beam
chamber connected to the lens chamber with the sample located on a feedthrough
at the end of the deceleration lenses, shown in the blowup. The bottom shows the
setup for Sample Location 3 in the deposition chamber (DC–3). This setup consists
of the ion beam and lens chambers connected to the deposition chamber with the
sample located on the manipulator, shown in the blowup.
120
just after the mass analyzer and mass-selecting aperture at the end of the ion beam
chamber. This is the location corresponding to the schematic setup shown in the
upper left section of Fig. 4.1 marked as “Sample Location 1: Ion Beam Chamber”,
which will be referred to as IC–1. The sample was mounted on an electrical vac-
uum feedthrough for deposition, which is shown in the transparent blowup of this
schematic. It consists of an isolated metal rod connecting two sides of a vacuum
flange. For the samples deposited at IC–1, the ion beamline was separated from
the other vacuum chambers just after the mass-selecting aperture. Initially this
setup was designed to maximize the total ion current onto the sample by position-
ing the sample before the electrostatic deceleration lenses. Additionally, this simple
configuration facilitated changing the mass-selecting aperture relatively quickly to
test apertures with different dimensions in an effort to optimize ion beam fluence
through it. This setup also had a higher background pressure during deposition due
to the beamline being disconnected from the rest of the vacuum system, which nor-
mally differentially pumps the sample location. Only a single turbo pump, marked
in Fig. 4.1, pumps this setup.
Next, the intermediate experimental configuration was used to deposit 28Si
samples in the deceleration lens chamber. This is the sample location correspond-
ing to the schematic setup shown in the upper right section of Fig. 4.1 marked as
“Sample Location 2: Lens Chamber”, which will be referred to as LC–2. In this
configuration, the ion beam chamber was reconnected to the deceleration lens cham-
ber (separated by a gate valve), and the samples were placed on a new sample stage
just after the lenses. This sample stage consisted, in part, of an electric feedthrough
121
to mount the sample, which is shown in the transparent blowup in this schematic.
This setup was used for almost all the samples deposited at room temperature. It
enabled better control of the ion beam by providing a means of focusing the beam
using the deceleration lenses. The lens chamber also has additional vacuum pump-
ing in the form of a second turbo pump, marked in Fig. 4.1 which differentially
pumped the sample location compared to the ion beam chamber. This resulted in
depositions at lower background pressures. This setup is also the same one used in
the Ne and C deposition experiments from Chapter 3. Depositing samples at LC–2
was advantageous because it provided relatively easy access to the sample stage.
This enabled moderately quick sample exchanges as well as the ability to perform
quick modifications to the sample stage in efforts to optimize the deposition process.
Photographs of the experimental setups for IC–1 and LC–2 are shown in Fig. B.9
and B.10 in Appendix B.
Lastly, the final experimental configuration used for depositing 28Si in this
work is one where the entire ion beamline, including the ion beam chamber and
the deceleration lens chamber, is connected to the deposition and analysis chamber.
This is the sample location corresponding to the schematic setup in the lower sec-
tion of Fig. 4.1 marked as “Sample Location 3: Deposition Chamber”, which will
be referred to as DC–3. In this setup, samples were placed on the 5-axis manip-
ulator for deposition of 28Si, shown in the transparent blowup of this schematic.
This configuration enabled lower background pressures, sample heating, and use of
analytic instruments, and it is discussed further in Chapter 5. This lower schematic
in Fig. 4.1 also marks the approximate locations of samples on the full combined
122
system for each of the three experimental configurations for depositing 28Si.
In Chapter 1, the timeline progression of the best sample enrichment values
achieved throughout this work using the ion beam deposition system was shown in
the enrichment progression timeline in Fig. 1.12. This timeline will be discussed
throughout this chapter as well as Chapter 5 in terms of the experimental changes
that led to improvements in the 28Si enrichment. A version of this timeline showing
the isotope fractions measured by SIMS of 29Si (squares) and 30Si (triangles) vs. de-
position date for just the 28Si samples is presented in Fig. 4.2 along with indicators of
the sample location during deposition. As mentioned in Chapter 1, isotope fractions
of a particular isotope are defined in a SIMS measurement as the average detected
counts of that isotope divided by the total average counts of the measurement. The
isotope fractions of Si are written as zSi/Sitot., where z is the mass number denoted
as 29 for 29Si average counts and similarly for 30Si and 28Si and Sitot. is the sum
of 28Si, 29Si, and 30Si average counts. Uncertainties in the isotope fractions are
derived from uncertainties in the SIMS measurements of those samples. Nine 28Si
samples out of a total of 61 produced in this work are represented on this timeline.
Each of the nine samples were the most highly enriched of any samples produced
up to that point on the timeline, that is, they represent new record enrichments
for 28Si samples achieved for this work. One can see that overall, the 29Si isotope
fraction was reduced from 2822(18) ppm in the initial sample, down to a minimum
of 127(29) ppb with an overall enrichment in 28Si of 99.9999819(35) % for the most
highly enriched sample produced in this work. This level of enrichment exceeds that
of all other known sources of 28Si and will be discussed in Chapter 5. This chapter
123
Figure 4.2: Enrichment progression timeline. A timeline of the progression of the
lowest residual isotope fractions of 29Si (squares) and 30Si (triangles), as measured
by SIMS. These were achieved for 28Si samples deposited over approximately three
and one half years. Groups of samples are labeled based on their deposition loca-
tions, IC–1 in the ion beam chamber, LC–2 in the lens chamber, and DC–3 in the
deposition chamber.
will focus on the enrichment of the samples in the first two sections of this timeline.
In total, 21 28Si samples were deposited at IC–1 and LC–2 combined.
Section 4.2 of this chapter will discuss samples deposited at IC–1 in the initial
experimental configuration and SIMS measurements of their enrichment. Section
4.3 will discuss samples deposited at LC–2 in the intermediate configuration, SIMS
measurements of their enrichment, and initial characterizations of their crystallinity
by TEM and chemical purity by SIMS and XPS. Finally, Section 4.4 will give a
brief summary of this chapter and the achieved enrichment values in the context
124
of the enrichment progression timeline. Some data presented in this chapter was
previously published in Ref. [84].
4.2 Si Deposition Proof of Principle: Ion Beam
Chamber Samples
4.2.1 Experimental Setup
This section discusses the initial deposition of 28Si samples in an experimental
configuration with the sample located in the ion beam chamber at IC–1 during de-
position. A schematic of the ion beam chamber in this setup is shown in the upper
left section of Fig. 4.1. Depositing 28Si at IC–1 was a proof of principle experiment
for the adaptation from depositing enriched C films to depositing enriched Si. One
experimental change implemented for the transition from C deposition to Si deposi-
tion was the initial use of the gas-mode Penning ion source described in Chapter 2.
This source was designed to more efficiently crack and ionize gas to generate an ion
beam as opposed to ions being generated by sputtering a solid target. Using this
ion source, an increase in the mass analyzed ion beam fluence was achieved from
an average of 0.55 µA for 12C depositions to an average of 0.92 µA for the initial
28Si samples described in this section. Additionally, the gas-mode source allows
a reduction in the source working pressure of nearly an order of magnitude from
≈ 1.3× 10−3 Pa (1.0× 10−5 Torr) used for the 12C deposition to ≈ 1.7× 10−4 Pa
(1.3× 10−6 Torr) for better integration with UHV deposition environments.
Samples at IC–1 were located immediately after the mass-selecting aperture,
125
and they were mounted on the end of an electrical feedthrough using a strip of
conductive carbon tape on the back of the chip. This feedthrough and the sample
location of IC–1 are shown in Fig. 4.1 in the blowup of the upper left section of
the figure. A ceramic electric break was used to isolate the mounting flange of
the electric vacuum feedthrough from the high voltage ion beam chamber. The
feedthrough and sample are both on axis with the ion beamline optics, but there
was no sample motion available in this configuration. Therefore, the ion beam must
be precisely tuned to be on axis as well. The purpose of this feedthrough was
to isolate the sample electrically from the chamber so that the ion beam current
could be monitored during beam tuning and deposition. A photograph of a sample
mounted on the vacuum feedthrough after a 28Si deposition is shown in Fig. B.9 in
Appendix B.
Using a small feedthrough to mount the sample in this way is advantageous in
that its simplicity allows for quick and easy sample loading and unloading. However,
it is disadvantageous in that there is no way to shield the sample substrate before
deposition, which means that it is exposed to and accumulates material from the ion
beam during initial tuning procedures and sweeps of the ion beam mass spectrum.
Additionally, no sample heating capabilities are available when using this setup,
which precludes in situ sample annealing and makes crystalline growth unlikely. The
mass-selecting aperture consisted of a slit approximately 1 mm in width, i.e. the
same direction that different mass ion beams are spatially separated. The aperture
was also approximately 15.25 mm tall and 2 mm thick. This aperture was expected
to provide an improvement in mass resolving power over the aperture used for the Ne
126
and C samples described in Chapter 3, which was an approximately 5 mm diameter
circular hole that was 16 mm thick.
As mentioned, the sample location was not differentially pumped in this con-
figuration, and so the base pressure and deposition pressures were that of the ion
beam chamber itself. The base pressure of the ion beam chamber for these sam-
ples ranged from approximately 6.5× 10−6 Pa to 1.3× 10−5 Pa (4.9× 10−8 Torr to
1.0× 10−7 Torr) before deposition. A residual gas mass spectra of the base pressure
of this chamber was shown in Fig. 2.22 in Chapter 2. This high base pressure is
due to ambient air leakage through the o-rings used to seal the ion source as well
as a gate valve, which are only rated for high vacuum, i.e. a minimum pressure of
roughly 1.3× 10−7 Pa (1.0× 10−9 Torr). As previously mentioned, SiH4 was used
as the source gas for 28Si deposition. SiH4 was injected into the ion source via a
UHV leak valve to generate a plasma that cracks and ionizes the SiH4, producing the
Si ion beam. The SiH4 gas used for these samples and all
28Si samples in this work
had a natural abundance of isotopes and a purity of 99.999 % according to the gas
vendor (Matheson Tri-Gas). During operation of the ion source, a working pressure
for the low pressure plasma mode typically around 1.7× 10−4 Pa (1.3× 10−6 Torr)
was chosen, as measured by the ion gauge. It should be noted that because there
is no differential pumping at the sample location and thus no outlet for gas diffu-
sion, there may be a slightly higher local pressure at the sample location. This is
because the ion beam itself carries a lot of hydrogen in the form of a H+2 beam as
well as H from SiH4 fragments. This fluence of particles is delivered close to the
sample before being blocked at the aperture where it can increase the local partial
127
pressures of H2 and SiH4. It is estimated that this effect may only contribute a
15 % higher total pressure at the sample, but it is nevertheless undesirable to have
the sample in an environment with unknown partial pressures which may lead to
increased adsorbates in the deposited film. SiH4 adsorption is of particular concern
in this work because, as described in Chapter 2, naturally abundant SiH4 (including
29SiH4) adsorbed onto the deposition surface becomes incorporated into the film
resulting in higher concentrations of 29Si and 30Si in the sample. This subject is
explored in great detail in Chapter 6.
4.2.2 Sample Preparation
For the initial 28Si deposition tests, very little ex situ sample preparation
occurred. Substrates consisted of “lightly doped” natural abundance commercial
Si(100) wafers that were cleaved by hand into approximately 1 cm by 1 cm chips.
The chips were handled with clean teflon tweezers and mounted on the end of the
feedthrough using a strip of carbon tape on the back of the wafer, and they were
loaded into the vacuum chamber with a native oxide. No further sample preparation
occurred in situ. The role of these substrates is to simply be a “catcher foil” to collect
the 28Si ions, and one substrate used here actually was a Ag foil. Typically, after
being loaded, samples sat several days in the vacuum chamber before deposition.
4.2.3 Deposition of 28Si
The background deposition pressure for these samples was roughly the operat-
ing pressure of the ion source, and so during deposition, the pressure in the chamber
128
rose to between approximately 1.6× 10−4 Pa and 2.3× 10−4 Pa (1.2× 10−6 Torr to
1.7× 10−6 Torr), as measured by an ion gauge in a different section of the chamber.
28Si ions were deposited onto the Si(100) substrates at room temperature. This tem-
perature was not measured but assumed to be similar to the ambient temperature
outside of the vacuum chamber, which was typically measured to be 21 ◦C ± 2 ◦C.
Initially, an average ion energy, Ei, at the sample of approximately 455 eV was used.
This value of Ei was chosen to be similar to the ion energy previously used for
12C
deposition. The energy was lowered to around 64 eV for the remainder of these
samples to reduce the sputter yield during deposition. Sputter yield values for Si
are shown in Fig. 2.19 in Chapter 2, as determined from TRIM calculations [69]. For
Si ions striking a Si target, the sputter yield is around 53 % at 455 eV. This means
that on average, every incident ion sputters 0.53 atoms from the surface leading to
a deposition rate that is effectively reduced by half. The sputter yield for Si at an
energy of 64 eV is reduced to only about 8 %. Sputtering is more significant for Si
deposition than for C deposition, which has sputter yields that are about half the
value of those of Si for a given ion energy.
The other significant aspect of the incident ion energies used here is that they
are in the hyperthermal energy regime, as discussed in Chapter 2. The energy pos-
sessed by the ions can be transferred into the depositing film promoting epitaxial
deposition. However, only amorphous deposition is expected for the samples dis-
cussed in this section due to the low substrate temperature and the presence of
a surface oxide on the substrates. For these samples, 28Si ion beam currents, Ii,
ranged from 770 nA to 1.1 µA. A mass spectrum from the measured ion current for
129
28Si and other SiH4 ions, which is representative of the configuration and ion beam
settings for samples deposited at sample location IC–1, is shown in Fig. 4.3. The
corresponding magnetic sector analyzer current used for the field sweep is shown
on the top axis of this figure. Ion current peaks on this semi-log plot (circles) are
observed between 28 u and 32 u. As discussed in Chapter 2, the 28 u current peak
is 28Si, and the 29 u current peak is 28SiH and approximately 6 % 29Si based on the
peak heights and the expected natural abundance. Ions of other Si hydrides up to
28SiH4 are also generated in the ion source. A sum of Gaussian fits to the peaks (line,
Eq. (2.14)) are also shown superimposed on the data, which they match fairly well.
As discussed in Chapter 2, the spacial distribution of ions in the beam is expected to
be roughly Gaussian for this system. The mass resolving power of the ion beam in
this configuration derived from this mass spectrum is m
∆m
≈ 38 (measured at 10 %
of the peak height). The high ion current level between peaks of more than one
tenth of the 28 u maximum (630 nA vs. 75 nA) possibly indicates a poorly focused
beam leading to a large overlap between peaks. This could limit the separation of
the 28Si ion beam from the 29Si ion beam and reduce the enrichment that is achieved
with the settings used for these samples. The separation of the peaks can also be
determined from the Gaussian fits, which give a standard deviation of the 29 u peak
of σ ≈ 0.14. This results in a peak separation between the 28 u and 29 u peaks of
approximately 7 σ.
The 28Si ion beam in this setup had a fairly large average spot size leading to a
deposition spot on the substrate of about 20 mm2. The 28Si spot is easily visible on
the substrates due to the difference in color from the underlying native oxide. This
130
Figure 4.3: SiH4 mass spectrum representative of the ion beam settings for samples
deposited at room temperature at IC–1. The ion current (circles) is recorded while
sweeping the mass analyzer current, and thus the magnetic field (top axis). The
28 u peak is 28Si and the 29 u peak is both 28SiH and ≈ 6 % 29Si. Other higher order
hydrides are also observed. Gaussian fits (line, Eq. (2.14)) to the peaks are shown
superimposed on the data. The centers of the 28 u and 29 u fits are separated by
≈ 7 σ.
large spot size is a result of not being able to use the focusing deceleration lenses in
this configuration. Despite relatively high ion currents achieved in this configuration,
the large spot size reduces the achievable growth rate for a given ion fluence, as
compared to a more compact spot. The resulting thickness, d, of one deposited film
was about 85 nm with a deposition rate, R, of 0.41 nm/min derived from dividing
the thickness by the deposition time. The thickness was measured by SEM cross-
sectional microscopy. A SEM micrograph of this 28Si sample is shown in Fig. 4.4. A
photograph of the sample as deposited is inset showing the Si(100) substrate with
131
Figure 4.4: SEM cross-sectional micrograph of an amorphous 28Si thin film deposited
at room temperature at IC–1. The Si(100) substrate is seen in the lower portion of
the image and the 28Si film is above it. A dashed line marks the interface between
the film and substrate. The thickness of the 28Si film here is approximately 86 nm.
Au-Pt is deposited on the top surface of the sample to to protect the film during
sample cleaving. Inset is a photograph of the sample after deposition showing the
roughly 6 mm wide 28Si spot on the substrate.
the 28Si deposition spot, which is roughly 6 mm wide. This SEM micrograph was
acquired in collaboration with Dr. Michael Stewart (NIST). The Si(100) substrate
is seen in the lower portion of the micrograph with the 28Si film appearing as a
slightly lighter region above it. A dashed line marks the interface between the film
and substrate. The thickness of the 28Si film seen here is approximately 86 nm. The
28Si appears to have a different texture and possibly structure than the substrate,
possibly indicating grains about 10 nm in size, although it is more likely that the
film is amorphous. Amorphous deposition is expected for these samples because
the presence of a native oxide prevents crystalline registration of the atoms in the
132
depositing film which could otherwise lead to epitaxy. Additionally, depositing at
room temperature means that the likelihood of forming polycrystalline grains is
low [85]. The thickness of two other samples was inferred from the calibration of the
SIMS depth profiles (discussed below). The thickness of one sample was estimated
to be only about 15 nm with a corresponding deposition rate of 0.07 nm/min, and
the other sample thickness ranged from approximately 75 nm to 92 nm across the
deposition spot with deposition rates between 0.33 nm/min and 0.40 nm/min.
To summarize, the typical deposition procedure for samples deposited at IC–1
was as follows:
1. a substrate is cleaved before being mounted onto the electrical feedthrough
and loaded into the ion beam chamber, which is pumped out for > 12 h,
2. SiH4 is then introduced into the chamber and the ion source is turned on. The
beam is tuned including characterization of the mass spectrum by collecting
the ion current on the substrate,
3. the ion beam is then tuned to the 28 u peak to commence deposition of 28Si
while monitoring the ion current onto the sample as a measure of the deposition
rate,
4. after typically three to four hours of deposition, the ion source is turned off
to end deposition, and the SiH4 leak valve is closed to reduce the ion beam
chamber pressure back to its base, and
5. the sample is then removed from the vacuum chamber for ex situ analyses by
venting the ion beam chamber.
A total of five 28Si samples were deposited at IC–1 and all used this procedure.
133
4.2.4 Enrichment Measurements via SIMS for IC–1
Samples
The enrichment of these samples and all samples discussed in this work were
measured ex situ by SIMS using a CAMECA IMS-1270E7 large geometry spec-
trometer as mentioned in Chapter 3. 28Si, 29Si, and 30Si isotopes were measured in
collaboration with Dr. David Simons (NIST) and Dr. Shinichiro Muramoto (NIST).
Dr. Simons collected the depth profile data for all but one of the samples discussed
in this thesis. The raw data was then mostly analyzed by myself. The basic SIMS
process for measuring Ne and C isotope ratios and isotope fractions was introduced
in Chapter 3, and the measurement is similar for Si isotope fractions.
28Si samples are bombarded with a primary ion beam of O+2 which sputters
the sample at a constant rate and is rastered across an area typically 50 µm across.
Ejected Si ions are collected into a secondary ion beam for analysis in a mass spec-
trometer. 28Si, 29Si, and 30Si ions are collected separately using an electron multiplier
and the counts are recorded for a given time interval, or SIMS cycle, to get a count
rate. With a mass resolving power m
∆m
= 6000 (measured at 10 % of the peak height)
for this instrument, 28SiH is easily distinguished from 29Si in these measurements.
An example SIMS mass spectrum (line) of these two separated peaks is shown in
Fig. 4.5. The ion masses are separated by only approximately 0.008 u, but are
well resolved in the SIMS instrument. This resolution is necessary for an accurate
measurement of the ratio of 29Si to 28Si. The count rates are used to determine
the isotope ratios 29Si/28Si and 30Si/28Si. To reduce discrete counting noise, isotope
134
Figure 4.5: SIMS mass spectrum of 29Si and 28SiH ion currents (line) showing that
the two peaks, which are separated by about 0.008 u, are well resolved in the SIMS
instrument. This resolution is necessary for an accurate measurement of the ratio
of 29Si to 28Si.
ratios for each cycle are then averaged together from the highly enriched portion
of the 28Si films after confirming no systematic trend to calculate the total average
isotope ratios for that measurement. Isotope count ratios are converted into isotope
fractions of the form (zSi/Sitot.), which was previously described. The uncertainty
of the isotope fractions was determined from the standard deviation of the mean of
the measurements.
The depth profiles are calibrated by measuring the depth of the crater formed
from sputtering the sample during the measurement. Crater depths were measured
using a stylus profilometer, and then this value was used to determine the depth
at each cycle by assuming a constant sputter rate. Because the measurement area
is much smaller than the typical deposition area, multiple SIMS measurements are
135
sometimes made on the same sample. Measurements are typically made in the
thickest portion of the 28Si film because better measurement statistics result from
analyzing more material within a single measurement. The results corresponding to
the highest individual enrichment measured in each sample is presented in this work
instead of an average of the multiple measurements. This is because the average of
multiple measurements depends on factors such as the number of measurements and
their location across the deposition spot, which are not consistent between different
samples. Different locations across the deposition spot have different enrichments
because the deposition rate and thus relative SiH4 adsorption is not constant across
the sample, which will be discussed further later. The multi-spot averages are there-
fore not a reliable metric for comparing the overall enrichments of samples, although
they can give an idea of the variation in enrichment across a single sample. Addi-
tionally, the best measured enrichment for a sample is a preferred metric because
it gives a lower bound on the best possible enrichment achievable by the deposition
system. Further details of the specific SIMS measurements used for the different
sets of samples discussed in this chapter can be found in Appendix E.
Of the five samples deposited at IC–1, three are represented on the enrichment
progression timeline in Fig. 4.2 and will be discussed here. A SIMS depth profile for
a 28Si sample deposited at IC–1 is shown in Fig. 4.6. 28Si (circles), 29Si (squares), and
30Si (triangles) isotope fractions are shown vs. sputter depth into the sample. This
measurement as done using a TOF-SIMS instrument. At very shallow depths below
the sample surface (0 nm to 10 nm), the isotope ratios are inflated to higher values
due primarily to surface contamination from the sample being exposed to the ambi-
136
Figure 4.6: SIMS depth profile for a 28Si sample deposited at room temperature at
IC–1. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope fractions are shown vs.
sputter depth into the sample. The average 29Si isotope fraction is 1130(14) ppm.
The natural abundance values for each isotope (dotted and dashed lines) are also
shown for reference. The 30Si data fluctuates between two values due to either zero
(not shown), one, or two counts being detected in each measurement cycle. At
92 nm, the isotope fractions return to their natural abundance values indicating the
interface with the substrate (shaded region). This measurement was done using a
TOF-SIMS instrument.
ent environment before the SIMS measurements. This initial “surface tail” artifact
is typical of SIMS and is seen in all the measurements discussed here. Isotope ratios
from each measurement cycle were averaged together from about 20 nm to 80 nm to
determine total isotope fractions for this sample. At a depth of around 92 nm, the
isotope fractions return to their natural abundance values (dotted and dashed lines)
indicating the interface with the natural abundance Si substrate (shaded region) and
giving an estimate of the 28Si film thickness. This value for the depth of the substrate
interface is determined as the point at which the 29Si and 30Si isotope fractions re-
137
turn to half of their natural abundance values. It is difficult to infer any quantitative
information about the width of the interface between the film and substrate from the
depth profile because SIMS measurements tend to exaggerate interface widths. This
can be due to interface roughness, which can be intrinsic to the sample or caused
by the SIMS sputtering process itself [86]. For the sample measured in Fig. 4.6,
the average measured 28Si isotope fraction in the 28Si film is 99.8850(14) %. The
average 29Si isotope fraction is 1.130(14)× 10−3 (1130(14) ppm), and the average
30Si isotope fraction is 2.03(14)× 10−5 (20.3(14) ppm). The 30Si signal appears in
two bands because of discrete counting fluctuations between one and two counts in
each SIMS data cycle and most 30Si data being zero counts in some cycles. This
measurement was the more highly enriched of two SIMS measurements performed
on different spots on this sample. The average 29Si isotope fraction of the two spots
for this sample is 1332(13) ppm.
As previously discussed in Chapter 3, the isotope reduction factor is another
useful parameter that describes the measured enrichment of a sample and, more
specifically, gives the amount by which an excluded isotope is reduced from the
natural abundance of that element. The reduction factor is determined by dividing
the natural abundance of an isotope of an element by the measured isotope fraction.
az represents the natural abundance of an isotope and, again, z is the mass number
of the isotope denoted as 29 for 29Si and 30 for 30Si. The isotope reduction factors
of the minor Si isotopes in the film are thus written as az/(
zSi/Sitot.). In the SIMS
measurement discussed here, the isotope reduction factor for 29Si is determined to
be 41.5(5) (i.e. approximately 41 times lower than the natural abundance of 29Si).
138
The reduction factor of 30Si in this measurement is higher than that of 29Si, having
a value of 1.5(1)× 103. This is likely due to the 30 u peak being farther away
spatially from the 28 u peak at the mass-selecting aperture. These measurements
indicate a realized mass selectivity for this configuration of approximately 45:1 for
29Si, which is lower than what was achieved for 12C. This sample is less enriched
than the 12C samples by a factor of about 30 despite the mass resolving power
determined from the mass spectrum being similar. This is perhaps not surprising
considering that the 28 u and 29 u peaks have a spatial separation (5.5 mm) at
the mass-selecting aperture approximately 44 % that of the 12 u and 13 u peaks
(12.4 mm). Additionally, the natural abundance of 29Si relative to 28Si is higher
than the natural abundance of 13C relative to 12C, so a 28Si sample deposited with a
given ion peak separation would be less enriched than a 12C sample deposited with
the same peak separation. A second 28Si sample deposited at IC–1 was measured
by SIMS as well, but it showed a 29Si isotope fraction of 2822(18) ppm, which
is approximately twice as large as the previous sample. This is potentially due
to slightly different beam tuning parameters and a slightly lower mass selectivity
achieved at the time of deposition.
After depositing these two samples and reviewing the SIMS data, a third 28Si
sample was deposited in an effort to better tune the ion beam parameters and
increase the mass spectrum geometric selectivity, i.e. reduce the overlap of the 29 u
peak onto the 28 u peak. This effort resulted in a lowering of the residual 29Si isotope
fraction by nearly a factor of 200. A SIMS depth profile of the most highly enriched
spot measured for this sample is shown in Fig. 4.7. 28Si (circles), 29Si (squares),
139
Figure 4.7: SIMS depth profile for the most highly enriched 28Si sample deposited
at room temperature at IC–1. 28Si (circles), 29Si (squares), and 30Si (triangles)
isotope fractions are shown vs. sputter depth. The average 29Si isotope fraction is
9.5(10) ppm. The natural abundance values for each isotope (dotted and dashed
lines) are also shown for reference. At a depth of 75 nm, the isotope fractions return
to their natural abundance values indicating the interface between the film and the
substrate (shaded region).
and 30Si (triangles) isotope fractions are shown vs. sputter depth. At a depth
of around 70 nm, the isotope fractions return to their natural abundance values
(dotted and dashed lines) giving an estimate of the 28Si film thickness. The substrate
is marked by the shaded region. The average measured 28Si isotope fraction in
the 28Si film is 99.99846(19) %. The average 29Si isotope fraction is 9.5(10)× 10−6
(9.5(10) ppm), and the average 30Si isotope fraction is 5.9(16)× 10−6 (5.9(16) ppm).
Isotope fractions were determined by averaging the data from depths between 20 nm
to 65 nm. The isotope fraction of 29Si in this measurement is a nearly a factor of
120 lower than that of the previous sample. These measurements indicate a much
140
larger average isotope reduction factor of approximately 5.1(8)× 103 for 29Si and
30Si. The large improvement in 29Si reduction factor but modest improvement in
30Si reduction factor shows that the ion beam was indeed tuned better for 29Si
rejection during deposition of this sample. For this sample, a total of four areas
across the deposition spot were measured by SIMS with similar results to the most
highly enriched one mentioned above. The average 29Si isotope fraction of the four
measurements from this sample is 11.28(54) ppm. This result shows that a very
high level of enrichment in 28Si is achievable, which surpasses the enrichment goal
stated at the beginning of this chapter.
It is difficult to determine the exact cause of the large reduction in the 29Si
isotope fraction, as seen in the enrichment progression timeline in Fig. 4.2. Pre-
sumably, the peak separation between the 28 u and 29 u ion current peaks in the
mass spectrum was significantly larger for this final sample than for earlier samples
resulting in a realized selectivity, although that data was not recorded.
4.2.5 Summary of Results for IC–1 Samples
The 28Si samples deposited at IC–1 in the initial experimental configuration
showed that the mass selected ion beam deposition method, demonstrated first in
in Chapter 3, could be adapted to produce enriched 28Si thin films. It established
the experimental parameters and methods needed for depositing 28Si from a nat-
ural abundance SiH4 source gas and subsequently characterizing the enrichment.
The most highly enriched 28Si sample produced in this configuration has an average
measured 28Si isotope fraction of 99.99846(19) %, an average residual 29Si isotope
141
fraction of 9.5(10) ppm, and an average 30Si isotope fraction of 5.9(16) ppm, similar
to that of the previous 12C samples. This level of enrichment not only matches
the enrichments of 28Si produced by other sources currently available in the QI
community, such as the IAC [32], but it surpasses them and meets the enrichment
goal stated at the beginning of this chapter. The 28Si enrichment of this sample
also surpasses the best reported enrichment of previous 28Si ion beam deposition by
Tsubouchi et al. [43]. However, the enrichment of some samples was significantly
worse, probably due to a lower geometric mass selectivity achieved for 28Si in these
experiments. It is obvious from these initial depositions that the geometric selec-
tivity of the ion beam needs to be consistently larger in order to, at a minimum,
consistently and predictably match the residual 29Si isotope fractions of 28Si mate-
rial produced by other sources including the IAC. Large variations and uncertainty
in the mass selectivity of the ion beam system is a clear drawback of this experi-
mental configuration. A cross section of a 28Si film was also analyzed using SEM as
a secondary method for measuring the film thickness, and it showed that the film
likely has a different structure than the substrate.
4.3 Achieving Highly Enriched 28Si: Lens
Chamber Samples
4.3.1 Experimental Setup
This section discusses the deposition of 28Si samples in an experimental con-
figuration with the sample located in the lens chamber at LC–2 during deposition.
142
A schematic of the ion beam and lens chambers in this setup is shown in the upper
right section of Fig. 4.1. These experiments expand upon the work of the previous
section by implementing a number of experimental improvements including recon-
necting the deceleration lens chamber to the ion beam chamber. The mass-selecting
aperture used between the ion beam and the lens chamber was the same 1 mm wide
slit discussed in the previous section.
In this configuration, samples were located after the deceleration lenses, and
they were mounted on the end of an electrical feedthrough which acted as a sam-
ple stage for this intermediate experimental setup. A sketch of the sample stage
feedthrough and sample location is shown in the blowup of the schematic of the
LC–2 setup in Fig. 4.1. Samples were mounted using a strip of conductive carbon
tape on the back side of the chip. As was the case with the previous setup, the pur-
pose of the feedthrough was to isolate the sample electrically from the chamber so
that the ion beam current could be monitored during beam tuning and deposition.
Unlike the previous setup, additional electrical vacuum feedthroughs on the sample
stage were used to mount a masking element above the sample location consisting
of a metal shim for collecting current. This was positioned between the sample
and the path of the ion beam and had a small circular aperture directly above the
sample. This mask and aperture was fixed in place for the duration of a deposition.
The fixed sample aperture was approximately 3 mm in diameter and provided a
mechanism to monitor the focusing of the ion beam on the sample by maximizing
the ion current detected on the sample while minimizing the current detected on the
sample mask. Additionally, the mask and sample aperture allowed for precise loca-
143
tion of the ion beam spot on the sample substrate. The feedthrough (and sample)
were positioned on axis with the ion beamline optics, and unlike the previous setup,
IC–1, the sample aperture allowed for the ion beam to be tuned and steered onto
the exact sample location under the aperture. However, a lack of sample motion
in this setup means that the ion beam was constrained to be on axis, which may
not have been the optimal tuning position. A photograph of a sample mounted on
the vacuum feedthrough on this intermediate sample stage with the sample mask
is shown in Fig. B.11 in Appendix B. As was the case with the previous setup, no
method for sample heating exists on this sample stage here. This precludes in situ
sample preparation and limits the available experimental phase space for achieving
epitaxial deposition.
For these experiments, the ion source was operated in the low pressure mode
with a working pressure of SiH4 similar to that which was used for the previous
setup of around 2.7× 10−4 Pa (2.0× 10−6 Torr). An additional turbo pump in the
lens chamber provides differential pumping at the sample location, which results in
lower partial pressures during deposition. Unlike the ion beam chamber, the lens
chamber is rated to UHV, although it was never baked prior to these experiments.
The base pressure of the lens chamber for these samples ranged from approximately
3.6× 10−8 Pa to 1.3× 10−6 Pa (2.7× 10−10 Torr to 1.0× 10−8 Torr) before deposi-
tion, although for the majority of samples, the base pressure was typically around
1.3× 10−7 Pa (1.0× 10−9 Torr).
The decelerating lenses themselves are described in detail in Chapter 2. As was
mentioned previously, the benefit of the lenses is that they provide focusing of the
144
28Si ion beam as it exits the mass-selecting aperture. The lenses help maintain a tight
beam spot as the ions are smoothly decelerated from the transport voltage (-4 kV) to
ground potential at the sample. A more focused beam spot is advantageous because
it corresponds to a higher ion flux, Fi, on the substrate and a higher deposition
rate. This results in a lower relative rate of adsorption of gaseous species from the
vacuum into the 28Si film for a given background pressure.
Another significant experimental change implemented for the experiments of
this section is the replacement and reconfiguration of several power supplies that
controlled the voltages of various electrostatic lens elements in the ion beamline.
These include the “arc” voltage, which defines the potential between the anode and
cathode in the source, the “extractor” element voltage, and the “focus” element volt-
age. The additional control and degrees of freedom provided by this reconfiguration
allowed for better ion beam tuning to maximize ion fluence as well as produce a
more confined beam before mass separation. This improved ion beam tuning yield-
ing a more consistent geometric selectivity similar to and exceeding that which was
achieved with the final sample deposited at IC–1. The circuit diagram of the power
supplies controlling the ion beam lens elements up to the sector mass analyzer is
shown in Fig. A.1 in Appendix A and discussed in Chapter 2.
4.3.2 Sample Preparation
Substrates used for depositing 28Si in this section consisted primarily of “lightly
doped” natural abundance commercial Si(100) wafers that were cleaved by hand
into approximately 1 cm by 1 cm chips. Additionally, several 28Si samples were
145
deposited onto silicon-on-insulator (SOI) chips which had a 40 nm Si “device” layer
on top of 400 nm of buried thermal oxide. The motivation for using these SOI
wafers was to address a potential difficulty in using electron microscopy to measure
the properties of a Si film deposited onto Si. If perfect epitaxy is achieved, then
the 28Si film would become indistinguishable from the substrate except by means
of isotope measurements such as SIMS. By introducing the buried oxide below an
ultra-thin Si surface layer of known thickness, the 28Si film-substrate interface, and
thus the film, would always be easily located in microscopy studies. Before being
loaded into the vacuum chamber, substrates were pre-treated with hydrofluoric acid
(HF) to remove the native oxide and provide a clean surface for deposition. This
treatment was necessary to enable the possibility of epitaxial deposition because in
situ substrate heating was not available to thermally desorb the surface oxide before
deposition. The substrates were only handled with clean teflon tweezers and were
mounted on the end of the sample feedthrough using a strip of carbon tape on the
back of the wafer. No further sample preparation occurred in situ. Typically, after
being loaded into the chamber, samples sat several days in the vacuum chamber
before deposition.
4.3.3 Deposition of 28Si
After tuning and focusing the ion beam according to the procedures described
in Chapter 2, 28Si ions were deposited onto the Si(100) substrates at room tem-
perature (≈ 21 ◦C, see Section 4.2.3). For the deposition of one 28Si sample, a
different sample stage was used that had a tungsten wire back heater for sample
146
heating. This was used to to deposit the 28Si film with a substrate temperature of
approximately 550 ◦C. This experiment, however, did not have a significant impact
on the work presented here and will not be discussed further. During deposition
of the 28Si films at room temperature, the pressure in the chamber rose to between
3.3× 10−6 Pa and 7.2× 10−6 Pa (2.5× 10−8 Torr to 5.4× 10−8 Torr) due to gas
diffusion from the operation of the ion source. These pressures are more than a
factor of 20 improvement over the previous setup. Ions were deposited with a range
of average ion energies at the sample between 50 eV and 170 eV. A lower value
of Ei generally produced a larger geometric selectivity as observed in mass spectra
during the tuning procedures prior to each deposition. For these samples, 28Si ion
beam currents ranged from 200 nA to 800 nA. A mass spectrum for an ion beam
with an average ion energy energy Ei ≈ 122 eV collected through the fixed sample
aperture prior to deposition of a 28Si sample is shown in Fig. 4.8. The corresponding
magnetic sector mass analyzer current used for the field sweep is shown on the top
axis. Ion current peaks (circles) corresponding to 28Si and Si hydrides are observed
from 28 u to 33 u. The 28Si ion peak at 28 u and the ion peak at 29 u, containing ≈
5 % 29Si, show a high degree of separation on this semi-log plot with no detectable
ion current signal occurring between the peaks. Secondary electrons generated by
the ion beam cause the current between the peaks to be ≈ -0.5 nA. Gaussian fits to
the data (line, Eq. (2.14)) are also shown for the 28 u and 29 u current peaks, and
one can see that the data matches the form of a Gaussian very well. This indicates
a symmetric and optimally tuned beam shape with minimal perturbations such as
scattering off of lens elements. The Gaussian fits give a separation of the 28 u peak
147
2 7 2 8 2 9 3 0 3 1 3 2 3 3 3 4
1 0 - 4
1 0 - 3
1 0 - 2
1 0 - 1
1 0 0  D a t a G a u s s i a n  F i t s
2 9 S i  +  2 8 S i H
Ion
 Cu
rren
t (µ
A)
M a s s  ( u )
2 8 S i
1 3 0 2 1 5 - 2 8 S i - S i
4 9 5 0 5 1 5 2 5 3 5 4 5 5M a g n e t  C u r r e n t  ( A )
Figure 4.8: SiH4 mass spectrum representative of the ion beam settings for samples
deposited at LC–2. The ion current (circles) is recorded while sweeping the the
analyzer current, and thus the magnetic field (top axis). The peak at 28 u is 28Si
and the peak at 29 u peak is 28SiH and ≈ 5 % 29Si. Several higher order hydrides
are also observed. Gaussian fits (line, Eq. (2.14)) to the 28 u and 29 u peaks are
superimposed on the data. The centers of the 28 u and 29 u fits are separated by
≈ 11 σ.
from the 29 u peak of approximately 11 σ. The mass resolving power of the ion
beam in this configuration derived from this mass spectrum is m
∆m
≈ 78 (measured
at 10 % of the peak height), which is significantly better than for the mass spectrum
shown in the previous section.
Focusing the 28Si ion beam with the deceleration lenses resulted in a more
compact beam spot size and an average deposition area of about 6 mm2. The beam
spot focusing procedure was not yet optimized for the initial sample deposited here
resulting in a larger deposition area of 16 mm2. The 28Si spots on these samples are
148
Figure 4.9: Optical micrograph of a 28Si sample deposited at room temperature at
LC–2. The Si(100) substrate consisting of a roughly 8.5 mm wide chip is seen with
the 28Si spot measuring about 3.8 mm long.
still visible to the naked eye although they appear different from the samples in the
previous section likely due to the native oxide being stripped prior to deposition.
A visible deposition spot indicates that the 28Si film is structurally different from
the substrate and probably not epitaxial. A representative optical micrograph of
a 28Si sample deposited at LC–2 is shown in Fig. 4.9. The Si(100) substrate is
approximately 8.5 mm wide and the 28Si spot is approximately 3.8 mm long. The
resulting thicknesses of these 28Si films ranged from approximately 50 nm to 350 nm
as inferred from SIMS depth profiles with corresponding deposition rates between
0.51 nm/min and 1.49 nm/min.
To summarize, the typical deposition procedure for samples deposited at LC–2
was as follows:
1. a substrate is cleaved and dipped in HF immediately before being mounted
onto the sample stage feedthrough and loaded into the deceleration lens cham-
ber which is pumped out for > 12 h,
2. after turning on the ion source, the gate valve to the ion beam chamber is
149
opened and the ion beam is tuned and characterized on the substrate using
the fixed sample aperture on the feedthrough. The mass spectrum, ion beam
energy, and beam spot focusing are analyzed and recorded,
3. the ion beam is then tuned to the 28 u peak to commence deposition of 28Si
while monitoring the ion current onto the sample as a measure of the deposition
rate,
4. after typically three to five hours of deposition, the gate valve to the ion beam
chamber is closed to end deposition and reduce the lens chamber pressure back
to its base, and
5. the sample is then removed from the vacuum chamber for ex situ analyses by
venting the lens chamber.
In total, 16 28Si samples were deposited at LC–2 under these conditions.
4.3.4 Enrichment Measurements via SIMS for LC–2
Samples
SIMS was used to assess the enrichment of several samples deposited at LC–
2 to determine if the experimental improvements including the higher degree of
control over the ion beam, which were discussed at the beginning of this section,
yielded more consistent and lower overall 29Si and 30Si isotope fractions. Of the
16 samples produced at LC–2, three are represented on the enrichment progression
timeline in Fig. 4.2, and will be discussed in this section. A SIMS depth profile of the
highest enrichment measured for the first 28Si sample deposited at LC–2 is shown
in Fig. 4.10. Measurements of the 28Si (circles), 29Si (squares), and 30Si (triangles)
isotope fractions are shown vs. the sputter depth into the sample. At a depth of
around 80 nm, the isotope fractions begin to increase to their natural abundance
150
Figure 4.10: SIMS depth profile of a 28Si sample deposited at room temperature
at LC–2. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope fractions are
shown vs. sputter depth. The average 29Si isotope fraction is 2.02(32) ppm, and
the average 30Si isotope fraction is 1.41(20) ppm. The natural abundance values
for each isotope (dotted and dashed lines) are also shown for reference. At a depth
of 80 nm, the isotope fractions increase to their natural abundance values in the
transition into the Si(100) substrate (shaded region). The film-substrate interface
is estimated to be at a depth of 105 nm (not shown).
values (dotted and dashed lines) in the transition into the Si(100) substrate, marked
by the shaded region. This relatively gradual increase in isotope fraction over 25 nm
is partially an artifact of the SIMS measurement, which was described in the previous
section. It is also probably due to beam tuning, including mass spectrum sweeps,
which would deposit all three Si isotopes on the substrate before 28Si deposition.
The film thickness is determined from the location of the interface between the film
and substrate, which is estimated to be at a depth of 105 nm (not shown).
The average measured 28Si isotope fraction in the 28Si film is 99.999657(38) %.
151
The average residual 29Si isotope fraction is 2.02(32)× 10−6 (2.02(32) ppm), and
the average 30Si isotope fraction is 1.41(20)× 10−6 (1.41(20) ppm). These mea-
surements indicate an average isotope reduction factor for 29Si and 30Si of nearly
2.3(2)× 104. A second area on this sample was also measured by SIMS giving an
average 29Si isotope fraction of these two measurements of 2.16(21) ppm. This level
of 29Si isotope fraction is nearly a factor of five lower than for sample with the
highest enrichment achieved at IC–1, as seen in the jump from IC–1 to LC–2 in the
enrichment progression timeline in Fig. 4.2. The similar reduction factors for 29Si
and 30Si indicate that the upgraded ion beam tuning control introduced in this sec-
tion enables more consistent geometric selectivites similar to and surpassing those
achieved for the final sample deposited at IC–1 in the last section. This sample
was deposited with a background pressure 20 times lower than previous samples,
and the resulting enrichment appears to support the conjecture discussed previously
that the enrichment is partially a consequence of the adsorption of SiH4 from the
background gas. Despite this improvement in enrichment, this initial sample had
a relatively large deposition spot size as was mentioned previously. This resulted
in the lowest deposition rate of any of the samples discussed in this section, which
may limit the enrichment that was achieved. This is because a lower 28Si deposition
rate will result in a higher relative adsorption rate of gaseous species for a given
background pressure during deposition. As was mentioned previously in Chapter
2 and in this chapter, part of this background pressure is natural abundance SiH4,
which, if adsorbed into the film, will result in a lower 28Si enrichment in the sample.
After depositing this initial sample, another 28Si sample was deposited under
152
similar conditions except with a more focused beam spot resulting in a smaller
deposition spot size of about 7 mm2. The SIMS measurement of the most highly
enriched region of this second sample is shown as a depth profile in Fig. 4.11. 28Si
(circles), 29Si (squares), and 30Si (triangles) isotope fractions are shown vs. sputter
depth. At a depth of around 340 nm, the isotope fractions begin to increase due
to ion beam tuning and mass spectrum sweeps prior to deposition of 28Si. Beyond
this, the isotope fractions continue to rise to their natural abundance values (dotted
and dashed lines) in the transition into the substrate, marked by the shaded region.
The film-substrate interface is estimated to be at a depth of 384 nm, giving a value
for the 28Si film thickness. The data appears to reside in two main bands because of
discrete counting fluctuations in the measurement, as mentioned for previous SIMS
measurements.
The average measured 28Si isotope fraction in the 28Si film for this sample is
99.9998308(82) %. The average 29Si isotope fraction is about a factor of two lower
than the previous sample at a value of 0.993(64)× 10−6 (0.993(64) ppm). The
average 30Si isotope fraction is 0.699(51)× 10−6 (0.699(51) ppm). These measure-
ments indicate an average isotope reduction factor for 29Si and 30Si of 4.6(2)× 104.
Achieving this very high enrichment potentially enables a robust measurement of
the dependance of electron coherence times on 29Si concentration in the single spin
regime, as discussed at the beginning of this chapter. Four other areas of the depo-
sition spot of this sample were also measured by SIMS giving an average 29Si isotope
fraction for the five measurements of this sample of 1.388(38) ppm.
This result again indicates that the measured enrichment could be due to the
153
Figure 4.11: SIMS depth profile of a 28Si sample deposited at room temperature at
LC–2. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope fractions are shown
vs. sputter depth. The average 29Si isotope fraction in the film is 0.993 ppm, and
the average 30Si isotope fraction is 0.699 ppm. The natural abundance values for
each isotope (dotted and dashed lines) are also shown for reference. At 340 nm, the
isotope fractions begin to increase to their natural abundance values in the transition
to the substrate (shaded region). The film-substrate interface is estimated to be at
a depth of 384 nm.
presence of a SiH4 background gas at the sample. The smaller spot size of this
sample led to the highest deposition rate for any of the samples discussed in this
section, and one that is almost three times higher than the previously discussed
sample. The higher deposition rate of this second sample results in a shorter time
for gaseous species to adsorb into the film for a given volume of material being
deposited and thus lower 29Si and 30Si residual isotope fractions.
This effect is evident not just when comparing the previous sample to this
one, but also within the five SIMS measurements of this sample. Due to the 28Si
154
ion beam flux being nonuniform throughout the beam spot with a roughly Gaussian
spatial distribution, the five areas on the deposition spot were measured to have
different thicknesses and thus different corresponding deposition rates. The SIMS
measurements of 29Si from this sample are shown vs. the deposition rate in Fig. 4.12.
As expected, the residual 29Si isotope fraction (squares) varies inversely with the
deposition rate for these five measurements. These deposition rates are derived from
the film thickness at each spot (top axis), which are inferred from the SIMS depth
profiles. The uncertainty of the deposition rates are determined from uncertainty
in the SIMS depth scales and deposition times. This shows that improvements to
either the background pressure during deposition or the deposition rate can improve
the enrichment of these 28Si samples.
Having establishing the experimental procedures and processes to produce this
last 28Si film with a residual 29Si isotope fraction of 0.993 ppm, several other 28Si
samples were deposited with isotope fractions of 29Si and 30Si consistently at or
below 1 ppm. One of these samples was deposited on a SOI substrate and was
measured by SIMS to have an average 28Si isotope fraction of 99.999863(16) %. The
average residual 29Si isotope fraction is 0.77(11)× 10−6 (0.77(11) ppm), and the
average 30Si isotope fraction is 0.60(11)× 10−6 (0.60(11) ppm). The most highly
enriched of these samples deposited at LC–2 has an average measured 28Si isotope
fraction of 99.999888(10) %. The average residual 29Si isotope fraction in this sam-
ple is 0.691(74)× 10−6 (0.691(74) ppm), and the average 30Si isotope fraction is
0.432(67)× 10−6 (0.432(67) ppm). The SIMS depth profile of this sample is shown
in Fig. 4.13. The isotope fraction values were determined by averaging the data be-
155
Figure 4.12: 29Si isotope fractions vs. the deposition rate for multiple SIMS mea-
surements made on a sample deposited at room temperature at LC–2. Five different
areas across the deposition spot experienced different deposition rates due to an in-
homogeneous ion beam flux. These rates were derived from the thickness at each
spot (top axis) inferred from the SIMS depth profiles. The 29Si isotope fractions
(squares) vary inversely with the deposition rate across the sample.
tween 25 nm and 100 nm. The data from the rest of the 28Si film was not included
because during the first half of the deposition, the pressure at LC–2 was being varied
as part of a separate experiment resulting in slightly increased 29Si and 30Si isotope
fractions due to SiH4 adsorption. The average isotope reduction factor for
29Si and
30Si in this sample is 7.0(7)× 104. At a depth of around 225 nm, the isotope frac-
tions begin to increase to the natural abundance values in the substrate (dotted and
dashed lines). The interface between the film and the substrate is estimated to be
at a depth of 249 nm and is marked by the shaded region.
Overall, SIMS measurements of samples deposited at LC–2 show that a reduc-
tion of nearly a factor of 14 in the 29Si isotope fraction of the most highly enriched
156
Figure 4.13: SIMS depth profile of the most highly enriched 28Si sample deposited
at room temperature at LC–2. 28Si (circles), 29Si (squares), and 30Si (triangles)
isotope fractions are shown vs. sputter depth. The average 29Si isotope fraction
in the film is 0.691 ppm, and the average 30Si isotope fraction is 0.432 ppm. The
natural abundance values for each isotope (dotted and dashed lines) are also shown
for reference. At a depth of 225 nm, the isotope fractions begin to increase to their
natural abundance values in the substrate (shaded region). The interface between
the film and the substrate is estimated to be at a depth of 249 nm.
sample was achieved compared to that of the previous most highly enriched sam-
ple deposited at IC–1, which can be seen in the enrichment progression timeline in
Fig. 4.2. Additionally, the 28Si enrichment values of these samples are consistently
high unlike those of the samples deposited at IC–1, which varied a great deal.
4.3.5 Crystallinity
Substrates were stripped of their native oxide ex situ in the experiments de-
scribed in this section to facilitate the possibility of polycrystalline or epitaxial
157
Figure 4.14: HR-TEM cross-sectional micrograph of a 28Si film deposited at room
temperature at LC–2. The Si substrate is seen in the lower right area of the image
in (a), a layer of glue (light) and Pd cap (dark) are seen in the upper left, and the
deposited 28Si film resides between them. (b) is a FFT of the 28Si region from box
(b) and shows that it is amorphous. By comparison, (c) is a FFT of the Si(100)
substrate from box (c) and it clearly shows a crystalline pattern. This image was
taken on the 〈001〉 zone axis.
deposition. However, depositing at room temperature severely limits Si epitaxy
to very thin layers if it occurs at all [55, 87], and Si solid phase epitaxy (SPE) is
expected to be negligible [88] making the formation of crystalline grains unlikely.
The suspected amorphous nature of these 28Si films is verified using TEM. A cross-
158
sectional TEM micrograph of a 28Si sample deposited at LC–2 is shown in Fig. 4.14.
This microscopy was done in collaboration with Dr. June Lau (NIST), and the
TEM specimen was prepared by mechanical polishing. Panel (a) is a high resolu-
tion (HR-TEM) image showing the deposited 28Si film between the substrate and
a Pd capping layer. The Si(100) substrate is the region in the lower right of the
micrograph and to the left of that is the 28Si film. In this area of the film, the 28Si
layer thickness is about 20 nm and the average thickness over all the areas surveyed
was 37 nm. Just to the left of the 28Si layer is a thin dark Pd capping layer deposited
to protect the film during the TEM specimen thinning process by Ar milling. In the
upper left of the micrograph is glue, also from specimen preparation. The sample
was tilted to the 〈001〉 silicon zone axis in this micrograph. One can see that the 28Si
film is amorphous because the lattice rows of the substrate do not continue into the
film region. This is made more clear by the fast fourier transform (FFT) analysis
of the film and substrate. Panel(b) in Fig. 4.14 is a FFT of the 28Si region (box
(b) in panel (a)) and corresponds to an amorphous pattern. In contrast, panel (c),
which is the FFT of the Si(100) substrate (box (c) in panel (a)), corresponds to a
crystalline pattern.
4.3.6 Chemical Purity
4.3.6.1 SIMS
To begin to assess the chemical purity of the 28Si samples in the context of the
materials goals laid out at the beginning of this chapter, SIMS was used to detect
159
C contamination in two samples deposited at LC–2. 12C and 13C were monitored in
the 28Si samples both as a marker for general chemical contaminants in the films,
but also because 13C possesses nuclear spin (I = 1/2), which will cause decoherence
of qubit spins in a quantum computing device in the same way that 29Si does. A
SIMS depth profile of the C atomic concentrations in a 28Si sample is shown in
Fig. 4.15. 12C (solid line) and 13C (circles and line) are plotted vs. sputter depth
into the sample. A profile of 30Si (triangles and line) measured at the same time
is shown as a reference to the boundary between the deposited 28Si film and the
natural Si(100) substrate (shaded region). The film-substrate interface is estimated
to be at a depth of around 425 nm. It is difficult to precisely determine the atomic
concentration of carbon in the silicon film because the carbon background due to
the SIMS instrument is not precisely known under the measurement conditions,
however it must be a factor of 10 lower than in the film because the C concentration
drops by that much in the substrate. For simplicity, C was monitored using the
same measurement conditions as for Si detection, and so the measurement was not
optimized for accurate C measurements. The slow roll off observed for the carbon
signals moving into the substrate is probably a measurement artifact. After apply-
ing a value obtained from literature of 0.007 for the relative sensitivity factor of
carbon to silicon for SIMS under similar analytical conditions [89], the average mea-
sured atomic fraction of 12C (measuring only Si and C isotopes) in the film is found
to be approximately 3.39(8) %. Considering the nominal atomic concentration for
amorphous Si (≈ 4.9× 1022) [90], the measured atomic fraction of 12C represents
an atomic concentration of 1.66(4)× 1021 cm−3. The average measured 13C atomic
160
Figure 4.15: SIMS depth profile showing the atomic concentration of 12C (solid
line), 13C (circles and line), and 30Si (triangles and line) in a 28Si sample deposited
at room temperature at LC–2. The sharp rise in 30Si concentration marks the
boundary between the deposited 28Si film and the natural Si(100) substrate (shaded
region). The slow roll off of the carbon profiles into the substrate is an artifact of
the SIMS measurement.
fraction is 0.031(1) % (310(10) ppm), which is about 110 times smaller than the
12C fraction and close to the natural abundance fraction of about 1.1 %. For refer-
ence, the 30Si atomic concentration in this particular measurement is approximately
1.6 ppm in the film.
The concentration of spins in the 28Si film due to 13C is over 300 times larger
than the residual 29Si spin concentration in this sample. The source of this C con-
tamination is probably predominately gaseous carbon containing compounds in the
vacuum such as CO and CO2, which can adsorb into the
28Si film during deposition.
A buildup of gaseous C compounds is also likely the source of the small spike in
161
C concentration which can be seen at the interface with the substrate because the
sample sat approximately one day in the chamber before deposition. In this ex-
perimental configuration there is no residual gas analyzer available to measure the
relative abundance of such compounds. It is also possible that some C adsorbs into
the amorphous film from the ambient environment after the sample is removed from
vacuum.
To examine the dependance of the C contamination in the 28Si film on back-
ground pressure, a sample was deposited using the nominal procedure described
above except that during deposition, the pressure in the chamber was raised for
about one third of the deposition time. This was achieved by closing the gate valve
to the turbo pump in the lens chamber. This caused the pressure in the chamber
to rise almost two orders of magnitude, which was partially due to SiH4 gas from
the ion source, but also probably due to an increase in partial pressures of typical
residual gasses found in high vacuum systems such as H2, CO, N, and CO2.
The C contamination of this sample was measured by SIMS in the same man-
ner as the last sample. A SIMS depth profile of the C concentration in this sample
is shown in Fig. 4.16. 12C (solid line) and 13C (circles and line) are plotted vs.
sputter depth into the sample. A profile for 30Si (triangles and line), which was
measured separately, is shown as a reference to the boundary between the deposited
28Si film and the natural Si(100) substrate. The film-substrate interface, marked
by the shaded region, is estimated to be at a depth of 240 nm. Also plotted is the
ratio of 28SiH/28Si (dotted line), which is typically monitored during SIMS mea-
surements and corresponds to the right axis in the figure. The 28SiH signal gives
162
Figure 4.16: SIMS depth profile showing the atomic concentration of 12C (solid line),
13C (circles and line), and 30Si (triangles and line) in a 28Si sample deposited at room
temperature at LC–2. The sharp rise in 30Si concentration marks the boundary
between the deposited 28Si film and the natural Si(100) substrate (shaded region).
Also shown is the measured SIMS ratio of 28SiH/28Si (dotted line) corresponding
to the right hand axis. The chamber pressure was increased almost two orders of
magnitude during the portion of the deposition bounded by the vertical dashed lines
between approximately 80 nm and 190 nm. The 12C, 13C, and 28SiH concentrations
increase in this region.
a qualitative idea of relative amounts of H in the different layers of the film. The
region of the film corresponding to the higher pressure during deposition is approx-
imately between 80 nm and 190 nm, bounded by the vertical dashed lines. In this
portion of the film, one can see that the C concentrations increase more than an
order of magnitude and the 28SiH signal increases slightly less than that. A spike in
C concentration is observed at the interface between the film and the substrate due
to carbon containing adsorbates accumulating when the sample sat for about three
days in vacuum before deposition.
163
In the low pressure region of the film, i.e. the top layer from 0 nm to 80 nm,
the average 12C atomic concentration is determined to be 2.67(3)× 1020 cm3, which
is 0.545(6) %, and the average 13C atomic concentration is 42(3) ppm. These val-
ues are around six times lower for this sample than for the previous one, prob-
ably due in part to the deposition pressure being slightly lower for this second
sample. In the high pressure region of the film, the 12C atomic concentration in-
creases to 3.2(2)× 1021 cm3 (6.6(4) %). The 13C atomic concentration increases to
610(40) ppm. For reference, the 30Si atomic concentration throughout the enriched
film is approximately 0.9 ppm. A lower bound on the total achievable Si chemical
purity for samples deposited at LC–2 can be determined from the low pressure re-
gion of this second sample to be approximately 99.451(6) %. This sample shows
that adsorption and incorporation of chemical contaminants from the vacuum back-
ground pressure into the depositing 28Si film is a significant issue in this experimental
configuration.
4.3.6.2 XPS
In addition to the chemical purity analysis provided by SIMS for the C con-
centrations in 28Si films, XPS was used to search for a broader range of chemical
contaminants. XPS spectra were acquired and analyzed in collaboration with Dr.
Kristen Steffens (NIST). The 28Si sample used for XPS analysis was deposited at
LC–2 under similar conditions as the previous samples described in this chapter.
Additionally, a control chip accompanied the sample through the deposition pro-
164
cess but was not irradiated by the ion beam. XPS spectra were collected after
sputter-cleaning the samples with Ar to remove surface contamination from the en-
vironment. Both low resolution survey scans (pass energy 160 eV, step size 0.5 eV)
and high resolution region scans (20 eV pass energy, step size 0.1 eV) for C 1s, N 1s,
Si 2p, and O 1s were performed on a Kratos Axis-Ultra DLD Photoelectron Spec-
trometer with a monochromated Al Kα x-ray source (1486.6 eV). Multiple spots
were measured on each sample to ensure consistency. Peak positions were calibrated
to the Si 2p1/2 peak at 99.3 eV.
Initially, prior to sputter cleaning, the 28Si film showed C, N and O peaks in
the XPS spectrum. However, after sputter cleaning, the N and the adventitious C
disappeared, but a C 1s peak due to SiC persisted. This 28Si XPS spectrum is shown
in Fig. 4.17 (upper spectrum) as a plot of count rate vs. electron binding energy.
The data for the 28Si sample was shifted up for clarity. The C 1s peak indicates a
relatively constant atomic fraction of approximately 3 %, consistent with the first
SIMS result. After sputtering, two small O 1s peaks also remain at 531.4 eV and
532.5 eV, corresponding to an atomic fraction of approximately 4 %. Also shown
are references for relevant elemental orbital level positions. In the Si 2p region, a
SiO2 peak from the native oxide was no longer present after sputtering, however,
in addition to the elemental Si peaks, a small shoulder attributed to SiC is seen
in finer scans. The control Si sample (lower spectrum) shows reduced C and O
peaks corresponding to atomic fractions of approximately 1 % to 2 % each. These
values are taken as upper limits which can give an indication of the instrumental
background because the actual C and O content of the wafer is expected to be
165
Figure 4.17: XPS spectra of a 28Si sample deposited at room temperature at LC–2
and a control Si sample. Count rates vs. electron binding energy for survey scans
are shown for the 28Si sample after Ar sputter cleaning (upper spectrum) and the
control chip (lower spectrum). The data for the 28Si sample was shifted up for
clarity. References for elemental orbital level positions are included above relevant
peaks. O 1s and C 1s peaks are visible with larger amplitudes in the 28Si film than
the control sample.
much lower. Both the 28Si and control scans show Ar peaks due to the Ar sputter
cleaning process. This XPS analysis agrees with the SIMS result that at least some
samples deposited at LC–2 can have as much as 3 % C throughout the 28Si film,
and in addition they can also contain an approximately equal amount of O. The Si
chemical purity for this sample measured by XPS is roughly 95 %. Although the
previous SIMS measurement showed that a sample can have lower level of C, it is
not known if the O contamination is similarly reduced in that sample, and so it is
difficult to place a bound on the total Si purity that is achievable for these samples.
166
4.4 Chapter 4 Summary
The 28Si samples deposited at IC–1 served as a successful proof of principle
for the adaptation of the in situ enrichment and ion beam deposition method from
12C deposition to 28Si. Experimental procedures for deposition using a natural
abundance SiH4 source gas and the characterization of the enrichment via SIMS
were established. The samples deposited at LC–2 showed that 28Si films could be
deposited with residual 29Si and 30Si isotope fractions consistently below 1 ppm. A
total of five 28Si samples were produced at IC–1 and 16 were produced at LC–2. The
overall improvement in 28Si enrichment achieved from the initial samples deposited
at IC–1 to the samples deposited at LC–2 can be seen in the enrichment progression
timeline in Fig. 4.2. The residual 29Si isotope fractions of these samples was reduced
from 2822(18) ppm to 0.691(74) ppm. This most highly enriched sample deposited
at LC–2 had a 28Si isotope fraction of 99.999888(10) %. The achieved reduction
in 29Si and 30Si isotope fractions in samples deposited at LC–2 was likely due to
several factors. First, better control of the ion beam tuning lead to consistently
higher geometric selectivities. Additionally, the lens chamber has significantly lower
background pressures resulting in less SiH4 adsorption during deposition. Finally,
the deceleration lenses enabled focusing of the ion beam spot resulting in smaller
deposition spots, higher growth rates, and less SiH4 adsorption. The enrichment
values of these samples meet the enrichment materials goal laid out at the beginning
of this chapter to achieve 28Si enrichments that surpass those of any other known
sources of 28Si including the IAC, and they do it consistently. Additionally, they
167
attain enrichments sufficient to enable a robust measurement of the dependance of
electron coherence time on 29Si concentration in the single spin regime and compare
it to theoretical predictions (see Fig. 1.9), as proposed in Chapter 1.
However, initial observations of 28Si film crystallinity and chemical purity show
that these samples deposited at IC–1 and LC–2 do not meet the second and third
materials goals stated at the beginning of this chapter of being crystalline and highly
chemically pure. SEM and TEM cross-sectional micrographs of two samples show
that the films are unsurprisingly amorphous, probably due to deposition occurring
at room temperature. Chemical analysis of several samples deposited at LC–2 by
both SIMS and XPS show that both C and O are present in these films at rela-
tively high atomic concentrations up to approximately 3 %. Achieving both single
crystal epitaxial deposition and eliminating chemical contaminants require further
experimental improvements, and will be the subject of Chapter 5.
168
Chapter 5
28Si Thin Film Deposition and
Characterization Phase II: Crystallinity
and Chemical Purity
5.1 Introduction
Chapter 4 demonstrated that extremely high levels of 28Si enrichment (< 1
ppm 29Si) were achievable for thin film samples produced via ion beam deposition
at LC–2. The most highly enriched sample deposited at LC–2 had a 28Si isotope
fraction of 99.999888(10) % and a residual 29Si isotope fraction of 0.691(74) ppm.
However, these samples were found to be amorphous, likely due to them being
deposited at room temperature, and they were measured to contain C and O in
atomic concentrations up to approximately 3 %. In order for the 28Si to be viable
for use in Si based solid state quantum computing, it must be of very high quality
and meet the three materials goals of this work discussed in Chapters 1 and 4. The
experiments in Chapter 4 produced samples which surpassed the first materials goal
of achieving enrichments with residual 29Si isotopic concentrations below 50 ppm
169
to match other sources of 28Si including the International Avogadro Coordination.
The residual 29Si isotope fractions of these samples was also potentially low enough
to facilitate a robust measurement of the dependence of electron coherence times
on 29Si concentration in the single spin regime, also discussed in Chapters 1 and 4.
Assessments of samples produced in the experiments discussed in Chapter 4 were
done in the context of the second and third materials goals of single-crystalline,
epitaxial thin films with dislocation densities below 1 × 106 cm−3 and chemical
impurity concentrations below 2× 1015 cm−3, however these goals were not pursued
further in those experiments.
Discussion of the ion beam deposition of 28Si films enriched in situ and the
characterization of their properties including enrichment, crystallinity, and chem-
ical purity is continued in this chapter. The experiments discussed here seek to
maintain and improve upon the already high level of achieved 28Si enrichment while
depositing samples in the final experimental configuration for this system used in
this work, which enables epitaxial deposition. This continued improvement in en-
richment and reduction in residual 29Si and 30Si isotope fractions is illustrated in
Fig. 4.2, the enrichment progression timeline. As previously described in Chapter 1,
the measured isotope fractions of Si are written as zSi/Sitot., representing the average
detected counts of an isotope divided by the total average counts of the SIMS mea-
surement. It shows the 29Si isotope fraction reduction at LC–2 from 0.691(74) ppm
down to a minimum at DC–3 of 127(29) ppb with an overall enrichment in 28Si of
99.9999819(35) % for the most highly enriched sample produced in this work. SIMS
measurements of the enrichment of 28Si presented in this chapter will be discussed in
170
terms of the progression of samples at DC–3 in Fig. 4.2 and the experimental factors
that resulted in those higher enrichments. The experiments discussed here also seek
to leverage the improved capabilities of this new experimental setup to facilitate
high-quality epitaxial deposition as well as study chemical contaminants to improve
the purity of the 28Si films. Chemical impurities present in 28Si films can act as
or induce scattering sites and charge traps [91, 92]. This reduces electron mobility
and other electronic properties that are important for the successful operation of QI
devices. Additionally, chemical impurities can possess nuclear spin, which will cause
decoherence of qubit spins in a quantum computing device in a similar manner as
the nuclear spin of 29Si in natural abundance Si. A total of 40 28Si samples were
produced at DC–3.
The experimental configuration used for depositing samples in this chapter is
the third setup discussed in Chapter 4 where samples are located at DC–3 in the
deposition chamber, as shown in the deposition chamber schematic in Fig. 4.1. This
setup had several advantages over the previous two including a further reduction in
the background pressure during deposition due to additional differential pumping.
In addition to a third turbo pump and an ion pump present in the deposition
chamber, an ion pump was added to the lens chamber for this setup, which are all
marked in Fig. 4.1. Lower partial pressures of gaseous species containing C and O
should result in higher chemical purities of the film. Depositing samples at DC–3
crucially enabled heating of the Si substrate for in situ preparation and heating
of the sample during 28Si deposition to facilitate epitaxial growth. This setup also
enables the use of the analytic instruments in the deposition chamber in conjunction
171
with 28Si deposition including RHEED, RGA, and STM, as discussed in Chapter 2.
These features will be discussed further in the next section.
Epitaxial deposition of 28Si thin films with a high degree of crystallinity, i.e.
a low defect density similar to electronics grade Si, is a key achievement towards
producing material suitable for use with solid state quantum computing devices. The
crystallinity of a film and the roughness of its surface depend on the characteristics
of the deposition and growth. Si MBE is ideally categorized as either Volmer-Weber
growth or Frank-van der Merwe growth [93]. Volmer-Weber growth is characterized
by the formation of 3D islands consisting of multiple layers growing at once as
islands coalesce into a continuous film. This type of growth is the result of the
ratio of the depositing atomic flux to the surface diffusivity being large such that
it is more likely that adatoms will cluster with each other and form islands before
they can diffuse to and over step edges [94]. These islands result in the buildup of
roughness on the surface. Conversely, Frank-van der Merwe growth is characterized
by smooth layer-by-layer growth where single layer 2D islands merge to ideally form
a continuous first monolayer before the second layer begins to form [95]. This type
of growth is the result of the ratio of atomic flux to surface diffusivity being small
such that adatoms are more likely to reach step edges as well as diffuse down to
lower steps before clustering to form second layer islands. With the presence of
atomic steps on the substrate, this growth mode can also lead to so-called step
flow growth where no islands form and all adatoms can diffuse to step edges before
any other interactions. In reality, most thin film deposition occurs between these
two ideal cases. In general, deposition dominated by 3D island growth can lead to
172
rougher films with more crystalline defects than deposition dominated by smooth
layer-by-layer growth.
Crystalline defects such as vacancies, dislocations, and stacking faults can de-
grade the performance of semiconductor electronic devices because they can act
as charge traps and scattering sites which reduces electron mobility. Additionally,
these types of defects have been theoretically predicted and experimentally shown
using electric dipole spin resonance (EDSR), EPR, and deep-level transient spec-
troscopy (DLTS) to introduce electric defect states in the band gap of Si [96–99].
Si divacancies can exist in multiple charge states and have multiple deep electronic
levels in the band gap, and stacking faults in a Si crystal lead to a defect state in
the band gap which is approximately 100 meV above the valence band [100].
Crystalline defects in 28Si are potentially more detrimental to quantum co-
herent devices because they introduce scattering centers and local time varying
electric and magnetic fields in the crystal that can contribute to the decoherence
of a quantum system nearby. Crystalline defects also introduce local strain fields
which are known to cause the appearance of unintentional quantum dots in Si wires
due to strain induced conduction band modulation [101]. Local strain fields around
a qubit, such as a 31P donor in Si, can also cause internal electric fields which have
been shown to Stark shift the donor’s electron energy levels and make the qubit spin
more sensitive to electric field noise [102, 103]. The nature and magnitude of the
effect of local crystalline defects on the performance of quantum coherent devices is
still an open area of research and a question that will need to be addressed as this
field progresses.
173
In any device comprised of multiple layers, smooth interfaces can be important
in addition to a low level of structural defects within the bulk of the layer. At a basic
level, fabricating additional layers or electric gates on top of a device layer that has
a rough surface can be challenging and is generally undesirable. Dangling bonds and
other defects present at a rough surface may result in an increased density of interface
charge traps typically seen at oxide interfaces. Surface or interface roughness has
also been predicted to affect the valley states of electrons in Si quantum dots and
Si/SiGe quantum wells by causing mixing of valley, spin, and orbital states as well
as random fluctuations in the phase of the valley-orbit coupling [104–108]. These
effects would vary across devices and be impossible to predict, making operation of
quantum devices utilizing the valley degree of freedom more difficult. The overall
smoothness or roughness of a deposited film can also be a general indicator of their
epitaxial quality or crystallinity, and so it is presented here as an important aspect
of the 28Si films discussed in this chapter.
Section 5.2 of this chapter discusses the specifics of the experimental setup for
samples produced at DC–3 in the final experimental configuration. Sections 5.3 and
5.4 go over the experimental methods for in situ and ex situ sample preparation
as well as the deposition conditions used for 28Si samples at DC–3. Section 5.5
discusses the results of SIMS measurements of the enrichment of significant 28Si
samples deposited at DC–3. Sections 5.6 explores experiments to deposit films
epitaxially at elevated temperatures and characterize their morphology by RHEED,
STM and SEM. Section 5.7 discusses measurements of the chemical purity of these
samples as measured by SIMS and XPS. Section 5.8 discusses the crystallinity of
174
deposited 28Si films observed by TEM. Finally, Section 5.9 summarizes these results.
5.2 Experimental Setup for Improving
Crystallinity and Chemical Purity:
Deposition Chamber Samples
This chapter discusses the deposition of 28Si samples in the final experimental
configuration for this ion beam deposition system as it was designed as a whole
and used in previous work [59]. In this experimental setup, samples are located at
DC–3 as shown in Fig. 4.1. A number of experimental improvements and use of
new analytic capabilities are enabled by connecting the ion beamline (including the
deceleration lens chamber) to the deposition chamber.
In this configuration, samples were located after the deceleration lenses, but
in the deposition chamber. As described in Chapter 2, the lenses protrude into the
deposition chamber and stop approximately 1 cm before the sample location. As
a consequence, the lens chamber and the deposition chamber are actually a single
connected vacuum environment. The ion beam chamber is separated from these
chambers just before the deceleration lenses by a gate valve. The mass-selecting
aperture used between the ion beam and the lens chamber here had a slightly differ-
ent geometry than the one used for the depositions in Chapter 4. The mass-selecting
aperture slit width was increased to 2 mm to allow a larger 28Si ion fluence to pass
into the deceleration lens section. Previously, a thinner slit width was chosen to
limit the range of mass values that can pass through from the beam thus increasing
175
the mass resolving power. However, the experiments at LC–2 in Chapter 4 demon-
strated better beam tuning and very high geometric selectivity as represented by
both SIMS measurements and mass spectra. Higher geometric selectivities allow for
the aperture slit width to be increased without impacting the realized selectivity.
The height of the slit was decreased from 15.25 mm to 12 mm. This was done to
decrease the conductance of gas from the higher pressure ion beam chamber into
the deposition chamber while maintaining the fluence of the ion beam, which has a
typical spot size at the aperture of < 10 mm in the slit height direction. The thick-
ness of the aperture was reduced to a knife edge at the opening to reduce potential
scattering of ions as well as sputtering of the aperture as they pass through.
A secondary aperture with dimensions 12.7 mm by 6.4 mm was also installed
at the beginning of the deceleration lenses for this configuration. This functioned
as a gas aperture to block the gas diffusing past the mass-selecting aperture from
entering the deceleration lens column. The only outlet for this gas would be at the
sample, but with the gas aperture in place, it is instead diverted around the lens
column where it may be better pumped away in the lens chamber. A photograph
of this secondary gas aperture is shown in Fig. B.8 in Appendix B.
Samples were mounted onto sample holders which were then introduced into
the vacuum chamber via the load lock and placed onto the 5-axis manipulator, all
of which are described in more detail in Chapter 2. The manipulator and sample
are positioned to face the deceleration lenses prior to deposition of 28Si. Similar
to the sample mounting setup described of Chapter 4 using an electrical vacuum
feedthrough at LC–2, samples discussed in this section are electrically isolated from
176
the chamber as they sit in the sample holder on the manipulator. This isolation
allows for the ion beam current to be monitored during deposition while controlling
the sample potential. In the previous experimental setup with samples at LC–2, the
fixed sample aperture that sat over the sample allowed for current collection and
focusing of the ion beam through the sample aperture (see Section 4.3.1). Here, an
interchangeable sample aperture, which can be inserted into the sample position and
then removed prior to deposition, provides similar capabilities. One advantage of
the interchangeable sample aperture is that the sample can be completely removed
from the path of the ion beam while beam tuning, focusing, and mass spectrum
sweeps are performed on the aperture, thus keeping the sample pristine prior to
deposition. A sample aperture which was 2.2 mm in diameter was used for most
samples, although other diameters including 3 mm, 2.5 mm, and 1 mm were also
used. Additionally, the 5-axis manipulator allows for sample motion in three spacial
dimensions and precise positioning in the vacuum chamber relative to the ion beam
optics. The advantage of this is that, unlike in the setups of Chapter 4, the ion
beam can be optimally tuned and the sample repositioned so that the beam spot is
located at any desired location on the sample, typically at the center. A photograph
of a sample mounted in a sample holder and on the manipulator at DC–3 was shown
in Fig. 2.23 in Chapter 2. Photographs of the interchangeable sample apertures can
be found in Fig. B.13 in Appendix B.
A significant experimental improvement enabled by this experimental configu-
ration with samples located at DC–3 is sample heating. As described in Chapter 2,
the 5-axis manipulator provides two methods of sample heating. First is a tungsten
177
wire back heater that sits behind the sample holder and can radiatively heat it up
to around 900 ◦C, although with a relatively slow response time of several minutes.
This is the RH method of sample heating referred to in Chapter 2. The second
method uses electrical contacts on the manipulator that interface with the sample
holder and provide the ability to pass current directly through the Si substrate,
heating it resistively. This is the DH method of sample heating referred to in Chap-
ter 2, and it can be used to heat the Si chip as high as its melting point with a
very fast response time of less than a second. Sample heating crucially enables in
situ substrate preparation and cleaning, deposition at elevated temperatures, and
post-deposition sample annealing. These abilities allow for control over a critical
experimental degree of freedom for achieving high-quality epitaxial deposition of
28Si [51] and will be discussed later in this chapter.
Connecting the ion beam and lens chambers to the deposition chamber pro-
vides additional differential pumping at the sample location from a turbo pump
and an ion pump, which result in a lower chamber base pressure and a lower back-
ground pressure during deposition. Additionally, for this configuration, an ion pump
was added to the lens chamber. The typical base pressure of the deposition cham-
ber for the experiments described in this chapter was approximately 6.7× 10−9 Pa
(5.0× 10−11 Torr). Heating the sample before deposition usually causes some de-
gassing that results in slightly elevated pressures. Typically, the pressure immedi-
ately before starting deposition was approximately 2.6× 10−8 Pa (2.0× 10−10 Torr),
although it varied almost an order of magnitude from approximately 7.1× 10−9 Pa
to 5.7× 10−8 Pa (5.3× 10−11 Torr to 4.3× 10−10 Torr). Additionally, the depo-
178
sition chamber and lens chamber, being of all UHV construction, were baked to
approximately 150 ◦C prior to these experiments to reduce the partial pressures of
water as well as carbon and oxygen containing compounds. The remaining domi-
nant partial pressure in these chambers was H2. The RGA in the deposition chamber
gives insight into the specific atomic and molecular species that make up the base
background and deposition pressures in this chamber. An example of the residual
gas mass spectrum corresponding to the base pressure of the deposition chamber
was shown in Fig. 2.22 in Chapter 2. For the experiments described in this chapter,
the ion source was operated with a working pressure of SiH4 between 1.0× 10−4 Pa
and 3.3× 10−4 Pa (7.5× 10−7 Torr to 2.5× 10−6 Torr), similar to that of samples
deposited at LC–2 in Chapter 4. The most common working pressure used was ap-
proximately 2.0× 10−4 Pa (1.5× 10−6 Torr). In addition to these typical low pres-
sure operating conditions, the high pressure mode of the ion source plasma was also
explored in which the typical operating pressure was approximately 1.3× 10−3 Pa
(1.0× 10−5 Torr). In this experimental setup, another RGA was also installed in the
ion beam chamber to diagnose the base pressure gas components and contaminants
in the SiH4 gas, both of which may diffuse into the deposition chamber. A residual
gas mass spectrum of the base pressure from the ion beam chamber was shown in
Fig. 2.3 in Chapter 2.
Another significant experimental advantage to depositing 28Si samples at DC–
3 vs. at IC–1 or LC–2 is having access to the various analytical and other features
of the deposition and analysis chamber, which are described in Chapter 2. The load
lock allows for quicker loading of multiple samples while maintaining the very high
179
vacuum level of the deposition chamber. Sample heating is critical for epitaxial
deposition and feedback mechanisms in the form of analysis tools in the deposi-
tion chamber are necessary for developing the correct experimental procedures to
achieve epitaxy. RHEED and the STM provide information on the state of the
pre-deposition substrate, the growth mode during deposition, and surface charac-
teristics of the deposited sample. The RGA and AES can provide ambient and
surface chemical information for samples. Finally, deposition parameters for 28Si
can be modified to produce higher quality films by comparing them to the natural
abundance Si films deposited from the electron beam evaporator (i.e. the EFM) in
the same system.
5.3 Sample Preparation
5.3.1 Ex Situ Cleaning
A variety of Si substrates were used for the 28Si samples discussed in this
chapter, but they were all natural abundance, electronic grade, single crystalline
commercial Si(100) wafers. A complete table of substrates used here is given in Ta-
ble C.1 in Appendix C. Si(100) has been shown to facilitate higher quality epitaxial
deposition at lower temperatures than other Si surface orientations such as Si(113),
Si(111) or Si(110) [87,109,110]. N-type, p-type, and undoped (intrinsic) wafers were
used. The p-type Si wafers were all boron-doped with resistivities between 1 Ω · cm
and 20 Ω · cm and thicknesses between 300 µm and 380 µm. These wafers were used
to deposit 22 samples at DC–3 and were obtained from both ITME and University
180
Wafer. Substrates from University Wafer have an unknown history because they
are reclaimed wafers. The n-type wafers were all phosphorous-doped, mostly with
resistivities of between 1 Ω · cm and 10 Ω · cm, and were 300 µm thick, although the
substrate used for one sample originated from 600 µm thick stock. These wafers were
used with eight samples and were obtained from University Wafer. Five samples
were deposited on intrinsic Si wafers in these experiments, and they were obtained
from University Wafer, were 380 µm thick, and had a resistivity of > 20 kΩ · cm.
The 28Si samples that were deposited using the high pressure plasma deposition
mode (a total of five) were deposited onto phosphorous-doped wafers with a resis-
tivity between 7 Ω · cm and 20 Ω · cm. They were 300 µm thick with a minimal
misalignment or miscut angle relative to the (100) plane of ± 0.05◦ and were ob-
tained from Virginia Semiconductor. The wafers from Virginia Semiconductor were
float-zone refined with an atomic concentration of O of < 9× 1017 cm−3 and an
atomic concentration of C of < 5× 1013 cm−3.
Initially, for about one quarter of the samples deposited at DC–3, there was
no ex situ cleaning performed on the substrates, which were loaded into the vacuum
chamber with a native oxide. This oxide was then thermally desorbed in situ in a
process described below. As a comparison to those samples with no cleaning, two
samples were treated with HF to strip off the oxide immediately prior to loading
them into the vacuum chamber. These initial samples were cleaved by hand into
approximately 5 mm by 12 mm chips, which is the approximate maximum sample
size that can be accepted by the sample holders used in these experiments. The
substrates used for the remainder of the samples deposited at DC–3 were cut using
181
a dicing saw into 4 mm by 10 mm chips after depositing a thin photoresist layer to
protect the surface from the accumulation of Si dust during dicing. Individual chips
were then cleaned using a more rigorous cleaning procedure designed for complemen-
tary metal-oxide-semiconductor (CMOS) technology to remove metals and organics
from Si surfaces. This clean consists of a piranha etch, HF strip, and “Standard
Clean 2” (SC–2) [111], and it was adopted to improve the substrate surface clean-
liness and quality after feedback from analysis of earlier samples. The full cleaning
procedure is as follows:
1. photoresist remover (PG Remover) at 70 ◦C for 10 min,
2. fresh PG Remover at room temperature for 5 min,
3. isopropanol (IPA) rinse for 1 min,
4. deionized water rinse, and N2 blow dry,
5. 6:1 H2SO4:H2O2 for 12 min with no deliberate heating (piranha etch),
6. deionized water rinse, and N2 blow dry,
7. 50:1 H2O:HF for 10 s,
8. deionized water rinse and, and N2 blow dry,
9. 5:1:1 H2O:HCl:H2O2 at 80
◦C for 12 min (SC–2),
10. deionized water rinse, and N2 blow dry.
PG Remover is a standard solvent used to remove the protective photoresist from
the substrates here. That is followed by a rinse in IPA and deionized water to remove
residual photoresist and PG Remover residue and prepare the substrate surface for
chemical cleaning. The first clean is a piranha etch designed to remove organics
from the surface through oxidation. This was chosen over the “Standard Clean 1”
182
(SC–1) solution, which also cleans organics, because (SC–1) is known to sometimes
roughen the wafer surface. A water rinse and N2 blow dry is used between each
cleaning step to remove chemical residues. The substrates are next etched with
HF to remove the surface oxide containing contaminants and also remove some
ionic metal contaminants. Finally, SC–2 solution is used to remove remaining ionic
metal contaminants from the surface before a final deionized water rinse. As a
result of the SC–2 clean, the chips are left with a thin protective oxide before being
mounted onto sample holders and loaded into the vacuum chamber via the load
lock. This protective oxide has been shown to reduce carbon contamination on the
Si(100) surface as compared to a substrate which was treated with HF to strip the
oxide prior to loading [112], and that result was confirmed in this work and will
be discussed later in this chapter. Substrates were only handled with clean teflon
tweezers. Typically, after being loaded, samples sat about a week in the vacuum
chamber while being prepared before deposition.
5.3.2 In Situ Preparation
After substrates are introduced into the vacuum chamber via the load lock,
they are prepared in situ using the degassing and flash annealing procedures first
described in Chapter 2. This well known UHV high temperature Si “flashing”
procedure is used to prepare an atomically clean Si(100) (2×1) reconstructed sur-
face [113,114], enabling epitaxial deposition. Before flashing, samples are loaded into
the deposition chamber after sitting in the load lock for approximately one day un-
til the pressure drops to around 1.3× 10−6 Pa (1.0× 10−8 Torr). Sample substrates
183
and the sample holders next need to be degassed slowly and thoroughly before they
can be flashed to high temperature in order to limit the pressure increase during
the flashing procedure. The magnitude of the pressure spike that occurs during
a high temperature flash is a critical parameter for forming a clean Si surface. A
bare Si surface should not be exposed to a background pressure higher than roughly
1.3× 10−7 Pa (1.0× 10−9 Torr), or it will become contaminated and more defec-
tive. This pressure is based on experience from this work and communications with
other labs as well as the literature [114]. Substrates are degassed using both the
RH back heater to heat the sample holder and DH power to heat the chip itself.
Degassing occurs at temperature of approximately 600 ◦C for at least 12 hours, and
the pressure in the deposition chamber is monitored during the temperature ramp
up to ensure that the pressure stays below 1.3× 10−7 Pa. This pressure criteria is
only precautionary for the degassing step if the substrate has a protective oxide.
After degassing, the DH power is used to rapidly flash anneal the sample in a
few seconds up to higher temperatures for a short period of time. Typically, a 1 min
flash to 1050 ◦C is initially used to desorb the thin oxide layer (if one is present)
that was either a native oxide or remained from the ex situ cleaning process [115].
This initial flash also degasses the sample holder more thoroughly before further
flashes to higher temperatures. After each flash anneal, the sample temperature is
dropped rapidly back to around 600 ◦C. The flash temperature is then sequentially
increased during the next several (one to five) flashes depending on the level of
outgassing during flashing. These initial higher temperature flashes have a duration
of approximately 15 s. The pressure spike that occurs from the higher temperature
184
outgassing (mostly from the sample holder) determines the duration and number of
times that each flash is repeated at a certain temperature. If the pressure during
flashing reaches the critical 1.3× 10−7 Pa value, the flash is interrupted to reduce
the pressure. This cycle is repeated until a final flash temperature of approximately
1150 ◦C to 1200 ◦C is reached. The sample sits at these higher temperatures for a
max of 10 s each. The purpose of the higher temperature flashes is to desorb any
trace amounts of oxide or other contaminants remaining on the surface and anneal
surface defects such as missing or buckled dimers to produce a clean, well ordered
Si(100) (2×1) surface. After the final flash, the sample is cooled slowly at a rate of
approximately 1 ◦C/s, which additionally helps to recrystallize the surface, down to
either the deposition temperature or near room temperature. Typically, as few as
five total flashes are needed to prepare a substrate, although some samples/sample
holders outgas more, requiring tens of flashes. A typical flash-anneal sequence for
substrates used for many of the 28Si samples deposited at DC–3 is as follows:
1. degas initially at ≈ 400 ◦C for 1 h,
2. degas at ≈ 600 ◦C for > 12 h,
3. flash to ≈ 1050 ◦C for 1 min then return to 600 ◦C,
4. flash to ≈ 1150 ◦C for 15 s then return to 600 ◦C,
5. repeat step 3,
6. flash to ≈ 1200 ◦C for 10 s then return to 600 ◦C,
7. repeat step 5,
8. flash to ≈ 1150 ◦C for 10 s then return to 850 ◦C,
9. ramp down the temperature at 1 ◦C/s to the desired final temperature.
185
Again, each flashing step is repeated until the pressure in the chamber remains less
than 1.3× 10−7 Pa for the maximum duration of the flash before the next flash step
proceeds.
In order to provide feedback on this high temperature in situ substrate prepa-
ration process and assess its effectiveness for each sample before deposition, two
analysis techniques are used: RHEED and STM. RHEED is used during the flash-
ing process to identify oxide remaining on the surface and verify the (2×1) recon-
struction of the Si surface. RHEED images are typically captured between flashes
when the substrate is at ≈ 600 ◦C, but not typically after step 6 above so as to
not introduce possible contamination due to the RHEED electron beam interacting
with the surface. The number of flashes required to prepare a sample is dictated by
the quality of the RHEED pattern in addition to the amount of outgassing. The
highest temperature flash may be repeated until a sufficient RHEED pattern typical
of a clean Si (2×1) surface is observed (discussed below). RHEED is also used to
screen for samples contaminated with silicon carbide (SiC) or those with particu-
larly rough surfaces. Figure 5.1 shows two RHEED images of a Si substrate before
and after flash annealing at DC–3 to prepare a clean surface. Panel (a) shows the
diffraction pattern from an intrinsic Si(100) substrate before flashing. Faint spots
are seen in a (1×1) pattern originating from diffraction to rods in reciprocal space,
but the image is dominated by the diffuse background caused by the amorphous
native SiO2 layer. This image was acquired at a substrate temperature of ≈ 600 ◦C
with the electron beam in the 〈110〉 direction. Panel (b) shows the sample substrate
after flash annealing seven times to a maximum temperature of ≈ 1197 ◦C for 8 s.
186
Figure 5.1: RHEED diffraction patterns showing the transition from Si with a native
oxide to a reconstructed Si(100) (2×1) surface during the substrate flash annealing
procedure. (a) Before flashing the substrate, a diffraction pattern showing weak Si
(1×1) spots on a diffuse background is seen which indicates the presence of the SiO2
surface layer. (b) After flashing the substrate to ≈ 1197 ◦C, the diffraction pattern
shows strong Si(100) (2×1) spots indicating a clean, reconstructed surface. Both
images were acquired at a substrate temperature of about 600 ◦C with the electron
beam in the 〈110〉 direction.
After flashing, strong Si (2×1) diffraction spots appear in the pattern between the
outer (1×1) spots, and the diffuse background is gone. These features plus the ab-
sence of additional spots due to chemical contaminants such as C indicates that the
substrate has a reasonably clean, crystalline, and reconstructed Si(100) surface.
RHEED serves as a quick feedback mechanism during the substrate flashing
procedure by indicating the general state and quality of surface, and a high-quality
187
RHEED pattern corresponding to a clean Si surface is a necessary condition for
producing an atomically clean Si(100) ideal for epitaxial deposition. A RHEED
pattern indicating a clean surface is not, however, a sufficient condition because the
RHEED pattern is an amalgam resulting from diffraction over a macroscopic area
(≈ 0.25 mm2) on the substrate. RHEED is thus insensitive to trace surface defects
and contaminants. To assess the surface quality of prepared substrates on an atomic
scale, STM imaging is used. The STM used in this system was briefly described in
Chapter 2.
After samples are flashed and cooled at a deliberate rate to approximately
250 ◦C using the DH power, they cool radiatively for at least 10 min to > 50 ◦C
before being transferred into the STM chamber. The typical base pressure in the
STM chamber is approximately 6.7× 10−9 Pa (5.0× 10−11 Torr), and it is separated
from the deposition chamber by a gate valve. The samples remain protected in the
isolated UHV environment of the STM during ion beam tuning in the deposition
chamber. The substrates are typically scanned both on relatively large areas (1 µm
× 1 µm) and smaller areas (50 nm × 50 nm) to screen for both larger particulates
or other features as well as atomic scale defects. Most STM images shown in this
chapter were acquired in conjunction with Hyun soo Kim, who assisted with some
aspects of these experiments. Several typical STM images of three different sub-
strate types after flash annealing are shown in Fig. 5.2. These STM topography
images all show clean Si(100) (2×1) reconstructed surfaces with atomic steps and
atomically flat terraces of various widths. Panels (a) and (b) show images of boron-
doped Si substrates with a relatively small average terrace width of approximately
188
Figure 5.2: STM topography filled state images of clean Si(100) (2×1) surfaces for
three substrate types prepared in situ by flash annealing. Images were typically
acquired with a tip bias ≈ -2 V and a tunneling current ≈ 100 pA. These images
show atomic steps on the Si surface and flat terraces of varying widths. Si (2×1)
dimer rows are also seen in (a), (b), (d), and (f) with a minimal density of dimer
defects (dark spots). (a) and (b) are boron-doped Si substrates with an average
terrace width of around 15 nm. (c) and (d) are intrinsic Si substrates with an average
terrace width of around 55 nm. (e) and (f) are phosphorous-doped Si substrates with
an average terrace width of about 100 nm.
189
15 nm. The terrace size is a result of a misalignment during manufacturing of the
cut direction of the wafer and the (100) crystal plane. When this miscut angle is
small, larger terraces result. Panels (c) and (d) show images of intrinsic Si substrates
with an average terrace width of approximately 55 nm. Finally, panels (e) and (f)
show images of the phosphorous-doped Si substrates with a small miscut angle of
± 0.05◦ resulting in a large average terrace width of approximately 100 nm. The
increase in average terrace width is apparent when noting the difference in scale
between panels (a), (c), and (e). Panel (c) shows an area roughly 25 times larger
than (a), and panel (e) shows an area four times larger than (c). Si (2×1) dimer
rows are also seen predominantly in panel (b) and are just visible in (a), (d), and
(f). A minimal density of dimer row defects (dark spots) are seen, which confirms
that the surface cleaning procedures were sufficient. These defects can simply be
missing surface atoms or due to chemical contaminants [116].
After the substrate surface is inspected via STM, and the ion beam is tuned
for deposition, the sample is moved back to the deposition chamber and flashed a
final time to remove possible adsorbates which may have accumulated during STM
scanning. The substrate temperature is then lowered to the deposition temperature
before 28Si deposition commences. This cool down typically takes around 30 min as
the substrate comes into equilibrium.
190
5.4 Deposition of 28Si
Once substrates are prepared and ion beam tuning similar to that of Chapter
4 is complete, 28Si is deposited in the deposition chamber at DC–3 using a range of
deposition conditions. As was mentioned previously, one of the critical degrees of
freedom for achieving epitaxial deposition that is enabled in this final experimental
setup is the substrate temperature. 28Si samples were deposited using a range of
substrate temperatures, T , from room temperature (≈ 21 ◦C) up to a maximum of
1080 ◦C and many temperatures in between. Depositing at room temperature allows
for comparison between these samples and those deposited in the previous setup at
LC–2. The most common sample deposition temperatures used were either around
450 ◦C or 700 ◦C. The range of temperatures was used to determine the effect of
temperature on both the epitaxial quality of the 28Si films and the adsorption of
SiH4 gas into samples, which then increases the
29Si and 30Si isotopic concentrations.
The first of these effects is described in a later section of this chapter, and the second
is described in Chapter 6.
For the samples deposited at DC–3, both the nominal low pressure working
mode of the ion source as well as the high pressure working mode were used, which
led to a range of deposition parameters. When using the low pressure mode, during
deposition, the pressure in the deposition chamber rose to between 1.6× 10−7 Pa
and 1.5× 10−6 Pa (1.2× 10−9 Torr to 1.1× 10−8 Torr) due to the SiH4 gas diffusion
from the ion beam chamber. Most samples, however, were deposited in a background
pressure of about 1.1× 10−6 Pa (8.0× 10−9 Torr). This pressure is a factor of three
191
lower than the low end of the range of background pressures measured during de-
position of samples at LC–2 in Chapter 4. The RGA in the deposition chamber
is routinely used to monitor the partial pressures of various chemical contaminants
present in the deposition chamber during 28Si deposition. A typical RGA mass
spectrum recorded while depositing a 28Si sample with a substrate temperature of
705 ◦C at DC–3 is shown in Fig. 5.3. H2 dominates the spectrum and is the result of
both residual H2 in the deposition chamber and also cracking of SiH4 gas in the ion
source releasing H2 which diffuses to the deposition chamber. SiH4 gas also diffuses
into the chamber, however, cracking from the RGA itself reduces the SiH4 signal
to mostly just the 28 u peak. This peak also contains N2 and CO. SiH4
2+ hydride
peaks do appear between 14 u and 16 u at 0.5 u increments. Partial pressures of
other chemicals of interest can be seen including C, H2O, F, and CO2. The peak
appearing at 26 u is not known, but it may be related to an alcohol. When using the
high pressure mode, the pressure in the deposition chamber rose to approximately
4.0× 10−6 Pa (3.0× 10−8 Torr), which is comparable to the samples deposited at
LC–2.
Ions were deposited with a range of average ion energies, Ei, at the sample
between 20 eV and 40 eV. This energy range is smaller with lower overall values
than the samples discussed in Chapter 4 in order to standardize the deposition
process and minimize sample sputtering. The estimated sputter yield for 30 eV 28Si
ions hitting a crystalline Si surface is approximately 2 % (see Fig. 2.19 in Chapter
2). The most common average ion anergy used here was approximately 35 eV. For
the low pressure mode, the average ion beam current achieved was approximately
192
Figure 5.3: Residual gas mass spectrum collected from the RGA in the deposition
chamber while operating the ion beam and depositing a 28Si sample at 705 ◦C at
DC–3. The spectrum is dominated by H2 which diffuses from the ion beam chamber
as SiH4 gas is cracked in the ion source. The SiH4 complex of peaks normally is
reduced to only the 28 u peak due to cracking of the molecules by the RGA itself.
N2 and CO also make up the 28 u peak. Doubly charged SiH4 hydrides do appear
at around 15 u. Other potential chemical contaminants are present including C, F,
H2O, and CO2.
0.55 µA. One major advantage of the high pressure plasma mode is that it yields
a much larger ion current on average, and for these samples, an ion current of
approximately 2.5 µA was achieved.
A mass spectrum for a Si ion beam with an ion energy of 37 eV generated
using the low pressure mode of the ion source is shown in Fig. 5.4. This spectrum
was collected using the interchangeable sample aperture at DC–3 prior to depositing
a 28Si sample. The corresponding magnet current used for the field sweep of the
magnetic sector mass analyzer is shown on the top axis. Ion current peaks (circles)
193
Figure 5.4: SiH4 mass spectrum representative of the ion beam settings for samples
deposited at DC–3 using the low pressure mode of the ion source. The ion current
(circles) is recorded while sweeping the the analyzer current, and thus the magnetic
field (top axis). The 28 u peak is 28Si and the 29 u peak is both 28SiH and ≈ 5 %
29Si. Several higher order hydrides are also shown. Gaussian fits (line, Eq. (2.14))
to the 28 u and 29 u peaks are shown superimposed on the data. The centers of the
28 u and 29 u fits are separated by ≈ 10 σ.
corresponding to 28Si and Si hydrides are observed between 28 u and 33 u. The ion
current peaks in this spectrum appear qualitatively similar to the mass spectrum
acquired in the previous experimental configuration (Fig. 4.8). The ion peak at
28 u is 28Si and the ion peak at 29 u contains ≈ 5 % 29Si based on the similar peak
height to that of 28 u and the expected natural abundance, as discussed in Chapter
2. These peaks show a high degree of separation with no detectable ion current
signal occurring between the peaks. Gaussian fits to the peaks (line, Eq. (2.14)) are
also shown superimposed on the data, and one can see only a slight asymmetry in
194
the beam profile and no shoulder peaks, which indicates a fairly optimally tuned
beam with minimal scattering off of lens elements. The mass resolving power of
the ion beam in this configuration derived from this mass spectrum is m
∆m
≈ 57
(measured at 10 % of the peak height). This value is similar to but slightly lower
than the mass resolving power achieved in the previous experimental configuration
perhaps due to the wider mass-selecting aperture used in this setup. However, no
current is detected between the peaks, and the geometric selectivity is still expected
to be sufficient to produce highly enriched 28Si films comparable to those produced
at LC–2. The Gaussian fits give a separation of the 28 u peak from the 29 u peak
of approximately 10 σ.
A mass spectrum for a Si ion beam with an ion energy of 31 eV generated
using the high pressure mode of the ion source is shown in Fig. 5.5. This spectrum
was also collected using the interchangeable sample aperture prior to depositing
a 28Si sample. Again, the corresponding current used for the field sweep for the
magnetic sector mass analyzer is shown on the top axis. The ion current peaks
(circles) observed between 28 u and 33 u show the peak shapes typical of the high
pressure mode. This spectrum shows a dramatic increase in the 28Si ion current over
the low pressure spectrum and a decrease in the 30 u and 31 u peaks. The 28Si ion
current becomes larger while the hydride peaks become smaller. This is because the
hydrides are cracked more efficiently in the high pressure ion source plasma mode, as
discussed in Chapter 2. The 28Si ion current peak at 28 u and the ion current peak
at 29 u show a fairly good separation but with a current level between the peaks of
about 8 nA, which is higher than for the low pressure mode. Using the high pressure
195
Figure 5.5: SiH4 mass spectrum representative of the ion beam settings for samples
deposited at DC–3 using the high pressure mode of the ion source. The ion current
(circles) is recorded while sweeping the mass analyzer current, and thus the magnetic
field (top axis). The 28 u peak is 28Si and the 29 u peak is both 28SiH and ≈ 24 %
29Si. Several higher order hydrides are also shown. Gaussian fits (line, Eq. (2.14))
to the 28 u and 29 u peaks are shown superimposed on the data. The centers of the
28 u and 29 u fits are separated by ≈ 8 σ.
mode, 29Si makes up approximately 24 % of the 29 u peak based on the peak heights
and the expected natural abundance, as discussed in Chapter 2. Gaussian fits to
the peaks (line, Eq. (2.14)) are also shown superimposed on the data as a sum. An
asymmetry in the lower portion of the peaks possibly indicates some beam scattering
to the lower mass side. The mass resolving power derived from this mass spectrum
is m
∆m
≈ 48 (measured at 10 % of the peak height), which is reduced further from
the previous low pressure mode mass spectrum. The asymmetric peak and the lower
mass resolving power may reduce the geometric selectivity and lower the achievable
196
enrichment level in 28Si samples deposited using this mode. However, the roughly
factor of five increase in 28Si current compared to the 29 u peak containing 29Si may
offset any deleterious effects of a lower observed peak separation. The Gaussian fits
give a separation of the 28 u peak from the 29 u peak of approximately 8 σ.
Tuning and focusing the ion beam onto the interchangeable sample aperture on
the sample manipulator resulted in deposition spot sizes similar to those of samples
deposited at LC–2. The average deposition area for samples deposited at DC–3 was
approximately 6 mm2 and as small as 2 mm2 for a few samples. During deposition,
RHEED was typically used to periodically monitor the structure of the depositing
film and verify the location of the deposition spot on the chip. The appearance
of the 28Si deposition spot on the substrate of samples deposited at DC–3 varied
depending on the deposition temperature. Optical micrographs of three samples
deposited at three different temperatures at DC–3 are shown in Fig. 5.6. These
micrographs show the Si(100) substrates with the 28Si deposition spot appearing
at the center as three visually distinct areas. Panel (a) shows 28Si deposited on an
intrinsic Si chip with a substrate temperature of approximately 249 ◦C in an area
about 4.7 mm long. The deposition spot is clearly visible as a discolored patch on the
chip and is visually similar to the samples deposited at room temperature at LC–2 in
Chapter 4. This may indicate poor epitaxial quality. Panel (b) shows 28Si deposited
on a phosphorous-doped Si chip at a substrate temperature of about 421 ◦C in an
area approximately 2.6 mm wide. The deposition spot is nearly indistinguishable
from the substrate and only a faint outline is visible. This may indicate that the
film is structurally similar to or even epitaxially matched with the substrate. The
197
Figure 5.6: Optical micrographs of three 28Si samples deposited at DC–3 at three
different temperatures. The Si(100) substrates are seen in all three micrographs with
the 28Si deposition spot at the center. (a) 28Si deposited on an intrinsic Si chip at
249 ◦C in an area 4.7 mm long is clearly visible. (b) 28Si deposited on a phosphorous-
doped Si chip at 421 ◦C in an area 2.6 mm wide is nearly indistinguishable from the
substrate. (c) 28Si deposited on a boron-doped Si chip at 705 ◦C in an area 3.7 mm
wide is clearly visible.
198
brightness and contrast of this micrograph was altered to highlight the edges of the
deposition spot, which is why the substrate appears lighter. Panel (c) shows 28Si
deposited with a substrate temperature of approximately 705 ◦C on a boron-doped
Si chip in an area approximately 3.7 mm wide. The deposition spot is clearly visible
on the chip and appears as a diffuse whitish color. Light scattering from the surface
of the deposited area is probably due to a higher degree of surface roughness, which
will be discussed later in this chapter.
The thicknesses of these 28Si films ranged from around 45 nm to 370 nm as
measured by TEM and/or inferred from SIMS depth profiling. The corresponding
deposition rates for samples deposited using the low pressure mode of the ion source
were between 0.32 nm/min and 1.41 nm/min with the most common deposition
rate around 0.80 nm/min. The high pressure mode of the ion source resulted in a
higher range of rates between 2.2 nm/min and 4.6 nm/min. These higher rates are
desirable because they enable production of thicker samples in a typical deposition
time (assuming similar spot sizes), and they lessen the relative adsorption rate of
gaseous species from the background pressure into the 28Si film during deposition.
To summarize, the typical deposition procedure for samples deposited at DC–3
was as follows:
1. a substrate is diced from a wafer and chemically cleaned ex situ before being
mounted onto a sample holder and loaded into the load lock which is pumped
out for > 12 h,
2. the substrate is moved into the deposition chamber and degassed on the sample
manipulator at 600 ◦C for > 12 h,
3. the substrate is flash annealed to ≈ 1200 ◦C to prepare a clean (2×1) surface
199
which is inspected by RHEED,
4. after a deliberate temperature ramp down to ≈ 250 ◦C following flash anneal-
ing, the substrate cools radiatively for at least 10 min to ≈ 50 ◦C and is then
transferred to the STM surface inspection,
5. with the sample in the STM, the ion beam is tuned and characterized using
the interchangeable sample aperture on the manipulator. The mass spectrum,
ion beam energy, beam spot focusing, and the deposition chamber RGA are
all analyzed and recorded,
6. the substrate is transferred back to the deposition chamber after the gate valve
to the ion beam is closed to reduce the background pressure where it is flashed
once and set to the deposition temperature and position,
7. the gate valve to the ion beam is opened to commence deposition of 28Si while
monitoring the ion current onto the sample as a measure of the deposition
rate,
8. RHEED is periodically used to monitor the crystallinity of the deposition and
locate the deposition spot,
9. after typically three to five hours of deposition, the gate valve to the ion beam
is closed to end deposition and reduce the deposition chamber pressure back
to its base, then the sample is cooled to room temperature,
10. the sample is then transferred to the STM to inspect the surface morphology
of the deposited 28Si film, and finally,
11. the sample may be annealed at around 600 ◦C back on the manipulator and
later inspected by STM again before ultimately being removed from the vac-
uum chamber via the load lock for ex situ analyses.
In total, 40 28Si samples were deposited at DC–3 including one room temper-
ature sample, 32 samples deposited at elevated temperatures using the low pressure
mode, and seven samples deposited at elevated temperatures using the high pressure
mode.
200
5.5 Enrichment Measurements via SIMS for
DC–3 Samples
5.5.1 Initial Tests at DC–3
SIMS was used to assess the enrichment of a large number of samples deposited
at DC–3. The basic procedure for SIMS isotope measurements of 28Si samples was
discussed in Chapter 4. Isotope ratio measurements by SIMS were used to calcu-
late the isotope fractions (zSi/Sitot.). SIMS depth profiles presented in this chapter
were acquired in collaboration with Dr. David Simons (NIST). Initial SIMS mea-
surements were done to verify that the high levels of 28Si enrichment achieved for
samples deposited at LC–2 in Chapter 4 were reproducible for samples deposited at
DC–3 after the significant experimental reconfiguration of connecting the ion beam
to the deposition chamber. Recall the mass-selecting aperture width was increased,
which may decrease the realized mass selectivity. Additionally, lower partial pres-
sures of SiH4 during deposition reduced the potential for adsorption of SiH4 into
the 28Si films, which results in increased 29Si and 30Si isotope fractions. Along those
same lines, it must be determined what role the substrate temperature plays in the
enrichment of the samples. While the substrate deposition temperature is increased
to facilitate epitaxial deposition, it may also change the kinetics and chemistry of
any possible SiH4 adsorption into the samples. Higher temperatures may reduce
gaseous adsorption or perhaps the added energy may enhance incorporation into
the depositing films. Addressing this topic required many SIMS measurements of
201
samples deposited at a large range of temperatures (some of which are shown here)
and is the subject of Chapter 6.
Depositing a sample at room temperature should exclude any potential sub-
strate temperature effects on the enrichment and provide a SIMS measurement more
comparable to those of the samples deposited at LC–2 to verify the enrichment level
of samples is no worse for this setup. A SIMS depth profile for a 28Si sample de-
posited at room temperature (≈ 21 ◦C) using the low pressure mode of the ion source
at DC–3 is shown in Fig. 5.7. Measurements of the 28Si (circles), 29Si (squares), and
30Si (triangles) isotope fractions are shown vs. the sputter depth into the sample.
This sample was relatively thin, and so the data density of the measurement is lower
than for other samples. Also, a fairly short region (≈ 20 nm) exists where the iso-
tope fractions reach a minimum and are averaged for the measurement. The count
rate for a number of SIMS measurement cycles in this region was zero and thus
that data does not appears on this semi-log plot. At a depth of around 40 nm, the
isotope fractions begin to increase to their natural abundance values (dotted and
dashed lines) in the transition into the Si(100) substrate, which is marked by the
shaded region. The interface between the film and the substrate is estimated to be
at a depth of about 53 nm, which is also the film thickness.
The average measured 28Si isotope fraction in this sample is 99.999898(35) %.
After some initial surface contamination, the average residual 29Si isotope fraction
measured between 13 nm and 41 nm is 0.58(26)× 10−6 (0.58(26) ppm), and the
average 30Si isotope fraction is 0.44(23)× 10−6 (0.44(23) ppm). Not only is the
measured level of enrichment not diminished after the experimental transition to
202
Figure 5.7: SIMS depth profile of a 28Si sample deposited at room temperature using
the low pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles)
isotope fractions are shown vs. sputter depth. The average 29Si isotope fraction
in the film is 0.58 ppm, and the average 30Si isotope fraction is 0.44 ppm (dashed
lines). Note that data are zero counts in some SIMS cycles. The natural abundance
values for each isotope (dotted and dashed lines) are also shown for reference. At a
depth of 40 nm, the isotope fractions begin to increase to their natural abundance
values in the Si(100) substrate. The interface between the film and the substrate
(shaded region) is estimated to be at a depth of 53 nm.
this final experimental configuration, it is slightly better than the most highly en-
riched sample deposited at LC–2 (0.691(74) ppm 29Si). This may be the result of
a combination of the background pressure during deposition of this sample being
roughly a factor of three lower than that of the sample deposited at LC–2 and the
deposition rate being nearly a factor of three higher. These counter acting condi-
tions would in fact result in only a slightly better enrichment. Also, apparently the
widening of the mass-selecting aperture did not measurably decrease the realized
selectivity in the 28Si samples.
203
Figure 5.8: SIMS depth profile of a 28Si sample deposited at 249 ◦C using the low
pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. The average 29Si isotope fraction in the
film is 0.79 ppm, and the average 30Si isotope fraction is 0.229 ppm (dashed lines).
Note that most data are zero counts in particular SIMS cycles for 30Si. The natural
abundance values for each isotope (dotted and dashed lines) are also shown for
reference. At a depth of about 280 nm, the isotope fractions begin to increase
to their natural abundance values in the substrate (shaded region). The interface
between the film and the substrate is estimated to be at a depth of 305 nm.
The room temperature sample can also be compared with samples deposited at
elevated temperatures. Figure 5.8 shows a SIMS depth profile for a sample deposited
at approximately 249 ◦C at DC–3 using the low pressure mode. The 28Si (circles),
29Si (squares), and 30Si (triangles) isotope fractions are shown vs. the sputter depth
into the sample. This much thicker sample shows a large flat region in the isotope
fraction profile and an abrupt increase up to the natural abundance values in the
substrate indicating a sharp interface. At a depth of about 280 nm, the isotope
fractions begin to increase to their natural abundance values (dotted and dashed
204
lines) in the transition into the substrate, which is marked by the shaded region.
The interface between the film and the substrate is estimated to be at a depth of
approximately 305 nm, giving a value for the film thickness.
For this sample, the average measured 28Si isotope fraction in the film is
99.999898(13) %. After sputtering through some initial surface contamination, the
average residual 29Si isotope fraction is 0.79(12)× 10−6 (0.79(12) ppm), and the av-
erage 30Si isotope fraction is 0.229(64)× 10−6 (0.229(64) ppm). For the 30Si data,
most of the SIMS measurement cycles in this region contained zero counts and thus
the data does not appears on this semi-log plot. These averages were calculated
using the data between 32 nm and 285 nm. The level of enrichment of this sample
is very similar to that of the room temperature sample. In fact, the 28Si isotope frac-
tion is identical with slightly different proportions of 29Si and 30Si. When this result
is combined with a SIMS measurement of a second area of this sample, the average
29Si isotope fraction of the two measurements is 0.90(10) ppm. These measurements
shows that 28Si samples can successfully be deposited at elevated temperatures us-
ing the low pressure mode of the ion source at DC–3 while maintaining a very high
level of enrichment.
5.5.2 Enrichment Progression Timeline Samples
Out of the 40 samples produced at DC–3, three are represented on the en-
richment progression timeline (Fig. 4.2) in Chapter 4 as new record enrichments
achieved in this work at the time of their deposition, and they will be discussed in
this section. In fact, these samples achieved the highest enrichments of any samples
205
produced in this work. A SIMS depth profile for the first of these 28Si samples
deposited using the low pressure mode of the ion source at DC–3 with a higher level
of enrichment than the previous ones deposited at LC–2 is shown in Fig. 5.9. This
sample was deposited at a substrate temperature of approximately 610 ◦C using the
low pressure mode of the ion source. The 28Si (circles), 29Si (squares), and 30Si (tri-
angles) isotope fractions are shown vs. the sputter depth into the sample. Beyond
some initial surface contamination, there is only a relatively thin region between
20 nm and 70 nm over which the 29Si isotope fraction reaches a sustained minimum
value. In this region of the film, the isotope fractions were averaged. At a depth
below 70 nm, the 29Si isotope fraction increases more than two orders of magnitude.
This increase is likely a result of extended ion beam tuning off of the 28 u peak after
deposition had commenced. At a depth of around 125 nm, the 29Si and 30Si isotope
fractions together begin to increase to their natural abundance values (dotted and
dashed lines) in the transition into the substrate, which is marked by the shaded
region. The interface between the film and the substrate is estimated to be at a
depth of approximately 162 nm, giving a value for the total film thickness. The
length over which the isotope fractions return to their natural abundance values is
relatively wide in this sample compared to the previous sample, possibly indicating
a larger degree of surface roughness.
Within the highly enriched portion of this sample, the average measured 28Si
isotope fraction is 99.9999570(70) %. The average residual 29Si isotope fraction in
the film is 3.00(60)× 10−7 (300(60) ppb), and the average 30Si isotope fraction is
1.30(37)× 10−7 (130(37) ppb). Some of the SIMS measurement cycles in this region
206
Figure 5.9: SIMS depth profile of a 28Si sample deposited at 610 ◦C using the low
pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. The isotope fraction values were averaged
between 20 nm and 70 nm. The average 29Si isotope fraction in this part of the film
is 300 ppb, and the average 30Si isotope fraction is 130 ppb (dashed lines). Note
that most data are zero counts in particular SIMS cycles. The natural abundance
values for each isotope (dotted and dashed lines) are also shown for reference. At a
depth of about 70 nm, the 29Si isotope fraction increases two orders of magnitude
likely as a result of ion beam tuning off of the 28 u peak. At a depth of 125 nm, the
29Si and 30Si isotope fractions begin to increase to their natural abundance values.
The interface between the film and the substrate (shaded region) is estimated to be
at a depth of 162 nm.
for the 29Si data and most of the measurement cycles for the 30Si data contained
zero counts and thus the data does not appears on this semi-log plot. These mea-
surements indicate an isotope reduction factor (az/(
zSi/Sitot.), discussed in Chapter
4) for 29Si of 1.6(3)× 105, i.e. the 29Si is approximately 1.6(3)× 105 times lower
than in natural abundance Si. The isotope reduction factor for 30Si is slightly higher
at 2.4(4)× 105, although it is similar to that of 29Si within the uncertainty of the
207
values. The 29Si isotope fraction of this sample is a little more than a factor of two
less than that of the most highly enriched sample deposited at LC–2 in Chapter 4,
as seen in the jump from LC–2 to DC–3 in Fig. 4.2. This decrease in 29Si and 30Si
isotope fractions was probably due to the significantly lowered background deposi-
tion pressure for these samples, which would result in a lower amount of adsorbed
SiH4. Additionally, this result shows that very high enrichments continue to be
achieved while the substrate deposition temperature is increased further compared
to the previous sample discussed in this section. Despite the high level of enrichment
achieved in a portion of the film, this sample is clearly not ideal due to the region
of elevated 29Si. Additional SIMS measurements of samples deposited under similar
conditions show the elevated 29Si to be an anomaly.
The second sample represented on the enrichment progression timeline in
Fig. 4.2, which was deposited at DC–3 and achieved an even higher level of en-
richment over the previous sample is one that was deposited with a substrate tem-
perature of approximately 712 ◦C using the low pressure mode. A SIMS depth
profile of the area of highest enrichment of this sample is shown in Fig. 5.10. The
28Si (circles), 29Si (squares), and 30Si (triangles) isotope fractions are shown vs. the
sputter depth into the sample. Beyond the initial signal due to surface contamina-
tion, the 29Si and 30Si isotope fractions show an extended minimum. At a depth of
around 227 nm, the isotope fractions begin to increase to their natural abundance
values (dotted and dashed lines) in the transition into the substrate, which is marked
by the shaded region. The interface between the film and the substrate is estimated
to be at a depth of approximately 256 nm, giving a value for the film thickness.
208
Figure 5.10: SIMS depth profile of a 28Si sample deposited at 712 ◦C using the low
pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. The average 29Si isotope fraction in the film
is 132 ppb, and the average 30Si isotope fraction is 70 ppb (dashed lines). Many data
in this region are zero counts in the SIMS cycles. The natural abundance values
for each isotope (dotted and dashed lines) are also shown for reference. The isotope
fraction values were averaged between 21 nm and 227 nm, where the 29Si and 30Si
isotope fractions begin to increase to their natural abundance values. The interface
between the film and the substrate (shaded region) is estimated to be at a depth of
256 nm.
The average measured 28Si isotope fraction in the highly enriched portion of
the film is 99.9999797(30) %. The average residual 29Si isotope fraction is mea-
sured to be 1.32(27)× 10−7 (132(27) ppb), and the average 30Si isotope fraction is
7.0(12)× 10−8 (70(12) ppb) (dashed lines). Many of the SIMS measurement cycles
in this region contained zero counts and thus the data does not appears on this semi-
log plot. The averages were calculated from the data between 21 nm and 227 nm.
These measurements indicate an isotope reduction factor for 29Si of 3.5(7)× 105.
209
The isotope reduction factor for 30Si is slightly higher at 4.4(8)× 105, although
they are similar within the uncertainty of the values. A second area of this sam-
ple was also measured by SIMS giving an average 29Si isotope fraction of the two
measurements of 163(18) ppb. As mentioned in Chapter 4, this variation between
measurements on the same sample is likely due to a difference in deposition rate at
the two areas of the film.
The enrichment of this sample represents yet another decrease of more than
a factor of two in the isotope fraction of 29Si from the 610 ◦C sample, which can
be seen in Fig. 4.2, the enrichment progression timeline. This reduction is possibly
due to a lower rate of incorporation of SiH4 into the film from a slight decrease in
deposition background pressure and also a nearly twofold increase in deposition rate
for this sample compared to the previous one. The deposition was approximately
1.22 nm/min, which is fairly high compared to other samples deposited using the
low pressure mode. This was a result of both a slightly higher average ion beam
current of about 740 nA and better beam spot tuning that led to a more compact
than average deposition spot of about 4.2 mm2. A complicating factor which may
effect the enrichment of this sample is an anomalous amount of nitrogen observed
in the ion beam mass spectrum before depositing this sample. This matter will be
addressed in the later sections of this chapter.
The third and final sample represented on the enrichment progression timeline
in Fig. 4.2 that was deposited at DC–3 is a sample deposited at approximately
502 ◦C using the low pressure mode. This 28Si sample is the most highly enriched
sample produced in this entire work. A SIMS depth profile of the area of highest
210
enrichment within this sample is shown in Fig. 5.11. The 28Si (circles), 29Si (squares),
and 30Si (triangles) isotope fractions are shown vs. the sputter depth into the sample.
Beyond the initial signal due to surface contamination, the 29Si and 30Si isotope
fractions show an extended minimum with very little variation in value. At a depth
of around 290 nm, the isotope fractions begin to increase to their natural abundance
values (dotted and dashed lines) in a fairly abruptly transition into the substrate,
which is marked by the shaded region, indicating a sharp interface. The interface
between the film and the substrate is estimated to be at a depth of approximately
321 nm, which also gives a value for the film thickness.
The average measured 28Si isotope fraction in this highly enriched film is
99.9999819(35) %. This value is the highest enrichment of any sample measured
in this work. Additionally, the average residual 29Si isotope fraction is the lowest
for any sample with a value of 1.27(29)× 10−7 (127(29) ppb), and the average 30Si
isotope fraction is 5.5(19)× 10−8 (55(19) ppb) (dashed lines). The averages were
calculated from the data between 30 nm and 290 nm. They lie well below where the
data appears because many of the SIMS measurement cycles in this region contained
zero counts and thus the data is not represented on this semi-log plot. These en-
richment measurements of 29Si and 30Si represent the best isotope reduction factors
for this entire work. The isotope reduction factor for 29Si is 3.7(8)× 105, and the
isotope reduction factor for 30Si is higher with a value of 5.6(19)× 105. These values
are actually similar to each other to within their uncertainties due to a large relative
uncertainty in the measurement of the isotope fraction of 30Si in this sample. The
level of enrichment of this sample is similar to that of the 712 ◦C sample, and while
211
Figure 5.11: SIMS depth profile of the most highly enriched 28Si sample deposited
at DC–3. This sample was deposited with a substrate temperature of 502 ◦C using
the low pressure mode. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. The average 29Si isotope fraction in the film
is 127 ppb, and the average 30Si isotope fraction is 55 ppb (dashed lines). Many data
in this region are zero counts in the SIMS cycles. The natural abundance values
for each isotope (dotted and dashed lines) are also shown for reference. The isotope
fraction values were averaged between 21 nm and 292 nm, where the 29Si and 30Si
isotope fractions begin to increase to their natural abundance values. The interface
between the film and the substrate (shaded region) is estimated to be at a depth of
321 nm.
the center value of this measurement is lower than that of the previous sample, they
agree within their uncertainties as seen in the enrichment progression timeline in
Fig. 4.2. Like the 712 ◦C sample, the extremely low isotope fractions of 29Si and
30Si observed in this sample are likely due to a lower rate of incorporation of SiH4
into the film. This is probably due in part to a relatively high deposition rate of
1.41 nm/min, which is actually the highest deposition rate for samples deposited
using the low pressure mode at DC–3. Again, the deposition temperature may also
212
be playing a part in the reduction of 29Si and 30Si although it is not clear from this
data alone.
A total of three SIMS measurements were made on different areas of this
sample resulting in an average 29Si isotope fraction for the three areas of 233(39) ppb.
The single area best value of 127(29) ppb 29Si residual isotope fraction is given more
weight in this discussion than the multiple measurement average because the average
depends on the distribution of the measurement areas across regions of the film with
different deposition rates. The single area best value also gives a lower bound on
the best possible enrichment achievable by the deposition system for the specific ion
beam selectivity, deposition rate, background pressure, and substrate temperature
used.
The samples already described in this chapter, which were deposited with el-
evated substrate temperatures, demonstrate enrichment levels similar to or better
than those of the highly enriched room temperature samples deposited at LC–2,
i.e. isotope fractions < 1 ppm of 29Si. However, some samples deposited at DC–3
with elevated substrate temperatures showed significantly higher isotope fractions
of 29Si and 30Si. One sample deposited at approximately 357 ◦C using the low
pressure mode was measured by SIMS to have a average 28Si isotope fraction of
99.999405(93) %. The average residual 29Si isotope fraction in the film is mea-
sured to be 4.18(70)× 10−6 (4.18(70) ppm), and the average 30Si isotope fraction is
1.77(61)× 10−6 (1.77(61) ppm). These higher isotope fractions are likely due in part
to a low deposition rate for this sample of 0.33 nm/min and a higher background
pressure during deposition of approximately 1.5× 10−6 Pa (1.1× 10−8 Torr), which
213
equals the highest pressure for any sample deposited at DC–3 using the low pressure
mode of the ion source.
Another sample deposited with a substrate temperature of 421 ◦C using the
low pressure mode was measured in one area to have an average 28Si isotope frac-
tion of 99.999812(25) %. The average residual 29Si isotope fraction measured for
this sample is 1.30(22)× 10−6 (1.30(22) ppm), and the average 30Si isotope fraction
is 5.8(12)× 10−7 (0.58(12) ppm). Averaging the measurements of three areas on
the deposition spot of this sample gives an overall average 29Si isotope fraction of
1.48(13) ppm. The average deposition rate corresponding to the three measured
areas of this sample was also lower than average at approximately 0.46 nm/min.
The background deposition pressure was similar to that of the 357 ◦C sample.
5.5.3 Samples with Deposition T > 600 ◦C
The highest residual 29Si and 30Si isotope fractions were measured in sam-
ples deposited with a substrate temperature above 600 ◦C, suggesting that higher
substrate deposition temperatures effect the enrichment. This temperature depen-
dence will be discussed in Chapter 6. SIMS measurements of a sample deposited
at approximately 705 ◦C using the low pressure mode show a maximum average
28Si isotope fraction of 99.999488(48) %. The average residual 29Si isotope frac-
tion is 3.30(25)× 10−6 (3.30(25) ppm), and the average 30Si isotope fraction is
1.82(42)× 10−6 (1.82(42) ppm). The 29Si isotope fraction averaged between two
measurements on this sample is 4.06(37) ppm. Unlike the 357 ◦C and 421 ◦C sam-
ples, this 705 ◦C sample had a higher estimated average deposition rate of approxi-
214
mately 0.74 nm/min, which is average for samples deposited at DC–3. This sample
was also deposited in a slightly lower background deposition pressure than the pre-
vious two sample. These results are counter intuitive when comparing them to
past results discussed throughout Chapter 4 and this chapter, and thus they seem
to provide supporting evidence for the substrate temperature having an affect the
enrichment.
Another similar sample with elevated 29Si and 30Si isotope fractions was de-
posited with a substrate temperature of approximately 812 ◦C using the low pres-
sure mode. The average 28Si isotope fraction measured by SIMS for this sam-
ple is 99.99907(10) %. The average residual 29Si isotope fraction in the film is
4.32(46)× 10−6 (4.32(46) ppm), and the average 30Si isotope fraction is slightly
higher at 4.96(93)× 10−6 (4.96(93) ppm). This sample was deposited in a slightly
lower background pressure than the 705 ◦C sample and had a slightly higher es-
timated deposition rate of 0.90 nm/min, which is again counter intuitive because
higher deposition rates were seen to lower the residual isotope fractions previously.
In addition to higher deposition temperatures, other differences exist between
these samples and others discussed in this chapter that may affect the measurement
of the isotope fractions. Unlike the SIMS measurements of the lower temperature
samples, the measurements of some of the samples deposited above 600 ◦C result
in only a very short range in the depth profile where the isotope fractions reach
a minimum value. A SIMS depth profile of the 812 ◦C sample discussed above is
shown in Fig. 5.12 and exhibits this effect. The 28Si (circles), 29Si (squares), and
30Si (triangles) isotope fractions are shown vs. the sputter depth into the sample.
215
Figure 5.12: SIMS depth profile of a 28Si sample deposited at 812 ◦C using the low
pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. After some surface contamination, the 29Si
and 30Si isotope fractions reach a minimum at a depth of 25 nm and then gradually
increase through the rest of the film to the natural values in the substrate. The
natural abundance values for each isotope (dotted and dashed lines) are shown
for reference. The 29Si and 30Si isotope fractions are averaged at this minimum
between 14 nm and 32 nm. The average 29Si isotope fraction at the film minimum
is 4.32 ppm, and the average 30Si isotope fraction is 4.96 ppm (dashed lines). The
interface between the film and the substrate (shaded region) is estimated to be at a
depth of 158 nm.
This depth profile is qualitatively very different from previously shown SIMS depth
profiles. After sputtering through the typical surface contamination, the 29Si and
30Si isotope fractions reach a minimum value at a depth of about 25 nm and then
gradually and immediately increase through the remaining 133 nm of deposited film
up to the natural values in the substrate (dotted and dashed lines). The interface
between the film and the substrate is estimated to be at a depth of approximately
158 nm, marked by the shaded region. The 29Si and 30Si isotope fraction averages
216
(dashed lines) were calculated for a small region where the signals reach its minimum
between 14 nm and 32 nm. SIMS depth profiles of other samples deposited at
temperatures of approximately 708 ◦C, 759 ◦C, 804 ◦C, and 1041 ◦C also exhibit
the same qualitative depth profile seen in Fig. 5.12.
The apparent gradual increase in 29Si and 30Si isotope fractions throughout the
enriched film of these samples may be explained by considering self-diffusion of 29Si
and 30Si isotopes from the naturally abundant Si substrate into the deposited film.
As the 28Si film is depositing, it is effectively being annealed by the elevated tem-
perature of the substrate. At deposition temperatures of about 700 ◦C and above,
the thermal activation of Si self-diffusion may be enough to produce the isotope
fraction gradients seen in the SIMS depth profiles. To explore this hypothesis, Si
diffusion profiles are calculated and compared to the measured SIMS depth profiles.
Si self-diffusion is believed to be dominated by self-interstitials above 800 ◦C with
an activation energy of approximately 4.75 eV [117]. The concentration of 29Si or
30Si in an enriched 28Si film, C(x), at a depth below the film surface, x, due to Si
isotope diffusion is given by
C(x) =
Csub + Cfilm
2
+
Csub − Cfilm
2
erf
(
x− d
2(DSDSi t)
0.5
)
, (5.1)
where Csub is the concentration of
29Si or 30Si in the substrate, Cfilm is the concen-
tration of 29Si or 30Si in the enriched film without diffusion, d is the film thickness,
t is the deposition time (annealing time), and the Si self-diffusion coefficient, DSDSi ,
is given by
DSDSi = (530) exp
(
− Ea
kBT
)
. (5.2)
217
Here, Ea is the activation energy, mentioned above. Equations (5.1) and (5.2)
as well as the exponential prefactor of Eq. (5.2), 530 cm2 · s−1, were taken from
Ref. [117]. These equations are used to calculate the expected concentration of
29Si from diffusion into enriched 28Si films and compare it to the measured SIMS
depth profile shown in Fig. 5.12. The calculated concentration profiles for 29Si
in a 28Si film resulting from several different deposition temperatures are shown
in Fig. 5.13. In these calculations, 29Si diffuses from the natural abundance Si
substrate (shaded region) into the 28Si film during deposition with elevated substrate
temperatures which anneal the film. The film thickness and thus film/substrate
interface is arbitrarily set at a depth of 100 nm below the film surface. Calculated
29Si concentration profiles are shown for deposition temperatures of 700 ◦C (solid
line), 750 ◦C (dotted line), 800 ◦C (dash-dot line), 850 ◦C (dash-dot-dot line), and
900 ◦C (dashed line). These calculations also used a total deposition time of 4 h,
which is an average deposition time for samples produced at DC–3, and a nominal
29Si isotope fraction in the 28Si film of 1 ppm. The 29Si profiles show that there
is relatively little diffusion expected from the substrate into the 28Si film for these
time and temperature combinations. Most of the samples deposited with substrate
temperatures within the range of these calculation were deposited at around 700 ◦C
or 800 ◦C, and the profiles for these temperatures show significant concentrations of
29Si only within approximately 1 nm of the interface. Even for the 900 ◦C profile, the
29Si concentration drops to less than double the nominal film concentration beyond
a distance of 10 nm from the substrate interface. This small amount of diffusion
would be not even be detected in the SIMS depth profiles because it is still smaller
218
Figure 5.13: Calculated 29Si concentration profiles in a 28Si film from isotope diffu-
sion at elevated temperatures. 29Si diffuses from the natural abundance Si substrate
(shaded region) at a depth of 100 nm into the 28Si film during deposition. Atomic
concentration profiles for five deposition temperatures from 700 ◦C (solid line) to
900 ◦C (dashed line). These profiles were calculated from Eq. (5.1) for a deposition
time of 4 h and a nominal 29Si isotope fraction in the 28Si film of 1 ppm.
than the typical 25 nm region over which the isotope signals transition from low
isotope fractions in the film to high isotope fractions in the substrate. Figure 5.13
shows that isotope self-diffusion from the substrate into the depositing 28Si films is
not responsible for the SIMS depth profile shape in Fig. 5.12.
Another possible explanation for the apparent gradual increase in 29Si and
30Si isotope fractions throughout the films of some samples deposited at higher
temperatures is that a large amount of surface roughness is causing an artifact
in the SIMS measurement. Producing an accurate SIMS depth profile relies on the
assumption that all of the sputtered atoms contributing to the signal at a given time
219
in the measurement originated from the same plane in the sample relative to the film-
substrate interface. The presence of surface roughness invalidates this assumption
because it results in the SIMS beam sputtering atoms from different depths at the
same time. Here, surface roughness is characterized by a surface width, ∆z, which
is the total width of the region at the surface between the highest peak and lowest
valley. This surface roughness will be transferred down through the film during the
SIMS measurement, and when the sputter beam reaches the interface between the
28Si film and the substrate, it will begin sampling the substrate in some areas while
still sampling the film in other areas. This effect artificially inflates the measured
29Si and 30Si isotope fractions near the substrate interface by mixing the higher 29Si
and 30Si signals from the substrate with the lower signals from the 28Si film. The
relationship between ∆z and the total film thickness, d, determine whether a reliable
SIMS measurement is possible. The range over which this artifact will manifest is
always similar to the size of ∆z, and it only manifests near the substrate interface.
So, for ∆z  d, the effect does not impact the ability to make a good measurement
of the film. However, for ∆z ∼ d, the signal from the measurement artifact would
be comparable or dominant to the signal of the true measurement of the film.
Surface roughness appears to be a reasonable explanation for the 28Si samples
deposited at higher temperatures that exhibit a SIMS depth profile qualitatively
similar to the one in Fig. 5.12, with a gradual 29Si and 30Si isotope fraction increase
through the depth profile of the film. Many of these samples are observed to have
much rougher surfaces as deposited than samples deposited at lower temperatures,
and for several samples, ∆z was of the order of d, as determined by SEM and
220
TEM cross-sectional microscopy. Details of this observed roughness is discussed in
a later section. A tilted SEM cross-sectional micrograph of a sample deposited with
a substrate temperature of approximately 708 ◦C using the low pressure mode is
shown in Fig. 5.14. This micrograph was acquired in collaboration with Dr. Joshua
Schumacher (NIST). The Si(100) substrate and the 28Si film both appear dark in
the lower half of this micrograph with the substrate at the bottom and the film
above it. The 28Si film is indistinguishable from from the substrate in this image
without obvious grains or other features. A dashed line represents the approximate
location of the interface between the substrate and the 28Si film, determined from
TEM cross-sectional microscopy. The light region above the film is Pt deposited
to protect the sample, which was cross-sectioned using a focused ion beam (FIB).
The top surface of the Pt is also visible because the sample is viewed at a tilt
angle of 52◦. The surface of the 28Si appears very rough with a maximum ∆z of
approximately 80 nm. The thickest areas of the film are approximately 120 nm. The
inset shows a cartoon of a rough 28Si film on a substrate being sputtered by SIMS in
an isotope measurement. SIMS sputter beams (vertical arrows) sample both a thick
and thin region of the film. Sputtering in the thick region results in isotope signals
from the 28Si film, while sputtering at the same time in the thin region results in
isotope signals partially originating from the natural abundance substrate. This is
the process that leads to the measurement artifact which inflates the 29Si and 30Si
isotope fractions above the nominal film values as the measurement approaches the
interface.
The measured 28Si enrichment values for all samples that exhibit the roughness
221
Figure 5.14: SEM tilted cross-sectional micrograph of a 28Si film deposited at 708 ◦C
at DC–3. The tilt angle is 52◦. The Si(100) substrate and 28Si film are indistin-
guishable and appear dark in the lower half of this micrograph with a dashed line
representing the film-substrate interface. The light region above the film is Pt de-
posited to protect the sample. The top surface of the Pt is visible in this tilted
image. The surface of the 28Si film appears very rough with a maximum surface
width ∆z ≈ 80 nm. The thickest areas of the film are almost 120 nm. Sputtering
this surface in a SIMS measurement introduces a measurement artifact. The inset
shows a cartoon of a rough 28Si film being sputtered by SIMS beams (arrows) in
a thick and thin region of the film. Sputtering in the thick region results in iso-
tope signals from the 28Si film, while sputtering in the thin region results in isotope
signals partially from the substrate.
induced SIMS measurement artifact need to be considered as a lower bound on the
true film enrichment. Conversely, the measured 29Si and 30Si isotope fraction for
these samples are an upper bound. This means that the isotope fractions are not
larger than the measured values, but they may be smaller due to the SIMS artifact
inflating the results. In this work, however, only SIMS measurements which are
believed to show an accurate minimum value for 29Si and 30Si isotope fractions are
reported.
222
The accuracy of the reported enrichment values of the 812 ◦C sample shown
above in Fig. 5.12 can be evaluated by comparing them to those of another sample
similar sample which does not show evidence for a large amount of surface roughness.
A sample deposited with a substrate temperature of approximately 808 ◦C using
the low pressure mode of the ion source was measured by SIMS and showed a more
typical depth profile without obvious effects from a measurement artifact. Also,
large scale roughness was not observed in a top down SEM micrograph of this sample
due to it being mostly amorphous from a higher than usual N content, which will
be discussed later. The SIMS depth profile of area of highest enrichment for this
sample is shown in Fig. 5.15. The 28Si (circles), 29Si (squares), and 30Si (triangles)
isotope fractions are shown vs. the sputter depth into the sample. Beyond the initial
signal due to some surface contamination, the 29Si and 30Si isotope fractions do show
an extended minimum between a depth of 22 nm and 81 nm. The average isotope
fraction values were calculated by averaging the data in this region. Beyond a depth
of about 81 nm, the isotope fractions begin to increase to their natural abundance
values (dotted and dashed lines) in the transition into the substrate, which is marked
by the shaded region. The interface between the film and the substrate is estimated
to be at a depth of approximately 112 nm, which also gives a value for the film
thickness.
Also shown for comparison is the 29Si isotope fraction depth profile (squares
and solid line) from the 812 ◦C sample shown above in Fig. 5.12. The depth scale
of this profile was shifted by compressing it so that the locations of the substrate
interfaces of this profile and the depth profiles of the 812 ◦C sample were aligned
223
Figure 5.15: SIMS depth profile of a 28Si sample deposited at 808 ◦C using the low
pressure mode at DC–3. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. After some surface contamination, the 29Si
and 30Si isotope fractions reach an extended minimum between a depth of 22 nm
and 81 nm where they are averaged. The average 29Si isotope fraction in the film
is 3.97 ppm, and the average 30Si isotope fraction is 2.23 ppm (dashed lines). At a
depth of 81 nm the 29Si and 30Si isotope fractions increase to their natural abundance
values in the substrate (shaded region). The natural abundance values for each
isotope (dotted and dashed lines) are also shown for reference. The interface between
the film and the substrate is estimated to be at a depth of 112 nm. For reference, the
29Si depth profile for the 812 ◦C sample (squares and solid line) is plotted showing
the effect of surface roughening on the SIMS measurement. The depth scale of this
profile was shifted to match the film thickness of the 808 ◦C sample.
with each other. Clearly, the depth profiles of the two sample are qualitatively
different. The 29Si and 30Si profiles of the 808 ◦C sample do not show signs of the
surface roughness induced SIMS measurement artifact. Unlike the 812 ◦C sample,
the 29Si and 30Si isotope fraction values of the 808 ◦C sample remain low through
most of the film thickness. These depth profiles seem to support the conclusion
that the depth profiles with elevated isotope fractions throughout the film, such
224
as those of the 812 ◦C sample, are not due to Si isotope self-diffusion at elevated
temperatures.
Despite the qualitative differences of the depth profiles of the two samples, the
minimum in 29Si isotope fractions for the two samples appears to be similar. The
average measured 28Si isotope fraction in this film is 99.999380(36) %. The average
residual 29Si isotope fraction is 3.97(31)× 10−6 (3.97(31) ppm), and the average
30Si isotope fraction is 2.23(19)× 10−6 (2.23(19) ppm) (dashed lines). A second
area of this sample was also measured giving an average 29Si isotope fraction for
the two areas on the sample of 4.33(23) ppm. The average 29Si isotope fraction in
the more highly enriched area of this sample is quite similar in value to the average
measured value of the 812 ◦C samples, which is 4.32(46) ppm, and they agree within
their uncertainties. This agreement is possibly partially coincidental because they
were deposited with slightly different deposition parameters. However, this result
does show that SIMS measurements of samples exhibiting the roughness induced
measurement artifact can give accurate enrichment values, with the caveat that
they still may be bounds on the true film enrichment values.
5.5.4 High Pressure Mode Sample
All of the previous samples discussed in this chapter were deposited using the
low pressure working mode of the ion source, but as mentioned in a previous section
discussing the deposition parameters of 28Si at DC–3, it is not clear if the benefit of
a significantly increased 28Si ion beam current produced in the high pressure mode is
offset by a lower mass resolving power. Figure 5.5 showed that the total 28Si current
225
of the high pressure mode was typically a factor of five larger than the currents
produced in the low pressure mode. An increased current is beneficial because it
allows for faster growth rates which enables production of thicker samples in a typical
deposition, and it reduces the relative adsorption rate of gaseous species into the 28Si
film, including SiH4. However, Fig. 5.5 also showed that the mass resolving power
was lower than had been observed using the low pressure mode with a measurable
ion current signal between the 28 u peak and the 29 u peaks of approximately 8 nA.
This current level may result in a lower geometric selectivity and thus lower realized
enrichments. Additionally, the high pressure mode requires higher pressures in the
ion source that result in a background pressure during deposition three times higher
than the highest pressures experienced by samples deposited using the low pressure
mode. Higher partial pressures of SiH4 may also decrease the enrichment level in
the samples.
In order to examine the viability of the high pressure mode as a source of
highly enrich 28Si and compare its effectiveness to the low pressure mode, SIMS
was used to measure the isotope fractions of a sample deposited with a substrate
temperature of approximately 421 ◦C at DC–3 using the high pressure mode of the
ion source. A SIMS depth profile of the area of highest enrichment of this sample
is shown in Fig. 5.16.
The 28Si (circles), 29Si (squares), and 30Si (triangles) isotope fractions are
shown vs. the sputter depth into the sample. Beyond an initial signal due to a
small amount of surface contamination, the 29Si and 30Si isotope fractions show an
extended minimum with very little variation in value. At a depth of around 290 nm,
226
Figure 5.16: SIMS depth profile of a 28Si sample deposited at 421 ◦C at DC–3 using
the high pressure mode. 28Si (circles), 29Si (squares), and 30Si (triangles) isotope
fractions are shown vs. sputter depth. The average 29Si isotope fraction in the
film is 0.303 ppm, and the average 30Si isotope fraction is 0.103 (dashed lines).
Many data in this region are zero counts in the SIMS cycles, especially for 30Si.
The natural abundance values for each isotope (dotted and dashed lines) are also
shown for reference. The isotope fraction values were averaged between 11 nm and
290 nm, where the 29Si and 30Si isotope fractions begin to increase to their natural
abundance values. The interface between the film and the substrate (shaded region)
is estimated to be at a depth of 315 nm.
the isotope fractions begin to increase to their natural abundance values (dotted and
dashed lines) in a fairly abruptly transition into the substrate, which is marked by
the shaded region. The interface between the film and the substrate is estimated
to be at a depth of approximately 315 nm, which also gives a value for the film
thickness.
The average measured 28Si isotope fraction in the film is 99.9999594(72) %.
The average residual 29Si isotope fraction is 3.03(58)× 10−7 (0.303(58) ppm), and
227
the average 30Si isotope fraction is 1.03(43)× 10−7 (0.103(43) ppm) (dashed lines).
These averages lie below where the data appears because many of the SIMS mea-
surement cycles in this region contained zero counts and thus the data does not
appears on this semi-log plot. This is especially true of the 30Si data. The averages
were calculated from the data between 11 nm and 290 nm. The enrichment level
of this sample is comparable to that of the most highly enriched samples deposited
at DC–3 that appear on the enrichment progression timeline in Fig. 4.2. The 29Si
isotope fraction of this sample is nearly identical to that of the 610 ◦C sample, al-
though the 30Si isotope fraction of this sample was lower, which resulted in the 28Si
isotope fraction for this 421 ◦C high pressure sample being slightly higher than that
of the 610 ◦C sample. A second area on this sample was measured by SIMS also
which resulted in an average 29Si isotope fraction between the two measurements of
0.355(41) ppm.
The 28Si ion current achieved for this sample was approximately 3.00 µA,
which is the highest for any sample in this work. This ion current, which was
only achieved using the high pressure mode, combined with a relatively small de-
position spot size of approximately 3.7 mm2 resulted in a high deposition rate of
3.94 nm/min. This high rate led to the very low 29Si and 30Si isotope fractions
measured in this sample despite the higher background pressure during deposition.
Additionally, the lowered mass resolving power and high overlap current between
peaks in the mass spectrum do not appear to reduce the realized mass selectivity
and enrichment in any significant (or at least measurable) way. This measurement
showed that the high pressure mode of the ion source is not only viable for pro-
228
ducing highly enriched 28Si films, but from the perspective of enrichment, it is the
preferred deposition mode because of the high deposition rates that are achiev-
able while maintaining extremely high levels of 28Si enrichment, and conversely,
extremely low levels of 29Si isotope fractions. These enrichments are comparable
to the most highly enriched 28Si samples produced in this work and are potentially
suitable for QI experiments such as a measurement of the dependance of electron
coherence times on 29Si concentration in the single spin regime, which was discussed
at the beginning of Chapter 4.
SIMS measurements of samples deposited at DC–3 show that overall, a re-
duction of more than a factor of five in the 29Si isotope fraction of the most highly
enriched sample was achieved compared to that of the previous most highly en-
riched sample deposited at LC–2, which can be seen in the enrichment progression
timeline in Fig. 4.2. These SIMS measurements also shows that these extremely
high enrichments are achievable both for samples deposited with elevated substrate
temperatures, and for samples deposited using the high pressure mode of the ion
source.
5.6 Epitaxial Deposition
5.6.1 Context
Epitaxial deposition of Si and other semiconductors at low temperature has
been extensively studied and characterized via molecular beam epitaxy (MBE) ex-
periments. Eaglesham et al. showed for the first time in 1990 that there exists a
229
critical thickness for epitaxy, hepi, at a given deposition temperature [118]. Beyond
this thickness, epitaxy breaks down and the film becomes amorphous. That study
found that for a deposition rate of around 0.4 nm/min and a substrate tempera-
ture of 200 ◦C, hepi = 25 nm for Si(100). At a deposition temperature of 300 ◦C,
hepi increased to 120 nm. hepi was found to increase exponentially with increasing
deposition temperature with an activation energy of about 0.4 eV, although this
value is dependent on the deposition rate. The concept of a critical epitaxial thick-
ness is only valid for these lower temperatures and typical deposition rates of a few
nm/min. Above a temperature of 500 ◦C, solid phase epitaxy (SPE) begins to dom-
inate the growth because the recrystallization front propagates faster (4.2 nm/min)
than the deposition rate, which effectively extends hepi to be infinite. This regime
is sometimes referred to as unlimited epitaxy.
Numerous research groups later studied low temperature Si epitaxy phenom-
ena, and some of these efforts were nicely summarized by Eaglesham [119]. For
low temperature MBE of Si and Ge and other similar deposition methods including
sputter deposition and ion assisted deposition (IAD), it is observed that an amor-
phous phase eventually develops after epitaxial growth. Prior to this amorphous
phase, an epitaxial but highly defective region forms in the film [109,110,120–122].
In this intermediate layer before amorphization occurs, the most apparent defects
observed by TEM are stacking faults and microtwins. The epitaxial, defective, and
amorphous regions of a Si film from Ref. [109] are seen in a cross-sectional TEM mi-
crograph in Fig. 5.17, separated by dashed lines. This Si(100) film was deposited at
270 ◦C by IAD. The bottom of the film (I) is epitaxial and defect free, then the film
230
Figure 5.17: TEM cross-sectional micrograph of a Si(100) film deposited at 270 ◦C
by IAD. Three visible regions are separated by dashed lines and include an epitaxial
region (I) above the substrate (substrate not shown), a defective region (II) approxi-
mately 80 nm thick with visible stacking faults, and an amorphous region (III) with
nanocrystallites. The arrow indicates several stacking faults on {111} planes. Insets
show magnified views of each region. (from Ref. [109])
becomes defective above that in the middle region (II). Stacking faults are visible
in this region (marked by the arrow) running along the {111} planes. The film then
becomes amorphous beyond hepi in the top region (III). The insets show magnified
views of the three regions where the crystallinity is clearly visible. The density
of stacking faults has been seen to generally increase with increasing deposition
temperature [110] until hepi becomes unlimited at higher temperatures. Another
morphological feature frequently seen in similar TEM micrographs of epitaxial thin
films is the formation of pyramidal structures bounded by the {111} stacking faults,
231
which result in a large roughness at the interface between the defective region and
the amorphous region. Locally, the transition to the amorphous region is fairly
abrupt, however.
There have been several proposed explanations for the existence of hepi and
the breakdown of an epitaxial layer into a defective layer and finally an amorphous
layer in low temperature Si deposition, which are summarized in Ref. [119]. One
model attributes the formation of the expitaxial layer to H incorporation during
deposition [123]. H then could segregate and accumulate at the growth surface
and disrupt the epitaxy by altering the nominal bonding pattern of the lattice.
The critical concentration of H for breakdown to occur is expected to be around
2× 1019 cm−3. Experimental observations of the effect of H coverage on epitaxy
seem to indicate, however, that H cannot solely be responsible for the epitaxial
breakdown in thin films [119,124] because epitaxy is possible on H terminated sur-
faces. Another explanation for the epitaxial breakdown is the accumulation of de-
fects in the depositing film until epitaxy is not sustainable [125]. This mechanism,
however, requires very large defect densities, possibly as high as 1× 1014 cm−2. Es-
timates of the density of extended defects in Si(100) put the number much lower
at around 1× 107 cm−2 [119], which seems to rule out defect buildup as the cause
of the amorphous transition. Finally, Eaglesham et al. propose that roughening
of the growth surface itself is the cause of the breakdown of an epitaxial film into
an amorphous one in low temperature Si epitaxy [126]. Roughening of the surface
during deposition may be due to several factors including the presence of impurities,
anisotropic surface diffusion, or faceting of the surface.
232
Si low temperature epitaxial deposition by MBE can be augmented using en-
ergetic (i.e. hyperthermal energy) ions with tens of eV of kinetic energy, as briefly
discussed in Chapter 1. The Si source target can be sputtered by ions to produce
a flux of Si with hyperthermal energy, or ions such as Ar or Xe can be used to
bombard the surface of the sample during Si epitaxial deposition. In addition to
ion assisted deposition, mentioned above, these techniques are also sometimes re-
ferred to as ion beam assisted deposition, ion enhanced deposition, or ion enhanced
epitaxy, but they will be collectively referred to as IAD here. These techniques use
hyperthermal energy ions to impart energy mostly in the form of momentum into
the sample during deposition, which has the effect of enhancing the epitaxial quality
of the film. Epitaxial deposition of Si and other elements has been demonstrated
using IAD with qualitatively similar results to MBE in terms of a limiting epitaxial
thickness and stacking fault formation, discussed above. However, IAD has been
shown to extend hepi to larger thicknesses for a given deposition temperature and
rate. Conversely, IAD can achieve the same hepi as MBE but at a lower tempera-
ture [109, 120, 121, 127–129]. Experimental examples of the benefits of IAD include
extending hepi to approximately 1 µm for a deposition at 300
◦C and a growth rate
of 6 nm/min [120], and lowering the temperature required for achieving unlimited
epitaxy to around 390 ◦C, even at very high deposition rates of 300 nm/min [109].
While Si IAD achieves enhanced epitaxy using a flux of hyperthermal ions of
a different element, ion beam epitaxy (IBE) achieves a similar enhancement of epi-
taxial deposition using ions of the material being deposited. This technique, which
typically generates the ions from an ion beamline, has been studied and used for de-
233
position of Si, Ge, and other materials, as discussed in Chapter 1. Research groups
studying IBE of 28Si have demonstrated epitaxial deposition of varying crystalline
quality on Si(100) with similar results to IAD. Al-Bayati et al. demonstrated IBE of
28Si with 50 eV ions [42]. Films deposited using a substrate temperature of 400 ◦C
were shown by TEM to be epitaxial but defective with stacking faults and twins.
Tsubouchi et al. used 40 eV ions and a deposition temperature of 600 ◦C [43] to
demonstrate epitaxy. TEM analysis showed that the films were epitaxial but likely
contained a high density of dislocations.
Rabalais et al. extensively studied the relationship between the epitaxial qual-
ity of 28Si films and the ion energy and substrate temperature [51]. They demon-
strated epitaxial deposition of varying quality with deposition temperatures between
40 ◦C and 290 ◦C and ion energies between 8 eV and 50 eV. RHEED and TEM
analysis showed that for 15 eV ions deposited at 160 ◦C and below, hepi was limited
to less than 15 nm. 28Si films deposited at 160 ◦C but an increased ion energy of
20 eV were found to be epitaxial with no visible stacking faults. Additionally, films
deposited with 15 eV ions but an increased substrate temperature of 290 ◦C were
also epitaxial. Their research showed that for a given deposition temperature, high-
quality epitaxial deposition occurred when using ions with energies within a certain
optimal range. The 28Si ion energy that resulted in the lowest temperature epitaxial
deposition was 20 eV.
Other IBE experiments also show optimal epitaxial deposition with ion ener-
gies around 20 eV including Matsuoka and Tohno, who observed that the highest
quality epitaxial growth occurred with 25 eV Si ions deposited at 400 ◦C, accord-
234
ing to RHEED measurements [130]. 50 eV ions produced epitaxial films, but with
stacking fault densities around 1× 1010 cm−2. Deposition using higher ion energies
produce more defective films, even at elevated deposition temperatures. Deposit-
ing Si at 740 ◦C with 200 eV ions resulted in epitaxial but highly defective films
with TEM showing stacking faults and twin structures [131]. Molecular Dynam-
ics simulations have also been used to predict a Si ion energy window for optimal
epitaxial deposition on Si(100) [132]. These simulations showed that epitaxial depo-
sition should be possible even for deposition temperatures below 200 ◦C when using
ions with energies between 20 eV and 25 eV. These theoretical results along with
the experimental results from Ref. [51] are represented on epitaxy phase diagrams
in Fig. 5.18. The quality of the epitaxy is represented in these figures by several
regions that occupy different portions of the deposition phase space of substrate
temperature, T , vs. ion energy, Ei.
Figure 5.18 (a) shows an epitaxy phase diagram from Ref. [132] for the be-
havior of Si epitaxy as predicted my molecular dynamics simulations. Regions of
unlimited epitaxy (I) and high-quality ion enhanced epitaxy (III) as well as regions
where epitaxy is defective and limited (i.e. a finite hepi) by lattice registry errors (II)
and vacancy formation (IV). Tepi is the temperature above which unlimited epitaxy
is achievable with conventional MBE. T∗epi is the temperature above which hyper-
thermal ions can enhance the films epitaxial quality, for a given deposition rate.
The phase boundary between limited and enhanced epitaxy at T∗epi occurs near Ed,
which is the energy threshold for lattice damage due to the ions and is about 25 eV
for Si deposited on Si(100). This indicates that damage caused by ions is beneficial
235
Figure 5.18: Theoretical, (a), and experimental, (b), phase diagrams for Si ion beam
epitaxy in the deposition phase space of substrate temperature, T , vs. ion energy,
Ei. (a) Si epitaxy phase diagram based on molecular dynamics simulations of the
qualitative behavior of the epitaxy for regions of unlimited (I) and high-quality ion
enhanced epitaxy (III) as well as regions of limited, defective epitaxy (II and IV).
Tepi is the temperature above which unlimited epitaxy is achievable with MBE, and
T∗epi is the temperature above which hyperthermal ions enhance epitaxy. Ed is the
energy threshold for lattice damage due to ions (from Ref. [132]). (b) Qualitative Si
epitaxy phase diagram based on experimental results. TEM and RHEED analysis of
28Si films deposited using various ion energies and substrate temperatures were used
to categorize deposition as either high-quality epitaxy (upper region) or defective
growth leading to an amorphous phase (lower region) (from Ref. [51]).
to the epitaxial quality.
Panel (b) of Fig. 5.18 shows an epitaxy phase diagram from Ref. [51] that is
similar to the one in (a) and is based on experimental results. TEM and RHEED
analysis of 28Si films deposited onto Si(100) using various ion energies and substrate
temperatures were used to define the phase boundary between high-quality and
defective epitaxy. The epitaxial deposition was categorized as either high-quality
epitaxy (upper region) or defective, limited epitaxy (lower region) with a finite hepi
leading to an amorphous phase. The boundary between the two epitaxial growth
236
modes is qualitatively similar to the one in panel (a) and shows a kink where high-
quality epitaxy is possible at a minimum substrate temperature of 160 ◦C and an
ion energy of 20 eV.
The energy deposited into the growth surface by hyperthermal ions produces
several effects beneficial to epitaxy. Ions can transfer energy to the film through neu-
tralization, which should be approximately the ionization potential, 8.15 eV [51].
This energy can excite nearby atoms and enhance their mobility. Molecular dy-
namics simulations show that hyperthermal ions can create vacancies that facilitate
adatom incorporation [133] during MBE and also suppress the formation of 3D is-
lands and step pinning from impurities [127,128]. The optimal ion energy of around
20 eV, observed to lead to higher quality epitaxy, seems to match both the critical
energy for defect formation in Si of 20 eV to 40 eV as well as an average Si displace-
ment energy of 13 eV to 15 eV [58,134,135]. Calculations show that approximately
70 % of 10 eV Si ions impinging on Si(111) surfaces penetrate about two layers into
the surface and stop in an intersticial site before diffusing to the surface to partici-
pate in the film growth. The impact and transfer of momentum from the ions leads
to the formation of dangling bonds and mobile defects such as Frenkel pairs. When
the concentration of these defects is high enough, they can facilitate ordered recrys-
tallization and epitaxial growth [136]. Simulations also show that ions with energies
greater than 50 eV begin to create more permanent defects and less of the mobile,
epitaxy-enhancing defects [51], which is supported by the experimental observations
mentioned previously.
237
5.6.2 Morphology of Films with Deposition T > 600 ◦C
28Si samples were deposited at sample location DC–3 with a range of substrate
temperatures, as was mentioned previously in this section. Initially, deposition
temperatures between 610 ◦C and 1041 ◦C were chosen in order to facilitate high-
quality epitaxial growth on Si(100) substrates. Depositing at these temperatures
using ions with a typical average ion energy Ei ≈ 33 eV for these samples should
result in a growth mode dominated by smooth layer-by-layer growth, in accordance
with the epitaxy phase diagram for IBE in Fig. 5.18 (b). Ten of these initial samples,
of which there were 12 in total, were deposited on boron-doped (University Wafer)
substrates that were not cleaned ex situ and were loaded into the vacuum chamber
with a native oxide. The final two 28Si samples deposited in this initial group were
prepared ex situ using an HF etch, and one of them was deposited on a phosphorous-
doped substrate.
5.6.2.1 RHEED
RHEED was used as an initial check on the epitaxial nature and surface mor-
phology of these samples immediately after deposition. The RHEED patterns of all
of the samples deposited with a substrate temperature above 600 ◦C showed that
they were crystalline and epitaxially aligned to the substrate due to the presence of
Si(100) diffractions spots. RHEED can only provide a limited view of the overall
epitaxial quality of the samples because it is only sensitive to the top few layers,
however, it was often used intermittently throughout the deposition to monitor the
238
Figure 5.19: RHEED diffraction pattern of a 28Si sample deposited at 708 ◦C at
DC–3. This image was acquired with the sample at the deposition temperature and
the electron beam in the 〈110〉 direction. The presence of (1×1) bulk Si diffraction
spots in this pattern indicate that the film is crystalline and aligned to the Si(100)
substrate. Additionally, this pattern of spots corresponds to a 3D transmission-type
pattern for diffraction from a rough surface. Faint (2×1) spots between some of the
(1×1) spots are visible and are likely due to part of the electron beam diffracting
from the substrate outside of the deposition spot.
crystallinity as a function of deposition thickness. The same crystalline RHEED
pattern was typically seen from the initial stages of deposition through to the end
for these samples. A typical RHEED pattern for these higher temperature 28Si sam-
ples is shown in Fig. 5.19. This sample was deposited at 708 ◦C and is about 120 nm
at its thickest. The image was acquired with the substrate at the deposition tem-
perature and the electron beam in the 〈110〉 direction. It is clear from the presence
of (1×1) bulk Si diffraction spots that the 28Si film is crystalline and aligned with
the Si(100) surface of the substrate. However, this pattern is quite different from
the RHEED pattern shown in Fig. 5.1 for a (2×1) reconstructed Si(100) surface
where each diffraction streak collapses into a point for a flat surface. This pattern
239
(Fig. 5.19) corresponds to a 3D transmission-type diffraction pattern where diffrac-
tion occurs not at rods, but at a 3D matrix of reciprocal space points. 3D diffraction
indicates that the surface of the deposited film is rough such that there is significant
transmission of the electron beam through raised features such as mounds or large
islands on the surface. Also seen in this RHEED pattern are faint (2×1) spots be-
tween some of the (1×1) spots, which are likely due to part of the RHEED electron
beam diffracting from an area of the substrate outside of the 28Si deposition spot
concurrent with diffraction from the film. A SEM cross-sectional micrograph of this
708 ◦C sample was previously shown in Fig. 5.14 where the rough surface is obvi-
ous. The surface roughness of these samples observed by RHEED and then by SEM,
which is discussed further below, is unexpected considering the phase diagrams for
smooth epitaxy in Fig. 5.18 and the comparatively high deposition temperatures
(i.e. > 600 ◦C) used for this initial set of samples.
RHEED also shows that the rough surfaces of these samples deposited with
substrate temperatures above 600 ◦C develop higher index microfacets. Figure 5.20
shows RHEED images for two 28Si samples deposited at DC–3 with clear signs of
faceting on a rough surface. The (1×1) bulk Si diffraction spots present in the
patterns of both of these samples indicate that the films are crystalline and aligned
to the Si(100) surface, similar to the sample represented in Fig. 5.19. The 3D
transmission patterns here indicate a rough surface. Figure 5.20 (a) is the diffraction
pattern for a sample deposited at 610 ◦C, and it was acquired with the sample at the
deposition temperature and the RHEED electron beam in the 〈110〉 direction. This
sample was measured to be approximately 162 nm thick. In the pattern, diffraction
240
Figure 5.20: RHEED diffraction patterns of two 28Si samples deposited at 610 ◦C,
(a), and 705 ◦C, (b), at DC–3. Both images were acquired with the samples at
the deposition temperatures and the electron beam in the 〈110〉 direction. The
(1×1) bulk Si diffraction spots present in both patterns indicate that the films
are crystalline and aligned to the Si(100) surface. They are also 3D transmission
patterns indicating a rough surface. In (a), additional lines forming a “chevron”
pattern emanate from several of the spots (solid arrows) indicating diffraction from
microfacets in the {113} family of planes. These lines are parallel to the dashed
arrow pointing from the (000) spot to the (113) spot, indicating the direction of
the facet face. In (b), additional lines can be seen connecting diffraction spots,
highlighted on the right by the dashed lines. These lines run along 〈111〉 directions,
as indicated by the arrow pointing from the (000) spot to the (111) spot, and they
are due to diffraction from {111} microfacets. Si(100) (2×1) spots are also seen in
(b) likely due to part of the electron beam diffracting from the substrate outside of
the deposition spot.
241
intensity from additional lines are observed emanating down and outward from some
of the spots, marked by the solid arrows. These lines form “chevron” patterns and
indicate diffraction from microfacets in the {113} family of planes. This classification
is evident when noting that these lines run parallel to the dashed arrow pointing
from the (000) spot to the (113) spot, indicating the plane of the facet face to be
{113}. These same {113} facet “chevron” patterns are seen on a number of other
RHEED patterns for samples deposited with deposition temperatures above 600 ◦C.
Panel (b) of Fig. 5.20 is the diffraction pattern for a sample deposited at
705 ◦C. It was acquired with the sample at the deposition temperature and the
RHEED electron beam in the 〈110〉 direction. This sample was measured to be
approximately 144 nm at its thickest. Similar to the pattern in panel (a), this
pattern shows diffraction intensity from additional lines connecting some of the
adjacent 3D diffraction spots. These lines are visible in the left side of the image and
are highlighted by the dashed lines in the right side of the image. The arrow pointing
from the (000) spot to the (111) spot indicates that these diffraction lines run along
〈111〉 directions and are due to diffraction from {111} microfacets. Superimposed
on the 3D pattern in this image is the nominal Si(100) (2×1) spot and rod pattern,
which is likely due to part of the electron beam diffracting from an area of the
substrate outside of the 28Si deposition spot. It should be noted that this sample
was actually deposited later and with a different preparation procedure than the
other high deposition temperature samples discussed so far in this section, but the
RHEED pattern of this sample is presented here because it is more illustrative of
{111} faceting than those of other samples. The presence of microfacets and their
242
orientation on the surfaces of these 28Si films are important pieces of information
that provide insight on the growth mechanisms leading to the unexpected surface
roughness.
5.6.2.2 STM
After deposition, the surface morphology of these samples was investigated
by in situ STM and ex situ SEM as a primary means of evaluating the quality of
the epitaxial growth as well as confirming and measuring the extent of the surface
roughness observed by RHEED. Figure 5.21 shows an example of a filled state STM
topography image of a 28Si sample deposited at approximately 708 ◦C at DC–3,
which had a RHEED pattern corresponding to a rough surface. The 28Si film was
measured by TEM to be 155 nm at the thickest. The roughness indicated in the
RHEED pattern of this and similar samples is supported by the presence of large
grain-like features visible in this micrograph. These features are approximately
200 nm wide and at least 1 µm long running diagonally from the bottom left to
the top right of the image. A measure of the roughness of this surface is given by
the total peak-to-valley surface width, i.e. the difference in height values between
the highest peak and lowest valley in the STM topography. This surface width
in this image is ∆z ≈ 12.8 nm. Other measurements of the value of the surface
width for this sample including those from a SEM cross-sectional micrograph show
that it is as much as 60 nm. The STM derived value may be smaller because the
STM tip is too large to fit in between the valleys seen in the image and thus it
243
Figure 5.21: STM topography filled state image of a 28Si sample deposited at 708 ◦C
at DC–3. The image was acquired with a tip bias ≈ -1.8 V and a tunneling current
≈ 100 pA. This film is about 155 nm at its thickest and was deposited on a Si(100)
substrate with no ex situ cleaning. Large grain-like features ≈ 200 nm wide are
visible on the surface running diagonally from the bottom left to the top right of
the image. The surface width determined from the topography is ∆z ≈ 12.8 nm.
cannot accurately measure the full range between valley and peak. Alternatively, it
may be that the film is thinner in the region scanned by the STM. The difficulty
STM has with accurately measuring the topography of very rough surfaces limits
its usefulness compared to SEM for evaluating the morphology of these samples.
5.6.2.3 SEM
SEM was used ex situ to survey the surface morphology of numerous 28Si
samples deposited with a substrate deposition temperature above 600 ◦C at DC–3.
SEM images presented in this section were acquired in collaboration with Dr. Joshua
244
Figure 5.22: SEM tilted micrograph of the surface of a 28Si film deposited at 708 ◦C
at DC–3. The tilt angle of this micrograph is 60◦. The substrate used for this sample
was not cleaned ex situ and was loaded into the chamber with a native oxide. This
film is almost 120 nm at its thickest. The surface morphology of the deposited
film appears to be extremely rough with mounds or grain-like features that have an
average length ≈ 440 nm. The longer side of some of the mounds appear to run
left-to-right in the image indicating orientation with a 〈110〉 direction.
Schumacher (NIST) and Dr. Vladimir Oleshko (NIST). A tilted SEM micrograph of
one of the first 28Si samples deposited at DC–3 is shown in Fig. 5.22. The tilt angle
of the image is 60◦. The substrate used for this deposition was boron-doped and
was not cleaned ex situ before being loaded into the vacuum chamber with a native
oxide and being prepared in situ in the usual manner. This sample was deposited
with a substrate temperature of approximately 708 ◦C and is almost 120 nm at its
thickest, as measured in cross-section. A SEM cross-sectional micrograph of this
sample was previously shown in Fig. 5.14. It is obvious from Fig. 5.22 that the
surface morphology of this sample is extremely rough. The surface of the film is
245
covered with large mounds or grain-like features. The rough surface indicated by
the RHEED image (Fig. 5.19) is clearly confirmed and understood to be due to
these tall mounds visible in the micrograph. The tilt of the image makes it difficult
to estimate the height of the mounds and their size in the direction running top-
to-bottom in the image, but their estimated average size in the direction running
left-to-right is approximately 440 nm. The longer side of some of the mounds appear
to directly run left-to-right in the image possibly indicating orientation with a 〈110〉
direction. The crystallographic directions in SEM micrographs presented here are
determined from the positioning of the samples in the SEM. This orientation of
mounds or grains on the surface of the films is consistent with the indication from
the RHEED images (Fig. 5.20) that {111} and {113} microfacets are present on the
surface.
All of the 28Si samples produced in this initial batch with substrate tem-
peratures above 600 ◦C which were inspected with SEM showed similarly rough
surfaces with mound formation, although with slightly varying morphologies. SEM
micrographs of six of these samples deposited at various temperatures are shown
in Fig. 5.23. The deposition temperatures for the samples in panels (a)–(f) were
610 ◦C, 708 ◦C, 804 ◦C, 812 ◦C, 920 ◦C, and 1041 ◦C respectively. All substrates
used for these samples were boron-doped except for the sample in panel (c), which
was phosphorous-doped. No ex situ cleaning was performed on the substrates used
for these samples except for the sample in panel (c), which was etched with HF prior
to being loaded into the vacuum chamber. Large mounds are visible on the surface
of the 610 ◦C sample in the top-down micrograph in panel (a). The thickness of
246
Figure 5.23: SEM top-down micrographs of the surface morphology of six 28Si films
deposited at DC–3 at 610 ◦C, (a), 708 ◦C, (b), 804 ◦C, (c), 812 ◦C, (d), 920 ◦C, (e),
and 1041 ◦C, (f). No ex situ cleaning was performed on these substrates except for
(c), which was etched with HF. (a) Mounds are apparent on a very rough surface
with an average length ≈ 320 nm and faceted sides oriented top-to-bottom or left-to-
right in 〈110〉 directions. (b) Irregularly shaped mound features are apparent with
an average size ≈ 450 nm. (c) Smaller grains are apparent with an average length
≈ 180 nm. (d) Mounds are visible with an average length ≈ 780 nm and align in
rows running left-to-right in a 〈110〉 direction. The larger grains in (d) compared to
(c) may be due to the different cleaning procedures, despite similar temperatures.
(e) 60◦ tilted image of hut-like mounds with faceted sides and an average length ≈
1105 nm. (f) Several hut-like mounds are apparent on a smooth surface with sizes
between ≈ 200 nm and 500 nm. These mounds have faceted sides oriented in 〈110〉
directions.
247
this sample measured by SIMS depth profiling was determined to be approximately
162 nm. The mounds appear generally elongated and are measured to have an aver-
age length ≈ 320 nm and an average width ≈ 175 nm. The length of the mounds is
defined as the longer dimension here. The length and width of mounds in the SEM
micrographs are determined from software analysis of the images using an autocorre-
lation function for finding repeating patterns. The values determined from multiple
images were then averaged together to produce the values reported here, which have
a 10 % relative uncertainty. Many of the mounds in panel (a) appear to be aligned
in similar directions. They also appear to have faceted sides where the edges of
the microfacets run top-to-bottom or left-to-right in the image indicating orienta-
tion with 〈110〉 directions. This observation of the presence of microfacets confirms
the indications from the RHEED images (Fig. 5.20). The observed orientation is
consistent with the microfacets being on {111} and {113} planes.
Panel (b) in Fig. 5.23 is a top-down micrograph showing the morphology of the
surface of the 708 ◦C sample, which consists of irregularly shaped mound features
with an average size ≈ 450 nm. The maximum thickness determined from SIMS
depth profiles of this sample is approximately 126 nm. Note that the area of panel
(b) is roughly four times larger than the area shown in (a). The mounds in panel
(b) do not seem to be oriented in any particular direction. The surface of the 804 ◦C
sample is shown in the top-down micrograph in panel (c). Smaller grain-like features
are seen covering the surface of the film. The scale of panel (c) is the same as that of
(a) and the average length of the grains in (c) is ≈ 180 nm. As mentioned above, the
substrate used for the 804 ◦C sample in panel (c) was the only one in Fig. 5.23 which
248
was etched with HF. Panel (d) shows a top-down micrograph of a sample deposited
at 812 ◦C, which has a maximum thickness determined from a SIMS depth profile
to be approximately 158 nm. Despite the similar deposition temperature to that
used for the sample in panel (c), the surface morphology of this sample is clearly
qualitatively different, possibly due to the different cleaning procedure. Much larger
mounds are seen on the surface, the edges of which are not as well defined as the
features of the other samples in the other micrographs. The mounds on this sample
appear to run together much more giving the surface more of a wavy appearance as
opposed to the granular appearance of the other micrographs. This could indicate
that the surface of this film is smoother than the others, however, the SIMS depth
profile of this sample (Fig. 5.12) showed the characteristic measurement artifact for
rough surfaces of a gradual increase in minor isotopes throughout the profile, as
determined in a previous section in this chapter. Although the boundaries of the
mounds in panel (d) are not as well defined as in the other micrographs in Fig. 5.23,
the autocorrelation of this image gives an average width ≈ 250 nm and an average
length ≈ 780 nm. These mounds also appear to approximately align in rows running
left-to-right in the image, which is consistent with alignment to a 〈110〉 direction.
Figure 5.23 (e) is a 60◦ tilted micrograph showing the surface morphology of a
sample deposited at 920 ◦C. Many hut-like mounds are visible on the surface with
an average width ≈ 560 nm and an average length ≈ 1105 nm. These measurements
were corrected for the tilt of the image. The thickness of this film was not directly
measured, but the height of some of the mounds is estimated from the image to
be roughly 400 nm. Many of these mounds do appear to have faceted sides giving
249
them the hut-like shape, and the long direction of several of them are clearly aligned
either top-to-bottom or left-to-right in the image. It can be reasonably presumed
that like in several of the other SEM micrographs shown here, the mounds in panel
(e) are aligned to a 〈110〉 direction, although this cannot be confirmed because the
positioning of the sample inside the SEM is not known. Finally, panel (f) is a
top-down micrograph showing the surface morphology of the 1041 ◦C sample. The
surface of this sample appears very different from the other samples shown here in
that the film does not appear to be continuous. Several hut-like mounds are seen
on an otherwise smooth surface. These mounds range in size from approximately
200 nm to 500 nm and were measured to be about 50 nm tall from a SIMS depth
profile and TEM cross-sectional imaging. The mounds have clearly faceted sides
with edges running from top-to-bottom and left-to-right in the image, indicating
orientation with 〈110〉 directions, as expected for {111} and {113} microfacets.
5.6.2.4 Step Pinning Induced Roughness
The results presented in Fig. 5.23 are not only unexpected because they show
that 28Si deposition at temperatures above 600 ◦C yields films with very rough
surfaces, but also because despite depositing over a range of temperatures up to
1041 ◦C, the surface roughness not only persists but appears to increase at higher
temperatures. Generally, in epitaxial deposition and growth, including Si MBE and
chemical vapor deposition (CVD), higher deposition temperatures result in higher
quality epitaxy and smoother films. Si CVD commonly uses deposition temperatures
250
above 600 ◦C or even in excess of 1100 ◦C both to achieve high growth rates from
decomposition from silane gas and to facilitate high-quality epitaxial growth of thin
films [137]. For Si MBE, above a deposition temperature of approximately 500 ◦C,
SPE is expected to dominate the growth [118], and IAD has been shown to produce
smooth, unlimited epitaxy at a deposition temperature as low as 390 ◦C [109]. As
mentioned previously, the epitaxy phase diagrams for IBE in Fig. 5.18 also predict
that higher deposition temperatures should lead to higher quality and unlimited
epitaxy, and IBE experiments have demonstrated epitaxial deposition of 28Si films
dominated by smooth, layer-by-layer growth [51]. Smooth, epitaxial growth pro-
ceeds in these cases because the dominant growth mode at higher temperatures is
2D layer-by-layer growth. The dominant growth for a given deposition flux is con-
trolled by the surface diffusivity, which is a thermally activated process. A high
surface diffusivity-to-flux ratio ideally leads 2D layer-by-layer or step flow growth
producing a smooth surface.
An explanation for the rough morphology characterized by large mounds seen
on the 28Si samples deposited at high temperature is presented here and is based
on the presence of contaminants interfering with smooth deposition. Contaminants
such as SiC or SiO2 can act as pinning sites for step movement during deposition in
a layer-by-layer growth mode at elevated temperatures. This leads to the formation
of pits in the growth surface, which are a manifestation of step pinning in the early
stages of thin film deposition. The continued movement of new steps around a pit
will lead to local step bunching, and inevitably a high enough step density will form
into larger microfacets in the film surface such as {111} and {113} microfacets. The
251
presence of strain in the substrate can also lead to increased roughness and step
bunching after thermal processing. Flash annealing strained SOI (sSOI) substrates
to between 900 ◦C and 1110 ◦C has been shown to produce a rough crosshatch
pattern of step bunching on the surface [138]. For nominally unstrained substrates,
strain can arise from the mounting of the chip in the sample holder. Strain could also
arise in a deposited 28Si film due to a small lattice constant mismatch between 28Si
and a natural abundance Si substrate. The lattice constant of 28Si is larger than that
of natural abundance Si by a relative value of roughly 1× 10−6 [29, 139, 140]. The
strain due to this difference is quite small compared to the typical strain of a sSOI
wafer of approximately 1 %, and so is unlikely to result in a large amount of surface
roughening. However, any lattice constant mismatch between a film and substrate
will still need to be accounted for by the development of dislocation in the film.
As step bunching and microfacets build up on the surface of a film, the roughness
increases, and they will come to dominate the growth and morphology of the film.
This process leads to the formation of mounds as raised microfacets meet forming
larger structures, similar to the faceted mounds seen in the SEM micrographs of
28Si samples deposited above 600 ◦C.
Numerous groups have studied the effects of roughness and faceting on the
critical thickness, hepi, and epitaxial quality of films produced by Si MBE, and
IBE [109, 110, 119, 120, 122]. hepi is found to be smallest on Si(111) surfaces, larger
on Si(113) surfaces, and significantly larger on Si(100) surfaces [109]. These studies
typically focus on the role of {111} microfacets in defect formation and transition
of the growth from epitaxial to amorphous. While they do not directly address
252
the crystalline growth of mounds on a film surface, they do show the critical effect
microfacets can have on the evolution of a depositing film and the important link
between {111} planes and defects. General roughness on a Si(100) surface is known
to develop into {111} microfacets [119], and both {111} and {113} microfacets
are commonly observed in low temperature Si epitaxy [122]. {111} planes have
been shown to form more easily on smaller terraces on and around islands, and
here it may be that the decreased terrace sizes at step bunches have the same
effect. Furthermore, incomplete filling of lattice sites by adatoms on step bunches
or existing {111} microfacets can lead to the growth of {113} or {115} planes during
deposition [110]. Thus, it is no surprise that {111} and {113}microfacets were found
on the 28Si samples discussed above.
One study showed that C contamination on a Si(100) surface can directly lead
to the formation of {113} microfacets after annealing the sample to above 950 ◦C.
This study concluded that impurities can effect formation of faceting, and that SiC,
for example, can act as a pinning site to step motion [141]. Moreover, once pinning
sites are formed from trace amounts of SiC, other C may migrate to the pinning
site and form larger, more stable clusters. In general, C and O impurities can cause
defects in a film during deposition that lead to pinning sites [122]. Experiments of Si
growth confined in bare opening of a patterned SiO2 surface layer have shown that
near boundaries, similar to step bunches but in this case the edge of the SiO2 pattern,
{113} and sometimes {111} develop from the formation of rebonded double steps
(DB) in the Si(100) surface [142,143]. Once {111} and {113} planes are established
in the growth surface, they tend to endure and subsequently expand because they
253
are more stable, i.e. energetically favorable, than the Si(100) surface. The Si(111)
surface is known to be the most stable with the lowest Si surface energy, and although
the Si(113) surface has a higher surface energy than Si(100), it becomes stable when
formed by the DB steps [143]. These surfaces are also more stable with the presence
of C contamination.
In addition to the {111} and {113} microfacets being highly stable, molecular
dynamics simulations show that the surface diffusion constant of surfaces such as
Si(111) is smaller than for Si(100) [144]. Similarly, total-energy calculations predict
that adatoms diffusing on a {113} surface encounter energy barriers at the rebonded
double steps which increases the activation energy for adatom diffusion and a lower
reactivity at the DB steps [143]. The result of this may be that the growth rate of
new layers on the Si(100) surfaces is higher because adatoms on the {111} and {113}
surfaces will preferentially hop to a neighboring {100} surface where movement is
easier and they are more likely to diffuse away from the facet boundary. The more
stable {111} and {113} surfaces may also inhibit new layer formation compared to
a {100} surface. A higher growth rate of the {100} surface would actually result
in the shrinking of that surface and simultaneous expansion of adjacent microfacet
surfaces as each new {100} layer encounters the {111} or {113} edges and adds to
them, expanding those surfaces. The {111} or {113} microfacets would grow in
time and come to dominate the surface morphology as the {100} surfaces shrink,
producing faceted mounds like those observed on the 28Si films deposited at high
temperature in this work. A similar explanation was used to describe the preferential
growth of {113} microfacets on Si(100) surfaces during Si CVD, where the {113}
254
surface has a slower CVD epitaxial growth rate than the {100} surface [142].
To better understand the mechanisms driving the observed rough morphol-
ogy, experiments were carried out to explore the different aspects of the proposed
sequence for mound formation. The development of step bunching and roughness
around pinning sites on a 28Si sample can be observed by inspecting a film thin
enough to exhibit the initial formation of these features in the form of pits on the
surface. A STM topography filled state image of a 28Si film deposited with a sub-
strate temperature of around 709 ◦C at DC–3 which exhibits pit formation is shown
in Fig. 5.24. The substrate used for this sample was prepared ex situ by an HF
etch. The film is estimated to be approximately 10 nm thick based on the ion flux
at the center of the deposition spot, although there is a large uncertainty on this
value. Thickness estimates of particular regions of very thin films in STM images
presented here may not be accurate because the ion flux across a deposition spot is
not uniform and the position of the STM scan area was not well known relative to
the center of the deposition spot.
A large area scan of the deposited film is shown in panel (a) of Fig. 5.24 with
a large number of dark, round, pit-like features in the surface. Large, flat, single
terraces roughly 100 nm wide appear bounded by the pits that act as pinning sites to
step movement and create step bunching. The single terraces between pits indicate
a predominantly layer-by-layer growth mode. Multiple bright spots are seen within
and around the pits and are probably clusters of chemical contaminants such as
SiC, SiO, or Si3N4. The average areal pit density of this sample is determined to
be 340 µm−2 ± 18 µm−2, and the total height scale of this image is about 2.1 nm.
255
Figure 5.24: STM topography filled state images of a 28Si film deposited at 709 ◦C
at DC–3 showing pits. The images were acquired with a tip bias ≈ -1.8 V and a
tunneling current ≈ 100 pA. This film was deposited after an ex situ HF etch of
the substrate, and it is estimated to be 10 nm thick. (a) Large area scan of the film
showing dark pits in the surface. Bounded by the pits and step bunching are large,
flat, single terraces about 100 nm wide. Bright spots are seen inside the pits and
are probably clusters of chemical contaminants, possibly SiC. The average areal pit
density of this sample is ≈ 340 µm−2 ± 18 µm−2. The total height scale of this
image is about 2.1 nm. (b) Small area scan of a pit on this sample. Numerous
contaminant clusters are seen in and near the pit and step bunching can be seen
around the pit, resulting in increased roughness. The height scale is about 1.5 nm.
Si (2×1) dimer rows can be seen on several terraces.
256
Panel (b) is a small area scan of a region near a pit from another area of this 28Si
sample showing numerous contaminant clusters in and near the pit. Step bunching
resulting from the pinning of steps by the clusters can be seen, resulting in increased
roughness. The total height scale of panel (b) is about 1.5 nm, which is mostly
accounted for by the apparent height of the clusters themselves. Si(100) (2×1)
dimer rows can be seen on several terraces surrounding the pit in this image.
To determine if the step pinning and pit formation in the initial stages of
deposition is related to certain aspects of the ion beam deposition process, a thin
natural abundance Si (natSi) film was deposited using the Si electron beam evapo-
rator (i.e. the EFM). Aspects of the ion beam excluded from the natSi deposition
which may affect the growth include the hyperthermal energy ions, the presence of
chemical contaminants in the ion beam that are transported ballistically as ions to
the sample, and the presence of SiH4 or other gases which diffuse from the ion beam
chamber during deposition. STM topography filled state images of a natSi thin film
deposited with a substrate temperature of approximately 713 ◦C at DC–3 are shown
in Fig. 5.25. Numerous dark pits are apparent on this sample, similar to the pitting
on the 28Si sample in Fig. 5.24. Panel (a) of Fig. 5.25 is a large area scan of the
film, which is estimated to be approximately 13 nm thick. Note that the scan area
of panel (a) is about four times larger than the displayed area of the 28Si film in
Fig. 5.24 (a). The pits in the natSi film are more ordered and square shaped than
the pits seen in the 28Si film. Bounded by the pits and step bunching are large,
flat, single terraces roughly 200 nm wide. The presence of these single terraces be-
tween pits indicates a predominantly layer-by-layer growth mode. Clearly the pits
257
Figure 5.25: STM topography filled state images of a natural abundance Si film
deposited at 713 ◦C at DC–3 showing pits. These images were acquired with a tip
bias ≈ -2 V and a tunneling current ≈ 150 pA. This film was deposited from the
natural abundance Si source after an ex situ HF etch of the substrate, and it is
estimated to be 13 nm thick. (a) Large area scan showing dark, square pits in the
film bordering large, flat, single terraces about 200 nm wide. Step bunching, seen
around the pits, results in increased surface roughness. Contaminant clusters are
seen as bright spots in several pits. The average areal pit density of this sample is
≈ 40 µm−2 ± 6 µm−2. The total height scale is about 5.6 nm. (b) Small area scan
of a pit in (a). The pit resembles an inverted pyramid with sides aligned with the
〈110〉 directions. A contaminant cluster, possibly SiC, can be seen at the bottom of
the pit with a height scale of about 3 nm. Si(100) (2×1) dimer rows can be seen on
the terraces.
258
are located at step pinning sites where step bunching during deposition has led to
increased roughness. The total height scale of this image is approximately 5.6 nm,
which is accounted for mostly by the pits and gives a reference for their depth.
Several of the pits in panel (a) contain brighter features at their center which are
probably contaminant clusters and possibly SiC. The average areal pit density of
this sample is determined to be 40 µm−2 ± 6 µm−2. This pit density is almost nine
times lower than that of the 28Si sample in Fig. 5.24 indicating that the roughness
observed in thicker films may be partially due to some factor inherent to the ion
beam deposition process. Clearly, though, the presence of any pits on the natSi film
shows that other factors not specific to the deposition sources also contribute to
development of pinning sites and roughness on these films.
Panel (b) in Fig. 5.25 is a small area scan of a pit in (a). The pit resembles
an inverted pyramid in the surface and is larger than the pit of the 28Si film in
Fig. 5.24 (b). Note that the scan area is approximately one quarter of the scan area
of image showing the natSi film in Fig. 5.25 (b). The sides of the pit can be seen to
align parallel or perpendicular with the Si (2×1) dimer rows that are visible on the
terraces surrounding the pit, i.e. the pit sides are aligned with 〈110〉 directions. A
contaminant cluster, possibly SiC, can be seen at the bottom of the pit in panel (b)
and is likely the source of the step pinning that formed this pit. The total height
scale of this image is about 3 nm from the bottom of the pit to the surrounding
terraces, making the pit quite shallow compared to its lateral size. The pits in both
the 28Si and natSi films as well as in other thin samples have a variety of depths.
This indicates that while some pits likely originate at the substrate surface, other
259
pits form during the deposition at different depths in the films. This indicates that
the deposition process itself can generate more pits.
It seems likely that the step pinning and pit formation observed in both the
thin 28Si (Fig. 5.24) and natSi (Fig. 5.25) films do lead to increased surface rough-
ness and eventually large scale mound formation for deposition occurring at higher
temperatures, i.e. above 600 ◦C. The step bunching and terrace formations that
develop around pinning sites and pits qualitatively resemble the shape, distribution,
and orientation of the mounds observed in SEM micrographs of the thicker 28Si
films (Fig. 5.23). These STM images suggest that mounds may form from the con-
tinued growth of the large flat terraces which separate from other nearby terraces by
groupings of pits where further growth is inhibited. Groups of pits may then form
the valleys between the mounds observed in SEM. It is also clear from the STM
image of the natSi film in Fig. 5.24 (a) and, to a lesser extent, the image showing the
28Si film in Fig. 5.25 (a) that the edges of the large terraces are aligned with 〈110〉
directions forming rectangular sections. This is consistent with the observation that
many of the large mounds visible in the SEM micrographs of 28Si samples have edges
oriented in 〈110〉 directions.
The features observed inside pits in STM images of both thin 28Si and natSi
films, which are presumed to be clusters of contaminants, likely act as nucleation
sites for the formation of the pits by pinning step motion. Several different types of
contaminants may play a role in step pinning during deposition, such as SiC. It is
important to identify contaminants in this system in order to better understand and
ultimately ameliorate the cause of mound formation and roughness on 28Si films.
260
Carbon contamination on substrates can originate from a number of sources
including the ambient environment before the substrate is loaded into the chamber,
carbon-containing adsorbates produced inside the vacuum chamber, or even ex situ
chemical cleaning procedures not targeted to remove organic compounds, such as HF
etching. SiC contamination in the form of clusters can be observed on the surface of
substrates directly by STM imaging, or it can be detected via the RHEED pattern of
substrates during in situ preparation by flash annealing. While the STM can observe
individual SiC clusters in very small areas on the substrate, observing the signature
of SiC in RHEED generally requires a significant amount of contamination over a
large area. SiC is indeed found to be present on substrates used for samples deposited
at DC–3 and is most often detected by RHEED during the flashing process. An
example of a RHEED image of a flashed Si(100) substrate displaying the signature
of SiC contamination is shown in Fig. 5.26. This substrate was prepared ex situ by
an HF etch. The RHEED image in panel (a) was acquired with the sample at 600 ◦C
and the electron beam in the 〈110〉 direction. A typical Si(100) (2×1) diffraction
pattern is seen consisting mainly of the five central streaks as well as bulk Si and
reconstruction spots indicated by the two inner arrows. Superimposed on the Si
pattern in this image is a SiC pattern. This pattern first appeared after flashing
the substrate to approximately 1040 ◦C for 20 s. Additional diffraction spots are
also visible just outside the bulk Si streaks, which correspond to SiC as indicated
by the two outer arrows. Other spots comprising the SiC pattern are visible on the
central rod as well as the far left and right edges of the image. This SiC pattern
is a transmission-type pattern, which indicates the SiC exists on the surface in 3D
261
Figure 5.26: RHEED diffraction pattern of a Si(100) substrate after an ex situ HF
etch and in situ flash annealing, which shows SiC contamination. (a) The RHEED
image was acquired with the sample at 600 ◦C and the electron beam in the 〈110〉
direction. A Si(100) (2×1) diffraction pattern is seen consisting of streaks and
bulk Si and reconstruction spots indicated by the two inner arrows. SiC diffraction
spots are also visible outside the bulk Si streaks, indicated by the outer arrows,
as well as on the central Si streak and the far left and right edges of the image.
The transmission-type pattern of the SiC spots indicates 3D clusters. (b) A line
profile of the diffraction intensity in arbitrary units is shown corresponding to the
dashed line in (a). The reciprocal space mapping of the profile was calibrated from
the separation of the bulk Si spots (dotted lines). The separation of the SiC spots
(dashed lines) is then calculated to be about 6.5 1/nm, corresponding to the 3C-SiC
lattice constant.
262
clusters aligned to the Si(100) surface.
Panel (b) of Fig. 5.26 shows a line profile of the RHEED diffraction intensity
in arbitrary units corresponding to the horizontal dashed line in (a). This profile
is plotted in reciprocal space in order to deduce the lattice constant of the SiC
clusters from the spacing of the diffraction spots. The reciprocal space mapping
was calibrated using the known separation of the bulk Si spots, indicated by the
dotted lines in panel (b). In reciprocal space, these spots should be separated by
approximately 5.2 1/nm, which is calculated from the distance between Si lattice
planes in the 〈110〉 direction and the Si lattice constant, a0 = 0.543 nm [145]. The
separation of the outer SiC spots, indicated by the dashed lines is then calculated to
be approximately 6.5 1/nm. Taking half of this value gives a distance of 3.25 1/nm
between one SiC spot and the central streak, which is difficult to measure directly
in this image. Then, inverting this value gives a real space distance between lattice
planes of the SiC cluster in the 〈110〉 direction of 0.31 nm. Multiplying this spacing
by 2/
√
2 gives the 〈100〉 lattice constant, 0.44 nm. This value agrees quite well with
the lattice constant of the 3C-SiC polytype, a0 = 0.436 nm [146]. 3C-SiC has a
zincblende crystal structure making it able to align to the Si diamond cubic lattice,
and this alignment is clear in the RHEED pattern of Fig. 5.26.
SiC diffraction spots in the RHEED patterns of substrates were commonly
seen during the flash annealing process of many samples in this work. This preva-
lence highlights the need for ex situ cleaning procedures to mitigate C and other
contaminants. Typically, indications of SiC appear after exceeding a temperature
of about 1000 ◦C, which is usually the minimum temperature of the initial flash in
263
the flashing sequence after keeping the substrate at about 600 ◦C or below (see the
in situ substrate preparation discussion in an previous section of this chapter). This
observation matches well with the known temperatures associated with SiC cluster
formation on Si surfaces [119]. SiC CVD experiments have shown that stoichiomet-
ric SiC begins to form on Si(100) at substrate temperatures above 700 ◦C [147], and
annealing C60 films on Si(100) substrates produces crystalline SiC between 800
◦C
and 900 ◦C [148]. Other experiments studying SiC contamination during vacuum
preparation of Si surfaces have seen similar results to those observed in this work.
Becker et al. used RHEED to demonstrate that SiC typically takes the form 3C-SiC
when clusters form on Si(100) and Si(111) surfaces [149]. Samples were cleaned ex
situ using HF and flashed in situ to between 800 ◦C and 1000 ◦C with background
pressures as high as 1.3× 10−5 Pa (1.0× 10−7 Torr) resulting in SiC contamination.
This study concluded that the contaminants originated from carbon-containing ad-
sorbates from the vacuum chamber which dissociate to release C during flashing,
including at lower pressures. Similarly, Henderson et al. showed that 3C-SiC forms
on Si(111) surfaces above temperatures of 800 ◦C and likely originates from both
carbon-containing adsorbates and the ex situ chemical cleaning procedure, which
included an HF etch [150]. These results indicate that SiC formation is not only
likely due to the flashing process, but that it possibly forms throughout a deposition
occurring at higher temperatures such as the ones described here above 600 ◦C.
Further, experiments by Mol et al. showed that pre-treating Si(100) substrates
with HF can directly result in SiC contamination [112]. Figure 5.27 shows a STM
topography filled state image from that work of a Si(100) substrate prepared ex
264
Figure 5.27: STM topography filled state image of a Si(100) substrate which was
prepared ex situ by an HF etch and in situ by a 900 ◦C anneal. The image was
acquired with a tip bias of -1.5 V and a tunneling current of 100 pA. Bright SiC
clusters about 15 nm across are seen covering the (2×1) reconstructed surface. Step
bunching is apparent in the upper left and lower right of the image. (from Ref. [112])
situ by an HF etch. The sample was loaded into the vacuum system with just the
H passivation layer from the etch before being annealed at approximately 900 ◦C.
Several bright SiC clusters approximately 15 nm wide can be seen on the annealed
surface. Below the clusters can be seen (2×1) dimer rows on short terraces with
steps appearing highly bunched in the upper right and lower left of the image, likely
due to the presence of the SiC. Another sample in this study was prepared with
a protective oxide layer resulting from the oxidation during ex situ cleaning with
HNO3. After the same anneal to 900
◦C to remove the oxide, the surface was found
to be flat and free of contaminants. Another group has shown that IBE deposition
of 28Si onto a substrate treated with HF produced a defective film with stacking
265
faults present [42].
Some of these previous experimental results regarding SiC discussed above
were reproduced in this current work. Approximately 90 % of substrates prepared ex
situ with an HF etch in this work showed signs of SiC contamination in the RHEED
pattern after initial flashes. Most of these patterns were completely dominated by
the SiC diffraction spots and showed only weak bulk Si spots or streaks, unlike
the more balanced combination of patterns observed in Fig. 5.26. This suggests
that the contamination was such that the surfaces of these substrates were mostly
if not completely covered in SiC clusters. By contrast, approximately 32 % of the
substrates which had no ex situ cleaning and were loaded into the chamber with a
native oxide showed signs of SiC in the RHEED pattern. Furthermore, the RHEED
patterns of these substrates typically only showed weak SiC diffraction spots with
much stronger Si (2×1) spots, likely indicating only trace amounts of SiC on the
surface. Another potential issue with cleaning substrates using only HF is that F
atoms in the solution or left behind as residue can etch the Si surface, leading to
roughening and potential pinning sites in film growth [150,151]. These observations
of the prevalence of SiC on substrates as well as the results of the experiment
shown in Fig. 5.27 highlight both the potential benefit of alternate chemical cleaning
combined with a protective oxide layer on substrates before in situ preparation as
well as the ineffectiveness of HF etches alone to mitigate contaminants.
After the appearance of SiC on the surface of prepared substrates, further heat
treatments in the form of higher temperature flash annealing typically results in the
disappearance of the SiC diffraction pattern. For almost all of the samples where a
266
RHEED diffraction pattern corresponding to SiC appears after initial flashing to at
least 1000 ◦C, flashing to between approximately 1150 ◦C and 1200 ◦C was required
to remove the SiC spots and recover a nominal Si (2×1) pattern. Experimental
results from other groups support this observation showing that SiC can be re-
moved from Si(100) and Si(111) surfaces only by heating it to between 1100 ◦C and
1200 ◦C [150,152]. It is also suggested that upon heating to these temperatures, the
C from the clusters goes into solution in the Si substrate, although it is not known
how this C may affect the growth of a film.
After RHEED is used to identify substrates with SiC contamination, STM
imaging is used to directly observe and inspect the SiC clusters to confirm their
presence and view their effect on the substrate. Additionally, STM inspection of
nominally clean substrates can reveal small contaminant clusters or other partic-
ulates in trace amounts below the detection capability of RHEED. These trace
contaminants may still result in pinning sites during film growth. Figure 5.28 shows
STM topography filled state images of two Si(100) substrates prepared in situ by
flash annealing that show contaminant clusters. Both these substrates were prepared
ex situ by an HF etch. These images were acquired with a tip bias ≈ -2 V and a tun-
neling current ≈ 150 pA. Panel (a) shows a phosphorous-doped Si substrate flashed
to approximately 1150 ◦C. This substrate was nominally clean after flashing and
although the RHEED pattern showed indication of SiC initially, no SiC signal was
present after the final, higher temperature flashes. Unknown contaminants appear
as bright spots in a cluster approximately 30 nm across on otherwise normal (2×1)
reconstructed terraces. SiC, SiO2, or other particulates are possible explanations
267
Figure 5.28: STM topography filled state images of Si(100) substrates prepared ex
situ by an HF etch and in situ by flash annealing. These images were acquired with a
tip bias ≈ -2 V and a tunneling current ≈ 150 pA. (a) Phosphorous-doped Si(100)
substrate flashed to around 1150 ◦C with contaminants appearing as a cluster of
bright spots on otherwise normal terraces. This substrate was nominally clean and
free of SiC as determined by RHEED. The height scale of this image is about 1.3 nm.
(b) Boron-doped Si(100) substrate flashed to around 1130 ◦C with two clusters of
contaminants appearing as bright areas. The RHEED pattern of this substrate
showed significant SiC contamination. The clusters act as step pinning sites causing
step bunching around them, which forms small mounds. The height scale of this
image is about 15 nm.
268
for clusters commonly seen during STM inspection on nominally clean substrates
such as that in panel (a). These features may also be due to Si atoms stuck on the
tip. During deposition of a thin film on such a surface, steps may become pinned
at this cluster, causing a buildup of roughness nearby. The total height scale of
this image is approximately 1.3 nm, which is mostly accounted for by the height of
the cluster. Some dark dimer row defects are also seen on this surface. Panel (b)
shows a boron-doped Si substrate flashed to around 1130 ◦C. The RHEED pattern
of this substrate showed significant SiC contamination. Two clusters of SiC appear
as bright areas approximately 25 nm across. The total height scale of this image
is approximately 15 nm. These clusters clearly act as step pinning sites causing
significant step bunching around them, which has resulted in the formation of small
mounds that the SiC clusters sit atop.
The in situ prepared Si(100) substrates used for the initial set of 28Si films
deposited above 600 ◦C can not only have SiC and other clusters of contaminants
and particulates on their surface, they can additionally exhibit signs of metal con-
tamination. Metal atoms on or just below Si(100) surfaces produce patterns of
dimer row defects observable in the STM, which are most often attributed to Ni
contamination [116,153–157]. The presence of Ni or possibly other metal impurities
including In, Ga, and Al on the Si(100) surface produces long chains or lines of
ordered dimer defects after high temperature heat treatments such as the typical
flash annealing process. Ni atoms are thought to reside just below the surface and
disrupt local bonds generating surface defects. These dimer vacancy lines (DVLs)
run perpendicular to the Si dimer rows on each terrace, forming (2×n) patterns
269
where typically n ≈ 8 and represents the number of normal Si dimers between the
DVLs [153,154,157]. Si surfaces with DVLs present appear striped in STM imaging.
The ordering of vacancies in DVLs is due to repulsive interactions between vacancies
along a dimer row and attractive interactions between vacancies in adjacent dimer
rows, which may be related to strain relaxation [153,158].
Figure 5.29 (a) shows an example of a STM topography filled state image of a
Si(100) surface with Ni contamination from Ref. [153]. This image was acquired with
a tip bias ≈ -2 V and a tunneling current ≈ 30 pA. This sample was deliberately
contaminated with Ni by contacting it with stainless steel tweezers ex situ before
it was loaded into the STM. That simple handling of the sample was enough to
contaminate the surface with Ni. After it was flash annealed to 1150 ◦C, the sample
then exhibited a highly ordered (2×n) reconstructed surface. Dark DVLs due to the
Ni contamination are ubiquitous on this surface and are seen running perpendicular
to the dimer row directions on each terrace. Typically, a surface coverage of < 1 %
Ni can produce the (2×n) surface seen in panel (a). A similar example of metal
contamination on a prepared substrate from this work is shown in Fig. 5.29 (b),
which is a STM topography filled state image of a Si(100) surface prepared by flash
annealing to approximately 1130 ◦C. This image was acquired with a tip bias ≈
-2 V and a tunneling current ≈ 150 pA. Although this boron-doped substrate was
not intentionally contaminated with Ni or other metals, its surface appears very
similar to the Si surface shown in panel (a). The surface has a (2×n) reconstruction
with dark DVLs running perpendicular to the dimer row directions, although the
concentration and length of DVLs is less than in panel (a).
270
Figure 5.29: STM topography filled state images of Si(100) substrates prepared in
situ by flash annealing. (a) A Si(100) substrate flash annealed to 1150 ◦C has a
(2×n) reconstructed surface with dark dimer vacancy lines due to Ni contamina-
tion. Vacancy lines run perpendicular to the dimer row direction on each terrace.
The substrate was intentionally contaminated by contacting it with stainless steel
tweezers ex situ. This image was acquired with a tip bias ≈ -2 V and a tunnel-
ing current ≈ 30 pA (from Ref. [153]). (b) A boron-doped Si(100) substrate flash
annealed to around 1130 ◦C has a (2×n) reconstructed surface with dark dimer va-
cancy lines due to metal although the substrate was not intentionally contaminated.
This image was acquired with a tip bias ≈ -2 V and a tunneling current ≈ 150 pA.
271
The concentration of DVLs seen in Fig. 5.29 (b) is around the highest ob-
served for substrates discussed to this point in this section, which were not chemi-
cally cleaned ex situ. Much more frequently, flashed substrates will exhibit a more
moderate concentration of DVLs with shorter lengths. Some substrates have had
a large number of individual or small clusters of dimer vacancy defects present on
their surface, although not aligned in DVLs. It is not obvious if all dimer vacancies
are related to metal contamination or what causes the dimer vacancies to form lines
on some substrates while remaining disordered on others. The formation of DVLs
is probably related to the level of contamination as well as differences in the specific
heating times and temperatures that each substrate experiences. It has been found
in this work that a minimal number of flashes to around 1200 ◦C followed by a quick
cool down to below 800 ◦C can eliminate DVLs from a surface. Other groups have
used similar procedures to reduce or eliminate DVLs [114, 154]. It is also observed
that prolonged annealing to temperatures between 600 ◦C and 800 ◦C can result
in the reappearance of DVLs even after their elimination by higher temperature
flashing. The solid solubility limit of Ni in Si at a temperature of around 1200 ◦C is
5.8× 1017 cm−3 [159]. So, at concentrations below this, Ni likely diffuses into the Si
bulk at higher temperatures and can return to the surface during lower temperature
annealing. While it is not ultimately desirable from a device point of view for 28Si
films to contain Ni or other metal impurities, it is not known exactly how this type of
contamination may effect the growth morphology or epitaxy of deposited 28Si films.
Clearly, Ni atoms disrupt the bonding structure of the surface potentially creating
strain and defects which may result in the formation of pinning sites or defects in
272
the epitaxy.
In addition to SiC or other particulates and metal contamination, there are
other candidates for contaminants on Si(100) substrates prepared by in situ flash
annealing which may lead to step pinning sites during deposition and 28Si film
growth. For the substrates which are not cleaned before they enter the vacuum
chamber with a native oxide, removal of the oxide may be incomplete during flash
annealing. The remaining, possibly non-stoichiometric, silicon oxide (SiOx) may
form clusters similar to the observed SiC, or molecules may remain isolated being
very difficult to detect using the STM. Even a small amount of SiOx on the surface
of a substrate may act as a pinning site. Within the deposition chamber, N2 is one of
the major residual components of the vacuum after H2 and CO2, as seen in the RGA
mass spectrum in Fig. 2.22 in Chapter 2. N-containing adsorbates on a Si substrate
may dissociate during thermal processes potentially leading to the formation of
silicon nitride clusters (Si3N4). As will be discussed in a following section, chemical
analysis shows that 28Si samples deposited after the ones discussed in this section
contain a relatively high concentration of N, originating from the vacuum or the
ion beam itself. Si3N4 clusters are thus a candidate for the nucleation of pinning
sites during deposition. The solid solubility limits of N, C, and O in Si are shown in
Fig. 5.42 and discussed in a later section. These solid solubility limits show, however,
that the measured concentrations of N, C, and O in the films can lead to contaminant
clusters that cause step pinning. The presence of H2 on Si surfaces is believed to lead
to increased roughness during epitaxial deposition [119]. Surface interaction with H2
is nearly unavoidable in a vacuum system especially when increased amounts of H2
273
and SiH4 are introduced into the deposition chamber from the ion beam chamber.
Finally, another potential contaminant coming from the vacuum during deposition is
F, as seen in the deposition chamber RGA mass spectrum in Fig. 5.3. As mentioned
previously, F atoms can also contaminate a substrate which was treated with HF
and etch the surface when they desorb with thermal processes in the vacuum [151].
Small areas etched by F on a Si substrate and the resulting defects may lead directly
to step pinning and increased roughness.
Another aspect to the formation of faceted mounds is the deposition tem-
perature. To restate from a previous discussion, the rough 28Si samples discussed
thus far in this section were deposited with substrate temperatures between 610 ◦C
and 1041 ◦C. At these temperatures, one would expect the deposition to occur
in a predominantly layer-by-layer 2D growth mode. Despite the extreme rough-
ness of these samples and morphology dominated by facet formation, the primary
mechanism of mass transport and step motion is likely still akin to layer-by-layer
growth, especially in the early stages of deposition when the surface is still fairly
smooth. The high temperature layer-by-layer growth likely enhances some aspects
of the mound formation. For example, the defects that form on {111} planes dur-
ing Si epitaxy are found to increase in quantity as the deposition temperature is
increased [120]. In general, layer-by-layer growth is much more likely to lead to the
formation of microfacets, especially the rebonded DB steps required for more stable
{113} planes [143]. It also leads to faster growth of {111} and {113} planes while the
{100} planes diminish, as mentioned previously. Probably the most important role
of predominantly layer-by-layer growth is that it allows for the flow of steps around
274
pinning sites which leads to the step bunching required for microfacet formation.
If layer-by-layer growth and Si surface diffusion are in fact important driving
aspects to the formation of mounds on the surface of 28Si samples, then the char-
acteristics of the mounds should reflect that. Another characteristic of the mounds
besides faceting which can inform on their origin is their size. Using the analysis
method described previously, the length and width of mounds on several samples
was analyzed from the SEM micrographs of the surface. Here, the length is defined
as the longer dimension of mounds (if applicable), while the width is the shorter
dimension. An autocorrelation function was used to determine these values for each
sample analyzed including several of the sample micrographs shown in Fig. 5.23 as
well as others.
The results of this analysis are shown in Fig. 5.30. The average length (trian-
gles) and width (squares) of mounds is plotted vs. the deposition temperature. Both
the length and width of mounds increase with increasing temperature. The widths
vary from approximately 175 nm for the 610 ◦C sample to approximately 560 nm
for the 920 ◦C sample. The mound lengths vary from approximately 320 nm for the
610 ◦C sample to approximately 1105 nm for the 920 ◦C sample. The relative un-
certainty in the lengths and widths is determined to be about 10 % from comparing
measurements of multiple SEM images from a single sample. The uncertainty in
the deposition temperature is due to uncertainty in the pyrometer readings and its
calibration, as discussed in Chapter 2. The length-to-width aspect ratio of mounds
formed on different samples is shown in the inset in Fig. 5.30 (open circles) to be
between 1.5:1 and 3.5:1. The average aspect ratio of the mounds from all samples
275
Figure 5.30: Average mound size vs. deposition temperature for rough 28Si films
deposited above 600 ◦C. The length (triangles) and width (squares) of large mounds
formed on the films increase with increasing temperature as calculated from SEM
micrographs of the surface morphology. The inset shows the ratios of the lengths
to the widths (open circles) of the mounds for the different samples, most of which
are about 2:1.
analyzed in this section is found to be ≈ 2:1. This value is similar to the aspect
ratio reported by Mo et al. of 2D Si islands deposited on Si(100) [160]. That study
used STM imaging to determine that the aspect ratio of the islands close to their
equilibrium shape was between 2:1 and 3:1, and attribute it to differences in adatom
incorporation and diffusion along the island edges.
The increase in mound size with increasing temperature suggests that the
mounds are formed from a process or mechanism that is thermally activated, such
as surface diffusion. Surface diffusion constants have the standard exponential form
276
of a thermally activated process:
D = D0 exp
(
− Ea
kBT
)
, (5.3)
where D0 is the exponential prefactor, kB is the Boltzmann constant, Ea is the
activation energy, and T is the substrate temperature. As the Si surface diffu-
sion constant increases exponentially with increasing temperature, the mechanisms
involving both the flow of steps leading to faceting and competing diffusion on dif-
ferent microfacets may be enhanced causing an increase in the characteristic size
of the mounds at a given temperature. Increased temperature also leads to in-
creased diffusion of adatoms over steps. This may lead to merging of different facets
around pinning sites at a higher rate and the formation of larger mounds at higher
temperatures.
It is important to note that other factors besides temperature may affect the
mound size. The deposition time, deposition rate, or more likely the final film
thickness may lead to different size mounds for a given temperature. However,
the mound sizes do not seem to vary as a function of film thickness and most of
the samples analyzed in Fig. 5.30 have similar thicknesses between approximately
110 nm and 160 nm. The thickness of the 920 ◦C sample was not measured, but
it is roughly estimated to be larger than the other samples at around 400 nm, and
this sample also exhibits the largest mounds. Another factor that may affect the
mound size independent of the temperature is the amount of contaminants present
on the substrate that would lead to pinning sites. A sample with a higher density
of contaminants and pinning sites may result in smaller mounds because the area
277
between pinning sites where mounds may form would be smaller. However, most of
the samples involved in this analysis were not cleaned ex situ, which should lead to
similar amounts of contaminants. So, contaminants and thus the size of the mounds
would not be expected to vary significantly across nominally similar samples. One
sample deposited at 804 ◦C and shown in Fig. 5.23 (c) was not included in the
mound size analysis because it was etched with HF, which is known to produce
more SiC contamination, and it exhibited smaller grains than the other samples.
This suggests that the amount of contaminants or other factors that lead to pinning
do affect the size of mounds or grains that form on the surface.
The link between mound size and surface diffusion is explored further by plot-
ting dependance of the mound size on deposition temperature in an Arrhenius form.
An effective activation energy can then be extracted that results from the activation
energy of the diffusion constant. This assumes that the change in the mound size
is proportional to the change in the diffusion constant over the temperature range
of the data. This proportionality makes sense when considering that an increased
surface diffusion would lead to an increase in adatom flux to the growing faceted
sides of the mounds and thus an increased area. In order to simplify this deter-
mination and not favor one particular dimension of the mounds, the mounds are
treated as rectangular and the lengths and widths are combined to yield the average
mound area, A. Then, taking A ∝ D and substituting A for D and a new expo-
nential prefactor for the mound area, A0, for D0 in Eq. (5.3), the area data can be
plotted to extract Ea. Taking the natural log of Eq. (5.3) with the aforementioned
278
Figure 5.31: Arrhenius plot of ln(A), the natural log of the average mound area
(circles), vs. inverse temperature in energy units for rough 28Si samples deposited
above 600 ◦C. The top axis shows the equivalent deposition temperature in ◦C.
The data are fit to a line (Eq. (5.4)) whose slope gives an activation energy Ea =
0.7(3) eV.
substitutions gives a linear equation,
ln(A) = ln(A0)− Ea
(
1
kBT
)
, (5.4)
where the slope of ln(A) as function of 1
kBT
is equal to Ea. In Fig. 5.31, ln(A) is
plotted vs. the inverse deposition temperature, which is modified by the Boltzmann
constant to have units of eV−1. The uncertainty in ln(A) is from the combined
uncertainties of the lengths and widths. The data is fit to a line using Eq. (5.4), the
slope of which gives a value for the activation energy for Si surface diffusion of Ea =
0.7(3) eV. The uncertainty in this value is the standard error from the fit. Again,
279
this assumes that the appropriate activation energy for diffusion would emerge from
this analysis of A, the mound areas, because A changes with T in proportion to D.
First-principles and experimental investigations of Si diffusion on Si(100) surfaces
find activation energies of 0.6 eV and 0.67 eV respectively for diffusion along dimer
rows, similar to the value reported here [161,162]. Those calculations also predict a
value of Ea = 1.0 eV for diffusion perpendicular to dimer rows. However, another
experiment found an average value of Ea ≈ 1.55 eV for diffusion across steps of
the Si(100) surface [163], and molecular dynamics simulations of Si diffusion on
Si(100) along dimer rows give a value of Ea = 0.2 eV [144]. While there is a large
range of values reported in the literature, it seems reasonable that the activation
energy determined here from Fig. 5.31 can be attributed to Si surface diffusion, an
important component in faceted mound formation.
5.6.3 Elimination Strategies for Step Pinning Sites
Several strategies were used to try to reduce the density of pinning sites on
substrates, which manifest during deposition as pits and step bunching. New clean-
ing protocols were established to reduce chemical contaminants, such as SiC, on
or near the substrates. An ex situ CMOS cleaning procedure was implemented for
preparing Si(100) substrates. This is the sample preparation cleaning procedure
described previously in this chapter in Section 5.3. This CMOS clean consists of
etching the substrate first with piranha solution, then etching the oxide that forms
with HF, and finally using SC–2 to clean and cap the substrate with a protective
oxide. This clean is designed to remove both organic impurities containing C as well
280
as metals including Ni from the substrate. An HF etch alone, while a standard Si
substrate preparation method for vacuum surface science experiments, was not cho-
sen because of the ample evidence presented previously that it results in a substrate
surface with far more SiC contamination than one that was protected with a native
oxide before in situ heat treatments. Clean Si substrates were only ever handled
with teflon tweezers that were also cleaned of metals using hydrochloric acid (HCl).
In addition to cleaning metal contamination off the substrates, the sample
manipulator in the deposition chamber as well as the in vacuum sample holders were
also cleaned. This was done to reduce cross-contamination onto the substrates of Ni
and other metals, which can migrate from these parts during sample heating. While
the sample holders are comprised of only Mo, several stainless steel components,
which can spread Ni, were removed from the Mo section of the manipulator that gets
hot during sample heating. Also, a chromel-alumel thermocouple, which is mostly
Ni, was removed from the manipulator. The Mo parts were cleaned first in “base
piranha”, which is a 1:1:3 mixture of ammonium hydroxide (NH4OH), hydrogen
peroxide (H2O2), and water. This solution gently etches Mo surfaces removing
contaminants, and it etches Group V elements. Then the Mo parts were treated
with HCl to remove metals. Mo tools used for manipulating the sample holders
were also cleaned in this manner. These vacuum component cleaning procedures
were adopted from Richardson [164].
Despite these efforts to eliminate contaminants that cause step pinning, indi-
cations of the presence of contaminants on substrates persists. The CMOS clean-
ing procedure is either not sufficient at removing the offending contaminants from
281
the substrate surface, or the substrates pick up additional contaminants after the
cleaning procedure. After preparing newly cleaned substrates in situ through flash
annealing, inspection by RHEED shows that SiC was present more often than be-
fore CMOS cleaning, although the level of contamination was still much less than
that of the substrates etched with HF alone. The frequency with which SiC was ob-
served depended on the source of the wafers used for substrates. Wafers that were
re-claimed wafers obtained from University Wafer and prepared with the CMOS
cleaning procedure exhibited signs of SiC in the RHEED pattern approximately
66 % of the time. Virginia Semiconductor is generally seen as a source of higher
quality wafers (e.g. lower impurities, smoother), and approximately 35 % of those
wafers exhibited signs of SiC in the RHEED pattern. That frequency of SiC for-
mation is similar to that of the older substrates which were not cleaned at all ex
situ. It is not clear why the samples from University Wafer exhibit more contami-
nation, but it is likely a combination of surface contamination from poor handling
by the company and some contaminants in solution from previous processing of the
re-claimed wafers.
Several samples were made to investigate any changes in contaminants, step
pinning, or morphology of 28Si samples deposited above 600 ◦C after implementing
the above described cleaning procedures. To view the formation of pits due to step
pinning sites, thinner films were deposited from both the ion beam and the natural
abundance Si EFM source on Si(100) substrates prepared by the CMOS cleaning
procedure. Figure 5.32 shows STM topography filled state images of thin films of
28Si, (a) and (b), and natSi, (c) and (d), which show the formation of pits in the
282
growth surface. Both of these films were deposited with substrate temperatures of
approximately 712 ◦C at DC–3, and both are estimated to be approximately 10 nm
thick. Panel (a) is a large area scan of the 28Si film. Numerous square shaped
pits can be seen in the growth surface surrounded by step bunching that bounds
elevated areas with large, flat, single terraces roughly 50 nm wide. Single terraces
between pits indicates a predominantly layer-by-layer growth mode. The pits are
qualitatively different from the pits previously seen in the 28Si film in Fig. 5.24 and
are more similar to the square pits seen in the natSi film in Fig. 5.25. The cause
of the qualitative difference in the pits seen in the two 28Si samples is not known,
although it is perhaps related to the new cleaning procedure for substrates. Also, the
native step density due to the wafer miscut of the substrate used for the second 28Si
sample was roughly five times higher than that of the first 28Si sample. The average
areal pit density of this 28Si sample in Fig. 5.32 is determined to be 580 µm−2 ±
24 µm−2, which is significantly more than the areal density of pits in the previous
28Si sample (340 µm−2) shown in Fig. 5.24, despite the more rigorous substrate
cleaning procedure used for this latter sample. Panel (b) of Fig. 5.32 shows a small
area scan of a pit in panel (a). Si (2×1) dimer rows are visible on terraces around
the pit, which appears as an inverted pyramid in the surface with sides aligned with
the 〈110〉 directions. The total height scale in panel (b) is approximately 1.4 nm,
giving an indication of the pit depth.
Panel (c) in Fig. 5.32 is a large area scan of the natSi film showing six pits
in the growth surface. Note that the scan area in panel (c) is roughly 25 times
larger than the area shown in panel (a). Surrounding the pits are triangular shaped
283
Figure 5.32: STM topography filled state images of 28Si, (a) and (b), and natural
abundance Si, (c) and (d), films deposited at 712 ◦C at DC–3. Si(100) substrates
were prepared ex situ by CMOS cleaning. Both films are estimated to be about
10 nm thick. (a) Large area scan of the 28Si film showing many dark, square pits
and step bunching in the surface bounding large, single terraces about 50 nm wide.
The average areal pit density of this sample is ≈ 580 µm−2 ± 24 µm−2. (b) Small
area scan of a pit in (a). Dimer rows are visible and the pit appears as an inverted
pyramid with sides aligned with the 〈110〉 directions. Images (a) and (b) were
acquired with a tip bias ≈ -1.8 V and a tunneling current ≈ 100 pA. (c) Large area
scan of the natSi film showing six dark pits in the surface. Many steps < 50 nm wide
are seen due to the wafer miscut, indicating step flow growth. Steps appear to flow
around the pinning sites. The average areal pit density of this sample is ≈ 1.5 µm−2.
(d) Small area scan of a square pit in (c). Dimer rows are visible around the pit
and a bright contaminant cluster, possibly SiC, appears at the center. Images (c)
and (d) were acquired with a tip bias ≈ -2 V and a tunneling current ≈ 150 pA. Si
(2×1) dimer rows are visible in (b) and (d), and their total height scales are about
1.4 nm and 1.6 nm respectively.
284
regions of step bunching where the step motion was pinned by and flowed around
the pits during deposition. A large number of steps with widths < 50 nm are seen
on the surface with step edges running top-to-bottom along the hypotenuse of the
pits in the image. These are a consequence of a large miscut away from {100} in
the plane of the wafer surface. The presence of steps matched to the underlying
substrate indicates that deposition on this sample proceeded in a step flow growth
mode. The average areal pit density of this sample is determined to be 1.5 µm−2
± 1.2 µm−2, which is less than the pit density of the previous natSi film (40 µm−2)
shown in Fig. 5.25. For the approximately ten natSi thin films deposited from the
EFM evaporation source that exhibited pit formation and were cleaned with the
CMOS procedure, the average areal density of pits varied between approximately
1 µm−2 and 30 µm−2. Panel (d) on Fig. 5.32 is a small area scan of one of the
pits in panel (a). Si (2×1) dimer rows are visible on terraces around the pit and
a bright contaminant cluster approximately 10 nm across, possibly SiC, appears at
the center. The flow of steps was right-to-left around the pit in this image with a
high degree of step bunching seen to the right of the pit. The total height scale in
panel (d) is approximately 1.6 nm, which is mostly accounted for by the height of
the pit.
The natSi films described above that exhibit step pinning and pits seem to
show that the overall areal pit density and thus the amount of contaminants present
on the substrates did decrease, if somewhat inconsistently, when implementing the
CMOS and other cleaning procedures. However, the same is not true of the 28Si film,
which appears to have a higher amount of contaminants. Considering these results,
285
the one to two orders of magnitude larger areal pit density of the 28Si films compared
to the natSi films suggests more strongly than before that many contaminants and
pinning sites are due in part or exacerbated by something intrinsic to the 28Si ion
beam deposition process. Two possibilities are contaminants in the ion beam and
contaminants in the gases that diffuse from the ion beam chamber to the sample.
To observe surface contaminants before thin film deposition, other Si sub-
strates were prepared ex situ by the CMOS cleaning procedure, flash annealed in
situ, and inspected with the STM. The normal flashing procedure of heating the
substrate up to a temperature of around 1200 ◦C was not used, however. Instead,
substrates were flashed to a lower temperature that should form SiC but not elim-
inate it from the surface, similar to the STM study of Si surface contaminants
represented in Fig. 5.27 showing SiC clusters after a 900 ◦C anneal. Figure 5.33
shows STM topography filled state images of two Si(100) substrates prepared in
this manner that show contaminant clusters. Panel (a) shows a phosphorous-doped
Si substrate flashed to a temperature of approximately 1060 ◦C two times. The
RHEED pattern of this sample did show faint signs of SiC on the surface after flash-
ing. Multiple contaminant clusters, probably SiC, are seen on the surface, appearing
as bright spots. Step movement during flashing caused the steps to recede around
the clusters, which act as pinning sites causing step bunching nearby. The total
height scale of this image is approximately 7.0 nm, which is mostly accounted for by
the height of the clusters. Panel (b) shows a boron-doped Si substrate flashed to a
temperature of approximately 1050 ◦C two times. It was unclear from the RHEED
pattern of this sample if SiC was present or not. A bright contaminant cluster, pos-
286
Figure 5.33: STM topography filled state images of Si(100) substrates prepared ex
situ by CMOS cleaning and in situ by flash annealing. These images were acquired
with a tip bias ≈ -2 V and a tunneling current ≈ 150 pA. (a) Phosphorous-doped
substrate flashed to around 1060 ◦C with contaminant clusters, probably SiC, ap-
pearing as bright spots. Step flow due to flashing appears pinned at the clusters
causing step bunching and increased roughness. The total height scale of this image
is about 7.0 nm. (b) Smaller area scan of a boron-doped substrate flashed to around
1050 ◦C with a contaminant cluster appearing in the upper right corner. The height
scale of this image is about 2.8 nm.
287
sibly SiC or SiOx from incomplete desorption due to the lower flash temperature,
is apparent in the upper right of the image. This cluster is approximately 20 nm
wide and approximately 60 nm long. The total height scale of this image is approx-
imately 2.8 nm, which is accounted for mostly by the height of the cluster. Short
terraces are also seen with step bunching occurring near the cluster and Si (2×1)
dimer rows visible on some terraces.
Finally, a thick 28Si film was deposited onto a substrate prepared with the
CMOS cleaning procedure to determine any changes in roughness or surface mor-
phology compared to the previous 28Si films shown in Fig. 5.23 that were not cleaned.
Figure 5.34 is a top-down SEM micrograph of a thick 28Si film deposited on a
clean, boron-doped Si(100) substrate with a substrate temperature of approximately
705 ◦C at DC–3. The maximum thickness of this film found by SIMS depth profiling
is approximately 144 nm. The surface morphology of the deposited film appears to
be very rough, similar to SEM micrographs of the previous 28Si films. Mounds cover
the surface with an average width ≈ 230 nm and an average length ≈ 495 nm. The
mounds have sharp, well defined, faceted edges predominantly running diagonally
in the micrograph, indicating that they are oriented with the 〈110〉 directions as
determined from the sample positioning in the SEM. The mounds on this sample
appear qualitatively different from those of the previous 28Si sample deposited with
a similar substrate temperature of 708 ◦C, which had more rounded and randomly
shaped mounds as opposed to the faceted, zig-zag pattern created by the mounds
in the 705 ◦C sample. It is not known if the cause of this qualitatively different
morphologies is due to the updated cleaning procedures.
288
Figure 5.34: SEM top-down micrograph of the surface of a 28Si film deposited at
705 ◦C at DC–3. The substrate used for this sample was prepared ex situ by CMOS
cleaning. The thickness of this film determined from SIMS depth profiling is about
144 nm. The surface morphology of the deposited film appears very rough with
mounds that have an average length ≈ 495 nm. The mounds have sharp, well de-
fined, faceted edges predominantly running diagonally in the micrograph, indicating
they are oriented with the 〈110〉 directions.
The above results show that the CMOS and other cleaning procedures im-
plemented to reduce substrate contamination and thus step pinning that leads to
mound formation and roughness were mostly unsuccessful. One notable aspect of
the as-prepared quality of substrates that seemed improved by the new cleaning pro-
cedures was the presence of Si (2×1) dimer row defects. Although DVLs similar to
those in Fig. 5.29 still appeared on some 28Si samples after deposition, very few were
seen on substrates prepared for deposition after flash annealing. Further, the surface
density of isolated dimer row defects was reduced on prepared substrates after the
new cleaning procedures. For substrates that were not cleaned ex situ, dimer row de-
289
fects were observed using STM to account for a wide range of areal densities between
approximately 1 % and 34 % of the surface, with an average value of approximately
9 % being typical. These areal dimer defect densities were determined using particle
detection software for the STM. By contrast, the higher quality substrates that were
cleaned with the CMOS procedure and prepared after the manipulator and sample
holder cleaning typically exhibited dimer row defects on approximately 2 % to 3 %
of the surface.
While the CMOS and other cleaning procedures mentioned above are ulti-
mately believed to be an important aspect in preparing clean substrates, it is ap-
parent that a potentially monumental effort would be required to sufficiently reduce
contaminants on substrates in this system. Considering this and the continued influ-
ence of contaminants and pinning sites on the growth morphology, surface roughness,
and possibly epitaxial quality of 28Si films deposited with substrate temperatures
above 600 ◦C, a strategy of lower temperature deposition was adopted to limit the
effects of the pinning sites. A predominantly layer-by-layer growth mode at higher
temperatures is believed to facilitate and enhance the formation of step bunching
at pinning sites and faceting at step bunching, as discussed previously. Depositing
with a substrate temperature below 600 ◦C in a 3D island growth mode prevents
step bunching and faceting from dominating the growth. While the formation of
3D multi-layer islands represents some intrinsic roughness in the growth surface, the
merging of islands in this growth mode leads to an overall smoother surface.
Another strategy adopted to improve the epitaxial quality of films and heal
potential defects resulting from deposition at lower temperatures is post-deposition
290
annealing. Thin film annealing has been used in both Si MBE and 28Si IBE ex-
periments to recrystalize amorphous layers and extend hepi, the epitaxial critical
thickness. Extending hepi for a given deposition temperature was demonstrated
by annealing at a minimum temperature of 500 ◦C, which is needed to break Si-H
bonds that can form at crystalline defects [165]. Another experiment found that
annealing at 600 ◦C results in SPE with a recrystallization front propagation speed
of approximately 60 nm/min, allowing crystalline 28Si layers to form from regions
amorphized by high energy ion impacts [166]. Finally, annealing at 510 ◦C has been
used to smooth thin films of Si deposited at low temperature and reduce roughness
due to surface islands [167]. A post-deposition anneal of 600 ◦C for 1 h was chosen
for most low temperature depositions in this work.
The strategy for eliminating pinning sites and producing smooth, epitaxial thin
films of 28Si consists primarily of preparing atomically clean Si substrates followed by
depositing at temperatures below 600 ◦C. Although the elimination of contaminants
on the substrates through ex situ cleaning is not fully experimentally realized, this
strategy can be summarized as follows:
1. prepare Si substrates free of organics and metals ex situ with a CMOS cleaning
procedure,
2. handle Si substrates with non-metal tools only,
3. manipulate and heat Si substrates in situ with only Mo parts cleaned of other
metals,
4. flash anneal Si substrates to approximately 1200 ◦C to prepare a clean surface,
verified by RHEED and STM,
5. deposit 28Si in a 3D island growth mode below 600 ◦C to reduce pinning and
step bunching,
291
6. and anneal the 28Si film at 600 ◦C for 1 h.
5.6.4 Morphology of Films with Deposition T < 600 ◦C
A total of 18 28Si samples were deposited at sample location DC–3 with a range
of substrate temperatures following the CMOS and other strategies laid out above
for producing smooth, epitaxial films. Samples were deposited with substrate tem-
peratures between approximately 249 ◦C and 502 ◦C on Si(100) substrates. Typical
average ion energies, Ei, between approximately 35 eV and 40 eV were used. The
epitaxy phase diagram for IBE in Fig. 5.18 predicts that a deposition temperature
of approximately 350 ◦C is needed with these ion energies to produce high-quality
epitaxial deposition, although the exact value would vary with deposition rate and
other factors. These samples were deposited on a variety of Si(100) substrates in-
cluding phosphorous-doped, boron-doped, and intrinsic wafers (University Wafer)
initially before transitioning to higher quality phosphorous-doped wafers (Virginia
Semiconductor).
5.6.4.1 RHEED
Initially, 28Si samples deposited at low temperature (i.e. below 600 ◦C) were
inspected using RHEED immediately following deposition to determine the epi-
taxial quality and morphology of the films. The first thicker sample produced at
low temperature was a 28Si film deposited at approximately 357 ◦C at DC–3. A
RHEED diffraction pattern of this sample after deposition is shown in Fig. 5.35.
This image was acquired with the sample at the deposition temperature and the
292
Figure 5.35: RHEED diffraction pattern of a 28Si sample deposited at 357 ◦C at
DC–3. This image was acquired with the sample at the deposition temperature and
the electron beam in the 〈110〉 direction. The presence of (1×1) and weak (2×1)
Si diffraction streaks indicate that the film is crystalline and aligned to the Si(100)
substrate. Additionally, the streaks of this pattern corresponds to diffraction from
very narrow terraces, likely due to a surface consisting of small, flat islands. A diffuse
3D transmission pattern is also visible superimposed on the streaks indicating some
surface roughness.
electron beam in the 〈110〉 direction. The diffraction pattern is clearly different
from those of the 28Si samples deposited with substrate temperatures above 600 ◦C
as in Fig. 5.19. In the pattern of this 357 ◦C sample, both (1×1) and (2×1) Si
diffraction streaks are visible indicating that the film is crystalline and epitaxially
aligned to the Si(100) substrate. The elongation of the nominal spots into streaks on
this pattern corresponds to diffraction terraces that are narrow, e.g. about 20 nm
wide, in two dimensions (parallel and perpendicular to the RHEED electron beam).
This diffraction pattern is likely due to a predominantly 2D surface consisting of
small, flat islands. This suggests that the low temperature deposition strategy was
indeed successful at producing a smooth film free of large mounds.
293
Although the well defined 3D transmission spot pattern of the high tempera-
ture samples is absent from this image, there does appear to be a diffuse transmis-
sion diffraction characteristic visible superimposed on the streaks, which indicates
the presence of some small amount of surface roughness, possibly due to the is-
lands themselves. Additionally, there is no diffraction attributable to microfacets
visible in this pattern, i.e. the “chevron” pattern or other lines connecting adja-
cent diffraction rods previously seen. With Ei ≈ 46 eV for this sample, the epitaxy
phase diagram for IBE in Fig. 5.18 would predict that a deposition temperature
above 450 ◦C would be needed to produce the epitaxy seen in the RHEED pattern.
However, this sample had a fairly slow growth rate of approximately 0.33 nm/min,
which may offset the effect of a lower temperature, and it was quite thin at only
approximately 50 nm as measured by SIMS depth profiling. This film may indeed
have been deposited in the limited epitaxy regime as Fig. 5.18 suggests but was too
thin to develop the amorphous phase at hepi, which has been shown to be as large
as 1 µm for deposition occurring above 300 ◦C [120].
Nearly all of the low temperature 28Si samples had RHEED diffraction pat-
terns similar to the one above in Fig. 5.35, although with varying intensities of
the transmission aspects of the patterns, indicating varying degrees of roughness
on the surface of the films. The most commonly used deposition temperatures for
these samples were nominally 400 ◦C or 450 ◦C. A summary of the evolution of the
RHEED diffraction patterns from these low temperature films to the high temper-
ature films is shown for a series of eight 28Si samples in Fig. 5.36. These samples
were deposited at DC–3 on Si(100) substrates with substrate temperatures of ap-
294
Figure 5.36: RHEED diffraction patterns of eight 28Si films deposited at DC–3 at
249 ◦C, (a), 349 ◦C, (b), 421 ◦C, (c), 502 ◦C, (d), 610 ◦C, (e), 708 ◦C, (f), 804 ◦C,
(g), and 920 ◦C, (h) on Si(100) substrates. These images were acquired with the
samples at the deposition temperatures and the electron beam in the 〈110〉 direc-
tion. (a) The 249 ◦C film has a diffuse pattern with faint (1×1) bulk Si streaks
indicating a partially disordered film and possibly a fully amorphous layer. (b)–(d)
The patterns of films deposited between 349 ◦C and 502 ◦C have (1×1) and (2×1) Si
diffraction streaks indicating crystalline films with small, flat islands on the surface.
Diffuse 3D transmission spots superimposed on (c) indicates some surface rough-
ness. (e)–(h) The patterns of films deposited between 610 ◦C and 920 ◦C show 3D
transmission spots indicating rough, crystalline surfaces. A faint “chevron” pattern
in (e) indicates diffraction from {311} microfacets. The (2×1) spots in (g) and (h)
are likely due to diffraction from the substrate outside of the deposition areas.
295
proximately 249 ◦C, (a), 349 ◦C, (b), 421 ◦C, (c), 502 ◦C, (d), 610 ◦C, (e), 708 ◦C,
(f), 804 ◦C, (g), and 920 ◦C, (h). These images were acquired with the samples at
the deposition temperatures and the electron beam in the 〈110〉 directions. The
diffraction pattern for the 249 ◦C sample in panel (a) is diffuse with faint (1×1)
bulk Si streaks. This indicates that the film is at least partially disordered and
possibly fully amorphous. The (1×1) streaks could be due to part of the RHEED
electron beam diffracting from the substrate outside of the amorphous deposition
area, although faint (2×1) streaks would probably be expected as well in that case.
(1×1) streaks could also be due to diffraction of crystalline Si below a thin disor-
dered layer or partial ordering within the film. This 249 ◦C sample was deposited
with Ei ≈ 38 eV and was determined by SIMS depth profiling to be approximately
305 nm in the thickest area measured. Given these parameters, the implication
of the RHEED pattern agrees with the epitaxy phase diagram for IBE shown in
Fig. 5.18 (b), which predicts limited epitaxy transitioning to an amorphous phase.
This sample differs from the 357 ◦C sample in that it had a much higher deposition
rate of approximately 1.25 nm/min, and it was much thicker.
The diffraction patterns of the films deposited with substrate temperatures
between 349 ◦C and 502 ◦C in panels (b)–(d) of Fig. 5.36 show (1×1) and (2×1)
Si diffraction streaks indicating crystalline films epitaxially aligned to the Si(100)
substrates with small, flat islands on a predominantly 2D surface, similar to the
pattern of the 357 ◦C sample in Fig. 5.35. Diffuse 3D transmission spots are visible
superimposed on the pattern of the 421 ◦C sample in panel (c) of Fig. 5.36 and to a
lesser extend on the pattern of the 502 ◦C sample in panel (d), indicates some small
296
surface roughness.
For samples with deposition temperatures higher than that of the 502 ◦C sam-
ple, there is an abrupt transition in the diffraction pattern from that of a 2D surface
with islands, to that of a rough 3D surface, and this latter pattern persists to the
highest deposition temperatures used here. The diffraction patterns of the films de-
posited with substrate temperatures between 610 ◦C and 920 ◦C in panels (e)–(h) of
Fig. 5.36 show 3D transmission spots indicating very rough but crystalline surfaces,
which known to be covered in mounds. These patterns are similar to that of the
708 ◦C sample shown in Fig. 5.19. A faint “chevron” pattern can be seen in panel
(e) of Fig. 5.36 indicating diffraction from {311} microfacets. The (2×1) spots seen
in panels (g) and (h) are likely due to diffraction from the substrate outside of the
deposition areas.
5.6.4.2 STM
In addition to observing 2D island growth diffraction patterns in RHEED for
28Si samples deposited at lower temperature, the STM was used to inspect the sur-
face morphology of these samples. The surface of a 28Si sample deposited at 357 ◦C
at DC–3, corresponding to the RHEED diffraction pattern shown in Fig. 5.35, is
shown in the STM topography filled state image in Fig. 5.37. This film is approxi-
mately 50 nm thick as determined from SIMS depth profiling, and it was deposited
on a Si(100) substrate that was prepared ex situ by the CMOS cleaning procedure.
After deposition and initial imaging in the STM, the sample was annealed at approx-
297
Figure 5.37: STM topography filled state images of a 28Si sample deposited at 357 ◦C
at DC–3. This film is about 50 nm thick determined from SIMS depth profiling and
was deposited on a Si(100) substrate after CMOS cleaning. The sample was annealed
at about 600 ◦C for 10 min. Images were acquired with a tip bias ≈ -1.8 V and a
tunneling current ≈ 100 pA. (a) Large area scan of the sample showing small, multi-
layer 3D islands comprising a smooth surface. Peaks and valleys are seen across the
surface due to islands merging. The surface width determined from the topography
is ∆z ≈ 1.6 nm or almost 12 monolayers. (b) Small area scan of an area in (a)
showing multi-layer islands roughly 10 nm wide with many steps visible. Si (2×1)
dimer rows are seen on the terraces making up the islands.
298
imately 600 ◦C for 10 min to form a more ordered surface before further imaging
shown here. Panel (a) is a large area scan of the sample showing many small,
multi-layer 3D islands comprising a relatively smooth, epitaxial surface. Peaks and
valleys are seen across the surface likely the result of various islands merging during
deposition. The topography of this image gives a surface width ∆z ≈ 1.6 nm for
the film. The height of a single atomic layer on a Si(100) surface is approximately
0.136 nm, and so the measured surface width of this sample equates to there being
approximately 13 crystalline layers exposed on the surface.
Panel (b) of Fig. 5.37 is a smaller scan of an area in panel (a). Again, multi-
layer islands roughly 10 nm wide are seen covering the surface with many steps and
terraces. Si (2×1) dimer rows are visible on the terraces forming the islands in this
image indicating epitaxial alignment with the substrate. The morphology observed
here matches the surface structure determined from the RHEED pattern for this
sample (Fig. 5.35). Further, the lack of mounds, large areas of step bunching, or
evidence of pinning sites visible in Fig. 5.37 affirms that lower temperature deposi-
tion, i.e. below 600 ◦C, does completely alter the growth morphology, resulting in a
smooth film. The smooth morphology of sample is in contrast to the rough surface
shown in Fig. 5.21 and the other samples with large mounds on the surface.
To replicate these results, several other 28Si samples were deposited with a
similar or slightly higher substrate temperature including nominal deposition tem-
peratures of 400 ◦C and 450 ◦C, which were most commonly used. The smooth
morphology of the 357 ◦C sample shown in Fig. 5.37 is generally found to be repro-
ducible for all of these depositions, which produce films with surfaces that appear
299
qualitatively similar in the STM. Four STM topography filled state images of four
different 28Si samples deposited at DC–3 with substrate temperatures of approxi-
mately 349 ◦C, (a), 417 ◦C, (b), and 421 ◦C, (c) and (d), are shown in Fig. 5.38.
These samples were deposited on Si(100) substrates that were prepared ex situ by
the CMOS cleaning procedure. After deposition, these samples were all annealed at
approximately 600 ◦C for 1 h to form more ordered surfaces. Similar to the 357 ◦C
sample, all of these images (panels (a)–(d)) show multi-layer 3D islands comprising
relatively smooth, epitaxial surfaces. Note that the scan areas in these four images
are the same to facilitate comparisons between them.
The film shown in panel (a) is approximately 206 nm thick with islands that
are roughly 20 nm wide. The surface width of this film determined from the STM
topography is ∆z ≈ 1.0 nm, which is approximately eight atomic layers exposed
to the surface. These eight layers are visible and can be counted in the image.
The film shown in panel (b) is approximately 250 nm thick with islands that are
roughly 15 nm wide. The surface width of this film determined from the topography
is ∆z ≈ 2.1 nm, which is approximately 16 atomic layers. The film show in panel
(c) is approximately 148 nm thick with islands that are roughly 30 nm wide. The
surface width of this film determined from the topography is ∆z ≈ 2.2 nm, which is
approximately 17 atomic layers. Finally, the film shown in panel (d) is approximately
320 nm thick with islands that are roughly 20 nm wide. The surface width of this
film determined from the topography is ∆z ≈ 1.0 nm, which is, again, approximately
eight atomic layers. The surface in panel (d) also shows some dark dimer row defects
on the terraces at a higher concentration than the other samples, possibly indicating
300
Figure 5.38: STM topography filled state images of four 28Si samples deposited at
DC–3 at 349 ◦C, (a), 417 ◦C, (b), and 421 ◦C, (c) and (d). These samples were
deposited on Si(100) substrates after CMOS cleaning and were annealed at about
600 ◦C for 1 h. Images were acquired with a tip bias ≈ -1.8 V and a tunneling
current ≈ 100 pA. Multi-layer 3D islands comprising smooth surfaces are seen on
all samples. Peaks and valleys are seen due to islands merging. (a) This film is about
206 nm thick with islands roughly 20 nm wide and the surface width determined
from the topography is ∆z ≈ 1.0 nm. (b) This film is about 250 nm thick with
islands roughly 15 nm wide and a surface width of ∆z ≈ 2.1 nm. (c) This film is
about 148 nm thick with islands roughly 30 nm wide and a surface width of ∆z ≈
2.2 nm. (d) This film is estimated to be about 320 nm thick with islands roughly
20 nm wide and a surface width of ∆z ≈ 1.0 nm. The surface in (d) also shows
some dark dimer row defects on the terraces. Si (2×1) dimer rows are visible on
the island terraces in (a)–(d). Film thicknesses were determined from SIMS depth
profiling except for that of the sample in (d), which is an estimate.
301
the presence of impurities in the film. These film thicknesses were the maximum
measured thicknesses determined from SIMS depth profiling except for that of the
sample in panel (d), which is an estimate based on the measured 28Si ion beam flux
during deposition. Si (2×1) dimer rows are visible on the terraces forming the island
in panels (a)–(d) indicating epitaxial alignment with the substrates.
The overall smoothness, i.e. the lack of extended step bunching and facet and
mound formation, of these 28Si films deposited with substrate temperatures below
600 ◦C is achievable for several reasons. As discusses previously, lower tempera-
ture deposition reduces step movement associated with higher temperature layer-
by-layer growth. This step movement can facilitate the formation of step bunching
and faceting, ultimately leading to mounds. Additionally, contaminants that cause
pinning sites and step bunching may cluster or become active at higher tempera-
tures. These effects are reduced during low temperature deposition in a 3D island
growth mode. However, 3D island growth introduces some intrinsic roughness and
step bunching around the islands, and defect formation that may act as pinning
sites can result from merging islands. Despite these aspects of 3D island growth,
films with a smooth morphology still occur due to the nature of island formation
on Si surfaces. Si island growth is anisotropic due to the difference in the diffusion
constants for adatom movement along or perpendicular to dimer rows. Anisotropic
growth is believed to result in the growth surface being constrained to a limited
number of atomic layers and thus a limited amount of surface roughness and step
bunching [160, 168]. This appears to be the case for the 28Si films deposited in a
3D island growth mode as in Fig. 5.38. The surface width of all of these samples is
302
fairly consistent with a typical value of ∆z ≈ 2 nm, which equates to approximately
16 atomic layers being exposed on the surface at any time. The surface width of
these samples does not seem to vary with deposition temperature or film thickness,
although there is not enough data to more strongly support these conclusions. This
possibly indicates that the measured ∆z values are indeed constant and intrinsic to
Si island growth.
Additionally, the hyperthermal energy of the 28Si ions may play a role in
producing a smooth surface during 3D island growth. Energetic ions are known to
suppress and dissociate 3D islands as well as defect clusters leading to smoother
growth. Mobile adatoms formed in the break up of islands may continually fill in
other inter-island trenches [120,127,169]. It is not clear if the 28Si ion energy has any
effect on the smoothing of films in this work because low temperature deposition
using the natural abundance Si EFM evaporator, which produces Si atoms with
only thermal energies, also results in smooth films. STM inspection of these natSi
films shows that they consist of small islands qualitatively similar to those of the
28Si films.
While most of the 28Si films deposited with substrate temperatures below
600 ◦C had surfaces that appear similar in the STM to those of the samples in
Fig. 5.38, several had surface morphologies that appeared different including two
samples with deposition temperatures of 502 ◦C and 249 ◦C. STM topography
filled state images of these samples and four others are shown in Fig. 5.39, which
serves to summarizes the variation in film morphology over the range of deposition
temperatures used in this work. 28Si samples were deposited at DC–3 with substrate
303
Figure 5.39: STM topography filled state images of six 28Si samples deposited at
DC–3 at 804 ◦C, (a), 705 ◦C, (b), 502 ◦C, (c), 421 ◦C, (d), 349 ◦C, (e), and 249 ◦C,
(f). Images were acquired with a tip bias≈ -1.8 V and a tunneling current≈ 100 pA.
All substrates used for these samples were prepared ex situ by CMOS cleaning except
for the sample in (a), which was etched with HF. (a) and (b) The morphology of
films deposited above 600 ◦C appear very rough with large mounds and grains.
The total height scale in (a) and (b) is about 45 nm and 50 nm respectively. (c)
The 502 ◦C sample morphology appears less rough with a large mound as well as
small island-like features visible. The height scale is about 13 nm. (d) and (e)
The morphology of films deposited between 349 ◦C and 421 ◦C appear smooth with
small, multi-layer 3D islands. The height scales in (d) and (e) are about 1.3 nm
and 3.6 nm respectively. (f) The 249 ◦C sample shows a smooth surface with small
features ≈ 3 nm in size. The height scale is about 1.0 nm.
304
temperatures of approximately 804 ◦C, (a), 705 ◦C, (b), 502 ◦C, (c), 421 ◦C, (d),
349 ◦C, (e), and 249 ◦C, (f). All Si(100) substrates used for these samples were
prepared ex situ by the CMOS cleaning procedure except for the sample in panel
(a), which was etched with HF. Note that the scan area displayed in panels (a)–(f)
are all the same to facilitate comparisons between them.
Panels (a) and (b) of Fig. 5.39 show the morphology of films deposited above
600 ◦C. These surfaces appear very rough with large mounds or grain-like features.
The total height scale in panels (a) and (b) is approximately 45 nm and 50 nm
respectively. For comparison, SEM micrographs of the 804 ◦C sample in panel
(a) and the 705 ◦C sample in panel (b) were shown in Fig. 5.23 (c) and Fig. 5.34
respectively. Panel (c) of Fig. 5.39 shows the morphology of the 502 ◦C sample,
which appears less rough than the samples in panels (a) and (b). This sample was
annealed at approximately 600 ◦C for 1 h after deposition. A large mound is seen
in the center of the image with smaller island-like features on it. The height scale
of this image is approximately 13 nm. The RHEED pattern of the 502 ◦C sample
(Fig. 5.36 (d)) indicated a smooth surface with small, flat islands, which does not
seem to be the case from the STM image. The nature of this discrepancy is not
known.
Panels (d) and (e) of Fig. 5.39 show the morphology of films deposited with
substrate temperatures of 349 ◦C and 421 ◦C, but they represent relatively smooth
films produces at deposition temperatures between 349 ◦C and 460 ◦C in this work.
These samples were annealed after deposition at approximately 600 ◦C for 1 h. The
surface of these samples appear smooth with small, flat, multi-layer 3D islands.
305
Smaller area scans of these samples show Si (2×1) dimer rows on the islands in-
dicating epitaxial deposition. The height scales in panels (d) and (e) are approxi-
mately 1.3 nm and 3.6 nm respectively. Finally, panel (f) shows the morphology of
the 249 ◦C sample, which has a smooth surface covered with small round features
approximately 3 nm in size that do not appear epitaxial. The height scale of this
image is approximately 1.0 nm. The RHEED pattern of this sample (Fig. 5.36 (a))
indicated an amorphous surface layer on the film, which may explain the round
features on the surface and absence of larger epitaxial islands.
It appears from this work that the smoothest epitaxial 28Si films are achieved
with deposition temperatures between approximately 349 ◦C and 460 ◦C. However,
although these smooth 28Si films were achieved with an apparent disappearance of
the pinning sites seen in the higher temperature samples, the defects and contam-
inants causing those pinning sites are likely still present in the film. While their
effects do not manifest in the film morphology at low deposition temperatures, they
may still affect the bulk crystallinity or even the electronic properties of the films.
5.7 Chemical Purity
5.7.1 Context
The 28Si films produced in this work need to have a high chemical purity
in order to be considered comparable to both commercially available electronics
grade natural abundance Si as well as the enriched 28Si available in the QI research
community. As mentioned at the beginning of this chapter, the third materials goal
306
for these 28Si films is to have chemical impurity concentrations including C and
O below 2 × 1015 cm−3. Chemical contaminates are undesirable in part because
of their detrimental effect on the smooth morphology and bulk epitaxy of thin
films, as discussed in detail above. Additionally, as mentioned previously, chemical
impurities present in the bulk of a 28Si film can act as or induce various scattering
sites and charge traps. This reduces electron mobility and other electronic properties
important for operating QI devices [91,92]. Further, chemical impurities can possess
nuclear spin including 13C, which has a nuclear spin I = 1/2, and 14N, which has
a nuclear spin I = 1. The presence of these nuclear spins will cause decoherence of
qubit spins in a quantum computing device, just as 29Si does.
Several sources of chemical contaminants exist within the deposition system
that may contribute impurities that become incorporated into the 28Si films. Con-
taminants can come from various molecular species comprising the background par-
tial pressures in the deposition chamber. A residual gas mass spectrum of the base
pressure of the deposition chamber acquired from the RGA is shown in Fig. 2.22
in Chapter 2 and gives insight into the contaminants present in the vacuum such
as N, C, O, and F, although quantitative partial pressures are not reliable. The
typical base pressure of the deposition chamber was approximately 6.7× 10−9 Pa
(5.0× 10−11 Torr). Other gaseous contaminants may diffuse into the deposition
chamber from the ion beam chamber during deposition. These can originate from
either the background base pressure of the ion beam chamber, which was typically
approximately 1.3× 10−5 Pa (1.0× 10−7 Torr), or the SiH4 gas source used during
deposition. Potential contaminants from the ion beam chamber including N, C, and
307
O can be seen in the residual gas mass spectrum of the base pressure acquired from
the RGA in the ion beam chamber, shown in Fig. 2.3 in Chapter 2. While the
SiH4 source bottle has a purity of 99.999 % according to the gas vendor (Matheson
Tri-Gas), the gas manifold used to load SiH4 into the ion source may contain much
higher levels of residual gas impurities because it is only pumped out to roughly
20 mTorr. When heating samples on the manipulator during deposition, the ele-
vated temperatures of the sample holder and other parts of the manipulator can
cause increases in the background pressure due to outgassing.
Probably the most significant source of chemical contaminants in the 28Si
films, specifically N, C, and O, is the ion beam itself. Selecting for 28Si+ ions in
the ion beam is the result of tuning the sector mass analyzer such that any ion
with a mass-to-charge ratio ≈ 28 u/e ± 0.18 u/e (at 28 u) will pass through the
mass-selecting aperture and propagate to the sample. The acceptance through the
aperture of a range of masses approximately 0.36 u wide (for a singly charged ion)
around 28 u is due to the width of the mass-selecting aperture (2 mm). This mass
range is also similar to the mass resolution at 28 u of 0.35 u, which is the smallest
mass that can be resolved by the system based on the maximum measured mass
resolving power m
∆m
≈ 78. Any ions with masses within this range will be trans-
ported to the sample along with 28Si. Two ionic species that match this criteria are
14N+2 and
12C16O+. The mass of 28Si is 27.97692653465(44) u, the mass of 12C16O is
27.99491461957(17) u, and the mass of 14N2 is 28.00614800886(40) u [170]. There-
fore, the separation in mass between 28Si and 12C16O is approximately 0.018 u, and
the separation in mass between 28Si and 14N2 is approximately 0.029 u, which are
308
both well within the range for transport past the aperture when the 28 u beam is
centered on it. A mass resolving power of m
∆m
≈ 1600 would be needed to signifi-
cantly separate 28Si from 12C16O and 14N2. The maximum achievable mass resolving
power for this system with an ion beam width due only to the energy spread from
the ion source, i.e. with no intrinsic beam width, would be approximately 350 at
mass 28 for a typical energy spread of ∆E = ± 6 eV. This means that for this
system, any N2 and CO ions generated in the ion source will be passed into the de-
position chamber and onto the sample along with 28Si. Further, these contaminant
species that have a mass separation below the mass resolution of the system are not
detectable in an ion beam mass spectrum.
5.7.2 XPS
XPS was used as an initial check on the chemical purity of 28Si samples de-
posited at elevated temperatures at DC–3 and to compare them to the previous
sample measured with XPS, which was deposited at room temperature at LC–2
(Fig. 4.17). This measurement was a search for gross chemical contaminants that
may have been introduced into the samples arising from the significant experimental
switch from 28Si deposition at LC–2 to DC–3 at elevated temperatures. XPS spectra
shown in this section were acquired and analyzed in collaboration with Dr. Kris-
ten Steffens (NIST). XPS spectra were collected after sputter-cleaning the sample
with Ar to remove surface contamination from the environment. The measurement
parameters for the spectra shown here are similar to those used for the previous
sample (see Section 4.3.6.2).
309
Figure 5.40: XPS spectra of a 28Si sample deposited at 812 ◦C at DC–3. Count
rates vs. electron binding energy for survey scans are shown for the 28Si sample
after Ar sputter cleaning (upper spectrum) as well as the substrate away from the
deposition area (lower spectrum). The data for the 28Si sample was shifted up for
clarity. References for elemental orbital levels are included above relevant peaks.
O 1s and C 1s peaks are visible with larger amplitudes in the 28Si film than the
substrate.
XPS spectra for the second 28Si sample deposited at DC–3 with a deposition
temperature of approximately 812 ◦C is shown in Fig. 5.40 (upper spectrum). Also
shown is a reference spectrum of the Si substrate away from the deposition area
(lower spectrum). References for relevant elemental orbital levels are shown at the
top.
A C 1s peak is visible at 285.2 eV in the 28Si spectrum corresponding to
an atomic fraction of approximately 3.4 %. Additionally, a O 1s peak is visible
at 532.7 eV corresponding to an atomic fraction of approximately 4.5 %. No C
or O peaks are visible above the noise in the spectrum for the Si substrate. A
310
small N 1s signal is also present at 400.9 eV corresponding to an atomic fraction of
approximately 1.7 %. Both the 28Si and the substrate spectra show Ar peaks due to
the Ar sputter cleaning process. The expected instrumental background for C and
O in these scans is approximately 1 % to 2 %. This measurement is similar to the
previous XPS measurement of a 28Si sample deposited at LC–2 with similar atomic
fractions of C and O. The small N concentration in the sample was not previously
detected. No other major contaminants were detected in this sample. Given the
large background signal for these measurements, it is difficult to place a bound on
the expected Si purity for this sample.
5.7.3 SIMS
SIMS was used not only to get a more accurate measurement of N, C, and
O concentrations in the 28Si films than XPS can provide but to also analyze the
films for a broad range of trace impurities. The SIMS measurements discussed here
were performed at EAG Laboratories unless otherwise stated, and are of the major
isotope of each element. A SIMS depth profile of the concentration of 22 different
potential contaminants in a 28Si film deposited with a substrate temperature of
approximately 460 ◦C at DC–3 is shown in Fig. 5.41. Atomic concentrations of
28Si (circles and line), 29Si (squares and line), and 30Si (triangles and line) are also
shown vs. sputter depth as an indicator of the enriched film where the 29Si and 30Si
values are reduced. The minimum detected 29Si and 30Si concentrations can only be
taken as bounds on the enrichment of this sample because they are limited by the
measurement noise floor. At a depth of approximately 145 nm they return to their
311
Figure 5.41: SIMS depth profile of the concentration of 22 atomic species in a 28Si
sample deposited at 460 ◦C at DC–3. Atomic concentrations of 28Si (circles and
line), 29Si (squares and line), and 30Si (triangles and line) are also shown vs. sputter
depth as an indicator of the enriched film. The minimum detected 29Si and 30Si con-
centrations are limited by the measurement noise floor. At a depth of about 145 nm
they return to their natural abundance values in the Si(100) substrate (shaded re-
gion), indicating the film thickness. Depth profiles of the atomic concentrations of
many elements including light gases and metals (open and closed symbols and lines)
in the enriched film and substrate are shown vs. sputter depth. N, C, O, F, and Al
were detected in the film while the remaining elements were not, indicated by the
box in the legend. The atomic concentration of N in the film is 7.1(1) ×1020 cm−3,
the concentration of C in the film is 4.4(2) ×1019 cm−3, and the concentration of
O in the film is 2.1(7) ×1019 cm−3. The signals for these elements drop to the
measurement detection limit in the substrate.
natural abundance values in the Si(100) substrate, which is marked by the shaded
region. This indicates the film interface with the substrate and gives a value for
the film thickness. Depth profiles of the atomic concentrations of many elements
that are potential contaminants in the 28Si films, including light gases and metals,
are shown vs. the sputter depth. N, C, O, F, and Al (solid lines) were all detected
312
in the film. The remaining elements Cl, Be, P, Na, K, Ti, V, Cr, Fe, Ni, Cu, Au,
Ag, As, Sb, Ta, and W (open and closed symbols and lines) were not detected in
the film or substrate down to the measurement detection limit, which is an atomic
concentration of approximately 1 ×1016 cm−3 for Sb, Au, As, and P, and is an
atomic concentration of approximately 5 ×1015 cm−3 for the remaining elements.
The average atomic concentration of N within the film was measured to be
7.1(1) ×1020 cm−3, or 1.42(2) %. The uncertainties of the values of atomic concen-
trations for the SIMS measurements discussed here are the standard deviations of
the means. This value is similar to the value of the N concentration of a 28Si sample
determined from XPS of approximately 1.7 %. The average atomic concentration of
C in the film was measured to be 4.4(2) ×1019 cm−3, or 880(40) ppm, and the aver-
age atomic concentration of O in the film was measured to be 2.1(7) ×1019 cm−3, or
420(140) ppm. These values are significantly less than the C and O values measured
by XPS. This may be due in part to the high measurement background in XPS for
C and O. The differences between the two samples is the ex situ cleaning procedure
(none vs. CMOS) and the deposition temperatures (812 ◦C vs. 460 ◦C). It may be
that an increase in the background pressure during deposition of the 812 ◦C sample
due to the high temperature of the sample and holder is responsible for increased C
and O concentrations. The signals for these elements drop to the detection limit of
the measurement in the substrate.
The structural relation between chemical impurities and the surrounding Si
crystal is dictated by the solid solubility of those impurities in Si. Small changes
in the atomic concentration of an impurity can greatly influence its effect on the
313
Figure 5.42: Phase diagrams for the N-Si, C-Si, and O-Si systems with low impurity
concentrations near the solid solubility limit. (a) The N-Si phase diagram near the
melting point shows the solid solubility of N in Si is about 4.5 ×1015 cm−3 and above
this, Si3N4 forms. (b) The C-Si phase diagram near the melting point shows the
solid solubility of C in Si is about 3.2 ×1017 cm−3 and above this, SiC forms. (c)
The O-Si phase diagram near the melting point shows the solid solubility of O in
Si is about 2.8 ×1018 cm−3 and above this, SiO forms. ((a) and (c) from Ref. [171],
(b) from Ref. [172])
structural properties of the host Si. Phase diagrams for the N-Si, C-Si, and O-Si
systems in the case of extremely low impurity concentrations near the solid solubility
are shown in Fig. 5.42. The N-Si phase diagram from Ref. [171] in panel (a) shows
that the solid solubility limit of N in Si is approximately 4.5 ×1015 cm−3. At atomic
concentrations of N that are higher than this, Si3N4 crystallites form in the Si.
Similarly, the C-Si phase diagram from Ref. [172] in panel (b) shows that the solid
solubility limit of C in Si is approximately 3.2 ×1017 cm−3. At atomic concentrations
314
of C that are higher than this, SiC crystallites form in the Si. Finally, the O-Si phase
diagram from Ref. [171] in panel (c) shows that the solid solubility limit of O in
Si is approximately 2.8 ×1018 cm−3. At atomic concentrations of O that are higher
than this, SiO clusters form in the Si. Based on these values, the measured atomic
concentrations of N, C, and O in the 460 ◦C sample analyzed in Fig. 5.41 are well
beyond their respective solid solubility limits in Si and so those contaminants likely
exist as Si3N4 and SiC crystallites, and SiO clusters in the
28Si film.
The average atomic concentration of F in the 460 ◦C sample analyzed in
Fig. 5.41 was measured to be 3.5(3) ×1016 cm−3, or 0.70(6) ppm. This F is be-
lieved to originate from the background vacuum in the deposition chamber and
incorporate into the depositing film. A partial pressure of F is always observed in
the RGA mass spectrum at a level larger than H2O, and can be seen in the RGA
spectrum recorded while operating the ion beam in Fig. 5.3. Several instruments
and apparatus components within the vacuum chamber contain teflon insulation
(PTFE), which is suspected to outgas F in UHV environments and contribute to
the observed partial pressure. These components include insulated wires on the
manipulator that supply power for sample heating, insulated wires on the STM tip
preparation tool, which heats STM tips, teflon support structures for the ion beam
deceleration lenses, and Viton seals on gate valves. The STM tip preparation tool
in particular was observed to significantly increase the partial pressure of F while
in use. Eliminating F-containing compounds within the vacuum system can reduce
the concentration of F in samples. The atomic concentration of Al detected in the
film is 6.2(3) ×1015 cm−3, or 0.12(2) ppm. The origin of this Al is an Al deposition
315
source sitting above the sample location that consists of an Al wire wrapped around
a tungsten heating element.
The total purity of this 28Si sample calculated from these measurements is
approximately 98.45(2) %, which is quite low compared to commercial Si wafers.
The concentrations of N, C, and O in the sample are far too high for the material
to be viable for use in quantum coherent devices or other QI related experiments.
The concentration of nuclear spins due to the 14N (I = 1) in the sample is much
larger than the residual 29Si isotopic concentration and is only roughly a factor of
three lower than that of 29Si in natural abundance Si. It is difficult to determine
if the N, C, and O detected in these films predominantly originates from the ion
beam or the background vacuum. It is unlikely that the N concentration is due to
adsorption from the vacuum because the partial pressure of N2 needed to account
for the measured concentration (with unity sticking) is higher than the measured
partial pressure of N2 in the chamber during deposition. This indicates that most,
i.e. likely > 90 %, of the N in this sample was introduced through the ion beam.
A second 28Si sample deposited at DC–3 was also analyzed by SIMS for chem-
ical contaminants after the first sample. This second sample had a deposition tem-
perature of approximately 421 ◦C. The only significant differences between the first
sample and this second one are that the deposition rate of the second sample was
roughly two times larger and the gas manifold was purged with Ar instead of N2
before deposition. It was thought that the previous N2 purge may have contributed
to the high atomic concentration of N measured in the first sample, although this
was not verified. The atomic concentrations of N, C, and O were all lower than
316
those of the first sample but not significantly, with slightly over half the total im-
purities. The average atomic concentration of N in this film was measured to be
4.16(5) ×1020 cm−3, or 0.83(1) %. The average atomic concentration of C in the film
was measured to be 1.81(2) ×1019 cm−3, or 363(4) ppm, and the average atomic con-
centration of O in the film was measured to be 6.93(6) ×1018 cm−3, or 139(1) ppm.
The total Si purity of this second sample is approximately 99.12(1) %.
In total, three SIMS measurements were made on this sample in three different
locations on the deposition spot. These three locations had three different measured
film thicknesses, which also means that they had three different local deposition
rates. The measurements of the chemical contaminants for the three locations can
then be compared for different deposition rates while other variables including the
background pressure are constant. Such an analysis can give insight into the source
of the contaminants. For adsorption from the background vacuum, an increasing
deposition rate would result in a decreasing contaminant concentration because the
gas flux is decreasing relative to the ion beam flux. Contaminants incorporated from
the ion beam, however, should remain constant for different deposition rates. The
three deposition rates determined from the SIMS depth profiles of the three spots
on this sample are approximately 0.11 nm/min, 0.32 nm/min, and 0.88 nm/min.
The N concentrations were found to not decrease with increasing deposition rate
and instead increased roughly 39 % from the area with the lowest deposition rate to
that with the highest. This again indicates that the detected N is not significantly
due to adsorption from the vacuum. C and O concentrations, however, were found to
decrease with increasing deposition rate. The C concentration decreased by roughly
317
71 % from the area with the lowest deposition rate to that with the highest. This
result indicates that the C and O concentrations measured in these films are due,
at least partially, to adsorption from the vacuum.
To determine the minimum N, C, and O contamination that is contributed
to samples by adsorption and incorporation during 28Si deposition, any effects of
the ion beam must be eliminated. To this end, a natural abundance Si sample was
deposited in the deposition chamber at DC–3 and analyzed by SIMS for chemical
contaminants. This sample was deposited by sublimating Si from a bare substrate
held at approximately 1150 ◦C and positioned over the surface of the target sub-
strate. The target Si(100) substrate was flash annealed before the deposition and
was nominally at room temperature during it, although it may have been heated
radiatively from the source substrate.
SIMS depth profiles of the atomic concentrations of N, C, and Cl (solid lines)
in the natSi film are shown vs. sputter depth in Fig. 5.43. An O profile did not
yield usable results due to atmospheric contamination. The profiles show slightly
increased values near the surface due to environmental surface contamination before
leveling off through the bulk of the film. The atomic concentrations of N and C
then dip just beyond 200 nm before peaking at approximately 227 nm. This peak
corresponds to the interface between the film and the substrate, which is marked
by the shaded region and indicates the film thickness. The N and C signals peak
at the interface because adsorbates will accumulate on the surface from the vacuum
before deposition begins. Initial heating of the sublimation source will also increase
the partial pressures of adsorbates, increasing the contaminant concentrations in
318
Figure 5.43: SIMS depth profile of a natSi film deposited at room temperature at
DC–3. This film was deposited on a Si(100) substrate by sublimating Si from a
second wafer held over the surface. Atomic concentrations of N, C, and Cl in the
deposited film and substrate are shown vs. sputter depth. The interface between the
film and substrate (shaded region) is indicated by the peak in the N and O signals
at a depth of about 227 nm, which also indicates the film thickness. The minimum
atomic concentration of N in the film at about 206 nm is 1.1(1) ×1017 cm−3, and
the minimum atomic concentration of C in the film is 1.43(3) ×1018 cm−3.
the film. The N, C, and Cl signals in the substrate are at the detection limit of the
measurement. The dip in the atomic concentrations of N and C near the interface
gives an upper bound on the minimum possible N and C concentrations that would
be expected in a 28Si film deposited without contamination from other sources such
as the ion beam.
The minimum atomic concentration of N in the natSi film was measured to
be 1.1(1) ×1017 cm−3, or 2.2(2) ppm. The minimum atomic concentration of C
in the film was measured to be 1.43(3) ×1018 cm−3, or 28.7(6) ppm. The higher
319
concentrations of N and C throughout the bulk of the film are likely due to increased
outgassing of these elements when the target sample began to be heated by the
source. While the detected atomic concentration of N in the natSi film is roughly
6.5 ×103 times lower than that measured in the first 28Si film, it is still more than
10 times higher than the 29Si isotopic concentration in the more highly enriched 28Si
samples produced in this work. Additional pumping or a faster growth rate may be
needed to reduce contaminants from the vacuum to concentrations that meet the
purity goal for 28Si films of 2× 1015 cm−3 stated at the beginning of this section.
Cl is also detected within this film with an average atomic concentration of
8.5(8) ×1016 cm−3, or 1.7(2) ppm. This Cl may be outgassing from the sample
holders or other elements near the sample that get hot because there is no peak in
the concentration near the surface due to build up from the vacuum. This means
that Cl only appears to be present once the sample heating was turned on to start
the deposition. Evidence of Cl in the vacuum chamber can be seen in the residual
gas mass spectrum of the RGA. During heating of samples and sample holders for
either degassing or sample flashing, several mass peaks can appear that are not
normally present in the chamber, particularly peaks at masses of 31 u and 50 u. An
example of a residual gas mass spectrum from the RGA that was recorded while a Si
substrate was being degassed is shown in Fig. 5.44. A spectrum of the base pressure
of the deposition chamber (diagonal line fill) recorded at a later date is given for
reference and shows major residual gas peaks corresponding to H2 (2 u), F (19 u),
CO and N2 (28 u), and CO2 (44 u). The Si substrate was degassed at approximately
600 ◦C while recording a RGA spectrum (solid fill). During degassing, the pressure
320
Figure 5.44: Residual gas mass spectra collected from the RGA in the deposition
chamber while degassing a Si substrate at about 600 ◦C. The base pressure of the
chamber (diagonal line fill) is shown for reference with peaks corresponding to H2,
F, CO and N2, and CO2 visible. When the substrate is heated for degassing, the
pressure increases and several new peaks appear (solid fill) including C and those
at masses 31 u and 50 u, which may indicate Cl-containing molecules.
rises from the typical base pressure, evidenced by the aforementioned peaks rising.
Additionally, several other peaks appear in the spectrum that are not present or
much smaller in the base pressure spectrum including a C peak as well as peaks at
masses of 31 u and 50 u.
There are several chemical compounds that are likely candidates for being as-
sociated with these masses in a vacuum environment, some of which contain Cl.
Chloromethane (CH3Cl) and related molecules Cl2F2 and CF3Cl all contribute sig-
nificant signals to mass 50 u if present in a vacuum. CF3Cl also appears at mass 31.
The compounds containing F are possible because F is known to be present in the
321
system, as seen in the base pressure spectra. Other possible candidates for the peak
at 31 u are ethanol and propanol. The source of these alcohols as well as Cl may be
the ex situ chemical cleaning procedure used on the substrates. The SC–2 solution
used contains HCl and so it may be that some residue from the clean resides on the
chip and is then released in the chamber during degassing and flashing procedures
as well as when samples are heated during deposition of Si films.
N2 and CO contamination in the
28Si ion beam can be measured in both the
mass spectrum and in a deposited 28Si sample if the concentrations are high enough.
A demonstration of this resulted from the case of a leak from air in the gas manifold
system that delivers SiH4 to the ion source. An increase in the concentration of
N2 and O2 in the ion source while it is running then leads to increased amounts of
14N+2 and
12C16O+ ions being created and extracted into the beamline. Tuning the
sector mass analyzer to select for 28Si then also selects these contaminants which
are measured in the SiH4 mass spectrum or deposited.
28Si samples which were
deposited with these conditions are referred to here as “N-contaminated” samples.
A mass spectrum of SiH4 that shows a large amount of N2 and/or CO contamination
in the ion beam is shown in Fig. 5.45. The ion currents shown in this figure (open
and closed circles) were recorded while sweeping the mass analyzer current, and thus
the magnetic field of the analyzer (top axes). Panel (a) shows a semi-log plot of a
mass spectrum used for depositing a 28Si sample at DC–3 using the low pressure
plasma mode of the ion source. The typical SiH4 current peaks are seen including
the 28 u peak, which is 28Si, and the 29 u peak, which is both 28SiH and 29Si. The
29 u peak is assumed to be ≈ 5 % 29Si based on previous mass spectrums. Several
322
Figure 5.45: SiH4 ion beam mass spectrum for N-contaminated
28Si samples de-
posited at DC–3 using the low pressure plasma mode of the ion source. The ion
currents (open and closed circles) are recorded while sweeping the mass analyzer
current, and thus the magnetic field (top axes). (a) Mass spectrum for samples
contaminated with large amounts of N shown on a semi-log scale with current peaks
typical of SiH4. The 28 u peak is
28Si and the 29 u peak is both 28SiH and 29Si,
assumed to be ≈ 5 % of the peak. Several higher order hydrides are also shown.
Gaussian fits (line, Eq. (2.14)) to the 28 u and 29 u peaks are shown superimposed
on the data. The 28 u peak appears higher than the 29 u peak, which is atypical.
(b) Comparison between the spectrum in (a) (closed circles and line) and a nominal
SiH4 spectrum (open circles and line) plotted on a linear scale. The difference in
signal of the 28 u peaks of the two spectra relative to the 29 u peak is clear. The
additional current in the 28 u peak beyond the nominal 28Si current is presumed to
be mostly N2.
323
higher order hydride peaks are also shown. Gaussian fits (line, Eq. (2.14)) to the
28 u and 29 u peaks are shown superimposed on the data. Unlike a typical SiH4
mass spectrum for the low pressure mode such as the one acquired before depositing
a 28Si sample shown in Fig. 5.4, the 28 u peak appears higher than the 29 u peak
here, indicated by the horizontal dashed lines. The increased current in the 28 u
peak is due to the contaminants.
The difference between the spectra is highlighted in panel (b) of Fig. 5.45,
which compares the contaminated mass spectrum (closed circles and line) to a nom-
inal SiH4 mass spectrum (open circles and line) from Fig. 5.4 on a linear current
scale. These two spectra are comparable because the current levels at the 29 u
peak are nearly identical. The difference in the current peak height of the contam-
inated spectrum compared to the nominal spectrum is clear here. The additional
current in the 28 u peak beyond the nominal 28Si current is presumed to be mostly
N2. This contamination comprises approximately 30 % of the 28 u current. For a
contaminant beam of only N2, this would actually amount to approximately 46 %
contamination in the sample because each N2 ion contains two N atoms.
A 28Si sample was then deposited using a contaminated ion beam similar
to that represented by Fig. 5.45, and it was analyzed by SIMS for chemical con-
taminants. This SIMS analysis was done in collaboration with Dr. David Simons
(NIST). SIMS depth profiles of the atomic concentrations of 14N, 12C, 16O, (lines)
and 13C (circles and line) in the N-contaminated 28Si sample deposited with a sub-
strate temperature of approximately 712 ◦C at DC–3 are shown vs. sputter depth
in Fig. 5.46. Also shown is a depth profile of the 29Si isotope fractions (squares
324
Figure 5.46: SIMS depth profiles showing the atomic concentration of 14N, 12C, 16O,
(lines) and 13C (circles and line) vs. sputter depth in a 28Si sample deposited at
712 ◦C at DC–3. Also shown is a depth profile of the 29Si isotope fraction (squares
and line) vs. sputter depth on the right axis as an indicator of the enriched film.
The average 29Si isotope fraction in the film is about 0.132(27) ppm. At a depth
of about 256 nm, the 29Si isotope fraction returns to the natural abundance value
in the Si(100) substrate (shaded region), which gives a value for the film thickness.
The atomic concentration of N in the 28Si film is greater than 1 ×1022 cm−3. This
value is not quantitatively accurate, but shows that N is present in the film at a
concentration of roughly 30 %. The tail of the N profile into the substrate is an
artifact of the SIMS measurement. The measurements of C and O are accurate and
show that the concentration of 12C in the film is 6.1(1) ×1019 cm−3, the concentra-
tion of 13C in the film is 1.46(3) ×1017 cm−3, and the concentration of 16O in the
film is 2.3(1) ×1019 cm−3. The signals for these elements in the substrate are due
to the measurement background.
and line) vs. sputter depth, which corresponds to the right axis as an indicator of
the deposited enriched film. The average measured 29Si isotope fraction in the film
is 0.132(27) ppm. At a depth of approximately 256 nm, the 29Si isotope fractions
return to the natural abundance value in the Si(100) substrate, which is indicated
by the shaded region and gives a value for the film thickness. The atomic concen-
325
tration of 14N detected in this film is extremely high. The measurement of such a
high concentration is not quantitatively accurate for the measurement conditions
used here, but it appears greater than 1 ×1022 cm−3. This measurement does show
that 14N is present in the film at a likely concentration of roughly 30 %. This rough
SIMS value for the N contamination is supported by other measurements of simi-
lar N-contaminated films including XPS and energy dispersive x-ray spectroscopy
(EDX), which give similar values. With this much N present in the 28Si film, it is
likely that a significant amount of silicon nitride (Si3N4) forms. The long tail of
the 14N profile that extends into the substrate is an artifact of the SIMS measure-
ment. 12C and 16O have concentrations much lower than that of 14N. The average
atomic concentration of 16O in the film was measured to be 2.3(1) ×1019 cm−3, or
470(20) ppm. This concentration of O is only slightly higher than that of the previ-
ous 28Si sample, which was not deposited with a contaminated beam. The average
atomic concentration of 12C in the film was measured to be 6.1(1) ×1019 cm−3, or
1240(20) ppm, and the average atomic concentration of 13C in the film was measured
to be 1.46(3) ×1017 cm−3, or 2.98(6) ppm. This concentration of C increased over
that of the previous 28Si sample by roughly the same amount that the concentration
of O increased, which is about 2 ×1019 cm−3. The signals for N, C, and O in the
substrate are due to the background level of the measurement.
The ratio of 12C to 13C in this sample can indicate its origin because only 12C
is selected through the ion beam while both isotopes are present in their natural
abundance coming from the vacuum. In this sample, the C ratio is measured to
be 12C/13C ≈ 420. This ratio is larger than the natural abundance ratio of ap-
326
proximately 89.9, meaning that the C is being enriched and thus must be at least
partially coming from the ion beam. Assuming that all the 13C originates from
the background pressure in the vacuum and using the C natural abundance ratio,
it is determined that the atomic concentration of 12C from the vacuum is approxi-
mately 1.3 ×1019 cm−3, and the atomic concentration of 12C from the ion beam is
approximately 4.8 ×1019 cm−3 in this sample.
In order to produce 28Si film with higher purities, several experimental im-
provements were made to the system, including the aforementioned switch to using
Ar to purge the gas manifold instead of N2. First, the Al deposition source was
relocated to another part of the chamber and away from the manipulator where
the sample sits during deposition at DC–3. Next, some in-vacuum components
that outgas F, specifically PTFE-coated wires on the manipulator and the STM
tip preparation tool, were removed from the system. These wires were replaced
with Kapton-coated wires. Then, the background pressure in both the deposition
chamber and the ion beam chamber were reduced to minimize contaminants ad-
sorbing from the vacuum during deposition. In the ion beam chamber, a gate
valve that was not rated for UHV was removed from the system. This resulted
in a reduction in the base pressure of the ion beam chamber from approximately
1.3 ×10−5 Pa (9.4 ×10−8 Torr) when the previously analyzed 28Si sample was de-
posited (Fig. 5.41) to approximately 3.9 ×10−6 Pa (2.9 ×10−8 Torr) when the first
28Si sample was deposited after these changes were made. The base pressure in
the deposition chamber was reduced by installing a new TSP to add more pumping
capacity. Also, it was baked more thoroughly than what was previously done and
327
at higher temperatures of 150 ◦C to 200 ◦C. These factors resulted in a reduction
in the base pressure from approximately 2.3 ×10−8 Pa (1.7 ×10−10 Torr) for the
previously analyzed sample to 8.3 ×10−9 Pa (6.2 ×10−11 Torr) for the next sample
deposited after these changes. Finally, for samples deposited after these experimen-
tal improvements, the high pressure plasma mode of the ion source was used. This
mode produces higher 28Si ion fluxes which resulted in increased growth rates of
samples. Depositing with higher growth rates should reduce the concentration of
contaminants adsorbed from the vacuum. Also, the higher cracking efficiency of the
high pressure mode may result in cracking of some N2 molecules in the ion source,
which would eliminate them from the 28 u ion beam.
After enacting these changes, another 28Si sample was deposited and analyzed
for chemical contaminants by SIMS. This sample was deposited with a substrate
temperature of approximately 460 ◦C at DC–3 using the high pressure mode of the
ion source. SIMS depth profiles of the atomic concentrations of N, C, O, F, Cl, Al,
and Mo (lines) in this 28Si sample are shown vs. sputter depth in Fig. 5.47. Mo was
analyzed in this sample to check for signs of contamination due to the Mo sample
holder that contacts the substrate. Also shown are depth profiles of 28Si (circles and
line), 29Si (squares and line), and 30Si (triangles and line), which correspond to the
right axis in arbitrary units related to the count rate that are roughly aligned to the
atomic concentrations on the left axis, as an indicator of the 28Si film. At a depth
of approximately 293 nm, the concentrations of the Si isotopes increase and return
to their natural abundance values in the Si(100) substrate, indicated by the shaded
region, and giving a value for the film thickness. The minimum detected 29Si and
328
Figure 5.47: SIMS depth profiles of the concentration of contaminants in a 28Si
film deposited at 460 ◦C at DC–3 using the high pressure mode of the ion source.
Atomic concentrations of 28Si (circles and line), 29Si (squares and line), and 30Si
(triangles and line) are also shown vs. sputter depth as an indicator of the 28Si film.
The Si concentrations correspond to the right axis, displayed in arbitrary units. The
minimum detected 29Si and 30Si concentrations are limited by the measurement noise
floor. At a depth of about 292 nm they return to their natural abundance values in
the Si(100) substrate (shaded region), indicating the film thickness. Depth profiles of
the atomic concentrations of N, C, O, F, Cl, Al, and Mo (lines) are shown vs. sputter
depth. N, C, O, and Cl were detected in the film. F and Al were not detected and
their signals are at the measurement detection limit. The concentration of N in the
film is 9.07(4) ×1018 cm−3, the concentration of C in the film is 8.27(3) ×1018 cm−3,
and the concentration of O in the film is 3.09(4) ×1018 cm−3. The signals for these
elements drop to the measurement detection limit in the substrate. It is unclear
why Cl is detected at two different concentrations in the film.
30Si concentrations are taken as bounds on the enrichment because they are limited
by the measurement noise floor, which for 29Si is approximately 20 ppm.
In this 28Si sample, N, C, O, and Cl are all detected in the film, while F and
Al are not detected. The signals for F and Al are at the detection limit of the
measurement in the film and substrate, which is approximately 1 ×1017 cm−3 for F
329
and 3 ×1015 cm−3 for Al. The elimination of Al in this sample compared to the pre-
vious sample is likely due to moving the Al deposition source away from the sample
location. The elimination of F in this sample compared to the previous sample may
be due in part to the removal of PTFE-coated wires in the chamber, although the
partial pressure of F in the chamber was still present after removing them. It was
unclear if the concentration of F in the vacuum was reduced because a different RGA
was used to measure residual gases after these experimental changes were made, and
it is difficult to compare the absolute values of the partial pressures to those of the
previous instrument. It is unclear why Cl is detected at two different concentra-
tions within the film. From the surface down to a depth of approximately 100 nm,
the average atomic concentration of Cl is 4.11(4) ×1017 cm−3, or 8.24(8) ppm. The
atomic concentration of Cl then drops to 8.6(3) ×1015 cm−3, or 0.172(6) ppm in
the remainder of the film. This indicates that something changed roughly midway
through the deposition. It may be that part of the sample holder started heating
slowly during the deposition which caused increased outgassing including a Cl com-
pound, as mentioned previously. Throughout most of the film the Mo signal is also
at the detection limit, although there may be a slight increase between a depth of
50 nm and 100 nm to an atomic concentration of 36(8) ppb.
The atomic concentrations of N, C, and O in this 28Si sample are clearly re-
duced from those of the previous sample shown in Fig. 5.41. Note that the scale
of the vertical axis of that figure and Fig. 5.47 are the same to facilitate easier
comparisons. The average atomic concentration of N in the film was measured to
be 9.07(4) ×1018 cm−3, or 181.8(8) ppm. N was reduced compared to the previ-
330
ous sample by a factor of roughly 78. The average atomic concentration of C in
the film was measured to be 8.27(3) ×1018 cm−3, or 165.7(6) ppm. C was reduced
compared to the previous sample by a factor of roughly 5. Finally, the average
atomic concentration of O in the film was measured to be 3.09(4) ×1018 cm−3, or
61.9(8) ppm, which is reduced by almost a factor of 10 compared to the previous
sample. The total purity for this 28Si sample determined from these measurements
is approximately 99.96(2) %. The atomic concentrations of N and C in this sample
are above their solid solubility limits, similar to the previous sample. Additionally,
the atomic concentration of O in this sample is also slightly over its solid solubility
limit in Si. This indicates that Si3N4, SiC, and SiO are all likely present in the
28Si
film.
The reduction of N in the second sample compared to the first seems likely
due to changes in the ion beam. This is because the partial pressure of N2 in the
deposition chamber during deposition appears to be similar for both of the sam-
ples, although it is difficult to determine precisely because of the overlap of N2 and
28Si at 28 u in the residual gas mass spectrum of the RGA as well as other uncer-
tainties. The deposition rate of the final sample deposited using the high pressure
mode (2.21 nm/min) was roughly six times higher than that of the first sample
(0.37 nm/min), which could account for some of the decrease in the N concentra-
tion if the N originated from the background vacuum. However, the reduction in
the N concentration is still more than an order of magnitude lower than what would
be expected just due to the increased deposition rate. So, the remaining reduction
may be due to the ion beam. Use of the high pressure mode of the ion source for
331
the final sample may lead to a lower concentration of N in the 28 u beam relative to
28Si. If the N2 ion flux for both the low pressure and high pressure modes is similar,
then the roughly factor of five increase in 28Si beam flux (ion current) for the final
sample compared to the first would result in a lower concentration of N relative
to 28Si. Additionally, the possibility of N2 molecules being cracked more efficiently
by the high pressure mode into atomic N and thus eliminated from the 28 u beam
would also reduce the relative N concentration. The results from several samples
do seem to support this hypothesis showing that the atomic concentration of N
measured in the samples is inversely related to the 28Si ion beam current, although
only one of those samples was deposited using the high pressure mode. This inverse
relationship can be seen in Fig. 5.48 showing the atomic concentration of N as well
as C and O vs. the total 28 u ion current used for deposition. The data with ion
currents around 0.5 µA correspond to the first two samples deposited using the low
pressure mode discussed above, while the data with an ion current of roughly 2.8 µA
corresponds to the final sample discussed in this section, which was deposited using
the high pressure mode. The uncertainties in the atomic concentrations (most are
smaller than the data symbols) are the standard deviations of the means, and an
uncertainty of 50 nA was assigned to the values of the ion current. The atomic
concentration of N (open squares) appears to have a strong inverse relationship to
the ion current, decreasing roughly a factor of 80 while the 28 u ion beam current
increases roughly a factor of five. This trend suggests that the higher concentrations
of N measured in the first samples are likely due to ballistic incorporation from the
ion beam. This also indicates that both increasing the 28Si ion current and use of
332
Figure 5.48: Atomic concentrations of N (open squares), C (open circles), and O
(open triangles) for three 28Si samples vs. the ion current used for deposition of
the samples. The data with ion currents around 0.5 µA correspond to the samples
deposited using the low pressure mode, while the data with an ion current of 2.8 µA
corresponds to the sample deposited using the high pressure mode. The concentra-
tions of N, C, and O all show an inverse relationship to the ion current, but N varies
much more strongly than C and O, indicating that N contamination is more likely
due to N2 in the ion beam.
the high pressure mode do reduce the relative N2 concentration in the beam. The
atomic concentrations of C (open circles) and O (open triangles) are also inversely
related to the ion current, but more weakly so than N. The decrease in the concen-
trations of C and O are likely mostly due to the increased deposition rates resulting
from the increased ion current, similar to the result discussed above for the three
measurements of the second analyzed sample.
The remaining N, C, and O in the final sample are difficult to attribute to
either adsorption from the vacuum or deposition from the ion beam, and their
presence is likely due to a combination of both. Either way, the solution to reducing
333
them further is a reduction of the partial pressures of these elements in the vacuum
because the contaminants in the ion beam ultimately originate from the vacuum in
the ion source. There also seems to be a significant amount of N2 or CO entering
the chamber with the SiH4 gas from the gas manifold, probably more than what is
due to the base pressures of the two chambers. Reducing N, C, and O concentration
in the deposition chamber, ion beam chamber, and the gas manifold is required
to reliably improve the purity of 28Si films deposited with this system in order to
achieve the third materials goal discussed here.
5.8 Crystallinity: Film Inspection via TEM
The 28Si samples deposited at elevated temperatures at sample location DC–3
exhibit different morphologies depending on the deposition temperature, as dis-
cussed previously in this chapter. Rough films produced with deposition tempera-
tures above 600 ◦C and smooth films produced with deposition temperatures below
600 ◦C both appear crystalline and epitaxially aligned to the underlying substrate
based on observations with RHEED and STM. While defects at the surface of a
depositing film will both cause and develop from step pinning and step bunching,
it is not clear what crystalline defects are formed in the bulk of the films for the
two deposition temperature ranges. Here, TEM is used to image and characterize
crystalline defects in 28Si films. TEM micrographs presented here were acquired in
collaboration with Dr. Alline Myers and Dr. Vladimir Oleshko.
TEM is a common analysis tool used for inspecting the epitaxial quality of
334
and defects in films produced by low temperature Si deposition including MBE and
IBE. A common type of defect observed in Si films deposited on Si(100) substrates is
{111} stacking faults. As seen in the TEM micrographs of a Si film deposited by IAD
in Fig. 5.17, {111} stacking faults and defects start building up in the epitaxial layer
until an amorphous phase develops. IBE experiments have demonstrated epitaxial
28Si films and used TEM to show, depending on the deposition conditions, both
defective films with {111} stacking faults and microtwins, as well as higher quality
epitaxial films without visible stacking faults or obvious dislocations. However, TEM
inspection shows that these films likely still contain defects and strain [43,51].
Chemical contaminants can not only affect the morphology of a depositing
film through surface defect formation, they can also cause structural defects and
amorphization in the bulk of a film when incorporated during deposition. Several
28Si IBE experiments have been done showing the effects of chemical contaminants
introduced through the ion beam during deposition on the epitaxial quality of the
film. One experiment found that introducing approximately 1 % N2 into the
28Si ion
beam during deposition caused the resulting film to be amorphous at a deposition
temperature of 350 ◦C [169]. Others found that N2 and CO in the ion beam resulted
in highly defective or amorphous films using a range of deposition temperatures [49,
50]. Finally, an experiment found that using 30Si ions to deposit a film resulted in
epitaxial growth that was less defective than 28Si, and concluded that trace amounts
of CO present in the 28Si beam, which are very difficult to eliminate, were the
cause [134]. These experiments show the importance of reducing the concentration
of N2 and CO in the ion beam, as discussed in the previous section.
335
The lattice constant mismatch between natural abundance Si and 28Si may also
introduce dislocations into the film, which may be observed by TEM. As mentioned
previously, however, the effect of the roughly 1× 10−6 relative difference in lattice
constants is likely too small compared to other defect causing mechanisms in these
films.
Initially, TEM was used to inspect the bulk crystallinity of rough 28Si films
deposited at higher substrate temperatures above 600 ◦C to determine the effect
on the crystallinity of the surface roughness. Additionally, it was used to confirm
the observations made using RHEED and STM that, despite the rough surface, the
films are still crystalline and epitaxially aligned to the Si(100) substrate. TEM cross-
sectional micrographs of a rough 28Si sample deposited with a substrate temperature
of approximately 708 ◦C at DC–3 are shown in Fig. 5.49. The substrate used for
this sample was not cleaned ex situ and was flashed annealed before deposition. The
TEM specimen was prepared using a FIB, and these images were taken on the 〈110〉
zone axis. Panel (a) is a bright field image showing the 28Si film above the Si(100)
substrate at the bottom of the micrograph. A protective, thin layer of C (light)
and a thicker layer of Pt (dark) are seen above the film. The 28Si film consists
of large mounds resulting in a very rough surface, as previously observed in the
SEM micrographs of this and other rough samples (Fig. 5.22). The maximum film
thickness (of the central mound) seen in this micrograph is approximately 116 nm
and the minimum thickness is approximately 31 nm. The surface of the mounds
comprising the film are faceted with predominately {113} and {111} microfacets
visible, indicated by the arrows. The surfaces of the {113} microfacets make an
336
Figure 5.49: TEM cross-sectional micrographs of a rough 28Si sample deposited at
708 ◦C at DC–3. This image was taken on the 〈110〉 zone axis. (a) Bright field image
with the Si(100) substrate is visible at the bottom of the image and the 28Si film
above it, which consists of large mounds. Protective layers of C (light) and Pt (dark)
were deposited on the 28Si film. The surface of the mounds are faceted with {113}
and {111} microfacets visible (arrows). Within the mounds, {111} stacking faults
are seen. (b) HR-TEM image of a group of {111} stacking faults and microtwins
that run from the interface of the film and the substrate up through the film. The
28Si film is seen to be crystalline and epitaxially aligned to the substrate, evidenced
by the continuation of 〈111〉 lattice rows across the interface.
337
angle with the surface of the substrate, or the substrate interface that matches the
expected angle between the 〈113〉 and 〈100〉 planes of approximately 25.2◦. Likewise,
surface of the {111} microfacets make an angle with the surface of the substrate that
matches the expected angle between the 〈111〉 and 〈100〉 planes of approximately
54.7◦. These microfacets match those that were observed in the RHEED diffraction
patterns for similar samples shown in Fig. 5.20. Also, within the mounds, multiple
stacking faults are visible running through the film along the 〈111〉 planes.
Panel (b) of Fig. 5.49 is an HR-TEM image that shows another region of the
same film in (a) at a higher magnification. The substrate is seen at the bottom
of the micrograph with the 28Si film above it. The 28Si film is crystalline and
epitaxially aligned to the substrate, as evidenced by the continuation of 〈111〉 lattice
rows across the substrate interface into the film. A group of {111} stacking faults
and microtwins are seen originating at the substrate interface and running through
the film. The presence of microtwins is evidenced by the dark fringes inside the
stacking fault appearing with a periodicity of three times the normal 〈111〉 lattice
row spacing, as is often observed for microtwins [173]. These are possibly due to
defects or contaminants such as SiC present on the surface at the beginning of the
growth that cause step bunching and defects to form on {111} planes.
Some areas of this sample appear with fewer stacking fault defects, and the
crystallinity can be inspected further and compared with that of the substrate using
fast fourier transform (FFT) analysis. Figure 5.50 shows an HR-TEM cross-sectional
micrograph of another area of the 708 ◦C sample analyzed using FFTs. Like with
the previous micrographs of this sample, this image was taken on the 〈110〉 zone
338
Figure 5.50: HR-TEM cross-sectional micrograph of a rough 28Si sample deposited
at 708 ◦C at DC–3. This image was taken on the 〈110〉 zone axis. (a) The Si(100)
substrate is at the bottom of the image, and the 28Si film is above it. Several {111}
stacking faults are visible running through the film. The 28Si film is seen to be
crystalline and epitaxially aligned to the substrate, evidenced by the continuation
of 〈111〉 lattice rows across the interface. (b) and (c) FFTs of the regions in the
boxes for the 28Si film and the substrate, respectively. The FFT of the film and
substrate show the same crystal pattern indicating that the film is aligned to the
substrate.
339
axis. Panel (a) shows the sample with the Si(100) substrate appearing dark at the
bottom of the micrograph and the 28Si film above it. Again, film appears epitaxially
aligned to the substrate, as evidenced by the continuation of 〈111〉 lattice rows across
the substrate interface into the film, indicated by the arrow. A few {111} stacking
faults are seen in the film as well. Panels (b) and (c) are FFTs of the regions marked
by the boxes in the substrate and 28Si film, respectively. These FFTs appear nearly
identical showing the sample crystal pattern and orientation, indicating again that
the film is epitaxially aligned to the substrate.
28Si samples made with lower deposition temperatures that were below 600 ◦C
were observed to be much smoother than those with high deposition temperatures,
and TEM was used to inspect these low deposition temperature samples for differ-
ences in the crystallinity compared to the samples with higher deposition temper-
atures. Samples with lower deposition temperatures were also prepared using the
revised cleaning procedures discussed previously in the chapter, which may affect
the bulk crystalline defects that develop in the film. TEM cross-sectional micro-
graphs at two magnifications of a smooth 28Si sample deposited with a substrate
temperature of approximately 460 ◦C at DC–3 are shown in Fig. 5.51. This sample
was prepared ex situ using the CMOS cleaning procedure and was flashed annealed
before being deposited using the low pressure mode of the ion source. These images
were taken on the 〈110〉 zone axis. Panel (a) shows a bright field image at lower
magnification with the Si(100) substrate in the lower left of the image appearing
lighter and the 28Si film to the right of that. The 28Si appears quite different from
the substrate with varying contrast and dark patches throughout the film. These
340
Figure 5.51: TEM cross-sectional micrographs of a smooth 28Si sample deposited
at 460 ◦C at DC–3 using the low pressure mode. These images were taken on the
〈110〉 zone axis. (a) Bright field image showing the Si(100) substrate in the lower left
appearing lighter and the 28Si film to the right of that with varied contrast. To the
upper right of the 28Si film are protective C (bright) and Pt (dark) layers. The 28Si
film thickness varies between about 105 nm and 110 nm here. (b) HR-TEM image
showing the substrate in the lower left and the 28Si film in the upper right. The 28Si
film is seen to be crystalline and epitaxially aligned to the substrate, evidenced by
the continuation of 〈111〉 lattice rows across the interface.
are likely defects causing local strain in the film. To the upper right of the 28Si are
layers of C (light) and Pt (dark) which were deposited to protect the sample during
preparation of the specimen, which was done using a FIB. The 28Si film thickness
varies between approximately 105 nm and 110 nm in this region of the film, meaning
there is a roughly 5 nm surface width here. Panel (b) shows an HR-TEM image
taken at much higher magnification (790 times). The crystalline Si(100) substrate
is seen in the lower left of the image and the 28Si film is in the upper right. The
dashed line representing the interface is only approximate because the true interface
is not clear at this magnification, and the 28Si appears similar to the substrate, un-
341
like in panel (a). The 28Si film in panel (b) is seen to be crystalline and epitaxially
aligned to the substrate, evidenced by the continuation of individual 〈111〉 lattice
rows throughout the image.
Another HR-TEM cross-sectional micrograph of this 28Si sample from Fig. 5.51
is shown in Fig. 5.52 and the crystallinity is analyzed using FFTs. This image was
taken on the 〈110〉 zone axis. Panel (a) shows the Si(100) substrate in the lower
left of the micrograph with the 28Si film in the upper right. The interface between
the substrate and the film is indicated by the dashed line and is less clear than in
the TEM micrographs of the previous sample. The film in this sample is crystalline
and epitaxially aligned to the substrate, which is evidenced by the continuation of
〈111〉 lattice rows across the interface, indicated by the arrow. No obvious stacking
faults are visible in the film in this micrograph or in any other areas of this film.
Several dark areas appear in the film but not the substrate, probably indicating
the presence of dislocation and other defects causing local strain fields in the film.
Also, the lattice rows in some areas of the film are not as clear as others or those
of the substrate, but these areas do not appear amorphous. Panels (b) and (c) are
FFTs of the regions marked by the boxes in the substrate and 28Si film, respectively.
These FFTs appear very similar showing the sample crystal pattern and orientation,
indicating that the film is, again, epitaxially aligned to the substrate.
This same 460 ◦C sample was analyzed by SIMS for chemical contaminants,
which was shown in Fig. 5.41, and found to have an atomic concentration of N of
approximately 7.1(1) ×1020 cm−3, or 1.42(2) %. It is thus surprising that the 28Si
film is not more defective given that it likely contains Si3N4 and SiC crystallites and
342
Figure 5.52: HR-TEM cross-sectional micrograph of a smooth 28Si sample deposited
at 460 ◦C at DC–3 using the low pressure mode. This image was taken on the 〈110〉
zone axis. (a) The Si(100) substrate is seen in the lower left and the 28Si film in
the upper right. The dashed line indicates the interface. The 28Si film is seen to be
crystalline and epitaxially aligned to the substrate, evidenced by the continuation
of 〈111〉 lattice rows across the interface. Several dark patches appear in the film
but not the substrate. (b) and (c) FFTs of the regions in the boxes for the substrate
and the 28Si film, respectively. The FFT of the film and substrate show the same
crystal pattern indicating that the film is aligned to the substrate.
343
that, as mentioned previously, 1 % N contamination can lead to an amorphous Si
film, although for a lower deposition temperature. The dark areas and other contrast
changes observed in the 28Si film may be related the Si3N4 and SiC compounds
present in the film and their related structural defects.
A reduction in the amount of contaminants was then achieved in the next
460 ◦C sample analyzed in Fig. 5.47 due to having lower background pressures in
the chambers and use of the high pressure mode of the ion source. A subsequent
sample deposited after these experimental changes and after analyzing the second
460 ◦C sample was then inspected using TEM. This was to determine if the reduction
in contaminants, particularly N and thus Si3N4 in the film, resulted in a reduction
of the dark patches in the film, seen in the TEM micrograph in Fig. 5.52. An HR-
TEM micrograph of this later 28Si sample deposited with a substrate temperature
of approximately 421 ◦C at DC–3 is shown in Fig. 5.53. This sample was prepared
ex situ using the CMOS cleaning procedure and was flashed annealed before being
deposited using the high pressure mode of the ion source. The TEM specimen was
prepared using a FIB, and this image was taken on the 〈110〉 zone axis. Panel (a)
shows the Si(100) substrate at the bottom of the micrograph and the 28Si film at
the top. The interface between the substrate and the film is roughly indicated by
the dashed line. Like the previous sample, the film in this sample is crystalline
and epitaxially aligned to the substrate, which is evidenced by the continuation of
〈111〉 lattice rows across the interface, indicated by the arrow. However, despite
the presumed reduction in chemical contaminants in this sample compared to that
shown in the previous TEM micrograph, this film appears much more defective.
344
Figure 5.53: HR-TEM cross-sectional micrograph of a smooth 28Si sample deposited
at 421 ◦C at DC–3 using the high pressure mode. This image was taken on the
〈110〉 zone axis. (a) The Si(100) substrate is seen at the bottom of the image
and the 28Si film is above it. The nearly horizontal dashed line roughly indicates
the interface. The 28Si film is seen to be crystalline and epitaxially aligned to the
substrate, evidenced by the continuation of 〈111〉 lattice rows across the interface.
{111} stacking faults are seen running through the film along with several dark
patches. (b) and (c) FFTs of the regions in the boxes for the substrate and the 28Si
film, respectively. The FFT of the film and substrate show the same crystal pattern
indicating that the film is aligned to the substrate.
345
Several {111} stacking faults are seen running through the film in this micrograph,
and further inspection of different areas of the film show that there are many more
stacking faults and other dislocations throughout the film. Dark patches are also
seen in this film, likely due to strain from defects. Panels (b) and (c) are FFTs of the
regions marked by the boxes in the substrate and 28Si film, respectively. These FFTs
appear very similar showing the sample crystal pattern and orientation, indicating
that the film is, again, epitaxially aligned to the substrate.
The more defective structure of the 421 ◦C sample compared to the previous
460 ◦C sample may be due to the deposition rate increasing from approximately
0.37 nm/min for the 460 ◦C sample to approximately 4.56 nm/min for the later
421 ◦C sample due to use of the high pressure deposition mode. These two factors
may move the quality of the film growth closer to the defective and strained region of
the epitaxy phase diagram for IBE, although there is no evidence of an amorphous
layer developing in this sample. In fact, none of the 28Si samples inspected with TEM
showed signs of a critical thickness, hepi, or an amorphous phase developing, and
they were always observed to be crystalline throughout the film including up to the
top surface. While the 28Si films inspected by TEM were crystalline and epitaxially
aligned to the substrate, the density of crystalline defects was too high to meet
the second materials goal stated at the beginning of this chapter. A reduction in
crystalline defects will likely require a reduction in chemical contaminants in these
film.
346
5.9 Chapter 5 Summary
The 28Si samples discussed in this chapter showed that 28Si films could be
deposited in the deposition chamber at sample location DC–3 while maintaining
residual 29Si and 30Si isotope fractions well below 1 ppm. Measurements also shows
that extremely high enrichments are achievable for samples deposited both with
elevated substrate temperatures and while using the high pressure plasma mode of
the ion source. In total, 40 28Si samples were produced at DC–3 with 39 of them
being deposited with elevated substrate temperatures. Additionally, seven of those
samples were deposited using the high pressure mode of the ion source. The residual
29Si isotope fraction of the most highly enriched sample deposited at DC–3 was
reduced by more than a factor of five compared to that of the previous most highly
enriched sample deposited at LC–2, going from 0.691(74) ppm to 127(29) ppb. This
progression can be seen in the enrichment progression timeline in Fig. 4.2.
The isotope reduction factor of 29Si for the most highly enriched sample de-
posited at DC–3 is 3.7(8)× 105. This value of the reduction factor along with the
values from all the other most highly enriched 28Si samples deposited at IC–1, LC–2,
and DC–3 are shown in an isotope reduction timeline, which is a progression of the
Si isotope reduction factors az/(
zSi/Sitot.), in Fig. 5.54. This timeline is a modified
version of the enrichment progression timeline, showing the isotope reduction fac-
tors vs. deposition date, where a larger reduction factor means a higher enrichment,
instead. Nine 28Si samples out of a total of 61 produced in this work are represented
on this timeline. As with the enrichment progression timeline, the nine samples
347
Figure 5.54: Isotope reduction timeline. A timeline of the progression of the isotope
reduction factors of the lowest residual isotope fractions of 29Si (squares) and 30Si
(triangles), as measured by SIMS. These were achieved for 28Si samples deposited
over approximately three and one half years. These results encompass samples
produced at IC–1, LC–2, and DC–3. Shown for comparison are the 29Si isotope
reduction factors of the 28Si epilayers and crystals produced by Isonics and Itoh
using 28SiH4 CVD (dash-double dotted line) from Ref. [36,37], the bulk
28Si material
produced by the IAC using 28SiH4 CVD (dash-dotted line) from Ref. [32], and the
28Si thin films produced by Tsubouchi et al. using 28Si IBE (dotted line) from
Ref. [43].
presented are those that achieved the best enrichment, or reduction factors of the
minor isotopes, of any sample deposited up to that point. Both the reduction factor
for 29Si (squares) and 30Si (triangles) are shown to increase in different samples over
time from 16.6(1) for the initial sample deposited at LC–1 to the afore mentioned
3.7(8)× 105 for the most highly enriched sample produced at DC–3. Uncertainties in
the reduction factors are derived from the uncertainties in the SIMS measurements
348
of the isotope fractions. Also shown for comparison are the 29Si isotope reduction
factors of three other sources of 28Si. The 28Si epilayers and crystals produced by
Isonics and Itoh using 28SiH4 CVD have a
29Si reduction factor of approximately
64 (dash-double dotted line), which is larger than only the first two 28Si samples
produced here at IC–1 [36, 37]. The bulk 28Si material produced by the IAC using
28SiH4 CVD has a
29Si reduction factor of 937 (dash-dotted line) [32], and the 28Si
thin films produced by Tsubouchi et al. using 28Si IBE has a 29Si reduction factor
of approximately 2.9 ×103 (dotted line) [43]. Both of these values are still below
the most highly enriched 28Si sample deposited at IC–1.
The overall 28Si isotope fraction of this most highly enriched sample deposited
at DC–3 and in this entire work was 99.9999819(35) %. These sample are more
highly enriched than any other known source of 28Si, including the IAC. These
samples also demonstrate enrichments sufficient to enable a robust measurement
of the dependance of electron coherence time on 29Si concentration in the single
spin regime and compare it to theoretical predictions (see Fig. 1.9), as proposed in
Chapter 1 [12].
The achieved reduction in 29Si and 30Si isotope fractions in samples deposited
at DC–3 was likely due to several factors. The deposition chamber had significantly
lower background pressures than the lens chamber at LC–2 resulting in less SiH4
adsorption during deposition. Also, higher deposition rates were generally achieved
for samples deposited at DC–3, including the highest rates achieved using the high
pressure mode of the ion source.
Depositing samples at DC–3 allowed for sample heating which was crucial in
349
achieving epitaxial deposition. Crystalline, epitaxial 28Si films were produced using
elevated deposition temperatures between 349 ◦C and 1041 ◦C, although samples
deposited above 600 ◦C were very rough. This was due to chemical contaminants
such as SiC at the growth surface that result in step pinning sites and lead to step
bunching, faceting on {111} and {113} planes, and large mound formation. 28Si
samples deposited at lower temperatures between approximately 349 ◦C and 460 ◦C
were found to be smooth with typical surface widths of ∆z = 2 nm.
Chemical contaminants in these 28Si films were measured by SIMS, which
detected N, C, O, F, Al, and Cl. The N, C, and O were detected at especially high
concentrations, initially all above 1× 1019 cm−3. F and Al were eliminated from a
second sample due to experimental alterations. By improving the vacuum in both
the deposition chamber and the ion beam chamber, and by using a higher 28Si ion
beam current generated in the high pressure mode of the ion source, N, C, and
O were all able to be reduced in the final sample analyzed by SIMS. The average
atomic concentration of N in the film was measured to be 9.07(4) ×1018 cm−3, or
181.8(8) ppm, the average atomic concentration of C in the film was measured to be
8.27(3) ×1018 cm−3, or 165.7(6) ppm, and the average atomic concentration of O in
the film was measured to be 3.09(4) ×1018 cm−3, or 61.9(8) ppm. Based on these
concentrations and the solid solubility of these elements in Si, it is likely that Si3N4,
SiC, and SiO exist within the 28Si films. The resulting total best purity for a 28Si
sample deposited at DC–3 and overall in this work is approximately 99.96(2) %.
Finally, TEM was used to confirm that 29Si films produced with both the higher
and lower deposition temperatures were crystalline and epitaxially aligned to the
350
substrates. All samples deposited with substrate temperatures above 600 ◦C were
observed to have {111} stacking faults and twinning present in the films. For lower
deposition temperature samples, one sample deposited using the low pressure mode
did not have any visible stacking faults but probably still contained other crystalline
defects. Another sample deposited with a slightly lower substrate temperature and
using the high pressure mode to generate a much higher deposition rate was observed
to have a lot of {111} stacking faults and likely other defects. Reduction of these
crystalline defects likely requires reduction of the chemical contaminants within the
28Si films. Overall, samples deposited at DC–3 enabled improvements and new
understanding regarding the second and third materials goals mentioned at the
beginning of this chapter in support of the broader 28Si effort, of which this work is
a part.
351
Chapter 6
Pressure and Temperature Dependent
Adsorption of 29Si and 30Si During 28Si
Deposition
6.1 Introduction
It was shown in Chapters 4 and 5 that an extremely high level of enrich-
ment was achieved for 28Si films deposited both amorphously at room temperature
and epitaxially at elevated temperatures. However, despite the fact that the 28Si
ion beam is well resolved and separated from the 29Si and 30Si ions, the residual
29Si and 30Si isotope fractions in these samples were not zero. Understanding this
discrepancy and how the concentration of isotopic contaminants are affected by dif-
ferent deposition parameters, such as substrate temperature, is necessary for the
further development of 28Si ion beam deposition. The objectives of the experiments
and analysis discussed in this chapter are as follows:
(1) understand the source of residual 29Si and 30Si in the 28Si films,
(2) determine the dependance of the residual isotope fractions on deposition tem-
352
perature, and
(3) understand the mechanism by which the residual isotope fractions depend on
temperature.
These objectives are part of the larger goals of this work set forth in Chapter 1,
which is to be able to produce 28Si samples with targeted levels of enrichment (29Si
isotope fractions). Targeting specific enrichments would facilitate a study of spin
coherence time as a function of 29Si concentration. Electron and nuclear T2 times of
single implanted 31P measured for a range of 29Si concentrations could be compared
to theoretical predictions (see Fig. 1.9), as mentioned in Chapter 1 [12].
The experiments described in this chapter were designed to test the hypothesis
that the source of 29Si and 30Si, measured in the samples by SIMS, is the natural
abundance SiH4 gas which diffuses from the ion source to the sample location during
deposition. This diffusion results in a partial pressure of SiH4 at the surface of the
28Si sample. The SiH4 molecules (some of which are
29SiH4 and
30SiH4) will stick to
the Si surface where they can be incorporated either through physisorption or in a
chemisorption reaction similar to that of CVD. The sticking and growth behavior of
SiH4 in Si CVD processes, described by the so-called reactive sticking coefficient, has
been studied extensively. The literature on this subject, however, is quite large and
diverse and results are often difficult to compare or reconcile because they depend
heavily on experimental conditions such as pressure, temperature, surface condition,
specific SiH4 species, and various systematic experimental uncertainties. Also, the
reaction describing the conversion of gaseous SiH4 to solid incorporated Si atoms is
more complex than one might naively guess because it can occur through multiple
353
decomposition channels. Several review articles by Comfort and Reif [174], Jasinski
and Gates [137], and Onischuk and Panfilov [175] have given summaries of both
reaction mechanisms of SiH4 CVD processes and experimental work measuring re-
active sticking coefficients and the associated activation energies for these processes.
The dominant reaction expected for SiH4 CVD at pressures ≤ 0.1 Pa (as is the case
in this work) is described by Jasinski and Gates to be
SiH4(g)→ Si(s) + 2H2(g), (6.1)
where (g) represents the gaseous phase and (s) represents the solid phase [137]. This
reaction is exothermic producing about 8.2 kcal/mol (0.36 eV), but energy in the
form of heat is required to overcome the kinetic barriers to the decomposition, which
is the activation energy. Figure 6.1 illustrates the reaction sequence and the role
of dangling bond sites ( ) in CVD reactions on the Si(100) surface. SiH4 initially
adsorbs on the surface at a double dangling bond site. The SiH4 then decomposes
into SiH3 on one dangling bond site and H on the other. By encountering further
dangling bond sites on the surface, the SiH3 decomposes further until a Si atom is
left along with four H atoms, which recombine into two H2 molecules and desorb.
The lone Si atom then becomes incorporated into the film. The desorption of H2
frees three dangling bond sites that are then cycled back into the reaction sequence
for the decomposition of other SiH4 molecules.
There are several differences between SiH4 based CVD and the sticking and/or
reaction of SiH4 being incorporated into the
28Si films discussed here. These dif-
ferences offer several advantages for this work over the typical experiments in the
354
Figure 6.1: SiH4 surface decomposition sequence during CVD growth for low pres-
sures. The SiH4 interacts with dangling bond sites ( ) on the Si(100) surface to
sequentially dissociate H which can then evaporate. (from Ref. [137])
literature, however these differences can make comparisons between the literature
and this work more difficult. Unlike the data presented here for samples deposited
with substrate temperatures ranging from room temperature up to 800 ◦C, CVD is
not typically studied with a growth temperature below 500 ◦C to 600 ◦C because the
growth rate drops dramatically below this range, depending on specific experimen-
tal conditions. This is partially due to increased H coverage at lower temperatures
which inhibits the CVD reaction. CVD typically uses high pressures of H2 with
SiH4 partial pressures as high as 130 Pa (975 mTorr) for UHV CVD, although some
experiments have used pressures as low as 1.3× 10−5 Pa (9.8× 10−8 Torr) [176].
These pressures are orders of magnitude higher than what is used for the samples
described in this chapter, and so H coverage is not believed to significantly influ-
ence the results reported here. The other major difference between CVD and the
355
ion beam deposited samples, is the 28Si ion beam itself, which could interact with
adsorbed SiH4. The deposition rate due to the ion beam is typically much larger,
i.e. > 10 times larger, than the growth rate due to the adsorbed species, therefore a
SiH4 molecule will (almost) always encounter a bare Si surface. Consequently, rate
limiting effects seen in CVD such as surface diffusion, SiH4 interaction with other
adsorbates like H, and surface reaction rates would not occur [177, 178]. Energetic
ions are also known to desorb H from Si(100) surfaces [121]. Finally, a unique fea-
ture of monoisotopic ion beam deposition in this context is that very small numbers
of adsorbed species can be measured. As seen in the analysis of the enrichment of
28Si samples in Chapters 4 and 5, SIMS is extremely sensitive to isotope ratios and
so trace amounts of 29Si and 30Si adsorbed in the 28Si films can be easily detected.
Section 6.2 of this chapter describes the experimental methods used to collect
and analyze the relevant sample parameters discussed in the remainder of the chap-
ter. In section 6.3, a model that describes the proposed SiH4 adsorption process is
introduced, and in section 6.4, it is used to analyze the data. Sections 6.5 and 6.6
explore the role of substrate temperature on the enrichment and calculated SiH4
incorporation fractions, respectively. Section 6.7 uses the incorporation fractions
to determine an activation energy for SiH4 adsorption. Section 6.8 provides a brief
summary. Analysis of some of the data discussed in this chapter was previously
published in Ref. [84].
356
6.2 Experimental Methods
6.2.1 28Si Samples and Enrichment Values
The general deposition parameters for 28Si samples similar to and including the
samples used in the analysis of this chapter were described in detail Chapters 4 and 5.
These samples were deposited on a variety of natural abundance Si(100) substrates
including p-type, n-type, and undoped (intrinsic) wafers, discussed in Chapters 4 and
5. For most samples discussed here and deposited at room temperature, substrates
were prepared ex situ by an HF etch to remove the native oxide and were not
prepared further in vacuum. These samples were deposited at sample location LC–
2 in the lens chamber. One room temperature sample was deposited at DC–3
in the deposition chamber, and was not prepared ex situ and were loaded with a
native oxide. For most of the samples deposited at elevated temperatures, substrates
were cleaned ex situ using the standard CMOS cleaning procedure described in
Chapter 5. Three samples deposited above 600 ◦C were not cleaned ex situ and
were loaded in the vacuum chamber with a native oxide. Substrates were then
prepared for deposition in situ by flash annealing them to 1200 ◦C for ≈ 10 s several
times to produce a clean (2×1) reconstructed Si(100) surface on which to deposit
28Si epitaxially. These samples were all deposited at sample location DC–3 in the
deposition chamber. As mentioned previously, the gas used in these experiments
to generate a 28Si ion beam was natural abundance SiH4 with a purity of 99.999 %
according to the gas vendor (Matheson Tri-Gas). To map out the temperature
357
dependance of the enrichment, samples were deposited with substrate temperatures
including room temperature (≈ 21 ◦C), 249 ◦C, 349 ◦C, 357 ◦C, 421 ◦C, 502 ◦C,
610 ◦C, 705 ◦C, 708 ◦C, and 812 ◦C.
28Si ions were deposited onto the substrates with an average ion energy, Ei,
at the sample of typically about 100 eV for the room temperature samples and ap-
proximately 35 eV for the samples deposited at elevated temperatures. Typical 28Si
ion beam currents, Ii, of around 500 nA were achieved over an area on the sub-
strate between about 3 mm2 and 16 mm2. For one sample deposited at 421 ◦C, a
higher ion beam current of approximately 3 µA was achieved. The resulting thick-
nesses, d, of the deposited films were inferred from the calibration of the SIMS depth
profiles and ranged from ≈ 50 nm to 350 nm. Dividing the thicknesses by the depo-
sition time for each sample gives an estimate for the deposition rates, R, of around
0.32 nm/min to 3.94 nm/min. Based on these rates, the corresponding average ion
flux,Fi, was then calculated for each sample. Fi varied from 2.70× 1013 cm−2 · s−1
to 3.4× 1014 cm−2 · s−1.
For the samples discussed in the analysis of this chapter, the total pressure
rise during deposition after subtracting the chamber base pressure ranged from
approximately 9.9× 10−7 Pa to 4.9× 10−6 Pa (7.5× 10−9 Torr to 3.7× 10−8 Torr)
for the samples deposited at room temperature, and it ranged from approximately
4.9× 10−7 Pa to 3.4× 10−6 Pa (3.7× 10−9 Torr to 2.5× 10−8 Torr) for the samples
deposited with elevated substrate temperatures. These pressures equate to a total
gas flux, F tot.g , on the surface of the sample during deposition. A more relevant
flux in this analysis is the flux due to the SiH4 partial pressure, Fg, which will be
358
discussed in a later section.
The 29Si and 30Si isotope fractions for these samples were measured by SIMS
in collaboration with Dr. David Simons (NIST) as described in Chapter 4. Iso-
tope fractions of a particular isotope of Si are defined in a SIMS measurement as
the detected average counts of that isotope divided by the total average counts of
the measurement and are written as zSi/Sitot. for an isotope with mass number z,
as previously discussed in Chapter 4. Measurements were performed by Dr. Si-
mons and the analysis presented here was done by myself. The raw measurements
show that at the low end of the deposition temperature range, the 249 ◦C sample
had a residual 29Si isotope fraction of 0.79(12)× 10−6 or 0.79(12) ppm. For the
sample deposited at the highest temperature, 812 ◦C, the 29Si isotope fraction was
4.32(46) ppm. This increase in isotope fraction with increasing substrate tempera-
ture is the focus of this discussion. The sample with the best enrichment and lowest
29Si isotope fractions in this study was deposited at 502 ◦C, as previously reported
in Chapter 5 (see Fig. 5.11). The measured 28Si isotope fraction of the most highly
enriched portion of this sample is 99.9999819(35) %, the average residual 29Si isotope
fraction is 127(29) ppb, and the average residual 30Si isotope fraction is 55(19) ppb.
The uncertainty of the isotope ratios was determined from the standard deviation
of the mean of the measurements. A list of the samples discussed in this chapter,
their deposition parameters, and measurement and analysis results can be found in
Tables D.10 to D.13 in Appendix D.
For the samples deposited at 705 ◦C, 708 ◦C, and 812 ◦C, the measured isotope
fractions have to be taken as an upper bound. This is because during deposition,
359
the samples deposited above 600 ◦C developed a large amount of surface roughness
on the order of the film thickness itself. As discussed in Chapter 5, this roughness
had the effect of artificially inflating the isotope fractions measured by SIMS (see
Fig. 5.12 and Fig. 5.14). The 610 ◦C sample is excluded from this caveat because the
SIMS measurement of it was found to be more trustworthy. Nominally the SIMS
depth profiles show a clear extended minimum in 29Si and 30Si isotope fractions
through the thickness of the film, but instead, in these higher deposition temperature
samples there is a gradual increase in isotope fractions up to the natural abundance
values in the substrate. This effect is a measurement artifact caused by the SIMS
sputter beam sampling the 28Si film and substrate at the sample time. The stated
enrichment values for the effected samples are considered upper bounds because the
measurement artifact would only increase the apparent isotope fractions but never
decrease them. Care was taken to exclude data that was clearly influenced by this
effect, however, it is possible that this artifact still played a small role in determining
the isotope ratios of the highest temperature samples.
6.2.2 SiH4 Mass Spectrum and Mass Selectivity
A key component of the analysis of SiH4 adsorption discussed in this chapter
is the assumption that the ion beam is 100 % pure 28Si. This assumption can be
justified in part by analysing the isolation of the 28Si ion beam as measured in the
SiH4 mass spectrum that is collected during operation of the ion beam with SiH4
gas. A portion of a SiH4 mass spectrum, representative of the ion beam conditions
used in Chapter 4 to deposit room temperature samples at sample location LC–2,
360
is presented in Fig. 6.2. Qualitatively similar mass spectrums were obtained for
samples deposited at elevated temperatures, as can be seen in Fig. 5.4 in Chapter 5.
In Fig. 6.2, the ion current peaks corresponding approximately to 28 u (28Si), 29 u
(29Si and 28SiH), 30 u (predominately 28SiH2), and 31 u (predominately
28SiH3) are
observed. Gaussian fits (Eq. (2.14)) to the 28 u (dashed line) and 29 u (solid line)
peaks are shown superimposed on the data. The 95 % confidence bands of the two
fits are also shown (dash-dotted lines). The applied current corresponding to the
sweep of the magnetic field of the mass analyzer is shown on the top axis. This
mass spectrum indicates a mass resolving power m
∆m
≈ 80 (measured at 10 % of the
peak height). The Gaussian fits give a separation of the 28 u peak from the 29 u
peak of about 11 σ (standard deviation). The Gaussian fits are used to determine
the approximate geometric mass selectivity (i.e. the amount of mass separation) of
the ion beam system to estimate the amount of 29Si potentially contaminating the
28Si beam. This is done by calculating the overlap of the 29 u and 28 u peaks using
the parameters G28(m) and G29(m), which are the values (calculated at mass m in
units of u) of the Gaussian fits to the current peaks at 28 u and 29 u respectively.
The overlap of the 29 u peak on the 28 u peak is then determined from
G29(28)
G28(28) +G29(28)
, (6.2)
where m = 28 for the above parameters signifying that the values of the Gaussian
fits are calculated at a mass of 28 u. The 95 % confidence band of the fit to the
28 u peak is used to calculate G28(m), and the 95 % upper confidence band of the
fit to the 29 u peak is used to calculate G29(m). When also taking into account
361
Figure 6.2: SiH4 ion beam mass spectrum (circles) showing the
28Si ion current peak
at 28 u, 29Si ion current at 29 u (≈ 5 % of the total current at 29 u), and two higher
mass Si hydride peaks. The top axis shows the corresponding current applied to
the magnetic sector mass analyzer to sweep the field. Gaussian fits (Eq. (2.14)) to
the 28 u (dashed line) and 29 u (solid line) peaks along with 95 % confidence bands
(dash-dotted lines) are plotted to calculate the overlap of the 29 u peak onto the
28 u peak. The centers of the 28 u and 29 u fits are separated by ≈ 11 σ.
that the 29 u peak consists of approximately 5 % 29Si (95 % 28SiH), as discussed
in Chapter 4, the 29Si contamination fraction of the 28 u ion current is calculated
to be ≈ 5× 10−26. This extreme estimate of the contamination fraction of the 28Si
peak from the 29Si peak is unphysical and is purely a measure of the geometric mass
selectivity of the ion beam system.
To get an estimate of the potential realized mass selectivity due to the 29Si peak
overlap, a gas scattering mechanism is considered that would likely be a dominant
contributing factor to the 29Si beam contamination. Inelastic scattering between ions
362
and gas molecules occurs along the flight path of the ions as they travel down the
beamline. This causes an ion at mass 29 u to lose sufficient energy to be incorporated
into the 28 u trajectory and pass through the mass-selecting aperture. This so-called
scattering tail effect, referred to as the abundance selectivity when considering the
tail contribution to an adjacent mass current peak, IS/I0, can be estimated from
IS
I0
∝ P∆x
( m
∆m
)n
, (6.3)
which was adapted from Ref. [179]. IS is the scattered ion current from a peak at
mass m to a peak at mass m + ∆m, where ∆m is an integer. For the scattering
contribution of an ion current peak to a current peak at an adjacent mass, ∆m = ±1.
I0 is the total ion current at mass m, P is the background gas pressure in the
beamline, ∆x is the width of the mass-selecting aperture, and n is a parameter
corresponding to the scattering cross sections and is ≈ 1.7 for similar ion beam
systems [179]. Figure 6.3 shows an experimental example of the scattering tail and
abundance selectivity for a mass spectrum of ThO+ from Ref. [180]. The current
of the ThO peak at mass 248 u drops quickly on either side of the peak, but then
levels off, illustrating the scattering tail effect, which contributes roughly 10−6 of
the 248 u peak at 247 u and 249 u.
From Eq. (6.3) and literature values of the abundance selectivity for a single
magnet system with an operating pressure within the beamline of approximately
1.3× 10−4 Pa (1.0× 10−6 Torr), a contribution of roughly 4× 10−6 of the higher
mass peak to the lower mass peak is expected at a mass of 28 u [180, 181]. This
peak tail current is not measurable in this ion beamline because the mass resolu-
363
Figure 6.3: ThO+ ion beam mass spectrum illustrating the scattering tail and abun-
dance selectivity at mass 248 u. The main peak initially drops quickly on either side
of the peak before leveling off as the ion current signal begins to be dominated by
the scatter tail. This tail decreases more gradually and contributes approximately
10−6 of the 248 u peak at ∆m = 1 u. (from Ref. [180])
tion and current sensitivity are too low. Combining the scattering fraction with the
29Si natural abundance and the fact that the 29 u peak is typically about the same
magnitude as the 28 u peak gives an estimate for an upper bound on the 29Si con-
centration in the 28Si beam of roughly 2×10−7, or 200 ppb. This concentration may
be significant for a few samples discussed here with the lowest 29Si isotope fractions
around 200 ppb, although the true value may be lower making it less significant.
Further, there is no evidence, e.g. significant and consistent attenuation of the 30Si
isotope fractions compared to 29Si, that this scattering limit is having a significant
effect on the measured enrichment. For example, the 502 ◦C sample with a mea-
sured 29Si isotope fraction of 127(29) ppb has a 29Si/30Si ratio of 2.31 ± 0.96. This
value agrees within the uncertainty with the natural abundance value of approxi-
mately 1.52. Further, the expected 29Si/30Si ratio of the scatter tails from Eq. (6.3)
364
is larger than the measured ratio at ≈ 4.6. Based on this analysis, the scattering tail
contribution is considered to be negligible of the purposes of the analysis discussed
in this chapter, and the ion beam is assumed to be pure 28Si. For the discussion
of the following sections, the difference between the expected (100 % enriched) and
measured enrichment is considered by identifying only the natural abundance SiH4
gas diffusing from the ion beam into the deposition chamber as the source of 29Si
and 30Si.
6.2.3 Determination of SiH4 Partial Pressures
An accurate estimate of the partial pressure of SiH4, present at the sample lo-
cation during 28Si deposition, is required to determine any correlations between this
partial pressure and enrichment levels (i.e. 29Si and 30Si isotope fractions). Total
pressure measurements were made using several different ion gauges in the system
located in two of the three sample deposition locations (the lens chamber, LC–2,
and the deposition chamber, DC–3) as described in Chapters 2 and 4. Comparing
samples with pressure readings taken from different gauges introduces some error
in the analysis because gauges may be calibrated differently. For the samples de-
posited at room temperature, and some samples deposited with elevated substrate
temperatures, the raw ion gauge readings from different gauges are assumed to be
comparable in this analysis, although with different known uncertainties in the read-
ings of different gauges, because a direct conversion was not performed at the time.
Another gauge used for some of the samples deposited at elevated temperatures had
a known offset which could be compared to a more accurate gauge at the same loca-
365
tion. A conversion was used to translate the pressure reading from this gauge to the
equivalent readings for the more accurate gauge. Typically the readings from these
types of ion gauges have a relative uncertainty of ≈ 20 %, while the more accurate
gauge has a relative uncertainty of ≈ 5 %. After determining the appropriate total
pressure during deposition for each sample, the base pressure at the relevant sample
location immediately prior to deposition was subtracted out to get the total pressure
increase due to gas diffusion from the ion source, which was typically a factor of
50 to 100 times higher than that base pressure. Pressure increases due to sample
heating were also taken into account.
To determine the SiH4 partial pressure component of the total pressure read-
ings, the RGA in the deposition chamber was used to take partial pressure mea-
surements from the residual gas mass spectrum while flowing SiH4 gas from the
ion source. Because the RGA is not necessarily calibrated the same way as the ion
gauges in the chamber, and the readings are influenced by the electron multiplier
settings in the detector of the RGA, the absolute partial pressure readings are not
reliable. Instead, the partial pressure readings of SiH4 were used to calculate the
approximate fraction of the total sum of partial pressure peaks in the residual gas
spectrum. Determining this SiH4 fraction is complicated by the fact that when
the ion source is in operation, it cracks SiH4 into lower order Si hydrides, SiHx
(1 < x < 4), and H2. The RGA filament itself further cracks SiHx. Figure 6.4 shows
RGA residual gas mass spectra for the base pressure of the deposition chamber,
which was about 8.3× 10−9 Pa (6.2× 10−11 Torr) (diagonal line fill), during SiH4
gas flow (horizontal line fill), and when the ion beam is on (solid line). When SiH4
366
Figure 6.4: Residual gas mass spectra collected from the RGA in the deposition
chamber for the chamber base pressure (diagonal line fill), while flowing SiH4 (hor-
izontal line fill), and with the ion beam in operation (solid line). The complex of
SiHx peaks are observed from 28 u to 33 u during SiH4 gas flow. H2 is also observed
to increase.
is flowed into the chamber with the ion source off, the SiHx peaks are observed
between 28 u and 33 u due to gas cracking from the RGA itself. The increased H2
signal is partially due to this process. When the ion source is operating, the SiH4
is initially cracked before diffusing to the RGA where the resulting SiHx is cracked
further so that the residual gas mass spectrum does not show any significant Si
hydride peaks. This is accompanied by an additional increase in H2 signal. This
reduction in SiHx signal with the ion beam on precludes a direct measurement of
the SiH4 partial pressure.
To get an estimate of the SiH4 partial pressure during ion beam operation,
the cracking efficiency of the ion source and RGA are assumed to be similar and
367
then measure the total SiH4 partial pressure fraction with the ion beam off. This
assumption is supported by the fact that the ratios of Si hydride peaks seen in the
ion beam mass spectrum above (Fig. 6.2) and the RGA residual gas mass spec-
trum (Fig. 6.4) are similar. From these measurements and estimates for the H2
and SiH4 gas sensitivity factors [182, 183], the SiHx partial pressures is estimated
to be roughly 28 % ± 5 % of the total pressure increase. This gives SiHx partial
pressures, PSiHx , that range from 2.8× 10−7 Pa to 1.4× 10−6 Pa (2.1× 10−9 Torr
to 1.1× 10−8 Torr) for the room temperature samples, and from 1.4× 10−7 Pa to
9.6× 10−7 Pa (1.1× 10−9 Torr to 7.2× 10−9 Torr) for the samples deposited at el-
evated temperatures.
The SiHx gas flux, Fg, which impinges on the sample during deposition is a
critical parameter to the analysis of this chapter. Fg is derived from the SiHx partial
pressures using the Hertz-Knudsen equation for gas flux [184,185],
Fg =
PSiHx√
2pimkBTg
, (6.4)
where PSiHx is the SiHx partial pressure, m is the molecular mass of SiH4 (≈
32 u), kB is the Boltzmann constant, and Tg is the temperature of the gas (≈
21 ◦C). This calculation gives SiH4 gas fluxes between 7.6× 1011 cm−2 · s−1 and
3.7× 1012 cm−2 · s−1 for the room temperature samples. For the higher tempera-
ture samples, the SiH4 gas fluxes are calculated to be between 3.7× 1011 cm−2 · s−1
and 2.6× 1012 cm−2 · s−1. The relative uncertainty of these estimates is ≈ 15 % to
20 %.
368
6.2.4 Substrate Temperature Calibration
The temperature of the samples were carefully measured during deposition to
ensure an accurate mapping of enrichment vs. temperature and determination of
the temperature dependance of the incorporation fraction, s, discussed in the fol-
lowing sections. As explained in Chapter 2, the samples were heated using direct
current heating (DH) through the substrate for all samples except the one deposited
at 249 ◦C, which was heated using the tungsten radiative back heater (RH). The
substrate temperatures were measured by the two previously discussed infrared py-
rometers. The Process Sensors pyrometer, which was calibrated for this system as
discussed in Chapter 2, was used for most samples discussed in this chapter, but
the un-calibrated Omega pyrometer was used for four of the samples. A correction
was applied to the Omega pyrometer readings, which can differ from the Process
Sensors readings by ≈ 25 ◦C. Including the uncertainties from the pyrometer cal-
ibration, the temperature readings of the substrate are estimated to have a 5 %
relative uncertainty due to fluctuations in the current used for sample heating as
well as temperature gradients across the sample. The exact temperatures of the
samples deposited at room temperature were not measured but instead assumed
to be similar to the typical measured ambient temperature outside of the vacuum
chamber, which was 21 ◦C ± 2 ◦C. For the analysis of these room temperature
samples, this value was used.
369
6.3 Temperature Dependent Gas Incorporation
Model
To correlate the effect of SiHx partial pressure on the incorporation or ad-
sorption of 29Si and 30Si during deposition of 28Si films at different temperatures, a
gas sticking deposition model is formulated that describes the different contributing
sources to the film deposition. The sticking of gaseous species is based simply on the
idea of idealized Langmuir adsorption for a flat surface with equivalent adsorption
sites [186]. This model is later compared to the sample enrichments measured by
SIMS.
The model describes two sources of Si atoms that contribute to deposition at
the substrate:
(1) the ion beam, which is presumed to be pure 28Si, deposits ions onto the sub-
strate, and
(2) the partial pressure of SiHx, which contains all three Si isotopes in their natural
abundance, can stick and become incorporated into the film.
Figure 6.5 is a cartoon representation of this two source deposition model. In this
model, the isotopic concentrations measured by SIMS is the fraction of the total
Si deposited that is due to 29Si or 30Si sticking from the SiHx background partial
pressure. The isotope fraction of 29Si and 30Si in a sample is described by the gas
sticking deposition model, cz (with z denoted as 29 for
29Si and 30 for 30Si), given
by
cz =
Fgazs
Fgs+ Fi
, (6.5)
370
Figure 6.5: Cartoon illustrating the 28Si thin film deposition process described by
the gas sticking deposition model. Two sources contributing Si to the deposition are
the 28Si ion beam and the SiH4 gas which diffuses from the ion source and contains
a natural abundance of isotopes. SiH4 can adsorb and react with the Si surface with
different probabilities.
where Fg is the SiHx gas flux, Fi is the
28Si ion flux, az is the isotopic abundance of
29Si or 30Si in the SiH4, which is assumed to have a natural abundance, and s is an
effective incorporation fraction (or sticking coefficient) for an average SiHx species
to be adsorbed into the surface. cz gives the calculated isotope fraction for a given
set of deposition conditions. cz can be re-written as a function that correlates the
isotope concentrations to the deposition conditions using a convenient deposition
parameter; the SiH4 flux ratio, k = Fg/Fi. cz written as a function of k is then
cz =
azsk
1 + sk
. (6.6)
371
Notice that cz increases with increasing SiH4 gas flux, and it decreases with increas-
ing ion beam flux. Additionally, when Fg  Fi, then cz ∝ k, and when Fg  Fi,
then cz ≈ az. Another convenient transformation of cz is to convert the isotope spe-
cific model of Eq. (6.6) into a general model for SiH4 sticking by dividing by each
isotope’s natural abundance so that 29Si and 30Si data can be fit together within
the same model. Dividing the measured isotope fraction of a sample by its natural
isotopic abundance gives an expected total adsorbed SiH4 fraction, which is then
described by the gas sticking deposition model giving the calculated SiH4 fraction,
ctot., for the total adsorbed SiH4 where
ctot. =
cz
az
=
sk
1 + sk
. (6.7)
Equation (6.7) allows the full statistical weight of all the data to be used for deter-
mining the incorporation fraction, s, for each sample deposited at different temper-
atures and get the trend of s vs. temperature, T .
Next, to describe the behavior of s vs. T , a temperature dependent incorpo-
ration model, s(T ), is defined that is described by two gas sticking terms; a sticking
probability resulting from physisorption, sc, and a higher temperature reactive stick-
ing coefficient, sr, resulting from chemisorption. sc and sr are both expected to be
activated by temperature, but sc decreases with increasing temperature, and sr
increases with increasing temperature. These components are defined to be
sc = 1− exp
(−Ep
kBT
)
(6.8)
and
sr = Ar exp
(−Ec
kBT
)
, (6.9)
372
where Ep is the activation energy for physisorption, Ec is the activation energy for
chemisorption, kB is the Boltzmann constant, and T is the substrate temperature
during deposition. The exponential prefactor Ar is left as a free parameter to account
for experimental uncertainties which may introduce a constant shift in the data, and
it will be discussed further in the following sections. The prefactor for sc is set to
1 because the sticking probability is expected to be close to unity as T approaches
zero. The total incorporation fraction at a given temperature is then described
by the temperature dependent incorporation model, which is the sum of the two
sticking components,
s(T ) = sc + sr = 1− exp
(−Ep
kBT
)
+ Ar exp
(−Ec
kBT
)
. (6.10)
6.4 Correlating Enrichment to SiH4 Partial
Pressure
To demonstrate the correlation between the SiHx partial pressure and the
28Si
sample enrichment, the raw SIMS isotope fraction data (zSi/Sitot.) for the room
temperature samples are plotted as a function of the SiH4 flux ratio, k = Fg/Fi in
Fig. 6.6, with 29Si (squares) and 30Si (triangles) isotope fractions plotted together.
The top axis in panel (a) shows the total gas flux ratio, F tot.g /Fi, using the gas flux
corresponding to the total measured pressure increase during deposition without
subtracting out the estimated H2 fraction in the gas. This difference only shifts
the axis laterally, and both the 29Si and 30Si isotope fractions have a strong linear
correlation with k, representing the deposition conditions. The resulting Pearson
373
Figure 6.6: Correlation plot of isotope fraction vs. k for 28Si samples deposited at
room temperature (≈ 21 ◦C). (a) SIMS measurements of 29Si (squares) and 30Si
(triangles) are shown. The top axis shows the total gas flux during deposition after
subtracting the background pressure flux. cz is fit (Eq. (6.6)) to the
29Si (solid line)
and 30Si (dashed line) data giving s = 6.8(3)× 10−4. c30 is also shown calculated
for two other values of s (dotted lines) to show the sensitivity of the model to s.
(b) cz fits from (a) asymptote to the natural abundance values (dash-dotted lines)
at large k (Fg  Fi).
374
correlation coefficient for the 29Si data is r = 0.95, and for the 30Si data r = 0.92.
This shows that an increase in SiH4 flux corresponds to an increased isotope fraction
in the sample.
The correlation between enrichment and k is modeled using Eq. (6.6) to get
c29 and c30, which are fit to the data with s as the only free parameter. These fits are
shown in Fig. 6.6 as solid and dashed lines respectively, and they are approximately
linear over the range of the data with a slope proportional to s. In panel (b), the
fits are plotted in an extended range to show the crossover to az at high values of
k. Above a k value of about 104, cz starts to asymptote to the natural abundance
values of 4.7 % for 29Si and 3.1 % for 30Si (dash-dotted lines). In other words, when
Fg/Fi  1, the contribution from the 28Si ion beam becomes negligible compared
to the gas flux, and so the film composition approaches the composition of the gas,
which is assumed in the model and expected in reality to have a natural abundance
of Si isotopes. The fits to the data give a room temperature incorporation fraction
of s = 6.8(3)×10−4. The uncertainty in this value is the standard error from the fit.
Also plotted for reference is c30 calculated for two other values of s (dotted lines),
2× 10−4 and 2× 10−3, which span an order of magnitude around the data. This is
to illustrate the sensitivity of the fit to s. Note that when viewed as a log-log plot as
in Fig. 6.6, cz does not change apparent slope as s is varied, it only changes vertical
offset.
To investigate this correlation over a larger range of k values and specifically
a larger range of SiH4 fluxes, i.e. pressures, samples deposited in the ion beam
chamber at sample location IC–1 were also analyzed. These samples, which were
375
described in Chapter 4, were also deposited at room temperature and with much
higher background pressures during deposition. These additional data are shown in
Fig. 6.7, which is an expanded version of Fig. 6.6. The 29Si and 30Si data for the
samples deposited at IC–1 (open squares and open triangles, respectively) exhibit
similar qualitative behavior to the data for the samples mostly deposited at LC–2
(closed squares and closed triangles, respectively) at lower k values from Fig. 6.6.
The fits of cz (Eq. (6.6)) from Fig. 6.6 are again shown, which are only fit to the
LC–2 sample 29Si and 30Si data (solid and dashed lines, respectively). The IC–1
samples generally increase in isotope fraction with increasing k, however, they tend
to deviate from the fit of cz exhibiting an apparently lower incorporation fraction.
This deviation can possibly be explained as due to several factors. First, because
there was no gas outlet via a vacuum pump at the sample location for these samples,
there is an unknown and likely large uncertainty on the total pressure at the sample
and its composition in terms of H2 and SiHx. The ion beam itself delivers gaseous
species to the location of the mass-selecting aperture, which is relatively close to the
sample, in the form of ionized SiHx that may locally increase the pressure around
the sample. This pressure increase can be roughly estimated to increase the H2
partial pressure to ≈ 85 % of the total pressure, which is an increase from 70 % in
the other experimental configurations. This correction cannot totally account for
the discrepancy between the data and the model fit, however.
Additionally, these samples were deposited with a total background pressure
increase during deposition between approximately 1.5 ×10−4 Pa and 2.1 ×10−4 Pa
(1.2 ×10−6 Torr to 1.6 ×10−6 Torr), which is at least a factor of 30 times higher
376
Figure 6.7: Correlation plot of isotope fraction vs. k for 28Si samples deposited at
room temperature (≈ 21 ◦C). The data from Fig. 6.6 of samples deposited mostly at
LC–2 (closed squares and closed triangles) are shown as well as SIMS measurements
of 29Si (open squares) and 30Si (open triangles) from samples deposited at IC–1 with
higher background pressures and thus larger Fg and k values. cz is fit (Eq. (6.6)) to
only the LC–2 29Si and 30Si data (solid and dashed lines, respectively). The vertical
dashed line indicates the boundary where k = 1. The LC–1 data lie to the right of
this where Fg > Fi in a different deposition regime than the LC–2 data.
than the other room temperature samples deposited at LC–2 and at most a factor
of 220 times higher. These pressures and corresponding higher gas flux results in
these samples being deposited in a regime where Fg > Fi. This is represented by
the vertical dashed line in Fig. 6.7. When Fg > Fi, an incident SiHx molecule
may not find an area of the surface with bare Si, and instead it may encounter
another adsorbed SiHx or H2 molecule, blocking further adsorption. H2 coverage is
known to decrease the effective reactive sticking coefficient of SiH4 [178], although
with a small sticking coefficient of ≈ 10−4 on Si, H2 coverage may not be significant
377
Figure 6.8: 29Si/30Si isotope ratios for all measured samples deposited at room
temperature (circles) and elevated temperatures (triangles). The ratios of these
samples agree with the natural abundance ratio of 1.52 (line) indicating that the
source of 29Si and 30Si is naturally abundant, probably the SiH4 gas.
here [169]. Energetic ions can also desorb H from Si(100) surfaces [121]. It is difficult
to accurately account for these high pressure effects and so the samples deposited
at IC–1 are excluded from the analysis of this chapter.
Further strong evidence that the SiH4 partial pressure is the source of the
measured residual isotope fraction of 29Si and 30Si can be derived from the measured
isotope ratios, 29Si/30Si, for each sample. If these isotopes were originating from
the ion beam, one would expect an attenuation of 30Si compared to 29Si which
would increase the 29Si/30Si ratio above the natural value. Instead, the measured
isotope ratios are found to be very close to the natural value of approximately
1.52. Figure 6.8 shows the 29Si/30Si isotope ratios for a large number of room
378
temperature (circles) and elevated temperature (triangles) samples. Also shown is a
line representing the natural abundance ratio. All of the data lie close to the natural
ratio within their error bars. This indicates that the source of 29Si and 30Si has a
natural abundance of Si isotopes, e.g., the SiH4 source gas. Measurement number
28 and 31, which lie above a 29Si/30Si ratio of four, suffer from discrete counting
noise in the SIMS measurements due to a total 30Si count < 10 through the entire
enriched film, which makes the ratio highly sensitive to single count fluctuations.
6.5 Temperature Dependence of 29Si and 30Si
Adsorption
The previous section showed that a strong correlation exists between the rela-
tive SiH4 flux at the sample and the measured enrichment. This pressure dependance
indicates that the natural abundance SiH4 gas diffusing from the ion beam is the
source of residual 29Si and 30Si in the samples. Next, because substrate deposition
temperature is a key parameter for facilitating and controlling the quality of epi-
taxial deposition as this work progresses further, it is important to understand the
effect of different substrate temperatures on the correlation between enrichment and
gas flux.
As reported previously in this chapter, the raw SIMS data show that the 29Si
isotope fractions increase rapidly in the deposition temperature range from 502 ◦C
(127 ppb) to 812 ◦C (4.32 ppm). However, the room temperature correlation plot
(Fig. 6.6) showed that the isotope fractions also depend on k, the SiH4 flux ratio
379
(i.e. the deposition conditions). In order to extract an accurate temperature depen-
dence of the enrichment, the raw SIMS measurements are adjusted to a common
set of deposition conditions (Fg and Fi). To perform this adjustment, the k value
matching the sample deposited at 502 ◦C is chosen, i.e. the sample with the lowest
measured isotope fractions of 29Si and 30Si. This adjustment suppresses the effect
on enrichment of varying deposition conditions across samples. By using the 502 ◦C
sample as a benchmark against which to compare the other samples, the change
in isotope fraction is mapped against temperature for the conditions that produced
the best measured enrichment.
To find the adjusted isotope fractions from the raw SIMS values, Eq. (6.6) is
first solved for s, using subscript T to denote the resulting value, sT , for a specific
data point with deposition temperature T (in ◦C). Equation (6.6) is then used
along with each calculated sT value to generate the gas sticking deposition model
curve, cz(sT ), for each data point. The k value of the 502
◦C sample, which is
the reference to which the raw SIMS data will be adjusted, is denoted as k502.
To get these adjusted values, cz(sT ) at k = k502 is then evaluated for each data
point. This gives cz(sT , k502), which is thus the isotope fraction of a sample with
deposition temperature T adjusted to the deposition conditions (SiH4 flux ratio)
of the 502 ◦C sample. For example, the adjusted isotope fraction of the 812 ◦C
sample is calculated as cz(s812, k502). In Fig. 6.9, the adjusted isotope fractions given
by cz(sT , k502) are plotted as a function of temperature showing the temperature
dependence of the enrichment, independent of variations in deposition conditions.
More specifically, the values in Fig. 6.9 are the expected isotope fractions of 29Si
380
Figure 6.9: Adjusted isotope fraction, cz(sT , k502), vs. temperature for
29Si (squares)
and 30Si (triangles). The raw isotope fractions are adjusted to the deposition condi-
tions (Fg and Fi) of the 502
◦C sample. The inset shows the same data on a semi-log
scale to highlight the behavior of the data below 500 ◦C.
(squares) and 30Si (triangles) for all samples had they been deposited with the same
SiH4 partial pressure and ion beam flux as the 502
◦C sample. The adjusted isotope
fractions are shown to trend downwards slightly from the 249 ◦C sample average of
about 0.76 ppm 29Si to a minimum at the 502 ◦C average of about 0.13 ppm 29Si.
This implies that if the substrate temperature of a sample similar to the 502 ◦C
sample is lowered to 249 ◦C during deposition, the 29Si isotope fraction is expected
to increase to approximately 0.76(17) ppm.
The room temperature samples do not seem to follow this trend and instead
have adjusted isotope fractions that are lower than expected. The deposition pres-
sure for these samples was measured using a different ion gauge configuration than
381
the later high temperature samples, which may affect the adjustment of these data.
Another difference is that the room temperature samples were all grown as amor-
phous films while the samples deposited at elevated temperatures were crystalline.
Surface orientation and crystallinity affect the adsorption of SiH4 on Si surfaces as
described by Comfort and Reif who note that for Si CVD, the growth rate on a
Si(100) surface is generally higher than on a Si(111) surface, which is itself higher
than the growth rate on a polycrystalline surface [174]. These effects may lead to
a lower effective sticking coefficient on the amorphous samples compared to that of
the crystalline samples.
Above 502 ◦C, the adjusted 29Si isotope fraction sharply increases up to a value
of 5.9 ppm 29Si at 812 ◦C. Again, this shows that if the substrate temperature of a
sample similar to the 502 ◦C sample was increased to 812 ◦C during deposition, the
29Si isotope fraction is expected to increase to 5.9(16) ppm. This increase is posited
to be due to an increase in s as a process similar to a CVD reaction becomes more
active, and the reactive sticking coefficient begins to dominate the total incorpora-
tion fraction. To examine this hypothesis, the values of s at each temperature need
to be determined and compared to the temperature dependent incorporation model
(Eq. (6.10)).
382
6.6 Temperature Dependence of the
Incorporation Fraction, s
The gas incorporation fraction, s, is determined at each sample temperature
using Eq. (6.6) shown in Fig. 6.6. It should be noted that the incorporation frac-
tions determined here are a total net sticking probability; i.e. a molecule was in-
corporated into the growing film, and remained there until detected by SIMS. A
convenient transformation in the analysis of this gas sticking deposition model is
to convert each isotope fraction for a given sample by dividing by their respective
natural abundance values, az, so both
29Si and 30Si can contribute statistical weight
together when fit with the model. This conversion follows the model of Eq. (6.7) for
calculating ctot. and gives the converted isotope fraction, i.e. the expected total SiH4
fraction adsorbed in the sample, (zSi/Sitot.)/az. Figure 6.10 (a) is a correlation plot
of the converted isotope fractions for several deposition temperatures: 812 ◦C (dia-
monds), 705 ◦C (left-pointing triangles), 249 ◦C (hexagons), 421 ◦C (right-pointing
triangles), and 502 ◦C (down-pointing triangles). These data are plotted vs. the
SiH4 flux ratio, k. Also plotted is ctot. (Eq. (6.7)), which is fit to each temperature
set with s as the only free parameter. The fits reported in this chapter are achieved
using an orthogonal distance regression method which accounts for uncertainty in
both coordinate values of each datum. Uncertainties reported with the fit values
are the standard error from the fits. Within the range of Fig. 6.10, ctot. is approx-
imately linear with a slope equal to s. Note that when ctot. is plotted on a linear
scale, it runs through the origin because zero SiH4 flux results in a calculated SiH4
383
Figure 6.10: Correlation plot of the converted isotope fractions vs. SiH4 flux ratio, k,
shown on a linear scale for samples deposited at several elevated temperatures. (a)
The raw SIMS isotope fractions for 29Si and 30Si are each converted to an expected
SiH4 fraction using their natural abundance, az. ctot., is fit (Eq. (6.7)) to the data
for each deposition temperature (solid, dashed, dotted lines) and is approximately
linear over this range with a slope of s. (b) ctot. fits from (a) asymptote to unity at
large values of k.
384
fraction (adsorbed SiH4) of zero. Figure 6.10 (b) illustrates the functional form of
the calculated SiH4 fraction, ctot., at large values of k = Fg/Fi where the SiH4 gas
flux dominates the ratio and ctot. asymptotes to unity.
From the fit values of ctot. at each temperature, the temperature dependence
of s is plotted in Fig. 6.11, which shows that s follows a similar trend to the 29Si
and 30Si isotope fractions in Fig. 6.9. In panel (a) of Fig. 6.11, s trends downwards
slightly from a value of 1.6(2)× 10−3 at 249 ◦C to a minimum of 2.9(4)× 10−4 at
502 ◦C. In this temperature range, the data appears to behave in a similar manner
to the sticking probability term resulting from physisorption, sc, which decreases
with increasing temperature. Then as T is increased more, s rapidly increases to
2.3(5)× 10−2 at 812 ◦C. This increase is expected qualitatively for the reactive stick-
ing coefficient term resulting from chemisorption, sr, which increases with increasing
temperature. A list of the samples discussed in this chapter, their deposition param-
eters, and measurement and analysis results can be found in Tables D.10 to D.13 in
Appendix D. These values of s are consistent with previously reported values of the
reactive sticking coefficient of silane species on Si surfaces, although there is a large
variation in the literature. Si CVD studies have shown sr to range from 5× 10−4 to
5× 10−3 for polycrystalline Si deposition at 600 ◦C to 800 ◦C [177], and it ranges
from 1× 10−3 to 3× 10−5 for Si(111) surfaces below 500 ◦C [187,188].
Next, the data in Fig. 6.11 is fit to the temperature dependent incorporation
model of Eq. (6.10), s(T ), which is the sum of the sticking terms, sc+sr. Ep, Ec, and
Ar are set as free parameters in the fit, which is shown in Fig. 6.11 (a) (line). Also
plotted separately are the sc (Eq. (6.8)) and sr (Eq. (6.9)) terms (dotted and dashed
385
Figure 6.11: (a) s (circles) vs. deposition temperature. The temperature dependent
incorporation model, s(T ), is fit to the data (Eq. (6.10)) and plotted (line) along
with the individual sticking terms calculated from the fit; sc (dotted line, Eq. (6.8)),
representing physisorption, and sr (dashed line, Eq. (6.9)), representing reactive
chemisorption. (b) Semi-log plot of the fits from (a) showing the crossover from
s(T ) dominated by sc to sr where s increases rapidly above 600
◦C.
386
lines, respectively), which are calculated from the fit parameters. Panel (b) shows
the fit and its modeled components on a semi-log plot to illustrate the crossover
from s(T ) being dominated by sc to s(T ) being dominated by sr. The fit of s(T )
matches the data fairly well at higher temperatures, while there is some deviation
in the lower temperature range. The fit also matches the minimum of s in the data
around 500 ◦C. The value of the activation energy for chemisorption, Ec, given by
this fit of Eq. (6.10) is 1.5(2) eV. The uncertainty in this value is the standard error
from the fit. The value of Ep generated by the fit, however, has a large uncertainty
indicating that the fit is under-constrained for the low temperature region, and so
the value is not necessarily reliable. Because the fit’s choice of Ep parameter can
also affect the high temperature part of the fit, the data dominated by Ec needs to
be isolated and analyzed separately, which will be discussed in the next section.
6.7 Determination of the Reactive Sticking
Activation Energy, Ec
To isolate the incorporation fraction data that is most sensitive to the value of
Ec, the data from Fig. 6.11 is plotted in an Arrhenius form for thermally activated
processes [189, 190], and the higher temperature data is fit to a line. This analysis
involves taking the natural log of s(T ) (Eq. (6.10)), the temperature dependent
incorporation model, giving
ln(s(T )) = ln
(
1− exp
(−Ep
kBT
)
+ Ar exp
(−Ec
kBT
))
. (6.11)
387
This equation in this form is not useful for this analysis because a linear term cannot
easily be extracted. However, for the higher temperature data, the reactive sticking
coefficient is expected to dominate, which means s ≈ sr, and so taking the natural
log of only the sr term (Eq. (6.9)) yields
ln(sr) = ln(Ar)− Ec
(
1
kBT
)
, (6.12)
which is the equation for a line with intercept ln(Ar) and slope Ec. Figure 6.12 is
a plot of ln(s) (circles) vs. inverse deposition temperature, 1
kBT
, modified by the
Boltzmann constant to have units of inverse energy. Also plotted is the natural log
of s(T ) (dotted line, Eq. (6.11)). The room temperature data resides outside the
graph window of Fig. 6.12 for clarity and to highlight the higher temperature data,
but it is included in the s(T ) fit. Using a linear fit to determine Ec is accurate
within the uncertainties in the linear region of the higher temperature data where
sr dominates. This linear fit (Eq. (6.12)) is also shown in Fig. 6.12 (solid line) for
the data between 502 ◦C and 812 ◦C. This fit gives a value for the chemisorption
activation energy of Ec = 1.1(1) eV. The uncertainty of this value is the standard
error from the fit. This value differs from the value of Ec obtained in the previous
section of 1.5(2) eV from the s(T ) fit. As was mentioned previously, the Arrhenius
fit is an alternative fitting method that gives a comparison to the previous fit of
s(T ) and is possibly more accurate because it allows analysis of Ec separate from
Ep and avoids convolution with the large uncertainty in the fit value of Ep. Both
of these values, however, are consistent with reported activation energies of SiH4
CVD between 600 ◦C and 800 ◦C. The literature values for SiH4 CVD activation
388
Figure 6.12: Arrhenius plot of ln(s) (circles) vs. inverse temperature in energy units.
The top axis shows the equivalent deposition temperature in ◦C. The data between
the 502 ◦C sample and the 812 ◦C sample are fit to a line (Eq. (6.12)) whose slope
gives an activation energy, Ec = 1.1(1) eV. Also plotted is the natural log of the fit
of s(T ) (dotted line, Eq. (6.11)).
energy can vary from about 0.4 eV to 2.2 eV depending heavily on experimental
conditions such as surface orientation, gas pressure, and hydride species [177, 187,
191–194]. An average value of the activation energies derived from data in Ref. [174]
for lower pressure CVD experiments with deposition temperatures between 500 ◦C
and 900 ◦C is approximately 1.1 eV.
The exponential prefactor, Ar, is required in the model for sr because it ac-
counts for the vertical offset of the data in the Arrhenius plot (Fig. 6.12) from the
389
expected intercept of zero. Leaving Ar as a free parameter allows for experimental
uncertainties that cannot reasonably be predicted such as the method used for es-
timating the SiH4 partial pressure having a systematic shift that propagates to the
calculated incorporation fraction values, which would affect Ar but not Ec. The pres-
ence of different silicon hydrides, (SiHx) or possibly disilane (Si2H6) which is known
to contaminate commercial silane and have higher sticking coefficients [188, 195],
in the background gas could also introduce errors because they may have different
reactive sticking coefficients leading to an average value being measured. Gates
has shown that the activation energy of Si CVD increases from around 1.17 eV to
2.04 eV for SiH3 and SiH respectively [194]. As stated previously, the prefactor for
sc is set to 1 because it is more reasonable that sc approaches unity at zero temper-
ature. Also, differences in reactive sticking of different molecules should not affect
the physisorption term as much, making it not as critical for the prefactor to be left
free.
6.8 Chapter 6 Summary
In this chapter, the measured enrichment (i.e. residual 29Si and 30Si isotope
fractions) of samples grown at both room temperature and with elevated substrate
temperatures ranging from 249 ◦C to 812 ◦C was analyzed to understand how en-
richment changes as a function of pressure and temperature due to SiH4 incorpo-
ration. The 29Si and 30Si isotope fraction were found to be highly correlated to
the background partial pressure of SiH4 during deposition. This showed that the
390
dominant, if not only, source of residual minor isotopes in 28Si samples is the nat-
ural abundance SiH4 gas which adsorbs into samples during growth with a room
temperature incorporation fraction, s = 6.8(3) × 10−4. From further analysis, the
temperature dependence of the incorporation fraction is determined and modeled
using two sticking terms in the temperature dependent incorporation model. A ph-
ysisorption sticking probability decreases weakly with increasing temperature while
a reactive sticking coefficient due to CVD-like chemisorption increases more strongly
with increasing temperature from a minimum of 2.9(4)× 10−4 for the 502 ◦C sample
up almost two orders of magnitude to 2.3(5)× 10−2 for the 812 ◦C sample. These
competing terms lead to a minimum in residual 29Si and 30Si isotopic concentration
at around 500 ◦C. The lowest 29Si isotopic concentration for a sample deposited
in that range was 127(29) ppb. As an alternative to the temperature dependent
incorporation model, an Arrhenius formalism was used to determine the activation
energy of the reactive sticking coefficient and found that Ec = 1.1(1) eV. Under-
standing the role of SiH4 gas sticking for a range of deposition temperatures allows
for better prediction of, and control over, the resulting enrichment. This knowledge
enables production of 28Si samples with targeted levels of enrichment (29Si isotope
fractions), which could facilitate a study of T2 coherence times as a function of
29Si
concentration.
391
Chapter 7
Summary of Results and Future
Experiments
7.1 Summary and Conclusions
In this work, a mass selected, hyperthermal energy ion beam deposition system
was used to achieve in situ isotopic enrichment of a total of three 22Ne samples, three
12C samples, and 61 28Si samples. Very high levels of enrichment were achieved,
especially for 28Si, which was also successfully deposited as epitaxial thin films.
Chapter 3 demonstrated successful proof of principle enrichment by implanting
the minor isotope 22Ne enriched with an isotope fraction of 99.455(36) % into Si as
well as depositing thin films of 12C enriched to an isotope fraction of 99.9961(4) %
using the mass selected ion beam system. Realized mass selectivities of 1785:1 for
22Ne and 276:1 for 12C were achieved. While these selectivities are not directly
translatable to 28Si enrichment, these experiments do show that very high levels of
in situ enrichment are possible in thin film deposition using this system.
Chapters 4 and 5 demonstrated that 28Si films could be deposited with residual
392
29Si and 30Si isotope fractions consistently below 1 ppm with both elevated substrate
temperatures and while using the high pressure plasma mode of the ion source. The
enrichment values of these samples meet the enrichment materials goal laid out at
the beginning of this thesis to achieve 28Si enrichments that surpass those of any
other known sources of 28Si including the IAC. The isotope reduction factor of 29Si
for the most highly enriched sample deposited at DC–3 is 3.7(8)× 105. This means
that there is approximately 3.7× 105 times less 29Si in that sample than in natural
abundance Si. This value of the reduction factor along with the reduction factor
values of the excluded isotopes from all the most highly enriched samples of 28Si,
22Ne, 12C are shown in the isotope reduction timeline of isotope reduction factors in
Fig. 7.1. A version of this figure was previously shown for Si samples in Fig. 5.54 in
Chapter 5.
Like that figure, Fig. 7.1 shows the isotope reduction factors vs. the sample
deposition date for samples with the highest enrichments achieved up to that point,
where a larger reduction factor is equivalent to a higher level of enrichment. As pre-
viously discussed, the isotope reduction factors are defined as the natural abundance
of an isotope of an element, az, divided by the measured isotope fraction of that iso-
tope. Here, the isotope fractions are written generally as zX/Xtot., where
zX refers
to the counts of a particular isotope in the measurement of a particular element,
and Xtot. is the sum of the counts of all isotopes being measured, as discussed in
Chapter 1. The isotope reduction factors are therefore written here as az/(
zX/Xtot.).
The isotope reduction factors for 22Ne (diamonds), 12C (circle), 29Si (squares) and
30Si (triangles) in Fig. 7.1 are shown to increase for samples deposited throughout
393
Figure 7.1: Isotope reduction timeline. Timeline of the progression of the isotope re-
duction factors for the lowest residual isotope fractions of 22Ne (diamonds), 12C (cir-
cle), 29Si (squares) and 30Si (triangles), as measured by SIMS. These were achieved
for 22Ne, 12C, and 28Si samples deposited over approximately five years.
this work. Uncertainties in the reduction factors are shown for all samples and are
derived from the SIMS measurements of the isotope fractions.
The overall 28Si isotope fraction of the most highly enriched 28Si sample de-
posited at DC–3 and in this entire work was 99.9999819(35) %. This demonstrates
an improvement over the initial 28Si sample deposited at IC–1 in the form of a re-
duction in 29Si of approximately 2.2 ×104 times. This and other similar samples
are more highly enriched than any other known source of 28Si, including that of
the IAC. These samples also demonstrate enrichments sufficient to enable a robust
measurement of the dependance of electron coherence times on 29Si concentration
394
in the single spin regime and compare it to theoretical predictions [12], as proposed
in Chapter 1 (see Fig. 1.9). The overall decrease in isotope fractions for isotopes
rejected during production of 22Ne, 12C, and 28Si samples can be seen in the en-
richment progression timeline for 20Ne (diamonds), 13C (circles), 29Si (squares) and
30Si (triangles) in Fig. 7.2. As mentioned previously, this timeline shows the SIMS
measurements of the isotope fractions vs. deposition date for the samples that were
the most highly enriched of any samples deposited up to that point. These are the
record enrichments of samples produced in this work. As previously discussed, the
isotope fractions are written generally here as zX/Xtot.. Uncertainties in the isotope
fractions are shown for all samples and are derived from the SIMS measurements,
as previously discussed. Also shown for comparison are the 29Si isotope fractions
of the 28Si epilayers and crystals produced by Isonics and Itoh using 28SiH4 CVD
(dash-double dotted line) from Ref. [36,37], the bulk 28Si material produced by the
IAC using 28SiH4 CVD (dash-dotted line) from Ref. [32], and the
28Si thin films
produced by Tsubouchi et al. using 28Si IBE (dotted line) from Ref. [43]. The 28Si
samples produced in this work have residual 29Si and 30Si isotope fraction that are
far less than those of these other sources.
Some of the key experimental changes that led to the improvements in en-
richment include improving the ion beam geometric mass selectivity through beam
tuning, depositing in a lower background pressure of SiH4, and achieving higher de-
position rates and smaller ion beam spot sizes through beam tuning. Most of these
factors relate to the adsorption of natural abundance SiH4 gas from the vacuum
during deposition, which lowers the resulting 28Si enrichment.
395
Figure 7.2: Enrichment progression timeline. A timeline of the progression of the
best residual isotope fractions of 20Ne (diamonds), 13C (circle), 29Si (squares), and
30Si (triangles), as measured by SIMS. These were achieved for 22Ne, 12C, and 28Si
samples deposited over approximately five years. Also shown for comparison are
the 29Si isotope fractions of the 28Si epilayers and crystals produced by Isonics and
Itoh using 28SiH4 CVD (dash-double dotted line) from Ref. [36, 37], the bulk
28Si
material produced by the IAC using 28SiH4 CVD (dash-dotted line) from Ref. [32],
and the 28Si thin films produced by Tsubouchi et al. using 28Si IBE (dotted line)
from Ref. [43].
Chapter 5 also demonstrated that sample heating was critical for achieving epi-
taxy. Crystalline, epitaxial 28Si films were produced using elevated deposition tem-
peratures between 349 ◦C and 1041 ◦C, although samples deposited above 600 ◦C
were very rough. This was due to chemical contaminants such as SiC at the growth
surface that result in step pinning sites that lead to step bunching, faceting on
{111} and {113} planes, and large mound formation. Smooth, epitaxial 28Si films
with typical surface widths of ∆z = 2 nm were achieved by depositing at lower
396
temperatures between 349 ◦C and 460 ◦C.
Chemical contaminants in these 28Si films were measured by SIMS, which de-
tected N, C, O, F, Al, and Cl. The N, C, and O were detected at especially high
concentrations, initially all above 1× 1019 cm−3. The dominant source of the N, C,
and O for this sample was the ion beam. N2 and CO ions were likely introduced
ballistically into the film from the 28 u beam. F and Al contamination were elimi-
nated for a later sample due to experimental alterations. By improving the vacuum
in both the deposition chamber and the ion beam chamber, and by using a higher
28Si ion beam current generated in the high pressure mode of the ion source, N, C,
and O were reduced in the final sample. The average atomic concentration of N
in the film was measured to be 9.07(4) ×1018 cm−3, or 181.8(8) ppm, the average
atomic concentration of C in the film was measured to be 8.27(3) ×1018 cm−3, or
165.7(6) ppm, and the average atomic concentration of O in the film was measured
to be 3.09(4) ×1018 cm−3, or 61.9(8) ppm. Based on these concentrations and the
solid solubility of these elements in Si, it is likely that Si3N4, SiC, and SiO exist
within the 28Si films. These remaining atomic concentrations were likely the result
of both adsorption from the background vacuum and ballistic transport from the
ion beam. The resulting best total chemical purity for a 28Si sample deposited at
DC–3 and overall in this work is approximately 99.96(2) %.
Finally, TEM was used to confirm that 29Si films produced with both the
higher and lower deposition temperatures were crystalline and epitaxially aligned
to the substrates. All samples deposited with substrate temperatures above 600 ◦C
were observed to have {111} stacking faults and twinning present in the films. For
397
lower deposition temperature samples, one sample deposited using the low pressure
mode did not have any visible stacking faults but probably still contained other crys-
talline defects. Another sample deposited using the high pressure mode to generate
a much higher deposition rate was observed to have a lot of {111} stacking faults
and likely other defects. This increased defect concentration may have been due
to the deposition rate for the second sample being over a factor of 12 larger while
the deposition temperature was slightly lower. In general, the crystalline defects in
these films are likely related to the presence of clusters of chemical contaminant.
Reduction of these crystalline defects likely requires reduction of the chemical con-
taminants within the 28Si films. The second and third materials goals for chemical
purity and crystallinity described in Chapter 1 were not pursued further within
the experiments depositing 28Si samples at DC–3, although improvements and new
understanding regarding the source of both chemical and crystalline defects were
made.
Chapter 6 explored and quantified the relation between the natural abundance
gas in the chamber and the resulting sample enrichments. The measured enrichment
(i.e. residual 29Si and 30Si isotope fractions) of samples grown at both room tem-
perature and with elevated substrate temperatures ranging from 249 ◦C to 812 ◦C
was analyzed to understand how enrichment changes as a function of pressure and
temperature due to SiH4 incorporation. The
29Si and 30Si isotope fractions were
found to be highly correlated to the background partial pressure of SiH4 during
deposition. This showed that the dominant, if not only, source of residual minor
isotopes in 28Si samples is the natural abundance SiH4 gas which adsorbs into sam-
398
ples during deposition with an incorporation fraction, s. From further analysis, the
temperature dependence of s is determined and modeled using two sticking terms in
a temperature dependent incorporation model. A physisorption sticking probability
decreases weakly with increasing temperature while a reactive sticking coefficient
due to CVD-like chemisorption increases more strongly with increasing tempera-
ture from a minimum of 2.9(4)× 10−4 for the 502 ◦C sample up almost two orders
of magnitude to 2.3(5)× 10−2 for the 812 ◦C sample. These competing terms lead
to a minimum in residual 29Si and 30Si isotopic concentrations at around 500 ◦C.
The lowest 29Si isotopic concentration for a sample deposited in that range was
127(29) ppb. As an alternative to the temperature dependent incorporation model,
an Arrhenius formalism was used to determine that the activation energy of the
reactive sticking coefficient is Ec = 1.1(1) eV. Understanding the role of SiH4 gas
sticking for a range of deposition temperatures allows for better prediction of, and
control over, the resulting enrichment. This knowledge can aid in enabling produc-
tion of 28Si samples with targeted levels of enrichment (29Si isotope fractions) that
could facilitate a study of T2 coherence times as a function of
29Si concentration, as
discussed in the next section.
The main impact of this work on the field of Si-based solid state quantum
computing is the demonstration of 28Si produced with extremely high levels of en-
richment that have never been produced before. This was achieved in a laboratory
setting as opposed to an industrial scale effort and was comparatively cheap. Ef-
fectively, a new material was engineered using a particular processing method, i.e.
ion beam in situ enrichment and deposition. Overall, the highest quality 28Si films
399
produced in this work, in terms of the materials goals stated in Chapter 1, were de-
posited with varied deposition conditions. However, 28Si films deposited using the
high pressure mode of the ion source with a deposition temperature of approximately
450 ◦C are most likely to result in the highest quality films. This temperature is
optimal for both enrichment and smooth epitaxial growth, and the high ion cur-
rent achieved in the high pressure mode should lead to a minimization of chemical
contaminants. The high growth rate however, may lead to diminished crystalline
quality.
7.2 Proposals for Future Experiments
7.2.1 Targeted Levels of Enrichment
One of the advantages of enriching material in situ nearly simultaneously with
deposition is the ability to more easily modulate the enrichment level throughout
the sample by switching the magnetic field of the sector mass analyzer to modulate
the isotope being deposited. As a demonstration of modulating the isotopic con-
centration of a material throughout its thickness, an isotope heterostructure that
alternates between layers of enriched 28Si and layers of near natural abundance Si
was produced. This sample was deposited at room temperature at LC–2. By switch-
ing the magnetic field of the sector mass analyzer from the 28 u beam to the 29 u
beam (which is composed of both 29Si and 28SiH), one can control which isotope
or combination of isotopes is deposited. In a typical SiH4 mass spectrum produced
when using the low pressure mode of the ion source, the ion current peak height at
400
Figure 7.3: SIMS depth profile of a Si isotope heterostructure deposited at room
temperature at LC–2. The isotope fractions are shown as a function of depth for
28Si (circles), 29Si (squares), and 30Si (triangles) in an isotope heterostructure. The
magnetic field of the sector mass analyzer was modulated to select 28 u or 29 u at
different depths, as noted on the top axis of the figure. The 29Si isotope fractions are
seen to be roughly the natural abundance within the 29 u regions and much lower
within the 28 u regions. The 29Si tail in 28 u regions is a measurement artifact.
The peak below a depth of 300 nm is due to ion beam tuning during the beginning
of the deposition. At a depth of about 351 nm, the isotope fractions return to
their natural abundance values, indicating the interface with the Si(100) substrate
(shaded region).
29 u is similar to the 28Si peak height, suggesting about 95 % of the 29 u peak is
composed of 28SiH, as previously discussed. The 29Si concentration when depositing
from the 29 u peak is then expected to be similar to the natural abundance value at
roughly 5 %. No 30Si is expected whether the analyzer is set to mass 28 u or 29 u.
SIMS was used to measure the isotope fractions of the isotope heterostructure.
A SIMS depth profile of this sample is shown in Fig. 7.3. Isotope fractions of 28Si
401
(circles), 29Si (squares), and 30Si (triangles) are shown as a function of sputter depth
into the film. The 29Si isotope fractions are seen to be roughly the natural abundance
within the 29 u regions and much lower within the 28 u regions. The 29Si tail in
28 u regions is a measurement artifact, and the true 29Si isotope fractions in these
regions is expected to be roughly 1× 10−6 (1 ppm), as seen at a depth of 50 nm.
The 30Si isotope fraction remains below roughly 1× 10−6 (1 ppm) throughout the
entire film, as expected. The peak in 29Si and 30Si below a depth of 300 nm is due
to ion beam tuning during the beginning of the deposition. At a depth of about
351 nm, the isotope fractions return to their natural abundance values, indicating
the interface with the Si(100) substrate, which is marked by the shaded region.
The ability to switch between 28Si and 29Si can enable deposition of 28Si sam-
ples with targeted enrichments, i.e. different isotopic concentrations of 29Si. To
produce a desired isotopic concentration of 29Si in a sample, one could deposit 28Si
normally and then during deposition, mass select for the 29 u ion beam for the ap-
propriate fraction of the total deposition time, taking into account the 28SiH in the
29 u peak. A duty cycle would be chosen to ensure sufficient mixing of the two iso-
topes, depending on the deposition rate. Samples with specific enrichments ranging
from natural abundance down to less than 1 ppm 29Si isotopic concentrations would
enable a measurement of T2 as a function of
29Si concentration for a wide range of
enrichment values, as discussed in Chapter 1. These measurements would begin to
address the question of “how good is good enough?” for the enrichment of 28Si in a
Si-based quantum computing architecture.
402
7.2.2 Enriched Si/Ge Deposition
Quantum wells in Si/SiGe heterostructures are a promising system for solid
state quantum information. Although 28Si has been used by several research groups
to produce quantum coherent devices, including a 28Si quantum well in SiGe [?],
fully enriched Si/SiGe devices have not yet been demonstrated. Like 28Si, 74Ge
has no net nuclear spin and is the most abundant of the Ge isotopes comprising
approximately 37 % of natural abundance Ge. The critical isotope to be removed
is 73Ge, which does have a nuclear spin of I = 9/2 and a natural abundance of
approximately 8 %.
The ability described previously to switch between different masses in the ion
beam during deposition enables the growth of heterostructures from isotopic mate-
rials. This is achieved by cycling between the constituents of compound materials
where both components are enriched, e.g. 28Si/28Si74Ge. To produce Ge ions, Ge
cathodes could be used in the solids-mode ion source designed for sputtering the
cathode material. The source of Si could either be SiH4 or a Si cathodes with Ar
being used as the working gas.
Magnetic peak switching could then be used, as described previously, to control
the relative concentrations of Si and Ge needed for SiGe deposition. A duty cycle
would be chosen to ensure that the compound is sufficiently mixed during deposition
to produced enriched 28Si74Ge. During deposition, a layer of pure 28Si would be
deposited as the quantum well layer followed by a capping layer of 28Si74Ge. Control
gates would then need to be patterned on the top surface of the heterostructure.
403
7.2.3 28Si sublimation
In order to facilitate collaborations with other research groups interested in
using enriched 28Si to fabricate quantum coherent devices, a useful experimental
technique would be thermal re-deposition of 28Si from a deposited thin film onto
a fabricated device as a capping layer. Quantum dot devices can benefit from not
only having a 28Si substrate, onto which the device is fabricated, but also a 28Si cap
to further isolate the qubit spins from 29Si. A demonstration of the re-deposition
of 28Si was attempted in this work by heating a 28Si thin film sample using the
DH method of heating to approximately 1100 ◦C to cause sublimation of the 28Si.
Another Si(100) substrate with a thermal oxide surface layer was used as a deposition
target. In order to attempt this transfer of 28Si, the sublimation rate needed to be
calibrated.
A calibration for three different sublimation temperatures was performed and
is shown in Fig. 7.4. The measured sublimation rates (triangles) at approximately
1100 ◦C, 1150 ◦C, and 1200 ◦C were determined from ellipsometry measurements
of the film thicknesses. Also shown are calculations for the expected sublimation
rate based on the Si vapor pressure at different temperatures (line). Taking into
account geometric factors of the deposition, a correction to the raw calculation was
applied. The corrected calculation (dashed line) has values that are 25 % of the raw
calculation and matches the measured rates well. The target substrate in the 28Si
sublimation test was measured by SIMS to detect the transferred 28Si film. However,
the presence of the thermal oxide and the fact that the film was very thin resulted
404
Figure 7.4: Si sublimation rates for a 28Si chip and calculated values. Measured rates
(triangles) at 1100 ◦C, 1150 ◦C, and 1200 ◦C are compared to calculations based on
Si vapor pressures (line) and a modified calculation (dashed line) whose values are
25 % of the raw calculation to account for geometric factors of the deposition.
in a null measurement with no 28Si being detected.
7.2.4 Al Dopant Devices with Hydrogen Lithography
STM hydrogen lithography is a promising technique for fabricating single atom
31P dopant devices for quantum information. A STM probe is used to selectively
remove H atoms from a H-terminated Si surface thus defining a lithographic pattern.
This surface is then exposed to phosphene gas which incorporates into the bare Si
within the pattern forming a single atom thick delta layer. Using this technique,
conducting wires and quantum dot islands comprised of 31P atoms, as well as single
atom islands can be fabricated [5]. Instead of using 31P, which is an electron donor,
it may be interesting to use an acceptor atom to create quantum coherent devices
based on the physics of hole transport. One candidate acceptor atom is Al. If Al
405
Figure 7.5: SIMS depth profile of an Al delta layer in Si showing atomic concen-
trations of Al, O, H, and P (lines) in the sample. Al is clearly concentrated in
a layer roughly 25 nm below the surface. Above this depth is the natSi capping
layer and below it is the Si(100) substrate. The peak atomic concentration of Al is
about 1× 1020 cm−3. The high concentration of O at the Al layer is likely due to
outgassing of the Al thermal deposition source.
can be selectively deposited in a delta layer onto Si(100) with a H pattern, then an
acceptor device may be possible.
To demonstrate the capability to deposit an Al delta layer, Al was deposited
on a Si(100) substrate with a roughly monolayer coverage and then overgrown with
natural abundance Si, natSi, from the evaporation source. The composition of this
sample was measured by SIMS in a depth profile, which is shown in Fig. 7.5. The
atomic concentrations of Al, O, H, and P (lines) are shown vs. the sputter depth
into the sample. One can see that there is a clear Al rich layer roughly 25 nm below
the surface. This first 25 nm represents the natSi capping layer. It is difficult to
406
determine from this measurement if the Al resides in a true delta layer, but the
Al likely spread out to some extent, possibly due to heating the sample during
natSi deposition. Beyond 25 nm represents the Si(100) substrate. The peak atomic
concentration of Al is approximately 1× 1020 cm−3, which is approximately 20 % of
the nominal density of crystalline Si. A large concentration of O is also observed at
the delta layer with a similar concentration as the Al, possibly due to the Al thermal
deposition source outgassing. This successful demonstration producing a confined
Al layer in Si is the first step towards patterned Al dopant devices defined by STM
hydrogen lithography. Selectivity of Al atoms deposited in a H pattern would need
to be tested further before production of a device.
7.2.5 Electrical Measurements and T2 in
28Si
Ultimately, the quality and usefulness of 28Si material produced in this work
to quantum information research will be determined by the T2 time of spins within
the material. Additionally, the electrical quality of the material is important for the
operation of quantum coherent devices. Several types of devices can be used to test
parameters related to the electrical quality including Schottky diodes, capacitors,
and transistors in Hall and Van der Pauw measurement geometries. A cartoon
schematic of a capacitor structure and measurement for a 28Si sample is shown in
Fig. 7.6. A 28Si film is seen on top of a natural abundance Si substrate. An isolation
gate oxide is grown from the 28Si film and then an Al gate 300 nm thick is deposited
on top of the oxide. Then, to perform measurements, a metal probe contacts the
top gate while the substrate contacts metal on the back side of the sample, and
407
Figure 7.6: Cartoon schematic of a 28Si capacitor. An isolating gate oxide is grown
on a 28Si film and an Al top gate is deposited on top of that, forming a capacitor. A
voltage sweep can then be applied between a metal probe contacting the gate and
a back metal contact behind the Si(100) wafer.
a voltage sweep is applied to perform a capacitance-voltage, C-V, measurement.
This measurement can determine the interface trap density and free charge carrier
density in the 28Si oxide. Transistors can also be fabricated on 28Si and used for
measuring electron mobility and carrier type within the enriched film.
While the 28Si films produced in this work typically do not comprise a large
enough volume of 28Si material to contain enough spins for traditional ESR mea-
surements, other specialized measurements can be used to measure T2 in these films.
For a sample implanted with 31P atoms, a small ESR probe can be fabricated onto
the surface of a 28Si thin film. Such a µm scale probe greatly increases the sensitivity
of the measurement and can measure a much smaller number of spins, potentially
as few as 10s of spins. A group led by Aharon Blank has tested this technique [40]
and has already made test devices on 28Si samples produced in this work. The
primary impediment of these measurements at the moment is the concentrations
of chemical contaminants within the 28Si samples. The concentration of N spins in
408
these samples exceeds that of the implanted 31P atoms, making a T2 measurement
impossible. Therefore, to enable these measurements, the purity of 28Si samples
must be improved and potentially can be by switching to a UHV compatible ion
source and a high purity gas feed line.
409
Appendix A: Ion Source and Beamline:
Additional Operating Parameters
This appendix gives additional operating parameters and analysis for the ion
source and ion beamline. A circuit diagram and schematic of the lens elements
of the beamline before the magnetic sector mass analyzer showing the relationship
between the controlling power supplies and measured currents is shown in Fig. A.1.
Operating parameter scans of the ion source and beamline using Ne in the gas mode
ion source are shown in Fig. A.2. These give the 20Ne ion current as a function
of source magnetic field, gas flow, arc voltage, and extractor voltage. Finally, the
operating parameters used for producing the most highly enriched 22Ne, 12C, and
28Si samples are given in Tables A.1 to A.3, respectively.
410
Figure A.1: Circuit diagram and schematic of the electrostatic lenses of the ion
beamline before the magnetic sector mass analyzer. Elements of the ion source
including the anode, cathodes and the source electromagnet coil are shown at the
left. The lens elements that form the potential landscape of the ion beam, depicted
from left to right, are the extractor, at negative voltage, VExt, a transport tube, at
negative transport voltage, VT , the focus, at negative voltage, VF , another transport
tube with X-Y deflectors biased positively with VXY , the skimmer, and the electron
suppressor, at negative voltage, Ve, which is set to a potential about 100 V more
negative than VT . VExt, VF , VXY , and Ve all float on top of VT . VA applies a
positive voltage to the anode, and Varc floats on top of the anode potential to apply
a negative voltage to the cathodes. The source electromagnet drives current ISM
through the solenoid and it floats on top of Varc due to the magnet housing not
being well isolated from the cathode housing. Iarc is the measured current flowing
between the anode and cathodes. All ion current that is neutralized on these lens
elements or the vacuum chamber is recorded as IT .
411
Figure A.2: Operating parameter scans of the gas mode ion source using Ne. (a)
The 20Ne ion current, Ii, and the arc current, Iarc, are shown as a function of the
source magnet field. The corresponding current, ISM , applied to the electromagnet
is shown on the top axis. The plasma ignites above 0.05 T and two peaks are
observed. (b) Ii and Iarc for
20Ne are shown as a function of the Ne gas flow, with
the corresponding pressure in the ion beam chamber displayed on the top axis. The
highest ion current is observed at a flow of about 6 mL/h. (c) Ii and Iarc for
20Ne
are shown as a function of the arc voltage, Varc. Ii and Iarc appear inversely related.
(d) Ii and the ion current on the skimmer, Isk, for
20Ne are shown as a function
of the extractor voltage, VExt. Ii was optimized at each value of Varc by tuning the
other lens elements. A peak in Ii at about -6.5 kV is observed.
412
22Ne Ion Beam Parameters
for 110504-22Ne-OxSi on 5/4/11
Source P VA Varc VC VT
(Pa) (V) (V) (V) (kV)
1.0× 10−2 501 500 0 -4.006
Sample Bias VExt VF ISM Iarc
(kV) (kV) (kV) (A) (mA)
-4.00 -7.0 -13.0 1.1 22
Table A.1: Ion source and beamline operating parameters for implanting 22Ne sample
110504-22Ne-OxSi on 5/4/11 using the solids-mode ion source. This sample was the most
highly enriched 22Ne sample produced in this work. Ne was used as the working gas for
this sample. VA is the anode voltage, and Varc is the arc voltage applied in reference to VA
to get the resulting cathode voltage, VC . VT is the transport voltage. A Si(100) substrate
was biased negatively for implantation. VExt is the extractor voltage and VF is the focus
voltage. The ion source plasma is ignited by setting the source electromagnet current,
ISM , producing an arc current, Iarc.
12C Ion Beam Parameters
for 120207-12C-OxSi on 2/7/12
Source P VA Varc VC VT
(Pa) (V) (kV) (V) (kV)
3.1× 10−3 608 1.147 -539 -4.00
VExt VF ISM Iarc Isk
(kV) (kV) (A) (mA) (µA)
-7.50 -15.0 1.45 25 32
Table A.2: Ion source and beamline operating parameters for depositing 12C sample
120207-12C-OxSi on 2/7/12 using the solids-mode ion source. This sample was the most
highly enriched 12C sample produced in this work. CO2 was used as the working gas for
this sample. VA is the anode voltage, and Varc is the arc voltage applied in reference to
VA to get the resulting cathode voltage, VC . VT is the transport voltage. VExt is the
extractor voltage and VF is the focus voltage. The ion source plasma is ignited by setting
the source electromagnet current, ISM , producing an arc current, Iarc. Isk is the ion
current measured on the skimmer.
413
28Si Ion Beam Parameters
for 151210-KD-7-i-28Si-500C on 12/18/15
Source P VA Varc VC VT
(Pa) (V) (kV) (kV) (kV)
2.1× 10−4 50 3.420 -3.370 -4.00
VExt VF ISM Iarc IT Isk
(kV) (kV) (A) (mA) (mA) (µA)
-11.0 -12.5 1.69 0.7 0.6 190
Deceleration Lenses
A2 A3 B2 B3 B4 X
(kV) (kV) (kV) (kV) (kV) (V)
-2.50 -0.49 -3.04 -0.70 -1.34 -14
Table A.3: Ion source and beamline operating parameters for depositing 28Si sample
151210-KD-7-i-28Si-500C on 12/18/15 using the gas-mode ion source. This sample was
the most highly enriched 28Si sample produced in this work. SiH4 was used as the working
gas for this sample. VA is the anode voltage, and Varc is the arc voltage applied in reference
to VA to get the resulting cathode voltage, VC . VT is the transport voltage. VExt is the
extractor voltage and VF is the focus voltage. The ion source plasma is ignited by setting
the source electromagnet current, ISM , producing an arc current, Iarc. IT is the ion
current measured on the transport voltage line, and Isk is the ion current measured on
the skimmer. A2, A3, B2, B3, B4, and X refer to deceleration lens elements.
414
Appendix B: Experimental Apparatus
Photographs
This appendix shows photographs of various experimental apparatus used for
producing enriched samples in this work including the hyperthermal energy ion beam
deposition system as a whole, the gas manifold, the two ion sources, the electrostatic
and magnetic lens elements comprising the beamline, the three mass-selecting aper-
tures, a gas aperture, experimental setups and sample stages for samples produced
at IC–1 in the ion beam chamber, LC–2 in the deceleration lens chamber, and DC–3
in the deposition chamber, and finally the sample apertures used for monitoring the
ion beam current.
415
Figure B.1: Photographs of the ion beam deposition system (a) Side view of the
ion beamline and deposition system. (b) Top-down view of the ion beamline. (c)
Side view of the deposition and analysis chamber showing the manipulator where
samples are located at DC–3.
416
Figure B.2: Photograph of the gas manifold. The natural abundance SiH4 gas
source used for depositing 28Si is connected at the left. The manifold contains two
gas reservoir tanks in the middle and a vacuum pumpout at the bottom. The
manifold connects to the ion source inlet out of view at the top right.
Figure B.3: Photographs of the ion source elements. (a) Inlet cathode housing of the
solids- mode ion source with a Cu cathode. (b) Anode housing of the solids-mode
ion source with a Cu anode disk. This housing fits over the cathode in (a). (c) Inlet
cathode housing of the gas-mode ion source with a Cu cathode. (d) Anode housing
of the gas-mode ion source with a cylindrical steel anode. The anode housing is seen
fitted over the cathode in (c).
417
Figure B.4: Ion beamline electrostatic elements and magnetic sector mass analyzer.
(a) The electrostatic elements used to generate an ion beam from the source are
seen with the extractor at the left, focus and X-Y deflectors in the middle, and
skimmer at the right. These reside between the ion source and the mass analyzer.
(b) Deceleration lenses after the mass analyzer focus ions onto the sample are seen
with alumina support rods and a teflon separator on top. (c) The bottom pole piece
and solenoid of the magnetic sector mass analyzer is seen with the entrance from
the beamline at the left and the exit leading to the mass-selecting aperture at the
bottom. There is a 90◦ bend in the vacuum pipe of the beamline.
418
Figure B.5: Photograph of the mass-selecting aperture used for 22Ne and 12C sam-
ples. The aperture consists of a 5 mm diameter hole in a standard 2.75 in size
(69.85 mm vacuum flange spacer 16 mm thick.
Figure B.6: Photographs of the mass-selecting aperture used for 28Si samples at IC–
1 and LC–2. The aperture consists of a 1 mm wide slit that is 15.25 mm tall and
2 mm thick in a Cu gasket. (a) A new aperture is seen at the exit of the magnetic
mass selecting magnet. (b) Closeup of a new aperture. (c) A used aperture had dark
deposited spots from the 30 u (far left of the aperture) and 29 u (left of the aperture)
ion beams. The separation between the 28Si beam centered on the aperture and the
29Si beam is about 6.2 mm.
419
Figure B.7: Photograph of the mass-selecting aperture used for 28Si samples at DC–
3. The aperture consists of a 2 mm wide slit that is 12 mm tall with a beveled edge
in a Cu gasket. The beveled edge is on the back side with the side shown facing the
ion beam.
Figure B.8: Photograph of a gas aperture on the inlet to the deceleration lenes in
the ion beam. The aperture consists of a 12.7 mm by 6.4 mm rectangular opening
in a stainless steel shim. The purpose of this gas aperture is to direct gas diffusing
from the ion beam chamber around the lenses to then be pumped away.
420
Figure B.9: Photographs of the experimental setup and sample stage for 28Si samples
deposited at IC–1 in the ion beam chamber. (a) An electrical vacuum feedthrough
at the left serves as a sample stage and is connected to an electric break at the exit
of the magnetic sector mass analyzer. (b) The feedthrough flange is seen. (c) A 28Si
sample is mounted on the end of the feedthrough using carbon tape.
Figure B.10: Photograph of the experimental setup of the deceleration lens chamber
for samples produced at LC–2. The ion beam enters the lens chamber from the right.
An electrical vacuum feedthrough serves as a sample stage at the right.
421
Figure B.11: Photographs of the sample stage for producing sample at LC–2. (a)
The sample stage flange with Cu electrical vacuum feedthrough for mounting sam-
ples is seen in the middle. Additional electrical feedthroughs are seen to the left
and right connected to the fixed sample aperture in the metal mask. (b) A stain-
less steel shim makes up the mask that covers the sample on the feedthrough. The
fixed sample aperture consists of a hole about 3 mm in diameter centered over the
feedthrough. (c) A Si(100) substrate is seen mounted on the end of the feedthrough
using carbon tape. (d) The Cu feedthrough is separated from the rest of the sample
stage with a 28Si sample mounted on the end.
422
Figure B.12: Photographs of the manipulator used for heating substates and samples
and depositing 28Si samples at DC–3 in the deposition chamber. (a) The manip-
ulator with the sample stage is seen in the deposition chamber with the ion beam
deceleration lenses behind it. (b) A closeup of the sample stage outside of the
chamber with the RH contacts visible at the left, the RH tungsten wire heater in
the middle below where the sample sits, and the DH contact visible at the right.
423
Figure B.13: Photographs of interchangeable sample apertures used to measure ion
beam currents prior to depositing 28Si samples at DC–3. (a) The side view of an
initial sample aperture consisting of a 2.2 mm diameter hole in a stainless steel shim
mounted on a sample holder. (b) The aperture in (a) sits on the manipulator and
faces the ion beamline to measure ion current. (c) The side view of a later sample
aperture consisting of a 2.5 mm diameter hole, and a 1 mm diameter hole in a Mo
shim mounted on a sample holder. (d) The aperture in (c) sits on the manipulator to
measure ion current. The 2.5 mm diameter aperture is at the center of the shim and
the 1 mm diameter aperture is offset to the right. Both apertures are electrically
isolated from the sample holders and connect to the DH contact.
424
Appendix C: Substrate Catalog
The table shown here gives the specifications and source of the various sub-
strates used for producing enriched samples in this work.
425
# Material Dopant
ρ Thickness
Supplier Identifier
(Ω · cm) (µm)
(1) Si(100) 200* lightly doped
(2) Ag 25 Kurt J. Lesker Company Pure Al Foil
(3) Si(100) boron 1 to 10 380 University Wafer Prime lot 12/0604
(4) Si(100) SOI boron 10 to 30 0.035† Dr. Neil Zimmerman (NIST) NIST-2
(5) Si(100) phosphorous 5 to 10 600 Dr. Mike Stewart (NIST) 55 nm thermal Ox
(6) Si(100) phosphorous 1 to 5 300 University Wafer 20141231-12
(7) Si(100) boron 7 to 20 340 University Wafer 4/15/15
(8) Si(100) undoped (intrinsic) 2× 104 380 University Wafer FZ SEMI Prime L849
(9) Si(100) boron 5 to 10 300 ITME 3611/110057
(10) Si(100) phosphorous 7 to 20 300 Virginia Semiconductor ± 0.05◦ 15-10973
Table C.1: Substrate catalog for substrates used for depositing enriched films. Substrate numbers are referenced in the sample catalogs.
*The wafer thickness of substrate (1) is an estimate. †The thickness of substrate (4) is the thickness of the Si device layer in the SOI
stack. This SOI wafer also possessed a 400 nm buried oxide layer. A 55 nm thermal oxide was grown on the surface of the wafer for
substrate (5). The wafer for substrate (10) was specified to have a miscut tolerance of ± 0.05◦ with the {100} surface.
426
Appendix D: Sample Catalogs
The sample catalogs presented here list all of the enriched samples produced in
this work and are organized several ways. The 22Ne sample implantation parameters
and measured enrichments are summarized in Table D.1. The 12C sample deposition
parameters and measured enrichments are summarized in Table D.2. 28Si samples
are organized by their deposition location, IC–1, the ion beam chamber, LC–2, the
deceleration lens chamber, and DC–3, the deposition and analysis chamber. 28Si
sample deposition parameters for samples deposited at IC–1 are summarized in Ta-
ble D.3. 28Si sample deposition parameters for samples deposited at LC–2 are sum-
marized in Tables D.4 and D.5. Finally, the 28Si sample deposition parameters for
samples deposited at DC–3 are summarized in Tables D.6 to D.9. Enrichment mea-
surements and parameters used in the modeling analysis discussed in Chapter 6 are
presented for 28Si samples deposited at room temperature in Tables D.10 and D.11.
Similarly, the enrichment measurements and modeling parameters for samples de-
posited with elevated temperatures are presented in Tables D.12 and D.13.
427
#
Sample Load Dep. Substrate Aperture Base P Implant P VA VT Ei
Name Date Date Type (mm) (Pa) (Pa) (V) (kV) (eV)
1 110301-22Ne-OxSi 2/28/11 3/1/11 Si(100):Ox (1) 5 1.1× 10−5 404 -3.885 ≈ 3000
2 110418-22Ne-OxSi 3/1/11 4/18/11 Si(100):Ox (1) 5 7.6× 10−7 3.3× 10−5 500 -4.002 4460
3 110504-22Ne-OxSi 4/28/11 5/4/11 Si(100):Ox (1) 5 3.1× 10−6 501 -4.006 4460
Isotope Fraction (zNe/Netot.)
#
Sample Ii Ion Dose Spot Size Implant Primary
22Ne 20Ne
Name (µA) (cm−2) (mm2) (nm) Measurement (%) (%)
1 110301-22Ne-OxSi 0.14* 7.7× 1016* 2* 9 SIMS 88.4(10) 15.2(10)
2 110418-22Ne-OxSi 0.14* 7.7× 1016* 5* 12.5
3 110504-22Ne-OxSi 0.14 7.7× 1016 2* 12.5 SIMS 99.455(36) 0.545(36)
Table D.1: 22Ne Sample Catalog (3/1/11–5/4/11). 22Ne samples were implanted at room temperature (≈ 21 ◦C) using Ne in the
solids-mode ion source at LC–2, the deceleration lens chamber. Substrates had a native oxide and were not prepared ex situ or in situ.
Aperture gives the diameter of the circular mass-selecting aperture. Pressures are raw reading of the lens chamber base and with the ion
source on. VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii is the average ion
current. Implant refers to the peak implantation depth. Isotope fraction measurements assume no 21Ne is present. Items marked with *
are estimates.
428
#
Sample Load Dep. Substrate Aperture Base P Dep. P VA VT Ei Ii
Name Date Date Type (mm) (Pa) (Pa) (V) (kV) (eV) (µA)
1 120125-12C-OxSi 1/24/12 1/25/12 Si(100):Ox (1) 5 4.4× 10−6 596 -4.00 543 0.55
2 120206-12C-OxSi 2/1/12 2/6/12 Si(100):Ox (1) 5 608 -4.00 554 0.40
3 120207-12C-OxSi 2/6/12 2/7/12 Si(100):Ox (1) 5 1.7× 10−6 2.1× 10−5 608 -4.00 554 0.70
Isotope Fraction (zC/Ctot.)
#
Sample C. I. Ion Dose Spot Size R d Film Mass Primary 12C 13C
Name (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement(s) (%) (ppm)
1 120125-12C-OxSi 0.01 2.7× 1018 2.3 0.68* 206* 0.95 SIMS 99.996107(88) 38.93(88)
2 120206-12C-OxSi 0.00012 4.7× 1016 1.6 0.77* 4* 0.01 SEM
3 120207-12C-OxSi 0.00555 2.9× 1018 1.2 0.76 100 0.52 SIMS, SEM 99.99608(13) 39.2(13)
Table D.2: 12C Sample Catalog (1/25/12–2/7/12). 12C samples were deposited at room temperature (≈ 21 ◦C) using CO2 in the
solids-mode ion source at LC–2, the deceleration lens chamber. Substrates had a native oxide and were not prepared ex situ or in situ.
Aperture gives the diameter of the circular mass-selecting aperture. Pressures are raw reading of the lens chamber base with the ion
source on. VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii is the average ion
current. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Isotope fraction measurements of sample
#1 were not corrected for instrumental error. Items marked with * are estimates.
429
#
Sample Load Dep. Substrate Aperture Base P Dep. P VA VT
Name Date Date Type (mm) (Pa) (Pa) (V) (kV)
1 120604-28Si-OxSi 6/1/12 6/4/12 Si(100):Ox (1) 1 6.5× 10−6 1.6× 10−4 514 -4.00
2 120613-28Si-OxSi 6/4/12 6/13/12 Si(100):Ox (1) 1 1.2× 10−5 1.7× 10−4 106 -4.315
3 120618-28Si-OxSi 6/14/12 6/18/12 Si(100):Ox (1) 1 1.3× 10−5 2.3× 10−4 106 -4.314
4 120627-28Si-Ag 6/19/12 6/27/12 Ag foil (2) 1 1.3× 10−5 2.1× 10−4 106 -4.312
5 120628-28Si-OxSi 6/27/12 6/28/12 Si(100):Ox (1) 1 1.3× 10−5 1.8× 10−4 106 -4.311
#
Sample Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement
1 120604-28Si-OxSi 455 0.95 0.0102 3.4× 1017 19 0.07 12 1.36 SIMS
2 120613-28Si-OxSi 64 1.10 0.0136 3.2× 1017 27 0.42 86 3.63 SEM
3 120618-28Si-OxSi 64 0.77 0.0114 4.8× 1017 15 0.40 92 3.04 TOF-SIMS
4 120627-28Si-Ag 64 0.95 0.0109 2.5× 1017 27 0.24* 46* 2.90
5 120628-28Si-OxSi 64 0.85 0.0075 2.9× 1017 16 0.59 86 1.98 SIMS
Table D.3: 28Si Sample Catalog: IC–1 (6/4/12–6/28/12). 28Si samples were deposited at room temperature (≈ 21 ◦C) using SiH4 with
the low pressure mode of the gas-mode ion source at IC–1, the ion beam chamber. Substrates had a native oxide and were not prepared
ex situ or in situ. Aperture gives the width of the mass-selecting aperture slit. Pressures are raw reading of the ion beam chamber base
and with the ion source on. VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii
is the average ion current. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and rates
correspond to the thickest measured film area. Items marked with * are estimates.
430
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P VA
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) (V)
6 130204-28Si-Si 2/1/13 2/4/13 Si(100) (1) HF etch 1 3.2× 10−7 5.9× 10−6 100
6a 151109-KD-2-P-Ox-28SiSub 11/9/15 1/1/16 Si(100):P:Ox (5) 2.3× 10−8 2.7× 10−7
7 130206-28Si-Si 2/5/13 2/6/13 Si(100) (1) HF etch 1 7.5× 10−7 7.2× 10−6 113
8 130208-28Si-Si 2/7/13 2/8/13 Si(100) (1) HF etch 1 1.3× 10−6 6.1× 10−6 150
9 130213-28Si-Si 2/11/13 2/13/13 Si(100) (1) HF etch 1 4.5× 10−7 6.4× 10−6 100
10 130215-28Si-Si 2/14/13 2/15/13 Si(100) (1) HF etch 1 7.6× 10−7 4.8× 10−6 150
11 130221-28Si-Si 2/19/13 2/21/13 Si(100) (1) HF etch 1 3.7× 10−7 4.9× 10−6 150
12 130227-28Si-Si 2/22/13 2/27/13 Si(100) (1) HF etch 1 2.0× 10−7 5.7× 10−6 100
13 130304-28Si-Si 3/1/13 3/4/13 Si(100) (1) HF etch 1 6.7× 10−7 6.0× 10−6 300
#
Sample VT Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (kV) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement
6 130204-28Si-Si -4.000 74 0.57 0.0067 2.6× 1017 16 0.56 110 1.76 SIMS
6a 151109-KD-2-P-Ox-28SiSub 22 0.56 50 SIMS
7 130206-28Si-Si -3.941 86 0.50 8 0.40* 80*
8 130208-28Si-Si -4.000 121 0.56 0.0081 7.6× 1017 7 1.49 384 1.96 SIMS
9 130213-28Si-Si -4.000 76 0.36 0.0046 2.9× 1017 10 0.17 37 1.20 TEM
10 130215-28Si-Si -3.997 122 0.36 0.0058 4.0× 1017 9 0.25* 67* 1.38 XPS
11 130221-28Si-Si -4.000 124 0.26 0.0035 3.1× 1017 7 0.23* 51* 0.83 SEM
12 130227-28Si-Si -3.998 74 0.5* 0.0021 5.4× 1017 2 1.41* 97* 0.54
13 130304-28Si-Si -4.000 160 0.65 0.0054 6.0× 1017 6 0.67 92* 1.20 SEM
Table D.4: 28Si Sample Catalog: LC–2: I (2/4/13–3/4/13). 28Si samples were deposited at room temperature (≈ 21 ◦C) using SiH4
with the low pressure mode of the gas-mode ion source at LC–2, the deceleration lens chamber. Substrates were etched with HF ex situ
but not prepared in situ. Aperture gives the width of the mass-selecting aperture slit. Pressures are raw reading of the lens chamber
base and with the ion source on. VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and
Ii is the average ion current. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and rates
correspond to the thickest measured film area. Items marked with * are estimates. Sample #6a was deposited at DC–3 by sublimating
sample #6 at about 1100 ◦C.
431
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P VA
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) (V)
14 130307-28Si-OxSi 3/4/13 3/7/13 Si(100) (1) HF etch 1 2.4× 10−7 4.0× 10−6 325*
15 130311-28Si-Si 3/8/13 3/11/13 Si(100) (1) HF etch 1 2.5× 10−7 5.5× 10−6 325*
16 130328-28Si(dP)-Si 3/25/13 3/28/13 Si(100) (1) HF etch 1 2.3× 10−7 3.7× 10
−6 (low)
142*
1.9× 10−4 (high)
17 130404-28Si(dP)-Si 4/1/13 4/4/13 Si(100) (1) HF etch 1 2.4× 10−7 3.9× 10
−6 (low)
130*
3.1× 10−4 (high)
18 130412-2829Si(28SiH)-Si 4/8/13 4/12/13 Si(100):B (3) HF etch 1 1.9× 10−7 3.6× 10−6 116
19 130520-28Si-SOI 4/15/13 5/20/13 Si(100) SOI (4) HF etch 1 1.1× 10−7 3.7× 10−6 229
20 130920-28Si(dP)-SOI 8/2/13 9/20/13 Si(100) SOI (4) HF etch 1 3.6× 10−8 5.1× 10
−6 (low)
100
5.3× 10−5 (high)
21 130924-28Si-Si-550C 9/23/13 9/24/13 Si(100):B (3) HF etch 1 1.1× 10−6 2.1× 10−5 200
#
Sample VT Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (kV) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement
14 130307-28Si-OxSi -4.000 230* 0.54 0.0089 6.0× 1017 9 1.52 415 1.94 SEM
15 130311-28Si-Si -4.000 230* 0.60 0.0066 4.3× 1017 10 0.35* 64* 1.45
16 130328-28Si(dP)-Si -3.995 64* 0.42 0.0070 2.7× 1017 6 0.93 249 2.00 SIMS
17 130404-28Si(dP)-Si -3.995 55* 0.43 0.0073 8.8× 1017 5 0.62* 175* 2.10 TOF-SIMS
18 130412-2829Si(28SiH)-Si -3.993 50 0.37 0.0070 9.1× 1017 5 0.84 351 1.92 SIMS
19 130520-28Si-SOI -4.000 179* 0.48 0.0040 3.8× 1016 7 1.07 149 0.89 SIMS
20 130920-28Si(dP)-SOI -4.000 80 0.62 0.0075 8.8× 1017 5 1.41 285 1.92 SIMS
21 130924-28Si-Si-550C -4.000 75 0.41 0.0007 1.5× 1017 3 0.74 22 0.19
Table D.5: 28Si Sample Catalog: LC–2: II (3/7/13–9/24/13). 28Si samples were deposited at room temperature (≈ 21 ◦C) using SiH4
with the low pressure mode of the gas-mode ion source at LC–2, the deceleration lens chamber. Sample #21 was deposited at about
550 ◦C. Substrates were etched with HF ex situ but not prepared in situ. Aperture gives the width of the mass-selecting aperture slit.
Pressures are raw reading of the lens chamber base and with the ion source on. For samples #16, #17, and #20, the pressure was
modulated from low to high to low. VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample,
and Ii is the average ion current. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and
rates correspond to the thickest measured film area. Items marked with * are estimates.
432
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P VA VT
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) (V) (kV)
22 140224-28Si-Si-750C 2/7/14 2/24/14 Si(100):B (3) 2 2.4× 10−8 2.1× 10−7 40 -4.000
23 140521-28Si-Si-800C 3/25/14 5/21/14 Si(100):B (3) 2 4.9× 10−9 2.4× 10−7 35 -4.000
24 140526-28Si-Si-700C 5/20/14 5/26/14 Si(100):B (3) 2 2.9× 10−9 2.0× 10−7 35 -4.000
25 140603-28Si-Si-700C 2/22/14 6/3/14 Si(100):B (3) 2 3.7× 10−9 2.0× 10−7 35 -4.000
26 140604-28Si-Si-600C 2/22/14 6/4/14 Si(100):B (3) 2 5.7× 10−9 2.0× 10−7 35 -4.000
27 140610-28Si-Si-900C 5/23/14 6/10/14 Si(100):B (3) 2 2.8× 10−9 2.0× 10−7 35 -4.000
28 140619-28Si-Si-700C 6/5/14 6/19/14 Si(100):B (3) 2 2.5× 10−9 2.0× 10−7 35 -4.001
29 140627-28Si-Si-1050C 5/28/14 6/27/14 Si(100):B (3) 2 2.8× 10−9 2.7× 10−7 35 -4.000
30 140710-28Si-Si-1010C 6/26/14 7/10/14 Si(100):B (3) 2 4.9× 10−9 1.5× 10−7 36 -4.000
31 140716-28Si-Si-1000C 6/26/14 7/16/14 Si(100):B (3) 2 5.1× 10−9 1.9× 10−7 50 -4.000
#
Sample T Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (◦C) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement(s)
22 140224-28Si-Si-750C 759 50 0.18 0.0012 28* 0.21* 22 SIMS, SEM
23 140521-28Si-Si-800C 812 33 0.42 0.0044 3.7× 1017 7 0.90 158 1.20 SIMS, SEM, XPS
24 140526-28Si-Si-700C 708 33 0.58 0.0054 4.1× 1017 8 0.77 120 1.50 SEM, TEM
25 140603-28Si-Si-700C 708 33 0.51 0.0032 4.5× 1017 4 1.20 126 0.90 SIMS, SEM, XPS
26 140604-28Si-Si-600C 610 33 0.50 0.0063 5.3× 1017 7 0.77 162 1.77 SIMS, SEM
27 140610-28Si-Si-900C 920 33 0.53 0.0049 8.3× 1017 4 2.60* 400* 1.38 SEM
28 140619-28Si-Si-700C 708 34 0.53 0.0057 9.9× 1017 4 0.84 150 1.60 SIMS, SEM, TEM
29 140627-28Si-Si-1050C 1085 33 0.60 0.0054 1* 1.50
30 140710-28Si-Si-1010C 1041 33 0.57 0.0060 0.29 50 1.77 SIMS, SEM, TEM
31 140716-28Si-Si-1000C 1030 31 0.55 0.0060 1.69
Table D.6: 28Si Sample Catalog: DC–3: I (2/7/14–6/26/14). 28Si samples were deposited using SiH4 with the low pressure mode of the
gas-mode ion source at DC–3, the deposition chamber. Substrates were not prepared ex situ but were flash annealed in situ. Aperture
gives the width of the mass-selecting aperture slit. Pressures are raw reading of the deposition chamber base and with the ion source on.
VA is the anode voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii is the average ion current. T
is the substrate deposition temperature. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses
and rates correspond to the thickest measured film area. Items marked with * are estimates.
433
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P VA VT
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) (V) (kV)
32 140828-28Si-Si 6/30/14 8/28/14 Si(100):B (4) 2 2.4× 10−9 1.2× 10−7 85 -4.00
33 150202-28Si-Si-800C 1/27/15 2/2/15 Si(100):P (5) HF etch 2 2.1× 10−8 1.0× 10−7 100 -4.00
34 150323-28Si-Si-700C 3/17/15 3/23/15 Si(100):B (4) HF etch 2 1.9× 10−8 3.2× 10−7 100 -4.00
35 150627-KD-7-P-28Si-600C 6/27/15 7/6/15 Si(100):P (6) CMOS clean 2 2.7× 10−8 6.5× 10−7 100 -4.00
36 150627-KD-3-P-28Si-700C 6/27/17 7/9/15 Si(100):P (6) CMOS clean 2 2.1× 10−8 5.5× 10−7 120 -4.00
37 150707-KD-1-P-28Si-800C 7/7/15 7/14/15 Si(100):P (6) CMOS clean 2 1.5× 10−8 7.6× 10−7 108 -4.00
38 150715-KD-5-P-28Si-800C 7/15/15 7/17/15 Si(100):P (6) CMOS clean 2 1.6× 10−8 1.2× 10−6 100 -4.00
39 150719-KD-2-B-28Si-700C 7/19/15 7/22/15 Si(100):B (7) CMOS clean 2 1.5× 10−8 1.0× 10−6 100 -4.00
40 150715-KD-9-B-28Si-700C 7/15/15 7/22/15 Si(100):B (7) CMOS clean 2 1.5× 10−8 1.0× 10−6 100 -4.00
41 150721-KD-5-P-28Si-700C 7/21/15 7/27/15 Si(100):P (6) CMOS clean 2 1.5× 10−8 1.0× 10−6 120 -4.00
#
Sample T Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (◦C) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement(s)
32 140828-28Si-Si 21* 38 0.25 0.0025 6.0× 1017 3 0.32 53 0.69 SIMS
33 150202-28Si-Si-800C 804 33 0.41* 2 1.4* 170* SIMS, SEM
34 150323-28Si-Si-700C 709 20 0.32 0.0002 8 0.06
35 150627-KD-7-P-28Si-600C 619 12 0.64 0.0004 0.11
36 150627-KD-3-P-28Si-700C 712 24 0.65 0.0101 1.3× 1018 5 0.54 140 2.89 SIMS, TEM, XPS
37 150707-KD-1-P-28Si-800C 808 34* 0.85 0.0071 1.1× 1018 4 1.57* 217* 1.97 SEM
38 150715-KD-5-P-28Si-800C 808 19 0.64* 0.0077* 2.3× 1018 2 0.60 120 2.13* SIMS, SEM
39 150719-KD-2-B-28Si-700C 712 18 0.72 0.0002 5* 0.05
40 150715-KD-9-B-28Si-700C 712 18 0.67 0.0001 5 0.04
41 150721-KD-5-P-28Si-700C 712 37 0.74 0.0093 1.4× 1018 4 1.22 256 2.60 SIMS
Table D.7: 28Si Sample Catalog: DC–3: II (8/28/14–7/27/15). 28Si samples were deposited using SiH4 with the low pressure mode
of the gas-mode ion source at DC–3, the deposition chamber. Substrates were flash annealed in situ. Aperture gives the width of the
mass-selecting aperture slit. Pressures are raw reading of the deposition chamber base and with the ion source on. VA is the anode
voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii is the average ion current. T is the substrate
deposition temperature. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and rates
correspond to the thickest measured film area. Items marked with * are estimates. Samples #35 to #41 contain roughly 30 % N.
434
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P VA VT
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) (V) (kV)
42 150920-KD-3-P-28Si-308C 9/20/15 10/2/15 Si(100):P (6) CMOS clean 2 1.7× 10−8 1.1× 10−6 45 -4.00
43 150926-KD-4-P-28Si-312C 9/26/15 10/5/15 Si(100):P (6) CMOS clean 2 1.6× 10−8 1.5× 10−6 50 -4.00
44 151020-KD-9-B-28Si-700C 10/20/15 10/22/15 Si(100):B (7) CMOS clean 2 2.3× 10−8 1.3× 10−6 50 -4.00
45 151019-KD-2-B-28Si-450C 10/19/15 10/22/15 Si(100):B (7) CMOS clean 2 2.3× 10−8 1.3× 10−6 50 -4.00
46 151025-KD-4-B-28Si-400C 10/25/15 11/2/15 Si(100):B (7) CMOS clean 2 1.6× 10−8 9.6× 10−7 50 -4.00
47 151028-KD-1-B-28Si-700C 10/28/15 11/20/15 Si(100):B (7) CMOS clean 2 1.5× 10−8 1.3× 10−6 50 -4.00
48 151102-KD-7-i-28Si-400C 11/2/15 11/30/15 Si(100):i (8) CMOS clean 2 1.5× 10−8 1.2× 10−6 50 -4.00
49 151202-KD-7-i-28Si-400C 12/2/15 12/7/15 Si(100):i (8) CMOS clean 2 1.5× 10−8 1.5× 10−6 50 -4.00
50 151109-KD-9-B-28Si-400C 11/9/15 12/10/15 Si(100):B (9) CMOS clean 2 1.7× 10−8 1.3× 10−6 50 -4.00
51 151109-KD-4-B-28Si-300C 11/9/15 12/14/15 Si(100):B (9) CMOS clean 2 1.5× 10−8 1.3× 10−6 50 -4.00
#
Sample T Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (◦C) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement(s)
42 150920-KD-3-P-28Si-308C 355 40 0.53 0.0003 1.14* 9* 0.07
43 150926-KD-4-P-28Si-312C 357 46 0.55 0.0050 1.1× 1018 3 0.33 50 1.36 SIMS, XPS
44 151020-KD-9-B-28Si-700C 712 43 0.54 0.0003 1 1.11* 11* 0.09
45 151019-KD-2-B-28Si-450C 460 43 0.54 0.0097 1.6× 1018 4 0.53 160 2.69 SIMS, TEM
46 151025-KD-4-B-28Si-400C 721 42 0.47* 0.0036 6.2× 1017 4 0.21 44 0.99* SIMS
47 151028-KD-1-B-28Si-700C 705 39 0.55 0.0058 4.6× 1017 8 0.82 144 1.62 SIMS, SEM
48 151102-KD-7-i-28Si-400C 417 40 0.55 0.0079 6.8× 1017 7 0.63 150 2.20 ESR
49 151202-KD-7-i-28Si-400C 417 40 0.54 0.0087 1.1× 1018 5 0.93 250 2.42 ESR
50 151109-KD-9-B-28Si-400C 421 38 0.54 0.0082 5.6× 1017* 9* 0.60 151 2.29 SIMS
51 151109-KD-4-B-28Si-300C 349 37 0.51 0.0074 8.5× 1017 5 0.87 210 2.05 SIMS
Table D.8: 28Si Sample Catalog: DC–3: III (9/20/15–11/9/15). 28Si samples were deposited using SiH4 with the low pressure mode
of the gas-mode ion source at DC–3, the deposition chamber. Substrates were flash annealed in situ. Aperture gives the width of the
mass-selecting aperture slit. Pressures are raw reading of the deposition chamber base and with the ion source on. VA is the anode
voltage and VT is the transport voltage. Ei is the average ion energy at the sample, and Ii is the average ion current. T is the substrate
deposition temperature. C. I. is the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and rates
correspond to the thickest measured film area. Items marked with * are estimates.
435
#
Sample Load Dep. Substrate Substrate Aperture Base P Dep. P Plasma VA
Name Date Date Type Prep. (ex situ) (mm) (Pa) (Pa) Mode (V)
52 151210-KD-7-i-28Si-500C 12/10/15 12/18/15 Si(100):i (8) CMOS clean 2 1.5× 10−8 1.2× 10−6 low P 50
53 151218-KD-4-i-28Si-200C 12/18/15 12/21/15 Si(100):i (8) CMOS clean 2 1.7× 10−8 1.2× 10−6 low P 57
54 151218-KD-9-B-28Si-400C 12/18/15 12/29/15 Si(100):B (9) CMOS clean 2 1.7× 10−8 1.2× 10−6 low P 69
55 160113-KD-4-P-28Si-400C 1/13/16 2/8/16 Si(100):P (10) CMOS clean 2 2.4× 10−8 5.3× 10−6 high P 55
56 160202-KD-7-P-28Si-400C 2/2/16 2/10/16 Si(100):P (10) CMOS clean 2 2.1× 10−8 5.2× 10−6 high P 55
57 160210-KD-1-P-28Si-400C 2/10/16 2/12/16 Si(100):P (10) CMOS clean 2 2.1× 10−8 5.1× 10−6 high P 55
58 160217-KD-7-P-28Si-450C 2/17/16 2/29/16 Si(100):P (10) CMOS clean 2 1.7× 10−8 5.5× 10−6 high P 55
59 160513-KD-7-i-28Si-450C 5/13/16 6/9/16 Si(100):i (8) CMOS clean 2 8.3× 10−9 3.6× 10−6 high P 52
60 160617-KD-3-P-28Si-400C 6/17/16 6/24/16 Si(100):P (10) CMOS clean 2 7.1× 10−9 3.4× 10−6 high P 52
61 160617-KD-7-P-28Si-403C 6/17/16 6/24/16 Si(100):P (10) CMOS clean 2 7.1× 10−9 2.9× 10−6 high P 50
#
Sample VT T Ei Ii C. I. Ion Dose Spot Size R d Film Mass Primary
Name (kV) (◦C) (eV) (µA) (C) (cm−2) (mm2) (nm/min) (nm) (µg) Measurement(s)
52 151210-KD-7-i-28Si-500C -4.00 502 37 0.52 0.0071 7.4× 1017 6 0.95 321 1.99 SIMS
53 151218-KD-4-i-28Si-200C -4.00 249 38 0.63 0.0091 6.6× 1017 9 1.27 305 2.54 SIMS
54 151218-KD-9-B-28Si-400C -4.00 421 40 0.69 0.0110 1.0× 1018 7 0.87 232 3.06 SIMS
55 160113-KD-4-P-28Si-400C -4.70 421 35 2.29 0.0178 1.5× 1018 7 2.48* 322* 4.99
56 160202-KD-7-P-28Si-400C -4.70 421 35 2.27 0.0190 1.4× 1018 9 2.57 359* 5.33
57 160210-KD-1-P-28Si-400C -4.70 421 36 2.28 0.0178 1.3× 1018 9 2.65* 345* 4.98
58 160217-KD-7-P-28Si-450C -4.70 460 37 2.15 0.0006 2.60* 13* 0.18
59 160513-KD-7-i-28Si-450C -4.70 460 34 2.77 0.0220 1.4× 1018 10 2.21 292 6.18 SIMS
60 160617-KD-3-P-28Si-400C -4.70 421 31 3.00 0.0144 2.4× 1018 4 4.56 365 4.07 SIMS, TEM
61 160617-KD-7-P-28Si-403C -4.70 423 32 2.50 0.0150 6.2× 1018 2 2.33* 233* 4.23
Table D.9: 28Si Sample Catalog: DC–3: IV (12/18/15–6/24/16). 28Si samples were deposited using SiH4 with the gas-mode ion source
at DC–3, the deposition chamber. Substrates were flash annealed in situ. Aperture gives the width of the mass-selecting aperture slit.
Pressures are raw reading of the deposition chamber base and with the ion source on. VA is the anode voltage and VT is the transport
voltage. Ei is the average ion energy at the sample, and Ii is the average ion current. T is the substrate deposition temperature. C. I. is
the ion current integral. R is the deposition rate, and d is the film thickness. Thicknesses and rates correspond to the thickest measured
film area. Items marked with * are estimates.
436
Sample Data Dep. Ei Ii R PSiHx Fi Fg
Name Set Location (eV) (µA) (nm/min) (Pa) (cm−2 · s−1) (cm−2 · s−1)
120604-28Si-OxSi 1 IC–1 455 0.95 0.07 2.4 ×10−5 1.1 ×1013 6.6 ×1013
120618-28Si-OxSi
Ar
IC–1 64 0.77
0.40
3.4 ×10−5 3.5 ×10
13
9.2 ×1013
SF5 0.33 2.8 ×1013
120628-28Si-OxSi
4
IC–1 64 0.85
0.5
2.6 ×10−5
4.4 ×1013
7.2 ×10135 0.5 4.4 ×10
13
7 0.47 4.2 ×1013
8 0.59 5.2 ×1013
130204-28Si-Si
1
LC–2 74 0.57
0.51
1.4 ×10−6 4.6 ×10
13
3.7 ×1012
2 0.56 5.1 ×1013
130208-28Si-Si
1
LC–2 121 0.56
1.49
1.4 ×10−6
1.5 ×1014
3.7 ×1012
2 1.05 1.1 ×1014
3 0.54 5.6 ×1013
4 0.53 5.4 ×1013
5 0.94 9.6 ×1013
130328-28Si(dP)-Si low-P LC–2 60 0.33 0.93 6.8 ×10−7 8.3 ×1013 1.8 ×1012
130412-2829Si(28SiH)-Si dip/film LC–2 50 0.31 0.84 6.8 ×10−7 7.3 ×1013 1.8 ×1012
130520-28Si-SOI 1 LC–2 180 0.48 1.07 8.3 ×10−7 1.2 ×1014 2.3 ×1012
140828-28Si-Si 1 DC–3 38 0.25 0.32 2.8 ×10−7 2.7 ×1013 7.6 ×1011
Table D.10: Deposition parameters of 28Si samples deposited at room temperature (≈ 21 ◦C) for the gas sticking deposition model
analysis discussed in Chapter 6. Samples deposited at IC–1 were excluded in that analysis. Some samples have multiple SIMS data sets
that are grouped together. Ei is the average ion energy at the sample. Ii is the average
28Si ion beam current. R is the deposition rate.
PSiHx is the SiHx partial pressure at the sample during deposition. Fi is the
28Si ion flux. Fg is the SiHx molecular flux at the sample.
437
Isotope Fraction (zSi/Sitot.) cz(sT , k502)
Sample Data Dep. d 28Si 29Si 30Si 29Si 30Si
Name Set Location (nm) (%) (ppm) (ppm) (ppm) (ppm)
120604-28Si-OxSi 1 IC–1 12 99.702(12) 2822(18) 156.1(38) 0.112
120618-28Si-OxSi
Ar
IC–1
92 99.8850(14) 1130(14) 20.3(14) 0.032
SF5 75 99.8448(21) 1534(21) 17.9(19) 0.0228
120628-28Si-OxSi
4
IC–1
73 99.99846(19) 9.5(10) 5.9(16) 0.043 0.027
5 73 99.99842(16) 10.5(13) 5.23(87) 0.047 0.024
7 69 99.99816(11) 12.3(10) 6.15(52) 0.052 0.026
8 86 99.99799(16) 12.8(10) 7.3(13) 0.068 0.039
130204-28Si-Si
1
LC–2
100 99.999657(38) 2.02(32) 1.41(20) 0.181 0.126
2 110 99.999592(36) 2.30(26) 1.78(24) 0.227 0.176
130208-28Si-Si
1
LC–2
384 99.9998308(82) 0.993(64) 0.699(51) 0.299 0.210
2 270 99.9998031(86) 1.252(72) 0.717(48) 0.265 0.152
3 140 99.999722(12) 1.724(97) 1.055(74) 0.189 0.116
4 136 99.999696(15) 1.85(11) 1.189(96) 0.197 0.127
5 241 99.999812(10) 1.122(76) 0.760(64) 0.212 0.144
130328-28Si(dP)-Si low-P LC–2 249 99.999888(10) 0.691(74) 0.432(67) 0.225 0.141
130412-2829Si(28SiH)-Si dip/film LC–2 351 99.999889(11) 0.78(10) 0.512(42) 0.223 0.147
130520-28Si-SOI 1 LC–2 149 99.999863(16) 0.77(11) 0.60(11) 0.289 0.225
140828-28Si-Si 1 DC–3 53 99.99990(11) 0.58(26) 0.44(23) 0.148 0.113
Table D.11: Enrichment measurements and gas sticking deposition model model analysis results for 28Si samples deposited at room
temperature (≈ 21 ◦C) discussed in Chapter 6. Some samples have multiple SIMS data sets that are grouped together. d is the 28Si film
thickness. The raw SIMS isotope fractions as well as those adjusted to the deposition conditions of the 502 ◦C sample, cz(sT , k502), are
listed.
438
Sample Data T Ei Ii R PSiHx Fi Fg
Name Set (◦C) (eV) (µA) (nm/min) (Pa) (cm−2 · s−1) (cm−2 · s−1)
151218-KD-4-i-28Si-200C
1
249 38 0.63
1.27
3.2 ×10−7 1.1 ×10
14
8.7 ×1011
2 1.11 9.6 ×1013
151109-KD-4-B-28Si-300C
3
349 37 0.51
0.80
3.6 ×10−7 6.9 ×10
13
9.8 ×1011
4 0.87 7.5 ×1013
150926-KD-4-P-28Si-312C 1 357 46 0.55 0.33 4.1 ×10−7 2.9 ×1013 1.1 ×1012
151109-KD-9-B-28Si-400C
1
421 38 0.54
0.43
3.7 ×10−7
3.7 ×1013
9.9 ×10114 0.36 3.1 ×1013
6 0.60 5.1 ×1013
160617-KD-3-P-28Si-400C
1
421 31 3.00
3.94
9.6 ×10−7 3.4 ×10
14
2.6 ×1012
2 3.36 2.9 ×1014
151210-KD-7-i-28Si-500C
2
502 37 0.52
0.95
3.3 ×10−7
8.1 ×1013
8.8 ×10113 0.43 3.7 ×1013
4 1.41 1.2 ×1014
140604-28Si-Si-600C 1 610 33 0.50 0.77 1.4 ×10−7 6.6 ×1013 3.7 ×1011
151028-KD-1-B-28Si-700C
1
705 39 0.55
0.82
3.5 ×10−7 7.0 ×10
13
9.5 ×1011
2 0.66 5.7 ×1013
140603-28Si-Si-700C
1
708 33 0.51
1.20
2.2 ×10−7 1.0 ×10
14
6.0 ×1011
2 0.98 8.3 ×1013
140521-28Si-Si-800C 1 812 33 0.42 0.90 1.5 ×10−7 7.7 ×1013 4.1 ×1011
Table D.12: Deposition parameters of 28Si samples deposited at elevated temperature for the gas sticking deposition model analysis
discussed in Chapter 6. Some samples have multiple SIMS data sets that are grouped together. T is the substrate deposition temperature.
Ei is the average ion energy at the sample. Ii is the average
28Si ion beam current. R is the deposition rate. PSiHx is the SiHx partial
pressure at the sample during deposition. Fi is the
28Si ion flux. Fg is the SiHx molecular flux at the sample.
439
Isotope Fraction (zSi/Sitot.) cz(sT , k502)
Sample Data T d 28Si 29Si 30Si 29Si 30Si s
Name Set (◦C) (nm) (%) (ppm) (ppm) (ppm) (ppm)
151218-KD-4-i-28Si-200C
1
249
305 99.999898(13) 0.79(12) 0.229(64) 0.72 0.208
1.6 ×10−3
2 267 99.999865(18) 1.01(17) 0.341(65) 0.80 0.271
151109-KD-4-B-28Si-300C
3
349
193 99.999884(18) 0.78(16) 0.386(88) 0.40 0.197
8 ×10−4
4 210 99.999924(17) 0.53(16) 0.230(46) 0.29 0.128
150926-KD-4-P-28Si-312C 1 357 50 99.999405(93) 4.18(70) 1.77(61) 0.80 0.34 2.0 ×10−3
151109-KD-9-B-28Si-400C
1
421
108 99.999762(27) 1.68(24) 0.70(14) 0.452 0.188
4 92 99.999812(25) 1.30(22) 0.58(12) 0.298 0.133
6 151 99.999818(26) 1.46(22) 0.35(15) 0.55 0.132
8.4 ×10−4
160617-KD-3-P-28Si-400C
1
421
315 99.9999594(72) 0.303(58) 0.103(43) 0.285 0.097
2 269 99.9999446(67) 0.407(59) 0.058(32) 0.327 0.068
151210-KD-7-i-28Si-500C
2
502
216 99.9999701(57) 0.259(53) 0.040(19) 0.174 0.027
2.9 ×10−43 99 99.999940(16) 0.31(10) 0.29(12) 0.096 0.089
4 321 99.9999819(35) 0.127(29) 0.055(19) 0.127 0.055
140604-28Si-Si-600C 1 610 162 99.9999570(70) 0.300(60) 0.130(37) 0.38 0.166 9 ×10−4
151028-KD-1-B-28Si-700C
1
705
144 99.999488(48) 3.30(25) 1.82(42) 1.77 0.98
5.6 ×10−3
2 117 99.999181(83) 4.82(70) 3.38(45) 2.11 1.48
140603-28Si-Si-700C
1
708
126 99.99986(11) 0.806(97) 0.61(37) 1.00 0.76
7 ×10−3
2 102 99.99947(18) 3.09(61) 2.17(42) 3.12 2.19
140521-28Si-Si-800C 1 812 158 99.99907(25) 4.32(46) 4.96(93) 5.9 6.8 2.3 ×10−2
Table D.13: Enrichment measurements and temperature dependant gas incorporation model analysis results for 28Si samples deposited
at elevated temperature discussed in Chapter 6. Some samples have multiple SIMS data sets that are grouped together. T is the substrate
deposition temperature. d is the 28Si film thickness. The raw SIMS isotope fractions as well as those adjusted to the deposition conditions
of the 502 ◦C sample, cz(sT , k502), are listed. s is the incorporation fraction at each deposition temperature.
440
Appendix E: SIMS Measurement Settings
22Ne:
The following is a statement of the SIMS measurement settings and tech-
niques used for assessing the enrichment of 22Ne samples implanted at LC–2. This
statement was adapted from one provided by Dr. Dave Simons (NIST) for sample
110504-22Ne-OxSi.
A Cameca IMS-1280 magnetic sector secondary ion mass spectrometer was
used to make the measurements. The primary ion beam was composed of Cs+ ions
with a current of 40 nA and an energy with respect to ground of 10 keV. The beam
was raster-scanned over a nominal 280 µm by 430 µm in size, and the ions were
only accepted for counting from a central region that represented about 40 % of
the area of the crater. At implantation doses used for these samples, it is expected
that the peak concentration of implanted species would occur at the surface and
would be constant roughly up to the range of the implanted ions. The maximum
concentration is governed by a balance between implantation and sputter removal
rates.
The most sensitive method to detect noble gases by SIMS is via Cs attach-
ment ions CsM+, where M is the noble gas isotope. Thus positive secondary ions
441
of 133Cs20Ne+, 133Cs22Ne+, and 133Cs29Si+ were accelerated to 5 keV, separated
by mass, and detected with a secondary electron multiplier operating in a pulse-
counting mode. These species were detected sequentially and repetitively over 50
cycles, with the Ne species being detected for 5 s each per cycle and the Si for
1 s. The sequence of data constitutes a depth profile since the sample surface is
eroded with time by ion sputtering. Depth profile data were taken both in the area
designated as having been implanted, as well as far from that area to serve as a
control.
12C:
The following is a statement of the SIMS measurement settings and techniques
used for assessing the enrichment of 12C samples deposited at LC–2. This statement
was adapted from one provided by Dr. David Simons (NIST) for sample 120207-
12C-OxSi.
The isotopic composition of deposited carbon films were measured with a
Cameca IMS-1270 large geometry secondary ion mass spectrometer. A Cs+ primary
ion beam with an impact energy of 20 keV was raster-scanned over a 150 µm by
150 µm area on the sample in the center of the deposited region. Secondary negative
ion signals of 12C and 13C were extracted from a gated area of 125 µm by 125 µm
into the mass spectrometer and alternately directed by magnetic peak-switching
into a secondary electron multiplier where their count rates were recorded. A mass
resolving power m
∆m
of 5000 at 10 % of peak amplitude was set by a combination of
entrance and exit slits so that a spectral interference of 12CH on 13C was effectively
442
excluded. A set of measurements were made by pre-sputtering the area for 90 s or
360 s with an 8 nA primary ion current and then recording 40 isotopic ratio pairs
with a beam current of 23 pA. The count rates were corrected for the dead time of
the electron multiplier.
The isotope profiles are the results of 8 successive sets of measurements of
13C/12C ratios from the same area of the carbon film. These values do not take
into account an instrumental mass fractionation for carbon that would make mea-
sured ratios smaller by about 5 %. A stylus profilometry is used to measure the
sputtered crater depth. To assess the degree of enrichment of 12C in the films, a
ratio measurement of a pyrolytic graphite disk with isotopically natural carbon was
made under similar analysis conditions. In this case the measured 13C/12C ratio was
1.046× 10−2.
28Si at IC–1:
The following is a statement of the SIMS measurement settings and techniques
used for assessing the enrichment of 28Si samples deposited at IC–1. This statement
was adapted from one provided by Dr. Dave Simons (NIST) for sample 120604-
28Si-OxSi.
The mass spectrometer was set to exclude 28SiH from the 29Si signal and was
used to first measure isotopic ratios in the Si substrate far from the film deposit.
The primary beam species was O− with an impact energy of 23 keV and a current
of about 0.4 nA. The analyzed area was 25 µm by 25 µm, defined by an aperture
in the ion optics. The beam was not rastered and the spot had an elliptical shape
443
about 50 µm on the major axis. The measured ratio of 29Si/28Si was 0.0497 and the
measured ratio of 30Si/28Si was 0.0323 in the Si substrate. These values are smaller
than the known ratios of 0.0508 for 29Si/28Si and 0.03353 for 30Si/28Si but that result
is expected since SIMS is known to ionize the lighter isotope with higher efficiency.
These measured values are used to calculate the degree to which the minor isotopes
have been reduced in the deposited film.
The film area was then analyzed in a region that appeared to be the center,
assumed to have the greatest thickness. 20 sequential analytical runs were made
on the same area, sputtering in total for nearly 4 h. The count rates of 28Si, 29Si
and 30Si were recorded sequentially. The count rates were corrected to account for
detector dead time. The crater depth was later measured with a stylus profilometer.
28Si at LC–2:
The following is a statement of the SIMS measurement settings and techniques
used for assessing the enrichment of 28Si samples deposited at LC–2. This statement
was adapted from one provided by Dr. Dave Simons (NIST) for sample 130208-28Si-
Si.
Isotopic measurements were made in a CAMECA IMS-1280 large geometry
secondary ion mass spectrometer. The sample was bombarded with a primary ion
beam of O+2 at an impact energy of 8 keV and a current of 0.3 nA. The beam was
focused to a probe size of a few micrometers diameter and it was raster-scanned over
a 50 µm by 50 µm area. Positive secondary ions were accepted for detection from the
central 25 µm by 25 µm portion of the rastered area. The secondary ions passed
444
through an entrance slit of 20 µm and an exit slit of 200 µm. These conditions
produce a mass resolving power of about 6000 ( m
∆m
at 10 % of peak maximum).
This resolving power is necessary to separate the 29Si peak from the 28SiH peak that
is produced during the SIMS process. The 28SiH signal is normally about 0.05 % to
0.1 % of the 28Si signal and the ratio can vary depending on the analytical conditions.
For one sample the 28SiH signal within the film was about 4000 times larger than the
29Si signal demonstrating why very good separation between these peaks is needed
to measure the 29Si/28Si ratio accurately.
Data were taken over 100 to 360 cycles in which the 28Si was measured for 1 s,
the 29Si for 10 s, 28SiH for 1 s and the 30Si for 10 s sequentially during each cycle by
switching the magnetic field. The runs took between 50 min and 100 min to sputter
through the film depending on the thickness of the film at the analysis location.
Quantitative isotopic ratio calculations were made by averaging the cycle-by-
cycle ratio measurements in the part of the profiles where the ratios were at a
minimum. These values were then corrected for instrumental mass fractionation
based on the differences between the measured ratios of the wafer silicon and the
accepted natural values. Uncertainties were determined from the standard deviation
of the mean of the measurements and were compared with a Poisson estimation
based on the total number of detected counts of the minor isotopes. The depths
of each crater were measured with a stylus profilometer and the film thickness was
defined as the depth where the 29Si/28Si ratio had risen to half of its natural value.
Carbon isotopic signals were monitored in a separate depth profile of the film.
30Si was also monitored as a marker of the film-substrate interface and as a normal-
445
ization signal for concentration estimates. The concentrations of the carbon isotopes
in the film were estimated by averaging their count rates over the cycles within the
film and applying a value obtained from literature of 0.007 for the relative sensitivity
factor of carbon to silicon by SIMS under similar analytical conditions. The SIMS
instrumental background for carbon analyzed under similar conditions is unknown
but it must be at least an order of magnitude lower than what is measured since the
carbon signals typically decrease by more than a factor of 10 in the Si substrate.
Trace carbon is not normally analyzed by SIMS under these conditions of oxygen ion
bombardment and positive ion detection but rather with cesium ion bombardment
and negative ion detection. In the present case the conditions were not changed to
simplify the measurement process.
28Si at DC–3:
The following is a statement of the SIMS measurement settings and techniques
used for assessing the enrichment of 28Si samples deposited at DC–3. This statement
was adapted from one provided by Dr. Dave Simons (NIST) as a general procedure
for measuring samples deposited at elevated temperatures.
Isotopic measurements of Si were made in a CAMECA IMS-1270E7 large
geometry secondary ion mass spectrometer. The samples were bombarded with a
primary ion beam of O+2 ions at an impact energy of 8 keV and a current of 1 nA.
The beam was focused to a probe size of a few micrometers diameter and it was
raster-scanned over a 50 µm by 50 µm area. Positive secondary ions were accepted
for detection from the central 20 µm by 20 µm portion of the rastered area as
446
defined by a field aperture in a focal plane of the mass spectrometer. The entrance
and exit slits of the spectrometer were selected to produce a mass resolving power
of about 6000 ( m
∆m
at 10 % of peak maximum). This resolving power is necessary
to separate cleanly the 29Si peak from the 28SiH peak that is produced during the
SIMS process. Under these conditions we estimate that less than 10−5 of the 28SiH
signal contributes to the 29Si measurement.
Depth profiles of the Si isotopes 28Si, 29Si and 30Si through deposited films
were acquired by monitoring 28Si for 1 s, 29Si for 10 s, 28SiH for 1 s and 30Si for 10 s
in each data cycle and collecting a sufficient number of data cycles until the profile
penetrated into the silicon substrate. The sputter rate as determined by measuring
the final crater depths with a stylus profilometer was approximately 0.15 nm/s under
these conditions.
Isotope ratios of 29Si/28Si and 30Si/28Si were calculated on a cycle-by-cycle ba-
sis. Average isotopic ratios for a film were calculated by averaging the cycle-by-cycle
ratio measurements in the portion of a profile where the ratios were at a relatively
constant minimum value. These values were then corrected for instrumental mass
fractionation based on the differences between the measured ratios from a Si wafer
and the accepted natural values. Uncertainties were determined from the standard
deviation of the mean of the measurements and were usually similar to Poisson
estimations based on the total number of detected counts of the minor isotopes.
447
Appendix F: 28Si Deposition Fun Facts
(1) the total mass of 28Si deposited in this work: ≈ 112 µg
(2) the total area deposited for 28Si samples: ≈ 460 mm2
(3) the total number of 28Si atoms deposited in this work: ≈ 2.6× 1018
(4) the total time depositing 28Si: ≈ 169 h ≈ 7 days
(5) the resulting film thickness if all 28Si samples were deposited at once with a
spot size of 7.5 mm2: ≈ 6.5 µm
(6) the average distance between 28Si ions along the beamline with an energy of
4050 eV for 0.5 µA of ion current: ≈ 54 nm
(7) the mean free path of SiH4 molecules at 21
◦C for a pressure of 1.3× 10−4 Pa
(1.0× 10−6 Torr): λ ≈ 200 m
448
Bibliography
[1] Jonathan Wood. “The top ten advances in materials science”. Mater. Today,
11(1-2):40–45, 2008.
[2] Floris A. Zwanenburg, Andrew S. Dzurak, Andrea Morello, Michelle Y. Sim-
mons, Lloyd C. L. Hollenberg, Gerhard Klimeck, Sven Rogge, Susan N. Cop-
persmith, and Mark A. Eriksson. “Silicon quantum electronics”. Rev. Mod.
Phys., 85(3):961–1019, 2013.
[3] B. E. Kane. “A silicon-based nuclear spin quantum computer”. Nature,
393(6681):133–137, 1998.
[4] VeldhorstM, J. C. C. Hwang, C. H. Yang, A. W. Leenstra, B. de Ronde,
J. P. Dehollain, J. T. Muhonen, F. E. Hudson, K. M. Itoh, MorelloA, and
A. S. Dzurak. “An addressable quantum dot qubit with fault-tolerant control-
fidelity”. Nat. Nano., 9(12):981–985, 2014.
[5] M. Fuechsle, S. Mahapatra, F. A. Zwanenburg, M. Friesen, M. A. Eriksson,
and M. Y. Simmons. “Spectroscopy of few-electron single-crystal silicon quan-
tum dots”. Nat. Nanotechnol., 5(7):502–505, 2010.
[6] M. Veldhorst, C. H. Yang, J. C. C. Hwang, W. Huang, J. P. Dehollain, J. T.
Muhonen, S. Simmons, A. Laucht, F. E. Hudson, K. M. Itoh, A. Morello, and
A. S. Dzurak. “A two-qubit logic gate in silicon”. Nature, 526(7573):410–414,
2015.
[7] Keiichi Edamatsu. “Quantum physics: swift control of a single spin”. Nature,
456(7219):182–183, 2008.
[8] John P. Ridgway. “Cardiovascular magnetic resonance physics for clinicians:
part I”. J. Cardiovasc. Magn. Reson., 12(1):71, 2010.
[9] M. Steger, K. Saeedi, M. L. W. Thewalt, J. J. L. Morton, H. Riemann, N. V.
Abrosimov, P. Becker, and H. J. Pohl. “Quantum information storage for
449
over 180 s using donor spins in a Si-28 ‘Semiconductor Vacuum’”. Science,
336(6086):1280–1283, 2012.
[10] Juha T. Muhonen, Juan P. Dehollain, Arne Laucht, Fay E. Hudson, Rach-
pon Kalra, Takeharu Sekiguchi, Kohei M. Itoh, David N. Jamieson, Jeffrey C.
McCallum, Andrew S. Dzurak, and Andrea Morello. “Storing quantum infor-
mation for 30 seconds in a nanoelectronic device”. Nat. Nano., 9(12):986–991,
2014.
[11] Eisuke Abe, Alexei M. Tyryshkin, Shinichi Tojo, John J. L. Morton, Wayne M.
Witzel, Akira Fujimoto, Joel W. Ager, Eugene E. Haller, Junichi Isoya,
Stephen A. Lyon, Mike L. W. Thewalt, and Kohei M. Itoh. “Electron spin
coherence of phosphorus donors in silicon: Effect of environmental nuclei”.
Phys. Rev. B, 82(12), 2010.
[12] Wayne M. Witzel, Malcolm S. Carroll,  Lukasz Cywin´ski, and S. Das Sarma.
“Quantum decoherence of the central spin in a sparse system of dipolar coupled
spins”. Phys. Rev. B, 86(3):35452, 2012.
[13] Alexei M. Tyryshkin, Shinichi Tojo, John J. L. Morton, Helge Riemann,
Nikolai V. Abrosimov, Peter Becker, Hans-Joachim Pohl, Thomas Schenkel,
Michael L. W. Thewalt, Kohei M. Itoh, and S. A. Lyon. “Electron spin coher-
ence exceeding seconds in high-purity silicon”. Nat. Mater., 11(2):143–147,
2012.
[14] Kamyar Saeedi, Stephanie Simmons, Jeff Z. Salvail, Phillip Dluhy, Helge
Riemann, Nikolai V. Abrosimov, Peter Becker, Hans-Joachim Pohl, John
J. L. Morton, and Mike L. W. Thewalt. “Room-temperature quantum bit
storage exceeding 39 minutes using ionized donors in silicon-28”. Science,
342(6160):830–833, 2013.
[15] S. Meiboom and D. Gill. “Modified spin-echo method for measuring nuclear
relaxation times”. Rev. Sci. Instrum., 29(8):688–691, 1958.
[16] Terry Gullion, David B. Baker, and Mark S. Conradi. “New, compensated
Carr-Purcell sequences”. J. Magn. Reson., 89(3):479–484, 1990.
[17] B. Weber, S. Mahapatra, T. F. Watson, and M. Y. Simmons. “Engineering in-
dependent electrostatic control of atomic-scale (similar to 4 nm) silicon double
quantum dots”. Nano Letters, 12(8):4001–4006, 2012.
[18] B. Weber, S. Mahapatra, H. Ryu, S. Lee, A. Fuhrer, T. C. G. Reusch, D. L.
Thompson, W. C. T. Lee, G. Klimeck, L. C. L. Hollenberg, and M. Y. Sim-
mons. “Ohm’s law survives to the atomic scale”. Science, 335(6064):64–67,
2012.
[19] H. Buch, S. Mahapatra, R. Rahman, A. Morello, and M. Y. Simmons. “Spin
readout and addressability of phosphorus-donor clusters in silicon”. Nature
Commun., 4:1038, 2013.
450
[20] J. A. Miwa, J. A. Mol, J. Salfi, S. Rogge, and M. Y. Simmons. “Transport
through a single donor in p-type silicon”. Appl. Phys. Lett., 103(4):043106,
2013.
[21] J. J. Pla, K. Y. Tan, J. P. Dehollain, W. H. Lim, J. J. L. Morton, F. A.
Zwanenburg, D. N. Jamieson, A. S. Dzurak, and A. Morello. “High-fidelity
readout and control of a nuclear spin qubit in silicon”. Nature, 496(7445):334–
338, 2013.
[22] C. H. Yang, A. Rossi, R. Ruskov, N. S. Lai, F. A. Mohiyaddin, S. Lee,
C. Tahan, G. Klimeck, A. Morello, and A. S. Dzurak. “Spin-valley lifetimes
in a silicon quantum dot with tunable valley splitting”. Nature Commun.,
4:3069, 2013.
[23] L. A. Tracy, T. M. Lu, N. C. Bishop, G. A. Ten Eyck, T. Pluym, J. R. Wendt,
M. P. Lilly, and M. S. Carroll. “Electron spin lifetime of a single antimony
donor in silicon”. Appl. Phys. Lett., 103(14):143115, 2013.
[24] A. Wild, J. Kierig, J. Sailer, J. W. Ager, E. E. Haller, G. Abstreiter, S. Ludwig,
and D. Bougeard. “Few electron double quantum dot in an isotopically purified
Si-28 quantum well”. Appl. Phys. Lett., 100(14):143110, 2012.
[25] J. Y. Li, C. T. Huang, L. P. Rokhinson, and J. C. Sturm. “Extremely high
electron mobility in isotopically-enriched Si-28 two-dimensional electron gases
grown by chemical vapor deposition”. Appl. Phys. Lett., 103(16):162105, 2013.
[26] C. C. Lo, A. Persaud, S. Dhuey, D. Olynick, F. Borondics, M. C. Martin,
H. A. Bechtel, J. Bokor, and T. Schenkel. “Device fabrication and transport
measurements of FinFETs built with Si-28 SOI wafers toward donor qubits in
silicon”. Semicond. Sci. Technol., 24(10):105022, 2009.
[27] John J. L. Morton, Alexei M. Tyryshkin, Richard M. Brown, Shyam Shankar,
Brendon W. Lovett, Arzhang Ardavan, Thomas Schenkel, Eugene E. Haller,
Joel W. Ager, and S. A. Lyon. “Solid-state quantum memory using the (31)P
nuclear spin”. Nature, 455(7216):1085–1088, 2008.
[28] L. A. Tracy, D. R. Luhman, S. M. Carr, N. C. Bishop, G. A. Ten Eyck,
T. Pluym, J. R. Wendt, M. P. Lilly, and M. S. Carroll. “Single shot spin
readout using a cryogenic high-electron-mobility transistor amplifier at sub-
Kelvin temperatures”. Appl. Phys. Lett., 108(6):63101, 2016.
[29] A. Yang, H. J. Lian, M. L. W. Thewalt, M. Uemura, A. Sagara, K. M. Itoh,
E. E. Haller, J. W. Ager, and S. A. Lyon. “Isotopic mass dependence of
the lattice parameter in silicon determined by measurement of strain-induced
splitting of impurity bound exciton transitions”. Phys. B-Condensed Matter,
376:54–56, 2006.
451
[30] O. E. Dial, M. D. Shulman, S. P. Harvey, H. Bluhm, V. Umansky, and A. Ya-
coby. “Charge Noise Spectroscopy Using Coherent Exchange Oscillations in a
Singlet-Triplet Qubit”. Phys. Rev. Lett., 110(14):146804, 2013.
[31] M. L. W. Thewalt, A. Yang, M. Steger, D. Karaiskaj, M. Cardona, H. Rie-
mann, N. V. Abrosimov, A. V. Gusev, A. D. Bulanov, I. D. Kovalev, A. K.
Kaliteevskii, O. N. Godisov, P. Becker, H. J. Pohl, E. E. Haller, J. W. Ager,
and K. M. Itoh. “Direct observation of the donor nuclear spin in a near-
gap bound exciton transition: 31P in highly enriched 28Si”. J. Appl. Phys.,
101(8):081724, 2007.
[32] P. Becker, H. J. Pohl, H. Riemann, and N. Abrosimov. “Enrichment of sil-
icon for a better kilogram”. Phys. Status Solidi a-Applications Mater. Sci.,
207(1):49–66, 2010.
[33] T. Ruf, R.W. Henn, M. Asen-Palmer, E. Gmelin, M. Cardona, H.-J. Pohl,
G.G. Devyatych, and P.G. Sennikov. “Thermal conductivity of isotopically
enriched silicon”. Solid State Commun., 115(5):243–247, 2000.
[34] P. Becker, H. Friedrich, K. Fujii, W. Giardini, G. Mana, A. Picard, H. J.
Pohl, H. Riemann, and S Valkiers. “The Avogadro constant determination
via enriched silicon-28”. Meas. Sci. Technol., 20(9), 2009.
[35] P. Becker and D. Schiel. “The Avogadro constant and a new definition of the
kilogram”. Int. J. Mass Spectrometry, 349:219–226, 2013.
[36] Shigeto Fukatsu, Tomonori Takahashi, Kohei M. Itoh, Masashi Uematsu,
Akira Fujiwara, Hiroyuki Kageshima, Yasuo Takahashi, Kenji Shiraishi, and
Ulrich Go¨sele. “Effect of the Si/SiO2 interface on self-diffusion of Si in
semiconductor-grade SiO2”. Appl. Phys. Lett., 83(19):3897–3899, 2003.
[37] K. Takyu, K. M. Itoh, K. Oka, N. Saito, and V. I. Ozhogin. “Growth and
characterization of the isotopically enriched Si-28 bulk single crystal”. Jpn. J.
Appl. Phys., 38(12B):L1493–L1495, 1999.
[38] S. A. Lyon, A. M. Tyryshkin, T. Schenkel, J. Bokor, J. Chu, W. Jantsch,
F. Schaffler, J. L. Truitt, S. N. Coppersmith, and M. A. Eriksson. “Electron
spin coherence in Si”. Physica E, 35(2), 2006.
[39] C. C. Lo, M. Urdampilleta, P. Ross, M. F. Gonzalez-Zalba, J. Mansir, S. A.
Lyon, M.L.W. Thewalt, and J. J. L. Morton. “Hybrid opticalelectrical detec-
tion of donor electron spins with bound excitons in silicon”. Nature Mater.,
14(5):490–494, 2015.
[40] Yaron Artzi, Ygal Twig, and Aharon Blank. “Induction-detection electron
spin resonance with spin sensitivity of a few tens of spins”. Appl. Phys. Lett.,
106(8):84104, 2015.
452
[41] A. Bienfait, J. J. Pla, Y. Kubo, M. Stern, X. Zhou, C. C. Lo, C. D. Weis,
T. Schenkel, M. L. W. Thewalt, D. Vion, D. Esteve, B. Julsgaard, K. Mølmer,
J. J. L. Morton, and P. Bertet. “Reaching the quantum limit of sensitivity in
electron spin resonance”. Nature Nanotech., 11(3):253–257, 2015.
[42] Kevin G. Orrman-Rossiter, D. R. G. Mitchell, S. E. Donnelly, C. J. Rossouw,
S. R. Glanvill, P. R. Miller, Amir H. Al-Bayati, J. A. van den Berg, and
D. G. Armour. “Evidence for competing growth phases in ion-beam-deposited
epitaxial silicon films”. Philos. Mag. Lett., 61(6):311–318, 1990.
[43] N. Tsubouchi, A. Chayahara, Y. Mokuno, A. Kinomura, and Y. Horino. “Epi-
taxial growth of pure Si-28 thin films using isotopically purified ion beams”.
Jpn. J. Appl. Phys. Part 2-Letters, 40(12A):L1283–L1285, 2001.
[44] Richard G. Hewlett and Oscar E. Jr. Anderson. The New World 1939/1946:
A History of the United States Atomic Energy Commission. The Pennsylvania
State University Press, University Park, Pa, volume 1 edition, 1962.
[45] Richard B. Fair. “Analysis and design of ion-beam deposition apparatus”. J.
Appl. Phys., 42(8):3176, 1971.
[46] Sol Aisenberg and Ronald Chabot. “Ion-Beam Deposition of Thin Films of
Diamondlike Carbon”. J. Appl. Phys., 42(7):2953, 1971.
[47] S. Shimizu, N. Sasaki, S. Ogata, and O. Tsukakoshi. “Ion beam decelera-
tion characteristics of a high-current, mass-separated, low-energy ion beam
deposition system”. Rev. Sci. Instrum., 67(10):3664–3671, 1996.
[48] N. Herbots, B.R. Appleton, T.S. Noggle, R.A. Zuhr, and S.J. Pennycook.
“Ion-solid interactions during ion beam deposition of 74Ge and 30Si on Si at
very low ion energies (0-200 eV range)”. Nucl. Instrum. Meth. Phys. Res.
Sect. B, 13(1):250–258, 1986.
[49] P. C. Zalm and L. J. Beckers. “Ion-beam epitaxy of silicon on Ge and Si at
temperatures of 400–K”. Appl. Phys. Lett., 41(2):167–169, 1982.
[50] K. Yagi, S. Tamura, and T. Tokuyama. “Germanium and silicon film growth
by low-energy ion-beam deposition”. Jpn. J. Appl. Phys., 16(2):245–251, 1977.
[51] J. W. Rabalais, A. H. Albayati, K. J. Boyd, D. Marton, J. Kulik, Z. Zhang,
and W. K. Chu. “Ion-energy effects in silicon ion-beam epitaxy”. Phys. Rev.
B, 53(16):10781–10792, 1996.
[52] A. H. Albayati, D. Marton, S. S. Todorov, K. J. Boyd, J. W. Rabalais, D. G.
Armour, J. S. Gordon, and G. Duller. “Performance of mass-analyzed, low-
energy, dual-ion beam system for materials research”. Rev. Sci. Instrum.,
65(8):2680–2692, 1994.
453
[53] Y. Horino, N. Tsubouchi, B. Enders, C. Heck, A. Chayahara, A. Kinomura,
and K. Fujii. “Development of a new ion-beam deposition technology for ultra-
high-purity film fabrication”. Phys. Status Solidi a-Applied Res., 160(2):583–
589, 1997.
[54] J. P. Zhao, D. X. Huang, Z. Y. Chen, W. K. Chu, B. Makarenkov, A. J.
Jacobson, B. Bahrim, and J. W. Rabalais. “Amorphous Ge quantum dots
embedded in SiO2 formed by low energy ion implantation”. J. Appl. Phys.,
103(12):124304, 2008.
[55] A. H. Al-Bayati, S. S. Todorov, K. J. Boyd, D. Marton, J. W. Rabalais, and
J. Kulik. “Homoepitaxy and controlled oxidation of silicon at low temperatures
using low-energy ion beams”. J. Vac. Sci. Technol. B Microelectron. Nanom.
Struct., 13(4):1639, 1995.
[56] N. Tsubouchi, A. Chayahara, Y. Mokuno, A. Kinomura, and Y. Horino. “3C-
SiC thin epilayer formation at low temperature using ion beams”. Appl. Surf.
Sci., 212:920–925, 2003.
[57] N. Tsubouchi, B. Enders, A. Chayahara, A. Kinomura, C. Heck, and
Y. Horino. “Optical properties of carbon and carbon nitride films prepared by
mass-separated energetic negative carbon and carbon nitrogen ions”. J. Vac.
Sci. Technol. A, 17(4):2384–2388, 1999.
[58] J. M. Pomeroy, J. Jacobsen, C. C. Hill, B. H. Cooper, and J. P. Sethna. “Ki-
netic Monte Carlo-molecular dynamics investigations of hyperthermal copper
deposition on Cu(111)”. Phys. Rev. B, 66(23):235411–235412, 2002.
[59] J. M. Pomeroy, A. J. Couture, M. V. R. Murty, E. N. Butler, and B. H.
Cooper. “Hyperthermal ion beam system optimized for studying the effects
of kinetic energy on thin-film growth”. Rev. Sci. Instrum., 73(11):3846–3852,
2002.
[60] R. C. Newman. “Defects in silicon”. Rep. Prog. Phys., 45(10):1163–1210,
1982.
[61] Stanley Wolf and Richard N. Tauber. Silicon Processing for the VLSI Era:
Volume 1 - Process Technology. Lattice Press, 2000.
[62] H. Baumann and K. Bethge. “The Frankfurt PIG ion source”. Nucl. Instrum.
Meth. Phys. Res., 189(1):107–110, 1981.
[63] P. Rohwer, H. Baumann, K. Bethge, and W. Schu¨tze. “Ion energy analysis
of a penning ion source using a high resolution mass spectrometer”. Nucl.
Instrum. Meth. Phys. Res., 204(1):245–248, 1982.
[64] P. Rohwer, H. Baumann, W. Schu¨tze, and K. Bethge. “Studies of the center
potential in a penning discharge”. Nucl. Instrum. Meth. Phys. Res., 211(2-
3):543–546, 1983.
454
[65] H. Baumann and K. Bethge. “PIG ion source with end extraction for multiply
charged ions”. Nucl. Instrum. Meth., 122:517–525, 1974.
[66] Z. Nouri, R. Li, R. A. Holt, and S. D. Rosner. “A Penning sputter ion
source with very low energy spread”. Nucl. Instrum. Meth. Phys. Res. Sect.
A, 614(2):174–178, 2010.
[67] Bernhard H. Wolf. Handbook of Ion Sources. CRC Press, 1995.
[68] C. E. S. Phillips. “The action of magnetised electrodes upon electrical dis-
charge phenomena in rarefied gases. Preliminary note”. Proc. R. Soc. Lond.,
64(1):172–176, 1898.
[69] J. P. Biersack and W. Eckstein. “Sputtering studies with the Monte Carlo
program TRIM.SP”. Appl. Phys. A Solids Surfaces, 34(2):73–94, 1984.
[70] V. I. Rudakov, V. V. Ovcharov, and V. P. Prigara. “Influence of optical
properties of the SOI structure on the wafer temperature during rapid thermal
annealing”. Russ. Microelectron., 41(1):15–24, 2012.
[71] J. Nulman, S. Antonio, and W. Blonigan. “Observation of silicon wafer emis-
sivity in rapid thermal processing chambers for pyrometric temperature mon-
itoring”. Appl. Phys. Lett., 56(25):2513, 1990.
[72] C. Y. Kuo and C. Gau. “Vaporsolidsolid growth of crystalline silicon nanowires
using anodic aluminum oxide template”. Thin Solid Films, 519(11):3603–
3607, 2011.
[73] P. C. Maurer, G. Kucsko, C. Latta, L. Jiang, N. Y. Yao, S. D. Bennett,
F. Pastawski, D. Hunger, N. Chisholm, M. Markham, D. J. Twitchen, J. I.
Cirac, and M. D. Lukin. “Room-temperature quantum bit memory exceeding
one second”. Science, 336(6086):1283–1286, 2012.
[74] K. D. Jahnke, B. Naydenov, T. Teraji, S. Koizumi, T. Umeda, J. Isoya, and
F. Jelezko. “Long coherence time of spin qubits in C-12 enriched polycrys-
talline chemical vapor deposition diamond”. Appl. Phys. Lett., 101(1), 2012.
[75] G. Balasubramanian, P. Neumann, D. Twitchen, M. Markham, R. Kolesov,
N. Mizuochi, J. Isoya, J. Achard, J. Beck, J. Tissler, V. Jacques, P. R. Hem-
mer, F. Jelezko, and J. Wrachtrup. “Ultralong spin coherence time in isotopi-
cally engineered diamond”. Nature Mater., 8(5):383–387, 2009.
[76] K. J. Dwyer, J. M. Pomeroy, and D. S. Simons. “99.996 % C-12 films isotopi-
cally enriched and deposited in situ”. Appl. Phys. Lett., 102(25), 2013.
[77] K. Ohno, F. J. Heremans, L. C. Bassett, B. A. Myers, D. M. Toyli, A. C. B.
Jayich, C. J. Palmstrom, and D. D. Awschalom. “Engineering shallow spins
in diamond with nitrogen delta-doping”. Appl. Phys. Lett., 101(8), 2012.
455
[78] T. Teraji, T. Taniguchi, S. Koizumi, K. Watanabe, M. Y. Liao, Y. Koide, and
J. Isoya. “Chemical vapor deposition of C-12 isotopically enriched polycrys-
talline diamond”. Jpn. J. Appl. Phys., 51(9), 2012.
[79] M. L. Markham, J. M. Dodson, G. A. Scarsbrook, D. J. Twitchen, G. Bala-
subramanian, F. Jelezko, and J. Wrachtrup. “CVD diamond for spintronics”.
Diamond Relat. Mater., 20(2):134–139, 2011.
[80] W. M. Lau, I. Bello, X. Feng, L. J. Huang, F. G. Qin, Z. Y. Yao, Z. Z. Ren,
and S. T. Lee. “Direct ion-beam deposition of carbon-films on silicon in the
ion energy-range of 15–500 eV”. J. Appl. Phys., 70(10):5623–5627, 1991.
[81] J. S. Hodges, L. Li, M. Lu, E. H. Chen, M. E. Trusheim, S. Allegri, X. Yao,
O. Gaathon, H. Bakhru, and D. Englund. “Long-lived NV-spin coherence in
high-purity diamond membranes”. New J. Phys., 14, 2012.
[82] F. Kuemmeth, H. O. H. Churchill, P. K. Herring, and C. M. Marcus. “Carbon
nanotubes for coherent spintronics”. Materials Today, 13(3):18–26, 2010.
[83] D. A. Alman and D. N. Ruzic. “Molecular dynamics calculation of car-
bon/hydrocarbon reflection coefficients on a hydrogenated graphite surface”.
J. Nucl. Mater., 313:182–186, 2003.
[84] K. J. Dwyer, J. M. Pomeroy, D. S. Simons, K. L. Steffens, and J. W. Lau.
“Enriching 28Si beyond 99.9998% for semiconductor quantum computing”. J.
Phys. D. Appl. Phys., 47(34):345105, 2014.
[85] J. K. Rath. “Low temperature polycrystalline silicon: a review on deposition,
physical properties and solar cell applications”. Sol. Energy Mater. Sol. Cells,
76(4):431–487, 2003.
[86] Siegfried Hofmann. “Atomic mixing, surface roughness and information depth
in high-resolution AES depth profiling of a GaAs/AlAs superlattice structure”.
Surf. Interface Anal., 21(9):673–678, 1994.
[87] T. de Jong, W. A. S. Douma, L. Smit, V. V. Korablev, and F. W. Saris.
“The use of pulsed laser irradiation in silicon molecular beam epitaxy: A
comparative low energy electron diffraction study”. J. Vac. Sci. Technol. B,
1(4):888, 1983.
[88] J. S. Williams. “Solid phase epitaxial regrowth phenomena in silicon”. Nucl.
Instrum. Meth. Phys. Res., 209:219–228, 1983.
[89] F. A. Stevie and R. G. Wilson. “Relative sensitivity factors for positive atomic
and molecular ions sputtered from Si and GaAs”. J. Vac. Sci. Technol. A,
9(6):3064, 1991.
456
[90] J. S. Custer, Michael O. Thompson, D. C. Jacobson, J. M. Poate, S. Roorda,
W. C. Sinke, and F. Spaepen. “Density of amorphous Si”. Appl. Phys. Lett.,
64(4):437, 1994.
[91] Donald Long and John Myers. “Ionized-impurity scattering mobility of elec-
trons in silicon”. Phys. Rev., 115(5):1107–1118, 1959.
[92] Dimitrie Culcer and Neil M. Zimmerman. “Dephasing of Si singlet-triplet
qubits due to charge and spin defects”. Appl. Phys. Lett., 102(23):232108,
2013.
[93] K. Oura, M. Katayama, A. V. Zotov, V. G. Lifshits, and A. A. Saranin. Surface
Science: An Introduction. Springer, Berlin, 2003.
[94] M. Volmer and A. Weber. “Keimbildung in u¨bersa¨ttigten Gebilden”. Z. Phys.
Chem., 119(277):227–301, 1926.
[95] F. C. Frank and J. H. van der Merwe. “One-dimensional dislocations. II.
Misfitting monolayers and oriented overgrowth”. Proc. R. Soc. London A
Math. Phys. Eng. Sci., 198(1053), 1949.
[96] Helmut Alexander. Dislocations in Solids. Pages 115–118. North-Holland,
Amsterdam, 1980.
[97] R. G. Humphreys, S. Brand, and M. Jaros. “Electronic structure of the diva-
cancy in silicon”. J. Phys. C Solid State Phys., 16(12):L337–L343, 1983.
[98] V. V. Kveder, T. Sekiguchi, and K. Sumino. “Electronic states associated
with dislocations in p-type silicon studied by means of electric-dipole spin
resonance and deep-level transient spectroscopy”. Phys. Rev. B. Condens.
Matter, 51(23):16721–16727, 1995.
[99] C. A. Londos. “Defect states in electron-bombarded n-type silicon”. Phys.
Status Solidi, 113(2):503–510, 1989.
[100] Niklas Lehto. “Shallow electron states of bounded intrinsic stacking faults in
silicon”. Phys. Rev. B, 55(23):15601–15607, 1997.
[101] Ted Thorbeck and Neil M. Zimmerman. “Formation of strain-induced quan-
tum dots in gated semiconductor nanostructures”. AIP Adv., 5(8):087107,
2015.
[102] A.J. Sigillito, A.M. Tyryshkin, and S.A. Lyon. “Anisotropic Stark effect and
electric-field noise suppression for phosphorus donor qubits in silicon”. Phys.
Rev. Lett., 114(21):217601, 2015.
[103] Giuseppe Pica, Gary Wolfowicz, Matias Urdampilleta, Mike L. W. Thewalt,
Helge Riemann, Nikolai V. Abrosimov, Peter Becker, Hans-Joachim Pohl,
John J. L. Morton, R. N. Bhatt, S. A. Lyon, and Brendon W. Lovett. “Hy-
perfine Stark effect of shallow donors in silicon”. Phys. Rev. B, 90, 2014.
457
[104] Dimitrie Culcer, Xuedong Hu, and S. Das Sarma. “Interface roughness, valley-
orbit coupling, and valley manipulation in quantum dots”. Phys. Rev. B,
82(20):205315, 2010.
[105] John King Gamble, M. A. Eriksson, S. N. Coppersmith, and Mark Friesen.
“Disorder-induced valley-orbit hybrid states in Si quantum dots”. Phys. Rev.
B, 88(3):035310, 2013.
[106] Xiaojie Hao, Rusko Ruskov, Ming Xiao, Charles Tahan, and HongWen Jiang.
“Electron spin resonance and spinvalley physics in a silicon double quantum
dot”. Nat. Commun., 5:344–347, 2014.
[107] S. Srinivasan, G. Klimeck, and L. P. Rokhinson. “Valley splitting in Si quan-
tum dots embedded in SiGe”. Appl. Phys. Lett., 93(11):112102, 2008.
[108] Neil Zimmerman, Peihao Huang, and Dimitrie Culcer. “Valley phase, inter-
face roughness and voltage control, and coherent manipulation in Si quantum
dots”. arXiv:1608.06881, 2016.
[109] M. Nerding, L. Oberbeck, T. A. Wagner, R. B. Bergmann, and H. P. Strunk.
“Single to polycrystalline transition in silicon growth by ion-assisted deposition
at low temperatures”. J. Appl. Phys., 93(5):2570, 2003.
[110] G. Xue, H. Z. Xiao, M.-A. Hasan, J. E. Greene, and H. K. Birnbaum. “Critical
epitaxial thicknesses for low-temperature (20-100 ◦C) Ge(001)2x1 growth by
molecular-beam epitaxy”. J. Appl. Phys., 74(4):2512, 1993.
[111] Werner Kern. Handbook of Semiconductor Wafer Cleaning Technology: Sci-
ence, Technology, and Applications. Page 623. Noyes Publications, 1993.
[112] J. A. Mol, S. P. C. Beentjes, and S. Rogge. “A low temperature surface
preparation method for STM nano-lithography on Si(100)”. Appl. Surf. Sci.,
256(16):5042–5045, 2010.
[113] R. J. Hamers, R. M. Tromp, and J. E. Demuth. “Scanning tunneling mi-
croscopy of Si(001)”. Phys. Rev. B, 34(8):5343–5357, 1986.
[114] K. Hata, T. Kimura, S. Ozawa, and H. Shigekawa. “How to fabricate a defect
free Si(001) surface”. J. Vac. Sci. Technol. A, 18(4):1933, 2000.
[115] E. Kasper, M. Bauer, and M. Oehme. “Quantitative secondary ion mass
spectrometry analysis of SiO2 desorption during in situ heat cleaning”. Thin
Solid Films, 321(1):148–152, 1998.
[116] V. A. Ukraintsev and John T. Yate. “The role of nickel in Si(001) roughening”.
Surf. Sci., 346(1):31–39, 1996.
[117] H. Bracht, E. E. Haller, and R. Clark-Phelps. “Silicon self-diffusion in isotope
heterostructures”. Phys. Rev. Lett., 81(2):393–396, 1998.
458
[118] D. Eaglesham, H.-J. Gossmann, and M. Cerullo. “Limiting thickness hepi for
epitaxial growth and room-temperature Si growth on Si(100)”. Phys. Rev.
Lett., 65(10):1227–1230, 1990.
[119] D. J. Eaglesham. “Semiconductor molecular-beam epitaxy at low tempera-
tures”. J. Appl. Phys., 77(8):3597, 1995.
[120] N.E. Lee, G. Xue, and J. E. Greene. “Epitaxial Si(001) grown at 80-750 ◦C
by ion-beam sputter deposition: crystal growth, doping, and electronic prop-
erties”. J. Appl. Phys., 80(2):769–780, 1996.
[121] M. V. Ramana Murty, Harry A. Atwater, A. J. Kellock, and J. E. E. Baglin.
“Very low temperature (400 ◦C) silicon molecular beam epitaxy: the role of
low energy ion irradiation”. Appl. Phys. Lett., 62(20):2566, 1993.
[122] M. V. Ramana Murty and Harry A. Atwater. “Crystal-stateamorphous-
state transition in low-temperature silicon homoepitaxy”. Phys. Rev. B,
49(12):8483–8486, 1994.
[123] Jack Thiesen, Howard M. Branz, and Richard S. Crandall. “Explanation of
the limiting thickness observed in low-temperature silicon epitaxy”. Appl.
Phys. Lett., 77(22):3589, 2000.
[124] D. J. Eaglesham, G. S. Higashi, and M. Cerullo. “370◦C clean for Si molecular
beam epitaxy using a HF dip”. Appl. Phys. Lett., 59(6):685, 1991.
[125] H. Jorke, H.-J. Herzog, and H. Kibbel. “Kinetics of ordered growth of Si on
Si(100) at low temperatures”. Phys. Rev. B, 40(3):2005–2008, 1989.
[126] D.J. Eaglesham, H.-J. Gossmann, M. Cerullo, L.N. Pfeiffer, and K.W. West.
“Limited thickness epitaxy of semiconductors and Si MBE down to room
temperature”. J. Cryst. Growth, 111(1):833–837, 1991.
[127] E. Chason, P. Bedrossian, K. M. Horn, J. Y. Tsao, and S. T. Picraux. “Ion
beam enhanced epitaxial growth of Ge(001)”. Appl. Phys. Lett., 57(17):1793,
1990.
[128] C.-H. Choi, R. Ai, and S. Barnett. “Suppression of three-dimensional island
nucleation during GaAs growth on Si(100)”. Phys. Rev. Lett., 67(20):2826–
2829, 1991.
[129] N.-E. Lee, G. A. Tomasch, and J. E. Greene. “Low-temperature Si(001) epi-
taxy using low-energy (E ' 18 eV) Si atoms”. Appl. Phys. Lett., 65(25):3236,
1994.
[130] Morito Matsuoka and Shun-ichi Tohno. “Ion beam epitaxy of silicon films in
an ultrahigh vacuum using a sputtering-type metal ion source”. J. Vac. Sci.
Technol. A, 13(2):305, 1995.
459
[131] K. Miyake and T. Tokuyama. “Germanium and silicon ion-beam deposition”.
Thin Solid Films, 92(1-2):123–129, 1982.
[132] K. J. Boyd, D. Marton, J. W. Rabalais, S. Uhlmann, and Th. Frauenheim.
“Semiquantitative subplantation model for low energy ion interactions with
solid surfaces. III. Ion beam homoepitaxy of Si”. J. Vac. Sci. Technol. A,
16(2):463, 1998.
[133] M. Kitabatake, P. Fons, and J. E. Greene. “Molecular dynamics and quasidy-
namics simulations of low-energy ion/surface interactions leading to decreased
epitaxial temperatures and increased dopant incorporation probabilities dur-
ing Si MBE”. J. Cryst. Growth, 111(1):870–875, 1991.
[134] Kevin G. Orrman-Rossiter, Amir H. Al-Bayati, D.G. Armour, S.E. Donnelly,
and J.A. van den Berg. “Ion beam deposited epitaxial thin silicon films”.
Nucl. Instrum. Meth. Phys. Res. Sect. B, 59:197–202, 1991.
[135] L. A. Miller, D. K. Brice, A. K. Prinja, and S. T. Picraux. “Displacement-
threshold energies in Si calculated by molecular dynamics”. Phys. Rev. B,
49(24):16953–16964, 1994.
[136] K. A. Jackson, L. Csepregi, W. K. Chu, H. Muller, J. W. Mayer, T. W. Sigmon,
G.L. Olson, S.A. Kokorowski, J.A. Roth, L.D. Hess, J. S. Williams, R. G.
Elliman, W. L. Brown, T. E. Seidel, J. Linnros, R. G. Elliman, W. L. Brown,
Jan Linnros, W. L. Brown, R. G. Elliman, J.P. Biersack, and L.G. Haggmark.
“A defect model for ion-induced crystallization and amorphization”. J. Mater.
Res., 3(06):1218–1226, 1988.
[137] Joseph M. Jasinski and Stephen M. Gates. “Silicon chemical vapor deposition
one step at a time: fundamental studies of silicon hydride chemistry”. Acc.
Chem. Res., 24(1):9–15, 1991.
[138] W. C. T. Lee, N. Bishop, D. L. Thompson, K. Xue, G. Scappucci, J. G.
Cederberg, J. K. Gray, S. M. Han, G. K. Celler, M. S. Carroll, and M. Y.
Simmons. “Thermal processing of strained silicon-on-insulator for atomically
precise silicon device fabrication”. Appl. Surf. Sci., 265:833–838, 2013.
[139] E. Massa, G. Mana, L. Ferroglio, E. G. Kessler, D. Schiel, and S. Zakel. “The
lattice parameter of the 28Si spheres in the determination of the Avogadro
constant”. Metrologia, 48(2):S44–S49, 2011.
[140] A. Yang, M. Steger, H. J. Lian, M. L. W. Thewalt, M. Uemura, A. Sagara,
K. M. Itoh, E. E. Haller, J. W. Ager III, S. A. Lyon, M. Konuma, and M. Car-
dona. “High-resolution photoluminescence measurement of the isotopic-mass
dependence of the lattice parameter of silicon”. Phys. Rev. B, 77(11), 2008.
[141] Yu-Nong Yang and Ellen D. Williams. “The role of carbon in the faceting
of silicon surfaces on the (111) to (001) azimuth”. J. Vac. Sci. Technol. A,
8(3):2481, 1990.
460
[142] Hiroyuki Hirayama, Masayuki Hiroi, and Takashi Ide. “{311} facets of selec-
tively grown epitaxial Si layers on SiO2–patterned Si(100) surfaces”. Phys.
Rev. B, 48(23):17331–17337, 1993.
[143] Atsushi Oshiyama. “Structures of steps and appearances of {311} facets on
Si(100) surfaces”. Phys. Rev. Lett., 74(1):130–133, 1995.
[144] P. Ashu, C. C. Matthai, and T.-H. Shen. “Dynamics of atoms on silicon
substrates”. Surf. Sci., 251:955–959, 1991.
[145] P. Becker, P. Scyfried, and H. Siegert. “The lattice parameter of highly pure
silicon single crystals”. Zeitschrift fu¨r Phys. B Condens. Matter, 48(1):17–21,
1982.
[146] O. Madelung, U. Rossler, and M. Schulz. “Silicon carbide (SiC), lattice
parameters, thermal expansion”. In Gr. IV Elem. IV-IV III-V Compd.
Part b - Electron. Transp. Opt. Other Prop. Pages 1–11. Springer-Verlag,
Berlin/Heidelberg, 41a1b edition, 2002.
[147] F. Bozso, J. T. Yates, W. J. Choyke, and L. Muehlhoff. “Studies of SiC for-
mation on Si (100) by chemical vapor deposition”. J. Appl. Phys., 57(8):2771,
1985.
[148] L. Moro, A. Paul, D. C. Lorents, R. Malhotra, R. S. Ruoff, P. Lazzeri,
L. Vanzetti, A. Lui, and S. Subramoney. “Silicon carbide formation by an-
nealing C60 films on silicon”. J. Appl. Phys., 81(9):6141, 1997.
[149] James P. Becker, Robert G. Long, and John E. Mahan. “Reflection high-
energy electron diffraction patterns of carbide-contaminated silicon surfaces”.
J. Vac. Sci. Technol. A, 12(1):174, 1994.
[150] R. C. Henderson, R. B. Marcus, and W. J. Polito. “Carbide contamination of
silicon surfaces”. J. Appl. Phys., 42(3):1208, 1971.
[151] J. R. Engstrom, M. M. Nelson, and T. Engel. “Thermal decomposition of
a silicon-fluoride adlayer: evidence for spatially inhomogeneous removal of a
single monolayer of the silicon substrate”. Phys. Rev. B, 37(11):6563–6566,
1988.
[152] M. Maeder, A. Ramı´rez, and A. Zehe. “A RHEED study of as-received and
atomically clean silicon surfaces in UHV-environment”. Superf. y vac´ıo, 9:242–
244, 1999.
[153] Jun-Zhong Wang, Jin-Feng Jia, Hong Liu, Jian-Long Li, Xi Liu, and Qi-Kun
Xue. “Spontaneous vacancy array formation on FeSi2 and CoSi2 formed on
Si(100) 2xn surface”. Appl. Phys. Lett., 80, 2002.
461
[154] Ja-Yong Koo, Jae-Yel Yi, Chanyong Hwang, Dal-Hyun Kim, Sekyung Lee,
and Dong-Hyuk Shin. “Dimer-vacancy defects on the Si(001)–2x1 and the
Ni-contaminated Si(001)–2xn surfaces”. Phys. Rev. B, 52(24):17269–17274,
1995.
[155] B. Kolditz and K. R. Roos. “Effect of ambient pressure and nickel contami-
nation on the dimer-dangling-bond surface state of Si(001)2x1”. J. Vac. Sci.
Technol. A, 25(4):721–725, 2007.
[156] Masamichi Yoshimura, Izumi Ono, and Kazuyuki Ueda. “Initial stages of
Ni reaction on Si(100) and H-terminated Si(100) surfaces”. Appl. Surf. Sci.,
130:276–281, 1998.
[157] W. Koczorowski, M. Bazarnik, and R. Czajka. “SPM characterization of
titanium disilicide nanostructures grown on a Ni-modified Si(100) substrate”.
ACTA Phys. Pol. A, 120(3):480–484, 2011.
[158] H. J. W. Zandvliet, H. K. Louwsma, P. E. Hegeman, and Bene Poelsema.
“Energetics of Ni-induced vacancy line defects on Si(001)”. Phys. Rev. Lett.,
75(21):3890–3893, 1995.
[159] Yoichiro Yamaguchi, Masayuki Yoshida, and Hiroyuki Aoki. “Solid solubility
of nickel in silicon determined by use of 63Ni as a tracer”. Jpn. J. Appl. Phys.,
2(11):714–718, 1963.
[160] Y.-W. Mo, R. Kariotis, B. S. Swartzentruber, M. B. Webb, and M. G. Lagally.
“Scanning tunneling microscopy study of diffusion, growth, and coarsening of
Si on Si(001)”. J. Vac. Sci. Technol. A, 8(1):201, 1990.
[161] Y. W. Mo, J. Kleiner, M. B. Webb, and M. G. Lagally. “Activation energy
for surface diffusion of Si on Si(001): a scanning-tunneling-microscopy study”.
Phys. Rev. Lett., 66(15):1998–2001, 1991.
[162] G. Brocks, P. J. Kelly, and R. Car. “Binding and diffusion of a Si adatom on
the Si(100) surface”. Phys. Rev. Lett., 66(13):1729–1732, 1991.
[163] Takahisa Doi, Masakazu Ichikawa, Shigeyuki Hosoki, and Ken Ninomiya.
“Anisotropic diffusion between the step-up and the step-down directions on a
Si(001) surface”. Phys. Rev. B, 53(24):16609–16614, 1996.
[164] Chris Richardson. Document: “MBE and UHV component etchants”. Private
communication, 2015.
[165] P. Asoka-Kumar, S. Szpala, B. Nielsen, Cs. Szeles, K. G. Lynn, W. A. Lanford,
C. A. Shepard, and H.-J. Gossmann. “Hydrogen-induced breakdown of low-
temperature molecular-beam epitaxy of Si”. Phys. Rev. B, 51(7):4630–4632,
1995.
462
[166] J. A. Roth, G. L. Olson, D. C. Jacobson, and J. M. Poate. “Kinetics of solid
phase epitaxy in thick amorphous Si layers formed by MeV ion implantation”.
Appl. Phys. Lett., 57(13):1340, 1990.
[167] L. Oberbeck, N. J. Curson, T. Hallam, M. Y. Simmons, G. Bilger, and R. G.
Clark. “Measurement of phosphorus segregation in silicon at the atomic scale
using scanning tunneling microscopy”. Appl. Phys. Lett., 85(8):1359–1361,
2004.
[168] R. J. Hamers, U. K. Ko¨hler, and J. E. Demuth. “Epitaxial growth of silicon on
Si(001) by scanning tunneling microscopy”. J. Vac. Sci. Technol. A, 8(1):195,
1990.
[169] A. H. Albayati, K. J. Boyd, D. Marton, S. S. Todorov, J. W. Rabalais, Z. H.
Zhang, and W. K. Chu. “Substrate-temperature dependence of homoepitax-
ial growth of Si using mass-selected ion-beam deposition”. J. Appl. Phys.,
76(7):4383–4389, 1994.
[170] M. Wang, G. Audi, A.H. Wapstra, F.G. Kondev, M. MacCormick, X. Xu,
and B. Pfeiffer. “The Ame2012 atomic mass evaluation”. Chinese Phys. C,
36(12):1603–2014, 2012.
[171] Y. Yatsurugi, N. Akiyama, Y. Endo, and T. Nozaki. “Concentration, solubil-
ity, and equilibrium distribution coefficient of nitrogen and oxygen in semi-
conductor silicon”. J. Electrochem. Soc., 120(7):975, 1973.
[172] Tadashi Nozaki, Yoshifumi Yatsurugi, and Nobuyuki Akiyama. “Concentra-
tion and behavior of carbon in semiconductor silicon”. J. Electrochem. Soc.,
117(12):1566, 1970.
[173] M. I. den Hertog, C. Cayron, P. Gentile, F. Dhalluin, F. Oehler, T. Baron,
and J. L. Rouviere. “Hidden defects in silicon nanowires”. Nanotechnology,
23(2):025701, 2012.
[174] James H. Comfort and Rafael Reif. “Chemical vapor deposition of epitaxial
silicon from silane at low temperatures”. J. Electrochem. Soc., 136(8):2386,
1989.
[175] Andrei A. Onischuk and Viktor N. O. Panfilov. “Mechanism of thermal de-
composition of silanes”. Russ. Chem. Rev., 70(4):321–332, 2001.
[176] B. A. Joyce, R. R. Bradley, and G. R. Booker. “A study of nucleation in
chemically grown epitaxial silicon films using molecular beam techniques III.
Nucleation rate measurements and the effect of oxygen on initial growth be-
haviour”. Philos. Mag., 15(138):1167–1187, 1967.
[177] R. J. Buss, P. Ho, W. G. Breiland, and M. E. Coltrin. “Reactive sticking
coefficient for silane and disilane on polycrystalline silicon”. J. Appl. Phys.,
63(8):2808–2819, 1988.
463
[178] M. Liehr, C. M. Greenlief, S. R. Kasi, and M. Offenberg. “Kinetics of silicon
epitaxy using SiH4 in a rapid thermal chemical vapor deposition reactor”.
Appl. Phys. Lett., 56(7):629, 1990.
[179] F. G. Ruedenauer. “Gas scattering as a limit to trace sensitivity in analytical
mass spectrometers”. Rev. Sci. Instrum., 41(10):1487, 1970.
[180] Trevor R. Ireland. “Invited Review Article: Recent developments in isotope-
ratio mass spectrometry for geochemistry and cosmochemistry”. Rev. Sci.
Instrum., 84(1):011101, 2013.
[181] N. J. Freeman, N. R. Daly, and R. E. Powell. “Retardation lens used to
improve the abundance sensitivity of a mass spectrometer”. Rev. Sci. Instrum.,
38(7):945, 1967.
[182] R. L. Summers. “Empirical observations on the sensitivity of hot cathode
ionization type vacuum gauges”. Technical report, National Aeronautics and
Space Administration, NASA TN D-5285, Washington, D.C., 1969.
[183] C. Y. Peng, A. A. Woodworth, and C. D. Stinespring. “Method for rapid deter-
mination of ion gauge sensitivity factors”. Rev. Sci. Instrum., 78(11):113908,
2007.
[184] H. Hertz. “On the evaporation of liquids, especially mercury, in vacuo”. Ann.
Phys. (N. Y.), 17:177, 1882.
[185] M. Knudesn. “Maximum rate of vaporization of mercury”. Ann. Phys. (N.
Y.), 47:697, 1915.
[186] Irving Langmuir. “The adsorption of gases on plane surfaces of glass, mica,
and platinum”. J. Am. Chem. Soc., 40(9):1361–1403, 1918.
[187] S. M. Gates, C. M. Greenlief, D. B. Beach, and R. R. Kunz. “Reactive stick-
ing coefficient of silane on the Si(111)-(7x7) surface”. Chem. Phys. Lett.,
154(6):505–510, 1989.
[188] Stephen M. Gates. “Adsorption kinetics of SiH4, Si2H6 and Si3H8 on the
Si(111)-(7x7) surface”. Surf. Sci., 195(1):307–329, 1988.
[189] S. R. Logan. “The origin and status of the Arrhenius equation”. J. Chem.
Educ., 59(4):279, 1982.
[190] Keith J. Laidler. “The development of the Arrhenius equation”. J. Chem.
Educ., 61(6):494, 1984.
[191] B. A. Scott, R. D. Estes, and J. M. Jasinski. “The role of surface reactions in
monosilane pyrolysis”. J. Chem. Phys., 89(4):2544–2549, 1988.
464
[192] Karl F. Roenigk, Klavs F. Jensen, and Robert W. Carr. “Rice-Ramsperger-
Kassel-Marcus theoretical prediction of high-pressure Arrhenius parameters
by nonlinear regression: application to silane and disilane decomposition”. J.
Phys. Chem., 91(22):5732–5739, 1987.
[193] F. Hirose. “Modeling growth in Si gas-source molecular beam epitaxy using
Si2H6”. J. Cryst. Growth, 179(1):108–114, 1997.
[194] S. M. Gates, C. M. Greenlief, S. K. Kulkarni, and H. H. Sawin. “Surface
reactions in Si chemical vapor deposition from silane”. J. Vac. Sci. Technol.
A Vacuum, Surfaces, Film., 8(3):2965, 1990.
[195] D. S. Lin, E. S. Hirschorn, T. C. Chiang, R. Tsu, D. Lubben, and J. E. Greene.
“Scanning-tunneling-microscopy studies of disilane adsorption and pyrolytic
growth on Si(100)-(2X1)”. Phys. Rev. B, 45(7):3494–3498, 1992.
465