The Growth and Characterization of Gallium Arsenide Nanowire Structures by Metal Organic Chemical Vapor Deposition

Home , Arsenide, Gallium arsenide, Trimethylgallium

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Nicholas G. Minutillo

Graduate Program in Physics

The Ohio State University

2014

Dissertation Committee:

Professor Fengyuan Yang, Advisor

Professor Jay A. Gupta

Professor Klaus Honscheid

Professor Mohit Randeria

Nicholas Gaetano Minutillo

2014

Abstract

Semiconductor nanowires hold a wealth of promise for studying the fundamental physics of electron behavior and interactions in a quasi-one dimensional environment as well as components in or the foundation of technological advancement in electronic and spintronic devices. Especially in the case of spintronic applications, the crystalline environment must be highly controlled. Unlike in electronic devices, predicated on the transport or storage of charges, spintronic devices often depend on relative phases of spin states. These phases are easily lost in an environment where scattering probabilities are high. In any material system, control of the material fabrication is the limiting factor to achieving the theoretical characteristics and operation. Still an active area of research, bottom-up synthesis of semiconductor nanowires has yet to reach the level of control required for wide spread adoption as a base system in condensed matter research. At this point in time, the material synthesis to meet the criteria for advanced applications remains a bottle neck in advancing the application of GaAs or any other semiconductor nanowires.

In this dissertation we discuss the vapor-liquid-solid (VLS) mechanism and its role in the growth of gallium arsenide and other III-V semiconductors. The VLS mechanism has become a foundation of bottom-up nanowire growth. We will further discuss metal organic chemical vapor deposition (MOCVD), an epitaxial technique ii developed for III-V semiconductor thin films that has risen to prominence in the field of nanowire growth. The physics that governs the VLS growth of GaAs nanowires is the subject of ongoing research.

We systematically analyze the effect of core-growth temperature on VLS, epitaxial GaAs/AlGaAs core/shell nanowires in MOCVD by photoluminescence characterization of nanowire ensembles as a function of core growth temperature. To our knowledge, a systematic study of the photoluminescence dependence on growth temperature prior to ours does not exist in the peer-reviewed literature. We demonstrate photoluminescence linewidths on ensembles of nanowires that are competitive with the best single-wire linewidths reported in the literature. We thus demonstrate wires of highly uniform characteristics across the entire growth surface. Our results also indicate that the effect of the core growth temperature is coupled to the crystal orientation of the substrate surface. At low growth temperatures, nanowires grown on a GaAs (100) surface exhibit a narrower photoluminescence peak at the band edge in a wider growth temperature window than do the wires grown on a GaAs (111)B surface. This is contrary to what might be expected given that all the wires grow in the <111> direction and display the same growth rate on both substrate surfaces. Under the conditions used, the window in growth temperature for a high optical quality gallium arsenide core nanowire is narrow compared to common conditions in thin film epitaxy by MOCVD.

We discuss our methods for the successful growth of a novel nanowire device structure by MOCVD. We grow vertical GaAs nanowires and embedded them in a continuous film of AlGaAs. This structure has thus far only been reported in molecular

iii beam epitaxy, which by its directional nature, more naturally lends itself growth in high aspect ratio channels. In addition to being relatively uncommon, this structure has several advantages. First, the 40 nm diameter GaAs nanowires are protected by the in situ

AlGaAs growth. Second, the geometry allows the use of thin film techniques for device processing to easily control the number of wires to be activated in the device by simply changing the area of the patterned contact. Development of this nanowire-thin-film geometry opens the door for the study of parallel ensembles of nanowires and nanowire heterostructures.

Dedication

To Gaetano Minutillo and Anthony Laudati

Acknowledgments

I must first thank my advisor, Professor Fengyuan Yang for his guidance, support, and patience as I navigated my way through graduate research. I am fortunate to have had the opportunity to work in his lab and learn about semiconductor physics and be granted access and exposure to many areas of semiconductor science in both the academic and industrial realm. Above all, I have learned that crystal growth requires a unique combination of determination, patience, impatience, and resilience. Many thanks to

Professor Zeke Johnston-Halperin, who has treated me like one of his own students throughout our collaboration and who has been a provider of helpful insights and perspectives. Thank you to Dr. John Carlin who taught me not only about semiconductor physics, but about the industry as well. His perspective as a scientist who straddles and succeeds in both worlds of semiconductor science remains invaluable to me as I move forward. Thank you for answering your phone at 2 am to help me bring the MOCVD out of an alarm state. Thank you to Yi-Hsin Chiu, my longest and most patient collaborator at

OSU. I thank her for all the conversations about physics and about life, for her level headed demeanor and for her tireless efforts to measure every sample I could throw at her.

I would like to further thank all the scientists who have lent me their invaluable time and expertise for my own edification and to the data contained herein. Thank you to

Rob Williams and Professor Dave McComb for working with us and imaging our nanowires with both of their best (and coolest) STEM instruments. Thank you also to Dr.

Camelia Selcu, who began our conductive AFM and gave me her expert advice as we continue on with the technique on our own. Thank you for your patience and hard work.

Thank you to Adam Hauser for always being a positive source of encouragement and acting as a role model worthy of emulation. Thanks to Brian Peters for our many late night/early morning conversations and helping me keep my sanity when things went awry. Thank you to Jeremy Lucy for having a remarkable strength of character. For saying what needs to be said, but also staying around to help fix whatever needs fixing, either broken instruments or broken morale. Thanks to James Gallagher for his unique and always sincere perspective. Thanks to Hailong who has proven the value of relentless hard work. Thank you to Greg Smith for his contributions to both this project and myself as a scientist. I wish you all the best in your new group and I hope that you carry with you the tradition of bad jokes I have tried so hard to instill.

Thank you to Mark Patrick, Megan Harberts, Ula Szafruga, Yaser Helal, and

Richelle Teeling-Smith for your friendships throughout this journey we call graduate school. Thank you to Tricia Meyer for all of your loving support. Thank you for being the voice of calm, reason, and compassion when I needed it the most and seeing in me what I might otherwise not have seen myself. Thank you to Professor Paul Angiolillo for believing in me from the moment I said I wanted to study physics, halfway through my

vii college career. Thank you for enabling me to make the step into graduate studies and for supporting me as a friend and a mentor the entire way. Thank you to Andrzej Latka for going on this adventure with me, for pushing me outside of my comfort zone, and for being the most loyal and true friend a person could hope to find. Finally thank you to my parents Angelo and Mary Ann and my sister Madeleine for supporting me in this and every chapter of my life. I could not have made it this far without your love and support, help through the growing pains and the triumphs. Thank you for keeping me grounded and giving me the confidence to push through the tough times. Thank you for helping me put everything in the proper perspective and allowing me to believe that my aspirations are possible. Above all thank you to my parents for every sacrifice you have made to provide us the environment and opportunity to achieve whatever level of education or career path we most desire.

viii

Vita

May 2004 ...... Saint Joseph’s Preparatory School

December 2008 ...... B.S. Physics, Saint Joseph’s University

August 2011 ...... M.S. Physics, The Ohio State University

September 2009 to June 2009 ...... Graduate Teaching Associate, Department

of Physics, The Ohio State University

June 2009 to August 2013 ...... Graduate Research Associate, Department

of Physics, The Ohio State University

September 2013 to present ...... Graduate Teaching Associate, Department

of Physics, The Ohio State University

Publications

Nicholas G. Minutillo, Yi-Hsin Chiu, Robert E.A. Williams, Greg J. Smith, David W.

McComb, John A. Carlin, Ezekiel Johnston-Halperin, Fengyuan Yang,

Photoluminescence and Morphology Evolution in GaAs/AlGaAs Core/Shell Nanowires

Grown by MOCVD: Effects of Core Growth Temperature – Submitted

Fields of Study

Major Field: Physics

Table of Contents

Abstract ...... ii

Dedication ...... v

Acknowledgments...... vi

Vita ...... ix

Table of Contents ...... xi

List of Tables ...... xvi

List of Figures ...... xvii

Chapter 1: Nanostructures & Semiconductors ...... 1

§1.1 – Introduction to Low Dimensional Semiconductor Structures & Nanowires ...... 2

§1.2 – Methods of Nanowire Synthesis: Bottom-Up Versus Top-Down ...... 4

§1.3 – III-V Semiconductors & Properties...... 7

§1.4 – GaAs/AlGaAs Core/Shell Nanowires ...... 9

§1.5 – Spintronics, Semiconductors, and Nanowires ...... 13

§1.5.1 – Overview of Spintronics ...... 13

§1.5.2 – Semiconductor Based Spintronics ...... 14

§1.6 – Optical Orientation and Measurement of Electron Spin in Semiconductors ..... 16

§1.6.1 – Optical Orientation ...... 17

§1.6.2 – Time Resolved Faraday Rotation ...... 18

Chapter 2: Nanowire Growth by Metal-Organic Chemical Vapor Deposition ...... 22

§2.1 – Overview of Metal Organic Chemical Vapor Deposition ...... 23

§2.1.1 – Solid Layer Formation in MOCVD ...... 28

§2.2 – Thermodynamics of Metal Organic Chemical Vapor Deposition ...... 29

§2.3 – Carbon Contamination and Precursor Decomposition ...... 32

§2.4 – Pyrometry in MOCVD Reactors ...... 35

Chapter 3: Equipment, Reactor Conditions, and In Situ Temperature Techniques for the

Experimental Growth of GaAs Nanowires by MOCVD ...... 39

§3.1 – Close Coupled Showerhead MOCVD Reactor and Standard Reactor Conditions

...... 39

§3.2 – Surface Temperature and Pyrometry in MOCVD...... 42

§3.2.1 – Temperature Balancing ...... 44

§3.2.2 – In Situ Pyrometry under Standard Growth Conditions ...... 45

§3.3 – Surface Pyrometry for Nanowire Growth ...... 46

Chapter 4: Vapor-Liquid-Solid Growth and III-V Nanowire Synthesis ...... 51

§4.1 – An Overview of the Vapor-Liquid-Solid Mechanism ...... 51

xii

§4.2 – Parameters in III-V Nanowire Growth by MOCVD ...... 55

§4.2.1 – Precursor Flux...... 56

§4.2.2 – Nanowire Growth Temperature ...... 56

§4.2.3 – V/III Ratio...... 59

§4.2.4 – Catalyst Spacing ...... 61

§4.2.5 – Seed Particle Diameter ...... 62

§4.3 – Nanowire Shell Growth ...... 62

§4.4 – Conformal Growth in Narrow Spaces ...... 64

§4.4.1 – Conformal AlGaAs Growth on Vertical Nanowires ...... 67

§4.4.2 – Long Term Use for a Coalesced GaAs Nanowire Structure ...... 69

§4.4.3 – Substrate Orientation in Conformal AlGaAs Growth ...... 70

Chapter 5: Characterization Techniques ...... 72

§5.1 – Electron Microscopy ...... 73

§5.1.1 – Scanning Electron Microscopy ...... 73

§5.1.2 – Scanning Transmission Electron Microscopy ...... 78

§5.2 – Photoluminescence Spectroscopy ...... 81

§5.3 – Atomic Force Microscopy ...... 86

§5.3.1 – Tapping Mode AFM ...... 88

§5.3.2 – Peak Force Tapping Mode AFM ...... 90

xiii

§5.3.3 – Conductive AFM ...... 92

Chapter 6: The Epitaxial Growth of GaAs/AlGaAs Core/Shell Nanowires by Metal-

Organic Chemical Vapor Deposition ...... 95

§6.1 – Structural Properties ...... 95

§6.1.1 – Nanowire VLS Growth Rate ...... 98

§6.1.2 – Crystal Structure and Twinning Defects ...... 101

§6.2 – Optical Properties ...... 109

§6.2.1 – Photoluminescence as a Function of Core Growth Temperature and

Substrate Orientation ...... 110

§6.2.2 – Peak Ratio and Low Temperature Growth ...... 114

Chapter 7: Surface Processing and Characteristics of GaAs Nanowires in a Continuous

AlGaAs Layer ...... 118

§7.1 – Topographical and Electrical Characteristics of Mechanically Polished Surfaces

...... 120

§7.1.1 – AlGaAs Removal by Ion Beam Milling ...... 123

§7.1.2 – The Future of GaAs Nanowire-Thin-Film Structures ...... 126

Conclusions ...... 128

References ...... 131

Appendix A: Preparation of the Gold Colloid ...... 139

Overview of VLS Surface Preparation...... 139 xiv

Nanowire Surface Prep in the Yang Lab ...... 141

List of Tables

Table 1: Full set of growth conditions used in temperature series. The conditions read chronologically left to right...... 97

xvi

List of Figures

Figure 1: A GaAs NW is made by (a) cleaving a high purity GaAs/AlGaAs QW (c) growing doping and gating material on the edge for (f) constraining the electrons to the edge of the well at the cleave. Adapted from Yacoby et al.2 ...... 5

Figure 2: InP nanowires grown on Si by masking with a SiO2 layer from Mohan et al.11 6

Figure 3: Plot of lattice constant versus band gap for III-V ternaries.14 Some version of this plot is often found hanging next to the periodic table above ones desk...... 8

Figure 4: Unit cell of GaAs. Ga atoms are grey and As atoms are blue...... 10

Figure 5: (a) Band structure of gallium arsenide calculated by Chelikowsky and Cohen in

1976.16 (b) Detailed view of the GaAs band edge adapted from Blakemore.17 ...... 11

Figure 6: (a) Schematic of the GaAs band structure near the Γ-point. (b) Schematic of valence to conduction band transitions by photon absorption denoted by arrows. The numbers indicate the relative transition rates by photons of that helicity. Figure from

Žutić et al.23 ...... 18

Figure 7: Schematic depiction of the TRFR measurement in the Voigt geometry from

Young et al. (2002).37 ...... 20

Figure 8: Aixtron CCSTM 3×2” MOCVD. Not depicted: gas cabinets for the group V precursors and exhaust scrubber. Image by Aixtron SE.42 ...... 24

Figure 9: The basic operation of an MOCVD System...... 25 xvii

Figure 10: MOCVD bubbler immersed in water for temperature stabilization...... 27

Figure 11: Different behaviors in epitaxial film growth. Adapted from Kuech 2010.46 .. 28

Figure 12: This GaAs surface, imaged by a Nomarski microscope, exhibits classic orange peel morphology along with faceted growths protruding from the surface due to crystalline contaminants that had fallen onto the surface before growth...... 29

Figure 13: Trimethylgallium: Gallium is covalently bonded to 3 methyl groups that are spaced at 120°. The carbon and gallium atoms are coplanar. Arsine: arsenic is covalently bonded to 3 hydrogens that are ~91° separated. The arsenic atom sits above the plane containing the hydrogen atoms.54,55 ...... 33

56 Figure 14: Composition of TMGa and AsH3 from DenBaars et al. The right plot is the relative concentration of AsH3 in H2 in the reactor...... 35

Figure 15: Radiance versus emission wavelength given by Equation 10...... 37

Figure 16: An Aixtron 3×2" CCS reactor identical to the reactor used here. The lid lifts up and the samples (here circular wafers) are place in pockets on the susceptor (grey plate). Image from Aixtron SE.61 ...... 40

Figure 17: Schematic of gas flow in our Aixtron CCS MOCVD reactor...... 41

Figure 18: The Yang group susceptor is 5” in diameter and has multiple pockets for growth. In this picture the pockets are occupied by dummy wafers that protect them when not in use...... 42

Figure 19: Schematic of the heater/susceptor assembly in our MOCVD. The susceptor is depicted as orange...... 43

xviii

Figure 20: Thee ports for monitoring susceptor surface temperature in heater zones A, B, and C...... 45

Figure 21: Cartoon representation of substrate placement for temperature monitoring during nanowire growth...... 47

Figure 22: Typical data for a core/shell growth. The hatched areas correspond to time intervals of stable temperatures. Growths are not started until the thermocouple reading is within 1°C of the targeted value. Note that the pyrometer output (bottom curve) does not output values below ~450°C while the thermocouple does...... 48

Figure 23: Pyrometer vs. thermocouple data for a clean graphite susceptor (▲), a Si surface during the first two temperature steps of the nanowire growth (■), and the targeted surface temperature for nanowire growth (□)...... 49

Figure 24: Gold-Silicon phase diagram. VLS growth happens at a temperature where a liquid alloy is formed but the silicon in the droplet increases and the system enters the liquid alloy + solid silicon region of the phase diagram...... 52

Figure 25: A schematic of core/shell growth for GaAs nanowires. (a) Au particle is dispersed on a GaAs {111} surface, (b) precursors are introduced into the reactor and a liquid Au-Ga-As alloy is formed, (c) a monolayer of solid GaAs forms at the substrate surface advancing the droplet up, (d) the reactor temperature is increased to temperatures conducive to epitaxial film growth and precursors are again introduced to grow an

AlxGa1-xAs shell on the sidewalls of the nanowires...... 53

xix

Figure 26: (a) GaAs nanowire segment with a ~15 nm AlGaAs shell sitting on a Si substrate; the Au droplet used for the core growth can been seen at the top of the wire (far right in the image). (b) magnification of the GaAs-alloy interface...... 54

Figure 28: Temperature dependence of GaAs thin film growth rate in MOCVD complied by G. B. Stringfellow.48 ...... 57

Figure 29: The V/III = 6.4 in this growth is too low and so the Ga and In have nucleated unwanted VLS growth all over the surface among the NWs seeded by Au...... 60

Figure 30: Schematic of TMGa diffusion in VLS growth...... 61

Figure 31: When the radii defined by the diffusion length of the group III precursor overlap, adjacent nanowires compete for precursors...... 62

Figure 32: (a) nanowire with no shell (b) a ~15 nm AlGaAs shell and (c) a thick ~1 μm shell. The nanowire in (c) has been partially cross sectioned and the GaAs core can be seen with brighter contrast at the base of the wire...... 63

Figure 33: 45° tilt view of 500 nm nanowires (a) on GaAs (111)B (b) with a coalesced

AlGaAs layer. The red dashed lines indicate the height of the nanowires above the substrate surface...... 65

Figure 34: Void formation (a) in Kusakabe et a.l77 (b) vertical nanowires...... 65

Figure 35: AlGaAs growth rate is much higher at the nanowire tip...... 66

Figure 36: Hillock of AlGaAs formed on a (111)B surface. The bead at the top is believed to be a sphere or group III metals due to insufficient As pressure...... 67

Figure 37: 45° tilted view of (a) GaAs nanowires ~500 nm tall, AlGaAs deposition for (b)

1 min, (c) 10 min, (d), 30 min imaged at 45°. Scale bars are 200 nm (a) & (b) and 400 nm

Figure 38: Spin LED schematic. Spin polarized carriers would recombine at the NW- substrate interface preferentially emitting left or right circularly polarized light depending on the spin orientation of the carriers...... 69

Figure 39: Spin LED used to measure spin coherence length by varying the i-GaAs thickness. From Ohno et al.82 ...... 70

Figure 40: General construction of an SEM. Reproduction of two figures from Goodhew et al.86 ...... 75

Figure 41: Diagram of the surface emissions as a result of the incident electron beam.

Reproduction of the figure found in Goodhew et al.86 ...... 76

Figure 42: Diagram of the interaction volume of the electron beam with the sample surface. Note that the backscattered electrons originate from deeper underneath the interface than do the secondary electrons. A reproduction of the figure found in Goodhew et al.86 ...... 77

Figure 43: Inner shell electrons are displaced by the incident beam and electron from an outer shell takes its place. The energy difference is in the x-ray band and is characteristic of the differences in energy levels of different nuclei. Figure from Williams & Carter.8778

Figure 44: This diagram depicts one condenser lens and one objective/projector lens but current STEM systems commonly have two condensers and 5 objective/projector lenses.86 ...... 79

xxi

Figure 45: Diagram of (a) direct gap transition and (b) indirect gap transitions from Y.P.

Varshi in Phys. Stat. Sol. (1967).89 Note that a direct transition can occur in indirect gap semiconductors but they are at a higher energy than the band gap...... 82

Figure 46: Schematic of possible radiative recombination pathways. (a) Band-to-band, (b) donor impurity to valence band, (c) conduction band to acceptor impurty, and (d) donor- to-acceptor impurities...... 83

Figure 47: Photoexcited electrons thermalize with the lattice towards the conduction band minimum before recombining with a corresponding hole. Diagram from Pavesi and

Guzzi.90 ...... 84

Figure 48: The PL from this nanowire growth shows a band edge peak at 821 nm as well as a defect peak near 830 nm. The defect is likely carbon...... 85

Figure 49: Diagram of AFM tip on surface with laser detection from Haugstad 2012.93 87

Figure 50: Schematic of tapping mode AFM on a sample surface...... 89

Figure 51: The Lennard-Jones potential between the tip and the surface as a function of their separation...... 89

Figure 52: Hysteresis loop in tip deflection as a function of height from the Bruker

Corporation.94 Image edited for clarity...... 91

Figure 53: Idealized force versus time plot for one cycle. This image is an adaptation of two images from literature provided by the Bruker Corporation94,95 edited for clarity. ... 92

Figure 54: Diagram of conductive AFM experimental geometry for the Bruker PeakForce

TUNA instrumentation...... 93

xxii

Figure 55: Relation in time of the tip oscillation, tip force, and detected current. This image is provided in literature published by the Bruker Corporation.96 ...... 94

Figure 56: GaAs/AlGaAs nanowires grown on (a) a (111)B GaAs substrate (b) a (100)

GaAs substrate. Note that the growth direction of the wires always <111>...... 96

Figure 57: Axial growth rate as a function of growth temperature for nanowires grown on a GaAs (100) and GaAs (111)B surface. Error bars not visible are inside the symbol. ... 98

Figure 58: Average axial growth rate (top) and the average tip diameter (bottom) of

GaAs/AlGaAs NWs grown on (111)B. The large standard deviations originate in part from differences in shell thickness and variation in Au particle diameter...... 100

Figure 59: Nanowires grown at 436°C (left) are notably longer with a smaller diameter than those grown at 472°C (right)...... 101

Figure 60: Top down view of nanowires cross sectioned close to their base in the plane of the c-axis. The green dot represents the 40 nm starting diameter of the Au seed...... 102

Figure 61: HR-STEM of nanowires grown on (111)B at 430°C. Contrast planes can be seen disrupting the crystal structure...... 103

Figure 62: HR-STEM of GaAs NWs with corresponding EDS intensity maps for each of the 4 expected elements...... 104

Figure 63: Atomically resolved image of a twin defect in a nanowire. Inset: the dumbbell structure of the GaAs is resolved...... 106

Figure 64: Comparison of a simulated twin in a ZB NW from Wood et al. with our HR-

STEM data...... 106

xxiii

Figure 65: FFT of the real space HR-STEM images above and below the boundary plane.

Indexing is explained in the text...... 107

Figure 66: 45° tilted view of a nanowires grown on a (111)B GaAs substrate...... 109

Figure 67: The evolution of the NW PL spectra with growth temperature on both substrates. The first and last temperatures are the two extremes in the temperature series.

...... 111

Figure 68: Peak ratio versus core growth temperature. Substrate surface orientation is indicated in the plot...... 115

Figure 69: Top down SEM image of a polished coalesced GaAs/AlGaAs surface. The sharp lines of contrast are due to multiple hillocks and nanowire walls growing laterally into each other. The irregular shapes are the bottom of valleys remaining from the as grown surface due to insufficient polishing depth...... 119

Figure 70: Topography of a polished coalesced NW surface by peak force tapping mode.

The large dip in height to the left is a void in the AlGaAs layer...... 120

Figure 71: Contact current data taken concurrently with the topography data in Figure 69.

...... 122

Figure 72: 45° view of silica particles remaining on the polished surface after sonication in solvent, scale bar 650 nm. Inset: Top down image of unidentified contaminants with much brighter contrast from the GaAs surface, scale bar 10 μm...... 123

Figure 73: 45° tilted views. Hexagonal patterns and pits can be seen all over the surface, scale 400 nm. Inset: Starfish feature with pit believed to be a GaAs NW that has etched faster than its surroundings. Scale 200 nm...... 124

xxiv

Figure 74: (a) High magnification SEM of surface feature resulting from Ar+ ion milling.

(b) TEM of a radial GaAs/AlGaAs structure from Zheng et al.113 (c) AFM of surface feature...... 125

Figure 75: Au colloids at two densities. The lower density tends to have a larger standard deviation in spacing...... 140

xxv

Chapter 1: Nanostructures & Semiconductors

In the last 50 years, semiconductor-based electronics have revolutionized technology and has markedly changed society and everyday life. Research in semiconductors has extended into every conceivable class of semiconductor and the systems and device applications are only limited by technological capabilities or the imagination of the scientist. While silicon is currently the dominant semiconductor in current electronic technology, many compound semiconductors play important roles in the realm of research and technology. Gallium arsenide is one of the most extensively studied and well developed semiconductor materials in bulk and thin films. Gallium arsenide and GaAs-alloys are currently used in a number of current technological applications. Owing to higher mobilities compared to silicon, gallium arsenide is employed in power amplifiers operating in the GHz range. Such amplifiers are often made of GaAs based heterojunction bipolar transistors or metal-semiconductor FETs and are commonly found in cell phones. GaAs and related alloys are also found in semiconductor laser diodes. While pure GaAs cannot emit in the visible spectrum, alloys can be synthesized to increase the emitted photon energy. GaAs and GaAs-alloy systems are further useful because their complimentary properties provide greater flexibility in

1 terms of what is possible to accomplish. Gallium arsenide based devices are also useful for space-based operations because it is more resilient to radiation damage than silicon.

§1.1 – Introduction to Low Dimensional Semiconductor Structures & Nanowires

A goal of physicists has been to study the behavior of electrons in systems that closely approximate low dimensional Hamiltonians due to the expected phenomena in low dimensions that are both scientifically interesting and technologically relevant. The first realization of this goal in the realm of condensed matter physics was the semiconductor quantum well. The semiconductor quantum well approximates a two dimensional environment by constraining the electrons in a well layer that is on the order of the electron wave function in the growth direction, but infinite laterally for all practical purposes. Electrons are constrained to the layer by sandwiching it in between layers of a different material with a larger band gap such that electrons cannot easily occupy states in the neighboring material. A classic example of this is the GaAs/AlGaAs quantum well where the active layer is a thin (~14 nm) gallium arsenide film sandwiched between two thicker aluminum gallium arsenide barrier layers.

This success fed the desire for other low dimensional material systems to study fundamental physics and create devices not possible in bulk materials. Recently, a new class of two dimensional (2D) materials, most noticeably graphene, has generated tremendous interest due to their promise in next-generation technologies and scientifically interesting properties. In zero dimensions (0D), semiconductor quantum dots, whose dimensions in all directions are on the order of the electron wave function,

2 have been fabricated in a number of ways and are seeing use in many scientific applications.1 Free-standing semiconductor nanowires are the “weak link” between 2D semiconductor quantum wells and 0D semiconductor quantum dots. The synthesis of these quasi-1D structures from vapor phase precursors has required a completely new growth paradigm and is still the subject of research since gaining traction in the 1990s thanks to the successful work the Lieber group at Harvard and others like the Samuelson group at Lund University. Semiconductor nanowires continue to open up new doors in fundamental research and they are excellent candidates for integration into already existing device structures as well as the basis for new devices. The semiconductor industry is constantly searching for the next method to shrink feature sizes, improve efficiency, reduce power cost, and increase speed. Nanowires show great potential for the future of the information industry and hold a wealth of scientific value from their synthesis to their unique properties.

A nanowire with a radius on the order of electron wave function in that material is a playground for physicists to study electron behavior in a quasi-one dimensional system.2 Nanowires possess advantages over thin films due to their nanometer scale and high aspect ratio. It is difficult to grow layered structures with materials that have a large lattice or crystal structure mismatch e.g. III-Vs on Si. Nanowires can accommodate larger strain because of their lateral dimensions are only 10s of nanometers. This allows the growth of heterostructures that are less feasible in a film structure.3 The high aspect ratio of the nanowire allows the strain in the III-V layer to relax much faster than in a film and nanowires of high crystalline quality have been grown on substrates with a large lattice

3 mismatch.4 Another important property of nanowires is the ability to further shrink dimensions of heterostructure devices while still maintaining feasibility of integration with conventional technology and also pushing forward towards new device structures.

Nanowires have been used to realize the advancement in a range of classical device structures from the practical devices like photodetectors5 and light-emitting diodes

(LEDs)6 to cutting edge and fascinating nano-devices like the nanowire laser7,8 and single electron transistors.9 The realm of possibilities is still being unlocked with every new investigation of this unique structure.

§1.2 – Methods of Nanowire Synthesis: Bottom-Up Versus Top-Down

There are two kinds of approaches to nanowire synthesis, “top-down” or “bottom- up.” They are distinguishable by their starting point with respect to the end result. A top- down approach is in essence like carving a statue out of a block of marble. While a poorly thought out display piece, nanowires synthesized this way benefit from the pre- existing mastery of bulk and film growth for high quality starting material. The challenge is in the post-growth processing to create an environment on the scale of electron scattering lengths or wave functions. The difficulties begin when we consider the lithography techniques required to create features on the tens of nanometers scale. A pseudo-one-dimensional structure such as that shown in Figure 1, which is created by growth on a cleaved quantum well edge, is not ideal for integration into a circuit. For the study of electronic behavior in 1D systems, devices such as this are excellent and benefit from the development of highly pure film growth techniques. However, for integration in

4 a circuit, it would be far more convenient if the dimensions of the nanowire were defined by the boundaries of the material.

The bottom-up approach seeks to create nanowires via chemical and thermodynamic processes using a template or growth along a preferred direction. In general terms, the bottom up approach to nanowires involves spatially anisotropic crystallization of a material at highly localized and well separated positions on a substrate to “grow” individual nanowires like blades of grass. The process mimics a natural crystallization of solids by enforcing the necessary thermodynamic or chemical conditions. Growth from vapor phase precursors is by far the dominant approach in semiconductor nanowire growth. This is largely because vapor phase techniques

5 dominate semiconductor thin film growth. Growing nanowires from vapor on a substrate via masking10 or seed particles are efficient ways to grow a few to a few hundred thousand nanowires in a single growth.

Figure 2: InP nanowires grown on Si by masking with a SiO2 layer from Mohan et al.11

In this work we will look extensively at bottom-up growth via the use of a gold nanoparticle as a seed for growth. This approach is popular due to the promise demonstrated in the works out of the Lieber Group12 and the deeper focus on the synthesis process notably by the Samuelson group13 and the Jagadish group at the

Australian National University.14 Selective area growth via masking always requires electron beam lithography, whereas to grow nanowires by seed particle the surface can be prepared without the need for lithography. Both growth techniques can and have been done in all the major epitaxial techniques e.g. molecular beam epitaxy, magnetron sputtering, pulsed laser deposition. We concentrate exclusively on growth by metal-

6 organic chemical vapor deposition, a leading technique for growing III-V epitaxial semiconductor films and nanowires.

§1.3 – III-V Semiconductors & Properties

Ashcroft and Mermin define a semiconductor as: “Solids that are insulators at T =

0, but whose energy gaps are of such a size that thermal excitation can lead to observable conductivity at temperatures below the melting point…”15 In practice this includes band gaps over a broad range such as gallium nitride with a band gap of 3.503 eV at 1.6K and indium arsenide with a band gap of 0.418 eV at 4.2K.16 Broadly speaking, research in semiconductor fabrication is primarily focused on the ability to grow uniform films or nanostructures, tightly control their doping and other properties (e.g. strain), and to combine different materials in compounds or layered structures. This is all for the sake of studying and controlling the electron behavior in the material. Semiconductors are often referred to by the groups in the periodic table of their constituent parts. For example

GaAs is a III-V semiconductor, ZnO is a II-VI, and Ge is a group IV. The most commonly discussed classes are group IV, III-V, and II-VI semiconductors.

The success of molecular beam epitaxy and other vapor phase growth methods has armed scientists with semiconductor materials of sufficient quality to push the boundaries of our fundamental understanding and find new directions for application. As chip features continue to shrink towards impracticality new paradigms of information technology are inevitable and semiconductor nanowires provide a promising pathway toward the future of information devices.

III-V and II-VI semiconductors have received much attention due to the ability to control their behavior though means outside of doping. They can be manipulated by varying the relative amounts of group III (II) and group V (VI) elements in binary (AB), ternary (AxB1-xC) and quaternary (AxB1-xCyD1-y) compounds.

Figure 3: Plot of lattice constant versus band gap for III-V ternaries.17

Figure 3 is a plot of band gap versus lattice constant for III-V ternaries.17 This plot is a standard reference for growth of III-V materials. Scientists have utilized the continuum of lattice constants and band gaps to integrate different materials for new or improved devices for industrial application or to create material systems to study 8 fundamental behaviors. The development of materials in one and zero dimensions has opened up further possibilities to integrate band gap engineering with novel device geometries and even unique material properties found in nanostructures.

In this dissertation we take an in-depth look at the growth and characterization of gallium arsenide and aluminum gallium arsenide in the GaAs/AlGaAs core/shell nanowire structure as well as a “nanowire film” geometry where vertical GaAs nanowires are embedded in a conformal layer of AlGaAs. While GaAs is the main material of interest, AlGaAs plays a vital role in both structures. It is therefore important to understand their properties and relationship to each other.

§1.4 – GaAs/AlGaAs Core/Shell Nanowires

Gallium arsenide and aluminum gallium arsenide enjoy a very complimentary relationship. Because of their very similar lattice constants, identical crystal structure, and different band gaps they are able to be combined to form structures like quantum wells used for devices like LEDs and lasers as well as in fundamental research. This is due to the development in film epitaxy aimed at perfecting the individual material quality and the techniques to accomplish important features such as abrupt interfaces. This relationship has been extended into the nanowire geometry in a very similar way.

GaAs is a III-V semiconductor compound with a zinc blende crystal structure, as shown in Figure 4. Zinc blende is a special case of the diamond structure where there is more than one atom in the unit cell. It is a cubic structure that can be described as two

1 1 1 interpenetrating face centered cubic sublattices of Ga and As offset by ( , , ) in units of 4 4 4

9 the lattice constant, a = 5.65325 Å. The Ga and As atoms are tetrahedrally-coordinated with covalent bonds.

Figure 4: Unit cell of GaAs. Ga atoms are grey and As atoms are blue.

AlAs is isostructural with GaAs with a slightly larger lattice parameter, a = 5.660

16 Å. The AlxGa1-xAs structure simply replaces gallium sites with aluminum in the crystal linearly with x. As with many ternary III-V compounds, the lattice constant of AlxGa1-

18 xAs follows Vagard’s law. As the concentration of aluminum increases, the lattice parameter increases linearly towards that of AlAs. Adachi gives the equation for the

18 lattice constant of AlxGa1-xAs.

푎(푥) = 5.6533 Å + 0.0078 푥 Å ( 1 )

Since the maximum percent difference in lattice constant is only 0.12%, GaAs and AlxGa1-xAs can be grown on top of one another with little difficulty due to nearly perfect lattice match. Thus in the nanowire geometry, an AlxGa1-xAs shell will not 10 introduce a significant strain on the GaAs core potentially altering its optical and electronic behavior.

Gallium arsenide is a direct gap semiconductor with a band gap of 1.519 eV at 0K and 1.424 eV at 300K in the bulk.16 The earliest accurate calculation of the band structure was done in the late 70s and is widely referenced today.19 The result of that calculation is the band structure in Figure 5(a).

(a) (b)

Figure 5: (a) Band structure of gallium arsenide calculated by Chelikowsky and Cohen in 1976.19 (b) Detailed view of the GaAs band edge adapted from Blakemore.20

The band structure of AlxGa1-xAs on the other hand depends on the composition.

AlAs is an indirect semiconductor with an indirect gap of 2.229 eV at 4K (2.153 eV at

16 300K) and a direct gap energy of 3.13 eV at 4K (3.03 eV at 300K). AlxGa1-xAs is therefore a direct gap for 0 < 푥 ≲ 0.35 and is fully indirect for 푥 ≳ 0.47. It should be noted that the transition from direct to indirect with increasing x is not sharp and the

11 literature varies in identifying at which composition it is no longer a direct gap material.

It is given in Chuang’s Physics of Optoelectronic Devices that the direct gap energy for x

< 0.4 can be written as21

퐸(훤) = 1.424 푒푉 + 1.247 푒푉 ∙ 푥 푎푡 300퐾 ( 2 )

퐸(훤) = 1.519 푒푉 + 1.447 푒푉 ∙ 푥 − 0.15 푒푉 ∙ 푥2 푎푡 0퐾 ( 3 )

Hence, AlxGa1-xAs can be grown radially on the nanowire as a shell to reduce

GaAs surface states without providing an alternate pathway for conduction electrons to travel in which is similar to a GaAs/AlGaAs quantum well.

Since GaAs is a direct gap semiconductor, it is widely used in optoelectronic applications. Hence, high efficiency of photon absorption and emission is important. For

GaAs nanowires, efficiency in photoemission is markedly lower compared to the bulk

푆 2 due to the high surface-to-volume ratio of the structure, = . This is because the 푉 휋푟 surface states of gallium arsenide mostly recombine non-radiatively.22,23 It has been shown that these surface states can be passivated if the nanowire is coated in a layer of

AlGaAs.10,23–25

To summarize the relationship between the GaAs core and its AlxGa1-xAs shell: the shell protects the GaAs core from oxidation and reduces the GaAs surface states which have poor optical efficiency. The electrons remain confined to the GaAs core due to the higher AlxGa1-xAs band gap and because of the lattice match, the interface is minimally strained. Additionally, the higher AlxGa1-xAs band gap implies that photoemission from the GaAs core will be well separated in energy from that of

AlxGa1-xAs. Thus optical studies of GaAs nanowires are altogether improved by the addition of the AlxGa1-xAs shell.

§1.5 – Spintronics, Semiconductors, and Nanowires

§1.5.1 – Overview of Spintronics

Spintronics utilizes the spin degree of freedom of the electron for processing, transferring, and storing information.26 In terms of its scientific interest, Bader and Parkin quite accurately explain that field of spintronics is “…an applied discipline that is so forward-looking that much of the research that supports it is at the center of basic condensed matter physics.”27

A seminal experimental result in metal spin-based devices is the discovery of giant magnetoresistance in France in 1988 and separately in Germany in 1989. For this

Albert Fert and Peter Grünberg were jointly awarded the Nobel Prize in Physics in

2007.28 In the paper out of France, Baibich et al. found that the resistance measured between two anti-aligned Fe layers separated by a 9 Å layer of Cr was reduced by a factor of 2 when the Fe layers where brought into parallel magnetic alignment.29 Soon after this discovery, the metal spacers were replaced with an insulating layer to create the magnetic tunnel junction (MTJ). The MTJ is now used in read heads for information storage.27,30

Today, spintronics encompasses a number of scientific disciplines and involves any material where the electron spin can be or could be controlled. Broadly speaking

13 spintronics can be divided into three areas: metal and magnetic material, semiconductor, and organic spintronics. Within these categories lie a multitude of different materials and applications.

§1.5.2 – Semiconductor Based Spintronics

Semiconductor spintronics holds the promise for natural integration with the current semiconductor based paradigm of the information industry.27 There have been a number of important achievements in semiconductor spintronics that have happened more recently. This is due in large part to the advances in high quality semiconductor growth not available to test the theories of the time. In 1971 the spin hall effect was predicted by D’yakonov and Perel’31 but was not experimentally observed until 2004 by

32 Kato et al. who observed this effect in GaAs and In0.07Ga0.93As grown by molecular beam epitaxy. The results of this work are ground breaking and the experiment itself demonstrates the use of a powerful method of detecting spin states in semiconductors via the Faraday Effect. We will discuss this method in more detail as it is a motivational force behind this work.

In 1990 Datta and Das laid out the theoretical basis for a device that modulates current by controlling the phase of spin polarized electrons across a narrow gap semiconductor.33 The proposed device creates a spin polarized current using an iron contact emitter (source) that traverses a channel to another iron contact reader (drain) on the other end. If the generated spin is pointed in the same direction as the reader, then it will enter the metal, if not it will scatter off of the interface.26,33 Their theoretical device

14 has since been called the Datta-Das field-effect transistor (FET) or a spin-FET and it has heavily influenced research in the physics of spin in semiconductors for spintronics. In the paper, Datta and Das propose using a narrow gap semiconductor like InGaAs as the channel because of the high energy splitting between spin-up and spin-down carriers in zero applied magnetic field. In this case the splitting is predominantly due to the standing electric field that arises in an inversion asymmetric crystal that is constrained in one dimension.34 A gate voltage applied to channel would modulate the precession frequency of the spins and therefore modulate the current by control of the direction of the spins with respect to the direction of the reader.26

A necessary characteristic of the semiconductor used as the channel is a long spin lifetime and a long coherence length as they propagate through the channel. Ideally this should occur at room temperature for any realistic application. Spin lifetimes of nanoseconds at room temperature were measured in two dimensional electron gas

(2DEG) systems by Kikkawa et al. in 1997.35 From the same group came the observation of coherent lateral drag of over 100 μm in GaAs.36 In comparison to dimensions of Si based processors, this is quite long. This observation was published in 1999, 10 years before Intel® launched processors based on 45 nm transistors.37

A natural follow up question is how long is the spin lifetime in a semiconductor nanowire? A self-assembled nanowire is by nature a narrow channel of controllable length. If the spin lifetime and coherence lengths are long, they would be valuable assets in spintronics devices. To achieve such characteristics, the material must be of uniform crystal structure and doping must be tightly controlled. Most questions regarding spin

15 lifetimes and relaxation mechanisms in semiconductor nanowires remain open. While large advances have been made in nanowire growth, more is needed to achieve the amount of control over material quality necessary for studying the electron spin dynamics.

§1.6 – Optical Orientation and Measurement of Electron Spin in Semiconductors

The theory of optical orientation describes the absorption process in a direct gap semiconductor such as GaAs that results in ensembles of spin polarized conduction band electrons. It is half of the theoretical foundation for the spin creation/measurement technique, Time Resolved Faraday Rotation. The theoretical framework leading to the process of optical orientation constructed over many years by Soviet and American physicists starting in the early 1950s into the mid-1970s. D.D. Awschalom and his group at UC Santa Barbara combined this idea with the theory of Faraday rotation of the polarization of light in a magnetic field to create a totally laser-based pump/probe technique that creates spin polarized states in semiconductors and tracks their time evolution through the relaxation time of the excited ensemble.

As with any study of this nature, the scientific goals and requirements of the measurement dictate the demands on the material quality and therefore directly influence the material fabrication. It is therefore instructive to the grower to enumerate the needs of the system based on the physics that lies behind the measurement.

§1.6.1 – Optical Orientation

In a direct gap semiconductor such as gallium arsenide, the dipole selection rules for transitions into the conduction band are such that excitation by photons of a uniform polarization will bias the final states to polarize the majority of spin states in the same direction with respect to an applied magnetic field, H, that breaks spatial symmetry. This excited state of spin polarized electrons corresponds to a net magnetization that precesses around H and decays in time. The net magnetization can be measured via the Faraday

Effect; a process where the polarization of a classical, linearly polarized light wave is rotated as it passes through a medium of non-zero net magnetic moment. The magnitude of the rotation is directly proportional to both the strength of the magnetization as well as the path length through the magnetized region.

The wave functions of valence band electrons in gallium arsenide close to k = 0

[Γ-point, see Figure 5 and Figure 6(a)] are p-wave Bloch functions. With spin orbit coupling, the total angular momentum of electrons in the system is given by J = L + S with conserved quantum number mj, the projection of the angular momentum along the

+z direction. The conduction band wave functions close to k = 0 can be expressed as s- wave Block functions also with total angular momentum J. For the p-wave electrons

(valence band) the magnitude of total angular momentum is then Jz = 3/2 while for the s-

38 wave (conduction band) electrons it is Jz = 1/2. Transitions across the band gap caused by absorption of a photon are represented by the dipole operator that corresponds to the photon with angular momentum S, Sx ±ħ, and frequency given by the correspondence

38 principle, ω = (Ef – Ei)/ħ. By conservation of angular momentum, these transitions must

17 satisfy the selection rule Δmj = ±1. If the photons are restricted to only one helicity, the calculation of transition probability amplitudes shows that transition into the spin up and spin down states in the conduction band are not equal (Figure 6). For σ+ helicity photons

- transition rates are 3 times higher into the mj = -1/2 state and for σ helicity photons the

26 transition rate is 3 times higher into mj = +1/2.

§1.6.2 – Time Resolved Faraday Rotation

Time resolved Faraday Rotation (TRFR) is a powerful all-optical, pump-probe technique for detecting the time evolution of spin polarized electrons in the conduction band of a semiconductor. TRFR in the Voigt geometry applies an external magnetic field,

H, in the plane of the surface normal. The pump is a circularly polarized laser pulse with wave vector, k, parallel to the surface normal (perpendicular to H). The conduction band 18 is hence pumped with electrons in a coherent superposition of basis states defined by the magnetic field that is perpendicular to the electrons’ total angular momentum. In the semi-classical picture of the electron spin, the magnetic moments precesses around H at the Larmor frequency.

Following the pump beam at a controlled time delay is a linearly polarized probe beam. The polarization of the probe is rotated as a function of the direction and magnitude of the total magnetization of the material and thus by varying the delay between pump and probe, the time evolution of the induced magnetization (direction and magnitude) can be measured by tracking this rotation.39 When detected in transmission, this rotation is called the Faraday Effect. The probe beam can also be measured in reflection, where its polarization will similarly be rotated proportional to the strength of the magnetic field of the surface. This analogous rotation is called the Kerr Effect and hence the measurement is called time resolved Kerr rotation (TRKR). This is particularly useful for extending this technique into systems that have small or zero transmission coefficients.

Figure 7: Schematic depiction of the TRFR measurement in the Voigt geometry from Young et al. (2002).40

The coherent electron states evolve in time with a relative phase according to their energy difference, ΔEt/ħ. For an ensemble of electrons, the individual moments constructively add to the net magnetization caused by the optical absorption. In time, the single electron states will become decoherent and the spin polarization between electrons will dephase. The net magnetization will hence decay and the corresponding decay

* 39 constant, T2 , is a quantity measurable by TRFR/TRKR.

This measurement has been successfully carried out in the aforementioned spin lifetime studies on GaAs/AlGaAs 2DEG structures by Kikkawa et al35,36 and numerous other studies of direct gap semiconductors and magnetically doped semiconductors for spintronics applications. As it does not require a large sample volume, TRKR is a good technique for measuring spin lifetimes in GaAs/AlGaAs nanowires. Because the measurement relies on the optical orientation phenomenon, the band structure of the material must be free of defect states within the band gap and the valence and conduction 20 band states must be sharp. A common contaminant in materials grown by trimethyl precursors in MOCVD is carbon. In GaAs, carbon incorporates into the lattice as an acceptor, opening up defect states within the band gap. Such states do not necessarily have well understood energies or transitions rules and they are not well controlled. Thus for a measurement to be made, characterization of the band gap by photoluminescence must demonstrate the absence of defect states within the band gap.

Chapter 2: Nanowire Growth by Metal-Organic Chemical Vapor Deposition

Nanowires have been grown by virtually every vapor phase method in existence.

In our own group we have previously had success with pulsed laser deposition41 and

UHV magnetron sputtering.42 Growth by molecular beam epitaxy (MBE) is a logical choice given its major contribution to growth of research-grade thin films. However, over the years metal-organic chemical vapor deposition (MOCVD) has become one of the most prominent methods of growing nanowires. This is in part due to its superior ability to grow phosphorous and arsenic containing compounds and much higher growth rates.

In the Yang group we began our investigation of nanowire growth by MOCVD shortly after the completion of its installation at Nanotech West in 2009. The details of MOCVD growth in the thin film and nanowire modes as well as the benefits and drawbacks will be discussed at length below. Our goal is to contribute to the ongoing efforts to understand and control nanowire growth. The fundamental truth for all condensed matter physics is that a well-controlled material system is absolutely everything.

§2.1 – Overview of Metal Organic Chemical Vapor Deposition

Metal Organic Chemical Vapor Deposition (MOCVD) is an epitaxial growth technique developed for the deposition of semiconducting thin films. It is the result of experimentation during the 1960s in growth techniques beyond liquid phase and chloride vapor-phase epitaxy. Researchers were looking toward the next step in semiconductor manufacturing when shrinking Si features beings to slow down and ultimately come to an end. The viability of a growth technique for cutting edge semiconductor application and research is measured by its ability to achieve high purity material, sharp interfaces, compositional control in the growth direction, and introduce and control dopants. This holds true for both thin film and nanostructure geometries. Finally, especially in the industrial setting, it needs to be low cost and high throughput. Super lattice structures are very difficult to grow with liquid phase epitaxy.43 These materials generally do not melt congruently, hence the solid formed is not in equilibrium with a liquid of the same composition.43 For applications of ternaries and quaternaries such as band gap engineering, epitaxial growth is the only method of materials synthesis. In academic research, molecular beam epitaxy (MBE) has long been the gold-standard for semiconductor thin film growth. This is an ultra-high vacuum (UHV) epitaxial growth technique where pure source materials traverse a mean free path larger than the chamber to deposit on a heated substrate. Relative to MOCVD, MBE matured quickly to achieve high purity materials with atomically sharp interfaces. The MBE technique opened up vast areas of research in semiconductors not previously accessible due to limitations in the materials synthesis.

Despite its revolutionary role in materials research, MBE does have limitations.

Primarily, it does not scale up well for industrial manufacturing. UHV must be broken often to replenish the source material and so a heavy time cost is incurred. Typically it is more difficult to grow compound semiconductors that contain phosphorous or arsenic by

MBE. It is because MOCVD held promise to be scaled up to industrial sizes and it can grow most III-V and II-VI compounds that research persisted and initial limitations such as carbon contamination and interface abruptness were overcome. To date, MOCVD growth has become an industry standard for high throughput III-V epitaxy and has grown in popularity for cutting edge research. MOCVD can produce high quality quantum well structures with interface regions less than 17Å and atomically sharp interfaces have been demonstrated.44

Figure 8: Aixtron CCSTM 3×2” MOCVD. Not depicted: gas cabinets for the group V precursors and exhaust scrubber. Image by Aixtron SE.45

Conceptually, MOCVD is a straightforward design. Gas phase precursors are piped into to a sealed reactor. The precursor gases flow over the hot growth surface where they pyrolize and the desired elements deposit onto the growth surface. Everything that does not deposit onto the surface or reactor walls is piped out the exhaust line.

Typical growth surface temperature range is 500 – 800°C. There are two main lines into the reactor, the metal organic line (MO line) and the hydride line, named for the precursors they carry into the reactor. A carrier gas, usually H2 or N2, flows through the two lines to the reactor, delivering the precursors that are injected into the stream. These two lines finally meet just before the gases enter the reactor.

Metal Organics Reactor

Hydrides Vent

Figure 9: The basic operation of an MOCVD System.

In contrast to MBE, there is no UHV and precursors can be replenished by simply exchanging a bubbler or pressurized gas bottle. Standard operating pressures for an

MOCVD reactor is on the order of 100 mbar (~75 torr) compared to MBE or magnetron

25 sputtering which generally operate in the 10-10 mbar (~7.5 10-11 torr) range. This higher base pressure is an operational advantage over a UHV technique. Two drawbacks are that the precursors are expensive and, especially in the case of the group V hydrides, they can be lethal.46,47 Special precautions must be taken to prevent accidental deaths associated with equipment malfunction or operator mistake. In all systems the precursor gas lines are double walled and designed to shut down all gas flow if an abrupt change in pressure is detected in either the inner or outer line. It is also standard the that room that houses the tool be fit with gas and pressures sensors that also shut down all gas flow in the event a precursor is detected above safe threshold or a sudden change in room pressure occurs.

The precursors are seen to be so lethal that in 2009 the Institute for Materials Research

Nanotech West facility, where the MOCVD is housed, was contacted by the United

States Department of Homeland Security to develop a plan for reducing their stores of group III precursors and limiting future storage volumes.

Due to early market availability, the two most common metal organic group III precursors are trimethyl or triethyl alkyls which are stored in liquid form at cool temperatures (4 – 22°C). The trimethyl molecules are favored because of their high vapor pressure and greater stability compared to the triethyl sources.46 Group V and Si precursors are typically hydrides stored at room temperature in pressurized bottles. The notable exception is Sb which is often stored as trimethylantimony. The metal organic precursors are stored in bottles called bubblers (see Figure 10) where the liquid and vapor phases of the molecule are in equilibrium. The carrier gas enters the inlet and passes into the precursor liquid. What leaves the bubbler is then a mixture of carrier gas and

26 precursor vapor. The vapor pressure in the bubbler is held constant by submerging it in a bath of water that is regulated to remain at a stable temperature (typically 4 – 20°C).

Thus, the flux of both metal organic and hydride precursors into the reactor is controlled via flow rate through the bubbler or from the bottle and into the main metal organic carrier line. Typically the carrier gas for work in materials development is H2 or N2, for work in the development of MOCVD itself, other carrier gases include He and deuterium

48 (D2).

Figure 10: MOCVD bubbler immersed in water for temperature stabilization.

The amount of precursor in the carrier gas ranges between 10 – 1000 μmol/min for the growths done in this work. Despite this, it is sometimes the case that a lower concentration of precursor material is necessary. When looking to further reduce the partial pressure of either source in the flow lines below the reliability of the mass flow controllers, a dilution source is employed. Here the precursor gas leaving the bubbler is

27 mixed with H2 in a separate line before entering the flow line to further dilute the gas.

The flow of this diluted mixture is then regulated into the main carrier line.

§2.1.1 – Solid Layer Formation in MOCVD

There are a number of different ways that a growing layer can form on a surface during a vapor phase growth (see Figure 11). For instance the gas could deposit layer-by- layer (Frank-van der Merwe) stacking in the growth direction or it could form islands that grow up and into each other (Volmer-Weber).49 For thin film growth of GaAs or similar

III-V materials by MOCVD, a step-flow growth mode often yields the highest quality layers. In this mode a layer of material deposits such that the growth front advances laterally across the wafer before the next layer is stacked on top.

Figure 11: Different behaviors in epitaxial film growth. Adapted from Kuech 2010.49

To promote the step-flow mode for deposition in the <001> direction, the substrate surface is cut slightly off the (001) axis by a few degrees. Hence an off-axis wafer is one where the surface normal and the normal to the indicated place are at some specified angle with each other. On-axis wafers often results in layers with “orange peel” morphology that is difficult to reduce via chamber conditions. Observation of this morphology motivated the move to growth on 6° GaAs (100) surfaces to avoid the morphology seen in Figure 12. In this figure we see the orange peel morphology of the top GaAs layer of an etch stop structure. The entire structure is

GaAs/InGaP/GaAs(buffer)/GaAs(100). Other growths have shown this morphology is propagated through all layers starting with the buffer layer.

§2.2 – Thermodynamics of Metal Organic Chemical Vapor Deposition

Before discussing nanowire growth by MOCVD and the vapor-liquid-solid (VLS) mechanism, it is useful to address the fundamentals of MOCVD thin film growth as well 29 as some of the relevant processes that will later contribute to the understanding of nanowire growth by MOCVD. Fifty years after its inception, the kinetics and thermodynamics of MOCVD thin film growth are not fully understood. The process is then further complicated by the introduction of a tiny, unassuming metal particle.

Because thin film epitaxy predates nanowire growth, there exists a large wealth of established knowledge. Researchers studying the VLS growth of nanowires commonly reference and discuss the results of MOCVD epitaxy to understand their own results as will be done here. Much exciting work in MOCVD has been done to advance the state of the art of materials synthesis and many interesting and promising materials systems are still to come.

The fundamental driving force of MOCVD is thermodynamic. Ultimately, by introducing gas phase precursors over a solid surface at high temperatures, the system is being forced out of thermodynamic equilibrium. This imbalance drives epitaxy to restore equilibrium.50 The equilibrium condition obtained by minimizing the Gibbs free energy is given by

1 휇표 + 휇표 = 휇표 . 퐺푎 4 퐴푠 퐺푎퐴푠 ( 4 ) For an ideal gas mixture50 (the gallium and arsenic in the reactor) the chemical potential can be written as:

표 푝푖 휇𝑖 = 휇𝑖 (푇) + 푅푇푙푛 [ 표]. ( 5 ) 푝푖

o o Where μi is the chemical potential and pi is the partial pressure of gas i at equilibrium. Correspondingly, for a solid (the GaAs surface) the chemical potential can be written as 30

표 휇𝑖 = 휇𝑖 (푇) + 푅푇푙푛[푎𝑖]. ( 6 )

51 Where 푎𝑖 = 푥𝑖훾𝑖 is the activity of a non-ideal solid. For this analysis, we will set the activity to unity and assume an ideal solid arguing that there is always plenty of solid phase GaAs available for the reaction.52 Finally we obtain the thermodynamic driving force of epitaxy50

1⁄4 푝 (푝 ) 표 표 1 표 퐺푎 퐴푠4 Δ휇 = 휇퐺푎퐴푠 − 휇퐺푎 − 휇퐴푠 = 푅푇푙푛 [ 1⁄4]. ( 7 ) 4 푝표 (푝표 ) 퐺푎 퐴푠4

Here, Δμ is the Gibbs free energy of the reaction and some algebra will recover the familiar form of the law of mass action. During growth, restoration of equilibrium is never achieved because precursors are continuously being replenished in the reactor hence material continues to deposit on the advancing surface. What we can immediately see is that the thermodynamic driving force and therefore the growth rate are proportional to the partial pressure of the reactants at the growth surface. For the growth of GaAs and other III-V compounds, the group V partial pressure is much larger than that of group III.

In the case of GaAs, at temperatures starting around 400°C and above, As will begin to sublimate out of the GaAs surface leaving behind areas of continuous Ga on the surface.

This process cannot in general be reversed by the subsequent addition of As into the environment and as a result the growth surface is destroyed. To preserve the integrity of the GaAs growth surface, the input partial pressure of AsH3 is always much higher than that of TMGa. What this means is that at the growth surface, the partial pressure of Ga is nearly depleted, while the partial pressure of As does not change and remains high. The implication is that the growth rate down to zero is directly proportional to the flux of Ga

31 to the growth surface. This has been observed experimentally for GaAs and other III-V materials both thin film and nanowire alike.53,54

As a brief but notable detour in the discussion, this As sublimation process is exploited to the advantage of the grower to improve the state of the surface before growth. Before epitaxy on a GaAs surface, the substrate will sit at high temperature (600

– 700°C ) for some prescribed amount of time under an arsine. During this time, the expulsion of As out of the surface will contribute to the removal of AsOx that will have formed on the surface during storage or exposure during processing. In its place, As from the vapor will adsorb onto the surface occupying the former oxide sites thereby improving the quality of the growth surface.

§2.3 – Carbon Contamination and Precursor Decomposition

Unintentional carbon doping was important obstacle to be overcome during the early days of MOCVD and it remains important in nanowire growth. In GaAs, carbon enters the lattice on the arsenic sites and is therefore an acceptor impurity, positively doping the material.55,56 Considering the chemical makeup of the group III precursors, it is not surprising that carbon is a source of contamination. TMGa – (CH3)3Ga – releases its 3 methane hydrocarbon radicals, CH3, when it fully pyrolizes.

If the radical occupies its remaining bond(s) with atomic hydrogen in the vapor to form CH4 (methane) the product will be harmlessly carried away from the growth surface in the exhaust. If it does not, the CHx radical can adsorb onto the growth surface and incorporate into the lattice resulting in carbon contamination.55 The decomposition process is written as59

3 (CH ) Ga + H → Ga + 3CH ( 8 ) 3 3 2 2 4

It has been found through experimentation with TMGa and AsH3 precursors that formation of CH4 occurs predominantly from the combination with H from the AsH3. The ideal reaction is given by equation 9.

(CH3)3Ga + AsH3 → GaAs + 3CH4 ( 9 )

This was verified by Kuech and Veuhoff by comparing growths using H2 and He carrier gases that were otherwise identical.55 They saw no change in C doping between

33 the two growths, indicating that formation of methane was not detectably increased by the presence of the H2 carrier as one might have naturally expected. Further evidence of their conclusion is in the increased carbon content of the GaAs growth at lower group V partial pressures on most of the growth surfaces where the group III partial pressure is held constant.

It has been found that the hydrogenolysis of the three methyl groups occurs incrementally as temperature is increased.59 In their 1986 paper on the decomposition of

TMGa and AsH3, DenBaars et al. found that TMGa loses the first two methyl groups at low temperatures (350 – 450°C) and does not lose the third until after 460°C. Further analysis of the precursor decomposition as a function of temperature and partial pressures reveals that while the decomposition of TMGa as a function of temperature is largely independent of the presence of arsine or a GaAs wafer, the same cannot be said for AsH3.

The decomposition temperature of arsine is strongly influenced by the partial pressure of

TMGa in the environment as well as the presence of a GaAs wafer (see Figure 14).59 This means that at the growth surface the V/III ratio – the ratio of group V to group III precursors – is a function of temperature, group III partial pressure, and the growth surface present.

59 Figure 14: Composition of TMGa and AsH3 from DenBaars et al. The right plot is the relative concentration of AsH3 in H2 in the reactor.

The importance of the surface orientation has been observed by a number of groups. Kuech and Veuhoff observed greater carbon incorporation on As terminated

GaAs (111) surfaces than on the Ga terminated (111) surfaces.55 It has been shown elsewhere that incorporation of carbon and other dopants is dependent on the growth surface.60–62 The AlGaAs shell that is deposited on the nanowire sidewall is growing on 6 different surface orientations, none of which are the standared (001) or (100) growth surface. Instead they are the six permutations of {1̅10}. Historically most work has been done on GaAs (001), GaAs (111)B, and GaAs (111)A. The growth on the nanowire sidewall surface is therefore a potential area of further exploration in terms of contribution to carbon doping.

§2.4 – Pyrometry in MOCVD Reactors

The monitoring of surface temperature during MOCVD growth is actually more challenging than one might expect. We will see that a standard thermocouple in our 35 reactor cannot provide accurate enough temperature data. Instead MOCVD reactors such as ours make heavy use of optical pyrometers. These instruments monitor the radiation emitted by various materials to determine their temperature based on Plank’s theory of black body (and grey body) radiation.

The energy density of radiated energy from a black body as a function of wavelength and temperature is famously given by Plank’s law

8휋ℎ 푑휆 휌(휆, 푇)푑휆 = ∙ . ( 10 ) 휆5 푒푐ℎ⁄휆푘퐵푇 − 1 The power emitted by the black body though some hole that is gathered by a detector per unit area per solid angle is given by the spectral radiance63

푐1퐿 푐 퐿푏(휆, 푇)푑휆 = 푑휆 = 휌(휆, 푇)푑휆 . ( 11 ) 휆5(푒푐ℎ⁄휆푘퐵푇 − 1) 4휋

2 Here C1L is the first radiation constant equal to 2c h and C2 is the second radiation constant equal to ch/kB from Plank’s law. As with the energy density given by Plank’s law, the spectra radiance as a function of wavelength do not cross for different temperatures. Hence, if the value of the radiance is known, the equation can be inverted to find the temperature of the black body. In fact, this what an optical pyrometer does.

Figure 15: Radiance versus emission wavelength given by Equation 10.

For maximum sensitivity to change in temperature the pyrometer should monitor a wavelength that yields a large percent change in radiance for a given percent change in temperature. The fractional change in radiance for a given change in temperature can be found by taking the partial derivative with respect to temperature

휕퐿 푐 푒푐2⁄휆푇 휕푇 푏 2 ( 12 ) = 푐 ⁄휆푇 . 퐿푏 휆푇 푒 2 − 1 푇 This equation implies that the highest sensitivity to changes in temperature will be found at short wavelengths. For a real surface and a real detector a balance must be made between high sensitivity and a noise floor that is below the desired sensitivity.63

In practice the object in question is always a grey body, hence the energy density and therefore the radiance are modified by the object’s emissivity

퐿(휆, 휃, 휙, 휎, 푇) = 퐿푏(휆, 푇)휖(휆, 휃, 휙, 휎, 푇) . ( 13 ) For a body in thermal equilibrium with its environment, the energy absorbed and the energy emitted must be identical. It is prior knowledge of the material’s emissivity 37 which allows the optical pyrometers employed in our systems to measure the temperature of various surfaces.

In many cases, the pyrometer used to measure the temperature of a wafer reflects light at two wavelengths (632 nm and 950 nm) off of the wafer surface in order to measure its temperature. The measurement is made assuming that the wafer is specular

(smooth), opaque, and that the reflectance does not depend on the azimuthal angle.63

These assumptions result in the following relationship between the wafer emissivity and the specular reflectance off of the surface63

휖(휆, 휃, 휙, 휎, 푇) = 훼(휆, 휃, 휙, 휎, 푇) = 1 − 푅(휆, 휃, 휎, 푇). ( 14 )

The expression for radiance of a smooth wafer is therefore given by

퐿(휆, 휃, 휙, 휎, 푇) = [1 − 푅(휆, 휃, 휎, 푇)]퐿푏(휆, 푇). ( 15 ) It is then this expression that can be inverted to find the wafer’s temperature using the power of the received wavelengths reflected off of the wafer surface.

Chapter 3: Equipment, Reactor Conditions, and In Situ Temperature Techniques for the

Experimental Growth of GaAs Nanowires by MOCVD

In the first part of this chapter we will describe the details of our MOCVD system used go grown all our core/shell and coalesced GaAs/AlGaAs nanowire structures. In the second part of the chapter we will devote significant detail to the in situ monitoring of surface temperatures during MOCVD growth and the technique we developed to overcome the technical limitations encountered while operating the system beyond its intended design.

§3.1 – Close Coupled Showerhead MOCVD Reactor and Standard Reactor Conditions

All growths were done in an Aixtron Close Coupled ShowerheadTM 3×2” reactor identical to the one depicted in Figure 8. The reactor (Figure 16 and Figure 17) is located inside a 5N nitrogen glove box. This protects the reactor and susceptor from oxygen and water contamination and provides another line of defense between the operator and precursors and their byproducts.

In this design, the gases enter the top of the reactor through a quartz shower head with ~100 inlets only a few millimeters in diameter. The gas flows down and over the growth surface and out the exhaust line at the bottom of the reactor (Figure 17). Both the reactor walls and the quartz showerhead are water cooled to remain around 55°C during operation. The shower head sits about 2” from the surface of the susceptor resulting in a relatively large temperature gradient pointing down from the showerhead.

Metal Organic Hydride Line Line

~50 C

600 C – 800 C

Vent

Figure 17: Schematic of gas flow in our Aixtron CCS MOCVD reactor.

In both the MO and the hydride lines (Figure 9) runs 3000 sccm of gas for a total of 6000 sccm (6 liters/min) into the reactor during growth. Earlier we noted that the flow of precursors into the system is on the order of 1 to 200 μmol/min. A 6 sccm flow of pure

H2 corresponds to ~0.268 moles/min. The precursor materials thus make up only a small fraction of the total gas flow.

Operating pressure inside the reactor is a constant 150 mbar for every growth.

During a growth different precursors can be switched on and off at any time. Dummy lines from the H2 source are employed to alleviate the pressure waves that would be created in the lines from simply turning a source on or off.

A standard practice in most vapor phase epitaxy procedures for III-V films and nanostructures is the constant overpressure of the group V at high temperatures. For our

GaAs based growths, a flow of no less than 200 sccm of AsH3 was kept in the chamber at

41 all temperatures above 300°C during the ramp up and ramp down of the system and all times in between.

§3.2 – Surface Temperature and Pyrometry in MOCVD

It will be demonstrated in Chapter 6 that the temperature at which the nanowire growth takes place is a key parameter to be controlled. We begin by discussing the various temperature readings in our system and how surface temperatures are controlled and monitored for standard thin film growth. We then discuss the obstacles of temperature control in nanowire growth and the procedure we developed during this project to regain control the surface temperature during nanowire growth.

Figure 18: The Yang group susceptor is 5” in diameter and has multiple pockets for growth. In this picture the pockets are occupied by dummy wafers that protect them when not in use.

In our reactor, substrates are placed in pockets on a SiC coated, graphite plate called a susceptor (Figure 18). The susceptor sits over three independently controlled,

42 concentric, resistive heating coils with a thermocouple positioned in the center of the assembly (Figure 19). The temperature inside the reactor is adjusted and maintained by set point control of the thermocouple under the susceptor. The thermocouple does not make contact with the susceptor because the susceptor spins during growth and would eventually drill a hole into the expensive graphite plate. Instead it sits about 2 mm below the bottom of the susceptor. Thus, the thermocouple is not measuring the temperature of the susceptor, but rather the temperature just under it. For growth, it is the temperature on the other side of the susceptor where the substrates sit that is of the utmost importance.

As discussed in Chapter 2, pyrometers are used to measure and monitor the surface temperature in the reactor.

Top Down View Cut-Away Side View

Susceptor Substrate Zone C

Zone B

Zone A Heaters TC TC

Figure 19: Schematic of the heater/susceptor assembly in our MOCVD. The susceptor is depicted as orange.

Our system is fit with two separate pyrometers that access the susceptor and substrates surfaces through O-ring sealed quartz windows situated on the top of the 43 reactor lid. The first is a free stand piece of equipment calibrated only to the emissivity of graphite while the second is equipped with software to communicate with the MOCVD computer so that it can provide real time temperature data for up to seven different materials during a growth.

§3.2.1 – Temperature Balancing

We use the free standing pyrometer to create a flat temperature profile across the susceptor surface in the range from 550 to 800°C. To do this the susceptor is rotated over the heaters at 50 RPM to ensure that the profile is azimuthally symmetric. Through three

O-ring sealed quartz ports positioned on the top of the susceptor, the pyrometer is used to map the radial temperature profile. The radial profile is adjusted through control of the power delivered to each of the three heater zones (coils). At each temperature set point, the system provides some amount of power to the heaters. The percentage of total power delivered to each zone is controlled by the operator. In this way, the radial temperature profile on the surface can be made uniform to within 1°C at the given thermocouple set point. For each susceptor the zone powers and pyrometer readings must be logged with their corresponding thermocouple set points. The temperature profile during a growth is thus inferred from the balancing data. As a result it is extremely important that the physical condition of the susceptor and reactor be closely monitored and zealously protected by the graduate student.

C B A

Figure 20: Thee ports for monitoring susceptor surface temperature in heater zones A, B, and C.

§3.2.2 – In Situ Pyrometry under Standard Growth Conditions

The second pyrometer used during growth sits over a separate port in the reactor lid. It is run by software that is able to track the rotation of the susceptor. And can assume different materials for any interval of the susceptor rotation. Through the port, it measures the intensity of radiation coming off the surface at 632 nm and 950 nm. Given the material being observed and the detected intensities, the computer references a preloaded grey body curve of that material to determine its current temperature. It is able to correct for the emissivity of several standard III-V materials in the standard deposition temperature range. The software outputs the surface temperature and the thermocouple set point data in real time during a growth (see Figure 22). If the surface temperature needs to be adjusted, the operator can change the thermocouple set point and the heater powers based on their tables until the pyrometer reads in the desired range.

This direct measurement of the surface temperature is crucial because the measured temperature of two nominally identical surfaces under nominally identical

45 conditions can be offset by 5°C (up to 10°C in extreme cases) due to drifts in the system.

The pyrometer output surface temperature data down to 480°C, after which it no longer responds. For the typical MOCVD growth range, this lower limit is sufficient. For nanowire growth, the majority of growth temperatures remain below that limit. To accomplish a consistent monitoring of the surface temperature during nanowire growth, we have to take a few extra steps to overcome this range limitation as well as some inherent limitations a nanowire surface presents to a pyrometry measurement.

§3.3 – Surface Pyrometry for Nanowire Growth

There are several problems with monitoring the surfaces prepared for nanowire growth by optical pyrometry. The first, being the aforementioned lower limit of the pyrometer. Most nanowire growths are typically done in the 390 – 500°C range, depending on the method and the desired structure. A second problem is that the emissivity of a GaAs wafer with Au particles on the surface is not known and could change depending on the preparation of the surface (e.g. surface density or particle diameter). This means that even if we were to try to track the surface temperature during baking, it would be difficult to guarantee self-consistent measurements of the surface where the temperature, uncorrected for emissivity, is replicated from growth to growth.

Finally, even in the absence of these two obstacles, the radiation signal from the growing surface would be lost during nanowire growth. As the nanowire layer grows it becomes increasingly more absorptive of radiation. Macroscopically, wires grown on a (111)B surface which are vertical on the surface appear black while wires grown on a (100)

46 surface which are at 35.3° from the surface appear dark brown. Hence the reflectance signal is likely to become diffuse and lower in intensity until it is eventually completely gone. Loss of intensity is seen in thin film growths where the deposited layer becomes rough over time.

Figure 21: Cartoon representation of substrate placement for temperature monitoring during nanowire growth.

To circumvent this limitation, a piece of Si is placed in the reactor with the prepared substrates as depicted in Figure 21. Its temperature is monitored and used to calculate a thermocouple set point for the targeted surface temperature. Each core/shell growth consists of 4 temperature steps: (1) the bake at 590°C, (2) the calibration step at

510°C, (3) the nanowire growth at the targeted temperature, and (4) the shell deposition at 650°C. The real time temperature data for a full core/shell growth is shown in Figure

22 where the four temperature steps are labeled.

During the first two steps, the temperature readings of the system from the thermocouple and of the Si from the pyrometer are recorded. The data are plotted against each other and a line is fit between the two points. The equation of this line is then used to calculate the thermocouple set point for the targeted surface temperature that lies outside the limits of the pyrometer. This procedure is represented by the blue data in

Figure 23. The solid blue squares are the data taken from the first two temperature steps.

The open blue square is calculated using the line defined by the previous two points. The desired surface temperature is input to find the correct thermocouple set point. Following the nanowire growth, the system is then ramped up to the shell growth temperature at

650°C. The pyrometer reading of the Si surface is not trustworthy at this point due to the low quality GaAs deposition on the surface from the previous step. Yet, this does not

48 pose a problem in terms of consistency of the growth conditions for the AlGaAs shell. At a deposition temperature of 650°C AlGaAs (and GaAs) films are comfortably in a mass transport limited growth regime and are not strongly sensitive to temperature changes by up to about 25°C.

It has been empirically shown that in this system, the relationship between the thermocouple data and pyrometer data is linear and that correcting for the system drift amounts to finding the slope and vertical intercept for the current growth. Using the data collected during several temperature balances of the bare susceptor, we verify that this linear relationship is valid from 800°C down to 400°C. Each of the four data points used in the calculation represents an average temperature in the same time interval. The data used to find an average are taken in a time interval where the thermocouple reading is the 49 most stable. The standard deviation for both data points is less than 1°C in both directions. Using basic error propagation analysis and the standard deviation values from the collected data, the error on the calculated surface temperatures is typically around

±1°C.

Chapter 4: Vapor-Liquid-Solid Growth and III-V Nanowire Synthesis

§4.1 – An Overview of the Vapor-Liquid-Solid Mechanism

The vapor-liquid-solid mechanism is the most popular approach to bottom up nanowire synthesis in recent years. The oft quoted and commonly accepted overview of how the mechanism works to result in low dimensional semiconductor growth is based on the explanation given by Wagner and Ellis in 1964 concerning the growth of “Si whiskers” from SiCl4 vapor through gold droplets. Small gold beads are placed on a Si

{111} surface and heated to 950°C. At the elevated temperatures, the gold particle and the silicon at the substrate surface diffuse to form a Au-Si alloy. As seen in Figure 24, the

Au-Si alloy melts at much lower temperatures than either pure gold or silicon.

Au-Si Phase Diagram

liquid

T (K) T alloy (l) + Si (s)

alloy (l) + Au (s)

solubility gap

Au xSi Si

SiCl4 precursor is introduced into the growth chamber and silicon that has broken away from its carriers is further absorbed into the Au-Si melt. If the temperature of the system is low enough, the system is pushed into the area labeled “alloy (l) + Si (s)” in the phase diagram of Figure 24. Silicon therefore nucleates out of the droplet and onto the surface forming a solid monolayer of crystalline Si, pushing the Au-Si droplet up the same amount. As this process continues, a column of crystalline silicon forms underneath the alloy, bounded by the radius of the droplet.65 The crux of this mechanism is that the

Au-Si liquid can be pushed into the region forming solid silicon under conditions where the silicon in the vapor cannot incorporate into the substrate surface. Therefore the growth is restricted to only the droplets.

(a) 430 C (b) 430 C (c) 430 C (d) TMGa 650 C TMAl AsH TMGa 3 AsH3 TMGa AsH Ga 3 AsH3 AsH3 AsH3 As Ga As As Ga Al

While the details of this process are understood to be different for III-V semiconductors – possibly even among different III-Vs – it is currently common to extrapolate this explanation to all semiconductors grown by a method where a metal

(almost always Au) is used as a local sink for precursors to grow semiconductor nanowires.66,67 It is conventionally stated that the growth happens similarly to the growth of silicon by a gold particle.68 In parallel with the explanation of Si growth by Wagner and Ellis, when the Au-Ga-As alloy becomes saturated at fixed temperature, it enters a phase of liquid alloy + solids GaAs forming a monolayer of solid GaAs at the droplet- substrate interface and pushing the droplet up. Typically, GaAs nanowires grown in

MOCVD are grow in the temperature range of 390°C to 500°C while GaAs films are generally deposited between 600°C and 800°C. As in the case of silicon, the growth through the metal droplet takes place at temperatures that are much lower than film deposition. Furthermore, for the majority of growth conditions, GaAs and other III-V nanowires grow in the <111> direction, independent of the substrate orientation.

Around 2004, researchers began to realized that there were significant problems with the assumptions made regarding the VLS growth of III-V nanowires.69,70 In a review of III-V nanowire growth, K. A. Dick points out that no stable liquid Au-III-V ternary or

Au-V binary has been reported above 600°C,67 which is around 100°C hotter than the typical VLS growth window in most vapor phase systems. Dubrovskii et al. point out in their review of nanowire growth by MBE that typical InAs growth takes place between

380 and 430°C yet the minimum eutectic temperature of Au-In is 454.3°C.71 Assuming that there is no liquid Au-Ga-As alloy during GaAs nanowire growth, it has been surmised that there could be a Au-Ga liquid, but how the As incorporates into the solid phase under the droplet is unclear. Perssons et al. suggest that in their growths by chemical beam epitaxy, the As could enter the wire at the seam between the particle and the nanowire70 (see Figure 26(b)). It is interesting to note that in our own experiments we

54 cannot detect the presence of As in the metal bead at the top of the nanowire after growth despite the fact that significant pressures of AsH3 remains in the reactor until the system is cooled to about 300°C.

Another ambiguity with the standard explanation of VLS involves the role of the

Au particle itself. It has been calculated in some experiments that the activation energy of

GaAs nanowire growth in MOCVD is comparable to the growth of GaAs film grown by the same precursors in the same kind of system.54 This suggests that the Au is not acting as a catalyst for precursor reaction to form solid GaAs. K.A. Dick points out in her review, however, that there are a number of groups who have reported lower activation energies. This discrepancy is at present still the subject of research.

While there is much to be learned about the process of III-V nanowire growth by a seed particle, this gap in understanding does not necessarily halt progress in nanowire synthesis. No matter what the state of the particle or the alloy formed, the growth still takes place. So long as the growth can be controlled, advancements in material synthesis can still be made and the precise mechanism at play can be slowly conquered through the hard work of trial and error. For the sake of clarity and consistency with the literature, we will continue to refer to the growth of GaAs nanowires by a gold seed particle as VLS growth.

§4.2 – Parameters in III-V Nanowire Growth by MOCVD

There are many important parameters in growth and many of them can be understood by evaluating their effect on the growth rate. Here we discuss the five most

55 important parameters in III-V nanowire growth. Important to both nanowires and thin films are precursor flux, growth temperature, and V/III ratio. Unique to nanowire growth are catalyst diameter and catalyst spacing. We start with precursor flux as it is the most straightforward.

§4.2.1 – Precursor Flux

Precursor flux is simply the delivery rate of precursor material into the reactor. It is generally true that flux into the reactor is proportional to flux to the growth surface.

The faster the vapor above the surface can be replenished or the more vapor phase elements available to keep the system out of thermodynamic equilibrium the faster the surface or nanowire will grow.72 As III-V compounds are always grown group III deficient in the input vapor (V/III >> 1) this effectively comes down to the flux of group

III precursor into the reactor.72

§4.2.2 – Nanowire Growth Temperature

Growth temperature is not only one of the most important parameters in MOCVD but it can also be one of the most difficult parameters to measure. What is really meant by growth temperature is the temperature at the surface where the phase transition from vapor to solid is taking place. Growth temperature affects the system and the growth rate in a more complex way than flux. There are three main growth modes that are controlled by temperature: the kinetically limited mode, the mass transport limited mode, and the thermodynamically limited mode. In each of these three modes, the growth rate is

56 affected differently by temperature. The figure below is a compilation of data by G.B.

Stringfellow51 from several authors demonstrating this temperature dependence in GaAs thin films. Each author uses the same precursors used here for growth of GaAs/AlGaAs nanowires, TMGa and AsH3.

Figure 27: Temperature dependence of GaAs thin film growth rate in MOCVD complied by G. B. Stringfellow.51

The plot naturally breaks down into three modes. Starting at the right and moving left is the kinetically limited growth mode, the mass transport limited mode, and the thermodynamically limited mode. In the kinetically limited growth mode, the vertical growth rate is directly proportional to the growth temperature. In this temperature regime, the precursors many not be fully pyrolized (see Figure 14) and various surface reaction rates may still be limited.51 The hallmark of the kinetically limited growth mode is a growth rate that is directly proportional to growth temperature. As pyrolysis increases 57 with temperature more atomic species are available to adsorb onto the surface and incorporate into the solid thus increasing the growth rate.

The next section of the plot, where the growth rate shows no dependence on the growth temperature is the mass transport regime. Here all precursors are pyrolized and all processes are being provided sufficient activation energy to take place. As suggested by its name, in this mode the growth rate is limited largely by how quickly nutrients can be delivered to the growing surface. This suggests that the surface is effectively in thermodynamic equilibrium with the vapor. Since this mode has a large range where the precise growth temperature is of little consequence, it is preferable to grow in this mode when possible there by reducing the number of variables and improving repeatability. In this work the AlGaAs used as a shell deposited on the GaAs nanowires were grown in this mode.

For the sake of completeness, we briefly discuss the thermodynamically limited mode, which will not enter into the discussion of nanowire growth. In this mode the growth rate decreases with increasing temperature. Due to the high temperatures, precursors are likely depositing upstream of the growth surface and possibly bonding in the gas phase producing GaAs molecules which do not incorporate into the surface due to the Soret effect.51 These additional reactions rob the system of precursors, effectively decreasing the precursor flux to the surface and therefore lowering the growth rate.

This knowledge of GaAs growth in thin films will later be called upon to understand the growth rate behavior in GaAs nanowires as a function of temperature.

Caution must be taken in the comparison because in the case of VLS grown nanowires,

58 there exist two competing growth modes, the VLS growth through the Au and the surface growth on the sidewalls of the nanowires as well as on the substrate surface.

§4.2.3 – V/III Ratio

The last parameter common to both grow modes is the V/III ratio. This is the ratio of the group V to group III input gas. The precise V/III ratio at the growth surface is sensitive to multiple reactor conditions as seen in Chapter 2 in the discussion of precursor pyrolysis and carbon contamination. If for the given reactor conditions, the V/III is too low, an additional phase of material will result. The group III elements will aggregate on the surface forming balls of metal and nucleate VLS growth on the surface even for conditions targeting epitaxial film growth. The behavior can be seen in Figure 28 where many irregular nanowires have been nucleated on the surface in addition to the wire we have intentionally grown through the Au seed particles. Typically VLS growth in the manner is difficult to control and the resulting structures are of poor crystalline quality.

Hence this is undesirable even in the case when VLS growth is the intended outcome.

Figure 28: The V/III = 6.4 in this growth is too low and so the Ga and In have nucleated unwanted VLS growth all over the surface among the NWs seeded by Au.

At higher V/III ratios which eliminate such VLS growth, a GaAs film grown using TMGa and AsH3 can suffer from carbon acceptor contamination. It was found by

Kuech and Veuhoff that carbon incorporation increased with decreasing V/III ratio.55

This dependence is found in nanowires as well but comes with additional considerations.73 As the V/III is increased, twinning densities along the c-axis decrease but if the V/III is too high, the wires will begin to kink. While reducing the number of twins in the crystal is desirable, kinked wires generally do not have a uniform crystal structure. In earlier discussion of both thermodynamics and flux it was stated that the Ga species is depleted at the growth surface and it is the flux of this species that dictates growth rate given all other conditions are fixed. It is then important to remember that the correct way to change the V/III ratio is by change the flow of the group V precursor into the reactor while the flow of the group III remains constant.

§4.2.4 – Catalyst Spacing

To understand the importance of the Au catalysts spacing, we begin by considering how precursors diffuse into the droplet. The precursors enter the Au in one of three ways: direct impingement into the Au, adsorption onto the substrate diffusing up the sidewall into the Au, and adsorption onto the sidewall diffusing into the Au.

Figure 29: Schematic of TMGa diffusion in VLS growth.

Given that typical growth conditions are such that growth is controlled by the group III precursor/element, it is common to focus on its diffusion in the system and assume the group V is always available.54,74,75 A given precursor has some diffusion length, λi, that often depends on the surface orientation e.g. (111)B on the substrate versus (11̅0) on one of the nanowire walls. Each catalyst collects precursors from a circular area whose radius is defined by the diffusion length λi. If the Au catalysts are arranged such that their collection areas overlap they will compete for precursors since the precursor can diffuse into more than one catalyst.

Figure 30: When the radii defined by the diffusion length of the group III precursor overlap, adjacent nanowires compete for precursors.

§4.2.5 – Seed Particle Diameter

Finally, it has been shown experimentally and modeled theoretically66,75,76, that the nanowire growth rate is dependent on the diameter of the seed particle. Furthermore is has been shown that the functional dependence of the growth rate on seed particle diameter is either proportional or inverse depending on the reactor conditions. While this is not a major topic of discussion in this dissertation, it is important to understand that catalyst diameters cannot simply be exchanged without consequence.

§4.3 – Nanowire Shell Growth

An AlGaAs shell is usually grown on the nanowires to passivate the surface for improved PL efficiency as well as to reduce polarization of incident and out-going radiation. To eliminate GaAs surface states, the shell need only be a few nanometers thick. For the contained studies, a typical “thin shell” is around 15 nm thick (30 nm total diameter increase). A thicker shell can be grown by increasing the growth time. In Figure

31(a) the GaAs directly under the metal droplet at the top of the wire does not extend beyond the diameter of the droplet while it is noticeably larger in (b). In Figure 31 (c) the nanowire has been cross sectioned along the growth axis from the substrate to about 2/3 up the length of the wire. The remaining 1/3 is the complete GaAs/AlGaAs structure with a large ball of AlGaAs at the top. Close inspection at the bottom of the wire reveals the

GaAs core which has a slightly brighter contrast than the surround AlGaAs shell.

(a) (b)

1 μm 300 nm

(c)

2 μm

Figure 31: (a) nanowire with no shell (b) a ~15 nm AlGaAs shell and (c) a thick ~1 μm shell. The nanowire in (c) has been partially cross sectioned and the GaAs core can be seen with brighter contrast at the base of the wire.

For the core growth temperature dependence study, the wires have a ~15 nm shell. The AlGaAs grows faster on the nanowire sidewalls because the reactor temperature is raised to a regime where surface growth will dominate the VLS mode. The

AlGaAs shells are typically grown at 650°C where the epitaxial growth is mass transport 63 limited. This is a standard growth temperature for GaAs or AlGaAs grown on a (100) substrate for our system. A thin shell is deposited by 3 minutes of growth time while a thick shell can be an hour or longer.

§4.4 – Conformal Growth in Narrow Spaces

An interesting and promising sample geometry achieved during this project is the growth of vertical GaAs nanowires embedded in a continuous matrix of AlGaAs. This is the structure in Figure 32(a). There are many examples of similar devices where the vacancy between the nanowires is filled post-growth with spin on glass (SOG) or an e- beam resist. Since SOG must often be cured at high temperatures, it is unattractive for

GaAs which should not be heated past ~400°C without some kind of arsenic overpressure else it will sublime out of the solid. The in situ growth of an epitaxial and conformal matrix of insulating material is advantageous because it ensures the highest quality interface between the nanowire sidewall and the insulator. This structure has been reported in MBE growth,77 but to the best of our knowledge, no such sample structure has been reported as grown by MOCVD.

(a) (b)

Figure 32: 45° tilt view of 500 nm nanowires (a) on GaAs (111)B (b) with a coalesced AlGaAs layer. The red dashed lines indicate the height of the nanowires above the substrate surface.

Initial attempts yielded growths with large voids in the AlGaAs that cannot be tolerated if a metal contact is to be patterned on the surface. Trench filling by MOCVD is a subject that has been addressed in publications largely pertaining to conformal coating of dielectrics or metals on microchip structures in the semiconductor industry. Shrinking, compacted features increased the aspect ratio of the space between features and it was discovered that layers would no longer deposit conformally, forming voids.78,79

1μm 2 μm (a) (b)

Figure 33: Void formation (a) in Kusakabe et a.l78 (b) vertical nanowires.

Under the standard MOCVD deposition conditions used to deposit metals such as tungsten or copper, the precursors’ reactive sticking coefficient was too high.78,79 The reactive sticking coefficient, Sc, is the probability a precursor will react and incorporate after impingement onto the surface.79 Due to insufficient absorption and reemission events79,80 metal would pile up at the top corners of the trenches until they grow into each other before sufficient amounts of precursors could absorb on the surface farther down the trench resulting in voids. This problem seems to be exacerbated in the nanowire case due to the Au catalyst that has been demonstrated here and elsewhere to still be active at high temperatures. Large spheres with radii larger than the nanowire can be seen at the top of virtually every growth with a shell deposition time larger than 5 minutes.

2 μm

Figure 34: AlGaAs growth rate is much higher at the nanowire tip.

Kusakabe et al.78 found that lower growth temperature, lower reactor pressure, and higher vapor pressure of reactants all improved the conformal coverage of trenches

66 with aspect ratios up to 4 with pressure making the biggest impact. Due to limitations in reactor design, the pressure during growth is not able to be lowered enough to warrant trials before other solutions are attempted. Lowering the temperature of the AlGaAs growth could yield new problems with the quality of the AlGaAs. Specifically it must remain more insulating than the nanowires. Further complicating matters is that the growth is epitaxial. In MOCVD GaAs is notoriously difficult to grow in the film geometry on a (111)B surface almost universally forming hillock structures on the surface.81 Figure 35 is an SEM image of a hillock that formed near the edge of a (111)B surface where the Au colloids did not adhere to the surface.

1μm

Figure 35: Hillock of AlGaAs formed on a (111)B surface. The bead at the top is believed to be a sphere or group III metals due to insufficient As pressure.

§4.4.1 – Conformal AlGaAs Growth on Vertical Nanowires

To achieve conformal growth, the aspect ratio of the vacancy between the wires is reduced from approximately 7 down to < 1 by growing wires that are ~500 nm tall at the standard inter-wire spacing of ~650 nm. This was done to avoid changing the inter-wire spacing which can compel the need to change the growth parameters if it is less than the 67 group III diffusion length as is the case for our standard surface density. Based on cross sectional imaging of different deposition times (Figure 36), it is observed that the

AlGaAs grows radially off of the nanowire walls as well as forming hillocks. As the growth continues the hillock grow into one another forming a continuous surface.

(a) (b)

Figure 36: 45° tilted view of (a) GaAs nanowires ~500 nm tall, AlGaAs deposition for (b) 1 min, (c) 10 min, (d), 30 min imaged at 45°. Scale bars are 200 nm (a) & (b) and 400 nm (c) & (d).

This structure offers the ability to take advantage of the patterning techniques developed for thin films. Once the top surface is smoothed, an electrical contact pad can be patterned by photo or electron beam lithography. Contact can be made on the bottom

68 of the wafer and a current can be passed through multiple nanowires in parallel, proportional to the area of the contact. Since the AlGaAs is more resistive than the GaAs nanowires, there exists a range of applied voltage where current will only pass through the wires down to the substrate.

§4.4.2 – Long Term Use for a Coalesced GaAs Nanowire Structure

A long term desire for this structure is to use it as an LED device to measure spin drag down the length of multiple nanowires in parallel. A p-n junction can be made by growing p-type wires on an n-type substrate or vice versa (see Figure 37). At the nanowire-substrate interface, spin polarized carriers would emit photons with angular momentum ±ħω, depending on the orientation of the spins. In an advanced stage of control, the spin coherence length for wires of different diameters could be measured by adjusting the length of the wires either from the growth or processing side. This idea borrows heavily from what has already been done in the thin film geometry.82,83

spin injection layers

n-GaAs NW

i-AlGaAs

p-GaAs(111)B

Figure 37: Spin LED schematic. Spin polarized carriers would recombine at the NW- substrate interface preferentially emitting left or right circularly polarized light depending on the spin orientation of the carriers.

In their 1999 publication in Letters to Nature, Ohno et al.83 measure the spin coherence length of polarized electrons through an insulating GaAs spacer (Figure 38). A p-doped GaMnAs layer injects spin polarized holes into the GaAs layer where they propagate and recombine in an InGaAs QW. The thickness of the GaAs spacers is varied to study the coherence length of the spin polarized holes in the GaAs layer. It is the author’s belief that an identical measurement could be done with a coalesced nanowire sample if the GaAs (111)B substrate were exchanged for GaAs (111)A.

Figure 38: Spin LED used to measure spin coherence length by varying the i-GaAs thickness. From Ohno et al.83

§4.4.3 – Substrate Orientation in Conformal AlGaAs Growth

Mao et al. were the first group to report specular GaAs and AlGaAs by MOCVD on a GaAs (111)A surface as well as a high quality GaAs/AlGaAs quantum well.84 In

2000 Kim et al.85 grew a 10 period InGaAs/GaAs quantum well on both GaAs (111)A and 2° GaAs (100). Off-axis GaAs (100) is well studied growth surface for GaAs based ternaries in MOCVD and is an obvious choice for a standard. In that study the quantum well grown on (111)A has a narrower PL signal than the 2° GaAs (100) by almost 10 70 meV. This indicates that high quality thin films are achievable by MOCVD such that the spin LED structure employed by Ohno et al.83 could be re-imagined by replacing the insulating GaAs spacer with a coalesced GaAs nanowire layer grown on top of an

InGaAs quantum well.

Specular AlGaAs layers have been demonstrated on on-axis GaAs (111)B by

Kato et al.81 at growth temperatures between 875 and 900°C. They further showed that hillocks can be reduced below 850°C by growing on misoriented substrates. Growth temperatures exceeding 800°C tend to cause thermocouple failures in our MOCVD system and a suitable replacement to avoid constant thermocouple exchange (a non- negligible time cost) has not been found thus far. Kato et al. report the formation of macrosteps on surfaces misoriented above 0.2°. Typical tolerances of off-axis wafers are

±0.5°. Off-axis wafers a made by physically cutting the ingot at the specified angle to its growth axis. Increasing the tolerance lowers the manufacturer’s yield driving up the cost per wafer for the consumer. More expensive wafers are a less attractive solution than simply using the (111)A surface that has been shown to produce high quality GaAs and

AlGaAs films.

Chapter 5: Characterization Techniques

As interest in synthesizing nanostructures for novel devices has grown, so too has interest in robust characterization. Techniques using commonly accessible equipment used in traditional bulk or thin film growth such as x-ray diffraction or Raman spectroscopy tend to require larger sample volumes than a typical growth will yield. Hall measurements or even straightforward I-V characterization can manifest themselves quite differently when applying them to nanostructures. For example, a hall measurement on a thin film sample may take hours of preparation to make good electrical contacts.

Conversely, in 2012 the data collected from a hall measurement made on a single nanowire was published in Nature Nanotechnology.86

The three major characterization methods utilized for nanowires in this work are scanning electron microscopy (SEM), photoluminescence spectroscopy (PL), and scanning transmission electron microscopy (STEM). Taken as a whole, the data provided from these three techniques are capable of revealing fundamental structural and optical properties of the nanowires as well as providing feedback for the growth process itself.

Atomic force microscopy (AFM) is a surface imaging technique that employs the use of a nanometer scale tip on the end of a cantilever that deflects in response to the topography of the surface as the cantilever moves in the x-y plane. AFM is capable of resolving changes in surface height on the nanometer scale at 10 nm or less lateral resolution.

Using a specialized experimental setup called conductive AFM (CAFM), the tip can be used to probe the resistance of the surface in addition to the topography of the surface.

This provides an additional map of relative resistance in the x-y plane.

§5.1 – Electron Microscopy

The invention of the electron microscope has opened up our collective eyes to the universe at the nanoscale. Without such capabilities, direct visual evidence of the nanostructure one has grown would not be possible and many fascinating images of successful and unsuccessful growth and processing attempts would go unseen. It is the opinion of many who have spent time working an SEM that it is the failures that often proved the most rewarding images. Of all the skills I’ve acquired during my time as a graduate student, getting to know my way around an SEM has been one of the most enjoyable. I am fortunate to have been afforded the opportunity.

§5.1.1 – Scanning Electron Microscopy

An SEM is an imaging system that takes advantage of the wave-particle theory of matter by using electrons instead of photons in the visible spectrum to image a surface.

Lord Rayleigh’s criterion for the minimum distance between two objects (points) says that if the maximum of one Airy disk coincides with the first minimum of the second

Airy disk, then the two objects are able to be distinguished from each other.87 This is the

Rayleigh criterion.

퐷 0.61 ∙ 휆 푟 = = ( 16 ) 2 휇 푠𝑖푛(훼)

Here r is the resolution limit, D is the width of the Airy disks, μ is the refractive index of the medium, and α is the half angle subtended between the aperture and the object. De Broglie proposed the relationship between wavelength and momentum of a particle that is the foundation for electron microscopy.

ℎ 휆 = ( 17 ) 푝

Hence the wavelength of a particle can be controlled by controlling the momentum of the particle. For accelerating voltages applicable to standard SEM imaging, relativistic corrections are unnecessary. The electron wavelength is thus given in terms of the accelerating voltage by straightforward substitution.

ℎ 휆 = ( 18 ) √2푚푒푒푉

Here the charge is the charge of the electron, e, and V is the accelerating voltage of the electron gun. Typical accelerating voltages for imaging are between 5 and 10 kV corresponding to an electron wavelength of 0.17 to 0.12 Å. Assuming everything else to be equivalent in the Rayleigh criterion, this improves the resolution of the microscope by

4 orders of magnitude.

Electron Gun

Condenser Lens

Objective Lens

Scan Coils

Aperture

Photomultiplier

Figure 39: General construction of an SEM. Reproduction of two figures from Goodhew et al.87

An SEM uses a tungsten filament to produce thermionic electrons. The filament is kept at a large negative potential wither respect to an anode with a small opening. The electrons are hence shot toward the sample surface at voltages ranging between 1 and 30 keV. The electron beam passes through one or more condensers to squeeze its diameter before passing through the final objective lens. The condensers and lens are finely engineered solenoids which produces a standing magnetic field over a short distance that forces the electrons into a tight helix around the optical axis of the microscope. After the objective, the beam is steered in a raster pattern on the surface by scanning coils. At the surface, the beam diameter can be down to 2 – 10 nm. In current SEMs, the scanning coils have been replaced by digital control of the beam position on the sample surface.87

Incident Beam Backscattered X-rays Electrons Secondary Cathodoluminescence Electrons

Sample

Figure 40: Diagram of the surface emissions as a result of the incident electron beam. Reproduction of the figure found in Goodhew et al.87

Breaking the analogue with a light microscope, most of the incident beam does not backscatter elastically off of the surface into a detector. Rather, the beam penetrates the surface in a teardrop shaped region called the interaction volume. Most incident electrons lose their energy and come to rest within the sample. Those electrons that scatter elastically off of the sample are called backscattered electrons. Most of the electrons available for detection are secondary electrons. A secondary electron is any electron that has escaped the sample with an energy ≤ 50 eV.87 Most of the secondary electrons are generated by the incident beam while far fewer are caused by backscattered electrons emerging from some point deeper under the sample surface. Because of the high yield from the incident beam, most secondary electrons originate from a region not much larger than the beam diameter. Compared with backscattered electrons, secondary electrons are more abundant, originate closer to the surface and have a smaller radius of origin. Hence for imaging, SEMs collect secondary electrons.87

Beam

Secondary electrons Backscattered electrons

X-rays

Figure 41: Diagram of the interaction volume of the electron beam with the sample surface. Note that the backscattered electrons originate from deeper underneath the interface than do the secondary electrons. A reproduction of the figure found in Goodhew et al.87

Interaction of the incident electrons with the material can also cause electrons bound to an atom in the sample be removed from its orbital. When the vacancy in the atomic shell is filled by another electron, the excited state of the atom is relaxed via photon emission. If the vacancy is in the valence shell, the photon energy will be in the visible range and this is called cathodoluminescence. If the vacancy is from an inner shell, the vacancy will be filled by an electron from an outer shell and x-ray frequency photons will be released. The analysis of those photons is called energy dispersive x-ray spectroscopy (EDS).87,88 In both cases, there are characteristic emission spectra associated with individual elements that aids the identification of the elemental composition of the sample. EDS is a standard technique in most SEM and STEM microscopes.

Figure 42: Inner shell electrons are displaced by the incident beam and electron from an outer shell takes its place. The energy difference is in the x-ray band and is characteristic of the differences in energy levels of different nuclei. Figure from Williams & Carter.88

We imaged all of our growths with one of two SEMs. The first is an FEI Helios

Nanolab 600 Dual Beam FIB/SEM. The second is a Zeiss Ultra 55 Plus FE-SEM.

Accelerating voltage on the FEI is typically 5 or 10 keV while on the Zeiss it is always kept at 5 keV.

§5.1.2 – Scanning Transmission Electron Microscopy

STEM is an imaging technique where a high energy electron beam is passed through a thin sample called a foil and projected onto a screen. In this geometry, the sample sits in-between the last condenser and the objective lens. After the objective lens, or lenses, sit multiple objective and projector lenses which magnify the image before it hits a screen.

Electron Gun

Condenser Lens(es)

Foil

Objective Lens(es)

Screen

Figure 43: This diagram depicts one condenser lens and one objective/projector lens but current STEM systems commonly have two condensers and 5 objective/projector lenses.87

In STEM the electron beam is accelerated at voltages that range from 60 to 300 keV. At the upper end of the range, relativistic corrections to the wavelength are no longer negligible. From introductory level special relativity, the relation between total energy and momentum is given by

2 2 2 2 퐸 = (푝푐) + (푚푒푐 ) . ( 19 ) And the relation between total energy and the kinetic energy of the electron is given by

2 퐸 = 푇 + 푚푒푐 . ( 20 ) The wavelength is thus given by

ℎ 휆 = . 2 ( 21 ) √ 푒푉 ( 푐 ) + 2푒푉푚푒

Where T has been replaced by eV. The wavelength of an electron accelerated through 300 kV is then 0.020 Å, while the lattice constant of GaAs is 5.65325 Å.

An STEM is used to project real space or reciprocal space image of the crystal structure of the foil onto the detector screen. X-rays produced by the incident beam can also be collected and mapped back to their point of origin for EDS. The real space images yield atomically resolved images of the sample foil and give clear evidence of a variety of crystallographic defects that may exist in the crystal structure. The capability to project the reciprocal lattice by electron diffraction is a very powerful technique. Real space projections of a crystal structure can be quantitatively verified by indexing reciprocal space data. It is a robust way to identify or verify the crystal structure of the foil. Because this technique does not require a large sample volume, it supplants x-ray diffraction spectroscopy for analysis of nanowires is most cases. For the temperature dependence study of this dissertation, electron diffraction patterns were not available, however, because of the beautiful real space images done by our colleagues at CEMAS, we are able to simulate the pattern by taking a Fast Fourier Transform of the images.

All HR-STEM data was taken by Dr. Robert E.A. Williams in our collaboration with he and Professor David McComb. High angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images and EDS maps were acquired

TM TM with their FEI Image-Corrected Titan 60-300 STEM equipped with Super EDX and also their FEI Probe-Corrected TitanTM 80-300 STEM.

§5.2 – Photoluminescence Spectroscopy

The technique of photoluminescence (PL) uses the radiative recombination of excited electron states (or electron-hole pairs) in a semiconductor crystal to probe the nature of the band structure. Photons incident upon a semiconductor surface are absorbed by electrons in the valence band causing them to jump the band gap into the conduction band. The photons can have energies that are several meV below the band gap energy to well above it.89–91 Removal of an electron creates a hole in the valence band that is in many cases bound to the excited electron through coulomb attraction. This electron-hole pair called an exciton. The energy of the photon released when the exciton recombines corresponds to the energy gap between the pair. The spectral distribution of emitted photons due to exciton recombination as a function of incident energy is a sensitive measure of characteristics that perturb the band structure e.g. strain, doping impurities, crystal defects. PL is a heavily utilized characterization technique for nanowires and is crucial to the main results of this work.

Excitons are created by hitting a semiconductor surface with a laser which is usually operated at powers that are non-destructive to the sample. PL is especially useful for characterization of nanowires because it does not require large sample volume. Many measurements have been done on only a single nanowire.23 The selection rules for transitions into the excited states reflect the conservation of crystal momentum of the excited electron.

퐤i − 퐤f = 퐤γ ≅ 0 ( 22 )

Where k is the wave number of the electron and the wave vector of the incident

90,92 photon, kγ, is taken to be zero. Since the absorption processes does not change the electron wave number, absorption of a photon only allows for changes in energy. This corresponds to vertical movement in the band diagram (see Figure 44).90

Figure 44: Diagram of (a) direct gap transition and (b) indirect gap transitions from Y.P. Varshi in Phys. Stat. Sol. (1967).90 Note that a direct transition can occur in indirect gap semiconductors but they are at a higher energy than the band gap.

The electron transitions into the conduction band are classified into two groups: direct transition and indirect transitions (Figure 44 (a) and (b)). In semiconductors such as gallium arsenide, the valence band maxima and conduction band minima lie at the same k. As a result the energy transition can occur by absorption of a photon. In a band diagram, the conduction band minimum is directly over the valence band maximum, hence they are called “direct gap” semiconductors. In the case of gallium arsenide the bands line up on the Γ-point – k = 0 at center of the Brillouin zone. Those whose maxima and minima are separated by some non-zero value of k are then “indirect gap” semiconductors. Transitions in these materials require the creation or destruction of a 82 phonon of energy Eg = ℏΩ in addition to absorption of a photon. The phonon provides the necessary change in crystal momentum such that the electron can occupy a different position in k-space. We will not need to consider indirect semiconductors in any further detail here.

The recombination pathways of interest generally involve photon energies at or just below the band gap. Peaks below the band gap energy often correspond to donor or acceptor states opened up within the band gap due to dopants or defects.

Figure 45: Schematic of possible radiative recombination pathways. (a) Band-to-band, (b) donor impurity to valence band, (c) conduction band to acceptor impurty, and (d) donor- to-acceptor impurities.

The initial and final states involved in exciton formation depend on the energy of the incident photons, 퐸훾 = ℏ휔. In general, photon energies for a standard PL measurement are larger than the band gap. In this case, photoexcited electrons acquire excess energy in the form of kinetic energy. Relaxation back down to the valence band generally occurs in three steps. After coming to thermal equilibrium amongst each other, the electrons then lose energy to the lattice by releasing optical phonons. These two processes combined take place on a time scale that is typically less than 100 ps. The final

83 step is the recombination of the electron-hole pair which takes place on the order of nanoseconds or more. Ideally, this is done by release of a photon, however, acoustic phonons can also be released.91

If non-radiative pathways dominate this last step, PL measurements are unhelpful for characterizing a material beyond the observation that the material has a poor optical response. This can be a symptom of poor material quality or a property of the material. It was mentioned in Chapter 1 that nanowires have a large surface-to-volume ratio and in the case of GaAs, the surface states tend to recombine non-radiatively which significantly lowers their PL efficiency.22,23 Such low signal strength can make characterization difficult especially if the material quality is not optimal which further reduces the PL efficiency.

Figure 46: Photoexcited electrons thermalize with the lattice towards the conduction band minimum before recombining with a corresponding hole. Diagram from Pavesi and Guzzi.91

Typically we have found a band edge transition near 1.513 eV (818 nm) at 4K for our aforementioned nanowires. Pavesi and Guzzi91 list 34 known lines due to point defects or dopants in GaAs. For the purposes of characterizing our nanowires, many of dopants can be ruled out because they have not been introduced into the growth chamber.

Carbon contamination in GaAs (~830 nm or 1.492 eV at 4K) is the classic contaminant in

MOCVD growth and is an important occurrence in GaAs nanowire growth that can hinder their use in optical studies. As shown in Figure 47, the replacement of arsenic sites with carbon results in a shallow donor state inside the band gap, where the energy difference, ΔE ≈ 23 meV, is the binding energy of carbon into the GaAs lattice.91

Figure 47: The PL from this nanowire growth shows a band edge peak at 821 nm as well as a defect peak near 830 nm. The defect is likely carbon.

As with any spectroscopic technique, to precisely interpret a spectrum, one must have complete theory of the transition mechanisms in hand. Broad peaks can be difficult to interpret in detail and are often most easily dealt with by trial and error elimination.

The PL spectra were acquired by Yi-Hsin Chui using an Nd:YVO4 continuous- wave laser, λ = 532 nm, power of 8.1 mW, and 100 μm diameter spot size. The nanowires are measured at 5K using a 0.3 m spectrometer with a LN2-cooled back illuminated CCD detector.

§5.3 – Atomic Force Microscopy

Atomic force microscopy (AFM) is a scanning probe microscopy (SPM) technique that uses a sharp tip on the end of a flexible cantilever to interact with a surface by “touching” it with the tip. The primary motivation for its development was as an alternative method to scanning tunneling microscopy (STM) for generating topographical images of a surface.93 The lateral resolution of an AFM is similar to that of an SEM, but it can more easily resolve sub nanometer topographical features and produce qualitative information (e.g. surface roughness) due to its fundamentally different operating principle. Surface features are resolved by detecting the deflection the cantilever in response to the interaction force between the tip and the surface. An image with three- dimensional information is made by raster scanning the surface in the x-y plane. In practice the tip can touch the surface or it can be brought close to the surface such that the attractive forces between the tip and the surface can be observed though changes in the

86 cantilever oscillation. AFM was used here to characterize the topography and electrical characteristics of the GaAs nanowires buried in a conformal layer of AlGaAs.

Figure 48: Diagram of AFM tip on surface with laser detection from Haugstad 2012.94

The cantilever and tip are often made of Si or Si3N4 and the entire probe is connected to a small chip of the same material roughly 10 mm by 5 mm and 1 or 2 mm thick. The cantilever-tip system is approximated to be a simple harmonic oscillator, modeled as an effective mass on a spring. Thus, the vertical deflection of the cantilever is related to the force applied to the surface by Hooke’s law.

−푭 훥풛 = ( 23 ) 푘 A typical cantilever for surface imaging is the Bruker TESPA. The cantilever is approximately 125 μm long, 40 μm wide, and 4 μm thick. The tip generally stands 10 –

15 μm tall and the radius is on the order of 10 nm. The spring constant is approximately

42 N/m with a natural frequency of 320 kHz.

A standard method to measure the deflection or change in oscillation of the cantilever is by reflecting a laser off the back of the cantilever into a 4 quadrant photodiode (see Figure 48). Before measurement, the laser spot is centered on the detector and the motion of the cantilever is detected by the movement of the spot in the detector.

There are two experimental approaches to imaging surface topography with an

AFM. The first is tapping mode which images the surface by oscillating the tip near the surface and detecting the change in amplitude or phase of the cantilever oscillation. This also goes by the names dynamic, soft-tapping, AC, vibrating, or non-contact mode. The second is a mode where the tip physically touches the surface in some capacity. One such mode is peak force tapping mode where the tip is oscillated by a vertical piezo and hits the surface during every cycle. This mode allows for the simultaneous collection of height and resistance information at each sampled position.

§5.3.1 – Tapping Mode AFM

In tapping mode, the cantilever is driven by a piezoelectric oscillator near (below) its natural frequency and lowered toward the sample surface. The tip does not touch the sample surface, but rather it oscillates very close such that it feels the attractive force between the tip and the surface. An advantage to this imaging mode is that it can be used to measure delicate samples that would otherwise be damaged or distorted by the tip. In the case of hard materials, the tip is not damaged by the harder surface and its lifetime is extended.

Probe Direction

Oscillation amplitude Fly Height

Figure 49: Schematic of tapping mode AFM on a sample surface.

The interaction potential between the tip and the surface can be described by the

Lennard-Jones potential, and the total force on the tip can be described by the sum of the spring force due to the cantilever and the force from the surface potential.94

Figure 50: The Lennard-Jones potential between the tip and the surface as a function of their separation.

The tip is held in the region where it feels the attractive van der Waals potential and the repulsive term of the L-J potential can be ignored. This addition to the restoring force of the cantilever modifies the natural frequency of the system which in turn modifies the amplitude of oscillation as the tip is brought closer to the surface. As the tip scans the surface the fly height (average height above the surface) is adjusted to achieve constant amplitude of oscillation.94

§5.3.2 – Peak Force Tapping Mode AFM

In peak force tapping mode the entire cantilever assembly is oscillated by the z piezo at a much lower frequency than the natural frequency of the cantilever. From some initial height, the base is moved toward the surface until a specified force between the tip and the surface is achieved. Once the set point force is achieved, the base is pulled away from the surface back to the starting height. The amplitude of oscillation is specified and topography of the surface is determined by the change in starting height needed to hit the set point force.

Figure 51: Hysteresis loop in tip deflection as a function of height from the Bruker Corporation.95 Image edited for clarity.

The van der Waals force between the tip and the surface results in a hysteresis loop in the plot of force versus base height. As the tip approaches the surface it will snap down onto the surface (step 2 in Figure 51) causing a sharp drop in force. After the tip has achieved the set point force (step 3) it will experience an attractive force from the surface lowering the detected force on the cantilever below neutral. Eventually the tip will snap off of the surface and the cantilever will return to the neutral position.

Figure 52: Idealized force versus time plot for one cycle. This image is an adaptation of two images from literature provided by the Bruker Corporation95,96 edited for clarity.

Driving the tip assembly below its natural frequency avoids noise from prolonged vibrations in the interval between the snap off and the next snap on. Oscillations can be interpreted as an engage of the surface if they deflect the cantilever at or more than the force set point. In cases where these signals persist, the tip will not be accurately tracking the surface and it can eventually drift away from the surface all together.

§5.3.3 – Conductive AFM

For conductive AFM a conducting tip is used in peak force tapping mode to probe the resistance between the tip and the sample. The sample stage is connected to a variable voltage source (±2V) while the tip is kept at virtual ground. The conducting tip is connected to a current amplifier and is capable of detecting picoamps of current.

Figure 53: Diagram of conductive AFM experimental geometry for the Bruker PeakForce TUNA instrumentation.

Ideally, current will flow to the amplifier for the entire duration the tip is touching the sample. The current per sample point can be displayed as an average over the entire time of contact or only the current at the peak force setpoint. The detected current with respect to the force on the probe is represented in a schematic plot from the Bruker

Corporation below.97 For conductive AFM, the cantilever used was a Bruker SCM-PIT.

This cantilever tends to be longer, ~225 μm, with a natural frequency around 65kHz. For electrical contact with the surface, the entire chip is coated in 20 nm of PtIr.

Figure 54: Relation in time of the tip oscillation, tip force, and detected current. This image is provided in literature published by the Bruker Corporation.97

Some inherent challenges with this technique are contamination or wear of the thin and fragile conductive coating on the tip as well as achieving a low enough resistance to produce a measureable current. This resistance is highly variable as it depends on the contact resistance between the tip and the sample surface. This contact resistance can be increased with the formation of a surface oxide layer or with the introduction of surface contaminants from the environment.

All AFM data was done with a Bruker AXS Dimension Icon Atom/Magnetic

Force Microscope. Topography and electrical data were acquired using a Bruker SCM-

PIT cantilever. This is a Si based cantilever coated with 20 nm of PtIr on either side. The tip is held in a Bruker DTRCH-AM cantilever holder which connects to the proprietary

Bruker current amplifier.

Chapter 6: The Epitaxial Growth of GaAs/AlGaAs Core/Shell Nanowires by Metal-

Organic Chemical Vapor Deposition

In this chapter we discuss the growth and characterization of GaAs/AlGaAs core/shell nanowires and the dependence of the morphology and photoluminescence spectra as a function of core growth temperature. We achieve growth conditions on GaAs

(100) and (111)B oriented surfaces that result in highly uniform photoluminescence spectra for the wires on each wafer. This is seen in our ensemble PL line widths of 12 meV and 10 meV on the two surfaces respectively. These are only a few meV wider than the narrowest single-wire lines seen in the literature. We have further shown that while the axial growth direction and rate are independent of the substrate orientation, the photoluminescence spectra are not. Nanowires grown on a GaAs (100) surface exhibit narrow line widths in a 30°C range while the nanowires grown on a GaAs (111)B surface exhibit the narrowest line width at 430°C.

§6.1 – Structural Properties

The nanowires here are grown on GaAs (100) or GaAs (111)B oriented surfaces.

The substrates in the temperature dependence study were purchased from MTI

Corporation and the substrates used for the coalesced growth in the next chapter were purchased from AXT Inc. Wafers are cleaved from 2” wafers into square pieces before

95 the gold particles are dispersed on the surface. This is so that they fit into our custom susceptor pockets.

Nanowires are seeded by 40 nm Au colloid suspended in deionized water capped with tannic acid purchased from Ted Pella Inc. The average Au particle spacing is about

650 nm on the surface. Particle spacing (surface density) is controlled by the time the colloidal suspension sits on the surface and diluting the colloidal solution from the bottle with deionized water (~18 MΩ). A detailed discussion of the surface preparation process can be found in Appendix A.

Typically wires grown on both GaAs (111)B and GaAs (100) surfaces will grow in the <111> direction. As shown in Figure 55, nanowires grow perpendicular to surface on GaAs (111)B and make a 54.7° angle with the surface normal on GaAs (100). It has been shown that the nanowire growth direction can be controlled with growth conditions,98 here we focus only on <111> oriented nanowires.

(a) (b)

2 μm 2 μm GaAs (111)B GaAs (100)

Figure 55: GaAs/AlGaAs nanowires grown on (a) a (111)B GaAs substrate (b) a (100) GaAs substrate. Note that the growth direction of the wires always <111>. 96

Two series of GaAs/AlGaAs core/shell nanowires were grown to investigate the

dependencies of the nanowire core on the growth conditions. Keeping all other conditions

constant, the core deposition temperature was varied in series between 410 and 472°C.

The first series covers 410 to 450°C and the second covers 435 to 472°C. In each, series

the next growth was started within 15 minutes of unloading the previous run. The

complete list of growth conditions are provided in Table 1 below.

While there are eight growth temperatures for the entire study, there are five

temperature data point points for each series. Two data points of the second series were

intentionally grown as repeats of the first series to check consistency and repeatability in

the system over the course of the week between the two.

Table 1: Full set of growth conditions used in temperature series. The conditions read chronologically left to right.

Bake Conditions GaAs Deposition AlGaAs Deposition

Surface Temp. (°C) 600 Surface Temp. (°C) 410 - 472 Surface Temp. (°C) 650 ± 10 Bake t (min) 15 GaAs Dep. t (min) 15 AlGaAs Dep. t (min) 3 Total Flow (sccm) 6000 V/III 26.1 V/III 49.4 AsH3 (sccm) 23.3 Total Flow (sccm) 6000 Total Flow (sccm) 6000 Au Diam. (nm) 40 TMGa (sccm) 2.15 TMGa (sccm) 6.5 Au:DIW 1:0 AsH3 (sccm) 5 AsH3 (sccm) 40 Substrate GaAs SiH4 (sccm) 0.0129 TMAl (sccm) 16 Orientation (100) & (111)B SiH4 (sccm) 0

§6.1.1 – Nanowire VLS Growth Rate

The axial (VLS) growth rate data for Figure 56 was gathered by measuring thirty or more nanowires for each growth temperature and substrate. The error bars are the standard deviation of the sampled population. Wires that are not straight or are deformed in some other way are intentionally excluded from the data.

Figure 56: Axial growth rate as a function of growth temperature for nanowires grown on a GaAs (100) and GaAs (111)B surface. Error bars not visible are inside the symbol.

At first glance, the curves for the two orientations look similar to the growth rate plotted in Figure 27 on page 57. However, closer inspection which accounts for the VLS growth reveals different behavior. As the core growth temperature is increased from

410°C, the axial growth rate also increases up to around 440°C. The proportional dependence on temperature indicates that the range between 410 and 440°C is the kinetically limited growth mode falling within as subset of the corresponding temperature

98 range in thin film epitaxy. As the precursors pyrolize more efficiently with the increasing temperature, the growth through the droplet increases.

After 440°C the growth rate beings to decrease with no clearly defined mass transport limited region (growth rate independent of the growth temperature). We recall that there are actually two growth modes involved in VLS growth, (1) the axial growth through the metal droplet, and (2) the growth on the nanowire sidewalls and substrate.

Starting at 450°C, the competition between the two existing growth modes become important and their effect can be seen in the VLS (axial) and sidewall (radial) growth rates.

Above 440°C, the surface growth rate on the sidewalls and on the surface becomes increasingly larger compared to the axial growth mode through the droplet. This competition for precursors between the two modes results in shorter wires with thicker sidewalls. Based on the data compiled by Stringfellow,51 the epitaxial growth mode should not level off until around 550°C. While comparing absolute temperatures among different growth systems can be difficult, a temperature discrepancy of 100°C is significant. The hypothesis of competing growth modes is reinforced by the measurement of the average nanowire diameter as a function of temperature plotted in Figure 57. The measurement of diameter is standardized to just under the Au particle at the tip of the wire to reduce the affect tapering would have on the statistics. These data were taken for

(111)B nanowires only. The averages are of a sample population of about 30 wires and the error bars represent the standard deviation.

Figure 57: Average axial growth rate (top) and the average tip diameter (bottom) of GaAs/AlGaAs NWs grown on (111)B. The large standard deviations originate in part from differences in shell thickness and variation in Au particle diameter.

The data in Figure 57 demonstrate a trend of increased nanowire diameter and decreased length as the temperature during the core growth is increased above 440°C. In these data, the average diameter at each temperature value is that of the core plus the shell. The low temperature data shows the largest standard deviation is in thickness is about 15% from the calculated average. This is believed to be due to a combination of the variation in the gold particle diameters as well variations in the shell thickness. This behavior is further visually verified by SEM in Figure 58. In the figure, the images are

100 standardized to the same magnification and those wires grown at the highest temperature,

472°C, are observably thicker and shorter than those grown at 463°C.

Figure 58: Nanowires grown at 436°C (left) are notably longer with a smaller diameter than those grown at 472°C (right).

§6.1.2 – Crystal Structure and Twinning Defects

A property of every GaAs nanowire grown with or without a shell is a hexagonal axial cross section as shown in Figure 59. The sidewall facets of the nanowires are comprised of six {1̅10} planes.

101

150 nm GaAs (111)B

Figure 59: Top down view of nanowires cross sectioned close to their base in the plane of the c-axis. The green dot represents the 40 nm starting diameter of the Au seed.

Using HR-STEM, we evaluate the crystal structure and elemental composition of the nanowires. The nanowires imaged by STEM were grown on the (111)B surface at

436°C. They are brush transferred onto a holey carbon TEM grid before delivery to our colleagues at CEMAS. As in Figure 60 we find that the wires have sections of uniform crystal structure that extend up to 100 nm or more along the growth axis and separated by defect planes along the axis. We believe these defect planes to be twins for the reasons explained below. It is notable that the planes extend all the way out to the edges of the wire suggesting that the shell does in fact grows epitaxially on the nanowire sidewall, even acquiring the same crystalline defects as the core structure while growing in an orthogonal direction.

102

100 nm

Figure 60: HR-STEM of nanowires grown on (111)B at 430°C. Contrast planes can be seen disrupting the crystal structure.

High resolution EDS data collected along with the HR-STEM image in Figure 61 confirms the spatial distribution of elements expected for a GaAs core with an AlGaAs shell. The X-ray intensity from aluminum is expected to be higher at the sides of the nanowire because there is a ~15 nm distance on either edge of the wire where the electron beam will only interact with the AlGaAs shell.

103

Al Ga As Au

100nm

Figure 61: HR-STEM of GaAs NWs with corresponding EDS intensity maps for each of the 4 expected elements.

EDS further shows that the center of the wire appears more aluminum heavy for a length of around 100 nm down the growth axis beginning under the seed particle. In the real space image (left, see also Figure 60), these sections of wire appear to have a very large density of contrast planes and contain a tapered region leading up to the particle.

Based on the higher aluminum content, we believe this region up to the taper is axial growth that takes place during the AlGaAs shell deposition. It is thought that the tapered region up to the metal seed is formed after the group III flow is terminated, but while the reactor remains at high temperature during ramp down from 650°C.

The EDS maps also demonstrate that the final gallium concentration of the metal alloy is below the detection limit of the EDS system, which should be around 0.1%.87

Since gallium is heavier than aluminum – bigger Z – it should be easier to detect which leads us to conclude that there is very little gallium in the seed. Taken with the high concentration of aluminum under the seed particle, this suggests that the aluminum may 104 be displacing much of the Ga in the droplet during the axial AlGaAs growth. Similar behavior has been seen in axial GaAs/InAs or InGaAs heterojunctions where indium easily replaces the gallium in the gold resulting in a sharp interface when switching from

GaAs to InAs and a compositional grading when going from InAs to GaAs.99 It is not clear why there should be such a large discrepancy between the detected amount of Ga in the regions of AlGaAs growth up to the seed particle and in the seed particle itself. We know that a solid AuGaAl compound can be formed at room temperature as there exist four known stable Au92GaxAly (x = 6, 4, or 2) ternary compounds that interestingly are superconducting.100 The reason the majority of the gallium is expelled from the metal seed could be related to the presence of the arsenic and its effect on the solubility of the gallium in the eutectic.99,13

Atomic resolution HR-STEM imaging of a nanowire shows the zinc blende crystal structure looking down one of the {1̅10} zones. In Figure 62, the dimer structure of the GaAs molecule is resolved. The structure exhibits the expected ABCABC stacking sequence. Since the crystal structure is mirrored on either side of the boundary, the defect is believed to be a twin boundary.

105

Figure 62: Atomically resolved image of a twin defect in a nanowire. Inset: the dumbbell structure of the GaAs is resolved.

Comparison of the image with simulated twin boundaries in zinc blende nanowires (Figure 63) further indicates that the observed defect is likely a twin boundary.101

Figure 63: Comparison of a simulated twin in a ZB NW from Wood et al. with our HR- STEM data.

106

In the absence of electron diffraction data, the crystal structure can be further characterized by performing a Fast Fourier Transform (FFT) on the HR-STEM real space image. The image in Figure 62 is cropped on either side of the defect plane such that only one zone axis is transformed per image. Using the open image processing program

ImageJ,102 a bandpass filter is used to smooth the image before transforming it with the built-in FFT function. The indexed results are Figure 64.

Above Below

Figure 64: FFT of the real space HR-STEM images above and below the boundary plane. Indexing is explained in the text.

Using the images at hand it is not possibly to index the FFT patterns with complete experimental certainty. Williams and Carter suggest that a complete indexing of a crystal structure by electron diffraction data should be done by rotating toward two other zones and recording the change in relative lengths and angles to check for consistency with the initial assumption.88 By indexing the pattern using our existing knowledge of the nanowires, we make a strong case that the crystal structure is zinc blende and the defect is a twinning defect, the most commonly reported defect in VLS 107

GaAs nanowire growth. We know the nanowire is growing along the [111] axis. We also know from imaging many brush transferred nanowires that they always lie on one of their sidewalls (e.g. Figure 26), in which case the zone axis is a sidewall axis. We assume the sidewalls are comprised of the family of {1̅10} planes. This assumption is based on comparison of the real space HR-STEM with modeled zinc blende nanowire crystal structure.

In a zinc blende structure, rotation around the [111] axis by (2n+1)60° results in a twin boundary which inverts the ABC stacking sequence. Rotation by (2n)60° returns the structure to its original orientation. Hence a rotation about [111] by 60°, 180°, or 300° will result in one of three symmetric zone axes. On one side of the boundary the zone axis is one of 6 choices and a twin implies that the zone axis on the other side of the boundary must be one of three possible zone axes of different symmetry than the first zone.

In a cubic system, the reciprocal lattice vectors, gi, are parallel to the lattice vectors, ri. Hence we can label the (1̅1̅1̅) and (111) peaks in Figure 64 which are in the direction of the nanowire growth axis in Figure 63. The ambiguity comes down to a negative sign. By calculating the relative lengths and angle between peaks and comparing to the allowed peaks for the 6 possible zone axes, we conclude that the patterns are consistent with assumption made that on the other side of the boundary the structure is

(2n+1)60° rotated and indicate that the boundary plane is a twin. Using only relative angles and lengths, it is not possible to identify the specific zone axis on either side as they are symmetric. Taking the negatives of the assumed (1̅1̅1̅) and (111) directions

108 simply puts a negative sign in front of the remaining peaks with no impact on relative angles or lengths.

§6.2 – Optical Properties

As discussed in Chapter 5, photoluminescence measurement of semiconductor nanowires is an important tool. This characterization is used to assess the quality of the

GaAs core as well as feedback for altering the growth conditions. Before measurement, nanowires from each growth temperature and each substrate are first brush transferred onto a Si or SiO2 surface before being placed in a liquid helium cryostat. Because the

GaAs nanowires are sitting on a GaAs surface (Figure 65) they cannot be measured as- grown because the response from the nanowires cannot be separated from that of the substrate.

Growth Surface

Cleaved Surface

Figure 65: 45° tilted view of a nanowires grown on a (111)B GaAs substrate.

109

§6.2.1 – Photoluminescence as a Function of Core Growth Temperature and Substrate

Orientation

In general, a sharp band edge transition is indicative of good crystalline quality. In bulk GaAs at 4K that peak should be centered around 817 nm. As carbon contamination is a common problem in MOCVD growth we look for a defect peak centered near 830 nm, which corresponds to the expected energy associated with carbon doping in GaAs.

Our PL results show that among the two substrates the optical response of the nanowires grown above 450°C is the same while there is a subtle difference in response between

410 and 430°C, indicating a dependence on the substrate orientation.

We analyze our PL data in two ways. First it is useful to examine the PL spectra of the individual core growth temperatures for the two substrates as they evolve with temperature. This is done in Figure 66. Each individual spectrum in Figure 66 is representative of 5 spectra taken at different spots on the sample. The curves are normalized to the band edge peak if it is present. The PL data at 436°C and 440°C was omitted because they do not add information to the understanding of the trend. They look similar to the low temperature curves shown.

110

Figure 66: The evolution of the NW PL spectra with growth temperature on both substrates. The first and last temperatures are the two extremes in the temperature series.

A sharp band edge peak is observed for samples grown on both substrates at and below 430°C. For wires grown on (100) the sharpest peaks can be found from 410°C up to 430°C centered around 818.8 nm, 819.0 nm, and 819.3nm respectively. In the (111)B

111 wires, there is a discernable low energy shoulder at 410°C . For those wires the sharpest band edge transition occurs at 430°C centered around 818.7 nm.

Figure 66 shows the difference in the spectra between samples with core growth temperatures above and below 450°C. Above 450°C the spectra become dominated by a broad peak centered around 830 nm. This peak is associated with carbon in MOCVD grown GaAs indicating that the wires are comprised of GaAs material that is unintentionally carbon doped. The results of growth rate analysis suggests that above

450°C it is the epitaxial GaAs growth on the sidewalls that is carbon doped instead of the core. Such behavior has been by Joyce et al. who found that reducing the sidewall growth rate – in that case controlled by group III flow instead of growth temperature – is correlated with lower carbon defect peaks in the PL.103 Due to the defect states in the band gap opened up by the carbon dopants, these wires are poor candidates for further study of spin lifetime by TRKR. The TRKR measurement requires a direct transition from the valence to conduction band. The carbon defect states are not well understood in the context of optical orientation nor are they necessarily well controlled.

Heavy carbon contamination in the epitaxial sidewall growth is supported by finding in the literature for growth on a (100) surface. Dapkus et al. found that carrier mobility drastically reduces for growth temperatures at and below 550°C compared to temperatures in the standard 600 – 800°C range.104 The impurity concentrations are not calculated for the lowest growth temperatures, presumably because of the poor morphology reported for those samples means they are not easily comparable to the other, specular growths. Conflicting statements in thin film MOCVD literature stating

112 that carbon contamination is reduced at decreasing growth temperatures are referencing ranges that do not extend much below 600°C as this usually the lower bound of the range of interest.105

It is possible that one of the contributions to carbon contamination is from the fact that the GaAs is growing on a {1̅10} surface, but comparison with thin film literature is difficult. The ideal complementary study would be growth of GaAs on a {1̅10} surface and a (100) surface using trimethylgallium and arsine. It was discussed earlier in Chapter

2 that carbon incorporation as an unintentional dopant is influenced by the growth surface.55 It has been shown elsewhere that incorporation of carbon and other species as intentional dopants is also influenced by the surface orientation.61,106 It therefore seems reasonable that the surface orientation is related to the amount of contamination. In those studies however, the surface normal is tilted from (100) in a direction that does not include a crystal face in the {1̅10} family of planes. Lum et al. have shown that incorporation due to 13C dopants is about the same between (100) and (011) but below the detection limit of 4 × 1014 cm-3 on a (111)A.107* Conclusions cannot be confidently made because the (011) surface not the same as a {1̅10} surfaces and it is well established that carbon contamination from group III precursor radicals is found in (111)A growth surfaces, suggesting that the surface chemistries are not comparable.

* The authors’ original notation has been altered to be consistent with the rest of the document, interpreting their brackets notation to mean a specific crystal face and not the family of planes. 113

§6.2.2 – Peak Ratio and Low Temperature Growth

We further analyze the PL behavior by quantifying the relative strength of the 819 nm band edge peak to the 830 nm defect peak and plotting the result against the growth temperature, Figure 67. This provides a density-independent measure of nanowire optical quality as a function of core growth temperature. This ratio, I819nm/I830nm, is calculated by integrating the intensity at 819 nm and 830 nm with a bandwidth of 4 nm. This is indicated by the grey hatched areas in Figure 66. Using this metric, we are able to observe more subtle behavior in Figure 67 not as readily apparent in the side-by-side comparison of individual spectra made in Figure 66. It is important to note that in this analysis, the intensity at 830 nm is being used as a benchmark for the amount of defect in the sample since transitions to or from defect states within the band gap are expected to release lower energies. Therefore we keep in mind that in the growths below 454°C, the defects are not necessarily the same as those identified in the results above 454°C. The line shapes in Figure 66 suggest that they are not.

114

Figure 67: Peak ratio versus core growth temperature. Substrate surface orientation is indicated in the plot.

This figure shows a very different low temperature trend between the two growth surfaces. By this metric, the optical quality of the samples grown on a (100) substrate appears to be level from 410 to 430°C followed by a steady decline down to 470°C.

Alternatively, the samples grown on a (111)B are clearly optimized at 430°C and suffer a noticeable decrease in peak ratio down to 410°C. As would be expected from comparison of the individual spectra, the high temperature data both follow the same decreasing trend down to 472°C.

The genesis of this difference at low temperatures in Figure 67 is difficult to address with the data in hand. It is often the case that growing a material at a different rate can increase or decrease the formation of crystalline defects or the incorporation of contaminants.55,103 However, our growth rate data shows that growth rates between the two substrates are comparable. Further, the SEM data represented by the images in

Figure 55 demonstrates that the wires all grow in the same direction crystal direction. In

115

MOCVD thin film growth, changing the surface orientation can greatly change the growth parameters for a good quality film.55 This situation is not commonly discussed in the nanowire growth literature because nanowires by VLS are almost universally grown in the < 1̅1̅1̅ > direction, Hiruma et al.68,108 being one of the few exceptions. It is possible that carbon contamination is being influenced by diffusion along the substrate surfaces. While the sidewalls are identical, precursors that hit the substrates first will be diffusing in different environments.

Another possibility is that the starting surface is influencing the twinning behavior. Twinning and the transition from zinc blende to wurtzite crystal structure has been observed to shift the center wavelength of PL spectra.109,110 An increase in twin plane densities in a crystal structure that remains zinc blende over all, might account for the low energy shoulder most prominent at 410°C. However more extensive STEM analysis is required to gather crystallographic information on the (111)B wires at 410°C and 422°C as well as all the (100) wires between 410 and 430°C to draw any meaningful conclusions.

The results of our best nanowire sample are competitive with the current state of the art. For the wires grown at 410°C on (100) the line width is λFWHM ≈ 7 nm or EFWHM ≈

12 meV. For the wires grown at 430°C on (111)B the line width is λFWHM ≈ 5 nm or

EFWHM ≈ 10 meV. These are very close to the range of the best line widths reported in the literature 7 – 10 meV.73,111 It is important to point out that our line widths are based on ensemble data while the referenced line widths are single wire data. It is expected that the

116 convolution of many nanowires will have a broader line width as fluctuations within the sampled population are expected in a total population of ~108 wires.

117

Chapter 7: Surface Processing and Characteristics of GaAs Nanowires in a Continuous

AlGaAs Layer

The surface of the coalesced GaAs/AlGaAs nanowire structure is not ready for

AFM experimentation directly after growth. The AlGaAs is grown over the GaAs cores and the surface is highly irregular thus not good for depositing layers of metal. In principle this could be avoided with reactor conditions that somehow suppress the growth on the nanowire sidewalls and through the gold particle in favor of growth on the substrate surface. Because the precursors are not delivered with a directional bias as in

MBE, this growth mode would need to be accomplished via surface chemistry. There is no literature that indicates such conditions are achievable in MOCVD. Instead, the surface is planarized by mechanical polishing to expose the GaAs nanowire tips and create a smooth surface suitable for patterning. Figure 68 is a top down view of the polished surface of a coalesced nanowire structure.

118

1 μm

Figure 68: Top down SEM image of a polished coalesced GaAs/AlGaAs surface. The sharp lines of contrast are due to multiple hillocks and nanowire walls growing laterally into each other. The irregular shapes are the bottom of valleys remaining from the as grown surface due to insufficient polishing depth.

A cloth made of soft, absorbing material is attached to a gear driven plate that spins horizontally at speeds up to 500 RPM. The cloth is covered in a suspension of particles (grit) which are of uniform size. Typically the grit is silica, alumina, or diamond in ascending order of hardness. The required size and hardness is determined by the material and the desired outcome. Alumina and diamond are typically for removing large amounts of material per unit time while silica is for finishing a surface. For our coalesced

GaAs/AlGaAs structures alumina is too hard and removes material too quickly. Instead

60 nm silica grit is preferred to remove the ~1μm of overgrown AlGaAs.

The sample is adhered to an aluminum plunger by hot wax. The plunger fits into the center of a wider but shorter aluminum cylinder with grooves cut into the bottom to allow for better grit flow to the sample. The sample is held on the polishing wheel until

119 the surface is brought down to the nanowire height. A topographical image by AFM of a polished surface is displayed in Figure 69.

§7.1 – Topographical and Electrical Characteristics of Mechanically Polished Surfaces

To characterize the polished surface and test the electrical behavior of the nanowires, we employ CAFM to simultaneously gather topography and resistance data of in an area of the sample surface. The RMS roughness of the scanned area in Figure 69 is approximately 1.6 nm after flattening the image and excluding the first ~250 nm starting at the far left (~1/8 of the total surface area). The abrupt drop in height is more than likely a void and hence skews the roughness information corresponding to mechanically polishing the surface.

5.4 nm

-5.8 nm

2 μm

Figure 69: Topography of a polished coalesced NW surface by peak force tapping mode. The large dip in height to the left is a void in the AlGaAs layer.

120

The tunneling AFM (TUNA) current map corresponding to the same area in

Figure 69 is shown in Figure 70 where the color scheme has been changed for clarity of interpretation. The dark spots scattered on the surface are regions of lower resistance with a cross sectional area on the order of the expected nanowire cross section ~40 nm.

Furthermore, the density of the dark dots in the image is reasonable given the typical nanowire distribution on a growth surface. Hence, the structure behaves as desired, with the AlGaAs matrix providing an insulating layer between the nanowires. The areas of lighter blue and white are regions of positive detected signal with a maximum of only

0.345 pA and where the color that corresponds to 0 A is indicated by the labeled arrow.

Excluding the areas that correspond to nanowires, the mean current values over the remaining areas are negative valued, ~-0.2 pA, with an RMS current on the order of

0.250 pA. We believe the small positive current is a result of noise due to charging on the surface from contaminants on the surface which result from the polishing procedure.

121

0.345 pA

0 pA

-1.5 pA

2 μm

Figure 70: TUNA current data taken concurrently with the topography data in Figure 69.

Experiments on other samples have not successfully reproduced the behavior of the sample above. Those samples have not produced a current signal above the ~4 pA noise. One possibility is that the resistance on those samples is extremely high. The use of, wax, polishing grit, and tap water for processing those surfaces introduces a large potential for contamination that could increase the total resistance. Under the SEM, surfaces have been imaged to reveal silica that was not removed by sonication after polishing as well as unidentified contrast that indicates some surfaces are not free of contamination. See Figure 71. Unidentified contamination could be drastically increasing the contact resistance. The area probed in Figure 69 and Figure 70 may have simply been free of contamination.

122

Figure 71: 45° view of silica particles remaining on the polished surface after sonication in solvent, scale bar 650 nm. Inset: Top down image of unidentified contaminants with much brighter contrast from the GaAs surface, scale bar 10 μm.

§7.1.1 – AlGaAs Removal by Ion Beam Milling

To avoid the problem of surface contamination as well as to try to gain tighter control of the polishing depth, the procedure was altered such that the as-grown surface is first planarized by mechanical polishing and the remaining AlGaAs is brought down to the nanowire level via use of Ar+ ion bombardment. For this we used the Kurt J. Lesker

Lab-18 Sputtering System. Based on the literature, we calculate that the penetration depth of the Ar+ at the standard Lab-18 bias of 400 kV would be ~12 nm eV at room temperature.112 In this calculation we have ignored that fact that plate to which the sample is fixed heats up to less than 100°C during the milling.

123

Figure 72: 45° tilted views. Hexagonal patterns and pits can be seen all over the surface, scale 400 nm. Inset: Starfish feature with pit believed to be a GaAs NW that has etched faster than its surroundings. Scale 200 nm.

Ion milling the AlGaAs surface resulted in a surface with a much higher RMS roughness than polishing by silica alone. 35 – 50 nm compared to 1 – 5 nm on a surface only polished by 60 nm silica colloids. Further, on the parts of the surface where material is removed down to the nanowire level, hexagonal “starfish” or “wagon wheel” features can been seen throughout the surface. This can be seen clearly in Figure 72. There are two general categories of features; the first is where the center is slightly higher than the surrounding AlGaAs and the second where it is lower. In the wagon wheel, the features that define the spokes and the hub of the wheel are slightly higher than the surrounding surface. They are understood to be the result of different etch rate due to an increase in Al content. As noted earlier, AlGaAs grows off of the sidewalls of the nanowires comprised of the 6 {1̅10} planes as well as growing in faceted hillock structures in flat areas of

GaAs (111)B (refer to §4.4.1 and Figure 36). In island growth modes or the like,

124 interfaces that grow into each other generally have different crystal structure and stoichiometry than the surrounding bulk. In this case the surfaces are in some sense

“growing into each other” in that on each of the six nanowire sidewalls there are six different surfaces growing simultaneously. Where they meet has been found to be aluminum rich in STEM and EDS analysis of GaAs/AlGaAs structures cross sectioned for imaging down the (1̅1̅1̅)/ (111) zone axes.113

(a) (b)

200 nm

(c)

180 nm

Figure 73: (a) High magnification SEM of surface feature resulting from Ar+ ion milling. (b) TEM of a radial GaAs/AlGaAs structure from Zheng et al.113 (c) AFM of surface feature.

Zheng et al. found that at the AlGaAs interfaces, the aluminum concentration increased while the arsenic concentration remained the same. Thus, at these interfaces, the structure should be heavily AlAs. Correspondingly, the resistivity on these lines and on the center features that are similarly raised should have a slightly higher resistivity compared to the rest of the AlGaAs. We further hypothesize that the centers that are

125 lower than the local surface could be the result of hitting a GaAs core and should have a slightly lower resistivity than their surroundings.

To our frustration, investigation of these features by CAFM has been inconclusive. However, recent review of the data indicates that the PtIr coating on the end of the tip in contact with the surface was likely stripped off before the scans on the features took place. Because peak force tapping mode is predicated on making contact with the surface, it can quickly reduce the lifetime of a tip. In particular the conducting tips can be easily stripped of their coating with peak force set points easily within the instrument’s normal operating range. It is therefore necessary to investigate further before definitive conclusions can be made about the electrical properties of the structure or its viability for further development.

§7.1.2 – The Future of GaAs Nanowire-Thin-Film Structures

Our initial CAFM measurements on the GaAs nanowires in an insulating

Al0.2Ga0.8As film demonstrate proof of concept. Future experimentation of these surfaces would likely see more successful measurements with improved measurement technique to reduce the amount of contact with the surface before starting to acquire conductivity data. Furthermore, changes to surface preparation should be made to produce a higher quality surface. The as-grown GaAs/AlGaAs surface should be planarized using a chemical-mechanical polishing (CMP) slurry where chemical reactions at the surface contribute to the removal of material and act to remove damage caused by physically tearing bonds apart with grit.114 This can be acquired commercially or it can be made in-

126 house. Hydrogen peroxide is commonly employed alone or in conjunction with ammonium hydroxide in bench top etching procedures of GaAs and AlGaAs.115 It has become favored over the use of Br2/methanol through demonstration of better final surface roughness.116,117 The final step in the process line should be treatment with concentrated HCl, which is known to improve the electronic and optical properties of

GaAs.118 Cheng et al. at IBM Watson recently demonstrated the use of HCl on processed

GaAs surfaces. Their results indicate that in the absence of an oxidizing agent, HCl can passivate a GaAs surface without etching it, in an manner analogous to the removal of native oxide layers and surface passivation on Si with HF.119

This coalesced GaAs nanowire structure would be an excellent geometry with which to study spin transport in a similar way as Ohno et al. discussed in §4.4.2. The major advantage of this structure is the ease with which hundreds of nanowires could be sampled in parallel. Ensemble measurements are complimentary to single wire data, but can also stand alone. With the mastery of a new nanowire structure, for instance an axial

GaAs/InGaAs heterostructure or an axial GaAs p-n junction, the coalesced structure can easily be produced for parallel ensemble measurements of those new systems.

Furthermore the number of nanowires involved in a measurement can be adjusted simply by changing the size of the top contact. With proper control of the nanowire location, a simple change in contact size can change an ensemble measurement to a single-wire measurement. Lastly, this structure lends itself better to creating TEM foils owing to the continuous layer.

127

Conclusions

We have studied the effect of growth temperature on both the morphology and optical characteristics of GaAs/AlGaAs core/shell nanowires grown on GaAs (100) and

(111)B surfaces by MOCVD. We have demonstrated that the optical characteristics of the core GaAs nanowires are dependent on the growth temperature as well as the substrate orientation. Within the temperature window for nanowire growth, nanowires grown on

(100) substrate exhibit comparably high optical quality from 430°C down to 410°C.

Wires grown on (111)B peak in optical quality at 430°C, with increased defect intensity on either side of growth temperature. The genesis of this difference remains unclear. At high growth temperatures, nanowires on both surfaces exhibit high levels of carbon doping. This conclusion is supported by our structure characterization that shows increased GaAs sidewall growth during the core deposition. Heavy carbon doping in epitaxial GaAs grown at these temperatures by MOCVD is supported by in the literature concerning both thin film GaAs and VLS grown GaAs nanowires. To the best of our knowledge, we are the first group to correlate the PL response of these structures with the core growth temperature.

We have shown through ensemble PL measurements that GaAs/AlGaAs core shell nanowires can be grow with high uniformity in optical characteristics through

128 comparison of our 10-12 meV linewidth to the 7-10 meV linewidths found in the literature for single-wire PL.

STEM data for real space images of the nanowires from a side view of the nanowires and their corresponding FFT patterns confirms that the III-V nanowires have the zinc blende structure with twinning defects normal to the growth axis. The high- resolution spatially resolved EDS data collected via STEM shows that the gallium is displaced from the Au-Al bead at the top of the wire after growth. Given the stability of the ternary Au-Al-Ga compounds at room temperature, this indicates that the arsenic may affect the solubility of the gallium in the presence of another group III atom such as aluminum. This is consistent with literature reporting on III-V axial heterostructures where gallium is easily displaced by an alternate group III while the gallium is slow to replace another group III already alloyed with the gold droplet.

We have shown that lowering the aspect ratio (height:diameter) to ≈0.77 will achieve a conformal, uniform growth of AlGaAs in between vertical standing nanowires.

We maintain that this structure holds great promise for use as ensembles of parallel nanowires devices. The post-growth surface processing must be improved to produce consistent surfaces of much lower resistivity that the majority of the samples probed by conductive AFM.

Lastly we have demonstrated that surface temperatures can be more accurately calculated during nanowire growth by tracking the surface temperature of a calibration sample of known emissivity. Using a piece of silicon, we can extrapolate to the surface temperature at values below the limit of our in situ pyrometer by simply fitting at line

129 between the thermocouple and pyrometer values in more reliable temperature ranges. It has been empirically demonstrated that this relationship is linear in the range of 400 to

700°C on the surface as measured by a pyrometer.

The ability to grow VLS nanowires with uniform optical characteristics across a substrate surface is progress towards creating a GaAs/AlGaAs core/shell nanowire suitable for optical spin lifetime experiments and experimentation with nanowire applications in advanced devices. Not only have we identified growth conditions to achieve high band edge transition efficiency, but we have demonstrated that changes to the growth parameters can create a uniform change in the global growth behavior resulting in a large number of high quality nanowires. The ability to grow a vertical GaAs nanowire in a continuous film of AlGaAs provides the first steps towards an advanced device structure that employs many GaAs nanowires in parallel that have a high quality sidewall and high PL efficiency due to the in situ AlGaAs deposition. Furthermore, this combination of techniques is advantageous because it enables the development of complimentary single-wire and ensemble-wire systems for comparative studies in electronic and spintronic behaviors between the two systems.

130

References

1 Y. Xia, P. Yang, Y. Sun, Y. Wu, B. Mayers, B. Gates, Y. Yin, F. Kim, and H. Yan, Adv. Mater. 15, 353 (2003).

2 A. Yacoby, H. Stormer, N. Wingreen, L. Pfeiffer, K. Baldwin, and K. West, Phys. Rev. Lett. 77, 4612 (1996).

3 E.P.A.M. Bakkers, M.T. Borgström, and M.A. Verheijen, MRS Bull. 32, 117 (2007).

4 O.D.D. Couto, D. Sercombe, J. Puebla, L. Otubo, I.J. Luxmoore, M. Sich, T.J. Elliott, E. a Chekhovich, L.R. Wilson, M.S. Skolnick, H.Y. Liu, and a I. Tartakovskii, Nano Lett. 12, 5269 (2012).

5 H. Pettersson, J. Trägårdh, A.I. Persson, L. Landin, D. Hessman, and L. Samuelson, Nano Lett. 6, 229 (2006).

6 S.D. Carnevale, T.F. Kent, P.J. Phillips, a T.M.G. Sarwar, C. Selcu, R.F. Klie, and R.C. Myers, Nano Lett. 13, 3029 (2013).

7 M.H. Huang, S. Mao, H. Feick, H. Yan, Y. Wu, H. Kind, E. Weber, R. Russo, and P. Yang, Science 292, 1897 (2001).

8 J.C. Johnson, H.-J. Choi, K.P. Knutsen, R.D. Schaller, P. Yang, and R.J. Saykally, Nat. Mater. 1, 106 (2002).

9 H. a Nilsson, T. Duty, S. Abay, C. Wilson, J.B. Wagner, C. Thelander, P. Delsing, and L. Samuelson, Nano Lett. 8, 872 (2008).

10 J. Motohisa, J. Takeda, M. Inari, J. Noborisaka, and T. Fukui, Phys. E Low- Dimensional Syst. Nanostructures 23, 298 (2004).

11 P. Mohan, J. Motohisa, and T. Fukui, Nanotechnology 16, 2903 (2005).

12 X. Duan, J. Wang, and C.M. Lieber, Appl. Phys. Lett. 76, 1116 (2000).

131

13 K.A. Dick, K. Deppert, L.S. Karlsson, L.R. Wallenberg, L. Samuelson, and W. Seifert, Adv. Funct. Mater. 15, 1603 (2005).

14 H.J. Joyce, Q. Gao, H.H. Tan, C. Jagadish, Y. Kim, X. Zhang, Y. Guo, and J. Zou, Nano Lett. 7, 921 (2007).

15 N.W. Ashcroft and N.D. Mermin, in Solid State Phys., 1st ed. (Brooks/Cole, Cengage Learning, Belmont, 1976), pp. 562–587.

16 O. Madelung and R. Poerschke, editors , Semiconductors Group IV Elements and III-V Compounds (Springer-Verlag, Heidelberg, 1991), pp. 1–164.

17 E.F. Schubert, Light-Emitting Diodes (2nd Edition) (Cambridge University Press, 2006).

18 S. Adachi, J. Appl. Phys. 58, R1 (1985).

19 J.R. Chelikowsky and M.L. Cohen, Phys. Rev. B 14, 556 (1976).

20 J.S. Blakemore, J. Appl. Phys. 53, R123 (1982).

21 S.L. Chuang, Physics of Optoelectronic Devices (John Wiley & Sons, Inc, New York, 1995).

22 S. Kodama, S. Koyanagi, T. Hashizume, and H. Hasegawa, J. Vac. Sci. Technol. B 13, 1794 (1995).

23 L. V. Titova, T.B. Hoang, H.E. Jackson, L.M. Smith, J.M. Yarrison-Rice, Y. Kim, H.J. Joyce, H.H. Tan, and C. Jagadish, Appl. Phys. Lett. 89, 173126 (2006).

24 J. Noborisaka, J. Motohisa, S. Hara, and T. Fukui, Appl. Phys. Lett. 87, 093109 (2005).

25 J. Noborisaka, J. Motohisa, and T. Fukui, Appl. Phys. Lett. 86, 213102 (2005).

26 I. Žutić, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, (2004).

27 S.D. Bader and S.S.P. Parkin, Annu. Rev. Condens. Matter Phys. 1, 71 (2010).

28 U. Björksten and E. Huss, Nobel Media AB (2014).

29 M.N. Baibich, J.M. Broto, A. Fert, F.N. Van Dau, F. Petroff, P. Eitenne, G. Creuzet, A. Friederich, and J. Chazelas, Phys. Rev. Lett. 61, 2472 (1988).

30 D.D. Awschalom and M.E. Flatté, Nat. Phys. 3, 153 (2007). 132

31 M.I. D’yakonov and V.I. Perel’, J. Exp. Theor. Phys. 13, 467 (1971).

32 Y.K. Kato, R.C. Myers, A.C. Gossard, and D.D. Awschalom, Science (80-. ). 306, 1910 (2004).

33 S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990).

34 J. Luo, H. Munekata, F. Fang, and P. Stiles, Phys. Rev. B 38, 10142 (1988).

35 J.M. Kikkawa, I.P. Smorchkova, N. Samarth, and D.D. Awschalom, Science (80-. ). 277, 1284 (1997).

36 J.M. Kikkawa and D.D. Awschalom, Nature 397, 139 (1999).

37White Paper Introduction to Intel’s 32nm Process Technology (2010), pp. 1–5.

38 F. Meier and B.P. Zakharchenya, editors , Optical Orientation (North-Holland Physics Publishing, Amsterdam, 1984).

39 D.D. Awschalom and N. Samarth, in Semicond. Spintron. Quantum Comput., edited by D.D. Awschalom, D. Loss, and N. Samarth (Springer-Verlag, Berlin, 2002).

40 D.K. Young, J.A. Gupta, E. Johnston-Halperin, R. Epstein, Y. Kato, and D.D. Awschalom, Semicond. Sci. Technol. 17, 275 (2002).

41 X.W. Zhao, A.J. Hauser, T.R. Lemberger, and F.Y. Yang, Nanotechnology 18, 485608 (2007).

42 X.W. Zhao and F.Y. Yang, J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 26, 675 (2008).

43 G.B. Stringfellow, in Organomet. Vap. Ep. Theory Pract., 1st ed. (Academic Press Inc., San Diego, 1989), pp. 1–14.

44 R.D. Dupuis, P.D. Dapkus, C.M. Garner, C.Y. Su, and W.E. Spicer, Appl. Phys. Lett. 34, 335 (1979).

45(2006).

46 G.B. Stringfellow, in Organomet. Vap. Ep. Theory Pract., 1st ed. (Academic Press Inc., San Diego, 1989), pp. 15–53.

47 G.B. Stringfellow, J. Cryst. Growth 115, 1 (1991).

133

48 G.B. Stringfellow, in Organomet. Vap. Ep. Theory Pract., 1st ed. (Academic Press Inc., San Diego, 1989), pp. 141–208.

49 T.F. Kuech, AIP Conf. Proc. 78, 78 (2010).

50 G.B. Stringfellow, AIP Conf. Proc. 916, 48 (2007).

51 G.B. Stringfellow, J. Cryst. Growth 68, 111 (1984).

52 D. V. Schroeder, in An Introd. to Therm. Phys. (Robin J. Heyden, 2000), pp. 214–215.

53 G.B. Stringfellow, in Organomet. Vap. Ep. Theory Pract., 1st ed. (San Diego, 1989), pp. 55–134.

54 M. Borgström, K. Deppert, L. Samuelson, and W. Seifert, J. Cryst. Growth 260, 18 (2004).

55 T.F. Kuech and E. Veuhoff, J. Cryst. Growth 68, 148 (1984).

56 B.J. Skromme and G.E. Stillman, Phys. Rev. B 29, 1982 (1984).

57 G. Graner, E. Hirota, T. Iijima, K. Kuchitsu, D.A. Ramsay, J. Vogt, and N. Vogt, Molecules Containing Three or Four Carbon Atoms (Springer-Verlag, Berlin/Heidelberg, 2001), p. 1974.

58 G. Graner, E. Hirota, T. Iijima, K. Kuchitsu, D.A. Ramsay, J. Vogt, and N. Vogt, Inorganic Molecules (Springer-Verlag, Berlin/Heidelberg, 1998).

59 S.P. DenBaars, B.Y. Maa, P.D. Dapkus, a. D. Danner, and H.C. Lee, J. Cryst. Growth 77, 188 (1986).

60 J. van de Ven, H.G. Schoot, and L.J. Giling, J. Appl. Phys. 60, 1648 (1986).

61 C. Caneau, R. Bhat, and M.A. Koza, J. Cryst. Growth 118, 467 (1992).

62 K. Tamamura, J. Ogawa, K. Akimoto, Y. Mori, and C. Kojima, Appl. Phys. Lett. 50, 1149 (1987).

63 W.G. Breiland, Reflectance-Correcting Pyrometry in Thin Film Deposition Applications (Albuquerque, 2003), pp. 1–85.

64(2006).

65 R. Wagner and W. Ellis, Appl. Phys. Lett. 4, 89 (1964). 134

66 V. Dubrovskii, G. Cirlin, I. Soshnikov, a. Tonkikh, N. Sibirev, Y. Samsonenko, and V. Ustinov, Phys. Rev. B 71, 205325 (2005).

67 K.A. Dick, Prog. Cryst. Growth Charact. Mater. 54, 138 (2008).

68 K. Hiruma, T. Katsuyama, K. Ogawa, M. Koguchi, H. Kakibayashi, and G.P. Morgan, Appl. Phys. Lett. 59, 431 (1991).

69 K.A. Dick, K. Deppert, T. Mårtensson, B. Mandl, L. Samuelson, and W. Seifert, Nano Lett. 5, 761 (2005).

70 A.I. Persson, M.W. Larsson, S. Stenström, B.J. Ohlsson, L. Samuelson, and L.R. Wallenberg, Nat. Mater. 3, 677 (2004).

71 V.G. Dubrovskii, G.E. Cirlin, and V.M. Ustinov, Semiconductors 43, 1539 (2009).

72 G.B. Stringfellow, in Organomet. Vap. Ep. Theory Pract., 2nd ed. (Academic Press Inc., San Diego, 1999), pp. 17–105.

73 H.J. Joyce, Q. Gao, H.H. Tan, C. Jagadish, Y. Kim, M.A. Fickenscher, S. Perera, T.B. Hoang, L.M. Smith, H.E. Jackson, J.M. Yarrison-Rice, X. Zhang, and J. Zou, Adv. Funct. Mater. 18, 3794 (2008).

74 V. Dubrovskii, N. Sibirev, G. Cirlin, I. Soshnikov, W.H. Chen, R. Larde, E. Cadel, P. Pareige, T. Xu, B. Grandidier, J.-P. Nys, D. Stievenard, M. Moewe, L. Chuang, and C. Chang-Hasnain, Phys. Rev. B 79, 205316 (2009).

75 J. Johansson, C.P.T. Svensson, T. Mårtensson, L. Samuelson, and W. Seifert, J. Phys. Chem. B 109, 13567 (2005).

76 E. Givargizov, J. Cryst. Growth 31, 20 (1975).

77 D. Lucot, F. Jabeen, J.-C. Harmand, G. Patriarche, R. Giraud, G. Faini, and D. Mailly, Appl. Phys. Lett. 98, 142114 (2011).

78 Y. Kusakabe, H. Ohnishi, T. Takahama, Y. Goto, and K. Machida, Appl. Surf. Sci. 71, 763 (1993).

79 A. Burke, G. Braeckelmann, D. Manger, E. Eisenbraun, A.E. Kaloyeros, J.P. McVittie, J. Han, D. Bang, J.F. Loan, and J.J. Sullivan, J. Appl. Phys. 82, 4651 (1997).

80 J.C. Rey, L.-Y. Cheng, J.P. McVittie, and K.C. Saraswat, J. Vac. Sci. Technol. A Vacuum, Surfaces, Film. 9, 1083 (1991).

135

81 K. Kato, Y. Hasumi, A. Kozen, and J. Temmyo, J. Appl. Phys. 65, 1947 (1989).

82 C. Adelmann, X. Lou, J. Strand, C. Palmstrøm, and P.A. Crowell, Phys. Rev. B 71, 121301 (2005).

83 Y. Ohno, D.K. Young, B. Beschoten, F. Matsukura, H. Ohno, and D.D. Awschalom, Nature 402, 790 (1999).

84 E. Mao, S. Dickey, A. Majerfeld, A. Sanz-Hervas, and B.W. Kim, Microelectronics J. 28, 727 (1997).

85 J. Kim, S. Cho, A. Sanz-Hervas, A. Majerfeld, and B.W. Kim, J. Cryst. Growth 221, 525 (2000).

86 K. Storm, F. Halvardsson, M. Heurlin, D. Lindgren, A. Gustafsson, P.M. Wu, B. Monemar, and L. Samuelson, Nat. Nanotechnol. 7, 718 (2012).

87 P.J. Goodhew, J. Humphreys, and R. Beanland, Electron Microscopy and Analysis, 3rd ed. (Taylor & Francis Inc., London, 2001).

88 D.B. Williams and C.B. Carter, Transmission Electron Microscopy A Textbook for Materials Science, 2nd ed. (Springer, New York, 2009), pp. 1–757.

89 R. Elliott, Phys. Rev. 108, 1384 (1957).

90 Y.P. Varshni, Phys. Status Solidi 19, 459 (1967).

91 L. Pavesi and M. Guzzi, J. Appl. Phys. 75, 4779 (1994).

92 C. Kittel, in Introd. to Solid State Phys., 8th ed. (John Wiley & Sons, Inc, 2005), pp. 185–222.

93 P. Eaton and P. West, Atomic Force Microscopy (Oxford University Press, Oxford, 2010).

94 G. Haugstad, Atomic Force Microscopy (A John Wiley & Sons, Inc., Hoboken, 2012).

95 S.B. Kaemmer, Application Note # 133 Introduction to Bruker’s ScanAsyst and PeakForce Tapping AFM Technology (Bruker Nano Surfaces Division, Santa Barbara, 2011), pp. 1–12.

96 B. Pittenger, Atomic Imaging with Peak Force Tapping (Bruker Nano Surfaces Division, Santa Barbara, 2012), pp. 1–22.

136

97 C. Li, S. Minne, B. Pittenger, and A. Mednick, Application Note # 132 Simultaneous Electrical and Mechanical Property Mapping at the Nanoscale with PeakForce TUNA (Bruker Nano Surfaces Division, Santa Barbara, 2011), pp. 1–12.

98 S.A. Fortuna, J. Wen, I.S. Chun, and X. Li, Nano Lett. 8, 4421 (2008).

99 K.A. Dick, J. Bolinsson, B.M. Borg, and J. Johansson, Nano Lett. 12, 3200 (2012).

100 R.. Hoyt and A.. Mota, J. Less Common Met. 62, 183 (1978).

101 E.L. Wood and F. Sansoz, Nanoscale 4, 5268 (2012).

102 W.S. Rasband, (n.d.).

103 H.J. Joyce, Q. Gao, H.H. Tan, C. Jagadish, Y. Kim, M.A. Fickenscher, S. Perera, T.B. Hoang, L.M. Smith, H.E. Jackson, J.M. Yarrison-Rice, X. Zhang, and J. Zou, Nano Lett. 9, 695 (2009).

104 P.D. Dapkus, H.M. Manasevit, K.L. Hess, T.S. Low, and G.E. Stillman, J. Cryst. Growth 55, 10 (1981).

105 G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy: Theory and Practice, 2nd ed. (Academic Press Inc., San Diego, 1999).

106 M. Kondo, C. Anayama, N. Okada, H. Sekiguchi, K. Domen, and T. Tanahashi, J. Appl. Phys. 76, 914 (1994).

107 R.M. Lum, J.K. Klingert, D.W. Kisker, S.M. Abys, and F.A. Stevie, J. Cryst. Growth 93, 120 (1988).

108 K. Hiruma, M. Yazawa, T. Katsuyama, K. Ogawa, K. Haraguchi, M. Koguchi, and H. Kakibayashi, Appl. Phys. Rev. 77, 447 (1995).

109 M. Heiss, S. Conesa-Boj, J. Ren, H.-H. Tseng, A. Gali, A. Rudolph, E. Uccelli, F. Peiró, J.R. Morante, D. Schuh, E. Reiger, E. Kaxiras, J. Arbiol, and A. Fontcuberta i Morral, Phys. Rev. B 83, 045303 (2011).

110 J. Bao, D.C. Bell, F. Capasso, J.B. Wagner, T. Mårtensson, J. Trägårdh, and L. Samuelson, Nano Lett. 8, 836 (2008).

111 S. Perera, M.A. Fickenscher, H.E. Jackson, L.M. Smith, J.M. Yarrison-Rice, H.J. Joyce, Q. Gao, H.H. Tan, C. Jagadish, X. Zhang, and J. Zou, Appl. Phys. Lett. 93, 053110 (2008).

137

112 J.M.E. Harper, J.J. Cuomo, and H.R. Kaufman, J. Vac. Sci. Technol. 21, 737 (1982).

113 C. Zheng, J. Wong-Leung, Q. Gao, H.H. Tan, C. Jagadish, and J. Etheridge, Nano Lett. 13, 3742 (2013).

114 V.S. Wang and R.J. Matyi, J. Electron. Mater. 21, 23 (1992).

115 J.S. Song, Y.C. Choi, S.H. Seo, D.C. Oh, M.W. Cho, T. Yao, and M.H. Oh, J. Cryst. Growth 264, 98 (2004).

116 L. Mcghee, S.G. McMeekin, I. Nicol, M.I. Robertson, and J.M. Winfield, J. Mater. Chem. 4, 29 (1994).

117 S.G. McMeekin, M. Robertson, L. McGhee, and J.M. Winfield, J. Mater. Chem. 2, 367 (1992).

118 Y. Ishikawa, H. Ishii, H. Hasegawa, and T. Fukui, J. Vac. Sci. Technol. B Microelectron. Nanom. Struct. 12, 2713 (1994).

119 C.-W. Cheng, K.-T. Shiu, N. Li, S.-J. Han, L. Shi, and D.K. Sadana, Nat. Commun. 4, 1 (2013).

138

Appendix A: Preparation of the Gold Colloid

Overview of VLS Surface Preparation

The Au is put down onto the growth surface in one of three ways (1) annealing a thin film, (2) drop-casting monodisperse colloids on the surface, or (3) templating a grid of gold onto the surface by electron beam lithography or photolithography. Each of these approaches has their advantages as well as drawbacks depending on the goals of the researcher. The annealing approach generally requires a film of Au on the order of 4 nm thick. This can be done with minimal interaction with the growth surface by placing the wafer directly into the evaporation chamber and then right into the growth chamber for annealing and growth. The resulting nanowires will have a large distribution in diameter as the diameter is controlled by the size of the bead created during the annealing of the

Au film. This can be advantageous if one desires to observe the growth behavior of many different diameters under identical growth conditions, but if one is looking to optimize the growth conditions of a single diameter, a different approach must be taken.

Au colloids suspended in deionized water are available commercially and are often used for the growth of nanowires of uniform diameter. These colloids are relatively cheap (electron beam time at OSU is ~$50/hr, the cost of one bottle of colloid with a shelf life of 6 months) and surface preparation is less time consuming compared to

139 patterning a grid, especially if that grid must be done via electron beam lithography. This method was exclusively employed during the course of this nanowire project and hence is the most familiar to the author. What this technique lacks is the ability to tightly control the spacing between nanowires. At higher surface densities, the average spacing between nanowires has a low standard deviation; however at lower surface densities there tends to be a larger spread in spacing where some colloids are nanometers apart while others can be several microns apart. This behavior is not easily controlled and indeed such behavior is an entire area of interest in the field of soft condensed matter. Nevertheless, this approach remains to be the most popular among researchers. Unless control of spacing down to the 10s of nm is crucial, what is gained in spacing control is lost in processing time and interaction with the growth surface. The lithography process will generally be more time consuming and potentially introduce unwanted organics compared to the simple act of dropping a colloidal suspension on the surface and then blowing it dry.

Figure 74: Au colloids at two densities. The lower density tends to have a larger standard deviation in spacing.

140

Nanowire Surface Prep in the Yang Lab

In this research, the surface preparation method has matured over time, but in its current state, it is described as follows. After the substrates are cut to fit the susceptor pockets, they are placed in a large beaker containing acetone. They are sonicated for 5 minutes and then placed in a secondary beaker full of isopropyl alcohol (IPA). The first beaker is rinsed with IPA and then more IPA is poured in and the substrates are replaced in the beaker. They are then sonicated another 5 minutes and the process is repeated with methanol. The substrates are then blown dry with an N2 gun and placed in a clean fluoroware wafer case. In the case, the substrates are brought to an ultraviolet ozone cleaner where they are placed for 8 minutes with 2 sccm of O2. The surface is then covered in poly-L-lysine (PLL), an organic that positively charges the surface, using a clean pipette to drop it onto the surface. The substrates sit with PLL on the surface covered under a beaker for 15 minutes. They are rinsed with DIW and blown dry. Using another clean pipette, the colloids are then dropped onto the surface and the substrates again sit under a beaker for 15 minutes. The substrates are then blown dry with an N2 gun and placed in their containers for transport to the MOCVD. SEM images of representative examples of prepared surfaces can be seen in Figure 74. In this process, all solvents are poured directly from the bottle into the beaker and all beakers and other containers are pre-cleaned using the same solvent chain process as for the substrates.

141